expand.c revision 194977
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[] = "@(#)expand.c	8.5 (Berkeley) 5/15/95";
36#endif
37#endif /* not lint */
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/bin/sh/expand.c 194977 2009-06-25 17:14:06Z jilles $");
40
41#include <sys/types.h>
42#include <sys/time.h>
43#include <sys/stat.h>
44#include <errno.h>
45#include <dirent.h>
46#include <unistd.h>
47#include <pwd.h>
48#include <stdlib.h>
49#include <limits.h>
50#include <stdio.h>
51#include <string.h>
52
53/*
54 * Routines to expand arguments to commands.  We have to deal with
55 * backquotes, shell variables, and file metacharacters.
56 */
57
58#include "shell.h"
59#include "main.h"
60#include "nodes.h"
61#include "eval.h"
62#include "expand.h"
63#include "syntax.h"
64#include "parser.h"
65#include "jobs.h"
66#include "options.h"
67#include "var.h"
68#include "input.h"
69#include "output.h"
70#include "memalloc.h"
71#include "error.h"
72#include "mystring.h"
73#include "arith.h"
74#include "show.h"
75
76/*
77 * Structure specifying which parts of the string should be searched
78 * for IFS characters.
79 */
80
81struct ifsregion {
82	struct ifsregion *next;	/* next region in list */
83	int begoff;		/* offset of start of region */
84	int endoff;		/* offset of end of region */
85	int inquotes;		/* search for nul bytes only */
86};
87
88
89STATIC char *expdest;			/* output of current string */
90STATIC struct nodelist *argbackq;	/* list of back quote expressions */
91STATIC struct ifsregion ifsfirst;	/* first struct in list of ifs regions */
92STATIC struct ifsregion *ifslastp;	/* last struct in list */
93STATIC struct arglist exparg;		/* holds expanded arg list */
94
95STATIC void argstr(char *, int);
96STATIC char *exptilde(char *, int);
97STATIC void expbackq(union node *, int, int);
98STATIC int subevalvar(char *, char *, int, int, int, int);
99STATIC char *evalvar(char *, int);
100STATIC int varisset(char *, int);
101STATIC void varvalue(char *, int, int, int);
102STATIC void recordregion(int, int, int);
103STATIC void removerecordregions(int);
104STATIC void ifsbreakup(char *, struct arglist *);
105STATIC void expandmeta(struct strlist *, int);
106STATIC void expmeta(char *, char *);
107STATIC void addfname(char *);
108STATIC struct strlist *expsort(struct strlist *);
109STATIC struct strlist *msort(struct strlist *, int);
110STATIC int pmatch(char *, char *, int);
111STATIC char *cvtnum(int, char *);
112STATIC int collate_range_cmp(int, int);
113
114STATIC int
115collate_range_cmp(int c1, int c2)
116{
117	static char s1[2], s2[2];
118
119	s1[0] = c1;
120	s2[0] = c2;
121	return (strcoll(s1, s2));
122}
123
124/*
125 * Expand shell variables and backquotes inside a here document.
126 *	union node *arg		the document
127 *	int fd;			where to write the expanded version
128 */
129
130void
131expandhere(union node *arg, int fd)
132{
133	herefd = fd;
134	expandarg(arg, (struct arglist *)NULL, 0);
135	xwrite(fd, stackblock(), expdest - stackblock());
136}
137
138
139/*
140 * Perform variable substitution and command substitution on an argument,
141 * placing the resulting list of arguments in arglist.  If EXP_FULL is true,
142 * perform splitting and file name expansion.  When arglist is NULL, perform
143 * here document expansion.
144 */
145
146void
147expandarg(union node *arg, struct arglist *arglist, int flag)
148{
149	struct strlist *sp;
150	char *p;
151
152	argbackq = arg->narg.backquote;
153	STARTSTACKSTR(expdest);
154	ifsfirst.next = NULL;
155	ifslastp = NULL;
156	argstr(arg->narg.text, flag);
157	if (arglist == NULL) {
158		return;			/* here document expanded */
159	}
160	STPUTC('\0', expdest);
161	p = grabstackstr(expdest);
162	exparg.lastp = &exparg.list;
163	/*
164	 * TODO - EXP_REDIR
165	 */
166	if (flag & EXP_FULL) {
167		ifsbreakup(p, &exparg);
168		*exparg.lastp = NULL;
169		exparg.lastp = &exparg.list;
170		expandmeta(exparg.list, flag);
171	} else {
172		if (flag & EXP_REDIR) /*XXX - for now, just remove escapes */
173			rmescapes(p);
174		sp = (struct strlist *)stalloc(sizeof (struct strlist));
175		sp->text = p;
176		*exparg.lastp = sp;
177		exparg.lastp = &sp->next;
178	}
179	while (ifsfirst.next != NULL) {
180		struct ifsregion *ifsp;
181		INTOFF;
182		ifsp = ifsfirst.next->next;
183		ckfree(ifsfirst.next);
184		ifsfirst.next = ifsp;
185		INTON;
186	}
187	*exparg.lastp = NULL;
188	if (exparg.list) {
189		*arglist->lastp = exparg.list;
190		arglist->lastp = exparg.lastp;
191	}
192}
193
194
195
196/*
197 * Perform variable and command substitution.  If EXP_FULL is set, output CTLESC
198 * characters to allow for further processing.  Otherwise treat
199 * $@ like $* since no splitting will be performed.
200 */
201
202STATIC void
203argstr(char *p, int flag)
204{
205	char c;
206	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);	/* do CTLESC */
207	int firsteq = 1;
208
209	if (*p == '~' && (flag & (EXP_TILDE | EXP_VARTILDE)))
210		p = exptilde(p, flag);
211	for (;;) {
212		switch (c = *p++) {
213		case '\0':
214		case CTLENDVAR: /* ??? */
215			goto breakloop;
216		case CTLQUOTEMARK:
217			/* "$@" syntax adherence hack */
218			if (p[0] == CTLVAR && p[2] == '@' && p[3] == '=')
219				break;
220			if ((flag & EXP_FULL) != 0)
221				STPUTC(c, expdest);
222			break;
223		case CTLESC:
224			if (quotes)
225				STPUTC(c, expdest);
226			c = *p++;
227			STPUTC(c, expdest);
228			break;
229		case CTLVAR:
230			p = evalvar(p, flag);
231			break;
232		case CTLBACKQ:
233		case CTLBACKQ|CTLQUOTE:
234			expbackq(argbackq->n, c & CTLQUOTE, flag);
235			argbackq = argbackq->next;
236			break;
237		case CTLENDARI:
238			expari(flag);
239			break;
240		case ':':
241		case '=':
242			/*
243			 * sort of a hack - expand tildes in variable
244			 * assignments (after the first '=' and after ':'s).
245			 */
246			STPUTC(c, expdest);
247			if (flag & EXP_VARTILDE && *p == '~') {
248				if (c == '=') {
249					if (firsteq)
250						firsteq = 0;
251					else
252						break;
253				}
254				p = exptilde(p, flag);
255			}
256			break;
257		default:
258			STPUTC(c, expdest);
259		}
260	}
261breakloop:;
262}
263
264STATIC char *
265exptilde(char *p, int flag)
266{
267	char c, *startp = p;
268	struct passwd *pw;
269	char *home;
270	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
271
272	while ((c = *p) != '\0') {
273		switch(c) {
274		case CTLESC:
275			return (startp);
276		case CTLQUOTEMARK:
277			return (startp);
278		case ':':
279			if (flag & EXP_VARTILDE)
280				goto done;
281			break;
282		case '/':
283			goto done;
284		}
285		p++;
286	}
287done:
288	*p = '\0';
289	if (*(startp+1) == '\0') {
290		if ((home = lookupvar("HOME")) == NULL)
291			goto lose;
292	} else {
293		if ((pw = getpwnam(startp+1)) == NULL)
294			goto lose;
295		home = pw->pw_dir;
296	}
297	if (*home == '\0')
298		goto lose;
299	*p = c;
300	while ((c = *home++) != '\0') {
301		if (quotes && SQSYNTAX[(int)c] == CCTL)
302			STPUTC(CTLESC, expdest);
303		STPUTC(c, expdest);
304	}
305	return (p);
306lose:
307	*p = c;
308	return (startp);
309}
310
311
312STATIC void
313removerecordregions(int endoff)
314{
315	if (ifslastp == NULL)
316		return;
317
318	if (ifsfirst.endoff > endoff) {
319		while (ifsfirst.next != NULL) {
320			struct ifsregion *ifsp;
321			INTOFF;
322			ifsp = ifsfirst.next->next;
323			ckfree(ifsfirst.next);
324			ifsfirst.next = ifsp;
325			INTON;
326		}
327		if (ifsfirst.begoff > endoff)
328			ifslastp = NULL;
329		else {
330			ifslastp = &ifsfirst;
331			ifsfirst.endoff = endoff;
332		}
333		return;
334	}
335
336	ifslastp = &ifsfirst;
337	while (ifslastp->next && ifslastp->next->begoff < endoff)
338		ifslastp=ifslastp->next;
339	while (ifslastp->next != NULL) {
340		struct ifsregion *ifsp;
341		INTOFF;
342		ifsp = ifslastp->next->next;
343		ckfree(ifslastp->next);
344		ifslastp->next = ifsp;
345		INTON;
346	}
347	if (ifslastp->endoff > endoff)
348		ifslastp->endoff = endoff;
349}
350
351/*
352 * Expand arithmetic expression.  Backup to start of expression,
353 * evaluate, place result in (backed up) result, adjust string position.
354 */
355void
356expari(int flag)
357{
358	char *p, *start;
359	arith_t result;
360	int begoff;
361	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
362	int quoted;
363
364
365	/*
366	 * This routine is slightly over-complicated for
367	 * efficiency.  First we make sure there is
368	 * enough space for the result, which may be bigger
369	 * than the expression if we add exponentiation.  Next we
370	 * scan backwards looking for the start of arithmetic.  If the
371	 * next previous character is a CTLESC character, then we
372	 * have to rescan starting from the beginning since CTLESC
373	 * characters have to be processed left to right.
374	 */
375	CHECKSTRSPACE(DIGITS(result) - 2, expdest);
376	USTPUTC('\0', expdest);
377	start = stackblock();
378	p = expdest - 2;
379	while (p >= start && *p != CTLARI)
380		--p;
381	if (p < start || *p != CTLARI)
382		error("missing CTLARI (shouldn't happen)");
383	if (p > start && *(p - 1) == CTLESC)
384		for (p = start; *p != CTLARI; p++)
385			if (*p == CTLESC)
386				p++;
387
388	if (p[1] == '"')
389		quoted=1;
390	else
391		quoted=0;
392	begoff = p - start;
393	removerecordregions(begoff);
394	if (quotes)
395		rmescapes(p+2);
396	result = arith(p+2);
397	fmtstr(p, DIGITS(result), ARITH_FORMAT_STR, result);
398	while (*p++)
399		;
400	if (quoted == 0)
401		recordregion(begoff, p - 1 - start, 0);
402	result = expdest - p + 1;
403	STADJUST(-result, expdest);
404}
405
406
407/*
408 * Expand stuff in backwards quotes.
409 */
410
411STATIC void
412expbackq(union node *cmd, int quoted, int flag)
413{
414	struct backcmd in;
415	int i;
416	char buf[128];
417	char *p;
418	char *dest = expdest;
419	struct ifsregion saveifs, *savelastp;
420	struct nodelist *saveargbackq;
421	char lastc;
422	int startloc = dest - stackblock();
423	char const *syntax = quoted? DQSYNTAX : BASESYNTAX;
424	int saveherefd;
425	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
426	int nnl;
427
428	INTOFF;
429	saveifs = ifsfirst;
430	savelastp = ifslastp;
431	saveargbackq = argbackq;
432	saveherefd = herefd;
433	herefd = -1;
434	p = grabstackstr(dest);
435	evalbackcmd(cmd, &in);
436	ungrabstackstr(p, dest);
437	ifsfirst = saveifs;
438	ifslastp = savelastp;
439	argbackq = saveargbackq;
440	herefd = saveherefd;
441
442	p = in.buf;
443	lastc = '\0';
444	nnl = 0;
445	/* Don't copy trailing newlines */
446	for (;;) {
447		if (--in.nleft < 0) {
448			if (in.fd < 0)
449				break;
450			while ((i = read(in.fd, buf, sizeof buf)) < 0 && errno == EINTR);
451			TRACE(("expbackq: read returns %d\n", i));
452			if (i <= 0)
453				break;
454			p = buf;
455			in.nleft = i - 1;
456		}
457		lastc = *p++;
458		if (lastc != '\0') {
459			if (quotes && syntax[(int)lastc] == CCTL)
460				STPUTC(CTLESC, dest);
461			if (lastc == '\n') {
462				nnl++;
463			} else {
464				while (nnl > 0) {
465					nnl--;
466					STPUTC('\n', dest);
467				}
468				STPUTC(lastc, dest);
469			}
470		}
471	}
472
473	if (in.fd >= 0)
474		close(in.fd);
475	if (in.buf)
476		ckfree(in.buf);
477	if (in.jp)
478		exitstatus = waitforjob(in.jp, (int *)NULL);
479	if (quoted == 0)
480		recordregion(startloc, dest - stackblock(), 0);
481	TRACE(("evalbackq: size=%d: \"%.*s\"\n",
482		(dest - stackblock()) - startloc,
483		(dest - stackblock()) - startloc,
484		stackblock() + startloc));
485	expdest = dest;
486	INTON;
487}
488
489
490
491STATIC int
492subevalvar(char *p, char *str, int strloc, int subtype, int startloc,
493  int varflags)
494{
495	char *startp;
496	char *loc = NULL;
497	char *q;
498	int c = 0;
499	int saveherefd = herefd;
500	struct nodelist *saveargbackq = argbackq;
501	int amount;
502
503	herefd = -1;
504	argstr(p, 0);
505	STACKSTRNUL(expdest);
506	herefd = saveherefd;
507	argbackq = saveargbackq;
508	startp = stackblock() + startloc;
509	if (str == NULL)
510	    str = stackblock() + strloc;
511
512	switch (subtype) {
513	case VSASSIGN:
514		setvar(str, startp, 0);
515		amount = startp - expdest;
516		STADJUST(amount, expdest);
517		varflags &= ~VSNUL;
518		if (c != 0)
519			*loc = c;
520		return 1;
521
522	case VSQUESTION:
523		if (*p != CTLENDVAR) {
524			outfmt(&errout, "%s\n", startp);
525			error((char *)NULL);
526		}
527		error("%.*s: parameter %snot set", (int)(p - str - 1),
528		      str, (varflags & VSNUL) ? "null or "
529					      : nullstr);
530		return 0;
531
532	case VSTRIMLEFT:
533		for (loc = startp; loc < str; loc++) {
534			c = *loc;
535			*loc = '\0';
536			if (patmatch(str, startp, varflags & VSQUOTE)) {
537				*loc = c;
538				goto recordleft;
539			}
540			*loc = c;
541			if ((varflags & VSQUOTE) && *loc == CTLESC)
542				loc++;
543		}
544		return 0;
545
546	case VSTRIMLEFTMAX:
547		for (loc = str - 1; loc >= startp;) {
548			c = *loc;
549			*loc = '\0';
550			if (patmatch(str, startp, varflags & VSQUOTE)) {
551				*loc = c;
552				goto recordleft;
553			}
554			*loc = c;
555			loc--;
556			if ((varflags & VSQUOTE) && loc > startp &&
557			    *(loc - 1) == CTLESC) {
558				for (q = startp; q < loc; q++)
559					if (*q == CTLESC)
560						q++;
561				if (q > loc)
562					loc--;
563			}
564		}
565		return 0;
566
567	case VSTRIMRIGHT:
568		for (loc = str - 1; loc >= startp;) {
569			if (patmatch(str, loc, varflags & VSQUOTE)) {
570				amount = loc - expdest;
571				STADJUST(amount, expdest);
572				return 1;
573			}
574			loc--;
575			if ((varflags & VSQUOTE) && loc > startp &&
576			    *(loc - 1) == CTLESC) {
577				for (q = startp; q < loc; q++)
578					if (*q == CTLESC)
579						q++;
580				if (q > loc)
581					loc--;
582			}
583		}
584		return 0;
585
586	case VSTRIMRIGHTMAX:
587		for (loc = startp; loc < str - 1; loc++) {
588			if (patmatch(str, loc, varflags & VSQUOTE)) {
589				amount = loc - expdest;
590				STADJUST(amount, expdest);
591				return 1;
592			}
593			if ((varflags & VSQUOTE) && *loc == CTLESC)
594				loc++;
595		}
596		return 0;
597
598
599	default:
600		abort();
601	}
602
603recordleft:
604	amount = ((str - 1) - (loc - startp)) - expdest;
605	STADJUST(amount, expdest);
606	while (loc != str - 1)
607		*startp++ = *loc++;
608	return 1;
609}
610
611
612/*
613 * Expand a variable, and return a pointer to the next character in the
614 * input string.
615 */
616
617STATIC char *
618evalvar(char *p, int flag)
619{
620	int subtype;
621	int varflags;
622	char *var;
623	char *val;
624	int patloc;
625	int c;
626	int set;
627	int special;
628	int startloc;
629	int varlen;
630	int easy;
631	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
632
633	varflags = (unsigned char)*p++;
634	subtype = varflags & VSTYPE;
635	var = p;
636	special = 0;
637	if (! is_name(*p))
638		special = 1;
639	p = strchr(p, '=') + 1;
640again: /* jump here after setting a variable with ${var=text} */
641	if (varflags & VSLINENO) {
642		set = 1;
643		special = 0;
644		val = var;
645		p[-1] = '\0';	/* temporarily overwrite '=' to have \0
646				   terminated string */
647	} else if (special) {
648		set = varisset(var, varflags & VSNUL);
649		val = NULL;
650	} else {
651		val = bltinlookup(var, 1);
652		if (val == NULL || ((varflags & VSNUL) && val[0] == '\0')) {
653			val = NULL;
654			set = 0;
655		} else
656			set = 1;
657	}
658	varlen = 0;
659	startloc = expdest - stackblock();
660	if (!set && uflag) {
661		switch (subtype) {
662		case VSNORMAL:
663		case VSTRIMLEFT:
664		case VSTRIMLEFTMAX:
665		case VSTRIMRIGHT:
666		case VSTRIMRIGHTMAX:
667		case VSLENGTH:
668			error("%.*s: parameter not set", (int)(p - var - 1),
669			    var);
670		}
671	}
672	if (set && subtype != VSPLUS) {
673		/* insert the value of the variable */
674		if (special) {
675			varvalue(var, varflags & VSQUOTE, subtype, flag);
676			if (subtype == VSLENGTH) {
677				varlen = expdest - stackblock() - startloc;
678				STADJUST(-varlen, expdest);
679			}
680		} else {
681			char const *syntax = (varflags & VSQUOTE) ? DQSYNTAX
682								  : BASESYNTAX;
683
684			if (subtype == VSLENGTH) {
685				for (;*val; val++)
686					varlen++;
687			}
688			else {
689				while (*val) {
690					if (quotes &&
691					    syntax[(int)*val] == CCTL)
692						STPUTC(CTLESC, expdest);
693					STPUTC(*val++, expdest);
694				}
695
696			}
697		}
698	}
699
700	if (subtype == VSPLUS)
701		set = ! set;
702
703	easy = ((varflags & VSQUOTE) == 0 ||
704		(*var == '@' && shellparam.nparam != 1));
705
706
707	switch (subtype) {
708	case VSLENGTH:
709		expdest = cvtnum(varlen, expdest);
710		goto record;
711
712	case VSNORMAL:
713		if (!easy)
714			break;
715record:
716		recordregion(startloc, expdest - stackblock(),
717			     varflags & VSQUOTE);
718		break;
719
720	case VSPLUS:
721	case VSMINUS:
722		if (!set) {
723			argstr(p, flag);
724			break;
725		}
726		if (easy)
727			goto record;
728		break;
729
730	case VSTRIMLEFT:
731	case VSTRIMLEFTMAX:
732	case VSTRIMRIGHT:
733	case VSTRIMRIGHTMAX:
734		if (!set)
735			break;
736		/*
737		 * Terminate the string and start recording the pattern
738		 * right after it
739		 */
740		STPUTC('\0', expdest);
741		patloc = expdest - stackblock();
742		if (subevalvar(p, NULL, patloc, subtype,
743			       startloc, varflags) == 0) {
744			int amount = (expdest - stackblock() - patloc) + 1;
745			STADJUST(-amount, expdest);
746		}
747		/* Remove any recorded regions beyond start of variable */
748		removerecordregions(startloc);
749		goto record;
750
751	case VSASSIGN:
752	case VSQUESTION:
753		if (!set) {
754			if (subevalvar(p, var, 0, subtype, startloc, varflags)) {
755				varflags &= ~VSNUL;
756				/*
757				 * Remove any recorded regions beyond
758				 * start of variable
759				 */
760				removerecordregions(startloc);
761				goto again;
762			}
763			break;
764		}
765		if (easy)
766			goto record;
767		break;
768
769	case VSERROR:
770		c = p - var - 1;
771		error("${%.*s%s}: Bad substitution", c, var,
772		    (c > 0 && *p != CTLENDVAR) ? "..." : "");
773
774	default:
775		abort();
776	}
777	p[-1] = '=';	/* recover overwritten '=' */
778
779	if (subtype != VSNORMAL) {	/* skip to end of alternative */
780		int nesting = 1;
781		for (;;) {
782			if ((c = *p++) == CTLESC)
783				p++;
784			else if (c == CTLBACKQ || c == (CTLBACKQ|CTLQUOTE)) {
785				if (set)
786					argbackq = argbackq->next;
787			} else if (c == CTLVAR) {
788				if ((*p++ & VSTYPE) != VSNORMAL)
789					nesting++;
790			} else if (c == CTLENDVAR) {
791				if (--nesting == 0)
792					break;
793			}
794		}
795	}
796	return p;
797}
798
799
800
801/*
802 * Test whether a specialized variable is set.
803 */
804
805STATIC int
806varisset(char *name, int nulok)
807{
808
809	if (*name == '!')
810		return backgndpid != -1;
811	else if (*name == '@' || *name == '*') {
812		if (*shellparam.p == NULL)
813			return 0;
814
815		if (nulok) {
816			char **av;
817
818			for (av = shellparam.p; *av; av++)
819				if (**av != '\0')
820					return 1;
821			return 0;
822		}
823	} else if (is_digit(*name)) {
824		char *ap;
825		int num = atoi(name);
826
827		if (num > shellparam.nparam)
828			return 0;
829
830		if (num == 0)
831			ap = arg0;
832		else
833			ap = shellparam.p[num - 1];
834
835		if (nulok && (ap == NULL || *ap == '\0'))
836			return 0;
837	}
838	return 1;
839}
840
841
842
843/*
844 * Add the value of a specialized variable to the stack string.
845 */
846
847STATIC void
848varvalue(char *name, int quoted, int subtype, int flag)
849{
850	int num;
851	char *p;
852	int i;
853	extern int oexitstatus;
854	char sep;
855	char **ap;
856	char const *syntax;
857
858#define STRTODEST(p) \
859	do {\
860	if (flag & (EXP_FULL | EXP_CASE) && subtype != VSLENGTH) { \
861		syntax = quoted? DQSYNTAX : BASESYNTAX; \
862		while (*p) { \
863			if (syntax[(int)*p] == CCTL) \
864				STPUTC(CTLESC, expdest); \
865			STPUTC(*p++, expdest); \
866		} \
867	} else \
868		while (*p) \
869			STPUTC(*p++, expdest); \
870	} while (0)
871
872
873	switch (*name) {
874	case '$':
875		num = rootpid;
876		goto numvar;
877	case '?':
878		num = oexitstatus;
879		goto numvar;
880	case '#':
881		num = shellparam.nparam;
882		goto numvar;
883	case '!':
884		num = backgndpid;
885numvar:
886		expdest = cvtnum(num, expdest);
887		break;
888	case '-':
889		for (i = 0 ; i < NOPTS ; i++) {
890			if (optlist[i].val)
891				STPUTC(optlist[i].letter, expdest);
892		}
893		break;
894	case '@':
895		if (flag & EXP_FULL && quoted) {
896			for (ap = shellparam.p ; (p = *ap++) != NULL ; ) {
897				STRTODEST(p);
898				if (*ap)
899					STPUTC('\0', expdest);
900			}
901			break;
902		}
903		/* FALLTHROUGH */
904	case '*':
905		if (ifsset())
906			sep = ifsval()[0];
907		else
908			sep = ' ';
909		for (ap = shellparam.p ; (p = *ap++) != NULL ; ) {
910			STRTODEST(p);
911			if (*ap && sep)
912				STPUTC(sep, expdest);
913		}
914		break;
915	case '0':
916		p = arg0;
917		STRTODEST(p);
918		break;
919	default:
920		if (is_digit(*name)) {
921			num = atoi(name);
922			if (num > 0 && num <= shellparam.nparam) {
923				p = shellparam.p[num - 1];
924				STRTODEST(p);
925			}
926		}
927		break;
928	}
929}
930
931
932
933/*
934 * Record the the fact that we have to scan this region of the
935 * string for IFS characters.
936 */
937
938STATIC void
939recordregion(int start, int end, int inquotes)
940{
941	struct ifsregion *ifsp;
942
943	if (ifslastp == NULL) {
944		ifsp = &ifsfirst;
945	} else {
946		if (ifslastp->endoff == start
947		    && ifslastp->inquotes == inquotes) {
948			/* extend previous area */
949			ifslastp->endoff = end;
950			return;
951		}
952		ifsp = (struct ifsregion *)ckmalloc(sizeof (struct ifsregion));
953		ifslastp->next = ifsp;
954	}
955	ifslastp = ifsp;
956	ifslastp->next = NULL;
957	ifslastp->begoff = start;
958	ifslastp->endoff = end;
959	ifslastp->inquotes = inquotes;
960}
961
962
963
964/*
965 * Break the argument string into pieces based upon IFS and add the
966 * strings to the argument list.  The regions of the string to be
967 * searched for IFS characters have been stored by recordregion.
968 */
969STATIC void
970ifsbreakup(char *string, struct arglist *arglist)
971{
972	struct ifsregion *ifsp;
973	struct strlist *sp;
974	char *start;
975	char *p;
976	char *q;
977	char *ifs;
978	const char *ifsspc;
979	int had_param_ch = 0;
980
981	start = string;
982
983	if (ifslastp == NULL) {
984		/* Return entire argument, IFS doesn't apply to any of it */
985		sp = (struct strlist *)stalloc(sizeof *sp);
986		sp->text = start;
987		*arglist->lastp = sp;
988		arglist->lastp = &sp->next;
989		return;
990	}
991
992	ifs = ifsset() ? ifsval() : " \t\n";
993
994	for (ifsp = &ifsfirst; ifsp != NULL; ifsp = ifsp->next) {
995		p = string + ifsp->begoff;
996		while (p < string + ifsp->endoff) {
997			q = p;
998			if (*p == CTLESC)
999				p++;
1000			if (ifsp->inquotes) {
1001				/* Only NULs (should be from "$@") end args */
1002				had_param_ch = 1;
1003				if (*p != 0) {
1004					p++;
1005					continue;
1006				}
1007				ifsspc = NULL;
1008			} else {
1009				if (!strchr(ifs, *p)) {
1010					had_param_ch = 1;
1011					p++;
1012					continue;
1013				}
1014				ifsspc = strchr(" \t\n", *p);
1015
1016				/* Ignore IFS whitespace at start */
1017				if (q == start && ifsspc != NULL) {
1018					p++;
1019					start = p;
1020					continue;
1021				}
1022				had_param_ch = 0;
1023			}
1024
1025			/* Save this argument... */
1026			*q = '\0';
1027			sp = (struct strlist *)stalloc(sizeof *sp);
1028			sp->text = start;
1029			*arglist->lastp = sp;
1030			arglist->lastp = &sp->next;
1031			p++;
1032
1033			if (ifsspc != NULL) {
1034				/* Ignore further trailing IFS whitespace */
1035				for (; p < string + ifsp->endoff; p++) {
1036					q = p;
1037					if (*p == CTLESC)
1038						p++;
1039					if (strchr(ifs, *p) == NULL) {
1040						p = q;
1041						break;
1042					}
1043					if (strchr(" \t\n", *p) == NULL) {
1044						p++;
1045						break;
1046					}
1047				}
1048			}
1049			start = p;
1050		}
1051	}
1052
1053	/*
1054	 * Save anything left as an argument.
1055	 * Traditionally we have treated 'IFS=':'; set -- x$IFS' as
1056	 * generating 2 arguments, the second of which is empty.
1057	 * Some recent clarification of the Posix spec say that it
1058	 * should only generate one....
1059	 */
1060	if (had_param_ch || *start != 0) {
1061		sp = (struct strlist *)stalloc(sizeof *sp);
1062		sp->text = start;
1063		*arglist->lastp = sp;
1064		arglist->lastp = &sp->next;
1065	}
1066}
1067
1068
1069
1070/*
1071 * Expand shell metacharacters.  At this point, the only control characters
1072 * should be escapes.  The results are stored in the list exparg.
1073 */
1074
1075STATIC char *expdir;
1076
1077
1078STATIC void
1079expandmeta(struct strlist *str, int flag __unused)
1080{
1081	char *p;
1082	struct strlist **savelastp;
1083	struct strlist *sp;
1084	char c;
1085	/* TODO - EXP_REDIR */
1086
1087	while (str) {
1088		if (fflag)
1089			goto nometa;
1090		p = str->text;
1091		for (;;) {			/* fast check for meta chars */
1092			if ((c = *p++) == '\0')
1093				goto nometa;
1094			if (c == '*' || c == '?' || c == '[' || c == '!')
1095				break;
1096		}
1097		savelastp = exparg.lastp;
1098		INTOFF;
1099		if (expdir == NULL) {
1100			int i = strlen(str->text);
1101			expdir = ckmalloc(i < 2048 ? 2048 : i); /* XXX */
1102		}
1103
1104		expmeta(expdir, str->text);
1105		ckfree(expdir);
1106		expdir = NULL;
1107		INTON;
1108		if (exparg.lastp == savelastp) {
1109			/*
1110			 * no matches
1111			 */
1112nometa:
1113			*exparg.lastp = str;
1114			rmescapes(str->text);
1115			exparg.lastp = &str->next;
1116		} else {
1117			*exparg.lastp = NULL;
1118			*savelastp = sp = expsort(*savelastp);
1119			while (sp->next != NULL)
1120				sp = sp->next;
1121			exparg.lastp = &sp->next;
1122		}
1123		str = str->next;
1124	}
1125}
1126
1127
1128/*
1129 * Do metacharacter (i.e. *, ?, [...]) expansion.
1130 */
1131
1132STATIC void
1133expmeta(char *enddir, char *name)
1134{
1135	char *p;
1136	char *q;
1137	char *start;
1138	char *endname;
1139	int metaflag;
1140	struct stat statb;
1141	DIR *dirp;
1142	struct dirent *dp;
1143	int atend;
1144	int matchdot;
1145
1146	metaflag = 0;
1147	start = name;
1148	for (p = name ; ; p++) {
1149		if (*p == '*' || *p == '?')
1150			metaflag = 1;
1151		else if (*p == '[') {
1152			q = p + 1;
1153			if (*q == '!' || *q == '^')
1154				q++;
1155			for (;;) {
1156				while (*q == CTLQUOTEMARK)
1157					q++;
1158				if (*q == CTLESC)
1159					q++;
1160				if (*q == '/' || *q == '\0')
1161					break;
1162				if (*++q == ']') {
1163					metaflag = 1;
1164					break;
1165				}
1166			}
1167		} else if (*p == '!' && p[1] == '!'	&& (p == name || p[-1] == '/')) {
1168			metaflag = 1;
1169		} else if (*p == '\0')
1170			break;
1171		else if (*p == CTLQUOTEMARK)
1172			continue;
1173		else if (*p == CTLESC)
1174			p++;
1175		if (*p == '/') {
1176			if (metaflag)
1177				break;
1178			start = p + 1;
1179		}
1180	}
1181	if (metaflag == 0) {	/* we've reached the end of the file name */
1182		if (enddir != expdir)
1183			metaflag++;
1184		for (p = name ; ; p++) {
1185			if (*p == CTLQUOTEMARK)
1186				continue;
1187			if (*p == CTLESC)
1188				p++;
1189			*enddir++ = *p;
1190			if (*p == '\0')
1191				break;
1192		}
1193		if (metaflag == 0 || lstat(expdir, &statb) >= 0)
1194			addfname(expdir);
1195		return;
1196	}
1197	endname = p;
1198	if (start != name) {
1199		p = name;
1200		while (p < start) {
1201			while (*p == CTLQUOTEMARK)
1202				p++;
1203			if (*p == CTLESC)
1204				p++;
1205			*enddir++ = *p++;
1206		}
1207	}
1208	if (enddir == expdir) {
1209		p = ".";
1210	} else if (enddir == expdir + 1 && *expdir == '/') {
1211		p = "/";
1212	} else {
1213		p = expdir;
1214		enddir[-1] = '\0';
1215	}
1216	if ((dirp = opendir(p)) == NULL)
1217		return;
1218	if (enddir != expdir)
1219		enddir[-1] = '/';
1220	if (*endname == 0) {
1221		atend = 1;
1222	} else {
1223		atend = 0;
1224		*endname++ = '\0';
1225	}
1226	matchdot = 0;
1227	p = start;
1228	while (*p == CTLQUOTEMARK)
1229		p++;
1230	if (*p == CTLESC)
1231		p++;
1232	if (*p == '.')
1233		matchdot++;
1234	while (! int_pending() && (dp = readdir(dirp)) != NULL) {
1235		if (dp->d_name[0] == '.' && ! matchdot)
1236			continue;
1237		if (patmatch(start, dp->d_name, 0)) {
1238			if (atend) {
1239				scopy(dp->d_name, enddir);
1240				addfname(expdir);
1241			} else {
1242				for (p = enddir, q = dp->d_name;
1243				     (*p++ = *q++) != '\0';)
1244					continue;
1245				p[-1] = '/';
1246				expmeta(p, endname);
1247			}
1248		}
1249	}
1250	closedir(dirp);
1251	if (! atend)
1252		endname[-1] = '/';
1253}
1254
1255
1256/*
1257 * Add a file name to the list.
1258 */
1259
1260STATIC void
1261addfname(char *name)
1262{
1263	char *p;
1264	struct strlist *sp;
1265
1266	p = stalloc(strlen(name) + 1);
1267	scopy(name, p);
1268	sp = (struct strlist *)stalloc(sizeof *sp);
1269	sp->text = p;
1270	*exparg.lastp = sp;
1271	exparg.lastp = &sp->next;
1272}
1273
1274
1275/*
1276 * Sort the results of file name expansion.  It calculates the number of
1277 * strings to sort and then calls msort (short for merge sort) to do the
1278 * work.
1279 */
1280
1281STATIC struct strlist *
1282expsort(struct strlist *str)
1283{
1284	int len;
1285	struct strlist *sp;
1286
1287	len = 0;
1288	for (sp = str ; sp ; sp = sp->next)
1289		len++;
1290	return msort(str, len);
1291}
1292
1293
1294STATIC struct strlist *
1295msort(struct strlist *list, int len)
1296{
1297	struct strlist *p, *q = NULL;
1298	struct strlist **lpp;
1299	int half;
1300	int n;
1301
1302	if (len <= 1)
1303		return list;
1304	half = len >> 1;
1305	p = list;
1306	for (n = half ; --n >= 0 ; ) {
1307		q = p;
1308		p = p->next;
1309	}
1310	q->next = NULL;			/* terminate first half of list */
1311	q = msort(list, half);		/* sort first half of list */
1312	p = msort(p, len - half);		/* sort second half */
1313	lpp = &list;
1314	for (;;) {
1315		if (strcmp(p->text, q->text) < 0) {
1316			*lpp = p;
1317			lpp = &p->next;
1318			if ((p = *lpp) == NULL) {
1319				*lpp = q;
1320				break;
1321			}
1322		} else {
1323			*lpp = q;
1324			lpp = &q->next;
1325			if ((q = *lpp) == NULL) {
1326				*lpp = p;
1327				break;
1328			}
1329		}
1330	}
1331	return list;
1332}
1333
1334
1335
1336/*
1337 * Returns true if the pattern matches the string.
1338 */
1339
1340int
1341patmatch(char *pattern, char *string, int squoted)
1342{
1343#ifdef notdef
1344	if (pattern[0] == '!' && pattern[1] == '!')
1345		return 1 - pmatch(pattern + 2, string);
1346	else
1347#endif
1348		return pmatch(pattern, string, squoted);
1349}
1350
1351
1352STATIC int
1353pmatch(char *pattern, char *string, int squoted)
1354{
1355	char *p, *q;
1356	char c;
1357
1358	p = pattern;
1359	q = string;
1360	for (;;) {
1361		switch (c = *p++) {
1362		case '\0':
1363			goto breakloop;
1364		case CTLESC:
1365			if (squoted && *q == CTLESC)
1366				q++;
1367			if (*q++ != *p++)
1368				return 0;
1369			break;
1370		case CTLQUOTEMARK:
1371			continue;
1372		case '?':
1373			if (squoted && *q == CTLESC)
1374				q++;
1375			if (*q++ == '\0')
1376				return 0;
1377			break;
1378		case '*':
1379			c = *p;
1380			while (c == CTLQUOTEMARK || c == '*')
1381				c = *++p;
1382			if (c != CTLESC &&  c != CTLQUOTEMARK &&
1383			    c != '?' && c != '*' && c != '[') {
1384				while (*q != c) {
1385					if (squoted && *q == CTLESC &&
1386					    q[1] == c)
1387						break;
1388					if (*q == '\0')
1389						return 0;
1390					if (squoted && *q == CTLESC)
1391						q++;
1392					q++;
1393				}
1394			}
1395			do {
1396				if (pmatch(p, q, squoted))
1397					return 1;
1398				if (squoted && *q == CTLESC)
1399					q++;
1400			} while (*q++ != '\0');
1401			return 0;
1402		case '[': {
1403			char *endp;
1404			int invert, found;
1405			char chr;
1406
1407			endp = p;
1408			if (*endp == '!' || *endp == '^')
1409				endp++;
1410			for (;;) {
1411				while (*endp == CTLQUOTEMARK)
1412					endp++;
1413				if (*endp == '\0')
1414					goto dft;		/* no matching ] */
1415				if (*endp == CTLESC)
1416					endp++;
1417				if (*++endp == ']')
1418					break;
1419			}
1420			invert = 0;
1421			if (*p == '!' || *p == '^') {
1422				invert++;
1423				p++;
1424			}
1425			found = 0;
1426			chr = *q++;
1427			if (squoted && chr == CTLESC)
1428				chr = *q++;
1429			if (chr == '\0')
1430				return 0;
1431			c = *p++;
1432			do {
1433				if (c == CTLQUOTEMARK)
1434					continue;
1435				if (c == CTLESC)
1436					c = *p++;
1437				if (*p == '-' && p[1] != ']') {
1438					p++;
1439					while (*p == CTLQUOTEMARK)
1440						p++;
1441					if (*p == CTLESC)
1442						p++;
1443					if (   collate_range_cmp(chr, c) >= 0
1444					    && collate_range_cmp(chr, *p) <= 0
1445					   )
1446						found = 1;
1447					p++;
1448				} else {
1449					if (chr == c)
1450						found = 1;
1451				}
1452			} while ((c = *p++) != ']');
1453			if (found == invert)
1454				return 0;
1455			break;
1456		}
1457dft:	        default:
1458			if (squoted && *q == CTLESC)
1459				q++;
1460			if (*q++ != c)
1461				return 0;
1462			break;
1463		}
1464	}
1465breakloop:
1466	if (*q != '\0')
1467		return 0;
1468	return 1;
1469}
1470
1471
1472
1473/*
1474 * Remove any CTLESC characters from a string.
1475 */
1476
1477void
1478rmescapes(char *str)
1479{
1480	char *p, *q;
1481
1482	p = str;
1483	while (*p != CTLESC && *p != CTLQUOTEMARK) {
1484		if (*p++ == '\0')
1485			return;
1486	}
1487	q = p;
1488	while (*p) {
1489		if (*p == CTLQUOTEMARK) {
1490			p++;
1491			continue;
1492		}
1493		if (*p == CTLESC)
1494			p++;
1495		*q++ = *p++;
1496	}
1497	*q = '\0';
1498}
1499
1500
1501
1502/*
1503 * See if a pattern matches in a case statement.
1504 */
1505
1506int
1507casematch(union node *pattern, char *val)
1508{
1509	struct stackmark smark;
1510	int result;
1511	char *p;
1512
1513	setstackmark(&smark);
1514	argbackq = pattern->narg.backquote;
1515	STARTSTACKSTR(expdest);
1516	ifslastp = NULL;
1517	argstr(pattern->narg.text, EXP_TILDE | EXP_CASE);
1518	STPUTC('\0', expdest);
1519	p = grabstackstr(expdest);
1520	result = patmatch(p, val, 0);
1521	popstackmark(&smark);
1522	return result;
1523}
1524
1525/*
1526 * Our own itoa().
1527 */
1528
1529STATIC char *
1530cvtnum(int num, char *buf)
1531{
1532	char temp[32];
1533	int neg = num < 0;
1534	char *p = temp + 31;
1535
1536	temp[31] = '\0';
1537
1538	do {
1539		*--p = num % 10 + '0';
1540	} while ((num /= 10) != 0);
1541
1542	if (neg)
1543		*--p = '-';
1544
1545	while (*p)
1546		STPUTC(*p++, buf);
1547	return buf;
1548}
1549
1550/*
1551 * Do most of the work for wordexp(3).
1552 */
1553
1554int
1555wordexpcmd(int argc, char **argv)
1556{
1557	size_t len;
1558	int i;
1559
1560	out1fmt("%08x", argc - 1);
1561	for (i = 1, len = 0; i < argc; i++)
1562		len += strlen(argv[i]);
1563	out1fmt("%08x", (int)len);
1564	for (i = 1; i < argc; i++) {
1565		out1str(argv[i]);
1566		out1c('\0');
1567	}
1568        return (0);
1569}
1570