parser.c revision 262951
1/*-
2 * Copyright (c) 1991, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Kenneth Almquist.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33#ifndef lint
34#if 0
35static char sccsid[] = "@(#)parser.c	8.7 (Berkeley) 5/16/95";
36#endif
37#endif /* not lint */
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: stable/10/bin/sh/parser.c 262951 2014-03-09 17:04:31Z jmmv $");
40
41#include <stdlib.h>
42#include <unistd.h>
43#include <stdio.h>
44
45#include "shell.h"
46#include "parser.h"
47#include "nodes.h"
48#include "expand.h"	/* defines rmescapes() */
49#include "syntax.h"
50#include "options.h"
51#include "input.h"
52#include "output.h"
53#include "var.h"
54#include "error.h"
55#include "memalloc.h"
56#include "mystring.h"
57#include "alias.h"
58#include "show.h"
59#include "eval.h"
60#include "exec.h"	/* to check for special builtins */
61#ifndef NO_HISTORY
62#include "myhistedit.h"
63#endif
64
65/*
66 * Shell command parser.
67 */
68
69#define	EOFMARKLEN	79
70#define	PROMPTLEN	128
71
72/* values of checkkwd variable */
73#define CHKALIAS	0x1
74#define CHKKWD		0x2
75#define CHKNL		0x4
76
77/* values returned by readtoken */
78#include "token.h"
79
80
81
82struct heredoc {
83	struct heredoc *next;	/* next here document in list */
84	union node *here;		/* redirection node */
85	char *eofmark;		/* string indicating end of input */
86	int striptabs;		/* if set, strip leading tabs */
87};
88
89struct parser_temp {
90	struct parser_temp *next;
91	void *data;
92};
93
94
95static struct heredoc *heredoclist;	/* list of here documents to read */
96static int doprompt;		/* if set, prompt the user */
97static int needprompt;		/* true if interactive and at start of line */
98static int lasttoken;		/* last token read */
99static int tokpushback;		/* last token pushed back */
100static char *wordtext;		/* text of last word returned by readtoken */
101static int checkkwd;
102static struct nodelist *backquotelist;
103static union node *redirnode;
104static struct heredoc *heredoc;
105static int quoteflag;		/* set if (part of) last token was quoted */
106static int startlinno;		/* line # where last token started */
107static int funclinno;		/* line # where the current function started */
108static struct parser_temp *parser_temp;
109
110
111static union node *list(int);
112static union node *andor(void);
113static union node *pipeline(void);
114static union node *command(void);
115static union node *simplecmd(union node **, union node *);
116static union node *makename(void);
117static union node *makebinary(int type, union node *n1, union node *n2);
118static void parsefname(void);
119static void parseheredoc(void);
120static int peektoken(void);
121static int readtoken(void);
122static int xxreadtoken(void);
123static int readtoken1(int, const char *, const char *, int);
124static int noexpand(char *);
125static void consumetoken(int);
126static void synexpect(int) __dead2;
127static void synerror(const char *) __dead2;
128static void setprompt(int);
129
130
131static void *
132parser_temp_alloc(size_t len)
133{
134	struct parser_temp *t;
135
136	INTOFF;
137	t = ckmalloc(sizeof(*t));
138	t->data = NULL;
139	t->next = parser_temp;
140	parser_temp = t;
141	t->data = ckmalloc(len);
142	INTON;
143	return t->data;
144}
145
146
147static void *
148parser_temp_realloc(void *ptr, size_t len)
149{
150	struct parser_temp *t;
151
152	INTOFF;
153	t = parser_temp;
154	if (ptr != t->data)
155		error("bug: parser_temp_realloc misused");
156	t->data = ckrealloc(t->data, len);
157	INTON;
158	return t->data;
159}
160
161
162static void
163parser_temp_free_upto(void *ptr)
164{
165	struct parser_temp *t;
166	int done = 0;
167
168	INTOFF;
169	while (parser_temp != NULL && !done) {
170		t = parser_temp;
171		parser_temp = t->next;
172		done = t->data == ptr;
173		ckfree(t->data);
174		ckfree(t);
175	}
176	INTON;
177	if (!done)
178		error("bug: parser_temp_free_upto misused");
179}
180
181
182static void
183parser_temp_free_all(void)
184{
185	struct parser_temp *t;
186
187	INTOFF;
188	while (parser_temp != NULL) {
189		t = parser_temp;
190		parser_temp = t->next;
191		ckfree(t->data);
192		ckfree(t);
193	}
194	INTON;
195}
196
197
198/*
199 * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
200 * valid parse tree indicating a blank line.)
201 */
202
203union node *
204parsecmd(int interact)
205{
206	int t;
207
208	/* This assumes the parser is not re-entered,
209	 * which could happen if we add command substitution on PS1/PS2.
210	 */
211	parser_temp_free_all();
212	heredoclist = NULL;
213
214	tokpushback = 0;
215	checkkwd = 0;
216	doprompt = interact;
217	if (doprompt)
218		setprompt(1);
219	else
220		setprompt(0);
221	needprompt = 0;
222	t = readtoken();
223	if (t == TEOF)
224		return NEOF;
225	if (t == TNL)
226		return NULL;
227	tokpushback++;
228	return list(1);
229}
230
231
232static union node *
233list(int nlflag)
234{
235	union node *ntop, *n1, *n2, *n3;
236	int tok;
237
238	checkkwd = CHKNL | CHKKWD | CHKALIAS;
239	if (!nlflag && tokendlist[peektoken()])
240		return NULL;
241	ntop = n1 = NULL;
242	for (;;) {
243		n2 = andor();
244		tok = readtoken();
245		if (tok == TBACKGND) {
246			if (n2 != NULL && n2->type == NPIPE) {
247				n2->npipe.backgnd = 1;
248			} else if (n2 != NULL && n2->type == NREDIR) {
249				n2->type = NBACKGND;
250			} else {
251				n3 = (union node *)stalloc(sizeof (struct nredir));
252				n3->type = NBACKGND;
253				n3->nredir.n = n2;
254				n3->nredir.redirect = NULL;
255				n2 = n3;
256			}
257		}
258		if (ntop == NULL)
259			ntop = n2;
260		else if (n1 == NULL) {
261			n1 = makebinary(NSEMI, ntop, n2);
262			ntop = n1;
263		}
264		else {
265			n3 = makebinary(NSEMI, n1->nbinary.ch2, n2);
266			n1->nbinary.ch2 = n3;
267			n1 = n3;
268		}
269		switch (tok) {
270		case TBACKGND:
271		case TSEMI:
272			tok = readtoken();
273			/* FALLTHROUGH */
274		case TNL:
275			if (tok == TNL) {
276				parseheredoc();
277				if (nlflag)
278					return ntop;
279			} else if (tok == TEOF && nlflag) {
280				parseheredoc();
281				return ntop;
282			} else {
283				tokpushback++;
284			}
285			checkkwd = CHKNL | CHKKWD | CHKALIAS;
286			if (!nlflag && tokendlist[peektoken()])
287				return ntop;
288			break;
289		case TEOF:
290			if (heredoclist)
291				parseheredoc();
292			else
293				pungetc();		/* push back EOF on input */
294			return ntop;
295		default:
296			if (nlflag)
297				synexpect(-1);
298			tokpushback++;
299			return ntop;
300		}
301	}
302}
303
304
305
306static union node *
307andor(void)
308{
309	union node *n;
310	int t;
311
312	n = pipeline();
313	for (;;) {
314		if ((t = readtoken()) == TAND) {
315			t = NAND;
316		} else if (t == TOR) {
317			t = NOR;
318		} else {
319			tokpushback++;
320			return n;
321		}
322		n = makebinary(t, n, pipeline());
323	}
324}
325
326
327
328static union node *
329pipeline(void)
330{
331	union node *n1, *n2, *pipenode;
332	struct nodelist *lp, *prev;
333	int negate, t;
334
335	negate = 0;
336	checkkwd = CHKNL | CHKKWD | CHKALIAS;
337	TRACE(("pipeline: entered\n"));
338	while (readtoken() == TNOT)
339		negate = !negate;
340	tokpushback++;
341	n1 = command();
342	if (readtoken() == TPIPE) {
343		pipenode = (union node *)stalloc(sizeof (struct npipe));
344		pipenode->type = NPIPE;
345		pipenode->npipe.backgnd = 0;
346		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
347		pipenode->npipe.cmdlist = lp;
348		lp->n = n1;
349		do {
350			prev = lp;
351			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
352			checkkwd = CHKNL | CHKKWD | CHKALIAS;
353			t = readtoken();
354			tokpushback++;
355			if (t == TNOT)
356				lp->n = pipeline();
357			else
358				lp->n = command();
359			prev->next = lp;
360		} while (readtoken() == TPIPE);
361		lp->next = NULL;
362		n1 = pipenode;
363	}
364	tokpushback++;
365	if (negate) {
366		n2 = (union node *)stalloc(sizeof (struct nnot));
367		n2->type = NNOT;
368		n2->nnot.com = n1;
369		return n2;
370	} else
371		return n1;
372}
373
374
375
376static union node *
377command(void)
378{
379	union node *n1, *n2;
380	union node *ap, **app;
381	union node *cp, **cpp;
382	union node *redir, **rpp;
383	int t;
384	int is_subshell;
385
386	checkkwd = CHKNL | CHKKWD | CHKALIAS;
387	is_subshell = 0;
388	redir = NULL;
389	n1 = NULL;
390	rpp = &redir;
391
392	/* Check for redirection which may precede command */
393	while (readtoken() == TREDIR) {
394		*rpp = n2 = redirnode;
395		rpp = &n2->nfile.next;
396		parsefname();
397	}
398	tokpushback++;
399
400	switch (readtoken()) {
401	case TIF:
402		n1 = (union node *)stalloc(sizeof (struct nif));
403		n1->type = NIF;
404		if ((n1->nif.test = list(0)) == NULL)
405			synexpect(-1);
406		consumetoken(TTHEN);
407		n1->nif.ifpart = list(0);
408		n2 = n1;
409		while (readtoken() == TELIF) {
410			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
411			n2 = n2->nif.elsepart;
412			n2->type = NIF;
413			if ((n2->nif.test = list(0)) == NULL)
414				synexpect(-1);
415			consumetoken(TTHEN);
416			n2->nif.ifpart = list(0);
417		}
418		if (lasttoken == TELSE)
419			n2->nif.elsepart = list(0);
420		else {
421			n2->nif.elsepart = NULL;
422			tokpushback++;
423		}
424		consumetoken(TFI);
425		checkkwd = CHKKWD | CHKALIAS;
426		break;
427	case TWHILE:
428	case TUNTIL:
429		t = lasttoken;
430		if ((n1 = list(0)) == NULL)
431			synexpect(-1);
432		consumetoken(TDO);
433		n1 = makebinary((t == TWHILE)? NWHILE : NUNTIL, n1, list(0));
434		consumetoken(TDONE);
435		checkkwd = CHKKWD | CHKALIAS;
436		break;
437	case TFOR:
438		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
439			synerror("Bad for loop variable");
440		n1 = (union node *)stalloc(sizeof (struct nfor));
441		n1->type = NFOR;
442		n1->nfor.var = wordtext;
443		while (readtoken() == TNL)
444			;
445		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
446			app = &ap;
447			while (readtoken() == TWORD) {
448				n2 = makename();
449				*app = n2;
450				app = &n2->narg.next;
451			}
452			*app = NULL;
453			n1->nfor.args = ap;
454			if (lasttoken != TNL && lasttoken != TSEMI)
455				synexpect(-1);
456		} else {
457			static char argvars[5] = {
458				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
459			};
460			n2 = (union node *)stalloc(sizeof (struct narg));
461			n2->type = NARG;
462			n2->narg.text = argvars;
463			n2->narg.backquote = NULL;
464			n2->narg.next = NULL;
465			n1->nfor.args = n2;
466			/*
467			 * Newline or semicolon here is optional (but note
468			 * that the original Bourne shell only allowed NL).
469			 */
470			if (lasttoken != TNL && lasttoken != TSEMI)
471				tokpushback++;
472		}
473		checkkwd = CHKNL | CHKKWD | CHKALIAS;
474		if ((t = readtoken()) == TDO)
475			t = TDONE;
476		else if (t == TBEGIN)
477			t = TEND;
478		else
479			synexpect(-1);
480		n1->nfor.body = list(0);
481		consumetoken(t);
482		checkkwd = CHKKWD | CHKALIAS;
483		break;
484	case TCASE:
485		n1 = (union node *)stalloc(sizeof (struct ncase));
486		n1->type = NCASE;
487		consumetoken(TWORD);
488		n1->ncase.expr = makename();
489		while (readtoken() == TNL);
490		if (lasttoken != TWORD || ! equal(wordtext, "in"))
491			synerror("expecting \"in\"");
492		cpp = &n1->ncase.cases;
493		checkkwd = CHKNL | CHKKWD, readtoken();
494		while (lasttoken != TESAC) {
495			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
496			cp->type = NCLIST;
497			app = &cp->nclist.pattern;
498			if (lasttoken == TLP)
499				readtoken();
500			for (;;) {
501				*app = ap = makename();
502				checkkwd = CHKNL | CHKKWD;
503				if (readtoken() != TPIPE)
504					break;
505				app = &ap->narg.next;
506				readtoken();
507			}
508			ap->narg.next = NULL;
509			if (lasttoken != TRP)
510				synexpect(TRP);
511			cp->nclist.body = list(0);
512
513			checkkwd = CHKNL | CHKKWD | CHKALIAS;
514			if ((t = readtoken()) != TESAC) {
515				if (t == TENDCASE)
516					;
517				else if (t == TFALLTHRU)
518					cp->type = NCLISTFALLTHRU;
519				else
520					synexpect(TENDCASE);
521				checkkwd = CHKNL | CHKKWD, readtoken();
522			}
523			cpp = &cp->nclist.next;
524		}
525		*cpp = NULL;
526		checkkwd = CHKKWD | CHKALIAS;
527		break;
528	case TLP:
529		n1 = (union node *)stalloc(sizeof (struct nredir));
530		n1->type = NSUBSHELL;
531		n1->nredir.n = list(0);
532		n1->nredir.redirect = NULL;
533		consumetoken(TRP);
534		checkkwd = CHKKWD | CHKALIAS;
535		is_subshell = 1;
536		break;
537	case TBEGIN:
538		n1 = list(0);
539		consumetoken(TEND);
540		checkkwd = CHKKWD | CHKALIAS;
541		break;
542	/* A simple command must have at least one redirection or word. */
543	case TBACKGND:
544	case TSEMI:
545	case TAND:
546	case TOR:
547	case TPIPE:
548	case TENDCASE:
549	case TFALLTHRU:
550	case TEOF:
551	case TNL:
552	case TRP:
553		if (!redir)
554			synexpect(-1);
555	case TWORD:
556		tokpushback++;
557		n1 = simplecmd(rpp, redir);
558		return n1;
559	default:
560		synexpect(-1);
561	}
562
563	/* Now check for redirection which may follow command */
564	while (readtoken() == TREDIR) {
565		*rpp = n2 = redirnode;
566		rpp = &n2->nfile.next;
567		parsefname();
568	}
569	tokpushback++;
570	*rpp = NULL;
571	if (redir) {
572		if (!is_subshell) {
573			n2 = (union node *)stalloc(sizeof (struct nredir));
574			n2->type = NREDIR;
575			n2->nredir.n = n1;
576			n1 = n2;
577		}
578		n1->nredir.redirect = redir;
579	}
580
581	return n1;
582}
583
584
585static union node *
586simplecmd(union node **rpp, union node *redir)
587{
588	union node *args, **app;
589	union node **orig_rpp = rpp;
590	union node *n = NULL;
591	int special;
592	int savecheckkwd;
593
594	/* If we don't have any redirections already, then we must reset */
595	/* rpp to be the address of the local redir variable.  */
596	if (redir == 0)
597		rpp = &redir;
598
599	args = NULL;
600	app = &args;
601	/*
602	 * We save the incoming value, because we need this for shell
603	 * functions.  There can not be a redirect or an argument between
604	 * the function name and the open parenthesis.
605	 */
606	orig_rpp = rpp;
607
608	savecheckkwd = CHKALIAS;
609
610	for (;;) {
611		checkkwd = savecheckkwd;
612		if (readtoken() == TWORD) {
613			n = makename();
614			*app = n;
615			app = &n->narg.next;
616			if (savecheckkwd != 0 && !isassignment(wordtext))
617				savecheckkwd = 0;
618		} else if (lasttoken == TREDIR) {
619			*rpp = n = redirnode;
620			rpp = &n->nfile.next;
621			parsefname();	/* read name of redirection file */
622		} else if (lasttoken == TLP && app == &args->narg.next
623					    && rpp == orig_rpp) {
624			/* We have a function */
625			consumetoken(TRP);
626			funclinno = plinno;
627			/*
628			 * - Require plain text.
629			 * - Functions with '/' cannot be called.
630			 * - Reject name=().
631			 * - Reject ksh extended glob patterns.
632			 */
633			if (!noexpand(n->narg.text) || quoteflag ||
634			    strchr(n->narg.text, '/') ||
635			    strchr("!%*+-=?@}~",
636				n->narg.text[strlen(n->narg.text) - 1]))
637				synerror("Bad function name");
638			rmescapes(n->narg.text);
639			if (find_builtin(n->narg.text, &special) >= 0 &&
640			    special)
641				synerror("Cannot override a special builtin with a function");
642			n->type = NDEFUN;
643			n->narg.next = command();
644			funclinno = 0;
645			return n;
646		} else {
647			tokpushback++;
648			break;
649		}
650	}
651	*app = NULL;
652	*rpp = NULL;
653	n = (union node *)stalloc(sizeof (struct ncmd));
654	n->type = NCMD;
655	n->ncmd.args = args;
656	n->ncmd.redirect = redir;
657	return n;
658}
659
660static union node *
661makename(void)
662{
663	union node *n;
664
665	n = (union node *)stalloc(sizeof (struct narg));
666	n->type = NARG;
667	n->narg.next = NULL;
668	n->narg.text = wordtext;
669	n->narg.backquote = backquotelist;
670	return n;
671}
672
673static union node *
674makebinary(int type, union node *n1, union node *n2)
675{
676	union node *n;
677
678	n = (union node *)stalloc(sizeof (struct nbinary));
679	n->type = type;
680	n->nbinary.ch1 = n1;
681	n->nbinary.ch2 = n2;
682	return (n);
683}
684
685void
686forcealias(void)
687{
688	checkkwd |= CHKALIAS;
689}
690
691void
692fixredir(union node *n, const char *text, int err)
693{
694	TRACE(("Fix redir %s %d\n", text, err));
695	if (!err)
696		n->ndup.vname = NULL;
697
698	if (is_digit(text[0]) && text[1] == '\0')
699		n->ndup.dupfd = digit_val(text[0]);
700	else if (text[0] == '-' && text[1] == '\0')
701		n->ndup.dupfd = -1;
702	else {
703
704		if (err)
705			synerror("Bad fd number");
706		else
707			n->ndup.vname = makename();
708	}
709}
710
711
712static void
713parsefname(void)
714{
715	union node *n = redirnode;
716
717	consumetoken(TWORD);
718	if (n->type == NHERE) {
719		struct heredoc *here = heredoc;
720		struct heredoc *p;
721		int i;
722
723		if (quoteflag == 0)
724			n->type = NXHERE;
725		TRACE(("Here document %d\n", n->type));
726		if (here->striptabs) {
727			while (*wordtext == '\t')
728				wordtext++;
729		}
730		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
731			synerror("Illegal eof marker for << redirection");
732		rmescapes(wordtext);
733		here->eofmark = wordtext;
734		here->next = NULL;
735		if (heredoclist == NULL)
736			heredoclist = here;
737		else {
738			for (p = heredoclist ; p->next ; p = p->next);
739			p->next = here;
740		}
741	} else if (n->type == NTOFD || n->type == NFROMFD) {
742		fixredir(n, wordtext, 0);
743	} else {
744		n->nfile.fname = makename();
745	}
746}
747
748
749/*
750 * Input any here documents.
751 */
752
753static void
754parseheredoc(void)
755{
756	struct heredoc *here;
757	union node *n;
758
759	while (heredoclist) {
760		here = heredoclist;
761		heredoclist = here->next;
762		if (needprompt) {
763			setprompt(2);
764			needprompt = 0;
765		}
766		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
767				here->eofmark, here->striptabs);
768		n = makename();
769		here->here->nhere.doc = n;
770	}
771}
772
773static int
774peektoken(void)
775{
776	int t;
777
778	t = readtoken();
779	tokpushback++;
780	return (t);
781}
782
783static int
784readtoken(void)
785{
786	int t;
787	struct alias *ap;
788#ifdef DEBUG
789	int alreadyseen = tokpushback;
790#endif
791
792	top:
793	t = xxreadtoken();
794
795	/*
796	 * eat newlines
797	 */
798	if (checkkwd & CHKNL) {
799		while (t == TNL) {
800			parseheredoc();
801			t = xxreadtoken();
802		}
803	}
804
805	/*
806	 * check for keywords and aliases
807	 */
808	if (t == TWORD && !quoteflag)
809	{
810		const char * const *pp;
811
812		if (checkkwd & CHKKWD)
813			for (pp = parsekwd; *pp; pp++) {
814				if (**pp == *wordtext && equal(*pp, wordtext))
815				{
816					lasttoken = t = pp - parsekwd + KWDOFFSET;
817					TRACE(("keyword %s recognized\n", tokname[t]));
818					goto out;
819				}
820			}
821		if (checkkwd & CHKALIAS &&
822		    (ap = lookupalias(wordtext, 1)) != NULL) {
823			pushstring(ap->val, strlen(ap->val), ap);
824			goto top;
825		}
826	}
827out:
828	if (t != TNOT)
829		checkkwd = 0;
830
831#ifdef DEBUG
832	if (!alreadyseen)
833	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
834	else
835	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
836#endif
837	return (t);
838}
839
840
841/*
842 * Read the next input token.
843 * If the token is a word, we set backquotelist to the list of cmds in
844 *	backquotes.  We set quoteflag to true if any part of the word was
845 *	quoted.
846 * If the token is TREDIR, then we set redirnode to a structure containing
847 *	the redirection.
848 * In all cases, the variable startlinno is set to the number of the line
849 *	on which the token starts.
850 *
851 * [Change comment:  here documents and internal procedures]
852 * [Readtoken shouldn't have any arguments.  Perhaps we should make the
853 *  word parsing code into a separate routine.  In this case, readtoken
854 *  doesn't need to have any internal procedures, but parseword does.
855 *  We could also make parseoperator in essence the main routine, and
856 *  have parseword (readtoken1?) handle both words and redirection.]
857 */
858
859#define RETURN(token)	return lasttoken = token
860
861static int
862xxreadtoken(void)
863{
864	int c;
865
866	if (tokpushback) {
867		tokpushback = 0;
868		return lasttoken;
869	}
870	if (needprompt) {
871		setprompt(2);
872		needprompt = 0;
873	}
874	startlinno = plinno;
875	for (;;) {	/* until token or start of word found */
876		c = pgetc_macro();
877		switch (c) {
878		case ' ': case '\t':
879			continue;
880		case '#':
881			while ((c = pgetc()) != '\n' && c != PEOF);
882			pungetc();
883			continue;
884		case '\\':
885			if (pgetc() == '\n') {
886				startlinno = ++plinno;
887				if (doprompt)
888					setprompt(2);
889				else
890					setprompt(0);
891				continue;
892			}
893			pungetc();
894			goto breakloop;
895		case '\n':
896			plinno++;
897			needprompt = doprompt;
898			RETURN(TNL);
899		case PEOF:
900			RETURN(TEOF);
901		case '&':
902			if (pgetc() == '&')
903				RETURN(TAND);
904			pungetc();
905			RETURN(TBACKGND);
906		case '|':
907			if (pgetc() == '|')
908				RETURN(TOR);
909			pungetc();
910			RETURN(TPIPE);
911		case ';':
912			c = pgetc();
913			if (c == ';')
914				RETURN(TENDCASE);
915			else if (c == '&')
916				RETURN(TFALLTHRU);
917			pungetc();
918			RETURN(TSEMI);
919		case '(':
920			RETURN(TLP);
921		case ')':
922			RETURN(TRP);
923		default:
924			goto breakloop;
925		}
926	}
927breakloop:
928	return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
929#undef RETURN
930}
931
932
933#define MAXNEST_static 8
934struct tokenstate
935{
936	const char *syntax; /* *SYNTAX */
937	int parenlevel; /* levels of parentheses in arithmetic */
938	enum tokenstate_category
939	{
940		TSTATE_TOP,
941		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
942		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
943		TSTATE_ARITH
944	} category;
945};
946
947
948/*
949 * Called to parse command substitutions.
950 */
951
952static char *
953parsebackq(char *out, struct nodelist **pbqlist,
954		int oldstyle, int dblquote, int quoted)
955{
956	struct nodelist **nlpp;
957	union node *n;
958	char *volatile str;
959	struct jmploc jmploc;
960	struct jmploc *const savehandler = handler;
961	size_t savelen;
962	int saveprompt;
963	const int bq_startlinno = plinno;
964	char *volatile ostr = NULL;
965	struct parsefile *const savetopfile = getcurrentfile();
966	struct heredoc *const saveheredoclist = heredoclist;
967	struct heredoc *here;
968
969	str = NULL;
970	if (setjmp(jmploc.loc)) {
971		popfilesupto(savetopfile);
972		if (str)
973			ckfree(str);
974		if (ostr)
975			ckfree(ostr);
976		heredoclist = saveheredoclist;
977		handler = savehandler;
978		if (exception == EXERROR) {
979			startlinno = bq_startlinno;
980			synerror("Error in command substitution");
981		}
982		longjmp(handler->loc, 1);
983	}
984	INTOFF;
985	savelen = out - stackblock();
986	if (savelen > 0) {
987		str = ckmalloc(savelen);
988		memcpy(str, stackblock(), savelen);
989	}
990	handler = &jmploc;
991	heredoclist = NULL;
992	INTON;
993        if (oldstyle) {
994                /* We must read until the closing backquote, giving special
995                   treatment to some slashes, and then push the string and
996                   reread it as input, interpreting it normally.  */
997                char *oout;
998                int c;
999                int olen;
1000
1001
1002                STARTSTACKSTR(oout);
1003		for (;;) {
1004			if (needprompt) {
1005				setprompt(2);
1006				needprompt = 0;
1007			}
1008			CHECKSTRSPACE(2, oout);
1009			switch (c = pgetc()) {
1010			case '`':
1011				goto done;
1012
1013			case '\\':
1014                                if ((c = pgetc()) == '\n') {
1015					plinno++;
1016					if (doprompt)
1017						setprompt(2);
1018					else
1019						setprompt(0);
1020					/*
1021					 * If eating a newline, avoid putting
1022					 * the newline into the new character
1023					 * stream (via the USTPUTC after the
1024					 * switch).
1025					 */
1026					continue;
1027				}
1028                                if (c != '\\' && c != '`' && c != '$'
1029                                    && (!dblquote || c != '"'))
1030                                        USTPUTC('\\', oout);
1031				break;
1032
1033			case '\n':
1034				plinno++;
1035				needprompt = doprompt;
1036				break;
1037
1038			case PEOF:
1039			        startlinno = plinno;
1040				synerror("EOF in backquote substitution");
1041 				break;
1042
1043			default:
1044				break;
1045			}
1046			USTPUTC(c, oout);
1047                }
1048done:
1049                USTPUTC('\0', oout);
1050                olen = oout - stackblock();
1051		INTOFF;
1052		ostr = ckmalloc(olen);
1053		memcpy(ostr, stackblock(), olen);
1054		setinputstring(ostr, 1);
1055		INTON;
1056        }
1057	nlpp = pbqlist;
1058	while (*nlpp)
1059		nlpp = &(*nlpp)->next;
1060	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1061	(*nlpp)->next = NULL;
1062
1063	if (oldstyle) {
1064		saveprompt = doprompt;
1065		doprompt = 0;
1066	}
1067
1068	n = list(0);
1069
1070	if (oldstyle) {
1071		if (peektoken() != TEOF)
1072			synexpect(-1);
1073		doprompt = saveprompt;
1074	} else
1075		consumetoken(TRP);
1076
1077	(*nlpp)->n = n;
1078        if (oldstyle) {
1079		/*
1080		 * Start reading from old file again, ignoring any pushed back
1081		 * tokens left from the backquote parsing
1082		 */
1083                popfile();
1084		tokpushback = 0;
1085	}
1086	STARTSTACKSTR(out);
1087	CHECKSTRSPACE(savelen + 1, out);
1088	INTOFF;
1089	if (str) {
1090		memcpy(out, str, savelen);
1091		STADJUST(savelen, out);
1092		ckfree(str);
1093		str = NULL;
1094	}
1095	if (ostr) {
1096		ckfree(ostr);
1097		ostr = NULL;
1098	}
1099	here = saveheredoclist;
1100	if (here != NULL) {
1101		while (here->next != NULL)
1102			here = here->next;
1103		here->next = heredoclist;
1104		heredoclist = saveheredoclist;
1105	}
1106	handler = savehandler;
1107	INTON;
1108	if (quoted)
1109		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1110	else
1111		USTPUTC(CTLBACKQ, out);
1112	return out;
1113}
1114
1115
1116/*
1117 * Called to parse a backslash escape sequence inside $'...'.
1118 * The backslash has already been read.
1119 */
1120static char *
1121readcstyleesc(char *out)
1122{
1123	int c, v, i, n;
1124
1125	c = pgetc();
1126	switch (c) {
1127	case '\0':
1128		synerror("Unterminated quoted string");
1129	case '\n':
1130		plinno++;
1131		if (doprompt)
1132			setprompt(2);
1133		else
1134			setprompt(0);
1135		return out;
1136	case '\\':
1137	case '\'':
1138	case '"':
1139		v = c;
1140		break;
1141	case 'a': v = '\a'; break;
1142	case 'b': v = '\b'; break;
1143	case 'e': v = '\033'; break;
1144	case 'f': v = '\f'; break;
1145	case 'n': v = '\n'; break;
1146	case 'r': v = '\r'; break;
1147	case 't': v = '\t'; break;
1148	case 'v': v = '\v'; break;
1149	case 'x':
1150		  v = 0;
1151		  for (;;) {
1152			  c = pgetc();
1153			  if (c >= '0' && c <= '9')
1154				  v = (v << 4) + c - '0';
1155			  else if (c >= 'A' && c <= 'F')
1156				  v = (v << 4) + c - 'A' + 10;
1157			  else if (c >= 'a' && c <= 'f')
1158				  v = (v << 4) + c - 'a' + 10;
1159			  else
1160				  break;
1161		  }
1162		  pungetc();
1163		  break;
1164	case '0': case '1': case '2': case '3':
1165	case '4': case '5': case '6': case '7':
1166		  v = c - '0';
1167		  c = pgetc();
1168		  if (c >= '0' && c <= '7') {
1169			  v <<= 3;
1170			  v += c - '0';
1171			  c = pgetc();
1172			  if (c >= '0' && c <= '7') {
1173				  v <<= 3;
1174				  v += c - '0';
1175			  } else
1176				  pungetc();
1177		  } else
1178			  pungetc();
1179		  break;
1180	case 'c':
1181		  c = pgetc();
1182		  if (c < 0x3f || c > 0x7a || c == 0x60)
1183			  synerror("Bad escape sequence");
1184		  if (c == '\\' && pgetc() != '\\')
1185			  synerror("Bad escape sequence");
1186		  if (c == '?')
1187			  v = 127;
1188		  else
1189			  v = c & 0x1f;
1190		  break;
1191	case 'u':
1192	case 'U':
1193		  n = c == 'U' ? 8 : 4;
1194		  v = 0;
1195		  for (i = 0; i < n; i++) {
1196			  c = pgetc();
1197			  if (c >= '0' && c <= '9')
1198				  v = (v << 4) + c - '0';
1199			  else if (c >= 'A' && c <= 'F')
1200				  v = (v << 4) + c - 'A' + 10;
1201			  else if (c >= 'a' && c <= 'f')
1202				  v = (v << 4) + c - 'a' + 10;
1203			  else
1204				  synerror("Bad escape sequence");
1205		  }
1206		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1207			  synerror("Bad escape sequence");
1208		  /* We really need iconv here. */
1209		  if (initial_localeisutf8 && v > 127) {
1210			  CHECKSTRSPACE(4, out);
1211			  /*
1212			   * We cannot use wctomb() as the locale may have
1213			   * changed.
1214			   */
1215			  if (v <= 0x7ff) {
1216				  USTPUTC(0xc0 | v >> 6, out);
1217				  USTPUTC(0x80 | (v & 0x3f), out);
1218				  return out;
1219			  } else if (v <= 0xffff) {
1220				  USTPUTC(0xe0 | v >> 12, out);
1221				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1222				  USTPUTC(0x80 | (v & 0x3f), out);
1223				  return out;
1224			  } else if (v <= 0x10ffff) {
1225				  USTPUTC(0xf0 | v >> 18, out);
1226				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1227				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1228				  USTPUTC(0x80 | (v & 0x3f), out);
1229				  return out;
1230			  }
1231		  }
1232		  if (v > 127)
1233			  v = '?';
1234		  break;
1235	default:
1236		  synerror("Bad escape sequence");
1237	}
1238	v = (char)v;
1239	/*
1240	 * We can't handle NUL bytes.
1241	 * POSIX says we should skip till the closing quote.
1242	 */
1243	if (v == '\0') {
1244		while ((c = pgetc()) != '\'') {
1245			if (c == '\\')
1246				c = pgetc();
1247			if (c == PEOF)
1248				synerror("Unterminated quoted string");
1249		}
1250		pungetc();
1251		return out;
1252	}
1253	if (SQSYNTAX[v] == CCTL)
1254		USTPUTC(CTLESC, out);
1255	USTPUTC(v, out);
1256	return out;
1257}
1258
1259
1260/*
1261 * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1262 * is not NULL, read a here document.  In the latter case, eofmark is the
1263 * word which marks the end of the document and striptabs is true if
1264 * leading tabs should be stripped from the document.  The argument firstc
1265 * is the first character of the input token or document.
1266 *
1267 * Because C does not have internal subroutines, I have simulated them
1268 * using goto's to implement the subroutine linkage.  The following macros
1269 * will run code that appears at the end of readtoken1.
1270 */
1271
1272#define CHECKEND()	{goto checkend; checkend_return:;}
1273#define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1274#define PARSESUB()	{goto parsesub; parsesub_return:;}
1275#define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1276
1277static int
1278readtoken1(int firstc, char const *initialsyntax, const char *eofmark,
1279    int striptabs)
1280{
1281	int c = firstc;
1282	char *out;
1283	int len;
1284	char line[EOFMARKLEN + 1];
1285	struct nodelist *bqlist;
1286	int quotef;
1287	int newvarnest;
1288	int level;
1289	int synentry;
1290	struct tokenstate state_static[MAXNEST_static];
1291	int maxnest = MAXNEST_static;
1292	struct tokenstate *state = state_static;
1293	int sqiscstyle = 0;
1294
1295	startlinno = plinno;
1296	quotef = 0;
1297	bqlist = NULL;
1298	newvarnest = 0;
1299	level = 0;
1300	state[level].syntax = initialsyntax;
1301	state[level].parenlevel = 0;
1302	state[level].category = TSTATE_TOP;
1303
1304	STARTSTACKSTR(out);
1305	loop: {	/* for each line, until end of word */
1306		CHECKEND();	/* set c to PEOF if at end of here document */
1307		for (;;) {	/* until end of line or end of word */
1308			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1309
1310			synentry = state[level].syntax[c];
1311
1312			switch(synentry) {
1313			case CNL:	/* '\n' */
1314				if (state[level].syntax == BASESYNTAX)
1315					goto endword;	/* exit outer loop */
1316				USTPUTC(c, out);
1317				plinno++;
1318				if (doprompt)
1319					setprompt(2);
1320				else
1321					setprompt(0);
1322				c = pgetc();
1323				goto loop;		/* continue outer loop */
1324			case CSBACK:
1325				if (sqiscstyle) {
1326					out = readcstyleesc(out);
1327					break;
1328				}
1329				/* FALLTHROUGH */
1330			case CWORD:
1331				USTPUTC(c, out);
1332				break;
1333			case CCTL:
1334				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1335					USTPUTC(CTLESC, out);
1336				USTPUTC(c, out);
1337				break;
1338			case CBACK:	/* backslash */
1339				c = pgetc();
1340				if (c == PEOF) {
1341					USTPUTC('\\', out);
1342					pungetc();
1343				} else if (c == '\n') {
1344					plinno++;
1345					if (doprompt)
1346						setprompt(2);
1347					else
1348						setprompt(0);
1349				} else {
1350					if (state[level].syntax == DQSYNTAX &&
1351					    c != '\\' && c != '`' && c != '$' &&
1352					    (c != '"' || (eofmark != NULL &&
1353						newvarnest == 0)) &&
1354					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1355						USTPUTC('\\', out);
1356					if ((eofmark == NULL ||
1357					    newvarnest > 0) &&
1358					    state[level].syntax == BASESYNTAX)
1359						USTPUTC(CTLQUOTEMARK, out);
1360					if (SQSYNTAX[c] == CCTL)
1361						USTPUTC(CTLESC, out);
1362					USTPUTC(c, out);
1363					if ((eofmark == NULL ||
1364					    newvarnest > 0) &&
1365					    state[level].syntax == BASESYNTAX &&
1366					    state[level].category == TSTATE_VAR_OLD)
1367						USTPUTC(CTLQUOTEEND, out);
1368					quotef++;
1369				}
1370				break;
1371			case CSQUOTE:
1372				USTPUTC(CTLQUOTEMARK, out);
1373				state[level].syntax = SQSYNTAX;
1374				sqiscstyle = 0;
1375				break;
1376			case CDQUOTE:
1377				USTPUTC(CTLQUOTEMARK, out);
1378				state[level].syntax = DQSYNTAX;
1379				break;
1380			case CENDQUOTE:
1381				if (eofmark != NULL && newvarnest == 0)
1382					USTPUTC(c, out);
1383				else {
1384					if (state[level].category == TSTATE_VAR_OLD)
1385						USTPUTC(CTLQUOTEEND, out);
1386					state[level].syntax = BASESYNTAX;
1387					quotef++;
1388				}
1389				break;
1390			case CVAR:	/* '$' */
1391				PARSESUB();		/* parse substitution */
1392				break;
1393			case CENDVAR:	/* '}' */
1394				if (level > 0 &&
1395				    ((state[level].category == TSTATE_VAR_OLD &&
1396				      state[level].syntax ==
1397				      state[level - 1].syntax) ||
1398				    (state[level].category == TSTATE_VAR_NEW &&
1399				     state[level].syntax == BASESYNTAX))) {
1400					if (state[level].category == TSTATE_VAR_NEW)
1401						newvarnest--;
1402					level--;
1403					USTPUTC(CTLENDVAR, out);
1404				} else {
1405					USTPUTC(c, out);
1406				}
1407				break;
1408			case CLP:	/* '(' in arithmetic */
1409				state[level].parenlevel++;
1410				USTPUTC(c, out);
1411				break;
1412			case CRP:	/* ')' in arithmetic */
1413				if (state[level].parenlevel > 0) {
1414					USTPUTC(c, out);
1415					--state[level].parenlevel;
1416				} else {
1417					if (pgetc() == ')') {
1418						if (level > 0 &&
1419						    state[level].category == TSTATE_ARITH) {
1420							level--;
1421							USTPUTC(CTLENDARI, out);
1422						} else
1423							USTPUTC(')', out);
1424					} else {
1425						/*
1426						 * unbalanced parens
1427						 *  (don't 2nd guess - no error)
1428						 */
1429						pungetc();
1430						USTPUTC(')', out);
1431					}
1432				}
1433				break;
1434			case CBQUOTE:	/* '`' */
1435				out = parsebackq(out, &bqlist, 1,
1436				    state[level].syntax == DQSYNTAX &&
1437				    (eofmark == NULL || newvarnest > 0),
1438				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1439				break;
1440			case CEOF:
1441				goto endword;		/* exit outer loop */
1442			case CIGN:
1443				break;
1444			default:
1445				if (level == 0)
1446					goto endword;	/* exit outer loop */
1447				USTPUTC(c, out);
1448			}
1449			c = pgetc_macro();
1450		}
1451	}
1452endword:
1453	if (state[level].syntax == ARISYNTAX)
1454		synerror("Missing '))'");
1455	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1456		synerror("Unterminated quoted string");
1457	if (state[level].category == TSTATE_VAR_OLD ||
1458	    state[level].category == TSTATE_VAR_NEW) {
1459		startlinno = plinno;
1460		synerror("Missing '}'");
1461	}
1462	if (state != state_static)
1463		parser_temp_free_upto(state);
1464	USTPUTC('\0', out);
1465	len = out - stackblock();
1466	out = stackblock();
1467	if (eofmark == NULL) {
1468		if ((c == '>' || c == '<')
1469		 && quotef == 0
1470		 && len <= 2
1471		 && (*out == '\0' || is_digit(*out))) {
1472			PARSEREDIR();
1473			return lasttoken = TREDIR;
1474		} else {
1475			pungetc();
1476		}
1477	}
1478	quoteflag = quotef;
1479	backquotelist = bqlist;
1480	grabstackblock(len);
1481	wordtext = out;
1482	return lasttoken = TWORD;
1483/* end of readtoken routine */
1484
1485
1486/*
1487 * Check to see whether we are at the end of the here document.  When this
1488 * is called, c is set to the first character of the next input line.  If
1489 * we are at the end of the here document, this routine sets the c to PEOF.
1490 */
1491
1492checkend: {
1493	if (eofmark) {
1494		if (striptabs) {
1495			while (c == '\t')
1496				c = pgetc();
1497		}
1498		if (c == *eofmark) {
1499			if (pfgets(line, sizeof line) != NULL) {
1500				const char *p, *q;
1501
1502				p = line;
1503				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1504				if ((*p == '\0' || *p == '\n') && *q == '\0') {
1505					c = PEOF;
1506					if (*p == '\n') {
1507						plinno++;
1508						needprompt = doprompt;
1509					}
1510				} else {
1511					pushstring(line, strlen(line), NULL);
1512				}
1513			}
1514		}
1515	}
1516	goto checkend_return;
1517}
1518
1519
1520/*
1521 * Parse a redirection operator.  The variable "out" points to a string
1522 * specifying the fd to be redirected.  The variable "c" contains the
1523 * first character of the redirection operator.
1524 */
1525
1526parseredir: {
1527	char fd = *out;
1528	union node *np;
1529
1530	np = (union node *)stalloc(sizeof (struct nfile));
1531	if (c == '>') {
1532		np->nfile.fd = 1;
1533		c = pgetc();
1534		if (c == '>')
1535			np->type = NAPPEND;
1536		else if (c == '&')
1537			np->type = NTOFD;
1538		else if (c == '|')
1539			np->type = NCLOBBER;
1540		else {
1541			np->type = NTO;
1542			pungetc();
1543		}
1544	} else {	/* c == '<' */
1545		np->nfile.fd = 0;
1546		c = pgetc();
1547		if (c == '<') {
1548			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1549				np = (union node *)stalloc(sizeof (struct nhere));
1550				np->nfile.fd = 0;
1551			}
1552			np->type = NHERE;
1553			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1554			heredoc->here = np;
1555			if ((c = pgetc()) == '-') {
1556				heredoc->striptabs = 1;
1557			} else {
1558				heredoc->striptabs = 0;
1559				pungetc();
1560			}
1561		} else if (c == '&')
1562			np->type = NFROMFD;
1563		else if (c == '>')
1564			np->type = NFROMTO;
1565		else {
1566			np->type = NFROM;
1567			pungetc();
1568		}
1569	}
1570	if (fd != '\0')
1571		np->nfile.fd = digit_val(fd);
1572	redirnode = np;
1573	goto parseredir_return;
1574}
1575
1576
1577/*
1578 * Parse a substitution.  At this point, we have read the dollar sign
1579 * and nothing else.
1580 */
1581
1582parsesub: {
1583	char buf[10];
1584	int subtype;
1585	int typeloc;
1586	int flags;
1587	char *p;
1588	static const char types[] = "}-+?=";
1589	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1590	int linno;
1591	int length;
1592	int c1;
1593
1594	c = pgetc();
1595	if (c == '(') {	/* $(command) or $((arith)) */
1596		if (pgetc() == '(') {
1597			PARSEARITH();
1598		} else {
1599			pungetc();
1600			out = parsebackq(out, &bqlist, 0,
1601			    state[level].syntax == DQSYNTAX &&
1602			    (eofmark == NULL || newvarnest > 0),
1603			    state[level].syntax == DQSYNTAX ||
1604			    state[level].syntax == ARISYNTAX);
1605		}
1606	} else if (c == '{' || is_name(c) || is_special(c)) {
1607		USTPUTC(CTLVAR, out);
1608		typeloc = out - stackblock();
1609		USTPUTC(VSNORMAL, out);
1610		subtype = VSNORMAL;
1611		flags = 0;
1612		if (c == '{') {
1613			bracketed_name = 1;
1614			c = pgetc();
1615			subtype = 0;
1616		}
1617varname:
1618		if (!is_eof(c) && is_name(c)) {
1619			length = 0;
1620			do {
1621				STPUTC(c, out);
1622				c = pgetc();
1623				length++;
1624			} while (!is_eof(c) && is_in_name(c));
1625			if (length == 6 &&
1626			    strncmp(out - length, "LINENO", length) == 0) {
1627				/* Replace the variable name with the
1628				 * current line number. */
1629				linno = plinno;
1630				if (funclinno != 0)
1631					linno -= funclinno - 1;
1632				snprintf(buf, sizeof(buf), "%d", linno);
1633				STADJUST(-6, out);
1634				STPUTS(buf, out);
1635				flags |= VSLINENO;
1636			}
1637		} else if (is_digit(c)) {
1638			if (bracketed_name) {
1639				do {
1640					STPUTC(c, out);
1641					c = pgetc();
1642				} while (is_digit(c));
1643			} else {
1644				STPUTC(c, out);
1645				c = pgetc();
1646			}
1647		} else if (is_special(c)) {
1648			c1 = c;
1649			c = pgetc();
1650			if (subtype == 0 && c1 == '#') {
1651				subtype = VSLENGTH;
1652				if (strchr(types, c) == NULL && c != ':' &&
1653				    c != '#' && c != '%')
1654					goto varname;
1655				c1 = c;
1656				c = pgetc();
1657				if (c1 != '}' && c == '}') {
1658					pungetc();
1659					c = c1;
1660					goto varname;
1661				}
1662				pungetc();
1663				c = c1;
1664				c1 = '#';
1665				subtype = 0;
1666			}
1667			USTPUTC(c1, out);
1668		} else {
1669			subtype = VSERROR;
1670			if (c == '}')
1671				pungetc();
1672			else if (c == '\n' || c == PEOF)
1673				synerror("Unexpected end of line in substitution");
1674			else
1675				USTPUTC(c, out);
1676		}
1677		if (subtype == 0) {
1678			switch (c) {
1679			case ':':
1680				flags |= VSNUL;
1681				c = pgetc();
1682				/*FALLTHROUGH*/
1683			default:
1684				p = strchr(types, c);
1685				if (p == NULL) {
1686					if (c == '\n' || c == PEOF)
1687						synerror("Unexpected end of line in substitution");
1688					if (flags == VSNUL)
1689						STPUTC(':', out);
1690					STPUTC(c, out);
1691					subtype = VSERROR;
1692				} else
1693					subtype = p - types + VSNORMAL;
1694				break;
1695			case '%':
1696			case '#':
1697				{
1698					int cc = c;
1699					subtype = c == '#' ? VSTRIMLEFT :
1700							     VSTRIMRIGHT;
1701					c = pgetc();
1702					if (c == cc)
1703						subtype++;
1704					else
1705						pungetc();
1706					break;
1707				}
1708			}
1709		} else if (subtype != VSERROR) {
1710			if (subtype == VSLENGTH && c != '}')
1711				subtype = VSERROR;
1712			pungetc();
1713		}
1714		STPUTC('=', out);
1715		if (state[level].syntax == DQSYNTAX ||
1716		    state[level].syntax == ARISYNTAX)
1717			flags |= VSQUOTE;
1718		*(stackblock() + typeloc) = subtype | flags;
1719		if (subtype != VSNORMAL) {
1720			if (level + 1 >= maxnest) {
1721				maxnest *= 2;
1722				if (state == state_static) {
1723					state = parser_temp_alloc(
1724					    maxnest * sizeof(*state));
1725					memcpy(state, state_static,
1726					    MAXNEST_static * sizeof(*state));
1727				} else
1728					state = parser_temp_realloc(state,
1729					    maxnest * sizeof(*state));
1730			}
1731			level++;
1732			state[level].parenlevel = 0;
1733			if (subtype == VSMINUS || subtype == VSPLUS ||
1734			    subtype == VSQUESTION || subtype == VSASSIGN) {
1735				/*
1736				 * For operators that were in the Bourne shell,
1737				 * inherit the double-quote state.
1738				 */
1739				state[level].syntax = state[level - 1].syntax;
1740				state[level].category = TSTATE_VAR_OLD;
1741			} else {
1742				/*
1743				 * The other operators take a pattern,
1744				 * so go to BASESYNTAX.
1745				 * Also, ' and " are now special, even
1746				 * in here documents.
1747				 */
1748				state[level].syntax = BASESYNTAX;
1749				state[level].category = TSTATE_VAR_NEW;
1750				newvarnest++;
1751			}
1752		}
1753	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1754		/* $'cstylequotes' */
1755		USTPUTC(CTLQUOTEMARK, out);
1756		state[level].syntax = SQSYNTAX;
1757		sqiscstyle = 1;
1758	} else {
1759		USTPUTC('$', out);
1760		pungetc();
1761	}
1762	goto parsesub_return;
1763}
1764
1765
1766/*
1767 * Parse an arithmetic expansion (indicate start of one and set state)
1768 */
1769parsearith: {
1770
1771	if (level + 1 >= maxnest) {
1772		maxnest *= 2;
1773		if (state == state_static) {
1774			state = parser_temp_alloc(
1775			    maxnest * sizeof(*state));
1776			memcpy(state, state_static,
1777			    MAXNEST_static * sizeof(*state));
1778		} else
1779			state = parser_temp_realloc(state,
1780			    maxnest * sizeof(*state));
1781	}
1782	level++;
1783	state[level].syntax = ARISYNTAX;
1784	state[level].parenlevel = 0;
1785	state[level].category = TSTATE_ARITH;
1786	USTPUTC(CTLARI, out);
1787	if (state[level - 1].syntax == DQSYNTAX)
1788		USTPUTC('"',out);
1789	else
1790		USTPUTC(' ',out);
1791	goto parsearith_return;
1792}
1793
1794} /* end of readtoken */
1795
1796
1797/*
1798 * Returns true if the text contains nothing to expand (no dollar signs
1799 * or backquotes).
1800 */
1801
1802static int
1803noexpand(char *text)
1804{
1805	char *p;
1806	char c;
1807
1808	p = text;
1809	while ((c = *p++) != '\0') {
1810		if ( c == CTLQUOTEMARK)
1811			continue;
1812		if (c == CTLESC)
1813			p++;
1814		else if (BASESYNTAX[(int)c] == CCTL)
1815			return 0;
1816	}
1817	return 1;
1818}
1819
1820
1821/*
1822 * Return true if the argument is a legal variable name (a letter or
1823 * underscore followed by zero or more letters, underscores, and digits).
1824 */
1825
1826int
1827goodname(const char *name)
1828{
1829	const char *p;
1830
1831	p = name;
1832	if (! is_name(*p))
1833		return 0;
1834	while (*++p) {
1835		if (! is_in_name(*p))
1836			return 0;
1837	}
1838	return 1;
1839}
1840
1841
1842int
1843isassignment(const char *p)
1844{
1845	if (!is_name(*p))
1846		return 0;
1847	p++;
1848	for (;;) {
1849		if (*p == '=')
1850			return 1;
1851		else if (!is_in_name(*p))
1852			return 0;
1853		p++;
1854	}
1855}
1856
1857
1858static void
1859consumetoken(int token)
1860{
1861	if (readtoken() != token)
1862		synexpect(token);
1863}
1864
1865
1866/*
1867 * Called when an unexpected token is read during the parse.  The argument
1868 * is the token that is expected, or -1 if more than one type of token can
1869 * occur at this point.
1870 */
1871
1872static void
1873synexpect(int token)
1874{
1875	char msg[64];
1876
1877	if (token >= 0) {
1878		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1879			tokname[lasttoken], tokname[token]);
1880	} else {
1881		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1882	}
1883	synerror(msg);
1884}
1885
1886
1887static void
1888synerror(const char *msg)
1889{
1890	if (commandname)
1891		outfmt(out2, "%s: %d: ", commandname, startlinno);
1892	outfmt(out2, "Syntax error: %s\n", msg);
1893	error((char *)NULL);
1894}
1895
1896static void
1897setprompt(int which)
1898{
1899	whichprompt = which;
1900
1901#ifndef NO_HISTORY
1902	if (!el)
1903#endif
1904	{
1905		out2str(getprompt(NULL));
1906		flushout(out2);
1907	}
1908}
1909
1910/*
1911 * called by editline -- any expansions to the prompt
1912 *    should be added here.
1913 */
1914char *
1915getprompt(void *unused __unused)
1916{
1917	static char ps[PROMPTLEN];
1918	char *fmt;
1919	const char *pwd;
1920	int i, trim;
1921	static char internal_error[] = "??";
1922
1923	/*
1924	 * Select prompt format.
1925	 */
1926	switch (whichprompt) {
1927	case 0:
1928		fmt = nullstr;
1929		break;
1930	case 1:
1931		fmt = ps1val();
1932		break;
1933	case 2:
1934		fmt = ps2val();
1935		break;
1936	default:
1937		return internal_error;
1938	}
1939
1940	/*
1941	 * Format prompt string.
1942	 */
1943	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1944		if (*fmt == '\\')
1945			switch (*++fmt) {
1946
1947				/*
1948				 * Hostname.
1949				 *
1950				 * \h specifies just the local hostname,
1951				 * \H specifies fully-qualified hostname.
1952				 */
1953			case 'h':
1954			case 'H':
1955				ps[i] = '\0';
1956				gethostname(&ps[i], PROMPTLEN - i);
1957				/* Skip to end of hostname. */
1958				trim = (*fmt == 'h') ? '.' : '\0';
1959				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1960					i++;
1961				break;
1962
1963				/*
1964				 * Working directory.
1965				 *
1966				 * \W specifies just the final component,
1967				 * \w specifies the entire path.
1968				 */
1969			case 'W':
1970			case 'w':
1971				pwd = lookupvar("PWD");
1972				if (pwd == NULL)
1973					pwd = "?";
1974				if (*fmt == 'W' &&
1975				    *pwd == '/' && pwd[1] != '\0')
1976					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
1977					    PROMPTLEN - i);
1978				else
1979					strlcpy(&ps[i], pwd, PROMPTLEN - i);
1980				/* Skip to end of path. */
1981				while (ps[i + 1] != '\0')
1982					i++;
1983				break;
1984
1985				/*
1986				 * Superuser status.
1987				 *
1988				 * '$' for normal users, '#' for root.
1989				 */
1990			case '$':
1991				ps[i] = (geteuid() != 0) ? '$' : '#';
1992				break;
1993
1994				/*
1995				 * A literal \.
1996				 */
1997			case '\\':
1998				ps[i] = '\\';
1999				break;
2000
2001				/*
2002				 * Emit unrecognized formats verbatim.
2003				 */
2004			default:
2005				ps[i++] = '\\';
2006				ps[i] = *fmt;
2007				break;
2008			}
2009		else
2010			ps[i] = *fmt;
2011	ps[i] = '\0';
2012	return (ps);
2013}
2014
2015
2016const char *
2017expandstr(const char *ps)
2018{
2019	union node n;
2020	struct jmploc jmploc;
2021	struct jmploc *const savehandler = handler;
2022	const int saveprompt = doprompt;
2023	struct parsefile *const savetopfile = getcurrentfile();
2024	struct parser_temp *const saveparser_temp = parser_temp;
2025	const char *result = NULL;
2026
2027	if (!setjmp(jmploc.loc)) {
2028		handler = &jmploc;
2029		parser_temp = NULL;
2030		setinputstring(ps, 1);
2031		doprompt = 0;
2032		readtoken1(pgetc(), DQSYNTAX, "\n\n", 0);
2033		if (backquotelist != NULL)
2034			error("Command substitution not allowed here");
2035
2036		n.narg.type = NARG;
2037		n.narg.next = NULL;
2038		n.narg.text = wordtext;
2039		n.narg.backquote = backquotelist;
2040
2041		expandarg(&n, NULL, 0);
2042		result = stackblock();
2043		INTOFF;
2044	}
2045	handler = savehandler;
2046	doprompt = saveprompt;
2047	popfilesupto(savetopfile);
2048	if (parser_temp != saveparser_temp) {
2049		parser_temp_free_all();
2050		parser_temp = saveparser_temp;
2051	}
2052	if (result != NULL) {
2053		INTON;
2054	} else if (exception == EXINT)
2055		raise(SIGINT);
2056	return result;
2057}
2058