xref: /freebsd/contrib/tcsh/glob.c (revision 2219fc0f)
1 /*
2  * Copyright (c) 1989 The Regents of the University of California.
3  * All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Guido van Rossum.
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  * 3. 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 #if defined(LIBC_SCCS) && !defined(lint)
33 static char sccsid[] = "@(#)glob.c	5.12 (Berkeley) 6/24/91";
34 #endif /* LIBC_SCCS and not lint */
35 /*
36  * Glob: the interface is a superset of the one defined in POSIX 1003.2,
37  * draft 9.
38  *
39  * The [!...] convention to negate a range is supported (SysV, Posix, ksh).
40  *
41  * Optional extra services, controlled by flags not defined by POSIX:
42  *
43  * GLOB_QUOTE:
44  *	Escaping convention: \ inhibits any special meaning the following
45  *	character might have (except \ at end of string is retained).
46  * GLOB_MAGCHAR:
47  *	Set in gl_flags if pattern contained a globbing character.
48  * GLOB_ALTNOT:
49  *	Use ^ instead of ! for "not".
50  * gl_matchc:
51  *	Number of matches in the current invocation of glob.
52  */
53 
54 #ifdef WINNT_NATIVE
55 	#pragma warning(disable:4244)
56 #endif /* WINNT_NATIVE */
57 
58 #define Char __Char
59 #include "sh.h"
60 #include "glob.h"
61 
62 #ifndef HAVE_MBLEN
63 #undef mblen
64 #define mblen(_s,_n)	mbrlen((_s),(_n),NULL)
65 #endif
66 
67 #undef Char
68 #undef QUOTE
69 #undef TILDE
70 #undef META
71 #undef ismeta
72 #undef Strchr
73 
74 #ifndef S_ISDIR
75 #define S_ISDIR(a)	(((a) & S_IFMT) == S_IFDIR)
76 #endif
77 
78 #if !defined(S_ISLNK) && defined(S_IFLNK)
79 #define S_ISLNK(a)	(((a) & S_IFMT) == S_IFLNK)
80 #endif
81 
82 #if !defined(S_ISLNK) && !defined(lstat)
83 #define lstat stat
84 #endif
85 
86 typedef unsigned short Char;
87 
88 static	int	 glob1 		(Char *, glob_t *, int);
89 static	int	 glob2		(struct strbuf *, const Char *, glob_t *, int);
90 static	int	 glob3		(struct strbuf *, const Char *, const Char *,
91 				 const Char *, glob_t *, int);
92 static	void	 globextend	(const char *, glob_t *);
93 static	int	 match		(const char *, const Char *, const Char *,
94 				 int);
95 static	int	 compare	(const void *, const void *);
96 static 	DIR	*Opendir	(const char *);
97 #ifdef S_IFLNK
98 static	int	 Lstat		(const char *, struct stat *);
99 #endif
100 static	int	 Stat		(const char *, struct stat *sb);
101 static 	Char 	*Strchr		(Char *, int);
102 #ifdef DEBUG
103 static	void	 qprintf	(const char *, const Char *);
104 #endif
105 
106 #define	DOLLAR		'$'
107 #define	DOT		'.'
108 #define	EOS		'\0'
109 #define	LBRACKET	'['
110 #define	NOT		'!'
111 #define ALTNOT		'^'
112 #define	QUESTION	'?'
113 #define	QUOTE		'\\'
114 #define	RANGE		'-'
115 #define	RBRACKET	']'
116 #define	SEP		'/'
117 #define	STAR		'*'
118 #define	TILDE		'~'
119 #define	UNDERSCORE	'_'
120 
121 #define	M_META		0x8000
122 #define M_PROTECT	0x4000
123 #define	M_MASK		0xffff
124 #define	M_ASCII		0x00ff
125 
126 #define	LCHAR(c)	((c)&M_ASCII)
127 #define	META(c)		((c)|M_META)
128 #define	M_ALL		META('*')
129 #define	M_END		META(']')
130 #define	M_NOT		META('!')
131 #define	M_ALTNOT	META('^')
132 #define	M_ONE		META('?')
133 #define	M_RNG		META('-')
134 #define	M_SET		META('[')
135 #define	ismeta(c)	(((c)&M_META) != 0)
136 
137 int
globcharcoll(__Char c1,__Char c2,int cs)138 globcharcoll(__Char c1, __Char c2, int cs)
139 {
140 #if defined(NLS) && defined(LC_COLLATE) && defined(HAVE_STRCOLL)
141 # if defined(WIDE_STRINGS)
142     wchar_t s1[2], s2[2];
143 
144     if (c1 == c2)
145 	return (0);
146     if (cs) {
147 	c1 = towlower(c1);
148 	c2 = towlower(c2);
149     } else {
150 #ifndef __FreeBSD__
151 	/* This should not be here, but I'll rather leave it in than engage in
152 	   a LC_COLLATE flamewar about a shell I don't use... */
153 	if (iswlower(c1) && iswupper(c2))
154 	    return (1);
155 	if (iswupper(c1) && iswlower(c2))
156 	    return (-1);
157 #endif
158     }
159     s1[0] = c1;
160     s2[0] = c2;
161     s1[1] = s2[1] = '\0';
162     return wcscoll(s1, s2);
163 # else /* not WIDE_STRINGS */
164     char s1[2], s2[2];
165 
166     if (c1 == c2)
167 	return (0);
168     /*
169      * From kevin lyda <kevin@suberic.net>:
170      * strcoll does not guarantee case sorting, so we pre-process now:
171      */
172     if (cs) {
173 	c1 = islower(c1) ? c1 : tolower(c1);
174 	c2 = islower(c2) ? c2 : tolower(c2);
175     } else {
176 	if (islower(c1) && isupper(c2))
177 	    return (1);
178 	if (isupper(c1) && islower(c2))
179 	    return (-1);
180     }
181     s1[0] = c1;
182     s2[0] = c2;
183     s1[1] = s2[1] = '\0';
184     return strcoll(s1, s2);
185 # endif
186 #else
187     return (c1 - c2);
188 #endif
189 }
190 
191 /*
192  * Need to dodge two kernel bugs:
193  * opendir("") != opendir(".")
194  * NAMEI_BUG: on plain files trailing slashes are ignored in some kernels.
195  *            POSIX specifies that they should be ignored in directories.
196  */
197 
198 static DIR *
Opendir(const char * str)199 Opendir(const char *str)
200 {
201 #if defined(hpux) || defined(__hpux)
202     struct stat st;
203 #endif
204 
205     if (!*str)
206 	return (opendir("."));
207 #if defined(hpux) || defined(__hpux)
208     /*
209      * Opendir on some device files hangs, so avoid it
210      */
211     if (stat(str, &st) == -1 || !S_ISDIR(st.st_mode))
212 	return NULL;
213 #endif
214     return opendir(str);
215 }
216 
217 #ifdef S_IFLNK
218 static int
Lstat(const char * fn,struct stat * sb)219 Lstat(const char *fn, struct stat *sb)
220 {
221     int st;
222 
223     st = lstat(fn, sb);
224 # ifdef NAMEI_BUG
225     if (*fn != 0 && strend(fn)[-1] == '/' && !S_ISDIR(sb->st_mode))
226 	st = -1;
227 # endif	/* NAMEI_BUG */
228     return st;
229 }
230 #else
231 #define Lstat Stat
232 #endif /* S_IFLNK */
233 
234 static int
Stat(const char * fn,struct stat * sb)235 Stat(const char *fn, struct stat *sb)
236 {
237     int st;
238 
239     st = stat(fn, sb);
240 #ifdef NAMEI_BUG
241     if (*fn != 0 && strend(fn)[-1] == '/' && !S_ISDIR(sb->st_mode))
242 	st = -1;
243 #endif /* NAMEI_BUG */
244     return st;
245 }
246 
247 static Char *
Strchr(Char * str,int ch)248 Strchr(Char *str, int ch)
249 {
250     do
251 	if (*str == ch)
252 	    return (str);
253     while (*str++);
254     return (NULL);
255 }
256 
257 #ifdef DEBUG
258 static void
qprintf(const char * pre,const Char * s)259 qprintf(const char *pre, const Char *s)
260 {
261     const Char *p;
262 
263     xprintf("%s", pre);
264     for (p = s; *p; p++)
265 	xprintf("%c", *p & 0xff);
266     xprintf("\n%s", pre);
267     for (p = s; *p; p++)
268 	xprintf("%c", *p & M_PROTECT ? '"' : ' ');
269     xprintf("\n%s", pre);
270     for (p = s; *p; p++)
271 	xprintf("%c", *p & M_META ? '_' : ' ');
272     xprintf("\n");
273 }
274 #endif /* DEBUG */
275 
276 static int
compare(const void * p,const void * q)277 compare(const void *p, const void *q)
278 {
279 #if defined(NLS) && defined(HAVE_STRCOLL)
280     return (strcoll(*(char *const *) p, *(char *const *) q));
281 #else
282     return (strcmp(*(char *const *) p, *(char *const *) q));
283 #endif /* NLS && HAVE_STRCOLL */
284 }
285 
286 /*
287  * The main glob() routine: compiles the pattern (optionally processing
288  * quotes), calls glob1() to do the real pattern matching, and finally
289  * sorts the list (unless unsorted operation is requested).  Returns 0
290  * if things went well, nonzero if errors occurred.  It is not an error
291  * to find no matches.
292  */
293 int
glob(const char * pattern,int flags,int (* errfunc)(const char *,int),glob_t * pglob)294 glob(const char *pattern, int flags, int (*errfunc) (const char *, int),
295      glob_t *pglob)
296 {
297     int     err, oldpathc;
298     Char *bufnext, m_not;
299     const unsigned char *patnext;
300     int     c, not;
301     Char *qpatnext, *patbuf;
302     int     no_match;
303 
304     patnext = (const unsigned char *) pattern;
305     if (!(flags & GLOB_APPEND)) {
306 	pglob->gl_pathc = 0;
307 	pglob->gl_pathv = NULL;
308 	if (!(flags & GLOB_DOOFFS))
309 	    pglob->gl_offs = 0;
310     }
311     pglob->gl_flags = flags & ~GLOB_MAGCHAR;
312     pglob->gl_errfunc = errfunc;
313     oldpathc = pglob->gl_pathc;
314     pglob->gl_matchc = 0;
315 
316     if (pglob->gl_flags & GLOB_ALTNOT) {
317 	not = ALTNOT;
318 	m_not = M_ALTNOT;
319     }
320     else {
321 	not = NOT;
322 	m_not = M_NOT;
323     }
324 
325     patbuf = xmalloc((strlen(pattern) + 1) * sizeof(*patbuf));
326     bufnext = patbuf;
327 
328     no_match = *patnext == not;
329     if (no_match)
330 	patnext++;
331 
332     if (flags & GLOB_QUOTE) {
333 	/* Protect the quoted characters */
334 	while ((c = *patnext++) != EOS) {
335 #ifdef WIDE_STRINGS
336 	    int len;
337 
338 	    len = mblen((const char *)(patnext - 1), MB_LEN_MAX);
339 	    if (len == -1)
340 		TCSH_IGNORE(mblen(NULL, 0));
341 	    else if (len > 1) {
342 		*bufnext++ = (Char) c;
343 		while (--len != 0)
344 		    *bufnext++ = (Char) (*patnext++ | M_PROTECT);
345 	    } else
346 #endif /* WIDE_STRINGS */
347 	    if (c == QUOTE) {
348 		if ((c = *patnext++) == EOS) {
349 		    c = QUOTE;
350 		    --patnext;
351 		}
352 		*bufnext++ = (Char) (c | M_PROTECT);
353 	    }
354 	    else
355 		*bufnext++ = (Char) c;
356 	}
357     }
358     else
359 	while ((c = *patnext++) != EOS)
360 	    *bufnext++ = (Char) c;
361     *bufnext = EOS;
362 
363     bufnext = patbuf;
364     qpatnext = patbuf;
365     while ((c = *qpatnext++) != EOS) {
366 	switch (c) {
367 	case LBRACKET:
368 	    c = *qpatnext;
369 	    if (c == not)
370 		++qpatnext;
371 	    if (*qpatnext == EOS ||
372 		Strchr(qpatnext + 1, RBRACKET) == NULL) {
373 		*bufnext++ = LBRACKET;
374 		if (c == not)
375 		    --qpatnext;
376 		break;
377 	    }
378 	    pglob->gl_flags |= GLOB_MAGCHAR;
379 	    *bufnext++ = M_SET;
380 	    if (c == not)
381 		*bufnext++ = m_not;
382 	    c = *qpatnext++;
383 	    do {
384 		*bufnext++ = LCHAR(c);
385 		if (*qpatnext == RANGE &&
386 		    (c = qpatnext[1]) != RBRACKET) {
387 		    *bufnext++ = M_RNG;
388 		    *bufnext++ = LCHAR(c);
389 		    qpatnext += 2;
390 		}
391 	    } while ((c = *qpatnext++) != RBRACKET);
392 	    *bufnext++ = M_END;
393 	    break;
394 	case QUESTION:
395 	    pglob->gl_flags |= GLOB_MAGCHAR;
396 	    *bufnext++ = M_ONE;
397 	    break;
398 	case STAR:
399 	    pglob->gl_flags |= GLOB_MAGCHAR;
400 	    /* collapse adjacent stars to one [or three if globstar],
401 	     * to avoid exponential behavior
402 	     */
403 	    if (bufnext == patbuf || bufnext[-1] != M_ALL ||
404 	       ((flags & GLOB_STAR) != 0 &&
405 		 (bufnext - 1 == patbuf || bufnext[-2] != M_ALL ||
406 		 bufnext - 2 == patbuf || bufnext[-3] != M_ALL)))
407 		*bufnext++ = M_ALL;
408 	    break;
409 	default:
410 	    *bufnext++ = LCHAR(c);
411 	    break;
412 	}
413     }
414     *bufnext = EOS;
415 #ifdef DEBUG
416     qprintf("patbuf=", patbuf);
417 #endif
418 
419     if ((err = glob1(patbuf, pglob, no_match)) != 0) {
420 	xfree(patbuf);
421 	return (err);
422     }
423 
424     /*
425      * If there was no match we are going to append the pattern
426      * if GLOB_NOCHECK was specified or if GLOB_NOMAGIC was specified
427      * and the pattern did not contain any magic characters
428      * GLOB_NOMAGIC is there just for compatibility with csh.
429      */
430     if (pglob->gl_pathc == oldpathc &&
431 	((flags & GLOB_NOCHECK) ||
432 	 ((flags & GLOB_NOMAGIC) && !(pglob->gl_flags & GLOB_MAGCHAR)))) {
433 	if (!(flags & GLOB_QUOTE))
434 	    globextend(pattern, pglob);
435 	else {
436 	    char *copy, *dest;
437 	    const char *src;
438 
439 	    /* copy pattern, interpreting quotes */
440 	    copy = xmalloc(strlen(pattern) + 1);
441 	    dest = copy;
442 	    src = pattern;
443 	    while (*src != EOS) {
444 		/* Don't interpret quotes. The spec does not say we should do */
445 		if (*src == QUOTE) {
446 		    if (*++src == EOS)
447 			--src;
448 		}
449 		*dest++ = *src++;
450 	    }
451 	    *dest = EOS;
452 	    globextend(copy, pglob);
453 	    xfree(copy);
454 	}
455 	xfree(patbuf);
456 	return 0;
457     }
458     else if (!(flags & GLOB_NOSORT) && (pglob->gl_pathc != oldpathc))
459 	qsort(pglob->gl_pathv + pglob->gl_offs + oldpathc,
460 	      pglob->gl_pathc - oldpathc, sizeof(char *), compare);
461     xfree(patbuf);
462     return (0);
463 }
464 
465 static int
glob1(Char * pattern,glob_t * pglob,int no_match)466 glob1(Char *pattern, glob_t *pglob, int no_match)
467 {
468     struct strbuf pathbuf = strbuf_INIT;
469     int err;
470 
471     /*
472      * a null pathname is invalid -- POSIX 1003.1 sect. 2.4.
473      */
474     if (*pattern == EOS)
475 	return (0);
476     err = glob2(&pathbuf, pattern, pglob, no_match);
477     xfree(pathbuf.s);
478     return err;
479 }
480 
481 /*
482  * functions glob2 and glob3 are mutually recursive; there is one level
483  * of recursion for each segment in the pattern that contains one or
484  * more meta characters.
485  */
486 static int
glob2(struct strbuf * pathbuf,const Char * pattern,glob_t * pglob,int no_match)487 glob2(struct strbuf *pathbuf, const Char *pattern, glob_t *pglob, int no_match)
488 {
489     struct stat sbuf;
490     int anymeta;
491     const Char *p;
492     size_t orig_len;
493 
494     /*
495      * loop over pattern segments until end of pattern or until segment with
496      * meta character found.
497      */
498     anymeta = 0;
499     for (;;) {
500 	if (*pattern == EOS) {	/* end of pattern? */
501 	    strbuf_terminate(pathbuf);
502 
503 	    if (Lstat(pathbuf->s, &sbuf))
504 		return (0);
505 
506 	    if (((pglob->gl_flags & GLOB_MARK) &&
507 		 pathbuf->s[pathbuf->len - 1] != SEP) &&
508 		(S_ISDIR(sbuf.st_mode)
509 #ifdef S_IFLNK
510 		 || (S_ISLNK(sbuf.st_mode) &&
511 		     (Stat(pathbuf->s, &sbuf) == 0) &&
512 		     S_ISDIR(sbuf.st_mode))
513 #endif
514 		 )) {
515 		strbuf_append1(pathbuf, SEP);
516 		strbuf_terminate(pathbuf);
517 	    }
518 	    ++pglob->gl_matchc;
519 	    globextend(pathbuf->s, pglob);
520 	    return 0;
521 	}
522 
523 	/* find end of next segment, tentatively copy to pathbuf */
524 	p = pattern;
525 	orig_len = pathbuf->len;
526 	while (*p != EOS && *p != SEP) {
527 	    if (ismeta(*p))
528 		anymeta = 1;
529 	    strbuf_append1(pathbuf, *p++);
530 	}
531 
532 	if (!anymeta) {		/* no expansion, do next segment */
533 	    pattern = p;
534 	    while (*pattern == SEP)
535 		strbuf_append1(pathbuf, *pattern++);
536 	}
537 	else {			/* need expansion, recurse */
538 	    pathbuf->len = orig_len;
539 	    return (glob3(pathbuf, pattern, p, pattern, pglob, no_match));
540 	}
541     }
542     /* NOTREACHED */
543 }
544 
545 static size_t
One_Char_mbtowc(__Char * pwc,const Char * s,size_t n)546 One_Char_mbtowc(__Char *pwc, const Char *s, size_t n)
547 {
548 #ifdef WIDE_STRINGS
549     char buf[MB_LEN_MAX], *p;
550 
551     if (n > MB_LEN_MAX)
552 	n = MB_LEN_MAX;
553     p = buf;
554     while (p < buf + n && (*p++ = LCHAR(*s++)) != 0)
555 	;
556     return one_mbtowc(pwc, buf, n);
557 #else
558     *pwc = *s & CHAR;
559     return 1;
560 #endif
561 }
562 
563 static int
glob3(struct strbuf * pathbuf,const Char * pattern,const Char * restpattern,const Char * pglobstar,glob_t * pglob,int no_match)564 glob3(struct strbuf *pathbuf, const Char *pattern, const Char *restpattern,
565       const Char *pglobstar, glob_t *pglob, int no_match)
566 {
567     DIR    *dirp;
568     struct dirent *dp;
569     struct stat sbuf;
570     int     err;
571     Char m_not = (pglob->gl_flags & GLOB_ALTNOT) ? M_ALTNOT : M_NOT;
572     size_t orig_len;
573     int globstar = 0;
574     int chase_symlinks = 0;
575     const Char *termstar = NULL;
576 
577     strbuf_terminate(pathbuf);
578     orig_len = pathbuf->len;
579     errno = err = 0;
580 
581     while (pglobstar < restpattern) {
582 	__Char wc;
583 	size_t width = One_Char_mbtowc(&wc, pglobstar, MB_LEN_MAX);
584 	if ((pglobstar[0] & M_MASK) == M_ALL &&
585 	    (pglobstar[width] & M_MASK) == M_ALL) {
586 	    globstar = 1;
587 	    chase_symlinks = (pglobstar[2 * width] & M_MASK) == M_ALL;
588 	    termstar = pglobstar + (2 + chase_symlinks) * width;
589 	    break;
590 	}
591         pglobstar += width;
592     }
593 
594     if (globstar) {
595 	err = pglobstar==pattern && termstar==restpattern ?
596 		*restpattern == EOS ?
597 		glob2(pathbuf, restpattern - 1, pglob, no_match) :
598 		glob2(pathbuf, restpattern + 1, pglob, no_match) :
599 		glob3(pathbuf, pattern, restpattern, termstar, pglob, no_match);
600 	if (err)
601 	    return err;
602 	pathbuf->len = orig_len;
603 	strbuf_terminate(pathbuf);
604     }
605 
606     if (*pathbuf->s && (Lstat(pathbuf->s, &sbuf) || !S_ISDIR(sbuf.st_mode)
607 #ifdef S_IFLINK
608 	     && ((globstar && !chase_symlinks) || !S_ISLNK(sbuf.st_mode))
609 #endif
610 	))
611 	return 0;
612 
613     if (!(dirp = Opendir(pathbuf->s))) {
614 	/* todo: don't call for ENOENT or ENOTDIR? */
615 	if ((pglob->gl_errfunc && (*pglob->gl_errfunc) (pathbuf->s, errno)) ||
616 	    (pglob->gl_flags & GLOB_ERR))
617 	    return (GLOB_ABEND);
618 	else
619 	    return (0);
620     }
621 
622     /* search directory for matching names */
623     while ((dp = readdir(dirp)) != NULL) {
624 	/* initial DOT must be matched literally */
625 	if (dp->d_name[0] == DOT && *pattern != DOT)
626 	    if (!(pglob->gl_flags & GLOB_DOT) || !dp->d_name[1] ||
627 		(dp->d_name[1] == DOT && !dp->d_name[2]))
628 		continue; /*unless globdot and not . or .. */
629 	pathbuf->len = orig_len;
630 	strbuf_append(pathbuf, dp->d_name);
631 	strbuf_terminate(pathbuf);
632 
633 	if (globstar) {
634 #ifdef S_IFLNK
635 	    if (!chase_symlinks &&
636 		(Lstat(pathbuf->s, &sbuf) || S_ISLNK(sbuf.st_mode)))
637 		    continue;
638 #endif
639 	    if (match(pathbuf->s + orig_len, pattern, termstar,
640 		(int)m_not) == no_match)
641 		    continue;
642 	    strbuf_append1(pathbuf, SEP);
643 	    strbuf_terminate(pathbuf);
644 	    if ((err = glob2(pathbuf, pglobstar, pglob, no_match)) != 0)
645 		break;
646 	} else {
647 	    if (match(pathbuf->s + orig_len, pattern, restpattern,
648 		(int) m_not) == no_match)
649 		continue;
650 	    if ((err = glob2(pathbuf, restpattern, pglob, no_match)) != 0)
651 		break;
652 	}
653     }
654     /* todo: check error from readdir? */
655     closedir(dirp);
656     return (err);
657 }
658 
659 
660 /*
661  * Extend the gl_pathv member of a glob_t structure to accomodate a new item,
662  * add the new item, and update gl_pathc.
663  *
664  * This assumes the BSD realloc, which only copies the block when its size
665  * crosses a power-of-two boundary; for v7 realloc, this would cause quadratic
666  * behavior.
667  *
668  * Return 0 if new item added, error code if memory couldn't be allocated.
669  *
670  * Invariant of the glob_t structure:
671  *	Either gl_pathc is zero and gl_pathv is NULL; or gl_pathc > 0 and
672  *	 gl_pathv points to (gl_offs + gl_pathc + 1) items.
673  */
674 static void
globextend(const char * path,glob_t * pglob)675 globextend(const char *path, glob_t *pglob)
676 {
677     char **pathv;
678     int i;
679     size_t newsize;
680 
681     newsize = sizeof(*pathv) * (2 + pglob->gl_pathc + pglob->gl_offs);
682     pathv = xrealloc(pglob->gl_pathv, newsize);
683 
684     if (pglob->gl_pathv == NULL && pglob->gl_offs > 0) {
685 	/* first time around -- clear initial gl_offs items */
686 	pathv += pglob->gl_offs;
687 	for (i = pglob->gl_offs; --i >= 0;)
688 	    *--pathv = NULL;
689     }
690     pglob->gl_pathv = pathv;
691 
692     pathv[pglob->gl_offs + pglob->gl_pathc++] = strsave(path);
693     pathv[pglob->gl_offs + pglob->gl_pathc] = NULL;
694 }
695 
696 /*
697  * pattern matching function for filenames.
698  */
699 static  int
match(const char * name,const Char * pat,const Char * patend,int m_not)700 match(const char *name, const Char *pat, const Char *patend, int m_not)
701 {
702     int ok, negate_range;
703     const Char *patNext;
704     const char *nameNext, *nameStart, *nameEnd;
705     Char c;
706 
707     patNext = pat;
708     nameStart = nameNext = name;
709     nameEnd = NULL;
710 
711     while (pat < patend || *name) {
712 	size_t lwk, pwk;
713 	__Char wc, wk, wc1;
714 
715 	c = *pat; /* Only for M_MASK bits */
716 	if (*name == EOS)
717 		nameEnd = name;
718 
719 	pwk = One_Char_mbtowc(&wc, pat, MB_LEN_MAX);
720 	lwk = one_mbtowc(&wk, name, MB_LEN_MAX);
721 	switch (c & M_MASK) {
722 	case M_ALL:
723 	    while ((*(pat + pwk) & M_MASK) == M_ALL) {
724 		pat += pwk;
725 		pwk = One_Char_mbtowc(&wc, pat, MB_LEN_MAX);
726 	    }
727 	    patNext = pat;
728 	    nameNext = name + lwk;
729 	    pat += pwk;
730 	    continue;
731 	case M_ONE:
732 	    if (*name == EOS)
733 		break;
734 	    name += lwk;
735 	    pat += pwk;
736 	    continue;
737 	case M_SET:
738 	    ok = 0;
739 	    if (*name == EOS)
740 		break;
741 	    pat += pwk;
742 	    pwk = One_Char_mbtowc(&wc, pat, MB_LEN_MAX);
743 	    name += lwk;
744 	    if ((negate_range = ((*pat & M_MASK) == m_not)) != 0) {
745 		pat += pwk;
746 		pwk = One_Char_mbtowc(&wc, pat, MB_LEN_MAX);
747 	    }
748 	    wc1 = wc;
749 	    while ((*pat & M_MASK) != M_END) {
750 		if ((*pat & M_MASK) == M_RNG) {
751 		    __Char wc2;
752 
753 		    pat += pwk;
754 		    pwk = One_Char_mbtowc(&wc2, pat, MB_LEN_MAX);
755 		    if (globcharcoll(wc1, wk, 0) <= 0 &&
756 			globcharcoll(wk, wc2, 0) <= 0)
757 			ok = 1;
758 		} else if (wc == wk)
759 		    ok = 1;
760 		pat += pwk;
761 		wc1 = wc;
762 		pwk = One_Char_mbtowc(&wc, pat, MB_LEN_MAX);
763 	    }
764 	    pat += pwk;
765 	    pwk = One_Char_mbtowc(&wc, pat, MB_LEN_MAX);
766 	    if (ok == negate_range)
767 		break;
768 	    continue;
769 	default:
770 	    if (*name == EOS || samecase(wk) != samecase(wc))
771 		break;
772 	    name += lwk;
773 	    pat += pwk;
774 	    continue;
775 	}
776 	if (nameNext != nameStart
777 	    && (nameEnd == NULL || nameNext <= nameEnd)) {
778 	    pat = patNext;
779 	    name = nameNext;
780 	    continue;
781 	}
782 	return 0;
783     }
784     return 1;
785 }
786 
787 /* free allocated data belonging to a glob_t structure */
788 void
globfree(glob_t * pglob)789 globfree(glob_t *pglob)
790 {
791     int i;
792     char **pp;
793 
794     if (pglob->gl_pathv != NULL) {
795 	pp = pglob->gl_pathv + pglob->gl_offs;
796 	for (i = pglob->gl_pathc; i--; ++pp)
797 	    if (*pp)
798 		xfree(*pp), *pp = NULL;
799 	xfree(pglob->gl_pathv), pglob->gl_pathv = NULL;
800     }
801 }
802