1This file contains information about GCC releases which has been generated 2automatically from the online release notes. It covers releases of GCC 3(and the former EGCS project) since EGCS 1.0, on the line of development 4that led to GCC 3. For information on GCC 2.8.1 and older releases of GCC 2, 5see ONEWS. 6 7====================================================================== 8http://gcc.gnu.org/gcc-5/index.html 9 10 GCC 5 Release Series 11 12 June 3, 2016 13 14 The [1]GNU project and the GCC developers are pleased to announce the 15 release of GCC 5.4. 16 17 This release is a bug-fix release, containing fixes for regressions in 18 GCC 5.3 relative to previous releases of GCC. 19 20Release History 21 22 GCC 5.4 23 June 3, 2016 ([2]changes, [3]documentation) 24 25 GCC 5.3 26 December 4, 2015 ([4]changes, [5]documentation) 27 28 GCC 5.2 29 July 16, 2015 ([6]changes, [7]documentation) 30 31 GCC 5.1 32 April 22, 2015 ([8]changes, [9]documentation) 33 34References and Acknowledgements 35 36 GCC used to stand for the GNU C Compiler, but since the compiler 37 supports several other languages aside from C, it now stands for the 38 GNU Compiler Collection. 39 40 A list of [10]successful builds is updated as new information becomes 41 available. 42 43 The GCC developers would like to thank the numerous people that have 44 contributed new features, improvements, bug fixes, and other changes as 45 well as test results to GCC. This [11]amazing group of volunteers is 46 what makes GCC successful. 47 48 For additional information about GCC please refer to the [12]GCC 49 project web site or contact the [13]GCC development mailing list. 50 51 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 52 53 54 For questions related to the use of GCC, please consult these web 55 pages and the [16]GCC manuals. If that fails, the 56 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 57 web pages and the development of GCC are welcome on our developer 58 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 59 archives. 60 61 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 62 distribution of this entire article is permitted in any medium, 63 provided this notice is preserved. 64 65 These pages are [21]maintained by the GCC team. Last modified 66 2016-06-03[22]. 67 68References 69 70 1. http://www.gnu.org/ 71 2. http://gcc.gnu.org/gcc-5/changes.html 72 3. http://gcc.gnu.org/onlinedocs/5.4.0/ 73 4. http://gcc.gnu.org/gcc-5/changes.html 74 5. http://gcc.gnu.org/onlinedocs/5.3.0/ 75 6. http://gcc.gnu.org/gcc-5/changes.html 76 7. http://gcc.gnu.org/onlinedocs/5.2.0/ 77 8. http://gcc.gnu.org/gcc-5/changes.html 78 9. http://gcc.gnu.org/onlinedocs/5.1.0/ 79 10. http://gcc.gnu.org/gcc-5/buildstat.html 80 11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 81 12. http://gcc.gnu.org/index.html 82 13. mailto:gcc@gcc.gnu.org 83 14. http://gcc.gnu.org/mirrors.html 84 15. http://gcc.gnu.org/svn.html 85 16. https://gcc.gnu.org/onlinedocs/ 86 17. mailto:gcc-help@gcc.gnu.org 87 18. mailto:gcc@gcc.gnu.org 88 19. https://gcc.gnu.org/lists.html 89 20. http://www.fsf.org/ 90 21. https://gcc.gnu.org/about.html 91 22. http://validator.w3.org/check/referer 92====================================================================== 93http://gcc.gnu.org/gcc-5/changes.html 94 95 GCC 5 Release Series 96 Changes, New Features, and Fixes 97 98Caveats 99 100 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 101 * The C++ runtime library (libstdc++) uses a new ABI by default (see 102 [1]below). 103 * The Graphite framework for loop optimizations no longer requires 104 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 105 The installation manual contains more information about 106 requirements to build GCC. 107 * The non-standard C++0x type traits has_trivial_default_constructor, 108 has_trivial_copy_constructor and has_trivial_copy_assign have been 109 deprecated and will be removed in a future version. The standard 110 C++11 traits is_trivially_default_constructible, 111 is_trivially_copy_constructible and is_trivially_copy_assignable 112 should be used instead. 113 * On AVR, support has been added for the devices 114 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 115 * The AVR port uses a new scheme to describe supported devices: For 116 each supported device the compiler provides a device-specific 117 [2]spec file. If the compiler is used together with AVR-LibC, this 118 requires at least GCC 5.2 and a version of AVR-LibC which 119 implements [3]feature #44574. 120 121General Optimizer Improvements 122 123 * Inter-procedural optimization improvements: 124 + An Identical Code Folding (ICF) pass (controlled via 125 -fipa-icf) has been added. Compared to the identical code 126 folding performed by the Gold linker this pass does not 127 require function sections. It also performs merging before 128 inlining, so inter-procedural optimizations are aware of the 129 code re-use. On the other hand not all unifications performed 130 by a linker are doable by GCC which must honor aliasing 131 information. During link-time optimization of Firefox, this 132 pass unifies about 31000 functions, that is 14% overall. 133 + The devirtualization pass was significantly improved by adding 134 better support for speculative devirtualization and dynamic 135 type detection. About 50% of virtual calls in Firefox are now 136 speculatively devirtualized during link-time optimization. 137 + A new comdat localization pass allows the linker to eliminate 138 more dead code in presence of C++ inline functions. 139 + Virtual tables are now optimized. Local aliases are used to 140 reduce dynamic linking time of C++ virtual tables on ELF 141 targets and data alignment has been reduced to limit data 142 segment bloat. 143 + A new -fno-semantic-interposition option can be used to 144 improve code quality of shared libraries where interposition 145 of exported symbols is not allowed. 146 + Write-only variables are now detected and optimized out. 147 + With profile feedback the function inliner can now bypass 148 --param inline-insns-auto and --param inline-insns-single 149 limits for hot calls. 150 + The IPA reference pass was significantly sped up making it 151 feasible to enable -fipa-reference with -fprofile-generate. 152 This also solves a bottleneck seen when building Chromium with 153 link-time optimization. 154 + The symbol table and call-graph API was reworked to C++ and 155 simplified. 156 + The interprocedural propagation of constants now also 157 propagates alignments of pointer parameters. This for example 158 means that the vectorizer often does not need to generate loop 159 prologues and epilogues to make up for potential 160 misalignments. 161 * Link-time optimization improvements: 162 + One Definition Rule based merging of C++ types has been 163 implemented. Type merging enables better devirtualization and 164 alias analysis. Streaming extra information needed to merge 165 types adds about 2-6% of memory size and object size increase. 166 This can be controlled by -flto-odr-type-merging. 167 + Command-line optimization and target options are now streamed 168 on a per-function basis and honored by the link-time 169 optimizer. This change makes link-time optimization a more 170 transparent replacement of per-file optimizations. It is now 171 possible to build projects that require different optimization 172 settings for different translation units (such as -ffast-math, 173 -mavx, or -finline). Contrary to earlier GCC releases, the 174 optimization and target options passed on the link command 175 line are ignored. 176 Note that this applies only to those command-line options that 177 can be passed to optimize and target attributes. Command-line 178 options affecting global code generation (such as -fpic), 179 warnings (such as -Wodr), optimizations affecting the way 180 static variables are optimized (such as -fcommon), debug 181 output (such as -g), and --param parameters can be applied 182 only to the whole link-time optimization unit. In these cases, 183 it is recommended to consistently use the same options at both 184 compile time and link time. 185 + GCC bootstrap now uses slim LTO object files. 186 + Memory usage and link times were improved. Tree merging was 187 sped up, memory usage of GIMPLE declarations and types was 188 reduced, and, support for on-demand streaming of variable 189 constructors was added. 190 * Feedback directed optimization improvements: 191 + A new auto-FDO mode uses profiles collected by low overhead 192 profiling tools (perf) instead of more expensive program 193 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 194 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 195 traditional feedback directed optimization. 196 + Profile precision was improved in presence of C++ inline and 197 extern inline functions. 198 + The new gcov-tool utility allows manipulating profiles. 199 + Profiles are now more tolerant to source file changes (this 200 can be controlled by --param profile-func-internal-id). 201 * Register allocation improvements: 202 + A new local register allocator (LRA) sub-pass, controlled by 203 -flra-remat, implements control-flow sensitive global register 204 rematerialization. Instead of spilling and restoring a 205 register value, it is recalculated if it is profitable. The 206 sub-pass improved SPEC2000 generated code by 1% and 0.5% 207 correspondingly on ARM and x86-64. 208 + Reuse of the PIC hard register, instead of using a fixed 209 register, was implemented on x86/x86-64 targets. This improves 210 generated PIC code performance as more hard registers can be 211 used. Shared libraries can significantly benefit from this 212 optimization. Currently it is switched on only for x86/x86-64 213 targets. As RA infrastructure is already implemented for PIC 214 register reuse, other targets might follow this in the future. 215 + A simple form of inter-procedural RA was implemented. When it 216 is known that a called function does not use caller-saved 217 registers, save/restore code is not generated around the call 218 for such registers. This optimization can be controlled by 219 -fipa-ra 220 + LRA is now much more effective at generating spills of general 221 registers into vector registers instead of memory on 222 architectures (e.g., modern Intel processors) where this is 223 profitable. 224 * UndefinedBehaviorSanitizer gained a few new sanitization options: 225 + -fsanitize=float-divide-by-zero: detect floating-point 226 division by zero; 227 + -fsanitize=float-cast-overflow: check that the result of 228 floating-point type to integer conversions do not overflow; 229 + -fsanitize=bounds: enable instrumentation of array bounds and 230 detect out-of-bounds accesses; 231 + -fsanitize=alignment: enable alignment checking, detect 232 various misaligned objects; 233 + -fsanitize=object-size: enable object size checking, detect 234 various out-of-bounds accesses. 235 + -fsanitize=vptr: enable checking of C++ member function calls, 236 member accesses and some conversions between pointers to base 237 and derived classes, detect if the referenced object does not 238 have the correct dynamic type. 239 * Pointer Bounds Checker, a bounds violation detector, has been added 240 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 241 instrumented with run-time checks of used pointers against their 242 bounds to detect pointer bounds violations (overflows). The Pointer 243 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 244 new ISA extension Intel MPX support. See the Pointer Bounds Checker 245 [4]Wiki page for more details. 246 247New Languages and Language specific improvements 248 249 * [5]OpenMP 4.0 specification offloading features are now supported 250 by the C, C++, and Fortran compilers. Generic changes: 251 + Infrastructure (suitable for any vendor). 252 + Testsuite which covers offloading from the [6]OpenMP 4.0 253 Examples document. 254 Specific for upcoming Intel Xeon Phi products: 255 + Run-time library. 256 + Card emulator. 257 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 258 specification. OpenACC is intended for programming accelerator 259 devices such as GPUs. See [7]the OpenACC wiki page for more 260 information. 261 262 C family 263 264 * The default setting of the -fdiagnostics-color= command-line option 265 is now [8]configurable when building GCC using configuration option 266 --with-diagnostics-color=. The possible values are: never, always, 267 auto and auto-if-env. The new default auto uses color only when the 268 standard error is a terminal. The default in GCC 4.9 was 269 auto-if-env, which is equivalent to auto if there is a non-empty 270 GCC_COLORS environment variable, and never otherwise. As in GCC 271 4.9, an empty GCC_COLORS variable in the environment will always 272 disable colors, no matter what the default is or what command-line 273 options are used. 274 * A new command-line option -Wswitch-bool has been added for the C 275 and C++ compilers, which warns whenever a switch statement has an 276 index of boolean type. 277 * A new command-line option -Wlogical-not-parentheses has been added 278 for the C and C++ compilers, which warns about "logical not" used 279 on the left hand side operand of a comparison. 280 * A new command-line option -Wsizeof-array-argument has been added 281 for the C and C++ compilers, which warns when the sizeof operator 282 is applied to a parameter that has been declared as an array in a 283 function definition. 284 * A new command-line option -Wbool-compare has been added for the C 285 and C++ compilers, which warns about boolean expressions compared 286 with an integer value different from true/false. 287 * Full support for [9]Cilk Plus has been added to the GCC compiler. 288 Cilk Plus is an extension to the C and C++ languages to support 289 data and task parallelism. 290 * A new attribute no_reorder prevents reordering of selected symbols 291 against other such symbols or inline assembler. This enables to 292 link-time optimize the Linux kernel without having to resort to 293 -fno-toplevel-reorder that disables several optimizations. 294 * New preprocessor constructs, __has_include and __has_include_next, 295 to test the availability of headers have been added. 296 This demonstrates a way to include the header <optional> only if it 297 is available: 298 299#ifdef __has_include 300# if __has_include(<optional>) 301# include <optional> 302# define have_optional 1 303# elif __has_include(<experimental/optional>) 304# include <experimental/optional> 305# define have_optional 1 306# define experimental_optional 307# else 308# define have_optional 0 309# endif 310#endif 311 312 The header search paths for __has_include and __has_include_next 313 are equivalent to those of the standard directive #include and the 314 extension #include_next respectively. 315 * A new built-in function-like macro to determine the existence of an 316 attribute, __has_attribute, has been added. The equivalent built-in 317 macro __has_cpp_attribute was added to C++ to support 318 [10]Feature-testing recommendations for C++. The macro 319 __has_attribute is added to all C-like languages as an extension: 320 321int 322#ifdef __has_attribute 323# if __has_attribute(__noinline__) 324 __attribute__((__noinline__)) 325# endif 326#endif 327foo(int x); 328 329 If an attribute exists, a nonzero constant integer is returned. For 330 standardized C++ attributes a date is returned, otherwise the 331 constant returned is 1. Both __has_attribute and 332 __has_cpp_attribute will add underscores to an attribute name if 333 necessary to resolve the name. For C++11 and onwards the attribute 334 may be scoped. 335 * A new set of built-in functions for arithmetics with overflow 336 checking has been added: __builtin_add_overflow, 337 __builtin_sub_overflow and __builtin_mul_overflow and for 338 compatibility with clang also other variants. These builtins have 339 two integral arguments (which don't need to have the same type), 340 the arguments are extended to infinite precision signed type, +, - 341 or * is performed on those, and the result is stored in an integer 342 variable pointed to by the last argument. If the stored value is 343 equal to the infinite precision result, the built-in functions 344 return false, otherwise true. The type of the integer variable that 345 will hold the result can be different from the types of the first 346 two arguments. The following snippet demonstrates how this can be 347 used in computing the size for the calloc function: 348 349void * 350calloc (size_t x, size_t y) 351{ 352 size_t sz; 353 if (__builtin_mul_overflow (x, y, &sz)) 354 return NULL; 355 void *ret = malloc (sz); 356 if (ret) memset (res, 0, sz); 357 return ret; 358} 359 360 On e.g. i?86 or x86-64 the above will result in a mul instruction 361 followed by a jump on overflow. 362 * The option -fextended-identifiers is now enabled by default for 363 C++, and for C99 and later C versions. Various bugs in the 364 implementation of extended identifiers have been fixed. 365 366 C 367 368 * The default mode has been changed to -std=gnu11. 369 * A new command-line option -Wc90-c99-compat has been added to warn 370 about features not present in ISO C90, but present in ISO C99. 371 * A new command-line option -Wc99-c11-compat has been added to warn 372 about features not present in ISO C99, but present in ISO C11. 373 * It is possible to disable warnings about conversions between 374 pointers that have incompatible types via a new warning option 375 -Wno-incompatible-pointer-types; warnings about implicit 376 incompatible integer to pointer and pointer to integer conversions 377 via a new warning option -Wno-int-conversion; and warnings about 378 qualifiers on pointers being discarded via a new warning option 379 -Wno-discarded-qualifiers. 380 * To allow proper use of const qualifiers with multidimensional 381 arrays, GCC will not warn about incompatible pointer types anymore 382 for conversions between pointers to arrays with and without const 383 qualifier (except when using -pedantic). Instead, a new warning is 384 emitted only if the const qualifier is lost. This can be controlled 385 with a new warning option -Wno-discarded-array-qualifiers. 386 * The C front end now generates more precise caret diagnostics. 387 * The -pg command-line option now only affects the current file in an 388 LTO build. 389 390 C++ 391 392 * G++ now supports [11]C++14 variable templates. 393 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 394 * Excessive template instantiation depth is now a fatal error. This 395 prevents excessive diagnostics that usually do not help to identify 396 the problem. 397 * G++ and libstdc++ now implement the feature-testing macros from 398 [12]Feature-testing recommendations for C++. 399 * G++ now allows typename in a template template parameter. 400 401template<template<typename> typename X> struct D; // OK 402 403 * G++ now supports [13]C++14 aggregates with non-static data member 404 initializers. 405 406struct A { int i, j = i; }; 407A a = { 42 }; // a.j is also 42 408 409 * G++ now supports [14]C++14 extended constexpr. 410 411constexpr int f (int i) 412{ 413 int j = 0; 414 for (; i > 0; --i) 415 ++j; 416 return j; 417} 418 419constexpr int i = f(42); // i is 42 420 421 * G++ now supports the [15]C++14 sized deallocation functions. 422 423void operator delete (void *, std::size_t) noexcept; 424void operator delete[] (void *, std::size_t) noexcept; 425 426 * A new One Definition Rule violation warning (controlled by -Wodr) 427 detects mismatches in type definitions and virtual table contents 428 during link-time optimization. 429 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 430 developers to annotate programs with final specifiers (or anonymous 431 namespaces) to improve code generation. These warnings can be used 432 at compile time, but they are more useful in combination with 433 link-time optimization. 434 * G++ no longer supports [16]N3639 variable length arrays, as they 435 were removed from the C++14 working paper prior to ratification. 436 GNU VLAs are still supported, so VLA support is now the same in 437 C++14 mode as in C++98 and C++11 modes. 438 * G++ now allows passing a non-trivially-copyable class via C 439 varargs, which is conditionally-supported with 440 implementation-defined semantics in the standard. This uses the 441 same calling convention as a normal value parameter. 442 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 443 various mangling bugs are fixed, but G++ will still emit aliases 444 with the old, wrong mangling where feasible. -Wabi=2 will warn 445 about differences between ABI version 2 and the current setting. 446 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 447 to be unaffected, but -Wabi=8 will warn about a non-static data 448 member with type std::nullptr_t which changes position due to this 449 change. 450 451 Runtime Library (libstdc++) 452 453 * A [17]Dual ABI is provided by the library. A new ABI is enabled by 454 default. The old ABI is still supported and can be used by defining 455 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 456 standard library headers. 457 * A new implementation of std::string is enabled by default, using 458 the small string optimization instead of copy-on-write reference 459 counting. 460 * A new implementation of std::list is enabled by default, with an 461 O(1) size() function; 462 * [18]Full support for C++11, including the following new features: 463 + std::deque and std::vector<bool> meet the allocator-aware 464 container requirements; 465 + movable and swappable iostream classes; 466 + support for std::align and std::aligned_union; 467 + type traits std::is_trivially_copyable, 468 std::is_trivially_constructible, std::is_trivially_assignable 469 etc.; 470 + I/O manipulators std::put_time, std::get_time, std::hexfloat 471 and std::defaultfloat; 472 + generic locale-aware std::isblank; 473 + locale facets for Unicode conversion; 474 + atomic operations for std::shared_ptr; 475 + std::notify_all_at_thread_exit() and functions for making 476 futures ready at thread exit. 477 * Support for the C++11 hexfloat manipulator changes how the num_put 478 facet formats floating point types when 479 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 480 This change affects all language modes, even though the C++98 481 standard gave no special meaning to that combination of flags. To 482 prevent the use of hexadecimal notation for floating point types 483 use str.unsetf(std::ios_base::floatfield) to clear the relevant 484 bits in str.flags(). 485 * [19]Full experimental support for C++14, including the following 486 new features: 487 + std::is_final type trait; 488 + heterogeneous comparison lookup in associative containers. 489 + global functions cbegin, cend, rbegin, rend, crbegin, and 490 crend for range access to containers, arrays and initializer 491 lists. 492 * [20]Improved experimental support for the Library Fundamentals TS, 493 including: 494 + class std::experimental::any; 495 + function template std::experimental::apply; 496 + function template std::experimental::sample; 497 + function template std::experimental::search and related 498 searcher types; 499 + variable templates for type traits; 500 + function template std::experimental::not_fn. 501 * New random number distributions logistic_distribution and 502 uniform_on_sphere_distribution as extensions. 503 * [21]GDB Xmethods for containers and std::unique_ptr. 504 505 Fortran 506 507 * Compatibility notice: 508 + The version of the module files (.mod) has been incremented. 509 + For free-form source files, [22]-Werror=line-truncation is now 510 enabled by default; note that comments exceeding the line 511 length are not diagnosed. (For fixed-form source code, the 512 same warning is available but turned off by default, such that 513 excess characters are ignored. -ffree-line-length-n and 514 -ffixed-line-length-n can be used to modify the default line 515 lengths of 132 and 72 columns, respectively.) 516 + The -Wtabs option is now more sensible: with -Wtabs the 517 compiler warns if it encounters tabs and with -Wno-tabs this 518 warning is turned off. Before, -Wno-tabs warned and -Wtabs 519 turned the warning off. As before, the warning is also enabled 520 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 521 options of -std=. 522 * Incomplete support for colorizing diagnostics emitted by gfortran 523 has been added. The option [23]-fdiagnostics-color controls when 524 color is used in diagnostics. The default value of this option can 525 be [24]configured when building GCC. The GCC_COLORS environment 526 variable can be used to customize the colors or disable coloring 527 completely. Sample diagnostics output: 528 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 529 test.f90:6:1: 530 531 0 continue 532 1 533 Error: Zero is not a valid statement label at (1) 534 test.f90:9:6: 535 536 USE foo 537 1 538 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 539 540 * The -Wuse-without-only option has been added to warn when a USE 541 statement has no ONLY qualifier and thus implicitly imports all 542 public entities of the used module. 543 * Formatted READ and WRITE statements now work correctly in 544 locale-aware programs. For more information and potential caveats, 545 see [25]Section 5.3 Thread-safety of the runtime library in the 546 manual. 547 * [26]Fortran 2003: 548 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 549 IEEE_ARITHMETIC) are now supported. 550 * [27]Fortran 2008: 551 + [28]Coarrays: Full experimental support of Fortran 2008's 552 coarrays with -fcoarray=lib except for allocatable/pointer 553 components of derived-type coarrays. GCC currently only ships 554 with a single-image library (libcaf_single), but multi-image 555 support based on MPI and GASNet is provided by the libraries 556 of the [29]OpenCoarrays project. 557 * TS18508 Additional Parallel Features in Fortran: 558 + Support for the collective intrinsic subroutines CO_MAX, 559 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 560 including -fcoarray=lib support. 561 + Support for the new atomic intrinsics has been added, 562 including -fcoarray=lib support. 563 * Fortran 2015: 564 + Support for IMPLICIT NONE (external, type). 565 + ERROR STOP is now permitted in pure procedures. 566 567 Go 568 569 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 570 * Building GCC 5 with Go enabled will install two new programs: 571 [30]go and [31]gofmt. 572 573libgccjit 574 575 New in GCC 5 is the ability to build GCC as a shared library for 576 embedding in other processes (such as interpreters), suitable for 577 Just-In-Time compilation to machine code. 578 579 The shared library has a [32]C API and a [33]C++ wrapper API providing 580 some "syntactic sugar". There are also bindings available from 3rd 581 parties for [34]Python and for [35]D. 582 583 For example, this library can be used by interpreters for [36]compiling 584 functions from bytecode to machine code. 585 586 The library can also be used for ahead-of-time compilation, enabling 587 GCC to be plugged into a pre-existing frontend. An example of using 588 this to build a compiler for an esoteric language we'll refer to as 589 "brainf" can be seen [37]here. 590 591 libgccjit is licensed under the GPLv3 (or at your option, any later 592 version) 593 594 It should be regarded as experimental at this time. 595 596New Targets and Target Specific Improvements 597 598 Reporting stack usage 599 600 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 601 -fstack-usage option. 602 603 AArch64 604 605 * Code generation for the ARM Cortex-A57 processor has been improved. 606 A more accurate instruction scheduling model for the processor is 607 now used, and a number of compiler tuning parameters have been set 608 to offer increased performance when compiling with -mcpu=cortex-a57 609 or -mtune=cortex-a57. 610 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 611 and can be enabled by giving the -mfix-cortex-a53-835769 option. 612 Alternatively it can be enabled by default by configuring GCC with 613 the --enable-fix-cortex-a53-835769 option. 614 * The optional cryptographic extensions to the ARMv8-A architecture 615 are no longer enabled by default when specifying the 616 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 617 options. To enable these extensions add +crypto to the value of 618 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 619 * Support has been added for the following processors (GCC 620 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 621 initial support for its big.LITTLE combination with the ARM 622 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 623 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 624 The GCC identifiers can be used as arguments to the -mcpu or -mtune 625 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 626 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 627 support for the Cortex-A72. 628 * The transitional options -mlra and -mno-lra have been removed. The 629 AArch64 backend now uses the local register allocator (LRA) only. 630 631 ARM 632 633 * Thumb-1 assembly code is now generated in unified syntax. The new 634 option -masm-syntax-unified specifies whether inline assembly code 635 is using unified syntax. By default the option is off which means 636 non-unified syntax is used. However this is subject to change in 637 future releases. Eventually the non-unified syntax will be 638 deprecated. 639 * It is now a configure-time error to use the --with-cpu configure 640 option with either of --with-tune or --with-arch. 641 * Code generation for the ARM Cortex-A57 processor has been improved. 642 A more accurate instruction scheduling model for the processor is 643 now used, and a number of compiler tuning parameters have been set 644 to offer increased performance when compiling with -mcpu=cortex-a57 645 or -mtune=cortex-a57. 646 * Support has been added for the following processors (GCC 647 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 648 initial support for its big.LITTLE combination with the ARM 649 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 650 initial support for its big.LITTLE combination with the ARM 651 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 652 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 653 The GCC identifiers can be used as arguments to the -mcpu or -mtune 654 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 655 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 656 support for the Cortex-A72. 657 * The deprecated option -mwords-little-endian has been removed. 658 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 659 -mtpcs-leaf-frame which are only applicable to the old ABI have 660 been deprecated. 661 * The transitional options -mlra and -mno-lra have been removed. The 662 ARM backend now uses the local register allocator (LRA) only. 663 664 AVR 665 666 * The compiler no more supports individual devices like ATmega8. 667 Specifying, say, -mmcu=atmega8 triggers the usage of the 668 device-specific [38]spec file specs-atmega8 which is part of the 669 installation and describes options for the sub-processes like 670 compiler proper, assembler and linker. You can add support for a 671 new device -mmcu=mydevice as follows: 672 1. In an empty directory /someplace, create a new directory 673 device-specs. 674 2. Copy a device spec file from the installed device-specs 675 folder, follow the comments in that file and then save it as 676 /someplace/device-specs/specs-mydevice. 677 3. Add -B /someplace -mmcu=mydevice to the compiler's 678 command-line options. Notice that /someplace must specify an 679 absolute path and that mydevice must not start with "avr". 680 4. Provided you have a device-specific library libmydevice.a 681 available, you can put it at /someplace, dito for a 682 device-specific startup file crtmydevice.o. 683 The contents of the device spec files depend on the compiler's 684 configuration, in particular on --with-avrlibc=no and whether or 685 not it is configured for RTEMS. 686 * A new command-line option -nodevicelib has been added. It prevents 687 the compiler from linking against AVR-LibC's device-specific 688 library libdevice.a. 689 * The following three command-line options have been added: 690 691 -mrmw 692 Set if the device supports the read-modify-write 693 instructions LAC, LAS, LAT and XCH. 694 695 -mn-flash=size 696 Specify the flash size of the device in units of 64 KiB, 697 rounded up to the next integer as needed. This option 698 affects the availability of the [39]AVR address-spaces. 699 700 -mskip-bug 701 Set if the device is affected by the respective silicon 702 bug. 703 704 In general, you don't need to set these options by hand. The new 705 device-specific spec file will set them as needed. 706 707 IA-32/x86-64 708 709 * New ISA extensions support [40]AVX-512{BW,DQ,VL,IFMA,VBMI} of 710 Intel's CPU codenamed Skylake Server was added to GCC. That 711 includes inline assembly support, new intrinsics, and basic 712 autovectorization. These new AVX-512 extensions are available via 713 the following GCC switches: AVX-512 Vector Length EVEX feature: 714 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 715 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 716 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 717 Instructions: -mavx512vbmi. 718 * New ISA extensions support [41]Intel MPX was added to GCC. This new 719 extension is available via the -mmpx compiler switch. Intel MPX is 720 a set of processor features which, with compiler, run-time library 721 and OS support, brings increased robustness to software by run-time 722 checking pointer references against their bounds. In GCC Intel MPX 723 is supported by Pointer Bounds Checker and libmpx run-time 724 libraries. 725 * The new -mrecord-mcount option for -pg generates a Linux kernel 726 style table of pointers to mcount or __fentry__ calls at the 727 beginning of functions. The new -mnop-mcount option in addition 728 also generates nops in place of the __fentry__ or mcount call, so 729 that a call per function can be later patched in. This can be used 730 for low overhead tracing or hot code patching. 731 * The new -malign-data option controls how GCC aligns variables. 732 -malign-data=compat uses increased alignment compatible with GCC 733 4.8 and earlier, -malign-data=abi uses alignment as specified by 734 the psABI, and -malign-data=cacheline uses increased alignment to 735 match the cache line size. -malign-data=compat is the default. 736 * The new -mskip-rax-setup option skips setting up the RAX register 737 when SSE is disabled and there are no variable arguments passed in 738 vector registers. This can be used to optimize the Linux kernel. 739 740 MIPS 741 742 * MIPS Releases 3 and 5 are now directly supported. Use the 743 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 744 to enable code-generation for these processors. 745 * The Imagination P5600 processor is now supported using the 746 -march=p5600 command-line option. 747 * The Cavium Octeon3 processor is now supported using the 748 -march=octeon3 command-line option. 749 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 750 command-line options. 751 * The o32 ABI has been modified and extended. The o32 64-bit 752 floating-point register support is now obsolete and has been 753 removed. It has been replaced by three ABI extensions FPXX, FP64A, 754 and FP64. The meaning of the -mfp64 command-line option has 755 changed. It is now used to enable the FP64A and FP64 ABI 756 extensions. 757 + The FPXX extension requires that code generated to access 758 double-precision values use even-numbered registers. Code that 759 adheres to this extension is link-compatible with all other 760 o32 double-precision ABI variants and will execute correctly 761 in all hardware FPU modes. The command-line options -mabi=32 762 -mfpxx can be used to enable this extension. MIPS II is the 763 minimum processor required. 764 + The o32 FP64A extension requires that floating-point registers 765 be 64-bit and odd-numbered single-precision registers are not 766 allowed. Code that adheres to the o32 FP64A variant is 767 link-compatible with all other o32 double-precision ABI 768 variants. The command-line options -mabi=32 -mfp64 769 -mno-odd-spreg can be used to enable this extension. MIPS32R2 770 is the minimum processor required. 771 + The o32 FP64 extension also requires that floating-point 772 registers be 64-bit, but permits the use of single-precision 773 registers. Code that adheres to the o32 FP64 variant is 774 link-compatible with o32 FPXX and o32 FP64A variants only, 775 i.e. it is not compatible with the original o32 776 double-precision ABI. The command-line options -mabi=32 -mfp64 777 -modd-spreg can be used to enable this extension. MIPS32R2 is 778 the minimum processor required. 779 The new ABI variants can be enabled by default using the configure 780 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 781 It is strongly recommended that all vendors begin to set o32 FPXX 782 as the default ABI. This will be required to run the generated code 783 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 784 MIPSR6 cores. 785 * GCC will now pass all floating-point options to the assembler if 786 GNU binutils 2.25 is used. As a result, any inline assembly code 787 that uses hard-float instructions should be amended to include a 788 .set directive to override the global assembler options when 789 compiling for soft-float targets. 790 791 NDS32 792 793 * The variadic function ABI implementation is now compatible with 794 past Andes toolchains where the caller uses registers to pass 795 arguments and the callee is in charge of pushing them on stack. 796 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 797 been removed since they are not yet available in the nds32 port of 798 GNU binutils. 799 * A new option -mcmodel=[small|medium|large] supports varied code 800 models on code generation. The -mgp-direct option became 801 meaningless and can be discarded. 802 803 RX 804 805 * A new command line option -mno-allow-string-insns can be used to 806 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 807 SWHILE and RMPA instructions. An erratum released by Renesas shows 808 that it is unsafe to use these instructions on addresses within the 809 I/O space of the processor. The new option can be used when the 810 programmer is concerned that the I/O space might be accessed. The 811 default is still to enable these instructions. 812 813 SH 814 815 * The compiler will now pass the appropriate --isa= option to the 816 assembler. 817 * The default handling for the GBR has been changed from call 818 clobbered to call preserved. The old behavior can be reinstated by 819 specifying the option -fcall-used-gbr. 820 * Support for the SH4A fpchg instruction has been added which will be 821 utilized when switching between single and double precision FPU 822 modes. 823 * The compiler no longer uses the __fpscr_values array for switching 824 between single and double FPU precision modes on non-SH4A targets. 825 Instead mode switching will now be performed by storing, modifying 826 and reloading the FPSCR, so that other FPSCR bits are preserved 827 across mode switches. The __fpscr_values array that is defined in 828 libgcc is still present for backwards compatibility, but it will 829 not be referenced by compiler generated code anymore. 830 * New builtin functions __builtin_sh_get_fpscr and 831 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 832 function will mask the specified bits in such a way that the SZ, PR 833 and FR mode bits will be preserved, while changing the other bits. 834 These new functions do not reference the __fpscr_values array. The 835 old functions __set_fpscr and __get_fpscr in libgcc which access 836 the __fpscr_values array are still present for backwards 837 compatibility, but their usage is highly discouraged. 838 * Some improvements to code generated for __atomic built-in 839 functions. 840 * When compiling for SH2E the compiler will no longer force the usage 841 of delay slots for conditional branch instructions bt and bf. The 842 old behavior can be reinstated (e.g. to work around a hardware bug 843 in the original SH7055) by specifying the new option 844 -mcbranch-force-delay-slot. 845 846Operating Systems 847 848 DragonFly BSD 849 850 * GCC now supports the DragonFly BSD operating system. 851 852 FreeBSD 853 854 * GCC now supports the FreeBSD operating system for the arm port 855 through the arm*-*-freebsd* target triplets. 856 857 VxWorks MILS 858 859 * GCC now supports the MILS (Multiple Independent Levels of Security) 860 variant of WindRiver's VxWorks operating system for PowerPC 861 targets. 862 863Other significant improvements 864 865 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 866 to set the compiler to use. 867 868 * When the new command-line option -freport-bug is used, GCC 869 automatically generates a developer-friendly reproducer whenever an 870 internal compiler error is encountered. 871 872GCC 5.2 873 874 This is the [42]list of problem reports (PRs) from GCC's bug tracking 875 system that are known to be fixed in the 5.2 release. This list might 876 not be complete (that is, it is possible that some PRs that have been 877 fixed are not listed here). 878 879Target Specific Changes 880 881 IA-32/x86-64 882 883 * Support for new AMD instructions monitorx and mwaitx has been 884 added. This includes new intrinsic and built-in support. It is 885 enabled through option -mmwaitx. The instructions monitorx and 886 mwaitx implement the same functionality as the old monitor and 887 mwait instructions. In addition mwaitx adds a configurable timer. 888 The timer value is received as third argument and stored in 889 register %ebx. 890 891 S/390, System z, IBM z Systems 892 893 * Support for the IBM z13 processor has been added. When using the 894 -march=z13 option, the compiler will generate code making use of 895 the new instructions and registers introduced with the vector 896 extension facility. The -mtune=z13 option enables z13 specific 897 instruction scheduling without making use of new instructions. 898 Compiling code with -march=z13 reduces the default alignment of 899 vector types bigger than 8 bytes to 8. This is an ABI change and 900 care must be taken when linking modules compiled with different 901 arch levels which interchange variables containing vector type 902 values. For newly compiled code the GNU linker will emit a warning. 903 * The -mzvector option enables a C/C++ language extension. This 904 extension provides a new keyword vector which can be used to define 905 vector type variables. (Note: This is not available when enforcing 906 strict standard compliance e.g. with -std=c99. Either enable GNU 907 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 908 Additionally a set of overloaded builtins is provided which is 909 partially compatible to the PowerPC Altivec builtins. In order to 910 make use of these builtins the vecintrin.h header file needs to be 911 included. 912 913GCC 5.3 914 915 This is the [43]list of problem reports (PRs) from GCC's bug tracking 916 system that are known to be fixed in the 5.3 release. This list might 917 not be complete (that is, it is possible that some PRs that have been 918 fixed are not listed here). 919 920Target Specific Changes 921 922 IA-32/x86-64 923 924 * GCC now supports the Intel CPU named Skylake with AVX-512 925 extensions through -march=skylake-avx512. The switch enables the 926 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 927 AVX-512DQ. 928 929 S/390, System z, IBM z Systems 930 931 * With this version of GCC IBM z Systems support has been added to 932 the GO runtime environment. GCC 5.3 has proven to be able to 933 compile larger GO applications on IBM z Systems. 934 935GCC 5.4 936 937 This is the [44]list of problem reports (PRs) from GCC's bug tracking 938 system that are known to be fixed in the 5.4 release. This list might 939 not be complete (that is, it is possible that some PRs that have been 940 fixed are not listed here). 941 942(Pending) GCC 5.5 943 944 This is the [45]list of problem reports (PRs) from GCC's bug tracking 945 system that are known to be fixed in the 5.5 release. This list might 946 not be complete (that is, it is possible that some PRs that have been 947 fixed are not listed here). 948 949 950 For questions related to the use of GCC, please consult these web 951 pages and the [46]GCC manuals. If that fails, the 952 [47]gcc-help@gcc.gnu.org mailing list might help. Comments on these 953 web pages and the development of GCC are welcome on our developer 954 list at [48]gcc@gcc.gnu.org. All of [49]our lists have public 955 archives. 956 957 Copyright (C) [50]Free Software Foundation, Inc. Verbatim copying and 958 distribution of this entire article is permitted in any medium, 959 provided this notice is preserved. 960 961 These pages are [51]maintained by the GCC team. Last modified 962 2016-06-03[52]. 963 964References 965 966 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 967 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 968 3. https://savannah.nongnu.org/bugs/?44574 969 4. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler 970 5. http://www.openmp.org/mp-documents/OpenMP4.0.0.pdf 971 6. http://openmp.org/mp-documents/OpenMP4.0.0.Examples.pdf 972 7. https://gcc.gnu.org/wiki/OpenACC 973 8. https://gcc.gnu.org/install/configure.html 974 9. https://www.cilkplus.org/ 975 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 976 11. http://gcc.gnu.org/projects/cxx1y.html 977 12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 978 13. http://gcc.gnu.org/projects/cxx1y.html 979 14. http://gcc.gnu.org/projects/cxx1y.html 980 15. http://gcc.gnu.org/projects/cxx1y.html 981 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 982 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 983 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 984 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 985 20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 986 21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 987 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 988 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 989 24. https://gcc.gnu.org/install/configure.html 990 25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 991 26. https://gcc.gnu.org/wiki/Fortran2003Status 992 27. https://gcc.gnu.org/wiki/Fortran2008Status 993 28. https://gcc.gnu.org/wiki/Coarray 994 29. http://www.opencoarrays.org/ 995 30. https://golang.org/cmd/go/ 996 31. https://golang.org/cmd/gofmt/ 997 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 998 33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 999 34. https://github.com/davidmalcolm/pygccjit 1000 35. https://github.com/ibuclaw/gccjitd 1001 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 1002 37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 1003 38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 1004 39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 1005 40. https://software.intel.com/sites/default/files/managed/0d/53/319433-023.pdf 1006 41. https://software.intel.com/sites/default/files/managed/0d/53/319433-022.pdf 1007 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 1008 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 1009 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 1010 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 1011 46. https://gcc.gnu.org/onlinedocs/ 1012 47. mailto:gcc-help@gcc.gnu.org 1013 48. mailto:gcc@gcc.gnu.org 1014 49. https://gcc.gnu.org/lists.html 1015 50. http://www.fsf.org/ 1016 51. https://gcc.gnu.org/about.html 1017 52. http://validator.w3.org/check/referer 1018====================================================================== 1019http://gcc.gnu.org/gcc-4.9/index.html 1020 1021 GCC 4.9 Release Series 1022 1023 June 26, 2015 1024 1025 The [1]GNU project and the GCC developers are pleased to announce the 1026 release of GCC 4.9.3. 1027 1028 This release is a bug-fix release, containing fixes for regressions in 1029 GCC 4.9.2 relative to previous releases of GCC. 1030 1031Release History 1032 1033 GCC 4.9.3 1034 June 26, 2015 ([2]changes, [3]documentation) 1035 1036 GCC 4.9.2 1037 October 30, 2014 ([4]changes, [5]documentation) 1038 1039 GCC 4.9.1 1040 July 16, 2014 ([6]changes, [7]documentation) 1041 1042 GCC 4.9.0 1043 April 22, 2014 ([8]changes, [9]documentation) 1044 1045References and Acknowledgements 1046 1047 GCC used to stand for the GNU C Compiler, but since the compiler 1048 supports several other languages aside from C, it now stands for the 1049 GNU Compiler Collection. 1050 1051 A list of [10]successful builds is updated as new information becomes 1052 available. 1053 1054 The GCC developers would like to thank the numerous people that have 1055 contributed new features, improvements, bug fixes, and other changes as 1056 well as test results to GCC. This [11]amazing group of volunteers is 1057 what makes GCC successful. 1058 1059 For additional information about GCC please refer to the [12]GCC 1060 project web site or contact the [13]GCC development mailing list. 1061 1062 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 1063 1064 1065 For questions related to the use of GCC, please consult these web 1066 pages and the [16]GCC manuals. If that fails, the 1067 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 1068 web pages and the development of GCC are welcome on our developer 1069 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 1070 archives. 1071 1072 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 1073 distribution of this entire article is permitted in any medium, 1074 provided this notice is preserved. 1075 1076 These pages are [21]maintained by the GCC team. Last modified 1077 2016-01-30[22]. 1078 1079References 1080 1081 1. http://www.gnu.org/ 1082 2. http://gcc.gnu.org/gcc-4.9/changes.html 1083 3. http://gcc.gnu.org/onlinedocs/4.9.3/ 1084 4. http://gcc.gnu.org/gcc-4.9/changes.html 1085 5. http://gcc.gnu.org/onlinedocs/4.9.2/ 1086 6. http://gcc.gnu.org/gcc-4.9/changes.html 1087 7. http://gcc.gnu.org/onlinedocs/4.9.1/ 1088 8. http://gcc.gnu.org/gcc-4.9/changes.html 1089 9. https://gcc.gnu.org/onlinedocs/4.9.0/ 1090 10. http://gcc.gnu.org/gcc-4.9/buildstat.html 1091 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1092 12. http://gcc.gnu.org/index.html 1093 13. mailto:gcc@gcc.gnu.org 1094 14. http://gcc.gnu.org/mirrors.html 1095 15. http://gcc.gnu.org/svn.html 1096 16. https://gcc.gnu.org/onlinedocs/ 1097 17. mailto:gcc-help@gcc.gnu.org 1098 18. mailto:gcc@gcc.gnu.org 1099 19. https://gcc.gnu.org/lists.html 1100 20. http://www.fsf.org/ 1101 21. https://gcc.gnu.org/about.html 1102 22. http://validator.w3.org/check/referer 1103====================================================================== 1104http://gcc.gnu.org/gcc-4.9/changes.html 1105 1106 GCC 4.9 Release Series 1107 Changes, New Features, and Fixes 1108 1109Caveats 1110 1111 * The mudflap run time checker has been removed. The mudflap options 1112 remain, but do nothing. 1113 * Support for a number of older systems and recently unmaintained or 1114 untested target ports of GCC has been declared obsolete in GCC 4.9. 1115 Unless there is activity to revive them, the next release of GCC 1116 will have their sources permanently removed. 1117 The following ports for individual systems on particular 1118 architectures have been obsoleted: 1119 + Solaris 9 (*-*-solaris2.9). Details can be found in the 1120 [1]announcement. 1121 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 1122 float64x1_t exported by arm_neon.h are defined to be the same as 1123 their base types. This results in incorrect application of 1124 parameter passing rules to arguments of types int64x1_t and 1125 uint64x1_t, with respect to the AAPCS64 ABI specification. In 1126 addition, names of C++ functions with parameters of these types 1127 (including float64x1_t) are not mangled correctly. The current 1128 typedef declarations also unintentionally allow implicit casting 1129 between singleton vector types and their base types. These issues 1130 will be resolved in a near future release. See [2]PR60825 for more 1131 information. 1132 1133 More information on porting to GCC 4.9 from previous versions of GCC 1134 can be found in the [3]porting guide for this release. 1135 1136General Optimizer Improvements 1137 1138 * AddressSanitizer, a fast memory error detector, is now available on 1139 ARM. 1140 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 1141 detector, has been added and can be enabled via 1142 -fsanitize=undefined. Various computations will be instrumented to 1143 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 1144 currently available for the C and C++ languages. 1145 * Link-time optimization (LTO) improvements: 1146 + Type merging was rewritten. The new implementation is 1147 significantly faster and uses less memory. 1148 + Better partitioning algorithm resulting in less streaming 1149 during link time. 1150 + Early removal of virtual methods reduces the size of object 1151 files and improves link-time memory usage and compile time. 1152 + Function bodies are now loaded on-demand and released early 1153 improving overall memory usage at link time. 1154 + C++ hidden keyed methods can now be optimized out. 1155 + When using a linker plugin, compiling with the -flto option 1156 now generates slim object files (.o) which only contain 1157 intermediate language representation for LTO. Use 1158 -ffat-lto-objects to create files which contain additionally 1159 the object code. To generate static libraries suitable for LTO 1160 processing, use gcc-ar and gcc-ranlib; to list symbols from a 1161 slim object file use gcc-nm. (This requires that ar, ranlib 1162 and nm have been compiled with plugin support.) 1163 Memory usage building Firefox with debug enabled was reduced from 1164 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 1165 * Inter-procedural optimization improvements: 1166 + New type inheritance analysis module improving 1167 devirtualization. Devirtualization now takes into account 1168 anonymous name-spaces and the C++11 final keyword. 1169 + New speculative devirtualization pass (controlled by 1170 -fdevirtualize-speculatively. 1171 + Calls that were speculatively made direct are turned back to 1172 indirect where direct call is not cheaper. 1173 + Local aliases are introduced for symbols that are known to be 1174 semantically equivalent across shared libraries improving 1175 dynamic linking times. 1176 * Feedback directed optimization improvements: 1177 + Profiling of programs using C++ inline functions is now more 1178 reliable. 1179 + New time profiling determines typical order in which functions 1180 are executed. 1181 + A new function reordering pass (controlled by 1182 -freorder-functions) significantly reduces startup time of 1183 large applications. Until binutils support is completed, it is 1184 effective only with link-time optimization. 1185 + Feedback driven indirect call removal and devirtualization now 1186 handle cross-module calls when link-time optimization is 1187 enabled. 1188 1189New Languages and Language specific improvements 1190 1191 * Version 4.0 of the [4]OpenMP specification is now supported in the 1192 C and C++ compilers and starting with the 4.9.1 release also in the 1193 Fortran compiler. The new -fopenmp-simd option can be used to 1194 enable OpenMP's SIMD directives while ignoring other OpenMP 1195 directives. The new [5]-fsimd-cost-model= option permits to tune 1196 the vectorization cost model for loops annotated with OpenMP and 1197 Cilk Plus simd directives. -Wopenmp-simd warns when the current 1198 cost model overrides simd directives set by the user. 1199 * The -Wdate-time option has been added for the C, C++ and Fortran 1200 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 1201 macros are used. Those macros might prevent bit-wise-identical 1202 reproducible compilations. 1203 1204 Ada 1205 1206 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 1207 1208 C family 1209 1210 * Support for colorizing diagnostics emitted by GCC has been added. 1211 The [6]-fdiagnostics-color=auto will enable it when outputting to 1212 terminals, -fdiagnostics-color=always unconditionally. The 1213 GCC_COLORS environment variable can be used to customize the colors 1214 or disable coloring. If GCC_COLORS variable is present in the 1215 environment, the default is -fdiagnostics-color=auto, otherwise 1216 -fdiagnostics-color=never. 1217 Sample diagnostics output: 1218 $ g++ -fdiagnostics-color=always -S -Wall test.C 1219 test.C: In function `int foo()': 1220 test.C:1:14: warning: no return statement in function returning non-void [-W 1221return-type] 1222 int foo () { } 1223 ^ 1224 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 1225 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>' 1226 template <int N> struct X { static const int value = X<N-1>::value; }; temp 1227late struct X<1000>; 1228 ^ 1229 test.C:2:46: recursively required from `const int X<999>::value' 1230 test.C:2:46: required from `const int X<1000>::value' 1231 test.C:2:88: required from here 1232 1233 test.C:2:46: error: incomplete type `X<100>' used in nested name specifier 1234 1235 * With the new [7]#pragma GCC ivdep, the user can assert that there 1236 are no loop-carried dependencies which would prevent concurrent 1237 execution of consecutive iterations using SIMD (single instruction 1238 multiple data) instructions. 1239 * Support for [8]Cilk Plus has been added and can be enabled with the 1240 -fcilkplus option. Cilk Plus is an extension to the C and C++ 1241 languages to support data and task parallelism. The present 1242 implementation follows ABI version 1.2; all features but _Cilk_for 1243 have been implemented. 1244 1245 C 1246 1247 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 1248 <stdatomic.h> header) are now supported. 1249 * ISO C11 generic selections (_Generic keyword) are now supported. 1250 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 1251 __thread) is now supported. 1252 * ISO C11 support is now at a similar level of completeness to ISO 1253 C99 support: substantially complete modulo bugs, extended 1254 identifiers (supported except for corner cases when 1255 -fextended-identifiers is used), floating-point issues (mainly but 1256 not entirely relating to optional C99 features from Annexes F and 1257 G) and the optional Annexes K (Bounds-checking interfaces) and L 1258 (Analyzability). 1259 * A new C extension __auto_type provides a subset of the 1260 functionality of C++11 auto in GNU C. 1261 1262 C++ 1263 1264 * The G++ implementation of [9]C++1y return type deduction for normal 1265 functions has been updated to conform to [10]N3638, the proposal 1266 accepted into the working paper. Most notably, it adds 1267 decltype(auto) for getting decltype semantics rather than the 1268 template argument deduction semantics of plain auto: 1269 1270int& f(); 1271 auto i1 = f(); // int 1272decltype(auto) i2 = f(); // int& 1273 1274 * G++ supports [11]C++1y lambda capture initializers: 1275 1276[x = 42]{ ... }; 1277 1278 Actually, they have been accepted since GCC 4.5, but now the 1279 compiler doesn't warn about them with -std=c++1y, and supports 1280 parenthesized and brace-enclosed initializers as well. 1281 * G++ supports [12]C++1y variable length arrays. G++ has supported 1282 GNU/C99-style VLAs for a long time, but now additionally supports 1283 initializers and lambda capture by reference. In C++1y mode G++ 1284 will complain about VLA uses that are not permitted by the draft 1285 standard, such as forming a pointer to VLA type or applying sizeof 1286 to a VLA variable. Note that it now appears that VLAs will not be 1287 part of C++14, but will be part of a separate document and then 1288 perhaps C++17. 1289 1290void f(int n) { 1291 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 1292 [&a]{ for (int i : a) { cout << i << endl; } }(); 1293 &a; // error, taking address of VLA 1294} 1295 1296 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in 1297 the underlying [[gnu::deprecated]] attribute. Classes and functions 1298 can be marked deprecated and a diagnostic message added: 1299 1300class A; 1301int bar(int n); 1302#if __cplusplus > 201103 1303class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 1304[[deprecated("bar is unsafe; use foo() instead")]] 1305int bar(int n); 1306 1307int foo(int n); 1308class B; 1309#endif 1310A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 1311int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 1312o() instead 1313 1314 * G++ supports [14]C++1y digit separators. Long numeric literals can 1315 be subdivided with a single quote ' to enhance readability: 1316 1317int i = 1048576; 1318int j = 1'048'576; 1319int k = 0x10'0000; 1320int m = 0'004'000'000; 1321int n = 0b0001'0000'0000'0000'0000'0000; 1322 1323double x = 1.602'176'565e-19; 1324double y = 1.602'176'565e-1'9; 1325 1326 * G++ supports [15]C++1y generic (polymorphic) lambdas. 1327 1328// a functional object that will increment any type 1329auto incr = [](auto x) { return x++; }; 1330 1331 * As a GNU extension, G++ supports explicit template parameter syntax 1332 for generic lambdas. This can be combined in the expected way with 1333 the standard auto syntax. 1334 1335// a functional object that will add two like-type objects 1336auto add = [] <typename T> (T a, T b) { return a + b; }; 1337 1338 * G++ supports unconstrained generic functions as specified by 1339 S:4.1.2 and S:5.1.1 of [16]N3889: Concepts Lite Specification. 1340 Briefly, auto may be used as a type-specifier in a parameter 1341 declaration of any function declarator in order to introduce an 1342 implicit function template parameter, akin to generic lambdas. 1343 1344// the following two function declarations are equivalent 1345auto incr(auto x) { return x++; } 1346template <typename T> 1347auto incr(T x) { return x++; } 1348 1349 Runtime Library (libstdc++) 1350 1351 * [17]Improved support for C++11, including: 1352 + support for <regex>; 1353 + The associative containers in <map> and <set> and the 1354 unordered associative containers in <unordered_map> and 1355 <unordered_set> meet the allocator-aware container 1356 requirements; 1357 * [18]Improved experimental support for the upcoming ISO C++ 1358 standard, C++14, including: 1359 + fixing constexpr member functions without const; 1360 + implementation of the std::exchange() utility function; 1361 + addressing tuples by type; 1362 + implemention of std::make_unique; 1363 + implemention of std::shared_lock; 1364 + making std::result_of SFINAE-friendly; 1365 + adding operator() to std::integral_constant; 1366 + adding user-defined literals for standard library types 1367 std::basic_string, std::chrono::duration, and std::complex; 1368 + adding two range overloads to non-modifying sequence oprations 1369 std::equal and std::mismatch; 1370 + adding IO manipulators for quoted strings; 1371 + adding constexpr members to <utility>, <complex>, <chrono>, 1372 and some containers; 1373 + adding compile-time std::integer_sequence; 1374 + adding cleaner transformation traits; 1375 + making <functional>s operator functors easier to use and more 1376 generic; 1377 * An implementation of std::experimental::optional. 1378 * An implementation of std::experimental::string_view. 1379 * The non-standard function std::copy_exception has been deprecated 1380 and will be removed in a future version. std::make_exception_ptr 1381 should be used instead. 1382 1383 Fortran 1384 1385 * Compatibility notice: 1386 + Module files: The version of the module files (.mod) has been 1387 incremented; additionally, module files are now compressed. 1388 Fortran MODULEs compiled by earlier GCC versions have to be 1389 recompiled, when they are USEd by files compiled with GCC 4.9. 1390 GCC 4.9 is not able to read .mod files of earlier GCC 1391 versions; attempting to do so gives an error message. Note: 1392 The ABI of the produced assembler data itself has not changed: 1393 object files and libraries are fully compatible with older 1394 versions (except as stated below). 1395 + ABI changes: 1396 o The [19]argument passing ABI has changed for scalar dummy 1397 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 1398 which have both the VALUE and the OPTIONAL attributes. 1399 o To support finalization the virtual table associated with 1400 polymorphic variables has changed. Code containing CLASS 1401 should be recompiled, including all files which define 1402 derived types involved in the type definition used by 1403 polymorphic variables. (Note: Due to the incremented 1404 module version, trying to mix old code with new code will 1405 usually give an error message.) 1406 + GNU Fortran no longer deallocates allocatable variables or 1407 allocatable components of variables declared in the main 1408 program. Since Fortran 2008, the standard explicitly states 1409 that variables declared in the Fortran main program 1410 automatically have the SAVE attribute. 1411 + When opening files, the close-on-exec flag is set if the 1412 system supports such a feature. This is generally considered 1413 good practice these days, but if there is a need to pass file 1414 descriptors to child processes the parent process must now 1415 remember to clear the close-on-exec flag by calling fcntl(), 1416 e.g. via ISO_C_BINDING, before executing the child process. 1417 * The deprecated command-line option -fno-whole-file has been 1418 removed. (-fwhole-file is the default since GCC 4.6.) 1419 -fwhole-file/-fno-whole-file continue to be accepted but do not 1420 influence the code generation. 1421 * The compiler no longer unconditionally warns about DO loops with 1422 zero iterations. This warning is now controlled by the -Wzerotrip 1423 option, which is implied by -Wall. 1424 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be 1425 used to disable the type-kind-rank (TKR) argument check for a dummy 1426 argument. The feature is similar to ISO/IEC TS 29133:2012's 1427 TYPE(*), except that it additionally also disables the rank check. 1428 Variables with NO_ARG_CHECK have to be dummy arguments and may only 1429 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 1430 to another NO_ARG_CHECK dummy argument; also the other constraints 1431 of TYPE(*) apply. The dummy arguments should be declared as scalar 1432 or assumed-size variable of type type(*) (recommended) - or of type 1433 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 1434 the data without further type or shape information is passed, 1435 similar to C's void*. Note that also TS 29113's 1436 type(*),dimension(..) accepts arguments of any type and rank; 1437 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 1438 descriptor which contains the array shape and stride of the 1439 argument. 1440 * [21]Fortran 2003: 1441 + Finalization is now supported. It is currently only done for a 1442 subset of those situations in which it should occur. 1443 + Experimental support for scalar character components with 1444 deferred length (i.e. allocatable string length) in derived 1445 types has been added. (Deferred-length character variables are 1446 supported since GCC 4.6.) 1447 * [22]Fortran 2008: 1448 + When STOP or ERROR STOP are used to terminate the execution 1449 and any exception (but inexact) is signaling, a warning is 1450 printed to ERROR_UNIT, indicating which exceptions are 1451 signaling. The [23]-ffpe-summary= command-line option can be 1452 used to fine-tune for which exceptions the warning should be 1453 shown. 1454 + Rounding on input (READ) is now handled on systems where 1455 strtod honours the rounding mode. (For output, rounding is 1456 supported since GCC 4.5.) Note that for input, the compatible 1457 rounding mode is handled as nearest (i.e., rounding to an even 1458 least significant [cf. IEC 60559:1989] for a tie, while 1459 compatible rounds away from zero in that case). 1460 1461 Go 1462 1463 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 1464 1465New Targets and Target Specific Improvements 1466 1467 AArch64 1468 1469 * The ARMv8-A crypto and CRC instructions are now supported through 1470 intrinsics. These are enabled when the architecture supports these 1471 and are available through the -march=armv8-a+crc and 1472 -march=armv8-a+crypto options. 1473 * Initial support for ILP32 has now been added to the compiler. This 1474 is now available through the command-line option -mabi=ilp32. 1475 Support for ILP32 is considered experimental as the ABI 1476 specification is still beta. 1477 * Coverage of more of the ISA including the SIMD extensions has been 1478 added. The Advanced SIMD intrinsics have also been improved. 1479 * The new local register allocator (LRA) is now on by default for the 1480 AArch64 backend. 1481 * The REE (Redundant extension elimination) pass has now been enabled 1482 by default for the AArch64 backend. 1483 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 1484 * Initial big.LITTLE tuning support for the combination of Cortex-A57 1485 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 1486 option. 1487 * A number of structural changes have been made to both the ARM and 1488 AArch64 backends to facilitate improved code-generation. 1489 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 1490 has been added and can be enabled by giving the 1491 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 1492 default by configuring GCC with the --enable-fix-cortex-a53-835769 1493 option. 1494 1495 ARC 1496 1497 * A port for Synopsys Designware ARC has been contributed by Embecosm 1498 and Synopsys Inc. 1499 1500 ARM 1501 1502 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 1503 disabled by default. This was found to generate better code in only 1504 a small number of cases. It can be turned back on with the 1505 -mneon-for-64bits option. 1506 * Further support for the ARMv8-A architecture, notably implementing 1507 the restriction around IT blocks in the Thumb32 instruction set has 1508 been added. The -mrestrict-it option can be used with 1509 -march=armv7-a or the -march=armv7ve options to make code 1510 generation fully compatible with the deprecated instructions in 1511 ARMv8-A. 1512 * Support has now been added for the ARMv7ve variant of the 1513 architecture. This can be used by the -march=armv7ve option. 1514 * The ARMv8-A crypto and CRC instructions are now supported through 1515 intrinsics and are available through the -march=armv8-a+crc and 1516 mfpu=crypto-neon-fp-armv8 options. 1517 * LRA is now on by default for the ARM target. This can be turned off 1518 using the -mno-lra option. This option is a purely transitionary 1519 command-line option and will be removed in a future release. We are 1520 interested in any bug reports regarding functional and performance 1521 regressions with LRA. 1522 * A new option -mslow-flash-data to improve performance of programs 1523 fetching data on slow flash memory has now been introduced for the 1524 ARMv7-M profile cores. 1525 * A new option -mpic-data-is-text-relative for targets that allows 1526 data segments to be relative to text segments has been added. This 1527 is on by default for all targets except VxWorks RTP. 1528 * A number of infrastructural changes have been made to both the ARM 1529 and AArch64 backends to facilitate improved code-generation. 1530 * GCC now supports Cortex-A12 and the Cortex-R7 through the 1531 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 1532 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 1533 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 1534 * Initial big.LITTLE tuning support for the combination of Cortex-A57 1535 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 1536 option. Similar support was added for the combination of Cortex-A15 1537 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 1538 * Further performance optimizations for the Cortex-A15 and the 1539 Cortex-M4 have been added. 1540 * A number of code generation improvements for Thumb2 to reduce code 1541 size when compiling for the M-profile processors. 1542 1543 IA-32/x86-64 1544 1545 * -mfpmath=sse is now implied by -ffast-math on all targets where 1546 SSE2 is supported. 1547 * Intel AVX-512 support was added to GCC. That includes inline 1548 assembly support, new registers and extending existing ones, new 1549 intrinsics (covered by corresponding testsuite), and basic 1550 autovectorization. AVX-512 instructions are available via the 1551 following GCC switches: AVX-512 foundation instructions: -mavx512f, 1552 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 1553 reciprocal instructions: -mavx512er, AVX-512 conflict detection 1554 instructions: -mavx512cd. 1555 * It is now possible to call x86 intrinsics from select functions in 1556 a file that are tagged with the corresponding target attribute 1557 without having to compile the entire file with the -mxxx option. 1558 This improves the usability of x86 intrinsics and is particularly 1559 useful when doing [24]Function Multiversioning. 1560 * GCC now supports the new Intel microarchitecture named Silvermont 1561 through -march=silvermont. 1562 * GCC now supports the new Intel microarchitecture named Broadwell 1563 through -march=broadwell. 1564 * Optimizing for other Intel microarchitectures have been renamed to 1565 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 1566 * -march=generic has been retuned for better support of Intel core 1567 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 1568 Pentium-M, and Pentium4 based CPUs is no longer considered 1569 important for generic. 1570 * -mtune=intel can now be used to generate code running well on the 1571 most current Intel processors, which are Haswell and Silvermont for 1572 GCC 4.9. 1573 * Support to encode 32-bit assembly instructions in 16-bit format is 1574 now available through the -m16 command-line option. 1575 * Better inlining of memcpy and memset that is aware of value ranges 1576 and produces shorter alignment prologues. 1577 * -mno-accumulate-outgoing-args is now honored when unwind 1578 information is output. Argument accumulation is also now turned off 1579 for portions of programs optimized for size. 1580 * Support for new AMD family 15h processors (Excavator core) is now 1581 available through the -march=bdver4 and -mtune=bdver4 options. 1582 1583 MSP430 1584 1585 * A new command-line option -mcpu= has been added to the MSP430 1586 backend. This option is used to specify the ISA to be used. 1587 Accepted values are msp430 (the default), msp430x and msp430xv2. 1588 The ISA is no longer deduced from the -mmcu= option as there are 1589 far too many different MCU names. The -mmcu= option is still 1590 supported, and this is still used to select linker scripts and 1591 generate a C preprocessor symbol that will be recognised by the 1592 msp430.h header file. 1593 1594 NDS32 1595 1596 * A new nds32 port supports the 32-bit architecture from Andes 1597 Technology Corporation. 1598 * The port provides initial support for the V2, V3, V3m instruction 1599 set architectures. 1600 1601 Nios II 1602 1603 * A port for the Altera Nios II has been contributed by Mentor 1604 Graphics. 1605 1606 PowerPC / PowerPC64 / RS6000 1607 1608 * GCC now supports Power ISA 2.07, which includes support for 1609 Hardware Transactional Memory (HTM), Quadword atomics and several 1610 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 1611 integer and decimal integer operations. 1612 * Support for the POWER8 processor is now available through the 1613 -mcpu=power8 and -mtune=power8 options. 1614 * The libitm library has been modified to add a HTM fastpath that 1615 automatically uses POWER's HTM hardware instructions when it is 1616 executing on a HTM enabled processor. 1617 * Support for the new powerpc64le-linux platform has been added. It 1618 defaults to generating code that conforms to the ELFV2 ABI. 1619 1620 S/390, System z 1621 1622 * Support for the Transactional Execution Facility included with the 1623 IBM zEnterprise zEC12 processor has been added. A set of GCC style 1624 builtins as well as XLC style builtins are provided. The builtins 1625 are enabled by default when using the -march=zEC12 option but can 1626 explicitly be disabled with -mno-htm. Using the GCC builtins also 1627 libitm supports hardware transactions on S/390. 1628 * The hotpatch features allows to prepare functions for hotpatching. 1629 A certain amount of bytes is reserved before the function entry 1630 label plus a NOP is inserted at its very beginning to implement a 1631 backward jump when applying a patch. The feature can either be 1632 enabled per compilation unit via the command-line option -mhotpatch 1633 or per function using the hotpatch attribute. 1634 * The shrink wrap optimization is now supported on S/390 and enabled 1635 by default. 1636 * A major rework of the routines to determine which registers need to 1637 be saved and restored in function prologue/epilogue now allow to 1638 use floating point registers as save slots. This will happen for 1639 certain leaf function with -march=z10 or higher. 1640 * The LRA rtl pass replaces reload by default on S/390. 1641 1642 RX 1643 1644 * The port now allows to specify the RX100, RX200, and RX600 1645 processors with the command-line options -mcpu=rx100, -mcpu=rx200 1646 and -mcpu=rx600. 1647 1648 SH 1649 1650 * Minor improvements to code generated for integer arithmetic and 1651 code that involves the T bit. 1652 * Added support for the SH2A clips and clipu instructions. The 1653 compiler will now try to utilize them for min/max expressions such 1654 as max (-128, min (127, x)). 1655 * Added support for the cmp/str instruction through built-in 1656 functions such as __builtin_strlen. When not optimizing for size, 1657 the compiler will now expand calls to e.g. strlen as an inlined 1658 sequences which utilize the cmp/str instruction. 1659 * Improved code generated around volatile memory loads and stores. 1660 * The option -mcbranchdi has been deprecated. Specifying it will 1661 result in a warning and will not influence code generation. 1662 * The option -mcmpeqdi has been deprecated. Specifying it will result 1663 in a warning and will not influence code generation. 1664 1665GCC 4.9.1 1666 1667 This is the [25]list of problem reports (PRs) from GCC's bug tracking 1668 system that are known to be fixed in the 4.9.1 release. This list might 1669 not be complete (that is, it is possible that some PRs that have been 1670 fixed are not listed here). 1671 1672 Version 4.0 of the [26]OpenMP specification is supported even in 1673 Fortran, not just C and C++. 1674 1675GCC 4.9.2 1676 1677 This is the [27]list of problem reports (PRs) from GCC's bug tracking 1678 system that are known to be fixed in the 4.9.2 release. This list might 1679 not be complete (that is, it is possible that some PRs that have been 1680 fixed are not listed here). 1681 1682GCC 4.9.3 1683 1684 This is the [28]list of problem reports (PRs) from GCC's bug tracking 1685 system that are known to be fixed in the 4.9.3 release. This list might 1686 not be complete (that is, it is possible that some PRs that have been 1687 fixed are not listed here). 1688 1689 1690 For questions related to the use of GCC, please consult these web 1691 pages and the [29]GCC manuals. If that fails, the 1692 [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these 1693 web pages and the development of GCC are welcome on our developer 1694 list at [31]gcc@gcc.gnu.org. All of [32]our lists have public 1695 archives. 1696 1697 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 1698 distribution of this entire article is permitted in any medium, 1699 provided this notice is preserved. 1700 1701 These pages are [34]maintained by the GCC team. Last modified 1702 2016-01-30[35]. 1703 1704References 1705 1706 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 1707 2. https://gcc.gnu.org/PR60825 1708 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 1709 4. http://openmp.org/wp/openmp-specifications/ 1710 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 1711 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 1712 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 1713 8. https://www.cilkplus.org/ 1714 9. http://gcc.gnu.org/projects/cxx1y.html 1715 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 1716 11. http://gcc.gnu.org/projects/cxx1y.html 1717 12. http://gcc.gnu.org/projects/cxx1y.html 1718 13. http://gcc.gnu.org/projects/cxx1y.html 1719 14. http://gcc.gnu.org/projects/cxx1y.html 1720 15. http://gcc.gnu.org/projects/cxx1y.html 1721 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 1722 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 1723 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 1724 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 1725 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 1726 21. https://gcc.gnu.org/wiki/Fortran2003Status 1727 22. https://gcc.gnu.org/wiki/Fortran2008Status 1728 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 1729 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 1730 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 1731 26. http://openmp.org/wp/openmp-specifications/ 1732 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 1733 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 1734 29. https://gcc.gnu.org/onlinedocs/ 1735 30. mailto:gcc-help@gcc.gnu.org 1736 31. mailto:gcc@gcc.gnu.org 1737 32. https://gcc.gnu.org/lists.html 1738 33. http://www.fsf.org/ 1739 34. https://gcc.gnu.org/about.html 1740 35. http://validator.w3.org/check/referer 1741====================================================================== 1742http://gcc.gnu.org/gcc-4.8/index.html 1743 1744 GCC 4.8 Release Series 1745 1746 June 23, 2015 1747 1748 The [1]GNU project and the GCC developers are pleased to announce the 1749 release of GCC 4.8.5. 1750 1751 This release is a bug-fix release, containing fixes for regressions in 1752 GCC 4.8.4 relative to previous releases of GCC. 1753 1754Release History 1755 1756 GCC 4.8.5 1757 June 23, 2015 ([2]changes, [3]documentation) 1758 1759 GCC 4.8.4 1760 December 19, 2014 ([4]changes, [5]documentation) 1761 1762 GCC 4.8.3 1763 May 22, 2014 ([6]changes, [7]documentation) 1764 1765 GCC 4.8.2 1766 October 16, 2013 ([8]changes, [9]documentation) 1767 1768 GCC 4.8.1 1769 May 31, 2013 ([10]changes, [11]documentation) 1770 1771 GCC 4.8.0 1772 March 22, 2013 ([12]changes, [13]documentation) 1773 1774References and Acknowledgements 1775 1776 GCC used to stand for the GNU C Compiler, but since the compiler 1777 supports several other languages aside from C, it now stands for the 1778 GNU Compiler Collection. 1779 1780 A list of [14]successful builds is updated as new information becomes 1781 available. 1782 1783 The GCC developers would like to thank the numerous people that have 1784 contributed new features, improvements, bug fixes, and other changes as 1785 well as test results to GCC. This [15]amazing group of volunteers is 1786 what makes GCC successful. 1787 1788 For additional information about GCC please refer to the [16]GCC 1789 project web site or contact the [17]GCC development mailing list. 1790 1791 To obtain GCC please use [18]our mirror sites or [19]our SVN server. 1792 1793 1794 For questions related to the use of GCC, please consult these web 1795 pages and the [20]GCC manuals. If that fails, the 1796 [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these 1797 web pages and the development of GCC are welcome on our developer 1798 list at [22]gcc@gcc.gnu.org. All of [23]our lists have public 1799 archives. 1800 1801 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 1802 distribution of this entire article is permitted in any medium, 1803 provided this notice is preserved. 1804 1805 These pages are [25]maintained by the GCC team. Last modified 1806 2016-01-30[26]. 1807 1808References 1809 1810 1. http://www.gnu.org/ 1811 2. http://gcc.gnu.org/gcc-4.8/changes.html 1812 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 1813 4. http://gcc.gnu.org/gcc-4.8/changes.html 1814 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 1815 6. http://gcc.gnu.org/gcc-4.8/changes.html 1816 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 1817 8. http://gcc.gnu.org/gcc-4.8/changes.html 1818 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 1819 10. http://gcc.gnu.org/gcc-4.8/changes.html 1820 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 1821 12. http://gcc.gnu.org/gcc-4.8/changes.html 1822 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 1823 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 1824 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1825 16. http://gcc.gnu.org/index.html 1826 17. mailto:gcc@gcc.gnu.org 1827 18. http://gcc.gnu.org/mirrors.html 1828 19. http://gcc.gnu.org/svn.html 1829 20. https://gcc.gnu.org/onlinedocs/ 1830 21. mailto:gcc-help@gcc.gnu.org 1831 22. mailto:gcc@gcc.gnu.org 1832 23. https://gcc.gnu.org/lists.html 1833 24. http://www.fsf.org/ 1834 25. https://gcc.gnu.org/about.html 1835 26. http://validator.w3.org/check/referer 1836====================================================================== 1837http://gcc.gnu.org/gcc-4.8/changes.html 1838 1839 GCC 4.8 Release Series 1840 Changes, New Features, and Fixes 1841 1842Caveats 1843 1844 GCC now uses C++ as its implementation language. This means that to 1845 build GCC from sources, you will need a C++ compiler that understands 1846 C++ 2003. For more details on the rationale and specific changes, 1847 please refer to the [1]C++ conversion page. 1848 1849 To enable the Graphite framework for loop optimizations you now need 1850 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 1851 the [2]GCC infrastructure directory. The installation manual contains 1852 more information about requirements to build GCC. 1853 1854 GCC now uses a more aggressive analysis to derive an upper bound for 1855 the number of iterations of loops using constraints imposed by language 1856 standards. This may cause non-conforming programs to no longer work as 1857 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 1858 option, -fno-aggressive-loop-optimizations, was added to disable this 1859 aggressive analysis. In some loops that have known constant number of 1860 iterations, but undefined behavior is known to occur in the loop before 1861 reaching or during the last iteration, GCC will warn about the 1862 undefined behavior in the loop instead of deriving lower upper bound of 1863 the number of iterations for the loop. The warning can be disabled with 1864 -Wno-aggressive-loop-optimizations. 1865 1866 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 1867 for the layout of vectors that could lead to wrong code being 1868 generated. Vectors larger than 8 bytes in size are now by default 1869 aligned to an 8-byte boundary. This is an ABI change: code that makes 1870 explicit use of vector types may be incompatible with binary objects 1871 built with older versions of GCC. Auto-vectorized code is not affected 1872 by this change. 1873 1874 On AVR, support has been removed for the command-line option 1875 -mshort-calls deprecated in GCC 4.7. 1876 1877 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 1878 is turned on per default for all non-RTEMS configurations. This option 1879 arranges for a better integration of [3]AVR Libc with avr-gcc. For 1880 technical details, see [4]PR54461. To turn off the option in non-RTEMS 1881 configurations, use --with-avrlibc=no. If the compiler is configured 1882 for RTEMS, the option is always turned off. 1883 1884 More information on porting to GCC 4.8 from previous versions of GCC 1885 can be found in the [5]porting guide for this release. 1886 1887General Optimizer Improvements (and Changes) 1888 1889 * DWARF4 is now the default when generating DWARF debug information. 1890 When -g is used on a platform that uses DWARF debugging 1891 information, GCC will now default to -gdwarf-4 1892 -fno-debug-types-section. 1893 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 1894 consumers support DWARF4 by default. Before GCC 4.8 the default 1895 version used was DWARF2. To make GCC 4.8 generate an older DWARF 1896 version use -g together with -gdwarf-2 or -gdwarf-3. The default 1897 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 1898 * A new general optimization level, -Og, has been introduced. It 1899 addresses the need for fast compilation and a superior debugging 1900 experience while providing a reasonable level of run-time 1901 performance. Overall experience for development should be better 1902 than the default optimization level -O0. 1903 * A new option -ftree-partial-pre was added to control the partial 1904 redundancy elimination (PRE) optimization. This option is enabled 1905 by default at the -O3 optimization level, and it makes PRE more 1906 aggressive. 1907 * The option -fconserve-space has been removed; it was no longer 1908 useful on most targets since GCC supports putting variables into 1909 BSS without making them common. 1910 * The struct reorg and matrix reorg optimizations (command-line 1911 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 1912 removed. They did not always work correctly, nor did they work with 1913 link-time optimization (LTO), hence were only applicable to 1914 programs consisting of a single translation unit. 1915 * Several scalability bottle-necks have been removed from GCC's 1916 optimization passes. Compilation of extremely large functions, e.g. 1917 due to the use of the flatten attribute in the "Eigen" C++ linear 1918 algebra templates library, is significantly faster than previous 1919 releases of GCC. 1920 * Link-time optimization (LTO) improvements: 1921 + LTO partitioning has been rewritten for better reliability and 1922 maintanibility. Several important bugs leading to link 1923 failures have been fixed. 1924 * Interprocedural optimization improvements: 1925 + A new symbol table has been implemented. It builds on existing 1926 callgraph and varpool modules and provide a new API. Unusual 1927 symbol visibilities and aliases are handled more consistently 1928 leading to, for example, more aggressive unreachable code 1929 removal with LTO. 1930 + The inline heuristic can now bypass limits on the size of of 1931 inlined functions when the inlining is particularly 1932 profitable. This happens, for example, when loop bounds or 1933 array strides get propagated. 1934 + Values passed through aggregates (either by value or 1935 reference) are now propagated at the inter-procedural level 1936 leading to better inlining decisions (for example in the case 1937 of Fortran array descriptors) and devirtualization. 1938 * [6]AddressSanitizer , a fast memory error detector, has been added 1939 and can be enabled via -fsanitize=address. Memory access 1940 instructions will be instrumented to detect heap-, stack-, and 1941 global-buffer overflow as well as use-after-free bugs. To get nicer 1942 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 1943 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 1944 x86-64 Darwin. 1945 * [7]ThreadSanitizer has been added and can be enabled via 1946 -fsanitize=thread. Instructions will be instrumented to detect data 1947 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 1948 * A new local register allocator (LRA) has been implemented, which 1949 replaces the 26 year old reload pass and improves generated code 1950 quality. For now it is active on the IA-32 and x86-64 targets. 1951 * Support for transactional memory has been implemented on the 1952 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 1953 Alpha. 1954 1955New Languages and Language specific improvements 1956 1957 C family 1958 1959 * Each diagnostic emitted now includes the original source line and a 1960 caret '^' indicating the column. The option 1961 -fno-diagnostics-show-caret suppresses this information. 1962 * The option -ftrack-macro-expansion=2 is now enabled by default. 1963 This allows the compiler to display the macro expansion stack in 1964 diagnostics. Combined with the caret information, an example 1965 diagnostic showing these two features is: 1966 1967t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float 1968') 1969 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 1970_b = (B); __a < __b ? __b : __a; }) 1971 1972 ^ 1973t.c:7:7: note: in expansion of macro 'MYMAX' 1974 X = MYMAX(P, F); 1975 ^ 1976 1977 * A new -Wsizeof-pointer-memaccess warning has been added (also 1978 enabled by -Wall) to warn about suspicious length parameters to 1979 certain string and memory built-in functions if the argument uses 1980 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 1981 (ptr)); if ptr is not an array, but a pointer, and suggests a 1982 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 1983 * The new option -Wpedantic is an alias for -pedantic, which is now 1984 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 1985 -Wno-error=pedantic work in the same way as for any other -W 1986 option. One caveat is that -Werror=pedantic is not equivalent to 1987 -pedantic-errors, since the latter makes into errors some warnings 1988 that are not controlled by -Wpedantic, and the former only affects 1989 diagnostics that are disabled when using -Wno-pedantic. 1990 * The option -Wshadow no longer warns if a declaration shadows a 1991 function declaration, unless the former declares a function or 1992 pointer to function, because this is [8]a common and valid case in 1993 real-world code. 1994 1995 C++ 1996 1997 * G++ now implements the [9]C++11 thread_local keyword; this differs 1998 from the GNU __thread keyword primarily in that it allows dynamic 1999 initialization and destruction semantics. Unfortunately, this 2000 support requires a run-time penalty for references to 2001 non-function-local thread_local variables defined in a different 2002 translation unit even if they don't need dynamic initialization, so 2003 users may want to continue to use __thread for TLS variables with 2004 static initialization semantics. 2005 If the programmer can be sure that no use of the variable in a 2006 non-defining TU needs to trigger dynamic initialization (either 2007 because the variable is statically initialized, or a use of the 2008 variable in the defining TU will be executed before any uses in 2009 another TU), they can avoid this overhead with the 2010 -fno-extern-tls-init option. 2011 OpenMP threadprivate variables now also support dynamic 2012 initialization and destruction by the same mechanism. 2013 * G++ now implements the [10]C++11 attribute syntax, e.g. 2014 2015[[noreturn]] void f(); 2016 2017 and also the alignment specifier, e.g. 2018 2019alignas(double) int i; 2020 2021 * G++ now implements [11]C++11 inheriting constructors, e.g. 2022 2023struct A { A(int); }; 2024struct B: A { using A::A; }; // defines B::B(int) 2025B b(42); // OK 2026 2027 * As of GCC 4.8.1, G++ implements the change to decltype semantics 2028 from [12]N3276. 2029 2030struct A f(); 2031decltype(f()) g(); // OK, return type of f() is not required to be complete. 2032 2033 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 2034 2035struct A { int f() &; }; 2036int i = A().f(); // error, f() requires an lvalue object 2037 2038 * G++ now supports a -std=c++1y option for experimentation with 2039 features proposed for the next revision of the standard, expected 2040 around 2014. Currently the only difference from -std=c++11 is 2041 support for return type deduction in normal functions, as proposed 2042 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 2043 [15]here. 2044 * The G++ namespace association extension, __attribute ((strong)), 2045 has been deprecated. Inline namespaces should be used instead. 2046 * G++ now supports a -fext-numeric-literal option to control whether 2047 GNU numeric literal suffixes are accepted as extensions or 2048 processed as C++11 user-defined numeric literal suffixes. The flag 2049 is on (use suffixes for GNU literals) by default for -std=gnu++*, 2050 and -std=c++98. The flag is off (use suffixes for user-defined 2051 literals) by default for -std=c++11 and later. 2052 2053 Runtime Library (libstdc++) 2054 2055 * [16]Improved experimental support for the new ISO C++ standard, 2056 C++11, including: 2057 + forward_list meets the allocator-aware container requirements; 2058 + this_thread::sleep_for(), this_thread::sleep_until() and 2059 this_thread::yield() are defined without requiring the 2060 configure option --enable-libstdcxx-time; 2061 * Improvements to <random>: 2062 + SSE optimized normal_distribution. 2063 + Use of hardware RNG instruction for random_device on new x86 2064 processors (requires the assembler to support the 2065 instruction.) 2066 and <ext/random>: 2067 + New random number engine simd_fast_mersenne_twister_engine 2068 with an optimized SSE implementation. 2069 + New random number distributions beta_distribution, 2070 normal_mv_distribution, rice_distribution, 2071 nakagami_distribution, pareto_distribution, k_distribution, 2072 arcsine_distribution, hoyt_distribution. 2073 * Added --disable-libstdcxx-verbose configure option to disable 2074 diagnostic messages issued when a process terminates abnormally. 2075 This may be useful for embedded systems to reduce the size of 2076 executables that link statically to the library. 2077 2078 Fortran 2079 2080 * Compatibility notice: 2081 + Module files: The version of module files (.mod) has been 2082 incremented. Fortran MODULEs compiled by earlier GCC versions 2083 have to be recompiled, when they are USEd by files compiled 2084 with GCC 4.8. GCC 4.8 is not able to read .mod files created 2085 by earlier versions; attempting to do so gives an error 2086 message. 2087 Note: The ABI of the produced assembler data itself has not 2088 changed; object files and libraries are fully compatible with 2089 older versions except as noted below. 2090 + ABI: Some internal names (used in the assembler/object file) 2091 have changed for symbols declared in the specification part of 2092 a module. If an affected module - or a file using it via use 2093 association - is recompiled, the module and all files which 2094 directly use such symbols have to be recompiled as well. This 2095 change only affects the following kind of module symbols: 2096 o Procedure pointers. Note: C-interoperable function 2097 pointers (type(c_funptr)) are not affected nor are 2098 procedure-pointer components. 2099 o Deferred-length character strings. 2100 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 2101 backtrace at an arbitrary place in user code; program execution 2102 continues normally afterwards. 2103 * The [18]-Wc-binding-type warning option has been added (disabled by 2104 default). It warns if the a variable might not be C interoperable; 2105 in particular, if the variable has been declared using an intrinsic 2106 type with default kind instead of using a kind parameter defined 2107 for C interoperability in the intrinsic ISO_C_Binding module. 2108 Before, this warning was always printed. The -Wc-binding-type 2109 option is enabled by -Wall. 2110 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 2111 options have been added, which diagnose when code is inserted for 2112 automatic (re)allocation of a variable during assignment. This 2113 option can be used to decide whether it is safe to use 2114 [20]-fno-realloc-lhs. Additionally, it can be used to find 2115 automatic (re)allocation in hot loops. (For arrays, replacing 2116 "var=" by "var(:)=" disables the automatic reallocation.) 2117 * The [21]-Wcompare-reals command-line option has been added. When 2118 this is set, warnings are issued when comparing REAL or COMPLEX 2119 types for equality and inequality; consider replacing a == b by 2120 abs(a -b) < eps with a suitable eps. -Wcompare-reals is enabled by 2121 -Wextra. 2122 * The [22]-Wtarget-lifetime command-line option has been added 2123 (enabled with -Wall), which warns if the pointer in a pointer 2124 assignment might outlive its target. 2125 * Reading floating point numbers which use "q" for the exponential 2126 (such as 4.0q0) is now supported as vendor extension for better 2127 compatibility with old data files. It is strongly recommended to 2128 use for I/O the equivalent but standard conforming "e" (such as 2129 4.0e0). 2130 (For Fortran source code, consider replacing the "q" in 2131 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 2132 suitable qp). Note that - in Fortran source code - replacing "q" by 2133 a simple "e" is not equivalent.) 2134 * The GFORTRAN_TMPDIR environment variable for specifying a 2135 non-default directory for files opened with STATUS="SCRATCH", is 2136 not used anymore. Instead gfortran checks the POSIX/GNU standard 2137 TMPDIR environment variable. If TMPDIR is not defined, gfortran 2138 falls back to other methods to determine the directory for 2139 temporary files as documented in the [23]user manual. 2140 * [24]Fortran 2003: 2141 + Support for unlimited polymorphic variables (CLASS(*)) has 2142 been added. Nonconstant character lengths are not yet 2143 supported. 2144 * [25]TS 29113: 2145 + Assumed types (TYPE(*)) are now supported. 2146 + Experimental support for assumed-rank arrays (dimension(..)) 2147 has been added. Note that currently gfortran's own array 2148 descriptor is used, which is different from the one defined in 2149 TS29113, see [26]gfortran's header file or use the [27]Chasm 2150 Language Interoperability Tools. 2151 2152 Go 2153 2154 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 2155 release. 2156 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 2157 release. The library support is not quite complete. 2158 * Go has been tested on GNU/Linux and Solaris platforms for various 2159 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 2160 work on other platforms as well. 2161 2162New Targets and Target Specific Improvements 2163 2164 AArch64 2165 2166 * A new port has been added to support AArch64, the new 64-bit 2167 architecture from ARM. Note that this is a separate port from the 2168 existing 32-bit ARM port. 2169 * The port provides initial support for the Cortex-A53 and the 2170 Cortex-A57 processors with the command line options 2171 -mcpu=cortex-a53 and -mcpu=cortex-a57. 2172 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 2173 has been added and can be enabled by giving the 2174 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 2175 default by configuring GCC with the --enable-fix-cortex-a53-835769 2176 option. 2177 2178 ARM 2179 2180 * Initial support has been added for the AArch32 extensions defined 2181 in the ARMv8 architecture. 2182 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 2183 * A new option, -mcpu=marvell-pj4, has been added to generate code 2184 for the Marvell PJ4 processor. 2185 * The compiler can now automatically generate the VFMA, VFMS, REVSH 2186 and REV16 instructions. 2187 * A new vectorizer cost model for Advanced SIMD configurations to 2188 improve the auto-vectorization strategies used. 2189 * The scheduler now takes into account the number of live registers 2190 to reduce the amount of spilling that can occur. This should 2191 improve code performance in large functions. The limit can be 2192 removed by using the option -fno-sched-pressure. 2193 * Improvements have been made to the Marvell iWMMX code generation 2194 and support for the iWMMX2 SIMD unit has been added. The option 2195 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 2196 * A number of code generation improvements for Thumb2 to reduce code 2197 size when compiling for the M-profile processors. 2198 * The RTEMS (arm-rtems) port has been updated to use the EABI. 2199 * Code generation support for the old FPA and Maverick floating-point 2200 architectures has been removed. Ports that previously relied on 2201 these features have also been removed. This includes the targets: 2202 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 2203 + arm*-*-elf (use arm*-*-eabi) 2204 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 2205 + arm*-*-ecos-elf (no alternative) 2206 + arm*-*-freebsd (no alternative) 2207 + arm*-wince-pe* (no alternative). 2208 2209 AVR 2210 2211 * Support for the "Embedded C" fixed-point has been added. For 2212 details, see the [28]GCC wiki and the [29]user manual. The support 2213 is not complete. 2214 * A new print modifier %r for register operands in inline assembler 2215 is supported. It will print the raw register number without the 2216 register prefix 'r': 2217 /* Return the most significant byte of 'val', a 64-bit value. */ 2218 2219 unsigned char msb (long long val) 2220 { 2221 unsigned char c; 2222 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 2223 return c; 2224 } 2225 The inline assembler in this example will generate code like 2226 mov r24, 8+7 2227 provided c is allocated to R24 and val is allocated to R8...R15. 2228 This works because the GNU assembler accepts plain register numbers 2229 without register prefix. 2230 * Static initializers with 3-byte symbols are supported now: 2231 extern const __memx char foo; 2232 const __memx void *pfoo = &foo; 2233 This requires at least Binutils 2.23. 2234 2235 IA-32/x86-64 2236 2237 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 2238 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 2239 stack alignment, this is ABI incompatible and intended to be used 2240 in controlled environments where stack space is an important 2241 limitation. This option will lead to wrong code when functions 2242 compiled with 16 byte stack alignment (such as functions from a 2243 standard library) are called with misaligned stack. In this case, 2244 SSE instructions may lead to misaligned memory access traps. In 2245 addition, variable arguments will be handled incorrectly for 16 2246 byte aligned objects (including x87 long double and __int128), 2247 leading to wrong results. You must build all modules with 2248 -mpreferred-stack-boundary=3, including any libraries. This 2249 includes the system libraries and startup modules. 2250 * Support for the new Intel processor codename Broadwell with RDSEED, 2251 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 2252 -mrdseed command-line options. 2253 * Support for the Intel RTM and HLE intrinsics, built-in functions 2254 and code generation is available via -mrtm and -mhle. 2255 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 2256 Intrinsics and built-in functions are available via -mfxsr, -mxsave 2257 and -mxsaveopt respectively. 2258 * New -maddress-mode=[short|long] options for x32. 2259 -maddress-mode=short overrides default 64-bit addresses to 32-bit 2260 by emitting the 0x67 address-size override prefix. This is the 2261 default address mode for x32. 2262 * New built-in functions to detect run-time CPU type and ISA: 2263 + A built-in function __builtin_cpu_is has been added to detect 2264 if the run-time CPU is of a particular type. It returns a 2265 positive integer on a match and zero otherwise. It accepts one 2266 string literal argument, the CPU name. For example, 2267 __builtin_cpu_is("westmere") returns a positive integer if the 2268 run-time CPU is an Intel Core i7 Westmere processor. Please 2269 refer to the [30]user manual for the list of valid CPU names 2270 recognized. 2271 + A built-in function __builtin_cpu_supports has been added to 2272 detect if the run-time CPU supports a particular ISA feature. 2273 It returns a positive integer on a match and zero otherwise. 2274 It accepts one string literal argument, the ISA feature. For 2275 example, __builtin_cpu_supports("ssse3") returns a positive 2276 integer if the run-time CPU supports SSSE3 instructions. 2277 Please refer to the [31]user manual for the list of valid ISA 2278 names recognized. 2279 Caveat: If these built-in functions are called before any static 2280 constructors are invoked, like during IFUNC initialization, then 2281 the CPU detection initialization must be explicitly run using this 2282 newly provided built-in function, __builtin_cpu_init. The 2283 initialization needs to be done only once. For example, this is how 2284 the invocation would look like inside an IFUNC initializer: 2285 static void (*some_ifunc_resolver(void))(void) 2286 { 2287 __builtin_cpu_init(); 2288 if (__builtin_cpu_is("amdfam10h") ... 2289 if (__builtin_cpu_supports("popcnt") ... 2290 } 2291 2292 * Function Multiversioning Support with G++: 2293 It is now possible to create multiple function versions each 2294 targeting a specific processor and/or ISA. Function versions have 2295 the same signature but different target attributes. For example, 2296 here is a program with function versions: 2297 __attribute__ ((target ("default"))) 2298 int foo(void) 2299 { 2300 return 1; 2301 } 2302 2303 __attribute__ ((target ("sse4.2"))) 2304 int foo(void) 2305 { 2306 return 2; 2307 } 2308 2309 int main (void) 2310 { 2311 int (*p) = &foo; 2312 assert ((*p)() == foo()); 2313 return 0; 2314 } 2315 2316 Please refer to this [32]wiki for more information. 2317 * The x86 back end has been improved to allow option -fschedule-insns 2318 to work reliably. This option can be used to schedule instructions 2319 better and leads to improved performace in certain cases. 2320 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 2321 from the Mingw-w64 trunk. 2322 * Support for new AMD family 15h processors (Steamroller core) is now 2323 available through the -march=bdver3 and -mtune=bdver3 options. 2324 * Support for new AMD family 16h processors (Jaguar core) is now 2325 available through the -march=btver2 and -mtune=btver2 options. 2326 2327 FRV 2328 2329 * This target now supports the -fstack-usage command-line option. 2330 2331 MIPS 2332 2333 * GCC can now generate code specifically for the R4700, Broadcom XLP 2334 and MIPS 34kn processors. The associated -march options are 2335 -march=r4700, -march=xlp and -march=34kn respectively. 2336 * GCC now generates better DSP code for MIPS 74k cores thanks to 2337 further scheduling optimizations. 2338 * The MIPS port now supports the -fstack-check option. 2339 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 2340 * Previous versions of GCC would silently accept -fpic and -fPIC for 2341 -mno-abicalls targets like mips*-elf. This combination was not 2342 intended or supported, and did not generate position-independent 2343 code. GCC 4.8 now reports an error when this combination is used. 2344 2345 PowerPC / PowerPC64 / RS6000 2346 2347 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 2348 restore or update the VRSAVE register by default. The respective 2349 operating systems manage the VRSAVE register directly. 2350 * Large TOC support has been added for AIX through the command line 2351 option -mcmodel=large. 2352 * Native Thread-Local Storage support has been added for AIX. 2353 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 2354 when targetting processors that support those hardware features on 2355 AIX 6.1 and above. 2356 2357 RX 2358 2359 * This target will now issue a warning message whenever multiple fast 2360 interrupt handlers are found in the same compilation unit. This 2361 feature can be turned off by the new 2362 -mno-warn-multiple-fast-interrupts command-line option. 2363 2364 S/390, System z 2365 2366 * Support for the IBM zEnterprise zEC12 processor has been added. 2367 When using the -march=zEC12 option, the compiler will generate code 2368 making use of the following new instructions: 2369 + load and trap instructions 2370 + 2 new compare and trap instructions 2371 + rotate and insert selected bits - without CC clobber 2372 The -mtune=zEC12 option enables zEC12 specific instruction 2373 scheduling without making use of new instructions. 2374 * Register pressure sensitive instruction scheduling is enabled by 2375 default. 2376 * The ifunc function attribute is enabled by default. 2377 * memcpy and memcmp invokations on big memory chunks or with run time 2378 lengths are not generated inline anymore when tuning for z10 or 2379 higher. The purpose is to make use of the IFUNC optimized versions 2380 in Glibc. 2381 2382 SH 2383 2384 * The default alignment settings have been reduced to be less 2385 aggressive. This results in more compact code for optimization 2386 levels other than -Os. 2387 * Improved support for the __atomic built-in functions: 2388 + A new option -matomic-model=model selects the model for the 2389 generated atomic sequences. The following models are 2390 supported: 2391 2392 soft-gusa 2393 Software gUSA sequences (SH3* and SH4* only). On 2394 SH4A targets this will now also partially utilize 2395 the movco.l and movli.l instructions. This is the 2396 default when the target is sh3*-*-linux* or 2397 sh4*-*-linux*. 2398 2399 hard-llcs 2400 Hardware movco.l / movli.l sequences (SH4A only). 2401 2402 soft-tcb 2403 Software thread control block sequences. 2404 2405 soft-imask 2406 Software interrupt flipping sequences (privileged 2407 mode only). This is the default when the target is 2408 sh1*-*-linux* or sh2*-*-linux*. 2409 2410 none 2411 Generates function calls to the respective __atomic 2412 built-in functions. This is the default for SH64 2413 targets or when the target is not sh*-*-linux*. 2414 2415 + The option -msoft-atomic has been deprecated. It is now an 2416 alias for -matomic-model=soft-gusa. 2417 + A new option -mtas makes the compiler generate the tas.b 2418 instruction for the __atomic_test_and_set built-in function 2419 regardless of the selected atomic model. 2420 + The __sync functions in libgcc now reflect the selected atomic 2421 model when building the toolchain. 2422 * Added support for the mov.b and mov.w instructions with 2423 displacement addressing. 2424 * Added support for the SH2A instructions movu.b and movu.w. 2425 * Various improvements to code generated for integer arithmetic. 2426 * Improvements to conditional branches and code that involves the T 2427 bit. A new option -mzdcbranch tells the compiler to favor 2428 zero-displacement branches. This is enabled by default for SH4* 2429 targets. 2430 * The pref instruction will now be emitted by the __builtin_prefetch 2431 built-in function for SH3* targets. 2432 * The fmac instruction will now be emitted by the fmaf standard 2433 function and the __builtin_fmaf built-in function. 2434 * The -mfused-madd option has been deprecated in favor of the 2435 machine-independent -ffp-contract option. Notice that the fmac 2436 instruction will now be generated by default for expressions like a 2437 * b + c. This is due to the compiler default setting 2438 -ffp-contract=fast. 2439 * Added new options -mfsrra and -mfsca to allow the compiler using 2440 the fsrra and fsca instructions on targets other than SH4A (where 2441 they are already enabled by default). 2442 * Added support for the __builtin_bswap32 built-in function. It is 2443 now expanded as a sequence of swap.b and swap.w instructions 2444 instead of a library function call. 2445 * The behavior of the -mieee option has been fixed and the negative 2446 form -mno-ieee has been added to control the IEEE conformance of 2447 floating point comparisons. By default -mieee is now enabled and 2448 the option -ffinite-math-only implicitly sets -mno-ieee. 2449 * Added support for the built-in functions __builtin_thread_pointer 2450 and __builtin_set_thread_pointer. This assumes that GBR is used to 2451 hold the thread pointer of the current thread. Memory loads and 2452 stores relative to the address returned by __builtin_thread_pointer 2453 will now also utilize GBR based displacement address modes. 2454 * The -mdiv= option for targets other than SHmedia has been fixed and 2455 documented. 2456 2457 SPARC 2458 2459 * Added optimized instruction scheduling for Niagara4. 2460 2461 TILE-Gx 2462 2463 * Added support for the -mcmodel=MODEL command-line option. The 2464 models supported are small and large. 2465 2466 V850 2467 2468 * This target now supports the E3V5 architecture via the use of the 2469 new -mv850e3v5 command-line option. It also has experimental 2470 support for the e3v5 LOOP instruction which can be enabled via the 2471 new -mloop command-line option. 2472 2473 XStormy16 2474 2475 * This target now supports the -fstack-usage command-line option. 2476 2477Operating Systems 2478 2479 Windows (Cygwin) 2480 2481 * Executables are now linked against shared libgcc by default. The 2482 previous default was to link statically, which can still be done by 2483 explicitly specifying -static or static-libgcc on the command line. 2484 However it is strongly advised against, as it will cause problems 2485 for any application that makes use of DLLs compiled by GCC. It 2486 should be alright for a monolithic stand-alone application that 2487 only links against the Windows DLLs, but offers little or no 2488 benefit. 2489 2490GCC 4.8.1 2491 2492 This is the [33]list of problem reports (PRs) from GCC's bug tracking 2493 system that are known to be fixed in the 4.8.1 release. This list might 2494 not be complete (that is, it is possible that some PRs that have been 2495 fixed are not listed here). 2496 2497 The C++11 <chrono> std::chrono::system_clock and 2498 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 2499 both are now separate (never typedefs of each other), both use 2500 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 2501 std::chrono::steady_clock is now finally monotonic, and both classes 2502 are mangled differently than in the previous GCC releases. 2503 std::chrono::system_clock::now() with std::chrono::microseconds resp. 2504 std::chrono::seconds resolution is still exported for backwards 2505 compatibility with default configured libstdc++. Note that libstdc++ 2506 configured with --enable-libstdcxx-time= used to be ABI incompatible 2507 with default configured libstdc++ for those two classes and no ABI 2508 compatibility can be offered for those configurations, so any C++11 2509 code that uses those classes and has been compiled and linked against 2510 libstdc++ configured with the non-default --enable-libstdcxx-time= 2511 configuration option needs to be recompiled. 2512 2513GCC 4.8.2 2514 2515 This is the [34]list of problem reports (PRs) from GCC's bug tracking 2516 system that are known to be fixed in the 4.8.2 release. This list might 2517 not be complete (that is, it is possible that some PRs that have been 2518 fixed are not listed here). 2519 2520GCC 4.8.3 2521 2522 This is the [35]list of problem reports (PRs) from GCC's bug tracking 2523 system that are known to be fixed in the 4.8.3 release. This list might 2524 not be complete (that is, it is possible that some PRs that have been 2525 fixed are not listed here). 2526 2527 Support for the new powerpc64le-linux platform has been added. It 2528 defaults to generating code that conforms to the ELFV2 ABI. 2529 2530GCC 4.8.4 2531 2532 This is the [36]list of problem reports (PRs) from GCC's bug tracking 2533 system that are known to be fixed in the 4.8.4 release. This list might 2534 not be complete (that is, it is possible that some PRs that have been 2535 fixed are not listed here). 2536 2537GCC 4.8.5 2538 2539 This is the [37]list of problem reports (PRs) from GCC's bug tracking 2540 system that are known to be fixed in the 4.8.5 release. This list might 2541 not be complete (that is, it is possible that some PRs that have been 2542 fixed are not listed here). 2543 2544 2545 For questions related to the use of GCC, please consult these web 2546 pages and the [38]GCC manuals. If that fails, the 2547 [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these 2548 web pages and the development of GCC are welcome on our developer 2549 list at [40]gcc@gcc.gnu.org. All of [41]our lists have public 2550 archives. 2551 2552 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 2553 distribution of this entire article is permitted in any medium, 2554 provided this notice is preserved. 2555 2556 These pages are [43]maintained by the GCC team. Last modified 2557 2016-05-28[44]. 2558 2559References 2560 2561 1. https://gcc.gnu.org/wiki/cxx-conversion 2562 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 2563 3. http://www.nongnu.org/avr-libc/ 2564 4. https://gcc.gnu.org/PR54461 2565 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 2566 6. https://github.com/google/sanitizers 2567 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 2568 8. https://lkml.org/lkml/2006/11/28/239 2569 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2570 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2571 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2572 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 2573 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2574 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 2575 15. http://gcc.gnu.org/projects/cxx1y.html 2576 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 2577 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 2578 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2579 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2580 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 2581 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2582 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2583 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 2584 24. https://gcc.gnu.org/wiki/Fortran2003Status 2585 25. https://gcc.gnu.org/wiki/TS29113Status 2586 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text%2Fplain&view=co 2587 27. http://chasm-interop.sourceforge.net/ 2588 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 2589 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 2590 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 2591 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 2592 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 2593 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 2594 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 2595 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 2596 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 2597 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 2598 38. https://gcc.gnu.org/onlinedocs/ 2599 39. mailto:gcc-help@gcc.gnu.org 2600 40. mailto:gcc@gcc.gnu.org 2601 41. https://gcc.gnu.org/lists.html 2602 42. http://www.fsf.org/ 2603 43. https://gcc.gnu.org/about.html 2604 44. http://validator.w3.org/check/referer 2605====================================================================== 2606http://gcc.gnu.org/gcc-4.7/index.html 2607 2608 GCC 4.7 Release Series 2609 2610 June 12, 2014 2611 2612 The [1]GNU project and the GCC developers are pleased to announce the 2613 release of GCC 4.7.4. 2614 2615 This release is a bug-fix release, containing fixes for regressions in 2616 GCC 4.7.3 relative to previous releases of GCC. 2617 2618Release History 2619 2620 GCC 4.7.4 2621 June 12, 2014 ([2]changes, [3]documentation) 2622 2623 GCC 4.7.3 2624 April 11, 2013 ([4]changes, [5]documentation) 2625 2626 GCC 4.7.2 2627 September 20, 2012 ([6]changes, [7]documentation) 2628 2629 GCC 4.7.1 2630 June 14, 2012 ([8]changes, [9]documentation) 2631 2632 GCC 4.7.0 2633 March 22, 2012 ([10]changes, [11]documentation) 2634 2635References and Acknowledgements 2636 2637 GCC used to stand for the GNU C Compiler, but since the compiler 2638 supports several other languages aside from C, it now stands for the 2639 GNU Compiler Collection. 2640 2641 A list of [12]successful builds is updated as new information becomes 2642 available. 2643 2644 The GCC developers would like to thank the numerous people that have 2645 contributed new features, improvements, bug fixes, and other changes as 2646 well as test results to GCC. This [13]amazing group of volunteers is 2647 what makes GCC successful. 2648 2649 For additional information about GCC please refer to the [14]GCC 2650 project web site or contact the [15]GCC development mailing list. 2651 2652 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 2653 2654 2655 For questions related to the use of GCC, please consult these web 2656 pages and the [18]GCC manuals. If that fails, the 2657 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 2658 web pages and the development of GCC are welcome on our developer 2659 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 2660 archives. 2661 2662 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 2663 distribution of this entire article is permitted in any medium, 2664 provided this notice is preserved. 2665 2666 These pages are [23]maintained by the GCC team. Last modified 2667 2016-01-30[24]. 2668 2669References 2670 2671 1. http://www.gnu.org/ 2672 2. http://gcc.gnu.org/gcc-4.7/changes.html 2673 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 2674 4. http://gcc.gnu.org/gcc-4.7/changes.html 2675 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 2676 6. http://gcc.gnu.org/gcc-4.7/changes.html 2677 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 2678 8. http://gcc.gnu.org/gcc-4.7/changes.html 2679 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 2680 10. http://gcc.gnu.org/gcc-4.7/changes.html 2681 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 2682 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 2683 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 2684 14. http://gcc.gnu.org/index.html 2685 15. mailto:gcc@gcc.gnu.org 2686 16. http://gcc.gnu.org/mirrors.html 2687 17. http://gcc.gnu.org/svn.html 2688 18. https://gcc.gnu.org/onlinedocs/ 2689 19. mailto:gcc-help@gcc.gnu.org 2690 20. mailto:gcc@gcc.gnu.org 2691 21. https://gcc.gnu.org/lists.html 2692 22. http://www.fsf.org/ 2693 23. https://gcc.gnu.org/about.html 2694 24. http://validator.w3.org/check/referer 2695====================================================================== 2696http://gcc.gnu.org/gcc-4.7/changes.html 2697 2698 GCC 4.7 Release Series 2699 Changes, New Features, and Fixes 2700 2701Caveats 2702 2703 * The -fconserve-space flag has been deprecated. The flag had no 2704 effect for most targets: only targets without a global .bss section 2705 and without support for switchable sections. Furthermore, the flag 2706 only had an effect for G++, where it could result in wrong 2707 semantics (please refer to the GCC manual for further details). The 2708 flag will be removed in GCC 4.8 2709 * Support for a number of older systems and recently unmaintained or 2710 untested target ports of GCC has been declared obsolete in GCC 4.7. 2711 Unless there is activity to revive them, the next release of GCC 2712 will have their sources permanently removed. 2713 All GCC ports for the following processor architectures have been 2714 declared obsolete: 2715 + picoChip (picochip-*) 2716 The following ports for individual systems on particular 2717 architectures have been obsoleted: 2718 + IRIX 6.5 (mips-sgi-irix6.5) 2719 + MIPS OpenBSD (mips*-*-openbsd*) 2720 + Solaris 8 (*-*-solaris2.8). Details can be found in the 2721 [1]announcement. 2722 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 2723 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 2724 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 2725 default, which for some sources generates code that accesses memory 2726 on unaligned addresses. This requires the kernel of those systems 2727 to enable such accesses (controlled by CP15 register c1, refer to 2728 ARM documentation). Alternatively, or for compatibility with 2729 kernels where unaligned accesses are not supported, all code has to 2730 be compiled with -mno-unaligned-access. Upstream Linux kernel 2731 releases have automatically and unconditionally supported unaligned 2732 accesses as emitted by GCC due to this option being active since 2733 version 2.6.28. 2734 * Support on ARM for the legacy floating-point accelerator (FPA) and 2735 the mixed-endian floating-point format that it used has been 2736 obsoleted. The ports that still use this format have been obsoleted 2737 as well. Many legacy ARM ports already provide an alternative that 2738 uses the VFP floating-point format. The obsolete ports will be 2739 deleted in the next release. 2740 The obsolete ports with alternatives are: 2741 + arm*-*-rtems (use arm*-*-rtemseabi) 2742 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 2743 + arm*-*-elf (use arm*-*-eabi) 2744 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 2745 Note, however, that these alternatives are not binary compatible 2746 with their legacy counterparts (although some can support running 2747 legacy applications). 2748 The obsolete ports that currently lack a modern alternative are: 2749 + arm*-*-ecos-elf 2750 + arm*-*-freebsd 2751 + arm*-wince-pe* 2752 New ports that support more recent versions of the architecture are 2753 welcome. 2754 * Support for the Maverick co-processor on ARM has been obsoleted. 2755 Code to support it will be deleted in the next release. 2756 * Support has been removed for Unix International threads on Solaris 2757 2, so the --enable-threads=solaris configure option and the 2758 -threads compiler option don't work any longer. 2759 * Support has been removed for the Solaris BSD Compatibility Package, 2760 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 2761 from Solaris 11, and was only intended as a migration aid from 2762 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 2763 recognized any longer. 2764 * The AVR port's libgcc has been improved and its multilib structure 2765 has been enhanced. As a result, all objects contributing to an 2766 application must either be compiled with GCC versions up to 4.6.x 2767 or with GCC versions 4.7.1 or later. If the compiler is used with 2768 AVR Libc, you need a version that supports the new layout, i.e. 2769 implements [2]#35407. 2770 * The AVR port's -mshort-calls command-line option has been 2771 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 2772 for a replacement. 2773 * The AVR port only references startup code that clears .bss and the 2774 common section resp. initializes the .data and .rodata section 2775 provided respective sections (or subsections thereof) are not 2776 empty, see [3]PR18145. Applications that put all static storage 2777 objects into non-standard sections and / or define all static 2778 storage objects in assembler modules, must reference __do_clear_bss 2779 resp. __do_copy_data by hand or undefine the symbol(s) by means of 2780 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 2781 * The ARM port's -mwords-little-endian option has been deprecated. It 2782 will be removed in a future release. 2783 * Support has been removed for the NetWare x86 configuration 2784 obsoleted in GCC 4.6. 2785 * It is no longer possible to use the "l" constraint in MIPS16 asm 2786 statements. 2787 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 2788 library which affected the ABI in C++11 mode: a data member was 2789 added to std::list changing its size and altering the definitions 2790 of some member functions, and std::pair's move constructor was 2791 non-trivial which altered the calling convention for functions with 2792 std::pair arguments or return types. The ABI incompatibilities have 2793 been fixed for GCC version 4.7.2 but as a result C++11 code 2794 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 2795 code compiled with different GCC versions and with C++98/C++03 code 2796 compiled with any version. 2797 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 2798 rules for the layout of vectors that could lead to wrong code being 2799 generated. Vectors larger than 8 bytes in size are now by default 2800 aligned to an 8-byte boundary. This is an ABI change: code that 2801 makes explicit use of vector types may be incompatible with binary 2802 objects built with older versions of GCC. Auto-vectorized code is 2803 not affected by this change. (This change affects GCC versions 2804 4.7.2 and later.) 2805 * More information on porting to GCC 4.7 from previous versions of 2806 GCC can be found in the [4]porting guide for this release. 2807 2808General Optimizer Improvements 2809 2810 * Support for a new parameter --param case-values-threshold=n was 2811 added to allow users to control the cutoff between doing switch 2812 statements as a series of if statements and using a jump table. 2813 * Link-time optimization (LTO) improvements: 2814 + Improved scalability and reduced memory usage. Link time 2815 optimization of Firefox now requires 3GB of RAM on a 64-bit 2816 system, while over 8GB was needed previously. Linking time has 2817 been improved, too. The serial stage of linking Firefox has 2818 been sped up by about a factor of 10. 2819 + Reduced size of object files and temporary storage used during 2820 linking. 2821 + Streaming performance (both outbound and inbound) has been 2822 improved. 2823 + ld -r is now supported with LTO. 2824 + Several bug fixes, especially in symbol table handling and 2825 merging. 2826 * Interprocedural optimization improvements: 2827 + Heuristics now take into account that after inlining code will 2828 be optimized out because of known values (or properties) of 2829 function parameters. For example: 2830void foo(int a) 2831{ 2832 if (a > 10) 2833 ... huge code ... 2834} 2835void bar (void) 2836{ 2837 foo (0); 2838} 2839 2840 The call of foo will be inlined into bar even when optimizing 2841 for code size. Constructs based on __builtin_constant_p are 2842 now understood by the inliner and code size estimates are 2843 evaluated a lot more realistically. 2844 + The representation of C++ virtual thunks and aliases (both 2845 implicit and defined via the alias attribute) has been 2846 re-engineered. Aliases no longer pose optimization barriers 2847 and calls to an alias can be inlined and otherwise optimized. 2848 + The inter-procedural constant propagation pass has been 2849 rewritten. It now performs generic function specialization. 2850 For example when compiling the following: 2851void foo(bool flag) 2852{ 2853 if (flag) 2854 ... do something ... 2855 else 2856 ... do something else ... 2857} 2858void bar (void) 2859{ 2860 foo (false); 2861 foo (true); 2862 foo (false); 2863 foo (true); 2864 foo (false); 2865 foo (true); 2866} 2867 2868 GCC will now produce two copies of foo. One with flag being 2869 true, while other with flag being false. This leads to 2870 performance improvements previously possible only by inlining 2871 all calls. Cloning causes a lot less code size growth. 2872 * A string length optimization pass has been added. It attempts to 2873 track string lengths and optimize various standard C string 2874 functions like strlen, strchr, strcpy, strcat, stpcpy and their 2875 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 2876 enabled by default at -O2 or above, unless optimizing for size, and 2877 can be disabled by the -fno-optimize-strlen option. The pass can 2878 e.g. optimize 2879char *bar (const char *a) 2880{ 2881 size_t l = strlen (a) + 2; 2882 char *p = malloc (l); if (p == NULL) return p; 2883 strcpy (p, a); strcat (p, "/"); return p; 2884} 2885 2886 into: 2887char *bar (const char *a) 2888{ 2889 size_t tmp = strlen (a); 2890 char *p = malloc (tmp + 2); if (p == NULL) return p; 2891 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 2892} 2893 2894 or for hosted compilations where stpcpy is available in the runtime 2895 and headers provide its prototype, e.g. 2896void foo (char *a, const char *b, const char *c, const char *d) 2897{ 2898 strcpy (a, b); strcat (a, c); strcat (a, d); 2899} 2900 2901 can be optimized into: 2902void foo (char *a, const char *b, const char *c, const char *d) 2903{ 2904 strcpy (stpcpy (stpcpy (a, b), c), d); 2905} 2906 2907New Languages and Language specific improvements 2908 2909 * Version 3.1 of the [5]OpenMP specification is now supported for the 2910 C, C++, and Fortran compilers. 2911 2912 Ada 2913 2914 * The command-line option -feliminate-unused-debug-types has been 2915 re-enabled by default, as it is for the other languages, leading to 2916 a reduction in debug info size of 12.5% and more for relevant 2917 cases, as well as to a small compilation speedup. 2918 2919 C family 2920 2921 * A new built-in, __builtin_assume_aligned, has been added, through 2922 which the compiler can be hinted about pointer alignment and can 2923 use it to improve generated code. 2924 * A new warning option -Wunused-local-typedefs was added for C, C++, 2925 Objective-C and Objective-C++. This warning diagnoses typedefs 2926 locally defined in a function, and otherwise not used. 2927 * A new experimental command-line option -ftrack-macro-expansion was 2928 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 2929 the compiler to emit diagnostic about the current macro expansion 2930 stack when a compilation error occurs in a macro expansion. 2931 * Experimental support for transactional memory has been added. It 2932 includes support in the compiler, as well as a supporting runtime 2933 library called libitm. To compile code with transactional memory 2934 constructs, use the -fgnu-tm option. 2935 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 2936 and 32-bit/64-bit x86 platforms. 2937 For more details on transactional memory see [6]the GCC WiKi. 2938 * Support for atomic operations specifying the C++11/C11 memory model 2939 has been added. These new __atomic routines replace the existing 2940 __sync built-in routines. 2941 Atomic support is also available for memory blocks. Lock-free 2942 instructions will be used if a memory block is the same size and 2943 alignment as a supported integer type. Atomic operations which do 2944 not have lock-free support are left as function calls. A set of 2945 library functions is available on the GCC atomic wiki in the 2946 "External Atomics Library" section. 2947 For more details on the memory models and features, see the 2948 [7]atomic wiki. 2949 * When a binary operation is performed on vector types and one of the 2950 operands is a uniform vector, it is possible to replace the vector 2951 with the generating element. For example: 2952typedef int v4si __attribute__ ((vector_size (16))); 2953v4si res, a = {1,2,3,4}; 2954int x; 2955 2956res = 2 + a; /* means {2,2,2,2} + a */ 2957res = a - x; /* means a - {x,x,x,x} */ 2958 2959 C 2960 2961 * There is support for some more features from the C11 revision of 2962 the ISO C standard. GCC now accepts the options -std=c11 and 2963 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 2964 + Unicode strings (previously supported only with options such 2965 as -std=gnu11, now supported with -std=c11), and the 2966 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 2967 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 2968 + Alignment support (_Alignas, _Alignof, max_align_t, 2969 <stdalign.h>). 2970 + A built-in function __builtin_complex is provided to support C 2971 library implementation of the CMPLX family of macros. 2972 2973 C++ 2974 2975 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 2976 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 2977 -Wc++0x-compat, respectively. 2978 * G++ now implements [8]C++11 extended friend syntax: 2979 2980template<class W> 2981class Q 2982{ 2983 static const int I = 2; 2984public: 2985 friend W; 2986}; 2987 2988struct B 2989{ 2990 int ar[Q<B>::I]; 2991}; 2992 2993 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit 2994 override control. 2995 2996struct B { 2997 virtual void f() const final; 2998 virtual void f(int); 2999}; 3000 3001struct D : B { 3002 void f() const; // error: D::f attempts to override final B::f 3003 void f(long) override; // error: doesn't override anything 3004 void f(int) override; // ok 3005}; 3006 3007struct E final { }; 3008struct F: E { }; // error: deriving from final class 3009 3010 * G++ now implements [10]C++11 non-static data member initializers. 3011 3012struct A { 3013 int i = 42; 3014} a; // initializes a.i to 42 3015 3016 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11 3017 user-defined literals. 3018 3019// Not actually a good approximation. :) 3020constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 3021long double pi = 180.0_degrees; 3022 3023 * G++ now implements [12]C++11 alias-declarations. 3024 3025template <class T> using Ptr = T*; 3026Ptr<int> ip; // decltype(ip) is int* 3027 3028 * Thanks to Ville Voutilainen and Pedro Lamarao, G++ now implements 3029 [13]C++11 delegating constructors. 3030 3031struct A { 3032 A(int); 3033 A(): A(42) { } // delegate to the A(int) constructor 3034}; 3035 3036 * G++ now fully implements C++11 atomic classes rather than just 3037 integer derived classes. 3038 3039class POD { 3040 int a; 3041 int b; 3042}; 3043std::atomic<POD> my_atomic_POD; 3044 3045 * G++ now sets the predefined macro __cplusplus to the correct value, 3046 199711L for C++98/03, and 201103L for C++11. 3047 * G++ now correctly implements the two-phase lookup rules such that 3048 an unqualified name used in a template must have an appropriate 3049 declaration found either in scope at the point of definition of the 3050 template or by argument-dependent lookup at the point of 3051 instantiation. As a result, code that relies on a second 3052 unqualified lookup at the point of instantiation to find functions 3053 declared after the template or in dependent bases will be rejected. 3054 The compiler will suggest ways to fix affected code, and using the 3055 -fpermissive compiler flag will allow the code to compile with a 3056 warning. 3057 3058template <class T> 3059void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 3060void g(int) { } // fix by moving this declaration before the declaration of f 3061 3062template <class T> 3063struct A: T { 3064 // error, B::g(B) not found by argument-dependent lookup 3065 void f() { g(T()); } // fix by using this->g or A::g 3066}; 3067 3068struct B { void g(B); }; 3069 3070int main() 3071{ 3072 f<int>(); 3073 A<B>().f(); 3074} 3075 3076 * G++ now properly re-uses stack space allocated for temporary 3077 objects when their lifetime ends, which can significantly lower 3078 stack consumption for some C++ functions. As a result of this, some 3079 code with undefined behavior will now break: 3080 3081const int &f(const int &i) { return i; } 3082.... 3083const int &x = f(1); 3084const int &y = f(2); 3085 3086 Here, x refers to the temporary allocated to hold the 1 argument, 3087 which only lives until the end of the initialization; it 3088 immediately becomes a dangling reference. So the next statement 3089 re-uses the stack slot to hold the 2 argument, and users of x get 3090 that value instead. 3091 Note that this should not cause any change of behavior for 3092 temporaries of types with non-trivial destructors, as they are 3093 already destroyed at end of full-expression; the change is that now 3094 the storage is released as well. 3095 * A new command-line option -Wdelete-non-virtual-dtor has been added 3096 to warn when delete is used to destroy an instance of a class which 3097 has virtual functions and non-virtual destructor. It is unsafe to 3098 delete an instance of a derived class through a pointer to a base 3099 class if the base class does not have a virtual destructor. This 3100 warning is enabled by -Wall. 3101 * A new command-line option -Wzero-as-null-pointer-constant has been 3102 added to warn when a literal '0' is used as null pointer constant. 3103 It can be useful to facilitate the conversion to nullptr in C++11. 3104 * As per C++98, access-declarations are now deprecated by G++. 3105 Using-declarations are to be used instead. Furthermore, some 3106 efforts have been made to improve the support of class scope 3107 using-declarations. In particular, using-declarations referring to 3108 a dependent type now work as expected ([14]bug c++/14258). 3109 * The ELF symbol visibility of a template instantiation is now 3110 properly constrained by the visibility of its template arguments 3111 ([15]bug c++/35688). 3112 3113 Runtime Library (libstdc++) 3114 3115 * [16]Improved experimental support for the new ISO C++ standard, 3116 C++11, including: 3117 + using noexcept in most of the library; 3118 + implementations of pointer_traits, allocator_traits and 3119 scoped_allocator_adaptor; 3120 + uses-allocator construction for tuple; 3121 + vector meets the allocator-aware container requirements; 3122 + replacing monotonic_clock with steady_clock; 3123 + enabling the thread support library on most POSIX targets; 3124 + many small improvements to conform to the FDIS. 3125 * Added --enable-clocale=newlib configure option. 3126 * Debug Mode iterators for unordered associative containers. 3127 * Avoid polluting the global namespace and do not include <unistd.h>. 3128 3129 Fortran 3130 3131 * The compile flag [17]-fstack-arrays has been added, which causes 3132 all local arrays to be put on stack memory. For some programs this 3133 will improve the performance significantly. If your program uses 3134 very large local arrays, it is possible that you will have to 3135 extend your runtime limits for stack memory. 3136 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and 3137 [20]-fstack-arrays. 3138 * Front-end optimizations can now be selected by the 3139 [21]-ffrontend-optimize option and deselected by the 3140 -fno-frontend-optimize option. 3141 * When front-end optimization removes a function call, 3142 [22]-Wfunction-elimination warns about that. 3143 * When performing front-end-optimization, the 3144 [23]-faggressive-function-elimination option allows the removal of 3145 duplicate function calls even for impure functions. 3146 * The flag [24]-Wreal-q-constant has been added, which warns if 3147 floating-point literals have been specified using q (such as 3148 1.0q0); the q marker is now supported as a vendor extension to 3149 denote quad precision (REAL(16) or, if not available, REAL(10)). 3150 Consider using a kind parameter (such as in 1.0_qp) instead, which 3151 can be obtained via [25]SELECTED_REAL_KIND. 3152 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 3153 Fortran now always prints error messages to standard error. If you 3154 wish to redirect standard error, please consult the manual for your 3155 OS, shell, batch environment etc. as appropriate. 3156 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 3157 variable have been removed. When encountering a serious error, 3158 gfortran will now always abort the program. Whether a core dump is 3159 generated depends on the user environment settings; see the ulimit 3160 -c setting for POSIX shells, limit coredumpsize for C shells, and 3161 the [26]WER user-mode dumps settings on Windows. 3162 * The [27]-fbacktrace option is now enabled by default. When 3163 encountering a fatal error, gfortran will attempt to print a 3164 backtrace to standard error before aborting. It can be disabled 3165 with -fno-backtrace. Note: On POSIX targets with the addr2line 3166 utility from GNU binutils, GNU Fortran can print a backtrace with 3167 function name, file name, line number information in addition to 3168 the addresses; otherwise only the addresses are printed. 3169 * [28]Fortran 2003: 3170 + Generic interface names which have the same name as derived 3171 types are now supported, which allows to write constructor 3172 functions. Note that Fortran does not support static 3173 constructor functions; only default initialization or an 3174 explicit structure-constructor initialization are available. 3175 + [29]Polymorphic (class) arrays are now supported. 3176 * [30]Fortran 2008: 3177 + Support for the DO CONCURRENT construct has been added, which 3178 allows the user to specify that individual loop iterations 3179 have no interdependencies. 3180 + [31]Coarrays: Full single-image support except for polymorphic 3181 coarrays. Additionally, preliminary support for multiple 3182 images via an MPI-based [32]coarray communication library has 3183 been added. Note: The library version is not yet usable as 3184 remote coarray access is not yet possible. 3185 * [33]TS 29113: 3186 + New flag [34]-std=f2008ts permits programs that are expected 3187 to conform to the Fortran 2008 standard and the draft 3188 Technical Specification (TS) 29113 on Further Interoperability 3189 of Fortran with C. 3190 + The OPTIONAL attribute is now allowed for dummy arguments of 3191 BIND(C) procedures. 3192 + The RANK intrinsic has been added. 3193 + The implementation of the ASYNCHRONOUS attribute in GCC is 3194 compatible with the candidate draft of TS 29113 (since GCC 3195 4.6). 3196 3197 Go 3198 3199 * GCC 4.7 implements the [35]Go 1 language standard. The library 3200 support in 4.7.0 is not quite complete, due to release timing. 3201 Release 4.7.1 includes complete support for Go 1. The Go library is 3202 from the Go 1.0.1 release. 3203 * Go has been tested on GNU/Linux and Solaris platforms. It may work 3204 on other platforms as well. 3205 3206New Targets and Target Specific Improvements 3207 3208 ARM 3209 3210 * GCC now supports the Cortex-A7 processor implementing the v7-a 3211 version of the architecture using the option -mcpu=cortex-a7. 3212 * The default vector size in auto-vectorization for NEON is now 128 3213 bits. If vectorization fails thusly, the vectorizer tries again 3214 with 64-bit vectors. 3215 * A new option -mvectorize-with-neon-double was added to allow users 3216 to change the vector size to 64 bits. 3217 3218 AVR 3219 3220 * GCC now supports the XMEGA architecture. This requires GNU binutils 3221 2.22 or later. 3222 * Support for the [36]named address spaces __flash, __flash1, ..., 3223 __flash5 and __memx has been added. These address spaces locate 3224 read-only data in flash memory and allow reading from flash memory 3225 by means of ordinary C code, i.e. without the need of (inline) 3226 assembler code: 3227 3228const __flash int values[] = { 42, 31 }; 3229 3230int add_values (const __flash int *p, int i) 3231{ 3232 return values[i] + *p; 3233} 3234 3235 * Support has been added for the AVR-specific configure option 3236 --with-avrlibc=yes in order to arrange for better integration of 3237 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 3238 and newer and will only take effect in non-RTEMS configurations. If 3239 avr-gcc is configured for RTEMS, the option will be ignored which 3240 is the same as specifying --with-avrlibc=no. See [38]PR54461 for 3241 more technical details. 3242 * Support for AVR-specific [39]built-in functions has been added. 3243 * Support has been added for the signed and unsigned 24-bit scalar 3244 integer types __int24 and __uint24. 3245 * New command-line options -maccumulate-args, -mbranch-cost=cost and 3246 -mstrict-X were added to allow better fine-tuning of code 3247 optimization. 3248 * The command option -fdata-sections now also takes affect on the 3249 section names of variables with the progmem attribute. 3250 * A new inline assembler print modifier %i to print a RAM address as 3251 I/O address has been added: 3252 3253#include <avr/io.h> /* Port Definitions from AVR-LibC */ 3254 3255void set_portb (uint8_t value) 3256{ 3257 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 3258} 3259 3260 The offset between an I/O address and the RAM address for that I/O 3261 location is device-specific. This offset is taken into account when 3262 printing a RAM address with the %i modifier so that the address is 3263 suitable to be used as operand in an I/O command. The address must 3264 be a constant integer known at compile time. 3265 * The inline assembler constraint "R" to represent integers in the 3266 range -6 ... 5 has been removed without replacement. 3267 * Many optimizations to: 3268 + 64-bit integer arithmetic 3269 + Widening multiplication 3270 + Integer division by a constant 3271 + Avoid constant reloading in multi-byte instructions. 3272 + Micro-optimizations for special instruction sequences. 3273 + Generic built-in functions like __builtin_ffs*, 3274 __builtin_clz*, etc. 3275 + If-else decision trees generated by switch instructions 3276 + Merging of data located in flash memory 3277 + New libgcc variants for devices with 8-bit wide stack pointer 3278 + ... 3279 * Better documentation: 3280 + Handling of EIND and indirect jumps on devices with more than 3281 128 KiB of program memory. 3282 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 3283 registers. 3284 + Function attributes OS_main and OS_task. 3285 + AVR-specific built-in macros. 3286 3287 C6X 3288 3289 * Support has been added for the Texas Instruments C6X family of 3290 processors. 3291 3292 CR16 3293 3294 * Support has been added for National Semiconductor's CR16 3295 architecture. 3296 3297 Epiphany 3298 3299 * Support has been added for Adapteva's Epiphany architecture. 3300 3301 IA-32/x86-64 3302 3303 * Support for Intel AVX2 intrinsics, built-in functions and code 3304 generation is available via -mavx2. 3305 * Support for Intel BMI2 intrinsics, built-in functions and code 3306 generation is available via -mbmi2. 3307 * Implementation and automatic generation of __builtin_clz* using the 3308 lzcnt instruction is available via -mlzcnt. 3309 * Support for Intel FMA3 intrinsics and code generation is available 3310 via -mfma. 3311 * A new -mfsgsbase command-line option is available that makes GCC 3312 generate new segment register read/write instructions through 3313 dedicated built-ins. 3314 * Support for the new Intel rdrnd instruction is available via 3315 -mrdrnd. 3316 * Two additional AVX vector conversion instructions are available via 3317 -mf16c. 3318 * Support for new Intel processor codename IvyBridge with RDRND, 3319 FSGSBASE and F16C is available through -march=core-avx-i. 3320 * Support for the new Intel processor codename Haswell with AVX2, 3321 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 3322 * Support for new AMD family 15h processors (Piledriver core) is now 3323 available through -march=bdver2 and -mtune=bdver2 options. 3324 * Support for [40]the x32 psABI is now available through the -mx32 3325 option. 3326 * Windows mingw targets are using the -mms-bitfields option by 3327 default. 3328 * Windows x86 targets are using the __thiscall calling convention for 3329 C++ class-member functions. 3330 * Support for the configure option --with-threads=posix for Windows 3331 mingw targets. 3332 3333 MIPS 3334 3335 * GCC now supports thread-local storage (TLS) for MIPS16. This 3336 requires GNU binutils 2.22 or later. 3337 * GCC can now generate code specifically for the Cavium Octeon+ and 3338 Octeon2 processors. The associated command-line options are 3339 -march=octeon+ and -march=octeon2 respectively. Both options 3340 require GNU binutils 2.22 or later. 3341 * GCC can now work around certain 24k errata, under the control of 3342 the command-line option -mfix-24k. These workarounds require GNU 3343 binutils 2.20 or later. 3344 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 3345 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 3346 toolchain that generates 32-bit code by default. Use the 3347 configure-time option --enable-targets=all to select these extra 3348 multilibs. 3349 * Passing -fno-delayed-branch now also stops the assembler from 3350 automatically filling delay slots. 3351 3352 PowerPC/PowerPC64 3353 3354 * Vectors of type vector long long or vector long are passed and 3355 returned using the same method as other vectors with the VSX 3356 instruction set. Previously GCC did not adhere to the ABI for 3357 128-bit vectors with 64-bit integer base types (PR 48857). This 3358 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 3359 * A new option -mno-pointers-to-nested-functions was added to allow 3360 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 3361 that the compiler should not load up the chain register (r11) 3362 before calling a function through a pointer. If you use this 3363 option, you cannot call nested functions through a pointer, or call 3364 other languages that might use the static chain. 3365 * A new option msave-toc-indirect was added to allow AIX 3366 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 3367 save the TOC in the prologue for indirect calls or generate the 3368 save inline. This can speed up some programs that call through a 3369 function pointer a lot, but it can slow down other functions that 3370 only call through a function pointer in exceptional cases. 3371 * The PowerPC port will now enable machine-specific built-in 3372 functions when the user switches the target machine using the 3373 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 3374 sequences. In addition, the target macros are updated. However, due 3375 to the way the -save-temps switch is implemented, you won't see the 3376 effect of these additional macros being defined in preprocessor 3377 output. 3378 3379 SH 3380 3381 * A new option -msoft-atomic has been added. When it is specified, 3382 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 3383 the new __atomic routines. 3384 * Since it is neither supported by GAS nor officially documented, 3385 code generation for little endian SH2A has been disabled. 3386 Specifying -ml with -m2a* will now result in a compiler error. 3387 * The defunct -mbranch-cost option has been fixed. 3388 * Some improvements to the generated code of: 3389 + Utilization of the tst #imm,R0 instruction. 3390 + Dynamic shift instructions on SH2A. 3391 + Integer absolute value calculations. 3392 * The -mdiv= option for targets other than SHmedia has been fixed and 3393 documented. 3394 3395 SPARC 3396 3397 * The option -mflat has been reinstated. When it is specified, the 3398 compiler will generate code for a single register window model. 3399 This is essentially a new implementation and the corresponding 3400 debugger support has been added to GDB 7.4. 3401 * Support for the options -mtune=native and -mcpu=native has been 3402 added on selected native platforms (GNU/Linux and Solaris). 3403 * Support for the SPARC T3 (Niagara 3) processor has been added. 3404 * VIS: 3405 + An intrinsics header visintrin.h has been added. 3406 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 3407 compare instructions have been added. 3408 + The little-endian version of alignaddr is now supported. 3409 + When possible, VIS builtins are marked const, which should 3410 increase the compiler's ability to optimize VIS operations. 3411 + The compiler now properly tracks the %gsr register and how it 3412 behaves as an input for various VIS instructions. 3413 + Akin to fzero, the compiler can now generate fone instructions 3414 in order to set all of the bits of a floating-point register 3415 to 1. 3416 + The documentation for the VIS intrinsics in the GCC manual has 3417 been brought up to date and many inaccuracies were fixed. 3418 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 3419 non-condition-code setting edge instructions have been added. 3420 Their availability is controlled by the new -mvis2 and 3421 -mno-vis2 options. They are enabled by default on 3422 UltraSPARC-III and later CPUs. 3423 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 3424 has been added. These instructions are enabled by default on SPARC 3425 T3 (Niagara 3) and later CPUs. 3426 3427 TILE-Gx/TILEPro 3428 3429 * Support has been added for the Tilera TILE-Gx and TILEPro families 3430 of processors. 3431 3432Other significant improvements 3433 3434 * A new option (-grecord-gcc-switches) was added that appends 3435 compiler command-line options that might affect code generation to 3436 the DW_AT_producer attribute string in the DWARF debugging 3437 information. 3438 * GCC now supports various new GNU extensions to the DWARF debugging 3439 information format, like [41]entry value and [42]call site 3440 information, [43]typed DWARF stack or [44]a more compact macro 3441 representation. Support for these extensions has been added to GDB 3442 7.4. They can be disabled through the -gstrict-dwarf command-line 3443 option. 3444 3445GCC 4.7.1 3446 3447 This is the [45]list of problem reports (PRs) from GCC's bug tracking 3448 system that are known to be fixed in the 4.7.1 release. This list might 3449 not be complete (that is, it is possible that some PRs that have been 3450 fixed are not listed here). 3451 3452 The Go frontend in the 4.7.1 release fully supports the [46]Go 1 3453 language standard. 3454 3455GCC 4.7.2 3456 3457 This is the [47]list of problem reports (PRs) from GCC's bug tracking 3458 system that are known to be fixed in the 4.7.2 release. This list might 3459 not be complete (that is, it is possible that some PRs that have been 3460 fixed are not listed here). 3461 3462GCC 4.7.3 3463 3464 This is the [48]list of problem reports (PRs) from GCC's bug tracking 3465 system that are known to be fixed in the 4.7.3 release. This list might 3466 not be complete (that is, it is possible that some PRs that have been 3467 fixed are not listed here). 3468 3469GCC 4.7.4 3470 3471 This is the [49]list of problem reports (PRs) from GCC's bug tracking 3472 system that are known to be fixed in the 4.7.4 release. This list might 3473 not be complete (that is, it is possible that some PRs that have been 3474 fixed are not listed here). 3475 3476 3477 For questions related to the use of GCC, please consult these web 3478 pages and the [50]GCC manuals. If that fails, the 3479 [51]gcc-help@gcc.gnu.org mailing list might help. Comments on these 3480 web pages and the development of GCC are welcome on our developer 3481 list at [52]gcc@gcc.gnu.org. All of [53]our lists have public 3482 archives. 3483 3484 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and 3485 distribution of this entire article is permitted in any medium, 3486 provided this notice is preserved. 3487 3488 These pages are [55]maintained by the GCC team. Last modified 3489 2016-02-29[56]. 3490 3491References 3492 3493 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 3494 2. http://savannah.nongnu.org/bugs/?35407 3495 3. https://gcc.gnu.org/PR18145 3496 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 3497 5. http://openmp.org/wp/openmp-specifications/ 3498 6. https://gcc.gnu.org/wiki/TransactionalMemory 3499 7. https://gcc.gnu.org/wiki/Atomic/GCCMM 3500 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3501 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3502 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3503 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3504 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3505 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3506 14. https://gcc.gnu.org/PR14258 3507 15. https://gcc.gnu.org/PR35688 3508 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 3509 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 3510 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 3511 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 3512 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 3513 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 3514 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 3515 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 3516 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 3517 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 3518 26. https://msdn.microsoft.com/en-us/library/bb787181%28v=vs.85%29.aspx 3519 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 3520 28. https://gcc.gnu.org/wiki/Fortran2003Status 3521 29. https://gcc.gnu.org/wiki/OOP 3522 30. https://gcc.gnu.org/wiki/Fortran2008Status 3523 31. https://gcc.gnu.org/wiki/Coarray 3524 32. https://gcc.gnu.org/wiki/CoarrayLib 3525 33. https://gcc.gnu.org/wiki/TS29113Status 3526 34. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 3527 35. https://golang.org/doc/go1 3528 36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 3529 37. http://nongnu.org/avr-libc/ 3530 38. https://gcc.gnu.org/PR54461 3531 39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html 3532 40. https://sites.google.com/site/x32abi/ 3533 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 3534 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 3535 43. http://www.dwarfstd.org/doc/040408.1.html 3536 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 3537 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 3538 46. https://golang.org/doc/go1 3539 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 3540 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 3541 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 3542 50. https://gcc.gnu.org/onlinedocs/ 3543 51. mailto:gcc-help@gcc.gnu.org 3544 52. mailto:gcc@gcc.gnu.org 3545 53. https://gcc.gnu.org/lists.html 3546 54. http://www.fsf.org/ 3547 55. https://gcc.gnu.org/about.html 3548 56. http://validator.w3.org/check/referer 3549====================================================================== 3550http://gcc.gnu.org/gcc-4.6/index.html 3551 3552 GCC 4.6 Release Series 3553 3554 April 12, 2013 3555 3556 The [1]GNU project and the GCC developers are pleased to announce the 3557 release of GCC 4.6.4. 3558 3559 This release is a bug-fix release, containing fixes for regressions in 3560 GCC 4.6.3 relative to previous releases of GCC. 3561 3562Release History 3563 3564 GCC 4.6.4 3565 April 12, 2013 ([2]changes, [3]documentation) 3566 3567 GCC 4.6.3 3568 March 1, 2012 ([4]changes, [5]documentation) 3569 3570 GCC 4.6.2 3571 October 26, 2011 ([6]changes, [7]documentation) 3572 3573 GCC 4.6.1 3574 June 27, 2011 ([8]changes, [9]documentation) 3575 3576 GCC 4.6.0 3577 March 25, 2011 ([10]changes, [11]documentation) 3578 3579References and Acknowledgements 3580 3581 GCC used to stand for the GNU C Compiler, but since the compiler 3582 supports several other languages aside from C, it now stands for the 3583 GNU Compiler Collection. 3584 3585 A list of [12]successful builds is updated as new information becomes 3586 available. 3587 3588 The GCC developers would like to thank the numerous people that have 3589 contributed new features, improvements, bug fixes, and other changes as 3590 well as test results to GCC. This [13]amazing group of volunteers is 3591 what makes GCC successful. 3592 3593 For additional information about GCC please refer to the [14]GCC 3594 project web site or contact the [15]GCC development mailing list. 3595 3596 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 3597 3598 3599 For questions related to the use of GCC, please consult these web 3600 pages and the [18]GCC manuals. If that fails, the 3601 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 3602 web pages and the development of GCC are welcome on our developer 3603 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 3604 archives. 3605 3606 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3607 distribution of this entire article is permitted in any medium, 3608 provided this notice is preserved. 3609 3610 These pages are [23]maintained by the GCC team. Last modified 3611 2016-01-30[24]. 3612 3613References 3614 3615 1. http://www.gnu.org/ 3616 2. http://gcc.gnu.org/gcc-4.6/changes.html 3617 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 3618 4. http://gcc.gnu.org/gcc-4.6/changes.html 3619 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 3620 6. http://gcc.gnu.org/gcc-4.6/changes.html 3621 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 3622 8. http://gcc.gnu.org/gcc-4.6/changes.html 3623 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 3624 10. http://gcc.gnu.org/gcc-4.6/changes.html 3625 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 3626 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 3627 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3628 14. http://gcc.gnu.org/index.html 3629 15. mailto:gcc@gcc.gnu.org 3630 16. http://gcc.gnu.org/mirrors.html 3631 17. http://gcc.gnu.org/svn.html 3632 18. https://gcc.gnu.org/onlinedocs/ 3633 19. mailto:gcc-help@gcc.gnu.org 3634 20. mailto:gcc@gcc.gnu.org 3635 21. https://gcc.gnu.org/lists.html 3636 22. http://www.fsf.org/ 3637 23. https://gcc.gnu.org/about.html 3638 24. http://validator.w3.org/check/referer 3639====================================================================== 3640http://gcc.gnu.org/gcc-4.6/changes.html 3641 3642 GCC 4.6 Release Series 3643 Changes, New Features, and Fixes 3644 3645Caveats 3646 3647 * The options -b <machine> and -V <version> have been removed because 3648 they were unreliable. Instead, users should directly run 3649 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 3650 run a different version of gcc. 3651 * GCC now has stricter checks for invalid command-line options. In 3652 particular, when gcc was called to link object files rather than 3653 compile source code, it would previously accept and ignore all 3654 options starting with --, including linker options such as 3655 --as-needed and --export-dynamic, although such options would 3656 result in errors if any source code was compiled. Such options, if 3657 unknown to the compiler, are now rejected in all cases; if the 3658 intent was to pass them to the linker, options such as 3659 -Wl,--as-needed should be used. 3660 * Versions of the GNU C library up to and including 2.11.1 included 3661 an [1]incorrect implementation of the cproj function. GCC optimizes 3662 its builtin cproj according to the behavior specified and allowed 3663 by the ISO C99 standard. If you want to avoid discrepancies between 3664 the C library and GCC's builtin transformations when using cproj in 3665 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 3666 and actually rely on the incorrect behavior of cproj, then you can 3667 disable GCC's transformations using -fno-builtin-cproj. 3668 * The C-only intermodule optimization framework (IMA, enabled by 3669 -combine) has been removed in favor of the new generic link-time 3670 optimization framework (LTO) introduced in [2]GCC 4.5.0. 3671 * GCC now ships with the LGPL-licensed libquadmath library, which 3672 provides quad-precision mathematical functions for targets with a 3673 __float128 datatype. __float128 is available for targets on 32-bit 3674 x86, x86-64 and Itanium architectures. The libquadmath library is 3675 automatically built on such targets when building the Fortran 3676 compiler. 3677 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 3678 warnings were added for C, C++, Objective-C and Objective-C++. 3679 These warnings diagnose variables respective parameters which are 3680 only set in the code and never otherwise used. Usually such 3681 variables are useless and often even the value assigned to them is 3682 computed needlessly, sometimes expensively. The 3683 -Wunused-but-set-variable warning is enabled by default by -Wall 3684 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 3685 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 3686 rules for the layout of vectors that could lead to wrong code being 3687 generated. Vectors larger than 8 bytes in size are now by default 3688 aligned to an 8-byte boundary. This is an ABI change: code that 3689 makes explicit use of vector types may be incompatible with binary 3690 objects built with older versions of GCC. Auto-vectorized code is 3691 not affected by this change. (This change affects GCC versions 3692 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 3693 * On AVR, variables with the progmem attribute to locate data in 3694 flash memory must be qualified as const. 3695 * Support for a number of older systems and recently unmaintained or 3696 untested target ports of GCC has been declared obsolete in GCC 4.6. 3697 Unless there is activity to revive them, the next release of GCC 3698 will have their sources permanently removed. 3699 All GCC ports for the following processor architectures have been 3700 declared obsolete: 3701 + Argonaut ARC (arc-*) 3702 + National Semiconductor CRX (crx-*) 3703 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 3704 m68hc12-*-*, m6812-*-*) 3705 + Sunplus S+core (score-*) 3706 The following ports for individual systems on particular 3707 architectures have been obsoleted: 3708 + Interix (i[34567]86-*-interix3*) 3709 + NetWare x86 (i[3456x]86-*-netware*) 3710 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 3711 + MCore PE (mcore-*-pe*) 3712 + SH SymbianOS (sh*-*-symbianelf*) 3713 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 3714 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 3715 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 3716 vax-*-netbsd*, but not *-*-netbsdelf*) 3717 The i[34567]86-*-pe alias for Cygwin targets has also been 3718 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 3719 Certain configure options to control the set of libraries built 3720 with GCC on some targets have been obsoleted. On ARM targets, the 3721 options --disable-fpu, --disable-26bit, --disable-underscore, 3722 --disable-interwork, --disable-biendian and --disable-nofmult have 3723 been obsoleted. On MIPS targets, the options 3724 --disable-single-float, --disable-biendian and --disable-softfloat 3725 have been obsoleted. 3726 * Support has been removed for all the [3]configurations obsoleted in 3727 GCC 4.5. 3728 * More information on porting to GCC 4.6 from previous versions of 3729 GCC can be found in the [4]porting guide for this release. 3730 3731General Optimizer Improvements 3732 3733 * A new general optimization level, -Ofast, has been introduced. It 3734 combines the existing optimization level -O3 with options that can 3735 affect standards compliance but result in better optimized code. 3736 For example, -Ofast enables -ffast-math. 3737 * Link-time optimization improvements: 3738 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 3739 stabilized to the point of being usable. It has become the 3740 default mode when using the LTO optimization model. Link time 3741 optimization can now split itself into multiple parallel 3742 compilations. Parallelism is controlled with -flto=n (where n 3743 specifies the number of compilations to execute in parallel). 3744 GCC can also cooperate with a GNU make job server by 3745 specifying the -flto=jobserver option and adding + to the 3746 beginning of the Makefile rule executing the linker. 3747 Classical LTO mode can be enforced by -flto-partition=none. 3748 This may result in small code quality improvements. 3749 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 3750 and other large applications can be built with LTO enabled. 3751 + The linker plugin support improvements 3752 o Linker plugin is now enabled by default when the linker 3753 is detected to have plugin support. This is the case for 3754 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 3755 the Gold linker on ELF targets. Plugin support of the 3756 Apple linker on Darwin is not compatible with GCC. The 3757 linker plugin can also be controlled by the 3758 -fuse-linker-plugin command line option. 3759 o Resolution information from the linker plugin is used to 3760 drive whole program assumptions. Use of the linker plugin 3761 results in more aggressive optimization on binaries and 3762 on shared libraries that use the hidden visibility 3763 attribute. Consequently the use of -fwhole-program is not 3764 necessary in addition to LTO. 3765 + Hidden symbols used from non-LTO objects now have to be 3766 explicitly annotated with externally_visible when the linker 3767 plugin is not used. 3768 + C++ inline functions and virtual tables are now privatized 3769 more aggressively, leading to better inter-procedural 3770 optimization and faster dynamic linking. 3771 + Memory usage and intermediate language streaming performance 3772 have been improved. 3773 + Static constructors and destructors from individual units are 3774 inlined into a single function. This can significantly improve 3775 startup times of large C++ applications where static 3776 constructors are very common. For example, static constructors 3777 are used when including the iostream header. 3778 + Support for the Ada language has been added. 3779 * Interprocedural optimization improvements 3780 + The interprocedural framework was re-tuned for link time 3781 optimization. Several scalability issues were resolved. 3782 + Improved auto-detection of const and pure functions. Newly, 3783 noreturn functions are auto-detected. 3784 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 3785 available that informs users when adding attributes to headers 3786 might improve code generation. 3787 + A number of inlining heuristic improvements. In particular: 3788 o Partial inlining is now supported and enabled by default 3789 at -O2 and greater. The feature can be controlled via 3790 -fpartial-inlining. 3791 Partial inlining splits functions with short hot path to 3792 return. This allows more aggressive inlining of the hot 3793 path leading to better performance and often to code size 3794 reductions (because cold parts of functions are not 3795 duplicated). 3796 o Scalability for large compilation units was improved 3797 significantly. 3798 o Inlining of callbacks is now more aggressive. 3799 o Virtual methods are considered for inlining when the 3800 caller is inlined and devirtualization is then possible. 3801 o Inlining when optimizing for size (either in cold regions 3802 of a program or when compiling with -Os) was improved to 3803 better handle C++ programs with larger abstraction 3804 penalty, leading to smaller and faster code. 3805 + The IPA reference optimization pass detecting global variables 3806 used or modified by functions was strengthened and sped up. 3807 + Functions whose address was taken are now optimized out when 3808 all references to them are dead. 3809 + A new inter-procedural static profile estimation pass detects 3810 functions that are executed once or unlikely to be executed. 3811 Unlikely executed functions are optimized for size. Functions 3812 executed once are optimized for size except for the inner 3813 loops. 3814 + On most targets with named section support, functions used 3815 only at startup (static constructors and main), functions used 3816 only at exit and functions detected to be cold are placed into 3817 separate text segment subsections. This extends the 3818 -freorder-functions feature and is controlled by the same 3819 switch. The goal is to improve the startup time of large C++ 3820 programs. 3821 Proper function placement requires linker support. GNU ld 3822 2.21.51 on ELF targets was updated to place those functions 3823 together within the text section leading to better code 3824 locality and faster startup times of large C++ programs. The 3825 feature is also supported in the Apple linker. Support in the 3826 gold linker is planned. 3827 * A new switch -fstack-usage has been added. It makes the compiler 3828 output stack usage information for the program, on a per-function 3829 basis, in an auxiliary file. 3830 * A new switch -fcombine-stack-adjustments has been added. It can be 3831 used to enable or disable the compiler's stack-slot combining pass 3832 which before was enabled automatically at -O1 and above, but could 3833 not be controlled on its own. 3834 * A new switch -fstrict-volatile-bitfields has been added. Using it 3835 indicates that accesses to volatile bitfields should use a single 3836 access of the width of the field's type. This option can be useful 3837 for precisely defining and accessing memory-mapped peripheral 3838 registers from C or C++. 3839 3840Compile time and memory usage improvements 3841 3842 * Datastructures used by the dataflow framework in GCC were 3843 reorganized for better memory usage and more cache locality. 3844 Compile time is improved especially on units with large functions 3845 (possibly resulting from a lot of inlining) not fitting into the 3846 processor cache. The compile time of the GCC C compiler binary with 3847 link-time optimization went down by over 10% (benchmarked on x86-64 3848 target). 3849 3850New Languages and Language specific improvements 3851 3852 Ada 3853 3854 * Stack checking has been improved on selected architectures (Alpha, 3855 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 3856 overflows in all cases on these architectures. 3857 * Initial support for Ada 2012 has been added. 3858 3859 C family 3860 3861 * A new warning, enabled by -Wdouble-promotion, has been added that 3862 warns about cases where a value of type float is implicitly 3863 promoted to double. This is especially helpful for CPUs that handle 3864 the former in hardware, but emulate the latter in software. 3865 * A new function attribute leaf was introduced. This attribute allows 3866 better inter-procedural optimization across calls to functions that 3867 return to the current unit only via returning or exception 3868 handling. This is the case for most library functions that have no 3869 callbacks. 3870 * Support for a new data type __int128 for targets having wide enough 3871 machine-mode support. 3872 * The new function attribute callee_pop_aggregate allows to specify 3873 if the caller or callee is responsible for popping the aggregate 3874 return pointer value from the stack. 3875 * Support for selectively enabling and disabling warnings via #pragma 3876 GCC diagnostic has been added. For instance: 3877#pragma GCC diagnostic error "-Wuninitialized" 3878 foo(a); /* error is given for this one */ 3879#pragma GCC diagnostic push 3880#pragma GCC diagnostic ignored "-Wuninitialized" 3881 foo(b); /* no diagnostic for this one */ 3882#pragma GCC diagnostic pop 3883 foo(c); /* error is given for this one */ 3884#pragma GCC diagnostic pop 3885 foo(d); /* depends on command line options */ 3886 3887 * The -fmax-errors=N option is now supported. Using this option 3888 causes the compiler to exit after N errors have been issued. 3889 3890 C 3891 3892 * There is now experimental support for some features from the 3893 upcoming C1X revision of the ISO C standard. This support may be 3894 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 3895 Note that this support is experimental and may change incompatibly 3896 in future releases for consistency with changes to the C1X standard 3897 draft. The following features are newly supported as described in 3898 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 3899 meeting); some other features were already supported with no 3900 compiler changes being needed, or have some support but not in full 3901 accord with N1539 (as amended). 3902 + Static assertions (_Static_assert keyword) 3903 + Typedef redefinition 3904 + New macros in <float.h> 3905 + Anonymous structures and unions 3906 * The new -fplan9-extensions option directs the compiler to support 3907 some extensions for anonymous struct fields which are implemented 3908 by the Plan 9 compiler. A pointer to a struct may be automatically 3909 converted to a pointer to an anonymous field when calling a 3910 function, in order to make the types match. An anonymous struct 3911 field whose type is a typedef name may be referred to using the 3912 typedef name. 3913 3914 C++ 3915 3916 * Improved [7]experimental support for the upcoming C++0x ISO C++ 3917 standard, including support for constexpr (thanks to Gabriel Dos 3918 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 3919 noexcept, unrestricted unions, range-based for loops (thanks to 3920 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 3921 Rodrigo), implicitly deleted functions and implicit move 3922 constructors. 3923 * When an extern declaration within a function does not match a 3924 declaration in the enclosing context, G++ now properly declares the 3925 name within the namespace of the function rather than the namespace 3926 which was open just before the function definition ([8]c++/43145). 3927 * GCC now warns by default when casting integers to larger pointer 3928 types. These warnings can be disabled with the option 3929 -Wno-int-to-pointer-cast, which is now also available in C++. 3930 * G++ no longer optimizes using the assumption that a value of 3931 enumeration type will fall within the range specified by the 3932 standard, since that assumption is easily violated with a 3933 conversion from integer type ([9]c++/43680). The old behavior can 3934 be restored with -fstrict-enums. 3935 * The new -fnothrow-opt flag changes the semantics of a throw() 3936 exception specification to match the proposed semantics of the 3937 noexcept specification: just call terminate if an exception tries 3938 to propagate out of a function with such an exception 3939 specification. This dramatically reduces or eliminates the code 3940 size overhead from adding the exception specification. 3941 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 3942 a function that the compiler can tell doesn't throw if it would 3943 change the value of a noexcept expression. 3944 * The -Wshadow option now warns if a local variable or type 3945 declaration shadows another type in C++. Note that the compiler 3946 will not warn if a local variable shadows a struct/class/enum, but 3947 will warn if it shadows an explicit typedef. 3948 * When an identifier is not found in the current scope, G++ now 3949 offers suggestions about which identifier might have been intended. 3950 * G++ now issues clearer diagnostics for missing semicolons after 3951 class, struct, and union definitions. 3952 * G++ now issues clearer diagnostics for missing semicolons after 3953 class member declarations. 3954 * G++ now issues clearer diagnostics when a colon is used in a place 3955 where a double-colon was intended. 3956 * G++ no longer accepts mutable on reference members ([10]c++/33558). 3957 Use -fpermissive to allow the old, non-conforming behaviour. 3958 * A few mangling fixes have been made, to attribute const/volatile on 3959 function pointer types, decltype of a plain decl, and use of a 3960 function parameter in the declaration of another parameter. By 3961 default the compiler still uses the old mangling, but emits aliases 3962 with the new mangling on targets that support strong aliases. Users 3963 can switch over entirely to the new mangling with -fabi-version=5 3964 or -fabi-version=0. -Wabi will now warn about code that uses the 3965 old mangling. 3966 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 3967 type to be default initialized unless the type has a user-declared 3968 default constructor. In 4.6.2 G++ implements the proposed 3969 resolution of [11]DR 253, so default initialization is allowed if 3970 it initializes all subobjects. Code that fails to compile can be 3971 fixed by providing an initializer e.g. 3972 struct A { A(); }; 3973 struct B : A { int i; }; 3974 const B b = B(); 3975 Use -fpermissive to allow the old, non-conforming behaviour. 3976 3977 Runtime Library (libstdc++) 3978 3979 * [12]Improved experimental support for the upcoming ISO C++ 3980 standard, C++0x, including using constexpr and nullptr. 3981 * Performance improvements to the [13]Debug Mode, thanks to Franc,ois 3982 Dumont. 3983 * Atomic operations used for reference-counting are annotated so that 3984 they can be understood by race detectors such as Helgrind, see 3985 [14]Data Race Hunting. 3986 * Most libstdc++ standard headers have been changed to no longer 3987 include the cstddef header as an implementation detail. Code that 3988 relied on that header being included as side-effect of including 3989 other standard headers will need to include cstddef explicitly. 3990 3991 Fortran 3992 3993 * On systems supporting the libquadmath library, GNU Fortran now also 3994 supports a quad-precision, kind=16 floating-point data type 3995 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 3996 hardware, calculations might be one to two orders of magnitude 3997 slower than with the 4, 8 or 10 bytes floating-point data types. 3998 This change does not affect systems which support REAL(16) in 3999 hardware nor those which do not support libquadmath. 4000 * Much improved compile time for large array constructors. 4001 * In order to reduce execution time and memory consumption, use of 4002 temporary arrays in assignment expressions is avoided for many 4003 cases. The compiler now reverses loops in order to avoid generating 4004 a temporary array where possible. 4005 * Improved diagnostics, especially with -fwhole-file. 4006 * The -fwhole-file flag is now enabled by default. This improves code 4007 generation and diagnostics. It can be disabled using the deprecated 4008 -fno-whole-file flag. 4009 * Support the generation of Makefile dependencies via the [15]-M... 4010 flags of GCC; you may need to specify the -cpp option in addition. 4011 The dependencies take modules, Fortran's include, and CPP's 4012 #include into account. Note: Using -M for the module path is no 4013 longer supported, use -J instead. 4014 * The flag -Wconversion has been modified to only issue warnings 4015 where a conversion leads to information loss. This drastically 4016 reduces the number of warnings; -Wconversion is thus now enabled 4017 with -Wall. The flag -Wconversion-extra has been added and also 4018 warns about other conversions; -Wconversion-extra typically issues 4019 a huge number of warnings, most of which can be ignored. 4020 * A new command-line option -Wunused-dummy-argument warns about 4021 unused dummy arguments and is included in -Wall. Before, 4022 -Wunused-variable also warned about unused dummy arguments. 4023 * Fortran 2003 support has been extended: 4024 + Improved support for polymorphism between libraries and 4025 programs and for complicated inheritance patterns (cf. 4026 [16]object-oriented programming). 4027 + Experimental support of the ASSOCIATE construct. 4028 + In pointer assignments it is now possible to specify the lower 4029 bounds of the pointer and, for a rank-1 or a simply contiguous 4030 data-target, to remap the bounds. 4031 + Automatic (re)allocation: In intrinsic assignments to 4032 allocatable variables the left-hand side will be automatically 4033 allocated (if unallocated) or reallocated (if the shape or 4034 type parameter is different). To avoid the small performance 4035 penalty, you can use a(:) = ... instead of a = ... for arrays 4036 and character strings - or disable the feature using -std=f95 4037 or -fno-realloc-lhs. 4038 + Deferred type parameter: For scalar allocatable and pointer 4039 variables the character length can be deferred. 4040 + Namelist variables with allocatable and pointer attribute and 4041 nonconstant length type parameter are supported. 4042 * Fortran 2008 support has been extended: 4043 + Experimental [17]coarray support (for one image only, i.e. 4044 num_images() == 1); use the [18]-fcoarray=single flag to 4045 enable it. 4046 + The STOP and the new ERROR STOP statements now support all 4047 constant expressions. 4048 + Support for the CONTIGUOUS attribute. 4049 + Support for ALLOCATE with MOLD. 4050 + Support for the STORAGE_SIZE intrinsic inquiry function. 4051 + Support of the NORM2 and PARITY intrinsic functions. 4052 + The following bit intrinsics were added: POPCNT and POPPAR for 4053 counting the number of 1 bits and returning the parity; BGE, 4054 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 4055 for combined left and right shifts, MASKL and MASKR for simple 4056 left and right justified masks, MERGE_BITS for a bitwise merge 4057 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 4058 and the transformational bit intrinsics IALL, IANY and 4059 IPARITY. 4060 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 4061 + Support for the IMPURE attribute for procedures, which allows 4062 for ELEMENTAL procedures without the restrictions of PURE. 4063 + Null pointers (including NULL()) and not allocated variables 4064 can be used as actual argument to optional non-pointer, 4065 non-allocatable dummy arguments, denoting an absent argument. 4066 + Non-pointer variables with TARGET attribute can be used as 4067 actual argument to POINTER dummies with INTENT(IN) 4068 + Pointers including procedure pointers and those in a derived 4069 type (pointer components) can now be initialized by a target 4070 instead of only by NULL. 4071 + The EXIT statement (with construct-name) can now be used to 4072 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 4073 SELECT CASE and SELECT TYPE constructs. 4074 + Internal procedures can now be used as actual argument. 4075 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 4076 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 4077 have been added; these arrays contain the supported kind 4078 values for the respective types. 4079 + The module procedures C_SIZEOF of the intrinsic module 4080 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 4081 ISO_FORTRAN_ENV have been implemented. 4082 + Minor changes: obsolescence diagnostics for ENTRY was added 4083 for -std=f2008; a line may start with a semicolon; for 4084 internal and module procedures END can be used instead of END 4085 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 4086 a RADIX argument; intrinsic types are supported for 4087 TYPE(intrinsic-type-spec); multiple type-bound procedures can 4088 be declared in a single PROCEDURE statement; implied-shape 4089 arrays are supported for named constants (PARAMETER). The 4090 transformational, three argument versions of BESSEL_JN and 4091 BESSEL_YN were added - the elemental, two-argument version had 4092 been added in GCC 4.4; note that the transformational 4093 functions use a recurrence algorithm. 4094 4095 Go 4096 4097 Support for the [19]Go programming language has been added to GCC. It 4098 is not enabled by default when you build GCC; use the 4099 --enable-languages configure option to build it. The driver program for 4100 compiling Go code is gccgo. 4101 4102 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 4103 is in progress. It may or may not work on other platforms. 4104 4105 Objective-C and Objective-C++ 4106 4107 * The -fobjc-exceptions flag is now required to enable Objective-C 4108 exception and synchronization syntax (introduced by the keywords 4109 @try, @catch, @finally and @synchronized). 4110 * A number of Objective-C 2.0 features and extensions are now 4111 supported by GCC. These features are enabled by default; you can 4112 disable them by using the new -fobjc-std=objc1 command-line option. 4113 * The Objective-C 2.0 dot-syntax is now supported. It is an 4114 alternative syntax for using getters and setters; object.count is 4115 automatically converted into [object count] or [object setCount: 4116 ...] depending on context; for example if (object.count > 0) is 4117 automatically compiled into the equivalent of if ([object count] > 4118 0) while object.count = 0; is automatically compiled into the 4119 equivalent ot [object setCount: 0];. The dot-syntax can be used 4120 with instance and class objects and with any setters or getters, no 4121 matter if they are part of a declared property or not. 4122 * Objective-C 2.0 declared properties are now supported. They are 4123 declared using the new @property keyword, and are most commonly 4124 used in conjunction with the new Objective-C 2.0 dot-syntax. The 4125 nonatomic, readonly, readwrite, assign, retain, copy, setter and 4126 getter attributes are all supported. Marking declared properties 4127 with __attribute__ ((deprecated)) is supported too. 4128 * The Objective-C 2.0 @synthesize and @dynamic keywords are 4129 supported. @synthesize causes the compiler to automatically 4130 synthesize a declared property, while @dynamic is used to disable 4131 all warnings for a declared property for which no implementation is 4132 provided at compile time. Synthesizing declared properties requires 4133 runtime support in most useful cases; to be able to use it with the 4134 GNU runtime, appropriate helper functions have been added to the 4135 GNU Objective-C runtime ABI, and are implemented by the GNU 4136 Objective-C runtime library shipped with GCC. 4137 * The Objective-C 2.0 fast enumeration syntax is supported in 4138 Objective-C. This is currently not yet available in Objective-C++. 4139 Fast enumeration requires support in the runtime, and such support 4140 has been added to the GNU Objective-C runtime library (shipped with 4141 GCC). 4142 * The Objective-C 2.0 @optional keyword is supported. It allows you 4143 to mark methods or properties in a protocol as optional as opposed 4144 to required. 4145 * The Objective-C 2.0 @package keyword is supported. It has currently 4146 the same effect as the @public keyword. 4147 * Objective-C 2.0 method attributes are supported. Currently the 4148 supported attributes are deprecated, sentinel, noreturn and format. 4149 * Objective-C 2.0 method argument attributes are supported. The most 4150 widely used attribute is unused, to mark an argument as unused in 4151 the implementation. 4152 * Objective-C 2.0 class and protocol attributes are supported. 4153 Currently the only supported attribute is deprecated. 4154 * Objective-C 2.0 class extensions are supported. A class extension 4155 has the same syntax as a category declaration with no category 4156 name, and the methods and properties declared in it are added 4157 directly to the main class. It is mostly used as an alternative to 4158 a category to add methods to a class without advertising them in 4159 the public headers, with the advantage that for class extensions 4160 the compiler checks that all the privately declared methods are 4161 actually implemented. 4162 * As a result of these enhancements, GCC can now be used to build 4163 Objective-C and Objective-C++ software that uses Foundation and 4164 other important system frameworks with the NeXT runtime on Darwin 9 4165 and Darwin 10 (OSX 10.5 and 10.6). 4166 * Many bugs in the compiler have been fixed in this release; in 4167 particular, LTO can now be used when compiling Objective-C and 4168 Objective-C++ and the parser is much more robust in dealing with 4169 invalid code. 4170 4171 Runtime Library (libobjc) 4172 4173 * The GNU Objective-C runtime library now defines the macro 4174 __GNU_LIBOBJC__ (with a value that is increased at every release 4175 where there is any change to the API) in objc/objc.h, making it 4176 easy to determine if the GNU Objective-C runtime library is being 4177 used, and if so, which version. Previous versions of the GNU 4178 Objective-C runtime library (and other Objective-C runtime 4179 libraries such as the Apple one) do not define this macro. 4180 * A new Objective-C 2.0 API, almost identical to the one implemented 4181 by the Apple Objective-C runtime, has been implemented in the GNU 4182 Objective-C runtime library. The new API hides the internals of 4183 most runtime structures but provides a more extensive set of 4184 functions to operate on them. It is much easier, for example, to 4185 create or modify classes at runtime. The new API also makes it 4186 easier to port software from Apple to GNU as almost no changes 4187 should be required. The old API is still supported for backwards 4188 compatibility; including the old objc/objc-api.h header file 4189 automatically selects the old API, while including the new 4190 objc/runtime.h header file automatically selects the new API. 4191 Support for the old API is being phased out and upgrading the 4192 software to use the new API is strongly recommended. To check for 4193 the availability of the new API, the __GNU_LIBOBJC__ macro can be 4194 used as older versions of the GNU Objective-C runtime library, 4195 which do not support the new API, do not define such a macro. 4196 * Runtime support for @synchronized has been added. 4197 * Runtime support for Objective-C 2.0 synthesized property accessors 4198 has been added. 4199 * Runtime support for Objective-C 2.0 fast enumeration has been 4200 added. 4201 4202New Targets and Target Specific Improvements 4203 4204 ARM 4205 4206 * GCC now supports the Cortex-M4 processor implementing the v7-em 4207 version of the architecture using the option -mcpu=cortex-m4. 4208 * Scheduling descriptions for the Cortex-M4, the Neon and the 4209 floating point units of the Cortex-A9 and a pipeline description 4210 for the Cortex-A5 have been added. 4211 * Synchronization primitives such as __sync_fetch_and_add and friends 4212 are now inlined for supported architectures rather than calling 4213 into a kernel helper function. 4214 * SSA loop prefetching is enabled by default for the Cortex-A9 at 4215 -O3. 4216 * Several improvements were committed to improve code generation for 4217 the ARM architecture including a rewritten implementation for load 4218 and store multiples. 4219 * Several enhancements were committed to improve SIMD code generation 4220 for NEON by adding support for widening instructions, misaligned 4221 loads and stores, vector conditionals and support for 64 bit 4222 arithmetic. 4223 * Support was added for the Faraday cores fa526, fa606te, fa626te, 4224 fmp626te, fmp626 and fa726te and can be used with the respective 4225 names as parameters to the -mcpu= option. 4226 * Basic support was added for Cortex-A15 and is available through 4227 -mcpu=cortex-a15. 4228 * GCC for AAPCS configurations now more closely adheres to the AAPCS 4229 specification by enabling -fstrict-volatile-bitfields by default. 4230 4231 IA-32/x86-64 4232 4233 * The new -fsplit-stack option permits programs to use a 4234 discontiguous stack. This is useful for threaded programs, in that 4235 it is no longer necessary to specify the maximum stack size when 4236 creating a thread. This feature is currently only implemented for 4237 32-bit and 64-bit x86 GNU/Linux targets. 4238 * Support for emitting profiler counter calls before function 4239 prologues. This is enabled via a new command-line option -mfentry. 4240 * Optimization for the Intel Core 2 processors is now available 4241 through the -march=core2 and -mtune=core2 options. 4242 * Support for Intel Core i3/i5/i7 processors is now available through 4243 the -march=corei7 and -mtune=corei7 options. 4244 * Support for Intel Core i3/i5/i7 processors with AVX is now 4245 available through the -march=corei7-avx and -mtune=corei7-avx 4246 options. 4247 * Support for AMD Bobcat (family 14) processors is now available 4248 through the -march=btver1 and -mtune=btver1 options. 4249 * Support for AMD Bulldozer (family 15) processors is now available 4250 through the -march=bdver1 and -mtune=bdver1 options. 4251 * The default setting (when not optimizing for size) for 32-bit 4252 GNU/Linux and Darwin x86 targets has been changed to 4253 -fomit-frame-pointer. The default can be reverted to 4254 -fno-omit-frame-pointer by configuring GCC with the 4255 --enable-frame-pointer configure option. 4256 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 4257 __float128 on 32-bit and 64-bit x86 targets. 4258 * AVX floating-point arithmetic can now be enabled by default at 4259 configure time with the new --with-fpmath=avx option. 4260 * The SSA loop prefetching pass is enabled when using -O3 when 4261 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 4262 than K6). 4263 * Support for TBM (Trailing Bit Manipulation) built-in functions and 4264 code generation is available via -mtbm. 4265 * Support for AMD's BMI (Bit Manipulation) built-in functions and 4266 code generation is available via -mbmi. 4267 4268 MicroBlaze 4269 4270 * Support has been added for the Xilinx MicroBlaze softcore processor 4271 (microblaze-elf) embedded target. This configurable processor is 4272 supported on several Xilinx Spartan and Virtex FPGAs. 4273 4274 MIPS 4275 4276 * GCC now supports the Loongson 3A processor. Its canonical -march= 4277 and -mtune= name is loongson3a. 4278 4279 MN10300 / AM33 4280 4281 * The inline assembly register constraint "A" has been renamed "c". 4282 This constraint is used to select a floating-point register that 4283 can be used as the destination of a multiply-accumulate 4284 instruction. 4285 * New inline assembly register constraints "A" and "D" have been 4286 added. These constraint letters resolve to all general registers 4287 when compiling for AM33, and resolve to address registers only or 4288 data registers only when compiling for MN10300. 4289 * The MDR register is represented in the compiler. One can access the 4290 register via the "z" constraint in inline assembly. It can be 4291 marked as clobbered or used as a local register variable via the 4292 "mdr" name. The compiler uses the RETF instruction if the function 4293 does not modify the MDR register, so it is important that inline 4294 assembly properly annotate any usage of the register. 4295 4296 PowerPC/PowerPC64 4297 4298 * GCC now supports the Applied Micro Titan processor with 4299 -mcpu=titan. 4300 * The -mrecip option has been added, which indicates whether the 4301 reciprocal and reciprocal square root instructions should be used. 4302 * The -mveclibabi=mass option can be used to enable the compiler to 4303 autovectorize mathematical functions using the Mathematical 4304 Acceleration Subsystem library. 4305 * The -msingle-pic-base option has been added, which instructs the 4306 compiler to avoid loading the PIC base register in function 4307 prologues. The PIC base register must be initialized by the runtime 4308 system. 4309 * The -mblock-move-inline-limit option has been added, which enables 4310 the user to control the maximum size of inlined memcpy calls and 4311 similar. 4312 * PowerPC64 GNU/Linux support for applications requiring a large TOC 4313 section has been improved. A new command-line option, 4314 -mcmodel=MODEL, controls this feature; valid values for MODEL are 4315 small, medium, or large. 4316 * The Altivec builtin functions vec_ld and vec_st have been modified 4317 to generate the Altivec memory instructions LVX and STVX, even if 4318 the -mvsx option is used. In the initial GCC 4.5 release, these 4319 builtin functions were changed to generate VSX memory reference 4320 instructions instead of Altivec memory instructions, but there are 4321 differences between the two instructions. If the VSX instruction 4322 set is available, you can now use the new builtin functions 4323 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 4324 instructions. 4325 * The GCC compiler on AIX now defaults to a process layout with a 4326 larger data space allowing larger programs to be compiled. 4327 * The GCC long double type on AIX 6.1 and above has reverted to 64 4328 bit double precision, matching the AIX XL compiler default, because 4329 of missing C99 symbols required by the GCC runtime. 4330 * The default processor scheduling model and tuning for PowerPC64 4331 GNU/Linux and for AIX 6.1 and above now is POWER7. 4332 * Starting with GCC 4.6.1, vectors of type vector long long or vector 4333 long are passed and returned in the same method as other vectors 4334 with the VSX instruction set. Previously the GCC compiler did not 4335 adhere to the ABI for 128-bit vectors with 64-bit integer base 4336 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 4337 4338 S/390, zSeries and System z9/z10, IBM zEnterprise z196 4339 4340 * Support for the zEnterprise z196 processor has been added. When 4341 using the -march=z196 option, the compiler will generate code 4342 making use of the following instruction facilities: 4343 + Conditional load/store 4344 + Distinct-operands 4345 + Floating-point-extension 4346 + Interlocked-access 4347 + Population-count 4348 The -mtune=z196 option avoids the compare and branch instructions 4349 as well as the load address instruction with an index register as 4350 much as possible and performs instruction scheduling appropriate 4351 for the new out-of-order pipeline architecture. 4352 * When using the -m31 -mzarch options the generated code still 4353 conforms to the 32-bit ABI but uses the general purpose registers 4354 as 64-bit registers internally. This requires a Linux kernel saving 4355 the whole 64-bit registers when doing a context switch. Kernels 4356 providing that feature indicate that by the 'highgprs' string in 4357 /proc/cpuinfo. 4358 * The SSA loop prefetching pass is enabled when using -O3. 4359 4360 SPARC 4361 4362 * GCC now supports the LEON series of SPARC V8 processors. The code 4363 generated by the compiler can either be tuned to it by means of the 4364 --with-tune=leon configure option and -mtune=leon compilation 4365 option, or the compiler can be built for the sparc-leon-{elf,linux} 4366 and sparc-leon3-{elf,linux} targets directly. 4367 * GCC has stopped sign/zero-extending parameter registers in the 4368 callee for functions taking parameters with sub-word size in 32-bit 4369 mode, since this is redundant with the specification of the ABI. 4370 GCC has never done so in 64-bit mode since this is also redundant. 4371 * The command line option -mfix-at697f has been added to enable the 4372 documented workaround for the single erratum of the Atmel AT697F 4373 processor. 4374 4375Operating Systems 4376 4377 Android 4378 4379 * GCC now supports the Bionic C library and provides a convenient way 4380 of building native libraries and applications for the Android 4381 platform. Refer to the documentation of the -mandroid and -mbionic 4382 options for details on building native code. At the moment, Android 4383 support is enabled only for ARM. 4384 4385 Darwin/Mac OS X 4386 4387 * General 4388 + Initial support for CFString types has been added. 4389 This allows GCC to build projects including the system Core 4390 Foundation frameworks. The GCC Objective-C family supports 4391 CFString "toll-free bridged" as per the Mac OS X system tools. 4392 CFString is also recognized in the context of format 4393 attributes and arguments (see the documentation for format 4394 attributes for limitations). At present, 8-bit character types 4395 are supported. 4396 + Object file size reduction. 4397 The Darwin zeroed memory allocators have been re-written to 4398 make more use of .zerofill sections. For non-debug code, this 4399 can reduce object file size significantly. 4400 + Objective-C family 64-bit support (NeXT ABI 2). 4401 Initial support has been added to support 64-bit Objective-C 4402 code using the Darwin/OS X native (NeXT) runtime. ABI version 4403 2 will be selected automatically when 64-bit code is built. 4404 + Objective-C family 32-bit ABI 1. 4405 For 32-bit code ABI 1 is also now also allowed. At present it 4406 must be selected manually using -fobjc-abi-version=1 where 4407 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 4408 * x86 Architecture 4409 + The -mdynamic-no-pic option has been enabled. 4410 Code supporting -mdynamic-no-pic optimization has been added 4411 and is applicable to -m32 builds. The compiler bootstrap uses 4412 the option where appropriate. 4413 + The default value for -mtune= has been changed. 4414 Since Darwin systems are primarily Xeon, Core-2 or similar the 4415 default tuning has been changed to -mtune=core2. 4416 + Enable 128-bit long double (__float128) support on Darwin. 4417 * PPC Architecture 4418 + Darwin64 ABI. 4419 Several significant bugs have been fixed, such that GCC now 4420 produces code compatible with the Darwin64 PowerPC ABI. 4421 + libffi and boehm-gc. 4422 The Darwin ports of the libffi and boehm-gc libraries have 4423 been upgraded to include a Darwin64 implementation. This means 4424 that powerpc*-*-darwin9 platforms may now, for example, build 4425 Java applications with -m64 enabled. 4426 + Plug-in support has been enabled. 4427 + The -fsection-anchors option is now available although, 4428 presently, not heavily tested. 4429 4430 Solaris 2 4431 4432 New Features 4433 4434 * Support symbol versioning with the Sun linker. 4435 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 4436 * Support thread-local storage (TLS) with the Sun assembler on 4437 Solaris 2/x86. 4438 * Support TLS on Solaris 8/9 if prerequisites are met. 4439 * Support COMDAT group with the GNU assembler and recent Sun linker. 4440 * Support the Sun assembler visibility syntax. 4441 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 4442 -march=pentiumpro (Solaris 8/9). 4443 * Don't use SSE on Solaris 8/9 x86 by default. 4444 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 4445 4446 ABI Change 4447 4448 * Change the ABI for returning 8-byte vectors like __m64 in MMX 4449 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 4450 compilers. This is an incompatible change. If you use such types, 4451 you must either recompile all your code with the new compiler or 4452 use the new -mvect8-ret-in-mem option to remain compatible with 4453 previous versions of GCC and Sun Studio. 4454 4455 Windows x86/x86_64 4456 4457 * Initial support for decimal floating point. 4458 * Support for the __thiscall calling-convention. 4459 * Support for hot-patchable function prologues via the 4460 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 4461 * Improvements of stack-probing and stack-allocation mechanisms. 4462 * Support of push/pop-macro pragma as preprocessor command. 4463 With #pragma push_macro("macro-name") the current definition of 4464 macro-name is saved and can be restored with #pragma 4465 pop_macro("macro-name") to its saved definition. 4466 * Enable 128-bit long double (__float128) support on MinGW and 4467 Cygwin. 4468 4469Other significant improvements 4470 4471 Installation changes 4472 4473 * An install-strip make target is provided that installs stripped 4474 executables, and may install libraries with unneeded or debugging 4475 sections stripped. 4476 * On Power7 systems, there is a potential problem if you build the 4477 GCC compiler with a host compiler using options that enable the VSX 4478 instruction set generation. If the host compiler has been patched 4479 so that the vec_ld and vec_st builtin functions generate Altivec 4480 memory instructions instead of VSX memory instructions, then you 4481 should be able to build the compiler with VSX instruction 4482 generation. 4483 4484Changes for GCC Developers 4485 4486 Note: these changes concern developers that develop GCC itself or 4487 software that integrates with GCC, such as plugins, and not the general 4488 GCC users. 4489 * The gengtype utility, which previously was internal to the GCC 4490 build process, has been enchanced to provide GC root information 4491 for plugins as necessary. 4492 * The old GC allocation interface of ggc_alloc and friends was 4493 replaced with a type-safe alternative. 4494 4495GCC 4.6.1 4496 4497 This is the [20]list of problem reports (PRs) from GCC's bug tracking 4498 system that are known to be fixed in the 4.6.1 release. This list might 4499 not be complete (that is, it is possible that some PRs that have been 4500 fixed are not listed here). 4501 4502GCC 4.6.2 4503 4504 This is the [21]list of problem reports (PRs) from GCC's bug tracking 4505 system that are known to be fixed in the 4.6.2 release. This list might 4506 not be complete (that is, it is possible that some PRs that have been 4507 fixed are not listed here). 4508 4509GCC 4.6.3 4510 4511 This is the [22]list of problem reports (PRs) from GCC's bug tracking 4512 system that are known to be fixed in the 4.6.3 release. This list might 4513 not be complete (that is, it is possible that some PRs that have been 4514 fixed are not listed here). 4515 4516GCC 4.6.4 4517 4518 This is the [23]list of problem reports (PRs) from GCC's bug tracking 4519 system that are known to be fixed in the 4.6.4 release. This list might 4520 not be complete (that is, it is possible that some PRs that have been 4521 fixed are not listed here). 4522 4523 4524 For questions related to the use of GCC, please consult these web 4525 pages and the [24]GCC manuals. If that fails, the 4526 [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these 4527 web pages and the development of GCC are welcome on our developer 4528 list at [26]gcc@gcc.gnu.org. All of [27]our lists have public 4529 archives. 4530 4531 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 4532 distribution of this entire article is permitted in any medium, 4533 provided this notice is preserved. 4534 4535 These pages are [29]maintained by the GCC team. Last modified 4536 2016-05-28[30]. 4537 4538References 4539 4540 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 4541 2. http://gcc.gnu.org/gcc-4.5/changes.html 4542 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 4543 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 4544 5. http://gcc.gnu.org/projects/lto/whopr.pdf 4545 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 4546 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 4547 8. https://gcc.gnu.org/PR43145 4548 9. https://gcc.gnu.org/PR43680 4549 10. https://gcc.gnu.org/PR33558 4550 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 4551 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 4552 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 4553 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 4554 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 4555 16. https://gcc.gnu.org/wiki/OOP 4556 17. https://gcc.gnu.org/wiki/Coarray 4557 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 4558 19. https://golang.org/ 4559 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 4560 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 4561 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 4562 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 4563 24. https://gcc.gnu.org/onlinedocs/ 4564 25. mailto:gcc-help@gcc.gnu.org 4565 26. mailto:gcc@gcc.gnu.org 4566 27. https://gcc.gnu.org/lists.html 4567 28. http://www.fsf.org/ 4568 29. https://gcc.gnu.org/about.html 4569 30. http://validator.w3.org/check/referer 4570====================================================================== 4571http://gcc.gnu.org/gcc-4.5/index.html 4572 4573 GCC 4.5 Release Series 4574 4575 Jul 2, 2012 4576 4577 The [1]GNU project and the GCC developers are pleased to announce the 4578 release of GCC 4.5.4. 4579 4580 This release is a bug-fix release, containing fixes for regressions in 4581 GCC 4.5.3 relative to previous releases of GCC. 4582 4583Release History 4584 4585 GCC 4.5.4 4586 Jul 2, 2012 ([2]changes) 4587 4588 GCC 4.5.3 4589 Apr 28, 2011 ([3]changes) 4590 4591 GCC 4.5.2 4592 Dec 16, 2010 ([4]changes) 4593 4594 GCC 4.5.1 4595 Jul 31, 2010 ([5]changes) 4596 4597 GCC 4.5.0 4598 April 14, 2010 ([6]changes) 4599 4600References and Acknowledgements 4601 4602 GCC used to stand for the GNU C Compiler, but since the compiler 4603 supports several other languages aside from C, it now stands for the 4604 GNU Compiler Collection. 4605 4606 A list of [7]successful builds is updated as new information becomes 4607 available. 4608 4609 The GCC developers would like to thank the numerous people that have 4610 contributed new features, improvements, bug fixes, and other changes as 4611 well as test results to GCC. This [8]amazing group of volunteers is 4612 what makes GCC successful. 4613 4614 For additional information about GCC please refer to the [9]GCC project 4615 web site or contact the [10]GCC development mailing list. 4616 4617 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 4618 4619 4620 For questions related to the use of GCC, please consult these web 4621 pages and the [13]GCC manuals. If that fails, the 4622 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 4623 web pages and the development of GCC are welcome on our developer 4624 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 4625 archives. 4626 4627 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 4628 distribution of this entire article is permitted in any medium, 4629 provided this notice is preserved. 4630 4631 These pages are [18]maintained by the GCC team. Last modified 4632 2016-01-30[19]. 4633 4634References 4635 4636 1. http://www.gnu.org/ 4637 2. http://gcc.gnu.org/gcc-4.5/changes.html 4638 3. http://gcc.gnu.org/gcc-4.5/changes.html 4639 4. http://gcc.gnu.org/gcc-4.5/changes.html 4640 5. http://gcc.gnu.org/gcc-4.5/changes.html 4641 6. http://gcc.gnu.org/gcc-4.5/changes.html 4642 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 4643 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 4644 9. http://gcc.gnu.org/index.html 4645 10. mailto:gcc@gcc.gnu.org 4646 11. http://gcc.gnu.org/mirrors.html 4647 12. http://gcc.gnu.org/svn.html 4648 13. https://gcc.gnu.org/onlinedocs/ 4649 14. mailto:gcc-help@gcc.gnu.org 4650 15. mailto:gcc@gcc.gnu.org 4651 16. https://gcc.gnu.org/lists.html 4652 17. http://www.fsf.org/ 4653 18. https://gcc.gnu.org/about.html 4654 19. http://validator.w3.org/check/referer 4655====================================================================== 4656http://gcc.gnu.org/gcc-4.5/changes.html 4657 4658 GCC 4.5 Release Series 4659 Changes, New Features, and Fixes 4660 4661Caveats 4662 4663 * GCC now requires the [1]MPC library in order to build. See the 4664 [2]prerequisites page for version requirements. 4665 * Support for a number of older systems and recently unmaintained or 4666 untested target ports of GCC has been declared obsolete in GCC 4.5. 4667 Unless there is activity to revive them, the next release of GCC 4668 will have their sources permanently removed. 4669 The following ports for individual systems on particular 4670 architectures have been obsoleted: 4671 + IRIX releases before 6.5 (mips-sgi-irix5*, 4672 mips-sgi-irix6.[0-4]) 4673 + Solaris 7 (*-*-solaris2.7) 4674 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 4675 alpha-dec-osf5.0*) 4676 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 4677 can be found in the [3]announcement. 4678 Support for the classic POWER architecture implemented in the 4679 original RIOS and RIOS2 processors of the old IBM RS/6000 product 4680 line has been obsoleted in the rs6000 port. This does not affect 4681 the new generation Power and PowerPC architectures. 4682 * Support has been removed for all the [4]configurations obsoleted in 4683 GCC 4.4. 4684 * Support has been removed for the protoize and unprotoize utilities, 4685 obsoleted in GCC 4.4. 4686 * Support has been removed for tuning for Itanium1 (Merced) variants. 4687 Note that code tuned for Itanium2 should also run correctly on 4688 Itanium1. 4689 * GCC now generates unwind info also for epilogues. DWARF debuginfo 4690 generated by GCC now uses more features of DWARF3 than before, and 4691 also some DWARF4 features. GDB older than 7.0 is not able to handle 4692 either of these, so to debug GCC 4.5 generated binaries or 4693 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 4694 features with the -gdwarf-3 -gstrict-dwarf options, or use 4695 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 4696 epilogue unwind info is emitted unconditionally whenever unwind 4697 info is emitted. 4698 * On x86 targets, code containing floating-point calculations may run 4699 significantly slower when compiled with GCC 4.5 in strict C99 4700 conformance mode than they did with earlier GCC versions. This is 4701 due to stricter standard conformance of the compiler and can be 4702 avoided by using the option -fexcess-precision=fast; also see 4703 [5]below. 4704 * The function attribute noinline no longer prevents GCC from cloning 4705 the function. A new attribute noclone has been introduced for this 4706 purpose. Cloning a function means that it is duplicated and the new 4707 copy is specialized for certain contexts (for example when a 4708 parameter is a known constant). 4709 4710General Optimizer Improvements 4711 4712 * The -save-temps now takes an optional argument. The -save-temps and 4713 -save-temps=cwd switches write the temporary files in the current 4714 working directory based on the original source file. The 4715 -save-temps=obj switch will write files into the directory 4716 specified with the -o option, and the intermediate filenames are 4717 based on the output file. This will allow the user to get the 4718 compiler intermediate files when doing parallel builds without two 4719 builds of the same filename located in different directories from 4720 interfering with each other. 4721 * Debugging dumps are now created in the same directory as the object 4722 file rather than in the current working directory. This allows the 4723 user to get debugging dumps when doing parallel builds without two 4724 builds of the same filename interfering with each other. 4725 * GCC has been integrated with the [6]MPC library. This allows GCC to 4726 evaluate complex arithmetic at compile time [7]more accurately. It 4727 also allows GCC to evaluate calls to complex built-in math 4728 functions having constant arguments and replace them at compile 4729 time with their mathematically equivalent results. In doing so, GCC 4730 can generate correct results regardless of the math library 4731 implementation or floating point precision of the host platform. 4732 This also allows GCC to generate identical results regardless of 4733 whether one compiles in native or cross-compile configurations to a 4734 particular target. The following built-in functions take advantage 4735 of this new capability: cacos, cacosh, casin, casinh, catan, 4736 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 4737 and ctanh. The float and long double variants of these functions 4738 (e.g. csinf and csinl) are also handled. 4739 * A new link-time optimizer has been added ([8]-flto). When this 4740 option is used, GCC generates a bytecode representation of each 4741 input file and writes it to specially-named sections in each object 4742 file. When the object files are linked together, all the function 4743 bodies are read from these named sections and instantiated as if 4744 they had been part of the same translation unit. This enables 4745 interprocedural optimizations to work across different files (and 4746 even different languages), potentially improving the performance of 4747 the generated code. To use the link-timer optimizer, -flto needs to 4748 be specified at compile time and during the final link. If the 4749 program does not require any symbols to be exported, it is possible 4750 to combine -flto and the experimental [9]-fwhopr with 4751 [10]-fwhole-program to allow the interprocedural optimizers to use 4752 more aggressive assumptions. 4753 * The automatic parallelization pass was enhanced to support 4754 parallelization of outer loops. 4755 * Automatic parallelization can be enabled as part of Graphite. In 4756 addition to -ftree-parallelize-loops=, specify 4757 -floop-parallelize-all to enable the Graphite-based optimization. 4758 * The infrastructure for optimizing based on [11]restrict qualified 4759 pointers has been rewritten and should result in code generation 4760 improvements. Optimizations based on restrict qualified pointers 4761 are now also available when using -fno-strict-aliasing. 4762 * There is a new optimization pass that attempts to change prototype 4763 of functions to avoid unused parameters, pass only relevant parts 4764 of structures and turn arguments passed by reference to arguments 4765 passed by value when possible. It is enabled by -O2 and above as 4766 well as -Os and can be manually invoked using the new command-line 4767 switch -fipa-sra. 4768 * GCC now optimize exception handling code. In particular cleanup 4769 regions that are proved to not have any effect are optimized out. 4770 4771New Languages and Language specific improvements 4772 4773 All languages 4774 4775 * The -fshow-column option is now on by default. This means error 4776 messages now have a column associated with them. 4777 4778 Ada 4779 4780 * Compilation of programs heavily using discriminated record types 4781 with variant parts has been sped up and generates more compact 4782 code. 4783 * Stack checking now works reasonably well on most plaforms. In some 4784 specific cases, stack overflows may still fail to be detected, but 4785 a compile-time warning will be issued for these cases. 4786 4787 C family 4788 4789 * If a header named in a #include directive is not found, the 4790 compiler exits immediately. This avoids a cascade of errors arising 4791 from declarations expected to be found in that header being 4792 missing. 4793 * A new built-in function __builtin_unreachable() has been added that 4794 tells the compiler that control will never reach that point. It may 4795 be used after asm statements that terminate by transferring control 4796 elsewhere, and in other places that are known to be unreachable. 4797 * The -Wlogical-op option now warns for logical expressions such as 4798 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 4799 mistakes. This option is disabled by default. 4800 * An asm goto feature has been added to allow asm statements that 4801 jump to C labels. 4802 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 4803 * The deprecated attribute now takes an optional string argument, for 4804 example, __attribute__((deprecated("text string"))), that will be 4805 printed together with the deprecation warning. 4806 4807 C 4808 4809 * The -Wenum-compare option, which warns when comparing values of 4810 different enum types, now works for C. It formerly only worked for 4811 C++. This warning is enabled by -Wall. It may be avoided by using a 4812 type cast. 4813 * The -Wcast-qual option now warns about casts which are unsafe in 4814 that they permit const-correctness to be violated without further 4815 warnings. Specifically, it warns about cases where a qualifier is 4816 added when all the lower types are not const. For example, it warns 4817 about a cast from char ** to const char **. 4818 * The -Wc++-compat option is significantly improved. It issues new 4819 warnings for: 4820 + Using C++ reserved operator names as identifiers. 4821 + Conversions to enum types without explicit casts. 4822 + Using va_arg with an enum type. 4823 + Using different enum types in the two branches of ?:. 4824 + Using ++ or -- on a variable of enum type. 4825 + Using the same name as both a struct, union or enum tag and a 4826 typedef, unless the typedef refers to the tagged type itself. 4827 + Using a struct, union, or enum which is defined within another 4828 struct or union. 4829 + A struct field defined using a typedef if there is a field in 4830 the struct, or an enclosing struct, whose name is the typedef 4831 name. 4832 + Duplicate definitions at file scope. 4833 + Uninitialized const variables. 4834 + A global variable with an anonymous struct, union, or enum 4835 type. 4836 + Using a string constant to initialize a char array whose size 4837 is the length of the string. 4838 * The new -Wjump-misses-init option warns about cases where a goto or 4839 switch skips the initialization of a variable. This sort of branch 4840 is an error in C++ but not in C. This warning is enabled by 4841 -Wc++-compat. 4842 * GCC now ensures that a C99-conforming <stdint.h> is present on most 4843 targets, and uses information about the types in this header to 4844 implement the Fortran bindings to those types. GCC does not ensure 4845 the presence of such a header, and does not implement the Fortran 4846 bindings, on the following targets: NetBSD, VxWorks, VMS, 4847 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 4848 * GCC now implements C90- and C99-conforming rules for constant 4849 expressions. This may cause warnings or errors for some code using 4850 expressions that can be folded to a constant but are not constant 4851 expressions as defined by ISO C. 4852 * All known target-independent C90 and C90 Amendment 1 conformance 4853 bugs, and all known target-independent C99 conformance bugs not 4854 related to floating point or extended identifiers, have been fixed. 4855 * The C decimal floating point support now includes support for the 4856 FLOAT_CONST_DECIMAL64 pragma. 4857 * The named address space feature from ISO/IEC TR 18037 is now 4858 supported. This is currently only implemented for the SPU 4859 processor. 4860 4861 C++ 4862 4863 * Improved [12]experimental support for the upcoming C++0x ISO C++ 4864 standard, including support for raw strings, lambda expressions and 4865 explicit type conversion operators. 4866 * When printing the name of a class template specialization, G++ will 4867 now omit any template arguments which come from default template 4868 arguments. This behavior (and the pretty-printing of function 4869 template specializations as template signature and arguments) can 4870 be disabled with the -fno-pretty-templates option. 4871 * Access control is now applied to typedef names used in a template, 4872 which may cause G++ to reject some ill-formed code that was 4873 accepted by earlier releases. The -fno-access-control option can be 4874 used as a temporary workaround until the code is corrected. 4875 * Compilation time for code that uses templates should now scale 4876 linearly with the number of instantiations rather than 4877 quadratically, as template instantiations are now looked up using 4878 hash tables. 4879 * Declarations of functions that look like builtin declarations of 4880 library functions are only considered to be redeclarations if they 4881 are declared with extern "C". This may cause problems with code 4882 that omits extern "C" on hand-written declarations of C library 4883 functions such as abort or memcpy. Such code is ill-formed, but was 4884 accepted by earlier releases. 4885 * Diagnostics that used to complain about passing non-POD types to 4886 ... or jumping past the declaration of a non-POD variable now check 4887 for triviality rather than PODness, as per C++0x. 4888 * In C++0x mode local and anonymous classes are now allowed as 4889 template arguments, and in declarations of variables and functions 4890 with linkage, so long as any such declaration that is used is also 4891 defined ([13]DR 757). 4892 * Labels may now have attributes, as has been permitted for a while 4893 in C. This is only permitted when the label definition and the 4894 attribute specifier is followed by a semicolon--i.e., the label 4895 applies to an empty statement. The only useful attribute for a 4896 label is unused. 4897 * G++ now implements [14]DR 176. Previously G++ did not support using 4898 the injected-class-name of a template base class as a type name, 4899 and lookup of the name found the declaration of the template in the 4900 enclosing scope. Now lookup of the name finds the 4901 injected-class-name, which can be used either as a type or as a 4902 template, depending on whether or not the name is followed by a 4903 template argument list. As a result of this change, some code that 4904 was previously accepted may be ill-formed because 4905 1. The injected-class-name is not accessible because it's from a 4906 private base, or 4907 2. The injected-class-name cannot be used as an argument for a 4908 template template parameter. 4909 In either of these cases, the code can be fixed by adding a 4910 nested-name-specifier to explicitly name the template. The first 4911 can be worked around with -fno-access-control; the second is only 4912 rejected with -pedantic. 4913 * A new standard mangling for SIMD vector types has been added, to 4914 avoid name clashes on systems with vectors of varying length. By 4915 default the compiler still uses the old mangling, but emits aliases 4916 with the new mangling on targets that support strong aliases. Users 4917 can switch over entirely to the new mangling with -fabi-version=4 4918 or -fabi-version=0. -Wabi will now warn about code that uses the 4919 old mangling. 4920 * The command-line option -ftemplate-depth-N is now written as 4921 -ftemplate-depth=N and the old form is deprecated. 4922 * Conversions between NULL and non-pointer types are now warned by 4923 default. The new option -Wno-conversion-null disables these 4924 warnings. Previously these warnings were only available when using 4925 -Wconversion explicitly. 4926 4927 Runtime Library (libstdc++) 4928 4929 * Improved experimental support for the upcoming ISO C++ standard, 4930 C++0x, including: 4931 + Support for <future>, <functional>, and <random>. 4932 + Existing facilities now exploit explicit operators and the 4933 newly implemented core C++0x features. 4934 + The header <cstdatomic> has been renamed to <atomic>. 4935 * An experimental [15]profile mode has been added. This is an 4936 implementation of many C++ standard library constructs with an 4937 additional analysis layer that gives performance improvement advice 4938 based on recognition of suboptimal usage patterns. For example, 4939#include <vector> 4940int main() 4941{ 4942 std::vector<int> v; 4943 for (int k = 0; k < 1024; ++k) 4944 v.insert(v.begin(), k); 4945} 4946 4947 When instrumented via the profile mode, can return suggestions 4948 about the initial size and choice of the container used as follows: 4949vector-to-list: improvement = 5: call stack = 0x804842c ... 4950 : advice = change std::vector to std::list 4951vector-size: improvement = 3: call stack = 0x804842c ... 4952 : advice = change initial container size from 0 to 1024 4953 4954 These constructs can be substituted for the normal libstdc++ 4955 constructs on a piecemeal basis, or all existing components can be 4956 transformed via the -D_GLIBCXX_PROFILE macro. 4957 * [16]Support for decimal floating-point arithmetic (aka ISO C++ TR 4958 24733) has been added. This support is in header file 4959 <decimal/decimal>, uses namespace std::decimal, and includes 4960 classes decimal32, decimal64, and decimal128. 4961 * Sources have been audited for application of function attributes 4962 nothrow, const, pure, and noreturn. 4963 * Python pretty-printers have been added for many standard library 4964 components that simplify the internal representation and present a 4965 more intuitive view of components when used with 4966 appropriately-advanced versions of GDB. For more information, 4967 please consult the more [17]detailed description. 4968 * The default behavior for comparing typeinfo names has changed, so 4969 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 4970 * The new -static-libstdc++ option directs g++ to link the C++ 4971 library statically, even if the default would normally be to link 4972 it dynamically. 4973 4974 Fortran 4975 4976 * The COMMON default padding has been changed - instead of adding the 4977 padding before a variable it is now added afterwards, which 4978 increases the compatibility with other vendors and helps to obtain 4979 the correct output in some cases. Cf. also the -falign-commons 4980 option ([18]added in 4.4). 4981 * The -finit-real= option now also supports the value snan for 4982 signaling not-a-number; to be effective, one additionally needs to 4983 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 4984 optimizations can turn a signaling NaN into a quiet one. 4985 * The new option -fcheck= has been added with the options bounds, 4986 array-temps, do, pointer, and recursive. The bounds and array-temps 4987 options are equivalent to -fbounds-check and 4988 -fcheck-array-temporaries. The do option checks for invalid 4989 modification of loop iteration variables, and the recursive option 4990 tests for recursive calls to subroutines/functions which are not 4991 marked as recursive. With pointer pointer association checks in 4992 calls are performed; however, neither undefined pointers nor 4993 pointers in expressions are handled. Using -fcheck=all enables all 4994 these run-time checks. 4995 * The run-time checking -fcheck=bounds now warns about invalid string 4996 lengths of character dummy arguments. Additionally, more 4997 compile-time checks have been added. 4998 * The new option [19]-fno-protect-parens has been added; if set, the 4999 compiler may reorder REAL and COMPLEX expressions without regard to 5000 parentheses. 5001 * GNU Fortran no longer links against libgfortranbegin. As before, 5002 MAIN__ (assembler symbol name) is the actual Fortran main program, 5003 which is invoked by the main function. However, main is now 5004 generated and put in the same object file as MAIN__. For the time 5005 being, libgfortranbegin still exists for backward compatibility. 5006 For details see the new [20]Mixed-Language Programming chapter in 5007 the manual. 5008 * The I/O library was restructured for performance and cleaner code. 5009 * Array assignments and WHERE are now run in parallel when OpenMP's 5010 WORKSHARE is used. 5011 * The experimental option -fwhole-file was added. The option allows 5012 whole-file checking of procedure arguments and allows for better 5013 optimizations. It can also be used with -fwhole-program, which is 5014 now also supported in gfortran. 5015 * More Fortran 2003 and Fortran 2008 mathematical functions can now 5016 be used as initialization expressions. 5017 * Some extended attributes such as STDCALL are now supported via the 5018 [21]GCC$ compiler directive. 5019 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 5020 intrinsic behaves now as if zero were always positive. 5021 * For legacy compatibiliy: On Cygwin and MinGW, the special files 5022 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 5023 supported. 5024 * Fortran 2003 support has been extended: 5025 + Procedure-pointer function results and procedure-pointer 5026 components (including PASS), 5027 + allocatable scalars (experimental), 5028 + DEFERRED type-bound procedures, 5029 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 5030 have been implemented. 5031 + The ALLOCATE statement supports type-specs and the SOURCE= 5032 argument. 5033 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 5034 type-bound procedure (i.e. as type-bound operators). 5035 + Rounding (ROUND=, RZ, ...) for output is now supported. 5036 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 5037 intrinsic module ISO_C_BINDING are now supported, except for 5038 the targets listed above as ones where GCC does not have 5039 <stdint.h> type information. 5040 + Extensible derived types with type-bound procedure or 5041 procedure pointer with PASS attribute now have to use CLASS in 5042 line with the Fortran 2003 standard; the workaround to use 5043 TYPE is no longer supported. 5044 + [22]Experimental, incomplete support for polymorphism, 5045 including CLASS, SELECT TYPE and dynamic dispatch of 5046 type-bound procedure calls. Some features do not work yet such 5047 as unlimited polymorphism (CLASS(*)). 5048 * Fortran 2008 support has been extended: 5049 + The OPEN statement now supports the NEWUNIT= option, which 5050 returns a unique file unit, thus preventing inadvertent use of 5051 the same unit in different parts of the program. 5052 + Support for unlimited format items has been added. 5053 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 5054 the intrinsic module ISO_FORTRAN_ENV are now supported. 5055 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 5056 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 5057 and ATANH have been added (for real and complex arguments) and 5058 ATAN(Y,X) is now an alias for ATAN2(Y,X). 5059 + The BLOCK construct has been implemented. 5060 5061New Targets and Target Specific Improvements 5062 5063 AIX 5064 5065 * Full cross-toolchain support now available with GNU Binutils 5066 5067 ARM 5068 5069 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 5070 * GCC now supports the ARM v7E-M architecture. 5071 * GCC now supports VFPv4-based FPUs and FPUs with 5072 single-precision-only VFP. 5073 * GCC has many improvements to optimization for other ARM processors, 5074 including scheduling support for the integer pipeline on Cortex-A9. 5075 * GCC now supports the IEEE 754-2008 half-precision floating-point 5076 type, and a variant ARM-specific half-precision type. This type is 5077 specified using __fp16, with the layout determined by 5078 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 5079 VFPv4 half-precision instructions will be used. 5080 * GCC now supports the variant of AAPCS that uses VFP registers for 5081 parameter passing and return values. 5082 5083 AVR 5084 5085 * The -mno-tablejump option has been removed because it has the same 5086 effect as the -fno-jump-tables option. 5087 * Added support for these new AVR devices: 5088 + ATmega8U2 5089 + ATmega16U2 5090 + ATmega32U2 5091 5092 IA-32/x86-64 5093 5094 * GCC now will set the default for -march= based on the configure 5095 target. 5096 * GCC now supports handling floating-point excess precision arising 5097 from use of the x87 floating-point unit in a way that conforms to 5098 ISO C99. This is enabled with -fexcess-precision=standard and with 5099 standards conformance options such as -std=c99, and may be disabled 5100 using -fexcess-precision=fast. 5101 * Support for the Intel Atom processor is now available through the 5102 -march=atom and -mtune=atom options. 5103 * A new -mcrc32 option is now available to enable crc32 intrinsics. 5104 * A new -mmovbe option is now available to enable GCC to use the 5105 movbe instruction to implement __builtin_bswap32 and 5106 __builtin_bswap64. 5107 * SSE math now can be enabled by default at configure time with the 5108 new --with-fpmath=sse option. 5109 * There is a new intrinsic header file, <x86intrin.h>. It should be 5110 included before using any IA-32/x86-64 intrinsics. 5111 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 5112 Orochi processors are now available with the -mxop, -mfma4, and 5113 -mlwp options. 5114 * The -mabm option enables GCC to use the popcnt and lzcnt 5115 instructions on AMD processors. 5116 * The -mpopcnt option enables GCC to use the popcnt instructions on 5117 both AMD and Intel processors. 5118 5119 M68K/ColdFire 5120 5121 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 5122 and 5441x devices. 5123 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 5124 processors. 5125 5126 MeP 5127 5128 Support has been added for the Toshiba Media embedded Processor (MeP, 5129 or mep-elf) embedded target. 5130 5131 MIPS 5132 5133 * GCC now supports MIPS 1004K processors. 5134 * GCC can now be configured with options --with-arch-32, 5135 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 5136 default optimization separately for 32-bit and 64-bit modes. 5137 * MIPS targets now support an alternative _mcount interface, in which 5138 register $12 points to the function's save slot for register $31. 5139 This interface is selected by the -mcount-ra-address option; see 5140 the documentation for more details. 5141 * GNU/Linux targets can now generate read-only .eh_frame sections. 5142 This optimization requires GNU binutils 2.20 or above, and is only 5143 available if GCC is configured with a suitable version of binutils. 5144 * GNU/Linux targets can now attach special relocations to indirect 5145 calls, so that the linker can turn them into direct jumps or 5146 branches. This optimization requires GNU binutils 2.20 or later, 5147 and is automatically selected if GCC is configured with an 5148 appropriate version of binutils. It can be explicitly enabled or 5149 disabled using the -mrelax-pic-calls command-line option. 5150 * GCC now generates more heavily-optimized atomic operations on 5151 Octeon processors. 5152 * MIPS targets now support the -fstack-protector option. 5153 * GCC now supports an -msynci option, which specifies that synci is 5154 enough to flush the instruction cache, without help from the 5155 operating system. GCC uses this information to optimize 5156 automatically-generated cache flush operations, such as those used 5157 for nested functions in C. There is also a --with-synci 5158 configure-time option, which makes -msynci the default. 5159 * GCC supports four new function attributes for interrupt handlers: 5160 interrupt, use_shadow_register_set, keep_interrupts_masked and 5161 use_debug_exception_return. See the documentation for more details 5162 about these attributes. 5163 5164 RS/6000 (POWER/PowerPC) 5165 5166 * GCC now supports the Power ISA 2.06, which includes the VSX 5167 instructions that add vector 64-bit floating point support, new 5168 population count instructions, and conversions between floating 5169 point and unsigned types. 5170 * Support for the power7 processor is now available through the 5171 -mcpu=power7 and -mtune=power7. 5172 * GCC will now vectorize loops that contain simple math functions 5173 like copysign when generating code for altivec or VSX targets. 5174 * Support for the A2 processor is now available through the -mcpu=a2 5175 and -mtune=a2 options. 5176 * Support for the 476 processor is now available through the 5177 -mcpu={476,476fp} and -mtune={476,476fp} options. 5178 * Support for the e500mc64 processor is now available through the 5179 -mcpu=e500mc64 and -mtune=e500mc64 options. 5180 * GCC can now be configured with options --with-cpu-32, 5181 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 5182 default optimization separately for 32-bit and 64-bit modes. 5183 * Starting with GCC 4.5.4, vectors of type vector long long or vector 5184 long are passed and returned in the same method as other vectors 5185 with the VSX instruction set. Previously the GCC compiler did not 5186 adhere to the ABI for 128-bit vectors with 64-bit integer base 5187 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 5188 5189 RX 5190 5191 Support has been added for the Renesas RX Processor (rx-elf) target. 5192 5193Operating Systems 5194 5195 Windows (Cygwin and MinGW) 5196 5197 * GCC now installs all the major language runtime libraries as DLLs 5198 when configured with the --enable-shared option. 5199 * GCC now makes use of the new support for aligned common variables 5200 in versions of binutils >= 2.20 to fix bugs in the support for SSE 5201 data types. 5202 * Improvements to the libffi support library increase the reliability 5203 of code generated by GCJ on all Windows platforms. Libgcj is 5204 enabled by default for the first time. 5205 * Libtool improvements simplify installation by placing the generated 5206 DLLs in the correct binaries directory. 5207 * Numerous other minor bugfixes and improvements, and substantial 5208 enhancements to the Fortran language support library. 5209 5210 > 5211 5212Other significant improvements 5213 5214 Plugins 5215 5216 * It is now possible to extend the compiler without having to modify 5217 its source code. A new option -fplugin=file.so tells GCC to load 5218 the shared object file.so and execute it as part of the compiler. 5219 The internal documentation describes the details on how plugins can 5220 interact with the compiler. 5221 5222 Installation changes 5223 5224 * The move to newer autotools changed default installation 5225 directories and switches to control them: The --with-datarootdir, 5226 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 5227 used any more. Instead, you can now use --datarootdir, --docdir, 5228 --htmldir, and --pdfdir. The default installation directories have 5229 changed as follows according to the GNU Coding Standards: 5230 5231 datarootdir read-only architecture-independent data root [PREFIX/share] 5232 localedir locale-specific message catalogs [DATAROOTDIR/locale] 5233 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 5234 htmldir html documentation [DOCDIR] 5235 dvidir dvi documentation [DOCDIR] 5236 pdfdir pdf documentation [DOCDIR] 5237 psdir ps documentation [DOCDIR] 5238 The following variables have new default values: 5239 5240 datadir read-only architecture-independent data [DATAROOTDIR] 5241 infodir info documentation [DATAROOTDIR/info] 5242 mandir man documentation [DATAROOTDIR/man] 5243 5244GCC 4.5.1 5245 5246 This is the [23]list of problem reports (PRs) from GCC's bug tracking 5247 system that are known to be fixed in the 4.5.1 release. This list might 5248 not be complete (that is, it is possible that some PRs that have been 5249 fixed are not listed here). 5250 5251 All languages 5252 5253 * GCC's new link-time optimizer ([24]-flto) now also works on a few 5254 non-ELF targets: 5255 + Cygwin (*-cygwin*) 5256 + MinGW (*-mingw*) 5257 + Darwin on x86-64 (x86_64-apple-darwin*) 5258 LTO is not enabled by default for these targets. To enable LTO, you 5259 should configure with the --enable-lto option. 5260 5261GCC 4.5.2 5262 5263 This is the [25]list of problem reports (PRs) from GCC's bug tracking 5264 system that are known to be fixed in the 4.5.2 release. This list might 5265 not be complete (that is, it is possible that some PRs that have been 5266 fixed are not listed here). 5267 5268GCC 4.5.3 5269 5270 This is the [26]list of problem reports (PRs) from GCC's bug tracking 5271 system that are known to be fixed in the 4.5.3 release. This list might 5272 not be complete (that is, it is possible that some PRs that have been 5273 fixed are not listed here). 5274 5275 On the PowerPC compiler, the Altivec builtin functions vec_ld and 5276 vec_st have been modified to generate the Altivec memory instructions 5277 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 5278 release, these builtin functions were changed to generate VSX memory 5279 reference instructions instead of Altivec memory instructions, but 5280 there are differences between the two instructions. If the VSX 5281 instruction set is available, you can now use the new builtin functions 5282 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 5283 instructions. 5284 5285GCC 4.5.4 5286 5287 This is the [27]list of problem reports (PRs) from GCC's bug tracking 5288 system that are known to be fixed in the 4.5.4 release. This list might 5289 not be complete (that is, it is possible that some PRs that have been 5290 fixed are not listed here). 5291 5292 5293 For questions related to the use of GCC, please consult these web 5294 pages and the [28]GCC manuals. If that fails, the 5295 [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these 5296 web pages and the development of GCC are welcome on our developer 5297 list at [30]gcc@gcc.gnu.org. All of [31]our lists have public 5298 archives. 5299 5300 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 5301 distribution of this entire article is permitted in any medium, 5302 provided this notice is preserved. 5303 5304 These pages are [33]maintained by the GCC team. Last modified 5305 2016-05-28[34]. 5306 5307References 5308 5309 1. http://www.multiprecision.org/ 5310 2. https://gcc.gnu.org/install/prerequisites.html 5311 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 5312 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 5313 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 5314 6. http://www.multiprecision.org/ 5315 7. https://gcc.gnu.org/PR30789 5316 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 5317 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 5318 10. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 5319 11. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 5320 12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 5321 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 5322 14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 5323 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 5324 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 5325 17. https://sourceware.org/gdb/wiki/STLSupport 5326 18. http://gcc.gnu.org/gcc-4.4/changes.html 5327 19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 5328 20. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 5329 21. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 5330 22. https://gcc.gnu.org/wiki/OOP 5331 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 5332 24. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 5333 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 5334 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 5335 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 5336 28. https://gcc.gnu.org/onlinedocs/ 5337 29. mailto:gcc-help@gcc.gnu.org 5338 30. mailto:gcc@gcc.gnu.org 5339 31. https://gcc.gnu.org/lists.html 5340 32. http://www.fsf.org/ 5341 33. https://gcc.gnu.org/about.html 5342 34. http://validator.w3.org/check/referer 5343====================================================================== 5344http://gcc.gnu.org/gcc-4.4/index.html 5345 5346 GCC 4.4 Release Series 5347 5348 March 13, 2012 5349 5350 The [1]GNU project and the GCC developers are pleased to announce the 5351 release of GCC 4.4.7. 5352 5353 This release is a bug-fix release, containing fixes for regressions in 5354 GCC 4.4.6 relative to previous releases of GCC. 5355 5356Release History 5357 5358 GCC 4.4.7 5359 March 13, 2012 ([2]changes) 5360 5361 GCC 4.4.6 5362 April 16, 2011 ([3]changes) 5363 5364 GCC 4.4.5 5365 October 1, 2010 ([4]changes) 5366 5367 GCC 4.4.4 5368 April 29, 2010 ([5]changes) 5369 5370 GCC 4.4.3 5371 January 21, 2010 ([6]changes) 5372 5373 GCC 4.4.2 5374 October 15, 2009 ([7]changes) 5375 5376 GCC 4.4.1 5377 July 22, 2009 ([8]changes) 5378 5379 GCC 4.4.0 5380 April 21, 2009 ([9]changes) 5381 5382References and Acknowledgements 5383 5384 GCC used to stand for the GNU C Compiler, but since the compiler 5385 supports several other languages aside from C, it now stands for the 5386 GNU Compiler Collection. 5387 5388 A list of [10]successful builds is updated as new information becomes 5389 available. 5390 5391 The GCC developers would like to thank the numerous people that have 5392 contributed new features, improvements, bug fixes, and other changes as 5393 well as test results to GCC. This [11]amazing group of volunteers is 5394 what makes GCC successful. 5395 5396 For additional information about GCC please refer to the [12]GCC 5397 project web site or contact the [13]GCC development mailing list. 5398 5399 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 5400 5401 5402 For questions related to the use of GCC, please consult these web 5403 pages and the [16]GCC manuals. If that fails, the 5404 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 5405 web pages and the development of GCC are welcome on our developer 5406 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 5407 archives. 5408 5409 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 5410 distribution of this entire article is permitted in any medium, 5411 provided this notice is preserved. 5412 5413 These pages are [21]maintained by the GCC team. Last modified 5414 2016-01-30[22]. 5415 5416References 5417 5418 1. http://www.gnu.org/ 5419 2. http://gcc.gnu.org/gcc-4.4/changes.html 5420 3. http://gcc.gnu.org/gcc-4.4/changes.html 5421 4. http://gcc.gnu.org/gcc-4.4/changes.html 5422 5. http://gcc.gnu.org/gcc-4.4/changes.html 5423 6. http://gcc.gnu.org/gcc-4.4/changes.html 5424 7. http://gcc.gnu.org/gcc-4.4/changes.html 5425 8. http://gcc.gnu.org/gcc-4.4/changes.html 5426 9. http://gcc.gnu.org/gcc-4.4/changes.html 5427 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 5428 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5429 12. http://gcc.gnu.org/index.html 5430 13. mailto:gcc@gcc.gnu.org 5431 14. http://gcc.gnu.org/mirrors.html 5432 15. http://gcc.gnu.org/svn.html 5433 16. https://gcc.gnu.org/onlinedocs/ 5434 17. mailto:gcc-help@gcc.gnu.org 5435 18. mailto:gcc@gcc.gnu.org 5436 19. https://gcc.gnu.org/lists.html 5437 20. http://www.fsf.org/ 5438 21. https://gcc.gnu.org/about.html 5439 22. http://validator.w3.org/check/referer 5440====================================================================== 5441http://gcc.gnu.org/gcc-4.4/changes.html 5442 5443 GCC 4.4 Release Series 5444 Changes, New Features, and Fixes 5445 5446 The latest release in the 4.4 release series is [1]GCC 4.4.7. 5447 5448Caveats 5449 5450 * __builtin_stdarg_start has been completely removed from GCC. 5451 Support for <varargs.h> had been deprecated since GCC 4.0. Use 5452 __builtin_va_start as a replacement. 5453 * Some of the errors issued by the C++ front end that could be 5454 downgraded to warnings in previous releases by using -fpermissive 5455 are now warnings by default. They can be converted into errors by 5456 using -pedantic-errors. 5457 * Use of the cpp assertion extension will now emit a warning when 5458 -Wdeprecated or -pedantic is used. This extension has been 5459 deprecated for many years, but never warned about. 5460 * Packed bit-fields of type char were not properly bit-packed on many 5461 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 5462 causes an ABI change. For example there is no longer a 4-bit 5463 padding between field a and b in this structure: 5464 struct foo 5465 { 5466 char a:4; 5467 char b:8; 5468 } __attribute__ ((packed)); 5469 There is a new warning to help identify fields that are affected: 5470 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 5471 The warning can be disabled with -Wno-packed-bitfield-compat. 5472 * On ARM EABI targets, the C++ mangling of the va_list type has been 5473 changed to conform to the current revision of the EABI. This does 5474 not affect the libstdc++ library included with GCC. 5475 * The SCOUNT and POS bits of the MIPS DSP control register are now 5476 treated as global. Previous versions of GCC treated these fields as 5477 call-clobbered instead. 5478 * The MIPS port no longer recognizes the h asm constraint. It was 5479 necessary to remove this constraint in order to avoid generating 5480 unpredictable code sequences. 5481 One of the main uses of the h constraint was to extract the high 5482 part of a multiplication on 64-bit targets. For example: 5483 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 5484 You can now achieve the same effect using 128-bit types: 5485 typedef unsigned int uint128_t __attribute__((mode(TI))); 5486 result = ((uint128_t) x * y) >> 64; 5487 The second sequence is better in many ways. For example, if x and y 5488 are constants, the compiler can perform the multiplication at 5489 compile time. If x and y are not constants, the compiler can 5490 schedule the runtime multiplication better than it can schedule an 5491 asm statement. 5492 * Support for a number of older systems and recently unmaintained or 5493 untested target ports of GCC has been declared obsolete in GCC 4.4. 5494 Unless there is activity to revive them, the next release of GCC 5495 will have their sources permanently removed. 5496 The following ports for individual systems on particular 5497 architectures have been obsoleted: 5498 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 5499 m68k-*-aout*) 5500 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 5501 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 5502 sh-*-*). This does not affect other more specific targets 5503 using the COFF object format on those architectures, or the 5504 more specific H8300 and SH targets (h8300-*-rtems*, 5505 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 5506 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 5507 + 2BSD on PDP-11 (pdp11-*-bsd) 5508 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 5509 powerpc-ibm-aix4.[12]*) 5510 + Tuning support for Itanium1 (Merced) variants. Note that code 5511 tuned for Itanium2 should also run correctly on Itanium1. 5512 * The protoize and unprotoize utilities have been obsoleted and will 5513 be removed in GCC 4.5. These utilities have not been installed by 5514 default since GCC 3.0. 5515 * Support has been removed for all the [2]configurations obsoleted in 5516 GCC 4.3. 5517 * Unknown -Wno-* options are now silently ignored by GCC if no other 5518 diagnostics are issued. If other diagnostics are issued, then GCC 5519 warns about the unknown options. 5520 * More information on porting to GCC 4.4 from previous versions of 5521 GCC can be found in the [3]porting guide for this release. 5522 5523General Optimizer Improvements 5524 5525 * A new command-line switch -findirect-inlining has been added. When 5526 turned on it allows the inliner to also inline indirect calls that 5527 are discovered to have known targets at compile time thanks to 5528 previous inlining. 5529 * A new command-line switch -ftree-switch-conversion has been added. 5530 This new pass turns simple initializations of scalar variables in 5531 switch statements into initializations from a static array, given 5532 that all the values are known at compile time and the ratio between 5533 the new array size and the original switch branches does not exceed 5534 the parameter --param switch-conversion-max-branch-ratio (default 5535 is eight). 5536 * A new command-line switch -ftree-builtin-call-dce has been added. 5537 This optimization eliminates unnecessary calls to certain builtin 5538 functions when the return value is not used, in cases where the 5539 calls can not be eliminated entirely because the function may set 5540 errno. This optimization is on by default at -O2 and above. 5541 * A new command-line switch -fconserve-stack directs the compiler to 5542 minimize stack usage even if it makes the generated code slower. 5543 This affects inlining decisions. 5544 * When the assembler supports it, the compiler will now emit unwind 5545 information using assembler .cfi directives. This makes it possible 5546 to use such directives in inline assembler code. The new option 5547 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 5548 directives. 5549 * The [4]Graphite branch has been merged. This merge has brought in a 5550 new framework for loop optimizations based on a polyhedral 5551 intermediate representation. These optimizations apply to all the 5552 languages supported by GCC. The following new code transformations 5553 are available in GCC 4.4: 5554 + -floop-interchange performs loop interchange transformations 5555 on loops. Interchanging two nested loops switches the inner 5556 and outer loops. For example, given a loop like: 5557 DO J = 1, M 5558 DO I = 1, N 5559 A(J, I) = A(J, I) * C 5560 ENDDO 5561 ENDDO 5562 5563 loop interchange will transform the loop as if the user had 5564 written: 5565 DO I = 1, N 5566 DO J = 1, M 5567 A(J, I) = A(J, I) * C 5568 ENDDO 5569 ENDDO 5570 5571 which can be beneficial when N is larger than the caches, 5572 because in Fortran, the elements of an array are stored in 5573 memory contiguously by column, and the original loop iterates 5574 over rows, potentially creating at each access a cache miss. 5575 + -floop-strip-mine performs loop strip mining transformations 5576 on loops. Strip mining splits a loop into two nested loops. 5577 The outer loop has strides equal to the strip size and the 5578 inner loop has strides of the original loop within a strip. 5579 For example, given a loop like: 5580 DO I = 1, N 5581 A(I) = A(I) + C 5582 ENDDO 5583 5584 loop strip mining will transform the loop as if the user had 5585 written: 5586 DO II = 1, N, 4 5587 DO I = II, min (II + 3, N) 5588 A(I) = A(I) + C 5589 ENDDO 5590 ENDDO 5591 5592 + -floop-block performs loop blocking transformations on loops. 5593 Blocking strip mines each loop in the loop nest such that the 5594 memory accesses of the element loops fit inside caches. For 5595 example, given a loop like: 5596 DO I = 1, N 5597 DO J = 1, M 5598 A(J, I) = B(I) + C(J) 5599 ENDDO 5600 ENDDO 5601 5602 loop blocking will transform the loop as if the user had 5603 written: 5604 DO II = 1, N, 64 5605 DO JJ = 1, M, 64 5606 DO I = II, min (II + 63, N) 5607 DO J = JJ, min (JJ + 63, M) 5608 A(J, I) = B(I) + C(J) 5609 ENDDO 5610 ENDDO 5611 ENDDO 5612 ENDDO 5613 5614 which can be beneficial when M is larger than the caches, 5615 because the innermost loop will iterate over a smaller amount 5616 of data that can be kept in the caches. 5617 * A new register allocator has replaced the old one. It is called 5618 integrated register allocator (IRA) because coalescing, register 5619 live range splitting, and hard register preferencing are done 5620 on-the-fly during coloring. It also has better integration with the 5621 reload pass. IRA is a regional register allocator which uses modern 5622 Chaitin-Briggs coloring instead of Chow's priority coloring used in 5623 the old register allocator. More info about IRA internals and 5624 options can be found in the GCC manuals. 5625 * A new instruction scheduler and software pipeliner, based on the 5626 selective scheduling approach, has been added. The new pass 5627 performs instruction unification, register renaming, substitution 5628 through register copies, and speculation during scheduling. The 5629 software pipeliner is able to pipeline non-countable loops. The new 5630 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 5631 it is available for the Intel Itanium platform working by default 5632 as the second scheduling pass (after register allocation) at the 5633 -O3 optimization level. 5634 * When using -fprofile-generate with a multi-threaded program, the 5635 profile counts may be slightly wrong due to race conditions. The 5636 new -fprofile-correction option directs the compiler to apply 5637 heuristics to smooth out the inconsistencies. By default the 5638 compiler will give an error message when it finds an inconsistent 5639 profile. 5640 * The new -fprofile-dir=PATH option permits setting the directory 5641 where profile data files are stored when using -fprofile-generate 5642 and friends, and the directory used when reading profile data files 5643 using -fprofile-use and friends. 5644 5645New warning options 5646 5647 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 5648 warning if any stack frame is larger than NUMBER bytes. This may be 5649 used to help ensure that code fits within a limited amount of stack 5650 space. 5651 * The command-line option -Wlarger-than-N is now written as 5652 -Wlarger-than=N and the old form is deprecated. 5653 * The new -Wno-mudflap option disables warnings about constructs 5654 which can not be instrumented when using -fmudflap. 5655 5656New Languages and Language specific improvements 5657 5658 * Version 3.0 of the [5]OpenMP specification is now supported for the 5659 C, C++, and Fortran compilers. 5660 * New character data types, per [6]TR 19769: New character types in 5661 C, are now supported for the C compiler in -std=gnu99 mode, as 5662 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 5663 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 5664 5665 C family 5666 5667 * A new optimize attribute was added to allow programmers to change 5668 the optimization level and particular optimization options for an 5669 individual function. You can also change the optimization options 5670 via the GCC optimize pragma for functions defined after the pragma. 5671 The GCC push_options pragma and the GCC pop_options pragma allow 5672 you temporarily save and restore the options used. The GCC 5673 reset_options pragma restores the options to what was specified on 5674 the command line. 5675 * Uninitialized warnings do not require enabling optimization 5676 anymore, that is, -Wuninitialized can be used together with -O0. 5677 Nonetheless, the warnings given by -Wuninitialized will probably be 5678 more accurate if optimization is enabled. 5679 * -Wparentheses now warns about expressions such as (!x | y) and (!x 5680 & y). Using explicit parentheses, such as in ((!x) | y), silences 5681 this warning. 5682 * -Wsequence-point now warns within if, while,do while and for 5683 conditions, and within for begin/end expressions. 5684 * A new option -dU is available to dump definitions of preprocessor 5685 macros that are tested or expanded. 5686 5687 C++ 5688 5689 * [7]Improved experimental support for the upcoming ISO C++ standard, 5690 C++0x. Including support for auto, inline namespaces, generalized 5691 initializer lists, defaulted and deleted functions, new character 5692 types, and scoped enums. 5693 * Those errors that may be downgraded to warnings to build legacy 5694 code now mention -fpermissive when -fdiagnostics-show-option is 5695 enabled. 5696 * -Wconversion now warns if the result of a static_cast to enumeral 5697 type is unspecified because the value is outside the range of the 5698 enumeral type. 5699 * -Wuninitialized now warns if a non-static reference or non-static 5700 const member appears in a class without constructors. 5701 * G++ now properly implements value-initialization, so objects with 5702 an initializer of () and an implicitly defined default constructor 5703 will be zero-initialized before the default constructor is called. 5704 5705 Runtime Library (libstdc++) 5706 5707 * Improved experimental support for the upcoming ISO C++ standard, 5708 C++0x, including: 5709 + Support for <chrono>, <condition_variable>, <cstdatomic>, 5710 <forward_list>, <initializer_list>, <mutex>, <ratio>, 5711 <system_error>, and <thread>. 5712 + unique_ptr, <algorithm> additions, exception propagation, and 5713 support for the new character types in <string> and <limits>. 5714 + Existing facilities now exploit initializer lists, defaulted 5715 and deleted functions, and the newly implemented core C++0x 5716 features. 5717 + Some standard containers are more efficient together with 5718 stateful allocators, i.e., no allocator is constructed on the 5719 fly at element construction time. 5720 * Experimental support for non-standard pointer types in containers. 5721 * The long standing libstdc++/30928 has been fixed for targets 5722 running glibc 2.10 or later. 5723 * As usual, many small and larger bug fixes, in particular quite a 5724 few corner cases in <locale>. 5725 5726 Fortran 5727 5728 * GNU Fortran now employs libcpp directly instead of using cc1 as an 5729 external preprocessor. The [8]-cpp option was added to allow manual 5730 invocation of the preprocessor without relying on filename 5731 extensions. 5732 * The [9]-Warray-temporaries option warns about array temporaries 5733 generated by the compiler, as an aid to optimization. 5734 * The [10]-fcheck-array-temporaries option has been added, printing a 5735 notification at run time, when an array temporary had to be created 5736 for an function argument. Contrary to -Warray-temporaries the 5737 warning is only printed if the array is noncontiguous. 5738 * Improved generation of DWARF debugging symbols 5739 * If using an intrinsic not part of the selected standard (via -std= 5740 and -fall-intrinsics) gfortran will now treat it as if this 5741 procedure were declared EXTERNAL and try to link to a user-supplied 5742 procedure. -Wintrinsics-std will warn whenever this happens. The 5743 now-useless option -Wnonstd-intrinsic was removed. 5744 * The flag -falign-commons has been added to control the alignment of 5745 variables in COMMON blocks, which is enabled by default in line 5746 with previous GCC version. Using -fno-align-commons one can force 5747 commons to be contiguous in memory as required by the Fortran 5748 standard, however, this slows down the memory access. The option 5749 -Walign-commons, which is enabled by default, warns when padding 5750 bytes were added for alignment. The proper solution is to sort the 5751 common objects by decreasing storage size, which avoids the 5752 alignment problems. 5753 * Fortran 2003 support has been extended: 5754 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 5755 now supported (except internal reads from/writes to wide 5756 strings). [11]-fbackslash now supports also \unnnn and 5757 \Unnnnnnnn to enter Unicode characters. 5758 + Asynchronous I/O (implemented as synchronous I/O) and the 5759 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 5760 are now supported in I/O statements. 5761 + Support for Fortran 2003 structure constructors and for array 5762 constructor with typespec has been added. 5763 + Procedure Pointers (but not yet as component in derived types 5764 and as function results) are now supported. 5765 + Abstract types, type extension, and type-bound procedures 5766 (both PROCEDURE and GENERIC but not as operators). Note: As 5767 CLASS/polymorphyic types are not implemented, type-bound 5768 procedures with PASS accept as non-standard extension TYPE 5769 arguments. 5770 * Fortran 2008 support has been added: 5771 + The -std=f2008 option and support for the file extensions 5772 .f2008 and .F2008 has been added. 5773 + The g0 format descriptor is now supported. 5774 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 5775 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 5776 are now available (some of them existed as GNU extension 5777 before). Note: The hyperbolic functions are not yet supporting 5778 complex arguments and the three- argument version of BESSEL_*N 5779 is not available. 5780 + The bit intrinsics LEADZ and TRAILZ have been added. 5781 5782 Java (GCJ) 5783 5784 Ada 5785 5786 * The Ada runtime now supports multilibs on many platforms including 5787 x86_64, SPARC and PowerPC. Their build is enabled by default. 5788 5789New Targets and Target Specific Improvements 5790 5791 ARM 5792 5793 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 5794 Cortex-R4F processors and has many other improvements to 5795 optimization for ARM processors. 5796 * GCC now supports the VFPv3 variant with 16 double-precision 5797 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 5798 renamed to -mfpu=vfpv3. 5799 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 5800 erratum on Cortex-M3 processors. 5801 * GCC now supports the __sync_* atomic operations for ARM EABI 5802 GNU/Linux. 5803 * The section anchors optimization is now enabled by default when 5804 optimizing for ARM. 5805 * GCC now uses a new EABI-compatible profiling interface for EABI 5806 targets. This requires a function __gnu_mcount_nc, which is 5807 provided by GNU libc versions 2.8 and later. 5808 5809 AVR 5810 5811 * The -mno-tablejump option has been deprecated because it has the 5812 same effect as the -fno-jump-tables option. 5813 * Added support for these new AVR devices: 5814 + ATA6289 5815 + ATtiny13A 5816 + ATtiny87 5817 + ATtiny167 5818 + ATtiny327 5819 + ATmega8C1 5820 + ATmega16C1 5821 + ATmega32C1 5822 + ATmega8M1 5823 + ATmega16M1 5824 + ATmega32M1 5825 + ATmega32U4 5826 + ATmega16HVB 5827 + ATmega4HVD 5828 + ATmega8HVD 5829 + ATmega64C1 5830 + ATmega64M1 5831 + ATmega16U4 5832 + ATmega32U6 5833 + ATmega128RFA1 5834 + AT90PWM81 5835 + AT90SCR100 5836 + M3000F 5837 + M3000S 5838 + M3001B 5839 5840 IA-32/x86-64 5841 5842 * Support for Intel AES built-in functions and code generation is 5843 available via -maes. 5844 * Support for Intel PCLMUL built-in function and code generation is 5845 available via -mpclmul. 5846 * Support for Intel AVX built-in functions and code generation is 5847 available via -mavx. 5848 * Automatically align the stack for local variables with alignment 5849 requirement. 5850 * GCC can now utilize the SVML library for vectorizing calls to a set 5851 of C99 functions if -mveclibabi=svml is specified and you link to 5852 an SVML ABI compatible library. 5853 * On x86-64, the ABI has been changed in the following cases to 5854 conform to the x86-64 ABI: 5855 + Passing/returning structures with flexible array member: 5856 struct foo 5857 { 5858 int i; 5859 int flex[]; 5860 }; 5861 + Passing/returning structures with complex float member: 5862 struct foo 5863 { 5864 int i; 5865 __complex__ float f; 5866 }; 5867 + Passing/returning unions with long double member: 5868 union foo 5869 { 5870 int x; 5871 long double ld; 5872 }; 5873 Code built with previous versions of GCC that uses any of these is 5874 not compatible with code built with GCC 4.4.0 or later. 5875 * A new target attribute was added to allow programmers to change the 5876 target options like -msse2 or -march=k8 for an individual function. 5877 You can also change the target options via the GCC target pragma 5878 for functions defined after the pragma. 5879 * GCC can now be configured with options --with-arch-32, 5880 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 5881 --with-tune-64 to control the default optimization separately for 5882 32-bit and 64-bit modes. 5883 5884 IA-32/IA64 5885 5886 * Support for __float128 (TFmode) IEEE quad type and corresponding 5887 TCmode IEEE complex quad type is available via the soft-fp library 5888 on IA-32/IA64 targets. This includes basic arithmetic operations 5889 (addition, subtraction, negation, multiplication and division) on 5890 __float128 real and TCmode complex values, the full set of IEEE 5891 comparisons between __float128 values, conversions to and from 5892 float, double and long double floating point types, as well as 5893 conversions to and from signed or unsigned integer, signed or 5894 unsigned long integer and signed or unsigned quad (TImode, IA64 5895 only) integer types. Additionally, all operations generate the full 5896 set of IEEE exceptions and support the full set of IEEE rounding 5897 modes. 5898 5899 M68K/ColdFire 5900 5901 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 5902 processors. (Scheduling support for ColdFire V2 processors was 5903 added in GCC 4.3.) 5904 * GCC now supports the -mxgot option to support programs requiring 5905 many GOT entries on ColdFire. 5906 * The m68k-*-linux-gnu target now builds multilibs by default. 5907 5908 MIPS 5909 5910 * MIPS Technologies have extended the original MIPS SVR4 ABI to 5911 include support for procedure linkage tables (PLTs) and copy 5912 relocations. These extensions allow GNU/Linux executables to use a 5913 significantly more efficient code model than the one defined by the 5914 original ABI. 5915 GCC support for this code model is available via a new command-line 5916 option, -mplt. There is also a new configure-time option, 5917 --with-mips-plt, to make -mplt the default. 5918 The new code model requires support from the assembler, the linker, 5919 and the runtime C library. This support is available in binutils 5920 2.19 and GLIBC 2.9. 5921 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 5922 and 32-bit GNU/Linux shared libraries. This feature requires GNU 5923 binutils 2.19 or above. 5924 * Support for RMI's XLR processor is now available through the 5925 -march=xlr and -mtune=xlr options. 5926 * 64-bit targets can now perform 128-bit multiplications inline, 5927 instead of relying on a libgcc function. 5928 * Native GNU/Linux toolchains now support -march=native and 5929 -mtune=native, which select the host processor. 5930 * GCC now supports the R10K, R12K, R14K and R16K processors. The 5931 canonical -march= and -mtune= names for these processors are 5932 r10000, r12000, r14000 and r16000 respectively. 5933 * GCC can now work around the side effects of speculative execution 5934 on R10K processors. Please see the documentation of the 5935 -mr10k-cache-barrier option for details. 5936 * Support for the MIPS64 Release 2 instruction set has been added. 5937 The option -march=mips64r2 enables generation of these 5938 instructions. 5939 * GCC now supports Cavium Networks' Octeon processor. This support is 5940 available through the -march=octeon and -mtune=octeon options. 5941 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 5942 canonical -march= and -mtune= names for these processors are 5943 loongson2e and loongson2f. 5944 5945 picochip 5946 5947 Picochip is a 16-bit processor. A typical picoChip contains over 250 5948 small cores, each with small amounts of memory. There are three 5949 processor variants (STAN, MEM and CTRL) with different instruction sets 5950 and memory configurations and they can be chosen using the -mae option. 5951 5952 This port is intended to be a "C" only port. 5953 5954 Power Architecture and PowerPC 5955 5956 * GCC now supports the e300c2, e300c3 and e500mc processors. 5957 * GCC now supports Xilinx processors with a single-precision FPU. 5958 * Decimal floating point is now supported for e500 processors. 5959 5960 S/390, zSeries and System z9/z10 5961 5962 * Support for the IBM System z10 EC/BC processor has been added. When 5963 using the -march=z10 option, the compiler will generate code making 5964 use of instructions provided by the General-Instruction-Extension 5965 Facility and the Execute-Extension Facility. 5966 5967 VxWorks 5968 5969 * GCC now supports the thread-local storage mechanism used on 5970 VxWorks. 5971 5972 Xtensa 5973 5974 * GCC now supports thread-local storage (TLS) for Xtensa processor 5975 configurations that include the Thread Pointer option. TLS also 5976 requires support from the assembler and linker; this support is 5977 provided in the GNU binutils beginning with version 2.19. 5978 5979Documentation improvements 5980 5981Other significant improvements 5982 5983GCC 4.4.1 5984 5985 This is the [12]list of problem reports (PRs) from GCC's bug tracking 5986 system that are known to be fixed in the 4.4.1 release. This list might 5987 not be complete (that is, it is possible that some PRs that have been 5988 fixed are not listed here). 5989 5990GCC 4.4.2 5991 5992 This is the [13]list of problem reports (PRs) from GCC's bug tracking 5993 system that are known to be fixed in the 4.4.2 release. This list might 5994 not be complete (that is, it is possible that some PRs that have been 5995 fixed are not listed here). 5996 5997GCC 4.4.3 5998 5999 This is the [14]list of problem reports (PRs) from GCC's bug tracking 6000 system that are known to be fixed in the 4.4.3 release. This list might 6001 not be complete (that is, it is possible that some PRs that have been 6002 fixed are not listed here). 6003 6004GCC 4.4.4 6005 6006 This is the [15]list of problem reports (PRs) from GCC's bug tracking 6007 system that are known to be fixed in the 4.4.4 release. This list might 6008 not be complete (that is, it is possible that some PRs that have been 6009 fixed are not listed here). 6010 6011GCC 4.4.5 6012 6013 This is the [16]list of problem reports (PRs) from GCC's bug tracking 6014 system that are known to be fixed in the 4.4.5 release. This list might 6015 not be complete (that is, it is possible that some PRs that have been 6016 fixed are not listed here). 6017 6018GCC 4.4.6 6019 6020 This is the [17]list of problem reports (PRs) from GCC's bug tracking 6021 system that are known to be fixed in the 4.4.6 release. This list might 6022 not be complete (that is, it is possible that some PRs that have been 6023 fixed are not listed here). 6024 6025GCC 4.4.7 6026 6027 This is the [18]list of problem reports (PRs) from GCC's bug tracking 6028 system that are known to be fixed in the 4.4.7 release. This list might 6029 not be complete (that is, it is possible that some PRs that have been 6030 fixed are not listed here). 6031 6032 6033 For questions related to the use of GCC, please consult these web 6034 pages and the [19]GCC manuals. If that fails, the 6035 [20]gcc-help@gcc.gnu.org mailing list might help. Comments on these 6036 web pages and the development of GCC are welcome on our developer 6037 list at [21]gcc@gcc.gnu.org. All of [22]our lists have public 6038 archives. 6039 6040 Copyright (C) [23]Free Software Foundation, Inc. Verbatim copying and 6041 distribution of this entire article is permitted in any medium, 6042 provided this notice is preserved. 6043 6044 These pages are [24]maintained by the GCC team. Last modified 6045 2016-01-30[25]. 6046 6047References 6048 6049 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 6050 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 6051 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 6052 4. https://gcc.gnu.org/wiki/Graphite 6053 5. http://openmp.org/wp/openmp-specifications/ 6054 6. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 6055 7. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 6056 8. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 6057 9. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 6058 10. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 6059 11. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 6060 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 6061 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 6062 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 6063 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 6064 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 6065 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 6066 18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 6067 19. https://gcc.gnu.org/onlinedocs/ 6068 20. mailto:gcc-help@gcc.gnu.org 6069 21. mailto:gcc@gcc.gnu.org 6070 22. https://gcc.gnu.org/lists.html 6071 23. http://www.fsf.org/ 6072 24. https://gcc.gnu.org/about.html 6073 25. http://validator.w3.org/check/referer 6074====================================================================== 6075http://gcc.gnu.org/gcc-4.3/index.html 6076 6077 GCC 4.3 Release Series 6078 6079 Jun 27, 2011 6080 6081 The [1]GNU project and the GCC developers are pleased to announce the 6082 release of GCC 4.3.6. 6083 6084 This release is a bug-fix release, containing fixes for regressions in 6085 GCC 4.3.5 relative to previous releases of GCC. 6086 6087Release History 6088 6089 GCC 4.3.6 6090 Jun 27, 2011 ([2]changes) 6091 6092 GCC 4.3.5 6093 May 22, 2010 ([3]changes) 6094 6095 GCC 4.3.4 6096 August 4, 2009 ([4]changes) 6097 6098 GCC 4.3.3 6099 January 24, 2009 ([5]changes) 6100 6101 GCC 4.3.2 6102 August 27, 2008 ([6]changes) 6103 6104 GCC 4.3.1 6105 June 6, 2008 ([7]changes) 6106 6107 GCC 4.3.0 6108 March 5, 2008 ([8]changes) 6109 6110References and Acknowledgements 6111 6112 GCC used to stand for the GNU C Compiler, but since the compiler 6113 supports several other languages aside from C, it now stands for the 6114 GNU Compiler Collection. 6115 6116 A list of [9]successful builds is updated as new information becomes 6117 available. 6118 6119 The GCC developers would like to thank the numerous people that have 6120 contributed new features, improvements, bug fixes, and other changes as 6121 well as test results to GCC. This [10]amazing group of volunteers is 6122 what makes GCC successful. 6123 6124 For additional information about GCC please refer to the [11]GCC 6125 project web site or contact the [12]GCC development mailing list. 6126 6127 To obtain GCC please use [13]our mirror sites or [14]our SVN server. 6128 6129 6130 For questions related to the use of GCC, please consult these web 6131 pages and the [15]GCC manuals. If that fails, the 6132 [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these 6133 web pages and the development of GCC are welcome on our developer 6134 list at [17]gcc@gcc.gnu.org. All of [18]our lists have public 6135 archives. 6136 6137 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 6138 distribution of this entire article is permitted in any medium, 6139 provided this notice is preserved. 6140 6141 These pages are [20]maintained by the GCC team. Last modified 6142 2016-01-30[21]. 6143 6144References 6145 6146 1. http://www.gnu.org/ 6147 2. http://gcc.gnu.org/gcc-4.3/changes.html 6148 3. http://gcc.gnu.org/gcc-4.3/changes.html 6149 4. http://gcc.gnu.org/gcc-4.3/changes.html 6150 5. http://gcc.gnu.org/gcc-4.3/changes.html 6151 6. http://gcc.gnu.org/gcc-4.3/changes.html 6152 7. http://gcc.gnu.org/gcc-4.3/changes.html 6153 8. http://gcc.gnu.org/gcc-4.3/changes.html 6154 9. http://gcc.gnu.org/gcc-4.3/buildstat.html 6155 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6156 11. http://gcc.gnu.org/index.html 6157 12. mailto:gcc@gcc.gnu.org 6158 13. http://gcc.gnu.org/mirrors.html 6159 14. http://gcc.gnu.org/svn.html 6160 15. https://gcc.gnu.org/onlinedocs/ 6161 16. mailto:gcc-help@gcc.gnu.org 6162 17. mailto:gcc@gcc.gnu.org 6163 18. https://gcc.gnu.org/lists.html 6164 19. http://www.fsf.org/ 6165 20. https://gcc.gnu.org/about.html 6166 21. http://validator.w3.org/check/referer 6167====================================================================== 6168http://gcc.gnu.org/gcc-4.3/changes.html 6169 6170 GCC 4.3 Release Series 6171 Changes, New Features, and Fixes 6172 6173 The latest release in the 4.3 release series is [1]GCC 4.3.5. 6174 6175Caveats 6176 6177 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 6178 various front-end languages it supports. See the [4]prerequisites 6179 page for version requirements. 6180 * ColdFire targets now treat long double as having the same format as 6181 double. In earlier versions of GCC, they used the 68881 long double 6182 format instead. 6183 * The m68k-uclinux target now uses the same calling conventions as 6184 m68k-linux-gnu. You can select the original calling conventions by 6185 configuring for m68k-uclinuxoldabi instead. Note that 6186 m68k-uclinuxoldabi also retains the original 80-bit long double on 6187 ColdFire targets. 6188 * The -fforce-mem option has been removed because it has had no 6189 effect in the last few GCC releases. 6190 * The i386 -msvr3-shlib option has been removed since it is no longer 6191 used. 6192 * Fastcall for i386 has been changed not to pass aggregate arguments 6193 in registers, following Microsoft compilers. 6194 * Support for the AOF assembler has been removed from the ARM back 6195 end; this affects only the targets arm-semi-aof and armel-semi-aof, 6196 which are no longer recognized. We removed these targets without a 6197 deprecation period because we discovered that they have been 6198 unusable since GCC 4.0.0. 6199 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 6200 has been removed. This support had been deprecated since GCC 4.0.0. 6201 * Support for a number of older systems and recently unmaintained or 6202 untested target ports of GCC has been declared obsolete in GCC 4.3. 6203 Unless there is activity to revive them, the next release of GCC 6204 will have their sources permanently removed. 6205 All GCC ports for the following processor architectures have been 6206 declared obsolete: 6207 + Morpho MT (mt-*) 6208 The following aliases for processor architectures have been 6209 declared obsolete. Users should use the indicated generic target 6210 names instead, with compile-time options such as -mcpu or 6211 configure-time options such as --with-cpu to control the 6212 configuration more precisely. 6213 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 6214 instead). 6215 + parisc*-*-* (use hppa*-*-* instead). 6216 + m680[012]0-*-* (use m68k-*-* instead). 6217 All GCC ports for the following operating systems have been 6218 declared obsolete: 6219 + BeOS (*-*-beos*) 6220 + kaOS (*-*-kaos*) 6221 + GNU/Linux using the a.out object format (*-*-linux*aout*) 6222 + GNU/Linux using version 1 of the GNU C Library 6223 (*-*-linux*libc1*) 6224 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 6225 *-*-solaris2.[0-6].*) 6226 + Miscellaneous System V (*-*-sysv*) 6227 + WindISS (*-*-windiss*) 6228 Also, those for some individual systems on particular architectures 6229 have been obsoleted: 6230 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 6231 + CRIS with a.out object format (cris-*-aout) 6232 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 6233 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 6234 + PRO on PA-RISC (hppa1.1-*-pro*) 6235 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 6236 i[34567]86-sequent-sysv4*) 6237 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 6238 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 6239 was previously [5]removed in 2001, leaving only the support 6240 for UWIN as a target now being deprecated) 6241 + ChorusOS on PowerPC (powerpc-*-chorusos*) 6242 + All VAX configurations apart from NetBSD and OpenBSD 6243 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 6244 * The [6]-Wconversion option has been modified. Its purpose now is to 6245 warn for implicit conversions that may alter a value. This new 6246 behavior is available for both C and C++. Warnings about 6247 conversions between signed and unsigned integers can be disabled by 6248 using -Wno-sign-conversion. In C++, they are disabled by default 6249 unless -Wsign-conversion is explicitly requested. The old behavior 6250 of -Wconversion, that is, warn for prototypes causing a type 6251 conversion that is different from what would happen to the same 6252 argument in the absence of a prototype, has been moved to a new 6253 option -Wtraditional-conversion, which is only available for C. 6254 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 6255 been removed because they were deprecated for more than 3 GCC major 6256 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 6257 -mtune=pentiumpro as a replacement. 6258 * The -funsafe-math-optimizations option now automatically turns on 6259 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 6260 reassociation and thus may introduce or remove traps. 6261 * The -ftree-vectorize option is now on by default under -O3. In 6262 order to generate code for a SIMD extension, it has to be enabled 6263 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 6264 i?86 and x86_64. 6265 * More information on porting to GCC 4.3 from previous versions of 6266 GCC can be found in the [7]porting guide for this release. 6267 6268General Optimizer Improvements 6269 6270 * The GCC middle-end has been integrated with the [8]MPFR library. 6271 This allows GCC to evaluate and replace at compile-time calls to 6272 built-in math functions having constant arguments with their 6273 mathematically equivalent results. In making use of [9]MPFR, GCC 6274 can generate correct results regardless of the math library 6275 implementation or floating point precision of the host platform. 6276 This also allows GCC to generate identical results regardless of 6277 whether one compiles in native or cross-compile configurations to a 6278 particular target. The following built-in functions take advantage 6279 of this new capability: acos, acosh, asin, asinh, atan2, atan, 6280 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 6281 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 6282 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 6283 tan, tanh, tgamma, y0, y1 and yn. The float and long double 6284 variants of these functions (e.g. sinf and sinl) are also handled. 6285 The sqrt and cabs functions with constant arguments were already 6286 optimized in prior GCC releases. Now they also use [10]MPFR. 6287 * A new forward propagation pass on RTL was added. The new pass 6288 replaces several slower transformations, resulting in compile-time 6289 improvements as well as better code generation in some cases. 6290 * A new command-line switch -frecord-gcc-switches has been added to 6291 GCC, although it is only enabled for some targets. The switch 6292 causes the command line that was used to invoke the compiler to be 6293 recorded into the object file that is being created. The exact 6294 format of this recording is target and binary file format 6295 dependent, but it usually takes the form of a note section 6296 containing ASCII text. The switch is related to the -fverbose-asm 6297 switch, but that one only records the information in the assembler 6298 output file as comments, so the information never reaches the 6299 object file. 6300 * The inliner heuristic is now aware of stack frame consumption. New 6301 command-line parameters --param large-stack-frame and --param 6302 large-stack-frame-growth can be used to limit stack frame size 6303 growth caused by inlining. 6304 * During feedback directed optimizations, the expected block size the 6305 memcpy, memset and bzero functions operate on is discovered and for 6306 cases of commonly used small sizes, specialized inline code is 6307 generated. 6308 * __builtin_expect no longer requires its argument to be a compile 6309 time constant. 6310 * Interprocedural optimization was reorganized to work on functions 6311 in SSA form. This enables more precise and cheaper dataflow 6312 analysis and makes writing interprocedural optimizations easier. 6313 The following improvements have been implemented on top of this 6314 framework: 6315 + Pre-inline optimization: Selected local optimization passes 6316 are run before the inliner (and other interprocedural passes) 6317 are executed. This significantly improves the accuracy of code 6318 growth estimates used by the inliner and reduces the overall 6319 memory footprint for large compilation units. 6320 + Early inlining (a simple bottom-up inliner pass inlining only 6321 functions whose body is smaller than the expected call 6322 overhead) is now executed with the early optimization passes, 6323 thus inlining already optimized function bodies into an 6324 unoptimized function that is subsequently optimized by early 6325 optimizers. This enables the compiler to quickly eliminate 6326 abstraction penalty in C++ programs. 6327 + Interprocedural constant propagation now operate on SSA form 6328 increasing accuracy of the analysis. 6329 * A new internal representation for GIMPLE statements has been 6330 contributed, resulting in compile-time memory savings. 6331 * The vectorizer was enhanced to support vectorization of outer 6332 loops, intra-iteration parallelism (loop-aware SLP), vectorization 6333 of strided accesses and loops with multiple data-types. Run-time 6334 dependency testing using loop versioning was added. The cost model, 6335 turned on by -fvect-cost-model, was developed. 6336 6337New Languages and Language specific improvements 6338 6339 * We have added new command-line options 6340 -finstrument-functions-exclude-function-list and 6341 -finstrument-functions-exclude-file-list. They provide more control 6342 over which functions are annotated by the -finstrument-functions 6343 option. 6344 6345 C family 6346 6347 * Implicit conversions between generic vector types are now only 6348 permitted when the two vectors in question have the same number of 6349 elements and compatible element types. (Note that the restriction 6350 involves compatible element types, not implicitly-convertible 6351 element types: thus, a vector type with element type int may not be 6352 implicitly converted to a vector type with element type unsigned 6353 int.) This restriction, which is in line with specifications for 6354 SIMD architectures such as AltiVec, may be relaxed using the flag 6355 -flax-vector-conversions. This flag is intended only as a 6356 compatibility measure and should not be used for new code. 6357 * -Warray-bounds has been added and is now enabled by default for 6358 -Wall . It produces warnings for array subscripts that can be 6359 determined at compile time to be always out of bounds. 6360 -Wno-array-bounds will disable the warning. 6361 * The constructor and destructor function attributes now accept 6362 optional priority arguments which control the order in which the 6363 constructor and destructor functions are run. 6364 * New [11]command-line options -Wtype-limits, 6365 -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body, 6366 -Wclobbered and -Wignored-qualifiers have been added for finer 6367 control of the diverse warnings enabled by -Wextra. 6368 * A new function attribute alloc_size has been added to mark up 6369 malloc style functions. For constant sized allocations this can be 6370 used to find out the size of the returned pointer using the 6371 __builtin_object_size() function for buffer overflow checking and 6372 similar. This supplements the already built-in malloc and calloc 6373 constant size handling. 6374 * Integer constants written in binary are now supported as a GCC 6375 extension. They consist of a prefix 0b or 0B, followed by a 6376 sequence of 0 and 1 digits. 6377 * A new predefined macro __COUNTER__ has been added. It expands to 6378 sequential integral values starting from 0. In conjunction with the 6379 ## operator, this provides a convenient means to generate unique 6380 identifiers. 6381 * A new command-line option -fdirectives-only has been added. It 6382 enables a special preprocessing mode which improves the performance 6383 of applications like distcc and ccache. 6384 * Fixed-point data types and operators have been added. They are 6385 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 6386 Currently, only MIPS targets are supported. 6387 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 6388 N1241, is now supported as a GCC extension to C for targets 6389 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 6390 and x86_64-*-linux-gnu. The feature introduces new data types 6391 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 6392 DD, and DL. 6393 6394 C++ 6395 6396 * [12]Experimental support for the upcoming ISO C++ standard, C++0x. 6397 * -Wc++0x-compat has been added and is now enabled by default for 6398 -Wall. It produces warnings for constructs whose meaning differs 6399 between ISO C++ 1998 and C++0x. 6400 * The -Wparentheses option now works for C++ as it does for C. It 6401 warns if parentheses are omitted when operators with confusing 6402 precedence are nested. It also warns about ambiguous else 6403 statements. Since -Wparentheses is enabled by -Wall, this may cause 6404 additional warnings with existing C++ code which uses -Wall. These 6405 new warnings may be disabled by using -Wall -Wno-parentheses. 6406 * The -Wmissing-declarations now works for C++ as it does for C. 6407 * The -fvisibility-ms-compat flag was added, to make it easier to 6408 port larger projects using shared libraries from Microsoft's Visual 6409 Studio to ELF and Mach-O systems. 6410 * C++ attribute handling has been overhauled for template arguments 6411 (ie dependent types). In particular, __attribute__((aligned(T))); 6412 works for C++ types. 6413 6414 Runtime Library (libstdc++) 6415 6416 * [13]Experimental support for the upcoming ISO C++ standard, C++0x. 6417 * Support for TR1 mathematical special functions and regular 6418 expressions. ([14]Implementation status of TR1) 6419 * Default what implementations give more elaborate exception strings 6420 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 6421 * Header dependencies have been streamlined, reducing unnecessary 6422 includes and pre-processed bloat. 6423 * Variadic template implementations of items in <tuple> and 6424 <functional>. 6425 * An experimental [15]parallel mode has been added. This is a 6426 parallel implementation of many C++ Standard library algorithms, 6427 like std::accumulate, std::for_each, std::transform, or std::sort, 6428 to give but four examples. These algorithms can be substituted for 6429 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 6430 or all existing algorithms can be transformed via the 6431 -D_GLIBCXX_PARALLEL macro. 6432 * Debug mode versions of classes in <unordered_set> and 6433 <unordered_map>. 6434 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 6435 now <backward/hash_set> and <backward/hash_map>. This code: 6436 #include <ext/hash_set> 6437 __gnu_cxx::hash_set<int> s; 6438 6439 Can be transformed (in order of preference) to: 6440 #include <tr1/unordered_set> 6441 std::tr1::unordered_set<int> s; 6442 6443 or 6444 #include <backward/hash_set> 6445 __gnu_cxx::hash_set<int> s; 6446 6447 Similar transformations apply to __gnu_cxx::hash_map, 6448 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 6449 __gnu_cxx::hash_multiset. 6450 6451 Fortran 6452 6453 * Due to the fact that the GMP and MPFR libraries are required for 6454 all languages, Fortran is no longer special in this regard and is 6455 available by default. 6456 * The [16]-fexternal-blas option has been added, which generates 6457 calls to BLAS routines for intrinsic matrix operations such as 6458 matmul rather than using the built-in algorithms. 6459 * Support to give a backtrace (compiler flag -fbacktrace or 6460 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 6461 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 6462 run-time error occured. 6463 * GNU Fortran now defines __GFORTRAN__ when it runs the C 6464 preprocessor (CPP). 6465 * The [17]-finit-local-zero, -finit-real, -finit-integer, 6466 -finit-character, and -finit-logical options have been added, which 6467 can be used to initialize local variables. 6468 * The intrinsic procedures [18]GAMMA and [19]LGAMMA have been added, 6469 which calculate the Gamma function and its logarithm. Use EXTERNAL 6470 gamma if you want to use your own gamma function. 6471 * GNU Fortran now regards the backslash character as literal (as 6472 required by the Fortran 2003 standard); using [20]-fbackslash GNU 6473 Fortran interprets backslashes as C-style escape characters. 6474 * The [21]interpretation of binary, octal and hexadecimal (BOZ) 6475 literal constants has been changed. Before they were always 6476 interpreted as integer; now they are bit-wise transferred as 6477 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 6478 2003 standard, and for real and complex variables in DATA 6479 statements or when directly assigned to real and complex variables. 6480 Everywhere else and especially in expressions they are still 6481 regarded as integer constants. 6482 * Fortran 2003 support has been extended: 6483 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 6484 + Pointer intent 6485 + Intrinsic module ISO_ENV_FORTRAN 6486 + Interoperability with C (ISO C Bindings) 6487 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 6488 attribute) 6489 + Fortran 2003 BOZ 6490 6491 Java (GCJ) 6492 6493 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 6494 This enables the use of all 1.5 language features, and fixes most 6495 existing front end bugs. 6496 * libgcj now supports all 1.5 language features which require runtime 6497 support: foreach, enum, annotations, generics, and auto-boxing. 6498 * We've made many changes to the tools shipped with gcj. 6499 + The old jv-scan tool has been removed. This tool never really 6500 worked properly. There is no replacement. 6501 + gcjh has been rewritten. Some of its more obscure options no 6502 longer work, but are still recognized in an attempt at 6503 compatibility. gjavah is a new program with similar 6504 functionality but different command-line options. 6505 + grmic and grmiregistry have been rewritten. grmid has been 6506 added. 6507 + gjar replaces the old fastjar. 6508 + gjarsigner (used for signing jars), gkeytool (used for key 6509 management), gorbd (for CORBA), gserialver (computes 6510 serialization UIDs), and gtnameserv (also for CORBA) are now 6511 installed. 6512 * The ability to dump the contents of the java run time heap to a 6513 file for off-line analysis has been added. The heap dumps may be 6514 analyzed with the new gc-analyze tool. They may be generated on 6515 out-of-memory conditions or on demand and are controlled by the new 6516 run time class gnu.gcj.util.GCInfo. 6517 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 6518 provide correct, updated, timezone information. This means that 6519 packagers no longer have to update libgcj when a time zone change 6520 is published. 6521 6522New Targets and Target Specific Improvements 6523 6524 IA-32/x86-64 6525 6526 * Tuning for Intel Core 2 processors is available via -mtune=core2 6527 and -march=core2. 6528 * Tuning for AMD Geode processors is available via -mtune=geode and 6529 -march=geode. 6530 * Code generation of block move (memcpy) and block set (memset) was 6531 rewritten. GCC can now pick the best algorithm (loop, unrolled 6532 loop, instruction with rep prefix or a library call) based on the 6533 size of the block being copied and the CPU being optimized for. A 6534 new option -minline-stringops-dynamically has been added. With this 6535 option string operations of unknown size are expanded such that 6536 small blocks are copied by in-line code, while for large blocks a 6537 library call is used. This results in faster code than 6538 -minline-all-stringops when the library implementation is capable 6539 of using cache hierarchy hints. The heuristic choosing the 6540 particular algorithm can be overwritten via -mstringop-strategy. 6541 Newly also memset of values different from 0 is inlined. 6542 * GCC no longer places the cld instruction before string operations. 6543 Both i386 and x86-64 ABI documents mandate the direction flag to be 6544 clear at the entry of a function. It is now invalid to set the flag 6545 in asm statement without reseting it afterward. 6546 * Support for SSSE3 built-in functions and code generation are 6547 available via -mssse3. 6548 * Support for SSE4.1 built-in functions and code generation are 6549 available via -msse4.1. 6550 * Support for SSE4.2 built-in functions and code generation are 6551 available via -msse4.2. 6552 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 6553 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 6554 allow explicit control of x87 floating point precision. 6555 * Support for __float128 (TFmode) IEEE quad type and corresponding 6556 TCmode IEEE complex quad type is available via the soft-fp library 6557 on x86_64 targets. This includes basic arithmetic operations 6558 (addition, subtraction, negation, multiplication and division) on 6559 __float128 real and TCmode complex values, the full set of IEEE 6560 comparisons between __float128 values, conversions to and from 6561 float, double and long double floating point types, as well as 6562 conversions to and from signed or unsigned integer, signed or 6563 unsigned long integer and signed or unsigned quad (TImode) integer 6564 types. Additionally, all operations generate the full set of IEEE 6565 exceptions and support the full set of IEEE rounding modes. 6566 * GCC can now utilize the ACML library for vectorizing calls to a set 6567 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 6568 link to an ACML ABI compatible library. 6569 6570 ARM 6571 6572 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 6573 has been added. 6574 6575 CRIS 6576 6577 New features 6578 6579 * Compiler and Library support for the CRIS v32 architecture, as 6580 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 6581 added. 6582 6583 Configuration changes 6584 6585 * The cris-*-elf target now includes support for CRIS v32, including 6586 libraries, through the -march=v32 option. 6587 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 6588 * A new crisv32-*-linux* target defaults to generate code for CRIS 6589 v32. 6590 * The cris-*-aout target has been obsoleted. 6591 6592 Improved support for built-in functions 6593 6594 * GCC can now use the lz and swapwbr instructions to implement the 6595 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 6596 * __builtin_bswap32 is now implemented using the swapwb instruction, 6597 when available. 6598 6599 m68k and ColdFire 6600 6601 New features 6602 6603 * Support for several new ColdFire processors has been added. You can 6604 generate code for them using the new -mcpu option. 6605 * All targets now support ColdFire processors. 6606 * m68k-uclinux targets have improved support for C++ constructors and 6607 destructors, and for shared libraries. 6608 * It is now possible to set breakpoints on the first or last line of 6609 a function, even if there are no statements on that line. 6610 6611 Optimizations 6612 6613 * Support for sibling calls has been added. 6614 * More use is now made of the ColdFire mov3q instruction. 6615 * __builtin_clz is now implemented using the ff1 ColdFire 6616 instruction, when available. 6617 * GCC now honors the -m68010 option. 68010 code now uses clr rather 6618 than move to zero volatile memory. 6619 * 68020 targets and above can now use symbol(index.size*scale) 6620 addresses for indexed array accesses. Earlier compilers would 6621 always load the symbol into a base register first. 6622 6623 Configuration changes 6624 6625 * All m68k and ColdFire targets now allow the default processor to be 6626 set at configure time using --with-cpu. 6627 * A --with-arch configuration option has been added. This option 6628 allows you to restrict a target to ColdFire or non-ColdFire 6629 processors. 6630 6631 Preprocessor macros 6632 6633 * An __mcfv*__ macro is now defined for all ColdFire targets. 6634 (Earlier versions of GCC only defined __mcfv4e__.) 6635 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 6636 * All targets now define __mc68010 and __mc68010__ when generating 6637 68010 code. 6638 6639 Command-line changes 6640 6641 * New command-line options -march, -mcpu, -mtune and -mhard-float 6642 have been added. These options apply to both m68k and ColdFire 6643 targets. 6644 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 6645 versions of -mshort, etc. 6646 * -fforce-addr has been removed. It is now ignored by the compiler. 6647 6648 Other improvements 6649 6650 * ColdFire targets now try to maintain a 4-byte-aligned stack where 6651 possible. 6652 * m68k-uclinux targets now try to avoid situations that lead to the 6653 load-time error: BINFMT_FLAT: reloc outside program. 6654 6655 MIPS 6656 6657 Changes to existing configurations 6658 6659 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 6660 and n64. Every GNU/Linux configuration now builds these libraries 6661 by default. 6662 * GNU/Linux configurations now generate -mno-shared code unless 6663 overridden by -fpic, -fPIC, -fpie or -fPIE. 6664 * mipsisa32*-linux-gnu configurations now generate hard-float code by 6665 default, just like other mipsisa32* and mips*-linux-gnu 6666 configurations. You can build a soft-float version of any 6667 mips*-linux-gnu configuration by passing --with-float=soft to 6668 configure. 6669 * mips-wrs-vxworks now supports run-time processes (RTPs). 6670 6671 Changes to existing command-line options 6672 6673 * The -march and -mtune options no longer accept 24k as a processor 6674 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 6675 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 6676 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 6677 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 6678 24kx, 24kex and 34kx. 6679 6680 New configurations 6681 6682 GCC now supports the following configurations: 6683 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 6684 default. Earlier releases also recognized this configuration, but 6685 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 6686 you can customize any mips*-linux-gnu* configuration to a 6687 particular ISA or processor by passing an appropriate --with-arch 6688 option to configure. 6689 * mipsisa*-sde-elf*, which provides compatibility with MIPS 6690 Technologies' SDE toolchains. The configuration uses the SDE 6691 libraries by default, but you can use it like other newlib-based 6692 ELF configurations by passing --with-newlib to configure. It is the 6693 only configuration besides mips64vr*-elf* to build MIPS16 as well 6694 as non-MIPS16 libraries. 6695 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 6696 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 6697 and 64-bit forms of the EABI. 6698 6699 New processors and application-specific extensions 6700 6701 * Support for the SmartMIPS ASE is available through the new 6702 -msmartmips option. 6703 * Support for revision 2 of the DSP ASE is available through the new 6704 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 6705 indicates the revision of the ASE in use. 6706 * Support for the 4KS and 74K families of processors is available 6707 through the -march and -mtune options. 6708 6709 Improved support for built-in functions 6710 6711 * GCC can now use load-linked, store-conditional and sync 6712 instructions to implement atomic built-in functions such as 6713 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 6714 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 6715 * GCC can now use the clz and dclz instructions to implement the 6716 __builtin_ctz and __builtin_ffs families of functions. 6717 * There is a new __builtin___clear_cache function for flushing the 6718 instruction cache. GCC expands this function inline on MIPS32 6719 revision 2 targets, otherwise it calls the function specified by 6720 -mcache-flush-func. 6721 6722 MIPS16 improvements 6723 6724 * GCC can now compile objects that contain a mixture of MIPS16 and 6725 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 6726 for specifying which mode a function should use. 6727 * A new option called -minterlink-mips16 makes non-MIPS16 code 6728 link-compatible with MIPS16 code. 6729 * After many bug fixes, the long-standing MIPS16 -mhard-float support 6730 should now work fairly reliably. 6731 * GCC can now use the MIPS16e save and restore instructions. 6732 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 6733 with -G0 -fsection-anchors is often smaller than code compiled with 6734 -G8. However, please note that you must usually compile all objects 6735 in your application with the same -G option; see the documentation 6736 of -G for details. 6737 * A new option called-mcode-readable specifies which instructions are 6738 allowed to load from the code segment. -mcode-readable=yes is the 6739 default and says that any instruction may load from the code 6740 segment. The other alternatives are -mcode-readable=pcrel, which 6741 says that only PC-relative MIPS16 instructions may load from the 6742 code segment, and -mcode-readable=no, which says that no 6743 instruction may do so. Please see the documentation for more 6744 details, including example uses. 6745 6746 Small-data improvements 6747 6748 There are three new options for controlling small data: 6749 * -mno-extern-sdata, which disables small-data accesses for 6750 externally-defined variables. Code compiled with -Gn 6751 -mno-extern-sdata will be link-compatible with any -G setting 6752 between -G0 and -Gn inclusive. 6753 * -mno-local-sdata, which disables the use of small-data sections for 6754 data that is not externally visible. This option can be a useful 6755 way of reducing small-data usage in less performance-critical parts 6756 of an application. 6757 * -mno-gpopt, which disables the use of the $gp register while still 6758 honoring the -G limit when placing externally-visible data. This 6759 option implies -mno-extern-sdata and -mno-local-sdata and it can be 6760 useful in situations where $gp does not necessarily hold the 6761 expected value. 6762 6763 Miscellaneous improvements 6764 6765 * There is a new option called -mbranch-cost for tweaking the 6766 perceived cost of branches. 6767 * If GCC is configured to use a version of GAS that supports the 6768 .gnu_attribute directive, it will use that directive to record 6769 certain properties of the output code. .gnu_attribute is new to GAS 6770 2.18. 6771 * There are two new function attributes, near and far, for overriding 6772 the command-line setting of -mlong-calls on a function-by-function 6773 basis. 6774 * -mfp64, which previously required a 64-bit target, now works with 6775 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 6776 mipsisa*-sde-elf* configurations provide suitable library support. 6777 * GCC now recognizes the -mdmx and -mmt options and passes them down 6778 to the assembler. It does nothing else with the options at present. 6779 6780 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 6781 (BEA) 6782 6783 * Support has been added for this new architecture. 6784 6785 RS6000 (POWER/PowerPC) 6786 6787 * Support for the PowerPC 750CL paired-single instructions has been 6788 added with a new powerpc-*-linux*paired* target configuration. It 6789 is enabled by an associated -mpaired option and can be accessed 6790 using new built-in functions. 6791 * Support for auto-detecting architecture and system configuration to 6792 auto-select processor optimization tuning. 6793 * Support for VMX on AIX 5.3 has been added. 6794 * Support for AIX Version 6.1 has been added. 6795 6796 S/390, zSeries and System z9 6797 6798 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 6799 added. When using the -march=z9-ec option, the compiler will 6800 generate code making use of instructions provided by the decimal 6801 floating point facility and the floating point conversion facility 6802 (pfpo). Besides the instructions used to implement decimal floating 6803 point operations these facilities also contain instructions to move 6804 between general purpose and floating point registers and to modify 6805 and copy the sign-bit of floating point values. 6806 * When the -march=z9-ec option is used the new 6807 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 6808 decimal floating point hardware instructions will be used or not. 6809 If none of them is given the hardware support is enabled by 6810 default. 6811 * The -mstack-guard option can now be omitted when using stack 6812 checking via -mstack-size in order to let GCC choose a sensible 6813 stack guard value according to the frame size of each function. 6814 * Various changes to improve performance of generated code have been 6815 implemented, including: 6816 + The condition code set by an add logical with carry 6817 instruction is now available for overflow checks like: a + b + 6818 carry < b. 6819 + The test data class instruction is now used to implement 6820 sign-bit and infinity checks of binary and decimal floating 6821 point numbers. 6822 6823 SPARC 6824 6825 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 6826 added. 6827 6828 Xtensa 6829 6830 * Stack unwinding for exception handling now uses by default a 6831 specialized version of DWARF unwinding. This is not 6832 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 6833 Xtensa with previous versions of GCC. 6834 * For Xtensa processors that include the Conditional Store option, 6835 the built-in functions for atomic memory access are now implemented 6836 using S32C1I instructions. 6837 * If the Xtensa NSA option is available, GCC will use it to implement 6838 the __builtin_ctz and __builtin_clz functions. 6839 6840Documentation improvements 6841 6842 * Existing libstdc++ documentation has been edited and restructured 6843 into a single DocBook XML manual. The results can be viewed online 6844 [22]here. 6845 6846Other significant improvements 6847 6848 * The compiler's --help command-line option has been extended so that 6849 it now takes an optional set of arguments. These arguments restrict 6850 the information displayed to specific classes of command-line 6851 options, and possibly only a subset of those options. It is also 6852 now possible to replace the descriptive text associated with each 6853 displayed option with an indication of its current value, or for 6854 binary options, whether it has been enabled or disabled. 6855 Here are some examples. The following will display all the options 6856 controlling warning messages: 6857 --help=warnings 6858 6859 Whereas this will display all the undocumented, target specific 6860 options: 6861 --help=target,undocumented 6862 6863 This sequence of commands will display the binary optimizations 6864 that are enabled by -O3: 6865 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 6866 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 6867 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 6868 6869 * The configure options --with-pkgversion and --with-bugurl have been 6870 added. These allow distributors of GCC to include a 6871 distributor-specific string in manuals and --version output and to 6872 specify the URL for reporting bugs in their versions of GCC. 6873 6874GCC 4.3.1 6875 6876 This is the [23]list of problem reports (PRs) from GCC's bug tracking 6877 system that are known to be fixed in the 4.3.1 release. This list might 6878 not be complete (that is, it is possible that some PRs that have been 6879 fixed are not listed here). 6880 6881Target Specific Changes 6882 6883 IA-32/x86-64 6884 6885 ABI changes 6886 6887 * Starting with GCC 4.3.1, decimal floating point variables are 6888 aligned to their natural boundaries when they are passed on the 6889 stack for i386. 6890 6891 Command-line changes 6892 6893 * Starting with GCC 4.3.1, the -mcld option has been added to 6894 automatically generate a cld instruction in the prologue of 6895 functions that use string instructions. This option is used for 6896 backward compatibility on some operating systems and can be enabled 6897 by default for 32-bit x86 targets by configuring GCC with the 6898 --enable-cld configure option. 6899 6900GCC 4.3.2 6901 6902 This is the [24]list of problem reports (PRs) from GCC's bug tracking 6903 system that are known to be fixed in the 4.3.2 release. This list might 6904 not be complete (that is, it is possible that some PRs that have been 6905 fixed are not listed here). 6906 6907GCC 4.3.3 6908 6909 This is the [25]list of problem reports (PRs) from GCC's bug tracking 6910 system that are known to be fixed in the 4.3.3 release. This list might 6911 not be complete (that is, it is possible that some PRs that have been 6912 fixed are not listed here). 6913 6914GCC 4.3.4 6915 6916 This is the [26]list of problem reports (PRs) from GCC's bug tracking 6917 system that are known to be fixed in the 4.3.4 release. This list might 6918 not be complete (that is, it is possible that some PRs that have been 6919 fixed are not listed here). 6920 6921GCC 4.3.5 6922 6923 This is the [27]list of problem reports (PRs) from GCC's bug tracking 6924 system that are known to be fixed in the 4.3.5 release. This list might 6925 not be complete (that is, it is possible that some PRs that have been 6926 fixed are not listed here). 6927 6928GCC 4.3.6 6929 6930 This is the [28]list of problem reports (PRs) from GCC's bug tracking 6931 system that are known to be fixed in the 4.3.6 release. This list might 6932 not be complete (that is, it is possible that some PRs that have been 6933 fixed are not listed here). 6934 6935 6936 For questions related to the use of GCC, please consult these web 6937 pages and the [29]GCC manuals. If that fails, the 6938 [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these 6939 web pages and the development of GCC are welcome on our developer 6940 list at [31]gcc@gcc.gnu.org. All of [32]our lists have public 6941 archives. 6942 6943 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 6944 distribution of this entire article is permitted in any medium, 6945 provided this notice is preserved. 6946 6947 These pages are [34]maintained by the GCC team. Last modified 6948 2016-01-30[35]. 6949 6950References 6951 6952 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 6953 2. https://gmplib.org/ 6954 3. http://www.mpfr.org/ 6955 4. https://gcc.gnu.org/install/prerequisites.html 6956 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 6957 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 6958 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 6959 8. http://www.mpfr.org/ 6960 9. http://www.mpfr.org/ 6961 10. http://www.mpfr.org/ 6962 11. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 6963 12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 6964 13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 6965 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#m anual.intro.status.standard.tr1 6966 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 6967 16. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 6968 17. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 6969 18. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 6970 19. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 6971 20. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 6972 21. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 6973 22. https://gcc.gnu.org/onlinedocs/libstdc++/ 6974 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 6975 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 6976 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 6977 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 6978 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 6979 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 6980 29. https://gcc.gnu.org/onlinedocs/ 6981 30. mailto:gcc-help@gcc.gnu.org 6982 31. mailto:gcc@gcc.gnu.org 6983 32. https://gcc.gnu.org/lists.html 6984 33. http://www.fsf.org/ 6985 34. https://gcc.gnu.org/about.html 6986 35. http://validator.w3.org/check/referer 6987====================================================================== 6988http://gcc.gnu.org/gcc-4.2/index.html 6989 6990 GCC 4.2 Release Series 6991 6992 May 19, 2008 6993 6994 The [1]GNU project and the GCC developers are pleased to announce the 6995 release of GCC 4.2.4. 6996 6997 This release is a bug-fix release, containing fixes for regressions in 6998 GCC 4.2.3 relative to previous releases of GCC. 6999 7000Release History 7001 7002 GCC 4.2.4 7003 May 19, 2008 ([2]changes) 7004 7005 GCC 4.2.3 7006 February 1, 2008 ([3]changes) 7007 7008 GCC 4.2.2 7009 October 7, 2007 ([4]changes) 7010 7011 GCC 4.2.1 7012 July 18, 2007 ([5]changes) 7013 7014 GCC 4.2.0 7015 May 13, 2007 ([6]changes) 7016 7017References and Acknowledgements 7018 7019 GCC used to stand for the GNU C Compiler, but since the compiler 7020 supports several other languages aside from C, it now stands for the 7021 GNU Compiler Collection. 7022 7023 A list of [7]successful builds is updated as new information becomes 7024 available. 7025 7026 The GCC developers would like to thank the numerous people that have 7027 contributed new features, improvements, bug fixes, and other changes as 7028 well as test results to GCC. This [8]amazing group of volunteers is 7029 what makes GCC successful. 7030 7031 For additional information about GCC please refer to the [9]GCC project 7032 web site or contact the [10]GCC development mailing list. 7033 7034 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 7035 7036 7037 For questions related to the use of GCC, please consult these web 7038 pages and the [13]GCC manuals. If that fails, the 7039 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 7040 web pages and the development of GCC are welcome on our developer 7041 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 7042 archives. 7043 7044 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 7045 distribution of this entire article is permitted in any medium, 7046 provided this notice is preserved. 7047 7048 These pages are [18]maintained by the GCC team. Last modified 7049 2016-01-30[19]. 7050 7051References 7052 7053 1. http://www.gnu.org/ 7054 2. http://gcc.gnu.org/gcc-4.2/changes.html 7055 3. http://gcc.gnu.org/gcc-4.2/changes.html 7056 4. http://gcc.gnu.org/gcc-4.2/changes.html 7057 5. http://gcc.gnu.org/gcc-4.2/changes.html 7058 6. http://gcc.gnu.org/gcc-4.2/changes.html 7059 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 7060 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7061 9. http://gcc.gnu.org/index.html 7062 10. mailto:gcc@gcc.gnu.org 7063 11. http://gcc.gnu.org/mirrors.html 7064 12. http://gcc.gnu.org/svn.html 7065 13. https://gcc.gnu.org/onlinedocs/ 7066 14. mailto:gcc-help@gcc.gnu.org 7067 15. mailto:gcc@gcc.gnu.org 7068 16. https://gcc.gnu.org/lists.html 7069 17. http://www.fsf.org/ 7070 18. https://gcc.gnu.org/about.html 7071 19. http://validator.w3.org/check/referer 7072====================================================================== 7073http://gcc.gnu.org/gcc-4.2/changes.html 7074 7075 GCC 4.2 Release Series 7076 Changes, New Features, and Fixes 7077 7078Caveats 7079 7080 * GCC no longer accepts the -fshared-data option. This option has had 7081 no effect in any GCC 4 release; the targets to which the option 7082 used to apply had been removed before GCC 4.0. 7083 7084General Optimizer Improvements 7085 7086 * New command-line options specify the possible relationships among 7087 parameters and between parameters and global data. For example, 7088 -fargument-noalias-anything specifies that arguments do not alias 7089 any other storage. 7090 Each language will automatically use whatever option is required by 7091 the language standard. You should not need to use these options 7092 yourself. 7093 7094New Languages and Language specific improvements 7095 7096 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 7097 * New command-line options -fstrict-overflow and -Wstrict-overflow 7098 have been added. -fstrict-overflow tells the compiler that it may 7099 assume that the program follows the strict signed overflow 7100 semantics permitted for the language: for C and C++ this means that 7101 the compiler may assume that signed overflow does not occur. For 7102 example, a loop like 7103 for (i = 1; i > 0; i *= 2) 7104 7105 is presumably intended to continue looping until i overflows. With 7106 -fstrict-overflow, the compiler may assume that signed overflow 7107 will not occur, and transform this into an infinite loop. 7108 -fstrict-overflow is turned on by default at -O2, and may be 7109 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 7110 be used to warn about cases where the compiler assumes that signed 7111 overflow will not occur. It takes five different levels: 7112 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 7113 -Wstrict-overflow=1 is enabled by -Wall. 7114 * The new command-line option -fno-toplevel-reorder directs GCC to 7115 emit top-level functions, variables, and asm statements in the same 7116 order that they appear in the input file. This is intended to 7117 support existing code which relies on a particular ordering (for 7118 example, code which uses top-level asm statements to switch 7119 sections). For new code, it is generally better to use function and 7120 variable attributes. The -fno-toplevel-reorder option may be used 7121 for most cases which currently use -fno-unit-at-a-time. The 7122 -fno-unit-at-a-time option will be removed in some future version 7123 of GCC. If you know of a case which requires -fno-unit-at-a-time 7124 which is not fixed by -fno-toplevel-reorder, please open a bug 7125 report. 7126 7127 C family 7128 7129 * The pragma redefine_extname will now macro expand its tokens for 7130 compatibility with SunPRO. 7131 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 7132 GCC to handle inline functions as specified in the C99 standard. In 7133 preparation for this, GCC 4.2 will warn about any use of non-static 7134 inline functions in gnu99 or c99 mode. This new warning may be 7135 disabled with the new gnu_inline function attribute or the new 7136 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 7137 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 7138 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 7139 in the current compilation. 7140 * A new command-line option -Waddress has been added to warn about 7141 suspicious uses of memory addresses as, for example, using the 7142 address of a function in a conditional expression, and comparisons 7143 against the memory address of a string literal. This warning is 7144 enabled by -Wall. 7145 7146 C++ 7147 7148 * C++ visibility handling has been overhauled. 7149 Restricted visiblity is propagated from classes to members, from 7150 functions to local statics, and from templates and template 7151 arguments to instantiations, unless the latter has explicitly 7152 declared visibility. 7153 The visibility attribute for a class must come between the 7154 class-key and the name, not after the closing brace. 7155 Attributes are now allowed for enums and elaborated-type-specifiers 7156 that only declare a type. 7157 Members of the anonymous namespace are now local to a particular 7158 translation unit, along with any other declarations which use them, 7159 though they are still treated as having external linkage for 7160 language semantics. 7161 * The (undocumented) extension which permitted templates with default 7162 arguments to be bound to template template parameters with fewer 7163 parameters has been removed. For example: 7164 template <template <typename> class C> 7165 void f(C<double>) {} 7166 7167 template <typename T, typename U = int> 7168 struct S {}; 7169 7170 template void f(S<double>); 7171 7172 is no longer accepted by G++. The reason this code is not accepted 7173 is that S is a template with two parameters; therefore, it cannot 7174 be bound to C which has only one parameter. 7175 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 7176 releases, have been removed. 7177 * The command-line option -fconst-strings, deprecated in previous GCC 7178 releases, has been removed. 7179 * The configure variable enable-__cxa_atexit is now enabled by 7180 default for more targets. Enabling this variable is necessary in 7181 order for static destructors to be executed in the correct order, 7182 but it depends upon the presence of a non-standard C library in the 7183 target library in order to work. The variable is now enabled for 7184 more targets which are known to have suitable C libraries. 7185 * -Wextra will produce warnings for if statements with a semicolon as 7186 the only body, to catch code like: 7187 if (a); 7188 return 1; 7189 return 0; 7190 7191 To suppress the warning in valid cases, use { } instead. 7192 * The C++ frontend now also produces strict aliasing warnings when 7193 -fstrict-aliasing -Wstrict-aliasing is in effect. 7194 7195 Runtime Library (libstdc++) 7196 7197 * Added support for TR1 <random>, <complex>, and C compatibility 7198 headers. In addition, a lock-free version of shared_ptr was 7199 contributed as part of Phillip Jordan's Google Summer of Code 7200 project on lock-free containers. ([3]Implementation status of TR1) 7201 * In association with the Summer of Code work on lock-free 7202 containers, the interface for atomic builtins was adjusted, 7203 creating simpler alternatives for non-threaded code paths. Also, 7204 usage was consolidated and all elements were moved from namespace 7205 std to namespace__gnu_cxx. Affected interfaces are the functions 7206 __exchange_and_add, __atomic_add, and the objects __mutex, 7207 __recursive_mutex, and __scoped_lock. 7208 * Support for versioning weak symbol names via namespace association 7209 was added. However, as this changes the names of exported symbols, 7210 this is turned off by default in the current ABI. Intrepid users 7211 can enable this feature by using 7212 --enable-symvers=gnu-versioned-namespace during configuration. 7213 * Revised, simplified, and expanded policy-based associative 7214 containers, including data types for tree and trie forms 7215 (basic_tree, tree, trie), lists (list_update), and both 7216 collision-chaining and probing hash-based containers 7217 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 7218 the [4]documentation. 7219 * The implementation of the debug mode was modified, whereby the 7220 debug namespaces were nested inside of namespace std and namespace 7221 __gnu_cxx in order to resolve some long standing corner cases 7222 involving name lookup. Debug functionality from the policy-based 7223 data structures was consolidated and enabled with the single macro, 7224 _GLIBCXX_DEBUG. See PR 26142 for more information. 7225 * Added extensions for type traits: __conditional_type, 7226 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 7227 * Added a typelist implementation for compile-time meta-programming. 7228 Elements for typelist construction and operation can be found 7229 within namespace __gnu_cxx::typelist. 7230 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 7231 exception-safety. 7232 * Enabled library-wide visibility control, allowing -fvisibility to 7233 be used. 7234 * Consolidated all nested namespaces and the conversion of 7235 __gnu_internal implementation-private details to anonymous 7236 namespaces whenever possible. 7237 * Implemented LWG resolutions DR 431 and DR 538. 7238 7239 Fortran 7240 7241 * Support for allocatable components has been added (TR 15581 and 7242 Fortran 2003). 7243 * Support for the Fortran 2003 streaming IO extension has been added. 7244 * The GNU Fortran compiler now uses 4-byte record markers by default 7245 for unformatted files to be compatible with g77 and most other 7246 compilers. The implementation allows for records greater than 2 GB 7247 and is compatible with several other compilers. Older versions of 7248 gfortran used 8-byte record markers by default (on most systems). 7249 In order to change the length of the record markers, e.g. to read 7250 unformatted files created by older gfortran versions, the 7251 [5]-frecord-marker=8 option can be used. 7252 7253 Java (GCJ) 7254 7255 * A new command-line option -static-libgcj has been added for targets 7256 that use a linker compatible with GNU Binutils. As its name 7257 implies, this causes libgcj to be linked statically. In some cases 7258 this causes the resulting executable to start faster and use less 7259 memory than if the shared version of libgcj were used. However 7260 caution should be used as it can also cause essential parts of the 7261 library to be omitted. Some of these issues are discussed in: 7262 [6]https://gcc.gnu.org/wiki/Statically_linking_libgcj 7263 * fastjar is no longer bundled with GCC. To build libgcj, you will 7264 need either InfoZIP (both zip and unzip) or an external jar 7265 program. In the former case, the GCC build will install a jar shell 7266 script that is based on InfoZIP and provides the same functionality 7267 as fastjar. 7268 7269New Targets and Target Specific Improvements 7270 7271 IA-32/x86-64 7272 7273 * -mtune=generic can now be used to generate code running well on 7274 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 7275 Pentium-M, Intel Pentium 4 and Intel Core 2. 7276 * -mtune=native and -march=native will produce code optimized for the 7277 host architecture as detected using the cpuid instruction. 7278 * Added a new command-line option -fstackrealign and and 7279 __attribute__ ((force_align_arg_pointer)) to realign the stack at 7280 runtime. This allows functions compiled with a vector-aligned stack 7281 to be invoked from legacy objects that keep only word-alignment. 7282 7283 SPARC 7284 7285 * The default CPU setting has been changed from V7 to V9 in 32-bit 7286 mode on Solaris 7 and above. This is already the case in 64-bit 7287 mode. It can be overridden by specifying --with-cpu at configure 7288 time. 7289 * Back-end support of built-in functions for atomic memory access has 7290 been implemented. 7291 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 7292 added. 7293 7294 M32C 7295 7296 * Various bug fixes have made some functions (notably, functions 7297 returning structures) incompatible with previous releases. 7298 Recompiling all libraries is recommended. Note that code quality 7299 has considerably improved since 4.1, making a recompile even more 7300 beneficial. 7301 7302 MIPS 7303 7304 * Added support for the Broadcom SB-1A core. 7305 7306 IA-64 7307 7308 * Added support for IA-64 data and control speculation. By default 7309 speculation is enabled only during second scheduler pass. A number 7310 of machine flags was introduced to control the usage of speculation 7311 for both scheduler passes. 7312 7313 HPPA 7314 7315 * Added Java language support (libffi and libjava) for 32-bit HP-UX 7316 11 target. 7317 7318Obsolete Systems 7319 7320Documentation improvements 7321 7322 PDF Documentation 7323 7324 * A make pdf target has been added to the top-level makefile, 7325 enabling automated production of PDF documentation files. 7326 (Front-ends external to GCC should modify their Make-lang.in file 7327 to add a lang.pdf: target.) 7328 7329Other significant improvements 7330 7331 Build system improvements 7332 7333 * All the components of the compiler are now bootstrapped by default. 7334 This improves the resilience to bugs in the system compiler or 7335 binary compatibility problems, as well as providing better testing 7336 of GCC 4.2 itself. In addition, if you build the compiler from a 7337 combined tree, the assembler, linker, etc. will also be 7338 bootstrapped (i.e. built with themselves). 7339 You can disable this behavior, and go back to the pre-GCC 4.2 set 7340 up, by configuring GCC with --disable-bootstrap. 7341 * The rules that configure follows to find target tools resemble more 7342 closely the locations that the built compiler will search. In 7343 addition, you can use the new configure option --with-target-tools 7344 to specify where to find the target tools used during the build, 7345 without affecting what the built compiler will use. 7346 This can be especially useful when building packages of GCC. For 7347 example, you may want to build GCC with GNU as or ld, even if the 7348 resulting compiler to work with the native assembler and linker. To 7349 do so, you can use --with-target-tools to point to the native 7350 tools. 7351 7352 Incompatible changes to the build system 7353 7354 * Front-ends external to GCC should modify their Make-lang.in file to 7355 replace double-colon rules (e.g. dvi::) with normal rules (like 7356 lang.dvi:). Front-end makefile hooks do not use double-colon rules 7357 anymore. 7358 * Up to GCC 4.1, a popular way to specify the target tools used 7359 during the build was to create directories named gas, binutils, 7360 etc. in the build tree, and create links to the tools from there. 7361 This does not work any more when the compiler is bootstrapped. The 7362 new configure option --with-target-tools provides a better way to 7363 achieve the same effect, and works for all native and cross 7364 settings. 7365 7366 7367 For questions related to the use of GCC, please consult these web 7368 pages and the [7]GCC manuals. If that fails, the 7369 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these 7370 web pages and the development of GCC are welcome on our developer 7371 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public 7372 archives. 7373 7374 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 7375 distribution of this entire article is permitted in any medium, 7376 provided this notice is preserved. 7377 7378 These pages are [12]maintained by the GCC team. Last modified 7379 2016-01-30[13]. 7380 7381References 7382 7383 1. http://gcc.gnu.org/projects/gomp/ 7384 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 7385 3. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.tr1 7386 4. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 7387 5. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 7388 6. https://gcc.gnu.org/wiki/Statically_linking_libgcj 7389 7. https://gcc.gnu.org/onlinedocs/ 7390 8. mailto:gcc-help@gcc.gnu.org 7391 9. mailto:gcc@gcc.gnu.org 7392 10. https://gcc.gnu.org/lists.html 7393 11. http://www.fsf.org/ 7394 12. https://gcc.gnu.org/about.html 7395 13. http://validator.w3.org/check/referer 7396====================================================================== 7397http://gcc.gnu.org/gcc-4.1/index.html 7398 7399 GCC 4.1 Release Series 7400 7401 February 13, 2007 7402 7403 The [1]GNU project and the GCC developers are pleased to announce the 7404 release of GCC 4.1.2. 7405 7406 This release is a bug-fix release, containing fixes for regressions in 7407 GCC 4.1.1 relative to previous releases of GCC. 7408 7409Release History 7410 7411 GCC 4.1.2 7412 February 13, 2007 ([2]changes) 7413 7414 GCC 4.1.1 7415 May 24, 2006 ([3]changes) 7416 7417 GCC 4.1.0 7418 February 28, 2006 ([4]changes) 7419 7420References and Acknowledgements 7421 7422 GCC used to stand for the GNU C Compiler, but since the compiler 7423 supports several other languages aside from C, it now stands for the 7424 GNU Compiler Collection. 7425 7426 A list of [5]successful builds is updated as new information becomes 7427 available. 7428 7429 The GCC developers would like to thank the numerous people that have 7430 contributed new features, improvements, bug fixes, and other changes as 7431 well as test results to GCC. This [6]amazing group of volunteers is 7432 what makes GCC successful. 7433 7434 For additional information about GCC please refer to the [7]GCC project 7435 web site or contact the [8]GCC development mailing list. 7436 7437 To obtain GCC please use [9]our mirror sites or [10]our SVN server. 7438 7439 7440 For questions related to the use of GCC, please consult these web 7441 pages and the [11]GCC manuals. If that fails, the 7442 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these 7443 web pages and the development of GCC are welcome on our developer 7444 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public 7445 archives. 7446 7447 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 7448 distribution of this entire article is permitted in any medium, 7449 provided this notice is preserved. 7450 7451 These pages are [16]maintained by the GCC team. Last modified 7452 2016-01-30[17]. 7453 7454References 7455 7456 1. http://www.gnu.org/ 7457 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 7458 3. http://gcc.gnu.org/gcc-4.1/changes.html 7459 4. http://gcc.gnu.org/gcc-4.1/changes.html 7460 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 7461 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7462 7. http://gcc.gnu.org/index.html 7463 8. mailto:gcc@gcc.gnu.org 7464 9. http://gcc.gnu.org/mirrors.html 7465 10. http://gcc.gnu.org/svn.html 7466 11. https://gcc.gnu.org/onlinedocs/ 7467 12. mailto:gcc-help@gcc.gnu.org 7468 13. mailto:gcc@gcc.gnu.org 7469 14. https://gcc.gnu.org/lists.html 7470 15. http://www.fsf.org/ 7471 16. https://gcc.gnu.org/about.html 7472 17. http://validator.w3.org/check/referer 7473====================================================================== 7474http://gcc.gnu.org/gcc-4.1/changes.html 7475 7476 GCC 4.1 Release Series 7477 Changes, New Features, and Fixes 7478 7479 The latest release in the 4.1 release series is [1]GCC 4.1.2. 7480 7481Caveats 7482 7483General Optimizer Improvements 7484 7485 * GCC now has infrastructure for inter-procedural optimizations and 7486 the following inter-procedural optimizations are implemented: 7487 + Profile guided inlining. When doing profile feedback guided 7488 optimization, GCC can now use the profile to make better 7489 informed decisions on whether inlining of a function is 7490 profitable or not. This means that GCC will no longer inline 7491 functions at call sites that are not executed very often, and 7492 that functions at hot call sites are more likely to be 7493 inlined. 7494 A new parameter min-inline-recursive-probability is also now 7495 available to throttle recursive inlining of functions with 7496 small average recursive depths. 7497 + Discovery of pure and const functions, a form of side-effects 7498 analysis. While older GCC releases could also discover such 7499 special functions, the new IPA-based pass runs earlier so that 7500 the results are available to more optimizers. The pass is also 7501 simply more powerful than the old one. 7502 + Analysis of references to static variables and type escape 7503 analysis, also forms of side-effects analysis. The results of 7504 these passes allow the compiler to be less conservative about 7505 call-clobbered variables and references. This results in more 7506 redundant loads being eliminated and in making static 7507 variables candidates for register promotion. 7508 + Improvement of RTL-based alias analysis. The results of type 7509 escape analysis are fed to the RTL type-based alias analyzer, 7510 allowing it to disambiguate more memory references. 7511 + Interprocedural constant propagation and function versioning. 7512 This pass looks for functions that are always called with the 7513 same constant value for one or more of the function arguments, 7514 and propagates those constants into those functions. 7515 + GCC will now eliminate static variables whose usage was 7516 optimized out. 7517 + -fwhole-program --combine can now be used to make all 7518 functions in program static allowing whole program 7519 optimization. As an exception, the main function and all 7520 functions marked with the new externally_visible attribute are 7521 kept global so that programs can link with runtime libraries. 7522 * GCC can now do a form of partial dead code elimination (PDCE) that 7523 allows code motion of expressions to the paths where the result of 7524 the expression is actually needed. This is not always a win, so the 7525 pass has been limited to only consider profitable cases. Here is an 7526 example: 7527 int foo (int *, int *); 7528 int 7529 bar (int d) 7530 { 7531 int a, b, c; 7532 b = d + 1; 7533 c = d + 2; 7534 a = b + c; 7535 if (d) 7536 { 7537 foo (&b, &c); 7538 a = b + c; 7539 } 7540 printf ("%d\n", a); 7541 } 7542 7543 The a = b + c can be sunk to right before the printf. Normal code 7544 sinking will not do this, it will sink the first one above into the 7545 else-branch of the conditional jump, which still gives you two 7546 copies of the code. 7547 * GCC now has a value range propagation pass. This allows the 7548 compiler to eliminate bounds checks and branches. The results of 7549 the pass can also be used to accurately compute branch 7550 probabilities. 7551 * The pass to convert PHI nodes to straight-line code (a form of 7552 if-conversion for GIMPLE) has been improved significantly. The two 7553 most significant improvements are an improved algorithm to 7554 determine the order in which the PHI nodes are considered, and an 7555 improvement that allow the pass to consider if-conversions of basic 7556 blocks with more than two predecessors. 7557 * Alias analysis improvements. GCC can now differentiate between 7558 different fields of structures in Tree-SSA's virtual operands form. 7559 This lets stores/loads from non-overlapping structure fields not 7560 conflict. A new algorithm to compute points-to sets was contributed 7561 that can allows GCC to see now that p->a and p->b, where p is a 7562 pointer to a structure, can never point to the same field. 7563 * Various enhancements to auto-vectorization: 7564 + Incrementally preserve SSA form when vectorizing. 7565 + Incrementally preserve loop-closed form when vectorizing. 7566 + Improvements to peeling for alignment: generate better code 7567 when the misalignment of an access is known at compile time, 7568 or when different accesses are known to have the same 7569 misalignment, even if the misalignment amount itself is 7570 unknown. 7571 + Consider dependence distance in the vectorizer. 7572 + Externalize generic parts of data reference analysis to make 7573 this analysis available to other passes. 7574 + Vectorization of conditional code. 7575 + Reduction support. 7576 * GCC can now partition functions in sections of hot and cold code. 7577 This can significantly improve performance due to better 7578 instruction cache locality. This feature works best together with 7579 profile feedback driven optimization. 7580 * A new pass to avoid saving of unneeded arguments to the stack in 7581 vararg functions if the compiler can prove that they will not be 7582 needed. 7583 * Transition of basic block profiling to tree level implementation 7584 has been completed. The new implementation should be considerably 7585 more reliable (hopefully avoiding profile mismatch errors when 7586 using -fprofile-use or -fbranch-probabilities) and can be used to 7587 drive higher level optimizations, such as inlining. 7588 The -ftree-based-profiling command-line option was removed and 7589 -fprofile-use now implies disabling old RTL level loop optimizer 7590 (-fno-loop-optimize). Speculative prefetching optimization 7591 (originally enabled by -fspeculative-prefetching) was removed. 7592 7593New Languages and Language specific improvements 7594 7595 C and Objective-C 7596 7597 * The old Bison-based C and Objective-C parser has been replaced by a 7598 new, faster hand-written recursive-descent parser. 7599 7600 Ada 7601 7602 * The build infrastructure for the Ada runtime library and tools has 7603 been changed to be better integrated with the rest of the build 7604 infrastructure of GCC. This should make doing cross builds of Ada a 7605 bit easier. 7606 7607 C++ 7608 7609 * ARM-style name-injection of friend declarations is no longer the 7610 default. For example: 7611 struct S { 7612 friend void f(); 7613 }; 7614 7615 void g() { f(); } 7616 will not be accepted; instead a declaration of f will need to be 7617 present outside of the scope of S. The new -ffriend-injection 7618 option will enable the old behavior. 7619 * The (undocumented) extension which permitted templates with default 7620 arguments to be bound to template template parameters with fewer 7621 parameters has been deprecated, and will be removed in the next 7622 major release of G++. For example: 7623 template <template <typename> class C> 7624 void f(C<double>) {} 7625 7626 template <typename T, typename U = int> 7627 struct S {}; 7628 7629 template void f(S<double>); 7630 7631 makes use of the deprecated extension. The reason this code is not 7632 valid ISO C++ is that S is a template with two parameters; 7633 therefore, it cannot be bound to C which has only one parameter. 7634 7635 Runtime Library (libstdc++) 7636 7637 * Optimization work: 7638 + A new implementation of std::search_n is provided, better 7639 performing in case of random access iterators. 7640 + Added further efficient specializations of istream functions, 7641 i.e., character array and string extractors. 7642 + Other smaller improvements throughout. 7643 * Policy-based associative containers, designed for high-performance, 7644 flexibility and semantic safety are delivered in ext/pb_assoc. 7645 * A versatile string class, __gnu_cxx::__versa_string, providing 7646 facilities conforming to the standard requirements for 7647 basic_string, is delivered in <ext/vstring.h>. In particular: 7648 + Two base classes are provided: the default one avoids 7649 reference counting and is optimized for short strings; the 7650 alternate one, still uses it while improving in a few low 7651 level areas (e.g., alignment). See vstring_fwd.h for some 7652 useful typedefs. 7653 + Various algorithms have been rewritten (e.g., replace), the 7654 code streamlined and simple optimizations added. 7655 + Option 3 of DR 431 is implemented for both available bases, 7656 thus improving the support for stateful allocators. 7657 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 7658 libstdc++/23953) and LWG resolutions put into effect for the first 7659 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 7660 Issue 6.19). The implementation status of TR1 is now tracked in the 7661 docs in tr1.html. 7662 7663 Objective-C++ 7664 7665 * A new language front end for Objective-C++ has been added. This 7666 language allows users to mix the object oriented features of 7667 Objective-C with those of C++. 7668 7669 Java (GCJ) 7670 7671 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 7672 features (plus some 0.20 bug-fixes) 7673 + Networking 7674 o The java.net.HttpURLConnection implementation no longer 7675 buffers the entire response body in memory. This means 7676 that response bodies larger than available memory can now 7677 be handled. 7678 + (N)IO 7679 o NIO FileChannel.map implementation, fast bulk put 7680 implementation for DirectByteBuffer (speeds up this 7681 method 10x). 7682 o FileChannel.lock() and FileChannel.force() implemented. 7683 + XML 7684 o gnu.xml fix for nodes created outside a namespace 7685 context. 7686 o Add support for output indenting and 7687 cdata-section-elements output instruction in 7688 xml.transform. 7689 o xml.xpath corrections for cases where elements/attributes 7690 might have been created in non-namespace-aware mode. 7691 Corrections to handling of XSL variables and minor 7692 conformance updates. 7693 + AWT 7694 o GNU JAWT implementation, the AWT Native Interface, which 7695 allows direct access to native screen resources from 7696 within a Canvas's paint method. GNU Classpath Examples 7697 comes with a Demo, see libjava/classpath/examples/README. 7698 o awt.datatransfer updated to 1.5 with support for 7699 FlavorEvents. The gtk+ awt peers now allow copy/paste of 7700 text, images, URIs/files and serialized objects with 7701 other applications and tracking clipboard change events 7702 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 7703 objects are supported). A GNU Classpath Examples 7704 datatransfer Demo was added to show the new 7705 functionality. 7706 o Split gtk+ awt peers event handling in two threads and 7707 improve gdk lock handling (solves several awt lock ups). 7708 o Speed up awt Image loading. 7709 o Better gtk+ scrollbar peer implementation when using gtk+ 7710 >= 2.6. 7711 o Handle image loading errors correctly for gdkpixbuf and 7712 MediaTracker. 7713 o Better handle GDK lock. Properly prefix gtkpeer native 7714 functions (cp_gtk). 7715 o GdkGraphics2D has been updated to use Cairo 0.5.x or 7716 higher. 7717 o BufferedImage and GtkImage rewrites. All image drawing 7718 operations should now work correctly (flipping requires 7719 gtk+ >= 2.6) 7720 o Future Graphics2D, image and text work is documented at: 7721 [2]http://developer.classpath.org/mediation/ClasspathGrap 7722 hicsImagesText 7723 o When gtk+ 2.6 or higher is installed the default log 7724 handler will produce stack traces whenever a WARNING, 7725 CRITICAL or ERROR message is produced. 7726 + Free Swing 7727 o The RepaintManager has been reworked for more efficient 7728 painting, especially for large GUIs. 7729 o The layout manager OverlayLayout has been implemented, 7730 the BoxLayout has been rewritten to make use of the 7731 SizeRequirements utility class and caching for more 7732 efficient layout. 7733 o Improved accessibility support. 7734 o Significant progress has been made in the implementation 7735 of the javax.swing.plaf.metal package, with most UI 7736 delegates in a working state now. Please test this with 7737 your own applications and provide feedback that will help 7738 us to improve this package. 7739 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 7740 extended to highlight various features in our Free Swing 7741 implementation. And it includes a look and feel switcher 7742 for Metal (default), Ocean and GNU themes. 7743 o The javax.swing.plaf.multi package is now implemented. 7744 o Editing and several key actions for JTree and JTable were 7745 implemented. 7746 o Lots of icons and look and feel improvements for Free 7747 Swing basic and metal themes were added. Try running the 7748 GNU Classpath Swing Demo in examples 7749 (gnu.classpath.examples.swing.Demo) with: 7750 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 7751 l or 7752 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 7753 l 7754 o Start of styled text capabilites for java.swing.text. 7755 o DefaultMutableTreeNode pre-order, post-order, depth-first 7756 and breadth-first traversal enumerations implemented. 7757 o JInternalFrame colors and titlebar draw properly. 7758 o JTree is working up to par (icons, selection and keyboard 7759 traversal). 7760 o JMenus were made more compatible in visual and 7761 programmatic behavior. 7762 o JTable changeSelection and multiple selections 7763 implemented. 7764 o JButton and JToggleButton change states work properly 7765 now. 7766 o JFileChooser fixes. 7767 o revalidate() and repaint() fixes which make Free Swing 7768 much more responsive. 7769 o MetalIconFactory implemented. 7770 o Free Swing Top-Level Compatibility. JFrame, JDialog, 7771 JApplet, JInternalFrame, and JWindow are now 1.5 7772 compatible in the sense that you can call add() and 7773 setLayout() directly on them, which will have the same 7774 effect as calling getContentPane().add() and 7775 getContentPane().setLayout(). 7776 o The JTree interface has been completed. JTrees now 7777 recognizes mouse clicks and selections work. 7778 o BoxLayout works properly now. 7779 o Fixed GrayFilter to actually work. 7780 o Metal SplitPane implemented. 7781 o Lots of Free Swing text and editor stuff work now. 7782 + Free RMI and Corba 7783 o Andrew Watson, Vice President and Technical Director of 7784 the Object Management Group, has officially assigned us 7785 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 7786 mark remote classpath-specific system exceptions. 7787 Obtaining the VMCID means that GNU Classpath now is a 7788 recogniseable type of node in a highly interoperable 7789 CORBA world. 7790 o GNU Classpath now includes the first working draft to 7791 support the RMI over IIOP protocol. The current 7792 implementation is capable of remote invocations, 7793 transferring various Serializables and Externalizables 7794 via RMI-IIOP protocol. It can flatten graphs and, at 7795 least for the simple cases, is interoperable with 1.5 7796 JDKs. 7797 o org.omg.PortableInterceptor and related functionality in 7798 other packages is now implemented: 7799 # The sever and client interceptors work as required 7800 since 1.4. 7801 # The IOR interceptor works as needed for 1.5. 7802 o The org.omg.DynamicAny package is completed and passes 7803 the prepared tests. 7804 o The Portable Object Adapter should now support the output 7805 of the recent IDL to java compilers. These compilers now 7806 generate servants and not CORBA objects as before, making 7807 the output depend on the existing POA implementation. 7808 Completing POA means that such code can already be tried 7809 to run on Classpath. Our POA is tested for the following 7810 usager scenarios: 7811 # POA converts servant to the CORBA object. 7812 # Servant provides to the CORBA object. 7813 # POA activates new CORBA object with the given Object 7814 Id (byte array) that is later accessible for the 7815 servant. 7816 # During the first call, the ServantActivator provides 7817 servant for this and all subsequent calls on the 7818 current object. 7819 # During each call, the ServantLocator provides 7820 servant for this call only. 7821 # ServantLocator or ServantActivator forwards call to 7822 another server. 7823 # POA has a single servant, responsible for all 7824 objects. 7825 # POA has a default servant, but some objects are 7826 explicitly connected to they specific servants. 7827 The POA is verified using tests from the former 7828 cost.omg.org. 7829 o The CORBA implementation is now a working prototype that 7830 should support features up to 1.3 inclusive. We invite 7831 groups writing CORBA dependent applications to try 7832 Classpath implementation, reporting any possible bugs. 7833 The CORBA prototype is interoperable with Sun's 7834 implementation v 1.4, transferring object references, 7835 primitive types, narrow and wide strings, arrays, 7836 structures, trees, abstract interfaces and value types 7837 (feature of CORBA 2.3) between these two platforms. 7838 Remote exceptions are transferred and handled correctly. 7839 The stringified object references (IORs) from various 7840 sources are parsed as required. The transient (for 7841 current session) and permanent (till jre restart) 7842 redirections work. Both Little and Big Endian encoded 7843 messages are accepted. The implementation is verified 7844 using tests from the former cost.omg.org. The current 7845 release includes working examples (see the examples 7846 directory), demonstrating the client-server 7847 communication, using either CORBA Request or IDL-based 7848 stub (usually generated by a IDL to java compiler). These 7849 examples also show how to use the Classpath CORBA naming 7850 service. The IDL to java compiler is not yet written, but 7851 as our library must be compatible, it naturally accepts 7852 the output of other idlj implementations. 7853 + Misc 7854 o Updated TimeZone data against Olson tzdata2005l. 7855 o Make zip and jar packages UTF-8 clean. 7856 o "native" code builds and compiles (warning free) on 7857 Darwin and Solaris. 7858 o java.util.logging.FileHandler now rotates files. 7859 o Start of a generic JDWP framework in gnu/classpath/jdwp. 7860 This is unfinished, but feedback (at classpath@gnu.org) 7861 from runtime hackers is greatly appreciated. Although 7862 most of the work is currently being done around gcj/gij 7863 we want this framework to be as VM neutral as possible. 7864 Early design is described in: 7865 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 7866 o QT4 AWT peers, enable by giving configure 7867 --enable-qt-peer. Included, but not ready for production 7868 yet. They are explicitly disabled and not supported. But 7869 if you want to help with the development of these new 7870 features we are interested in feedback. You will have to 7871 explicitly enable them to try them out (and they will 7872 most likely contain bugs). 7873 o Documentation fixes all over the place. See 7874 [4]http://developer.classpath.org/doc/ 7875 7876New Targets and Target Specific Improvements 7877 7878 IA-32/x86-64 7879 7880 * The x86-64 medium model (that allows building applications whose 7881 data segment exceeds 4GB) was redesigned to match latest ABI draft. 7882 New implementation split large datastructures into separate segment 7883 improving performance of accesses to small datastructures and also 7884 allows linking of small model libraries into medium model programs 7885 as long as the libraries are not accessing the large datastructures 7886 directly. Medium model is also supported in position independent 7887 code now. 7888 The ABI change results in partial incompatibility among medium 7889 model objects. Linking medium model libraries (or objects) compiled 7890 with new compiler into medium model program compiled with older 7891 will likely result in exceeding ranges of relocations. 7892 Binutils 2.16.91 or newer are required for compiling medium model 7893 now. 7894 7895 RS6000 (POWER/PowerPC) 7896 7897 * The AltiVec vector primitives in <altivec.h> are now implemented in 7898 a way that puts a smaller burden on the preprocessor, instead 7899 processing the "overloading" in the front ends. This should benefit 7900 compilation speed on AltiVec vector code. 7901 * AltiVec initializers now are generated more efficiently. 7902 * The popcountb instruction available on POWER5 now is generated. 7903 * The floating point round to integer instructions available on 7904 POWER5+ now is generated. 7905 * Floating point divides can be synthesized using the floating point 7906 reciprocal estimate instructions. 7907 * Double precision floating point constants are initialized as single 7908 precision values if they can be represented exactly. 7909 7910 S/390, zSeries and System z9 7911 7912 * Support for the IBM System z9 109 processor has been added. When 7913 using the -march=z9-109 option, the compiler will generate code 7914 making use of instructions provided by the extended immediate 7915 facility. 7916 * Support for 128-bit IEEE floating point has been added. When using 7917 the -mlong-double-128 option, the compiler will map the long double 7918 data type to 128-bit IEEE floating point. Using this option 7919 constitutes an ABI change, and requires glibc support. 7920 * Various changes to improve performance of generated code have been 7921 implemented, including: 7922 + In functions that do not require a literal pool, register %r13 7923 (which is traditionally reserved as literal pool pointer), can 7924 now be freely used for other purposes by the compiler. 7925 + More precise tracking of register use allows the compiler to 7926 generate more efficient function prolog and epilog code in 7927 certain cases. 7928 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 7929 instructions are now used to implement C string functions. 7930 + The MOVE CHARACTER instruction with single byte overlap is now 7931 used to implement the memset function with non-zero fill byte. 7932 + The LOAD ZERO instructions are now used where appropriate. 7933 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 7934 and INSERT IMMEDIATE instructions are now used more frequently 7935 to optimize bitfield operations. 7936 + The BRANCH ON COUNT instruction is now used more frequently. 7937 In particular, the fact that a loop contains a subroutine call 7938 no longer prevents the compiler from using this instruction. 7939 + The compiler is now aware that all shift and rotate 7940 instructions implicitly truncate the shift count to six bits. 7941 * Back-end support for the following generic features has been 7942 implemented: 7943 + The full set of [5]built-in functions for atomic memory 7944 access. 7945 + The -fstack-protector feature. 7946 + The optimization pass avoiding unnecessary stores of incoming 7947 argument registers in functions with variable argument list. 7948 7949 SPARC 7950 7951 * The default code model in 64-bit mode has been changed from 7952 Medium/Anywhere to Medium/Middle on Solaris. 7953 * TLS support is disabled by default on Solaris prior to release 10. 7954 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 7955 and later) by specifying --enable-tls at configure time. 7956 7957 MorphoSys 7958 7959 * Support has been added for this new architecture. 7960 7961Obsolete Systems 7962 7963Documentation improvements 7964 7965Other significant improvements 7966 7967 * GCC can now emit code for protecting applications from 7968 stack-smashing attacks. The protection is realized by buffer 7969 overflow detection and reordering of stack variables to avoid 7970 pointer corruption. 7971 * Some built-in functions have been fortified to protect them against 7972 various buffer overflow (and format string) vulnerabilities. 7973 Compared to the mudflap bounds checking feature, the safe builtins 7974 have far smaller overhead. This means that programs built using 7975 safe builtins should not experience any measurable slowdown. 7976 7977GCC 4.1.2 7978 7979 This is the [6]list of problem reports (PRs) from GCC's bug tracking 7980 system that are known to be fixed in the 4.1.2 release. This list might 7981 not be complete (that is, it is possible that some PRs that have been 7982 fixed are not listed here). 7983 7984 When generating code for a shared library, GCC now recognizes that 7985 global functions may be replaced when the program runs. Therefore, it 7986 is now more conservative in deducing information from the bodies of 7987 functions. For example, in this example: 7988 void f() {} 7989 void g() { 7990 try { f(); } 7991 catch (...) { 7992 cout << "Exception"; 7993 } 7994 } 7995 7996 G++ would previously have optimized away the catch clause, since it 7997 would have concluded that f cannot throw exceptions. Because users may 7998 replace f with another function in the main body of the program, this 7999 optimization is unsafe, and is no longer performed. If you wish G++ to 8000 continue to optimize as before, you must add a throw() clause to the 8001 declaration of f to make clear that it does not throw exceptions. 8002 8003 8004 For questions related to the use of GCC, please consult these web 8005 pages and the [7]GCC manuals. If that fails, the 8006 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these 8007 web pages and the development of GCC are welcome on our developer 8008 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public 8009 archives. 8010 8011 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 8012 distribution of this entire article is permitted in any medium, 8013 provided this notice is preserved. 8014 8015 These pages are [12]maintained by the GCC team. Last modified 8016 2016-01-30[13]. 8017 8018References 8019 8020 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 8021 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 8022 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 8023 4. http://developer.classpath.org/doc/ 8024 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 8025 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 8026 7. https://gcc.gnu.org/onlinedocs/ 8027 8. mailto:gcc-help@gcc.gnu.org 8028 9. mailto:gcc@gcc.gnu.org 8029 10. https://gcc.gnu.org/lists.html 8030 11. http://www.fsf.org/ 8031 12. https://gcc.gnu.org/about.html 8032 13. http://validator.w3.org/check/referer 8033====================================================================== 8034http://gcc.gnu.org/gcc-4.0/index.html 8035 8036 GCC 4.0 Release Series 8037 8038 January 31, 2007 8039 8040 The [1]GNU project and the GCC developers are pleased to announce the 8041 release of GCC 4.0.4. 8042 8043 This release is a bug-fix release, containing fixes for regressions in 8044 GCC 4.0.3 relative to previous releases of GCC. 8045 8046Release History 8047 8048 GCC 4.0.4 8049 January 31, 2007 ([2]changes) 8050 8051 GCC 4.0.3 8052 March 10, 2006 ([3]changes) 8053 8054 GCC 4.0.2 8055 September 28, 2005 ([4]changes) 8056 8057 GCC 4.0.1 8058 July 7, 2005 ([5]changes) 8059 8060 GCC 4.0.0 8061 April 20, 2005 ([6]changes) 8062 8063References and Acknowledgements 8064 8065 GCC used to stand for the GNU C Compiler, but since the compiler 8066 supports several other languages aside from C, it now stands for the 8067 GNU Compiler Collection. 8068 8069 A list of [7]successful builds is updated as new information becomes 8070 available. 8071 8072 The GCC developers would like to thank the numerous people that have 8073 contributed new features, improvements, bug fixes, and other changes as 8074 well as test results to GCC. This [8]amazing group of volunteers is 8075 what makes GCC successful. 8076 8077 For additional information about GCC please refer to the [9]GCC project 8078 web site or contact the [10]GCC development mailing list. 8079 8080 To obtain GCC please use [11]our mirror sites, or [12]our SVN server. 8081 8082 8083 For questions related to the use of GCC, please consult these web 8084 pages and the [13]GCC manuals. If that fails, the 8085 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 8086 web pages and the development of GCC are welcome on our developer 8087 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 8088 archives. 8089 8090 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 8091 distribution of this entire article is permitted in any medium, 8092 provided this notice is preserved. 8093 8094 These pages are [18]maintained by the GCC team. Last modified 8095 2016-01-30[19]. 8096 8097References 8098 8099 1. http://www.gnu.org/ 8100 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 8101 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 8102 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 8103 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 8104 6. http://gcc.gnu.org/gcc-4.0/changes.html 8105 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 8106 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8107 9. http://gcc.gnu.org/index.html 8108 10. mailto:gcc@gcc.gnu.org 8109 11. http://gcc.gnu.org/mirrors.html 8110 12. http://gcc.gnu.org/svn.html 8111 13. https://gcc.gnu.org/onlinedocs/ 8112 14. mailto:gcc-help@gcc.gnu.org 8113 15. mailto:gcc@gcc.gnu.org 8114 16. https://gcc.gnu.org/lists.html 8115 17. http://www.fsf.org/ 8116 18. https://gcc.gnu.org/about.html 8117 19. http://validator.w3.org/check/referer 8118====================================================================== 8119http://gcc.gnu.org/gcc-4.0/changes.html 8120 8121 GCC 4.0 Release Series 8122 Changes, New Features, and Fixes 8123 8124 The latest release in the 4.0 release series is [1]GCC 4.0.4. 8125 8126Caveats 8127 8128 * GCC now generates location lists by default when compiling with 8129 debug info and optimization. 8130 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 8131 or later is needed to debug binaries containing location 8132 lists. 8133 + When you are trying to view a value of a variable in a part of 8134 a function where it has no location (for example when the 8135 variable is no longer used and thus its location was used for 8136 something else) GDB will say that it is not available. 8137 You can disable generating location lists by -fno-var-tracking. 8138 * GCC no longer accepts the -fwritable-strings option. Use named 8139 character arrays when you need a writable string. 8140 * The options -freduce-all-givs and -fmove-all-movables have been 8141 discontinued. They were used to circumvent a shortcoming in the 8142 heuristics of the old loop optimization code with respect to common 8143 Fortran constructs. The new (tree) loop optimizer works differently 8144 and doesn't need those work-arounds. 8145 * The graph-coloring register allocator, formerly enabled by the 8146 option -fnew-ra, has been discontinued. 8147 * -I- has been deprecated. -iquote is meant to replace the need for 8148 this option. 8149 * The MIPS -membedded-pic and -mrnames options have been removed. 8150 * All MIPS targets now require the GNU assembler. In particular, IRIX 8151 configurations can no longer use the MIPSpro assemblers, although 8152 they do still support the MIPSpro linkers. 8153 * The SPARC option -mflat has been removed. 8154 * English-language diagnostic messages will now use Unicode quotation 8155 marks in UTF-8 locales. (Non-English messages already used the 8156 quotes appropriate for the language in previous releases.) If your 8157 terminal does not support UTF-8 but you are using a UTF-8 locale 8158 (such locales are the default on many GNU/Linux systems) then you 8159 should set LC_CTYPE=C in the environment to disable that locale. 8160 Programs that parse diagnostics and expect plain ASCII 8161 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 8162 explanation of Unicode quotation marks for more information. 8163 * The specs file is no longer installed on most platforms. Most users 8164 will be totally unaffected. However, if you are accustomed to 8165 editing the specs file yourself, you will now have to use the 8166 -dumpspecs option to generate the specs file, and then edit the 8167 resulting file. 8168 8169General Optimizer Improvements 8170 8171 * The [3]tree ssa branch has been merged. This merge has brought in a 8172 completely new optimization framework based on a higher level 8173 intermediate representation than the existing RTL representation. 8174 Numerous new code transformations based on the new framework are 8175 available in GCC 4.0, including: 8176 + Scalar replacement of aggregates 8177 + Constant propagation 8178 + Value range propagation 8179 + Partial redundancy elimination 8180 + Load and store motion 8181 + Strength reduction 8182 + Dead store elimination 8183 + Dead and unreachable code elimination 8184 + [4]Autovectorization 8185 + Loop interchange 8186 + Tail recursion by accumulation 8187 Many of these passes outperform their counterparts from previous 8188 GCC releases. 8189 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 8190 scheduling optimization intended for loops that perform heavy 8191 computations. 8192 8193New Languages and Language specific improvements 8194 8195 C family 8196 8197 * The sentinel attribute has been added to GCC. This function 8198 attribute allows GCC to warn when variadic functions such as execl 8199 are not NULL terminated. See the GCC manual for a complete 8200 description of its behavior. 8201 * Given __attribute__((alias("target"))) it is now an error if target 8202 is not a symbol, defined in the same translation unit. This also 8203 applies to aliases created by #pragma weak alias=target. This is 8204 because it's meaningless to define an alias to an undefined symbol. 8205 On Solaris, the native assembler would have caught this error, but 8206 GNU as does not. 8207 8208 C and Objective-C 8209 8210 * The -Wstrict-aliasing=2 option has been added. This warning catches 8211 all unsafe cases, but it may also give a warning for some cases 8212 that are safe. 8213 * The cast-as-lvalue, conditional-expression-as-lvalue and 8214 compound-expression-as-lvalue extensions, which were deprecated in 8215 3.3.4 and 3.4, have been removed. 8216 * The -fwritable-strings option, which was deprecated in 3.4, has 8217 been removed. 8218 * #pragma pack() semantics have been brought closer to those used by 8219 other compilers. This also applies to C++. 8220 * Taking the address of a variable with register storage is invalid 8221 in C. GCC now issues an error instead of a warning. 8222 * Arrays of incomplete element type are invalid in C. GCC now issues 8223 an error for such arrays. Declarations such as extern struct s x[]; 8224 (where struct s has not been defined) can be moved after the 8225 definition of struct s. Function parameters declared as arrays of 8226 incomplete type can instead be declared as pointers. 8227 8228 C++ 8229 8230 * When compiling without optimizations (-O0), the C++ frontend is 8231 much faster than in any previous versions of GCC. Independent 8232 testers have measured speed-ups up to 25% in real-world production 8233 code, compared to the 3.4 family (which was already the fastest 8234 version to date). Upgrading from older versions might show even 8235 bigger improvements. 8236 * ELF visibility attributes can now be applied to a class type, so 8237 that it affects every member function of a class at once, without 8238 having to specify each individually: 8239class __attribute__ ((visibility("hidden"))) Foo 8240{ 8241 int foo1(); 8242 void foo2(); 8243}; 8244 The syntax is deliberately similar to the __declspec() system used 8245 by Microsoft Windows based compilers, allowing cross-platform 8246 projects to easily reuse their existing macro system for denoting 8247 exports and imports. By explicitly marking internal classes never 8248 used outside a binary as hidden, one can completely avoid PLT 8249 indirection overheads during their usage by the compiler. You can 8250 find out more about the advantages of this at 8251 [6]https://www.akkadia.org/drepper/dsohowto.pdf 8252 * The -fvisibility-inlines-hidden option has been added which marks 8253 all inlineable functions as having hidden ELF visibility, thus 8254 removing their symbol and typeinfo from the exported symbol table 8255 of the output ELF binary. Using this option can reduce the exported 8256 symbol count of template-heavy code by up to 40% with no code 8257 change at all, thus notably improving link and load times for the 8258 binary as well as a reduction in size of up to 10%. Also, check the 8259 new [7]-fvisibility option. 8260 * The compiler now uses the library interface specified by the [8]C++ 8261 ABI for thread-safe initialization of function-scope static 8262 variables. Most users should leave this alone, but embedded 8263 programmers may want to disable this by specifying 8264 -fno-threadsafe-statics for a small savings in code size. 8265 * Taking the address of an explicit register variable is no longer 8266 supported. Note that C++ allows taking the address of variables 8267 with register storage so this will continue to compile with a 8268 warning. For example, assuming that r0 is a machine register: 8269register int foo asm ("r0"); 8270register int bar; 8271&foo; // error, no longer accepted 8272&bar; // OK, with a warning 8273 * G++ has an undocumented extension to virtual function covariancy 8274 rules that allowed the overrider to return a type that was 8275 implicitly convertable to the overridden function's return type. 8276 For instance a function returning void * could be overridden by a 8277 function returning T *. This is now deprecated and will be removed 8278 in a future release. 8279 * The G++ minimum and maximum operators (<? and >?) and their 8280 compound forms (<?=) and >?=) have been deprecated and will be 8281 removed in a future version. Code using these operators should be 8282 modified to use std::min and std::max instead. 8283 * Declaration of nested classes of class templates as friends are 8284 supported: 8285template <typename T> struct A { 8286 class B {}; 8287}; 8288class C { 8289 template <typename T> friend class A<T>::B; 8290}; 8291 This complements the feature member functions of class templates as 8292 friends introduced in GCC 3.4.0. 8293 * When declaring a friend class using an unqualified name, classes 8294 outside the innermost non-class scope are not searched: 8295class A; 8296namespace N { 8297 class B { 8298 friend class A; // Refer to N::A which has not been declared yet 8299 // because name outside namespace N are not searched 8300 friend class ::A; // Refer to ::A 8301 }; 8302} 8303 Hiding the friend name until declaration is still not implemented. 8304 * Friends of classes defined outside their namespace are correctly 8305 handled: 8306namespace N { 8307 class A; 8308} 8309class N::A { 8310 friend class B; // Refer to N::B in GCC 4.0.0 8311 // but ::B in earlier versions of GCC 8312}; 8313 8314 Runtime Library (libstdc++) 8315 8316 * Optimization work: 8317 + Added efficient specializations of istream functions for char 8318 and wchar_t. 8319 + Further performance tuning of strings, in particular wrt 8320 single-char append and getline. 8321 + iter_swap - and therefore most of the mutating algorithms - 8322 now makes an unqualified call to swap when the value_type of 8323 the two iterators is the same. 8324 * A large subset of the features in Technical Report 1 (TR1 for 8325 short) is experimentally delivered (i.e., no guarantees about the 8326 implementation are provided. In particular it is not promised that 8327 the library will remain link-compatible when code using TR1 is 8328 used): 8329 + General utilities such as reference_wrapper and shared_ptr. 8330 + Function objects, i.e., result_of, mem_fn, bind, function. 8331 + Support for metaprogramming. 8332 + New containers such as tuple, array, unordered_set, 8333 unordered_map, unordered_multiset, unordered_multimap. 8334 * As usual, many bugs have been fixed and LWG resolutions implemented 8335 for the first time (e.g., DR 409). 8336 8337 Java 8338 8339 * In order to prevent naming conflicts with other implementations of 8340 these tools, some GCJ binaries have been renamed: 8341 + rmic is now grmic, 8342 + rmiregistry is now grmiregistry, and 8343 + jar is now fastjar. 8344 In particular, these names were problematic for the jpackage.org 8345 packaging conventions which install symlinks in /usr/bin that point 8346 to the preferred versions of these tools. 8347 * The -findirect-dispatch argument to the compiler now works and 8348 generates code following a new "binary compatibility" ABI. Code 8349 compiled this way follows the binary compatibility rules of the 8350 Java Language Specification. 8351 * libgcj now has support for using GCJ as a JIT, using the 8352 gnu.gcj.jit family of system properties. 8353 * libgcj can now find a shared library corresponding to the bytecode 8354 representation of a class. See the documentation for the new 8355 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 8356 property. 8357 * There have been many improvements to the class library. Here are 8358 some highlights: 8359 + Much more of AWT and Swing exist. 8360 + Many new packages and classes were added, including 8361 java.util.regex, java.net.URI, javax.crypto, 8362 javax.crypto.interfaces, javax.crypto.spec, javax.net, 8363 javax.net.ssl, javax.security.auth, 8364 javax.security.auth.callback, javax.security.auth.login, 8365 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 8366 javax.imageio, javax.imageio.event, javax.imageio.spi, 8367 javax.print, javax.print.attribute, 8368 javax.print.attribute.standard, javax.print.event, and 8369 javax.xml 8370 + Updated SAX and DOM, and imported GNU JAXP 8371 8372 Fortran 8373 8374 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 8375 front end. The new front end supports Fortran 90 and Fortran 95. It 8376 may not yet be as stable as the old Fortran front end. 8377 8378 Ada 8379 8380 * Ada (with tasking and Zero Cost Exceptions) is now available on 8381 many more targets, including but not limited to: alpha-linux, 8382 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 8383 s390x-linux, sparc-linux. 8384 * Some of the new Ada 2005 features are now implemented like 8385 Wide_Wide_Character and Ada.Containers. 8386 * Many bugs have been fixed, tools and documentation improved. 8387 * To compile Ada from the sources, install an older working Ada 8388 compiler and then use --enable-languages=ada at configuration time, 8389 since the Ada frontend is not currently activated by default. See 8390 the [10]Installing GCC for details. 8391 8392New Targets and Target Specific Improvements 8393 8394 H8/300 8395 8396 * The frame layout has changed. In the new layout, the prologue of a 8397 function first saves registers and then allocate space for locals, 8398 resulting in an 1% improvement on code size. 8399 8400 IA-32/x86-64 (AMD64) 8401 8402 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 8403 log1p, log2, logb and tan mathematical builtins (and their float 8404 and long double variants) are now implemented as inline x87 8405 intrinsics when using -ffast-math. 8406 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 8407 (and their float and long double variants) are now implemented as 8408 inline x87 intrinsics when using -ffast-math. 8409 * The x87's fsincos instruction is now used automatically with 8410 -ffast-math when calculating both the sin and cos of the same 8411 argument. 8412 * Instruction selection for multiplication and division by constants 8413 has been improved. 8414 8415 IA-64 8416 8417 * Floating point division, integer division and sqrt are now inlined, 8418 resulting in significant performance improvements on some codes. 8419 8420 MIPS 8421 8422 * Division by zero checks now use conditional traps if the target 8423 processor supports them. This decreases code size by one word per 8424 division operation. The old behavior (branch and break) can be 8425 obtained either at configure time by passing --with-divide=breaks 8426 to configure or at runtime by passing -mdivide-breaks to GCC. 8427 * Support for MIPS64 paired-single instructions has been added. It is 8428 enabled by -mpaired-single and can be accessed using both the 8429 target-independent vector extensions and new MIPS-specific built-in 8430 functions. 8431 * Support for the MIPS-3D ASE has been added. It is enabled by 8432 -mips3d and provides new MIPS-3D-specific built-in functions. 8433 * The -mexplicit-relocs option now supports static n64 code (as is 8434 used, for example, in 64-bit linux kernels). -mexplicit-relocs 8435 should now be feature-complete and is enabled by default when GCC 8436 is configured to use a compatible assembler. 8437 * Support for the NEC VR4130 series has been added. This support 8438 includes the use of VR-specific instructions and a new VR4130 8439 scheduler. Full VR4130 support can be selected with -march=vr4130 8440 while code for any ISA can be tuned for the VR4130 using 8441 -mtune=vr4130. There is also a new -mvr4130-align option that 8442 produces better schedules at the cost of increased code size. 8443 * Support for the Broadcom SB-1 has been extended. There is now an 8444 SB-1 scheduler as well as support for the SB-1-specific 8445 paired-single instructions. Full SB-1 support can be selected with 8446 -march=sb1 while code for any ISA can be optimized for the SB-1 8447 using -mtune=sb1. 8448 * The compiler can now work around errata in R4000, R4400, VR4120 and 8449 VR4130 processors. These workarounds are enabled by -mfix-r4000, 8450 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 8451 and VR4130 workarounds need binutils 2.16 or above. 8452 * IRIX shared libraries are now installed into the standard library 8453 directories: o32 libraries go into lib/, n32 libraries go into 8454 lib32/ and n64 libraries go into lib64/. 8455 * The compiler supports a new -msym32 option. It can be used to 8456 optimize n64 code in which all symbols are known to have 32-bit 8457 values. 8458 8459 S/390 and zSeries 8460 8461 * New command-line options help to generate code intended to run in 8462 an environment where stack space is restricted, e.g. Linux kernel 8463 code: 8464 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 8465 warnings for single functions that require large or dynamic 8466 stack frames. 8467 + -mstack-size and -mstack-guard generate code that checks for 8468 stack overflow at run time. 8469 + -mpacked-stack generates code that reduces the stack frame 8470 size of many functions by reusing unneeded parts of the stack 8471 bias area. 8472 * The -msoft-float option now ensures that generated code never 8473 accesses floating point registers. 8474 * The s390x-ibm-tpf target now fully supports C++, including 8475 exceptions and threads. 8476 * Various changes to improve performance of the generated code have 8477 been implemented, including: 8478 + GCC now uses sibling calls where possible. 8479 + Condition code handling has been optimized, allowing GCC to 8480 omit redundant comparisons in certain cases. 8481 + The cost function guiding many optimizations has been refined 8482 to more accurately represent the z900 and z990 processors. 8483 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 8484 instructions are now used to avoid conditional branches in 8485 certain cases. 8486 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 8487 optimize address arithmetic required to access large stack 8488 frames. 8489 + GCC now makes more efficient use of memory-to-memory type 8490 instructions (MVC, CLC, ...). 8491 + More precise tracking of special register use allows better 8492 instruction scheduling, in particular of the function prologue 8493 and epilogue sequences. 8494 + The Java front end now generates inline code to implement 8495 integer division, instead of calling library routines. 8496 8497 SPARC 8498 8499 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 8500 -mf934 have been removed. They have been replaced with -mcpu=xxx. 8501 * The internal model used to estimate the relative cost of each 8502 instruction has been updated. It is expected to give better results 8503 on recent UltraSPARC processors. 8504 * Code generation for function prologues and epilogues has been 8505 improved, resulting in better scheduling and allowing multiple exit 8506 points in functions. 8507 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 8508 It is enabled by -mvis and provides new built-in functions for VIS 8509 instructions on UltraSPARC processors. 8510 * The option -mapp-regs has been turned on by default on Solaris too. 8511 8512 NetWare 8513 8514 * Novell NetWare (on ix86, no other hardware platform was ever really 8515 supported by this OS) has been re-enabled and the ABI supported by 8516 GCC has been brought into sync with that of MetroWerks CodeWarrior 8517 (the ABI previously supported was that of some Unix systems, which 8518 NetWare never tried to support). 8519 8520Obsolete Systems 8521 8522 Support for a number of older systems has been declared obsolete in GCC 8523 4.0. Unless there is activity to revive them, the next release of GCC 8524 will have their sources permanently removed. 8525 8526 All GCC ports for the following processor architectures have been 8527 declared obsolete: 8528 * Intel i860 8529 * Ubicom IP2022 8530 * National Semiconductor NS32K 8531 * Texas Instruments TMS320C[34]x 8532 8533 Also, those for some individual systems have been obsoleted: 8534 * SPARC family 8535 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 8536 sparc86x-*-elf) 8537 + OpenBSD 32-bit (sparc-*-openbsd*) 8538 8539Documentation improvements 8540 8541Other significant improvements 8542 8543 * Location lists are now generated by default when compiling with 8544 debug info and optimization. Location lists provide more accurate 8545 debug info about locations of variables and they allow debugging 8546 code compiled with -fomit-frame-pointer. 8547 * The -fvisibility option has been added which allows the default ELF 8548 visibility of all symbols to be set per compilation and the new 8549 #pragma GCC visibility preprocessor command allows the setting of 8550 default ELF visibility for a region of code. Using 8551 -fvisibility=hidden especially in combination with the new 8552 -fvisibility-inlines-hidden can yield substantial improvements in 8553 output binary quality including avoiding PLT indirection overheads, 8554 reduction of the exported symbol count by up to 60% (with resultant 8555 improvements to link and load times), better scope for the 8556 optimizer to improve code and up to a 20% reduction in binary size. 8557 Using these options correctly yields a binary with a similar symbol 8558 count to a Windows DLL. 8559 Perhaps more importantly, this new feature finally allows (with 8560 careful planning) complete avoidance of symbol clashes when 8561 manually loading shared objects with RTLD_GLOBAL, thus finally 8562 solving problems many projects such as python were forced to use 8563 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 8564 can find more information about using these options at 8565 [11]https://gcc.gnu.org/wiki/Visibility. 8566 __________________________________________________________________ 8567 8568GCC 4.0.1 8569 8570 This is the [12]list of problem reports (PRs) from GCC's bug tracking 8571 system that are known to be fixed in the 4.0.1 release. This list might 8572 not be complete (that is, it is possible that some PRs that have been 8573 fixed are not listed here). 8574 8575GCC 4.0.2 8576 8577 This is the [13]list of problem reports (PRs) from GCC's bug tracking 8578 system that are known to be fixed in the 4.0.2 release. This list might 8579 not be complete (that is, it is possible that some PRs that have been 8580 fixed are not listed here). 8581 8582 Unfortunately, due to a release engineering failure, this release has a 8583 regression on Solaris that will affect some C++ programs. We suggest 8584 that Solaris users apply a [14]patch that corrects the problem. Users 8585 who do not wish to apply the patch should explicitly link C++ programs 8586 with the -pthreads option, even if they do not use threads. This 8587 problem has been corrected in the current 4.0 branch sources and will 8588 not be present in GCC 4.0.3. 8589 8590GCC 4.0.3 8591 8592 Starting with this release, the function getcontext is recognized by 8593 the compiler as having the same semantics as the setjmp function. In 8594 particular, the compiler will ensure that all registers are dead before 8595 calling such a function and will emit a warning about the variables 8596 that may be clobbered after the second return from the function. 8597 8598GCC 4.0.4 8599 8600 This is the [15]list of problem reports (PRs) from GCC's bug tracking 8601 system that are known to be fixed in the 4.0.4 release. This list might 8602 not be complete (that is, it is possible that some PRs that have been 8603 fixed are not listed here). 8604 8605 The 4.0.4 release is provided for those that require a high degree of 8606 binary compatibility with previous 4.0.x releases. For most users, the 8607 GCC team recommends that version 4.1.1 or later be used instead." 8608 8609 8610 For questions related to the use of GCC, please consult these web 8611 pages and the [16]GCC manuals. If that fails, the 8612 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 8613 web pages and the development of GCC are welcome on our developer 8614 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 8615 archives. 8616 8617 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 8618 distribution of this entire article is permitted in any medium, 8619 provided this notice is preserved. 8620 8621 These pages are [21]maintained by the GCC team. Last modified 8622 2016-05-27[22]. 8623 8624References 8625 8626 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 8627 2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 8628 3. http://gcc.gnu.org/projects/tree-ssa/ 8629 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 8630 5. http://gcc.gnu.org/news/sms.html 8631 6. https://www.akkadia.org/drepper/dsohowto.pdf 8632 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 8633 8. http://mentorembedded.github.io/cxx-abi/ 8634 9. http://gcc.gnu.org/fortran/ 8635 10. https://gcc.gnu.org/install/ 8636 11. https://gcc.gnu.org/wiki/Visibility 8637 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 8638 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 8639 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 8640 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 8641 16. https://gcc.gnu.org/onlinedocs/ 8642 17. mailto:gcc-help@gcc.gnu.org 8643 18. mailto:gcc@gcc.gnu.org 8644 19. https://gcc.gnu.org/lists.html 8645 20. http://www.fsf.org/ 8646 21. https://gcc.gnu.org/about.html 8647 22. http://validator.w3.org/check/referer 8648====================================================================== 8649http://gcc.gnu.org/gcc-3.4/index.html 8650 8651 GCC 3.4 Release Series 8652 8653 May 26, 2006 8654 8655 The [1]GNU project and the GCC developers are pleased to announce the 8656 release of GCC 3.4.6. 8657 8658 This release is a bug-fix release, containing fixes for regressions in 8659 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 8660 3.4.x series. 8661 8662 The GCC 3.4 release series includes numerous [2]new features, 8663 improvements, bug fixes, and other changes, thanks to an [3]amazing 8664 group of volunteers. 8665 8666Release History 8667 8668 GCC 3.4.6 8669 March 6, 2006 ([4]changes) 8670 8671 GCC 3.4.5 8672 November 30, 2005 ([5]changes) 8673 8674 GCC 3.4.4 8675 May 18, 2005 ([6]changes) 8676 8677 GCC 3.4.3 8678 November 4, 2004 ([7]changes) 8679 8680 GCC 3.4.2 8681 September 6, 2004 ([8]changes) 8682 8683 GCC 3.4.1 8684 July 1, 2004 ([9]changes) 8685 8686 GCC 3.4.0 8687 April 18, 2004 ([10]changes) 8688 8689References and Acknowledgements 8690 8691 GCC used to stand for the GNU C Compiler, but since the compiler 8692 supports several other languages aside from C, it now stands for the 8693 GNU Compiler Collection. 8694 8695 A list of [11]successful builds is updated as new information becomes 8696 available. 8697 8698 The GCC developers would like to thank the numerous people that have 8699 contributed new features, improvements, bug fixes, and other changes as 8700 well as test results to GCC. This [12]amazing group of volunteers is 8701 what makes GCC successful. 8702 8703 For additional information about GCC please refer to the [13]GCC 8704 project web site or contact the [14]GCC development mailing list. 8705 8706 To obtain GCC please use [15]our mirror sites, or [16]our SVN server. 8707 8708 8709 For questions related to the use of GCC, please consult these web 8710 pages and the [17]GCC manuals. If that fails, the 8711 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these 8712 web pages and the development of GCC are welcome on our developer 8713 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public 8714 archives. 8715 8716 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 8717 distribution of this entire article is permitted in any medium, 8718 provided this notice is preserved. 8719 8720 These pages are [22]maintained by the GCC team. Last modified 8721 2016-01-30[23]. 8722 8723References 8724 8725 1. http://www.gnu.org/ 8726 2. http://gcc.gnu.org/gcc-3.4/changes.html 8727 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8728 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 8729 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 8730 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 8731 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 8732 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 8733 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 8734 10. http://gcc.gnu.org/gcc-3.4/changes.html 8735 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 8736 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8737 13. http://gcc.gnu.org/index.html 8738 14. mailto:gcc@gcc.gnu.org 8739 15. http://gcc.gnu.org/mirrors.html 8740 16. http://gcc.gnu.org/svn.html 8741 17. https://gcc.gnu.org/onlinedocs/ 8742 18. mailto:gcc-help@gcc.gnu.org 8743 19. mailto:gcc@gcc.gnu.org 8744 20. https://gcc.gnu.org/lists.html 8745 21. http://www.fsf.org/ 8746 22. https://gcc.gnu.org/about.html 8747 23. http://validator.w3.org/check/referer 8748====================================================================== 8749http://gcc.gnu.org/gcc-3.4/changes.html 8750 8751 GCC 3.4 Release Series 8752 Changes, New Features, and Fixes 8753 8754 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 8755 is now closed. 8756 8757 GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting 8758 a bug, please make sure it's really GCC, and not your code, that is 8759 broken. 8760 8761Caveats 8762 8763 * GNU Make is now required to build GCC. 8764 * With -nostdinc the preprocessor used to ignore both standard 8765 include paths and include paths contained in environment variables. 8766 It was neither documented nor intended that environment variable 8767 paths be ignored, so this has been corrected. 8768 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 8769 -fvolatile-static. It is unlikely that they worked correctly in any 8770 3.x release. 8771 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 8772 * Support for all the systems [3]obsoleted in GCC 3.3 has been 8773 removed from GCC 3.4. See below for a [4]list of systems which are 8774 obsoleted in this release. 8775 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 8776 compilers will not work. 8777 * The implementation of the [5]MIPS ABIs has changed. As a result, 8778 the code generated for certain MIPS targets will not be binary 8779 compatible with earlier releases. 8780 * In previous releases, the MIPS port had a fake "hilo" register with 8781 the user-visible name accum. This register has been removed. 8782 * The implementation of the [6]SPARC ABIs has changed. As a result, 8783 the code generated will not be binary compatible with earlier 8784 releases in certain cases. 8785 * The configure option --enable-threads=pthreads has been removed; 8786 use --enable-threads=posix instead, which should have the same 8787 effect. 8788 * Code size estimates used by inlining heuristics for C, Objective-C, 8789 C++ and Java have been redesigned significantly. As a result the 8790 parameters of -finline-insns, --param max-inline-insns-single and 8791 --param max-inline-insns-auto need to be reconsidered. 8792 * --param max-inline-slope and --param min-inline-insns have been 8793 removed; they are not needed for the new bottom-up inlining 8794 heuristics. 8795 * The new unit-at-a-time compilation scheme has several compatibility 8796 issues: 8797 + The order in which functions, variables, and top-level asm 8798 statements are emitted may have changed. Code relying on some 8799 particular ordering needs to be updated. The majority of such 8800 top-level asm statements can be replaced by section 8801 attributes. 8802 + Unreferenced static variables and functions are removed. This 8803 may result in undefined references when an asm statement 8804 refers to the variable/function directly. In that case either 8805 the variable/function shall be listed in asm statement operand 8806 or in the case of top-level asm statements the attribute used 8807 shall be used to force function/variable to be always output 8808 and considered as a possibly used by unknown code. 8809 For variables the attribute is accepted only by GCC 3.4 and 8810 newer, while for earlier versions it is sufficient to use 8811 unused to silence warnings about the variables not being 8812 referenced. To keep code portable across different GCC 8813 versions, you can use appropriate preprocessor conditionals. 8814 + Static functions now can use non-standard passing conventions 8815 that may break asm statements calling functions directly. 8816 Again the attribute used shall be used to prevent this 8817 behavior. 8818 As a temporary workaround, -fno-unit-at-a-time can be used, but 8819 this scheme may not be supported by future releases of GCC. 8820 * GCC 3.4 automatically places zero-initialized variables in the .bss 8821 section on some operating systems. Versions of GNU Emacs up to (and 8822 including) 21.3 will not work correctly when using this 8823 optimization; you can use -fno-zero-initialized-in-bss to disable 8824 it. 8825 * If GCC 3.4 is configured with --enable-threads=posix (the default 8826 on most targets that support pthreads) then _REENTRANT will be 8827 defined unconditionally by some libstdc++ headers. C++ code which 8828 relies on that macro to detect whether multi-threaded code is being 8829 compiled might change in meaning, possibly resulting in linker 8830 errors for single-threaded programs. Affected users of [7]Boost 8831 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 8832 See Bugzilla for [8]more information. 8833 8834General Optimizer Improvements 8835 8836 * Usability of the profile feedback and coverage testing has been 8837 improved. 8838 + Performance of profiled programs has been improved by faster 8839 profile merging code. 8840 + Better use of the profile feedback for optimization (loop 8841 unrolling and loop peeling). 8842 + File locking support allowing fork() calls and parallel runs 8843 of profiled programs. 8844 + Coverage file format has been redesigned. 8845 + gcov coverage tool has been improved. 8846 + make profiledbootstrap available to build a faster compiler. 8847 Experiments made on i386 hardware showed an 11% speedup on -O0 8848 and a 7.5% speedup on -O2 compilation of a [9]large C++ 8849 testcase. 8850 + New value profiling pass enabled via -fprofile-values 8851 + New value profile transformations pass enabled via -fvpt aims 8852 to optimize some code sequences by exploiting knowledge about 8853 value ranges or other properties of the operands. At the 8854 moment a conversion of expensive divisions into cheaper 8855 operations has been implemented. 8856 + New -fprofile-generate and -fprofile-use command-line options 8857 to simplify the use of profile feedback. 8858 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 8859 Java which is enabled via -funit-at-a-time (and implied by -O2). In 8860 this scheme a whole file is parsed first and optimized later. The 8861 following basic inter-procedural optimizations are implemented: 8862 + Removal of unreachable functions and variables 8863 + Discovery of local functions (functions with static linkage 8864 whose address is never taken) 8865 + On i386, these local functions use register parameter passing 8866 conventions. 8867 + Reordering of functions in topological order of the call graph 8868 to enable better propagation of optimizing hints (such as the 8869 stack alignments needed by functions) in the back end. 8870 + Call graph based out-of-order inlining heuristics which allows 8871 to limit overall compilation unit growth (--param 8872 inline-unit-growth). 8873 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 8874 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 8875 CPU). 8876 * More realistic code size estimates used by inlining for C, 8877 Objective-C, C++ and Java. The growth of large functions can now be 8878 limited via --param large-function-insns and --param 8879 large-function-growth. 8880 * A new cfg-level loop optimizer pass replaces the old loop unrolling 8881 pass and adds two other loop transformations -- loop peeling and 8882 loop unswitching -- and also uses the profile feedback to limit 8883 code growth. (The three optimizations are enabled by 8884 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 8885 respectively). 8886 The old loop unroller still can be enabled by -fold-unroll-loops 8887 and may produce better code in some cases, especially when the 8888 webizer optimization pass is not run. 8889 * A new web construction pass enabled via -fweb (and implied by -O3) 8890 improves the quality of register allocation, CSE, first scheduling 8891 pass and some other optimization passes by avoiding re-use of 8892 pseudo registers with non-overlapping live ranges. The pass almost 8893 always improves code quality but does make debugging difficult and 8894 thus is not enabled by default by -O2 8895 The pass is especially effective as cleanup after code duplication 8896 passes, such as the loop unroller or the tracer. 8897 * Experimental implementations of superblock or trace scheduling in 8898 the second scheduling pass can be enabled via 8899 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 8900 8901New Languages and Language specific improvements 8902 8903 Ada 8904 8905 * The Ada front end has been updated to include numerous bug fixes 8906 and enhancements. These include: 8907 + Improved project file support 8908 + Additional set of warnings about potential wrong code 8909 + Improved error messages 8910 + Improved code generation 8911 + Improved cross reference information 8912 + Improved inlining 8913 + Better run-time check elimination 8914 + Better error recovery 8915 + More efficient implementation of unbounded strings 8916 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 8917 ... 8918 + New GNAT.xxxx packages (e.g. GNAT.Strings, 8919 GNAT.Exception_Action) 8920 + New pragmas 8921 + New -gnatS switch replacing gnatpsta 8922 + Implementation of new Ada features (in particular limited 8923 with, limited aggregates) 8924 8925 C/Objective-C/C++ 8926 8927 * Precompiled headers are now supported. Precompiled headers can 8928 dramatically speed up compilation of some projects. There are some 8929 known defects in the current precompiled header implementation that 8930 will result in compiler crashes in relatively rare situations. 8931 Therefore, precompiled headers should be considered a "technology 8932 preview" in this release. Read the manual for details about how to 8933 use precompiled headers. 8934 * File handling in the preprocessor has been rewritten. GCC no longer 8935 gets confused by symlinks and hardlinks, and now has a correct 8936 implementation of #import and #pragma once. These two directives 8937 have therefore been un-deprecated. 8938 * The undocumented extension that allowed C programs to have a label 8939 at the end of a compound statement, which has been deprecated since 8940 GCC 3.0, has been removed. 8941 * The cast-as-lvalue extension has been removed for C++ and 8942 deprecated for C and Objective-C. In particular, code like this: 8943 int i; 8944 (char) i = 5; 8945 8946 or this: 8947 char *p; 8948 ((int *) p)++; 8949 8950 is no longer accepted for C++ and will not be accepted for C and 8951 Objective-C in a future version. 8952 * The conditional-expression-as-lvalue extension has been deprecated 8953 for C and Objective-C. In particular, code like this: 8954 int a, b, c; 8955 (a ? b : c) = 2; 8956 8957 will not be accepted for C and Objective-C in a future version. 8958 * The compound-expression-as-lvalue extension has been deprecated for 8959 C and Objective-C. In particular, code like this: 8960 int a, b; 8961 (a, b) = 2; 8962 8963 will not be accepted for C and Objective-C in a future version. A 8964 possible non-intrusive workaround is the following: 8965 (*(a, &b)) = 2; 8966 8967 * Several [10]built-in functions such as __builtin_popcount for 8968 counting bits, finding the highest and lowest bit in a word, and 8969 parity have been added. 8970 * The -fwritable-strings option has been deprecated and will be 8971 removed. 8972 * Many C math library functions are now recognized as built-ins and 8973 optimized. 8974 * The C, C++, and Objective-C compilers can now handle source files 8975 written in any character encoding supported by the host C library. 8976 The default input character set is taken from the current locale, 8977 and may be overridden with the -finput-charset command line option. 8978 In the future we will add support for inline encoding markers. 8979 8980 C++ 8981 8982 * G++ is now much closer to full conformance to the ISO/ANSI C++ 8983 standard. This means, among other things, that a lot of invalid 8984 constructs which used to be accepted in previous versions will now 8985 be rejected. It is very likely that existing C++ code will need to 8986 be fixed. This document lists some of the most common issues. 8987 * A hand-written recursive-descent C++ parser has replaced the 8988 YACC-derived C++ parser from previous GCC releases. The new parser 8989 contains much improved infrastructure needed for better parsing of 8990 C++ source codes, handling of extensions, and clean separation 8991 (where possible) between proper semantics analysis and parsing. The 8992 new parser fixes many bugs that were found in the old parser. 8993 * You must now use the typename and template keywords to disambiguate 8994 dependent names, as required by the C++ standard. 8995 struct K { 8996 typedef int mytype_t; 8997 }; 8998 8999 template <class T1> struct A { 9000 template <class T2> struct B { 9001 void callme(void); 9002 }; 9003 9004 template <int N> void bar(void) 9005 { 9006 // Use 'typename' to tell the parser that T1::mytype_t names 9007 // a type. This is needed because the name is dependent (in 9008 // this case, on template parameter T1). 9009 typename T1::mytype_t x; 9010 x = 0; 9011 } 9012 }; 9013 9014 template <class T> void template_func(void) 9015 { 9016 // Use 'template' to prefix member templates within 9017 // dependent types (a has type A<T>, which depends on 9018 // the template parameter T). 9019 A<T> a; 9020 a.template bar<0>(); 9021 9022 // Use 'template' to tell the parser that B is a nested 9023 // template class (dependent on template parameter T), and 9024 // 'typename' because the whole A<T>::B<int> is 9025 // the name of a type (again, dependent). 9026 typename A<T>::template B<int> b; 9027 b.callme(); 9028 } 9029 9030 void non_template_func(void) 9031 { 9032 // Outside of any template class or function, no names can be 9033 // dependent, so the use of the keyword 'typename' and 'template' 9034 // is not needed (and actually forbidden). 9035 A<K> a; 9036 a.bar<0>(); 9037 A<K>::B<float> b; 9038 b.callme(); 9039 } 9040 * In a template definition, unqualified names will no longer find 9041 members of a dependent base (as specified by [temp.dep]/3 in the 9042 C++ standard). For example, 9043 template <typename T> struct B { 9044 int m; 9045 int n; 9046 int f (); 9047 int g (); 9048 }; 9049 int n; 9050 int g (); 9051 template <typename T> struct C : B<T> { 9052 void h () 9053 { 9054 m = 0; // error 9055 f (); // error 9056 n = 0; // ::n is modified 9057 g (); // ::g is called 9058 } 9059 }; 9060 You must make the names dependent, e.g. by prefixing them with 9061 this->. Here is the corrected definition of C<T>::h, 9062 template <typename T> void C<T>::h () 9063 { 9064 this->m = 0; 9065 this->f (); 9066 this->n = 0 9067 this->g (); 9068 } 9069 As an alternative solution (unfortunately not backwards compatible 9070 with GCC 3.3), you may use using declarations instead of this->: 9071 template <typename T> struct C : B<T> { 9072 using B<T>::m; 9073 using B<T>::f; 9074 using B<T>::n; 9075 using B<T>::g; 9076 void h () 9077 { 9078 m = 0; 9079 f (); 9080 n = 0; 9081 g (); 9082 } 9083 }; 9084 * In templates, all non-dependent names are now looked up and bound 9085 at definition time (while parsing the code), instead of later when 9086 the template is instantiated. For instance: 9087 void foo(int); 9088 9089 template <int> struct A { 9090 static void bar(void){ 9091 foo('a'); 9092 } 9093 }; 9094 9095 void foo(char); 9096 9097 int main() 9098 { 9099 A<0>::bar(); // Calls foo(int), used to call foo(char). 9100 } 9101 9102 * In an explicit instantiation of a class template, you must use 9103 class or struct before the template-id: 9104 template <int N> 9105 class A {}; 9106 9107 template A<0>; // error, not accepted anymore 9108 template class A<0>; // OK 9109 * The "named return value" and "implicit typename" extensions have 9110 been removed. 9111 * Default arguments in function types have been deprecated and will 9112 be removed. 9113 * ARM-style name-injection of friend declarations has been deprecated 9114 and will be removed. For example: struct S { friend void f(); }; 9115 void g() { f(); } will not be accepted by future versions of G++; 9116 instead a declaration of "f" will need to be present outside of the 9117 scope of "S". 9118 * Covariant returns are implemented for all but varadic functions 9119 that require an adjustment. 9120 * When -pedantic is used, G++ now issues errors about spurious 9121 semicolons. For example, 9122 namespace N {}; // Invalid semicolon. 9123 void f() {}; // Invalid semicolon. 9124 * G++ no longer accepts attributes for a declarator after the 9125 initializer associated with that declarator. For example, 9126 X x(1) __attribute__((...)); 9127 is no longer accepted. Instead, use: 9128 X x __attribute__((...)) (1); 9129 * Inside the scope of a template class, the name of the class itself 9130 can be treated as either a class or a template. So GCC used to 9131 accept the class name as argument of type template, and template 9132 template parameter. However this is not C++ standard compliant. Now 9133 the name is not treated as a valid template template argument 9134 unless you qualify the name by its scope. For example, the code 9135 below no longer compiles. 9136 template <template <class> class TT> class X {}; 9137 template <class T> class Y { 9138 X<Y> x; // Invalid, Y is always a type template parameter. 9139 }; 9140 The valid code for the above example is 9141 X< ::Y> x; // Valid. 9142 (Notice the space between < and : to prevent GCC to interpret this 9143 as a digraph for [.) 9144 * Friend declarations that refer to template specializations are 9145 rejected if the template has not already been declared. For 9146 example, 9147 template <typename T> 9148 class C { 9149 friend void f<> (C&); 9150 }; 9151 is rejected. You must first declare f as a template, 9152 template <typename T> 9153 void f(T); 9154 * In case of friend declarations, every name used in the friend 9155 declaration must be accessible at the point of that declaration. 9156 Previous versions of G++ used to be less strict about this and 9157 allowed friend declarations for private class members, for example. 9158 See the ISO C++ Standard Committee's [11]defect report #209 for 9159 details. 9160 * Declaration of member functions of class templates as friends are 9161 supported. For example, 9162 template <typename T> struct A { 9163 void f(); 9164 }; 9165 class C { 9166 template <typename T> friend void A<T>::f(); 9167 }; 9168 * You must use template <> to introduce template specializations, as 9169 required by the standard. For example, 9170 template <typename T> 9171 struct S; 9172 9173 struct S<int> { }; 9174 is rejected. You must write, 9175 template <> struct S<int> {}; 9176 * G++ used to accept code like this, 9177 struct S { 9178 int h(); 9179 void f(int i = g()); 9180 int g(int i = h()); 9181 }; 9182 This behavior is not mandated by the standard. Now G++ issues an 9183 error about this code. To avoid the error, you must move the 9184 declaration of g before the declaration of f. The default arguments 9185 for g must be visible at the point where it is called. 9186 * The C++ ABI Section 3.3.3 specifications for the array construction 9187 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 9188 NULL when the allocator argument returns NULL. These changes are 9189 incorporated into the libstdc++ runtime library. 9190 * Using a name introduced by a typedef in a friend declaration or in 9191 an explicit instantiation is now rejected, as specified by the ISO 9192 C++ standard. 9193 class A; 9194 typedef A B; 9195 class C { 9196 friend class B; // error, no typedef name here 9197 friend B; // error, friend always needs class/struct/enum 9198 friend class A; // OK 9199 }; 9200 9201 template <int> class Q {}; 9202 typedef Q<0> R; 9203 template class R; // error, no typedef name here 9204 template class Q<0>; // OK 9205 * When allocating an array with a new expression, GCC used to allow 9206 parentheses around the type name. This is actually ill-formed and 9207 it is now rejected: 9208 int* a = new (int)[10]; // error, not accepted anymore 9209 int* a = new int[10]; // OK 9210 * When binding an rvalue of class type to a reference, the copy 9211 constructor of the class must be accessible. For instance, consider 9212 the following code: 9213 class A 9214 { 9215 public: 9216 A(); 9217 9218 private: 9219 A(const A&); // private copy ctor 9220 }; 9221 9222 A makeA(void); 9223 void foo(const A&); 9224 9225 void bar(void) 9226 { 9227 foo(A()); // error, copy ctor is not accessible 9228 foo(makeA()); // error, copy ctor is not accessible 9229 9230 A a1; 9231 foo(a1); // OK, a1 is a lvalue 9232 } 9233 This might be surprising at first sight, especially since most 9234 popular compilers do not correctly implement this rule ([12]further 9235 details). 9236 * When forming a pointer to member or a pointer to member function, 9237 access checks for class visibility (public, protected, private) are 9238 now performed using the qualifying scope of the name itself. This 9239 is better explained with an example: 9240 class A 9241 { 9242 public: 9243 void pub_func(); 9244 protected: 9245 void prot_func(); 9246 private: 9247 void priv_func(); 9248 }; 9249 9250 class B : public A 9251 { 9252 public: 9253 void foo() 9254 { 9255 &A::pub_func; // OK, pub_func is accessible through A 9256 &A::prot_func; // error, cannot access prot_func through A 9257 &A::priv_func; // error, cannot access priv_func through A 9258 9259 &B::pub_func; // OK, pub_func is accessible through B 9260 &B::prot_func; // OK, can access prot_func through B (within B) 9261 &B::priv_func; // error, cannot access priv_func through B 9262 } 9263 }; 9264 9265 Runtime Library (libstdc++) 9266 9267 * Optimization work: 9268 + Streamlined streambuf, filebuf, separate synched with C 9269 Standard I/O streambuf. 9270 + All formatted I/O now uses cached locale information. 9271 + STL optimizations (memory/speed for list, red-black trees as 9272 used by sets and maps). 9273 + More use of GCC builtins. 9274 + String optimizations (avoid contention on 9275 increment/decrement-and-test of the reference count in the 9276 empty-string object, constructor from input_iterators 9277 speedup). 9278 * Static linkage size reductions. 9279 * Large File Support (files larger than 2 GB on 32-bit systems). 9280 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 9281 * Generic character traits. 9282 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 9283 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 9284 * The allocator class is now standard-conformant, and two additional 9285 extension allocators have been added, mt_alloc and 9286 bitmap_allocator. 9287 * PCH support: -include bits/stdc++.h (2x compile speedup). 9288 * Rewrote __cxa_demangle with support for C++ style allocators. 9289 * New debug modes for STL containers and iterators. 9290 * Testsuite rewrite: five times as many tests, plus increasingly 9291 sophisticated tests, including I/O, MT, multi-locale, wide and 9292 narrow characters. 9293 * Use current versions of GNU "autotools" for build/configuration. 9294 9295 Objective-C 9296 9297 * The Objective-C front end has been updated to include the numerous 9298 bug fixes and enhancements previously available only in Apple's 9299 version of GCC. These include: 9300 + Structured exception (@try... @catch... @finally, @throw) and 9301 synchronization (@synchronized) support. These are accessible 9302 via the -fobjc-exceptions switch; as of this writing, they may 9303 only be used in conjunction with -fnext-runtime on Mac OS X 9304 10.3 and later. See [13]Options Controlling Objective-C 9305 Dialect for more information. 9306 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 9307 may now be encoded as 'B'. In addition, the back-end/codegen 9308 dependencies have been removed. 9309 + An overhaul of message dispatch construction, ensuring that 9310 the various receiver types (and casts thereof) are handled 9311 properly, and that correct diagnostics are issued. 9312 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 9313 (-freplace-objc-classes) debugging modes, currently available 9314 on Mac OS X 10.3 and later. See [14]Options Controlling 9315 Objective-C Dialect for more information. 9316 + Access to optimized runtime entry points (-fno-nil-receivers ) 9317 on the assumption that message receivers are never nil. This 9318 is currently available on Mac OS X 10.3 and later. See 9319 [15]Options Controlling Objective-C Dialect for more 9320 information. 9321 9322 Java 9323 9324 * Compiling a .jar file will now cause non-.class entries to be 9325 automatically compiled as resources. 9326 * libgcj has been ported to Darwin. 9327 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 9328 to gcj. 9329 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 9330 code from shared libraries. 9331 * libgcj has been much more completely merged with [16]GNU Classpath. 9332 * Class loading is now much more correct; in particular the caller's 9333 class loader is now used when that is required. 9334 * [17]Eclipse 2.x will run out of the box using gij. 9335 * Parts of java.nio have been implemented. Direct and indirect 9336 buffers work, as do fundamental file and socket operations. 9337 * java.awt has been improved, though it is still not ready for 9338 general use. 9339 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 9340 method. 9341 * The MinGW port has matured. Enhancements include socket timeout 9342 support, thread interruption, improved Runtime.exec() handling and 9343 support for accented characters in filenames. 9344 9345 Fortran 9346 9347 * Fortran improvements are listed in the [18]Fortran documentation. 9348 9349New Targets and Target Specific Improvements 9350 9351 Alpha 9352 9353 * Several [19]built-in functions have been added such as 9354 __builtin_alpha_zap to allow utilizing the more obscure 9355 instructions of the CPU. 9356 * Parameter passing of complex arguments has changed to match the 9357 [20]ABI. This change is incompatible with previous GCC versions, 9358 but does fix compatibility with the Tru64 compiler and several 9359 corner cases where GCC was incompatible with itself. 9360 9361 ARM 9362 9363 * Nicolas Pitre has contributed his hand-coded floating-point support 9364 code for ARM. It is both significantly smaller and faster than the 9365 existing C-based implementation, even when building applications 9366 for Thumb. The arm-elf configuration has been converted to use the 9367 new code. 9368 * Support for the Intel's iWMMXt architecture, a second generation 9369 XScale processor, has been added. Enabled at run time with the 9370 -mcpu=iwmmxt command line switch. 9371 * A new ARM target has been added: arm-wince-pe. This is similar to 9372 the arm-pe target, but it defaults to using the APCS32 ABI. 9373 * The existing ARM pipeline description has been converted to the use 9374 the [21]DFA processor pipeline model. There is not much change in 9375 code performance, but the description is now [22]easier to 9376 understand. 9377 * Support for the Cirrus EP9312 Maverick floating point co-processor 9378 added. Enabled at run time with the -mcpu=ep9312 command line 9379 switch. Note however that the multilibs to support this chip are 9380 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 9381 enable their production you will have to uncomment the entries in 9382 that file. 9383 9384 H8/300 9385 9386 * Support for long long has been added. 9387 * Support for saveall attribute has been added. 9388 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 9389 for H8/300H and H8S, which is much faster than the previous 9390 implementation. 9391 * A lot of small performance improvements. 9392 9393 IA-32/AMD64 (x86-64) 9394 9395 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 9396 -march=k8 and -mcpu=k8. 9397 * Scalar SSE code generation carefully avoids reformatting penalties, 9398 hidden dependencies and minimizes the number of uops generated on 9399 both Intel and AMD CPUs. 9400 * Vector MMX and SSE operands are now passed in registers to improve 9401 performance and match the argument passing convention used by the 9402 Intel C++ Compiler. As a result it is not possible to call 9403 functions accepting vector arguments compiled by older GCC version. 9404 * Conditional jump elimination is now more aggressive on modern CPUs. 9405 * The Athlon ports has been converted to use the DFA processor 9406 pipeline description. 9407 * Optimization of indirect tail calls is now possible in a similar 9408 fashion as direct sibcall optimization. 9409 * Further small performance improvements. 9410 * -m128bit-long-double is now less buggy. 9411 * __float128 support in 64-bit compilation. 9412 * Support for data structures exceeding 2GB in 64-bit mode. 9413 * -mcpu has been renamed to -mtune. 9414 9415 IA-64 9416 9417 * Tuning code for the Itanium 2 processor has been added. The 9418 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 9419 enabled by default now. To generate code tuned for Itanium 1 the 9420 option -mtune=itanium1 should be used. 9421 * [23]DFA processor pipeline descriptions for the IA-64 processors 9422 have been added. This resulted in about 3% improvement on the 9423 SPECInt2000 benchmark for Itanium 2. 9424 * Instruction bundling for the IA-64 processors has been rewritten 9425 using the DFA pipeline hazard recognizer. It resulted in about 60% 9426 compiler speedup on the SPECInt2000 C programs. 9427 9428 M32R 9429 9430 * Support for the M32R/2 processor has been added by Renesas. 9431 * Support for an M32R GNU/Linux target and PIC code generation has 9432 been added by Renesas. 9433 9434 M68000 9435 9436 * Bernardo Innocenti (Develer S.r.l.) has contributed the 9437 m68k-uclinux target, based on former work done by Paul Dale 9438 (SnapGear Inc.). Code generation for the ColdFire processors family 9439 has been enhanced and extended to support the MCF 53xx and MCF 54xx 9440 cores, integrating former work done by Peter Barada (Motorola). 9441 9442 MIPS 9443 9444 Processor-specific changes 9445 9446 * Support for the RM7000 and RM9000 processors has been added. It can 9447 be selected using the -march compiler option and should work with 9448 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 9449 * Support for revision 2 of the MIPS32 ISA has been added. It can be 9450 selected with the command-line option -march=mips32r2. 9451 * There is a new option, -mfix-sb1, to work around certain SB-1 9452 errata. 9453 9454 Configuration 9455 9456 * It is possible to customize GCC using the following configure-time 9457 options: 9458 + --with-arch, which specifies the default value of the -march 9459 option. 9460 + --with-tune, which specifies the default value of the -mtune 9461 option. 9462 + --with-abi, which specifies the default ABI. 9463 + --with-float=soft, which tells GCC to use software floating 9464 point by default. 9465 + --with-float=hard, which tells GCC to use hardware floating 9466 point by default. 9467 * A 64-bit GNU/Linux port has been added. The associated 9468 configurations are mips64-linux-gnu and mips64el-linux-gnu. 9469 * The 32-bit GNU/Linux port now supports Java. 9470 * The IRIX 6 configuration now supports the o32 ABI and will build 9471 o32 multilibs by default. This support is compatible with both 9472 binutils and the SGI tools, but note that several features, 9473 including debugging information and DWARF2 exception handling, are 9474 only available when using the GNU assembler. Use of the GNU 9475 assembler and linker (version 2.15 or above) is strongly 9476 recommended. 9477 * The IRIX 6 configuration now supports 128-bit long doubles. 9478 * There are two new RTEMS-specific configurations, mips-rtems and 9479 mipsel-rtems. 9480 * There are two new *-elf configurations, mipsisa32r2-elf and 9481 mipsisa32r2el-elf. 9482 9483 General 9484 9485 * Several [24]ABI bugs have been fixed. Unfortunately, these changes 9486 will break binary compatibility with earlier releases. 9487 * GCC can now use explicit relocation operators when generating 9488 -mabicalls code. This behavior is controlled by -mexplicit-relocs 9489 and can have several performance benefits. For example: 9490 + It allows for more optimization of GOT accesses, including 9491 better scheduling and redundancy elimination. 9492 + It allows sibling calls to be implemented as jumps. 9493 + n32 and n64 leaf functions can use a call-clobbered global 9494 pointer instead of $28. 9495 + The code to set up $gp can be removed from functions that 9496 don't need it. 9497 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 9498 option is equivalent to the assembler's -xgot option and should be 9499 used instead of -Wa,-xgot. 9500 * Frame pointer elimination is now supported when generating 64-bit 9501 MIPS16 code. 9502 * Inline block moves have been optimized to take more account of 9503 alignment information. 9504 * Many internal changes have been made to the MIPS port, mostly aimed 9505 at reducing the reliance on assembler macros. 9506 9507 PowerPC 9508 9509 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 9510 [25]ABI incompatibilities regarding the way parameters are passed 9511 during functions calls. These changes may result in incompatibility 9512 between code compiled with GCC 3.3 and GCC 3.4. 9513 9514 PowerPC Darwin 9515 9516 * Support for shared/dylib gcc libraries has been added. It is 9517 enabled by default on powerpc-apple-darwin7.0.0 and up. 9518 * Libgcj is enabled by default. On systems older than 9519 powerpc-apple-darwin7.0.0 you need to install dlcompat. 9520 * 128-bit IBM extended precision format support added for long 9521 double. 9522 9523 PowerPC64 GNU/Linux 9524 9525 * By default, PowerPC64 GNU/Linux now uses natural alignment of 9526 structure elements. The old four byte alignment for double, with 9527 special rules for a struct starting with a double, can be chosen 9528 with -malign-power. This change may result in incompatibility 9529 between code compiled with GCC 3.3 and GCC 3.4. 9530 * -mabi=altivec is now the default rather than -mabi=no-altivec. 9531 * 128-bit IBM extended precision format support added for long 9532 double. 9533 9534 S/390 and zSeries 9535 9536 * New command-line options allow to specify the intended execution 9537 environment for generated code: 9538 + -mesa/-mzarch allows to specify whether to generate code 9539 running in ESA/390 mode or in z/Architecture mode (this is 9540 applicable to 31-bit code only). 9541 + -march allows to specify a minimum processor architecture 9542 level (g5, g6, z900, or z990). 9543 + -mtune allows to specify which processor to tune for. 9544 * It is possible to customize GCC using the following configure-time 9545 options: 9546 + --with-mode, which specifies whether to default to assuming 9547 ESA/390 or z/Architecture mode. 9548 + --with-arch, which specifies the default value of the -march 9549 option. 9550 + --with-tune, which specifies the default value of the -mtune 9551 option. 9552 * Support for the z990 processor has been added, and can be selected 9553 using -march=z990 or -mtune=z990. This includes instruction 9554 scheduling tuned for the superscalar instruction pipeline of the 9555 z990 processor as well as support for all new instructions provided 9556 by the long-displacement facility. 9557 * Support to generate 31-bit code optimized for zSeries processors 9558 (running in ESA/390 or in z/Architecture mode) has been added. This 9559 can be selected using -march=z900 and -mzarch respectively. 9560 * Instruction scheduling for the z900 and z990 processors now uses 9561 the DFA pipeline hazard recognizer. 9562 * GCC no longer generates code to maintain a stack backchain, 9563 previously used to generate stack backtraces for debugging 9564 purposes. As replacement that does not incur runtime overhead, 9565 DWARF-2 call frame information is provided by GCC; this is 9566 supported by GDB 6.1. The old behavior can be restored using the 9567 -mbackchain option. 9568 * The stack frame size of functions may now exceed 2 GB in 64-bit 9569 code. 9570 * A port for the 64-bit IBM TPF operating system has been added; the 9571 configuration is s390x-ibm-tpf. This configuration is supported as 9572 cross-compilation target only. 9573 * Various changes to improve the generated code have been 9574 implemented, including: 9575 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 9576 instructions to significantly speed up many floating-point 9577 applications. 9578 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 9579 WITH BORROW instructions to speed up long long arithmetic. 9580 + GCC now uses the SEARCH STRING instruction to implement 9581 strlen(). 9582 + In many cases, function call overhead for 31-bit code has been 9583 reduced by placing the literal pool after the function code 9584 instead of after the function prolog. 9585 + Register 14 is no longer reserved in 64-bit code. 9586 + Handling of global register variables has been improved. 9587 9588 SPARC 9589 9590 * The option -mflat is deprecated. 9591 * Support for large (> 2GB) frames has been added to the 64-bit port. 9592 * Several [26]ABI bugs have been fixed. Unfortunately, these changes 9593 will break binary compatibility with earlier releases. 9594 * The default debugging format has been switched from STABS to 9595 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 9596 the default debugging format for 64-bit code on Solaris. 9597 9598 SuperH 9599 9600 * Support for the SH2E processor has been added. Enabled at run time 9601 with the -m2e command line switch, or at configure time by 9602 specifying sh2e as the machine part of the target triple. 9603 9604 V850 9605 9606 * Support for the Mitsubishi V850E1 processor has been added. This is 9607 a variant of the V850E processor with some additional debugging 9608 instructions. 9609 9610 Xtensa 9611 9612 * Several ABI bugs have been fixed. Unfortunately, these changes 9613 break binary compatibility with earlier releases. 9614 + For big-endian processors, the padding of aggregate return 9615 values larger than a word has changed. If the size of an 9616 aggregate return value is not a multiple of 32 bits, previous 9617 versions of GCC inserted padding in the most-significant bytes 9618 of the first return value register. Aggregates larger than a 9619 word are now padded in the least-significant bytes of the last 9620 return value register used. Aggregates smaller than a word are 9621 still padded in the most-significant bytes. The return value 9622 padding has not changed for little-endian processors. 9623 + Function arguments with 16-byte alignment are now properly 9624 aligned. 9625 + The implementation of the va_list type has changed. A va_list 9626 value created by va_start from a previous release cannot be 9627 used with va_arg from this release, or vice versa. 9628 * More processor configuration options for Xtensa processors are 9629 supported: 9630 + the ABS instruction is now optional; 9631 + the ADDX* and SUBX* instructions are now optional; 9632 + an experimental CONST16 instruction can be used to synthesize 9633 constants instead of loading them from constant pools. 9634 These and other Xtensa processor configuration options can no 9635 longer be enabled or disabled by command-line options; the 9636 processor configuration must be specified by the xtensa-config.h 9637 header file when building GCC. Additionally, the 9638 -mno-serialize-volatile option is no longer supported. 9639 9640Obsolete Systems 9641 9642 Support for a number of older systems has been declared obsolete in GCC 9643 3.4. Unless there is activity to revive them, the next release of GCC 9644 will have their sources permanently removed. 9645 9646 All configurations of the following processor architectures have been 9647 declared obsolete: 9648 * Mitsubishi D30V, d30v-* 9649 * AT&T DSP1600 and DSP1610, dsp16xx-* 9650 * Intel 80960, i960 9651 9652 Also, some individual systems have been obsoleted: 9653 * ARM Family 9654 + Support for generating code for operation in APCS/26 mode 9655 (-mapcs-26). 9656 * IBM ESA/390 9657 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 9658 maintained and supported.) 9659 * Intel 386 family 9660 + MOSS, i?86-moss-msdos and i?86-*-moss* 9661 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 9662 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 9663 i?86-*-freebsd2* 9664 + GNU/Linux with a.out object format, i?86-linux*aout* 9665 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 9666 + Interix versions before Interix 3, i?86-*-interix 9667 + Mach microkernel, i?86-mach* 9668 + SCO UnixWare with UDK, i?86-*-udk* 9669 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 9670 + VSTa microkernel, i386-*-vsta 9671 * Motorola M68000 family 9672 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 9673 + NetBSD with a.out object format (before NetBSD 1.4), 9674 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 9675 + Generic System V r.4, m68k-*-sysv4* 9676 * VAX 9677 + Generic VAX, vax-*-* (This is generic VAX only; we have not 9678 obsoleted any VAX triples for specific operating systems.) 9679 9680Documentation improvements 9681 9682Other significant improvements 9683 9684 * The build system has undergone several significant cleanups. 9685 Subdirectories will only be configured if they are being built, and 9686 all subdirectory configures are run from the make command. The top 9687 level has been autoconfiscated. 9688 * Building GCC no longer writes to its source directory. This should 9689 help those wishing to share a read-only source directory over NFS 9690 or build from a CD. The exceptions to this feature are if you 9691 configure with either --enable-maintainer-mode or 9692 --enable-generated-files-in-srcdir. 9693 * The -W warning option has been renamed to -Wextra, which is more 9694 easily understood. The older spelling will be retained for 9695 backwards compatibility. 9696 * Substantial improvements in compile time have been made, 9697 particularly for non-optimizing compilations. 9698 __________________________________________________________________ 9699 9700GCC 3.4.0 9701 9702 Bug Fixes 9703 9704 A vast number of bugs have been fixed in 3.4.0, too many to publish a 9705 complete list here. [27]Follow this link to query the Bugzilla database 9706 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 9707 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 9708 regressions. 9709 __________________________________________________________________ 9710 9711GCC 3.4.1 9712 9713 Bug Fixes 9714 9715 This section lists the problem reports (PRs) from GCC's bug tracking 9716 system that are known to be fixed in the 3.4.1 release. This list might 9717 not be complete (that is, it is possible that some PRs that have been 9718 fixed are not listed here). 9719 9720 Bootstrap failures 9721 9722 * [28]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 9723 emitted - PIC related 9724 * [29]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 9725 * [30]14760 A bug in configure.in prevents using both 9726 --program-suffix and --program-prefix 9727 * [31]14671 [hppa64] bootstrap fails: ICE in 9728 save_call_clobbered_regs, in caller_save.c 9729 * [32]15093 [alpha][Java] make bootstrap fails to configure libffi on 9730 Alpha 9731 * [33]15178 Solaris 9/x86 fails linking after stage 3 9732 9733 Multi-platform internal compiler errors (ICEs) 9734 9735 * [34]12753 (preprocessor) Memory corruption in preprocessor on bad 9736 input 9737 * [35]13985 ICE in gcc.c-torture/compile/930621-1.c 9738 * [36]14810 (c++) tree check failures with invalid code involving 9739 templates 9740 * [37]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 9741 cp/parser.c 9742 * [38]15044 (c++) ICE on syntax error, template header 9743 * [39]15057 (c++) Compiling of conditional value throw constructs 9744 cause a segmentation violation 9745 * [40]15064 (c++) typeid of template parameter gives ICE 9746 * [41]15142 (c++) ICE when passing a string where a char* is expected 9747 in a throw statement 9748 * [42]15159 ICE in rtl_verify_flow_info_1 9749 * [43]15165 (c++) ICE in instantiate_template 9750 * [44]15193 Unary minus using pointer to V4SF vector causes 9751 -fforce-mem to exhaust all memory 9752 * [45]15209 (c++) Runs out of memory with packed structs 9753 * [46]15227 (c++) Trouble with invalid function definition 9754 * [47]15285 (c++) instantiate_type ICE when forming pointer to 9755 template function 9756 * [48]15299 (c++) ICE in resolve_overloaded_unification 9757 * [49]15329 (c++) ICE on constructor of member template 9758 * [50]15550 ICE in extract_insn, in recog.c 9759 * [51]15554 (c++) ICE in tsubst_copy, in cp/pt.c 9760 * [52]15640 (c++) ICE on invalid code in arg_assoc, in 9761 cp/name-lookup.c 9762 * [53]15666 [unit-at-a-time] Gcc abort on valid code 9763 * [54]15696 (c++) ICE with bad pointer-to-member code 9764 * [55]15701 (c++) ICE with friends and template template parameter 9765 * [56]15761 ICE in do_SUBST, in combine.c 9766 * [57]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 9767 9768 Ada 9769 9770 * [58]14538 All RTEMS targets broken for gnat 9771 9772 C front end 9773 9774 * [59]12391 missing warning about assigning to an incomplete type 9775 * [60]14649 atan(1.0) should not be a constant expression 9776 * [61]15004 [unit-at-a-time] no warning for unused paramater in 9777 static function 9778 * [62]15749 --pedantic-errors behaves differently from --pedantic 9779 with C-compiler on GNU/Linux 9780 9781 C++ compiler and library 9782 9783 * [63]10646 non-const reference is incorrectly matched in a "const T" 9784 partial specialization 9785 * [64]12077 wcin.rdbuf()->in_avail() return value too high 9786 * [65]13598 enc_filebuf doesn't work 9787 * [66]14211 const_cast returns lvalue but should be rvalue 9788 * [67]14220 num_put::do_put() undesired float/double behavior 9789 * [68]14245 problem with user-defined allocators in std::basic_string 9790 * [69]14340 libstdc++ Debug mode: failure to convert iterator to 9791 const_iterator 9792 * [70]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 9793 FILE* 9794 * [71]14668 no warning anymore for reevaluation of declaration 9795 * [72]14775 LFS (large file support) tests missing 9796 * [73]14821 Duplicate namespace alias declaration should not conflict 9797 * [74]14930 Friend declaration ignored 9798 * [75]14932 cannot use offsetof to get offsets of array elements in 9799 g++ 3.4.0 9800 * [76]14950 [non unit-at-a-time] always_inline does not mix with 9801 templates and -O0 9802 * [77]14962 g++ ignores #pragma redefine_extname 9803 * [78]14975 Segfault on low-level write error during imbue 9804 * [79]15002 Linewise stream input is unusably slow (std::string slow) 9805 * [80]15025 compiler accepts redeclaration of template as 9806 non-template 9807 * [81]15046 [arm] Math functions misdetected by cross configuration 9808 * [82]15069 a bit test on a variable of enum type is miscompiled 9809 * [83]15074 g++ -lsupc++ still links against libstdc++ 9810 * [84]15083 spurious "statement has no effect" warning 9811 * [85]15096 parse error with templates and pointer to const member 9812 * [86]15287 combination of operator[] and operator .* fails in 9813 templates 9814 * [87]15317 __attribute__ unused in first parameter of constructor 9815 gives error 9816 * [88]15337 sizeof on incomplete type diagnostic 9817 * [89]15361 bitset<>::_Find_next fails 9818 * [90]15412 _GLIBCXX_ symbols symbols defined and used in different 9819 namespaces 9820 * [91]15427 valid code results in incomplete type error 9821 * [92]15471 Incorrect member pointer offsets in anonymous 9822 structs/unions 9823 * [93]15503 nested template problem 9824 * [94]15507 compiler hangs while laying out union 9825 * [95]15542 operator & and template definitions 9826 * [96]15565 SLES9: leading + sign for unsigned int with showpos 9827 * [97]15625 friend defined inside a template fails to find static 9828 function 9829 * [98]15629 Function templates, overloads, and friend name injection 9830 * [99]15742 'noreturn' attribute ignored in method of template 9831 functions. 9832 * [100]15775 Allocator::pointer consistently ignored 9833 * [101]15821 Duplicate namespace alias within namespace rejected 9834 * [102]15862 'enum yn' fails (confict with undeclared builtin) 9835 * [103]15875 rejects pointer to member in template 9836 * [104]15877 valid code using templates and anonymous enums is 9837 rejected 9838 * [105]15947 Puzzling error message for wrong destructor declaration 9839 in template class 9840 * [106]16020 cannot copy __gnu_debug::bitset 9841 * [107]16154 input iterator concept too restrictive 9842 * [108]16174 deducing top-level consts 9843 9844 Java 9845 9846 * [109]14315 Java compiler is not parallel make safe 9847 9848 Fortran 9849 9850 * [110]15151 [g77] incorrect logical i/o in 64-bit mode 9851 9852 Objective-C 9853 9854 * [111]7993 private variables cannot be shadowed in subclasses 9855 9856 Optimization bugs 9857 9858 * [112]15228 useless copies of floating point operands 9859 * [113]15345 [non-unit-at-a-time] unreferenced nested inline 9860 functions not optimized away 9861 * [114]15945 Incorrect floating point optimization 9862 * [115]15526 ftrapv aborts on 0 * (-1) 9863 * [116]14690 Miscompiled POOMA tests 9864 * [117]15112 GCC generates code to write to unchanging memory 9865 9866 Preprocessor 9867 9868 * [118]15067 Minor glitch in the source of cpp 9869 9870 Main driver program bugs 9871 9872 * [119]1963 collect2 interprets -oldstyle_liblookup as -o 9873 ldstyle_liblookup 9874 9875 x86-specific (Intel/AMD) 9876 9877 * [120]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 9878 section} 9879 9880 HPPA-specific 9881 9882 * [121]14782 GCC produces an unaligned data access at -O2 9883 * [122]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 9884 * [123]15202 ICE in reload_cse_simplify_operands, in postreload.c 9885 9886 IA64-specific 9887 9888 * [124]14610 __float80 constants incorrectly emitted 9889 * [125]14813 init_array sections are initialized in the wrong order 9890 * [126]14857 GCC segfault on duplicated asm statement 9891 * [127]15598 Gcc 3.4 ICE on valid code 9892 * [128]15653 Gcc 3.4 ICE on valid code 9893 9894 MIPS-specific 9895 9896 * [129]15189 wrong filling of delay slot with -march=mips1 -G0 9897 -mno-split-addresses -mno-explicit-relocs 9898 * [130]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 9899 2.14.91 9900 * [131]16144 Bogus reference to __divdf3 when -O1 9901 * [132]16176 Miscompilation of unaligned data in MIPS backend 9902 9903 PowerPC-specific 9904 9905 * [133]11591 ICE in gcc.dg/altivec-5.c 9906 * [134]12028 powerpc-eabispe produces bad sCOND operation 9907 * [135]14478 rs6000 geu/ltu patterns generate incorrect code 9908 * [136]14567 long double and va_arg complex args 9909 * [137]14715 Altivec stack layout may overlap gpr save with stack 9910 temps 9911 * [138]14902 (libstdc++) Stream checking functions fail when -pthread 9912 option is used. 9913 * [139]14924 Compiler ICE on valid code 9914 * [140]14960 -maltivec affects vector return with -mabi=no-altivec 9915 * [141]15106 vector varargs failure passing from altivec to 9916 non-altivec code for -m32 9917 * [142]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 9918 half-word operation 9919 * [143]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 9920 and stvx 9921 * [144]15662 Segmentation fault when an exception is thrown - even if 9922 try and catch are specified 9923 9924 s390-specific 9925 9926 * [145]15054 Bad code due to overlapping stack temporaries 9927 9928 SPARC-specific 9929 9930 * [146]15783 ICE with union assignment in 64-bit mode 9931 * [147]15626 GCC 3.4 emits "ld: warning: relocation error: 9932 R_SPARC_UA32" 9933 9934 x86-64-specific 9935 9936 * [148]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 9937 * [149]14723 Backported -march=nocona from mainline 9938 * [150]15290 __float128 failed to pass to function properly 9939 9940 Cygwin/Mingw32-specific 9941 9942 * [151]15250 Option -mms-bitfields support on GCC 3.4 is not 9943 conformant to MS layout 9944 * [152]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 9945 worker on windows32 targets 9946 9947 Bugs specific to embedded processors 9948 9949 * [153]8309 [m68k] -m5200 produces erroneous SImode set of short 9950 varaible on stack 9951 * [154]13250 [SH] Gcc code for rotation clobbers the register, but 9952 gcc continues to use the register as if it was not clobbered 9953 * [155]13803 [coldfire] movqi operand constraints too restrictivefor 9954 TARGET_COLDFIRE 9955 * [156]14093 [SH] ICE for code when using -mhitachi option in SH 9956 * [157]14457 [m6811hc] ICE with simple c++ source 9957 * [158]14542 [m6811hc] ICE on simple source 9958 * [159]15100 [SH] cc1plus got hang-up on 9959 libstdc++-v3/testsuite/abi_check.cc 9960 * [160]15296 [CRIS] Delayed branch scheduling causing invalid code on 9961 cris-* 9962 * [161]15396 [SH] ICE with -O2 -fPIC 9963 * [162]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 9964 ColdFire 9965 9966 Testsuite problems (compiler not affected) 9967 9968 * [163]11610 libstdc++ testcases 27_io/* don't work properly remotely 9969 * [164]15488 (libstdc++) possibly insufficient file permissions for 9970 executing test suite 9971 * [165]15489 (libstdc++) testsuite_files determined incorrectly 9972 9973 Documentation bugs 9974 9975 * [166]13928 (libstdc++) no whatis info in some man pages generated 9976 by doxygen 9977 * [167]14150 Ada documentation out of date 9978 * [168]14949 (c++) Need to document method visibility changes 9979 * [169]15123 libstdc++-doc: Allocators.3 manpage is empty 9980 __________________________________________________________________ 9981 9982GCC 3.4.2 9983 9984 Bug Fixes 9985 9986 This section lists the problem reports (PRs) from GCC's bug tracking 9987 system that are known to be fixed in the 3.4.2 release. This list might 9988 not be complete (that is, it is possible that some PRs that have been 9989 fixed are not listed here). 9990 9991 Bootstrap failures and issues 9992 9993 * [170]16469 [mips-sgi-irix5.3] bootstrap fails in 9994 libstdc++-v3/testsuite 9995 * [171]16344 [hppa-linux-gnu] libstdc++'s PCH built by 9996 profiledbootstrap does not work with the built compiler 9997 * [172]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 9998 9999 Multi-platform internal compiler errors (ICEs) 10000 10001 * [173]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 10002 cp_parser_class_specifier, in cp/parser.c 10003 * [174]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 10004 * [175]15461 (c++) ICE due to NRV and inlining 10005 * [176]15890 (c++) ICE in c_expand_expr, in c-common.c 10006 * [177]16180 ICE: segmentation fault in RTL optimization 10007 * [178]16224 (c++) ICE in write_unscoped_name (template/namespace) 10008 * [179]16408 ICE: in delete_insn, in cfgrtl.c 10009 * [180]16529 (c++) ICE for: namespace-alias shall not be declared as 10010 the name of any other entity 10011 * [181]16698 (c++) ICE with exceptions and declaration of __cxa_throw 10012 * [182]16706 (c++) ICE in finish_member_declaration, in 10013 cp/semantics.c 10014 * [183]16810 (c++) Legal C++ program with cast gives ICE in 10015 build_ptrmemfunc 10016 * [184]16851 (c++) ICE when throwing a comma expression 10017 * [185]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 10018 * [186]16904 (c++) ICE in finish_class_member_access_expr, in 10019 cp/typeck.c 10020 * [187]16905 (c++) ICE (segfault) with exceptions 10021 * [188]16964 (c++) ICE in cp_parser_class_specifier due to 10022 redefinition 10023 * [189]17068 (c++) ICE: tree check: expected class 'd', have 'x' 10024 (identifier_node) in dependent_template_p, in cp/pt.c 10025 10026 Preprocessor bugs 10027 10028 * [190]16366 Preprocessor option -remap causes memory corruption 10029 10030 Optimization 10031 10032 * [191]15345 unreferenced nested inline functions not optimized away 10033 * [192]16590 Incorrect execution when compiling with -O2 10034 * [193]16693 Bitwise AND is lost when used within a cast to an enum 10035 of the same precision 10036 * [194]17078 Jump into if(0) substatement fails 10037 10038 Problems in generated debug information 10039 10040 * [195]13956 incorrect stabs for nested local variables 10041 10042 C front end bugs 10043 10044 * [196]16684 GCC should not warn about redundant redeclarations of 10045 built-ins 10046 10047 C++ compiler and library 10048 10049 * [197]12658 Thread safety problems in locale::global() and 10050 locale::locale() 10051 * [198]13092 g++ accepts invalid pointer-to-member conversion 10052 * [199]15320 Excessive memory consumption 10053 * [200]16246 Incorrect template argument deduction 10054 * [201]16273 Memory exhausted when using nested classes and virtual 10055 functions 10056 * [202]16401 ostringstream in gcc 3.4.x very slow for big data 10057 * [203]16411 undefined reference to 10058 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 10059 >::file() 10060 * [204]16489 G++ incorrectly rejects use of a null constant integral 10061 expression as a null constant pointer 10062 * [205]16618 offsetof fails with constant member 10063 * [206]16637 syntax error reported for valid input code 10064 * [207]16717 __attribute__((constructor)) broken in C++ 10065 * [208]16813 compiler error in DEBUG version of range insertion 10066 std::map::insert 10067 * [209]16853 pointer-to-member initialization from incompatible one 10068 accepted 10069 * [210]16889 ambiguity is not detected 10070 * [211]16959 Segmentation fault in ios_base::sync_with_stdio 10071 10072 Java compiler and library 10073 10074 * [212]7587 direct threaded interpreter not thread-safe 10075 * [213]16473 ServerSocket accept() leaks file descriptors 10076 * [214]16478 Hash synchronization deadlock with finalizers 10077 10078 Alpha-specific 10079 10080 * [215]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 10081 * [216]16974 could not split insn (ice in final_scan_insn, in 10082 final.c) 10083 10084 x86-specific 10085 10086 * [217]16298 ICE in output_operand 10087 * [218]17113 ICE with SSE2 intrinsics 10088 10089 x86-64 specific 10090 10091 * [219]14697 libstdc++ couldn't find 32bit libgcc_s 10092 10093 MIPS-specific 10094 10095 * [220]15869 [mips64] No NOP after LW (with -mips1 -O0) 10096 * [221]16325 [mips64] value profiling clobbers gp on mips 10097 * [222]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 10098 char[]s 10099 * [223]16380 [mips64] Use of uninitialised register after dbra 10100 conversion 10101 * [224]16407 [mips64] Unaligned access to local variables 10102 * [225]16643 [mips64] verify_local_live_at_start ICE after 10103 crossjumping & cfgcleanup 10104 10105 ARM-specific 10106 10107 * [226]15927 THUMB -O2: strength-reduced iteration variable ends up 10108 off by 1 10109 * [227]15948 THUMB: ICE with non-commutative cbranch 10110 * [228]17019 THUMB: bad switch statement in md code for 10111 addsi3_cbranch_scratch 10112 10113 IA64-specific 10114 10115 * [229]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 10116 (-mtune=merced) 10117 * [230]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 10118 (-mtune=itanium) 10119 * [231]16278 Gcc failed to build Linux kernel with -mtune=merced 10120 * [232]16414 ICE on valid code: typo in comparison of asm_noperands 10121 result 10122 * [233]16445 ICE on valid code: don't count ignored insns 10123 * [234]16490 ICE (segfault) while compiling with -fprofile-use 10124 * [235]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 10125 10126 PowerPC-specific 10127 10128 * [236]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 10129 * [237]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 10130 issue) 10131 10132 SPARC-specific 10133 10134 * [238]16199 ICE while compiling apache 2.0.49 10135 * [239]16416 -m64 doesn't imply -mcpu=v9 anymore 10136 * [240]16430 ICE when returning non-C aggregates larger than 16 bytes 10137 10138 Bugs specific to embedded processors 10139 10140 * [241]16379 [m32r] can't output large model function call of memcpy 10141 * [242]17093 [m32r] ICE with -msdata=use -O0 10142 * [243]17119 [m32r] ICE at switch case 0x8000 10143 10144 DJGPP-specific 10145 10146 * [244]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 10147 10148 Alpha Tru64-specific 10149 10150 * [245]16210 libstdc++ gratuitously omits "long long" I/O 10151 10152 Testsuite, documentation issues (compiler is not affected): 10153 10154 * [246]15488 (libstdc++) possibly insufficient file permissions for 10155 executing test suite 10156 * [247]16250 ada/doctools runs makeinfo even in release tarball 10157 __________________________________________________________________ 10158 10159GCC 3.4.3 10160 10161 This is the [248]list of problem reports (PRs) from GCC's bug tracking 10162 system that are known to be fixed in the 3.4.3 release. This list might 10163 not be complete (that is, it is possible that some PRs that have been 10164 fixed are not listed here). 10165 10166 Bootstrap failures 10167 10168 * [249]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 10169 * [250]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 10170 when undeclared 10171 10172 Internal compiler errors (ICEs) affecting multiple platforms 10173 10174 * [251]13948 (java) GCJ segmentation fault while compiling GL4Java 10175 .class files 10176 * [252]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 10177 * [253]16301 (c++) ICE when "strong" attribute is attached to a using 10178 directive 10179 * [254]16566 ICE with flexible arrays 10180 * [255]17023 ICE with nested functions in parameter declaration 10181 * [256]17027 ICE with noreturn function in loop at -O2 10182 * [257]17524 ICE in grokdeclarator, in cp/decl.c 10183 * [258]17826 (c++) ICE in cp_tree_equal 10184 10185 C and optimization bugs 10186 10187 * [259]15526 -ftrapv aborts on 0 * (-1) 10188 * [260]16999 #ident stopped working 10189 * [261]17503 quadratic behaviour in invalid_mode_change_p 10190 * [262]17581 Long long arithmetic fails inside a switch/case 10191 statement when compiled with -O2 10192 * [263]18129 -fwritable-strings doesn't work 10193 10194 C++ compiler and library bugs 10195 10196 * [264]10975 incorrect initial ostringstream::tellp() 10197 * [265]11722 Unbuffered filebuf::sgetn is slow 10198 * [266]14534 Unrecognizing static function as a template parameter 10199 when its return value is also templated 10200 * [267]15172 Copy constructor optimization in aggregate 10201 initialization 10202 * [268]15786 Bad error message for frequently occuring error. 10203 * [269]16162 Rejects valid member-template-definition 10204 * [270]16612 empty basic_strings can't live in shared memory 10205 * [271]16715 std::basic_iostream is instantiated when used, even 10206 though instantiations are already contained in libstdc++ 10207 * [272]16848 code in /ext/demangle.h appears broken 10208 * [273]17132 GCC fails to eliminate function template specialization 10209 when argument deduction fails 10210 * [274]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 10211 in ropeimpl.h 10212 * [275]17327 use of `enumeral_type' in template type unification 10213 * [276]17393 "unused variable '._0'" warning with -Wall 10214 * [277]17501 Confusion with member templates 10215 * [278]17537 g++ not passing -lstdc++ to linker when all command line 10216 arguments are libraries 10217 * [279]17585 usage of unqualified name of static member from within 10218 class not allowed 10219 * [280]17821 Poor diagnostic for using "." instead of "->" 10220 * [281]17829 wrong error: call of overloaded function is ambiguous 10221 * [282]17851 Misleading diagnostic for invalid function declarations 10222 with undeclared types 10223 * [283]17976 Destructor is called twice 10224 * [284]18020 rejects valid definition of enum value in template 10225 * [285]18093 bogus conflict in namespace aliasing 10226 * [286]18140 C++ parser bug when using >> in templates 10227 10228 Fortran 10229 10230 * [287]17541 data statements with double precision constants fail 10231 10232 x86-specific 10233 10234 * [288]17853 -O2 ICE for MMX testcase 10235 10236 SPARC-specific 10237 10238 * [289]17245 ICE compiling gsl-1.5 statistics/lag1.c 10239 10240 Darwin-specific 10241 10242 * [290]17167 FATAL:Symbol L_foo$stub already defined. 10243 10244 AIX-specific 10245 10246 * [291]17277 could not catch an exception when specified -maix64 10247 10248 Solaris-specific 10249 10250 * [292]17505 <cmath> calls acosf(), ceilf(), and other functions 10251 missing from system libraries 10252 10253 HP/UX specific: 10254 10255 * [293]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 10256 10257 ARM-specific 10258 10259 * [294]17384 ICE with mode attribute on structures 10260 10261 MIPS-specific 10262 10263 * [295]17770 No NOP after LWL with -mips1 10264 10265 Other embedded target specific 10266 10267 * [296]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 10268 * [297]14064 [avr-elf] -fdata-sections triggers ICE 10269 * [298]14678 [m68hc11-elf] gcc ICE 10270 * [299]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 10271 * [300]15790 [i686-coff] Alignment error building gcc with i686-coff 10272 target 10273 * [301]15886 [SH] Miscompilation with -O2 -fPIC 10274 * [302]16884 [avr-elf] [fweb related] bug while initializing 10275 variables 10276 10277 Bugs relating to debugger support 10278 10279 * [303]13841 missing debug info for _Complex function arguments 10280 * [304]15860 [big-endian targets] No DW_AT_location debug info is 10281 emitted for formal arguments to a function that uses "register" 10282 qualifiers 10283 10284 Testsuite issues (compiler not affected) 10285 10286 * [305]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 10287 * [306]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 10288 * [307]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 10289 testsuite 10290 10291 Documentation 10292 10293 * [308]15498 typo in gcc manual: non-existing locale example en_UK, 10294 should be en_GB 10295 * [309]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 10296 document broken shell 10297 * [310]16406 USE_LD_AS_NEEDED undocumented 10298 __________________________________________________________________ 10299 10300GCC 3.4.4 10301 10302 This is the [311]list of problem reports (PRs) from GCC's bug tracking 10303 system that are known to be fixed in the 3.4.4 release. This list might 10304 not be complete (that is, it is possible that some PRs that have been 10305 fixed are not listed here). 10306 __________________________________________________________________ 10307 10308GCC 3.4.5 10309 10310 This is the [312]list of problem reports (PRs) from GCC's bug tracking 10311 system that are known to be fixed in the 3.4.5 release. This list might 10312 not be complete (that is, it is possible that some PRs that have been 10313 fixed are not listed here). 10314 10315 Bootstrap issues 10316 10317 * [313]24688 sco_math fixincl breaks math.h 10318 10319 C compiler bugs 10320 10321 * [314]17188 struct Foo { } redefinition 10322 * [315]20187 wrong code for ((unsigned char)(unsigned long 10323 long)((a?a:1)&(a*b)))?0:1) 10324 * [316]21873 infinite warning loop on bad array initializer 10325 * [317]21899 enum definition accepts values to be overriden 10326 * [318]22061 ICE in find_function_data, in function.c 10327 * [319]22308 Failure to diagnose violation of constraint 6.516p2 10328 * [320]22458 ICE on missing brace 10329 * [321]22589 ICE casting to long long 10330 * [322]24101 Segfault with preprocessed source 10331 10332 C++ compiler and library bugs 10333 10334 * [323]10611 operations on vector mode not recognized in C++ 10335 * [324]13377 unexpected behavior of namespace usage directive 10336 * [325]16002 Strange error message with new parser 10337 * [326]17413 local classes as template argument 10338 * [327]17609 spurious error message after using keyword 10339 * [328]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 10340 * [329]18124 ICE with invalid template template parameter 10341 * [330]18155 typedef in template declaration not rejected 10342 * [331]18177 ICE with const_cast for undeclared variable 10343 * [332]18368 C++ error message regression 10344 * [333]16378 ICE when returning a copy of a packed member 10345 * [334]18466 int ::i; accepted 10346 * [335]18512 ICE on invalid usage of template base class 10347 * [336]18454 ICE when returning undefined type 10348 * [337]18738 typename not allowed with non-dependent qualified name 10349 * [338]18803 rejects access to operator() in template 10350 * [339]19004 ICE in uses_template_parms, in cp/pt.c 10351 * [340]19208 Spurious error about variably modified type 10352 * [341]18253 bad error message / ICE for invalid template parameter 10353 * [342]19608 ICE after friend function definition in local class 10354 * [343]19884 ICE on explicit instantiation of a non-template 10355 constructor 10356 * [344]20153 ICE when C++ template function contains anonymous union 10357 * [345]20563 Infinite loop in diagnostic (and ice after error 10358 message) 10359 * [346]20789 ICE with incomplete type in template 10360 * [347]21336 Internal compiler error when using custom new operators 10361 * [348]21768 ICE in error message due to violation of coding 10362 conventions 10363 * [349]21853 constness of pointer to data member ignored 10364 * [350]21903 Default argument of template function causes a 10365 compile-time error 10366 * [351]21983 multiple diagnostics 10367 * [352]21987 New testsuite failure 10368 g++.dg/warn/conversion-function-1.C 10369 * [353]22153 ICE on invalid template specialization 10370 * [354]22172 Internal compiler error, seg fault. 10371 * [355]21286 filebuf::xsgetn vs pipes 10372 * [356]22233 ICE with wrong number of template parameters 10373 * [357]22508 ICE after invalid operator new 10374 * [358]22545 ICE with pointer to class member & user defined 10375 conversion operator 10376 * [359]23528 Wrong default allocator in ext/hash_map 10377 * [360]23550 char_traits requirements/1.cc test bad math 10378 * [361]23586 Bad diagnostic for invalid namespace-name 10379 * [362]23624 ICE in invert_truthvalue, in fold-const.c 10380 * [363]23639 Bad error message: not a member of '<declaration error>' 10381 * [364]23797 ICE on typename outside template 10382 * [365]23965 Bogus error message: no matching function for call to 10383 'foo(<type error>)' 10384 * [366]24052 &#`label_decl' not supported by dump_expr#<expression 10385 error> 10386 * [367]24580 virtual base class cause exception not to be caught 10387 10388 Problems in generated debug information 10389 10390 * [368]24267 Bad DWARF for altivec vectors 10391 10392 Optimizations issues 10393 10394 * [369]17810 ICE in verify_local_live_at_start 10395 * [370]17860 Wrong generated code for loop with varying bound 10396 * [371]21709 ICE on compile-time complex NaN 10397 * [372]21964 broken tail call at -O2 or more 10398 * [373]22167 Strange optimization bug when using -Os 10399 * [374]22619 Compilation failure for real_const_1.f and 10400 real_const_2.f90 10401 * [375]23241 Invalid code generated for comparison of uchar to 255 10402 * [376]23478 Miscompilation due to reloading of a var that is also 10403 used in EH pad 10404 * [377]24470 segmentation fault in cc1plus when compiling with -O 10405 * [378]24950 ICE in operand_subword_force 10406 10407 Precompiled headers problems 10408 10409 * [379]14400 Cannot compile qt-x11-free-3.3.0 10410 * [380]14940 PCH largefile test fails on various platforms 10411 10412 Preprocessor bugs 10413 10414 * [381]20239 ICE on empty preprocessed input 10415 * [382]15220 "gcc -E -MM -MG" reports missing system headers in 10416 source directory 10417 10418 Testsuite issues 10419 10420 * [383]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 10421 i686-pc-linux-gnu 10422 10423 Alpha specific 10424 10425 * [384]21888 bootstrap failure with linker relaxation enabled 10426 10427 ARM specific 10428 10429 * [385]15342 [arm-linux]: ICE in verify_local_live_at_start 10430 * [386]23985 Memory aliasing information incorrect in inlined memcpy 10431 10432 ColdFile specific 10433 10434 * [387]16719 Illegal move of byte into address register causes 10435 compiler to ICE 10436 10437 HPPA specific 10438 10439 * [388]21723 ICE while building libgfortran 10440 * [389]21841 -mhp-ld/-mgnu-ld documentation 10441 10442 IA-64 specific 10443 10444 * [390]23644 IA-64 hardware models and configuration options 10445 documentation error 10446 * [391]24718 Shared libgcc not used for linking by default 10447 10448 M68000 specific 10449 10450 * [392]18421 ICE in reload_cse_simplify_operands, in postreload.c 10451 10452 MIPS specific 10453 10454 * [393]20621 ICE in change_address_1, in emit-rtl.c 10455 10456 PowerPC and PowerPC64 specific 10457 10458 * [394]18583 error on valid code: const 10459 __attribute__((altivec(vector__))) doesn't work in arrays 10460 * [395]20191 ICE in reload_cse_simplify_operands 10461 * [396]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 10462 * [397]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 10463 * [398]23404 gij trashes args of functions with more than 8 fp args 10464 * [399]23539 C & C++ compiler generating misaligned references 10465 regardless of compiler flags 10466 * [400]24102 floatdisf2_internal2 broken 10467 * [401]24465 -mminimal-toc miscompilation of __thread vars 10468 10469 Solaris specific 10470 10471 * [402]19933 Problem with define of HUGE_VAL in math_c99 10472 * [403]21889 Native Solaris assembler cannot grok DTP-relative debug 10473 symbols 10474 10475 SPARC specific 10476 10477 * [404]19300 PCH failures on sparc-linux 10478 * [405]20301 Assembler labels have a leading "-" 10479 * [406]20673 C PCH testsuite assembly comparison failure 10480 10481 x86 and x86_64 specific 10482 10483 * [407]18582 ICE with arrays of type V2DF 10484 * [408]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 10485 -fsched2-use-traces 10486 * [409]21716 ICE in reg-stack.c's swap_rtx_condition 10487 * [410]24315 amd64 fails -fpeephole2 10488 __________________________________________________________________ 10489 10490GCC 3.4.6 10491 10492 This is the [411]list of problem reports (PRs) from GCC's bug tracking 10493 system that are known to be fixed in the 3.4.6 release. This list might 10494 not be complete (that is, it is possible that some PRs that have been 10495 fixed are not listed here). 10496 10497 10498 For questions related to the use of GCC, please consult these web 10499 pages and the [412]GCC manuals. If that fails, the 10500 [413]gcc-help@gcc.gnu.org mailing list might help. Comments on these 10501 web pages and the development of GCC are welcome on our developer 10502 list at [414]gcc@gcc.gnu.org. All of [415]our lists have public 10503 archives. 10504 10505 Copyright (C) [416]Free Software Foundation, Inc. Verbatim copying and 10506 distribution of this entire article is permitted in any medium, 10507 provided this notice is preserved. 10508 10509 These pages are [417]maintained by the GCC team. Last modified 10510 2016-01-30[418]. 10511 10512References 10513 10514 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 10515 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 10516 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 10517 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 10518 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 10519 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 10520 7. http://www.boost.org/ 10521 8. https://gcc.gnu.org/PR11953 10522 9. https://gcc.gnu.org/PR8361 10523 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins 10524 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 10525 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 10526 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 10527 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 10528 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 10529 16. http://www.gnu.org/software/classpath/ 10530 17. http://www.eclipse.org/ 10531 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 10532 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 10533 20. http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51A_HTML/ARH9MBTE/DTMNPLTN.HTM#normal-argument-list-structure 10534 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 10535 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 10536 23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 10537 24. http://gcc.gnu.org/gcc-3.4/mips-abi.html 10538 25. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 10539 26. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 10540 27. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=%5C%5B3%5C.4.*%5BRr%5Degression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 10541 28. https://gcc.gnu.org/PR10129 10542 29. https://gcc.gnu.org/PR14576 10543 30. https://gcc.gnu.org/PR14760 10544 31. https://gcc.gnu.org/PR14671 10545 32. https://gcc.gnu.org/PR15093 10546 33. https://gcc.gnu.org/PR15178 10547 34. https://gcc.gnu.org/PR12753 10548 35. https://gcc.gnu.org/PR13985 10549 36. https://gcc.gnu.org/PR14810 10550 37. https://gcc.gnu.org/PR14883 10551 38. https://gcc.gnu.org/PR15044 10552 39. https://gcc.gnu.org/PR15057 10553 40. https://gcc.gnu.org/PR15064 10554 41. https://gcc.gnu.org/PR15142 10555 42. https://gcc.gnu.org/PR15159 10556 43. https://gcc.gnu.org/PR15165 10557 44. https://gcc.gnu.org/PR15193 10558 45. https://gcc.gnu.org/PR15209 10559 46. https://gcc.gnu.org/PR15227 10560 47. https://gcc.gnu.org/PR15285 10561 48. https://gcc.gnu.org/PR15299 10562 49. https://gcc.gnu.org/PR15329 10563 50. https://gcc.gnu.org/PR15550 10564 51. https://gcc.gnu.org/PR15554 10565 52. https://gcc.gnu.org/PR15640 10566 53. https://gcc.gnu.org/PR15666 10567 54. https://gcc.gnu.org/PR15696 10568 55. https://gcc.gnu.org/PR15701 10569 56. https://gcc.gnu.org/PR15761 10570 57. https://gcc.gnu.org/PR15829 10571 58. https://gcc.gnu.org/PR14538 10572 59. https://gcc.gnu.org/PR12391 10573 60. https://gcc.gnu.org/PR14649 10574 61. https://gcc.gnu.org/PR15004 10575 62. https://gcc.gnu.org/PR15749 10576 63. https://gcc.gnu.org/PR10646 10577 64. https://gcc.gnu.org/PR12077 10578 65. https://gcc.gnu.org/PR13598 10579 66. https://gcc.gnu.org/PR14211 10580 67. https://gcc.gnu.org/PR14220 10581 68. https://gcc.gnu.org/PR14245 10582 69. https://gcc.gnu.org/PR14340 10583 70. https://gcc.gnu.org/PR14600 10584 71. https://gcc.gnu.org/PR14668 10585 72. https://gcc.gnu.org/PR14775 10586 73. https://gcc.gnu.org/PR14821 10587 74. https://gcc.gnu.org/PR14930 10588 75. https://gcc.gnu.org/PR14932 10589 76. https://gcc.gnu.org/PR14950 10590 77. https://gcc.gnu.org/PR14962 10591 78. https://gcc.gnu.org/PR14975 10592 79. https://gcc.gnu.org/PR15002 10593 80. https://gcc.gnu.org/PR15025 10594 81. https://gcc.gnu.org/PR15046 10595 82. https://gcc.gnu.org/PR15069 10596 83. https://gcc.gnu.org/PR15074 10597 84. https://gcc.gnu.org/PR15083 10598 85. https://gcc.gnu.org/PR15096 10599 86. https://gcc.gnu.org/PR15287 10600 87. https://gcc.gnu.org/PR15317 10601 88. https://gcc.gnu.org/PR15337 10602 89. https://gcc.gnu.org/PR15361 10603 90. https://gcc.gnu.org/PR15412 10604 91. https://gcc.gnu.org/PR15427 10605 92. https://gcc.gnu.org/PR15471 10606 93. https://gcc.gnu.org/PR15503 10607 94. https://gcc.gnu.org/PR15507 10608 95. https://gcc.gnu.org/PR15542 10609 96. https://gcc.gnu.org/PR15565 10610 97. https://gcc.gnu.org/PR15625 10611 98. https://gcc.gnu.org/PR15629 10612 99. https://gcc.gnu.org/PR15742 10613 100. https://gcc.gnu.org/PR15775 10614 101. https://gcc.gnu.org/PR15821 10615 102. https://gcc.gnu.org/PR15862 10616 103. https://gcc.gnu.org/PR15875 10617 104. https://gcc.gnu.org/PR15877 10618 105. https://gcc.gnu.org/PR15947 10619 106. https://gcc.gnu.org/PR16020 10620 107. https://gcc.gnu.org/PR16154 10621 108. https://gcc.gnu.org/PR16174 10622 109. https://gcc.gnu.org/PR14315 10623 110. https://gcc.gnu.org/PR15151 10624 111. https://gcc.gnu.org/PR7993 10625 112. https://gcc.gnu.org/PR15228 10626 113. https://gcc.gnu.org/PR15345 10627 114. https://gcc.gnu.org/PR15945 10628 115. https://gcc.gnu.org/PR15526 10629 116. https://gcc.gnu.org/PR14690 10630 117. https://gcc.gnu.org/PR15112 10631 118. https://gcc.gnu.org/PR15067 10632 119. https://gcc.gnu.org/PR1963 10633 120. https://gcc.gnu.org/PR15717 10634 121. https://gcc.gnu.org/PR14782 10635 122. https://gcc.gnu.org/PR14828 10636 123. https://gcc.gnu.org/PR15202 10637 124. https://gcc.gnu.org/PR14610 10638 125. https://gcc.gnu.org/PR14813 10639 126. https://gcc.gnu.org/PR14857 10640 127. https://gcc.gnu.org/PR15598 10641 128. https://gcc.gnu.org/PR15653 10642 129. https://gcc.gnu.org/PR15189 10643 130. https://gcc.gnu.org/PR15331 10644 131. https://gcc.gnu.org/PR16144 10645 132. https://gcc.gnu.org/PR16176 10646 133. https://gcc.gnu.org/PR11591 10647 134. https://gcc.gnu.org/PR12028 10648 135. https://gcc.gnu.org/PR14478 10649 136. https://gcc.gnu.org/PR14567 10650 137. https://gcc.gnu.org/PR14715 10651 138. https://gcc.gnu.org/PR14902 10652 139. https://gcc.gnu.org/PR14924 10653 140. https://gcc.gnu.org/PR14960 10654 141. https://gcc.gnu.org/PR15106 10655 142. https://gcc.gnu.org/PR16026 10656 143. https://gcc.gnu.org/PR15191 10657 144. https://gcc.gnu.org/PR15662 10658 145. https://gcc.gnu.org/PR15054 10659 146. https://gcc.gnu.org/PR15783 10660 147. https://gcc.gnu.org/PR15626 10661 148. https://gcc.gnu.org/PR14326 10662 149. https://gcc.gnu.org/PR14723 10663 150. https://gcc.gnu.org/PR15290 10664 151. https://gcc.gnu.org/PR15250 10665 152. https://gcc.gnu.org/PR15551 10666 153. https://gcc.gnu.org/PR8309 10667 154. https://gcc.gnu.org/PR13250 10668 155. https://gcc.gnu.org/PR13803 10669 156. https://gcc.gnu.org/PR14093 10670 157. https://gcc.gnu.org/PR14457 10671 158. https://gcc.gnu.org/PR14542 10672 159. https://gcc.gnu.org/PR15100 10673 160. https://gcc.gnu.org/PR15296 10674 161. https://gcc.gnu.org/PR15396 10675 162. https://gcc.gnu.org/PR15782 10676 163. https://gcc.gnu.org/PR11610 10677 164. https://gcc.gnu.org/PR15488 10678 165. https://gcc.gnu.org/PR15489 10679 166. https://gcc.gnu.org/PR13928 10680 167. https://gcc.gnu.org/PR14150 10681 168. https://gcc.gnu.org/PR14949 10682 169. https://gcc.gnu.org/PR15123 10683 170. https://gcc.gnu.org/PR16469 10684 171. https://gcc.gnu.org/PR16344 10685 172. https://gcc.gnu.org/PR16842 10686 173. https://gcc.gnu.org/PR12608 10687 174. https://gcc.gnu.org/PR14492 10688 175. https://gcc.gnu.org/PR15461 10689 176. https://gcc.gnu.org/PR15890 10690 177. https://gcc.gnu.org/PR16180 10691 178. https://gcc.gnu.org/PR16224 10692 179. https://gcc.gnu.org/PR16408 10693 180. https://gcc.gnu.org/PR16529 10694 181. https://gcc.gnu.org/PR16698 10695 182. https://gcc.gnu.org/PR16706 10696 183. https://gcc.gnu.org/PR16810 10697 184. https://gcc.gnu.org/PR16851 10698 185. https://gcc.gnu.org/PR16870 10699 186. https://gcc.gnu.org/PR16904 10700 187. https://gcc.gnu.org/PR16905 10701 188. https://gcc.gnu.org/PR16964 10702 189. https://gcc.gnu.org/PR17068 10703 190. https://gcc.gnu.org/PR16366 10704 191. https://gcc.gnu.org/PR15345 10705 192. https://gcc.gnu.org/PR16590 10706 193. https://gcc.gnu.org/PR16693 10707 194. https://gcc.gnu.org/PR17078 10708 195. https://gcc.gnu.org/PR13956 10709 196. https://gcc.gnu.org/PR16684 10710 197. https://gcc.gnu.org/PR12658 10711 198. https://gcc.gnu.org/PR13092 10712 199. https://gcc.gnu.org/PR15320 10713 200. https://gcc.gnu.org/PR16246 10714 201. https://gcc.gnu.org/PR16273 10715 202. https://gcc.gnu.org/PR16401 10716 203. https://gcc.gnu.org/PR16411 10717 204. https://gcc.gnu.org/PR16489 10718 205. https://gcc.gnu.org/PR16618 10719 206. https://gcc.gnu.org/PR16637 10720 207. https://gcc.gnu.org/PR16717 10721 208. https://gcc.gnu.org/PR16813 10722 209. https://gcc.gnu.org/PR16853 10723 210. https://gcc.gnu.org/PR16889 10724 211. https://gcc.gnu.org/PR16959 10725 212. https://gcc.gnu.org/PR7587 10726 213. https://gcc.gnu.org/PR16473 10727 214. https://gcc.gnu.org/PR16478 10728 215. https://gcc.gnu.org/PR10695 10729 216. https://gcc.gnu.org/PR16974 10730 217. https://gcc.gnu.org/PR16298 10731 218. https://gcc.gnu.org/PR17113 10732 219. https://gcc.gnu.org/PR14697 10733 220. https://gcc.gnu.org/PR15869 10734 221. https://gcc.gnu.org/PR16325 10735 222. https://gcc.gnu.org/PR16357 10736 223. https://gcc.gnu.org/PR16380 10737 224. https://gcc.gnu.org/PR16407 10738 225. https://gcc.gnu.org/PR16643 10739 226. https://gcc.gnu.org/PR15927 10740 227. https://gcc.gnu.org/PR15948 10741 228. https://gcc.gnu.org/PR17019 10742 229. https://gcc.gnu.org/PR16130 10743 230. https://gcc.gnu.org/PR16142 10744 231. https://gcc.gnu.org/PR16278 10745 232. https://gcc.gnu.org/PR16414 10746 233. https://gcc.gnu.org/PR16445 10747 234. https://gcc.gnu.org/PR16490 10748 235. https://gcc.gnu.org/PR16683 10749 236. https://gcc.gnu.org/PR16195 10750 237. https://gcc.gnu.org/PR16239 10751 238. https://gcc.gnu.org/PR16199 10752 239. https://gcc.gnu.org/PR16416 10753 240. https://gcc.gnu.org/PR16430 10754 241. https://gcc.gnu.org/PR16379 10755 242. https://gcc.gnu.org/PR17093 10756 243. https://gcc.gnu.org/PR17119 10757 244. https://gcc.gnu.org/PR15928 10758 245. https://gcc.gnu.org/PR16210 10759 246. https://gcc.gnu.org/PR15488 10760 247. https://gcc.gnu.org/PR16250 10761 248. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 10762 249. https://gcc.gnu.org/PR17369 10763 250. https://gcc.gnu.org/PR17850 10764 251. https://gcc.gnu.org/PR13948 10765 252. https://gcc.gnu.org/PR14492 10766 253. https://gcc.gnu.org/PR16301 10767 254. https://gcc.gnu.org/PR16566 10768 255. https://gcc.gnu.org/PR17023 10769 256. https://gcc.gnu.org/PR17027 10770 257. https://gcc.gnu.org/PR17524 10771 258. https://gcc.gnu.org/PR17826 10772 259. https://gcc.gnu.org/PR15526 10773 260. https://gcc.gnu.org/PR16999 10774 261. https://gcc.gnu.org/PR17503 10775 262. https://gcc.gnu.org/PR17581 10776 263. https://gcc.gnu.org/PR18129 10777 264. https://gcc.gnu.org/PR10975 10778 265. https://gcc.gnu.org/PR11722 10779 266. https://gcc.gnu.org/PR14534 10780 267. https://gcc.gnu.org/PR15172 10781 268. https://gcc.gnu.org/PR15786 10782 269. https://gcc.gnu.org/PR16162 10783 270. https://gcc.gnu.org/PR16612 10784 271. https://gcc.gnu.org/PR16715 10785 272. https://gcc.gnu.org/PR16848 10786 273. https://gcc.gnu.org/PR17132 10787 274. https://gcc.gnu.org/PR17259 10788 275. https://gcc.gnu.org/PR17327 10789 276. https://gcc.gnu.org/PR17393 10790 277. https://gcc.gnu.org/PR17501 10791 278. https://gcc.gnu.org/PR17537 10792 279. https://gcc.gnu.org/PR17585 10793 280. https://gcc.gnu.org/PR17821 10794 281. https://gcc.gnu.org/PR17829 10795 282. https://gcc.gnu.org/PR17851 10796 283. https://gcc.gnu.org/PR17976 10797 284. https://gcc.gnu.org/PR18020 10798 285. https://gcc.gnu.org/PR18093 10799 286. https://gcc.gnu.org/PR18140 10800 287. https://gcc.gnu.org/PR17541 10801 288. https://gcc.gnu.org/PR17853 10802 289. https://gcc.gnu.org/PR17245 10803 290. https://gcc.gnu.org/PR17167 10804 291. https://gcc.gnu.org/PR17277 10805 292. https://gcc.gnu.org/PR17505 10806 293. https://gcc.gnu.org/PR17684 10807 294. https://gcc.gnu.org/PR17384 10808 295. https://gcc.gnu.org/PR17770 10809 296. https://gcc.gnu.org/PR11476 10810 297. https://gcc.gnu.org/PR14064 10811 298. https://gcc.gnu.org/PR14678 10812 299. https://gcc.gnu.org/PR15583 10813 300. https://gcc.gnu.org/PR15790 10814 301. https://gcc.gnu.org/PR15886 10815 302. https://gcc.gnu.org/PR16884 10816 303. https://gcc.gnu.org/PR13841 10817 304. https://gcc.gnu.org/PR15860 10818 305. https://gcc.gnu.org/PR17465 10819 306. https://gcc.gnu.org/PR17469 10820 307. https://gcc.gnu.org/PR18138 10821 308. https://gcc.gnu.org/PR15498 10822 309. https://gcc.gnu.org/PR15747 10823 310. https://gcc.gnu.org/PR16406 10824 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 10825 312. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 10826 313. https://gcc.gnu.org/PR24688 10827 314. https://gcc.gnu.org/PR17188 10828 315. https://gcc.gnu.org/PR20187 10829 316. https://gcc.gnu.org/PR21873 10830 317. https://gcc.gnu.org/PR21899 10831 318. https://gcc.gnu.org/PR22061 10832 319. https://gcc.gnu.org/PR22208 10833 320. https://gcc.gnu.org/PR22458 10834 321. https://gcc.gnu.org/PR22589 10835 322. https://gcc.gnu.org/PR24101 10836 323. https://gcc.gnu.org/PR10611 10837 324. https://gcc.gnu.org/PR13377 10838 325. https://gcc.gnu.org/PR16002 10839 326. https://gcc.gnu.org/PR17413 10840 327. https://gcc.gnu.org/PR17609 10841 328. https://gcc.gnu.org/PR17618 10842 329. https://gcc.gnu.org/PR18124 10843 330. https://gcc.gnu.org/PR18155 10844 331. https://gcc.gnu.org/PR18177 10845 332. https://gcc.gnu.org/PR18368 10846 333. https://gcc.gnu.org/PR18378 10847 334. https://gcc.gnu.org/PR18466 10848 335. https://gcc.gnu.org/PR18512 10849 336. https://gcc.gnu.org/PR18545 10850 337. https://gcc.gnu.org/PR18738 10851 338. https://gcc.gnu.org/PR18803 10852 339. https://gcc.gnu.org/PR19004 10853 340. https://gcc.gnu.org/PR19208 10854 341. https://gcc.gnu.org/PR19253 10855 342. https://gcc.gnu.org/PR19608 10856 343. https://gcc.gnu.org/PR19884 10857 344. https://gcc.gnu.org/PR20153 10858 345. https://gcc.gnu.org/PR20563 10859 346. https://gcc.gnu.org/PR20789 10860 347. https://gcc.gnu.org/PR21336 10861 348. https://gcc.gnu.org/PR21768 10862 349. https://gcc.gnu.org/PR21853 10863 350. https://gcc.gnu.org/PR21903 10864 351. https://gcc.gnu.org/PR21983 10865 352. https://gcc.gnu.org/PR21987 10866 353. https://gcc.gnu.org/PR22153 10867 354. https://gcc.gnu.org/PR22172 10868 355. https://gcc.gnu.org/PR21286 10869 356. https://gcc.gnu.org/PR22233 10870 357. https://gcc.gnu.org/PR22508 10871 358. https://gcc.gnu.org/PR22545 10872 359. https://gcc.gnu.org/PR23528 10873 360. https://gcc.gnu.org/PR23550 10874 361. https://gcc.gnu.org/PR23586 10875 362. https://gcc.gnu.org/PR23624 10876 363. https://gcc.gnu.org/PR23639 10877 364. https://gcc.gnu.org/PR23797 10878 365. https://gcc.gnu.org/PR23965 10879 366. https://gcc.gnu.org/PR24052 10880 367. https://gcc.gnu.org/PR24580 10881 368. https://gcc.gnu.org/PR24267 10882 369. https://gcc.gnu.org/PR17810 10883 370. https://gcc.gnu.org/PR17860 10884 371. https://gcc.gnu.org/PR21709 10885 372. https://gcc.gnu.org/PR21964 10886 373. https://gcc.gnu.org/PR22167 10887 374. https://gcc.gnu.org/PR22619 10888 375. https://gcc.gnu.org/PR23241 10889 376. https://gcc.gnu.org/PR23478 10890 377. https://gcc.gnu.org/PR24470 10891 378. https://gcc.gnu.org/PR24950 10892 379. https://gcc.gnu.org/PR14400 10893 380. https://gcc.gnu.org/PR14940 10894 381. https://gcc.gnu.org/PR20239 10895 382. https://gcc.gnu.org/PR15220 10896 383. https://gcc.gnu.org/PR19275 10897 384. https://gcc.gnu.org/PR21888 10898 385. https://gcc.gnu.org/PR15342 10899 386. https://gcc.gnu.org/PR23985 10900 387. https://gcc.gnu.org/PR16719 10901 388. https://gcc.gnu.org/PR21723 10902 389. https://gcc.gnu.org/PR21841 10903 390. https://gcc.gnu.org/PR23644 10904 391. https://gcc.gnu.org/PR24718 10905 392. https://gcc.gnu.org/PR18421 10906 393. https://gcc.gnu.org/PR20621 10907 394. https://gcc.gnu.org/PR18583 10908 395. https://gcc.gnu.org/PR20191 10909 396. https://gcc.gnu.org/PR22083 10910 397. https://gcc.gnu.org/PR23070 10911 398. https://gcc.gnu.org/PR23404 10912 399. https://gcc.gnu.org/PR23539 10913 400. https://gcc.gnu.org/PR24102 10914 401. https://gcc.gnu.org/PR24465 10915 402. https://gcc.gnu.org/PR19933 10916 403. https://gcc.gnu.org/PR21889 10917 404. https://gcc.gnu.org/PR19300 10918 405. https://gcc.gnu.org/PR20301 10919 406. https://gcc.gnu.org/PR20673 10920 407. https://gcc.gnu.org/PR18582 10921 408. https://gcc.gnu.org/PR19340 10922 409. https://gcc.gnu.org/PR21716 10923 410. https://gcc.gnu.org/PR24315 10924 411. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 10925 412. https://gcc.gnu.org/onlinedocs/ 10926 413. mailto:gcc-help@gcc.gnu.org 10927 414. mailto:gcc@gcc.gnu.org 10928 415. https://gcc.gnu.org/lists.html 10929 416. http://www.fsf.org/ 10930 417. https://gcc.gnu.org/about.html 10931 418. http://validator.w3.org/check/referer 10932====================================================================== 10933http://gcc.gnu.org/gcc-3.3/index.html 10934 10935 GCC 3.3 Release Series 10936 10937 May 03, 2005 10938 10939 The [1]GNU project and the GCC developers are pleased to announce the 10940 release of GCC 3.3.6. 10941 10942 This release is a bug-fix release, containing fixes for regressions in 10943 GCC 3.3.5 relative to previous releases of GCC. 10944 10945 This release is the last of the series 3.3.x. 10946 10947 The GCC 3.3 release series includes numerous [2]new features, 10948 improvements, bug fixes, and other changes, thanks to an [3]amazing 10949 group of volunteers. 10950 10951Release History 10952 10953 GCC 3.3.6 10954 May 3, 2005 ([4]changes) 10955 10956 GCC 3.3.5 10957 September 30, 2004 ([5]changes) 10958 10959 GCC 3.3.4 10960 May 31, 2004 ([6]changes) 10961 10962 GCC 3.3.3 10963 February 14, 2004 ([7]changes) 10964 10965 GCC 3.3.2 10966 October 16, 2003 ([8]changes) 10967 10968 GCC 3.3.1 10969 August 8, 2003 ([9]changes) 10970 10971 GCC 3.3 10972 May 14, 2003 ([10]changes) 10973 10974References and Acknowledgements 10975 10976 GCC used to stand for the GNU C Compiler, but since the compiler 10977 supports several other languages aside from C, it now stands for the 10978 GNU Compiler Collection. 10979 10980 A list of [11]successful builds is updated as new information becomes 10981 available. 10982 10983 The GCC developers would like to thank the numerous people that have 10984 contributed new features, improvements, bug fixes, and other changes as 10985 well as test results to GCC. This [12]amazing group of volunteers is 10986 what makes GCC successful. 10987 10988 For additional information about GCC please refer to the [13]GCC 10989 project web site or contact the [14]GCC development mailing list. 10990 10991 To obtain GCC please use [15]our mirror sites, or our CVS server. 10992 10993 10994 For questions related to the use of GCC, please consult these web 10995 pages and the [16]GCC manuals. If that fails, the 10996 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 10997 web pages and the development of GCC are welcome on our developer 10998 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 10999 archives. 11000 11001 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 11002 distribution of this entire article is permitted in any medium, 11003 provided this notice is preserved. 11004 11005 These pages are [21]maintained by the GCC team. Last modified 11006 2016-01-30[22]. 11007 11008References 11009 11010 1. http://www.gnu.org/ 11011 2. http://gcc.gnu.org/gcc-3.3/changes.html 11012 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11013 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 11014 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 11015 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 11016 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 11017 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 11018 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 11019 10. http://gcc.gnu.org/gcc-3.3/changes.html 11020 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 11021 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11022 13. http://gcc.gnu.org/index.html 11023 14. mailto:gcc@gcc.gnu.org 11024 15. http://gcc.gnu.org/mirrors.html 11025 16. https://gcc.gnu.org/onlinedocs/ 11026 17. mailto:gcc-help@gcc.gnu.org 11027 18. mailto:gcc@gcc.gnu.org 11028 19. https://gcc.gnu.org/lists.html 11029 20. http://www.fsf.org/ 11030 21. https://gcc.gnu.org/about.html 11031 22. http://validator.w3.org/check/referer 11032====================================================================== 11033http://gcc.gnu.org/gcc-3.3/changes.html 11034 11035 GCC 3.3 Release Series 11036 Changes, New Features, and Fixes 11037 11038 The latest release in the 3.3 release series is [1]GCC 3.3.6. 11039 11040Caveats 11041 11042 * The preprocessor no longer accepts multi-line string literals. They 11043 were deprecated in 3.0, 3.1, and 3.2. 11044 * The preprocessor no longer supports the -A- switch when appearing 11045 alone. -A- followed by an assertion is still supported. 11046 * Support for all the systems [2]obsoleted in GCC 3.1 has been 11047 removed from GCC 3.3. See below for a [3]list of systems which are 11048 obsoleted in this release. 11049 * Checking for null format arguments has been decoupled from the rest 11050 of the format checking mechanism. Programs which use the format 11051 attribute may regain this functionality by using the new [4]nonnull 11052 function attribute. Note that all functions for which GCC has a 11053 built-in format attribute, an appropriate built-in nonnull 11054 attribute is also applied. 11055 * The DWARF (version 1) debugging format has been deprecated and will 11056 be removed in a future version of GCC. Version 2 of the DWARF 11057 debugging format will continue to be supported for the foreseeable 11058 future. 11059 * The C and Objective-C compilers no longer accept the "Naming Types" 11060 extension (typedef foo = bar); it was already unavailable in C++. 11061 Code which uses it will need to be changed to use the "typeof" 11062 extension instead: typedef typeof(bar) foo. (We have removed this 11063 extension without a period of deprecation because it has caused the 11064 compiler to crash since version 3.0 and no one noticed until very 11065 recently. Thus we conclude it is not in widespread use.) 11066 * The -traditional C compiler option has been removed. It was 11067 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 11068 available.) The <varargs.h> header, used for writing variadic 11069 functions in traditional C, still exists but will produce an error 11070 message if used. 11071 * GCC 3.3.1 automatically places zero-initialized variables in the 11072 .bss section on some operating systems. Versions of GNU Emacs up to 11073 (and including) 21.3 will not work correctly when using this 11074 optimization; you can use -fno-zero-initialized-in-bss to disable 11075 it. 11076 11077General Optimizer Improvements 11078 11079 * A new scheme for accurately describing processor pipelines, the 11080 [5]DFA scheduler, has been added. 11081 * Pavel Nejedly, Charles University Prague, has contributed new file 11082 format used by the edge coverage profiler (-fprofile-arcs). 11083 The new format is robust and diagnoses common mistakes where 11084 profiles from different versions (or compilations) of the program 11085 are combined resulting in nonsensical profiles and slow code to 11086 produced with profile feedback. Additionally this format allows 11087 extra data to be gathered. Currently, overall statistics are 11088 produced helping optimizers to identify hot spots of a program 11089 globally replacing the old intra-procedural scheme and resulting in 11090 better code. Note that the gcov tool from older GCC versions will 11091 not be able to parse the profiles generated by GCC 3.3 and vice 11092 versa. 11093 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 11094 pass enabled using -ftracer. This pass simplifies the control flow 11095 of functions allowing other optimizations to do better job. 11096 He also contributed the function reordering pass 11097 (-freorder-functions) to optimize function placement using profile 11098 feedback. 11099 11100New Languages and Language specific improvements 11101 11102 C/ObjC/C++ 11103 11104 * The preprocessor now accepts directives within macro arguments. It 11105 processes them just as if they had not been within macro arguments. 11106 * The separate ISO and traditional preprocessors have been completely 11107 removed. The front end handles either type of preprocessed output 11108 if necessary. 11109 * In C99 mode preprocessor arithmetic is done in the precision of the 11110 target's intmax_t, as required by that standard. 11111 * The preprocessor can now copy comments inside macros to the output 11112 file when the macro is expanded. This feature, enabled using the 11113 -CC option, is intended for use by applications which place 11114 metadata or directives inside comments, such as lint. 11115 * The method of constructing the list of directories to be searched 11116 for header files has been revised. If a directory named by a -I 11117 option is a standard system include directory, the option is 11118 ignored to ensure that the default search order for system 11119 directories and the special treatment of system header files are 11120 not defeated. 11121 * A few more [6]ISO C99 features now work correctly. 11122 * A new function attribute, nonnull, has been added which allows 11123 pointer arguments to functions to be specified as requiring a 11124 non-null value. The compiler currently uses this information to 11125 issue a warning when it detects a null value passed in such an 11126 argument slot. 11127 * A new type attribute, may_alias, has been added. Accesses to 11128 objects with types with this attribute are not subjected to 11129 type-based alias analysis, but are instead assumed to be able to 11130 alias any other type of objects, just like the char type. 11131 11132 C++ 11133 11134 * Type based alias analysis has been implemented for C++ aggregate 11135 types. 11136 11137 Objective-C 11138 11139 * Generate an error if Objective-C objects are passed by value in 11140 function and method calls. 11141 * When -Wselector is used, check the whole list of selectors at the 11142 end of compilation, and emit a warning if a @selector() is not 11143 known. 11144 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 11145 * No longer need to include objc/objc-class.h to compile self calls 11146 in class methods (NeXT runtime only). 11147 * New -Wundeclared-selector option. 11148 * Removed selector bloating which was causing object files to be 10% 11149 bigger on average (GNU runtime only). 11150 * Using at run time @protocol() objects has been fixed in certain 11151 situations (GNU runtime only). 11152 * Type checking has been fixed and improved in many situations 11153 involving protocols. 11154 11155 Java 11156 11157 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 11158 1.4) API. 11159 * The JDK 1.4 assert facility has been implemented. 11160 * The bytecode interpreter is now direct threaded and thus faster. 11161 11162 Fortran 11163 11164 * Fortran improvements are listed in [7]the Fortran documentation. 11165 11166 Ada 11167 11168 * Ada tasking now works with glibc 2.3.x threading libraries. 11169 11170New Targets and Target Specific Improvements 11171 11172 * The following changes have been made to the HP-PA port: 11173 + The port now defaults to scheduling for the PA8000 series of 11174 processors. 11175 + Scheduling support for the PA7300 processor has been added. 11176 + The 32-bit port now supports weak symbols under HP-UX 11. 11177 + The handling of initializers and finalizers has been improved 11178 under HP-UX 11. The 64-bit port no longer uses collect2. 11179 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 11180 + ABI fixes to correct the passing of small structures by value. 11181 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 11182 use the DFA processor pipeline description. 11183 * The following NetBSD configurations for the SuperH processor family 11184 have been added: 11185 + SH3, big-endian, sh-*-netbsdelf* 11186 + SH3, little-endian, shle-*-netbsdelf* 11187 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 11188 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 11189 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 11190 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 11191 * The following changes have been made to the IA-32/x86-64 port: 11192 + SSE2 and 3dNOW! intrinsics are now supported. 11193 + Support for thread local storage has been added to the IA-32 11194 and x86-64 ports. 11195 + The x86-64 port has been significantly improved. 11196 * The following changes have been made to the MIPS port: 11197 + All configurations now accept the -mabi switch. Note that you 11198 will need appropriate multilibs for this option to work 11199 properly. 11200 + ELF configurations will always pass an ABI flag to the 11201 assembler, except when the MIPS EABI is selected. 11202 + -mabi=64 no longer selects MIPS IV code. 11203 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 11204 been removed from this release. 11205 + -march now changes the core ISA level. In previous releases, 11206 it would change the use of processor-specific extensions, but 11207 would leave the core ISA unchanged. For example, mips64-elf 11208 -march=r8000 will now generate MIPS IV code. 11209 + Under most configurations, -mipsN now acts as a synonym for 11210 -march. 11211 + There are some new preprocessor macros to describe the -march 11212 and -mtune settings. See the documentation of those options 11213 for details. 11214 + Support for the NEC VR-Series processors has been added. This 11215 includes the 54xx, 5500, and 41xx series. 11216 + Support for the Sandcraft sr71k processor has been added. 11217 * The following changes have been made to the S/390 port: 11218 + Support to build the Java runtime libraries has been added. 11219 Java is now enabled by default on s390-*-linux* and 11220 s390x-*-linux* targets. 11221 + Multilib support for the s390x-*-linux* target has been added; 11222 this allows to build 31-bit binaries using the -m31 option. 11223 + Support for thread local storage has been added. 11224 + Inline assembler code may now use the 'Q' constraint to 11225 specify memory operands without index register. 11226 + Various platform-specific performance improvements have been 11227 implemented; in particular, the compiler now uses the BRANCH 11228 ON COUNT family of instructions and makes more frequent use of 11229 the TEST UNDER MASK family of instructions. 11230 * The following changes have been made to the PowerPC port: 11231 + Support for IBM Power4 processor added. 11232 + Support for Motorola e500 SPE added. 11233 + Support for AIX 5.2 added. 11234 + Function and Data sections now supported on AIX. 11235 + Sibcall optimizations added. 11236 * The support for H8 Tiny is added to the H8/300 port with -mn. 11237 11238Obsolete Systems 11239 11240 Support for a number of older systems has been declared obsolete in GCC 11241 3.3. Unless there is activity to revive them, the next release of GCC 11242 will have their sources permanently removed. 11243 11244 All configurations of the following processor architectures have been 11245 declared obsolete: 11246 * Matsushita MN10200, mn10200-*-* 11247 * Motorola 88000, m88k-*-* 11248 * IBM ROMP, romp-*-* 11249 11250 Also, some individual systems have been obsoleted: 11251 * Alpha 11252 + Interix, alpha*-*-interix* 11253 + Linux libc1, alpha*-*-linux*libc1* 11254 + Linux ECOFF, alpha*-*-linux*ecoff* 11255 * ARM 11256 + Generic a.out, arm*-*-aout* 11257 + Conix, arm*-*-conix* 11258 + "Old ABI," arm*-*-oabi 11259 + StrongARM/COFF, strongarm-*-coff* 11260 * HPPA (PA-RISC) 11261 + Generic OSF, hppa1.0-*-osf* 11262 + Generic BSD, hppa1.0-*-bsd* 11263 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 11264 + HiUX, hppa*-*-hiux* 11265 + Mach Lites, hppa*-*-lites* 11266 * Intel 386 family 11267 + Windows NT 3.x, i?86-*-win32 11268 * MC68000 family 11269 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 11270 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 11271 m68k-sun-mach* 11272 + AT&T systems, m68000-att-sysv* 11273 + Atari systems, m68k-atari-sysv* 11274 + Motorola systems, m68k-motorola-sysv* 11275 + NCR systems, m68k-ncr-sysv* 11276 + Plexus systems, m68k-plexus-sysv* 11277 + Commodore systems, m68k-cbm-sysv* 11278 + Citicorp TTI, m68k-tti-* 11279 + Unos, m68k-crds-unos* 11280 + Concurrent RTU, m68k-ccur-rtu* 11281 + Linux a.out, m68k-*-linux*aout* 11282 + Linux libc1, m68k-*-linux*libc1* 11283 + pSOS, m68k-*-psos* 11284 * MIPS 11285 + Generic ECOFF, mips*-*-ecoff* 11286 + SINIX, mips-sni-sysv4 11287 + Orion RTEMS, mips64orion-*-rtems* 11288 * National Semiconductor 32000 11289 + OpenBSD, ns32k-*-openbsd* 11290 * POWER (aka RS/6000) and PowerPC 11291 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 11292 + Bull BOSX, rs6000-bull-bosx 11293 + Generic Mach, rs6000-*-mach* 11294 + Generic SysV, powerpc*-*-sysv* 11295 + Linux libc1, powerpc*-*-linux*libc1* 11296 * Sun SPARC 11297 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 11298 sparclite-*-aout*, and sparc86x-*-aout* 11299 + NetBSD a.out, sparc-*-netbsd*aout* 11300 + Generic BSD, sparc-*-bsd* 11301 + ChorusOS, sparc-*-chorusos* 11302 + Linux a.out, sparc-*-linux*aout* 11303 + Linux libc1, sparc-*-linux*libc1* 11304 + LynxOS, sparc-*-lynxos* 11305 + Solaris on HAL hardware, sparc-hal-solaris2* 11306 + SunOS versions 3 and 4, sparc-*-sunos[34]* 11307 * NEC V850 11308 + RTEMS, v850-*-rtems* 11309 * VAX 11310 + VMS, vax-*-vms* 11311 11312Documentation improvements 11313 11314Other significant improvements 11315 11316 * Almost all front-end dependencies in the compiler have been 11317 separated out into a set of language hooks. This should make adding 11318 a new front end clearer and easier. 11319 * One effect of removing the separate preprocessor is a small 11320 increase in the robustness of the compiler in general, and the 11321 maintainability of target descriptions. Previously target-specific 11322 built-in macros and others, such as __FAST_MATH__, had to be 11323 handled with so-called specs that were hard to maintain. Often they 11324 would fail to behave properly when conflicting options were 11325 supplied on the command line, and define macros in the user's 11326 namespace even when strict ISO compliance was requested. 11327 Integrating the preprocessor has cleanly solved these issues. 11328 * The Makefile suite now supports redirection of make install by 11329 means of the variable DESTDIR. 11330 __________________________________________________________________ 11331 11332GCC 3.3 11333 11334 Detailed release notes for the GCC 3.3 release follow. 11335 11336 Bug Fixes 11337 11338 bootstrap failures 11339 11340 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 11341 [9]10198,[10]10338) 11342 11343 Internal compiler errors (multi-platform) 11344 11345 * [11]3581 large string causes segmentation fault in cc1 11346 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 11347 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 11348 init, invalid_op) 11349 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 11350 * [15]6412 (c++) ICE in retrieve_specialization 11351 * [16]6620 (c++) partial template specialization causes an ICE 11352 (segmentation fault) 11353 * [17]6663 (c++) ICE with attribute aligned 11354 * [18]7068 ICE with incomplete types 11355 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 11356 * [20]7647 (c++) ICE when data member has the name of the enclosing 11357 class 11358 * [21]7675 ICE in fixup_var_refs_1 11359 * [22]7718 'complex' template instantiation causes ICE 11360 * [23]8116 (c++) ICE in member template function 11361 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 11362 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 11363 fault 11364 * [26]8564 (c++) ICE in find_function_data, in function.c 11365 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 11366 * [28]8766 (c++) ICE after failed initialization of static template 11367 variable 11368 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 11369 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 11370 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 11371 definition 11372 * [32]9216 (c++) ICE on missing template parameter 11373 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 11374 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 11375 loop 11376 * [35]9429 (c++) ICE in template instantiation with a pointered new 11377 operator 11378 * [36]9516 Internal error when using a big array 11379 * [37]9600 (c++) ICE with typedefs in template class 11380 * [38]9629 (c++) virtual inheritance segfault 11381 * [39]9672 (c++) ICE: Error reporting routines re-entered 11382 * [40]9749 (c++) ICE in write_expression on invalid function 11383 prototype 11384 * [41]9794 (fortran) ICE: floating point exception during constant 11385 folding 11386 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 11387 * [43]9916 (c++) ICE with noreturn function in ?: statement 11388 * [44]9936 ICE with local function and variable-length 2d array 11389 * [45]10262 (c++) cc1plus crashes with large generated code 11390 * [46]10278 (c++) ICE in parser for invalid code 11391 * [47]10446 (c++) ICE on definition of nonexistent member function of 11392 nested class in a class template 11393 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 11394 declaration 11395 * [49]10506 (c++) ICE in build_new at cp/init.c with 11396 -fkeep-inline-functions and multiple inheritance 11397 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 11398 precision of the declared type 11399 11400 Optimization bugs 11401 11402 * [51]2001 Inordinately long compile times in reload CSE regs 11403 * [52]2391 Exponential compilation time explosion in combine 11404 * [53]2960 Duplicate loop conditions even with -Os 11405 * [54]4046 redundant conditional branch 11406 * [55]6405 Loop-unrolling related performance regressions 11407 * [56]6798 very long compile time with large case-statement 11408 * [57]6871 const objects shouldn't be moved to .bss 11409 * [58]6909 problem w/ -Os on modified loop-2c.c test case 11410 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 11411 non-void function'' warning 11412 * [60]7642 optimization problem with signbit() 11413 * [61]8634 incorrect code for inlining of memcpy under -O2 11414 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 11415 regular function call 11416 11417 C front end 11418 11419 * [63]2161 long if-else cascade overflows parser stack 11420 * [64]4319 short accepted on typedef'd char 11421 * [65]8602 incorrect line numbers in warning messages when using 11422 inline functions 11423 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 11424 AST nodes and breaks debugging dumps 11425 * [67]9853 miscompilation of non-constant structure initializer 11426 11427 c++ compiler and library 11428 11429 * [68]45 legal template specialization code is rejected (DUP: 11430 [69]3784) 11431 * [70]764 lookup failure: friend operator and dereferencing a pointer 11432 and templates (DUP: [71]5116) 11433 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 11434 2863) 11435 * [73]3663 G++ doesn't check access control during template 11436 instantiation 11437 * [74]3797 gcc fails to emit explicit specialization of a template 11438 member 11439 * [75]3948 Two destructors are called when no copy destructor is 11440 defined (ABI change) 11441 * [76]4137 Conversion operator within template is not accepted 11442 * [77]4361 bogus ambiguity taking the address of a member template 11443 * [78]4802 g++ accepts illegal template code (access to private 11444 member; DUP: [79]5837) 11445 * [80]4803 inline function is used but never defined, and g++ does 11446 not object 11447 * [81]5094 Partial specialization cannot be friend? 11448 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 11449 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 11450 time 11451 * [84]7015 certain __asm__ constructs rejected 11452 * [85]7086 compile time regression (quadratic behavior in 11453 fixup_var_refs) 11454 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 11455 std::abort 11456 * [87]7247 copy constructor missing when inlining enabled (invalid 11457 optimization?) 11458 * [88]7441 string array initialization compilation time regression 11459 from seconds to minutes 11460 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 11461 * [90]7804 bad printing of floating point constant in warning message 11462 * [91]8099 Friend classes and template specializations 11463 * [92]8117 member function pointers and multiple inheritance 11464 * [93]8205 using declaration and multiple inheritance 11465 * [94]8645 unnecessary non-zero checks in stl_tree.h 11466 * [95]8724 explicit destructor call for incomplete class allowed 11467 * [96]8805 compile time regression with many member variables 11468 * [97]8691 -O3 and -fno-implicit-templates are incompatible 11469 * [98]8700 unhelpful error message for binding temp to reference 11470 * [99]8724 explicit destructor call for incomplete class allowed 11471 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 11472 * [101]9016 Failure to consistently constant fold "constant" C++ 11473 objects 11474 * [102]9053 g++ confused about ambiguity of overloaded function 11475 templates 11476 * [103]9152 undefined virtual thunks 11477 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 11478 * [105]9297 data corruption due to codegen bug (when copying.) 11479 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 11480 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 11481 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 11482 local classes 11483 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 11484 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 11485 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 11486 and <iostream.h> 11487 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 11488 [114][DR 231] 11489 * [115]9555 ostream inserters fail to set badbit on exception 11490 * [116]9561 ostream inserters rethrow exception of wrong type 11491 * [117]9563 ostream::sentry returns true after a failed preparation 11492 * [118]9582 one-definition rule violation in std::allocator 11493 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 11494 * [120]9683 bug in initialization chains for static const variables 11495 from template classes 11496 * [121]9791 -Woverloaded-virtual reports hiding of destructor 11497 * [122]9817 collate::compare doesn't handle nul characters 11498 * [123]9825 filebuf::sputbackc breaks sbumpc 11499 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 11500 with custom traits 11501 * [125]9924 Multiple using statements for builtin functions not 11502 allowed 11503 * [126]9946 destructor is not called for temporary object 11504 * [127]9964 filebuf::close() sometimes fails to close file 11505 * [128]9988 filebuf::overflow writes EOF to file 11506 * [129]10033 optimization breaks polymorphic references w/ typeid 11507 operator 11508 * [130]10097 filebuf::underflow drops characters 11509 * [131]10132 filebuf destructor can throw exceptions 11510 * [132]10180 gcc fails to warn about non-inlined function 11511 * [133]10199 method parametrized by template does not work everywhere 11512 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 11513 * [135]10427 Stack corruption with variable-length automatic arrays 11514 and virtual destructors 11515 * [136]10503 Compilation never stops in fixed_type_or_null 11516 11517 Objective-C 11518 11519 * [137]5956 selectors aren't matched properly when added to the 11520 selector table 11521 11522 Fortran compiler and library 11523 11524 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 11525 detect 11526 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 11527 info requested 11528 * [140]5634 doc: explain that configure --prefix=~/... does not work 11529 * [141]6367 multiple repeat counts confuse namelist read into array 11530 * [142]6491 Logical operations error on logicals when using 11531 -fugly-logint 11532 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 11533 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 11534 on irix6.5 11535 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 11536 assume a direct access file 11537 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 11538 -fno-automatic) 11539 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 11540 * [148]7388 Incorrect output with 0-based array of characters 11541 * [149]8587 Double complex zero ** double precision number -> NaN 11542 instead of zero 11543 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 11544 unknown register name line-length-none 11545 * [151]10197 Direct access files not unformatted by default 11546 11547 Java compiler and library 11548 11549 * [152]6005 gcj fails to build rhug on alpha 11550 * [153]6389 System.getProperty("") should always throw an 11551 IllegalArgumentException 11552 * [154]6576 java.util.ResourceBundle.getResource ignores locale 11553 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 11554 * [156]7060 getMethod() doesn't search super interface 11555 * [157]7073 bytecode interpreter gives wrong answer for interface 11556 getSuperclass() 11557 * [158]7180 possible bug in 11558 javax.naming.spi.NamingManager.getPlusPath() 11559 * [159]7416 java.security startup refs "GNU libgcj.security" 11560 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 11561 env (DUP: [161]7578) 11562 * [162]7611 Internal error while compiling libjava with -O 11563 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 11564 * [164]7766 ZipInputStream.available returns 0 immediately after 11565 construction 11566 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 11567 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 11568 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 11569 * [168]8234 ZipInputStream chokes when InputStream.read() returns 11570 small chunks 11571 * [169]8415 reflection bug: exception info for Method 11572 * [170]8481 java.Random.nextInt(int) may return negative 11573 * [171]8593 Error reading GZIPped files with BufferedReader 11574 * [172]8759 java.beans.Introspector has no flushCaches() or 11575 flushFromCaches() methods 11576 * [173]8997 spin() calls Thread.sleep 11577 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 11578 instead of the root content of C: 11579 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 11580 wrong return codes 11581 * [176]9271 Severe bias in java.security.SecureRandom 11582 11583 Ada compiler and library 11584 11585 * [177]6767 make gnatlib-shared fails on -laddr2line 11586 * [178]9911 gnatmake fails to link when GCC configured with 11587 --with-sjlj-exceptions=yes 11588 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 11589 * [180]10546 Ada tasking not working on Red Hat 9 11590 11591 preprocessor 11592 11593 * [181]7029 preprocessor should ignore #warning with -M 11594 11595 ARM-specific 11596 11597 * [182]2903 [arm] Optimization bug with long long arithmetic 11598 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 11599 11600 FreeBSD-specific 11601 11602 * [184]7680 float functions undefined in math.h/cmath with #define 11603 _XOPEN_SOURCE 11604 11605 HP-UX or HP-PA-specific 11606 11607 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 11608 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 11609 fputc_unlocked 11610 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 11611 11612 m68hc11-specific 11613 11614 * [188]6744 Bad assembler code generated: reference to pseudo 11615 register z 11616 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 11617 in reload1.c 11618 11619 MIPS-specific 11620 11621 * [190]9496 [mips-linux] bug in optimizer? 11622 11623 PowerPC-specific 11624 11625 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 11626 space 11627 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 11628 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 11629 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 11630 11631 SPARC-specific 11632 11633 * [195]10267 (documentation) Wrong build instructions for 11634 *-*-solaris2* 11635 11636 x86-specific (Intel/AMD) 11637 11638 * [196]7916 ICE in instantiate_virtual_register_1 11639 * [197]7926 (c++) i486 instructions in header files make c++ programs 11640 crash on i386 11641 * [198]8555 ICE in gen_split_1231 11642 * [199]8994 ICE with -O -march=pentium4 11643 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 11644 * [201]9806 ICE in inline assembly with -fPIC flag 11645 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 11646 regs 11647 * [203]10233 64-bit comparison only comparing bottom 32-bits 11648 * [204]10286 type-punning doesn't work with __m64 and -O 11649 * [205]10308 [x86] ICE with -O -fgcse or -O2 11650 __________________________________________________________________ 11651 11652GCC 3.3.1 11653 11654 Bug Fixes 11655 11656 This section lists the problem reports (PRs) from GCC's bug tracking 11657 system that are known to be fixed in the 3.3.1 release. This list might 11658 not be complete (that is, it is possible that some PRs that have been 11659 fixed are not listed here). 11660 11661 Bootstrap failures 11662 11663 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 11664 11665 Internal compiler errors (multi-platform) 11666 11667 * [207]5754 ICE on invalid nested template class 11668 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 11669 and --enable-checking 11670 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 11671 * [210]7053 (c++) ICE when declaring a function already defined as a 11672 friend method of a template class 11673 * [211]8164 (c++) ICE when using different const expressions as 11674 template parameter 11675 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 11676 * [213]9559 (c++) ICE with invalid initialization of a static const 11677 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 11678 when redeclaring a static member variable 11679 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 11680 dwarfout.c with -g -O -finline-functions 11681 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 11682 * [217]10475 ICE in subreg_highpart_offset for code with long long 11683 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 11684 from a void pointer 11685 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 11686 instantiating static member variables 11687 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 11688 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 11689 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 11690 MAX_INT_64BIT 11691 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 11692 sched.c 11693 * [224]10939 (c++) ICE with template code 11694 * [225]10956 (c++) ICE when specializing a template member function 11695 of a template class, in tsubst, in cp/pt.c 11696 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 11697 defined) 11698 * [227]11059 (c++) ICE with empty union 11699 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 11700 -O2 -fnon-call-exceptions 11701 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 11702 * [230]11149 (c++) ICE on error when instantiation with call function 11703 of a base type 11704 * [231]11228 (c++) ICE on new-expression using array operator new and 11705 default-initialization 11706 * [232]11282 (c++) Infinite memory usage after syntax error 11707 * [233]11301 (fortran) ICE with -fno-globals 11708 * [234]11308 (c++) ICE when using an enum type name as if it were a 11709 class or namespace 11710 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 11711 an empty struct 11712 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 11713 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 11714 template member functions 11715 11716 Optimization bugs 11717 11718 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 11719 problem) 11720 * [239]11304 Wrong code production with -fomit-frame-pointer 11721 * [240]11381 volatile memory access optimized away 11722 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 11723 * [242]11557 constant folding bug generates wrong code 11724 11725 C front end 11726 11727 * [243]5897 No warning for statement after return 11728 * [244]11279 DWARF-2 output mishandles large enums 11729 11730 Preprocessor bugs 11731 11732 * [245]11022 no warning for non-compatible macro redefinition 11733 11734 C++ compiler and library 11735 11736 * [246]2330 static_cast<>() to a private base is allowed 11737 * [247]5388 Incorrect message "operands to ?: have different types" 11738 * [248]5390 Libiberty fails to demangle multi-digit template 11739 parameters 11740 * [249]7877 Incorrect parameter passing to specializations of member 11741 function templates 11742 * [250]9393 Anonymous namespaces and compiling the same file twice 11743 * [251]10032 -pedantic converts some errors to warnings 11744 * [252]10468 const typeof(x) is non-const, but only in templates 11745 * [253]10527 confused error message with "new int()" parameter 11746 initializer 11747 * [254]10679 parameter MIN_INLINE_INSNS is not honored 11748 * [255]10682 gcc chokes on a typedef for an enum inside a class 11749 template 11750 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 11751 0. 11752 * [257]10845 template member function (with nested template as 11753 parameter) cannot be called anymore if another unrelated template 11754 member function is defined 11755 * [258]10849 Cannot define an out-of-class specialization of a 11756 private nested template class 11757 * [259]10888 Suppress -Winline warnings for system headers 11758 * [260]10929 -Winline warns about functions for which no definition 11759 is visible 11760 * [261]10931 valid conversion static_cast<const unsigned 11761 int&>(lvalue-of-type-int) is rejected 11762 * [262]10940 Bad code with explicit specialization 11763 * [263]10968 If member function implicitly instantiated, explicit 11764 instantiation of class fails to instantiate it 11765 * [264]10990 Cannot convert with dynamic_cast<> to a private base 11766 class from within a member function 11767 * [265]11039 Bad interaction between implicit typename deprecation 11768 and friendship 11769 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 11770 "__unused__" instead 11771 * [267]11095 C++ iostream manipulator causes segfault when called 11772 with negative argument 11773 * [268]11098 g++ doesn't emit complete debugging information for 11774 local variables in destructors 11775 * [269]11137 GNU/Linux shared library constructors not called unless 11776 there's one global object 11777 * [270]11154 spurious ambiguity report for template class 11778 specialization 11779 * [271]11329 Compiler cannot find user defined implicit typecast 11780 * [272]11332 Spurious error with casts in ?: expression 11781 * [273]11431 static_cast behavior with subclasses when default 11782 constructor available 11783 * [274]11528 money_get facet does not accept "$.00" as valid 11784 * [275]11546 Type lookup problems in out-of-line definition of a 11785 class doubly nested from a template class 11786 * [276]11567 C++ code containing templated member function with same 11787 name as pure virtual member function results in linking failure 11788 * [277]11645 Failure to deal with using and private inheritance 11789 11790 Java compiler and library 11791 11792 * [278]5179 Qualified static field access doesn't initialize its 11793 class 11794 * [279]8204 gcj -O2 to native reorders certain instructions 11795 improperly 11796 * [280]10838 java.io.ObjectInputStream syntax error 11797 * [281]10886 The RMI registry that comes with GCJ does not work 11798 correctly 11799 * [282]11349 JNDI URL context factories not located correctly 11800 11801 x86-specific (Intel/AMD) 11802 11803 * [283]4823 ICE on inline assembly code 11804 * [284]8878 miscompilation with -O and SSE 11805 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 11806 -masm=intel 11807 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 11808 in reload1.c 11809 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 11810 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 11811 * [289]11044 [x86] out of range loop instructions for FP code on K6 11812 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 11813 built-ins 11814 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 11815 is used 11816 11817 SPARC- or Solaris- specific 11818 11819 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 11820 * [293]10142 [SPARC64] gcc produces wrong code when passing 11821 structures by value 11822 * [294]10663 New configure check aborts with Sun tools. 11823 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 11824 * [296]10876 ICE in calculate_giv_inc when building KDE 11825 * [297]10955 wrong code at -O3 for structure argument in context of 11826 structure return 11827 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 11828 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 11829 Linux kernel 11830 11831 ia64 specific 11832 11833 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 11834 * [301]11320 scheduler bug (in machine depended reorganization pass) 11835 * [302]11599 bug with conditional and __builtin_prefetch 11836 11837 PowerPC specific 11838 11839 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 11840 during loop) 11841 * [304]10871 error in rs6000_stack_info save_size computation 11842 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 11843 cures it 11844 11845 m68k-specific 11846 11847 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 11848 * [307]10557 [m68k] ICE in subreg_offset_representable_p 11849 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 11850 11851 ARM-specific 11852 11853 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 11854 functions with __attribute__ ((interrupt ("IRQ"))) 11855 * [310]10842 [arm] Clobbered link register is copied to pc under 11856 certain circumstances 11857 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 11858 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 11859 (3.4) 11860 11861 MIPS-specific 11862 11863 * [313]11084 ICE in propagate_one_insn, in flow.c 11864 11865 SH-specific 11866 11867 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 11868 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 11869 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 11870 C++ files 11871 11872 GNU/Linux (or Hurd?) specific 11873 11874 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 11875 11876 UnixWare specific 11877 11878 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 11879 7.1.1 11880 11881 Cygwin (or mingw) specific 11882 11883 * [319]5287 ICE with dllimport attribute 11884 * [320]10148 [MingW/CygWin] Compiler dumps core 11885 11886 DJGPP specific 11887 11888 * [321]8787 GCC fails to emit .intel_syntax when invoked with 11889 -masm=intel on DJGPP 11890 11891 Darwin (and MacOS X) specific 11892 11893 * [322]10900 trampolines crash 11894 11895 Documentation 11896 11897 * [323]1607 (c++) Format attributes on methods undocumented 11898 * [324]4252 Invalid option `-fdump-translation-unit' 11899 * [325]4490 Clarify restrictions on -m96bit-long-double, 11900 -m128bit-long-double 11901 * [326]10355 document an issue with regparm attribute on some systems 11902 (e.g. Solaris) 11903 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 11904 (Unix)" is wrong 11905 * [328]10805 document bug in old version of Sun assembler 11906 * [329]10815 warn against GNU binutils on AIX 11907 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 11908 * [331]11280 Manual incorrect with respect to -freorder-blocks 11909 * [332]11466 Document -mlittle-endian and its restrictions for the 11910 sparc64 port 11911 11912 Testsuite bugs (compiler itself is not affected) 11913 11914 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 11915 report failure 11916 * [334]10810 gcc-3.3 fails make check: buffer overrun in 11917 test_demangle.c 11918 __________________________________________________________________ 11919 11920GCC 3.3.2 11921 11922 Bug Fixes 11923 11924 This section lists the problem reports (PRs) from GCC's bug tracker 11925 that are known to be fixed in the 3.3.2 release. This list might not be 11926 complete (that is, it is possible that some PRs that have been fixed 11927 are not listed here). 11928 11929 Bootstrap failures and problems 11930 11931 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 11932 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 11933 --enable-threads=posix 11934 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 11935 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 11936 7.1.1) 11937 * [339]11687 xstormy16-elf build fails in libf2c 11938 * [340]12263 [SGI IRIX] bootstrap fails during compile of 11939 libf2c/libI77/backspace.c 11940 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 11941 fix-header processing) 11942 11943 Internal compiler errors (multi-platform) 11944 11945 * [342]7277 Casting integers to vector types causes ICE 11946 * [343]7939 (c++) ICE on invalid function template specialization 11947 * [344]11063 (c++) ICE on parsing initialization list of const array 11948 member 11949 * [345]11207 ICE with negative index in array element designator 11950 * [346]11522 (fortran) g77 dwarf-2 ICE in 11951 add_abstract_origin_attribute 11952 * [347]11595 (c++) ICE on duplicate label definition 11953 * [348]11646 (c++) ICE in commit_one_edge_insertion with 11954 -fnon-call-exceptions -fgcse -O 11955 * [349]11665 ICE in struct initializer when taking address 11956 * [350]11852 (c++) ICE with bad struct initializer. 11957 * [351]11878 (c++) ICE in cp_expr_size 11958 * [352]11883 ICE with any -O on mercury-generated C code 11959 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 11960 cp/typeck2.c when applying typeid operator to template template 11961 parameter 11962 * [354]12146 ICE in lookup_template_function, in cp/pt.c 11963 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 11964 -fno-gcse -O2 11965 * [356]12369 (c++) ICE with templates and friends 11966 * [357]12446 ICE in emit_move_insn on complicated array reference 11967 * [358]12510 ICE in final_scan_insn 11968 * [359]12544 ICE with large parameters used in nested functions 11969 11970 C and optimization bugs 11971 11972 * [360]9862 spurious warnings with -W -finline-functions 11973 * [361]10962 lookup_field is a linear search on a linked list (can be 11974 slow if large struct) 11975 * [362]11370 -Wunreachable-code gives false complaints 11976 * [363]11637 invalid assembly with -fnon-call-exceptions 11977 * [364]11885 Problem with bitfields in packed structs 11978 * [365]12082 Inappropriate unreachable code warnings 11979 * [366]12180 Inline optimization fails for variadic function 11980 * [367]12340 loop unroller + gcse produces wrong code 11981 11982 C++ compiler and library 11983 11984 * [368]3907 nested template parameter collides with member name 11985 * [369]5293 confusing message when binding a temporary to a reference 11986 * [370]5296 [DR115] Pointers to functions and to template functions 11987 behave differently in deduction 11988 * [371]7939 ICE on function template specialization 11989 * [372]8656 Unable to assign function with __attribute__ and pointer 11990 return type to an appropriate variable 11991 * [373]10147 Confusing error message for invalid template function 11992 argument 11993 * [374]11400 std::search_n() makes assumptions about Size parameter 11994 * [375]11409 issues with using declarations, overloading, and 11995 built-in functions 11996 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 11997 multiple bits in mask 11998 * [377]11786 operator() call on variable in other namespace not 11999 recognized 12000 * [378]11867 static_cast ignores ambiguity 12001 * [379]11928 bug with conversion operators that are typedefs 12002 * [380]12114 Uninitialized memory accessed in dtor 12003 * [381]12163 static_cast + explicit constructor regression 12004 * [382]12181 Wrong code with comma operator and c++ 12005 * [383]12236 regparm and fastcall messes up parameters 12006 * [384]12266 incorrect instantiation of unneeded template during 12007 overload resolution 12008 * [385]12296 istream::peek() doesn't set eofbit 12009 * [386]12298 [sjlj exceptions] Stack unwind destroys 12010 not-yet-constructed object 12011 * [387]12369 ICE with templates and friends 12012 * [388]12337 apparently infinite loop in g++ 12013 * [389]12344 stdcall attribute ignored if function returns a pointer 12014 * [390]12451 missing(late) class forward declaration in cxxabi.h 12015 * [391]12486 g++ accepts invalid use of a qualified name 12016 12017 x86 specific (Intel/AMD) 12018 12019 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 12020 builtins 12021 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 12022 -O2 12023 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 12024 * [395]12116 [k6] Invalid assembly output values with X-MAME code 12025 * [396]12070 ICE converting between double and long double with 12026 -msoft-float 12027 12028 ia64-specific 12029 12030 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 12031 * [398]11535 __builtin_return_address may not work on ia64 12032 * [399]11693 [ia64] ICE in gen_nop_type 12033 * [400]12224 [ia64] Thread-local storage doesn't work 12034 12035 PowerPC-specific 12036 12037 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 12038 kernel 12039 * [402]11319 loop miscompiled on ppc32 12040 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 12041 12042 SPARC-specific 12043 12044 * [404]11662 wrong code for expr. with cast to long long and 12045 exclusive or 12046 * [405]11965 invalid assembler code for a shift < 32 operation 12047 * [406]12301 (c++) stack corruption when a returned expression throws 12048 an exception 12049 12050 Alpha-specific 12051 12052 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 12053 kernel 2.4.22-pre8 12054 12055 HPUX-specific 12056 12057 * [408]11313 problem with #pragma weak and static inline functions 12058 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 12059 12060 Solaris specific 12061 12062 * [410]12166 Profiled programs crash if PROFDIR is set 12063 12064 Solaris-x86 specific 12065 12066 * [411]12101 i386 Solaris no longer works with GNU as? 12067 12068 Miscellaneous embedded target-specific bugs 12069 12070 * [412]10988 [m32r-elf] wrong blockmove code with -O3 12071 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 12072 -O2 12073 * [414]11902 [sh4] spec file improperly inserts rpath even when none 12074 needed 12075 * [415]11903 [sh4] -pthread fails to link due to error in spec file 12076 on sh4 12077 __________________________________________________________________ 12078 12079GCC 3.3.3 12080 12081 Minor features 12082 12083 In addition to the bug fixes documented below, this release contains 12084 few minor features such as: 12085 * Support for --with-sysroot 12086 * Support for automatic detection of executable stacks 12087 * Support for SSE3 instructions 12088 * Support for thread local storage debugging under GDB on S390 12089 12090 Bug Fixes 12091 12092 This section lists the problem reports (PRs) from GCC's bug tracker 12093 that are known to be fixed in the 3.3.3 release. This list might not be 12094 complete (that is, it is possible that some PRs that have been fixed 12095 are not listed here). 12096 12097 Bootstrap failures and issues 12098 12099 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 12100 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 12101 unable to infer tagged configuration 12102 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 12103 subdirectories properly 12104 12105 Internal compiler errors (multi-platform) 12106 12107 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 12108 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 12109 * [420]10555 (c++) ICE on undefined template argument 12110 * [421]10706 (c++) ICE in mangle_class_name_for_template 12111 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 12112 active 12113 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 12114 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 12115 * [425]12632 (fortran) -fbounds-check ICE 12116 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 12117 3.3.2 12118 * [427]12726 (c++) ICE (segfault) on trivial code 12119 * [428]12890 (c++) ICE on compilation of class with throwing method 12120 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 12121 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 12122 correct code with -O2 -fno-force-mem 12123 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 12124 template 12125 * [432]13318 ICE: floating point exception in the loop optimizer 12126 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 12127 except.c 12128 * [434]13574 (c++) invalid array default initializer in class lets 12129 gcc consume all memory and die 12130 * [435]13475 ICE on SIMD variables with partial value initialization 12131 * [436]13797 (c++) ICE on invalid template parameter 12132 * [437]13824 (java) gcj SEGV with simple .java program 12133 12134 C and optimization bugs 12135 12136 * [438]8776 loop invariants are not removed (most likely) 12137 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 12138 strncmp by memcmp 12139 * [440]11350 undefined labels with -Os -fPIC 12140 * [441]12826 Optimizer removes reference through volatile pointer 12141 * [442]12500 stabs debug info: void no longer a predefined / builtin 12142 type 12143 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 12144 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 12145 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 12146 * [446]13507 spurious printf format warning 12147 * [447]13382 Type information for const pointer disappears during 12148 optimization. 12149 * [448]13394 noreturn attribute ignored on recursive invokation 12150 * [449]13400 Compiled code crashes storing to read-only location 12151 * [450]13521 Endless loop in calculate_global_regs_live 12152 12153 C++ compiler and library 12154 12155 Some of the bug fixes in this list were made to implement decisions 12156 that the ISO C++ standards committee has made concerning several defect 12157 reports (DRs). Links in the list below point to detailed discussion of 12158 the relevant defect report. 12159 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 12160 unification 12161 * [452]2294 using declaration confusion 12162 * [453]5050 template instantiation depth exceeds limit: recursion 12163 problem? 12164 * [454]9371 Bad exception handling in 12165 i/ostream::operator>>/<<(streambuf*) 12166 * [455]9546 bad exception handling in ostream members 12167 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 12168 face of unknown locales 12169 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 12170 * [459]10095 istream::operator>>(int&) sets ios::badbit when 12171 ios::failbit is set. 12172 * [460]11554 Warning about reordering of initializers doesn't mention 12173 location of constructor 12174 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 12175 * [462]12352 Exception safety problems in src/localename.cc 12176 * [463]12438 Memory leak in locale::combine() 12177 * [464]12540 Memory leak in locale::locale(const char*) 12178 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 12179 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 12180 * [470]12696 memory eating infinite loop in diagnostics (error 12181 recovery problem) 12182 * [471]12815 Code compiled with optimization behaves unexpectedly 12183 * [472]12862 Conflicts between typedefs/enums and namespace member 12184 declarations 12185 * [473]12926 Wrong value after assignment in initialize list using 12186 bit-fields 12187 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 12188 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 12189 * [478]13007 basic_streambuf::pubimbue, imbue wrong 12190 * [479]13009 Implicitly-defined assignment operator writes to wrong 12191 memory 12192 * [480]13057 regparm attribute not applied to destructor 12193 * [481]13070 -Wformat option ignored in g++ 12194 * [482]13081 forward template declarations in <complex> let inlining 12195 fail 12196 * [483]13239 Assertion does not seem to work correctly anymore 12197 * [484]13262 "xxx is private within this context" when initializing a 12198 self-contained template class 12199 * [485]13290 simple typo in concept checking for std::generate_n 12200 * [486]13323 Template code does not compile in presence of typedef 12201 * [487]13369 __verify_grouping (and __add_grouping?) not correct 12202 * [488]13371 infinite loop with packed struct and inlining 12203 * [489]13445 Template argument replacement "dereferences" a typedef 12204 * [490]13461 Fails to access protected-ctor from public constant 12205 * [491]13462 Non-standard-conforming type set::pointer 12206 * [492]13478 gcc uses wrong constructor to initialize a const 12207 reference 12208 * [493]13544 "conflicting types" for enums in different scopes 12209 * [494]13650 string::compare should not (always) use 12210 traits_type::length() 12211 * [495]13683 bogus warning about passing non-PODs through ellipsis 12212 * [496]13688 Derived class is denied access to protected base class 12213 member class 12214 * [497]13774 Member variable cleared in virtual multiple inheritance 12215 class 12216 * [498]13884 Protect sstream.tcc from extern template use 12217 12218 Java compiler and library 12219 12220 * [499]10746 [win32] garbage collection crash in GCJ 12221 12222 Objective-C compiler and library 12223 12224 * [500]11433 Crash due to dereferencing null pointer when querying 12225 protocol 12226 12227 Fortran compiler and library 12228 12229 * [501]12633 logical expression gives incorrect result with 12230 -fugly-logint option 12231 * [502]13037 [gcse-lm] g77 generates incorrect code 12232 * [503]13213 Hex constant problem when compiling with -fugly-logint 12233 and -ftypeless-boz 12234 12235 x86-specific (Intel/AMD) 12236 12237 * [504]4490 ICE with -m128bit-long-double 12238 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 12239 `reg' in make_field_assignment, in combine.c 12240 * [506]12441 ICE: can't find a register to spill 12241 * [507]12943 array static-init failure under -fpic, -fPIC 12242 * [508]13608 Incorrect code with -O3 -ffast-math 12243 12244 PowerPC-specific 12245 12246 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 12247 __attribute__((aligned(16))) 12248 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 12249 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 12250 altivec.md) 12251 * [512]12537 g++ generates writeable text sections 12252 12253 SPARC-specific 12254 12255 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 12256 -m64 12257 * [514]12865 mprotect call to make trampoline executable may fail 12258 * [515]13354 ICE in sparc_emit_set_const32 12259 12260 ARM-specific 12261 12262 * [516]10467 [arm] ICE in pre_insert_copy_insn, 12263 12264 ia64-specific 12265 12266 * [517]11226 ICE passing struct arg with two floats 12267 * [518]11227 ICE for _Complex float, _Complex long double args 12268 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 12269 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 12270 * Various fixes for libunwind 12271 12272 Alpha-specific 12273 12274 * [521]12654 Incorrect comparison code generated for Alpha 12275 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 12276 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 12277 12278 HPPA-specific 12279 12280 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 12281 * [525]12158 [hppa] compilation does not terminate at -O1 12282 12283 S390-specific 12284 12285 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 12286 (1<<24)-1 possible for CLCL-Instruction 12287 12288 SH-specific 12289 12290 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 12291 * [528]10392 optimizer generates faulty array indexing 12292 * [529]11322 SH profiler outputs multiple definitions of symbol 12293 * [530]13069 gcc/config/sh/rtems.h broken 12294 * [531]13302 Putting a va_list in a struct causes seg fault 12295 * [532]13585 Incorrect optimization of call to sfunc 12296 * Fix inappropriately exported libgcc functions from the shared 12297 library 12298 12299 Other embedded target specific 12300 12301 * [533]8916 [mcore] unsigned char assign gets hosed. 12302 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 12303 * [535]13122 [h8300] local variable gets corrupted by function call 12304 when -fomit-frame-pointer is given 12305 * [536]13256 [cris] strict_low_part mistreated in delay slots 12306 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 12307 -fexpensive-optimizations produces wrong code on mcore 12308 12309 GNU HURD-specific 12310 12311 * [538]12561 gcc/config/t-gnu needs updating to work with 12312 --with-sysroot 12313 12314 Tru64 Unix specific 12315 12316 * [539]6243 testsuite fails almost all tests due to no libintl in 12317 LD_LIBRARY_PATH during test. 12318 * [540]11397 weak aliases broken on Tru64 UNIX 12319 12320 AIX-specific 12321 12322 * [541]12505 build failure due to defines of uchar in cpphash.h and 12323 sys/types.h 12324 * [542]13150 WEAK symbols not exported by collect2 12325 12326 IRIX-specific 12327 12328 * [543]12666 fixincludes problem on IRIX 6.5.19m 12329 12330 Solaris-specific 12331 12332 * [544]12969 Including sys/byteorder.h breaks configure checks 12333 12334 Testsuite problems (compiler is not affected) 12335 12336 * [545]10819 testsuite creates CR+LF on compiler version lines in 12337 test summary files 12338 * [546]11612 abi_check not finding correct libgcc_s.so.1 12339 12340 Miscellaneous 12341 12342 * [547]13211 using -###, incorrect warnings about unused linker file 12343 are produced 12344 __________________________________________________________________ 12345 12346GCC 3.3.4 12347 12348 This is the [548]list of problem reports (PRs) from GCC's bug tracking 12349 system that are known to be fixed in the 3.3.4 release. This list might 12350 not be complete (that is, it is possible that some PRs that have been 12351 fixed are not listed here). 12352 __________________________________________________________________ 12353 12354GCC 3.3.5 12355 12356 This is the [549]list of problem reports (PRs) from GCC's bug tracking 12357 system that are known to be fixed in the 3.3.5 release. This list might 12358 not be complete (that is, it is possible that some PRs that have been 12359 fixed are not listed here). 12360 __________________________________________________________________ 12361 12362GCC 3.3.6 12363 12364 This is the [550]list of problem reports (PRs) from GCC's bug tracking 12365 system that are known to be fixed in the 3.3.6 release. This list might 12366 not be complete (that is, it is possible that some PRs that have been 12367 fixed are not listed here). 12368 12369 12370 For questions related to the use of GCC, please consult these web 12371 pages and the [551]GCC manuals. If that fails, the 12372 [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these 12373 web pages and the development of GCC are welcome on our developer 12374 list at [553]gcc@gcc.gnu.org. All of [554]our lists have public 12375 archives. 12376 12377 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 12378 distribution of this entire article is permitted in any medium, 12379 provided this notice is preserved. 12380 12381 These pages are [556]maintained by the GCC team. Last modified 12382 2016-01-30[557]. 12383 12384References 12385 12386 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 12387 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 12388 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 12389 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 12390 5. http://gcc.gnu.org/news/dfa.html 12391 6. http://gcc.gnu.org/c99status.html 12392 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 12393 8. https://gcc.gnu.org/PR10140 12394 9. https://gcc.gnu.org/PR10198 12395 10. https://gcc.gnu.org/PR10338 12396 11. https://gcc.gnu.org/PR3581 12397 12. https://gcc.gnu.org/PR4382 12398 13. https://gcc.gnu.org/PR5533 12399 14. https://gcc.gnu.org/PR6387 12400 15. https://gcc.gnu.org/PR6412 12401 16. https://gcc.gnu.org/PR6620 12402 17. https://gcc.gnu.org/PR6663 12403 18. https://gcc.gnu.org/PR7068 12404 19. https://gcc.gnu.org/PR7083 12405 20. https://gcc.gnu.org/PR7647 12406 21. https://gcc.gnu.org/PR7675 12407 22. https://gcc.gnu.org/PR7718 12408 23. https://gcc.gnu.org/PR8116 12409 24. https://gcc.gnu.org/PR8358 12410 25. https://gcc.gnu.org/PR8511 12411 26. https://gcc.gnu.org/PR8564 12412 27. https://gcc.gnu.org/PR8660 12413 28. https://gcc.gnu.org/PR8766 12414 29. https://gcc.gnu.org/PR8803 12415 30. https://gcc.gnu.org/PR8846 12416 31. https://gcc.gnu.org/PR8906 12417 32. https://gcc.gnu.org/PR9216 12418 33. https://gcc.gnu.org/PR9261 12419 34. https://gcc.gnu.org/PR9263 12420 35. https://gcc.gnu.org/PR9429 12421 36. https://gcc.gnu.org/PR9516 12422 37. https://gcc.gnu.org/PR9600 12423 38. https://gcc.gnu.org/PR9629 12424 39. https://gcc.gnu.org/PR9672 12425 40. https://gcc.gnu.org/PR9749 12426 41. https://gcc.gnu.org/PR9794 12427 42. https://gcc.gnu.org/PR9829 12428 43. https://gcc.gnu.org/PR9916 12429 44. https://gcc.gnu.org/PR9936 12430 45. https://gcc.gnu.org/PR10262 12431 46. https://gcc.gnu.org/PR10278 12432 47. https://gcc.gnu.org/PR10446 12433 48. https://gcc.gnu.org/PR10451 12434 49. https://gcc.gnu.org/PR10506 12435 50. https://gcc.gnu.org/PR10549 12436 51. https://gcc.gnu.org/PR2001 12437 52. https://gcc.gnu.org/PR2391 12438 53. https://gcc.gnu.org/PR2960 12439 54. https://gcc.gnu.org/PR4046 12440 55. https://gcc.gnu.org/PR6405 12441 56. https://gcc.gnu.org/PR6798 12442 57. https://gcc.gnu.org/PR6871 12443 58. https://gcc.gnu.org/PR6909 12444 59. https://gcc.gnu.org/PR7189 12445 60. https://gcc.gnu.org/PR7642 12446 61. https://gcc.gnu.org/PR8634 12447 62. https://gcc.gnu.org/PR8750 12448 63. https://gcc.gnu.org/PR2161 12449 64. https://gcc.gnu.org/PR4319 12450 65. https://gcc.gnu.org/PR8602 12451 66. https://gcc.gnu.org/PR9177 12452 67. https://gcc.gnu.org/PR9853 12453 68. https://gcc.gnu.org/PR45 12454 69. https://gcc.gnu.org/PR3784 12455 70. https://gcc.gnu.org/PR764 12456 71. https://gcc.gnu.org/PR5116 12457 72. https://gcc.gnu.org/PR2862 12458 73. https://gcc.gnu.org/PR3663 12459 74. https://gcc.gnu.org/PR3797 12460 75. https://gcc.gnu.org/PR3948 12461 76. https://gcc.gnu.org/PR4137 12462 77. https://gcc.gnu.org/PR4361 12463 78. https://gcc.gnu.org/PR4802 12464 79. https://gcc.gnu.org/PR5837 12465 80. https://gcc.gnu.org/PR4803 12466 81. https://gcc.gnu.org/PR5094 12467 82. https://gcc.gnu.org/PR5730 12468 83. https://gcc.gnu.org/PR6713 12469 84. https://gcc.gnu.org/PR7015 12470 85. https://gcc.gnu.org/PR7086 12471 86. https://gcc.gnu.org/PR7099 12472 87. https://gcc.gnu.org/PR7247 12473 88. https://gcc.gnu.org/PR7441 12474 89. https://gcc.gnu.org/PR7768 12475 90. https://gcc.gnu.org/PR7804 12476 91. https://gcc.gnu.org/PR8099 12477 92. https://gcc.gnu.org/PR8117 12478 93. https://gcc.gnu.org/PR8205 12479 94. https://gcc.gnu.org/PR8645 12480 95. https://gcc.gnu.org/PR8724 12481 96. https://gcc.gnu.org/PR8805 12482 97. https://gcc.gnu.org/PR8691 12483 98. https://gcc.gnu.org/PR8700 12484 99. https://gcc.gnu.org/PR8724 12485 100. https://gcc.gnu.org/PR8949 12486 101. https://gcc.gnu.org/PR9016 12487 102. https://gcc.gnu.org/PR9053 12488 103. https://gcc.gnu.org/PR9152 12489 104. https://gcc.gnu.org/PR9182 12490 105. https://gcc.gnu.org/PR9297 12491 106. https://gcc.gnu.org/PR9318 12492 107. https://gcc.gnu.org/PR9320 12493 108. https://gcc.gnu.org/PR9400 12494 109. https://gcc.gnu.org/PR9424 12495 110. https://gcc.gnu.org/PR9425 12496 111. https://gcc.gnu.org/PR9439 12497 112. https://gcc.gnu.org/PR9474 12498 113. https://gcc.gnu.org/PR9548 12499 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 12500 115. https://gcc.gnu.org/PR9555 12501 116. https://gcc.gnu.org/PR9561 12502 117. https://gcc.gnu.org/PR9563 12503 118. https://gcc.gnu.org/PR9582 12504 119. https://gcc.gnu.org/PR9622 12505 120. https://gcc.gnu.org/PR9683 12506 121. https://gcc.gnu.org/PR9791 12507 122. https://gcc.gnu.org/PR9817 12508 123. https://gcc.gnu.org/PR9825 12509 124. https://gcc.gnu.org/PR9826 12510 125. https://gcc.gnu.org/PR9924 12511 126. https://gcc.gnu.org/PR9946 12512 127. https://gcc.gnu.org/PR9964 12513 128. https://gcc.gnu.org/PR9988 12514 129. https://gcc.gnu.org/PR10033 12515 130. https://gcc.gnu.org/PR10097 12516 131. https://gcc.gnu.org/PR10132 12517 132. https://gcc.gnu.org/PR10180 12518 133. https://gcc.gnu.org/PR10199 12519 134. https://gcc.gnu.org/PR10300 12520 135. https://gcc.gnu.org/PR10427 12521 136. https://gcc.gnu.org/PR10503 12522 137. https://gcc.gnu.org/PR5956 12523 138. https://gcc.gnu.org/PR1832 12524 139. https://gcc.gnu.org/PR3924 12525 140. https://gcc.gnu.org/PR5634 12526 141. https://gcc.gnu.org/PR6367 12527 142. https://gcc.gnu.org/PR6491 12528 143. https://gcc.gnu.org/PR6742 12529 144. https://gcc.gnu.org/PR7113 12530 145. https://gcc.gnu.org/PR7236 12531 146. https://gcc.gnu.org/PR7278 12532 147. https://gcc.gnu.org/PR7384 12533 148. https://gcc.gnu.org/PR7388 12534 149. https://gcc.gnu.org/PR8587 12535 150. https://gcc.gnu.org/PR9038 12536 151. https://gcc.gnu.org/PR10197 12537 152. https://gcc.gnu.org/PR6005 12538 153. https://gcc.gnu.org/PR6389 12539 154. https://gcc.gnu.org/PR6576 12540 155. https://gcc.gnu.org/PR6652 12541 156. https://gcc.gnu.org/PR7060 12542 157. https://gcc.gnu.org/PR7073 12543 158. https://gcc.gnu.org/PR7180 12544 159. https://gcc.gnu.org/PR7416 12545 160. https://gcc.gnu.org/PR7570 12546 161. https://gcc.gnu.org/PR7578 12547 162. https://gcc.gnu.org/PR7611 12548 163. https://gcc.gnu.org/PR7709 12549 164. https://gcc.gnu.org/PR7766 12550 165. https://gcc.gnu.org/PR7785 12551 166. https://gcc.gnu.org/PR7786 12552 167. https://gcc.gnu.org/PR8142 12553 168. https://gcc.gnu.org/PR8234 12554 169. https://gcc.gnu.org/PR8415 12555 170. https://gcc.gnu.org/PR8481 12556 171. https://gcc.gnu.org/PR8593 12557 172. https://gcc.gnu.org/PR8759 12558 173. https://gcc.gnu.org/PR8997 12559 174. https://gcc.gnu.org/PR9253 12560 175. https://gcc.gnu.org/PR9254 12561 176. https://gcc.gnu.org/PR9271 12562 177. https://gcc.gnu.org/PR6767 12563 178. https://gcc.gnu.org/PR9911 12564 179. https://gcc.gnu.org/PR10020 12565 180. https://gcc.gnu.org/PR10546 12566 181. https://gcc.gnu.org/PR7029 12567 182. https://gcc.gnu.org/PR2903 12568 183. https://gcc.gnu.org/PR7873 12569 184. https://gcc.gnu.org/PR7680 12570 185. https://gcc.gnu.org/PR8705 12571 186. https://gcc.gnu.org/PR9986 12572 187. https://gcc.gnu.org/PR10056 12573 188. https://gcc.gnu.org/PR6744 12574 189. https://gcc.gnu.org/PR7361 12575 190. https://gcc.gnu.org/PR9496 12576 191. https://gcc.gnu.org/PR7067 12577 192. https://gcc.gnu.org/PR8480 12578 193. https://gcc.gnu.org/PR8784 12579 194. https://gcc.gnu.org/PR10315 12580 195. https://gcc.gnu.org/PR10267 12581 196. https://gcc.gnu.org/PR7916 12582 197. https://gcc.gnu.org/PR7926 12583 198. https://gcc.gnu.org/PR8555 12584 199. https://gcc.gnu.org/PR8994 12585 200. https://gcc.gnu.org/PR9426 12586 201. https://gcc.gnu.org/PR9806 12587 202. https://gcc.gnu.org/PR10077 12588 203. https://gcc.gnu.org/PR10233 12589 204. https://gcc.gnu.org/PR10286 12590 205. https://gcc.gnu.org/PR10308 12591 206. https://gcc.gnu.org/PR11272 12592 207. https://gcc.gnu.org/PR5754 12593 208. https://gcc.gnu.org/PR6597 12594 209. https://gcc.gnu.org/PR6949 12595 210. https://gcc.gnu.org/PR7053 12596 211. https://gcc.gnu.org/PR8164 12597 212. https://gcc.gnu.org/PR8384 12598 213. https://gcc.gnu.org/PR9559 12599 214. https://gcc.gnu.org/PR9649 12600 215. https://gcc.gnu.org/PR9864 12601 216. https://gcc.gnu.org/PR10432 12602 217. https://gcc.gnu.org/PR10475 12603 218. https://gcc.gnu.org/PR10635 12604 219. https://gcc.gnu.org/PR10661 12605 220. https://gcc.gnu.org/PR10700 12606 221. https://gcc.gnu.org/PR10712 12607 222. https://gcc.gnu.org/PR10796 12608 223. https://gcc.gnu.org/PR10890 12609 224. https://gcc.gnu.org/PR10939 12610 225. https://gcc.gnu.org/PR10956 12611 226. https://gcc.gnu.org/PR11041 12612 227. https://gcc.gnu.org/PR11059 12613 228. https://gcc.gnu.org/PR11083 12614 229. https://gcc.gnu.org/PR11105 12615 230. https://gcc.gnu.org/PR11149 12616 231. https://gcc.gnu.org/PR11228 12617 232. https://gcc.gnu.org/PR11282 12618 233. https://gcc.gnu.org/PR11301 12619 234. https://gcc.gnu.org/PR11308 12620 235. https://gcc.gnu.org/PR11473 12621 236. https://gcc.gnu.org/PR11503 12622 237. https://gcc.gnu.org/PR11513 12623 238. https://gcc.gnu.org/PR11198 12624 239. https://gcc.gnu.org/PR11304 12625 240. https://gcc.gnu.org/PR11381 12626 241. https://gcc.gnu.org/PR11536 12627 242. https://gcc.gnu.org/PR11557 12628 243. https://gcc.gnu.org/PR5897 12629 244. https://gcc.gnu.org/PR11279 12630 245. https://gcc.gnu.org/PR11022 12631 246. https://gcc.gnu.org/PR2330 12632 247. https://gcc.gnu.org/PR5388 12633 248. https://gcc.gnu.org/PR5390 12634 249. https://gcc.gnu.org/PR7877 12635 250. https://gcc.gnu.org/PR9393 12636 251. https://gcc.gnu.org/PR10032 12637 252. https://gcc.gnu.org/PR10468 12638 253. https://gcc.gnu.org/PR10527 12639 254. https://gcc.gnu.org/PR10679 12640 255. https://gcc.gnu.org/PR10682 12641 256. https://gcc.gnu.org/PR10689 12642 257. https://gcc.gnu.org/PR10845 12643 258. https://gcc.gnu.org/PR10849 12644 259. https://gcc.gnu.org/PR10888 12645 260. https://gcc.gnu.org/PR10929 12646 261. https://gcc.gnu.org/PR10931 12647 262. https://gcc.gnu.org/PR10940 12648 263. https://gcc.gnu.org/PR10968 12649 264. https://gcc.gnu.org/PR10990 12650 265. https://gcc.gnu.org/PR11039 12651 266. https://gcc.gnu.org/PR11062 12652 267. https://gcc.gnu.org/PR11095 12653 268. https://gcc.gnu.org/PR11098 12654 269. https://gcc.gnu.org/PR11137 12655 270. https://gcc.gnu.org/PR11154 12656 271. https://gcc.gnu.org/PR11329 12657 272. https://gcc.gnu.org/PR11332 12658 273. https://gcc.gnu.org/PR11431 12659 274. https://gcc.gnu.org/PR11528 12660 275. https://gcc.gnu.org/PR11546 12661 276. https://gcc.gnu.org/PR11567 12662 277. https://gcc.gnu.org/PR11645 12663 278. https://gcc.gnu.org/PR5179 12664 279. https://gcc.gnu.org/PR8204 12665 280. https://gcc.gnu.org/PR10838 12666 281. https://gcc.gnu.org/PR10886 12667 282. https://gcc.gnu.org/PR11349 12668 283. https://gcc.gnu.org/PR4823 12669 284. https://gcc.gnu.org/PR8878 12670 285. https://gcc.gnu.org/PR9815 12671 286. https://gcc.gnu.org/PR10402 12672 287. https://gcc.gnu.org/PR10504 12673 288. https://gcc.gnu.org/PR10673 12674 289. https://gcc.gnu.org/PR11044 12675 290. https://gcc.gnu.org/PR11089 12676 291. https://gcc.gnu.org/PR11420 12677 292. https://gcc.gnu.org/PR9362 12678 293. https://gcc.gnu.org/PR10142 12679 294. https://gcc.gnu.org/PR10663 12680 295. https://gcc.gnu.org/PR10835 12681 296. https://gcc.gnu.org/PR10876 12682 297. https://gcc.gnu.org/PR10955 12683 298. https://gcc.gnu.org/PR11018 12684 299. https://gcc.gnu.org/PR11556 12685 300. https://gcc.gnu.org/PR10907 12686 301. https://gcc.gnu.org/PR11320 12687 302. https://gcc.gnu.org/PR11599 12688 303. https://gcc.gnu.org/PR9745 12689 304. https://gcc.gnu.org/PR10871 12690 305. https://gcc.gnu.org/PR11440 12691 306. https://gcc.gnu.org/PR7594 12692 307. https://gcc.gnu.org/PR10557 12693 308. https://gcc.gnu.org/PR11054 12694 309. https://gcc.gnu.org/PR10834 12695 310. https://gcc.gnu.org/PR10842 12696 311. https://gcc.gnu.org/PR11052 12697 312. https://gcc.gnu.org/PR11183 12698 313. https://gcc.gnu.org/PR11084 12699 314. https://gcc.gnu.org/PR10331 12700 315. https://gcc.gnu.org/PR10413 12701 316. https://gcc.gnu.org/PR11096 12702 317. https://gcc.gnu.org/PR2873 12703 318. https://gcc.gnu.org/PR3163 12704 319. https://gcc.gnu.org/PR5287 12705 320. https://gcc.gnu.org/PR10148 12706 321. https://gcc.gnu.org/PR8787 12707 322. https://gcc.gnu.org/PR10900 12708 323. https://gcc.gnu.org/PR1607 12709 324. https://gcc.gnu.org/PR4252 12710 325. https://gcc.gnu.org/PR4490 12711 326. https://gcc.gnu.org/PR10355 12712 327. https://gcc.gnu.org/PR10726 12713 328. https://gcc.gnu.org/PR10805 12714 329. https://gcc.gnu.org/PR10815 12715 330. https://gcc.gnu.org/PR10877 12716 331. https://gcc.gnu.org/PR11280 12717 332. https://gcc.gnu.org/PR11466 12718 333. https://gcc.gnu.org/PR10737 12719 334. https://gcc.gnu.org/PR10810 12720 335. https://gcc.gnu.org/PR8336 12721 336. https://gcc.gnu.org/PR9330 12722 337. https://gcc.gnu.org/PR9631 12723 338. https://gcc.gnu.org/PR9877 12724 339. https://gcc.gnu.org/PR11687 12725 340. https://gcc.gnu.org/PR12263 12726 341. https://gcc.gnu.org/PR12490 12727 342. https://gcc.gnu.org/PR7277 12728 343. https://gcc.gnu.org/PR7939 12729 344. https://gcc.gnu.org/PR11063 12730 345. https://gcc.gnu.org/PR11207 12731 346. https://gcc.gnu.org/PR11522 12732 347. https://gcc.gnu.org/PR11595 12733 348. https://gcc.gnu.org/PR11646 12734 349. https://gcc.gnu.org/PR11665 12735 350. https://gcc.gnu.org/PR11852 12736 351. https://gcc.gnu.org/PR11878 12737 352. https://gcc.gnu.org/PR11883 12738 353. https://gcc.gnu.org/PR11991 12739 354. https://gcc.gnu.org/PR12146 12740 355. https://gcc.gnu.org/PR12215 12741 356. https://gcc.gnu.org/PR12369 12742 357. https://gcc.gnu.org/PR12446 12743 358. https://gcc.gnu.org/PR12510 12744 359. https://gcc.gnu.org/PR12544 12745 360. https://gcc.gnu.org/PR9862 12746 361. https://gcc.gnu.org/PR10962 12747 362. https://gcc.gnu.org/PR11370 12748 363. https://gcc.gnu.org/PR11637 12749 364. https://gcc.gnu.org/PR11885 12750 365. https://gcc.gnu.org/PR12082 12751 366. https://gcc.gnu.org/PR12180 12752 367. https://gcc.gnu.org/PR12340 12753 368. https://gcc.gnu.org/PR3907 12754 369. https://gcc.gnu.org/PR5293 12755 370. https://gcc.gnu.org/PR5296 12756 371. https://gcc.gnu.org/PR7939 12757 372. https://gcc.gnu.org/PR8656 12758 373. https://gcc.gnu.org/PR10147 12759 374. https://gcc.gnu.org/PR11400 12760 375. https://gcc.gnu.org/PR11409 12761 376. https://gcc.gnu.org/PR11740 12762 377. https://gcc.gnu.org/PR11786 12763 378. https://gcc.gnu.org/PR11867 12764 379. https://gcc.gnu.org/PR11928 12765 380. https://gcc.gnu.org/PR12114 12766 381. https://gcc.gnu.org/PR12163 12767 382. https://gcc.gnu.org/PR12181 12768 383. https://gcc.gnu.org/PR12236 12769 384. https://gcc.gnu.org/PR12266 12770 385. https://gcc.gnu.org/PR12296 12771 386. https://gcc.gnu.org/PR12298 12772 387. https://gcc.gnu.org/PR12369 12773 388. https://gcc.gnu.org/PR12337 12774 389. https://gcc.gnu.org/PR12344 12775 390. https://gcc.gnu.org/PR12451 12776 391. https://gcc.gnu.org/PR12486 12777 392. https://gcc.gnu.org/PR8869 12778 393. https://gcc.gnu.org/PR9786 12779 394. https://gcc.gnu.org/PR11689 12780 395. https://gcc.gnu.org/PR12116 12781 396. https://gcc.gnu.org/PR12070 12782 397. https://gcc.gnu.org/PR11184 12783 398. https://gcc.gnu.org/PR11535 12784 399. https://gcc.gnu.org/PR11693 12785 400. https://gcc.gnu.org/PR12224 12786 401. https://gcc.gnu.org/PR11087 12787 402. https://gcc.gnu.org/PR11319 12788 403. https://gcc.gnu.org/PR11949 12789 404. https://gcc.gnu.org/PR11662 12790 405. https://gcc.gnu.org/PR11965 12791 406. https://gcc.gnu.org/PR12301 12792 407. https://gcc.gnu.org/PR11717 12793 408. https://gcc.gnu.org/PR11313 12794 409. https://gcc.gnu.org/PR11712 12795 410. https://gcc.gnu.org/PR12166 12796 411. https://gcc.gnu.org/PR12101 12797 412. https://gcc.gnu.org/PR10988 12798 413. https://gcc.gnu.org/PR11805 12799 414. https://gcc.gnu.org/PR11902 12800 415. https://gcc.gnu.org/PR11903 12801 416. https://gcc.gnu.org/PR11890 12802 417. https://gcc.gnu.org/PR12399 12803 418. https://gcc.gnu.org/PR13068 12804 419. https://gcc.gnu.org/PR10060 12805 420. https://gcc.gnu.org/PR10555 12806 421. https://gcc.gnu.org/PR10706 12807 422. https://gcc.gnu.org/PR11496 12808 423. https://gcc.gnu.org/PR11741 12809 424. https://gcc.gnu.org/PR12440 12810 425. https://gcc.gnu.org/PR12632 12811 426. https://gcc.gnu.org/PR12712 12812 427. https://gcc.gnu.org/PR12726 12813 428. https://gcc.gnu.org/PR12890 12814 429. https://gcc.gnu.org/PR12900 12815 430. https://gcc.gnu.org/PR13060 12816 431. https://gcc.gnu.org/PR13289 12817 432. https://gcc.gnu.org/PR13318 12818 433. https://gcc.gnu.org/PR13392 12819 434. https://gcc.gnu.org/PR13574 12820 435. https://gcc.gnu.org/PR13475 12821 436. https://gcc.gnu.org/PR13797 12822 437. https://gcc.gnu.org/PR13824 12823 438. https://gcc.gnu.org/PR8776 12824 439. https://gcc.gnu.org/PR10339 12825 440. https://gcc.gnu.org/PR11350 12826 441. https://gcc.gnu.org/PR12826 12827 442. https://gcc.gnu.org/PR12500 12828 443. https://gcc.gnu.org/PR12941 12829 444. https://gcc.gnu.org/PR12953 12830 445. https://gcc.gnu.org/PR13041 12831 446. https://gcc.gnu.org/PR13507 12832 447. https://gcc.gnu.org/PR13382 12833 448. https://gcc.gnu.org/PR13394 12834 449. https://gcc.gnu.org/PR13400 12835 450. https://gcc.gnu.org/PR13521 12836 451. https://gcc.gnu.org/PR2094 12837 452. https://gcc.gnu.org/PR2294 12838 453. https://gcc.gnu.org/PR5050 12839 454. https://gcc.gnu.org/PR9371 12840 455. https://gcc.gnu.org/PR9546 12841 456. https://gcc.gnu.org/PR10081 12842 457. https://gcc.gnu.org/PR10093 12843 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 12844 459. https://gcc.gnu.org/PR10095 12845 460. https://gcc.gnu.org/PR11554 12846 461. https://gcc.gnu.org/PR12297 12847 462. https://gcc.gnu.org/PR12352 12848 463. https://gcc.gnu.org/PR12438 12849 464. https://gcc.gnu.org/PR12540 12850 465. https://gcc.gnu.org/PR12594 12851 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 12852 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 12853 468. https://gcc.gnu.org/PR12657 12854 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 12855 470. https://gcc.gnu.org/PR12696 12856 471. https://gcc.gnu.org/PR12815 12857 472. https://gcc.gnu.org/PR12862 12858 473. https://gcc.gnu.org/PR12926 12859 474. https://gcc.gnu.org/PR12967 12860 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 12861 476. https://gcc.gnu.org/PR12971 12862 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 12863 478. https://gcc.gnu.org/PR13007 12864 479. https://gcc.gnu.org/PR13009 12865 480. https://gcc.gnu.org/PR13057 12866 481. https://gcc.gnu.org/PR13070 12867 482. https://gcc.gnu.org/PR13081 12868 483. https://gcc.gnu.org/PR13239 12869 484. https://gcc.gnu.org/PR13262 12870 485. https://gcc.gnu.org/PR13290 12871 486. https://gcc.gnu.org/PR13323 12872 487. https://gcc.gnu.org/PR13369 12873 488. https://gcc.gnu.org/PR13371 12874 489. https://gcc.gnu.org/PR13445 12875 490. https://gcc.gnu.org/PR13461 12876 491. https://gcc.gnu.org/PR13462 12877 492. https://gcc.gnu.org/PR13478 12878 493. https://gcc.gnu.org/PR13544 12879 494. https://gcc.gnu.org/PR13650 12880 495. https://gcc.gnu.org/PR13683 12881 496. https://gcc.gnu.org/PR13688 12882 497. https://gcc.gnu.org/PR13774 12883 498. https://gcc.gnu.org/PR13884 12884 499. https://gcc.gnu.org/PR10746 12885 500. https://gcc.gnu.org/PR11433 12886 501. https://gcc.gnu.org/PR12633 12887 502. https://gcc.gnu.org/PR13037 12888 503. https://gcc.gnu.org/PR13213 12889 504. https://gcc.gnu.org/PR4490 12890 505. https://gcc.gnu.org/PR12292 12891 506. https://gcc.gnu.org/PR12441 12892 507. https://gcc.gnu.org/PR12943 12893 508. https://gcc.gnu.org/PR13608 12894 509. https://gcc.gnu.org/PR11598 12895 510. https://gcc.gnu.org/PR11793 12896 511. https://gcc.gnu.org/PR12467 12897 512. https://gcc.gnu.org/PR12537 12898 513. https://gcc.gnu.org/PR12496 12899 514. https://gcc.gnu.org/PR12865 12900 515. https://gcc.gnu.org/PR13354 12901 516. https://gcc.gnu.org/PR10467 12902 517. https://gcc.gnu.org/PR11226 12903 518. https://gcc.gnu.org/PR11227 12904 519. https://gcc.gnu.org/PR12644 12905 520. https://gcc.gnu.org/PR13149 12906 521. https://gcc.gnu.org/PR12654 12907 522. https://gcc.gnu.org/PR12965 12908 523. https://gcc.gnu.org/PR13031 12909 524. https://gcc.gnu.org/PR11634 12910 525. https://gcc.gnu.org/PR12158 12911 526. https://gcc.gnu.org/PR11992 12912 527. https://gcc.gnu.org/PR9365 12913 528. https://gcc.gnu.org/PR10392 12914 529. https://gcc.gnu.org/PR11322 12915 530. https://gcc.gnu.org/PR13069 12916 531. https://gcc.gnu.org/PR13302 12917 532. https://gcc.gnu.org/PR13585 12918 533. https://gcc.gnu.org/PR8916 12919 534. https://gcc.gnu.org/PR11576 12920 535. https://gcc.gnu.org/PR13122 12921 536. https://gcc.gnu.org/PR13256 12922 537. https://gcc.gnu.org/PR13373 12923 538. https://gcc.gnu.org/PR12561 12924 539. https://gcc.gnu.org/PR6243 12925 540. https://gcc.gnu.org/PR11397 12926 541. https://gcc.gnu.org/PR12505 12927 542. https://gcc.gnu.org/PR13150 12928 543. https://gcc.gnu.org/PR12666 12929 544. https://gcc.gnu.org/PR12969 12930 545. https://gcc.gnu.org/PR10819 12931 546. https://gcc.gnu.org/PR11612 12932 547. https://gcc.gnu.org/PR13211 12933 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 12934 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 12935 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 12936 551. https://gcc.gnu.org/onlinedocs/ 12937 552. mailto:gcc-help@gcc.gnu.org 12938 553. mailto:gcc@gcc.gnu.org 12939 554. https://gcc.gnu.org/lists.html 12940 555. http://www.fsf.org/ 12941 556. https://gcc.gnu.org/about.html 12942 557. http://validator.w3.org/check/referer 12943====================================================================== 12944http://gcc.gnu.org/gcc-3.2/index.html 12945 12946 GCC 3.2 Release Series 12947 12948 April 25, 2003 12949 12950 The [1]GNU project and the GCC developers are pleased to announce the 12951 release of GCC 3.2.3. 12952 12953 The purpose of the GCC 3.2 release series is to provide a stable 12954 platform for OS distributors to use building their next releases. A 12955 primary objective was to stabilize the C++ ABI; we believe that the 12956 interface to the compiler and the C++ standard library are now 12957 relatively stable. 12958 12959 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 12960 interoperate with code compiled by GCC 3.1.1 or earlier. 12961 12962 Please refer to our [2]detailed list of news, caveats, and bug-fixes 12963 for further information. 12964 12965Release History 12966 12967 GCC 3.2.3 12968 April 25, 2003 ([3]changes) 12969 12970 GCC 3.2.2 12971 February 5, 2003 ([4]changes) 12972 12973 GCC 3.2.1 12974 November 19, 2002 ([5]changes) 12975 12976 GCC 3.2 12977 August 14, 2002 ([6]changes) 12978 12979References and Acknowledgements 12980 12981 GCC used to stand for the GNU C Compiler, but since the compiler 12982 supports several other languages aside from C, it now stands for the 12983 GNU Compiler Collection. 12984 12985 A list of [7]successful builds is updated as new information becomes 12986 available. 12987 12988 The GCC developers would like to thank the numerous people that have 12989 contributed new features, improvements, bug fixes, and other changes as 12990 well as test results to GCC. This [8]amazing group of volunteers is 12991 what makes GCC successful. 12992 12993 For additional information about GCC please refer to the [9]GCC project 12994 web site or contact the [10]GCC development mailing list. 12995 12996 To obtain GCC please use [11]our mirror sites, or our CVS server. 12997 12998 12999 For questions related to the use of GCC, please consult these web 13000 pages and the [12]GCC manuals. If that fails, the 13001 [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these 13002 web pages and the development of GCC are welcome on our developer 13003 list at [14]gcc@gcc.gnu.org. All of [15]our lists have public 13004 archives. 13005 13006 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 13007 distribution of this entire article is permitted in any medium, 13008 provided this notice is preserved. 13009 13010 These pages are [17]maintained by the GCC team. Last modified 13011 2016-01-30[18]. 13012 13013References 13014 13015 1. http://www.gnu.org/ 13016 2. http://gcc.gnu.org/gcc-3.2/changes.html 13017 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 13018 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 13019 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 13020 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 13021 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 13022 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13023 9. http://gcc.gnu.org/index.html 13024 10. mailto:gcc@gcc.gnu.org 13025 11. http://gcc.gnu.org/mirrors.html 13026 12. https://gcc.gnu.org/onlinedocs/ 13027 13. mailto:gcc-help@gcc.gnu.org 13028 14. mailto:gcc@gcc.gnu.org 13029 15. https://gcc.gnu.org/lists.html 13030 16. http://www.fsf.org/ 13031 17. https://gcc.gnu.org/about.html 13032 18. http://validator.w3.org/check/referer 13033====================================================================== 13034http://gcc.gnu.org/gcc-3.2/changes.html 13035 13036 GCC 3.2 Release Series 13037 Changes, New Features, and Fixes 13038 13039 The latest release in the 3.2 release series is [1]GCC 3.2.3. 13040 13041Caveats and New Features 13042 13043 Caveats 13044 13045 * The C++ compiler does not correctly zero-initialize 13046 pointers-to-data members. You must explicitly initialize them. For 13047 example: int S::*m(0); will work, but depending on 13048 default-initialization to zero will not work. This bug cannot be 13049 fixed in GCC 3.2 without inducing unacceptable risks. It will be 13050 fixed in GCC 3.3. 13051 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 13052 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 13053 a number of C++ ABI fixes which make its C++ compiler generate 13054 binary code which is incompatible with the C++ compilers found in 13055 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 13056 13057 Frontend Enhancements 13058 13059 C/C++/Objective-C 13060 13061 * The method of constructing the list of directories to be searched 13062 for header files has been revised. If a directory named by a -I 13063 option is a standard system include directory, the option is 13064 ignored to ensure that the default search order for system 13065 directories and the special treatment of system header files are 13066 not defeated. 13067 * The C and Objective-C compilers no longer accept the "Naming Types" 13068 extension (typedef foo = bar); it was already unavailable in C++. 13069 Code which uses it will need to be changed to use the "typeof" 13070 extension instead: typedef typeof(bar) foo. (We have removed this 13071 extension without a period of deprecation because it has caused the 13072 compiler to crash since version 3.0 and no one noticed until very 13073 recently. Thus we conclude it is not in widespread use.) 13074 13075 C++ 13076 13077 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 13078 in GCC and the multi-vendor standard, but more have been found 13079 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 13080 code which is affected by these bugs. We will fix these bugs in 13081 some future release, once we are confident that all have been 13082 found; until then, it is our intention to make changes to the ABI 13083 only if they are necessary for correct compilation of C++, as 13084 opposed to conformance to the ABI documents. 13085 * For details on how to build an ABI compliant compiler for GNU/Linux 13086 systems, check the [3]common C++ ABI page. 13087 13088 New Targets and Target Specific Improvements 13089 13090 IA-32 13091 13092 * Fixed a number of bugs in SSE and MMX intrinsics. 13093 * Fixed common compiler crashes with SSE instruction set enabled 13094 (implied by -march=pentium3, pentium4, athlon-xp) 13095 * __m128 and __m128i is not 128bit aligned when used in structures. 13096 13097 x86-64 13098 13099 * A bug whereby the compiler could generate bad code for bzero has 13100 been fixed. 13101 * ABI fixes (implying ABI incompatibilities with previous version in 13102 some corner cases) 13103 * Fixed prefetch code generation 13104 __________________________________________________________________ 13105 13106GCC 3.2.3 13107 13108 3.2.3 is a bug fix release only; there are no new features that were 13109 not present in GCC 3.2.2. 13110 13111 Bug Fixes 13112 13113 This section lists the problem reports (PRs) from GCC's bug tracking 13114 system that are known to be fixed in the 3.2.3 release. This list might 13115 not be complete (that is, it is possible that some PRs that have been 13116 fixed are not listed here), and some of the titles have been changed to 13117 make them more clear. 13118 13119 Internal Compiler Errors (multi-platform) 13120 13121 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 13122 cc1plus 13123 * [5]6440: (c++) template specializations cause ICE 13124 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 13125 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 13126 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 13127 * [9]8068: exceedingly high (infinite) memory usage 13128 * [10]8178: ICE with __builtin_ffs 13129 * [11]8396: ICE in copy_to_mode_reg, in explow.c 13130 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 13131 * [13]9768: ICE when optimizing inline code at -O2 13132 * [14]9798: (c++) Infinite recursion (segfault) in 13133 cp/decl.c:push_using_directive with recursive using directives 13134 * [15]9799: mismatching structure initializer with nested flexible 13135 array member: ICE 13136 * [16]9928: ICE on duplicate enum declaration 13137 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 13138 sparc, alpha) 13139 * [18]10352: ICE in find_reloads_toplev 13140 * [19]10336: ICE with -Wunreachable-code 13141 13142 C/optimizer bugs: 13143 13144 * [20]8224: Incorrect joining of signed and unsigned division 13145 * [21]8613: -O2 produces wrong code with builtin strlen and 13146 postincrements 13147 * [22]8828: gcc reports some code is unreachable when it is not 13148 * [23]9226: GCSE breaking argument passing 13149 * [24]9853: miscompilation of non-constant structure initializer 13150 * [25]9797: C99-style struct initializers are miscompiled 13151 * [26]9967: Some standard C function calls should not be replaced 13152 when optimizing for size 13153 * [27]10116: ce2: invalid merge of join_bb in the context of switch 13154 statements 13155 * [28]10171: wrong code for inlined function 13156 * [29]10175: -Wunreachable-code doesn't work for single lines 13157 13158 C++ compiler and library: 13159 13160 * [30]8316: Confusing diagnostic for code that misuses conversion 13161 operators 13162 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 13163 * [32]9420: incomplete type incorrectly reported 13164 * [33]9459: typeof in return type specification of template not 13165 supported 13166 * [34]9507: filebuf::open handles ios_base::ate incorrectly 13167 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 13168 * [36]9602: Total confusion about template/friend/virtual/abstract 13169 * [37]9993: destructor not called for local object created within and 13170 returned from infinite loop 13171 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 13172 system 13173 13174 Java compiler and library: 13175 13176 * [39]9652: libgcj build fails on irix6.5.1[78] 13177 * [40]10144: gas on solaris complains about bad .stabs lines for 13178 java, native as unaffected 13179 13180 x86-specific (Intel/AMD): 13181 13182 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 13183 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 13184 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 13185 failed 13186 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 13187 failed 13188 13189 SPARC-specific: 13190 13191 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 13192 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 13193 unroll.c 13194 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 13195 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 13196 execute/loop-2d.c 13197 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 13198 * [50]9414: Scheduling bug on Ultrasparc 13199 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 13200 13201 m68k-specific: 13202 13203 * [52]7248: broken "inclusive or" code 13204 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 13205 13206 PowerPC-specific: 13207 13208 * [54]9732: Wrong code with -O2 -fPIC 13209 * [55]10073: ICE: powerpc cannot split insn 13210 13211 Alpha-specific: 13212 13213 * [56]7702: optimization problem on a DEC alpha under OSF1 13214 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 13215 13216 HP-specific: 13217 13218 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 13219 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 13220 (missing symbol) 13221 * [60]10271: Floating point args don't get reloaded across function 13222 calls with -O2 13223 13224 MIPS specific: 13225 13226 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 13227 compile/920501-4.c 13228 13229 CRIS specific: 13230 13231 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 13232 13233 Miscellaneous and minor bugs: 13234 13235 * [63]6955: collect2 says "core dumped" when there is no core 13236 __________________________________________________________________ 13237 13238GCC 3.2.2 13239 13240 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 13241 install by means of the DESTDIR variable. Parts of the GCC tree have 13242 featured that support long before, but now it is available even from 13243 the top level. 13244 13245 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 13246 features that were not present in GCC 3.2.1. 13247 13248 Bug Fixes 13249 13250 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 13251 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 13252 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 13253 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 13254 change, and thus restores ABI-compatibility with previous releases 13255 (except GCC 3.2.1) on these platforms. 13256 13257 This section lists the problem reports (PRs) from GCC's bug tracking 13258 system that are known to be fixed in the 3.2.2 release. This list might 13259 not be complete (that is, it is possible that some PRs that have been 13260 fixed are not listed here) and some of the titles have been changed to 13261 make them more clear. 13262 13263 Internal Compiler Errors (multi-platform) 13264 13265 * [64]5919: (c++) ICE when passing variable array to template 13266 function 13267 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 13268 * [66]7507: ICE with -O2 when address of called function is a 13269 complicated expression 13270 * [67]7622: ICE with nested inline functions if function's address is 13271 taken 13272 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 13273 [69]9258) 13274 * [70]8031: (c++) ICE in code comparing typeids and casting from 13275 virtual base 13276 * [71]8275: ICE in simplify_subreg 13277 * [72]8332: (c++) builtin strlen/template interaction causes ICE 13278 * [73]8372: (c++) ICE on explicit call of destructor 13279 * [74]8439: (c, not c++) empty struct causes ICE 13280 * [75]8442: (c++) ICE with nested template classes 13281 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 13282 * [77]8615: (c++) ICE with out-of-range character constant template 13283 argument 13284 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 13285 * [79]8799: (c++) ICE: error reporting routines re-entered 13286 * [80]9328: (c++) ICE with typeof(X) for overloaded X 13287 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 13288 13289 C++ (compiler and library) bugs 13290 13291 * [82]47: scoping in nested classes is broken 13292 * [83]6745: problems with iostream rdbuf() member function 13293 * [84]8214: conversion from const char* const to char* sometimes 13294 accepted illegally 13295 * [85]8493: builtin strlen and overload resolution (same bug as 13296 [86]8332) 13297 * [87]8503: strange behaviour of function types 13298 * [88]8727: compiler confused by inheritance from an anonymous struct 13299 * [89]7445: poor performance of std::locale::classic() in 13300 multi-threaded applications 13301 * [90]8230: mishandling of overflow in vector<T>::resize 13302 * [91]8399: sync_with_stdio(false) breaks unformatted input 13303 * [92]8662: illegal access of private member of unnamed class is 13304 accepted 13305 * [93]8707: "make distclean" fails in libstdc++-v3 directory 13306 * [94]8708: __USE_MALLOC doesn't work 13307 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 13308 * [96]8887: Bug in date formats with --enable-clocale=generic 13309 * [97]9076: Call Frame Instructions are not handled correctly during 13310 unwind operation 13311 * [98]9151: std::setprecision limited to 16 digits when outputting a 13312 double to a stream 13313 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 13314 * [100]9269: libstdc++ headers: explicit specialization of function 13315 must precede its first use 13316 * [101]9322: return value of basic_streambuf<>::getloc affected by 13317 locale::global 13318 * [102]9433: segfault in runtime support for dynamic_cast 13319 13320 C and optimizer bugs 13321 13322 * [103]8032: GCC incorrectly initializes static structs that have 13323 flexible arrays 13324 * [104]8639: simple arithmetic expression broken 13325 * [105]8794: optimization improperly eliminates certain expressions 13326 * [106]8832: traditional "asm volatile" code is illegally optimized 13327 * [107]8988: loop optimizer bug: with -O2, code is generated that 13328 segfaults (found on i386, bug present for all platforms) 13329 * [108]9492: structure copy clobbers subsequent stores to structure 13330 13331 Objective-C bugs 13332 13333 * [109]9267: Objective-C parser won't build with newer bison versions 13334 (e.g. 1.875) 13335 13336 Ada bugs 13337 13338 * [110]8344: Ada build problem due to conflict between gcc/final.o, 13339 gcc/ada/final.o 13340 13341 Preprocessor bugs 13342 13343 * [111]8524: _Pragma within macros is improperly expanded 13344 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 13345 -fshort-wchar 13346 13347 ARM-specific 13348 13349 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 13350 13351 x86-specific (Intel/AMD) 13352 13353 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 13354 * [115]8599: loop unroll bug with -march=k6-3 13355 * [116]9506: ABI breakage in structure return (affects BSD and 13356 Cygwin, but not GNU/Linux) 13357 13358 FreeBSD 5.0 specific 13359 13360 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 13361 13362 RTEMS-specific 13363 13364 * [118]9292: hppa1.1-rtems configurery problems 13365 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 13366 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 13367 * [121]9296: gthr-rtems regression 13368 * [122]9316: powerpc-rtems: extending multilibs 13369 13370 HP-PA specific 13371 13372 * [123]9493: ICE with -O2 when building a simple function 13373 13374 Documentation 13375 13376 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 13377 * [125]8947: Please add a warning about "-malign-double" in docs 13378 * [126]7448, [127]8882: typo cleanups 13379 __________________________________________________________________ 13380 13381GCC 3.2.1 13382 13383 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 13384 generates code that is known not to be binary-compatible with the 13385 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 13386 in the distribution, for details. 13387 13388 This release also removes an old GCC extension, "naming types", and the 13389 documentation now directs users to use a different GCC extension, 13390 __typeof__, instead. The feature had evidently been broken for a while. 13391 13392 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 13393 the new warning there are no new features that were not present in GCC 13394 3.2. 13395 13396 In addition, the previous fix for [128]PR 7445 (poor performance of 13397 std::locale::classic() in multi-threaded applications) was reverted 13398 ("unfixed"), because the "fix" was not thread-safe. 13399 13400 Bug Fixes 13401 13402 This section lists the problem reports (PRs) from GCC's bug tracking 13403 system that are known to be fixed in the 3.2.1 release. This list might 13404 not be complete (that is, it is possible that some PRs that have been 13405 fixed are not listed here). As you can see, the number of bug fixes is 13406 quite large, so it is strongly recommended that users of earlier GCC 13407 3.x releases upgrade to GCC 3.2.1. 13408 13409 Internal Compiler Errors (multi-platform) 13410 13411 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 13412 * [130]5661: (c++) ICE instantiating template on array of unknown 13413 size (bad code) 13414 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 13415 64-bit platforms 13416 * [132]6994: (c++) ICE in find_function_data 13417 * [133]7150: preprocessor: GCC -dM -E gives an ICE 13418 * [134]7160: ICE when optimizing branches without a return value 13419 * [135]7228: (c++) ICE when using member template and template 13420 function 13421 * [136]7266: (c++) ICE with -pedantic on missing typename 13422 * [137]7353: ICE from use of "Naming Types" extension, see above 13423 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 13424 * [139]7478: (c++) ICE on static_cast inside template 13425 * [140]7526: preprocessor core dump when _Pragma implies #pragma 13426 dependency 13427 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 13428 is a duplicate) 13429 * [143]7754: (c++) ICE on union with template parameter 13430 * [144]7788: (c++) redeclaring a definition as an incomplete class 13431 causes ICE 13432 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 13433 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 13434 kernel 13435 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 13436 variables 13437 * [148]8134: (c++) ICE in force_store_init_value on legal code 13438 * [149]8149: (c++) ICE on incomplete type 13439 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 13440 initialization 13441 13442 C++ (compiler and library) bugs 13443 13444 * [151]5607: No pointer adjustment in covariant return types 13445 * [152]6579: Infinite loop with statement expressions in member 13446 initialization 13447 * [153]6803: Default copy constructor bug in GCC 3.1 13448 * [154]7176: g++ confused by friend and static member with same name 13449 * [155]7188: Segfault with template class and recursive (incorrect) 13450 initializer list 13451 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 13452 inheritance if a method has a variable number of arguments 13453 * [157]7461: ctype<char>::classic_table() returns offset array on 13454 Cygwin 13455 * [158]7524: f(const float arg[3]) fails 13456 * [159]7584: Erroneous ambiguous base error on using declaration 13457 * [160]7676: Member template overloading problem 13458 * [161]7679: infinite loop when a right parenthesis is missing 13459 * [162]7811: default locale not taken from environment 13460 * [163]7961: compare( char *) implemented incorrectly in 13461 basic_string<> 13462 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 13463 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 13464 [166]6745) 13465 * [167]8096: deque::at() throws std::range_error instead of 13466 std::out_of_range 13467 * [168]8127: cout << cin.rdbuf() infinite loop 13468 * [169]8218: Excessively large memory consumed for classes with large 13469 array members 13470 * [170]8287: GCC 3.2: Destructor called for non-constructed local 13471 object 13472 * [171]8347: empty vector range used in string construction causes 13473 core dump 13474 * [172]8348: fail() flag is set in istringstream when eof() flag is 13475 set 13476 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 13477 13478 C and optimizer bugs 13479 13480 * [174]6627: -fno-align-functions doesn't seem to disable function 13481 alignment 13482 * [175]6631: life_analysis misoptimizes code to initialize fields of 13483 a structure 13484 * [176]7102: unsigned char division results in floating exception 13485 * [177]7120: Run once loop should *always* be unrolled 13486 (pessimization) 13487 * [178]7209: Bug involving array referencing and ?: operator 13488 * [179]7515: invalid inlining of global function with -O3 13489 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 13490 * [181]8467: bug in sibling call optimization 13491 13492 Preprocessor bugs 13493 13494 * [182]4890: incorrect line markers from the traditional preprocessor 13495 * [183]7357: -M option omits system headers files (making it the same 13496 as -MM) 13497 * [184]7358: Changes to Sun's make Dependencies 13498 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 13499 C headers 13500 * [186]7862: Interrupting GCC -MD removes .d file but not .o 13501 * [187]8190: Failed compilation deletes -MD dependency file 13502 * [188]8524: _Pragma within macro is improperly expanded 13503 13504 x86 specific (Intel/AMD) 13505 13506 * [189]5351: (i686-only) function pass-by-value structure copy 13507 corrupts stack ([190]7591 is a duplicate) 13508 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 13509 -march=pentium3/pentium2/athlon (these are all the same underlying 13510 bug, in MMX register use) 13511 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 13512 as above?) 13513 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 13514 * [199]6981: wrong code in 64-bit manipulation on x86 13515 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 13516 macro 13517 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 13518 intrinsics are broken 13519 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 13520 -march=pentium4 13521 * [203]7693: Typo in i386 mmintrin.h header 13522 * [204]7723: ICE - Pentium3 sse - GCC 3.2 13523 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 13524 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 13525 13526 PowerPC specific 13527 13528 * [207]5967: GCC bug when profiling nested functions on powerpc 13529 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 13530 loop on PowerPC 13531 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 13532 * [210]7130: miscompiled code for GCC-3.1 on 13533 powerpc-unknown-linux-gnu with -funroll-all-loops 13534 * [211]7133: PowerPC ICE: unrecognizable insn 13535 * [212]7380: ICE in extract_insn, at recog.c:2148 13536 * [213]8252: ICE on Altivec code with optimization turned on 13537 * [214]8451: Altivec ICE in GCC 3.2 13538 13539 HP/PA specific 13540 13541 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 13542 13543 SPARC specific 13544 13545 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 13546 in the wrong place on sparc-solaris 13547 * [217]7151: ICE when compiling for UltraSPARC 13548 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 13549 double and -O1 13550 * [219]7842: [REGRESSION] SPARC code gen bug 13551 13552 ARM specific 13553 13554 * [220]7856: [arm] invalid offset in constant pool reference 13555 * [221]7967: optimization produces wrong code (ARM) 13556 13557 Alpha specific 13558 13559 * [222]7374: __builtin_fabsl broken on alpha 13560 13561 IBM s390 specific 13562 13563 * [223]7370: ICE in fixup_var_refs_1 on s390x 13564 * [224]7409: loop optimization bug on s390x-linux-gnu 13565 * [225]8232: s390x: ICE when using bcmp with int length argument 13566 13567 SCO specific 13568 13569 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 13570 symbol: BITS_PER_UNIT 13571 13572 m68k/Coldfire specific 13573 13574 * [227]8314: crtbegin, crtend need to be multilib'ed for this 13575 platform 13576 13577 Documentation 13578 13579 * [228]761: Document some undocumented options 13580 * [229]5610: Fix documentation about invoking SSE instructions 13581 (-mfpmath=sse) 13582 * [230]7484: List -Wmissing-declarations as C-only option 13583 * [231]7531: -mcmodel not documented for x86-64 13584 * [232]8120: Update documentation of bad use of ## 13585 __________________________________________________________________ 13586 13587GCC 3.2 13588 13589 3.2 is a small bug fix release, but there is a change to the 13590 application binary interface (ABI), hence the change to the second part 13591 of the version number. 13592 13593 The main purpose of the 3.2 release is to correct a couple of problems 13594 in the C++ ABI, with the intention of providing a stable interface 13595 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 13596 13597 Bug Fixes 13598 13599 C++ 13600 13601 * [233]7320: g++ 3.2 relocation problem 13602 * [234]7470: vtable: virtual function pointers not in declaration 13603 order 13604 13605 libstdc++ 13606 13607 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 13608 * [236]6503, [237]6642, [238]7186: Problems with comparing or 13609 subtracting various types of const and non-const iterators 13610 * [239]7216: ambiguity with basic_iostream::traits_type 13611 * [240]7220: problem with basic_istream::ignore(0,delimiter) 13612 * [241]7222: locale::operator==() doesn't work on std::locale("") 13613 * [242]7286: placement operator delete issue 13614 * [243]7442: cxxabi.h does not match the C++ ABI 13615 * [244]7445: poor performance of std::locale::classic() in 13616 multi-threaded applications 13617 13618 x86-64 specific 13619 13620 * [245]7291: off-by-one in generated inline bzero code for x86-64 13621 13622 13623 For questions related to the use of GCC, please consult these web 13624 pages and the [246]GCC manuals. If that fails, the 13625 [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these 13626 web pages and the development of GCC are welcome on our developer 13627 list at [248]gcc@gcc.gnu.org. All of [249]our lists have public 13628 archives. 13629 13630 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 13631 distribution of this entire article is permitted in any medium, 13632 provided this notice is preserved. 13633 13634 These pages are [251]maintained by the GCC team. Last modified 13635 2016-01-30[252]. 13636 13637References 13638 13639 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 13640 2. http://gcc.gnu.org/gcc-3.1/changes.html 13641 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 13642 4. https://gcc.gnu.org/PR3782 13643 5. https://gcc.gnu.org/PR6440 13644 6. https://gcc.gnu.org/PR7050 13645 7. https://gcc.gnu.org/PR7741 13646 8. https://gcc.gnu.org/PR7982 13647 9. https://gcc.gnu.org/PR8068 13648 10. https://gcc.gnu.org/PR8178 13649 11. https://gcc.gnu.org/PR8396 13650 12. https://gcc.gnu.org/PR8674 13651 13. https://gcc.gnu.org/PR9768 13652 14. https://gcc.gnu.org/PR9798 13653 15. https://gcc.gnu.org/PR9799 13654 16. https://gcc.gnu.org/PR9928 13655 17. https://gcc.gnu.org/PR10114 13656 18. https://gcc.gnu.org/PR10352 13657 19. https://gcc.gnu.org/PR10336 13658 20. https://gcc.gnu.org/PR8224 13659 21. https://gcc.gnu.org/PR8613 13660 22. https://gcc.gnu.org/PR8828 13661 23. https://gcc.gnu.org/PR9226 13662 24. https://gcc.gnu.org/PR9853 13663 25. https://gcc.gnu.org/PR9797 13664 26. https://gcc.gnu.org/PR9967 13665 27. https://gcc.gnu.org/PR10116 13666 28. https://gcc.gnu.org/PR10171 13667 29. https://gcc.gnu.org/PR10175 13668 30. https://gcc.gnu.org/PR8316 13669 31. https://gcc.gnu.org/PR9169 13670 32. https://gcc.gnu.org/PR9420 13671 33. https://gcc.gnu.org/PR9459 13672 34. https://gcc.gnu.org/PR9507 13673 35. https://gcc.gnu.org/PR9538 13674 36. https://gcc.gnu.org/PR9602 13675 37. https://gcc.gnu.org/PR9993 13676 38. https://gcc.gnu.org/PR10167 13677 39. https://gcc.gnu.org/PR9652 13678 40. https://gcc.gnu.org/PR10144 13679 41. https://gcc.gnu.org/PR8746 13680 42. https://gcc.gnu.org/PR9888 13681 43. https://gcc.gnu.org/PR9638 13682 44. https://gcc.gnu.org/PR9954 13683 45. https://gcc.gnu.org/PR7784 13684 46. https://gcc.gnu.org/PR7796 13685 47. https://gcc.gnu.org/PR8281 13686 48. https://gcc.gnu.org/PR8366 13687 49. https://gcc.gnu.org/PR8726 13688 50. https://gcc.gnu.org/PR9414 13689 51. https://gcc.gnu.org/PR10067 13690 52. https://gcc.gnu.org/PR7248 13691 53. https://gcc.gnu.org/PR8343 13692 54. https://gcc.gnu.org/PR9732 13693 55. https://gcc.gnu.org/PR10073 13694 56. https://gcc.gnu.org/PR7702 13695 57. https://gcc.gnu.org/PR9671 13696 58. https://gcc.gnu.org/PR8694 13697 59. https://gcc.gnu.org/PR9953 13698 60. https://gcc.gnu.org/PR10271 13699 61. https://gcc.gnu.org/PR6362 13700 62. https://gcc.gnu.org/PR10377 13701 63. https://gcc.gnu.org/PR6955 13702 64. https://gcc.gnu.org/PR5919 13703 65. https://gcc.gnu.org/PR7129 13704 66. https://gcc.gnu.org/PR7507 13705 67. https://gcc.gnu.org/PR7622 13706 68. https://gcc.gnu.org/PR7681 13707 69. https://gcc.gnu.org/PR9528 13708 70. https://gcc.gnu.org/PR8031 13709 71. https://gcc.gnu.org/PR8275 13710 72. https://gcc.gnu.org/PR8332 13711 73. https://gcc.gnu.org/PR8372 13712 74. https://gcc.gnu.org/PR8439 13713 75. https://gcc.gnu.org/PR8442 13714 76. https://gcc.gnu.org/PR8518 13715 77. https://gcc.gnu.org/PR8615 13716 78. https://gcc.gnu.org/PR8663 13717 79. https://gcc.gnu.org/PR8799 13718 80. https://gcc.gnu.org/PR9328 13719 81. https://gcc.gnu.org/PR9465 13720 82. https://gcc.gnu.org/PR47 13721 83. https://gcc.gnu.org/PR6745 13722 84. https://gcc.gnu.org/PR8214 13723 85. https://gcc.gnu.org/PR8493 13724 86. https://gcc.gnu.org/PR8332 13725 87. https://gcc.gnu.org/PR8503 13726 88. https://gcc.gnu.org/PR8727 13727 89. https://gcc.gnu.org/PR7445 13728 90. https://gcc.gnu.org/PR8230 13729 91. https://gcc.gnu.org/PR8399 13730 92. https://gcc.gnu.org/PR8662 13731 93. https://gcc.gnu.org/PR8707 13732 94. https://gcc.gnu.org/PR8708 13733 95. https://gcc.gnu.org/PR8790 13734 96. https://gcc.gnu.org/PR8887 13735 97. https://gcc.gnu.org/PR9076 13736 98. https://gcc.gnu.org/PR9151 13737 99. https://gcc.gnu.org/PR9168 13738 100. https://gcc.gnu.org/PR9269 13739 101. https://gcc.gnu.org/PR9322 13740 102. https://gcc.gnu.org/PR9433 13741 103. https://gcc.gnu.org/PR8032 13742 104. https://gcc.gnu.org/PR8639 13743 105. https://gcc.gnu.org/PR8794 13744 106. https://gcc.gnu.org/PR8832 13745 107. https://gcc.gnu.org/PR8988 13746 108. https://gcc.gnu.org/PR9492 13747 109. https://gcc.gnu.org/PR9267 13748 110. https://gcc.gnu.org/PR8344 13749 111. https://gcc.gnu.org/PR8524 13750 112. https://gcc.gnu.org/PR8880 13751 113. https://gcc.gnu.org/PR9090 13752 114. https://gcc.gnu.org/PR8588 13753 115. https://gcc.gnu.org/PR8599 13754 116. https://gcc.gnu.org/PR9506 13755 117. https://gcc.gnu.org/PR9484 13756 118. https://gcc.gnu.org/PR9292 13757 119. https://gcc.gnu.org/PR9293 13758 120. https://gcc.gnu.org/PR9295 13759 121. https://gcc.gnu.org/PR9296 13760 122. https://gcc.gnu.org/PR9316 13761 123. https://gcc.gnu.org/PR9493 13762 124. https://gcc.gnu.org/PR7341 13763 125. https://gcc.gnu.org/PR8947 13764 126. https://gcc.gnu.org/PR7448 13765 127. https://gcc.gnu.org/PR8882 13766 128. https://gcc.gnu.org/PR7445 13767 129. https://gcc.gnu.org/PR2521 13768 130. https://gcc.gnu.org/PR5661 13769 131. https://gcc.gnu.org/PR6419 13770 132. https://gcc.gnu.org/PR6994 13771 133. https://gcc.gnu.org/PR7150 13772 134. https://gcc.gnu.org/PR7160 13773 135. https://gcc.gnu.org/PR7228 13774 136. https://gcc.gnu.org/PR7266 13775 137. https://gcc.gnu.org/PR7353 13776 138. https://gcc.gnu.org/PR7411 13777 139. https://gcc.gnu.org/PR7478 13778 140. https://gcc.gnu.org/PR7526 13779 141. https://gcc.gnu.org/PR7721 13780 142. https://gcc.gnu.org/PR7803 13781 143. https://gcc.gnu.org/PR7754 13782 144. https://gcc.gnu.org/PR7788 13783 145. https://gcc.gnu.org/PR8031 13784 146. https://gcc.gnu.org/PR8055 13785 147. https://gcc.gnu.org/PR8067 13786 148. https://gcc.gnu.org/PR8134 13787 149. https://gcc.gnu.org/PR8149 13788 150. https://gcc.gnu.org/PR8160 13789 151. https://gcc.gnu.org/PR5607 13790 152. https://gcc.gnu.org/PR6579 13791 153. https://gcc.gnu.org/PR6803 13792 154. https://gcc.gnu.org/PR7176 13793 155. https://gcc.gnu.org/PR7188 13794 156. https://gcc.gnu.org/PR7306 13795 157. https://gcc.gnu.org/PR7461 13796 158. https://gcc.gnu.org/PR7524 13797 159. https://gcc.gnu.org/PR7584 13798 160. https://gcc.gnu.org/PR7676 13799 161. https://gcc.gnu.org/PR7679 13800 162. https://gcc.gnu.org/PR7811 13801 163. https://gcc.gnu.org/PR7961 13802 164. https://gcc.gnu.org/PR8071 13803 165. https://gcc.gnu.org/PR8127 13804 166. https://gcc.gnu.org/PR6745 13805 167. https://gcc.gnu.org/PR8096 13806 168. https://gcc.gnu.org/PR8127 13807 169. https://gcc.gnu.org/PR8218 13808 170. https://gcc.gnu.org/PR8287 13809 171. https://gcc.gnu.org/PR8347 13810 172. https://gcc.gnu.org/PR8348 13811 173. https://gcc.gnu.org/PR8391 13812 174. https://gcc.gnu.org/PR6627 13813 175. https://gcc.gnu.org/PR6631 13814 176. https://gcc.gnu.org/PR7102 13815 177. https://gcc.gnu.org/PR7120 13816 178. https://gcc.gnu.org/PR7209 13817 179. https://gcc.gnu.org/PR7515 13818 180. https://gcc.gnu.org/PR7814 13819 181. https://gcc.gnu.org/PR8467 13820 182. https://gcc.gnu.org/PR4890 13821 183. https://gcc.gnu.org/PR7357 13822 184. https://gcc.gnu.org/PR7358 13823 185. https://gcc.gnu.org/PR7602 13824 186. https://gcc.gnu.org/PR7862 13825 187. https://gcc.gnu.org/PR8190 13826 188. https://gcc.gnu.org/PR8524 13827 189. https://gcc.gnu.org/PR5351 13828 190. https://gcc.gnu.org/PR7591 13829 191. https://gcc.gnu.org/PR6845 13830 192. https://gcc.gnu.org/PR7034 13831 193. https://gcc.gnu.org/PR7124 13832 194. https://gcc.gnu.org/PR7174 13833 195. https://gcc.gnu.org/PR7134 13834 196. https://gcc.gnu.org/PR7375 13835 197. https://gcc.gnu.org/PR7390 13836 198. https://gcc.gnu.org/PR6890 13837 199. https://gcc.gnu.org/PR6981 13838 200. https://gcc.gnu.org/PR7242 13839 201. https://gcc.gnu.org/PR7396 13840 202. https://gcc.gnu.org/PR7630 13841 203. https://gcc.gnu.org/PR7693 13842 204. https://gcc.gnu.org/PR7723 13843 205. https://gcc.gnu.org/PR7951 13844 206. https://gcc.gnu.org/PR8146 13845 207. https://gcc.gnu.org/PR5967 13846 208. https://gcc.gnu.org/PR6984 13847 209. https://gcc.gnu.org/PR7114 13848 210. https://gcc.gnu.org/PR7130 13849 211. https://gcc.gnu.org/PR7133 13850 212. https://gcc.gnu.org/PR7380 13851 213. https://gcc.gnu.org/PR8252 13852 214. https://gcc.gnu.org/PR8451 13853 215. https://gcc.gnu.org/PR7250 13854 216. https://gcc.gnu.org/PR6668 13855 217. https://gcc.gnu.org/PR7151 13856 218. https://gcc.gnu.org/PR7335 13857 219. https://gcc.gnu.org/PR7842 13858 220. https://gcc.gnu.org/PR7856 13859 221. https://gcc.gnu.org/PR7967 13860 222. https://gcc.gnu.org/PR7374 13861 223. https://gcc.gnu.org/PR7370 13862 224. https://gcc.gnu.org/PR7409 13863 225. https://gcc.gnu.org/PR8232 13864 226. https://gcc.gnu.org/PR7623 13865 227. https://gcc.gnu.org/PR8314 13866 228. https://gcc.gnu.org/PR761 13867 229. https://gcc.gnu.org/PR5610 13868 230. https://gcc.gnu.org/PR7484 13869 231. https://gcc.gnu.org/PR7531 13870 232. https://gcc.gnu.org/PR8120 13871 233. https://gcc.gnu.org/PR7320 13872 234. https://gcc.gnu.org/PR7470 13873 235. https://gcc.gnu.org/PR6410 13874 236. https://gcc.gnu.org/PR6503 13875 237. https://gcc.gnu.org/PR6642 13876 238. https://gcc.gnu.org/PR7186 13877 239. https://gcc.gnu.org/PR7216 13878 240. https://gcc.gnu.org/PR7220 13879 241. https://gcc.gnu.org/PR7222 13880 242. https://gcc.gnu.org/PR7286 13881 243. https://gcc.gnu.org/PR7442 13882 244. https://gcc.gnu.org/PR7445 13883 245. https://gcc.gnu.org/PR7291 13884 246. https://gcc.gnu.org/onlinedocs/ 13885 247. mailto:gcc-help@gcc.gnu.org 13886 248. mailto:gcc@gcc.gnu.org 13887 249. https://gcc.gnu.org/lists.html 13888 250. http://www.fsf.org/ 13889 251. https://gcc.gnu.org/about.html 13890 252. http://validator.w3.org/check/referer 13891====================================================================== 13892http://gcc.gnu.org/gcc-3.1/index.html 13893 13894 GCC 3.1 13895 13896 July 27, 2002 13897 13898 The [1]GNU project and the GCC developers are pleased to announce the 13899 release of GCC 3.1.1. 13900 13901 The links below still apply to GCC 3.1.1. 13902 13903 May 15, 2002 13904 13905 The [2]GNU project and the GCC developers are pleased to announce the 13906 release of GCC 3.1. 13907 13908 GCC used to stand for the GNU C Compiler, but since the compiler 13909 supports several other languages aside from C, it now stands for the 13910 GNU Compiler Collection. 13911 13912 A list of [3]successful builds is updated as new information becomes 13913 available. 13914 13915 The GCC developers would like to thank the numerous people that have 13916 contributed [4]new features, improvements, bug fixes, and other changes 13917 as well as test results to GCC. This [5]amazing group of volunteers is 13918 what makes GCC successful. 13919 13920 For additional information about GCC please refer to the [6]GCC project 13921 web site or contact the [7]GCC development mailing list. 13922 13923 To obtain GCC please use [8]our mirror sites, or our CVS server. 13924 __________________________________________________________________ 13925 13926 13927 For questions related to the use of GCC, please consult these web 13928 pages and the [9]GCC manuals. If that fails, the 13929 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these 13930 web pages and the development of GCC are welcome on our developer 13931 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public 13932 archives. 13933 13934 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 13935 distribution of this entire article is permitted in any medium, 13936 provided this notice is preserved. 13937 13938 These pages are [14]maintained by the GCC team. Last modified 13939 2016-01-30[15]. 13940 13941References 13942 13943 1. http://www.gnu.org/ 13944 2. http://www.gnu.org/ 13945 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 13946 4. http://gcc.gnu.org/gcc-3.1/changes.html 13947 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13948 6. http://gcc.gnu.org/index.html 13949 7. mailto:gcc@gcc.gnu.org 13950 8. http://gcc.gnu.org/mirrors.html 13951 9. https://gcc.gnu.org/onlinedocs/ 13952 10. mailto:gcc-help@gcc.gnu.org 13953 11. mailto:gcc@gcc.gnu.org 13954 12. https://gcc.gnu.org/lists.html 13955 13. http://www.fsf.org/ 13956 14. https://gcc.gnu.org/about.html 13957 15. http://validator.w3.org/check/referer 13958====================================================================== 13959http://gcc.gnu.org/gcc-3.1/changes.html 13960 13961 GCC 3.1 Release Series 13962 Changes, New Features, and Fixes 13963 13964Additional changes in GCC 3.1.1 13965 13966 * A bug related to how structures and unions are returned has been 13967 fixed for powerpc-*-netbsd*. 13968 * An important bug in the implementation of -fprefetch-loop-arrays 13969 has been fixed. Previously the optimization prefetched random 13970 blocks of memory for most targets except for i386. 13971 * The Java compiler now compiles Java programs much faster and also 13972 works with parallel make. 13973 * Nested functions have been fixed for mips*-*-netbsd*. 13974 * Some missing floating point support routines have beed added for 13975 mips*-*-netbsd*. 13976 * This [1]message gives additional information about the bugs fixed 13977 in this release. 13978 13979Caveats 13980 13981 * The -traditional C compiler option has been deprecated and will be 13982 removed in GCC 3.3. (It remains possible to preprocess non-C code 13983 with the traditional preprocessor.) 13984 * The default debugging format for most ELF platforms (including 13985 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 13986 from stabs to DWARF2. This requires GDB 5.1.1 or later. 13987 13988General Optimizer Improvements 13989 13990 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 13991 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 13992 for profile driven optimizations. 13993 Options -fprofile-arcs and -fbranch-probabilities can now be used 13994 to improve speed of the generated code by profiling the actual 13995 program behaviour on typical runs. In the absence of profile info 13996 the compiler attempts to guess the profile statically. 13997 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 13998 monitor performance of the generated code. 13999 According to the SPECInt2000 results on an AMD Athlon CPU, the code 14000 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 14001 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 14002 is about 2.1% faster compared to 2.95.3. Tests were done using the 14003 -O2 -march=athlon command-line options. 14004 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 14005 infrastructure developed by CodeSourcery, LLC for the C++ front 14006 end, so that it is now used in the C front end too. Inlining 14007 functions as trees exposes them earlier to the compiler, giving it 14008 more opportunities for optimization. 14009 * Support for data prefetching instructions has been added to the GCC 14010 back end and several targets. A new __builtin_prefetch intrinsic is 14011 available to explicitly insert prefetch instructions and 14012 experimental support for loop array prefetching has been added (see 14013 -fprefetch-loop-array documentation). 14014 * Support for emitting debugging information for macros has been 14015 added for DWARF2. It is activated using -g3. 14016 14017New Languages and Language specific improvements 14018 14019 C/C++ 14020 14021 * A few more [4]ISO C99 features. 14022 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 14023 * The preprocessor's symbol table has been merged with the symbol 14024 table of the C, C++ and Objective-C front ends. 14025 * The preprocessor consumes less memory than the preprocessor in GCC 14026 3.0, often significantly so. On normal input files, it typically 14027 consumes less memory than pre-3.0 cccp-based GCC, too. 14028 14029 C++ 14030 14031 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 14032 was a workaround to allow std compliant code to work with the 14033 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 14034 * The C++ ABI has been fixed so that void (A::*)() const is mangled 14035 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 14036 affects pointer to cv-qualified member function types. 14037 * The C++ ABI has been changed to correctly handle this code: 14038 struct A { 14039 void operator delete[] (void *, size_t); 14040 }; 14041 14042 struct B : public A { 14043 }; 14044 14045 new B[10]; 14046 14047 The amount of storage allocated for the array will be greater than 14048 it was in 3.0, in order to store the number of elements in the 14049 array, so that the correct size can be passed to operator delete[] 14050 when the array is deleted. Previously, the value passed to operator 14051 delete[] was unpredictable. 14052 This change will only affect code that declares a two-argument 14053 operator delete[] with a second parameter of type size_t in a base 14054 class, and does not override that definition in a derived class. 14055 * The C++ ABI has been changed so that: 14056 struct A { 14057 void operator delete[] (void *, size_t); 14058 void operator delete[] (void *); 14059 }; 14060 14061 does not cause unnecessary storage to be allocated when an array of 14062 A objects is allocated. 14063 This change will only affect code that declares both of these forms 14064 of operator delete[], and declared the two-argument form before the 14065 one-argument form. 14066 * The C++ ABI has been changed so that when a parameter is passed by 14067 value, any cleanup for that parameter is performed in the caller, 14068 as specified by the ia64 C++ ABI, rather than the called function 14069 as before. As a result, classes with a non-trivial destructor but a 14070 trivial copy constructor will be passed and returned by invisible 14071 reference, rather than by bitwise copy as before. 14072 * G++ now supports the "named return value optimization": for code 14073 like 14074 A f () { 14075 A a; 14076 ... 14077 return a; 14078 } 14079 14080 G++ will allocate a in the return value slot, so that the return 14081 becomes a no-op. For this to work, all return statements in the 14082 function must return the same variable. 14083 * Improvements to the C++ library are listed in [5]the libstdc++-v3 14084 FAQ. 14085 14086 Objective-C 14087 14088 * Annoying linker warnings (due to incorrect code being generated) 14089 have been fixed. 14090 * If a class method cannot be found, the compiler no longer issues a 14091 warning if a corresponding instance method exists in the root 14092 class. 14093 * Forward @protocol declarations have been fixed. 14094 * Loading of categories has been fixed in certain situations (GNU run 14095 time only). 14096 * The class lookup in the run-time library has been rewritten so that 14097 class method dispatch is more than twice as fast as it used to be 14098 (GNU run time only). 14099 14100 Java 14101 14102 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 14103 javax.transaction. 14104 * Property files and other system resources can be compiled into 14105 executables which use libgcj using the new gcj --resource feature. 14106 * libgcj has been ported to more platforms. In particular there is 14107 now a mostly-functional mingw32 (Windows) target port. 14108 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 14109 Java code can now be called from a C/C++ application. 14110 * gcj can now use builtin functions for certain known methods, for 14111 instance Math.cos. 14112 * gcj can now automatically remove redundant array-store checks in 14113 some common cases. 14114 * The --no-store-checks optimization option was added. This can be 14115 used to omit runtime store checks for code which is known not to 14116 throw ArrayStoreException 14117 * The following third party interface standards were added to libgcj: 14118 org.w3c.dom and org.xml.sax. 14119 * java.security has been merged with GNU Classpath. The new package 14120 is now JDK 1.2 compliant, and much more complete. 14121 * A bytecode verifier was added to the libgcj interpreter. 14122 * java.lang.Character was rewritten to comply with the Unicode 3.0 14123 standard, and improve performance. 14124 * Partial support for many more locales was added to libgcj. 14125 * Socket timeouts have been implemented. 14126 * libgcj has been merged into a single shared library. There are no 14127 longer separate shared libraries for the garbage collector and 14128 zlib. 14129 * Several performance improvements were made to gcj and libgcj: 14130 + Hash synchronization (thin locks) 14131 + A special allocation path for finalizer-free objects 14132 + Thread-local allocation 14133 + Parallel GC, and other GC tweaks 14134 14135 Fortran 14136 14137 Fortran improvements are listed in [6]the Fortran documentation. 14138 14139 Ada 14140 14141 [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front 14142 end and associated tools. The GNAT compiler fully implements the Ada 14143 language as defined by the ISO/IEC 8652 standard. 14144 14145 Please note that the integration of the Ada front end is still work in 14146 progress. 14147 14148New Targets and Target Specific Improvements 14149 14150 * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU 14151 architecture used in new editions of Donald E. Knuth's The Art of 14152 Computer Programming. 14153 * [9]Axis Communications has contributed its port to the CRIS CPU 14154 architecture, used in the ETRAX system-on-a-chip series. See 14155 [10]Axis' developer site for technical information. 14156 * Alexandre Oliva, of Red Hat, has contributed a port to the 14157 [11]SuperH SH5 64-bit RISC microprocessor architecture, extending 14158 the existing SH port. 14159 * UltraSPARC is fully supported in 64-bit mode. The option -m64 14160 enables it. 14161 * For compatibility with the Sun compiler #pragma redefine_extname 14162 has been implemented on Solaris. 14163 * The x86 back end has had some noticeable work done to it. 14164 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 14165 Jaeger have contributed a port to the AMD x86-64 architecture. 14166 For more information on x86-64 see [12]http://www.x86-64.org. 14167 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 14168 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 14169 enable the respective instruction sets. Intel C++ compatible 14170 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 14171 will be added in next major release. 14172 + Following those improvements, targets for Pentium MMX, K6-2, 14173 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 14174 added. Refer to the documentation on -march= and -mcpu= 14175 options for details. 14176 + For those targets that support it, -mfpmath=sse will cause the 14177 compiler to generate SSE/SSE2 instructions for floating point 14178 math instead of x87 instructions. Usually, this will lead to 14179 quicker code -- especially on the Pentium 4. Note that only 14180 scalar floating point instructions are used and GCC does not 14181 exploit SIMD features yet. 14182 + Prefetch support has been added to the Pentium III, Pentium 4, 14183 K6-2, K6-3, and Athlon series. 14184 + Code generated for floating point to integer conversions has 14185 been improved leading to better performance of many 3D 14186 applications. 14187 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 14188 * C++ support for AIX has been improved. 14189 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 14190 PowerPC port supporting the AltiVec programming model (SIMD). The 14191 support, though presently useful, is experimental and is expected 14192 to stabilize for 3.2. The support is written to conform to 14193 Motorola's AltiVec specs. See -maltivec. 14194 14195Obsolete Systems 14196 14197 Support for a number of older systems has been declared obsolete in GCC 14198 3.1. Unless there is activity to revive them, the next release of GCC 14199 will have their sources permanently removed. 14200 14201 All configurations of the following processor architectures have been 14202 declared obsolete: 14203 * MIL-STD-1750A, 1750a-*-* 14204 * AMD A29k, a29k-*-* 14205 * Convex, c*-convex-* 14206 * Clipper, clipper-*-* 14207 * Elxsi, elxsi-*-* 14208 * Intel i860, i860-*-* 14209 * Sun picoJava, pj-*-* and pjl-*-* 14210 * Western Electric 32000, we32k-*-* 14211 14212 Most configurations of the following processor architectures have been 14213 declared obsolete, but we are preserving a few systems which may have 14214 active developers. It is unlikely that the remaining systems will 14215 survive much longer unless we see definite signs of port activity. 14216 * Motorola 88000 except 14217 + Generic a.out, m88k-*-aout* 14218 + Generic SVR4, m88k-*-sysv4 14219 + OpenBSD, m88k-*-openbsd* 14220 * NS32k except 14221 + NetBSD, ns32k-*-netbsd* 14222 + OpenBSD, ns32k-*-openbsd*. 14223 * ROMP except 14224 + OpenBSD, romp-*-openbsd*. 14225 14226 Finally, only some configurations of these processor architectures are 14227 being obsoleted. 14228 * Alpha: 14229 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 14230 alpha*-*-osf[45], are still supported.) 14231 * ARM: 14232 + RISCiX, arm-*-riscix*. 14233 * i386: 14234 + 386BSD, i?86-*-bsd* 14235 + Chorus, i?86-*-chorusos* 14236 + DG/UX, i?86-*-dgux* 14237 + FreeBSD 1.x, i?86-*-freebsd1.* 14238 + IBM AIX, i?86-*-aix* 14239 + ISC UNIX, i?86-*-isc* 14240 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 14241 + NEXTstep, i?86-next-* 14242 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 14243 + RTEMS/coff, i?86-*-rtemscoff* 14244 + RTEMS/go32, i?86-go32-rtems* 14245 + Sequent/BSD, i?86-sequent-bsd* 14246 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 14247 i?86-sequent-sysv3* 14248 + SunOS, i?86-*-sunos* 14249 * Motorola 68000: 14250 + Altos, m68[k0]*-altos-* 14251 + Apollo, m68[k0]*-apollo-* 14252 + Apple A/UX, m68[k0]*-apple-* 14253 + Bull, m68[k0]*-bull-* 14254 + Convergent, m68[k0]*-convergent-* 14255 + Generic SVR3, m68[k0]*-*-sysv3* 14256 + ISI, m68[k0]*-isi-* 14257 + LynxOS, m68[k0]*-*-lynxos* 14258 + NEXT, m68[k0]*-next-* 14259 + RTEMS/coff, m68[k0]*-*-rtemscoff* 14260 + Sony, m68[k0]*-sony-* 14261 * MIPS: 14262 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 14263 + Generic BSD, mips-*-bsd* 14264 + Generic System V, mips-*-sysv* 14265 + IRIX before version 5, mips-sgi-irix[1234]* 14266 + RiscOS, mips-*-riscos* 14267 + Sony, mips-sony-* 14268 + Tandem, mips-tandem-* 14269 * SPARC: 14270 + RTEMS/a.out, sparc-*-rtemsaout*. 14271 14272Documentation improvements 14273 14274 * The old manual ("Using and Porting the GNU Compiler Collection") 14275 has been replaced by a users manual ("Using the GNU Compiler 14276 Collection") and a separate internals reference manual ("GNU 14277 Compiler Collection Internals"). 14278 * More complete and much improved documentation about GCC's internal 14279 representation used by the C and C++ front ends. 14280 * Many cleanups and improvements in general. 14281 14282 14283 For questions related to the use of GCC, please consult these web 14284 pages and the [13]GCC manuals. If that fails, the 14285 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14286 web pages and the development of GCC are welcome on our developer 14287 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 14288 archives. 14289 14290 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 14291 distribution of this entire article is permitted in any medium, 14292 provided this notice is preserved. 14293 14294 These pages are [18]maintained by the GCC team. Last modified 14295 2016-01-30[19]. 14296 14297References 14298 14299 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 14300 2. http://gcc.gnu.org/news/profiledriven.html 14301 3. http://gcc.gnu.org/benchmarks/ 14302 4. http://gcc.gnu.org/c99status.html 14303 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 14304 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 14305 7. http://www.adacore.com/ 14306 8. http://www-cs-faculty.stanford.edu/~uno/mmix.html 14307 9. http://www.axis.com/ 14308 10. http://developer.axis.com/ 14309 11. http://www.superh.com/ 14310 12. http://www.x86-64.org/ 14311 13. https://gcc.gnu.org/onlinedocs/ 14312 14. mailto:gcc-help@gcc.gnu.org 14313 15. mailto:gcc@gcc.gnu.org 14314 16. https://gcc.gnu.org/lists.html 14315 17. http://www.fsf.org/ 14316 18. https://gcc.gnu.org/about.html 14317 19. http://validator.w3.org/check/referer 14318====================================================================== 14319http://gcc.gnu.org/gcc-3.0/index.html 14320 14321 GCC 3.0.4 14322 14323 February 20, 2002 14324 14325 The [1]GNU project and the GCC developers are pleased to announce the 14326 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 14327 series. 14328 14329 GCC used to stand for the GNU C Compiler, but since the compiler 14330 supports several other languages aside from C, it now stands for the 14331 GNU Compiler Collection. 14332 14333 GCC 3.0.x has several new optimizations, new targets, new languages and 14334 many other new features, relative to GCC 2.95.x. See the [2]new 14335 features page for a more complete list. 14336 14337 A list of [3]successful builds is updated as new information becomes 14338 available. 14339 14340 The GCC developers would like to thank the numerous people that have 14341 contributed new features, test results, bug fixes, etc to GCC. This 14342 [4]amazing group of volunteers is what makes GCC successful. 14343 14344 And finally, we can't in good conscience fail to mention some 14345 [5]caveats to using GCC 3.0.x. 14346 14347 For additional information about GCC please refer to the [6]GCC project 14348 web site or contact the [7]GCC development mailing list. 14349 14350 To obtain GCC please use [8]our mirror sites, or our CVS server. 14351 __________________________________________________________________ 14352 14353Previous 3.0.x Releases 14354 14355 December 20, 2001: GCC 3.0.3 has been released. 14356 October 25, 2001: GCC 3.0.2 has been released. 14357 August 20, 2001: GCC 3.0.1 has been released. 14358 June 18, 2001: GCC 3.0 has been released. 14359 14360 14361 For questions related to the use of GCC, please consult these web 14362 pages and the [9]GCC manuals. If that fails, the 14363 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14364 web pages and the development of GCC are welcome on our developer 14365 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public 14366 archives. 14367 14368 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 14369 distribution of this entire article is permitted in any medium, 14370 provided this notice is preserved. 14371 14372 These pages are [14]maintained by the GCC team. Last modified 14373 2016-01-30[15]. 14374 14375References 14376 14377 1. http://www.gnu.org/ 14378 2. http://gcc.gnu.org/gcc-3.0/features.html 14379 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 14380 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14381 5. http://gcc.gnu.org/gcc-3.0/caveats.html 14382 6. http://gcc.gnu.org/index.html 14383 7. mailto:gcc@gcc.gnu.org 14384 8. http://gcc.gnu.org/mirrors.html 14385 9. https://gcc.gnu.org/onlinedocs/ 14386 10. mailto:gcc-help@gcc.gnu.org 14387 11. mailto:gcc@gcc.gnu.org 14388 12. https://gcc.gnu.org/lists.html 14389 13. http://www.fsf.org/ 14390 14. https://gcc.gnu.org/about.html 14391 15. http://validator.w3.org/check/referer 14392====================================================================== 14393http://gcc.gnu.org/gcc-3.0/features.html 14394 14395 GCC 3.0 New Features 14396 14397Additional changes in GCC 3.0.4 14398 14399 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 14400 system, which use the ELF object file format, on x86 processors. 14401 * Correct debugging information is generated from functions that have 14402 lines from multiple files (e.g. yacc output). 14403 * A fix for whitespace handling in the -traditional preprocessor, 14404 which can affect Fortran. 14405 * Fixes to the exception handling runtime. 14406 * More fixes for bad code generation in C++. 14407 * A fix for shared library generation under AIX 4.3. 14408 * Documentation updates. 14409 * Port of GCC to Tensilica's Xtensa processor contributed. 14410 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 14411 14412Additional changes in GCC 3.0.3 14413 14414 * A fix to correct an accidental change to the PowerPC ABI. 14415 * Fixes for bad code generation on a variety of architectures. 14416 * Improvements to the debugging information generated for C++ 14417 classes. 14418 * Fixes for bad code generation in C++. 14419 * A fix to avoid crashes in the C++ demangler. 14420 * A fix to the C++ standard library to avoid buffer overflows. 14421 * Miscellaneous improvements for a variety of architectures. 14422 14423Additional changes in GCC 3.0.2 14424 14425 * Fixes for bad code generation during loop unrolling. 14426 * Fixes for bad code generation by the sibling call optimization. 14427 * Minor improvements to x86 code generation. 14428 * Implementation of function descriptors in C++ vtables for IA64. 14429 * Numerous minor bug-fixes. 14430 14431Additional changes in GCC 3.0.1 14432 14433 * C++ fixes for incorrect code-generation. 14434 * Improved cross-compiling support for the C++ standard library. 14435 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 14436 in GCC 3.0. 14437 * Fixes for various exception-handling bugs. 14438 * A port to the S/390 architecture. 14439 14440General Optimizer Improvements 14441 14442 * [2]Basic block reordering pass. 14443 * New if-conversion pass with support for conditional (predicated) 14444 execution. 14445 * New tail call and sibling call elimination optimizations. 14446 * New register renaming pass. 14447 * New (experimental) [3]static single assignment (SSA) representation 14448 support. 14449 * New dead-code elimination pass implemented using the SSA 14450 representation. 14451 * [4]Global null pointer test elimination. 14452 * [5]Global code hoisting/unification. 14453 * More builtins and optimizations for stdio.h, string.h and old BSD 14454 functions, as well as for ISO C99 functions. 14455 * New builtin __builtin_expect for giving hints to the branch 14456 predictor. 14457 14458New Languages and Language specific improvements 14459 14460 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 14461 and supported, including the run-time library containing most 14462 common non-GUI Java classes, a bytecode interpreter, and the Boehm 14463 conservative garbage collector. Many bugs have been fixed. GCJ can 14464 compile Java source or Java bytecodes to either native code or Java 14465 class files, and supports native methods written in either the 14466 standard JNI or the more efficient and convenient CNI. 14467 * Here is a [6]partial list of C++ improvements, both new features 14468 and those no longer supported. 14469 * New C++ ABI. On the IA-64 platform GCC is capable of 14470 inter-operating with other IA-64 compilers. 14471 * The new ABI also significantly reduces the size of symbol and debug 14472 information. 14473 * New [7]C++ support library and many C++ bug fixes, vastly improving 14474 our conformance to the ISO C++ standard. 14475 * New [8]inliner for C++. 14476 * Rewritten C preprocessor, integrated into the C, C++ and Objective 14477 C compilers, with very many improvements including ISO C99 support 14478 and [9]improvements to dependency generation. 14479 * Support for more [10]ISO C99 features. 14480 * Many improvements to support for checking calls to format functions 14481 such as printf and scanf, including support for ISO C99 format 14482 features, extensions from the Single Unix Specification and GNU 14483 libc 2.2, checking of strfmon formats and features to assist in 14484 auditing for format string security bugs. 14485 * New warnings for C code that may have undefined semantics because 14486 of violations of sequence point rules in the C standard (such as a 14487 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 14488 * Additional warning option -Wfloat-equal. 14489 * Improvements to -Wtraditional. 14490 * Fortran improvements are listed in [11]the Fortran documentation. 14491 14492New Targets and Target Specific Improvements 14493 14494 * New x86 back end, generating much improved code. 14495 * Support for a generic i386-elf target contributed. 14496 * New option to emit x86 assembly code using Intel style syntax 14497 (-mintel-syntax). 14498 * HPUX 11 support contributed. 14499 * Improved PowerPC code generation, including scheduled prologue and 14500 epilogue. 14501 * Port of GCC to Intel's IA-64 processor contributed. 14502 * Port of GCC to Motorola's MCore 210 and 340 contributed. 14503 * New unified back-end for Arm, Thumb and StrongArm contributed. 14504 * Port of GCC to Intel's XScale processor contributed. 14505 * Port of GCC to Atmel's AVR microcontrollers contributed. 14506 * Port of GCC to Mitsubishi's D30V processor contributed. 14507 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 14508 processor family) contributed. 14509 * Port of GCC to Fujitsu's FR30 processor contributed. 14510 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 14511 * Port of GCC to Sun's picoJava processor core contributed. 14512 14513Documentation improvements 14514 14515 * Substantially rewritten and improved C preprocessor manual. 14516 * Many improvements to other documentation. 14517 * Manpages for gcc, cpp and gcov are now generated automatically from 14518 the master Texinfo manual, eliminating the problem of manpages 14519 being out of date. (The generated manpages are only extracts from 14520 the full manual, which is provided in Texinfo form, from which 14521 info, HTML, other formats and a printed manual can be generated.) 14522 * Generated info files are included in the release tarballs alongside 14523 their Texinfo sources, avoiding problems on some platforms with 14524 building makeinfo as part of the GCC distribution. 14525 14526Other significant improvements 14527 14528 * Garbage collection used internally by the compiler for most memory 14529 allocation instead of obstacks. 14530 * Lengauer and Tarjan algorithm used for computing dominators in the 14531 CFG. This algorithm can be significantly faster and more space 14532 efficient than our older algorithm. 14533 * gccbug script provided to assist in submitting bug reports to our 14534 bug tracking system. (Bug reports previously submitted directly to 14535 our mailing lists, for which you received no bug tracking number, 14536 should be submitted again using gccbug if you can reproduce the 14537 problem with GCC 3.0.) 14538 * The internal libgcc library is [12]built as a shared library on 14539 systems that support it. 14540 * Extensive testsuite included with GCC, with many new tests. In 14541 addition to tests for GCC bugs that have been fixed, many tests 14542 have been added for language features, compiler warnings and 14543 builtin functions. 14544 * Additional language-independent warning options -Wpacked, -Wpadded, 14545 -Wunreachable-code and -Wdisabled-optimization. 14546 * Target-independent options -falign-functions, -falign-loops and 14547 -falign-jumps. 14548 14549 Plus a great many bug fixes and almost all the [13]features found in 14550 GCC 2.95. 14551 14552 14553 For questions related to the use of GCC, please consult these web 14554 pages and the [14]GCC manuals. If that fails, the 14555 [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14556 web pages and the development of GCC are welcome on our developer 14557 list at [16]gcc@gcc.gnu.org. All of [17]our lists have public 14558 archives. 14559 14560 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 14561 distribution of this entire article is permitted in any medium, 14562 provided this notice is preserved. 14563 14564 These pages are [19]maintained by the GCC team. Last modified 14565 2016-01-30[20]. 14566 14567References 14568 14569 1. http://www.netbsd.org/ 14570 2. http://gcc.gnu.org/news/reorder.html 14571 3. http://gcc.gnu.org/news/ssa.html 14572 4. http://gcc.gnu.org/news/null.html 14573 5. http://gcc.gnu.org/news/unify.html 14574 6. http://gcc.gnu.org/gcc-3.0/c++features.html 14575 7. http://gcc.gnu.org/libstdc++/ 14576 8. http://gcc.gnu.org/news/inlining.html 14577 9. http://gcc.gnu.org/news/dependencies.html 14578 10. http://gcc.gnu.org/c99status.html 14579 11. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 14580 12. http://gcc.gnu.org/gcc-3.0/libgcc.html 14581 13. http://gcc.gnu.org/gcc-2.95/features.html 14582 14. https://gcc.gnu.org/onlinedocs/ 14583 15. mailto:gcc-help@gcc.gnu.org 14584 16. mailto:gcc@gcc.gnu.org 14585 17. https://gcc.gnu.org/lists.html 14586 18. http://www.fsf.org/ 14587 19. https://gcc.gnu.org/about.html 14588 20. http://validator.w3.org/check/referer 14589====================================================================== 14590http://gcc.gnu.org/gcc-3.0/caveats.html 14591 14592 GCC 3.0 Caveats 14593 14594 * -fstrict-aliasing is now part of -O2 and higher optimization 14595 levels. This allows the compiler to assume the strictest aliasing 14596 rules applicable to the language being compiled. For C and C++, 14597 this activates optimizations based on the type of expressions. This 14598 optimization may thus break old, non-compliant code. 14599 * Enumerations are now properly promoted to int in function 14600 parameters and function returns. Normally this change is not 14601 visible, but when using -fshort-enums this is an ABI change. 14602 * The undocumented extension that allowed C programs to have a label 14603 at the end of a compound statement has been deprecated and may be 14604 removed in a future version. Programs that now generate a warning 14605 about this may be fixed by adding a null statement (a single 14606 semicolon) after the label. 14607 * The poorly documented extension that allowed string constants in C, 14608 C++ and Objective C to contain unescaped newlines has been 14609 deprecated and may be removed in a future version. Programs using 14610 this extension may be fixed in several ways: the bare newline may 14611 be replaced by \n, or preceded by \n\, or string concatenation may 14612 be used with the bare newline preceded by \n" and " placed at the 14613 start of the next line. 14614 * The Chill compiler is not included in GCC 3.0, because of the lack 14615 of a volunteer to convert it to use garbage collection. 14616 * Certain non-standard iostream methods from earlier versions of 14617 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 14618 ostream::form, and istream::gets. 14619 * The new C++ ABI is not yet fully supported by current (as of 14620 2001-07-01) releases and development versions of GDB, or any 14621 earlier versions. There is a problem setting breakpoints by line 14622 number, and other related issues that have been fixed in GCC 3.0 14623 but not yet handled in GDB: 14624 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 14625 14626 14627 For questions related to the use of GCC, please consult these web 14628 pages and the [2]GCC manuals. If that fails, the 14629 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14630 web pages and the development of GCC are welcome on our developer 14631 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives. 14632 14633 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 14634 distribution of this entire article is permitted in any medium, 14635 provided this notice is preserved. 14636 14637 These pages are [7]maintained by the GCC team. Last modified 14638 2016-01-30[8]. 14639 14640References 14641 14642 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 14643 2. https://gcc.gnu.org/onlinedocs/ 14644 3. mailto:gcc-help@gcc.gnu.org 14645 4. mailto:gcc@gcc.gnu.org 14646 5. https://gcc.gnu.org/lists.html 14647 6. http://www.fsf.org/ 14648 7. https://gcc.gnu.org/about.html 14649 8. http://validator.w3.org/check/referer 14650====================================================================== 14651http://gcc.gnu.org/gcc-2.95/index.html 14652 14653 GCC 2.95 14654 14655 March 16, 2001: The GNU project and the GCC developers are pleased to 14656 announce the release of GCC version 2.95.3. 14657 14658Release History 14659 14660 GCC 2.95.3 14661 March 16, 2001 14662 14663 GCC 2.95.2 14664 October 27, 1999 14665 14666 GCC 2.95.1 14667 August 19, 1999 14668 14669 GCC 2.95 14670 July 31, 1999. This is the first release of GCC since the April 14671 1999 GCC/EGCS reunification and includes nearly a year's worth 14672 of new development and bugfixes. 14673 14674References and Acknowledgements 14675 14676 GCC used to stand for the GNU C Compiler, but since the compiler 14677 supports several other languages aside from C, it now stands for the 14678 GNU Compiler Collection. 14679 14680 The whole suite has been extensively [1]regression tested and 14681 [2]package tested. It should be reliable and suitable for widespread 14682 use. 14683 14684 The compiler has several new optimizations, new targets, new languages 14685 and other new features. See the [3]new features page for a more 14686 complete list of new features found in the GCC 2.95 releases. 14687 14688 The sources include installation instructions in both HTML and 14689 plaintext forms in the install directory in the distribution. However, 14690 the most up to date installation instructions and [4]build/test status 14691 are on the web pages. We will update those pages as new information 14692 becomes available. 14693 14694 The GCC developers would like to thank the numerous people that have 14695 contributed new features, test results, bugfixes, etc to GCC. This 14696 [5]amazing group of volunteers is what makes GCC successful. 14697 14698 And finally, we can't in good conscience fail to mention some 14699 [6]caveats to using GCC 2.95. 14700 14701 Download GCC 2.95 from one of our many [7]mirror sites. 14702 14703 For additional information about GCC please see the [8]GCC project web 14704 server or contact the [9]GCC development mailing list. 14705 14706 14707 For questions related to the use of GCC, please consult these web 14708 pages and the [10]GCC manuals. If that fails, the 14709 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14710 web pages and the development of GCC are welcome on our developer 14711 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public 14712 archives. 14713 14714 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 14715 distribution of this entire article is permitted in any medium, 14716 provided this notice is preserved. 14717 14718 These pages are [15]maintained by the GCC team. Last modified 14719 2016-01-30[16]. 14720 14721References 14722 14723 1. http://gcc.gnu.org/gcc-2.95/regress.html 14724 2. http://gcc.gnu.org/gcc-2.95/othertest.html 14725 3. http://gcc.gnu.org/gcc-2.95/features.html 14726 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 14727 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14728 6. http://gcc.gnu.org/gcc-2.95/caveats.html 14729 7. http://gcc.gnu.org/mirrors.html 14730 8. http://gcc.gnu.org/index.html 14731 9. mailto:gcc@gcc.gnu.org 14732 10. https://gcc.gnu.org/onlinedocs/ 14733 11. mailto:gcc-help@gcc.gnu.org 14734 12. mailto:gcc@gcc.gnu.org 14735 13. https://gcc.gnu.org/lists.html 14736 14. http://www.fsf.org/ 14737 15. https://gcc.gnu.org/about.html 14738 16. http://validator.w3.org/check/referer 14739====================================================================== 14740http://gcc.gnu.org/gcc-2.95/features.html 14741 14742 GCC 2.95 New Features 14743 14744 * General Optimizer Improvements: 14745 + [1]Localized register spilling to improve speed and code 14746 density especially on small register class machines. 14747 + [2]Global CSE using lazy code motion algorithms. 14748 + [3]Improved global constant/copy propagation. 14749 + [4]Improved control flow graph analysis and manipulation. 14750 + [5]Local dead store elimination. 14751 + [6]Memory Load hoisting/store sinking in loops. 14752 + [7]Type based alias analysis is enabled by default. Note this 14753 feature will expose bugs in the Linux kernel. Please refer to 14754 the FAQ (as shipped with GCC 2.95) for additional information 14755 on this issue. 14756 + Major revamp of GIV detection, combination and simplification 14757 to improve loop performance. 14758 + Major improvements to register allocation and reloading. 14759 * New Languages and Language specific improvements 14760 + [8]Many C++ improvements. 14761 + [9]Many Fortran improvements. 14762 + [10]Java front-end has been integrated. [11]runtime library is 14763 available separately. 14764 + [12]ISO C99 support 14765 + [13]Chill front-end and runtime has been integrated. 14766 + Boehm garbage collector support in libobjc. 14767 + More support for various pragmas which appear in vendor 14768 include files 14769 * New Targets and Target Specific Improvements 14770 + [14]SPARC backend rewrite. 14771 + -mschedule=8000 will optimize code for PA8000 class 14772 processors; -mpa-risc-2-0 will generate code for PA2.0 14773 processors 14774 + Various micro-optimizations for the ia32 port. K6 14775 optimizations 14776 + Compiler will attempt to align doubles in the stack on the 14777 ia32 port 14778 + Alpha EV6 support 14779 + PowerPC 750 14780 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 14781 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 14782 + c3x, c4x 14783 + HyperSPARC 14784 + SparcLite86x 14785 + sh4 14786 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 14787 arm-linux) 14788 + vxWorks targets include support for vxWorks threads 14789 + StrongARM 110 and ARM9 support added. ARM Scheduling 14790 parameters rewritten. 14791 + Various changes to the MIPS port to avoid assembler macros, 14792 which in turn improves performance 14793 + Various performance improvements to the i960 port. 14794 + Major rewrite of ns32k port 14795 * Other significant improvements 14796 + [15]Ability to dump cfg information and display it using vcg. 14797 + The new faster scheme for fixing vendor header files is 14798 enabled by default. 14799 + Experimental internationalization support. 14800 + multibyte character support 14801 + Some compile-time speedups for pathological problems 14802 + Better support for complex types 14803 * Plus the usual mountain of bugfixes 14804 * Core compiler is based on the gcc2 development tree from Sept 30, 14805 1998, so we have all of the [16]features found in GCC 2.8. 14806 14807Additional Changes in GCC 2.95.1 14808 14809 * Generic bugfixes and improvements 14810 + Various documentation fixes related to the GCC/EGCS merger. 14811 + Fix memory management bug which could lead to spurious aborts, 14812 core dumps or random parsing errors in the compiler. 14813 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 14814 support. 14815 + Fix infinite loop in the CSE optimizer. 14816 + Avoid undefined behavior in compiler FP emulation code 14817 + Fix install problem when prefix is overridden on the make 14818 install command. 14819 + Fix problem with unwanted installation of assert.h on some 14820 systems. 14821 + Fix problem with finding the wrong assembler in a single tree 14822 build. 14823 + Avoid increasing the known alignment of a register that is 14824 already known to be a pointer. 14825 * Platform specific bugfixes and improvements 14826 + Codegen bugfix for prologue/epilogue for cpu32 target. 14827 + Fix long long code generation bug for the Coldfire target. 14828 + Fix various aborts in the SH compiler. 14829 + Fix bugs in libgcc support library for the SH. 14830 + Fix alpha ev6 code generation bug. 14831 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 14832 AIX platforms. 14833 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 14834 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 14835 targets. 14836 + Fix weak symbol handling for rs6000/ppc svr4 targets. 14837 + Fix various problems with 64bit code generation for the 14838 rs6000/ppc port. 14839 + Fix codegen bug which caused tetex to be mis-compiled on the 14840 x86. 14841 + Fix compiler abort in new cfg code exposed by x86 port. 14842 + Fix out of range array reference in code convert flat 14843 registers to the x87 stacked FP register file. 14844 + Fix minor vxworks configuration bug. 14845 + Fix return type of bsearch for SunOS 4.x. 14846 * Language & Runtime specific fixes. 14847 + The G++ signature extension has been deprecated. It will be 14848 removed in the next major release of G++. Use of signatures 14849 will result in a warning from the compiler. 14850 + Several bugs relating to templates and namespaces were fixed. 14851 + A bug that caused crashes when combining templates with -g on 14852 DWARF1 platforms was fixed. 14853 + Pointers-to-members, virtual functions, and multiple 14854 inheritance should now work together correctly. 14855 + Some code-generation bugs relating to function try blocks were 14856 fixed. 14857 + G++ is a little bit more lenient with certain archaic 14858 constructs than in GCC 2.95. 14859 + Fix to prevent shared library version #s from bring truncated 14860 to 1 digit 14861 + Fix missing std:: in the libstdc++ library. 14862 + Fix stream locking problems in libio. 14863 + Fix problem in java compiler driver. 14864 14865Additional Changes in GCC 2.95.2 14866 14867 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 14868 the optimizations performed by -fstrict-aliasing are valid according to 14869 the C and C++ standards, the optimization have caused some problems, 14870 particularly with old non-conforming code. 14871 14872 The GCC developers are experimenting with ways to warn users about code 14873 which violates the C/C++ standards, but those warnings are not ready 14874 for widespread use at this time. Rather than wait for those warnings 14875 the GCC developers have chosen to disable -fstrict-aliasing by default 14876 for the GCC 2.95.2 release. 14877 14878 We strongly encourage developers to find and fix code which violates 14879 the C/C++ standards as -fstrict-aliasing may be enabled by default in 14880 future releases. Use the option -fstrict-aliasing to re-enable these 14881 optimizations. 14882 * Generic bugfixes and improvements 14883 + Fix incorrectly optimized memory reference in global common 14884 subexpression elimination (GCSE) optimization pass. 14885 + Fix code generation bug in regmove.c in which it could 14886 incorrectly change a "const" value. 14887 + Fix bug in optimization of conditionals involving volatile 14888 memory references. 14889 + Avoid over-allocation of stack space for some procedures. 14890 + Fixed bug in the compiler which caused incorrect optimization 14891 of an obscure series of bit manipulations, shifts and 14892 arithmetic. 14893 + Fixed register allocator bug which caused teTeX to be 14894 mis-compiled on SPARC targets. 14895 + Avoid incorrect optimization of degenerate case statements for 14896 certain targets such as the ARM. 14897 + Fix out of range memory reference in the jump optimizer. 14898 + Avoid dereferencing null pointer in fix-header. 14899 + Fix test for GCC specific features so that it is possible to 14900 bootstrap with gcc-2.6.2 and older versions of GCC. 14901 + Fix typo in scheduler which could potentially cause out of 14902 range memory accesses. 14903 + Avoid incorrect loop reversal which caused incorrect code for 14904 certain loops on PowerPC targets. 14905 + Avoid incorrect optimization of switch statements on certain 14906 targets (for example the ARM). 14907 * Platform specific bugfixes and improvements 14908 + Work around bug in Sun V5.0 compilers which caused bootstrap 14909 comparison failures on SPARC targets. 14910 + Fix SPARC backend bug which caused aborts in final.c. 14911 + Fix sparc-hal-solaris2* configuration fragments. 14912 + Fix bug in sparc block profiling. 14913 + Fix obscure code generation bug for the PARISC targets. 14914 + Define __STDC_EXT__ for HPUX configurations. 14915 + Various POWERPC64 code generation bugfixes. 14916 + Fix abort for PPC targets using ELF (ex GNU/Linux). 14917 + Fix collect2 problems for AIX targets. 14918 + Correct handling of .file directive for PPC targets. 14919 + Fix bug in fix_trunc x86 patterns. 14920 + Fix x86 port to correctly pop the FP stack for functions that 14921 return structures in memory. 14922 + Fix minor bug in strlen x86 pattern. 14923 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 14924 + Fix template repository code to handle leading underscore in 14925 mangled names. 14926 + Fix weak/weak alias support for OpenBSD. 14927 + GNU/Linux for the ARM has C++ compatible include files. 14928 * Language & Runtime specific fixes. 14929 + Fix handling of constructor attribute in the C front-end which 14930 caused problems building the Chill runtime library on some 14931 targets. 14932 + Fix minor problem merging type qualifiers in the C front-end. 14933 + Fix aliasing bug for pointers and references (C/C++). 14934 + Fix incorrect "non-constant initializer bug" when -traditional 14935 or -fwritable-strings is enabled. 14936 + Fix build error for Chill front-end on SunOS. 14937 + Do not complain about duplicate instantiations when using 14938 -frepo (C++). 14939 + Fix array bounds handling in C++ front-end which caused 14940 problems with dwarf debugging information in some 14941 circumstances. 14942 + Fix minor namespace problem. 14943 + Fix problem linking java programs. 14944 14945Additional Changes in GCC 2.95.3 14946 14947 * Generic bugfixes and improvements 14948 + Fix numerous problems that caused incorrect optimization in 14949 the register reloading code. 14950 + Fix numerous problems that caused incorrect optimization in 14951 the loop optimizer. 14952 + Fix aborts in the functions build_insn_chain and scan_loops 14953 under some circumstances. 14954 + Fix an alias analysis bug. 14955 + Fix an infinite compilation bug in the combiner. 14956 + A few problems with complex number support have been fixed. 14957 + It is no longer possible for gcc to act as a fork bomb when 14958 installed incorrectly. 14959 + The -fpack-struct option should be recognized now. 14960 + Fixed a bug that caused incorrect code to be generated due to 14961 a lost stack adjustment. 14962 * Platform specific bugfixes and improvements 14963 + Support building ARM toolchains hosted on Windows. 14964 + Fix attribute calculations in ARM toolchains. 14965 + arm-linux support has been improved. 14966 + Fix a PIC failure on sparc targets. 14967 + On ix86 targets, the regparm attribute should now work 14968 reliably. 14969 + Several updates for the h8300 port. 14970 + Fix problem building libio with glibc 2.2. 14971 14972 14973 For questions related to the use of GCC, please consult these web 14974 pages and the [17]GCC manuals. If that fails, the 14975 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14976 web pages and the development of GCC are welcome on our developer 14977 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public 14978 archives. 14979 14980 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 14981 distribution of this entire article is permitted in any medium, 14982 provided this notice is preserved. 14983 14984 These pages are [22]maintained by the GCC team. Last modified 14985 2016-01-30[23]. 14986 14987References 14988 14989 1. http://gcc.gnu.org/news/spill.html 14990 2. http://gcc.gnu.org/news/lcm.html 14991 3. http://gcc.gnu.org/news/cprop.html 14992 4. http://gcc.gnu.org/news/cfg.html 14993 5. http://gcc.gnu.org/news/dse.html 14994 6. http://gcc.gnu.org/news/hoist.html 14995 7. http://gcc.gnu.org/news/alias.html 14996 8. http://gcc.gnu.org/gcc-2.95/c++features.html 14997 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 14998 10. http://gcc.gnu.org/java/gcj-announce.txt 14999 11. http://gcc.gnu.org/news/javaannounce.html 15000 12. http://gcc.gnu.org/c99status.html 15001 13. http://gcc.gnu.org/news/chill.html 15002 14. http://gcc.gnu.org/news/sparc.html 15003 15. http://gcc.gnu.org/news/egcs-vcg.html 15004 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 15005 17. https://gcc.gnu.org/onlinedocs/ 15006 18. mailto:gcc-help@gcc.gnu.org 15007 19. mailto:gcc@gcc.gnu.org 15008 20. https://gcc.gnu.org/lists.html 15009 21. http://www.fsf.org/ 15010 22. https://gcc.gnu.org/about.html 15011 23. http://validator.w3.org/check/referer 15012====================================================================== 15013http://gcc.gnu.org/gcc-2.95/caveats.html 15014 15015 GCC 2.95 Caveats 15016 15017 * GCC 2.95 will issue an error for invalid asm statements that had 15018 been silently accepted by earlier versions of the compiler. This is 15019 particularly noticeable when compiling older versions of the Linux 15020 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 15021 for more information on this issue. 15022 * GCC 2.95 implements type based alias analysis to disambiguate 15023 memory references. Some programs, particularly the Linux kernel 15024 violate ANSI/ISO aliasing rules and therefore may not operate 15025 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 15026 shipped with GCC 2.95) for more information on this issue. 15027 * GCC 2.95 has a known bug in its handling of complex variables for 15028 64bit targets. Instead of silently generating incorrect code, GCC 15029 2.95 will issue a fatal error for situations it can not handle. 15030 This primarily affects the Fortran community as Fortran makes more 15031 use of complex variables than C or C++. 15032 * GCC 2.95 has an integrated libstdc++, but does not have an 15033 integrated libg++. Furthermore old libg++ releases will not work 15034 with GCC 2.95. You can retrieve a recent copy of libg++ from the 15035 [1]GCC ftp server. 15036 Note most C++ programs only need libstdc++. 15037 * Exception handling may not work with shared libraries, particularly 15038 on alphas, hppas, rs6000/powerpc and mips based platforms. 15039 Exception handling is known to work on x86 GNU/Linux platforms with 15040 shared libraries. 15041 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 15042 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 15043 or EGCS 1.1. As a result it may be necessary to fix C++ code before 15044 it will compile with GCC 2.95. 15045 * G++ is also converting toward the ISO C++ standard; as a result 15046 code which was previously valid (and thus accepted by other 15047 compilers and older versions of g++) may no longer be accepted. The 15048 flag -fpermissive may allow some non-conforming code to compile 15049 with GCC 2.95. 15050 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 15051 1.1.x, EGCS 1.0.x or GCC 2.8.x. 15052 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 15053 between Sept 30, 1998 and April 30, 1999 (the official end of the 15054 GCC 2.8 project). Future GCC releases will include all the changes 15055 from the defunct GCC 2.8 sources. 15056 15057 15058 For questions related to the use of GCC, please consult these web 15059 pages and the [2]GCC manuals. If that fails, the 15060 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these 15061 web pages and the development of GCC are welcome on our developer 15062 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives. 15063 15064 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 15065 distribution of this entire article is permitted in any medium, 15066 provided this notice is preserved. 15067 15068 These pages are [7]maintained by the GCC team. Last modified 15069 2016-01-30[8]. 15070 15071References 15072 15073 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 15074 2. https://gcc.gnu.org/onlinedocs/ 15075 3. mailto:gcc-help@gcc.gnu.org 15076 4. mailto:gcc@gcc.gnu.org 15077 5. https://gcc.gnu.org/lists.html 15078 6. http://www.fsf.org/ 15079 7. https://gcc.gnu.org/about.html 15080 8. http://validator.w3.org/check/referer 15081====================================================================== 15082http://gcc.gnu.org/egcs-1.1/index.html 15083 15084 EGCS 1.1 15085 15086 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 15087 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 15088 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 15089 15090 EGCS is a free software project to further the development of the GNU 15091 compilers using an open development environment. 15092 15093 EGCS 1.1 is a major new release of the EGCS compiler system. It has 15094 been [1]extensively tested and is believed to be stable and suitable 15095 for widespread use. 15096 15097 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 15098 development sources; it contains all of the new features found in GCC 15099 2.8.1 as well as all new development from GCC up to June 6, 1998. 15100 15101 EGCS 1.1 also contains many improvements and features not found in GCC 15102 or in older versions of EGCS: 15103 * Global common subexpression elimination and global constant/copy 15104 propagation (aka [2]gcse) 15105 * Ongoing improvements to the [3]alias analysis support to allow for 15106 better optimizations throughout the compiler. 15107 * Vastly improved [4]C++ compiler and integrated C++ runtime 15108 libraries. 15109 * Fixes for the /tmp symlink race security problems. 15110 * New targets including mips16, arm-thumb and 64 bit PowerPC. 15111 * Improvements to GNU Fortran (g77) compiler and runtime library made 15112 since g77 version 0.5.23. 15113 15114 See the [5]new features page for a more complete list of new features 15115 found in EGCS 1.1 releases. 15116 15117 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 15118 1.1: 15119 * General improvements and fixes 15120 + Avoid some stack overflows when compiling large functions. 15121 + Avoid incorrect loop invariant code motions. 15122 + Fix some core dumps on Linux kernel code. 15123 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 15124 + Fix code generation problem in gcse. 15125 + Various documentation related fixes. 15126 * g++/libstdc++ improvements and fixes 15127 + MT safe EH fix for setjmp/longjmp based exception handling. 15128 + Fix a few bad interactions between optimization and exception 15129 handling. 15130 + Fixes for demangling of template names starting with "__". 15131 + Fix a bug that would fail to run destructors in some cases 15132 with -O2. 15133 + Fix 'new' of classes with virtual bases. 15134 + Fix crash building Qt on the Alpha. 15135 + Fix failure compiling WIFEXITED macro on GNU/Linux. 15136 + Fix some -frepo failures. 15137 * g77 and libf2c improvements and fixes 15138 + Various documentation fixes. 15139 + Avoid compiler crash on RAND intrinsic. 15140 + Fix minor bugs in makefiles exposed by BSD make programs. 15141 + Define _XOPEN_SOURCE for libI77 build to avoid potential 15142 problems on some 64-bit systems. 15143 + Fix problem with implicit endfile on rewind. 15144 + Fix spurious recursive I/O errors. 15145 * platform specific improvements and fixes 15146 + Match all versions of UnixWare7. 15147 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 15148 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 15149 from unsigned ints to double precision floats. 15150 + Fix ARM ABI issue with NetBSD. 15151 + Fix a few arm code generation bugs. 15152 + Fixincludes will fix additional broken SCO OpenServer header 15153 files. 15154 + Fix a m68k backend bug which caused invalid offsets in reg+d 15155 addresses. 15156 + Fix problems with 64bit AIX 4.3 support. 15157 + Fix handling of long longs for varargs/stdarg functions on the 15158 ppc. 15159 + Minor fixes to CPP predefines for Windows. 15160 + Fix code generation problems with gpr<->fpr copies for 64bit 15161 ppc. 15162 + Fix a few coldfire code generation bugs. 15163 + Fix some more header file problems on SunOS 4.x. 15164 + Fix assert.h handling for RTEMS. 15165 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 15166 + Fix x86 compiler abort in reg-stack pass. 15167 + Fix cygwin/windows problem with section attributes. 15168 + Fix Alpha code generation problem exposed by SMP Linux 15169 kernels. 15170 + Fix typo in m68k 32->64bit integer conversion. 15171 + Make sure target libraries build with -fPIC for PPC & Alpha 15172 targets. 15173 15174 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 15175 1.1.1: 15176 * General improvements and fixes 15177 + Fix bug in loop optimizer which caused the SPARC (and 15178 potentially other) ports to segfault. 15179 + Fix infinite recursion in alias analysis and combiner code. 15180 + Fix bug in regclass preferencing. 15181 + Fix incorrect loop reversal which caused incorrect code to be 15182 generated for several targets. 15183 + Fix return value for builtin memcpy. 15184 + Reduce compile time for certain loops which exposed quadratic 15185 behavior in the loop optimizer. 15186 + Fix bug which caused volatile memory to be written multiple 15187 times when only one write was needed/desired. 15188 + Fix compiler abort in caller-save.c 15189 + Fix combiner bug which caused incorrect code generation for 15190 certain division by constant operations. 15191 + Fix incorrect code generation due to a bug in range check 15192 optimizations. 15193 + Fix incorrect code generation due to mis-handling of clobbered 15194 values in CSE. 15195 + Fix compiler abort/segfault due to incorrect register 15196 splitting when unrolling loops. 15197 + Fix code generation involving autoincremented addresses with 15198 ternary operators. 15199 + Work around bug in the scheduler which caused qt to be 15200 mis-compiled on some platforms. 15201 + Fix code generation problems with -fshort-enums. 15202 + Tighten security for temporary files. 15203 + Improve compile time for codes which make heavy use of 15204 overloaded functions. 15205 + Fix multiply defined constructor/destructor symbol problems. 15206 + Avoid setting bogus RPATH environment variable during 15207 bootstrap. 15208 + Avoid GNU-make dependencies in the texinfo subdir. 15209 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 15210 --enable-cpp=<dirname> can be used to specify an additional 15211 install directory for the cpp wrapper script. 15212 + Fix CSE bug which caused incorrect label-label refs to appear 15213 on some platforms. 15214 + Avoid linking in EH routines from libgcc if they are not 15215 needed. 15216 + Avoid obscure bug in aliasing code. 15217 + Fix bug in weak symbol handling. 15218 * Platform-specific improvements and fixes 15219 + Fix detection of PPro/PII on Unixware 7. 15220 + Fix compiler segfault when building spec99 and other programs 15221 for SPARC targets. 15222 + Fix code-generation bugs for integer and floating point 15223 conditional move instructions on the PPro/PII. 15224 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 15225 + Fix build failure for the arc port. 15226 + Fix floating point format configuration for i?86-gnu port. 15227 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 15228 threads are enabled. 15229 + Fix coldfire code generation bugs. 15230 + Fix "unrecognized insn" problems for Alpha and PPC ports. 15231 + Fix h8/300 code generation problem with floating point values 15232 in memory. 15233 + Fix unrecognized insn problems for the m68k port. 15234 + Fix namespace-pollution problem for the x86 port. 15235 + Fix problems with old assembler on x86 NeXT systems. 15236 + Fix PIC code-generation problems for the SPARC port. 15237 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 15238 + Fix minor ISO namespace violation in Alpha varargs/stdarg 15239 support. 15240 + Fix incorrect "braf" instruction usage for the SH port. 15241 + Fix minor bug in va-sh which prevented its use with -ansi. 15242 + Fix problems recognizing and supporting FreeBSD. 15243 + Handle OpenBSD systems correctly. 15244 + Minor fixincludes fix for Digital UNIX 4.0B. 15245 + Fix problems with ctors/dtors in SCO shared libraries. 15246 + Abort instead of generating incorrect code for PPro/PII 15247 floating point conditional moves. 15248 + Avoid multiply defined symbols on GNU/Linux systems using 15249 libc-5.4.xx. 15250 + Fix abort in alpha compiler. 15251 * Fortran-specific fixes 15252 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 15253 is in the documented, non-Y2K-compliant range of 0-99, instead 15254 of being returned as 100 in the year 2000. 15255 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 15256 milliseconds value properly in Values(8). 15257 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 15258 information properly in SArray(7). 15259 15260 Each release includes installation instructions in both HTML and 15261 plaintext forms (see the INSTALL directory in the toplevel directory of 15262 the distribution). However, we also keep the most up to date 15263 installation instructions and [6]build/test status on our web page. We 15264 will update those pages as new information becomes available. 15265 15266 The EGCS project would like to thank the numerous people that have 15267 contributed new features, test results, bugfixes, etc. This [7]amazing 15268 group of volunteers is what makes EGCS successful. 15269 15270 And finally, we can't in good conscience fail to mention some 15271 [8]caveats to using EGCS 1.1. 15272 15273 Download EGCS from egcs.cygnus.com (USA California). 15274 15275 The EGCS 1.1 release is also available on many mirror sites. 15276 [9]Goto mirror list to find a closer site. 15277 15278 15279 For questions related to the use of GCC, please consult these web 15280 pages and the [10]GCC manuals. If that fails, the 15281 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these 15282 web pages and the development of GCC are welcome on our developer 15283 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public 15284 archives. 15285 15286 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 15287 distribution of this entire article is permitted in any medium, 15288 provided this notice is preserved. 15289 15290 These pages are [15]maintained by the GCC team. Last modified 15291 2016-01-30[16]. 15292 15293References 15294 15295 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 15296 2. http://gcc.gnu.org/news/gcse.html 15297 3. http://gcc.gnu.org/news/alias.html 15298 4. http://gcc.gnu.org/egcs-1.1/c++features.html 15299 5. http://gcc.gnu.org/egcs-1.1/features.html 15300 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 15301 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15302 8. http://gcc.gnu.org/egcs-1.1/caveats.html 15303 9. http://gcc.gnu.org/mirrors.html 15304 10. https://gcc.gnu.org/onlinedocs/ 15305 11. mailto:gcc-help@gcc.gnu.org 15306 12. mailto:gcc@gcc.gnu.org 15307 13. https://gcc.gnu.org/lists.html 15308 14. http://www.fsf.org/ 15309 15. https://gcc.gnu.org/about.html 15310 16. http://validator.w3.org/check/referer 15311====================================================================== 15312http://gcc.gnu.org/egcs-1.1/features.html 15313 15314 EGCS 1.1 new features 15315 15316 * Integrated GNU Fortran (g77) compiler and runtime library with 15317 improvements, based on g77 version 0.5.23. 15318 * Vast improvements in the C++ compiler; so many they have [1]page of 15319 their own! 15320 * Compiler implements [2]global common subexpression elimination and 15321 global copy/constant propagation. 15322 * More major improvements in the [3]alias analysis code. 15323 * More major improvements in the exception handling code to improve 15324 performance, lower static overhead and provide the infrastructure 15325 for future improvements. 15326 * The infamous /tmp symlink race security problems have been fixed. 15327 * The regmove optimization pass has been nearly completely rewritten 15328 to improve performance of generated code. 15329 * The compiler now recomputes register usage information before local 15330 register allocation. By providing more accurate information to the 15331 priority based allocator, we get better register allocation. 15332 * The register reloading phase of the compiler optimizes spill code 15333 much better than in previous releases. 15334 * Some bad interactions between the register allocator and 15335 instruction scheduler have been fixed, resulting in much better 15336 code for certain programs. Additionally, we have tuned the 15337 scheduler in various ways to improve performance of generated code 15338 for some architectures. 15339 * The compiler's branch shortening algorithms have been significantly 15340 improved to work better on targets which align jump targets. 15341 * The compiler now supports -Os to prefer optimizing for code space 15342 over optimizing for code speed. 15343 * The compiler will now totally eliminate library calls which compute 15344 constant values. This primarily helps targets with no integer 15345 div/mul support and targets without floating point support. 15346 * The compiler now supports an extensive "--help" option. 15347 * cpplib has been greatly improved and may be suitable for limited 15348 use. 15349 * Memory footprint for the compiler has been significantly reduced 15350 for some pathological cases. 15351 * The time to build EGCS has been improved for certain targets 15352 (particularly the alpha and mips platforms). 15353 * Many infrastructure improvements throughout the compiler, plus the 15354 usual mountain of bugfixes and minor improvements. 15355 * Target dependent improvements: 15356 + SPARC port now includes V8 plus and V9 support as well as 15357 performance tuning for Ultra class machines. The SPARC port 15358 now uses the Haifa scheduler. 15359 + Alpha port has been tuned for the EV6 processor and has an 15360 optimized expansion of memcpy/bzero. The Alpha port now uses 15361 the Haifa scheduler. 15362 + RS6000/PowerPC: support for the Power64 architecture and AIX 15363 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 15364 + x86: Alignment of static store data and jump targets is per 15365 Intel recommendations now. Various improvements throughout the 15366 x86 port to improve performance on Pentium processors 15367 (including improved epilogue sequences for Pentium chips and 15368 backend improvements which should help register allocation on 15369 all x86 variants. Conditional move support has been fixed and 15370 enabled for PPro processors. The x86 port also better supports 15371 64bit operations now. Unixware 7, a System V Release 5 target, 15372 is now supported and SCO OpenServer targets can support GAS. 15373 + MIPS has improved multiply/multiply-add support and now 15374 includes mips16 ISA support. 15375 + M68k has many micro-optimizations and Coldfire fixes. 15376 * Core compiler is based on the GCC development tree from June 9, 15377 1998, so we have all of the [4]features found in GCC 2.8. 15378 15379 15380 For questions related to the use of GCC, please consult these web 15381 pages and the [5]GCC manuals. If that fails, the 15382 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these 15383 web pages and the development of GCC are welcome on our developer 15384 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives. 15385 15386 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 15387 distribution of this entire article is permitted in any medium, 15388 provided this notice is preserved. 15389 15390 These pages are [10]maintained by the GCC team. Last modified 15391 2016-01-30[11]. 15392 15393References 15394 15395 1. http://gcc.gnu.org/egcs-1.1/c++features.html 15396 2. http://gcc.gnu.org/news/gcse.html 15397 3. http://gcc.gnu.org/news/alias.html 15398 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 15399 5. https://gcc.gnu.org/onlinedocs/ 15400 6. mailto:gcc-help@gcc.gnu.org 15401 7. mailto:gcc@gcc.gnu.org 15402 8. https://gcc.gnu.org/lists.html 15403 9. http://www.fsf.org/ 15404 10. https://gcc.gnu.org/about.html 15405 11. http://validator.w3.org/check/referer 15406====================================================================== 15407http://gcc.gnu.org/egcs-1.1/caveats.html 15408 15409 EGCS 1.1 Caveats 15410 15411 * EGCS has an integrated libstdc++, but does not have an integrated 15412 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 15413 Lu has made a libg++-2.8.1.2 snapshot available which may work with 15414 EGCS. 15415 Note most C++ programs only need libstdc++. 15416 * Exception handling may not work with shared libraries, particularly 15417 on alphas, hppas, rs6000/powerpc and mips based platforms. 15418 Exception handling is known to work on x86-linux platforms with 15419 shared libraries. 15420 * Some versions of the Linux kernel have bugs which prevent them from 15421 being compiled or from running when compiled by EGCS. See the FAQ 15422 (as shipped with EGCS 1.1) for additional information. 15423 * In general, EGCS is more rigorous about rejecting invalid C++ code 15424 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 15425 a result it may be necessary to fix C++ code before it will compile 15426 with EGCS. 15427 * G++ is also converting toward the ISO C++ standard; as a result 15428 code which was previously valid (and thus accepted by other 15429 compilers and older versions of g++) may no longer be accepted. 15430 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 15431 or GCC 2.8.x due to changes necessary to support thread safe 15432 exception handling. 15433 15434 15435 For questions related to the use of GCC, please consult these web 15436 pages and the [1]GCC manuals. If that fails, the 15437 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these 15438 web pages and the development of GCC are welcome on our developer 15439 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives. 15440 15441 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 15442 distribution of this entire article is permitted in any medium, 15443 provided this notice is preserved. 15444 15445 These pages are [6]maintained by the GCC team. Last modified 15446 2016-01-30[7]. 15447 15448References 15449 15450 1. https://gcc.gnu.org/onlinedocs/ 15451 2. mailto:gcc-help@gcc.gnu.org 15452 3. mailto:gcc@gcc.gnu.org 15453 4. https://gcc.gnu.org/lists.html 15454 5. http://www.fsf.org/ 15455 6. https://gcc.gnu.org/about.html 15456 7. http://validator.w3.org/check/referer 15457====================================================================== 15458http://gcc.gnu.org/egcs-1.0/index.html 15459 15460 EGCS 1.0 15461 15462 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 15463 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 15464 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 15465 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 15466 15467 EGCS is a collaborative effort involving several groups of hackers 15468 using an open development model to accelerate development and testing 15469 of GNU compilers and runtime libraries. 15470 15471 An important goal of EGCS is to allow wide scale testing of 15472 experimental features and optimizations; therefore, EGCS contains some 15473 features and optimizations which are still under development. However, 15474 EGCS has been carefully tested and should be comparable in quality to 15475 most GCC releases. 15476 15477 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 15478 development sources; it contains nearly all of the new features found 15479 in GCC 2.8. 15480 15481 EGCS 1.0 also contains many improvements and features not found in GCC 15482 2.7 and even the GCC 2.8 series (which was released after the original 15483 EGCS 1.0 release). 15484 * Integrated C++ runtime libraries, including support for most major 15485 GNU/Linux systems! 15486 * The integrated libstdc++ library includes a verbatim copy of SGI's 15487 STL release. 15488 * Integrated GNU Fortran compiler. 15489 * New instruction scheduler. 15490 * New alias analysis code. 15491 15492 See the [1]new features page for a more complete list of new features. 15493 15494 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 15495 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 15496 EGCS 1.0 release: 15497 * Add support for Red Hat 5.0 Linux and better support for Linux 15498 systems using glibc2. 15499 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 15500 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 15501 fix these problems. 15502 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 15503 handling interfaces. 15504 To avoid future compatibility problems, we strongly urge anyone who 15505 is planning on distributing shared libraries that contain C++ code 15506 to upgrade to EGCS 1.0.1 first. 15507 Soon after EGCS 1.0 was released, the GCC developers made some 15508 incompatible changes in libgcc's exception handling interfaces. 15509 These changes were needed to solve problems on some platforms. This 15510 means that GCC 2.8.0, when released, will not be seamlessly 15511 compatible with shared libraries built by EGCS 1.0. The reason is 15512 that the libgcc.a in GCC 2.8.0 will not contain a function needed 15513 by the old interface. 15514 The result of this is that there may be compatibility problems with 15515 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 15516 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 15517 and libgcc.a has the support routines for both the old and the new 15518 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 15519 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 15520 The maintainers of GCC 2.x have decided against including seamless 15521 support for the old interface in 2.8.0, since it was never 15522 "official", so to avoid future compatibility problems we recommend 15523 against distributing any shared libraries built by EGCS 1.0 that 15524 contain C++ code (upgrade to 1.0.1 and use that). 15525 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 15526 The x86 changes fix code generation errors exposed when building 15527 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 15528 The hppa change fixes a compiler abort when configured for use with 15529 RTEMS. 15530 The MIPS changes fix problems with the definition of LONG_MAX on 15531 newer systems, allow for command line selection of the target ABI, 15532 and fix one code generation problem. 15533 The rs6000/ppc change fixes some problems with passing structures 15534 to varargs/stdarg functions. 15535 * A few machine independent bugfixes, mostly to fix code generation 15536 errors when building Linux kernels or glibc. 15537 * Fix a few critical exception handling and template bugs in the C++ 15538 compiler. 15539 * Fix Fortran namelist bug on alphas. 15540 * Fix build problems on x86-solaris systems. 15541 15542 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 15543 serious problems in EGCS 1.0.1. 15544 * General improvements and fixes 15545 + Memory consumption significantly reduced, especially for 15546 templates and inline functions. 15547 + Fix various problems with glibc2.1. 15548 + Fix loop optimization bug exposed by rs6000/ppc port. 15549 + Fix to avoid potential code generation problems in jump.c. 15550 + Fix some undefined symbol problems in dwarf1 debug support. 15551 * g++/libstdc++ improvements and fixes 15552 + libstdc++ in the EGCS release has been updated and should be 15553 link compatible with libstdc++-2.8. 15554 + Various fixes in libio/libstdc++ to work better on GNU/Linux 15555 systems. 15556 + Fix problems with duplicate symbols on systems that do not 15557 support weak symbols. 15558 + Memory corruption bug and undefined symbols in bastring have 15559 been fixed. 15560 + Various exception handling fixes. 15561 + Fix compiler abort for very long thunk names. 15562 * g77 improvements and fixes 15563 + Fix compiler crash for omitted bound in Fortran CASE 15564 statement. 15565 + Add missing entries to g77 lang-options. 15566 + Fix problem with -fpedantic in the g77 compiler. 15567 + Fix "backspace" problem with g77 on alphas. 15568 + Fix x86 backend problem with Fortran literals and -fpic. 15569 + Fix some of the problems with negative subscripts for g77 on 15570 alphas. 15571 + Fixes for Fortran builds on cygwin32/mingw32. 15572 * platform specific improvements and fixes 15573 + Fix long double problems on x86 (exposed by glibc). 15574 + x86 ports define i386 again to keep imake happy. 15575 + Fix exception handling support on NetBSD ports. 15576 + Several changes to collect2 to fix many problems with AIX. 15577 + Define __ELF__ for GNU/Linux on rs6000. 15578 + Fix -mcall-linux problem on GNU/Linux on rs6000. 15579 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 15580 + Allow autoconf to select a proper install problem on AIX 3.1. 15581 + m68k port support includes -mcpu32 option as well as cpu32 15582 multilibs. 15583 + Fix stdarg bug for irix6. 15584 + Allow EGCS to build on irix5 without the gnu assembler. 15585 + Fix problem with static linking on sco5. 15586 + Fix bootstrap on sco5 with native compiler. 15587 + Fix for abort building newlib on H8 target. 15588 + Fix fixincludes handling of math.h on SunOS. 15589 + Minor fix for Motorola 3300 m68k systems. 15590 15591 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 15592 problems reported by Red Hat for builds of Red Hat 5.1. 15593 * Generic bugfixes: 15594 + Fix a typo in the libio library which resulted in incorrect 15595 behavior of istream::get. 15596 + Fix the Fortran negative array index problem. 15597 + Fix a major problem with the ObjC runtime thread support 15598 exposed by glibc2. 15599 + Reduce memory consumption of the Haifa scheduler. 15600 * Target specific bugfixes: 15601 + Fix one x86 floating point code generation bug exposed by 15602 glibc2 builds. 15603 + Fix one x86 internal compiler error exposed by glibc2 builds. 15604 + Fix profiling bugs on the Alpha. 15605 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 15606 + Fix rs6000/ppc bug when converting values from integer types 15607 to floating point types. 15608 15609 The EGCS 1.0 releases include installation instructions in both HTML 15610 and plaintext forms (see the INSTALL directory in the toplevel 15611 directory of the distribution). However, we also keep the most up to 15612 date installation instructions and [2]build/test status on our web 15613 page. We will update those pages as new information becomes available. 15614 15615 And, we can't in good conscience fail to mention some [3]caveats to 15616 using EGCS. 15617 15618 Update: Big thanks to Stanford for providing a high speed link for 15619 downloading EGCS (go.cygnus.com)! 15620 15621 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 15622 (USA California -- High speed link provided by Stanford). 15623 15624 The EGCS 1.0 release is also available many mirror sites. 15625 [4]Goto mirror list to find a closer site 15626 15627 We'd like to thank the numerous people that have contributed new 15628 features, test results, bugfixes, etc. Unfortunately, they're far too 15629 numerous to mention by name. 15630 15631 15632 For questions related to the use of GCC, please consult these web 15633 pages and the [5]GCC manuals. If that fails, the 15634 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these 15635 web pages and the development of GCC are welcome on our developer 15636 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives. 15637 15638 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 15639 distribution of this entire article is permitted in any medium, 15640 provided this notice is preserved. 15641 15642 These pages are [10]maintained by the GCC team. Last modified 15643 2016-01-30[11]. 15644 15645References 15646 15647 1. http://gcc.gnu.org/egcs-1.0/features.html 15648 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 15649 3. http://gcc.gnu.org/egcs-1.0/caveats.html 15650 4. http://gcc.gnu.org/mirrors.html 15651 5. https://gcc.gnu.org/onlinedocs/ 15652 6. mailto:gcc-help@gcc.gnu.org 15653 7. mailto:gcc@gcc.gnu.org 15654 8. https://gcc.gnu.org/lists.html 15655 9. http://www.fsf.org/ 15656 10. https://gcc.gnu.org/about.html 15657 11. http://validator.w3.org/check/referer 15658====================================================================== 15659http://gcc.gnu.org/egcs-1.0/features.html 15660 15661 EGCS 1.0 features 15662 15663 * Core compiler is based on the gcc2 development tree from Aug 2, 15664 1997, so we have most of the [1]features found in GCC 2.8. 15665 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 15666 * Vast improvements in the C++ compiler; so many they have [2]page of 15667 their own! 15668 * Integrated C++ runtime libraries, including support for most major 15669 GNU/Linux systems! 15670 * New instruction scheduler from IBM Haifa which includes support for 15671 function wide instruction scheduling as well as superscalar 15672 scheduling. 15673 * Significantly improved alias analysis code. 15674 * Improved register allocation for two address machines. 15675 * Significant code generation improvements for Fortran code on 15676 Alphas. 15677 * Various optimizations from the g77 project as well as improved loop 15678 optimizations. 15679 * Dwarf2 debug format support for some targets. 15680 * egcs libstdc++ includes the SGI STL implementation without changes. 15681 * As a result of these and other changes, egcs libstc++ is not binary 15682 compatible with previous releases of libstdc++. 15683 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 15684 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 15685 1.1), Support for RTEMS on several embedded targets, Support for 15686 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 15687 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 15688 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 15689 * RS6000/PowerPC ports generate code which can run on all 15690 RS6000/PowerPC variants by default. 15691 * -mcpu= and -march= switches for the x86 port to allow better 15692 control over how the x86 port generates code. 15693 * Includes the template repository patch (aka repo patch); note the 15694 new template code makes repo obsolete for ELF systems using gnu-ld 15695 such as GNU/Linux. 15696 * Plus the usual assortment of bugfixes and improvements. 15697 15698 15699 For questions related to the use of GCC, please consult these web 15700 pages and the [3]GCC manuals. If that fails, the 15701 [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these 15702 web pages and the development of GCC are welcome on our developer 15703 list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives. 15704 15705 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 15706 distribution of this entire article is permitted in any medium, 15707 provided this notice is preserved. 15708 15709 These pages are [8]maintained by the GCC team. Last modified 15710 2016-01-30[9]. 15711 15712References 15713 15714 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 15715 2. http://gcc.gnu.org/egcs-1.0/c++features.html 15716 3. https://gcc.gnu.org/onlinedocs/ 15717 4. mailto:gcc-help@gcc.gnu.org 15718 5. mailto:gcc@gcc.gnu.org 15719 6. https://gcc.gnu.org/lists.html 15720 7. http://www.fsf.org/ 15721 8. https://gcc.gnu.org/about.html 15722 9. http://validator.w3.org/check/referer 15723====================================================================== 15724http://gcc.gnu.org/egcs-1.0/caveats.html 15725 15726 EGCS 1.0 Caveats 15727 15728 * EGCS has an integrated libstdc++, but does not have an integrated 15729 libg++. Furthermore old libg++ releases will not work with egc; HJ 15730 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 15731 Note most C++ programs only need libstdc++. 15732 * Note that using -pedantic or -Wreturn-type can cause an explosion 15733 in the amount of memory needed for template-heavy C++ code, such as 15734 code that uses STL. Also note that -Wall includes -Wreturn-type, so 15735 if you use -Wall you will need to specify -Wno-return-type to turn 15736 it off. 15737 * Exception handling may not work with shared libraries, particularly 15738 on alphas, hppas, and mips based platforms. Exception handling is 15739 known to work on x86-linux platforms with shared libraries. 15740 * Some versions of the Linux kernel have bugs which prevent them from 15741 being compiled or from running when compiled by EGCS. See the FAQ 15742 (as shipped with EGCS 1.0) for additional information. 15743 * In general, EGCS is more rigorous about rejecting invalid C++ code 15744 or deprecated C++ constructs than G++ 2.7. As a result it may be 15745 necessary to fix C++ code before it will compile with EGCS. 15746 * G++ is also aggressively tracking the C++ standard; as a result 15747 code which was previously valid (and thus accepted by other 15748 compilers and older versions of G++) may no longer be accepted. 15749 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 15750 1.0.x and later releases should work with Red Hat Linux 5.0. 15751 15752 15753 For questions related to the use of GCC, please consult these web 15754 pages and the [1]GCC manuals. If that fails, the 15755 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these 15756 web pages and the development of GCC are welcome on our developer 15757 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives. 15758 15759 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 15760 distribution of this entire article is permitted in any medium, 15761 provided this notice is preserved. 15762 15763 These pages are [6]maintained by the GCC team. Last modified 15764 2016-01-30[7]. 15765 15766References 15767 15768 1. https://gcc.gnu.org/onlinedocs/ 15769 2. mailto:gcc-help@gcc.gnu.org 15770 3. mailto:gcc@gcc.gnu.org 15771 4. https://gcc.gnu.org/lists.html 15772 5. http://www.fsf.org/ 15773 6. https://gcc.gnu.org/about.html 15774 7. http://validator.w3.org/check/referer 15775====================================================================== 15776