2.3.90-12
[glibc.git] / posix / fnmatch_loop.c
blob449e5e02b49000a756f0efd0e078bf198bdc6709
1 /* Copyright (C) 1991,1992,1993,1996,1997,1998,1999,2000,2001,2003,2004
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 second = hash % (table_size - 2);
506 while (symb_table[2 * elem] != 0)
508 /* First compare the hashing value. */
509 if (symb_table[2 * elem] == hash
510 && c1 == extra[symb_table[2 * elem + 1]]
511 && memcmp (str,
512 &extra[symb_table[2 * elem + 1]
513 + 1], c1) == 0)
515 /* Yep, this is the entry. */
516 idx = symb_table[2 * elem + 1];
517 idx += 1 + extra[idx];
518 break;
521 /* Next entry. */
522 elem += second;
525 if (symb_table[2 * elem] != 0)
527 /* Compare the byte sequence but only if
528 this is not part of a range. */
529 # ifdef WIDE_CHAR_VERSION
530 int32_t *wextra;
532 idx += 1 + extra[idx];
533 /* Adjust for the alignment. */
534 idx = (idx + 3) & ~3;
536 wextra = (int32_t *) &extra[idx + 4];
537 # endif
539 if (! is_range)
541 # ifdef WIDE_CHAR_VERSION
542 for (c1 = 0;
543 (int32_t) c1 < wextra[idx];
544 ++c1)
545 if (n[c1] != wextra[1 + c1])
546 break;
548 if ((int32_t) c1 == wextra[idx])
549 goto matched;
550 # else
551 for (c1 = 0; c1 < extra[idx]; ++c1)
552 if (n[c1] != extra[1 + c1])
553 break;
555 if (c1 == extra[idx])
556 goto matched;
557 # endif
560 /* Get the collation sequence value. */
561 is_seqval = 1;
562 # ifdef WIDE_CHAR_VERSION
563 cold = wextra[1 + wextra[idx]];
564 # else
565 /* Adjust for the alignment. */
566 idx += 1 + extra[idx];
567 idx = (idx + 3) & ~4;
568 cold = *((int32_t *) &extra[idx]);
569 # endif
571 c = *p++;
573 else if (c1 == 1)
575 /* No valid character. Match it as a
576 single byte. */
577 if (!is_range && *n == str[0])
578 goto matched;
580 cold = str[0];
581 c = *p++;
583 else
584 return FNM_NOMATCH;
587 else
588 # undef str
589 #endif
591 c = FOLD (c);
592 normal_bracket:
594 /* We have to handling the symbols differently in
595 ranges since then the collation sequence is
596 important. */
597 is_range = (*p == L('-') && p[1] != L('\0')
598 && p[1] != L(']'));
600 if (!is_range && c == fn)
601 goto matched;
603 /* This is needed if we goto normal_bracket; from
604 outside of is_seqval's scope. */
605 is_seqval = 0;
606 cold = c;
607 c = *p++;
610 if (c == L('-') && *p != L(']'))
612 #if _LIBC
613 /* We have to find the collation sequence
614 value for C. Collation sequence is nothing
615 we can regularly access. The sequence
616 value is defined by the order in which the
617 definitions of the collation values for the
618 various characters appear in the source
619 file. A strange concept, nowhere
620 documented. */
621 uint32_t fcollseq;
622 uint32_t lcollseq;
623 UCHAR cend = *p++;
625 # ifdef WIDE_CHAR_VERSION
626 /* Search in the `names' array for the characters. */
627 fcollseq = __collseq_table_lookup (collseq, fn);
628 if (fcollseq == ~((uint32_t) 0))
629 /* XXX We don't know anything about the character
630 we are supposed to match. This means we are
631 failing. */
632 goto range_not_matched;
634 if (is_seqval)
635 lcollseq = cold;
636 else
637 lcollseq = __collseq_table_lookup (collseq, cold);
638 # else
639 fcollseq = collseq[fn];
640 lcollseq = is_seqval ? cold : collseq[(UCHAR) cold];
641 # endif
643 is_seqval = 0;
644 if (cend == L('[') && *p == L('.'))
646 uint32_t nrules =
647 _NL_CURRENT_WORD (LC_COLLATE,
648 _NL_COLLATE_NRULES);
649 const CHAR *startp = p;
650 size_t c1 = 0;
652 while (1)
654 c = *++p;
655 if (c == L('.') && p[1] == L(']'))
657 p += 2;
658 break;
660 if (c == '\0')
661 return FNM_NOMATCH;
662 ++c1;
665 if (nrules == 0)
667 /* There are no names defined in the
668 collation data. Therefore we only
669 accept the trivial names consisting
670 of the character itself. */
671 if (c1 != 1)
672 return FNM_NOMATCH;
674 cend = startp[1];
676 else
678 int32_t table_size;
679 const int32_t *symb_table;
680 # ifdef WIDE_CHAR_VERSION
681 char str[c1];
682 unsigned int strcnt;
683 # else
684 # define str (startp + 1)
685 # endif
686 const unsigned char *extra;
687 int32_t idx;
688 int32_t elem;
689 int32_t second;
690 int32_t hash;
692 # ifdef WIDE_CHAR_VERSION
693 /* We have to convert the name to a single-byte
694 string. This is possible since the names
695 consist of ASCII characters and the internal
696 representation is UCS4. */
697 for (strcnt = 0; strcnt < c1; ++strcnt)
698 str[strcnt] = startp[1 + strcnt];
699 # endif
701 table_size =
702 _NL_CURRENT_WORD (LC_COLLATE,
703 _NL_COLLATE_SYMB_HASH_SIZEMB);
704 symb_table = (const int32_t *)
705 _NL_CURRENT (LC_COLLATE,
706 _NL_COLLATE_SYMB_TABLEMB);
707 extra = (const unsigned char *)
708 _NL_CURRENT (LC_COLLATE,
709 _NL_COLLATE_SYMB_EXTRAMB);
711 /* Locate the character in the hashing
712 table. */
713 hash = elem_hash (str, c1);
715 idx = 0;
716 elem = hash % table_size;
717 second = hash % (table_size - 2);
718 while (symb_table[2 * elem] != 0)
720 /* First compare the hashing value. */
721 if (symb_table[2 * elem] == hash
722 && (c1
723 == extra[symb_table[2 * elem + 1]])
724 && memcmp (str,
725 &extra[symb_table[2 * elem + 1]
726 + 1], c1) == 0)
728 /* Yep, this is the entry. */
729 idx = symb_table[2 * elem + 1];
730 idx += 1 + extra[idx];
731 break;
734 /* Next entry. */
735 elem += second;
738 if (symb_table[2 * elem] != 0)
740 /* Compare the byte sequence but only if
741 this is not part of a range. */
742 # ifdef WIDE_CHAR_VERSION
743 int32_t *wextra;
745 idx += 1 + extra[idx];
746 /* Adjust for the alignment. */
747 idx = (idx + 3) & ~4;
749 wextra = (int32_t *) &extra[idx + 4];
750 # endif
751 /* Get the collation sequence value. */
752 is_seqval = 1;
753 # ifdef WIDE_CHAR_VERSION
754 cend = wextra[1 + wextra[idx]];
755 # else
756 /* Adjust for the alignment. */
757 idx += 1 + extra[idx];
758 idx = (idx + 3) & ~4;
759 cend = *((int32_t *) &extra[idx]);
760 # endif
762 else if (symb_table[2 * elem] != 0 && c1 == 1)
764 cend = str[0];
765 c = *p++;
767 else
768 return FNM_NOMATCH;
770 # undef str
772 else
774 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
775 cend = *p++;
776 if (cend == L('\0'))
777 return FNM_NOMATCH;
778 cend = FOLD (cend);
781 /* XXX It is not entirely clear to me how to handle
782 characters which are not mentioned in the
783 collation specification. */
784 if (
785 # ifdef WIDE_CHAR_VERSION
786 lcollseq == 0xffffffff ||
787 # endif
788 lcollseq <= fcollseq)
790 /* We have to look at the upper bound. */
791 uint32_t hcollseq;
793 if (is_seqval)
794 hcollseq = cend;
795 else
797 # ifdef WIDE_CHAR_VERSION
798 hcollseq =
799 __collseq_table_lookup (collseq, cend);
800 if (hcollseq == ~((uint32_t) 0))
802 /* Hum, no information about the upper
803 bound. The matching succeeds if the
804 lower bound is matched exactly. */
805 if (lcollseq != fcollseq)
806 goto range_not_matched;
808 goto matched;
810 # else
811 hcollseq = collseq[cend];
812 # endif
815 if (lcollseq <= hcollseq && fcollseq <= hcollseq)
816 goto matched;
818 # ifdef WIDE_CHAR_VERSION
819 range_not_matched:
820 # endif
821 #else
822 /* We use a boring value comparison of the character
823 values. This is better than comparing using
824 `strcoll' since the latter would have surprising
825 and sometimes fatal consequences. */
826 UCHAR cend = *p++;
828 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
829 cend = *p++;
830 if (cend == L('\0'))
831 return FNM_NOMATCH;
833 /* It is a range. */
834 if (cold <= fn && fn <= cend)
835 goto matched;
836 #endif
838 c = *p++;
842 if (c == L(']'))
843 break;
846 if (!not)
847 return FNM_NOMATCH;
848 break;
850 matched:
851 /* Skip the rest of the [...] that already matched. */
854 ignore_next:
855 c = *p++;
857 if (c == L('\0'))
858 /* [... (unterminated) loses. */
859 return FNM_NOMATCH;
861 if (!(flags & FNM_NOESCAPE) && c == L('\\'))
863 if (*p == L('\0'))
864 return FNM_NOMATCH;
865 /* XXX 1003.2d11 is unclear if this is right. */
866 ++p;
868 else if (c == L('[') && *p == L(':'))
870 int c1 = 0;
871 const CHAR *startp = p;
873 while (1)
875 c = *++p;
876 if (++c1 == CHAR_CLASS_MAX_LENGTH)
877 return FNM_NOMATCH;
879 if (*p == L(':') && p[1] == L(']'))
880 break;
882 if (c < L('a') || c >= L('z'))
884 p = startp;
885 goto ignore_next;
888 p += 2;
889 c = *p++;
891 else if (c == L('[') && *p == L('='))
893 c = *++p;
894 if (c == L('\0'))
895 return FNM_NOMATCH;
896 c = *++p;
897 if (c != L('=') || p[1] != L(']'))
898 return FNM_NOMATCH;
899 p += 2;
900 c = *p++;
902 else if (c == L('[') && *p == L('.'))
904 ++p;
905 while (1)
907 c = *++p;
908 if (c == '\0')
909 return FNM_NOMATCH;
911 if (*p == L('.') && p[1] == L(']'))
912 break;
914 p += 2;
915 c = *p++;
918 while (c != L(']'));
919 if (not)
920 return FNM_NOMATCH;
922 break;
924 case L('+'):
925 case L('@'):
926 case L('!'):
927 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
929 int res;
931 res = EXT (c, p, n, string_end, no_leading_period, flags);
932 if (res != -1)
933 return res;
935 goto normal_match;
937 case L('/'):
938 if (NO_LEADING_PERIOD (flags))
940 if (n == string_end || c != (UCHAR) *n)
941 return FNM_NOMATCH;
943 new_no_leading_period = 1;
944 break;
946 /* FALLTHROUGH */
947 default:
948 normal_match:
949 if (n == string_end || c != FOLD ((UCHAR) *n))
950 return FNM_NOMATCH;
953 no_leading_period = new_no_leading_period;
954 ++n;
957 if (n == string_end)
958 return 0;
960 if ((flags & FNM_LEADING_DIR) && n != string_end && *n == L('/'))
961 /* The FNM_LEADING_DIR flag says that "foo*" matches "foobar/frobozz". */
962 return 0;
964 return FNM_NOMATCH;
968 static const CHAR *
969 internal_function
970 END (const CHAR *pattern)
972 const CHAR *p = pattern;
974 while (1)
975 if (*++p == L('\0'))
976 /* This is an invalid pattern. */
977 return pattern;
978 else if (*p == L('['))
980 /* Handle brackets special. */
981 if (posixly_correct == 0)
982 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
984 /* Skip the not sign. We have to recognize it because of a possibly
985 following ']'. */
986 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
987 ++p;
988 /* A leading ']' is recognized as such. */
989 if (*p == L(']'))
990 ++p;
991 /* Skip over all characters of the list. */
992 while (*p != L(']'))
993 if (*p++ == L('\0'))
994 /* This is no valid pattern. */
995 return pattern;
997 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
998 || *p == L('!')) && p[1] == L('('))
999 p = END (p + 1);
1000 else if (*p == L(')'))
1001 break;
1003 return p + 1;
1007 static int
1008 internal_function
1009 EXT (INT opt, const CHAR *pattern, const CHAR *string, const CHAR *string_end,
1010 int no_leading_period, int flags)
1012 const CHAR *startp;
1013 int level;
1014 struct patternlist
1016 struct patternlist *next;
1017 CHAR str[0];
1018 } *list = NULL;
1019 struct patternlist **lastp = &list;
1020 size_t pattern_len = STRLEN (pattern);
1021 const CHAR *p;
1022 const CHAR *rs;
1024 /* Parse the pattern. Store the individual parts in the list. */
1025 level = 0;
1026 for (startp = p = pattern + 1; level >= 0; ++p)
1027 if (*p == L('\0'))
1028 /* This is an invalid pattern. */
1029 return -1;
1030 else if (*p == L('['))
1032 /* Handle brackets special. */
1033 if (posixly_correct == 0)
1034 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
1036 /* Skip the not sign. We have to recognize it because of a possibly
1037 following ']'. */
1038 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
1039 ++p;
1040 /* A leading ']' is recognized as such. */
1041 if (*p == L(']'))
1042 ++p;
1043 /* Skip over all characters of the list. */
1044 while (*p != L(']'))
1045 if (*p++ == L('\0'))
1046 /* This is no valid pattern. */
1047 return -1;
1049 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
1050 || *p == L('!')) && p[1] == L('('))
1051 /* Remember the nesting level. */
1052 ++level;
1053 else if (*p == L(')'))
1055 if (level-- == 0)
1057 /* This means we found the end of the pattern. */
1058 #define NEW_PATTERN \
1059 struct patternlist *newp; \
1061 if (opt == L('?') || opt == L('@')) \
1062 newp = alloca (sizeof (struct patternlist) \
1063 + (pattern_len * sizeof (CHAR))); \
1064 else \
1065 newp = alloca (sizeof (struct patternlist) \
1066 + ((p - startp + 1) * sizeof (CHAR))); \
1067 *((CHAR *) MEMPCPY (newp->str, startp, p - startp)) = L('\0'); \
1068 newp->next = NULL; \
1069 *lastp = newp; \
1070 lastp = &newp->next
1071 NEW_PATTERN;
1074 else if (*p == L('|'))
1076 if (level == 0)
1078 NEW_PATTERN;
1079 startp = p + 1;
1082 assert (list != NULL);
1083 assert (p[-1] == L(')'));
1084 #undef NEW_PATTERN
1086 switch (opt)
1088 case L('*'):
1089 if (FCT (p, string, string_end, no_leading_period, flags) == 0)
1090 return 0;
1091 /* FALLTHROUGH */
1093 case L('+'):
1096 for (rs = string; rs <= string_end; ++rs)
1097 /* First match the prefix with the current pattern with the
1098 current pattern. */
1099 if (FCT (list->str, string, rs, no_leading_period,
1100 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0
1101 /* This was successful. Now match the rest with the rest
1102 of the pattern. */
1103 && (FCT (p, rs, string_end,
1104 rs == string
1105 ? no_leading_period
1106 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1107 flags & FNM_FILE_NAME
1108 ? flags : flags & ~FNM_PERIOD) == 0
1109 /* This didn't work. Try the whole pattern. */
1110 || (rs != string
1111 && FCT (pattern - 1, rs, string_end,
1112 rs == string
1113 ? no_leading_period
1114 : (rs[-1] == '/' && NO_LEADING_PERIOD (flags)
1115 ? 1 : 0),
1116 flags & FNM_FILE_NAME
1117 ? flags : flags & ~FNM_PERIOD) == 0)))
1118 /* It worked. Signal success. */
1119 return 0;
1121 while ((list = list->next) != NULL);
1123 /* None of the patterns lead to a match. */
1124 return FNM_NOMATCH;
1126 case L('?'):
1127 if (FCT (p, string, string_end, no_leading_period, flags) == 0)
1128 return 0;
1129 /* FALLTHROUGH */
1131 case L('@'):
1133 /* I cannot believe it but `strcat' is actually acceptable
1134 here. Match the entire string with the prefix from the
1135 pattern list and the rest of the pattern following the
1136 pattern list. */
1137 if (FCT (STRCAT (list->str, p), string, string_end,
1138 no_leading_period,
1139 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0)
1140 /* It worked. Signal success. */
1141 return 0;
1142 while ((list = list->next) != NULL);
1144 /* None of the patterns lead to a match. */
1145 return FNM_NOMATCH;
1147 case L('!'):
1148 for (rs = string; rs <= string_end; ++rs)
1150 struct patternlist *runp;
1152 for (runp = list; runp != NULL; runp = runp->next)
1153 if (FCT (runp->str, string, rs, no_leading_period,
1154 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0)
1155 break;
1157 /* If none of the patterns matched see whether the rest does. */
1158 if (runp == NULL
1159 && (FCT (p, rs, string_end,
1160 rs == string
1161 ? no_leading_period
1162 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1163 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD)
1164 == 0))
1165 /* This is successful. */
1166 return 0;
1169 /* None of the patterns together with the rest of the pattern
1170 lead to a match. */
1171 return FNM_NOMATCH;
1173 default:
1174 assert (! "Invalid extended matching operator");
1175 break;
1178 return -1;
1182 #undef FOLD
1183 #undef CHAR
1184 #undef UCHAR
1185 #undef INT
1186 #undef FCT
1187 #undef EXT
1188 #undef END
1189 #undef MEMPCPY
1190 #undef MEMCHR
1191 #undef STRCOLL
1192 #undef STRLEN
1193 #undef STRCAT
1194 #undef L
1195 #undef BTOWC