xref: /freebsd/bin/sh/expand.c (revision 7661de35d15f582ab33e3bd6b8d909601557e436)
1 /*-
2  * Copyright (c) 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * Copyright (c) 1997-2005
5  *	Herbert Xu <herbert@gondor.apana.org.au>.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Kenneth Almquist.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 #ifndef lint
36 #if 0
37 static char sccsid[] = "@(#)expand.c	8.5 (Berkeley) 5/15/95";
38 #endif
39 #endif /* not lint */
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include <sys/types.h>
44 #include <sys/time.h>
45 #include <sys/stat.h>
46 #include <dirent.h>
47 #include <errno.h>
48 #include <inttypes.h>
49 #include <limits.h>
50 #include <pwd.h>
51 #include <stdio.h>
52 #include <stdlib.h>
53 #include <string.h>
54 #include <unistd.h>
55 #include <wchar.h>
56 #include <wctype.h>
57 
58 /*
59  * Routines to expand arguments to commands.  We have to deal with
60  * backquotes, shell variables, and file metacharacters.
61  */
62 
63 #include "shell.h"
64 #include "main.h"
65 #include "nodes.h"
66 #include "eval.h"
67 #include "expand.h"
68 #include "syntax.h"
69 #include "parser.h"
70 #include "jobs.h"
71 #include "options.h"
72 #include "var.h"
73 #include "input.h"
74 #include "output.h"
75 #include "memalloc.h"
76 #include "error.h"
77 #include "mystring.h"
78 #include "arith.h"
79 #include "show.h"
80 #include "builtins.h"
81 
82 /*
83  * Structure specifying which parts of the string should be searched
84  * for IFS characters.
85  */
86 
87 struct ifsregion {
88 	struct ifsregion *next;	/* next region in list */
89 	int begoff;		/* offset of start of region */
90 	int endoff;		/* offset of end of region */
91 	int inquotes;		/* search for nul bytes only */
92 };
93 
94 
95 static char *expdest;			/* output of current string */
96 static struct nodelist *argbackq;	/* list of back quote expressions */
97 static struct ifsregion ifsfirst;	/* first struct in list of ifs regions */
98 static struct ifsregion *ifslastp;	/* last struct in list */
99 static struct arglist exparg;		/* holds expanded arg list */
100 
101 static void argstr(char *, int);
102 static char *exptilde(char *, int);
103 static void expari(int);
104 static void expbackq(union node *, int, int);
105 static int subevalvar(char *, char *, int, int, int, int, int);
106 static char *evalvar(char *, int);
107 static int varisset(char *, int);
108 static void varvalue(char *, int, int, int);
109 static void recordregion(int, int, int);
110 static void removerecordregions(int);
111 static void ifsbreakup(char *, struct arglist *);
112 static void expandmeta(struct strlist *, int);
113 static void expmeta(char *, char *);
114 static void addfname(char *);
115 static struct strlist *expsort(struct strlist *);
116 static struct strlist *msort(struct strlist *, int);
117 static int patmatch(const char *, const char *, int);
118 static char *cvtnum(int, char *);
119 static int collate_range_cmp(wchar_t, wchar_t);
120 
121 static int
122 collate_range_cmp(wchar_t c1, wchar_t c2)
123 {
124 	static wchar_t s1[2], s2[2];
125 
126 	s1[0] = c1;
127 	s2[0] = c2;
128 	return (wcscoll(s1, s2));
129 }
130 
131 static char *
132 stputs_quotes(const char *data, const char *syntax, char *p)
133 {
134 	while (*data) {
135 		CHECKSTRSPACE(2, p);
136 		if (syntax[(int)*data] == CCTL)
137 			USTPUTC(CTLESC, p);
138 		USTPUTC(*data++, p);
139 	}
140 	return (p);
141 }
142 #define STPUTS_QUOTES(data, syntax, p) p = stputs_quotes((data), syntax, p)
143 
144 /*
145  * Perform expansions on an argument, placing the resulting list of arguments
146  * in arglist.  Parameter expansion, command substitution and arithmetic
147  * expansion are always performed; additional expansions can be requested
148  * via flag (EXP_*).
149  * The result is left in the stack string.
150  * When arglist is NULL, perform here document expansion.
151  *
152  * Caution: this function uses global state and is not reentrant.
153  * However, a new invocation after an interrupted invocation is safe
154  * and will reset the global state for the new call.
155  */
156 void
157 expandarg(union node *arg, struct arglist *arglist, int flag)
158 {
159 	struct strlist *sp;
160 	char *p;
161 
162 	argbackq = arg->narg.backquote;
163 	STARTSTACKSTR(expdest);
164 	ifsfirst.next = NULL;
165 	ifslastp = NULL;
166 	argstr(arg->narg.text, flag);
167 	if (arglist == NULL) {
168 		STACKSTRNUL(expdest);
169 		return;			/* here document expanded */
170 	}
171 	STPUTC('\0', expdest);
172 	p = grabstackstr(expdest);
173 	exparg.lastp = &exparg.list;
174 	/*
175 	 * TODO - EXP_REDIR
176 	 */
177 	if (flag & EXP_FULL) {
178 		ifsbreakup(p, &exparg);
179 		*exparg.lastp = NULL;
180 		exparg.lastp = &exparg.list;
181 		expandmeta(exparg.list, flag);
182 	} else {
183 		if (flag & EXP_REDIR) /*XXX - for now, just remove escapes */
184 			rmescapes(p);
185 		sp = (struct strlist *)stalloc(sizeof (struct strlist));
186 		sp->text = p;
187 		*exparg.lastp = sp;
188 		exparg.lastp = &sp->next;
189 	}
190 	while (ifsfirst.next != NULL) {
191 		struct ifsregion *ifsp;
192 		INTOFF;
193 		ifsp = ifsfirst.next->next;
194 		ckfree(ifsfirst.next);
195 		ifsfirst.next = ifsp;
196 		INTON;
197 	}
198 	*exparg.lastp = NULL;
199 	if (exparg.list) {
200 		*arglist->lastp = exparg.list;
201 		arglist->lastp = exparg.lastp;
202 	}
203 }
204 
205 
206 
207 /*
208  * Perform parameter expansion, command substitution and arithmetic
209  * expansion, and tilde expansion if requested via EXP_TILDE/EXP_VARTILDE.
210  * Processing ends at a CTLENDVAR character as well as '\0'.
211  * This is used to expand word in ${var+word} etc.
212  * If EXP_FULL, EXP_CASE or EXP_REDIR are set, keep and/or generate CTLESC
213  * characters to allow for further processing.
214  * If EXP_FULL is set, also preserve CTLQUOTEMARK characters.
215  */
216 static void
217 argstr(char *p, int flag)
218 {
219 	char c;
220 	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);	/* do CTLESC */
221 	int firsteq = 1;
222 	int split_lit;
223 	int lit_quoted;
224 
225 	split_lit = flag & EXP_SPLIT_LIT;
226 	lit_quoted = flag & EXP_LIT_QUOTED;
227 	flag &= ~(EXP_SPLIT_LIT | EXP_LIT_QUOTED);
228 	if (*p == '~' && (flag & (EXP_TILDE | EXP_VARTILDE)))
229 		p = exptilde(p, flag);
230 	for (;;) {
231 		CHECKSTRSPACE(2, expdest);
232 		switch (c = *p++) {
233 		case '\0':
234 		case CTLENDVAR:
235 			goto breakloop;
236 		case CTLQUOTEMARK:
237 			lit_quoted = 1;
238 			/* "$@" syntax adherence hack */
239 			if (p[0] == CTLVAR && p[2] == '@' && p[3] == '=')
240 				break;
241 			if ((flag & EXP_FULL) != 0)
242 				USTPUTC(c, expdest);
243 			break;
244 		case CTLQUOTEEND:
245 			lit_quoted = 0;
246 			break;
247 		case CTLESC:
248 			if (quotes)
249 				USTPUTC(c, expdest);
250 			c = *p++;
251 			USTPUTC(c, expdest);
252 			if (split_lit && !lit_quoted)
253 				recordregion(expdest - stackblock() -
254 				    (quotes ? 2 : 1),
255 				    expdest - stackblock(), 0);
256 			break;
257 		case CTLVAR:
258 			p = evalvar(p, flag);
259 			break;
260 		case CTLBACKQ:
261 		case CTLBACKQ|CTLQUOTE:
262 			expbackq(argbackq->n, c & CTLQUOTE, flag);
263 			argbackq = argbackq->next;
264 			break;
265 		case CTLENDARI:
266 			expari(flag);
267 			break;
268 		case ':':
269 		case '=':
270 			/*
271 			 * sort of a hack - expand tildes in variable
272 			 * assignments (after the first '=' and after ':'s).
273 			 */
274 			USTPUTC(c, expdest);
275 			if (split_lit && !lit_quoted)
276 				recordregion(expdest - stackblock() - 1,
277 				    expdest - stackblock(), 0);
278 			if (flag & EXP_VARTILDE && *p == '~' &&
279 			    (c != '=' || firsteq)) {
280 				if (c == '=')
281 					firsteq = 0;
282 				p = exptilde(p, flag);
283 			}
284 			break;
285 		default:
286 			USTPUTC(c, expdest);
287 			if (split_lit && !lit_quoted)
288 				recordregion(expdest - stackblock() - 1,
289 				    expdest - stackblock(), 0);
290 		}
291 	}
292 breakloop:;
293 }
294 
295 /*
296  * Perform tilde expansion, placing the result in the stack string and
297  * returning the next position in the input string to process.
298  */
299 static char *
300 exptilde(char *p, int flag)
301 {
302 	char c, *startp = p;
303 	struct passwd *pw;
304 	char *home;
305 	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
306 
307 	while ((c = *p) != '\0') {
308 		switch(c) {
309 		case CTLESC: /* This means CTL* are always considered quoted. */
310 		case CTLVAR:
311 		case CTLBACKQ:
312 		case CTLBACKQ | CTLQUOTE:
313 		case CTLARI:
314 		case CTLENDARI:
315 		case CTLQUOTEMARK:
316 			return (startp);
317 		case ':':
318 			if (flag & EXP_VARTILDE)
319 				goto done;
320 			break;
321 		case '/':
322 		case CTLENDVAR:
323 			goto done;
324 		}
325 		p++;
326 	}
327 done:
328 	*p = '\0';
329 	if (*(startp+1) == '\0') {
330 		if ((home = lookupvar("HOME")) == NULL)
331 			goto lose;
332 	} else {
333 		if ((pw = getpwnam(startp+1)) == NULL)
334 			goto lose;
335 		home = pw->pw_dir;
336 	}
337 	if (*home == '\0')
338 		goto lose;
339 	*p = c;
340 	if (quotes)
341 		STPUTS_QUOTES(home, SQSYNTAX, expdest);
342 	else
343 		STPUTS(home, expdest);
344 	return (p);
345 lose:
346 	*p = c;
347 	return (startp);
348 }
349 
350 
351 static void
352 removerecordregions(int endoff)
353 {
354 	if (ifslastp == NULL)
355 		return;
356 
357 	if (ifsfirst.endoff > endoff) {
358 		while (ifsfirst.next != NULL) {
359 			struct ifsregion *ifsp;
360 			INTOFF;
361 			ifsp = ifsfirst.next->next;
362 			ckfree(ifsfirst.next);
363 			ifsfirst.next = ifsp;
364 			INTON;
365 		}
366 		if (ifsfirst.begoff > endoff)
367 			ifslastp = NULL;
368 		else {
369 			ifslastp = &ifsfirst;
370 			ifsfirst.endoff = endoff;
371 		}
372 		return;
373 	}
374 
375 	ifslastp = &ifsfirst;
376 	while (ifslastp->next && ifslastp->next->begoff < endoff)
377 		ifslastp=ifslastp->next;
378 	while (ifslastp->next != NULL) {
379 		struct ifsregion *ifsp;
380 		INTOFF;
381 		ifsp = ifslastp->next->next;
382 		ckfree(ifslastp->next);
383 		ifslastp->next = ifsp;
384 		INTON;
385 	}
386 	if (ifslastp->endoff > endoff)
387 		ifslastp->endoff = endoff;
388 }
389 
390 /*
391  * Expand arithmetic expression.  Backup to start of expression,
392  * evaluate, place result in (backed up) result, adjust string position.
393  */
394 static void
395 expari(int flag)
396 {
397 	char *p, *q, *start;
398 	arith_t result;
399 	int begoff;
400 	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
401 	int quoted;
402 
403 	/*
404 	 * This routine is slightly over-complicated for
405 	 * efficiency.  First we make sure there is
406 	 * enough space for the result, which may be bigger
407 	 * than the expression.  Next we
408 	 * scan backwards looking for the start of arithmetic.  If the
409 	 * next previous character is a CTLESC character, then we
410 	 * have to rescan starting from the beginning since CTLESC
411 	 * characters have to be processed left to right.
412 	 */
413 	CHECKSTRSPACE(DIGITS(result) - 2, expdest);
414 	USTPUTC('\0', expdest);
415 	start = stackblock();
416 	p = expdest - 2;
417 	while (p >= start && *p != CTLARI)
418 		--p;
419 	if (p < start || *p != CTLARI)
420 		error("missing CTLARI (shouldn't happen)");
421 	if (p > start && *(p - 1) == CTLESC)
422 		for (p = start; *p != CTLARI; p++)
423 			if (*p == CTLESC)
424 				p++;
425 
426 	if (p[1] == '"')
427 		quoted=1;
428 	else
429 		quoted=0;
430 	begoff = p - start;
431 	removerecordregions(begoff);
432 	if (quotes)
433 		rmescapes(p+2);
434 	q = grabstackstr(expdest);
435 	result = arith(p+2);
436 	ungrabstackstr(q, expdest);
437 	fmtstr(p, DIGITS(result), ARITH_FORMAT_STR, result);
438 	while (*p++)
439 		;
440 	if (quoted == 0)
441 		recordregion(begoff, p - 1 - start, 0);
442 	result = expdest - p + 1;
443 	STADJUST(-result, expdest);
444 }
445 
446 
447 /*
448  * Perform command substitution.
449  */
450 static void
451 expbackq(union node *cmd, int quoted, int flag)
452 {
453 	struct backcmd in;
454 	int i;
455 	char buf[128];
456 	char *p;
457 	char *dest = expdest;
458 	struct ifsregion saveifs, *savelastp;
459 	struct nodelist *saveargbackq;
460 	char lastc;
461 	int startloc = dest - stackblock();
462 	char const *syntax = quoted? DQSYNTAX : BASESYNTAX;
463 	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
464 	size_t nnl;
465 
466 	INTOFF;
467 	saveifs = ifsfirst;
468 	savelastp = ifslastp;
469 	saveargbackq = argbackq;
470 	p = grabstackstr(dest);
471 	evalbackcmd(cmd, &in);
472 	ungrabstackstr(p, dest);
473 	ifsfirst = saveifs;
474 	ifslastp = savelastp;
475 	argbackq = saveargbackq;
476 
477 	p = in.buf;
478 	lastc = '\0';
479 	nnl = 0;
480 	/* Don't copy trailing newlines */
481 	for (;;) {
482 		if (--in.nleft < 0) {
483 			if (in.fd < 0)
484 				break;
485 			while ((i = read(in.fd, buf, sizeof buf)) < 0 && errno == EINTR);
486 			TRACE(("expbackq: read returns %d\n", i));
487 			if (i <= 0)
488 				break;
489 			p = buf;
490 			in.nleft = i - 1;
491 		}
492 		lastc = *p++;
493 		if (lastc != '\0') {
494 			if (lastc == '\n') {
495 				nnl++;
496 			} else {
497 				CHECKSTRSPACE(nnl + 2, dest);
498 				while (nnl > 0) {
499 					nnl--;
500 					USTPUTC('\n', dest);
501 				}
502 				if (quotes && syntax[(int)lastc] == CCTL)
503 					USTPUTC(CTLESC, dest);
504 				USTPUTC(lastc, dest);
505 			}
506 		}
507 	}
508 
509 	if (in.fd >= 0)
510 		close(in.fd);
511 	if (in.buf)
512 		ckfree(in.buf);
513 	if (in.jp)
514 		exitstatus = waitforjob(in.jp, (int *)NULL);
515 	if (quoted == 0)
516 		recordregion(startloc, dest - stackblock(), 0);
517 	TRACE(("expbackq: size=%td: \"%.*s\"\n",
518 		((dest - stackblock()) - startloc),
519 		(int)((dest - stackblock()) - startloc),
520 		stackblock() + startloc));
521 	expdest = dest;
522 	INTON;
523 }
524 
525 
526 
527 static int
528 subevalvar(char *p, char *str, int strloc, int subtype, int startloc,
529   int varflags, int quotes)
530 {
531 	char *startp;
532 	char *loc = NULL;
533 	char *q;
534 	int c = 0;
535 	struct nodelist *saveargbackq = argbackq;
536 	int amount;
537 
538 	argstr(p, (subtype == VSTRIMLEFT || subtype == VSTRIMLEFTMAX ||
539 	    subtype == VSTRIMRIGHT || subtype == VSTRIMRIGHTMAX ?
540 	    EXP_CASE : 0) | EXP_TILDE);
541 	STACKSTRNUL(expdest);
542 	argbackq = saveargbackq;
543 	startp = stackblock() + startloc;
544 	if (str == NULL)
545 	    str = stackblock() + strloc;
546 
547 	switch (subtype) {
548 	case VSASSIGN:
549 		setvar(str, startp, 0);
550 		amount = startp - expdest;
551 		STADJUST(amount, expdest);
552 		varflags &= ~VSNUL;
553 		return 1;
554 
555 	case VSQUESTION:
556 		if (*p != CTLENDVAR) {
557 			outfmt(out2, "%s\n", startp);
558 			error((char *)NULL);
559 		}
560 		error("%.*s: parameter %snot set", (int)(p - str - 1),
561 		      str, (varflags & VSNUL) ? "null or "
562 					      : nullstr);
563 		return 0;
564 
565 	case VSTRIMLEFT:
566 		for (loc = startp; loc < str; loc++) {
567 			c = *loc;
568 			*loc = '\0';
569 			if (patmatch(str, startp, quotes)) {
570 				*loc = c;
571 				goto recordleft;
572 			}
573 			*loc = c;
574 			if (quotes && *loc == CTLESC)
575 				loc++;
576 		}
577 		return 0;
578 
579 	case VSTRIMLEFTMAX:
580 		for (loc = str - 1; loc >= startp;) {
581 			c = *loc;
582 			*loc = '\0';
583 			if (patmatch(str, startp, quotes)) {
584 				*loc = c;
585 				goto recordleft;
586 			}
587 			*loc = c;
588 			loc--;
589 			if (quotes && loc > startp && *(loc - 1) == CTLESC) {
590 				for (q = startp; q < loc; q++)
591 					if (*q == CTLESC)
592 						q++;
593 				if (q > loc)
594 					loc--;
595 			}
596 		}
597 		return 0;
598 
599 	case VSTRIMRIGHT:
600 		for (loc = str - 1; loc >= startp;) {
601 			if (patmatch(str, loc, quotes)) {
602 				amount = loc - expdest;
603 				STADJUST(amount, expdest);
604 				return 1;
605 			}
606 			loc--;
607 			if (quotes && loc > startp && *(loc - 1) == CTLESC) {
608 				for (q = startp; q < loc; q++)
609 					if (*q == CTLESC)
610 						q++;
611 				if (q > loc)
612 					loc--;
613 			}
614 		}
615 		return 0;
616 
617 	case VSTRIMRIGHTMAX:
618 		for (loc = startp; loc < str - 1; loc++) {
619 			if (patmatch(str, loc, quotes)) {
620 				amount = loc - expdest;
621 				STADJUST(amount, expdest);
622 				return 1;
623 			}
624 			if (quotes && *loc == CTLESC)
625 				loc++;
626 		}
627 		return 0;
628 
629 
630 	default:
631 		abort();
632 	}
633 
634 recordleft:
635 	amount = ((str - 1) - (loc - startp)) - expdest;
636 	STADJUST(amount, expdest);
637 	while (loc != str - 1)
638 		*startp++ = *loc++;
639 	return 1;
640 }
641 
642 
643 /*
644  * Expand a variable, and return a pointer to the next character in the
645  * input string.
646  */
647 
648 static char *
649 evalvar(char *p, int flag)
650 {
651 	int subtype;
652 	int varflags;
653 	char *var;
654 	char *val;
655 	int patloc;
656 	int c;
657 	int set;
658 	int special;
659 	int startloc;
660 	int varlen;
661 	int varlenb;
662 	int easy;
663 	int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR);
664 
665 	varflags = (unsigned char)*p++;
666 	subtype = varflags & VSTYPE;
667 	var = p;
668 	special = 0;
669 	if (! is_name(*p))
670 		special = 1;
671 	p = strchr(p, '=') + 1;
672 again: /* jump here after setting a variable with ${var=text} */
673 	if (varflags & VSLINENO) {
674 		set = 1;
675 		special = 0;
676 		val = var;
677 		p[-1] = '\0';	/* temporarily overwrite '=' to have \0
678 				   terminated string */
679 	} else if (special) {
680 		set = varisset(var, varflags & VSNUL);
681 		val = NULL;
682 	} else {
683 		val = bltinlookup(var, 1);
684 		if (val == NULL || ((varflags & VSNUL) && val[0] == '\0')) {
685 			val = NULL;
686 			set = 0;
687 		} else
688 			set = 1;
689 	}
690 	varlen = 0;
691 	startloc = expdest - stackblock();
692 	if (!set && uflag && *var != '@' && *var != '*') {
693 		switch (subtype) {
694 		case VSNORMAL:
695 		case VSTRIMLEFT:
696 		case VSTRIMLEFTMAX:
697 		case VSTRIMRIGHT:
698 		case VSTRIMRIGHTMAX:
699 		case VSLENGTH:
700 			error("%.*s: parameter not set", (int)(p - var - 1),
701 			    var);
702 		}
703 	}
704 	if (set && subtype != VSPLUS) {
705 		/* insert the value of the variable */
706 		if (special) {
707 			varvalue(var, varflags & VSQUOTE, subtype, flag);
708 			if (subtype == VSLENGTH) {
709 				varlenb = expdest - stackblock() - startloc;
710 				varlen = varlenb;
711 				if (localeisutf8) {
712 					val = stackblock() + startloc;
713 					for (;val != expdest; val++)
714 						if ((*val & 0xC0) == 0x80)
715 							varlen--;
716 				}
717 				STADJUST(-varlenb, expdest);
718 			}
719 		} else {
720 			char const *syntax = (varflags & VSQUOTE) ? DQSYNTAX
721 								  : BASESYNTAX;
722 
723 			if (subtype == VSLENGTH) {
724 				for (;*val; val++)
725 					if (!localeisutf8 ||
726 					    (*val & 0xC0) != 0x80)
727 						varlen++;
728 			}
729 			else {
730 				if (quotes)
731 					STPUTS_QUOTES(val, syntax, expdest);
732 				else
733 					STPUTS(val, expdest);
734 
735 			}
736 		}
737 	}
738 
739 	if (subtype == VSPLUS)
740 		set = ! set;
741 
742 	easy = ((varflags & VSQUOTE) == 0 ||
743 		(*var == '@' && shellparam.nparam != 1));
744 
745 
746 	switch (subtype) {
747 	case VSLENGTH:
748 		expdest = cvtnum(varlen, expdest);
749 		goto record;
750 
751 	case VSNORMAL:
752 		if (!easy)
753 			break;
754 record:
755 		recordregion(startloc, expdest - stackblock(),
756 		    varflags & VSQUOTE || (ifsset() && ifsval()[0] == '\0' &&
757 		    (*var == '@' || *var == '*')));
758 		break;
759 
760 	case VSPLUS:
761 	case VSMINUS:
762 		if (!set) {
763 			argstr(p, flag | (flag & EXP_FULL ? EXP_SPLIT_LIT : 0) |
764 			    (varflags & VSQUOTE ? EXP_LIT_QUOTED : 0));
765 			break;
766 		}
767 		if (easy)
768 			goto record;
769 		break;
770 
771 	case VSTRIMLEFT:
772 	case VSTRIMLEFTMAX:
773 	case VSTRIMRIGHT:
774 	case VSTRIMRIGHTMAX:
775 		if (!set)
776 			break;
777 		/*
778 		 * Terminate the string and start recording the pattern
779 		 * right after it
780 		 */
781 		STPUTC('\0', expdest);
782 		patloc = expdest - stackblock();
783 		if (subevalvar(p, NULL, patloc, subtype,
784 		    startloc, varflags, quotes) == 0) {
785 			int amount = (expdest - stackblock() - patloc) + 1;
786 			STADJUST(-amount, expdest);
787 		}
788 		/* Remove any recorded regions beyond start of variable */
789 		removerecordregions(startloc);
790 		goto record;
791 
792 	case VSASSIGN:
793 	case VSQUESTION:
794 		if (!set) {
795 			if (subevalvar(p, var, 0, subtype, startloc, varflags,
796 			    quotes)) {
797 				varflags &= ~VSNUL;
798 				/*
799 				 * Remove any recorded regions beyond
800 				 * start of variable
801 				 */
802 				removerecordregions(startloc);
803 				goto again;
804 			}
805 			break;
806 		}
807 		if (easy)
808 			goto record;
809 		break;
810 
811 	case VSERROR:
812 		c = p - var - 1;
813 		error("${%.*s%s}: Bad substitution", c, var,
814 		    (c > 0 && *p != CTLENDVAR) ? "..." : "");
815 
816 	default:
817 		abort();
818 	}
819 	p[-1] = '=';	/* recover overwritten '=' */
820 
821 	if (subtype != VSNORMAL) {	/* skip to end of alternative */
822 		int nesting = 1;
823 		for (;;) {
824 			if ((c = *p++) == CTLESC)
825 				p++;
826 			else if (c == CTLBACKQ || c == (CTLBACKQ|CTLQUOTE)) {
827 				if (set)
828 					argbackq = argbackq->next;
829 			} else if (c == CTLVAR) {
830 				if ((*p++ & VSTYPE) != VSNORMAL)
831 					nesting++;
832 			} else if (c == CTLENDVAR) {
833 				if (--nesting == 0)
834 					break;
835 			}
836 		}
837 	}
838 	return p;
839 }
840 
841 
842 
843 /*
844  * Test whether a specialized variable is set.
845  */
846 
847 static int
848 varisset(char *name, int nulok)
849 {
850 
851 	if (*name == '!')
852 		return backgndpidset();
853 	else if (*name == '@' || *name == '*') {
854 		if (*shellparam.p == NULL)
855 			return 0;
856 
857 		if (nulok) {
858 			char **av;
859 
860 			for (av = shellparam.p; *av; av++)
861 				if (**av != '\0')
862 					return 1;
863 			return 0;
864 		}
865 	} else if (is_digit(*name)) {
866 		char *ap;
867 		int num = atoi(name);
868 
869 		if (num > shellparam.nparam)
870 			return 0;
871 
872 		if (num == 0)
873 			ap = arg0;
874 		else
875 			ap = shellparam.p[num - 1];
876 
877 		if (nulok && (ap == NULL || *ap == '\0'))
878 			return 0;
879 	}
880 	return 1;
881 }
882 
883 static void
884 strtodest(const char *p, int flag, int subtype, int quoted)
885 {
886 	if (flag & (EXP_FULL | EXP_CASE) && subtype != VSLENGTH)
887 		STPUTS_QUOTES(p, quoted ? DQSYNTAX : BASESYNTAX, expdest);
888 	else
889 		STPUTS(p, expdest);
890 }
891 
892 /*
893  * Add the value of a specialized variable to the stack string.
894  */
895 
896 static void
897 varvalue(char *name, int quoted, int subtype, int flag)
898 {
899 	int num;
900 	char *p;
901 	int i;
902 	char sep;
903 	char **ap;
904 
905 	switch (*name) {
906 	case '$':
907 		num = rootpid;
908 		goto numvar;
909 	case '?':
910 		num = oexitstatus;
911 		goto numvar;
912 	case '#':
913 		num = shellparam.nparam;
914 		goto numvar;
915 	case '!':
916 		num = backgndpidval();
917 numvar:
918 		expdest = cvtnum(num, expdest);
919 		break;
920 	case '-':
921 		for (i = 0 ; i < NOPTS ; i++) {
922 			if (optlist[i].val)
923 				STPUTC(optlist[i].letter, expdest);
924 		}
925 		break;
926 	case '@':
927 		if (flag & EXP_FULL && quoted) {
928 			for (ap = shellparam.p ; (p = *ap++) != NULL ; ) {
929 				strtodest(p, flag, subtype, quoted);
930 				if (*ap)
931 					STPUTC('\0', expdest);
932 			}
933 			break;
934 		}
935 		/* FALLTHROUGH */
936 	case '*':
937 		if (ifsset())
938 			sep = ifsval()[0];
939 		else
940 			sep = ' ';
941 		for (ap = shellparam.p ; (p = *ap++) != NULL ; ) {
942 			strtodest(p, flag, subtype, quoted);
943 			if (!*ap)
944 				break;
945 			if (sep || (flag & EXP_FULL && !quoted && **ap != '\0'))
946 				STPUTC(sep, expdest);
947 		}
948 		break;
949 	case '0':
950 		p = arg0;
951 		strtodest(p, flag, subtype, quoted);
952 		break;
953 	default:
954 		if (is_digit(*name)) {
955 			num = atoi(name);
956 			if (num > 0 && num <= shellparam.nparam) {
957 				p = shellparam.p[num - 1];
958 				strtodest(p, flag, subtype, quoted);
959 			}
960 		}
961 		break;
962 	}
963 }
964 
965 
966 
967 /*
968  * Record the fact that we have to scan this region of the
969  * string for IFS characters.
970  */
971 
972 static void
973 recordregion(int start, int end, int inquotes)
974 {
975 	struct ifsregion *ifsp;
976 
977 	if (ifslastp == NULL) {
978 		ifsp = &ifsfirst;
979 	} else {
980 		if (ifslastp->endoff == start
981 		    && ifslastp->inquotes == inquotes) {
982 			/* extend previous area */
983 			ifslastp->endoff = end;
984 			return;
985 		}
986 		ifsp = (struct ifsregion *)ckmalloc(sizeof (struct ifsregion));
987 		ifslastp->next = ifsp;
988 	}
989 	ifslastp = ifsp;
990 	ifslastp->next = NULL;
991 	ifslastp->begoff = start;
992 	ifslastp->endoff = end;
993 	ifslastp->inquotes = inquotes;
994 }
995 
996 
997 
998 /*
999  * Break the argument string into pieces based upon IFS and add the
1000  * strings to the argument list.  The regions of the string to be
1001  * searched for IFS characters have been stored by recordregion.
1002  * CTLESC characters are preserved but have little effect in this pass
1003  * other than escaping CTL* characters.  In particular, they do not escape
1004  * IFS characters: that should be done with the ifsregion mechanism.
1005  * CTLQUOTEMARK characters are used to preserve empty quoted strings.
1006  * This pass treats them as a regular character, making the string non-empty.
1007  * Later, they are removed along with the other CTL* characters.
1008  */
1009 static void
1010 ifsbreakup(char *string, struct arglist *arglist)
1011 {
1012 	struct ifsregion *ifsp;
1013 	struct strlist *sp;
1014 	char *start;
1015 	char *p;
1016 	char *q;
1017 	const char *ifs;
1018 	const char *ifsspc;
1019 	int had_param_ch = 0;
1020 
1021 	start = string;
1022 
1023 	if (ifslastp == NULL) {
1024 		/* Return entire argument, IFS doesn't apply to any of it */
1025 		sp = (struct strlist *)stalloc(sizeof *sp);
1026 		sp->text = start;
1027 		*arglist->lastp = sp;
1028 		arglist->lastp = &sp->next;
1029 		return;
1030 	}
1031 
1032 	ifs = ifsset() ? ifsval() : " \t\n";
1033 
1034 	for (ifsp = &ifsfirst; ifsp != NULL; ifsp = ifsp->next) {
1035 		p = string + ifsp->begoff;
1036 		while (p < string + ifsp->endoff) {
1037 			q = p;
1038 			if (*p == CTLESC)
1039 				p++;
1040 			if (ifsp->inquotes) {
1041 				/* Only NULs (should be from "$@") end args */
1042 				had_param_ch = 1;
1043 				if (*p != 0) {
1044 					p++;
1045 					continue;
1046 				}
1047 				ifsspc = NULL;
1048 			} else {
1049 				if (!strchr(ifs, *p)) {
1050 					had_param_ch = 1;
1051 					p++;
1052 					continue;
1053 				}
1054 				ifsspc = strchr(" \t\n", *p);
1055 
1056 				/* Ignore IFS whitespace at start */
1057 				if (q == start && ifsspc != NULL) {
1058 					p++;
1059 					start = p;
1060 					continue;
1061 				}
1062 				had_param_ch = 0;
1063 			}
1064 
1065 			/* Save this argument... */
1066 			*q = '\0';
1067 			sp = (struct strlist *)stalloc(sizeof *sp);
1068 			sp->text = start;
1069 			*arglist->lastp = sp;
1070 			arglist->lastp = &sp->next;
1071 			p++;
1072 
1073 			if (ifsspc != NULL) {
1074 				/* Ignore further trailing IFS whitespace */
1075 				for (; p < string + ifsp->endoff; p++) {
1076 					q = p;
1077 					if (*p == CTLESC)
1078 						p++;
1079 					if (strchr(ifs, *p) == NULL) {
1080 						p = q;
1081 						break;
1082 					}
1083 					if (strchr(" \t\n", *p) == NULL) {
1084 						p++;
1085 						break;
1086 					}
1087 				}
1088 			}
1089 			start = p;
1090 		}
1091 	}
1092 
1093 	/*
1094 	 * Save anything left as an argument.
1095 	 * Traditionally we have treated 'IFS=':'; set -- x$IFS' as
1096 	 * generating 2 arguments, the second of which is empty.
1097 	 * Some recent clarification of the Posix spec say that it
1098 	 * should only generate one....
1099 	 */
1100 	if (had_param_ch || *start != 0) {
1101 		sp = (struct strlist *)stalloc(sizeof *sp);
1102 		sp->text = start;
1103 		*arglist->lastp = sp;
1104 		arglist->lastp = &sp->next;
1105 	}
1106 }
1107 
1108 
1109 static char expdir[PATH_MAX];
1110 #define expdir_end (expdir + sizeof(expdir))
1111 
1112 /*
1113  * Perform pathname generation and remove control characters.
1114  * At this point, the only control characters should be CTLESC and CTLQUOTEMARK.
1115  * The results are stored in the list exparg.
1116  */
1117 static void
1118 expandmeta(struct strlist *str, int flag __unused)
1119 {
1120 	char *p;
1121 	struct strlist **savelastp;
1122 	struct strlist *sp;
1123 	char c;
1124 	/* TODO - EXP_REDIR */
1125 
1126 	while (str) {
1127 		if (fflag)
1128 			goto nometa;
1129 		p = str->text;
1130 		for (;;) {			/* fast check for meta chars */
1131 			if ((c = *p++) == '\0')
1132 				goto nometa;
1133 			if (c == '*' || c == '?' || c == '[')
1134 				break;
1135 		}
1136 		savelastp = exparg.lastp;
1137 		INTOFF;
1138 		expmeta(expdir, str->text);
1139 		INTON;
1140 		if (exparg.lastp == savelastp) {
1141 			/*
1142 			 * no matches
1143 			 */
1144 nometa:
1145 			*exparg.lastp = str;
1146 			rmescapes(str->text);
1147 			exparg.lastp = &str->next;
1148 		} else {
1149 			*exparg.lastp = NULL;
1150 			*savelastp = sp = expsort(*savelastp);
1151 			while (sp->next != NULL)
1152 				sp = sp->next;
1153 			exparg.lastp = &sp->next;
1154 		}
1155 		str = str->next;
1156 	}
1157 }
1158 
1159 
1160 /*
1161  * Do metacharacter (i.e. *, ?, [...]) expansion.
1162  */
1163 
1164 static void
1165 expmeta(char *enddir, char *name)
1166 {
1167 	const char *p;
1168 	const char *q;
1169 	const char *start;
1170 	char *endname;
1171 	int metaflag;
1172 	struct stat statb;
1173 	DIR *dirp;
1174 	struct dirent *dp;
1175 	int atend;
1176 	int matchdot;
1177 	int esc;
1178 	int namlen;
1179 
1180 	metaflag = 0;
1181 	start = name;
1182 	for (p = name; esc = 0, *p; p += esc + 1) {
1183 		if (*p == '*' || *p == '?')
1184 			metaflag = 1;
1185 		else if (*p == '[') {
1186 			q = p + 1;
1187 			if (*q == '!' || *q == '^')
1188 				q++;
1189 			for (;;) {
1190 				while (*q == CTLQUOTEMARK)
1191 					q++;
1192 				if (*q == CTLESC)
1193 					q++;
1194 				if (*q == '/' || *q == '\0')
1195 					break;
1196 				if (*++q == ']') {
1197 					metaflag = 1;
1198 					break;
1199 				}
1200 			}
1201 		} else if (*p == '\0')
1202 			break;
1203 		else if (*p == CTLQUOTEMARK)
1204 			continue;
1205 		else {
1206 			if (*p == CTLESC)
1207 				esc++;
1208 			if (p[esc] == '/') {
1209 				if (metaflag)
1210 					break;
1211 				start = p + esc + 1;
1212 			}
1213 		}
1214 	}
1215 	if (metaflag == 0) {	/* we've reached the end of the file name */
1216 		if (enddir != expdir)
1217 			metaflag++;
1218 		for (p = name ; ; p++) {
1219 			if (*p == CTLQUOTEMARK)
1220 				continue;
1221 			if (*p == CTLESC)
1222 				p++;
1223 			*enddir++ = *p;
1224 			if (*p == '\0')
1225 				break;
1226 			if (enddir == expdir_end)
1227 				return;
1228 		}
1229 		if (metaflag == 0 || lstat(expdir, &statb) >= 0)
1230 			addfname(expdir);
1231 		return;
1232 	}
1233 	endname = name + (p - name);
1234 	if (start != name) {
1235 		p = name;
1236 		while (p < start) {
1237 			while (*p == CTLQUOTEMARK)
1238 				p++;
1239 			if (*p == CTLESC)
1240 				p++;
1241 			*enddir++ = *p++;
1242 			if (enddir == expdir_end)
1243 				return;
1244 		}
1245 	}
1246 	if (enddir == expdir) {
1247 		p = ".";
1248 	} else if (enddir == expdir + 1 && *expdir == '/') {
1249 		p = "/";
1250 	} else {
1251 		p = expdir;
1252 		enddir[-1] = '\0';
1253 	}
1254 	if ((dirp = opendir(p)) == NULL)
1255 		return;
1256 	if (enddir != expdir)
1257 		enddir[-1] = '/';
1258 	if (*endname == 0) {
1259 		atend = 1;
1260 	} else {
1261 		atend = 0;
1262 		*endname = '\0';
1263 		endname += esc + 1;
1264 	}
1265 	matchdot = 0;
1266 	p = start;
1267 	while (*p == CTLQUOTEMARK)
1268 		p++;
1269 	if (*p == CTLESC)
1270 		p++;
1271 	if (*p == '.')
1272 		matchdot++;
1273 	while (! int_pending() && (dp = readdir(dirp)) != NULL) {
1274 		if (dp->d_name[0] == '.' && ! matchdot)
1275 			continue;
1276 		if (patmatch(start, dp->d_name, 0)) {
1277 			namlen = dp->d_namlen;
1278 			if (enddir + namlen + 1 > expdir_end)
1279 				continue;
1280 			memcpy(enddir, dp->d_name, namlen + 1);
1281 			if (atend)
1282 				addfname(expdir);
1283 			else {
1284 				if (dp->d_type != DT_UNKNOWN &&
1285 				    dp->d_type != DT_DIR &&
1286 				    dp->d_type != DT_LNK)
1287 					continue;
1288 				if (enddir + namlen + 2 > expdir_end)
1289 					continue;
1290 				enddir[namlen] = '/';
1291 				enddir[namlen + 1] = '\0';
1292 				expmeta(enddir + namlen + 1, endname);
1293 			}
1294 		}
1295 	}
1296 	closedir(dirp);
1297 	if (! atend)
1298 		endname[-esc - 1] = esc ? CTLESC : '/';
1299 }
1300 
1301 
1302 /*
1303  * Add a file name to the list.
1304  */
1305 
1306 static void
1307 addfname(char *name)
1308 {
1309 	char *p;
1310 	struct strlist *sp;
1311 	size_t len;
1312 
1313 	len = strlen(name);
1314 	p = stalloc(len + 1);
1315 	memcpy(p, name, len + 1);
1316 	sp = (struct strlist *)stalloc(sizeof *sp);
1317 	sp->text = p;
1318 	*exparg.lastp = sp;
1319 	exparg.lastp = &sp->next;
1320 }
1321 
1322 
1323 /*
1324  * Sort the results of file name expansion.  It calculates the number of
1325  * strings to sort and then calls msort (short for merge sort) to do the
1326  * work.
1327  */
1328 
1329 static struct strlist *
1330 expsort(struct strlist *str)
1331 {
1332 	int len;
1333 	struct strlist *sp;
1334 
1335 	len = 0;
1336 	for (sp = str ; sp ; sp = sp->next)
1337 		len++;
1338 	return msort(str, len);
1339 }
1340 
1341 
1342 static struct strlist *
1343 msort(struct strlist *list, int len)
1344 {
1345 	struct strlist *p, *q = NULL;
1346 	struct strlist **lpp;
1347 	int half;
1348 	int n;
1349 
1350 	if (len <= 1)
1351 		return list;
1352 	half = len >> 1;
1353 	p = list;
1354 	for (n = half ; --n >= 0 ; ) {
1355 		q = p;
1356 		p = p->next;
1357 	}
1358 	q->next = NULL;			/* terminate first half of list */
1359 	q = msort(list, half);		/* sort first half of list */
1360 	p = msort(p, len - half);		/* sort second half */
1361 	lpp = &list;
1362 	for (;;) {
1363 		if (strcmp(p->text, q->text) < 0) {
1364 			*lpp = p;
1365 			lpp = &p->next;
1366 			if ((p = *lpp) == NULL) {
1367 				*lpp = q;
1368 				break;
1369 			}
1370 		} else {
1371 			*lpp = q;
1372 			lpp = &q->next;
1373 			if ((q = *lpp) == NULL) {
1374 				*lpp = p;
1375 				break;
1376 			}
1377 		}
1378 	}
1379 	return list;
1380 }
1381 
1382 
1383 
1384 static wchar_t
1385 get_wc(const char **p)
1386 {
1387 	wchar_t c;
1388 	int chrlen;
1389 
1390 	chrlen = mbtowc(&c, *p, 4);
1391 	if (chrlen == 0)
1392 		return 0;
1393 	else if (chrlen == -1)
1394 		c = 0;
1395 	else
1396 		*p += chrlen;
1397 	return c;
1398 }
1399 
1400 
1401 /*
1402  * See if a character matches a character class, starting at the first colon
1403  * of "[:class:]".
1404  * If a valid character class is recognized, a pointer to the next character
1405  * after the final closing bracket is stored into *end, otherwise a null
1406  * pointer is stored into *end.
1407  */
1408 static int
1409 match_charclass(const char *p, wchar_t chr, const char **end)
1410 {
1411 	char name[20];
1412 	const char *nameend;
1413 	wctype_t cclass;
1414 
1415 	*end = NULL;
1416 	p++;
1417 	nameend = strstr(p, ":]");
1418 	if (nameend == NULL || (size_t)(nameend - p) >= sizeof(name) ||
1419 	    nameend == p)
1420 		return 0;
1421 	memcpy(name, p, nameend - p);
1422 	name[nameend - p] = '\0';
1423 	*end = nameend + 2;
1424 	cclass = wctype(name);
1425 	/* An unknown class matches nothing but is valid nevertheless. */
1426 	if (cclass == 0)
1427 		return 0;
1428 	return iswctype(chr, cclass);
1429 }
1430 
1431 
1432 /*
1433  * Returns true if the pattern matches the string.
1434  */
1435 
1436 static int
1437 patmatch(const char *pattern, const char *string, int squoted)
1438 {
1439 	const char *p, *q, *end;
1440 	const char *bt_p, *bt_q;
1441 	char c;
1442 	wchar_t wc, wc2;
1443 
1444 	p = pattern;
1445 	q = string;
1446 	bt_p = NULL;
1447 	bt_q = NULL;
1448 	for (;;) {
1449 		switch (c = *p++) {
1450 		case '\0':
1451 			if (*q != '\0')
1452 				goto backtrack;
1453 			return 1;
1454 		case CTLESC:
1455 			if (squoted && *q == CTLESC)
1456 				q++;
1457 			if (*q++ != *p++)
1458 				goto backtrack;
1459 			break;
1460 		case CTLQUOTEMARK:
1461 			continue;
1462 		case '?':
1463 			if (squoted && *q == CTLESC)
1464 				q++;
1465 			if (*q == '\0')
1466 				return 0;
1467 			if (localeisutf8) {
1468 				wc = get_wc(&q);
1469 				/*
1470 				 * A '?' does not match invalid UTF-8 but a
1471 				 * '*' does, so backtrack.
1472 				 */
1473 				if (wc == 0)
1474 					goto backtrack;
1475 			} else
1476 				wc = (unsigned char)*q++;
1477 			break;
1478 		case '*':
1479 			c = *p;
1480 			while (c == CTLQUOTEMARK || c == '*')
1481 				c = *++p;
1482 			/*
1483 			 * If the pattern ends here, we know the string
1484 			 * matches without needing to look at the rest of it.
1485 			 */
1486 			if (c == '\0')
1487 				return 1;
1488 			/*
1489 			 * First try the shortest match for the '*' that
1490 			 * could work. We can forget any earlier '*' since
1491 			 * there is no way having it match more characters
1492 			 * can help us, given that we are already here.
1493 			 */
1494 			bt_p = p;
1495 			bt_q = q;
1496 			break;
1497 		case '[': {
1498 			const char *endp;
1499 			int invert, found;
1500 			wchar_t chr;
1501 
1502 			endp = p;
1503 			if (*endp == '!' || *endp == '^')
1504 				endp++;
1505 			for (;;) {
1506 				while (*endp == CTLQUOTEMARK)
1507 					endp++;
1508 				if (*endp == 0)
1509 					goto dft;		/* no matching ] */
1510 				if (*endp == CTLESC)
1511 					endp++;
1512 				if (*++endp == ']')
1513 					break;
1514 			}
1515 			invert = 0;
1516 			if (*p == '!' || *p == '^') {
1517 				invert++;
1518 				p++;
1519 			}
1520 			found = 0;
1521 			if (squoted && *q == CTLESC)
1522 				q++;
1523 			if (*q == '\0')
1524 				return 0;
1525 			if (localeisutf8) {
1526 				chr = get_wc(&q);
1527 				if (chr == 0)
1528 					goto backtrack;
1529 			} else
1530 				chr = (unsigned char)*q++;
1531 			c = *p++;
1532 			do {
1533 				if (c == CTLQUOTEMARK)
1534 					continue;
1535 				if (c == '[' && *p == ':') {
1536 					found |= match_charclass(p, chr, &end);
1537 					if (end != NULL)
1538 						p = end;
1539 				}
1540 				if (c == CTLESC)
1541 					c = *p++;
1542 				if (localeisutf8 && c & 0x80) {
1543 					p--;
1544 					wc = get_wc(&p);
1545 					if (wc == 0) /* bad utf-8 */
1546 						return 0;
1547 				} else
1548 					wc = (unsigned char)c;
1549 				if (*p == '-' && p[1] != ']') {
1550 					p++;
1551 					while (*p == CTLQUOTEMARK)
1552 						p++;
1553 					if (*p == CTLESC)
1554 						p++;
1555 					if (localeisutf8) {
1556 						wc2 = get_wc(&p);
1557 						if (wc2 == 0) /* bad utf-8 */
1558 							return 0;
1559 					} else
1560 						wc2 = (unsigned char)*p++;
1561 					if (   collate_range_cmp(chr, wc) >= 0
1562 					    && collate_range_cmp(chr, wc2) <= 0
1563 					   )
1564 						found = 1;
1565 				} else {
1566 					if (chr == wc)
1567 						found = 1;
1568 				}
1569 			} while ((c = *p++) != ']');
1570 			if (found == invert)
1571 				goto backtrack;
1572 			break;
1573 		}
1574 dft:	        default:
1575 			if (squoted && *q == CTLESC)
1576 				q++;
1577 			if (*q == '\0')
1578 				return 0;
1579 			if (*q++ == c)
1580 				break;
1581 backtrack:
1582 			/*
1583 			 * If we have a mismatch (other than hitting the end
1584 			 * of the string), go back to the last '*' seen and
1585 			 * have it match one additional character.
1586 			 */
1587 			if (bt_p == NULL)
1588 				return 0;
1589 			if (squoted && *bt_q == CTLESC)
1590 				bt_q++;
1591 			if (*bt_q == '\0')
1592 				return 0;
1593 			bt_q++;
1594 			p = bt_p;
1595 			q = bt_q;
1596 			break;
1597 		}
1598 	}
1599 }
1600 
1601 
1602 
1603 /*
1604  * Remove any CTLESC and CTLQUOTEMARK characters from a string.
1605  */
1606 
1607 void
1608 rmescapes(char *str)
1609 {
1610 	char *p, *q;
1611 
1612 	p = str;
1613 	while (*p != CTLESC && *p != CTLQUOTEMARK && *p != CTLQUOTEEND) {
1614 		if (*p++ == '\0')
1615 			return;
1616 	}
1617 	q = p;
1618 	while (*p) {
1619 		if (*p == CTLQUOTEMARK || *p == CTLQUOTEEND) {
1620 			p++;
1621 			continue;
1622 		}
1623 		if (*p == CTLESC)
1624 			p++;
1625 		*q++ = *p++;
1626 	}
1627 	*q = '\0';
1628 }
1629 
1630 
1631 
1632 /*
1633  * See if a pattern matches in a case statement.
1634  */
1635 
1636 int
1637 casematch(union node *pattern, const char *val)
1638 {
1639 	struct stackmark smark;
1640 	int result;
1641 	char *p;
1642 
1643 	setstackmark(&smark);
1644 	argbackq = pattern->narg.backquote;
1645 	STARTSTACKSTR(expdest);
1646 	ifslastp = NULL;
1647 	argstr(pattern->narg.text, EXP_TILDE | EXP_CASE);
1648 	STPUTC('\0', expdest);
1649 	p = grabstackstr(expdest);
1650 	result = patmatch(p, val, 0);
1651 	popstackmark(&smark);
1652 	return result;
1653 }
1654 
1655 /*
1656  * Our own itoa().
1657  */
1658 
1659 static char *
1660 cvtnum(int num, char *buf)
1661 {
1662 	char temp[32];
1663 	int neg = num < 0;
1664 	char *p = temp + 31;
1665 
1666 	temp[31] = '\0';
1667 
1668 	do {
1669 		*--p = num % 10 + '0';
1670 	} while ((num /= 10) != 0);
1671 
1672 	if (neg)
1673 		*--p = '-';
1674 
1675 	STPUTS(p, buf);
1676 	return buf;
1677 }
1678 
1679 /*
1680  * Do most of the work for wordexp(3).
1681  */
1682 
1683 int
1684 wordexpcmd(int argc, char **argv)
1685 {
1686 	size_t len;
1687 	int i;
1688 
1689 	out1fmt("%08x", argc - 1);
1690 	for (i = 1, len = 0; i < argc; i++)
1691 		len += strlen(argv[i]);
1692 	out1fmt("%08x", (int)len);
1693 	for (i = 1; i < argc; i++)
1694 		outbin(argv[i], strlen(argv[i]) + 1, out1);
1695         return (0);
1696 }
1697