unifdef.c revision 279639
1/* 2 * Copyright (c) 2002 - 2014 Tony Finch <dot@dotat.at> 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 */ 25 26/* 27 * unifdef - remove ifdef'ed lines 28 * 29 * This code was derived from software contributed to Berkeley by Dave Yost. 30 * It was rewritten to support ANSI C by Tony Finch. The original version 31 * of unifdef carried the 4-clause BSD copyright licence. None of its code 32 * remains in this version (though some of the names remain) so it now 33 * carries a more liberal licence. 34 * 35 * Wishlist: 36 * provide an option which will append the name of the 37 * appropriate symbol after #else's and #endif's 38 * provide an option which will check symbols after 39 * #else's and #endif's to see that they match their 40 * corresponding #ifdef or #ifndef 41 * 42 * These require better buffer handling, which would also make 43 * it possible to handle all "dodgy" directives correctly. 44 */ 45 46#include "unifdef.h" 47 48static const char copyright[] = 49 "@(#) $Version: unifdef-2.10 $\n" 50 "@(#) $FreeBSD: stable/10/usr.bin/unifdef/unifdef.c 279639 2015-03-05 09:39:29Z hselasky $\n" 51 "@(#) $Author: Tony Finch (dot@dotat.at) $\n" 52 "@(#) $URL: http://dotat.at/prog/unifdef $\n" 53; 54 55/* types of input lines: */ 56typedef enum { 57 LT_TRUEI, /* a true #if with ignore flag */ 58 LT_FALSEI, /* a false #if with ignore flag */ 59 LT_IF, /* an unknown #if */ 60 LT_TRUE, /* a true #if */ 61 LT_FALSE, /* a false #if */ 62 LT_ELIF, /* an unknown #elif */ 63 LT_ELTRUE, /* a true #elif */ 64 LT_ELFALSE, /* a false #elif */ 65 LT_ELSE, /* #else */ 66 LT_ENDIF, /* #endif */ 67 LT_DODGY, /* flag: directive is not on one line */ 68 LT_DODGY_LAST = LT_DODGY + LT_ENDIF, 69 LT_PLAIN, /* ordinary line */ 70 LT_EOF, /* end of file */ 71 LT_ERROR, /* unevaluable #if */ 72 LT_COUNT 73} Linetype; 74 75static char const * const linetype_name[] = { 76 "TRUEI", "FALSEI", "IF", "TRUE", "FALSE", 77 "ELIF", "ELTRUE", "ELFALSE", "ELSE", "ENDIF", 78 "DODGY TRUEI", "DODGY FALSEI", 79 "DODGY IF", "DODGY TRUE", "DODGY FALSE", 80 "DODGY ELIF", "DODGY ELTRUE", "DODGY ELFALSE", 81 "DODGY ELSE", "DODGY ENDIF", 82 "PLAIN", "EOF", "ERROR" 83}; 84 85#define linetype_if2elif(lt) ((Linetype)(lt - LT_IF + LT_ELIF)) 86#define linetype_2dodgy(lt) ((Linetype)(lt + LT_DODGY)) 87 88/* state of #if processing */ 89typedef enum { 90 IS_OUTSIDE, 91 IS_FALSE_PREFIX, /* false #if followed by false #elifs */ 92 IS_TRUE_PREFIX, /* first non-false #(el)if is true */ 93 IS_PASS_MIDDLE, /* first non-false #(el)if is unknown */ 94 IS_FALSE_MIDDLE, /* a false #elif after a pass state */ 95 IS_TRUE_MIDDLE, /* a true #elif after a pass state */ 96 IS_PASS_ELSE, /* an else after a pass state */ 97 IS_FALSE_ELSE, /* an else after a true state */ 98 IS_TRUE_ELSE, /* an else after only false states */ 99 IS_FALSE_TRAILER, /* #elifs after a true are false */ 100 IS_COUNT 101} Ifstate; 102 103static char const * const ifstate_name[] = { 104 "OUTSIDE", "FALSE_PREFIX", "TRUE_PREFIX", 105 "PASS_MIDDLE", "FALSE_MIDDLE", "TRUE_MIDDLE", 106 "PASS_ELSE", "FALSE_ELSE", "TRUE_ELSE", 107 "FALSE_TRAILER" 108}; 109 110/* state of comment parser */ 111typedef enum { 112 NO_COMMENT = false, /* outside a comment */ 113 C_COMMENT, /* in a comment like this one */ 114 CXX_COMMENT, /* between // and end of line */ 115 STARTING_COMMENT, /* just after slash-backslash-newline */ 116 FINISHING_COMMENT, /* star-backslash-newline in a C comment */ 117 CHAR_LITERAL, /* inside '' */ 118 STRING_LITERAL /* inside "" */ 119} Comment_state; 120 121static char const * const comment_name[] = { 122 "NO", "C", "CXX", "STARTING", "FINISHING", "CHAR", "STRING" 123}; 124 125/* state of preprocessor line parser */ 126typedef enum { 127 LS_START, /* only space and comments on this line */ 128 LS_HASH, /* only space, comments, and a hash */ 129 LS_DIRTY /* this line can't be a preprocessor line */ 130} Line_state; 131 132static char const * const linestate_name[] = { 133 "START", "HASH", "DIRTY" 134}; 135 136/* 137 * Minimum translation limits from ISO/IEC 9899:1999 5.2.4.1 138 */ 139#define MAXDEPTH 64 /* maximum #if nesting */ 140#define MAXLINE 4096 /* maximum length of line */ 141#define MAXSYMS 16384 /* maximum number of symbols */ 142 143/* 144 * Sometimes when editing a keyword the replacement text is longer, so 145 * we leave some space at the end of the tline buffer to accommodate this. 146 */ 147#define EDITSLOP 10 148 149/* 150 * Globals. 151 */ 152 153static bool compblank; /* -B: compress blank lines */ 154static bool lnblank; /* -b: blank deleted lines */ 155static bool complement; /* -c: do the complement */ 156static bool debugging; /* -d: debugging reports */ 157static bool inplace; /* -m: modify in place */ 158static bool iocccok; /* -e: fewer IOCCC errors */ 159static bool strictlogic; /* -K: keep ambiguous #ifs */ 160static bool killconsts; /* -k: eval constant #ifs */ 161static bool lnnum; /* -n: add #line directives */ 162static bool symlist; /* -s: output symbol list */ 163static bool symdepth; /* -S: output symbol depth */ 164static bool text; /* -t: this is a text file */ 165 166static const char *symname[MAXSYMS]; /* symbol name */ 167static const char *value[MAXSYMS]; /* -Dsym=value */ 168static bool ignore[MAXSYMS]; /* -iDsym or -iUsym */ 169static int nsyms; /* number of symbols */ 170 171static FILE *input; /* input file pointer */ 172static const char *filename; /* input file name */ 173static int linenum; /* current line number */ 174static const char *linefile; /* file name for #line */ 175static FILE *output; /* output file pointer */ 176static const char *ofilename; /* output file name */ 177static const char *backext; /* backup extension */ 178static char *tempname; /* avoid splatting input */ 179 180static char tline[MAXLINE+EDITSLOP];/* input buffer plus space */ 181static char *keyword; /* used for editing #elif's */ 182 183/* 184 * When processing a file, the output's newline style will match the 185 * input's, and unifdef correctly handles CRLF or LF endings whatever 186 * the platform's native style. The stdio streams are opened in binary 187 * mode to accommodate platforms whose native newline style is CRLF. 188 * When the output isn't a processed input file (when it is error / 189 * debug / diagnostic messages) then unifdef uses native line endings. 190 */ 191 192static const char *newline; /* input file format */ 193static const char newline_unix[] = "\n"; 194static const char newline_crlf[] = "\r\n"; 195 196static Comment_state incomment; /* comment parser state */ 197static Line_state linestate; /* #if line parser state */ 198static Ifstate ifstate[MAXDEPTH]; /* #if processor state */ 199static bool ignoring[MAXDEPTH]; /* ignore comments state */ 200static int stifline[MAXDEPTH]; /* start of current #if */ 201static int depth; /* current #if nesting */ 202static int delcount; /* count of deleted lines */ 203static unsigned blankcount; /* count of blank lines */ 204static unsigned blankmax; /* maximum recent blankcount */ 205static bool constexpr; /* constant #if expression */ 206static bool zerosyms; /* to format symdepth output */ 207static bool firstsym; /* ditto */ 208 209static int exitmode; /* exit status mode */ 210static int exitstat; /* program exit status */ 211 212static void addsym1(bool, bool, char *); 213static void addsym2(bool, const char *, const char *); 214static char *astrcat(const char *, const char *); 215static void cleantemp(void); 216static void closeio(void); 217static void debug(const char *, ...); 218static void debugsym(const char *, int); 219static bool defundef(void); 220static void defundefile(const char *); 221static void done(void); 222static void error(const char *); 223static int findsym(const char **); 224static void flushline(bool); 225static void hashline(void); 226static void help(void); 227static Linetype ifeval(const char **); 228static void ignoreoff(void); 229static void ignoreon(void); 230static void indirectsym(void); 231static void keywordedit(const char *); 232static const char *matchsym(const char *, const char *); 233static void nest(void); 234static Linetype parseline(void); 235static void process(void); 236static void processinout(const char *, const char *); 237static const char *skipargs(const char *); 238static const char *skipcomment(const char *); 239static const char *skiphash(void); 240static const char *skipline(const char *); 241static const char *skipsym(const char *); 242static void state(Ifstate); 243static void unnest(void); 244static void usage(void); 245static void version(void); 246static const char *xstrdup(const char *, const char *); 247 248#define endsym(c) (!isalnum((unsigned char)c) && c != '_') 249 250/* 251 * The main program. 252 */ 253int 254main(int argc, char *argv[]) 255{ 256 int opt; 257 258 while ((opt = getopt(argc, argv, "i:D:U:f:I:M:o:x:bBcdehKklmnsStV")) != -1) 259 switch (opt) { 260 case 'i': /* treat stuff controlled by these symbols as text */ 261 /* 262 * For strict backwards-compatibility the U or D 263 * should be immediately after the -i but it doesn't 264 * matter much if we relax that requirement. 265 */ 266 opt = *optarg++; 267 if (opt == 'D') 268 addsym1(true, true, optarg); 269 else if (opt == 'U') 270 addsym1(true, false, optarg); 271 else 272 usage(); 273 break; 274 case 'D': /* define a symbol */ 275 addsym1(false, true, optarg); 276 break; 277 case 'U': /* undef a symbol */ 278 addsym1(false, false, optarg); 279 break; 280 case 'I': /* no-op for compatibility with cpp */ 281 break; 282 case 'b': /* blank deleted lines instead of omitting them */ 283 case 'l': /* backwards compatibility */ 284 lnblank = true; 285 break; 286 case 'B': /* compress blank lines around removed section */ 287 compblank = true; 288 break; 289 case 'c': /* treat -D as -U and vice versa */ 290 complement = true; 291 break; 292 case 'd': 293 debugging = true; 294 break; 295 case 'e': /* fewer errors from dodgy lines */ 296 iocccok = true; 297 break; 298 case 'f': /* definitions file */ 299 defundefile(optarg); 300 break; 301 case 'h': 302 help(); 303 break; 304 case 'K': /* keep ambiguous #ifs */ 305 strictlogic = true; 306 break; 307 case 'k': /* process constant #ifs */ 308 killconsts = true; 309 break; 310 case 'm': /* modify in place */ 311 inplace = true; 312 break; 313 case 'M': /* modify in place and keep backup */ 314 inplace = true; 315 backext = optarg; 316 break; 317 case 'n': /* add #line directive after deleted lines */ 318 lnnum = true; 319 break; 320 case 'o': /* output to a file */ 321 ofilename = optarg; 322 break; 323 case 's': /* only output list of symbols that control #ifs */ 324 symlist = true; 325 break; 326 case 'S': /* list symbols with their nesting depth */ 327 symlist = symdepth = true; 328 break; 329 case 't': /* don't parse C comments */ 330 text = true; 331 break; 332 case 'V': 333 version(); 334 break; 335 case 'x': 336 exitmode = atoi(optarg); 337 if(exitmode < 0 || exitmode > 2) 338 usage(); 339 break; 340 default: 341 usage(); 342 } 343 argc -= optind; 344 argv += optind; 345 if (compblank && lnblank) 346 errx(2, "-B and -b are mutually exclusive"); 347 if (symlist && (ofilename != NULL || inplace || argc > 1)) 348 errx(2, "-s only works with one input file"); 349 if (argc > 1 && ofilename != NULL) 350 errx(2, "-o cannot be used with multiple input files"); 351 if (argc > 1 && !inplace) 352 errx(2, "multiple input files require -m or -M"); 353 if (argc == 0) 354 argc = 1; 355 if (argc == 1 && !inplace && ofilename == NULL) 356 ofilename = "-"; 357 indirectsym(); 358 359 atexit(cleantemp); 360 if (ofilename != NULL) 361 processinout(*argv, ofilename); 362 else while (argc-- > 0) { 363 processinout(*argv, *argv); 364 argv++; 365 } 366 switch(exitmode) { 367 case(0): exit(exitstat); 368 case(1): exit(!exitstat); 369 case(2): exit(0); 370 default: abort(); /* bug */ 371 } 372} 373 374/* 375 * File logistics. 376 */ 377static void 378processinout(const char *ifn, const char *ofn) 379{ 380 struct stat st; 381 382 if (ifn == NULL || strcmp(ifn, "-") == 0) { 383 filename = "[stdin]"; 384 linefile = NULL; 385 input = fbinmode(stdin); 386 } else { 387 filename = ifn; 388 linefile = ifn; 389 input = fopen(ifn, "rb"); 390 if (input == NULL) 391 err(2, "can't open %s", ifn); 392 } 393 if (strcmp(ofn, "-") == 0) { 394 output = fbinmode(stdout); 395 process(); 396 return; 397 } 398 if (stat(ofn, &st) < 0) { 399 output = fopen(ofn, "wb"); 400 if (output == NULL) 401 err(2, "can't create %s", ofn); 402 process(); 403 return; 404 } 405 406 tempname = astrcat(ofn, ".XXXXXX"); 407 output = mktempmode(tempname, st.st_mode); 408 if (output == NULL) 409 err(2, "can't create %s", tempname); 410 411 process(); 412 413 if (backext != NULL) { 414 char *backname = astrcat(ofn, backext); 415 if (rename(ofn, backname) < 0) 416 err(2, "can't rename \"%s\" to \"%s\"", ofn, backname); 417 free(backname); 418 } 419 if (replace(tempname, ofn) < 0) 420 err(2, "can't rename \"%s\" to \"%s\"", tempname, ofn); 421 free(tempname); 422 tempname = NULL; 423} 424 425/* 426 * For cleaning up if there is an error. 427 */ 428static void 429cleantemp(void) 430{ 431 if (tempname != NULL) 432 remove(tempname); 433} 434 435/* 436 * Self-identification functions. 437 */ 438 439static void 440version(void) 441{ 442 const char *c = copyright; 443 for (;;) { 444 while (*++c != '$') 445 if (*c == '\0') 446 exit(0); 447 while (*++c != '$') 448 putc(*c, stderr); 449 putc('\n', stderr); 450 } 451} 452 453static void 454synopsis(FILE *fp) 455{ 456 fprintf(fp, 457 "usage: unifdef [-bBcdehKkmnsStV] [-x{012}] [-Mext] [-opath] \\\n" 458 " [-[i]Dsym[=val]] [-[i]Usym] [-fpath] ... [file] ...\n"); 459} 460 461static void 462usage(void) 463{ 464 synopsis(stderr); 465 exit(2); 466} 467 468static void 469help(void) 470{ 471 synopsis(stdout); 472 printf( 473 " -Dsym=val define preprocessor symbol with given value\n" 474 " -Dsym define preprocessor symbol with value 1\n" 475 " -Usym preprocessor symbol is undefined\n" 476 " -iDsym=val \\ ignore C strings and comments\n" 477 " -iDsym ) in sections controlled by these\n" 478 " -iUsym / preprocessor symbols\n" 479 " -fpath file containing #define and #undef directives\n" 480 " -b blank lines instead of deleting them\n" 481 " -B compress blank lines around deleted section\n" 482 " -c complement (invert) keep vs. delete\n" 483 " -d debugging mode\n" 484 " -e ignore multiline preprocessor directives\n" 485 " -h print help\n" 486 " -Ipath extra include file path (ignored)\n" 487 " -K disable && and || short-circuiting\n" 488 " -k process constant #if expressions\n" 489 " -Mext modify in place and keep backups\n" 490 " -m modify input files in place\n" 491 " -n add #line directives to output\n" 492 " -opath output file name\n" 493 " -S list #if control symbols with nesting\n" 494 " -s list #if control symbols\n" 495 " -t ignore C strings and comments\n" 496 " -V print version\n" 497 " -x{012} exit status mode\n" 498 ); 499 exit(0); 500} 501 502/* 503 * A state transition function alters the global #if processing state 504 * in a particular way. The table below is indexed by the current 505 * processing state and the type of the current line. 506 * 507 * Nesting is handled by keeping a stack of states; some transition 508 * functions increase or decrease the depth. They also maintain the 509 * ignore state on a stack. In some complicated cases they have to 510 * alter the preprocessor directive, as follows. 511 * 512 * When we have processed a group that starts off with a known-false 513 * #if/#elif sequence (which has therefore been deleted) followed by a 514 * #elif that we don't understand and therefore must keep, we edit the 515 * latter into a #if to keep the nesting correct. We use memcpy() to 516 * overwrite the 4 byte token "elif" with "if " without a '\0' byte. 517 * 518 * When we find a true #elif in a group, the following block will 519 * always be kept and the rest of the sequence after the next #elif or 520 * #else will be discarded. We edit the #elif into a #else and the 521 * following directive to #endif since this has the desired behaviour. 522 * 523 * "Dodgy" directives are split across multiple lines, the most common 524 * example being a multi-line comment hanging off the right of the 525 * directive. We can handle them correctly only if there is no change 526 * from printing to dropping (or vice versa) caused by that directive. 527 * If the directive is the first of a group we have a choice between 528 * failing with an error, or passing it through unchanged instead of 529 * evaluating it. The latter is not the default to avoid questions from 530 * users about unifdef unexpectedly leaving behind preprocessor directives. 531 */ 532typedef void state_fn(void); 533 534/* report an error */ 535static void Eelif (void) { error("Inappropriate #elif"); } 536static void Eelse (void) { error("Inappropriate #else"); } 537static void Eendif(void) { error("Inappropriate #endif"); } 538static void Eeof (void) { error("Premature EOF"); } 539static void Eioccc(void) { error("Obfuscated preprocessor control line"); } 540/* plain line handling */ 541static void print (void) { flushline(true); } 542static void drop (void) { flushline(false); } 543/* output lacks group's start line */ 544static void Strue (void) { drop(); ignoreoff(); state(IS_TRUE_PREFIX); } 545static void Sfalse(void) { drop(); ignoreoff(); state(IS_FALSE_PREFIX); } 546static void Selse (void) { drop(); state(IS_TRUE_ELSE); } 547/* print/pass this block */ 548static void Pelif (void) { print(); ignoreoff(); state(IS_PASS_MIDDLE); } 549static void Pelse (void) { print(); state(IS_PASS_ELSE); } 550static void Pendif(void) { print(); unnest(); } 551/* discard this block */ 552static void Dfalse(void) { drop(); ignoreoff(); state(IS_FALSE_TRAILER); } 553static void Delif (void) { drop(); ignoreoff(); state(IS_FALSE_MIDDLE); } 554static void Delse (void) { drop(); state(IS_FALSE_ELSE); } 555static void Dendif(void) { drop(); unnest(); } 556/* first line of group */ 557static void Fdrop (void) { nest(); Dfalse(); } 558static void Fpass (void) { nest(); Pelif(); } 559static void Ftrue (void) { nest(); Strue(); } 560static void Ffalse(void) { nest(); Sfalse(); } 561/* variable pedantry for obfuscated lines */ 562static void Oiffy (void) { if (!iocccok) Eioccc(); Fpass(); ignoreon(); } 563static void Oif (void) { if (!iocccok) Eioccc(); Fpass(); } 564static void Oelif (void) { if (!iocccok) Eioccc(); Pelif(); } 565/* ignore comments in this block */ 566static void Idrop (void) { Fdrop(); ignoreon(); } 567static void Itrue (void) { Ftrue(); ignoreon(); } 568static void Ifalse(void) { Ffalse(); ignoreon(); } 569/* modify this line */ 570static void Mpass (void) { memcpy(keyword, "if ", 4); Pelif(); } 571static void Mtrue (void) { keywordedit("else"); state(IS_TRUE_MIDDLE); } 572static void Melif (void) { keywordedit("endif"); state(IS_FALSE_TRAILER); } 573static void Melse (void) { keywordedit("endif"); state(IS_FALSE_ELSE); } 574 575static state_fn * const trans_table[IS_COUNT][LT_COUNT] = { 576/* IS_OUTSIDE */ 577{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Eendif, 578 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eendif, 579 print, done, abort }, 580/* IS_FALSE_PREFIX */ 581{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Strue, Sfalse,Selse, Dendif, 582 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Eioccc,Eioccc,Eioccc,Eioccc, 583 drop, Eeof, abort }, 584/* IS_TRUE_PREFIX */ 585{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Dfalse,Dfalse,Dfalse,Delse, Dendif, 586 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, 587 print, Eeof, abort }, 588/* IS_PASS_MIDDLE */ 589{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Pelif, Mtrue, Delif, Pelse, Pendif, 590 Oiffy, Oiffy, Fpass, Oif, Oif, Pelif, Oelif, Oelif, Pelse, Pendif, 591 print, Eeof, abort }, 592/* IS_FALSE_MIDDLE */ 593{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Pelif, Mtrue, Delif, Pelse, Pendif, 594 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, 595 drop, Eeof, abort }, 596/* IS_TRUE_MIDDLE */ 597{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Melif, Melif, Melif, Melse, Pendif, 598 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Pendif, 599 print, Eeof, abort }, 600/* IS_PASS_ELSE */ 601{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Pendif, 602 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Pendif, 603 print, Eeof, abort }, 604/* IS_FALSE_ELSE */ 605{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Dendif, 606 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Eioccc, 607 drop, Eeof, abort }, 608/* IS_TRUE_ELSE */ 609{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Dendif, 610 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eioccc, 611 print, Eeof, abort }, 612/* IS_FALSE_TRAILER */ 613{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Dendif, 614 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Eioccc, 615 drop, Eeof, abort } 616/*TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF 617 TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF (DODGY) 618 PLAIN EOF ERROR */ 619}; 620 621/* 622 * State machine utility functions 623 */ 624static void 625ignoreoff(void) 626{ 627 if (depth == 0) 628 abort(); /* bug */ 629 ignoring[depth] = ignoring[depth-1]; 630} 631static void 632ignoreon(void) 633{ 634 ignoring[depth] = true; 635} 636static void 637keywordedit(const char *replacement) 638{ 639 snprintf(keyword, tline + sizeof(tline) - keyword, 640 "%s%s", replacement, newline); 641 print(); 642} 643static void 644nest(void) 645{ 646 if (depth > MAXDEPTH-1) 647 abort(); /* bug */ 648 if (depth == MAXDEPTH-1) 649 error("Too many levels of nesting"); 650 depth += 1; 651 stifline[depth] = linenum; 652} 653static void 654unnest(void) 655{ 656 if (depth == 0) 657 abort(); /* bug */ 658 depth -= 1; 659} 660static void 661state(Ifstate is) 662{ 663 ifstate[depth] = is; 664} 665 666/* 667 * The last state transition function. When this is called, 668 * lineval == LT_EOF, so the process() loop will terminate. 669 */ 670static void 671done(void) 672{ 673 if (incomment) 674 error("EOF in comment"); 675 closeio(); 676} 677 678/* 679 * Write a line to the output or not, according to command line options. 680 * If writing fails, closeio() will print the error and exit. 681 */ 682static void 683flushline(bool keep) 684{ 685 if (symlist) 686 return; 687 if (keep ^ complement) { 688 bool blankline = tline[strspn(tline, " \t\r\n")] == '\0'; 689 if (blankline && compblank && blankcount != blankmax) { 690 delcount += 1; 691 blankcount += 1; 692 } else { 693 if (lnnum && delcount > 0) 694 hashline(); 695 if (fputs(tline, output) == EOF) 696 closeio(); 697 delcount = 0; 698 blankmax = blankcount = blankline ? blankcount + 1 : 0; 699 } 700 } else { 701 if (lnblank && fputs(newline, output) == EOF) 702 closeio(); 703 exitstat = 1; 704 delcount += 1; 705 blankcount = 0; 706 } 707 if (debugging && fflush(output) == EOF) 708 closeio(); 709} 710 711/* 712 * Format of #line directives depends on whether we know the input filename. 713 */ 714static void 715hashline(void) 716{ 717 int e; 718 719 if (linefile == NULL) 720 e = fprintf(output, "#line %d%s", linenum, newline); 721 else 722 e = fprintf(output, "#line %d \"%s\"%s", 723 linenum, linefile, newline); 724 if (e < 0) 725 closeio(); 726} 727 728/* 729 * Flush the output and handle errors. 730 */ 731static void 732closeio(void) 733{ 734 /* Tidy up after findsym(). */ 735 if (symdepth && !zerosyms) 736 printf("\n"); 737 if (output != NULL && (ferror(output) || fclose(output) == EOF)) 738 err(2, "%s: can't write to output", filename); 739 fclose(input); 740} 741 742/* 743 * The driver for the state machine. 744 */ 745static void 746process(void) 747{ 748 Linetype lineval = LT_PLAIN; 749 /* When compressing blank lines, act as if the file 750 is preceded by a large number of blank lines. */ 751 blankmax = blankcount = 1000; 752 zerosyms = true; 753 newline = NULL; 754 linenum = 0; 755 while (lineval != LT_EOF) { 756 lineval = parseline(); 757 trans_table[ifstate[depth]][lineval](); 758 debug("process line %d %s -> %s depth %d", 759 linenum, linetype_name[lineval], 760 ifstate_name[ifstate[depth]], depth); 761 } 762} 763 764/* 765 * Parse a line and determine its type. We keep the preprocessor line 766 * parser state between calls in the global variable linestate, with 767 * help from skipcomment(). 768 */ 769static Linetype 770parseline(void) 771{ 772 const char *cp; 773 int cursym; 774 Linetype retval; 775 Comment_state wascomment; 776 777 wascomment = incomment; 778 cp = skiphash(); 779 if (cp == NULL) 780 return (LT_EOF); 781 if (newline == NULL) { 782 if (strrchr(tline, '\n') == strrchr(tline, '\r') + 1) 783 newline = newline_crlf; 784 else 785 newline = newline_unix; 786 } 787 if (*cp == '\0') { 788 retval = LT_PLAIN; 789 goto done; 790 } 791 keyword = tline + (cp - tline); 792 if ((cp = matchsym("ifdef", keyword)) != NULL || 793 (cp = matchsym("ifndef", keyword)) != NULL) { 794 cp = skipcomment(cp); 795 if ((cursym = findsym(&cp)) < 0) 796 retval = LT_IF; 797 else { 798 retval = (keyword[2] == 'n') 799 ? LT_FALSE : LT_TRUE; 800 if (value[cursym] == NULL) 801 retval = (retval == LT_TRUE) 802 ? LT_FALSE : LT_TRUE; 803 if (ignore[cursym]) 804 retval = (retval == LT_TRUE) 805 ? LT_TRUEI : LT_FALSEI; 806 } 807 } else if ((cp = matchsym("if", keyword)) != NULL) 808 retval = ifeval(&cp); 809 else if ((cp = matchsym("elif", keyword)) != NULL) 810 retval = linetype_if2elif(ifeval(&cp)); 811 else if ((cp = matchsym("else", keyword)) != NULL) 812 retval = LT_ELSE; 813 else if ((cp = matchsym("endif", keyword)) != NULL) 814 retval = LT_ENDIF; 815 else { 816 cp = skipsym(keyword); 817 /* no way can we deal with a continuation inside a keyword */ 818 if (strncmp(cp, "\\\r\n", 3) == 0 || 819 strncmp(cp, "\\\n", 2) == 0) 820 Eioccc(); 821 cp = skipline(cp); 822 retval = LT_PLAIN; 823 goto done; 824 } 825 cp = skipcomment(cp); 826 if (*cp != '\0') { 827 cp = skipline(cp); 828 if (retval == LT_TRUE || retval == LT_FALSE || 829 retval == LT_TRUEI || retval == LT_FALSEI) 830 retval = LT_IF; 831 if (retval == LT_ELTRUE || retval == LT_ELFALSE) 832 retval = LT_ELIF; 833 } 834 /* the following can happen if the last line of the file lacks a 835 newline or if there is too much whitespace in a directive */ 836 if (linestate == LS_HASH) { 837 long len = cp - tline; 838 if (fgets(tline + len, MAXLINE - len, input) == NULL) { 839 if (ferror(input)) 840 err(2, "can't read %s", filename); 841 /* append the missing newline at eof */ 842 strcpy(tline + len, newline); 843 cp += strlen(newline); 844 linestate = LS_START; 845 } else { 846 linestate = LS_DIRTY; 847 } 848 } 849 if (retval != LT_PLAIN && (wascomment || linestate != LS_START)) { 850 retval = linetype_2dodgy(retval); 851 linestate = LS_DIRTY; 852 } 853done: 854 debug("parser line %d state %s comment %s line", linenum, 855 comment_name[incomment], linestate_name[linestate]); 856 return (retval); 857} 858 859/* 860 * These are the binary operators that are supported by the expression 861 * evaluator. 862 */ 863static Linetype op_strict(long *p, long v, Linetype at, Linetype bt) { 864 if(at == LT_IF || bt == LT_IF) return (LT_IF); 865 return (*p = v, v ? LT_TRUE : LT_FALSE); 866} 867static Linetype op_lt(long *p, Linetype at, long a, Linetype bt, long b) { 868 return op_strict(p, a < b, at, bt); 869} 870static Linetype op_gt(long *p, Linetype at, long a, Linetype bt, long b) { 871 return op_strict(p, a > b, at, bt); 872} 873static Linetype op_le(long *p, Linetype at, long a, Linetype bt, long b) { 874 return op_strict(p, a <= b, at, bt); 875} 876static Linetype op_ge(long *p, Linetype at, long a, Linetype bt, long b) { 877 return op_strict(p, a >= b, at, bt); 878} 879static Linetype op_eq(long *p, Linetype at, long a, Linetype bt, long b) { 880 return op_strict(p, a == b, at, bt); 881} 882static Linetype op_ne(long *p, Linetype at, long a, Linetype bt, long b) { 883 return op_strict(p, a != b, at, bt); 884} 885static Linetype op_or(long *p, Linetype at, long a, Linetype bt, long b) { 886 if (!strictlogic && (at == LT_TRUE || bt == LT_TRUE)) 887 return (*p = 1, LT_TRUE); 888 return op_strict(p, a || b, at, bt); 889} 890static Linetype op_and(long *p, Linetype at, long a, Linetype bt, long b) { 891 if (!strictlogic && (at == LT_FALSE || bt == LT_FALSE)) 892 return (*p = 0, LT_FALSE); 893 return op_strict(p, a && b, at, bt); 894} 895 896/* 897 * An evaluation function takes three arguments, as follows: (1) a pointer to 898 * an element of the precedence table which lists the operators at the current 899 * level of precedence; (2) a pointer to an integer which will receive the 900 * value of the expression; and (3) a pointer to a char* that points to the 901 * expression to be evaluated and that is updated to the end of the expression 902 * when evaluation is complete. The function returns LT_FALSE if the value of 903 * the expression is zero, LT_TRUE if it is non-zero, LT_IF if the expression 904 * depends on an unknown symbol, or LT_ERROR if there is a parse failure. 905 */ 906struct ops; 907 908typedef Linetype eval_fn(const struct ops *, long *, const char **); 909 910static eval_fn eval_table, eval_unary; 911 912/* 913 * The precedence table. Expressions involving binary operators are evaluated 914 * in a table-driven way by eval_table. When it evaluates a subexpression it 915 * calls the inner function with its first argument pointing to the next 916 * element of the table. Innermost expressions have special non-table-driven 917 * handling. 918 */ 919struct op { 920 const char *str; 921 Linetype (*fn)(long *, Linetype, long, Linetype, long); 922}; 923struct ops { 924 eval_fn *inner; 925 struct op op[5]; 926}; 927static const struct ops eval_ops[] = { 928 { eval_table, { { "||", op_or } } }, 929 { eval_table, { { "&&", op_and } } }, 930 { eval_table, { { "==", op_eq }, 931 { "!=", op_ne } } }, 932 { eval_unary, { { "<=", op_le }, 933 { ">=", op_ge }, 934 { "<", op_lt }, 935 { ">", op_gt } } } 936}; 937 938/* Current operator precedence level */ 939static long prec(const struct ops *ops) 940{ 941 return (ops - eval_ops); 942} 943 944/* 945 * Function for evaluating the innermost parts of expressions, 946 * viz. !expr (expr) number defined(symbol) symbol 947 * We reset the constexpr flag in the last two cases. 948 */ 949static Linetype 950eval_unary(const struct ops *ops, long *valp, const char **cpp) 951{ 952 const char *cp; 953 char *ep; 954 int sym; 955 bool defparen; 956 Linetype lt; 957 958 cp = skipcomment(*cpp); 959 if (*cp == '!') { 960 debug("eval%d !", prec(ops)); 961 cp++; 962 lt = eval_unary(ops, valp, &cp); 963 if (lt == LT_ERROR) 964 return (LT_ERROR); 965 if (lt != LT_IF) { 966 *valp = !*valp; 967 lt = *valp ? LT_TRUE : LT_FALSE; 968 } 969 } else if (*cp == '(') { 970 cp++; 971 debug("eval%d (", prec(ops)); 972 lt = eval_table(eval_ops, valp, &cp); 973 if (lt == LT_ERROR) 974 return (LT_ERROR); 975 cp = skipcomment(cp); 976 if (*cp++ != ')') 977 return (LT_ERROR); 978 } else if (isdigit((unsigned char)*cp)) { 979 debug("eval%d number", prec(ops)); 980 *valp = strtol(cp, &ep, 0); 981 if (ep == cp) 982 return (LT_ERROR); 983 lt = *valp ? LT_TRUE : LT_FALSE; 984 cp = ep; 985 } else if (matchsym("defined", cp) != NULL) { 986 cp = skipcomment(cp+7); 987 if (*cp == '(') { 988 cp = skipcomment(cp+1); 989 defparen = true; 990 } else { 991 defparen = false; 992 } 993 sym = findsym(&cp); 994 cp = skipcomment(cp); 995 if (defparen && *cp++ != ')') { 996 debug("eval%d defined missing ')'", prec(ops)); 997 return (LT_ERROR); 998 } 999 if (sym < 0) { 1000 debug("eval%d defined unknown", prec(ops)); 1001 lt = LT_IF; 1002 } else { 1003 debug("eval%d defined %s", prec(ops), symname[sym]); 1004 *valp = (value[sym] != NULL); 1005 lt = *valp ? LT_TRUE : LT_FALSE; 1006 } 1007 constexpr = false; 1008 } else if (!endsym(*cp)) { 1009 debug("eval%d symbol", prec(ops)); 1010 sym = findsym(&cp); 1011 if (sym < 0) { 1012 lt = LT_IF; 1013 cp = skipargs(cp); 1014 } else if (value[sym] == NULL) { 1015 *valp = 0; 1016 lt = LT_FALSE; 1017 } else { 1018 *valp = strtol(value[sym], &ep, 0); 1019 if (*ep != '\0' || ep == value[sym]) 1020 return (LT_ERROR); 1021 lt = *valp ? LT_TRUE : LT_FALSE; 1022 cp = skipargs(cp); 1023 } 1024 constexpr = false; 1025 } else { 1026 debug("eval%d bad expr", prec(ops)); 1027 return (LT_ERROR); 1028 } 1029 1030 *cpp = cp; 1031 debug("eval%d = %d", prec(ops), *valp); 1032 return (lt); 1033} 1034 1035/* 1036 * Table-driven evaluation of binary operators. 1037 */ 1038static Linetype 1039eval_table(const struct ops *ops, long *valp, const char **cpp) 1040{ 1041 const struct op *op; 1042 const char *cp; 1043 long val; 1044 Linetype lt, rt; 1045 1046 debug("eval%d", prec(ops)); 1047 cp = *cpp; 1048 lt = ops->inner(ops+1, valp, &cp); 1049 if (lt == LT_ERROR) 1050 return (LT_ERROR); 1051 for (;;) { 1052 cp = skipcomment(cp); 1053 for (op = ops->op; op->str != NULL; op++) 1054 if (strncmp(cp, op->str, strlen(op->str)) == 0) 1055 break; 1056 if (op->str == NULL) 1057 break; 1058 cp += strlen(op->str); 1059 debug("eval%d %s", prec(ops), op->str); 1060 rt = ops->inner(ops+1, &val, &cp); 1061 if (rt == LT_ERROR) 1062 return (LT_ERROR); 1063 lt = op->fn(valp, lt, *valp, rt, val); 1064 } 1065 1066 *cpp = cp; 1067 debug("eval%d = %d", prec(ops), *valp); 1068 debug("eval%d lt = %s", prec(ops), linetype_name[lt]); 1069 return (lt); 1070} 1071 1072/* 1073 * Evaluate the expression on a #if or #elif line. If we can work out 1074 * the result we return LT_TRUE or LT_FALSE accordingly, otherwise we 1075 * return just a generic LT_IF. 1076 */ 1077static Linetype 1078ifeval(const char **cpp) 1079{ 1080 Linetype ret; 1081 long val = 0; 1082 1083 debug("eval %s", *cpp); 1084 constexpr = killconsts ? false : true; 1085 ret = eval_table(eval_ops, &val, cpp); 1086 debug("eval = %d", val); 1087 return (constexpr ? LT_IF : ret == LT_ERROR ? LT_IF : ret); 1088} 1089 1090/* 1091 * Read a line and examine its initial part to determine if it is a 1092 * preprocessor directive. Returns NULL on EOF, or a pointer to a 1093 * preprocessor directive name, or a pointer to the zero byte at the 1094 * end of the line. 1095 */ 1096static const char * 1097skiphash(void) 1098{ 1099 const char *cp; 1100 1101 linenum++; 1102 if (fgets(tline, MAXLINE, input) == NULL) { 1103 if (ferror(input)) 1104 err(2, "can't read %s", filename); 1105 else 1106 return (NULL); 1107 } 1108 cp = skipcomment(tline); 1109 if (linestate == LS_START && *cp == '#') { 1110 linestate = LS_HASH; 1111 return (skipcomment(cp + 1)); 1112 } else if (*cp == '\0') { 1113 return (cp); 1114 } else { 1115 return (skipline(cp)); 1116 } 1117} 1118 1119/* 1120 * Mark a line dirty and consume the rest of it, keeping track of the 1121 * lexical state. 1122 */ 1123static const char * 1124skipline(const char *cp) 1125{ 1126 if (*cp != '\0') 1127 linestate = LS_DIRTY; 1128 while (*cp != '\0') 1129 cp = skipcomment(cp + 1); 1130 return (cp); 1131} 1132 1133/* 1134 * Skip over comments, strings, and character literals and stop at the 1135 * next character position that is not whitespace. Between calls we keep 1136 * the comment state in the global variable incomment, and we also adjust 1137 * the global variable linestate when we see a newline. 1138 * XXX: doesn't cope with the buffer splitting inside a state transition. 1139 */ 1140static const char * 1141skipcomment(const char *cp) 1142{ 1143 if (text || ignoring[depth]) { 1144 for (; isspace((unsigned char)*cp); cp++) 1145 if (*cp == '\n') 1146 linestate = LS_START; 1147 return (cp); 1148 } 1149 while (*cp != '\0') 1150 /* don't reset to LS_START after a line continuation */ 1151 if (strncmp(cp, "\\\r\n", 3) == 0) 1152 cp += 3; 1153 else if (strncmp(cp, "\\\n", 2) == 0) 1154 cp += 2; 1155 else switch (incomment) { 1156 case NO_COMMENT: 1157 if (strncmp(cp, "/\\\r\n", 4) == 0) { 1158 incomment = STARTING_COMMENT; 1159 cp += 4; 1160 } else if (strncmp(cp, "/\\\n", 3) == 0) { 1161 incomment = STARTING_COMMENT; 1162 cp += 3; 1163 } else if (strncmp(cp, "/*", 2) == 0) { 1164 incomment = C_COMMENT; 1165 cp += 2; 1166 } else if (strncmp(cp, "//", 2) == 0) { 1167 incomment = CXX_COMMENT; 1168 cp += 2; 1169 } else if (strncmp(cp, "\'", 1) == 0) { 1170 incomment = CHAR_LITERAL; 1171 linestate = LS_DIRTY; 1172 cp += 1; 1173 } else if (strncmp(cp, "\"", 1) == 0) { 1174 incomment = STRING_LITERAL; 1175 linestate = LS_DIRTY; 1176 cp += 1; 1177 } else if (strncmp(cp, "\n", 1) == 0) { 1178 linestate = LS_START; 1179 cp += 1; 1180 } else if (strchr(" \r\t", *cp) != NULL) { 1181 cp += 1; 1182 } else 1183 return (cp); 1184 continue; 1185 case CXX_COMMENT: 1186 if (strncmp(cp, "\n", 1) == 0) { 1187 incomment = NO_COMMENT; 1188 linestate = LS_START; 1189 } 1190 cp += 1; 1191 continue; 1192 case CHAR_LITERAL: 1193 case STRING_LITERAL: 1194 if ((incomment == CHAR_LITERAL && cp[0] == '\'') || 1195 (incomment == STRING_LITERAL && cp[0] == '\"')) { 1196 incomment = NO_COMMENT; 1197 cp += 1; 1198 } else if (cp[0] == '\\') { 1199 if (cp[1] == '\0') 1200 cp += 1; 1201 else 1202 cp += 2; 1203 } else if (strncmp(cp, "\n", 1) == 0) { 1204 if (incomment == CHAR_LITERAL) 1205 error("unterminated char literal"); 1206 else 1207 error("unterminated string literal"); 1208 } else 1209 cp += 1; 1210 continue; 1211 case C_COMMENT: 1212 if (strncmp(cp, "*\\\r\n", 4) == 0) { 1213 incomment = FINISHING_COMMENT; 1214 cp += 4; 1215 } else if (strncmp(cp, "*\\\n", 3) == 0) { 1216 incomment = FINISHING_COMMENT; 1217 cp += 3; 1218 } else if (strncmp(cp, "*/", 2) == 0) { 1219 incomment = NO_COMMENT; 1220 cp += 2; 1221 } else 1222 cp += 1; 1223 continue; 1224 case STARTING_COMMENT: 1225 if (*cp == '*') { 1226 incomment = C_COMMENT; 1227 cp += 1; 1228 } else if (*cp == '/') { 1229 incomment = CXX_COMMENT; 1230 cp += 1; 1231 } else { 1232 incomment = NO_COMMENT; 1233 linestate = LS_DIRTY; 1234 } 1235 continue; 1236 case FINISHING_COMMENT: 1237 if (*cp == '/') { 1238 incomment = NO_COMMENT; 1239 cp += 1; 1240 } else 1241 incomment = C_COMMENT; 1242 continue; 1243 default: 1244 abort(); /* bug */ 1245 } 1246 return (cp); 1247} 1248 1249/* 1250 * Skip macro arguments. 1251 */ 1252static const char * 1253skipargs(const char *cp) 1254{ 1255 const char *ocp = cp; 1256 int level = 0; 1257 cp = skipcomment(cp); 1258 if (*cp != '(') 1259 return (cp); 1260 do { 1261 if (*cp == '(') 1262 level++; 1263 if (*cp == ')') 1264 level--; 1265 cp = skipcomment(cp+1); 1266 } while (level != 0 && *cp != '\0'); 1267 if (level == 0) 1268 return (cp); 1269 else 1270 /* Rewind and re-detect the syntax error later. */ 1271 return (ocp); 1272} 1273 1274/* 1275 * Skip over an identifier. 1276 */ 1277static const char * 1278skipsym(const char *cp) 1279{ 1280 while (!endsym(*cp)) 1281 ++cp; 1282 return (cp); 1283} 1284 1285/* 1286 * Skip whitespace and take a copy of any following identifier. 1287 */ 1288static const char * 1289getsym(const char **cpp) 1290{ 1291 const char *cp = *cpp, *sym; 1292 1293 cp = skipcomment(cp); 1294 cp = skipsym(sym = cp); 1295 if (cp == sym) 1296 return NULL; 1297 *cpp = cp; 1298 return (xstrdup(sym, cp)); 1299} 1300 1301/* 1302 * Check that s (a symbol) matches the start of t, and that the 1303 * following character in t is not a symbol character. Returns a 1304 * pointer to the following character in t if there is a match, 1305 * otherwise NULL. 1306 */ 1307static const char * 1308matchsym(const char *s, const char *t) 1309{ 1310 while (*s != '\0' && *t != '\0') 1311 if (*s != *t) 1312 return (NULL); 1313 else 1314 ++s, ++t; 1315 if (*s == '\0' && endsym(*t)) 1316 return(t); 1317 else 1318 return(NULL); 1319} 1320 1321/* 1322 * Look for the symbol in the symbol table. If it is found, we return 1323 * the symbol table index, else we return -1. 1324 */ 1325static int 1326findsym(const char **strp) 1327{ 1328 const char *str; 1329 int symind; 1330 1331 str = *strp; 1332 *strp = skipsym(str); 1333 if (symlist) { 1334 if (*strp == str) 1335 return (-1); 1336 if (symdepth && firstsym) 1337 printf("%s%3d", zerosyms ? "" : "\n", depth); 1338 firstsym = zerosyms = false; 1339 printf("%s%.*s%s", 1340 symdepth ? " " : "", 1341 (int)(*strp-str), str, 1342 symdepth ? "" : "\n"); 1343 /* we don't care about the value of the symbol */ 1344 return (0); 1345 } 1346 for (symind = 0; symind < nsyms; ++symind) { 1347 if (matchsym(symname[symind], str) != NULL) { 1348 debugsym("findsym", symind); 1349 return (symind); 1350 } 1351 } 1352 return (-1); 1353} 1354 1355/* 1356 * Resolve indirect symbol values to their final definitions. 1357 */ 1358static void 1359indirectsym(void) 1360{ 1361 const char *cp; 1362 int changed, sym, ind; 1363 1364 do { 1365 changed = 0; 1366 for (sym = 0; sym < nsyms; ++sym) { 1367 if (value[sym] == NULL) 1368 continue; 1369 cp = value[sym]; 1370 ind = findsym(&cp); 1371 if (ind == -1 || ind == sym || 1372 *cp != '\0' || 1373 value[ind] == NULL || 1374 value[ind] == value[sym]) 1375 continue; 1376 debugsym("indir...", sym); 1377 value[sym] = value[ind]; 1378 debugsym("...ectsym", sym); 1379 changed++; 1380 } 1381 } while (changed); 1382} 1383 1384/* 1385 * Add a symbol to the symbol table, specified with the format sym=val 1386 */ 1387static void 1388addsym1(bool ignorethis, bool definethis, char *symval) 1389{ 1390 const char *sym, *val; 1391 1392 sym = symval; 1393 val = skipsym(sym); 1394 if (definethis && *val == '=') { 1395 symval[val - sym] = '\0'; 1396 val = val + 1; 1397 } else if (*val == '\0') { 1398 val = definethis ? "1" : NULL; 1399 } else { 1400 usage(); 1401 } 1402 addsym2(ignorethis, sym, val); 1403} 1404 1405/* 1406 * Add a symbol to the symbol table. 1407 */ 1408static void 1409addsym2(bool ignorethis, const char *sym, const char *val) 1410{ 1411 const char *cp = sym; 1412 int symind; 1413 1414 symind = findsym(&cp); 1415 if (symind < 0) { 1416 if (nsyms >= MAXSYMS) 1417 errx(2, "too many symbols"); 1418 symind = nsyms++; 1419 } 1420 ignore[symind] = ignorethis; 1421 symname[symind] = sym; 1422 value[symind] = val; 1423 debugsym("addsym", symind); 1424} 1425 1426static void 1427debugsym(const char *why, int symind) 1428{ 1429 debug("%s %s%c%s", why, symname[symind], 1430 value[symind] ? '=' : ' ', 1431 value[symind] ? value[symind] : "undef"); 1432} 1433 1434/* 1435 * Add symbols to the symbol table from a file containing 1436 * #define and #undef preprocessor directives. 1437 */ 1438static void 1439defundefile(const char *fn) 1440{ 1441 filename = fn; 1442 input = fopen(fn, "rb"); 1443 if (input == NULL) 1444 err(2, "can't open %s", fn); 1445 linenum = 0; 1446 while (defundef()) 1447 ; 1448 if (ferror(input)) 1449 err(2, "can't read %s", filename); 1450 else 1451 fclose(input); 1452 if (incomment) 1453 error("EOF in comment"); 1454} 1455 1456/* 1457 * Read and process one #define or #undef directive 1458 */ 1459static bool 1460defundef(void) 1461{ 1462 const char *cp, *kw, *sym, *val, *end; 1463 1464 cp = skiphash(); 1465 if (cp == NULL) 1466 return (false); 1467 if (*cp == '\0') 1468 goto done; 1469 /* strip trailing whitespace, and do a fairly rough check to 1470 avoid unsupported multi-line preprocessor directives */ 1471 end = cp + strlen(cp); 1472 while (end > tline && strchr(" \t\n\r", end[-1]) != NULL) 1473 --end; 1474 if (end > tline && end[-1] == '\\') 1475 Eioccc(); 1476 1477 kw = cp; 1478 if ((cp = matchsym("define", kw)) != NULL) { 1479 sym = getsym(&cp); 1480 if (sym == NULL) 1481 error("missing macro name in #define"); 1482 if (*cp == '(') { 1483 val = "1"; 1484 } else { 1485 cp = skipcomment(cp); 1486 val = (cp < end) ? xstrdup(cp, end) : ""; 1487 } 1488 debug("#define"); 1489 addsym2(false, sym, val); 1490 } else if ((cp = matchsym("undef", kw)) != NULL) { 1491 sym = getsym(&cp); 1492 if (sym == NULL) 1493 error("missing macro name in #undef"); 1494 cp = skipcomment(cp); 1495 debug("#undef"); 1496 addsym2(false, sym, NULL); 1497 } else { 1498 error("unrecognized preprocessor directive"); 1499 } 1500 skipline(cp); 1501done: 1502 debug("parser line %d state %s comment %s line", linenum, 1503 comment_name[incomment], linestate_name[linestate]); 1504 return (true); 1505} 1506 1507/* 1508 * Concatenate two strings into new memory, checking for failure. 1509 */ 1510static char * 1511astrcat(const char *s1, const char *s2) 1512{ 1513 char *s; 1514 int len; 1515 size_t size; 1516 1517 len = snprintf(NULL, 0, "%s%s", s1, s2); 1518 if (len < 0) 1519 err(2, "snprintf"); 1520 size = (size_t)len + 1; 1521 s = (char *)malloc(size); 1522 if (s == NULL) 1523 err(2, "malloc"); 1524 snprintf(s, size, "%s%s", s1, s2); 1525 return (s); 1526} 1527 1528/* 1529 * Duplicate a segment of a string, checking for failure. 1530 */ 1531static const char * 1532xstrdup(const char *start, const char *end) 1533{ 1534 size_t n; 1535 char *s; 1536 1537 if (end < start) abort(); /* bug */ 1538 n = (size_t)(end - start) + 1; 1539 s = malloc(n); 1540 if (s == NULL) 1541 err(2, "malloc"); 1542 snprintf(s, n, "%s", start); 1543 return (s); 1544} 1545 1546/* 1547 * Diagnostics. 1548 */ 1549static void 1550debug(const char *msg, ...) 1551{ 1552 va_list ap; 1553 1554 if (debugging) { 1555 va_start(ap, msg); 1556 vwarnx(msg, ap); 1557 va_end(ap); 1558 } 1559} 1560 1561static void 1562error(const char *msg) 1563{ 1564 if (depth == 0) 1565 warnx("%s: %d: %s", filename, linenum, msg); 1566 else 1567 warnx("%s: %d: %s (#if line %d depth %d)", 1568 filename, linenum, msg, stifline[depth], depth); 1569 closeio(); 1570 errx(2, "output may be truncated"); 1571} 1572