* sysdeps/generic/strcpy_chk.c (__strcpy_chk): Speed up by checking
[glibc.git] / posix / fnmatch_loop.c
blobfb56bd426bc280f17b238b846c1c54d019e057c8
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 cold = c;
604 c = *p++;
607 if (c == L('-') && *p != L(']'))
609 #if _LIBC
610 /* We have to find the collation sequence
611 value for C. Collation sequence is nothing
612 we can regularly access. The sequence
613 value is defined by the order in which the
614 definitions of the collation values for the
615 various characters appear in the source
616 file. A strange concept, nowhere
617 documented. */
618 uint32_t fcollseq;
619 uint32_t lcollseq;
620 UCHAR cend = *p++;
622 # ifdef WIDE_CHAR_VERSION
623 /* Search in the `names' array for the characters. */
624 fcollseq = __collseq_table_lookup (collseq, fn);
625 if (fcollseq == ~((uint32_t) 0))
626 /* XXX We don't know anything about the character
627 we are supposed to match. This means we are
628 failing. */
629 goto range_not_matched;
631 if (is_seqval)
632 lcollseq = cold;
633 else
634 lcollseq = __collseq_table_lookup (collseq, cold);
635 # else
636 fcollseq = collseq[fn];
637 lcollseq = is_seqval ? cold : collseq[(UCHAR) cold];
638 # endif
640 is_seqval = 0;
641 if (cend == L('[') && *p == L('.'))
643 uint32_t nrules =
644 _NL_CURRENT_WORD (LC_COLLATE,
645 _NL_COLLATE_NRULES);
646 const CHAR *startp = p;
647 size_t c1 = 0;
649 while (1)
651 c = *++p;
652 if (c == L('.') && p[1] == L(']'))
654 p += 2;
655 break;
657 if (c == '\0')
658 return FNM_NOMATCH;
659 ++c1;
662 if (nrules == 0)
664 /* There are no names defined in the
665 collation data. Therefore we only
666 accept the trivial names consisting
667 of the character itself. */
668 if (c1 != 1)
669 return FNM_NOMATCH;
671 cend = startp[1];
673 else
675 int32_t table_size;
676 const int32_t *symb_table;
677 # ifdef WIDE_CHAR_VERSION
678 char str[c1];
679 unsigned int strcnt;
680 # else
681 # define str (startp + 1)
682 # endif
683 const unsigned char *extra;
684 int32_t idx;
685 int32_t elem;
686 int32_t second;
687 int32_t hash;
689 # ifdef WIDE_CHAR_VERSION
690 /* We have to convert the name to a single-byte
691 string. This is possible since the names
692 consist of ASCII characters and the internal
693 representation is UCS4. */
694 for (strcnt = 0; strcnt < c1; ++strcnt)
695 str[strcnt] = startp[1 + strcnt];
696 # endif
698 table_size =
699 _NL_CURRENT_WORD (LC_COLLATE,
700 _NL_COLLATE_SYMB_HASH_SIZEMB);
701 symb_table = (const int32_t *)
702 _NL_CURRENT (LC_COLLATE,
703 _NL_COLLATE_SYMB_TABLEMB);
704 extra = (const unsigned char *)
705 _NL_CURRENT (LC_COLLATE,
706 _NL_COLLATE_SYMB_EXTRAMB);
708 /* Locate the character in the hashing
709 table. */
710 hash = elem_hash (str, c1);
712 idx = 0;
713 elem = hash % table_size;
714 second = hash % (table_size - 2);
715 while (symb_table[2 * elem] != 0)
717 /* First compare the hashing value. */
718 if (symb_table[2 * elem] == hash
719 && (c1
720 == extra[symb_table[2 * elem + 1]])
721 && memcmp (str,
722 &extra[symb_table[2 * elem + 1]
723 + 1], c1) == 0)
725 /* Yep, this is the entry. */
726 idx = symb_table[2 * elem + 1];
727 idx += 1 + extra[idx];
728 break;
731 /* Next entry. */
732 elem += second;
735 if (symb_table[2 * elem] != 0)
737 /* Compare the byte sequence but only if
738 this is not part of a range. */
739 # ifdef WIDE_CHAR_VERSION
740 int32_t *wextra;
742 idx += 1 + extra[idx];
743 /* Adjust for the alignment. */
744 idx = (idx + 3) & ~4;
746 wextra = (int32_t *) &extra[idx + 4];
747 # endif
748 /* Get the collation sequence value. */
749 is_seqval = 1;
750 # ifdef WIDE_CHAR_VERSION
751 cend = wextra[1 + wextra[idx]];
752 # else
753 /* Adjust for the alignment. */
754 idx += 1 + extra[idx];
755 idx = (idx + 3) & ~4;
756 cend = *((int32_t *) &extra[idx]);
757 # endif
759 else if (symb_table[2 * elem] != 0 && c1 == 1)
761 cend = str[0];
762 c = *p++;
764 else
765 return FNM_NOMATCH;
767 # undef str
769 else
771 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
772 cend = *p++;
773 if (cend == L('\0'))
774 return FNM_NOMATCH;
775 cend = FOLD (cend);
778 /* XXX It is not entirely clear to me how to handle
779 characters which are not mentioned in the
780 collation specification. */
781 if (
782 # ifdef WIDE_CHAR_VERSION
783 lcollseq == 0xffffffff ||
784 # endif
785 lcollseq <= fcollseq)
787 /* We have to look at the upper bound. */
788 uint32_t hcollseq;
790 if (is_seqval)
791 hcollseq = cend;
792 else
794 # ifdef WIDE_CHAR_VERSION
795 hcollseq =
796 __collseq_table_lookup (collseq, cend);
797 if (hcollseq == ~((uint32_t) 0))
799 /* Hum, no information about the upper
800 bound. The matching succeeds if the
801 lower bound is matched exactly. */
802 if (lcollseq != fcollseq)
803 goto range_not_matched;
805 goto matched;
807 # else
808 hcollseq = collseq[cend];
809 # endif
812 if (lcollseq <= hcollseq && fcollseq <= hcollseq)
813 goto matched;
815 # ifdef WIDE_CHAR_VERSION
816 range_not_matched:
817 # endif
818 #else
819 /* We use a boring value comparison of the character
820 values. This is better than comparing using
821 `strcoll' since the latter would have surprising
822 and sometimes fatal consequences. */
823 UCHAR cend = *p++;
825 if (!(flags & FNM_NOESCAPE) && cend == L('\\'))
826 cend = *p++;
827 if (cend == L('\0'))
828 return FNM_NOMATCH;
830 /* It is a range. */
831 if (cold <= fn && fn <= cend)
832 goto matched;
833 #endif
835 c = *p++;
839 if (c == L(']'))
840 break;
843 if (!not)
844 return FNM_NOMATCH;
845 break;
847 matched:
848 /* Skip the rest of the [...] that already matched. */
851 ignore_next:
852 c = *p++;
854 if (c == L('\0'))
855 /* [... (unterminated) loses. */
856 return FNM_NOMATCH;
858 if (!(flags & FNM_NOESCAPE) && c == L('\\'))
860 if (*p == L('\0'))
861 return FNM_NOMATCH;
862 /* XXX 1003.2d11 is unclear if this is right. */
863 ++p;
865 else if (c == L('[') && *p == L(':'))
867 int c1 = 0;
868 const CHAR *startp = p;
870 while (1)
872 c = *++p;
873 if (++c1 == CHAR_CLASS_MAX_LENGTH)
874 return FNM_NOMATCH;
876 if (*p == L(':') && p[1] == L(']'))
877 break;
879 if (c < L('a') || c >= L('z'))
881 p = startp;
882 goto ignore_next;
885 p += 2;
886 c = *p++;
888 else if (c == L('[') && *p == L('='))
890 c = *++p;
891 if (c == L('\0'))
892 return FNM_NOMATCH;
893 c = *++p;
894 if (c != L('=') || p[1] != L(']'))
895 return FNM_NOMATCH;
896 p += 2;
897 c = *p++;
899 else if (c == L('[') && *p == L('.'))
901 ++p;
902 while (1)
904 c = *++p;
905 if (c == '\0')
906 return FNM_NOMATCH;
908 if (*p == L('.') && p[1] == L(']'))
909 break;
911 p += 2;
912 c = *p++;
915 while (c != L(']'));
916 if (not)
917 return FNM_NOMATCH;
919 break;
921 case L('+'):
922 case L('@'):
923 case L('!'):
924 if (__builtin_expect (flags & FNM_EXTMATCH, 0) && *p == '(')
926 int res;
928 res = EXT (c, p, n, string_end, no_leading_period, flags);
929 if (res != -1)
930 return res;
932 goto normal_match;
934 case L('/'):
935 if (NO_LEADING_PERIOD (flags))
937 if (n == string_end || c != (UCHAR) *n)
938 return FNM_NOMATCH;
940 new_no_leading_period = 1;
941 break;
943 /* FALLTHROUGH */
944 default:
945 normal_match:
946 if (n == string_end || c != FOLD ((UCHAR) *n))
947 return FNM_NOMATCH;
950 no_leading_period = new_no_leading_period;
951 ++n;
954 if (n == string_end)
955 return 0;
957 if ((flags & FNM_LEADING_DIR) && n != string_end && *n == L('/'))
958 /* The FNM_LEADING_DIR flag says that "foo*" matches "foobar/frobozz". */
959 return 0;
961 return FNM_NOMATCH;
965 static const CHAR *
966 internal_function
967 END (const CHAR *pattern)
969 const CHAR *p = pattern;
971 while (1)
972 if (*++p == L('\0'))
973 /* This is an invalid pattern. */
974 return pattern;
975 else if (*p == L('['))
977 /* Handle brackets special. */
978 if (posixly_correct == 0)
979 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
981 /* Skip the not sign. We have to recognize it because of a possibly
982 following ']'. */
983 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
984 ++p;
985 /* A leading ']' is recognized as such. */
986 if (*p == L(']'))
987 ++p;
988 /* Skip over all characters of the list. */
989 while (*p != L(']'))
990 if (*p++ == L('\0'))
991 /* This is no valid pattern. */
992 return pattern;
994 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
995 || *p == L('!')) && p[1] == L('('))
996 p = END (p + 1);
997 else if (*p == L(')'))
998 break;
1000 return p + 1;
1004 static int
1005 internal_function
1006 EXT (INT opt, const CHAR *pattern, const CHAR *string, const CHAR *string_end,
1007 int no_leading_period, int flags)
1009 const CHAR *startp;
1010 int level;
1011 struct patternlist
1013 struct patternlist *next;
1014 CHAR str[0];
1015 } *list = NULL;
1016 struct patternlist **lastp = &list;
1017 size_t pattern_len = STRLEN (pattern);
1018 const CHAR *p;
1019 const CHAR *rs;
1021 /* Parse the pattern. Store the individual parts in the list. */
1022 level = 0;
1023 for (startp = p = pattern + 1; level >= 0; ++p)
1024 if (*p == L('\0'))
1025 /* This is an invalid pattern. */
1026 return -1;
1027 else if (*p == L('['))
1029 /* Handle brackets special. */
1030 if (posixly_correct == 0)
1031 posixly_correct = getenv ("POSIXLY_CORRECT") != NULL ? 1 : -1;
1033 /* Skip the not sign. We have to recognize it because of a possibly
1034 following ']'. */
1035 if (*++p == L('!') || (posixly_correct < 0 && *p == L('^')))
1036 ++p;
1037 /* A leading ']' is recognized as such. */
1038 if (*p == L(']'))
1039 ++p;
1040 /* Skip over all characters of the list. */
1041 while (*p != L(']'))
1042 if (*p++ == L('\0'))
1043 /* This is no valid pattern. */
1044 return -1;
1046 else if ((*p == L('?') || *p == L('*') || *p == L('+') || *p == L('@')
1047 || *p == L('!')) && p[1] == L('('))
1048 /* Remember the nesting level. */
1049 ++level;
1050 else if (*p == L(')'))
1052 if (level-- == 0)
1054 /* This means we found the end of the pattern. */
1055 #define NEW_PATTERN \
1056 struct patternlist *newp; \
1058 if (opt == L('?') || opt == L('@')) \
1059 newp = alloca (sizeof (struct patternlist) \
1060 + (pattern_len * sizeof (CHAR))); \
1061 else \
1062 newp = alloca (sizeof (struct patternlist) \
1063 + ((p - startp + 1) * sizeof (CHAR))); \
1064 *((CHAR *) MEMPCPY (newp->str, startp, p - startp)) = L('\0'); \
1065 newp->next = NULL; \
1066 *lastp = newp; \
1067 lastp = &newp->next
1068 NEW_PATTERN;
1071 else if (*p == L('|'))
1073 if (level == 0)
1075 NEW_PATTERN;
1076 startp = p + 1;
1079 assert (list != NULL);
1080 assert (p[-1] == L(')'));
1081 #undef NEW_PATTERN
1083 switch (opt)
1085 case L('*'):
1086 if (FCT (p, string, string_end, no_leading_period, flags) == 0)
1087 return 0;
1088 /* FALLTHROUGH */
1090 case L('+'):
1093 for (rs = string; rs <= string_end; ++rs)
1094 /* First match the prefix with the current pattern with the
1095 current pattern. */
1096 if (FCT (list->str, string, rs, no_leading_period,
1097 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0
1098 /* This was successful. Now match the rest with the rest
1099 of the pattern. */
1100 && (FCT (p, rs, string_end,
1101 rs == string
1102 ? no_leading_period
1103 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1104 flags & FNM_FILE_NAME
1105 ? flags : flags & ~FNM_PERIOD) == 0
1106 /* This didn't work. Try the whole pattern. */
1107 || (rs != string
1108 && FCT (pattern - 1, rs, string_end,
1109 rs == string
1110 ? no_leading_period
1111 : (rs[-1] == '/' && NO_LEADING_PERIOD (flags)
1112 ? 1 : 0),
1113 flags & FNM_FILE_NAME
1114 ? flags : flags & ~FNM_PERIOD) == 0)))
1115 /* It worked. Signal success. */
1116 return 0;
1118 while ((list = list->next) != NULL);
1120 /* None of the patterns lead to a match. */
1121 return FNM_NOMATCH;
1123 case L('?'):
1124 if (FCT (p, string, string_end, no_leading_period, flags) == 0)
1125 return 0;
1126 /* FALLTHROUGH */
1128 case L('@'):
1130 /* I cannot believe it but `strcat' is actually acceptable
1131 here. Match the entire string with the prefix from the
1132 pattern list and the rest of the pattern following the
1133 pattern list. */
1134 if (FCT (STRCAT (list->str, p), string, string_end,
1135 no_leading_period,
1136 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0)
1137 /* It worked. Signal success. */
1138 return 0;
1139 while ((list = list->next) != NULL);
1141 /* None of the patterns lead to a match. */
1142 return FNM_NOMATCH;
1144 case L('!'):
1145 for (rs = string; rs <= string_end; ++rs)
1147 struct patternlist *runp;
1149 for (runp = list; runp != NULL; runp = runp->next)
1150 if (FCT (runp->str, string, rs, no_leading_period,
1151 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD) == 0)
1152 break;
1154 /* If none of the patterns matched see whether the rest does. */
1155 if (runp == NULL
1156 && (FCT (p, rs, string_end,
1157 rs == string
1158 ? no_leading_period
1159 : rs[-1] == '/' && NO_LEADING_PERIOD (flags) ? 1 : 0,
1160 flags & FNM_FILE_NAME ? flags : flags & ~FNM_PERIOD)
1161 == 0))
1162 /* This is successful. */
1163 return 0;
1166 /* None of the patterns together with the rest of the pattern
1167 lead to a match. */
1168 return FNM_NOMATCH;
1170 default:
1171 assert (! "Invalid extended matching operator");
1172 break;
1175 return -1;
1179 #undef FOLD
1180 #undef CHAR
1181 #undef UCHAR
1182 #undef INT
1183 #undef FCT
1184 #undef EXT
1185 #undef END
1186 #undef MEMPCPY
1187 #undef MEMCHR
1188 #undef STRCOLL
1189 #undef STRLEN
1190 #undef STRCAT
1191 #undef L
1192 #undef BTOWC