* locales/en_US: Add first_weekday and first_workday.
[glibc.git] / posix / fnmatch_loop.c
blob2bdd83718459217cf57a57aac6088cfbf9e459ea
1 /* Copyright (C) 1991,1992,1993,1996,1997,1998,1999,2000,2001,2003,2004,2005,
2 2007 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 struct STRUCT
22 const CHAR *pattern;
23 const CHAR *string;
24 int no_leading_period;
27 /* Match STRING against the filename pattern PATTERN, returning zero if
28 it matches, nonzero if not. */
29 static int FCT (const CHAR *pattern, const CHAR *string,
30 const CHAR *string_end, int no_leading_period, int flags,
31 struct STRUCT *ends)
32 internal_function;
33 static int EXT (INT opt, const CHAR *pattern, const CHAR *string,
34 const CHAR *string_end, int no_leading_period, int flags)
35 internal_function;
36 static const CHAR *END (const CHAR *patternp) internal_function;
38 static int
39 internal_function
40 FCT (pattern, string, string_end, no_leading_period, flags, ends)
41 const CHAR *pattern;
42 const CHAR *string;
43 const CHAR *string_end;
44 int no_leading_period;
45 int flags;
46 struct STRUCT *ends;
48 register const CHAR *p = pattern, *n = string;
49 register UCHAR c;
50 #ifdef _LIBC
51 # if WIDE_CHAR_VERSION
52 const char *collseq = (const char *)
53 _NL_CURRENT(LC_COLLATE, _NL_COLLATE_COLLSEQWC);
54 # else
55 const UCHAR *collseq = (const UCHAR *)
56 _NL_CURRENT(LC_COLLATE, _NL_COLLATE_COLLSEQMB);
57 # endif
58 #endif
60 while ((c = *p++) != L('\0'))
62 int new_no_leading_period = 0;
63 c = FOLD (c);
65 switch (c)
67 case L('?'):
68 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
70 int res;
72 res = EXT (c, p, n, string_end, no_leading_period,
73 flags);
74 if (res != -1)
75 return res;
78 if (n == string_end)
79 return FNM_NOMATCH;
80 else if (*n == L('/') && (flags & FNM_FILE_NAME))
81 return FNM_NOMATCH;
82 else if (*n == L('.') && no_leading_period)
83 return FNM_NOMATCH;
84 break;
86 case L('\\'):
87 if (!(flags & FNM_NOESCAPE))
89 c = *p++;
90 if (c == L('\0'))
91 /* Trailing \ loses. */
92 return FNM_NOMATCH;
93 c = FOLD (c);
95 if (n == string_end || FOLD ((UCHAR) *n) != c)
96 return FNM_NOMATCH;
97 break;
99 case L('*'):
100 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
102 int res;
104 res = EXT (c, p, n, string_end, no_leading_period,
105 flags);
106 if (res != -1)
107 return res;
109 else if (ends != NULL)
111 ends->pattern = p - 1;
112 ends->string = n;
113 ends->no_leading_period = no_leading_period;
114 return 0;
117 if (n != string_end && *n == L('.') && no_leading_period)
118 return FNM_NOMATCH;
120 for (c = *p++; c == L('?') || c == L('*'); c = *p++)
122 if (*p == L('(') && (flags & FNM_EXTMATCH) != 0)
124 const CHAR *endp = END (p);
125 if (endp != p)
127 /* This is a pattern. Skip over it. */
128 p = endp;
129 continue;
133 if (c == L('?'))
135 /* A ? needs to match one character. */
136 if (n == string_end)
137 /* There isn't another character; no match. */
138 return FNM_NOMATCH;
139 else if (*n == L('/')
140 && __builtin_expect (flags & FNM_FILE_NAME, 0))
141 /* A slash does not match a wildcard under
142 FNM_FILE_NAME. */
143 return FNM_NOMATCH;
144 else
145 /* One character of the string is consumed in matching
146 this ? wildcard, so *??? won't match if there are
147 less than three characters. */
148 ++n;
152 if (c == L('\0'))
153 /* The wildcard(s) is/are the last element of the pattern.
154 If the name is a file name and contains another slash
155 this means it cannot match, unless the FNM_LEADING_DIR
156 flag is set. */
158 int result = (flags & FNM_FILE_NAME) == 0 ? 0 : FNM_NOMATCH;
160 if (flags & FNM_FILE_NAME)
162 if (flags & FNM_LEADING_DIR)
163 result = 0;
164 else
166 if (MEMCHR (n, L('/'), string_end - n) == NULL)
167 result = 0;
171 return result;
173 else
175 const CHAR *endp;
176 struct STRUCT end;
178 end.pattern = NULL;
179 endp = MEMCHR (n, (flags & FNM_FILE_NAME) ? L('/') : L('\0'),
180 string_end - n);
181 if (endp == NULL)
182 endp = string_end;
184 if (c == L('[')
185 || (__builtin_expect (flags & FNM_EXTMATCH, 0) != 0
186 && (c == L('@') || c == L('+') || c == L('!'))
187 && *p == L('(')))
189 int flags2 = ((flags & FNM_FILE_NAME)
190 ? flags : (flags & ~FNM_PERIOD));
192 for (--p; n < endp; ++n, no_leading_period = 0)
193 if (FCT (p, n, string_end, no_leading_period, flags2,
194 &end) == 0)
195 goto found;
197 else if (c == L('/') && (flags & FNM_FILE_NAME))
199 while (n < string_end && *n != L('/'))
200 ++n;
201 if (n < string_end && *n == L('/')
202 && (FCT (p, n + 1, string_end, flags & FNM_PERIOD, flags,
203 NULL) == 0))
204 return 0;
206 else
208 int flags2 = ((flags & FNM_FILE_NAME)
209 ? flags : (flags & ~FNM_PERIOD));
211 if (c == L('\\') && !(flags & FNM_NOESCAPE))
212 c = *p;
213 c = FOLD (c);
214 for (--p; n < endp; ++n, no_leading_period = 0)
215 if (FOLD ((UCHAR) *n) == c
216 && (FCT (p, n, string_end, no_leading_period, flags2,
217 &end) == 0))
219 found:
220 if (end.pattern == NULL)
221 return 0;
222 break;
224 if (end.pattern != NULL)
226 p = end.pattern;
227 n = end.string;
228 no_leading_period = end.no_leading_period;
229 continue;
234 /* If we come here no match is possible with the wildcard. */
235 return FNM_NOMATCH;
237 case L('['):
239 /* Nonzero if the sense of the character class is inverted. */
240 register int not;
241 CHAR cold;
242 UCHAR fn;
244 if (posixly_correct == 0)
245 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
247 if (n == string_end)
248 return FNM_NOMATCH;
250 if (*n == L('.') && no_leading_period)
251 return FNM_NOMATCH;
253 if (*n == L('/') && (flags & FNM_FILE_NAME))
254 /* `/' cannot be matched. */
255 return FNM_NOMATCH;
257 not = (*p == L('!') || (posixly_correct < 0 && *p == L('^')));
258 if (not)
259 ++p;
261 fn = FOLD ((UCHAR) *n);
263 c = *p++;
264 for (;;)
266 if (!(flags & FNM_NOESCAPE) && c == L('\\'))
268 if (*p == L('\0'))
269 return FNM_NOMATCH;
270 c = FOLD ((UCHAR) *p);
271 ++p;
273 goto normal_bracket;
275 else if (c == L('[') && *p == L(':'))
277 /* Leave room for the null. */
278 CHAR str[CHAR_CLASS_MAX_LENGTH + 1];
279 size_t c1 = 0;
280 #if defined _LIBC || (defined HAVE_WCTYPE_H && defined HAVE_WCHAR_H)
281 wctype_t wt;
282 #endif
283 const CHAR *startp = p;
285 for (;;)
287 if (c1 == CHAR_CLASS_MAX_LENGTH)
288 /* The name is too long and therefore the pattern
289 is ill-formed. */
290 return FNM_NOMATCH;
292 c = *++p;
293 if (c == L(':') && p[1] == L(']'))
295 p += 2;
296 break;
298 if (c < L('a') || c >= L('z'))
300 /* This cannot possibly be a character class name.
301 Match it as a normal range. */
302 p = startp;
303 c = L('[');
304 goto normal_bracket;
306 str[c1++] = c;
308 str[c1] = L('\0');
310 #if defined _LIBC || (defined HAVE_WCTYPE_H && defined HAVE_WCHAR_H)
311 wt = IS_CHAR_CLASS (str);
312 if (wt == 0)
313 /* Invalid character class name. */
314 return FNM_NOMATCH;
316 # if defined _LIBC && ! WIDE_CHAR_VERSION
317 /* The following code is glibc specific but does
318 there a good job in speeding up the code since
319 we can avoid the btowc() call. */
320 if (_ISCTYPE ((UCHAR) *n, wt))
321 goto matched;
322 # else
323 if (ISWCTYPE (BTOWC ((UCHAR) *n), wt))
324 goto matched;
325 # endif
326 #else
327 if ((STREQ (str, L("alnum")) && ISALNUM ((UCHAR) *n))
328 || (STREQ (str, L("alpha")) && ISALPHA ((UCHAR) *n))
329 || (STREQ (str, L("blank")) && ISBLANK ((UCHAR) *n))
330 || (STREQ (str, L("cntrl")) && ISCNTRL ((UCHAR) *n))
331 || (STREQ (str, L("digit")) && ISDIGIT ((UCHAR) *n))
332 || (STREQ (str, L("graph")) && ISGRAPH ((UCHAR) *n))
333 || (STREQ (str, L("lower")) && ISLOWER ((UCHAR) *n))
334 || (STREQ (str, L("print")) && ISPRINT ((UCHAR) *n))
335 || (STREQ (str, L("punct")) && ISPUNCT ((UCHAR) *n))
336 || (STREQ (str, L("space")) && ISSPACE ((UCHAR) *n))
337 || (STREQ (str, L("upper")) && ISUPPER ((UCHAR) *n))
338 || (STREQ (str, L("xdigit")) && ISXDIGIT ((UCHAR) *n)))
339 goto matched;
340 #endif
341 c = *p++;
343 #ifdef _LIBC
344 else if (c == L('[') && *p == L('='))
346 UCHAR str[1];
347 uint32_t nrules =
348 _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
349 const CHAR *startp = p;
351 c = *++p;
352 if (c == L('\0'))
354 p = startp;
355 c = L('[');
356 goto normal_bracket;
358 str[0] = c;
360 c = *++p;
361 if (c != L('=') || p[1] != L(']'))
363 p = startp;
364 c = L('[');
365 goto normal_bracket;
367 p += 2;
369 if (nrules == 0)
371 if ((UCHAR) *n == str[0])
372 goto matched;
374 else
376 const int32_t *table;
377 # if WIDE_CHAR_VERSION
378 const int32_t *weights;
379 const int32_t *extra;
380 # else
381 const unsigned char *weights;
382 const unsigned char *extra;
383 # endif
384 const int32_t *indirect;
385 int32_t idx;
386 const UCHAR *cp = (const UCHAR *) str;
388 /* This #include defines a local function! */
389 # if WIDE_CHAR_VERSION
390 # include <locale/weightwc.h>
391 # else
392 # include <locale/weight.h>
393 # endif
395 # if WIDE_CHAR_VERSION
396 table = (const int32_t *)
397 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEWC);
398 weights = (const int32_t *)
399 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTWC);
400 extra = (const int32_t *)
401 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAWC);
402 indirect = (const int32_t *)
403 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTWC);
404 # else
405 table = (const int32_t *)
406 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB);
407 weights = (const unsigned char *)
408 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTMB);
409 extra = (const unsigned char *)
410 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB);
411 indirect = (const int32_t *)
412 _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTMB);
413 # endif
415 idx = findidx (&cp);
416 if (idx != 0)
418 /* We found a table entry. Now see whether the
419 character we are currently at has the same
420 equivalance class value. */
421 int len = weights[idx];
422 int32_t idx2;
423 const UCHAR *np = (const UCHAR *) n;
425 idx2 = findidx (&np);
426 if (idx2 != 0 && len == weights[idx2])
428 int cnt = 0;
430 while (cnt < len
431 && (weights[idx + 1 + cnt]
432 == weights[idx2 + 1 + cnt]))
433 ++cnt;
435 if (cnt == len)
436 goto matched;
441 c = *p++;
443 #endif
444 else if (c == L('\0'))
445 /* [ (unterminated) loses. */
446 return FNM_NOMATCH;
447 else
449 int is_range = 0;
451 #ifdef _LIBC
452 int is_seqval = 0;
454 if (c == L('[') && *p == L('.'))
456 uint32_t nrules =
457 _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
458 const CHAR *startp = p;
459 size_t c1 = 0;
461 while (1)
463 c = *++p;
464 if (c == L('.') && p[1] == L(']'))
466 p += 2;
467 break;
469 if (c == '\0')
470 return FNM_NOMATCH;
471 ++c1;
474 /* We have to handling the symbols differently in
475 ranges since then the collation sequence is
476 important. */
477 is_range = *p == L('-') && p[1] != L('\0');
479 if (nrules == 0)
481 /* There are no names defined in the collation
482 data. Therefore we only accept the trivial
483 names consisting of the character itself. */
484 if (c1 != 1)
485 return FNM_NOMATCH;
487 if (!is_range && *n == startp[1])
488 goto matched;
490 cold = startp[1];
491 c = *p++;
493 else
495 int32_t table_size;
496 const int32_t *symb_table;
497 # ifdef WIDE_CHAR_VERSION
498 char str[c1];
499 unsigned int strcnt;
500 # else
501 # define str (startp + 1)
502 # endif
503 const unsigned char *extra;
504 int32_t idx;
505 int32_t elem;
506 int32_t second;
507 int32_t hash;
509 # ifdef WIDE_CHAR_VERSION
510 /* We have to convert the name to a single-byte
511 string. This is possible since the names
512 consist of ASCII characters and the internal
513 representation is UCS4. */
514 for (strcnt = 0; strcnt < c1; ++strcnt)
515 str[strcnt] = startp[1 + strcnt];
516 #endif
518 table_size =
519 _NL_CURRENT_WORD (LC_COLLATE,
520 _NL_COLLATE_SYMB_HASH_SIZEMB);
521 symb_table = (const int32_t *)
522 _NL_CURRENT (LC_COLLATE,
523 _NL_COLLATE_SYMB_TABLEMB);
524 extra = (const unsigned char *)
525 _NL_CURRENT (LC_COLLATE,
526 _NL_COLLATE_SYMB_EXTRAMB);
528 /* Locate the character in the hashing table. */
529 hash = elem_hash (str, c1);
531 idx = 0;
532 elem = hash % table_size;
533 if (symb_table[2 * elem] != 0)
535 second = hash % (table_size - 2) + 1;
539 /* First compare the hashing value. */
540 if (symb_table[2 * elem] == hash
541 && (c1
542 == extra[symb_table[2 * elem + 1]])
543 && memcmp (str,
544 &extra[symb_table[2 * elem
545 + 1]
546 + 1], c1) == 0)
548 /* Yep, this is the entry. */
549 idx = symb_table[2 * elem + 1];
550 idx += 1 + extra[idx];
551 break;
554 /* Next entry. */
555 elem += second;
557 while (symb_table[2 * elem] != 0);
560 if (symb_table[2 * elem] != 0)
562 /* Compare the byte sequence but only if
563 this is not part of a range. */
564 # ifdef WIDE_CHAR_VERSION
565 int32_t *wextra;
567 idx += 1 + extra[idx];
568 /* Adjust for the alignment. */
569 idx = (idx + 3) & ~3;
571 wextra = (int32_t *) &extra[idx + 4];
572 # endif
574 if (! is_range)
576 # ifdef WIDE_CHAR_VERSION
577 for (c1 = 0;
578 (int32_t) c1 < wextra[idx];
579 ++c1)
580 if (n[c1] != wextra[1 + c1])
581 break;
583 if ((int32_t) c1 == wextra[idx])
584 goto matched;
585 # else
586 for (c1 = 0; c1 < extra[idx]; ++c1)
587 if (n[c1] != extra[1 + c1])
588 break;
590 if (c1 == extra[idx])
591 goto matched;
592 # endif
595 /* Get the collation sequence value. */
596 is_seqval = 1;
597 # ifdef WIDE_CHAR_VERSION
598 cold = wextra[1 + wextra[idx]];
599 # else
600 /* Adjust for the alignment. */
601 idx += 1 + extra[idx];
602 idx = (idx + 3) & ~4;
603 cold = *((int32_t *) &extra[idx]);
604 # endif
606 c = *p++;
608 else if (c1 == 1)
610 /* No valid character. Match it as a
611 single byte. */
612 if (!is_range && *n == str[0])
613 goto matched;
615 cold = str[0];
616 c = *p++;
618 else
619 return FNM_NOMATCH;
622 else
623 # undef str
624 #endif
626 c = FOLD (c);
627 normal_bracket:
629 /* We have to handling the symbols differently in
630 ranges since then the collation sequence is
631 important. */
632 is_range = (*p == L('-') && p[1] != L('\0')
633 && p[1] != L(']'));
635 if (!is_range && c == fn)
636 goto matched;
638 /* This is needed if we goto normal_bracket; from
639 outside of is_seqval's scope. */
640 is_seqval = 0;
641 cold = c;
642 c = *p++;
645 if (c == L('-') && *p != L(']'))
647 #if _LIBC
648 /* We have to find the collation sequence
649 value for C. Collation sequence is nothing
650 we can regularly access. The sequence
651 value is defined by the order in which the
652 definitions of the collation values for the
653 various characters appear in the source
654 file. A strange concept, nowhere
655 documented. */
656 uint32_t fcollseq;
657 uint32_t lcollseq;
658 UCHAR cend = *p++;
660 # ifdef WIDE_CHAR_VERSION
661 /* Search in the `names' array for the characters. */
662 fcollseq = __collseq_table_lookup (collseq, fn);
663 if (fcollseq == ~((uint32_t) 0))
664 /* XXX We don't know anything about the character
665 we are supposed to match. This means we are
666 failing. */
667 goto range_not_matched;
669 if (is_seqval)
670 lcollseq = cold;
671 else
672 lcollseq = __collseq_table_lookup (collseq, cold);
673 # else
674 fcollseq = collseq[fn];
675 lcollseq = is_seqval ? cold : collseq[(UCHAR) cold];
676 # endif
678 is_seqval = 0;
679 if (cend == L('[') && *p == L('.'))
681 uint32_t nrules =
682 _NL_CURRENT_WORD (LC_COLLATE,
683 _NL_COLLATE_NRULES);
684 const CHAR *startp = p;
685 size_t c1 = 0;
687 while (1)
689 c = *++p;
690 if (c == L('.') && p[1] == L(']'))
692 p += 2;
693 break;
695 if (c == '\0')
696 return FNM_NOMATCH;
697 ++c1;
700 if (nrules == 0)
702 /* There are no names defined in the
703 collation data. Therefore we only
704 accept the trivial names consisting
705 of the character itself. */
706 if (c1 != 1)
707 return FNM_NOMATCH;
709 cend = startp[1];
711 else
713 int32_t table_size;
714 const int32_t *symb_table;
715 # ifdef WIDE_CHAR_VERSION
716 char str[c1];
717 unsigned int strcnt;
718 # else
719 # define str (startp + 1)
720 # endif
721 const unsigned char *extra;
722 int32_t idx;
723 int32_t elem;
724 int32_t second;
725 int32_t hash;
727 # ifdef WIDE_CHAR_VERSION
728 /* We have to convert the name to a single-byte
729 string. This is possible since the names
730 consist of ASCII characters and the internal
731 representation is UCS4. */
732 for (strcnt = 0; strcnt < c1; ++strcnt)
733 str[strcnt] = startp[1 + strcnt];
734 # endif
736 table_size =
737 _NL_CURRENT_WORD (LC_COLLATE,
738 _NL_COLLATE_SYMB_HASH_SIZEMB);
739 symb_table = (const int32_t *)
740 _NL_CURRENT (LC_COLLATE,
741 _NL_COLLATE_SYMB_TABLEMB);
742 extra = (const unsigned char *)
743 _NL_CURRENT (LC_COLLATE,
744 _NL_COLLATE_SYMB_EXTRAMB);
746 /* Locate the character in the hashing
747 table. */
748 hash = elem_hash (str, c1);
750 idx = 0;
751 elem = hash % table_size;
752 if (symb_table[2 * elem] != 0)
754 second = hash % (table_size - 2) + 1;
758 /* First compare the hashing value. */
759 if (symb_table[2 * elem] == hash
760 && (c1
761 == extra[symb_table[2 * elem + 1]])
762 && memcmp (str,
763 &extra[symb_table[2 * elem + 1]
764 + 1], c1) == 0)
766 /* Yep, this is the entry. */
767 idx = symb_table[2 * elem + 1];
768 idx += 1 + extra[idx];
769 break;
772 /* Next entry. */
773 elem += second;
775 while (symb_table[2 * elem] != 0);
778 if (symb_table[2 * elem] != 0)
780 /* Compare the byte sequence but only if
781 this is not part of a range. */
782 # ifdef WIDE_CHAR_VERSION
783 int32_t *wextra;
785 idx += 1 + extra[idx];
786 /* Adjust for the alignment. */
787 idx = (idx + 3) & ~4;
789 wextra = (int32_t *) &extra[idx + 4];
790 # endif
791 /* Get the collation sequence value. */
792 is_seqval = 1;
793 # ifdef WIDE_CHAR_VERSION
794 cend = wextra[1 + wextra[idx]];
795 # else
796 /* Adjust for the alignment. */
797 idx += 1 + extra[idx];
798 idx = (idx + 3) & ~4;
799 cend = *((int32_t *) &extra[idx]);
800 # endif
802 else if (symb_table[2 * elem] != 0 && c1 == 1)
804 cend = str[0];
805 c = *p++;
807 else
808 return FNM_NOMATCH;
810 # undef str
812 else
814 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
815 cend = *p++;
816 if (cend == L('\0'))
817 return FNM_NOMATCH;
818 cend = FOLD (cend);
821 /* XXX It is not entirely clear to me how to handle
822 characters which are not mentioned in the
823 collation specification. */
824 if (
825 # ifdef WIDE_CHAR_VERSION
826 lcollseq == 0xffffffff ||
827 # endif
828 lcollseq <= fcollseq)
830 /* We have to look at the upper bound. */
831 uint32_t hcollseq;
833 if (is_seqval)
834 hcollseq = cend;
835 else
837 # ifdef WIDE_CHAR_VERSION
838 hcollseq =
839 __collseq_table_lookup (collseq, cend);
840 if (hcollseq == ~((uint32_t) 0))
842 /* Hum, no information about the upper
843 bound. The matching succeeds if the
844 lower bound is matched exactly. */
845 if (lcollseq != fcollseq)
846 goto range_not_matched;
848 goto matched;
850 # else
851 hcollseq = collseq[cend];
852 # endif
855 if (lcollseq <= hcollseq && fcollseq <= hcollseq)
856 goto matched;
858 # ifdef WIDE_CHAR_VERSION
859 range_not_matched:
860 # endif
861 #else
862 /* We use a boring value comparison of the character
863 values. This is better than comparing using
864 `strcoll' since the latter would have surprising
865 and sometimes fatal consequences. */
866 UCHAR cend = *p++;
868 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
869 cend = *p++;
870 if (cend == L('\0'))
871 return FNM_NOMATCH;
873 /* It is a range. */
874 if (cold <= fn && fn <= cend)
875 goto matched;
876 #endif
878 c = *p++;
882 if (c == L(']'))
883 break;
886 if (!not)
887 return FNM_NOMATCH;
888 break;
890 matched:
891 /* Skip the rest of the [...] that already matched. */
894 ignore_next:
895 c = *p++;
897 if (c == L('\0'))
898 /* [... (unterminated) loses. */
899 return FNM_NOMATCH;
901 if (!(flags & FNM_NOESCAPE) && c == L('\\'))
903 if (*p == L('\0'))
904 return FNM_NOMATCH;
905 /* XXX 1003.2d11 is unclear if this is right. */
906 ++p;
908 else if (c == L('[') && *p == L(':'))
910 int c1 = 0;
911 const CHAR *startp = p;
913 while (1)
915 c = *++p;
916 if (++c1 == CHAR_CLASS_MAX_LENGTH)
917 return FNM_NOMATCH;
919 if (*p == L(':') && p[1] == L(']'))
920 break;
922 if (c < L('a') || c >= L('z'))
924 p = startp;
925 goto ignore_next;
928 p += 2;
929 c = *p++;
931 else if (c == L('[') && *p == L('='))
933 c = *++p;
934 if (c == L('\0'))
935 return FNM_NOMATCH;
936 c = *++p;
937 if (c != L('=') || p[1] != L(']'))
938 return FNM_NOMATCH;
939 p += 2;
940 c = *p++;
942 else if (c == L('[') && *p == L('.'))
944 ++p;
945 while (1)
947 c = *++p;
948 if (c == '\0')
949 return FNM_NOMATCH;
951 if (*p == L('.') && p[1] == L(']'))
952 break;
954 p += 2;
955 c = *p++;
958 while (c != L(']'));
959 if (not)
960 return FNM_NOMATCH;
962 break;
964 case L('+'):
965 case L('@'):
966 case L('!'):
967 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
969 int res;
971 res = EXT (c, p, n, string_end, no_leading_period, flags);
972 if (res != -1)
973 return res;
975 goto normal_match;
977 case L('/'):
978 if (NO_LEADING_PERIOD (flags))
980 if (n == string_end || c != (UCHAR) *n)
981 return FNM_NOMATCH;
983 new_no_leading_period = 1;
984 break;
986 /* FALLTHROUGH */
987 default:
988 normal_match:
989 if (n == string_end || c != FOLD ((UCHAR) *n))
990 return FNM_NOMATCH;
993 no_leading_period = new_no_leading_period;
994 ++n;
997 if (n == string_end)
998 return 0;
1000 if ((flags & FNM_LEADING_DIR) && n != string_end && *n == L('/'))
1001 /* The FNM_LEADING_DIR flag says that "foo*" matches "foobar/frobozz". */
1002 return 0;
1004 return FNM_NOMATCH;
1008 static const CHAR *
1009 internal_function
1010 END (const CHAR *pattern)
1012 const CHAR *p = pattern;
1014 while (1)
1015 if (*++p == L('\0'))
1016 /* This is an invalid pattern. */
1017 return pattern;
1018 else if (*p == L('['))
1020 /* Handle brackets special. */
1021 if (posixly_correct == 0)
1022 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
1024 /* Skip the not sign. We have to recognize it because of a possibly
1025 following ']'. */
1026 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
1027 ++p;
1028 /* A leading ']' is recognized as such. */
1029 if (*p == L(']'))
1030 ++p;
1031 /* Skip over all characters of the list. */
1032 while (*p != L(']'))
1033 if (*p++ == L('\0'))
1034 /* This is no valid pattern. */
1035 return pattern;
1037 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
1038 || *p == L('!')) && p[1] == L('('))
1039 p = END (p + 1);
1040 else if (*p == L(')'))
1041 break;
1043 return p + 1;
1047 static int
1048 internal_function
1049 EXT (INT opt, const CHAR *pattern, const CHAR *string, const CHAR *string_end,
1050 int no_leading_period, int flags)
1052 const CHAR *startp;
1053 int level;
1054 struct patternlist
1056 struct patternlist *next;
1057 CHAR str[0];
1058 } *list = NULL;
1059 struct patternlist **lastp = &list;
1060 size_t pattern_len = STRLEN (pattern);
1061 const CHAR *p;
1062 const CHAR *rs;
1064 /* Parse the pattern. Store the individual parts in the list. */
1065 level = 0;
1066 for (startp = p = pattern + 1; level >= 0; ++p)
1067 if (*p == L('\0'))
1068 /* This is an invalid pattern. */
1069 return -1;
1070 else if (*p == L('['))
1072 /* Handle brackets special. */
1073 if (posixly_correct == 0)
1074 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
1076 /* Skip the not sign. We have to recognize it because of a possibly
1077 following ']'. */
1078 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
1079 ++p;
1080 /* A leading ']' is recognized as such. */
1081 if (*p == L(']'))
1082 ++p;
1083 /* Skip over all characters of the list. */
1084 while (*p != L(']'))
1085 if (*p++ == L('\0'))
1086 /* This is no valid pattern. */
1087 return -1;
1089 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
1090 || *p == L('!')) && p[1] == L('('))
1091 /* Remember the nesting level. */
1092 ++level;
1093 else if (*p == L(')'))
1095 if (level-- == 0)
1097 /* This means we found the end of the pattern. */
1098 #define NEW_PATTERN \
1099 struct patternlist *newp; \
1101 if (opt == L('?') || opt == L('@')) \
1102 newp = alloca (sizeof (struct patternlist) \
1103 + (pattern_len * sizeof (CHAR))); \
1104 else \
1105 newp = alloca (sizeof (struct patternlist) \
1106 + ((p - startp + 1) * sizeof (CHAR))); \
1107 *((CHAR *) MEMPCPY (newp->str, startp, p - startp)) = L('\0'); \
1108 newp->next = NULL; \
1109 *lastp = newp; \
1110 lastp = &newp->next
1111 NEW_PATTERN;
1114 else if (*p == L('|'))
1116 if (level == 0)
1118 NEW_PATTERN;
1119 startp = p + 1;
1122 assert (list != NULL);
1123 assert (p[-1] == L(')'));
1124 #undef NEW_PATTERN
1126 switch (opt)
1128 case L('*'):
1129 if (FCT (p, string, string_end, no_leading_period, flags, NULL) == 0)
1130 return 0;
1131 /* FALLTHROUGH */
1133 case L('+'):
1136 for (rs = string; rs <= string_end; ++rs)
1137 /* First match the prefix with the current pattern with the
1138 current pattern. */
1139 if (FCT (list->str, string, rs, no_leading_period,
1140 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD,
1141 NULL) == 0
1142 /* This was successful. Now match the rest with the rest
1143 of the pattern. */
1144 && (FCT (p, rs, string_end,
1145 rs == string
1146 ? no_leading_period
1147 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1148 flags & FNM_FILE_NAME
1149 ? flags : flags & ~FNM_PERIOD, NULL) == 0
1150 /* This didn't work. Try the whole pattern. */
1151 || (rs != string
1152 && FCT (pattern - 1, rs, string_end,
1153 rs == string
1154 ? no_leading_period
1155 : (rs[-1] == '/' && NO_LEADING_PERIOD (flags)
1156 ? 1 : 0),
1157 flags & FNM_FILE_NAME
1158 ? flags : flags & ~FNM_PERIOD, NULL) == 0)))
1159 /* It worked. Signal success. */
1160 return 0;
1162 while ((list = list->next) != NULL);
1164 /* None of the patterns lead to a match. */
1165 return FNM_NOMATCH;
1167 case L('?'):
1168 if (FCT (p, string, string_end, no_leading_period, flags, NULL) == 0)
1169 return 0;
1170 /* FALLTHROUGH */
1172 case L('@'):
1174 /* I cannot believe it but `strcat' is actually acceptable
1175 here. Match the entire string with the prefix from the
1176 pattern list and the rest of the pattern following the
1177 pattern list. */
1178 if (FCT (STRCAT (list->str, p), string, string_end,
1179 no_leading_period,
1180 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD,
1181 NULL) == 0)
1182 /* It worked. Signal success. */
1183 return 0;
1184 while ((list = list->next) != NULL);
1186 /* None of the patterns lead to a match. */
1187 return FNM_NOMATCH;
1189 case L('!'):
1190 for (rs = string; rs <= string_end; ++rs)
1192 struct patternlist *runp;
1194 for (runp = list; runp != NULL; runp = runp->next)
1195 if (FCT (runp->str, string, rs, no_leading_period,
1196 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD,
1197 NULL) == 0)
1198 break;
1200 /* If none of the patterns matched see whether the rest does. */
1201 if (runp == NULL
1202 && (FCT (p, rs, string_end,
1203 rs == string
1204 ? no_leading_period
1205 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1206 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD,
1207 NULL) == 0))
1208 /* This is successful. */
1209 return 0;
1212 /* None of the patterns together with the rest of the pattern
1213 lead to a match. */
1214 return FNM_NOMATCH;
1216 default:
1217 assert (! "Invalid extended matching operator");
1218 break;
1221 return -1;
1225 #undef FOLD
1226 #undef CHAR
1227 #undef UCHAR
1228 #undef INT
1229 #undef FCT
1230 #undef EXT
1231 #undef END
1232 #undef STRUCT
1233 #undef MEMPCPY
1234 #undef MEMCHR
1235 #undef STRCOLL
1236 #undef STRLEN
1237 #undef STRCAT
1238 #undef L
1239 #undef BTOWC