[BZ #1978]
[glibc.git] / posix / fnmatch_loop.c
blob02a4e9030c79aac8678668ce159bc9b1415ccc14
1 /* Copyright (C) 1991,1992,1993,1996,1997,1998,1999,2000,2001,2003,2004,2005
2 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
5 The GNU C Library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 2.1 of the License, or (at your option) any later version.
10 The GNU C Library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
15 You should have received a copy of the GNU Lesser General Public
16 License along with the GNU C Library; if not, write to the Free
17 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
18 02111-1307 USA. */
20 /* Match STRING against the filename pattern PATTERN, returning zero if
21 it matches, nonzero if not. */
22 static int FCT (const CHAR *pattern, const CHAR *string,
23 const CHAR *string_end, int no_leading_period, int flags)
24 internal_function;
25 static int EXT (INT opt, const CHAR *pattern, const CHAR *string,
26 const CHAR *string_end, int no_leading_period, int flags)
27 internal_function;
28 static const CHAR *END (const CHAR *patternp) internal_function;
30 static int
31 internal_function
32 FCT (pattern, string, string_end, no_leading_period, flags)
33 const CHAR *pattern;
34 const CHAR *string;
35 const CHAR *string_end;
36 int no_leading_period;
37 int flags;
39 register const CHAR *p = pattern, *n = string;
40 register UCHAR c;
41 #ifdef _LIBC
42 # if WIDE_CHAR_VERSION
43 const char *collseq = (const char *)
44 _NL_CURRENT(LC_COLLATE, _NL_COLLATE_COLLSEQWC);
45 # else
46 const UCHAR *collseq = (const UCHAR *)
47 _NL_CURRENT(LC_COLLATE, _NL_COLLATE_COLLSEQMB);
48 # endif
49 #endif
51 while ((c = *p++) != L('\0'))
53 int new_no_leading_period = 0;
54 c = FOLD (c);
56 switch (c)
58 case L('?'):
59 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
61 int res;
63 res = EXT (c, p, n, string_end, no_leading_period,
64 flags);
65 if (res != -1)
66 return res;
69 if (n == string_end)
70 return FNM_NOMATCH;
71 else if (*n == L('/') && (flags & FNM_FILE_NAME))
72 return FNM_NOMATCH;
73 else if (*n == L('.') && no_leading_period)
74 return FNM_NOMATCH;
75 break;
77 case L('\\'):
78 if (!(flags & FNM_NOESCAPE))
80 c = *p++;
81 if (c == L('\0'))
82 /* Trailing \ loses. */
83 return FNM_NOMATCH;
84 c = FOLD (c);
86 if (n == string_end || FOLD ((UCHAR) *n) != c)
87 return FNM_NOMATCH;
88 break;
90 case L('*'):
91 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
93 int res;
95 res = EXT (c, p, n, string_end, no_leading_period,
96 flags);
97 if (res != -1)
98 return res;
101 if (n != string_end && *n == L('.') && no_leading_period)
102 return FNM_NOMATCH;
104 for (c = *p++; c == L('?') || c == L('*'); c = *p++)
106 if (*p == L('(') && (flags & FNM_EXTMATCH) != 0)
108 const CHAR *endp = END (p);
109 if (endp != p)
111 /* This is a pattern. Skip over it. */
112 p = endp;
113 continue;
117 if (c == L('?'))
119 /* A ? needs to match one character. */
120 if (n == string_end)
121 /* There isn't another character; no match. */
122 return FNM_NOMATCH;
123 else if (*n == L('/')
124 && __builtin_expect (flags & FNM_FILE_NAME, 0))
125 /* A slash does not match a wildcard under
126 FNM_FILE_NAME. */
127 return FNM_NOMATCH;
128 else
129 /* One character of the string is consumed in matching
130 this ? wildcard, so *??? won't match if there are
131 less than three characters. */
132 ++n;
136 if (c == L('\0'))
137 /* The wildcard(s) is/are the last element of the pattern.
138 If the name is a file name and contains another slash
139 this means it cannot match, unless the FNM_LEADING_DIR
140 flag is set. */
142 int result = (flags & FNM_FILE_NAME) == 0 ? 0 : FNM_NOMATCH;
144 if (flags & FNM_FILE_NAME)
146 if (flags & FNM_LEADING_DIR)
147 result = 0;
148 else
150 if (MEMCHR (n, L('/'), string_end - n) == NULL)
151 result = 0;
155 return result;
157 else
159 const CHAR *endp;
161 endp = MEMCHR (n, (flags & FNM_FILE_NAME) ? L('/') : L('\0'),
162 string_end - n);
163 if (endp == NULL)
164 endp = string_end;
166 if (c == L('[')
167 || (__builtin_expect (flags & FNM_EXTMATCH, 0) != 0
168 && (c == L('@') || c == L('+') || c == L('!'))
169 && *p == L('(')))
171 int flags2 = ((flags & FNM_FILE_NAME)
172 ? flags : (flags & ~FNM_PERIOD));
173 int no_leading_period2 = no_leading_period;
175 for (--p; n < endp; ++n, no_leading_period2 = 0)
176 if (FCT (p, n, string_end, no_leading_period2, flags2)
177 == 0)
178 return 0;
180 else if (c == L('/') && (flags & FNM_FILE_NAME))
182 while (n < string_end && *n != L('/'))
183 ++n;
184 if (n < string_end && *n == L('/')
185 && (FCT (p, n + 1, string_end, flags & FNM_PERIOD, flags)
186 == 0))
187 return 0;
189 else
191 int flags2 = ((flags & FNM_FILE_NAME)
192 ? flags : (flags & ~FNM_PERIOD));
193 int no_leading_period2 = no_leading_period;
195 if (c == L('\\') && !(flags & FNM_NOESCAPE))
196 c = *p;
197 c = FOLD (c);
198 for (--p; n < endp; ++n, no_leading_period2 = 0)
199 if (FOLD ((UCHAR) *n) == c
200 && (FCT (p, n, string_end, no_leading_period2, flags2)
201 == 0))
202 return 0;
206 /* If we come here no match is possible with the wildcard. */
207 return FNM_NOMATCH;
209 case L('['):
211 /* Nonzero if the sense of the character class is inverted. */
212 register int not;
213 CHAR cold;
214 UCHAR fn;
216 if (posixly_correct == 0)
217 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
219 if (n == string_end)
220 return FNM_NOMATCH;
222 if (*n == L('.') && no_leading_period)
223 return FNM_NOMATCH;
225 if (*n == L('/') && (flags & FNM_FILE_NAME))
226 /* `/' cannot be matched. */
227 return FNM_NOMATCH;
229 not = (*p == L('!') || (posixly_correct < 0 && *p == L('^')));
230 if (not)
231 ++p;
233 fn = FOLD ((UCHAR) *n);
235 c = *p++;
236 for (;;)
238 if (!(flags & FNM_NOESCAPE) && c == L('\\'))
240 if (*p == L('\0'))
241 return FNM_NOMATCH;
242 c = FOLD ((UCHAR) *p);
243 ++p;
245 goto normal_bracket;
247 else if (c == L('[') && *p == L(':'))
249 /* Leave room for the null. */
250 CHAR str[CHAR_CLASS_MAX_LENGTH + 1];
251 size_t c1 = 0;
252 #if defined _LIBC || (defined HAVE_WCTYPE_H && defined HAVE_WCHAR_H)
253 wctype_t wt;
254 #endif
255 const CHAR *startp = p;
257 for (;;)
259 if (c1 == CHAR_CLASS_MAX_LENGTH)
260 /* The name is too long and therefore the pattern
261 is ill-formed. */
262 return FNM_NOMATCH;
264 c = *++p;
265 if (c == L(':') && p[1] == L(']'))
267 p += 2;
268 break;
270 if (c < L('a') || c >= L('z'))
272 /* This cannot possibly be a character class name.
273 Match it as a normal range. */
274 p = startp;
275 c = L('[');
276 goto normal_bracket;
278 str[c1++] = c;
280 str[c1] = L('\0');
282 #if defined _LIBC || (defined HAVE_WCTYPE_H && defined HAVE_WCHAR_H)
283 wt = IS_CHAR_CLASS (str);
284 if (wt == 0)
285 /* Invalid character class name. */
286 return FNM_NOMATCH;
288 # if defined _LIBC && ! WIDE_CHAR_VERSION
289 /* The following code is glibc specific but does
290 there a good job in speeding up the code since
291 we can avoid the btowc() call. */
292 if (_ISCTYPE ((UCHAR) *n, wt))
293 goto matched;
294 # else
295 if (ISWCTYPE (BTOWC ((UCHAR) *n), wt))
296 goto matched;
297 # endif
298 #else
299 if ((STREQ (str, L("alnum")) && ISALNUM ((UCHAR) *n))
300 || (STREQ (str, L("alpha")) && ISALPHA ((UCHAR) *n))
301 || (STREQ (str, L("blank")) && ISBLANK ((UCHAR) *n))
302 || (STREQ (str, L("cntrl")) && ISCNTRL ((UCHAR) *n))
303 || (STREQ (str, L("digit")) && ISDIGIT ((UCHAR) *n))
304 || (STREQ (str, L("graph")) && ISGRAPH ((UCHAR) *n))
305 || (STREQ (str, L("lower")) && ISLOWER ((UCHAR) *n))
306 || (STREQ (str, L("print")) && ISPRINT ((UCHAR) *n))
307 || (STREQ (str, L("punct")) && ISPUNCT ((UCHAR) *n))
308 || (STREQ (str, L("space")) && ISSPACE ((UCHAR) *n))
309 || (STREQ (str, L("upper")) && ISUPPER ((UCHAR) *n))
310 || (STREQ (str, L("xdigit")) && ISXDIGIT ((UCHAR) *n)))
311 goto matched;
312 #endif
313 c = *p++;
315 #ifdef _LIBC
316 else if (c == L('[') && *p == L('='))
318 UCHAR str[1];
319 uint32_t nrules =
320 _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
321 const CHAR *startp = p;
323 c = *++p;
324 if (c == L('\0'))
326 p = startp;
327 c = L('[');
328 goto normal_bracket;
330 str[0] = c;
332 c = *++p;
333 if (c != L('=') || p[1] != L(']'))
335 p = startp;
336 c = L('[');
337 goto normal_bracket;
339 p += 2;
341 if (nrules == 0)
343 if ((UCHAR) *n == str[0])
344 goto matched;
346 else
348 const int32_t *table;
349 # if WIDE_CHAR_VERSION
350 const int32_t *weights;
351 const int32_t *extra;
352 # else
353 const unsigned char *weights;
354 const unsigned char *extra;
355 # endif
356 const int32_t *indirect;
357 int32_t idx;
358 const UCHAR *cp = (const UCHAR *) str;
360 /* This #include defines a local function! */
361 # if WIDE_CHAR_VERSION
362 # include <locale/weightwc.h>
363 # else
364 # include <locale/weight.h>
365 # endif
367 # if WIDE_CHAR_VERSION
368 table = (const int32_t *)
369 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEWC);
370 weights = (const int32_t *)
371 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTWC);
372 extra = (const int32_t *)
373 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAWC);
374 indirect = (const int32_t *)
375 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTWC);
376 # else
377 table = (const int32_t *)
378 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB);
379 weights = (const unsigned char *)
380 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTMB);
381 extra = (const unsigned char *)
382 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB);
383 indirect = (const int32_t *)
384 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTMB);
385 # endif
387 idx = findidx (&cp);
388 if (idx != 0)
390 /* We found a table entry. Now see whether the
391 character we are currently at has the same
392 equivalance class value. */
393 int len = weights[idx];
394 int32_t idx2;
395 const UCHAR *np = (const UCHAR *) n;
397 idx2 = findidx (&np);
398 if (idx2 != 0 && len == weights[idx2])
400 int cnt = 0;
402 while (cnt < len
403 && (weights[idx + 1 + cnt]
404 == weights[idx2 + 1 + cnt]))
405 ++cnt;
407 if (cnt == len)
408 goto matched;
413 c = *p++;
415 #endif
416 else if (c == L('\0'))
417 /* [ (unterminated) loses. */
418 return FNM_NOMATCH;
419 else
421 int is_range = 0;
423 #ifdef _LIBC
424 int is_seqval = 0;
426 if (c == L('[') && *p == L('.'))
428 uint32_t nrules =
429 _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
430 const CHAR *startp = p;
431 size_t c1 = 0;
433 while (1)
435 c = *++p;
436 if (c == L('.') && p[1] == L(']'))
438 p += 2;
439 break;
441 if (c == '\0')
442 return FNM_NOMATCH;
443 ++c1;
446 /* We have to handling the symbols differently in
447 ranges since then the collation sequence is
448 important. */
449 is_range = *p == L('-') && p[1] != L('\0');
451 if (nrules == 0)
453 /* There are no names defined in the collation
454 data. Therefore we only accept the trivial
455 names consisting of the character itself. */
456 if (c1 != 1)
457 return FNM_NOMATCH;
459 if (!is_range && *n == startp[1])
460 goto matched;
462 cold = startp[1];
463 c = *p++;
465 else
467 int32_t table_size;
468 const int32_t *symb_table;
469 # ifdef WIDE_CHAR_VERSION
470 char str[c1];
471 unsigned int strcnt;
472 # else
473 # define str (startp + 1)
474 # endif
475 const unsigned char *extra;
476 int32_t idx;
477 int32_t elem;
478 int32_t second;
479 int32_t hash;
481 # ifdef WIDE_CHAR_VERSION
482 /* We have to convert the name to a single-byte
483 string. This is possible since the names
484 consist of ASCII characters and the internal
485 representation is UCS4. */
486 for (strcnt = 0; strcnt < c1; ++strcnt)
487 str[strcnt] = startp[1 + strcnt];
488 #endif
490 table_size =
491 _NL_CURRENT_WORD (LC_COLLATE,
492 _NL_COLLATE_SYMB_HASH_SIZEMB);
493 symb_table = (const int32_t *)
494 _NL_CURRENT (LC_COLLATE,
495 _NL_COLLATE_SYMB_TABLEMB);
496 extra = (const unsigned char *)
497 _NL_CURRENT (LC_COLLATE,
498 _NL_COLLATE_SYMB_EXTRAMB);
500 /* Locate the character in the hashing table. */
501 hash = elem_hash (str, c1);
503 idx = 0;
504 elem = hash % table_size;
505 if (symb_table[2 * elem] != 0)
507 second = hash % (table_size - 2) + 1;
511 /* First compare the hashing value. */
512 if (symb_table[2 * elem] == hash
513 && (c1
514 == extra[symb_table[2 * elem + 1]])
515 && memcmp (str,
516 &extra[symb_table[2 * elem
517 + 1]
518 + 1], c1) == 0)
520 /* Yep, this is the entry. */
521 idx = symb_table[2 * elem + 1];
522 idx += 1 + extra[idx];
523 break;
526 /* Next entry. */
527 elem += second;
529 while (symb_table[2 * elem] != 0);
532 if (symb_table[2 * elem] != 0)
534 /* Compare the byte sequence but only if
535 this is not part of a range. */
536 # ifdef WIDE_CHAR_VERSION
537 int32_t *wextra;
539 idx += 1 + extra[idx];
540 /* Adjust for the alignment. */
541 idx = (idx + 3) & ~3;
543 wextra = (int32_t *) &extra[idx + 4];
544 # endif
546 if (! is_range)
548 # ifdef WIDE_CHAR_VERSION
549 for (c1 = 0;
550 (int32_t) c1 < wextra[idx];
551 ++c1)
552 if (n[c1] != wextra[1 + c1])
553 break;
555 if ((int32_t) c1 == wextra[idx])
556 goto matched;
557 # else
558 for (c1 = 0; c1 < extra[idx]; ++c1)
559 if (n[c1] != extra[1 + c1])
560 break;
562 if (c1 == extra[idx])
563 goto matched;
564 # endif
567 /* Get the collation sequence value. */
568 is_seqval = 1;
569 # ifdef WIDE_CHAR_VERSION
570 cold = wextra[1 + wextra[idx]];
571 # else
572 /* Adjust for the alignment. */
573 idx += 1 + extra[idx];
574 idx = (idx + 3) & ~4;
575 cold = *((int32_t *) &extra[idx]);
576 # endif
578 c = *p++;
580 else if (c1 == 1)
582 /* No valid character. Match it as a
583 single byte. */
584 if (!is_range && *n == str[0])
585 goto matched;
587 cold = str[0];
588 c = *p++;
590 else
591 return FNM_NOMATCH;
594 else
595 # undef str
596 #endif
598 c = FOLD (c);
599 normal_bracket:
601 /* We have to handling the symbols differently in
602 ranges since then the collation sequence is
603 important. */
604 is_range = (*p == L('-') && p[1] != L('\0')
605 && p[1] != L(']'));
607 if (!is_range && c == fn)
608 goto matched;
610 /* This is needed if we goto normal_bracket; from
611 outside of is_seqval's scope. */
612 is_seqval = 0;
613 cold = c;
614 c = *p++;
617 if (c == L('-') && *p != L(']'))
619 #if _LIBC
620 /* We have to find the collation sequence
621 value for C. Collation sequence is nothing
622 we can regularly access. The sequence
623 value is defined by the order in which the
624 definitions of the collation values for the
625 various characters appear in the source
626 file. A strange concept, nowhere
627 documented. */
628 uint32_t fcollseq;
629 uint32_t lcollseq;
630 UCHAR cend = *p++;
632 # ifdef WIDE_CHAR_VERSION
633 /* Search in the `names' array for the characters. */
634 fcollseq = __collseq_table_lookup (collseq, fn);
635 if (fcollseq == ~((uint32_t) 0))
636 /* XXX We don't know anything about the character
637 we are supposed to match. This means we are
638 failing. */
639 goto range_not_matched;
641 if (is_seqval)
642 lcollseq = cold;
643 else
644 lcollseq = __collseq_table_lookup (collseq, cold);
645 # else
646 fcollseq = collseq[fn];
647 lcollseq = is_seqval ? cold : collseq[(UCHAR) cold];
648 # endif
650 is_seqval = 0;
651 if (cend == L('[') && *p == L('.'))
653 uint32_t nrules =
654 _NL_CURRENT_WORD (LC_COLLATE,
655 _NL_COLLATE_NRULES);
656 const CHAR *startp = p;
657 size_t c1 = 0;
659 while (1)
661 c = *++p;
662 if (c == L('.') && p[1] == L(']'))
664 p += 2;
665 break;
667 if (c == '\0')
668 return FNM_NOMATCH;
669 ++c1;
672 if (nrules == 0)
674 /* There are no names defined in the
675 collation data. Therefore we only
676 accept the trivial names consisting
677 of the character itself. */
678 if (c1 != 1)
679 return FNM_NOMATCH;
681 cend = startp[1];
683 else
685 int32_t table_size;
686 const int32_t *symb_table;
687 # ifdef WIDE_CHAR_VERSION
688 char str[c1];
689 unsigned int strcnt;
690 # else
691 # define str (startp + 1)
692 # endif
693 const unsigned char *extra;
694 int32_t idx;
695 int32_t elem;
696 int32_t second;
697 int32_t hash;
699 # ifdef WIDE_CHAR_VERSION
700 /* We have to convert the name to a single-byte
701 string. This is possible since the names
702 consist of ASCII characters and the internal
703 representation is UCS4. */
704 for (strcnt = 0; strcnt < c1; ++strcnt)
705 str[strcnt] = startp[1 + strcnt];
706 # endif
708 table_size =
709 _NL_CURRENT_WORD (LC_COLLATE,
710 _NL_COLLATE_SYMB_HASH_SIZEMB);
711 symb_table = (const int32_t *)
712 _NL_CURRENT (LC_COLLATE,
713 _NL_COLLATE_SYMB_TABLEMB);
714 extra = (const unsigned char *)
715 _NL_CURRENT (LC_COLLATE,
716 _NL_COLLATE_SYMB_EXTRAMB);
718 /* Locate the character in the hashing
719 table. */
720 hash = elem_hash (str, c1);
722 idx = 0;
723 elem = hash % table_size;
724 if (symb_table[2 * elem] != 0)
726 second = hash % (table_size - 2) + 1;
730 /* First compare the hashing value. */
731 if (symb_table[2 * elem] == hash
732 && (c1
733 == extra[symb_table[2 * elem + 1]])
734 && memcmp (str,
735 &extra[symb_table[2 * elem + 1]
736 + 1], c1) == 0)
738 /* Yep, this is the entry. */
739 idx = symb_table[2 * elem + 1];
740 idx += 1 + extra[idx];
741 break;
744 /* Next entry. */
745 elem += second;
747 while (symb_table[2 * elem] != 0);
750 if (symb_table[2 * elem] != 0)
752 /* Compare the byte sequence but only if
753 this is not part of a range. */
754 # ifdef WIDE_CHAR_VERSION
755 int32_t *wextra;
757 idx += 1 + extra[idx];
758 /* Adjust for the alignment. */
759 idx = (idx + 3) & ~4;
761 wextra = (int32_t *) &extra[idx + 4];
762 # endif
763 /* Get the collation sequence value. */
764 is_seqval = 1;
765 # ifdef WIDE_CHAR_VERSION
766 cend = wextra[1 + wextra[idx]];
767 # else
768 /* Adjust for the alignment. */
769 idx += 1 + extra[idx];
770 idx = (idx + 3) & ~4;
771 cend = *((int32_t *) &extra[idx]);
772 # endif
774 else if (symb_table[2 * elem] != 0 && c1 == 1)
776 cend = str[0];
777 c = *p++;
779 else
780 return FNM_NOMATCH;
782 # undef str
784 else
786 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
787 cend = *p++;
788 if (cend == L('\0'))
789 return FNM_NOMATCH;
790 cend = FOLD (cend);
793 /* XXX It is not entirely clear to me how to handle
794 characters which are not mentioned in the
795 collation specification. */
796 if (
797 # ifdef WIDE_CHAR_VERSION
798 lcollseq == 0xffffffff ||
799 # endif
800 lcollseq <= fcollseq)
802 /* We have to look at the upper bound. */
803 uint32_t hcollseq;
805 if (is_seqval)
806 hcollseq = cend;
807 else
809 # ifdef WIDE_CHAR_VERSION
810 hcollseq =
811 __collseq_table_lookup (collseq, cend);
812 if (hcollseq == ~((uint32_t) 0))
814 /* Hum, no information about the upper
815 bound. The matching succeeds if the
816 lower bound is matched exactly. */
817 if (lcollseq != fcollseq)
818 goto range_not_matched;
820 goto matched;
822 # else
823 hcollseq = collseq[cend];
824 # endif
827 if (lcollseq <= hcollseq && fcollseq <= hcollseq)
828 goto matched;
830 # ifdef WIDE_CHAR_VERSION
831 range_not_matched:
832 # endif
833 #else
834 /* We use a boring value comparison of the character
835 values. This is better than comparing using
836 `strcoll' since the latter would have surprising
837 and sometimes fatal consequences. */
838 UCHAR cend = *p++;
840 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
841 cend = *p++;
842 if (cend == L('\0'))
843 return FNM_NOMATCH;
845 /* It is a range. */
846 if (cold <= fn && fn <= cend)
847 goto matched;
848 #endif
850 c = *p++;
854 if (c == L(']'))
855 break;
858 if (!not)
859 return FNM_NOMATCH;
860 break;
862 matched:
863 /* Skip the rest of the [...] that already matched. */
866 ignore_next:
867 c = *p++;
869 if (c == L('\0'))
870 /* [... (unterminated) loses. */
871 return FNM_NOMATCH;
873 if (!(flags & FNM_NOESCAPE) && c == L('\\'))
875 if (*p == L('\0'))
876 return FNM_NOMATCH;
877 /* XXX 1003.2d11 is unclear if this is right. */
878 ++p;
880 else if (c == L('[') && *p == L(':'))
882 int c1 = 0;
883 const CHAR *startp = p;
885 while (1)
887 c = *++p;
888 if (++c1 == CHAR_CLASS_MAX_LENGTH)
889 return FNM_NOMATCH;
891 if (*p == L(':') && p[1] == L(']'))
892 break;
894 if (c < L('a') || c >= L('z'))
896 p = startp;
897 goto ignore_next;
900 p += 2;
901 c = *p++;
903 else if (c == L('[') && *p == L('='))
905 c = *++p;
906 if (c == L('\0'))
907 return FNM_NOMATCH;
908 c = *++p;
909 if (c != L('=') || p[1] != L(']'))
910 return FNM_NOMATCH;
911 p += 2;
912 c = *p++;
914 else if (c == L('[') && *p == L('.'))
916 ++p;
917 while (1)
919 c = *++p;
920 if (c == '\0')
921 return FNM_NOMATCH;
923 if (*p == L('.') && p[1] == L(']'))
924 break;
926 p += 2;
927 c = *p++;
930 while (c != L(']'));
931 if (not)
932 return FNM_NOMATCH;
934 break;
936 case L('+'):
937 case L('@'):
938 case L('!'):
939 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
941 int res;
943 res = EXT (c, p, n, string_end, no_leading_period, flags);
944 if (res != -1)
945 return res;
947 goto normal_match;
949 case L('/'):
950 if (NO_LEADING_PERIOD (flags))
952 if (n == string_end || c != (UCHAR) *n)
953 return FNM_NOMATCH;
955 new_no_leading_period = 1;
956 break;
958 /* FALLTHROUGH */
959 default:
960 normal_match:
961 if (n == string_end || c != FOLD ((UCHAR) *n))
962 return FNM_NOMATCH;
965 no_leading_period = new_no_leading_period;
966 ++n;
969 if (n == string_end)
970 return 0;
972 if ((flags & FNM_LEADING_DIR) && n != string_end && *n == L('/'))
973 /* The FNM_LEADING_DIR flag says that "foo*" matches "foobar/frobozz". */
974 return 0;
976 return FNM_NOMATCH;
980 static const CHAR *
981 internal_function
982 END (const CHAR *pattern)
984 const CHAR *p = pattern;
986 while (1)
987 if (*++p == L('\0'))
988 /* This is an invalid pattern. */
989 return pattern;
990 else if (*p == L('['))
992 /* Handle brackets special. */
993 if (posixly_correct == 0)
994 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
996 /* Skip the not sign. We have to recognize it because of a possibly
997 following ']'. */
998 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
999 ++p;
1000 /* A leading ']' is recognized as such. */
1001 if (*p == L(']'))
1002 ++p;
1003 /* Skip over all characters of the list. */
1004 while (*p != L(']'))
1005 if (*p++ == L('\0'))
1006 /* This is no valid pattern. */
1007 return pattern;
1009 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
1010 || *p == L('!')) && p[1] == L('('))
1011 p = END (p + 1);
1012 else if (*p == L(')'))
1013 break;
1015 return p + 1;
1019 static int
1020 internal_function
1021 EXT (INT opt, const CHAR *pattern, const CHAR *string, const CHAR *string_end,
1022 int no_leading_period, int flags)
1024 const CHAR *startp;
1025 int level;
1026 struct patternlist
1028 struct patternlist *next;
1029 CHAR str[0];
1030 } *list = NULL;
1031 struct patternlist **lastp = &list;
1032 size_t pattern_len = STRLEN (pattern);
1033 const CHAR *p;
1034 const CHAR *rs;
1036 /* Parse the pattern. Store the individual parts in the list. */
1037 level = 0;
1038 for (startp = p = pattern + 1; level >= 0; ++p)
1039 if (*p == L('\0'))
1040 /* This is an invalid pattern. */
1041 return -1;
1042 else if (*p == L('['))
1044 /* Handle brackets special. */
1045 if (posixly_correct == 0)
1046 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
1048 /* Skip the not sign. We have to recognize it because of a possibly
1049 following ']'. */
1050 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
1051 ++p;
1052 /* A leading ']' is recognized as such. */
1053 if (*p == L(']'))
1054 ++p;
1055 /* Skip over all characters of the list. */
1056 while (*p != L(']'))
1057 if (*p++ == L('\0'))
1058 /* This is no valid pattern. */
1059 return -1;
1061 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
1062 || *p == L('!')) && p[1] == L('('))
1063 /* Remember the nesting level. */
1064 ++level;
1065 else if (*p == L(')'))
1067 if (level-- == 0)
1069 /* This means we found the end of the pattern. */
1070 #define NEW_PATTERN \
1071 struct patternlist *newp; \
1073 if (opt == L('?') || opt == L('@')) \
1074 newp = alloca (sizeof (struct patternlist) \
1075 + (pattern_len * sizeof (CHAR))); \
1076 else \
1077 newp = alloca (sizeof (struct patternlist) \
1078 + ((p - startp + 1) * sizeof (CHAR))); \
1079 *((CHAR *) MEMPCPY (newp->str, startp, p - startp)) = L('\0'); \
1080 newp->next = NULL; \
1081 *lastp = newp; \
1082 lastp = &newp->next
1083 NEW_PATTERN;
1086 else if (*p == L('|'))
1088 if (level == 0)
1090 NEW_PATTERN;
1091 startp = p + 1;
1094 assert (list != NULL);
1095 assert (p[-1] == L(')'));
1096 #undef NEW_PATTERN
1098 switch (opt)
1100 case L('*'):
1101 if (FCT (p, string, string_end, no_leading_period, flags) == 0)
1102 return 0;
1103 /* FALLTHROUGH */
1105 case L('+'):
1108 for (rs = string; rs <= string_end; ++rs)
1109 /* First match the prefix with the current pattern with the
1110 current pattern. */
1111 if (FCT (list->str, string, rs, no_leading_period,
1112 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0
1113 /* This was successful. Now match the rest with the rest
1114 of the pattern. */
1115 && (FCT (p, rs, string_end,
1116 rs == string
1117 ? no_leading_period
1118 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1119 flags & FNM_FILE_NAME
1120 ? flags : flags & ~FNM_PERIOD) == 0
1121 /* This didn't work. Try the whole pattern. */
1122 || (rs != string
1123 && FCT (pattern - 1, rs, string_end,
1124 rs == string
1125 ? no_leading_period
1126 : (rs[-1] == '/' && NO_LEADING_PERIOD (flags)
1127 ? 1 : 0),
1128 flags & FNM_FILE_NAME
1129 ? flags : flags & ~FNM_PERIOD) == 0)))
1130 /* It worked. Signal success. */
1131 return 0;
1133 while ((list = list->next) != NULL);
1135 /* None of the patterns lead to a match. */
1136 return FNM_NOMATCH;
1138 case L('?'):
1139 if (FCT (p, string, string_end, no_leading_period, flags) == 0)
1140 return 0;
1141 /* FALLTHROUGH */
1143 case L('@'):
1145 /* I cannot believe it but `strcat' is actually acceptable
1146 here. Match the entire string with the prefix from the
1147 pattern list and the rest of the pattern following the
1148 pattern list. */
1149 if (FCT (STRCAT (list->str, p), string, string_end,
1150 no_leading_period,
1151 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0)
1152 /* It worked. Signal success. */
1153 return 0;
1154 while ((list = list->next) != NULL);
1156 /* None of the patterns lead to a match. */
1157 return FNM_NOMATCH;
1159 case L('!'):
1160 for (rs = string; rs <= string_end; ++rs)
1162 struct patternlist *runp;
1164 for (runp = list; runp != NULL; runp = runp->next)
1165 if (FCT (runp->str, string, rs, no_leading_period,
1166 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0)
1167 break;
1169 /* If none of the patterns matched see whether the rest does. */
1170 if (runp == NULL
1171 && (FCT (p, rs, string_end,
1172 rs == string
1173 ? no_leading_period
1174 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1175 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD)
1176 == 0))
1177 /* This is successful. */
1178 return 0;
1181 /* None of the patterns together with the rest of the pattern
1182 lead to a match. */
1183 return FNM_NOMATCH;
1185 default:
1186 assert (! "Invalid extended matching operator");
1187 break;
1190 return -1;
1194 #undef FOLD
1195 #undef CHAR
1196 #undef UCHAR
1197 #undef INT
1198 #undef FCT
1199 #undef EXT
1200 #undef END
1201 #undef MEMPCPY
1202 #undef MEMCHR
1203 #undef STRCOLL
1204 #undef STRLEN
1205 #undef STRCAT
1206 #undef L
1207 #undef BTOWC