merge emacs-23
[emacs.git] / src / regex.c
blob85ff9b99604465bd4a28d8697508db212eae9b2e
1 /* Extended regular expression matching and search library, version
2 0.12. (Implements POSIX draft P1003.2/D11.2, except for some of the
3 internationalization features.)
5 Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001,
6 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
7 Free Software Foundation, Inc.
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 3, or (at your option)
12 any later version.
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program; if not, write to the Free Software
21 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301,
22 USA. */
24 /* TODO:
25 - structure the opcode space into opcode+flag.
26 - merge with glibc's regex.[ch].
27 - replace (succeed_n + jump_n + set_number_at) with something that doesn't
28 need to modify the compiled regexp so that re_match can be reentrant.
29 - get rid of on_failure_jump_smart by doing the optimization in re_comp
30 rather than at run-time, so that re_match can be reentrant.
33 /* AIX requires this to be the first thing in the file. */
34 #if defined _AIX && !defined REGEX_MALLOC
35 #pragma alloca
36 #endif
38 #ifdef HAVE_CONFIG_H
39 # include <config.h>
40 #endif
42 #if defined STDC_HEADERS && !defined emacs
43 # include <stddef.h>
44 #else
45 /* We need this for `regex.h', and perhaps for the Emacs include files. */
46 # include <sys/types.h>
47 #endif
49 /* Whether to use ISO C Amendment 1 wide char functions.
50 Those should not be used for Emacs since it uses its own. */
51 #if defined _LIBC
52 #define WIDE_CHAR_SUPPORT 1
53 #else
54 #define WIDE_CHAR_SUPPORT \
55 (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC && !emacs)
56 #endif
58 /* For platform which support the ISO C amendement 1 functionality we
59 support user defined character classes. */
60 #if WIDE_CHAR_SUPPORT
61 /* Solaris 2.5 has a bug: <wchar.h> must be included before <wctype.h>. */
62 # include <wchar.h>
63 # include <wctype.h>
64 #endif
66 #ifdef _LIBC
67 /* We have to keep the namespace clean. */
68 # define regfree(preg) __regfree (preg)
69 # define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef)
70 # define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags)
71 # define regerror(err_code, preg, errbuf, errbuf_size) \
72 __regerror(err_code, preg, errbuf, errbuf_size)
73 # define re_set_registers(bu, re, nu, st, en) \
74 __re_set_registers (bu, re, nu, st, en)
75 # define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \
76 __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
77 # define re_match(bufp, string, size, pos, regs) \
78 __re_match (bufp, string, size, pos, regs)
79 # define re_search(bufp, string, size, startpos, range, regs) \
80 __re_search (bufp, string, size, startpos, range, regs)
81 # define re_compile_pattern(pattern, length, bufp) \
82 __re_compile_pattern (pattern, length, bufp)
83 # define re_set_syntax(syntax) __re_set_syntax (syntax)
84 # define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \
85 __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop)
86 # define re_compile_fastmap(bufp) __re_compile_fastmap (bufp)
88 /* Make sure we call libc's function even if the user overrides them. */
89 # define btowc __btowc
90 # define iswctype __iswctype
91 # define wctype __wctype
93 # define WEAK_ALIAS(a,b) weak_alias (a, b)
95 /* We are also using some library internals. */
96 # include <locale/localeinfo.h>
97 # include <locale/elem-hash.h>
98 # include <langinfo.h>
99 #else
100 # define WEAK_ALIAS(a,b)
101 #endif
103 /* This is for other GNU distributions with internationalized messages. */
104 #if HAVE_LIBINTL_H || defined _LIBC
105 # include <libintl.h>
106 #else
107 # define gettext(msgid) (msgid)
108 #endif
110 #ifndef gettext_noop
111 /* This define is so xgettext can find the internationalizable
112 strings. */
113 # define gettext_noop(String) String
114 #endif
116 /* The `emacs' switch turns on certain matching commands
117 that make sense only in Emacs. */
118 #ifdef emacs
120 # include <setjmp.h>
121 # include "lisp.h"
122 # include "buffer.h"
124 /* Make syntax table lookup grant data in gl_state. */
125 # define SYNTAX_ENTRY_VIA_PROPERTY
127 # include "syntax.h"
128 # include "character.h"
129 # include "category.h"
131 # ifdef malloc
132 # undef malloc
133 # endif
134 # define malloc xmalloc
135 # ifdef realloc
136 # undef realloc
137 # endif
138 # define realloc xrealloc
139 # ifdef free
140 # undef free
141 # endif
142 # define free xfree
144 /* Converts the pointer to the char to BEG-based offset from the start. */
145 # define PTR_TO_OFFSET(d) POS_AS_IN_BUFFER (POINTER_TO_OFFSET (d))
146 # define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object)))
148 # define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte)
149 # define RE_TARGET_MULTIBYTE_P(bufp) ((bufp)->target_multibyte)
150 # define RE_STRING_CHAR(p, multibyte) \
151 (multibyte ? (STRING_CHAR (p)) : (*(p)))
152 # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) \
153 (multibyte ? (STRING_CHAR_AND_LENGTH (p, len)) : ((len) = 1, *(p)))
155 # define RE_CHAR_TO_MULTIBYTE(c) UNIBYTE_TO_CHAR (c)
157 # define RE_CHAR_TO_UNIBYTE(c) CHAR_TO_BYTE_SAFE (c)
159 /* Set C a (possibly converted to multibyte) character before P. P
160 points into a string which is the virtual concatenation of STR1
161 (which ends at END1) or STR2 (which ends at END2). */
162 # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
163 do { \
164 if (target_multibyte) \
166 re_char *dtemp = (p) == (str2) ? (end1) : (p); \
167 re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \
168 while (dtemp-- > dlimit && !CHAR_HEAD_P (*dtemp)); \
169 c = STRING_CHAR (dtemp); \
171 else \
173 (c = ((p) == (str2) ? (end1) : (p))[-1]); \
174 (c) = RE_CHAR_TO_MULTIBYTE (c); \
176 } while (0)
178 /* Set C a (possibly converted to multibyte) character at P, and set
179 LEN to the byte length of that character. */
180 # define GET_CHAR_AFTER(c, p, len) \
181 do { \
182 if (target_multibyte) \
183 (c) = STRING_CHAR_AND_LENGTH (p, len); \
184 else \
186 (c) = *p; \
187 len = 1; \
188 (c) = RE_CHAR_TO_MULTIBYTE (c); \
190 } while (0)
192 #else /* not emacs */
194 /* If we are not linking with Emacs proper,
195 we can't use the relocating allocator
196 even if config.h says that we can. */
197 # undef REL_ALLOC
199 # if defined STDC_HEADERS || defined _LIBC
200 # include <stdlib.h>
201 # else
202 char *malloc ();
203 char *realloc ();
204 # endif
206 /* When used in Emacs's lib-src, we need xmalloc and xrealloc. */
208 void *
209 xmalloc (size)
210 size_t size;
212 register void *val;
213 val = (void *) malloc (size);
214 if (!val && size)
216 write (2, "virtual memory exhausted\n", 25);
217 exit (1);
219 return val;
222 void *
223 xrealloc (block, size)
224 void *block;
225 size_t size;
227 register void *val;
228 /* We must call malloc explicitly when BLOCK is 0, since some
229 reallocs don't do this. */
230 if (! block)
231 val = (void *) malloc (size);
232 else
233 val = (void *) realloc (block, size);
234 if (!val && size)
236 write (2, "virtual memory exhausted\n", 25);
237 exit (1);
239 return val;
242 # ifdef malloc
243 # undef malloc
244 # endif
245 # define malloc xmalloc
246 # ifdef realloc
247 # undef realloc
248 # endif
249 # define realloc xrealloc
251 /* When used in Emacs's lib-src, we need to get bzero and bcopy somehow.
252 If nothing else has been done, use the method below. */
253 # ifdef INHIBIT_STRING_HEADER
254 # if !(defined HAVE_BZERO && defined HAVE_BCOPY)
255 # if !defined bzero && !defined bcopy
256 # undef INHIBIT_STRING_HEADER
257 # endif
258 # endif
259 # endif
261 /* This is the normal way of making sure we have memcpy, memcmp and bzero.
262 This is used in most programs--a few other programs avoid this
263 by defining INHIBIT_STRING_HEADER. */
264 # ifndef INHIBIT_STRING_HEADER
265 # if defined HAVE_STRING_H || defined STDC_HEADERS || defined _LIBC
266 # include <string.h>
267 # ifndef bzero
268 # ifndef _LIBC
269 # define bzero(s, n) (memset (s, '\0', n), (s))
270 # else
271 # define bzero(s, n) __bzero (s, n)
272 # endif
273 # endif
274 # else
275 # include <strings.h>
276 # ifndef memcmp
277 # define memcmp(s1, s2, n) bcmp (s1, s2, n)
278 # endif
279 # ifndef memcpy
280 # define memcpy(d, s, n) (bcopy (s, d, n), (d))
281 # endif
282 # endif
283 # endif
285 /* Define the syntax stuff for \<, \>, etc. */
287 /* Sword must be nonzero for the wordchar pattern commands in re_match_2. */
288 enum syntaxcode { Swhitespace = 0, Sword = 1, Ssymbol = 2 };
290 # define SWITCH_ENUM_CAST(x) (x)
292 /* Dummy macros for non-Emacs environments. */
293 # define BASE_LEADING_CODE_P(c) (0)
294 # define CHAR_CHARSET(c) 0
295 # define CHARSET_LEADING_CODE_BASE(c) 0
296 # define MAX_MULTIBYTE_LENGTH 1
297 # define RE_MULTIBYTE_P(x) 0
298 # define RE_TARGET_MULTIBYTE_P(x) 0
299 # define WORD_BOUNDARY_P(c1, c2) (0)
300 # define CHAR_HEAD_P(p) (1)
301 # define SINGLE_BYTE_CHAR_P(c) (1)
302 # define SAME_CHARSET_P(c1, c2) (1)
303 # define MULTIBYTE_FORM_LENGTH(p, s) (1)
304 # define PREV_CHAR_BOUNDARY(p, limit) ((p)--)
305 # define STRING_CHAR(p) (*(p))
306 # define RE_STRING_CHAR(p, multibyte) STRING_CHAR (p)
307 # define CHAR_STRING(c, s) (*(s) = (c), 1)
308 # define STRING_CHAR_AND_LENGTH(p, actual_len) ((actual_len) = 1, *(p))
309 # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) STRING_CHAR_AND_LENGTH (p, len)
310 # define RE_CHAR_TO_MULTIBYTE(c) (c)
311 # define RE_CHAR_TO_UNIBYTE(c) (c)
312 # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
313 (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1)))
314 # define GET_CHAR_AFTER(c, p, len) \
315 (c = *p, len = 1)
316 # define MAKE_CHAR(charset, c1, c2) (c1)
317 # define BYTE8_TO_CHAR(c) (c)
318 # define CHAR_BYTE8_P(c) (0)
319 # define CHAR_LEADING_CODE(c) (c)
321 #endif /* not emacs */
323 #ifndef RE_TRANSLATE
324 # define RE_TRANSLATE(TBL, C) ((unsigned char)(TBL)[C])
325 # define RE_TRANSLATE_P(TBL) (TBL)
326 #endif
328 /* Get the interface, including the syntax bits. */
329 #include "regex.h"
331 /* isalpha etc. are used for the character classes. */
332 #include <ctype.h>
334 #ifdef emacs
336 /* 1 if C is an ASCII character. */
337 # define IS_REAL_ASCII(c) ((c) < 0200)
339 /* 1 if C is a unibyte character. */
340 # define ISUNIBYTE(c) (SINGLE_BYTE_CHAR_P ((c)))
342 /* The Emacs definitions should not be directly affected by locales. */
344 /* In Emacs, these are only used for single-byte characters. */
345 # define ISDIGIT(c) ((c) >= '0' && (c) <= '9')
346 # define ISCNTRL(c) ((c) < ' ')
347 # define ISXDIGIT(c) (((c) >= '0' && (c) <= '9') \
348 || ((c) >= 'a' && (c) <= 'f') \
349 || ((c) >= 'A' && (c) <= 'F'))
351 /* This is only used for single-byte characters. */
352 # define ISBLANK(c) ((c) == ' ' || (c) == '\t')
354 /* The rest must handle multibyte characters. */
356 # define ISGRAPH(c) (SINGLE_BYTE_CHAR_P (c) \
357 ? (c) > ' ' && !((c) >= 0177 && (c) <= 0237) \
358 : 1)
360 # define ISPRINT(c) (SINGLE_BYTE_CHAR_P (c) \
361 ? (c) >= ' ' && !((c) >= 0177 && (c) <= 0237) \
362 : 1)
364 # define ISALNUM(c) (IS_REAL_ASCII (c) \
365 ? (((c) >= 'a' && (c) <= 'z') \
366 || ((c) >= 'A' && (c) <= 'Z') \
367 || ((c) >= '0' && (c) <= '9')) \
368 : SYNTAX (c) == Sword)
370 # define ISALPHA(c) (IS_REAL_ASCII (c) \
371 ? (((c) >= 'a' && (c) <= 'z') \
372 || ((c) >= 'A' && (c) <= 'Z')) \
373 : SYNTAX (c) == Sword)
375 # define ISLOWER(c) (LOWERCASEP (c))
377 # define ISPUNCT(c) (IS_REAL_ASCII (c) \
378 ? ((c) > ' ' && (c) < 0177 \
379 && !(((c) >= 'a' && (c) <= 'z') \
380 || ((c) >= 'A' && (c) <= 'Z') \
381 || ((c) >= '0' && (c) <= '9'))) \
382 : SYNTAX (c) != Sword)
384 # define ISSPACE(c) (SYNTAX (c) == Swhitespace)
386 # define ISUPPER(c) (UPPERCASEP (c))
388 # define ISWORD(c) (SYNTAX (c) == Sword)
390 #else /* not emacs */
392 /* Jim Meyering writes:
394 "... Some ctype macros are valid only for character codes that
395 isascii says are ASCII (SGI's IRIX-4.0.5 is one such system --when
396 using /bin/cc or gcc but without giving an ansi option). So, all
397 ctype uses should be through macros like ISPRINT... If
398 STDC_HEADERS is defined, then autoconf has verified that the ctype
399 macros don't need to be guarded with references to isascii. ...
400 Defining isascii to 1 should let any compiler worth its salt
401 eliminate the && through constant folding."
402 Solaris defines some of these symbols so we must undefine them first. */
404 # undef ISASCII
405 # if defined STDC_HEADERS || (!defined isascii && !defined HAVE_ISASCII)
406 # define ISASCII(c) 1
407 # else
408 # define ISASCII(c) isascii(c)
409 # endif
411 /* 1 if C is an ASCII character. */
412 # define IS_REAL_ASCII(c) ((c) < 0200)
414 /* This distinction is not meaningful, except in Emacs. */
415 # define ISUNIBYTE(c) 1
417 # ifdef isblank
418 # define ISBLANK(c) (ISASCII (c) && isblank (c))
419 # else
420 # define ISBLANK(c) ((c) == ' ' || (c) == '\t')
421 # endif
422 # ifdef isgraph
423 # define ISGRAPH(c) (ISASCII (c) && isgraph (c))
424 # else
425 # define ISGRAPH(c) (ISASCII (c) && isprint (c) && !isspace (c))
426 # endif
428 # undef ISPRINT
429 # define ISPRINT(c) (ISASCII (c) && isprint (c))
430 # define ISDIGIT(c) (ISASCII (c) && isdigit (c))
431 # define ISALNUM(c) (ISASCII (c) && isalnum (c))
432 # define ISALPHA(c) (ISASCII (c) && isalpha (c))
433 # define ISCNTRL(c) (ISASCII (c) && iscntrl (c))
434 # define ISLOWER(c) (ISASCII (c) && islower (c))
435 # define ISPUNCT(c) (ISASCII (c) && ispunct (c))
436 # define ISSPACE(c) (ISASCII (c) && isspace (c))
437 # define ISUPPER(c) (ISASCII (c) && isupper (c))
438 # define ISXDIGIT(c) (ISASCII (c) && isxdigit (c))
440 # define ISWORD(c) ISALPHA(c)
442 # ifdef _tolower
443 # define TOLOWER(c) _tolower(c)
444 # else
445 # define TOLOWER(c) tolower(c)
446 # endif
448 /* How many characters in the character set. */
449 # define CHAR_SET_SIZE 256
451 # ifdef SYNTAX_TABLE
453 extern char *re_syntax_table;
455 # else /* not SYNTAX_TABLE */
457 static char re_syntax_table[CHAR_SET_SIZE];
459 static void
460 init_syntax_once ()
462 register int c;
463 static int done = 0;
465 if (done)
466 return;
468 bzero (re_syntax_table, sizeof re_syntax_table);
470 for (c = 0; c < CHAR_SET_SIZE; ++c)
471 if (ISALNUM (c))
472 re_syntax_table[c] = Sword;
474 re_syntax_table['_'] = Ssymbol;
476 done = 1;
479 # endif /* not SYNTAX_TABLE */
481 # define SYNTAX(c) re_syntax_table[(c)]
483 #endif /* not emacs */
485 #ifndef NULL
486 # define NULL (void *)0
487 #endif
489 /* We remove any previous definition of `SIGN_EXTEND_CHAR',
490 since ours (we hope) works properly with all combinations of
491 machines, compilers, `char' and `unsigned char' argument types.
492 (Per Bothner suggested the basic approach.) */
493 #undef SIGN_EXTEND_CHAR
494 #if __STDC__
495 # define SIGN_EXTEND_CHAR(c) ((signed char) (c))
496 #else /* not __STDC__ */
497 /* As in Harbison and Steele. */
498 # define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128)
499 #endif
501 /* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we
502 use `alloca' instead of `malloc'. This is because using malloc in
503 re_search* or re_match* could cause memory leaks when C-g is used in
504 Emacs; also, malloc is slower and causes storage fragmentation. On
505 the other hand, malloc is more portable, and easier to debug.
507 Because we sometimes use alloca, some routines have to be macros,
508 not functions -- `alloca'-allocated space disappears at the end of the
509 function it is called in. */
511 #ifdef REGEX_MALLOC
513 # define REGEX_ALLOCATE malloc
514 # define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize)
515 # define REGEX_FREE free
517 #else /* not REGEX_MALLOC */
519 /* Emacs already defines alloca, sometimes. */
520 # ifndef alloca
522 /* Make alloca work the best possible way. */
523 # ifdef __GNUC__
524 # define alloca __builtin_alloca
525 # else /* not __GNUC__ */
526 # ifdef HAVE_ALLOCA_H
527 # include <alloca.h>
528 # endif /* HAVE_ALLOCA_H */
529 # endif /* not __GNUC__ */
531 # endif /* not alloca */
533 # define REGEX_ALLOCATE alloca
535 /* Assumes a `char *destination' variable. */
536 # define REGEX_REALLOCATE(source, osize, nsize) \
537 (destination = (char *) alloca (nsize), \
538 memcpy (destination, source, osize))
540 /* No need to do anything to free, after alloca. */
541 # define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */
543 #endif /* not REGEX_MALLOC */
545 /* Define how to allocate the failure stack. */
547 #if defined REL_ALLOC && defined REGEX_MALLOC
549 # define REGEX_ALLOCATE_STACK(size) \
550 r_alloc (&failure_stack_ptr, (size))
551 # define REGEX_REALLOCATE_STACK(source, osize, nsize) \
552 r_re_alloc (&failure_stack_ptr, (nsize))
553 # define REGEX_FREE_STACK(ptr) \
554 r_alloc_free (&failure_stack_ptr)
556 #else /* not using relocating allocator */
558 # ifdef REGEX_MALLOC
560 # define REGEX_ALLOCATE_STACK malloc
561 # define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize)
562 # define REGEX_FREE_STACK free
564 # else /* not REGEX_MALLOC */
566 # define REGEX_ALLOCATE_STACK alloca
568 # define REGEX_REALLOCATE_STACK(source, osize, nsize) \
569 REGEX_REALLOCATE (source, osize, nsize)
570 /* No need to explicitly free anything. */
571 # define REGEX_FREE_STACK(arg) ((void)0)
573 # endif /* not REGEX_MALLOC */
574 #endif /* not using relocating allocator */
577 /* True if `size1' is non-NULL and PTR is pointing anywhere inside
578 `string1' or just past its end. This works if PTR is NULL, which is
579 a good thing. */
580 #define FIRST_STRING_P(ptr) \
581 (size1 && string1 <= (ptr) && (ptr) <= string1 + size1)
583 /* (Re)Allocate N items of type T using malloc, or fail. */
584 #define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t)))
585 #define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t)))
586 #define RETALLOC_IF(addr, n, t) \
587 if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t)
588 #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t)))
590 #define BYTEWIDTH 8 /* In bits. */
592 #define STREQ(s1, s2) ((strcmp (s1, s2) == 0))
594 #undef MAX
595 #undef MIN
596 #define MAX(a, b) ((a) > (b) ? (a) : (b))
597 #define MIN(a, b) ((a) < (b) ? (a) : (b))
599 /* Type of source-pattern and string chars. */
600 typedef const unsigned char re_char;
602 typedef char boolean;
603 #define false 0
604 #define true 1
606 static int re_match_2_internal _RE_ARGS ((struct re_pattern_buffer *bufp,
607 re_char *string1, int size1,
608 re_char *string2, int size2,
609 int pos,
610 struct re_registers *regs,
611 int stop));
613 /* These are the command codes that appear in compiled regular
614 expressions. Some opcodes are followed by argument bytes. A
615 command code can specify any interpretation whatsoever for its
616 arguments. Zero bytes may appear in the compiled regular expression. */
618 typedef enum
620 no_op = 0,
622 /* Succeed right away--no more backtracking. */
623 succeed,
625 /* Followed by one byte giving n, then by n literal bytes. */
626 exactn,
628 /* Matches any (more or less) character. */
629 anychar,
631 /* Matches any one char belonging to specified set. First
632 following byte is number of bitmap bytes. Then come bytes
633 for a bitmap saying which chars are in. Bits in each byte
634 are ordered low-bit-first. A character is in the set if its
635 bit is 1. A character too large to have a bit in the map is
636 automatically not in the set.
638 If the length byte has the 0x80 bit set, then that stuff
639 is followed by a range table:
640 2 bytes of flags for character sets (low 8 bits, high 8 bits)
641 See RANGE_TABLE_WORK_BITS below.
642 2 bytes, the number of pairs that follow (upto 32767)
643 pairs, each 2 multibyte characters,
644 each multibyte character represented as 3 bytes. */
645 charset,
647 /* Same parameters as charset, but match any character that is
648 not one of those specified. */
649 charset_not,
651 /* Start remembering the text that is matched, for storing in a
652 register. Followed by one byte with the register number, in
653 the range 0 to one less than the pattern buffer's re_nsub
654 field. */
655 start_memory,
657 /* Stop remembering the text that is matched and store it in a
658 memory register. Followed by one byte with the register
659 number, in the range 0 to one less than `re_nsub' in the
660 pattern buffer. */
661 stop_memory,
663 /* Match a duplicate of something remembered. Followed by one
664 byte containing the register number. */
665 duplicate,
667 /* Fail unless at beginning of line. */
668 begline,
670 /* Fail unless at end of line. */
671 endline,
673 /* Succeeds if at beginning of buffer (if emacs) or at beginning
674 of string to be matched (if not). */
675 begbuf,
677 /* Analogously, for end of buffer/string. */
678 endbuf,
680 /* Followed by two byte relative address to which to jump. */
681 jump,
683 /* Followed by two-byte relative address of place to resume at
684 in case of failure. */
685 on_failure_jump,
687 /* Like on_failure_jump, but pushes a placeholder instead of the
688 current string position when executed. */
689 on_failure_keep_string_jump,
691 /* Just like `on_failure_jump', except that it checks that we
692 don't get stuck in an infinite loop (matching an empty string
693 indefinitely). */
694 on_failure_jump_loop,
696 /* Just like `on_failure_jump_loop', except that it checks for
697 a different kind of loop (the kind that shows up with non-greedy
698 operators). This operation has to be immediately preceded
699 by a `no_op'. */
700 on_failure_jump_nastyloop,
702 /* A smart `on_failure_jump' used for greedy * and + operators.
703 It analyses the loop before which it is put and if the
704 loop does not require backtracking, it changes itself to
705 `on_failure_keep_string_jump' and short-circuits the loop,
706 else it just defaults to changing itself into `on_failure_jump'.
707 It assumes that it is pointing to just past a `jump'. */
708 on_failure_jump_smart,
710 /* Followed by two-byte relative address and two-byte number n.
711 After matching N times, jump to the address upon failure.
712 Does not work if N starts at 0: use on_failure_jump_loop
713 instead. */
714 succeed_n,
716 /* Followed by two-byte relative address, and two-byte number n.
717 Jump to the address N times, then fail. */
718 jump_n,
720 /* Set the following two-byte relative address to the
721 subsequent two-byte number. The address *includes* the two
722 bytes of number. */
723 set_number_at,
725 wordbeg, /* Succeeds if at word beginning. */
726 wordend, /* Succeeds if at word end. */
728 wordbound, /* Succeeds if at a word boundary. */
729 notwordbound, /* Succeeds if not at a word boundary. */
731 symbeg, /* Succeeds if at symbol beginning. */
732 symend, /* Succeeds if at symbol end. */
734 /* Matches any character whose syntax is specified. Followed by
735 a byte which contains a syntax code, e.g., Sword. */
736 syntaxspec,
738 /* Matches any character whose syntax is not that specified. */
739 notsyntaxspec
741 #ifdef emacs
742 ,before_dot, /* Succeeds if before point. */
743 at_dot, /* Succeeds if at point. */
744 after_dot, /* Succeeds if after point. */
746 /* Matches any character whose category-set contains the specified
747 category. The operator is followed by a byte which contains a
748 category code (mnemonic ASCII character). */
749 categoryspec,
751 /* Matches any character whose category-set does not contain the
752 specified category. The operator is followed by a byte which
753 contains the category code (mnemonic ASCII character). */
754 notcategoryspec
755 #endif /* emacs */
756 } re_opcode_t;
758 /* Common operations on the compiled pattern. */
760 /* Store NUMBER in two contiguous bytes starting at DESTINATION. */
762 #define STORE_NUMBER(destination, number) \
763 do { \
764 (destination)[0] = (number) & 0377; \
765 (destination)[1] = (number) >> 8; \
766 } while (0)
768 /* Same as STORE_NUMBER, except increment DESTINATION to
769 the byte after where the number is stored. Therefore, DESTINATION
770 must be an lvalue. */
772 #define STORE_NUMBER_AND_INCR(destination, number) \
773 do { \
774 STORE_NUMBER (destination, number); \
775 (destination) += 2; \
776 } while (0)
778 /* Put into DESTINATION a number stored in two contiguous bytes starting
779 at SOURCE. */
781 #define EXTRACT_NUMBER(destination, source) \
782 do { \
783 (destination) = *(source) & 0377; \
784 (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \
785 } while (0)
787 #ifdef DEBUG
788 static void extract_number _RE_ARGS ((int *dest, re_char *source));
789 static void
790 extract_number (dest, source)
791 int *dest;
792 re_char *source;
794 int temp = SIGN_EXTEND_CHAR (*(source + 1));
795 *dest = *source & 0377;
796 *dest += temp << 8;
799 # ifndef EXTRACT_MACROS /* To debug the macros. */
800 # undef EXTRACT_NUMBER
801 # define EXTRACT_NUMBER(dest, src) extract_number (&dest, src)
802 # endif /* not EXTRACT_MACROS */
804 #endif /* DEBUG */
806 /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number.
807 SOURCE must be an lvalue. */
809 #define EXTRACT_NUMBER_AND_INCR(destination, source) \
810 do { \
811 EXTRACT_NUMBER (destination, source); \
812 (source) += 2; \
813 } while (0)
815 #ifdef DEBUG
816 static void extract_number_and_incr _RE_ARGS ((int *destination,
817 re_char **source));
818 static void
819 extract_number_and_incr (destination, source)
820 int *destination;
821 re_char **source;
823 extract_number (destination, *source);
824 *source += 2;
827 # ifndef EXTRACT_MACROS
828 # undef EXTRACT_NUMBER_AND_INCR
829 # define EXTRACT_NUMBER_AND_INCR(dest, src) \
830 extract_number_and_incr (&dest, &src)
831 # endif /* not EXTRACT_MACROS */
833 #endif /* DEBUG */
835 /* Store a multibyte character in three contiguous bytes starting
836 DESTINATION, and increment DESTINATION to the byte after where the
837 character is stored. Therefore, DESTINATION must be an lvalue. */
839 #define STORE_CHARACTER_AND_INCR(destination, character) \
840 do { \
841 (destination)[0] = (character) & 0377; \
842 (destination)[1] = ((character) >> 8) & 0377; \
843 (destination)[2] = (character) >> 16; \
844 (destination) += 3; \
845 } while (0)
847 /* Put into DESTINATION a character stored in three contiguous bytes
848 starting at SOURCE. */
850 #define EXTRACT_CHARACTER(destination, source) \
851 do { \
852 (destination) = ((source)[0] \
853 | ((source)[1] << 8) \
854 | ((source)[2] << 16)); \
855 } while (0)
858 /* Macros for charset. */
860 /* Size of bitmap of charset P in bytes. P is a start of charset,
861 i.e. *P is (re_opcode_t) charset or (re_opcode_t) charset_not. */
862 #define CHARSET_BITMAP_SIZE(p) ((p)[1] & 0x7F)
864 /* Nonzero if charset P has range table. */
865 #define CHARSET_RANGE_TABLE_EXISTS_P(p) ((p)[1] & 0x80)
867 /* Return the address of range table of charset P. But not the start
868 of table itself, but the before where the number of ranges is
869 stored. `2 +' means to skip re_opcode_t and size of bitmap,
870 and the 2 bytes of flags at the start of the range table. */
871 #define CHARSET_RANGE_TABLE(p) (&(p)[4 + CHARSET_BITMAP_SIZE (p)])
873 /* Extract the bit flags that start a range table. */
874 #define CHARSET_RANGE_TABLE_BITS(p) \
875 ((p)[2 + CHARSET_BITMAP_SIZE (p)] \
876 + (p)[3 + CHARSET_BITMAP_SIZE (p)] * 0x100)
878 /* Test if C is listed in the bitmap of charset P. */
879 #define CHARSET_LOOKUP_BITMAP(p, c) \
880 ((c) < CHARSET_BITMAP_SIZE (p) * BYTEWIDTH \
881 && (p)[2 + (c) / BYTEWIDTH] & (1 << ((c) % BYTEWIDTH)))
883 /* Return the address of end of RANGE_TABLE. COUNT is number of
884 ranges (which is a pair of (start, end)) in the RANGE_TABLE. `* 2'
885 is start of range and end of range. `* 3' is size of each start
886 and end. */
887 #define CHARSET_RANGE_TABLE_END(range_table, count) \
888 ((range_table) + (count) * 2 * 3)
890 /* Test if C is in RANGE_TABLE. A flag NOT is negated if C is in.
891 COUNT is number of ranges in RANGE_TABLE. */
892 #define CHARSET_LOOKUP_RANGE_TABLE_RAW(not, c, range_table, count) \
893 do \
895 re_wchar_t range_start, range_end; \
896 re_char *p; \
897 re_char *range_table_end \
898 = CHARSET_RANGE_TABLE_END ((range_table), (count)); \
900 for (p = (range_table); p < range_table_end; p += 2 * 3) \
902 EXTRACT_CHARACTER (range_start, p); \
903 EXTRACT_CHARACTER (range_end, p + 3); \
905 if (range_start <= (c) && (c) <= range_end) \
907 (not) = !(not); \
908 break; \
912 while (0)
914 /* Test if C is in range table of CHARSET. The flag NOT is negated if
915 C is listed in it. */
916 #define CHARSET_LOOKUP_RANGE_TABLE(not, c, charset) \
917 do \
919 /* Number of ranges in range table. */ \
920 int count; \
921 re_char *range_table = CHARSET_RANGE_TABLE (charset); \
923 EXTRACT_NUMBER_AND_INCR (count, range_table); \
924 CHARSET_LOOKUP_RANGE_TABLE_RAW ((not), (c), range_table, count); \
926 while (0)
928 /* If DEBUG is defined, Regex prints many voluminous messages about what
929 it is doing (if the variable `debug' is nonzero). If linked with the
930 main program in `iregex.c', you can enter patterns and strings
931 interactively. And if linked with the main program in `main.c' and
932 the other test files, you can run the already-written tests. */
934 #ifdef DEBUG
936 /* We use standard I/O for debugging. */
937 # include <stdio.h>
939 /* It is useful to test things that ``must'' be true when debugging. */
940 # include <assert.h>
942 static int debug = -100000;
944 # define DEBUG_STATEMENT(e) e
945 # define DEBUG_PRINT1(x) if (debug > 0) printf (x)
946 # define DEBUG_PRINT2(x1, x2) if (debug > 0) printf (x1, x2)
947 # define DEBUG_PRINT3(x1, x2, x3) if (debug > 0) printf (x1, x2, x3)
948 # define DEBUG_PRINT4(x1, x2, x3, x4) if (debug > 0) printf (x1, x2, x3, x4)
949 # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \
950 if (debug > 0) print_partial_compiled_pattern (s, e)
951 # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \
952 if (debug > 0) print_double_string (w, s1, sz1, s2, sz2)
955 /* Print the fastmap in human-readable form. */
957 void
958 print_fastmap (fastmap)
959 char *fastmap;
961 unsigned was_a_range = 0;
962 unsigned i = 0;
964 while (i < (1 << BYTEWIDTH))
966 if (fastmap[i++])
968 was_a_range = 0;
969 putchar (i - 1);
970 while (i < (1 << BYTEWIDTH) && fastmap[i])
972 was_a_range = 1;
973 i++;
975 if (was_a_range)
977 printf ("-");
978 putchar (i - 1);
982 putchar ('\n');
986 /* Print a compiled pattern string in human-readable form, starting at
987 the START pointer into it and ending just before the pointer END. */
989 void
990 print_partial_compiled_pattern (start, end)
991 re_char *start;
992 re_char *end;
994 int mcnt, mcnt2;
995 re_char *p = start;
996 re_char *pend = end;
998 if (start == NULL)
1000 fprintf (stderr, "(null)\n");
1001 return;
1004 /* Loop over pattern commands. */
1005 while (p < pend)
1007 fprintf (stderr, "%d:\t", p - start);
1009 switch ((re_opcode_t) *p++)
1011 case no_op:
1012 fprintf (stderr, "/no_op");
1013 break;
1015 case succeed:
1016 fprintf (stderr, "/succeed");
1017 break;
1019 case exactn:
1020 mcnt = *p++;
1021 fprintf (stderr, "/exactn/%d", mcnt);
1024 fprintf (stderr, "/%c", *p++);
1026 while (--mcnt);
1027 break;
1029 case start_memory:
1030 fprintf (stderr, "/start_memory/%d", *p++);
1031 break;
1033 case stop_memory:
1034 fprintf (stderr, "/stop_memory/%d", *p++);
1035 break;
1037 case duplicate:
1038 fprintf (stderr, "/duplicate/%d", *p++);
1039 break;
1041 case anychar:
1042 fprintf (stderr, "/anychar");
1043 break;
1045 case charset:
1046 case charset_not:
1048 register int c, last = -100;
1049 register int in_range = 0;
1050 int length = CHARSET_BITMAP_SIZE (p - 1);
1051 int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1);
1053 fprintf (stderr, "/charset [%s",
1054 (re_opcode_t) *(p - 1) == charset_not ? "^" : "");
1056 if (p + *p >= pend)
1057 fprintf (stderr, " !extends past end of pattern! ");
1059 for (c = 0; c < 256; c++)
1060 if (c / 8 < length
1061 && (p[1 + (c/8)] & (1 << (c % 8))))
1063 /* Are we starting a range? */
1064 if (last + 1 == c && ! in_range)
1066 fprintf (stderr, "-");
1067 in_range = 1;
1069 /* Have we broken a range? */
1070 else if (last + 1 != c && in_range)
1072 fprintf (stderr, "%c", last);
1073 in_range = 0;
1076 if (! in_range)
1077 fprintf (stderr, "%c", c);
1079 last = c;
1082 if (in_range)
1083 fprintf (stderr, "%c", last);
1085 fprintf (stderr, "]");
1087 p += 1 + length;
1089 if (has_range_table)
1091 int count;
1092 fprintf (stderr, "has-range-table");
1094 /* ??? Should print the range table; for now, just skip it. */
1095 p += 2; /* skip range table bits */
1096 EXTRACT_NUMBER_AND_INCR (count, p);
1097 p = CHARSET_RANGE_TABLE_END (p, count);
1100 break;
1102 case begline:
1103 fprintf (stderr, "/begline");
1104 break;
1106 case endline:
1107 fprintf (stderr, "/endline");
1108 break;
1110 case on_failure_jump:
1111 extract_number_and_incr (&mcnt, &p);
1112 fprintf (stderr, "/on_failure_jump to %d", p + mcnt - start);
1113 break;
1115 case on_failure_keep_string_jump:
1116 extract_number_and_incr (&mcnt, &p);
1117 fprintf (stderr, "/on_failure_keep_string_jump to %d", p + mcnt - start);
1118 break;
1120 case on_failure_jump_nastyloop:
1121 extract_number_and_incr (&mcnt, &p);
1122 fprintf (stderr, "/on_failure_jump_nastyloop to %d", p + mcnt - start);
1123 break;
1125 case on_failure_jump_loop:
1126 extract_number_and_incr (&mcnt, &p);
1127 fprintf (stderr, "/on_failure_jump_loop to %d", p + mcnt - start);
1128 break;
1130 case on_failure_jump_smart:
1131 extract_number_and_incr (&mcnt, &p);
1132 fprintf (stderr, "/on_failure_jump_smart to %d", p + mcnt - start);
1133 break;
1135 case jump:
1136 extract_number_and_incr (&mcnt, &p);
1137 fprintf (stderr, "/jump to %d", p + mcnt - start);
1138 break;
1140 case succeed_n:
1141 extract_number_and_incr (&mcnt, &p);
1142 extract_number_and_incr (&mcnt2, &p);
1143 fprintf (stderr, "/succeed_n to %d, %d times", p - 2 + mcnt - start, mcnt2);
1144 break;
1146 case jump_n:
1147 extract_number_and_incr (&mcnt, &p);
1148 extract_number_and_incr (&mcnt2, &p);
1149 fprintf (stderr, "/jump_n to %d, %d times", p - 2 + mcnt - start, mcnt2);
1150 break;
1152 case set_number_at:
1153 extract_number_and_incr (&mcnt, &p);
1154 extract_number_and_incr (&mcnt2, &p);
1155 fprintf (stderr, "/set_number_at location %d to %d", p - 2 + mcnt - start, mcnt2);
1156 break;
1158 case wordbound:
1159 fprintf (stderr, "/wordbound");
1160 break;
1162 case notwordbound:
1163 fprintf (stderr, "/notwordbound");
1164 break;
1166 case wordbeg:
1167 fprintf (stderr, "/wordbeg");
1168 break;
1170 case wordend:
1171 fprintf (stderr, "/wordend");
1172 break;
1174 case symbeg:
1175 fprintf (stderr, "/symbeg");
1176 break;
1178 case symend:
1179 fprintf (stderr, "/symend");
1180 break;
1182 case syntaxspec:
1183 fprintf (stderr, "/syntaxspec");
1184 mcnt = *p++;
1185 fprintf (stderr, "/%d", mcnt);
1186 break;
1188 case notsyntaxspec:
1189 fprintf (stderr, "/notsyntaxspec");
1190 mcnt = *p++;
1191 fprintf (stderr, "/%d", mcnt);
1192 break;
1194 # ifdef emacs
1195 case before_dot:
1196 fprintf (stderr, "/before_dot");
1197 break;
1199 case at_dot:
1200 fprintf (stderr, "/at_dot");
1201 break;
1203 case after_dot:
1204 fprintf (stderr, "/after_dot");
1205 break;
1207 case categoryspec:
1208 fprintf (stderr, "/categoryspec");
1209 mcnt = *p++;
1210 fprintf (stderr, "/%d", mcnt);
1211 break;
1213 case notcategoryspec:
1214 fprintf (stderr, "/notcategoryspec");
1215 mcnt = *p++;
1216 fprintf (stderr, "/%d", mcnt);
1217 break;
1218 # endif /* emacs */
1220 case begbuf:
1221 fprintf (stderr, "/begbuf");
1222 break;
1224 case endbuf:
1225 fprintf (stderr, "/endbuf");
1226 break;
1228 default:
1229 fprintf (stderr, "?%d", *(p-1));
1232 fprintf (stderr, "\n");
1235 fprintf (stderr, "%d:\tend of pattern.\n", p - start);
1239 void
1240 print_compiled_pattern (bufp)
1241 struct re_pattern_buffer *bufp;
1243 re_char *buffer = bufp->buffer;
1245 print_partial_compiled_pattern (buffer, buffer + bufp->used);
1246 printf ("%ld bytes used/%ld bytes allocated.\n",
1247 bufp->used, bufp->allocated);
1249 if (bufp->fastmap_accurate && bufp->fastmap)
1251 printf ("fastmap: ");
1252 print_fastmap (bufp->fastmap);
1255 printf ("re_nsub: %d\t", bufp->re_nsub);
1256 printf ("regs_alloc: %d\t", bufp->regs_allocated);
1257 printf ("can_be_null: %d\t", bufp->can_be_null);
1258 printf ("no_sub: %d\t", bufp->no_sub);
1259 printf ("not_bol: %d\t", bufp->not_bol);
1260 printf ("not_eol: %d\t", bufp->not_eol);
1261 printf ("syntax: %lx\n", bufp->syntax);
1262 fflush (stdout);
1263 /* Perhaps we should print the translate table? */
1267 void
1268 print_double_string (where, string1, size1, string2, size2)
1269 re_char *where;
1270 re_char *string1;
1271 re_char *string2;
1272 int size1;
1273 int size2;
1275 int this_char;
1277 if (where == NULL)
1278 printf ("(null)");
1279 else
1281 if (FIRST_STRING_P (where))
1283 for (this_char = where - string1; this_char < size1; this_char++)
1284 putchar (string1[this_char]);
1286 where = string2;
1289 for (this_char = where - string2; this_char < size2; this_char++)
1290 putchar (string2[this_char]);
1294 #else /* not DEBUG */
1296 # undef assert
1297 # define assert(e)
1299 # define DEBUG_STATEMENT(e)
1300 # define DEBUG_PRINT1(x)
1301 # define DEBUG_PRINT2(x1, x2)
1302 # define DEBUG_PRINT3(x1, x2, x3)
1303 # define DEBUG_PRINT4(x1, x2, x3, x4)
1304 # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e)
1305 # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2)
1307 #endif /* not DEBUG */
1309 /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can
1310 also be assigned to arbitrarily: each pattern buffer stores its own
1311 syntax, so it can be changed between regex compilations. */
1312 /* This has no initializer because initialized variables in Emacs
1313 become read-only after dumping. */
1314 reg_syntax_t re_syntax_options;
1317 /* Specify the precise syntax of regexps for compilation. This provides
1318 for compatibility for various utilities which historically have
1319 different, incompatible syntaxes.
1321 The argument SYNTAX is a bit mask comprised of the various bits
1322 defined in regex.h. We return the old syntax. */
1324 reg_syntax_t
1325 re_set_syntax (syntax)
1326 reg_syntax_t syntax;
1328 reg_syntax_t ret = re_syntax_options;
1330 re_syntax_options = syntax;
1331 return ret;
1333 WEAK_ALIAS (__re_set_syntax, re_set_syntax)
1335 /* Regexp to use to replace spaces, or NULL meaning don't. */
1336 static re_char *whitespace_regexp;
1338 void
1339 re_set_whitespace_regexp (regexp)
1340 const char *regexp;
1342 whitespace_regexp = (re_char *) regexp;
1344 WEAK_ALIAS (__re_set_syntax, re_set_syntax)
1346 /* This table gives an error message for each of the error codes listed
1347 in regex.h. Obviously the order here has to be same as there.
1348 POSIX doesn't require that we do anything for REG_NOERROR,
1349 but why not be nice? */
1351 static const char *re_error_msgid[] =
1353 gettext_noop ("Success"), /* REG_NOERROR */
1354 gettext_noop ("No match"), /* REG_NOMATCH */
1355 gettext_noop ("Invalid regular expression"), /* REG_BADPAT */
1356 gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */
1357 gettext_noop ("Invalid character class name"), /* REG_ECTYPE */
1358 gettext_noop ("Trailing backslash"), /* REG_EESCAPE */
1359 gettext_noop ("Invalid back reference"), /* REG_ESUBREG */
1360 gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */
1361 gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */
1362 gettext_noop ("Unmatched \\{"), /* REG_EBRACE */
1363 gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */
1364 gettext_noop ("Invalid range end"), /* REG_ERANGE */
1365 gettext_noop ("Memory exhausted"), /* REG_ESPACE */
1366 gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */
1367 gettext_noop ("Premature end of regular expression"), /* REG_EEND */
1368 gettext_noop ("Regular expression too big"), /* REG_ESIZE */
1369 gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */
1370 gettext_noop ("Range striding over charsets") /* REG_ERANGEX */
1373 /* Avoiding alloca during matching, to placate r_alloc. */
1375 /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the
1376 searching and matching functions should not call alloca. On some
1377 systems, alloca is implemented in terms of malloc, and if we're
1378 using the relocating allocator routines, then malloc could cause a
1379 relocation, which might (if the strings being searched are in the
1380 ralloc heap) shift the data out from underneath the regexp
1381 routines.
1383 Here's another reason to avoid allocation: Emacs
1384 processes input from X in a signal handler; processing X input may
1385 call malloc; if input arrives while a matching routine is calling
1386 malloc, then we're scrod. But Emacs can't just block input while
1387 calling matching routines; then we don't notice interrupts when
1388 they come in. So, Emacs blocks input around all regexp calls
1389 except the matching calls, which it leaves unprotected, in the
1390 faith that they will not malloc. */
1392 /* Normally, this is fine. */
1393 #define MATCH_MAY_ALLOCATE
1395 /* The match routines may not allocate if (1) they would do it with malloc
1396 and (2) it's not safe for them to use malloc.
1397 Note that if REL_ALLOC is defined, matching would not use malloc for the
1398 failure stack, but we would still use it for the register vectors;
1399 so REL_ALLOC should not affect this. */
1400 #if defined REGEX_MALLOC && defined emacs
1401 # undef MATCH_MAY_ALLOCATE
1402 #endif
1405 /* Failure stack declarations and macros; both re_compile_fastmap and
1406 re_match_2 use a failure stack. These have to be macros because of
1407 REGEX_ALLOCATE_STACK. */
1410 /* Approximate number of failure points for which to initially allocate space
1411 when matching. If this number is exceeded, we allocate more
1412 space, so it is not a hard limit. */
1413 #ifndef INIT_FAILURE_ALLOC
1414 # define INIT_FAILURE_ALLOC 20
1415 #endif
1417 /* Roughly the maximum number of failure points on the stack. Would be
1418 exactly that if always used TYPICAL_FAILURE_SIZE items each time we failed.
1419 This is a variable only so users of regex can assign to it; we never
1420 change it ourselves. We always multiply it by TYPICAL_FAILURE_SIZE
1421 before using it, so it should probably be a byte-count instead. */
1422 # if defined MATCH_MAY_ALLOCATE
1423 /* Note that 4400 was enough to cause a crash on Alpha OSF/1,
1424 whose default stack limit is 2mb. In order for a larger
1425 value to work reliably, you have to try to make it accord
1426 with the process stack limit. */
1427 size_t re_max_failures = 40000;
1428 # else
1429 size_t re_max_failures = 4000;
1430 # endif
1432 union fail_stack_elt
1434 re_char *pointer;
1435 /* This should be the biggest `int' that's no bigger than a pointer. */
1436 long integer;
1439 typedef union fail_stack_elt fail_stack_elt_t;
1441 typedef struct
1443 fail_stack_elt_t *stack;
1444 size_t size;
1445 size_t avail; /* Offset of next open position. */
1446 size_t frame; /* Offset of the cur constructed frame. */
1447 } fail_stack_type;
1449 #define FAIL_STACK_EMPTY() (fail_stack.frame == 0)
1450 #define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size)
1453 /* Define macros to initialize and free the failure stack.
1454 Do `return -2' if the alloc fails. */
1456 #ifdef MATCH_MAY_ALLOCATE
1457 # define INIT_FAIL_STACK() \
1458 do { \
1459 fail_stack.stack = (fail_stack_elt_t *) \
1460 REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * TYPICAL_FAILURE_SIZE \
1461 * sizeof (fail_stack_elt_t)); \
1463 if (fail_stack.stack == NULL) \
1464 return -2; \
1466 fail_stack.size = INIT_FAILURE_ALLOC; \
1467 fail_stack.avail = 0; \
1468 fail_stack.frame = 0; \
1469 } while (0)
1471 # define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack)
1472 #else
1473 # define INIT_FAIL_STACK() \
1474 do { \
1475 fail_stack.avail = 0; \
1476 fail_stack.frame = 0; \
1477 } while (0)
1479 # define RESET_FAIL_STACK() ((void)0)
1480 #endif
1483 /* Double the size of FAIL_STACK, up to a limit
1484 which allows approximately `re_max_failures' items.
1486 Return 1 if succeeds, and 0 if either ran out of memory
1487 allocating space for it or it was already too large.
1489 REGEX_REALLOCATE_STACK requires `destination' be declared. */
1491 /* Factor to increase the failure stack size by
1492 when we increase it.
1493 This used to be 2, but 2 was too wasteful
1494 because the old discarded stacks added up to as much space
1495 were as ultimate, maximum-size stack. */
1496 #define FAIL_STACK_GROWTH_FACTOR 4
1498 #define GROW_FAIL_STACK(fail_stack) \
1499 (((fail_stack).size * sizeof (fail_stack_elt_t) \
1500 >= re_max_failures * TYPICAL_FAILURE_SIZE) \
1501 ? 0 \
1502 : ((fail_stack).stack \
1503 = (fail_stack_elt_t *) \
1504 REGEX_REALLOCATE_STACK ((fail_stack).stack, \
1505 (fail_stack).size * sizeof (fail_stack_elt_t), \
1506 MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \
1507 ((fail_stack).size * sizeof (fail_stack_elt_t) \
1508 * FAIL_STACK_GROWTH_FACTOR))), \
1510 (fail_stack).stack == NULL \
1511 ? 0 \
1512 : ((fail_stack).size \
1513 = (MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \
1514 ((fail_stack).size * sizeof (fail_stack_elt_t) \
1515 * FAIL_STACK_GROWTH_FACTOR)) \
1516 / sizeof (fail_stack_elt_t)), \
1517 1)))
1520 /* Push a pointer value onto the failure stack.
1521 Assumes the variable `fail_stack'. Probably should only
1522 be called from within `PUSH_FAILURE_POINT'. */
1523 #define PUSH_FAILURE_POINTER(item) \
1524 fail_stack.stack[fail_stack.avail++].pointer = (item)
1526 /* This pushes an integer-valued item onto the failure stack.
1527 Assumes the variable `fail_stack'. Probably should only
1528 be called from within `PUSH_FAILURE_POINT'. */
1529 #define PUSH_FAILURE_INT(item) \
1530 fail_stack.stack[fail_stack.avail++].integer = (item)
1532 /* Push a fail_stack_elt_t value onto the failure stack.
1533 Assumes the variable `fail_stack'. Probably should only
1534 be called from within `PUSH_FAILURE_POINT'. */
1535 #define PUSH_FAILURE_ELT(item) \
1536 fail_stack.stack[fail_stack.avail++] = (item)
1538 /* These three POP... operations complement the three PUSH... operations.
1539 All assume that `fail_stack' is nonempty. */
1540 #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer
1541 #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer
1542 #define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail]
1544 /* Individual items aside from the registers. */
1545 #define NUM_NONREG_ITEMS 3
1547 /* Used to examine the stack (to detect infinite loops). */
1548 #define FAILURE_PAT(h) fail_stack.stack[(h) - 1].pointer
1549 #define FAILURE_STR(h) (fail_stack.stack[(h) - 2].pointer)
1550 #define NEXT_FAILURE_HANDLE(h) fail_stack.stack[(h) - 3].integer
1551 #define TOP_FAILURE_HANDLE() fail_stack.frame
1554 #define ENSURE_FAIL_STACK(space) \
1555 while (REMAINING_AVAIL_SLOTS <= space) { \
1556 if (!GROW_FAIL_STACK (fail_stack)) \
1557 return -2; \
1558 DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", (fail_stack).size);\
1559 DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\
1562 /* Push register NUM onto the stack. */
1563 #define PUSH_FAILURE_REG(num) \
1564 do { \
1565 char *destination; \
1566 ENSURE_FAIL_STACK(3); \
1567 DEBUG_PRINT4 (" Push reg %d (spanning %p -> %p)\n", \
1568 num, regstart[num], regend[num]); \
1569 PUSH_FAILURE_POINTER (regstart[num]); \
1570 PUSH_FAILURE_POINTER (regend[num]); \
1571 PUSH_FAILURE_INT (num); \
1572 } while (0)
1574 /* Change the counter's value to VAL, but make sure that it will
1575 be reset when backtracking. */
1576 #define PUSH_NUMBER(ptr,val) \
1577 do { \
1578 char *destination; \
1579 int c; \
1580 ENSURE_FAIL_STACK(3); \
1581 EXTRACT_NUMBER (c, ptr); \
1582 DEBUG_PRINT4 (" Push number %p = %d -> %d\n", ptr, c, val); \
1583 PUSH_FAILURE_INT (c); \
1584 PUSH_FAILURE_POINTER (ptr); \
1585 PUSH_FAILURE_INT (-1); \
1586 STORE_NUMBER (ptr, val); \
1587 } while (0)
1589 /* Pop a saved register off the stack. */
1590 #define POP_FAILURE_REG_OR_COUNT() \
1591 do { \
1592 int reg = POP_FAILURE_INT (); \
1593 if (reg == -1) \
1595 /* It's a counter. */ \
1596 /* Here, we discard `const', making re_match non-reentrant. */ \
1597 unsigned char *ptr = (unsigned char*) POP_FAILURE_POINTER (); \
1598 reg = POP_FAILURE_INT (); \
1599 STORE_NUMBER (ptr, reg); \
1600 DEBUG_PRINT3 (" Pop counter %p = %d\n", ptr, reg); \
1602 else \
1604 regend[reg] = POP_FAILURE_POINTER (); \
1605 regstart[reg] = POP_FAILURE_POINTER (); \
1606 DEBUG_PRINT4 (" Pop reg %d (spanning %p -> %p)\n", \
1607 reg, regstart[reg], regend[reg]); \
1609 } while (0)
1611 /* Check that we are not stuck in an infinite loop. */
1612 #define CHECK_INFINITE_LOOP(pat_cur, string_place) \
1613 do { \
1614 int failure = TOP_FAILURE_HANDLE (); \
1615 /* Check for infinite matching loops */ \
1616 while (failure > 0 \
1617 && (FAILURE_STR (failure) == string_place \
1618 || FAILURE_STR (failure) == NULL)) \
1620 assert (FAILURE_PAT (failure) >= bufp->buffer \
1621 && FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \
1622 if (FAILURE_PAT (failure) == pat_cur) \
1624 cycle = 1; \
1625 break; \
1627 DEBUG_PRINT2 (" Other pattern: %p\n", FAILURE_PAT (failure)); \
1628 failure = NEXT_FAILURE_HANDLE(failure); \
1630 DEBUG_PRINT2 (" Other string: %p\n", FAILURE_STR (failure)); \
1631 } while (0)
1633 /* Push the information about the state we will need
1634 if we ever fail back to it.
1636 Requires variables fail_stack, regstart, regend and
1637 num_regs be declared. GROW_FAIL_STACK requires `destination' be
1638 declared.
1640 Does `return FAILURE_CODE' if runs out of memory. */
1642 #define PUSH_FAILURE_POINT(pattern, string_place) \
1643 do { \
1644 char *destination; \
1645 /* Must be int, so when we don't save any registers, the arithmetic \
1646 of 0 + -1 isn't done as unsigned. */ \
1648 DEBUG_STATEMENT (nfailure_points_pushed++); \
1649 DEBUG_PRINT1 ("\nPUSH_FAILURE_POINT:\n"); \
1650 DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail); \
1651 DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\
1653 ENSURE_FAIL_STACK (NUM_NONREG_ITEMS); \
1655 DEBUG_PRINT1 ("\n"); \
1657 DEBUG_PRINT2 (" Push frame index: %d\n", fail_stack.frame); \
1658 PUSH_FAILURE_INT (fail_stack.frame); \
1660 DEBUG_PRINT2 (" Push string %p: `", string_place); \
1661 DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, size2);\
1662 DEBUG_PRINT1 ("'\n"); \
1663 PUSH_FAILURE_POINTER (string_place); \
1665 DEBUG_PRINT2 (" Push pattern %p: ", pattern); \
1666 DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern, pend); \
1667 PUSH_FAILURE_POINTER (pattern); \
1669 /* Close the frame by moving the frame pointer past it. */ \
1670 fail_stack.frame = fail_stack.avail; \
1671 } while (0)
1673 /* Estimate the size of data pushed by a typical failure stack entry.
1674 An estimate is all we need, because all we use this for
1675 is to choose a limit for how big to make the failure stack. */
1676 /* BEWARE, the value `20' is hard-coded in emacs.c:main(). */
1677 #define TYPICAL_FAILURE_SIZE 20
1679 /* How many items can still be added to the stack without overflowing it. */
1680 #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail)
1683 /* Pops what PUSH_FAIL_STACK pushes.
1685 We restore into the parameters, all of which should be lvalues:
1686 STR -- the saved data position.
1687 PAT -- the saved pattern position.
1688 REGSTART, REGEND -- arrays of string positions.
1690 Also assumes the variables `fail_stack' and (if debugging), `bufp',
1691 `pend', `string1', `size1', `string2', and `size2'. */
1693 #define POP_FAILURE_POINT(str, pat) \
1694 do { \
1695 assert (!FAIL_STACK_EMPTY ()); \
1697 /* Remove failure points and point to how many regs pushed. */ \
1698 DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \
1699 DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \
1700 DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \
1702 /* Pop the saved registers. */ \
1703 while (fail_stack.frame < fail_stack.avail) \
1704 POP_FAILURE_REG_OR_COUNT (); \
1706 pat = POP_FAILURE_POINTER (); \
1707 DEBUG_PRINT2 (" Popping pattern %p: ", pat); \
1708 DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \
1710 /* If the saved string location is NULL, it came from an \
1711 on_failure_keep_string_jump opcode, and we want to throw away the \
1712 saved NULL, thus retaining our current position in the string. */ \
1713 str = POP_FAILURE_POINTER (); \
1714 DEBUG_PRINT2 (" Popping string %p: `", str); \
1715 DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \
1716 DEBUG_PRINT1 ("'\n"); \
1718 fail_stack.frame = POP_FAILURE_INT (); \
1719 DEBUG_PRINT2 (" Popping frame index: %d\n", fail_stack.frame); \
1721 assert (fail_stack.avail >= 0); \
1722 assert (fail_stack.frame <= fail_stack.avail); \
1724 DEBUG_STATEMENT (nfailure_points_popped++); \
1725 } while (0) /* POP_FAILURE_POINT */
1729 /* Registers are set to a sentinel when they haven't yet matched. */
1730 #define REG_UNSET(e) ((e) == NULL)
1732 /* Subroutine declarations and macros for regex_compile. */
1734 static reg_errcode_t regex_compile _RE_ARGS ((re_char *pattern, size_t size,
1735 reg_syntax_t syntax,
1736 struct re_pattern_buffer *bufp));
1737 static void store_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc, int arg));
1738 static void store_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc,
1739 int arg1, int arg2));
1740 static void insert_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc,
1741 int arg, unsigned char *end));
1742 static void insert_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc,
1743 int arg1, int arg2, unsigned char *end));
1744 static boolean at_begline_loc_p _RE_ARGS ((re_char *pattern,
1745 re_char *p,
1746 reg_syntax_t syntax));
1747 static boolean at_endline_loc_p _RE_ARGS ((re_char *p,
1748 re_char *pend,
1749 reg_syntax_t syntax));
1750 static re_char *skip_one_char _RE_ARGS ((re_char *p));
1751 static int analyse_first _RE_ARGS ((re_char *p, re_char *pend,
1752 char *fastmap, const int multibyte));
1754 /* Fetch the next character in the uncompiled pattern, with no
1755 translation. */
1756 #define PATFETCH(c) \
1757 do { \
1758 int len; \
1759 if (p == pend) return REG_EEND; \
1760 c = RE_STRING_CHAR_AND_LENGTH (p, len, multibyte); \
1761 p += len; \
1762 } while (0)
1765 /* If `translate' is non-null, return translate[D], else just D. We
1766 cast the subscript to translate because some data is declared as
1767 `char *', to avoid warnings when a string constant is passed. But
1768 when we use a character as a subscript we must make it unsigned. */
1769 #ifndef TRANSLATE
1770 # define TRANSLATE(d) \
1771 (RE_TRANSLATE_P (translate) ? RE_TRANSLATE (translate, (d)) : (d))
1772 #endif
1775 /* Macros for outputting the compiled pattern into `buffer'. */
1777 /* If the buffer isn't allocated when it comes in, use this. */
1778 #define INIT_BUF_SIZE 32
1780 /* Make sure we have at least N more bytes of space in buffer. */
1781 #define GET_BUFFER_SPACE(n) \
1782 while ((size_t) (b - bufp->buffer + (n)) > bufp->allocated) \
1783 EXTEND_BUFFER ()
1785 /* Make sure we have one more byte of buffer space and then add C to it. */
1786 #define BUF_PUSH(c) \
1787 do { \
1788 GET_BUFFER_SPACE (1); \
1789 *b++ = (unsigned char) (c); \
1790 } while (0)
1793 /* Ensure we have two more bytes of buffer space and then append C1 and C2. */
1794 #define BUF_PUSH_2(c1, c2) \
1795 do { \
1796 GET_BUFFER_SPACE (2); \
1797 *b++ = (unsigned char) (c1); \
1798 *b++ = (unsigned char) (c2); \
1799 } while (0)
1802 /* As with BUF_PUSH_2, except for three bytes. */
1803 #define BUF_PUSH_3(c1, c2, c3) \
1804 do { \
1805 GET_BUFFER_SPACE (3); \
1806 *b++ = (unsigned char) (c1); \
1807 *b++ = (unsigned char) (c2); \
1808 *b++ = (unsigned char) (c3); \
1809 } while (0)
1812 /* Store a jump with opcode OP at LOC to location TO. We store a
1813 relative address offset by the three bytes the jump itself occupies. */
1814 #define STORE_JUMP(op, loc, to) \
1815 store_op1 (op, loc, (to) - (loc) - 3)
1817 /* Likewise, for a two-argument jump. */
1818 #define STORE_JUMP2(op, loc, to, arg) \
1819 store_op2 (op, loc, (to) - (loc) - 3, arg)
1821 /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */
1822 #define INSERT_JUMP(op, loc, to) \
1823 insert_op1 (op, loc, (to) - (loc) - 3, b)
1825 /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */
1826 #define INSERT_JUMP2(op, loc, to, arg) \
1827 insert_op2 (op, loc, (to) - (loc) - 3, arg, b)
1830 /* This is not an arbitrary limit: the arguments which represent offsets
1831 into the pattern are two bytes long. So if 2^15 bytes turns out to
1832 be too small, many things would have to change. */
1833 # define MAX_BUF_SIZE (1L << 15)
1835 #if 0 /* This is when we thought it could be 2^16 bytes. */
1836 /* Any other compiler which, like MSC, has allocation limit below 2^16
1837 bytes will have to use approach similar to what was done below for
1838 MSC and drop MAX_BUF_SIZE a bit. Otherwise you may end up
1839 reallocating to 0 bytes. Such thing is not going to work too well.
1840 You have been warned!! */
1841 #if defined _MSC_VER && !defined WIN32
1842 /* Microsoft C 16-bit versions limit malloc to approx 65512 bytes. */
1843 # define MAX_BUF_SIZE 65500L
1844 #else
1845 # define MAX_BUF_SIZE (1L << 16)
1846 #endif
1847 #endif /* 0 */
1849 /* Extend the buffer by twice its current size via realloc and
1850 reset the pointers that pointed into the old block to point to the
1851 correct places in the new one. If extending the buffer results in it
1852 being larger than MAX_BUF_SIZE, then flag memory exhausted. */
1853 #if __BOUNDED_POINTERS__
1854 # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated)
1855 # define MOVE_BUFFER_POINTER(P) \
1856 (__ptrlow (P) = new_buffer + (__ptrlow (P) - old_buffer), \
1857 SET_HIGH_BOUND (P), \
1858 __ptrvalue (P) = new_buffer + (__ptrvalue (P) - old_buffer))
1859 # define ELSE_EXTEND_BUFFER_HIGH_BOUND \
1860 else \
1862 SET_HIGH_BOUND (b); \
1863 SET_HIGH_BOUND (begalt); \
1864 if (fixup_alt_jump) \
1865 SET_HIGH_BOUND (fixup_alt_jump); \
1866 if (laststart) \
1867 SET_HIGH_BOUND (laststart); \
1868 if (pending_exact) \
1869 SET_HIGH_BOUND (pending_exact); \
1871 #else
1872 # define MOVE_BUFFER_POINTER(P) ((P) = new_buffer + ((P) - old_buffer))
1873 # define ELSE_EXTEND_BUFFER_HIGH_BOUND
1874 #endif
1875 #define EXTEND_BUFFER() \
1876 do { \
1877 unsigned char *old_buffer = bufp->buffer; \
1878 if (bufp->allocated == MAX_BUF_SIZE) \
1879 return REG_ESIZE; \
1880 bufp->allocated <<= 1; \
1881 if (bufp->allocated > MAX_BUF_SIZE) \
1882 bufp->allocated = MAX_BUF_SIZE; \
1883 RETALLOC (bufp->buffer, bufp->allocated, unsigned char); \
1884 if (bufp->buffer == NULL) \
1885 return REG_ESPACE; \
1886 /* If the buffer moved, move all the pointers into it. */ \
1887 if (old_buffer != bufp->buffer) \
1889 unsigned char *new_buffer = bufp->buffer; \
1890 MOVE_BUFFER_POINTER (b); \
1891 MOVE_BUFFER_POINTER (begalt); \
1892 if (fixup_alt_jump) \
1893 MOVE_BUFFER_POINTER (fixup_alt_jump); \
1894 if (laststart) \
1895 MOVE_BUFFER_POINTER (laststart); \
1896 if (pending_exact) \
1897 MOVE_BUFFER_POINTER (pending_exact); \
1899 ELSE_EXTEND_BUFFER_HIGH_BOUND \
1900 } while (0)
1903 /* Since we have one byte reserved for the register number argument to
1904 {start,stop}_memory, the maximum number of groups we can report
1905 things about is what fits in that byte. */
1906 #define MAX_REGNUM 255
1908 /* But patterns can have more than `MAX_REGNUM' registers. We just
1909 ignore the excess. */
1910 typedef int regnum_t;
1913 /* Macros for the compile stack. */
1915 /* Since offsets can go either forwards or backwards, this type needs to
1916 be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */
1917 /* int may be not enough when sizeof(int) == 2. */
1918 typedef long pattern_offset_t;
1920 typedef struct
1922 pattern_offset_t begalt_offset;
1923 pattern_offset_t fixup_alt_jump;
1924 pattern_offset_t laststart_offset;
1925 regnum_t regnum;
1926 } compile_stack_elt_t;
1929 typedef struct
1931 compile_stack_elt_t *stack;
1932 unsigned size;
1933 unsigned avail; /* Offset of next open position. */
1934 } compile_stack_type;
1937 #define INIT_COMPILE_STACK_SIZE 32
1939 #define COMPILE_STACK_EMPTY (compile_stack.avail == 0)
1940 #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size)
1942 /* The next available element. */
1943 #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail])
1945 /* Explicit quit checking is only used on NTemacs and whenever we
1946 use polling to process input events. */
1947 #if defined emacs && (defined WINDOWSNT || defined SYNC_INPUT) && defined QUIT
1948 extern int immediate_quit;
1949 # define IMMEDIATE_QUIT_CHECK \
1950 do { \
1951 if (immediate_quit) QUIT; \
1952 } while (0)
1953 #else
1954 # define IMMEDIATE_QUIT_CHECK ((void)0)
1955 #endif
1957 /* Structure to manage work area for range table. */
1958 struct range_table_work_area
1960 int *table; /* actual work area. */
1961 int allocated; /* allocated size for work area in bytes. */
1962 int used; /* actually used size in words. */
1963 int bits; /* flag to record character classes */
1966 /* Make sure that WORK_AREA can hold more N multibyte characters.
1967 This is used only in set_image_of_range and set_image_of_range_1.
1968 It expects WORK_AREA to be a pointer.
1969 If it can't get the space, it returns from the surrounding function. */
1971 #define EXTEND_RANGE_TABLE(work_area, n) \
1972 do { \
1973 if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \
1975 extend_range_table_work_area (&work_area); \
1976 if ((work_area).table == 0) \
1977 return (REG_ESPACE); \
1979 } while (0)
1981 #define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \
1982 (work_area).bits |= (bit)
1984 /* Bits used to implement the multibyte-part of the various character classes
1985 such as [:alnum:] in a charset's range table. */
1986 #define BIT_WORD 0x1
1987 #define BIT_LOWER 0x2
1988 #define BIT_PUNCT 0x4
1989 #define BIT_SPACE 0x8
1990 #define BIT_UPPER 0x10
1991 #define BIT_MULTIBYTE 0x20
1993 /* Set a range (RANGE_START, RANGE_END) to WORK_AREA. */
1994 #define SET_RANGE_TABLE_WORK_AREA(work_area, range_start, range_end) \
1995 do { \
1996 EXTEND_RANGE_TABLE ((work_area), 2); \
1997 (work_area).table[(work_area).used++] = (range_start); \
1998 (work_area).table[(work_area).used++] = (range_end); \
1999 } while (0)
2001 /* Free allocated memory for WORK_AREA. */
2002 #define FREE_RANGE_TABLE_WORK_AREA(work_area) \
2003 do { \
2004 if ((work_area).table) \
2005 free ((work_area).table); \
2006 } while (0)
2008 #define CLEAR_RANGE_TABLE_WORK_USED(work_area) ((work_area).used = 0, (work_area).bits = 0)
2009 #define RANGE_TABLE_WORK_USED(work_area) ((work_area).used)
2010 #define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits)
2011 #define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i])
2014 /* Set the bit for character C in a list. */
2015 #define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH))
2018 #ifdef emacs
2020 /* Store characters in the range FROM to TO in the bitmap at B (for
2021 ASCII and unibyte characters) and WORK_AREA (for multibyte
2022 characters) while translating them and paying attention to the
2023 continuity of translated characters.
2025 Implementation note: It is better to implement these fairly big
2026 macros by a function, but it's not that easy because macros called
2027 in this macro assume various local variables already declared. */
2029 /* Both FROM and TO are ASCII characters. */
2031 #define SETUP_ASCII_RANGE(work_area, FROM, TO) \
2032 do { \
2033 int C0, C1; \
2035 for (C0 = (FROM); C0 <= (TO); C0++) \
2037 C1 = TRANSLATE (C0); \
2038 if (! ASCII_CHAR_P (C1)) \
2040 SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \
2041 if ((C1 = RE_CHAR_TO_UNIBYTE (C1)) < 0) \
2042 C1 = C0; \
2044 SET_LIST_BIT (C1); \
2046 } while (0)
2049 /* Both FROM and TO are unibyte characters (0x80..0xFF). */
2051 #define SETUP_UNIBYTE_RANGE(work_area, FROM, TO) \
2052 do { \
2053 int C0, C1, C2, I; \
2054 int USED = RANGE_TABLE_WORK_USED (work_area); \
2056 for (C0 = (FROM); C0 <= (TO); C0++) \
2058 C1 = RE_CHAR_TO_MULTIBYTE (C0); \
2059 if (CHAR_BYTE8_P (C1)) \
2060 SET_LIST_BIT (C0); \
2061 else \
2063 C2 = TRANSLATE (C1); \
2064 if (C2 == C1 \
2065 || (C1 = RE_CHAR_TO_UNIBYTE (C2)) < 0) \
2066 C1 = C0; \
2067 SET_LIST_BIT (C1); \
2068 for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \
2070 int from = RANGE_TABLE_WORK_ELT (work_area, I); \
2071 int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \
2073 if (C2 >= from - 1 && C2 <= to + 1) \
2075 if (C2 == from - 1) \
2076 RANGE_TABLE_WORK_ELT (work_area, I)--; \
2077 else if (C2 == to + 1) \
2078 RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \
2079 break; \
2082 if (I < USED) \
2083 SET_RANGE_TABLE_WORK_AREA ((work_area), C2, C2); \
2086 } while (0)
2089 /* Both FROM and TO are multibyte characters. */
2091 #define SETUP_MULTIBYTE_RANGE(work_area, FROM, TO) \
2092 do { \
2093 int C0, C1, C2, I, USED = RANGE_TABLE_WORK_USED (work_area); \
2095 SET_RANGE_TABLE_WORK_AREA ((work_area), (FROM), (TO)); \
2096 for (C0 = (FROM); C0 <= (TO); C0++) \
2098 C1 = TRANSLATE (C0); \
2099 if ((C2 = RE_CHAR_TO_UNIBYTE (C1)) >= 0 \
2100 || (C1 != C0 && (C2 = RE_CHAR_TO_UNIBYTE (C0)) >= 0)) \
2101 SET_LIST_BIT (C2); \
2102 if (C1 >= (FROM) && C1 <= (TO)) \
2103 continue; \
2104 for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \
2106 int from = RANGE_TABLE_WORK_ELT (work_area, I); \
2107 int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \
2109 if (C1 >= from - 1 && C1 <= to + 1) \
2111 if (C1 == from - 1) \
2112 RANGE_TABLE_WORK_ELT (work_area, I)--; \
2113 else if (C1 == to + 1) \
2114 RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \
2115 break; \
2118 if (I < USED) \
2119 SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \
2121 } while (0)
2123 #endif /* emacs */
2125 /* Get the next unsigned number in the uncompiled pattern. */
2126 #define GET_UNSIGNED_NUMBER(num) \
2127 do { \
2128 if (p == pend) \
2129 FREE_STACK_RETURN (REG_EBRACE); \
2130 else \
2132 PATFETCH (c); \
2133 while ('0' <= c && c <= '9') \
2135 int prev; \
2136 if (num < 0) \
2137 num = 0; \
2138 prev = num; \
2139 num = num * 10 + c - '0'; \
2140 if (num / 10 != prev) \
2141 FREE_STACK_RETURN (REG_BADBR); \
2142 if (p == pend) \
2143 FREE_STACK_RETURN (REG_EBRACE); \
2144 PATFETCH (c); \
2147 } while (0)
2149 #if ! WIDE_CHAR_SUPPORT
2151 /* Map a string to the char class it names (if any). */
2152 re_wctype_t
2153 re_wctype (str)
2154 re_char *str;
2156 const char *string = str;
2157 if (STREQ (string, "alnum")) return RECC_ALNUM;
2158 else if (STREQ (string, "alpha")) return RECC_ALPHA;
2159 else if (STREQ (string, "word")) return RECC_WORD;
2160 else if (STREQ (string, "ascii")) return RECC_ASCII;
2161 else if (STREQ (string, "nonascii")) return RECC_NONASCII;
2162 else if (STREQ (string, "graph")) return RECC_GRAPH;
2163 else if (STREQ (string, "lower")) return RECC_LOWER;
2164 else if (STREQ (string, "print")) return RECC_PRINT;
2165 else if (STREQ (string, "punct")) return RECC_PUNCT;
2166 else if (STREQ (string, "space")) return RECC_SPACE;
2167 else if (STREQ (string, "upper")) return RECC_UPPER;
2168 else if (STREQ (string, "unibyte")) return RECC_UNIBYTE;
2169 else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE;
2170 else if (STREQ (string, "digit")) return RECC_DIGIT;
2171 else if (STREQ (string, "xdigit")) return RECC_XDIGIT;
2172 else if (STREQ (string, "cntrl")) return RECC_CNTRL;
2173 else if (STREQ (string, "blank")) return RECC_BLANK;
2174 else return 0;
2177 /* True if CH is in the char class CC. */
2178 boolean
2179 re_iswctype (ch, cc)
2180 int ch;
2181 re_wctype_t cc;
2183 switch (cc)
2185 case RECC_ALNUM: return ISALNUM (ch);
2186 case RECC_ALPHA: return ISALPHA (ch);
2187 case RECC_BLANK: return ISBLANK (ch);
2188 case RECC_CNTRL: return ISCNTRL (ch);
2189 case RECC_DIGIT: return ISDIGIT (ch);
2190 case RECC_GRAPH: return ISGRAPH (ch);
2191 case RECC_LOWER: return ISLOWER (ch);
2192 case RECC_PRINT: return ISPRINT (ch);
2193 case RECC_PUNCT: return ISPUNCT (ch);
2194 case RECC_SPACE: return ISSPACE (ch);
2195 case RECC_UPPER: return ISUPPER (ch);
2196 case RECC_XDIGIT: return ISXDIGIT (ch);
2197 case RECC_ASCII: return IS_REAL_ASCII (ch);
2198 case RECC_NONASCII: return !IS_REAL_ASCII (ch);
2199 case RECC_UNIBYTE: return ISUNIBYTE (ch);
2200 case RECC_MULTIBYTE: return !ISUNIBYTE (ch);
2201 case RECC_WORD: return ISWORD (ch);
2202 case RECC_ERROR: return false;
2203 default:
2204 abort();
2208 /* Return a bit-pattern to use in the range-table bits to match multibyte
2209 chars of class CC. */
2210 static int
2211 re_wctype_to_bit (cc)
2212 re_wctype_t cc;
2214 switch (cc)
2216 case RECC_NONASCII: case RECC_PRINT: case RECC_GRAPH:
2217 case RECC_MULTIBYTE: return BIT_MULTIBYTE;
2218 case RECC_ALPHA: case RECC_ALNUM: case RECC_WORD: return BIT_WORD;
2219 case RECC_LOWER: return BIT_LOWER;
2220 case RECC_UPPER: return BIT_UPPER;
2221 case RECC_PUNCT: return BIT_PUNCT;
2222 case RECC_SPACE: return BIT_SPACE;
2223 case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL:
2224 case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0;
2225 default:
2226 abort();
2229 #endif
2231 /* Filling in the work area of a range. */
2233 /* Actually extend the space in WORK_AREA. */
2235 static void
2236 extend_range_table_work_area (work_area)
2237 struct range_table_work_area *work_area;
2239 work_area->allocated += 16 * sizeof (int);
2240 if (work_area->table)
2241 work_area->table
2242 = (int *) realloc (work_area->table, work_area->allocated);
2243 else
2244 work_area->table
2245 = (int *) malloc (work_area->allocated);
2248 #if 0
2249 #ifdef emacs
2251 /* Carefully find the ranges of codes that are equivalent
2252 under case conversion to the range start..end when passed through
2253 TRANSLATE. Handle the case where non-letters can come in between
2254 two upper-case letters (which happens in Latin-1).
2255 Also handle the case of groups of more than 2 case-equivalent chars.
2257 The basic method is to look at consecutive characters and see
2258 if they can form a run that can be handled as one.
2260 Returns -1 if successful, REG_ESPACE if ran out of space. */
2262 static int
2263 set_image_of_range_1 (work_area, start, end, translate)
2264 RE_TRANSLATE_TYPE translate;
2265 struct range_table_work_area *work_area;
2266 re_wchar_t start, end;
2268 /* `one_case' indicates a character, or a run of characters,
2269 each of which is an isolate (no case-equivalents).
2270 This includes all ASCII non-letters.
2272 `two_case' indicates a character, or a run of characters,
2273 each of which has two case-equivalent forms.
2274 This includes all ASCII letters.
2276 `strange' indicates a character that has more than one
2277 case-equivalent. */
2279 enum case_type {one_case, two_case, strange};
2281 /* Describe the run that is in progress,
2282 which the next character can try to extend.
2283 If run_type is strange, that means there really is no run.
2284 If run_type is one_case, then run_start...run_end is the run.
2285 If run_type is two_case, then the run is run_start...run_end,
2286 and the case-equivalents end at run_eqv_end. */
2288 enum case_type run_type = strange;
2289 int run_start, run_end, run_eqv_end;
2291 Lisp_Object eqv_table;
2293 if (!RE_TRANSLATE_P (translate))
2295 EXTEND_RANGE_TABLE (work_area, 2);
2296 work_area->table[work_area->used++] = (start);
2297 work_area->table[work_area->used++] = (end);
2298 return -1;
2301 eqv_table = XCHAR_TABLE (translate)->extras[2];
2303 for (; start <= end; start++)
2305 enum case_type this_type;
2306 int eqv = RE_TRANSLATE (eqv_table, start);
2307 int minchar, maxchar;
2309 /* Classify this character */
2310 if (eqv == start)
2311 this_type = one_case;
2312 else if (RE_TRANSLATE (eqv_table, eqv) == start)
2313 this_type = two_case;
2314 else
2315 this_type = strange;
2317 if (start < eqv)
2318 minchar = start, maxchar = eqv;
2319 else
2320 minchar = eqv, maxchar = start;
2322 /* Can this character extend the run in progress? */
2323 if (this_type == strange || this_type != run_type
2324 || !(minchar == run_end + 1
2325 && (run_type == two_case
2326 ? maxchar == run_eqv_end + 1 : 1)))
2328 /* No, end the run.
2329 Record each of its equivalent ranges. */
2330 if (run_type == one_case)
2332 EXTEND_RANGE_TABLE (work_area, 2);
2333 work_area->table[work_area->used++] = run_start;
2334 work_area->table[work_area->used++] = run_end;
2336 else if (run_type == two_case)
2338 EXTEND_RANGE_TABLE (work_area, 4);
2339 work_area->table[work_area->used++] = run_start;
2340 work_area->table[work_area->used++] = run_end;
2341 work_area->table[work_area->used++]
2342 = RE_TRANSLATE (eqv_table, run_start);
2343 work_area->table[work_area->used++]
2344 = RE_TRANSLATE (eqv_table, run_end);
2346 run_type = strange;
2349 if (this_type == strange)
2351 /* For a strange character, add each of its equivalents, one
2352 by one. Don't start a range. */
2355 EXTEND_RANGE_TABLE (work_area, 2);
2356 work_area->table[work_area->used++] = eqv;
2357 work_area->table[work_area->used++] = eqv;
2358 eqv = RE_TRANSLATE (eqv_table, eqv);
2360 while (eqv != start);
2363 /* Add this char to the run, or start a new run. */
2364 else if (run_type == strange)
2366 /* Initialize a new range. */
2367 run_type = this_type;
2368 run_start = start;
2369 run_end = start;
2370 run_eqv_end = RE_TRANSLATE (eqv_table, run_end);
2372 else
2374 /* Extend a running range. */
2375 run_end = minchar;
2376 run_eqv_end = RE_TRANSLATE (eqv_table, run_end);
2380 /* If a run is still in progress at the end, finish it now
2381 by recording its equivalent ranges. */
2382 if (run_type == one_case)
2384 EXTEND_RANGE_TABLE (work_area, 2);
2385 work_area->table[work_area->used++] = run_start;
2386 work_area->table[work_area->used++] = run_end;
2388 else if (run_type == two_case)
2390 EXTEND_RANGE_TABLE (work_area, 4);
2391 work_area->table[work_area->used++] = run_start;
2392 work_area->table[work_area->used++] = run_end;
2393 work_area->table[work_area->used++]
2394 = RE_TRANSLATE (eqv_table, run_start);
2395 work_area->table[work_area->used++]
2396 = RE_TRANSLATE (eqv_table, run_end);
2399 return -1;
2402 #endif /* emacs */
2404 /* Record the image of the range start..end when passed through
2405 TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end)
2406 and is not even necessarily contiguous.
2407 Normally we approximate it with the smallest contiguous range that contains
2408 all the chars we need. However, for Latin-1 we go to extra effort
2409 to do a better job.
2411 This function is not called for ASCII ranges.
2413 Returns -1 if successful, REG_ESPACE if ran out of space. */
2415 static int
2416 set_image_of_range (work_area, start, end, translate)
2417 RE_TRANSLATE_TYPE translate;
2418 struct range_table_work_area *work_area;
2419 re_wchar_t start, end;
2421 re_wchar_t cmin, cmax;
2423 #ifdef emacs
2424 /* For Latin-1 ranges, use set_image_of_range_1
2425 to get proper handling of ranges that include letters and nonletters.
2426 For a range that includes the whole of Latin-1, this is not necessary.
2427 For other character sets, we don't bother to get this right. */
2428 if (RE_TRANSLATE_P (translate) && start < 04400
2429 && !(start < 04200 && end >= 04377))
2431 int newend;
2432 int tem;
2433 newend = end;
2434 if (newend > 04377)
2435 newend = 04377;
2436 tem = set_image_of_range_1 (work_area, start, newend, translate);
2437 if (tem > 0)
2438 return tem;
2440 start = 04400;
2441 if (end < 04400)
2442 return -1;
2444 #endif
2446 EXTEND_RANGE_TABLE (work_area, 2);
2447 work_area->table[work_area->used++] = (start);
2448 work_area->table[work_area->used++] = (end);
2450 cmin = -1, cmax = -1;
2452 if (RE_TRANSLATE_P (translate))
2454 int ch;
2456 for (ch = start; ch <= end; ch++)
2458 re_wchar_t c = TRANSLATE (ch);
2459 if (! (start <= c && c <= end))
2461 if (cmin == -1)
2462 cmin = c, cmax = c;
2463 else
2465 cmin = MIN (cmin, c);
2466 cmax = MAX (cmax, c);
2471 if (cmin != -1)
2473 EXTEND_RANGE_TABLE (work_area, 2);
2474 work_area->table[work_area->used++] = (cmin);
2475 work_area->table[work_area->used++] = (cmax);
2479 return -1;
2481 #endif /* 0 */
2483 #ifndef MATCH_MAY_ALLOCATE
2485 /* If we cannot allocate large objects within re_match_2_internal,
2486 we make the fail stack and register vectors global.
2487 The fail stack, we grow to the maximum size when a regexp
2488 is compiled.
2489 The register vectors, we adjust in size each time we
2490 compile a regexp, according to the number of registers it needs. */
2492 static fail_stack_type fail_stack;
2494 /* Size with which the following vectors are currently allocated.
2495 That is so we can make them bigger as needed,
2496 but never make them smaller. */
2497 static int regs_allocated_size;
2499 static re_char ** regstart, ** regend;
2500 static re_char **best_regstart, **best_regend;
2502 /* Make the register vectors big enough for NUM_REGS registers,
2503 but don't make them smaller. */
2505 static
2506 regex_grow_registers (num_regs)
2507 int num_regs;
2509 if (num_regs > regs_allocated_size)
2511 RETALLOC_IF (regstart, num_regs, re_char *);
2512 RETALLOC_IF (regend, num_regs, re_char *);
2513 RETALLOC_IF (best_regstart, num_regs, re_char *);
2514 RETALLOC_IF (best_regend, num_regs, re_char *);
2516 regs_allocated_size = num_regs;
2520 #endif /* not MATCH_MAY_ALLOCATE */
2522 static boolean group_in_compile_stack _RE_ARGS ((compile_stack_type
2523 compile_stack,
2524 regnum_t regnum));
2526 /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX.
2527 Returns one of error codes defined in `regex.h', or zero for success.
2529 Assumes the `allocated' (and perhaps `buffer') and `translate'
2530 fields are set in BUFP on entry.
2532 If it succeeds, results are put in BUFP (if it returns an error, the
2533 contents of BUFP are undefined):
2534 `buffer' is the compiled pattern;
2535 `syntax' is set to SYNTAX;
2536 `used' is set to the length of the compiled pattern;
2537 `fastmap_accurate' is zero;
2538 `re_nsub' is the number of subexpressions in PATTERN;
2539 `not_bol' and `not_eol' are zero;
2541 The `fastmap' field is neither examined nor set. */
2543 /* Insert the `jump' from the end of last alternative to "here".
2544 The space for the jump has already been allocated. */
2545 #define FIXUP_ALT_JUMP() \
2546 do { \
2547 if (fixup_alt_jump) \
2548 STORE_JUMP (jump, fixup_alt_jump, b); \
2549 } while (0)
2552 /* Return, freeing storage we allocated. */
2553 #define FREE_STACK_RETURN(value) \
2554 do { \
2555 FREE_RANGE_TABLE_WORK_AREA (range_table_work); \
2556 free (compile_stack.stack); \
2557 return value; \
2558 } while (0)
2560 static reg_errcode_t
2561 regex_compile (pattern, size, syntax, bufp)
2562 re_char *pattern;
2563 size_t size;
2564 reg_syntax_t syntax;
2565 struct re_pattern_buffer *bufp;
2567 /* We fetch characters from PATTERN here. */
2568 register re_wchar_t c, c1;
2570 /* A random temporary spot in PATTERN. */
2571 re_char *p1;
2573 /* Points to the end of the buffer, where we should append. */
2574 register unsigned char *b;
2576 /* Keeps track of unclosed groups. */
2577 compile_stack_type compile_stack;
2579 /* Points to the current (ending) position in the pattern. */
2580 #ifdef AIX
2581 /* `const' makes AIX compiler fail. */
2582 unsigned char *p = pattern;
2583 #else
2584 re_char *p = pattern;
2585 #endif
2586 re_char *pend = pattern + size;
2588 /* How to translate the characters in the pattern. */
2589 RE_TRANSLATE_TYPE translate = bufp->translate;
2591 /* Address of the count-byte of the most recently inserted `exactn'
2592 command. This makes it possible to tell if a new exact-match
2593 character can be added to that command or if the character requires
2594 a new `exactn' command. */
2595 unsigned char *pending_exact = 0;
2597 /* Address of start of the most recently finished expression.
2598 This tells, e.g., postfix * where to find the start of its
2599 operand. Reset at the beginning of groups and alternatives. */
2600 unsigned char *laststart = 0;
2602 /* Address of beginning of regexp, or inside of last group. */
2603 unsigned char *begalt;
2605 /* Place in the uncompiled pattern (i.e., the {) to
2606 which to go back if the interval is invalid. */
2607 re_char *beg_interval;
2609 /* Address of the place where a forward jump should go to the end of
2610 the containing expression. Each alternative of an `or' -- except the
2611 last -- ends with a forward jump of this sort. */
2612 unsigned char *fixup_alt_jump = 0;
2614 /* Work area for range table of charset. */
2615 struct range_table_work_area range_table_work;
2617 /* If the object matched can contain multibyte characters. */
2618 const boolean multibyte = RE_MULTIBYTE_P (bufp);
2620 /* If a target of matching can contain multibyte characters. */
2621 const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp);
2623 /* Nonzero if we have pushed down into a subpattern. */
2624 int in_subpattern = 0;
2626 /* These hold the values of p, pattern, and pend from the main
2627 pattern when we have pushed into a subpattern. */
2628 re_char *main_p;
2629 re_char *main_pattern;
2630 re_char *main_pend;
2632 #ifdef DEBUG
2633 debug++;
2634 DEBUG_PRINT1 ("\nCompiling pattern: ");
2635 if (debug > 0)
2637 unsigned debug_count;
2639 for (debug_count = 0; debug_count < size; debug_count++)
2640 putchar (pattern[debug_count]);
2641 putchar ('\n');
2643 #endif /* DEBUG */
2645 /* Initialize the compile stack. */
2646 compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t);
2647 if (compile_stack.stack == NULL)
2648 return REG_ESPACE;
2650 compile_stack.size = INIT_COMPILE_STACK_SIZE;
2651 compile_stack.avail = 0;
2653 range_table_work.table = 0;
2654 range_table_work.allocated = 0;
2656 /* Initialize the pattern buffer. */
2657 bufp->syntax = syntax;
2658 bufp->fastmap_accurate = 0;
2659 bufp->not_bol = bufp->not_eol = 0;
2660 bufp->used_syntax = 0;
2662 /* Set `used' to zero, so that if we return an error, the pattern
2663 printer (for debugging) will think there's no pattern. We reset it
2664 at the end. */
2665 bufp->used = 0;
2667 /* Always count groups, whether or not bufp->no_sub is set. */
2668 bufp->re_nsub = 0;
2670 #if !defined emacs && !defined SYNTAX_TABLE
2671 /* Initialize the syntax table. */
2672 init_syntax_once ();
2673 #endif
2675 if (bufp->allocated == 0)
2677 if (bufp->buffer)
2678 { /* If zero allocated, but buffer is non-null, try to realloc
2679 enough space. This loses if buffer's address is bogus, but
2680 that is the user's responsibility. */
2681 RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char);
2683 else
2684 { /* Caller did not allocate a buffer. Do it for them. */
2685 bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char);
2687 if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE);
2689 bufp->allocated = INIT_BUF_SIZE;
2692 begalt = b = bufp->buffer;
2694 /* Loop through the uncompiled pattern until we're at the end. */
2695 while (1)
2697 if (p == pend)
2699 /* If this is the end of an included regexp,
2700 pop back to the main regexp and try again. */
2701 if (in_subpattern)
2703 in_subpattern = 0;
2704 pattern = main_pattern;
2705 p = main_p;
2706 pend = main_pend;
2707 continue;
2709 /* If this is the end of the main regexp, we are done. */
2710 break;
2713 PATFETCH (c);
2715 switch (c)
2717 case ' ':
2719 re_char *p1 = p;
2721 /* If there's no special whitespace regexp, treat
2722 spaces normally. And don't try to do this recursively. */
2723 if (!whitespace_regexp || in_subpattern)
2724 goto normal_char;
2726 /* Peek past following spaces. */
2727 while (p1 != pend)
2729 if (*p1 != ' ')
2730 break;
2731 p1++;
2733 /* If the spaces are followed by a repetition op,
2734 treat them normally. */
2735 if (p1 != pend
2736 && (*p1 == '*' || *p1 == '+' || *p1 == '?'
2737 || (*p1 == '\\' && p1 + 1 != pend && p1[1] == '{')))
2738 goto normal_char;
2740 /* Replace the spaces with the whitespace regexp. */
2741 in_subpattern = 1;
2742 main_p = p1;
2743 main_pend = pend;
2744 main_pattern = pattern;
2745 p = pattern = whitespace_regexp;
2746 pend = p + strlen (p);
2747 break;
2750 case '^':
2752 if ( /* If at start of pattern, it's an operator. */
2753 p == pattern + 1
2754 /* If context independent, it's an operator. */
2755 || syntax & RE_CONTEXT_INDEP_ANCHORS
2756 /* Otherwise, depends on what's come before. */
2757 || at_begline_loc_p (pattern, p, syntax))
2758 BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? begbuf : begline);
2759 else
2760 goto normal_char;
2762 break;
2765 case '$':
2767 if ( /* If at end of pattern, it's an operator. */
2768 p == pend
2769 /* If context independent, it's an operator. */
2770 || syntax & RE_CONTEXT_INDEP_ANCHORS
2771 /* Otherwise, depends on what's next. */
2772 || at_endline_loc_p (p, pend, syntax))
2773 BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? endbuf : endline);
2774 else
2775 goto normal_char;
2777 break;
2780 case '+':
2781 case '?':
2782 if ((syntax & RE_BK_PLUS_QM)
2783 || (syntax & RE_LIMITED_OPS))
2784 goto normal_char;
2785 handle_plus:
2786 case '*':
2787 /* If there is no previous pattern... */
2788 if (!laststart)
2790 if (syntax & RE_CONTEXT_INVALID_OPS)
2791 FREE_STACK_RETURN (REG_BADRPT);
2792 else if (!(syntax & RE_CONTEXT_INDEP_OPS))
2793 goto normal_char;
2797 /* 1 means zero (many) matches is allowed. */
2798 boolean zero_times_ok = 0, many_times_ok = 0;
2799 boolean greedy = 1;
2801 /* If there is a sequence of repetition chars, collapse it
2802 down to just one (the right one). We can't combine
2803 interval operators with these because of, e.g., `a{2}*',
2804 which should only match an even number of `a's. */
2806 for (;;)
2808 if ((syntax & RE_FRUGAL)
2809 && c == '?' && (zero_times_ok || many_times_ok))
2810 greedy = 0;
2811 else
2813 zero_times_ok |= c != '+';
2814 many_times_ok |= c != '?';
2817 if (p == pend)
2818 break;
2819 else if (*p == '*'
2820 || (!(syntax & RE_BK_PLUS_QM)
2821 && (*p == '+' || *p == '?')))
2823 else if (syntax & RE_BK_PLUS_QM && *p == '\\')
2825 if (p+1 == pend)
2826 FREE_STACK_RETURN (REG_EESCAPE);
2827 if (p[1] == '+' || p[1] == '?')
2828 PATFETCH (c); /* Gobble up the backslash. */
2829 else
2830 break;
2832 else
2833 break;
2834 /* If we get here, we found another repeat character. */
2835 PATFETCH (c);
2838 /* Star, etc. applied to an empty pattern is equivalent
2839 to an empty pattern. */
2840 if (!laststart || laststart == b)
2841 break;
2843 /* Now we know whether or not zero matches is allowed
2844 and also whether or not two or more matches is allowed. */
2845 if (greedy)
2847 if (many_times_ok)
2849 boolean simple = skip_one_char (laststart) == b;
2850 unsigned int startoffset = 0;
2851 re_opcode_t ofj =
2852 /* Check if the loop can match the empty string. */
2853 (simple || !analyse_first (laststart, b, NULL, 0))
2854 ? on_failure_jump : on_failure_jump_loop;
2855 assert (skip_one_char (laststart) <= b);
2857 if (!zero_times_ok && simple)
2858 { /* Since simple * loops can be made faster by using
2859 on_failure_keep_string_jump, we turn simple P+
2860 into PP* if P is simple. */
2861 unsigned char *p1, *p2;
2862 startoffset = b - laststart;
2863 GET_BUFFER_SPACE (startoffset);
2864 p1 = b; p2 = laststart;
2865 while (p2 < p1)
2866 *b++ = *p2++;
2867 zero_times_ok = 1;
2870 GET_BUFFER_SPACE (6);
2871 if (!zero_times_ok)
2872 /* A + loop. */
2873 STORE_JUMP (ofj, b, b + 6);
2874 else
2875 /* Simple * loops can use on_failure_keep_string_jump
2876 depending on what follows. But since we don't know
2877 that yet, we leave the decision up to
2878 on_failure_jump_smart. */
2879 INSERT_JUMP (simple ? on_failure_jump_smart : ofj,
2880 laststart + startoffset, b + 6);
2881 b += 3;
2882 STORE_JUMP (jump, b, laststart + startoffset);
2883 b += 3;
2885 else
2887 /* A simple ? pattern. */
2888 assert (zero_times_ok);
2889 GET_BUFFER_SPACE (3);
2890 INSERT_JUMP (on_failure_jump, laststart, b + 3);
2891 b += 3;
2894 else /* not greedy */
2895 { /* I wish the greedy and non-greedy cases could be merged. */
2897 GET_BUFFER_SPACE (7); /* We might use less. */
2898 if (many_times_ok)
2900 boolean emptyp = analyse_first (laststart, b, NULL, 0);
2902 /* The non-greedy multiple match looks like
2903 a repeat..until: we only need a conditional jump
2904 at the end of the loop. */
2905 if (emptyp) BUF_PUSH (no_op);
2906 STORE_JUMP (emptyp ? on_failure_jump_nastyloop
2907 : on_failure_jump, b, laststart);
2908 b += 3;
2909 if (zero_times_ok)
2911 /* The repeat...until naturally matches one or more.
2912 To also match zero times, we need to first jump to
2913 the end of the loop (its conditional jump). */
2914 INSERT_JUMP (jump, laststart, b);
2915 b += 3;
2918 else
2920 /* non-greedy a?? */
2921 INSERT_JUMP (jump, laststart, b + 3);
2922 b += 3;
2923 INSERT_JUMP (on_failure_jump, laststart, laststart + 6);
2924 b += 3;
2928 pending_exact = 0;
2929 break;
2932 case '.':
2933 laststart = b;
2934 BUF_PUSH (anychar);
2935 break;
2938 case '[':
2940 CLEAR_RANGE_TABLE_WORK_USED (range_table_work);
2942 if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
2944 /* Ensure that we have enough space to push a charset: the
2945 opcode, the length count, and the bitset; 34 bytes in all. */
2946 GET_BUFFER_SPACE (34);
2948 laststart = b;
2950 /* We test `*p == '^' twice, instead of using an if
2951 statement, so we only need one BUF_PUSH. */
2952 BUF_PUSH (*p == '^' ? charset_not : charset);
2953 if (*p == '^')
2954 p++;
2956 /* Remember the first position in the bracket expression. */
2957 p1 = p;
2959 /* Push the number of bytes in the bitmap. */
2960 BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH);
2962 /* Clear the whole map. */
2963 bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH);
2965 /* charset_not matches newline according to a syntax bit. */
2966 if ((re_opcode_t) b[-2] == charset_not
2967 && (syntax & RE_HAT_LISTS_NOT_NEWLINE))
2968 SET_LIST_BIT ('\n');
2970 /* Read in characters and ranges, setting map bits. */
2971 for (;;)
2973 boolean escaped_char = false;
2974 const unsigned char *p2 = p;
2975 re_wchar_t ch, c2;
2977 if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
2979 /* Don't translate yet. The range TRANSLATE(X..Y) cannot
2980 always be determined from TRANSLATE(X) and TRANSLATE(Y)
2981 So the translation is done later in a loop. Example:
2982 (let ((case-fold-search t)) (string-match "[A-_]" "A")) */
2983 PATFETCH (c);
2985 /* \ might escape characters inside [...] and [^...]. */
2986 if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\')
2988 if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
2990 PATFETCH (c);
2991 escaped_char = true;
2993 else
2995 /* Could be the end of the bracket expression. If it's
2996 not (i.e., when the bracket expression is `[]' so
2997 far), the ']' character bit gets set way below. */
2998 if (c == ']' && p2 != p1)
2999 break;
3002 /* See if we're at the beginning of a possible character
3003 class. */
3005 if (!escaped_char &&
3006 syntax & RE_CHAR_CLASSES && c == '[' && *p == ':')
3008 /* Leave room for the null. */
3009 unsigned char str[CHAR_CLASS_MAX_LENGTH + 1];
3010 const unsigned char *class_beg;
3012 PATFETCH (c);
3013 c1 = 0;
3014 class_beg = p;
3016 /* If pattern is `[[:'. */
3017 if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
3019 for (;;)
3021 PATFETCH (c);
3022 if ((c == ':' && *p == ']') || p == pend)
3023 break;
3024 if (c1 < CHAR_CLASS_MAX_LENGTH)
3025 str[c1++] = c;
3026 else
3027 /* This is in any case an invalid class name. */
3028 str[0] = '\0';
3030 str[c1] = '\0';
3032 /* If isn't a word bracketed by `[:' and `:]':
3033 undo the ending character, the letters, and
3034 leave the leading `:' and `[' (but set bits for
3035 them). */
3036 if (c == ':' && *p == ']')
3038 re_wctype_t cc;
3039 int limit;
3041 cc = re_wctype (str);
3043 if (cc == 0)
3044 FREE_STACK_RETURN (REG_ECTYPE);
3046 /* Throw away the ] at the end of the character
3047 class. */
3048 PATFETCH (c);
3050 if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
3052 #ifndef emacs
3053 for (ch = 0; ch < (1 << BYTEWIDTH); ++ch)
3054 if (re_iswctype (btowc (ch), cc))
3056 c = TRANSLATE (ch);
3057 if (c < (1 << BYTEWIDTH))
3058 SET_LIST_BIT (c);
3060 #else /* emacs */
3061 /* Most character classes in a multibyte match
3062 just set a flag. Exceptions are is_blank,
3063 is_digit, is_cntrl, and is_xdigit, since
3064 they can only match ASCII characters. We
3065 don't need to handle them for multibyte.
3066 They are distinguished by a negative wctype. */
3068 /* Setup the gl_state object to its buffer-defined
3069 value. This hardcodes the buffer-global
3070 syntax-table for ASCII chars, while the other chars
3071 will obey syntax-table properties. It's not ideal,
3072 but it's the way it's been done until now. */
3073 SETUP_SYNTAX_TABLE (BEGV, 0);
3075 for (ch = 0; ch < 256; ++ch)
3077 c = RE_CHAR_TO_MULTIBYTE (ch);
3078 if (! CHAR_BYTE8_P (c)
3079 && re_iswctype (c, cc))
3081 SET_LIST_BIT (ch);
3082 c1 = TRANSLATE (c);
3083 if (c1 == c)
3084 continue;
3085 if (ASCII_CHAR_P (c1))
3086 SET_LIST_BIT (c1);
3087 else if ((c1 = RE_CHAR_TO_UNIBYTE (c1)) >= 0)
3088 SET_LIST_BIT (c1);
3091 SET_RANGE_TABLE_WORK_AREA_BIT
3092 (range_table_work, re_wctype_to_bit (cc));
3093 #endif /* emacs */
3094 /* In most cases the matching rule for char classes
3095 only uses the syntax table for multibyte chars,
3096 so that the content of the syntax-table it is not
3097 hardcoded in the range_table. SPACE and WORD are
3098 the two exceptions. */
3099 if ((1 << cc) & ((1 << RECC_SPACE) | (1 << RECC_WORD)))
3100 bufp->used_syntax = 1;
3102 /* Repeat the loop. */
3103 continue;
3105 else
3107 /* Go back to right after the "[:". */
3108 p = class_beg;
3109 SET_LIST_BIT ('[');
3111 /* Because the `:' may starts the range, we
3112 can't simply set bit and repeat the loop.
3113 Instead, just set it to C and handle below. */
3114 c = ':';
3118 if (p < pend && p[0] == '-' && p[1] != ']')
3121 /* Discard the `-'. */
3122 PATFETCH (c1);
3124 /* Fetch the character which ends the range. */
3125 PATFETCH (c1);
3126 #ifdef emacs
3127 if (CHAR_BYTE8_P (c1)
3128 && ! ASCII_CHAR_P (c) && ! CHAR_BYTE8_P (c))
3129 /* Treat the range from a multibyte character to
3130 raw-byte character as empty. */
3131 c = c1 + 1;
3132 #endif /* emacs */
3134 else
3135 /* Range from C to C. */
3136 c1 = c;
3138 if (c > c1)
3140 if (syntax & RE_NO_EMPTY_RANGES)
3141 FREE_STACK_RETURN (REG_ERANGEX);
3142 /* Else, repeat the loop. */
3144 else
3146 #ifndef emacs
3147 /* Set the range into bitmap */
3148 for (; c <= c1; c++)
3150 ch = TRANSLATE (c);
3151 if (ch < (1 << BYTEWIDTH))
3152 SET_LIST_BIT (ch);
3154 #else /* emacs */
3155 if (c < 128)
3157 ch = MIN (127, c1);
3158 SETUP_ASCII_RANGE (range_table_work, c, ch);
3159 c = ch + 1;
3160 if (CHAR_BYTE8_P (c1))
3161 c = BYTE8_TO_CHAR (128);
3163 if (c <= c1)
3165 if (CHAR_BYTE8_P (c))
3167 c = CHAR_TO_BYTE8 (c);
3168 c1 = CHAR_TO_BYTE8 (c1);
3169 for (; c <= c1; c++)
3170 SET_LIST_BIT (c);
3172 else if (multibyte)
3174 SETUP_MULTIBYTE_RANGE (range_table_work, c, c1);
3176 else
3178 SETUP_UNIBYTE_RANGE (range_table_work, c, c1);
3181 #endif /* emacs */
3185 /* Discard any (non)matching list bytes that are all 0 at the
3186 end of the map. Decrease the map-length byte too. */
3187 while ((int) b[-1] > 0 && b[b[-1] - 1] == 0)
3188 b[-1]--;
3189 b += b[-1];
3191 /* Build real range table from work area. */
3192 if (RANGE_TABLE_WORK_USED (range_table_work)
3193 || RANGE_TABLE_WORK_BITS (range_table_work))
3195 int i;
3196 int used = RANGE_TABLE_WORK_USED (range_table_work);
3198 /* Allocate space for COUNT + RANGE_TABLE. Needs two
3199 bytes for flags, two for COUNT, and three bytes for
3200 each character. */
3201 GET_BUFFER_SPACE (4 + used * 3);
3203 /* Indicate the existence of range table. */
3204 laststart[1] |= 0x80;
3206 /* Store the character class flag bits into the range table.
3207 If not in emacs, these flag bits are always 0. */
3208 *b++ = RANGE_TABLE_WORK_BITS (range_table_work) & 0xff;
3209 *b++ = RANGE_TABLE_WORK_BITS (range_table_work) >> 8;
3211 STORE_NUMBER_AND_INCR (b, used / 2);
3212 for (i = 0; i < used; i++)
3213 STORE_CHARACTER_AND_INCR
3214 (b, RANGE_TABLE_WORK_ELT (range_table_work, i));
3217 break;
3220 case '(':
3221 if (syntax & RE_NO_BK_PARENS)
3222 goto handle_open;
3223 else
3224 goto normal_char;
3227 case ')':
3228 if (syntax & RE_NO_BK_PARENS)
3229 goto handle_close;
3230 else
3231 goto normal_char;
3234 case '\n':
3235 if (syntax & RE_NEWLINE_ALT)
3236 goto handle_alt;
3237 else
3238 goto normal_char;
3241 case '|':
3242 if (syntax & RE_NO_BK_VBAR)
3243 goto handle_alt;
3244 else
3245 goto normal_char;
3248 case '{':
3249 if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES)
3250 goto handle_interval;
3251 else
3252 goto normal_char;
3255 case '\\':
3256 if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
3258 /* Do not translate the character after the \, so that we can
3259 distinguish, e.g., \B from \b, even if we normally would
3260 translate, e.g., B to b. */
3261 PATFETCH (c);
3263 switch (c)
3265 case '(':
3266 if (syntax & RE_NO_BK_PARENS)
3267 goto normal_backslash;
3269 handle_open:
3271 int shy = 0;
3272 regnum_t regnum = 0;
3273 if (p+1 < pend)
3275 /* Look for a special (?...) construct */
3276 if ((syntax & RE_SHY_GROUPS) && *p == '?')
3278 PATFETCH (c); /* Gobble up the '?'. */
3279 while (!shy)
3281 PATFETCH (c);
3282 switch (c)
3284 case ':': shy = 1; break;
3285 case '0':
3286 /* An explicitly specified regnum must start
3287 with non-0. */
3288 if (regnum == 0)
3289 FREE_STACK_RETURN (REG_BADPAT);
3290 case '1': case '2': case '3': case '4':
3291 case '5': case '6': case '7': case '8': case '9':
3292 regnum = 10*regnum + (c - '0'); break;
3293 default:
3294 /* Only (?:...) is supported right now. */
3295 FREE_STACK_RETURN (REG_BADPAT);
3301 if (!shy)
3302 regnum = ++bufp->re_nsub;
3303 else if (regnum)
3304 { /* It's actually not shy, but explicitly numbered. */
3305 shy = 0;
3306 if (regnum > bufp->re_nsub)
3307 bufp->re_nsub = regnum;
3308 else if (regnum > bufp->re_nsub
3309 /* Ideally, we'd want to check that the specified
3310 group can't have matched (i.e. all subgroups
3311 using the same regnum are in other branches of
3312 OR patterns), but we don't currently keep track
3313 of enough info to do that easily. */
3314 || group_in_compile_stack (compile_stack, regnum))
3315 FREE_STACK_RETURN (REG_BADPAT);
3317 else
3318 /* It's really shy. */
3319 regnum = - bufp->re_nsub;
3321 if (COMPILE_STACK_FULL)
3323 RETALLOC (compile_stack.stack, compile_stack.size << 1,
3324 compile_stack_elt_t);
3325 if (compile_stack.stack == NULL) return REG_ESPACE;
3327 compile_stack.size <<= 1;
3330 /* These are the values to restore when we hit end of this
3331 group. They are all relative offsets, so that if the
3332 whole pattern moves because of realloc, they will still
3333 be valid. */
3334 COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer;
3335 COMPILE_STACK_TOP.fixup_alt_jump
3336 = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0;
3337 COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer;
3338 COMPILE_STACK_TOP.regnum = regnum;
3340 /* Do not push a start_memory for groups beyond the last one
3341 we can represent in the compiled pattern. */
3342 if (regnum <= MAX_REGNUM && regnum > 0)
3343 BUF_PUSH_2 (start_memory, regnum);
3345 compile_stack.avail++;
3347 fixup_alt_jump = 0;
3348 laststart = 0;
3349 begalt = b;
3350 /* If we've reached MAX_REGNUM groups, then this open
3351 won't actually generate any code, so we'll have to
3352 clear pending_exact explicitly. */
3353 pending_exact = 0;
3354 break;
3357 case ')':
3358 if (syntax & RE_NO_BK_PARENS) goto normal_backslash;
3360 if (COMPILE_STACK_EMPTY)
3362 if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
3363 goto normal_backslash;
3364 else
3365 FREE_STACK_RETURN (REG_ERPAREN);
3368 handle_close:
3369 FIXUP_ALT_JUMP ();
3371 /* See similar code for backslashed left paren above. */
3372 if (COMPILE_STACK_EMPTY)
3374 if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
3375 goto normal_char;
3376 else
3377 FREE_STACK_RETURN (REG_ERPAREN);
3380 /* Since we just checked for an empty stack above, this
3381 ``can't happen''. */
3382 assert (compile_stack.avail != 0);
3384 /* We don't just want to restore into `regnum', because
3385 later groups should continue to be numbered higher,
3386 as in `(ab)c(de)' -- the second group is #2. */
3387 regnum_t regnum;
3389 compile_stack.avail--;
3390 begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset;
3391 fixup_alt_jump
3392 = COMPILE_STACK_TOP.fixup_alt_jump
3393 ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1
3394 : 0;
3395 laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset;
3396 regnum = COMPILE_STACK_TOP.regnum;
3397 /* If we've reached MAX_REGNUM groups, then this open
3398 won't actually generate any code, so we'll have to
3399 clear pending_exact explicitly. */
3400 pending_exact = 0;
3402 /* We're at the end of the group, so now we know how many
3403 groups were inside this one. */
3404 if (regnum <= MAX_REGNUM && regnum > 0)
3405 BUF_PUSH_2 (stop_memory, regnum);
3407 break;
3410 case '|': /* `\|'. */
3411 if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR)
3412 goto normal_backslash;
3413 handle_alt:
3414 if (syntax & RE_LIMITED_OPS)
3415 goto normal_char;
3417 /* Insert before the previous alternative a jump which
3418 jumps to this alternative if the former fails. */
3419 GET_BUFFER_SPACE (3);
3420 INSERT_JUMP (on_failure_jump, begalt, b + 6);
3421 pending_exact = 0;
3422 b += 3;
3424 /* The alternative before this one has a jump after it
3425 which gets executed if it gets matched. Adjust that
3426 jump so it will jump to this alternative's analogous
3427 jump (put in below, which in turn will jump to the next
3428 (if any) alternative's such jump, etc.). The last such
3429 jump jumps to the correct final destination. A picture:
3430 _____ _____
3431 | | | |
3432 | v | v
3433 a | b | c
3435 If we are at `b', then fixup_alt_jump right now points to a
3436 three-byte space after `a'. We'll put in the jump, set
3437 fixup_alt_jump to right after `b', and leave behind three
3438 bytes which we'll fill in when we get to after `c'. */
3440 FIXUP_ALT_JUMP ();
3442 /* Mark and leave space for a jump after this alternative,
3443 to be filled in later either by next alternative or
3444 when know we're at the end of a series of alternatives. */
3445 fixup_alt_jump = b;
3446 GET_BUFFER_SPACE (3);
3447 b += 3;
3449 laststart = 0;
3450 begalt = b;
3451 break;
3454 case '{':
3455 /* If \{ is a literal. */
3456 if (!(syntax & RE_INTERVALS)
3457 /* If we're at `\{' and it's not the open-interval
3458 operator. */
3459 || (syntax & RE_NO_BK_BRACES))
3460 goto normal_backslash;
3462 handle_interval:
3464 /* If got here, then the syntax allows intervals. */
3466 /* At least (most) this many matches must be made. */
3467 int lower_bound = 0, upper_bound = -1;
3469 beg_interval = p;
3471 GET_UNSIGNED_NUMBER (lower_bound);
3473 if (c == ',')
3474 GET_UNSIGNED_NUMBER (upper_bound);
3475 else
3476 /* Interval such as `{1}' => match exactly once. */
3477 upper_bound = lower_bound;
3479 if (lower_bound < 0 || upper_bound > RE_DUP_MAX
3480 || (upper_bound >= 0 && lower_bound > upper_bound))
3481 FREE_STACK_RETURN (REG_BADBR);
3483 if (!(syntax & RE_NO_BK_BRACES))
3485 if (c != '\\')
3486 FREE_STACK_RETURN (REG_BADBR);
3487 if (p == pend)
3488 FREE_STACK_RETURN (REG_EESCAPE);
3489 PATFETCH (c);
3492 if (c != '}')
3493 FREE_STACK_RETURN (REG_BADBR);
3495 /* We just parsed a valid interval. */
3497 /* If it's invalid to have no preceding re. */
3498 if (!laststart)
3500 if (syntax & RE_CONTEXT_INVALID_OPS)
3501 FREE_STACK_RETURN (REG_BADRPT);
3502 else if (syntax & RE_CONTEXT_INDEP_OPS)
3503 laststart = b;
3504 else
3505 goto unfetch_interval;
3508 if (upper_bound == 0)
3509 /* If the upper bound is zero, just drop the sub pattern
3510 altogether. */
3511 b = laststart;
3512 else if (lower_bound == 1 && upper_bound == 1)
3513 /* Just match it once: nothing to do here. */
3516 /* Otherwise, we have a nontrivial interval. When
3517 we're all done, the pattern will look like:
3518 set_number_at <jump count> <upper bound>
3519 set_number_at <succeed_n count> <lower bound>
3520 succeed_n <after jump addr> <succeed_n count>
3521 <body of loop>
3522 jump_n <succeed_n addr> <jump count>
3523 (The upper bound and `jump_n' are omitted if
3524 `upper_bound' is 1, though.) */
3525 else
3526 { /* If the upper bound is > 1, we need to insert
3527 more at the end of the loop. */
3528 unsigned int nbytes = (upper_bound < 0 ? 3
3529 : upper_bound > 1 ? 5 : 0);
3530 unsigned int startoffset = 0;
3532 GET_BUFFER_SPACE (20); /* We might use less. */
3534 if (lower_bound == 0)
3536 /* A succeed_n that starts with 0 is really a
3537 a simple on_failure_jump_loop. */
3538 INSERT_JUMP (on_failure_jump_loop, laststart,
3539 b + 3 + nbytes);
3540 b += 3;
3542 else
3544 /* Initialize lower bound of the `succeed_n', even
3545 though it will be set during matching by its
3546 attendant `set_number_at' (inserted next),
3547 because `re_compile_fastmap' needs to know.
3548 Jump to the `jump_n' we might insert below. */
3549 INSERT_JUMP2 (succeed_n, laststart,
3550 b + 5 + nbytes,
3551 lower_bound);
3552 b += 5;
3554 /* Code to initialize the lower bound. Insert
3555 before the `succeed_n'. The `5' is the last two
3556 bytes of this `set_number_at', plus 3 bytes of
3557 the following `succeed_n'. */
3558 insert_op2 (set_number_at, laststart, 5, lower_bound, b);
3559 b += 5;
3560 startoffset += 5;
3563 if (upper_bound < 0)
3565 /* A negative upper bound stands for infinity,
3566 in which case it degenerates to a plain jump. */
3567 STORE_JUMP (jump, b, laststart + startoffset);
3568 b += 3;
3570 else if (upper_bound > 1)
3571 { /* More than one repetition is allowed, so
3572 append a backward jump to the `succeed_n'
3573 that starts this interval.
3575 When we've reached this during matching,
3576 we'll have matched the interval once, so
3577 jump back only `upper_bound - 1' times. */
3578 STORE_JUMP2 (jump_n, b, laststart + startoffset,
3579 upper_bound - 1);
3580 b += 5;
3582 /* The location we want to set is the second
3583 parameter of the `jump_n'; that is `b-2' as
3584 an absolute address. `laststart' will be
3585 the `set_number_at' we're about to insert;
3586 `laststart+3' the number to set, the source
3587 for the relative address. But we are
3588 inserting into the middle of the pattern --
3589 so everything is getting moved up by 5.
3590 Conclusion: (b - 2) - (laststart + 3) + 5,
3591 i.e., b - laststart.
3593 We insert this at the beginning of the loop
3594 so that if we fail during matching, we'll
3595 reinitialize the bounds. */
3596 insert_op2 (set_number_at, laststart, b - laststart,
3597 upper_bound - 1, b);
3598 b += 5;
3601 pending_exact = 0;
3602 beg_interval = NULL;
3604 break;
3606 unfetch_interval:
3607 /* If an invalid interval, match the characters as literals. */
3608 assert (beg_interval);
3609 p = beg_interval;
3610 beg_interval = NULL;
3612 /* normal_char and normal_backslash need `c'. */
3613 c = '{';
3615 if (!(syntax & RE_NO_BK_BRACES))
3617 assert (p > pattern && p[-1] == '\\');
3618 goto normal_backslash;
3620 else
3621 goto normal_char;
3623 #ifdef emacs
3624 /* There is no way to specify the before_dot and after_dot
3625 operators. rms says this is ok. --karl */
3626 case '=':
3627 BUF_PUSH (at_dot);
3628 break;
3630 case 's':
3631 laststart = b;
3632 PATFETCH (c);
3633 BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]);
3634 break;
3636 case 'S':
3637 laststart = b;
3638 PATFETCH (c);
3639 BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]);
3640 break;
3642 case 'c':
3643 laststart = b;
3644 PATFETCH (c);
3645 BUF_PUSH_2 (categoryspec, c);
3646 break;
3648 case 'C':
3649 laststart = b;
3650 PATFETCH (c);
3651 BUF_PUSH_2 (notcategoryspec, c);
3652 break;
3653 #endif /* emacs */
3656 case 'w':
3657 if (syntax & RE_NO_GNU_OPS)
3658 goto normal_char;
3659 laststart = b;
3660 BUF_PUSH_2 (syntaxspec, Sword);
3661 break;
3664 case 'W':
3665 if (syntax & RE_NO_GNU_OPS)
3666 goto normal_char;
3667 laststart = b;
3668 BUF_PUSH_2 (notsyntaxspec, Sword);
3669 break;
3672 case '<':
3673 if (syntax & RE_NO_GNU_OPS)
3674 goto normal_char;
3675 BUF_PUSH (wordbeg);
3676 break;
3678 case '>':
3679 if (syntax & RE_NO_GNU_OPS)
3680 goto normal_char;
3681 BUF_PUSH (wordend);
3682 break;
3684 case '_':
3685 if (syntax & RE_NO_GNU_OPS)
3686 goto normal_char;
3687 laststart = b;
3688 PATFETCH (c);
3689 if (c == '<')
3690 BUF_PUSH (symbeg);
3691 else if (c == '>')
3692 BUF_PUSH (symend);
3693 else
3694 FREE_STACK_RETURN (REG_BADPAT);
3695 break;
3697 case 'b':
3698 if (syntax & RE_NO_GNU_OPS)
3699 goto normal_char;
3700 BUF_PUSH (wordbound);
3701 break;
3703 case 'B':
3704 if (syntax & RE_NO_GNU_OPS)
3705 goto normal_char;
3706 BUF_PUSH (notwordbound);
3707 break;
3709 case '`':
3710 if (syntax & RE_NO_GNU_OPS)
3711 goto normal_char;
3712 BUF_PUSH (begbuf);
3713 break;
3715 case '\'':
3716 if (syntax & RE_NO_GNU_OPS)
3717 goto normal_char;
3718 BUF_PUSH (endbuf);
3719 break;
3721 case '1': case '2': case '3': case '4': case '5':
3722 case '6': case '7': case '8': case '9':
3724 regnum_t reg;
3726 if (syntax & RE_NO_BK_REFS)
3727 goto normal_backslash;
3729 reg = c - '0';
3731 if (reg > bufp->re_nsub || reg < 1
3732 /* Can't back reference to a subexp before its end. */
3733 || group_in_compile_stack (compile_stack, reg))
3734 FREE_STACK_RETURN (REG_ESUBREG);
3736 laststart = b;
3737 BUF_PUSH_2 (duplicate, reg);
3739 break;
3742 case '+':
3743 case '?':
3744 if (syntax & RE_BK_PLUS_QM)
3745 goto handle_plus;
3746 else
3747 goto normal_backslash;
3749 default:
3750 normal_backslash:
3751 /* You might think it would be useful for \ to mean
3752 not to translate; but if we don't translate it
3753 it will never match anything. */
3754 goto normal_char;
3756 break;
3759 default:
3760 /* Expects the character in `c'. */
3761 normal_char:
3762 /* If no exactn currently being built. */
3763 if (!pending_exact
3765 /* If last exactn not at current position. */
3766 || pending_exact + *pending_exact + 1 != b
3768 /* We have only one byte following the exactn for the count. */
3769 || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH
3771 /* If followed by a repetition operator. */
3772 || (p != pend && (*p == '*' || *p == '^'))
3773 || ((syntax & RE_BK_PLUS_QM)
3774 ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?')
3775 : p != pend && (*p == '+' || *p == '?'))
3776 || ((syntax & RE_INTERVALS)
3777 && ((syntax & RE_NO_BK_BRACES)
3778 ? p != pend && *p == '{'
3779 : p + 1 < pend && p[0] == '\\' && p[1] == '{')))
3781 /* Start building a new exactn. */
3783 laststart = b;
3785 BUF_PUSH_2 (exactn, 0);
3786 pending_exact = b - 1;
3789 GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH);
3791 int len;
3793 if (multibyte)
3795 c = TRANSLATE (c);
3796 len = CHAR_STRING (c, b);
3797 b += len;
3799 else
3801 c1 = RE_CHAR_TO_MULTIBYTE (c);
3802 if (! CHAR_BYTE8_P (c1))
3804 re_wchar_t c2 = TRANSLATE (c1);
3806 if (c1 != c2 && (c1 = RE_CHAR_TO_UNIBYTE (c2)) >= 0)
3807 c = c1;
3809 *b++ = c;
3810 len = 1;
3812 (*pending_exact) += len;
3815 break;
3816 } /* switch (c) */
3817 } /* while p != pend */
3820 /* Through the pattern now. */
3822 FIXUP_ALT_JUMP ();
3824 if (!COMPILE_STACK_EMPTY)
3825 FREE_STACK_RETURN (REG_EPAREN);
3827 /* If we don't want backtracking, force success
3828 the first time we reach the end of the compiled pattern. */
3829 if (syntax & RE_NO_POSIX_BACKTRACKING)
3830 BUF_PUSH (succeed);
3832 /* We have succeeded; set the length of the buffer. */
3833 bufp->used = b - bufp->buffer;
3835 #ifdef DEBUG
3836 if (debug > 0)
3838 re_compile_fastmap (bufp);
3839 DEBUG_PRINT1 ("\nCompiled pattern: \n");
3840 print_compiled_pattern (bufp);
3842 debug--;
3843 #endif /* DEBUG */
3845 #ifndef MATCH_MAY_ALLOCATE
3846 /* Initialize the failure stack to the largest possible stack. This
3847 isn't necessary unless we're trying to avoid calling alloca in
3848 the search and match routines. */
3850 int num_regs = bufp->re_nsub + 1;
3852 if (fail_stack.size < re_max_failures * TYPICAL_FAILURE_SIZE)
3854 fail_stack.size = re_max_failures * TYPICAL_FAILURE_SIZE;
3856 if (! fail_stack.stack)
3857 fail_stack.stack
3858 = (fail_stack_elt_t *) malloc (fail_stack.size
3859 * sizeof (fail_stack_elt_t));
3860 else
3861 fail_stack.stack
3862 = (fail_stack_elt_t *) realloc (fail_stack.stack,
3863 (fail_stack.size
3864 * sizeof (fail_stack_elt_t)));
3867 regex_grow_registers (num_regs);
3869 #endif /* not MATCH_MAY_ALLOCATE */
3871 FREE_STACK_RETURN (REG_NOERROR);
3872 } /* regex_compile */
3874 /* Subroutines for `regex_compile'. */
3876 /* Store OP at LOC followed by two-byte integer parameter ARG. */
3878 static void
3879 store_op1 (op, loc, arg)
3880 re_opcode_t op;
3881 unsigned char *loc;
3882 int arg;
3884 *loc = (unsigned char) op;
3885 STORE_NUMBER (loc + 1, arg);
3889 /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */
3891 static void
3892 store_op2 (op, loc, arg1, arg2)
3893 re_opcode_t op;
3894 unsigned char *loc;
3895 int arg1, arg2;
3897 *loc = (unsigned char) op;
3898 STORE_NUMBER (loc + 1, arg1);
3899 STORE_NUMBER (loc + 3, arg2);
3903 /* Copy the bytes from LOC to END to open up three bytes of space at LOC
3904 for OP followed by two-byte integer parameter ARG. */
3906 static void
3907 insert_op1 (op, loc, arg, end)
3908 re_opcode_t op;
3909 unsigned char *loc;
3910 int arg;
3911 unsigned char *end;
3913 register unsigned char *pfrom = end;
3914 register unsigned char *pto = end + 3;
3916 while (pfrom != loc)
3917 *--pto = *--pfrom;
3919 store_op1 (op, loc, arg);
3923 /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */
3925 static void
3926 insert_op2 (op, loc, arg1, arg2, end)
3927 re_opcode_t op;
3928 unsigned char *loc;
3929 int arg1, arg2;
3930 unsigned char *end;
3932 register unsigned char *pfrom = end;
3933 register unsigned char *pto = end + 5;
3935 while (pfrom != loc)
3936 *--pto = *--pfrom;
3938 store_op2 (op, loc, arg1, arg2);
3942 /* P points to just after a ^ in PATTERN. Return true if that ^ comes
3943 after an alternative or a begin-subexpression. We assume there is at
3944 least one character before the ^. */
3946 static boolean
3947 at_begline_loc_p (pattern, p, syntax)
3948 re_char *pattern, *p;
3949 reg_syntax_t syntax;
3951 re_char *prev = p - 2;
3952 boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\';
3954 return
3955 /* After a subexpression? */
3956 (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash))
3957 /* After an alternative? */
3958 || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash))
3959 /* After a shy subexpression? */
3960 || ((syntax & RE_SHY_GROUPS) && prev - 2 >= pattern
3961 && prev[-1] == '?' && prev[-2] == '('
3962 && (syntax & RE_NO_BK_PARENS
3963 || (prev - 3 >= pattern && prev[-3] == '\\')));
3967 /* The dual of at_begline_loc_p. This one is for $. We assume there is
3968 at least one character after the $, i.e., `P < PEND'. */
3970 static boolean
3971 at_endline_loc_p (p, pend, syntax)
3972 re_char *p, *pend;
3973 reg_syntax_t syntax;
3975 re_char *next = p;
3976 boolean next_backslash = *next == '\\';
3977 re_char *next_next = p + 1 < pend ? p + 1 : 0;
3979 return
3980 /* Before a subexpression? */
3981 (syntax & RE_NO_BK_PARENS ? *next == ')'
3982 : next_backslash && next_next && *next_next == ')')
3983 /* Before an alternative? */
3984 || (syntax & RE_NO_BK_VBAR ? *next == '|'
3985 : next_backslash && next_next && *next_next == '|');
3989 /* Returns true if REGNUM is in one of COMPILE_STACK's elements and
3990 false if it's not. */
3992 static boolean
3993 group_in_compile_stack (compile_stack, regnum)
3994 compile_stack_type compile_stack;
3995 regnum_t regnum;
3997 int this_element;
3999 for (this_element = compile_stack.avail - 1;
4000 this_element >= 0;
4001 this_element--)
4002 if (compile_stack.stack[this_element].regnum == regnum)
4003 return true;
4005 return false;
4008 /* analyse_first.
4009 If fastmap is non-NULL, go through the pattern and fill fastmap
4010 with all the possible leading chars. If fastmap is NULL, don't
4011 bother filling it up (obviously) and only return whether the
4012 pattern could potentially match the empty string.
4014 Return 1 if p..pend might match the empty string.
4015 Return 0 if p..pend matches at least one char.
4016 Return -1 if fastmap was not updated accurately. */
4018 static int
4019 analyse_first (p, pend, fastmap, multibyte)
4020 re_char *p, *pend;
4021 char *fastmap;
4022 const int multibyte;
4024 int j, k;
4025 boolean not;
4027 /* If all elements for base leading-codes in fastmap is set, this
4028 flag is set true. */
4029 boolean match_any_multibyte_characters = false;
4031 assert (p);
4033 /* The loop below works as follows:
4034 - It has a working-list kept in the PATTERN_STACK and which basically
4035 starts by only containing a pointer to the first operation.
4036 - If the opcode we're looking at is a match against some set of
4037 chars, then we add those chars to the fastmap and go on to the
4038 next work element from the worklist (done via `break').
4039 - If the opcode is a control operator on the other hand, we either
4040 ignore it (if it's meaningless at this point, such as `start_memory')
4041 or execute it (if it's a jump). If the jump has several destinations
4042 (i.e. `on_failure_jump'), then we push the other destination onto the
4043 worklist.
4044 We guarantee termination by ignoring backward jumps (more or less),
4045 so that `p' is monotonically increasing. More to the point, we
4046 never set `p' (or push) anything `<= p1'. */
4048 while (p < pend)
4050 /* `p1' is used as a marker of how far back a `on_failure_jump'
4051 can go without being ignored. It is normally equal to `p'
4052 (which prevents any backward `on_failure_jump') except right
4053 after a plain `jump', to allow patterns such as:
4054 0: jump 10
4055 3..9: <body>
4056 10: on_failure_jump 3
4057 as used for the *? operator. */
4058 re_char *p1 = p;
4060 switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
4062 case succeed:
4063 return 1;
4064 continue;
4066 case duplicate:
4067 /* If the first character has to match a backreference, that means
4068 that the group was empty (since it already matched). Since this
4069 is the only case that interests us here, we can assume that the
4070 backreference must match the empty string. */
4071 p++;
4072 continue;
4075 /* Following are the cases which match a character. These end
4076 with `break'. */
4078 case exactn:
4079 if (fastmap)
4081 /* If multibyte is nonzero, the first byte of each
4082 character is an ASCII or a leading code. Otherwise,
4083 each byte is a character. Thus, this works in both
4084 cases. */
4085 fastmap[p[1]] = 1;
4086 if (! multibyte)
4088 /* For the case of matching this unibyte regex
4089 against multibyte, we must set a leading code of
4090 the corresponding multibyte character. */
4091 int c = RE_CHAR_TO_MULTIBYTE (p[1]);
4093 fastmap[CHAR_LEADING_CODE (c)] = 1;
4096 break;
4099 case anychar:
4100 /* We could put all the chars except for \n (and maybe \0)
4101 but we don't bother since it is generally not worth it. */
4102 if (!fastmap) break;
4103 return -1;
4106 case charset_not:
4107 if (!fastmap) break;
4109 /* Chars beyond end of bitmap are possible matches. */
4110 for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH;
4111 j < (1 << BYTEWIDTH); j++)
4112 fastmap[j] = 1;
4115 /* Fallthrough */
4116 case charset:
4117 if (!fastmap) break;
4118 not = (re_opcode_t) *(p - 1) == charset_not;
4119 for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH - 1, p++;
4120 j >= 0; j--)
4121 if (!!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) ^ not)
4122 fastmap[j] = 1;
4124 #ifdef emacs
4125 if (/* Any leading code can possibly start a character
4126 which doesn't match the specified set of characters. */
4129 /* If we can match a character class, we can match any
4130 multibyte characters. */
4131 (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2])
4132 && CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0))
4135 if (match_any_multibyte_characters == false)
4137 for (j = MIN_MULTIBYTE_LEADING_CODE;
4138 j <= MAX_MULTIBYTE_LEADING_CODE; j++)
4139 fastmap[j] = 1;
4140 match_any_multibyte_characters = true;
4144 else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2])
4145 && match_any_multibyte_characters == false)
4147 /* Set fastmap[I] to 1 where I is a leading code of each
4148 multibyte characer in the range table. */
4149 int c, count;
4150 unsigned char lc1, lc2;
4152 /* Make P points the range table. `+ 2' is to skip flag
4153 bits for a character class. */
4154 p += CHARSET_BITMAP_SIZE (&p[-2]) + 2;
4156 /* Extract the number of ranges in range table into COUNT. */
4157 EXTRACT_NUMBER_AND_INCR (count, p);
4158 for (; count > 0; count--, p += 3)
4160 /* Extract the start and end of each range. */
4161 EXTRACT_CHARACTER (c, p);
4162 lc1 = CHAR_LEADING_CODE (c);
4163 p += 3;
4164 EXTRACT_CHARACTER (c, p);
4165 lc2 = CHAR_LEADING_CODE (c);
4166 for (j = lc1; j <= lc2; j++)
4167 fastmap[j] = 1;
4170 #endif
4171 break;
4173 case syntaxspec:
4174 case notsyntaxspec:
4175 if (!fastmap) break;
4176 #ifndef emacs
4177 not = (re_opcode_t)p[-1] == notsyntaxspec;
4178 k = *p++;
4179 for (j = 0; j < (1 << BYTEWIDTH); j++)
4180 if ((SYNTAX (j) == (enum syntaxcode) k) ^ not)
4181 fastmap[j] = 1;
4182 break;
4183 #else /* emacs */
4184 /* This match depends on text properties. These end with
4185 aborting optimizations. */
4186 return -1;
4188 case categoryspec:
4189 case notcategoryspec:
4190 if (!fastmap) break;
4191 not = (re_opcode_t)p[-1] == notcategoryspec;
4192 k = *p++;
4193 for (j = (1 << BYTEWIDTH); j >= 0; j--)
4194 if ((CHAR_HAS_CATEGORY (j, k)) ^ not)
4195 fastmap[j] = 1;
4197 /* Any leading code can possibly start a character which
4198 has or doesn't has the specified category. */
4199 if (match_any_multibyte_characters == false)
4201 for (j = MIN_MULTIBYTE_LEADING_CODE;
4202 j <= MAX_MULTIBYTE_LEADING_CODE; j++)
4203 fastmap[j] = 1;
4204 match_any_multibyte_characters = true;
4206 break;
4208 /* All cases after this match the empty string. These end with
4209 `continue'. */
4211 case before_dot:
4212 case at_dot:
4213 case after_dot:
4214 #endif /* !emacs */
4215 case no_op:
4216 case begline:
4217 case endline:
4218 case begbuf:
4219 case endbuf:
4220 case wordbound:
4221 case notwordbound:
4222 case wordbeg:
4223 case wordend:
4224 case symbeg:
4225 case symend:
4226 continue;
4229 case jump:
4230 EXTRACT_NUMBER_AND_INCR (j, p);
4231 if (j < 0)
4232 /* Backward jumps can only go back to code that we've already
4233 visited. `re_compile' should make sure this is true. */
4234 break;
4235 p += j;
4236 switch (SWITCH_ENUM_CAST ((re_opcode_t) *p))
4238 case on_failure_jump:
4239 case on_failure_keep_string_jump:
4240 case on_failure_jump_loop:
4241 case on_failure_jump_nastyloop:
4242 case on_failure_jump_smart:
4243 p++;
4244 break;
4245 default:
4246 continue;
4248 /* Keep `p1' to allow the `on_failure_jump' we are jumping to
4249 to jump back to "just after here". */
4250 /* Fallthrough */
4252 case on_failure_jump:
4253 case on_failure_keep_string_jump:
4254 case on_failure_jump_nastyloop:
4255 case on_failure_jump_loop:
4256 case on_failure_jump_smart:
4257 EXTRACT_NUMBER_AND_INCR (j, p);
4258 if (p + j <= p1)
4259 ; /* Backward jump to be ignored. */
4260 else
4261 { /* We have to look down both arms.
4262 We first go down the "straight" path so as to minimize
4263 stack usage when going through alternatives. */
4264 int r = analyse_first (p, pend, fastmap, multibyte);
4265 if (r) return r;
4266 p += j;
4268 continue;
4271 case jump_n:
4272 /* This code simply does not properly handle forward jump_n. */
4273 DEBUG_STATEMENT (EXTRACT_NUMBER (j, p); assert (j < 0));
4274 p += 4;
4275 /* jump_n can either jump or fall through. The (backward) jump
4276 case has already been handled, so we only need to look at the
4277 fallthrough case. */
4278 continue;
4280 case succeed_n:
4281 /* If N == 0, it should be an on_failure_jump_loop instead. */
4282 DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0));
4283 p += 4;
4284 /* We only care about one iteration of the loop, so we don't
4285 need to consider the case where this behaves like an
4286 on_failure_jump. */
4287 continue;
4290 case set_number_at:
4291 p += 4;
4292 continue;
4295 case start_memory:
4296 case stop_memory:
4297 p += 1;
4298 continue;
4301 default:
4302 abort (); /* We have listed all the cases. */
4303 } /* switch *p++ */
4305 /* Getting here means we have found the possible starting
4306 characters for one path of the pattern -- and that the empty
4307 string does not match. We need not follow this path further. */
4308 return 0;
4309 } /* while p */
4311 /* We reached the end without matching anything. */
4312 return 1;
4314 } /* analyse_first */
4316 /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in
4317 BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible
4318 characters can start a string that matches the pattern. This fastmap
4319 is used by re_search to skip quickly over impossible starting points.
4321 Character codes above (1 << BYTEWIDTH) are not represented in the
4322 fastmap, but the leading codes are represented. Thus, the fastmap
4323 indicates which character sets could start a match.
4325 The caller must supply the address of a (1 << BYTEWIDTH)-byte data
4326 area as BUFP->fastmap.
4328 We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in
4329 the pattern buffer.
4331 Returns 0 if we succeed, -2 if an internal error. */
4334 re_compile_fastmap (bufp)
4335 struct re_pattern_buffer *bufp;
4337 char *fastmap = bufp->fastmap;
4338 int analysis;
4340 assert (fastmap && bufp->buffer);
4342 bzero (fastmap, 1 << BYTEWIDTH); /* Assume nothing's valid. */
4343 bufp->fastmap_accurate = 1; /* It will be when we're done. */
4345 analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used,
4346 fastmap, RE_MULTIBYTE_P (bufp));
4347 bufp->can_be_null = (analysis != 0);
4348 return 0;
4349 } /* re_compile_fastmap */
4351 /* Set REGS to hold NUM_REGS registers, storing them in STARTS and
4352 ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use
4353 this memory for recording register information. STARTS and ENDS
4354 must be allocated using the malloc library routine, and must each
4355 be at least NUM_REGS * sizeof (regoff_t) bytes long.
4357 If NUM_REGS == 0, then subsequent matches should allocate their own
4358 register data.
4360 Unless this function is called, the first search or match using
4361 PATTERN_BUFFER will allocate its own register data, without
4362 freeing the old data. */
4364 void
4365 re_set_registers (bufp, regs, num_regs, starts, ends)
4366 struct re_pattern_buffer *bufp;
4367 struct re_registers *regs;
4368 unsigned num_regs;
4369 regoff_t *starts, *ends;
4371 if (num_regs)
4373 bufp->regs_allocated = REGS_REALLOCATE;
4374 regs->num_regs = num_regs;
4375 regs->start = starts;
4376 regs->end = ends;
4378 else
4380 bufp->regs_allocated = REGS_UNALLOCATED;
4381 regs->num_regs = 0;
4382 regs->start = regs->end = (regoff_t *) 0;
4385 WEAK_ALIAS (__re_set_registers, re_set_registers)
4387 /* Searching routines. */
4389 /* Like re_search_2, below, but only one string is specified, and
4390 doesn't let you say where to stop matching. */
4393 re_search (bufp, string, size, startpos, range, regs)
4394 struct re_pattern_buffer *bufp;
4395 const char *string;
4396 int size, startpos, range;
4397 struct re_registers *regs;
4399 return re_search_2 (bufp, NULL, 0, string, size, startpos, range,
4400 regs, size);
4402 WEAK_ALIAS (__re_search, re_search)
4404 /* Head address of virtual concatenation of string. */
4405 #define HEAD_ADDR_VSTRING(P) \
4406 (((P) >= size1 ? string2 : string1))
4408 /* End address of virtual concatenation of string. */
4409 #define STOP_ADDR_VSTRING(P) \
4410 (((P) >= size1 ? string2 + size2 : string1 + size1))
4412 /* Address of POS in the concatenation of virtual string. */
4413 #define POS_ADDR_VSTRING(POS) \
4414 (((POS) >= size1 ? string2 - size1 : string1) + (POS))
4416 /* Using the compiled pattern in BUFP->buffer, first tries to match the
4417 virtual concatenation of STRING1 and STRING2, starting first at index
4418 STARTPOS, then at STARTPOS + 1, and so on.
4420 STRING1 and STRING2 have length SIZE1 and SIZE2, respectively.
4422 RANGE is how far to scan while trying to match. RANGE = 0 means try
4423 only at STARTPOS; in general, the last start tried is STARTPOS +
4424 RANGE.
4426 In REGS, return the indices of the virtual concatenation of STRING1
4427 and STRING2 that matched the entire BUFP->buffer and its contained
4428 subexpressions.
4430 Do not consider matching one past the index STOP in the virtual
4431 concatenation of STRING1 and STRING2.
4433 We return either the position in the strings at which the match was
4434 found, -1 if no match, or -2 if error (such as failure
4435 stack overflow). */
4438 re_search_2 (bufp, str1, size1, str2, size2, startpos, range, regs, stop)
4439 struct re_pattern_buffer *bufp;
4440 const char *str1, *str2;
4441 int size1, size2;
4442 int startpos;
4443 int range;
4444 struct re_registers *regs;
4445 int stop;
4447 int val;
4448 re_char *string1 = (re_char*) str1;
4449 re_char *string2 = (re_char*) str2;
4450 register char *fastmap = bufp->fastmap;
4451 register RE_TRANSLATE_TYPE translate = bufp->translate;
4452 int total_size = size1 + size2;
4453 int endpos = startpos + range;
4454 boolean anchored_start;
4455 /* Nonzero if we are searching multibyte string. */
4456 const boolean multibyte = RE_TARGET_MULTIBYTE_P (bufp);
4458 /* Check for out-of-range STARTPOS. */
4459 if (startpos < 0 || startpos > total_size)
4460 return -1;
4462 /* Fix up RANGE if it might eventually take us outside
4463 the virtual concatenation of STRING1 and STRING2.
4464 Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */
4465 if (endpos < 0)
4466 range = 0 - startpos;
4467 else if (endpos > total_size)
4468 range = total_size - startpos;
4470 /* If the search isn't to be a backwards one, don't waste time in a
4471 search for a pattern anchored at beginning of buffer. */
4472 if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0)
4474 if (startpos > 0)
4475 return -1;
4476 else
4477 range = 0;
4480 #ifdef emacs
4481 /* In a forward search for something that starts with \=.
4482 don't keep searching past point. */
4483 if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0)
4485 range = PT_BYTE - BEGV_BYTE - startpos;
4486 if (range < 0)
4487 return -1;
4489 #endif /* emacs */
4491 /* Update the fastmap now if not correct already. */
4492 if (fastmap && !bufp->fastmap_accurate)
4493 re_compile_fastmap (bufp);
4495 /* See whether the pattern is anchored. */
4496 anchored_start = (bufp->buffer[0] == begline);
4498 #ifdef emacs
4499 gl_state.object = re_match_object;
4501 int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (startpos));
4503 SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1);
4505 #endif
4507 /* Loop through the string, looking for a place to start matching. */
4508 for (;;)
4510 /* If the pattern is anchored,
4511 skip quickly past places we cannot match.
4512 We don't bother to treat startpos == 0 specially
4513 because that case doesn't repeat. */
4514 if (anchored_start && startpos > 0)
4516 if (! ((startpos <= size1 ? string1[startpos - 1]
4517 : string2[startpos - size1 - 1])
4518 == '\n'))
4519 goto advance;
4522 /* If a fastmap is supplied, skip quickly over characters that
4523 cannot be the start of a match. If the pattern can match the
4524 null string, however, we don't need to skip characters; we want
4525 the first null string. */
4526 if (fastmap && startpos < total_size && !bufp->can_be_null)
4528 register re_char *d;
4529 register re_wchar_t buf_ch;
4531 d = POS_ADDR_VSTRING (startpos);
4533 if (range > 0) /* Searching forwards. */
4535 register int lim = 0;
4536 int irange = range;
4538 if (startpos < size1 && startpos + range >= size1)
4539 lim = range - (size1 - startpos);
4541 /* Written out as an if-else to avoid testing `translate'
4542 inside the loop. */
4543 if (RE_TRANSLATE_P (translate))
4545 if (multibyte)
4546 while (range > lim)
4548 int buf_charlen;
4550 buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen);
4551 buf_ch = RE_TRANSLATE (translate, buf_ch);
4552 if (fastmap[CHAR_LEADING_CODE (buf_ch)])
4553 break;
4555 range -= buf_charlen;
4556 d += buf_charlen;
4558 else
4559 while (range > lim)
4561 register re_wchar_t ch, translated;
4563 buf_ch = *d;
4564 ch = RE_CHAR_TO_MULTIBYTE (buf_ch);
4565 translated = RE_TRANSLATE (translate, ch);
4566 if (translated != ch
4567 && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0)
4568 buf_ch = ch;
4569 if (fastmap[buf_ch])
4570 break;
4571 d++;
4572 range--;
4575 else
4577 if (multibyte)
4578 while (range > lim)
4580 int buf_charlen;
4582 buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen);
4583 if (fastmap[CHAR_LEADING_CODE (buf_ch)])
4584 break;
4585 range -= buf_charlen;
4586 d += buf_charlen;
4588 else
4589 while (range > lim && !fastmap[*d])
4591 d++;
4592 range--;
4595 startpos += irange - range;
4597 else /* Searching backwards. */
4599 if (multibyte)
4601 buf_ch = STRING_CHAR (d);
4602 buf_ch = TRANSLATE (buf_ch);
4603 if (! fastmap[CHAR_LEADING_CODE (buf_ch)])
4604 goto advance;
4606 else
4608 register re_wchar_t ch, translated;
4610 buf_ch = *d;
4611 ch = RE_CHAR_TO_MULTIBYTE (buf_ch);
4612 translated = TRANSLATE (ch);
4613 if (translated != ch
4614 && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0)
4615 buf_ch = ch;
4616 if (! fastmap[TRANSLATE (buf_ch)])
4617 goto advance;
4622 /* If can't match the null string, and that's all we have left, fail. */
4623 if (range >= 0 && startpos == total_size && fastmap
4624 && !bufp->can_be_null)
4625 return -1;
4627 val = re_match_2_internal (bufp, string1, size1, string2, size2,
4628 startpos, regs, stop);
4630 if (val >= 0)
4631 return startpos;
4633 if (val == -2)
4634 return -2;
4636 advance:
4637 if (!range)
4638 break;
4639 else if (range > 0)
4641 /* Update STARTPOS to the next character boundary. */
4642 if (multibyte)
4644 re_char *p = POS_ADDR_VSTRING (startpos);
4645 re_char *pend = STOP_ADDR_VSTRING (startpos);
4646 int len = MULTIBYTE_FORM_LENGTH (p, pend - p);
4648 range -= len;
4649 if (range < 0)
4650 break;
4651 startpos += len;
4653 else
4655 range--;
4656 startpos++;
4659 else
4661 range++;
4662 startpos--;
4664 /* Update STARTPOS to the previous character boundary. */
4665 if (multibyte)
4667 re_char *p = POS_ADDR_VSTRING (startpos) + 1;
4668 re_char *p0 = p;
4669 re_char *phead = HEAD_ADDR_VSTRING (startpos);
4671 /* Find the head of multibyte form. */
4672 PREV_CHAR_BOUNDARY (p, phead);
4673 range += p0 - 1 - p;
4674 if (range > 0)
4675 break;
4677 startpos -= p0 - 1 - p;
4681 return -1;
4682 } /* re_search_2 */
4683 WEAK_ALIAS (__re_search_2, re_search_2)
4685 /* Declarations and macros for re_match_2. */
4687 static int bcmp_translate _RE_ARGS((re_char *s1, re_char *s2,
4688 register int len,
4689 RE_TRANSLATE_TYPE translate,
4690 const int multibyte));
4692 /* This converts PTR, a pointer into one of the search strings `string1'
4693 and `string2' into an offset from the beginning of that string. */
4694 #define POINTER_TO_OFFSET(ptr) \
4695 (FIRST_STRING_P (ptr) \
4696 ? ((regoff_t) ((ptr) - string1)) \
4697 : ((regoff_t) ((ptr) - string2 + size1)))
4699 /* Call before fetching a character with *d. This switches over to
4700 string2 if necessary.
4701 Check re_match_2_internal for a discussion of why end_match_2 might
4702 not be within string2 (but be equal to end_match_1 instead). */
4703 #define PREFETCH() \
4704 while (d == dend) \
4706 /* End of string2 => fail. */ \
4707 if (dend == end_match_2) \
4708 goto fail; \
4709 /* End of string1 => advance to string2. */ \
4710 d = string2; \
4711 dend = end_match_2; \
4714 /* Call before fetching a char with *d if you already checked other limits.
4715 This is meant for use in lookahead operations like wordend, etc..
4716 where we might need to look at parts of the string that might be
4717 outside of the LIMITs (i.e past `stop'). */
4718 #define PREFETCH_NOLIMIT() \
4719 if (d == end1) \
4721 d = string2; \
4722 dend = end_match_2; \
4725 /* Test if at very beginning or at very end of the virtual concatenation
4726 of `string1' and `string2'. If only one string, it's `string2'. */
4727 #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2)
4728 #define AT_STRINGS_END(d) ((d) == end2)
4731 /* Test if D points to a character which is word-constituent. We have
4732 two special cases to check for: if past the end of string1, look at
4733 the first character in string2; and if before the beginning of
4734 string2, look at the last character in string1. */
4735 #define WORDCHAR_P(d) \
4736 (SYNTAX ((d) == end1 ? *string2 \
4737 : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \
4738 == Sword)
4740 /* Disabled due to a compiler bug -- see comment at case wordbound */
4742 /* The comment at case wordbound is following one, but we don't use
4743 AT_WORD_BOUNDARY anymore to support multibyte form.
4745 The DEC Alpha C compiler 3.x generates incorrect code for the
4746 test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of
4747 AT_WORD_BOUNDARY, so this code is disabled. Expanding the
4748 macro and introducing temporary variables works around the bug. */
4750 #if 0
4751 /* Test if the character before D and the one at D differ with respect
4752 to being word-constituent. */
4753 #define AT_WORD_BOUNDARY(d) \
4754 (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \
4755 || WORDCHAR_P (d - 1) != WORDCHAR_P (d))
4756 #endif
4758 /* Free everything we malloc. */
4759 #ifdef MATCH_MAY_ALLOCATE
4760 # define FREE_VAR(var) if (var) { REGEX_FREE (var); var = NULL; } else
4761 # define FREE_VARIABLES() \
4762 do { \
4763 REGEX_FREE_STACK (fail_stack.stack); \
4764 FREE_VAR (regstart); \
4765 FREE_VAR (regend); \
4766 FREE_VAR (best_regstart); \
4767 FREE_VAR (best_regend); \
4768 } while (0)
4769 #else
4770 # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */
4771 #endif /* not MATCH_MAY_ALLOCATE */
4774 /* Optimization routines. */
4776 /* If the operation is a match against one or more chars,
4777 return a pointer to the next operation, else return NULL. */
4778 static re_char *
4779 skip_one_char (p)
4780 re_char *p;
4782 switch (SWITCH_ENUM_CAST (*p++))
4784 case anychar:
4785 break;
4787 case exactn:
4788 p += *p + 1;
4789 break;
4791 case charset_not:
4792 case charset:
4793 if (CHARSET_RANGE_TABLE_EXISTS_P (p - 1))
4795 int mcnt;
4796 p = CHARSET_RANGE_TABLE (p - 1);
4797 EXTRACT_NUMBER_AND_INCR (mcnt, p);
4798 p = CHARSET_RANGE_TABLE_END (p, mcnt);
4800 else
4801 p += 1 + CHARSET_BITMAP_SIZE (p - 1);
4802 break;
4804 case syntaxspec:
4805 case notsyntaxspec:
4806 #ifdef emacs
4807 case categoryspec:
4808 case notcategoryspec:
4809 #endif /* emacs */
4810 p++;
4811 break;
4813 default:
4814 p = NULL;
4816 return p;
4820 /* Jump over non-matching operations. */
4821 static re_char *
4822 skip_noops (p, pend)
4823 re_char *p, *pend;
4825 int mcnt;
4826 while (p < pend)
4828 switch (SWITCH_ENUM_CAST ((re_opcode_t) *p))
4830 case start_memory:
4831 case stop_memory:
4832 p += 2; break;
4833 case no_op:
4834 p += 1; break;
4835 case jump:
4836 p += 1;
4837 EXTRACT_NUMBER_AND_INCR (mcnt, p);
4838 p += mcnt;
4839 break;
4840 default:
4841 return p;
4844 assert (p == pend);
4845 return p;
4848 /* Non-zero if "p1 matches something" implies "p2 fails". */
4849 static int
4850 mutually_exclusive_p (bufp, p1, p2)
4851 struct re_pattern_buffer *bufp;
4852 re_char *p1, *p2;
4854 re_opcode_t op2;
4855 const boolean multibyte = RE_MULTIBYTE_P (bufp);
4856 unsigned char *pend = bufp->buffer + bufp->used;
4858 assert (p1 >= bufp->buffer && p1 < pend
4859 && p2 >= bufp->buffer && p2 <= pend);
4861 /* Skip over open/close-group commands.
4862 If what follows this loop is a ...+ construct,
4863 look at what begins its body, since we will have to
4864 match at least one of that. */
4865 p2 = skip_noops (p2, pend);
4866 /* The same skip can be done for p1, except that this function
4867 is only used in the case where p1 is a simple match operator. */
4868 /* p1 = skip_noops (p1, pend); */
4870 assert (p1 >= bufp->buffer && p1 < pend
4871 && p2 >= bufp->buffer && p2 <= pend);
4873 op2 = p2 == pend ? succeed : *p2;
4875 switch (SWITCH_ENUM_CAST (op2))
4877 case succeed:
4878 case endbuf:
4879 /* If we're at the end of the pattern, we can change. */
4880 if (skip_one_char (p1))
4882 DEBUG_PRINT1 (" End of pattern: fast loop.\n");
4883 return 1;
4885 break;
4887 case endline:
4888 case exactn:
4890 register re_wchar_t c
4891 = (re_opcode_t) *p2 == endline ? '\n'
4892 : RE_STRING_CHAR (p2 + 2, multibyte);
4894 if ((re_opcode_t) *p1 == exactn)
4896 if (c != RE_STRING_CHAR (p1 + 2, multibyte))
4898 DEBUG_PRINT3 (" '%c' != '%c' => fast loop.\n", c, p1[2]);
4899 return 1;
4903 else if ((re_opcode_t) *p1 == charset
4904 || (re_opcode_t) *p1 == charset_not)
4906 int not = (re_opcode_t) *p1 == charset_not;
4908 /* Test if C is listed in charset (or charset_not)
4909 at `p1'. */
4910 if (! multibyte || IS_REAL_ASCII (c))
4912 if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH
4913 && p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
4914 not = !not;
4916 else if (CHARSET_RANGE_TABLE_EXISTS_P (p1))
4917 CHARSET_LOOKUP_RANGE_TABLE (not, c, p1);
4919 /* `not' is equal to 1 if c would match, which means
4920 that we can't change to pop_failure_jump. */
4921 if (!not)
4923 DEBUG_PRINT1 (" No match => fast loop.\n");
4924 return 1;
4927 else if ((re_opcode_t) *p1 == anychar
4928 && c == '\n')
4930 DEBUG_PRINT1 (" . != \\n => fast loop.\n");
4931 return 1;
4934 break;
4936 case charset:
4938 if ((re_opcode_t) *p1 == exactn)
4939 /* Reuse the code above. */
4940 return mutually_exclusive_p (bufp, p2, p1);
4942 /* It is hard to list up all the character in charset
4943 P2 if it includes multibyte character. Give up in
4944 such case. */
4945 else if (!multibyte || !CHARSET_RANGE_TABLE_EXISTS_P (p2))
4947 /* Now, we are sure that P2 has no range table.
4948 So, for the size of bitmap in P2, `p2[1]' is
4949 enough. But P1 may have range table, so the
4950 size of bitmap table of P1 is extracted by
4951 using macro `CHARSET_BITMAP_SIZE'.
4953 In a multibyte case, we know that all the character
4954 listed in P2 is ASCII. In a unibyte case, P1 has only a
4955 bitmap table. So, in both cases, it is enough to test
4956 only the bitmap table of P1. */
4958 if ((re_opcode_t) *p1 == charset)
4960 int idx;
4961 /* We win if the charset inside the loop
4962 has no overlap with the one after the loop. */
4963 for (idx = 0;
4964 (idx < (int) p2[1]
4965 && idx < CHARSET_BITMAP_SIZE (p1));
4966 idx++)
4967 if ((p2[2 + idx] & p1[2 + idx]) != 0)
4968 break;
4970 if (idx == p2[1]
4971 || idx == CHARSET_BITMAP_SIZE (p1))
4973 DEBUG_PRINT1 (" No match => fast loop.\n");
4974 return 1;
4977 else if ((re_opcode_t) *p1 == charset_not)
4979 int idx;
4980 /* We win if the charset_not inside the loop lists
4981 every character listed in the charset after. */
4982 for (idx = 0; idx < (int) p2[1]; idx++)
4983 if (! (p2[2 + idx] == 0
4984 || (idx < CHARSET_BITMAP_SIZE (p1)
4985 && ((p2[2 + idx] & ~ p1[2 + idx]) == 0))))
4986 break;
4988 if (idx == p2[1])
4990 DEBUG_PRINT1 (" No match => fast loop.\n");
4991 return 1;
4996 break;
4998 case charset_not:
4999 switch (SWITCH_ENUM_CAST (*p1))
5001 case exactn:
5002 case charset:
5003 /* Reuse the code above. */
5004 return mutually_exclusive_p (bufp, p2, p1);
5005 case charset_not:
5006 /* When we have two charset_not, it's very unlikely that
5007 they don't overlap. The union of the two sets of excluded
5008 chars should cover all possible chars, which, as a matter of
5009 fact, is virtually impossible in multibyte buffers. */
5010 break;
5012 break;
5014 case wordend:
5015 return ((re_opcode_t) *p1 == syntaxspec && p1[1] == Sword);
5016 case symend:
5017 return ((re_opcode_t) *p1 == syntaxspec
5018 && (p1[1] == Ssymbol || p1[1] == Sword));
5019 case notsyntaxspec:
5020 return ((re_opcode_t) *p1 == syntaxspec && p1[1] == p2[1]);
5022 case wordbeg:
5023 return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == Sword);
5024 case symbeg:
5025 return ((re_opcode_t) *p1 == notsyntaxspec
5026 && (p1[1] == Ssymbol || p1[1] == Sword));
5027 case syntaxspec:
5028 return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == p2[1]);
5030 case wordbound:
5031 return (((re_opcode_t) *p1 == notsyntaxspec
5032 || (re_opcode_t) *p1 == syntaxspec)
5033 && p1[1] == Sword);
5035 #ifdef emacs
5036 case categoryspec:
5037 return ((re_opcode_t) *p1 == notcategoryspec && p1[1] == p2[1]);
5038 case notcategoryspec:
5039 return ((re_opcode_t) *p1 == categoryspec && p1[1] == p2[1]);
5040 #endif /* emacs */
5042 default:
5046 /* Safe default. */
5047 return 0;
5051 /* Matching routines. */
5053 #ifndef emacs /* Emacs never uses this. */
5054 /* re_match is like re_match_2 except it takes only a single string. */
5057 re_match (bufp, string, size, pos, regs)
5058 struct re_pattern_buffer *bufp;
5059 const char *string;
5060 int size, pos;
5061 struct re_registers *regs;
5063 int result = re_match_2_internal (bufp, NULL, 0, (re_char*) string, size,
5064 pos, regs, size);
5065 return result;
5067 WEAK_ALIAS (__re_match, re_match)
5068 #endif /* not emacs */
5070 #ifdef emacs
5071 /* In Emacs, this is the string or buffer in which we
5072 are matching. It is used for looking up syntax properties. */
5073 Lisp_Object re_match_object;
5074 #endif
5076 /* re_match_2 matches the compiled pattern in BUFP against the
5077 the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1
5078 and SIZE2, respectively). We start matching at POS, and stop
5079 matching at STOP.
5081 If REGS is non-null and the `no_sub' field of BUFP is nonzero, we
5082 store offsets for the substring each group matched in REGS. See the
5083 documentation for exactly how many groups we fill.
5085 We return -1 if no match, -2 if an internal error (such as the
5086 failure stack overflowing). Otherwise, we return the length of the
5087 matched substring. */
5090 re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
5091 struct re_pattern_buffer *bufp;
5092 const char *string1, *string2;
5093 int size1, size2;
5094 int pos;
5095 struct re_registers *regs;
5096 int stop;
5098 int result;
5100 #ifdef emacs
5101 int charpos;
5102 gl_state.object = re_match_object;
5103 charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (pos));
5104 SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1);
5105 #endif
5107 result = re_match_2_internal (bufp, (re_char*) string1, size1,
5108 (re_char*) string2, size2,
5109 pos, regs, stop);
5110 return result;
5112 WEAK_ALIAS (__re_match_2, re_match_2)
5115 /* This is a separate function so that we can force an alloca cleanup
5116 afterwards. */
5117 static int
5118 re_match_2_internal (bufp, string1, size1, string2, size2, pos, regs, stop)
5119 struct re_pattern_buffer *bufp;
5120 re_char *string1, *string2;
5121 int size1, size2;
5122 int pos;
5123 struct re_registers *regs;
5124 int stop;
5126 /* General temporaries. */
5127 int mcnt;
5128 size_t reg;
5129 boolean not;
5131 /* Just past the end of the corresponding string. */
5132 re_char *end1, *end2;
5134 /* Pointers into string1 and string2, just past the last characters in
5135 each to consider matching. */
5136 re_char *end_match_1, *end_match_2;
5138 /* Where we are in the data, and the end of the current string. */
5139 re_char *d, *dend;
5141 /* Used sometimes to remember where we were before starting matching
5142 an operator so that we can go back in case of failure. This "atomic"
5143 behavior of matching opcodes is indispensable to the correctness
5144 of the on_failure_keep_string_jump optimization. */
5145 re_char *dfail;
5147 /* Where we are in the pattern, and the end of the pattern. */
5148 re_char *p = bufp->buffer;
5149 re_char *pend = p + bufp->used;
5151 /* We use this to map every character in the string. */
5152 RE_TRANSLATE_TYPE translate = bufp->translate;
5154 /* Nonzero if BUFP is setup from a multibyte regex. */
5155 const boolean multibyte = RE_MULTIBYTE_P (bufp);
5157 /* Nonzero if STRING1/STRING2 are multibyte. */
5158 const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp);
5160 /* Failure point stack. Each place that can handle a failure further
5161 down the line pushes a failure point on this stack. It consists of
5162 regstart, and regend for all registers corresponding to
5163 the subexpressions we're currently inside, plus the number of such
5164 registers, and, finally, two char *'s. The first char * is where
5165 to resume scanning the pattern; the second one is where to resume
5166 scanning the strings. */
5167 #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */
5168 fail_stack_type fail_stack;
5169 #endif
5170 #ifdef DEBUG
5171 unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0;
5172 #endif
5174 #if defined REL_ALLOC && defined REGEX_MALLOC
5175 /* This holds the pointer to the failure stack, when
5176 it is allocated relocatably. */
5177 fail_stack_elt_t *failure_stack_ptr;
5178 #endif
5180 /* We fill all the registers internally, independent of what we
5181 return, for use in backreferences. The number here includes
5182 an element for register zero. */
5183 size_t num_regs = bufp->re_nsub + 1;
5185 /* Information on the contents of registers. These are pointers into
5186 the input strings; they record just what was matched (on this
5187 attempt) by a subexpression part of the pattern, that is, the
5188 regnum-th regstart pointer points to where in the pattern we began
5189 matching and the regnum-th regend points to right after where we
5190 stopped matching the regnum-th subexpression. (The zeroth register
5191 keeps track of what the whole pattern matches.) */
5192 #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
5193 re_char **regstart, **regend;
5194 #endif
5196 /* The following record the register info as found in the above
5197 variables when we find a match better than any we've seen before.
5198 This happens as we backtrack through the failure points, which in
5199 turn happens only if we have not yet matched the entire string. */
5200 unsigned best_regs_set = false;
5201 #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
5202 re_char **best_regstart, **best_regend;
5203 #endif
5205 /* Logically, this is `best_regend[0]'. But we don't want to have to
5206 allocate space for that if we're not allocating space for anything
5207 else (see below). Also, we never need info about register 0 for
5208 any of the other register vectors, and it seems rather a kludge to
5209 treat `best_regend' differently than the rest. So we keep track of
5210 the end of the best match so far in a separate variable. We
5211 initialize this to NULL so that when we backtrack the first time
5212 and need to test it, it's not garbage. */
5213 re_char *match_end = NULL;
5215 #ifdef DEBUG
5216 /* Counts the total number of registers pushed. */
5217 unsigned num_regs_pushed = 0;
5218 #endif
5220 DEBUG_PRINT1 ("\n\nEntering re_match_2.\n");
5222 INIT_FAIL_STACK ();
5224 #ifdef MATCH_MAY_ALLOCATE
5225 /* Do not bother to initialize all the register variables if there are
5226 no groups in the pattern, as it takes a fair amount of time. If
5227 there are groups, we include space for register 0 (the whole
5228 pattern), even though we never use it, since it simplifies the
5229 array indexing. We should fix this. */
5230 if (bufp->re_nsub)
5232 regstart = REGEX_TALLOC (num_regs, re_char *);
5233 regend = REGEX_TALLOC (num_regs, re_char *);
5234 best_regstart = REGEX_TALLOC (num_regs, re_char *);
5235 best_regend = REGEX_TALLOC (num_regs, re_char *);
5237 if (!(regstart && regend && best_regstart && best_regend))
5239 FREE_VARIABLES ();
5240 return -2;
5243 else
5245 /* We must initialize all our variables to NULL, so that
5246 `FREE_VARIABLES' doesn't try to free them. */
5247 regstart = regend = best_regstart = best_regend = NULL;
5249 #endif /* MATCH_MAY_ALLOCATE */
5251 /* The starting position is bogus. */
5252 if (pos < 0 || pos > size1 + size2)
5254 FREE_VARIABLES ();
5255 return -1;
5258 /* Initialize subexpression text positions to -1 to mark ones that no
5259 start_memory/stop_memory has been seen for. Also initialize the
5260 register information struct. */
5261 for (reg = 1; reg < num_regs; reg++)
5262 regstart[reg] = regend[reg] = NULL;
5264 /* We move `string1' into `string2' if the latter's empty -- but not if
5265 `string1' is null. */
5266 if (size2 == 0 && string1 != NULL)
5268 string2 = string1;
5269 size2 = size1;
5270 string1 = 0;
5271 size1 = 0;
5273 end1 = string1 + size1;
5274 end2 = string2 + size2;
5276 /* `p' scans through the pattern as `d' scans through the data.
5277 `dend' is the end of the input string that `d' points within. `d'
5278 is advanced into the following input string whenever necessary, but
5279 this happens before fetching; therefore, at the beginning of the
5280 loop, `d' can be pointing at the end of a string, but it cannot
5281 equal `string2'. */
5282 if (pos >= size1)
5284 /* Only match within string2. */
5285 d = string2 + pos - size1;
5286 dend = end_match_2 = string2 + stop - size1;
5287 end_match_1 = end1; /* Just to give it a value. */
5289 else
5291 if (stop < size1)
5293 /* Only match within string1. */
5294 end_match_1 = string1 + stop;
5295 /* BEWARE!
5296 When we reach end_match_1, PREFETCH normally switches to string2.
5297 But in the present case, this means that just doing a PREFETCH
5298 makes us jump from `stop' to `gap' within the string.
5299 What we really want here is for the search to stop as
5300 soon as we hit end_match_1. That's why we set end_match_2
5301 to end_match_1 (since PREFETCH fails as soon as we hit
5302 end_match_2). */
5303 end_match_2 = end_match_1;
5305 else
5306 { /* It's important to use this code when stop == size so that
5307 moving `d' from end1 to string2 will not prevent the d == dend
5308 check from catching the end of string. */
5309 end_match_1 = end1;
5310 end_match_2 = string2 + stop - size1;
5312 d = string1 + pos;
5313 dend = end_match_1;
5316 DEBUG_PRINT1 ("The compiled pattern is: ");
5317 DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend);
5318 DEBUG_PRINT1 ("The string to match is: `");
5319 DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2);
5320 DEBUG_PRINT1 ("'\n");
5322 /* This loops over pattern commands. It exits by returning from the
5323 function if the match is complete, or it drops through if the match
5324 fails at this starting point in the input data. */
5325 for (;;)
5327 DEBUG_PRINT2 ("\n%p: ", p);
5329 if (p == pend)
5330 { /* End of pattern means we might have succeeded. */
5331 DEBUG_PRINT1 ("end of pattern ... ");
5333 /* If we haven't matched the entire string, and we want the
5334 longest match, try backtracking. */
5335 if (d != end_match_2)
5337 /* 1 if this match ends in the same string (string1 or string2)
5338 as the best previous match. */
5339 boolean same_str_p = (FIRST_STRING_P (match_end)
5340 == FIRST_STRING_P (d));
5341 /* 1 if this match is the best seen so far. */
5342 boolean best_match_p;
5344 /* AIX compiler got confused when this was combined
5345 with the previous declaration. */
5346 if (same_str_p)
5347 best_match_p = d > match_end;
5348 else
5349 best_match_p = !FIRST_STRING_P (d);
5351 DEBUG_PRINT1 ("backtracking.\n");
5353 if (!FAIL_STACK_EMPTY ())
5354 { /* More failure points to try. */
5356 /* If exceeds best match so far, save it. */
5357 if (!best_regs_set || best_match_p)
5359 best_regs_set = true;
5360 match_end = d;
5362 DEBUG_PRINT1 ("\nSAVING match as best so far.\n");
5364 for (reg = 1; reg < num_regs; reg++)
5366 best_regstart[reg] = regstart[reg];
5367 best_regend[reg] = regend[reg];
5370 goto fail;
5373 /* If no failure points, don't restore garbage. And if
5374 last match is real best match, don't restore second
5375 best one. */
5376 else if (best_regs_set && !best_match_p)
5378 restore_best_regs:
5379 /* Restore best match. It may happen that `dend ==
5380 end_match_1' while the restored d is in string2.
5381 For example, the pattern `x.*y.*z' against the
5382 strings `x-' and `y-z-', if the two strings are
5383 not consecutive in memory. */
5384 DEBUG_PRINT1 ("Restoring best registers.\n");
5386 d = match_end;
5387 dend = ((d >= string1 && d <= end1)
5388 ? end_match_1 : end_match_2);
5390 for (reg = 1; reg < num_regs; reg++)
5392 regstart[reg] = best_regstart[reg];
5393 regend[reg] = best_regend[reg];
5396 } /* d != end_match_2 */
5398 succeed_label:
5399 DEBUG_PRINT1 ("Accepting match.\n");
5401 /* If caller wants register contents data back, do it. */
5402 if (regs && !bufp->no_sub)
5404 /* Have the register data arrays been allocated? */
5405 if (bufp->regs_allocated == REGS_UNALLOCATED)
5406 { /* No. So allocate them with malloc. We need one
5407 extra element beyond `num_regs' for the `-1' marker
5408 GNU code uses. */
5409 regs->num_regs = MAX (RE_NREGS, num_regs + 1);
5410 regs->start = TALLOC (regs->num_regs, regoff_t);
5411 regs->end = TALLOC (regs->num_regs, regoff_t);
5412 if (regs->start == NULL || regs->end == NULL)
5414 FREE_VARIABLES ();
5415 return -2;
5417 bufp->regs_allocated = REGS_REALLOCATE;
5419 else if (bufp->regs_allocated == REGS_REALLOCATE)
5420 { /* Yes. If we need more elements than were already
5421 allocated, reallocate them. If we need fewer, just
5422 leave it alone. */
5423 if (regs->num_regs < num_regs + 1)
5425 regs->num_regs = num_regs + 1;
5426 RETALLOC (regs->start, regs->num_regs, regoff_t);
5427 RETALLOC (regs->end, regs->num_regs, regoff_t);
5428 if (regs->start == NULL || regs->end == NULL)
5430 FREE_VARIABLES ();
5431 return -2;
5435 else
5437 /* These braces fend off a "empty body in an else-statement"
5438 warning under GCC when assert expands to nothing. */
5439 assert (bufp->regs_allocated == REGS_FIXED);
5442 /* Convert the pointer data in `regstart' and `regend' to
5443 indices. Register zero has to be set differently,
5444 since we haven't kept track of any info for it. */
5445 if (regs->num_regs > 0)
5447 regs->start[0] = pos;
5448 regs->end[0] = POINTER_TO_OFFSET (d);
5451 /* Go through the first `min (num_regs, regs->num_regs)'
5452 registers, since that is all we initialized. */
5453 for (reg = 1; reg < MIN (num_regs, regs->num_regs); reg++)
5455 if (REG_UNSET (regstart[reg]) || REG_UNSET (regend[reg]))
5456 regs->start[reg] = regs->end[reg] = -1;
5457 else
5459 regs->start[reg]
5460 = (regoff_t) POINTER_TO_OFFSET (regstart[reg]);
5461 regs->end[reg]
5462 = (regoff_t) POINTER_TO_OFFSET (regend[reg]);
5466 /* If the regs structure we return has more elements than
5467 were in the pattern, set the extra elements to -1. If
5468 we (re)allocated the registers, this is the case,
5469 because we always allocate enough to have at least one
5470 -1 at the end. */
5471 for (reg = num_regs; reg < regs->num_regs; reg++)
5472 regs->start[reg] = regs->end[reg] = -1;
5473 } /* regs && !bufp->no_sub */
5475 DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n",
5476 nfailure_points_pushed, nfailure_points_popped,
5477 nfailure_points_pushed - nfailure_points_popped);
5478 DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed);
5480 mcnt = POINTER_TO_OFFSET (d) - pos;
5482 DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt);
5484 FREE_VARIABLES ();
5485 return mcnt;
5488 /* Otherwise match next pattern command. */
5489 switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
5491 /* Ignore these. Used to ignore the n of succeed_n's which
5492 currently have n == 0. */
5493 case no_op:
5494 DEBUG_PRINT1 ("EXECUTING no_op.\n");
5495 break;
5497 case succeed:
5498 DEBUG_PRINT1 ("EXECUTING succeed.\n");
5499 goto succeed_label;
5501 /* Match the next n pattern characters exactly. The following
5502 byte in the pattern defines n, and the n bytes after that
5503 are the characters to match. */
5504 case exactn:
5505 mcnt = *p++;
5506 DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt);
5508 /* Remember the start point to rollback upon failure. */
5509 dfail = d;
5511 #ifndef emacs
5512 /* This is written out as an if-else so we don't waste time
5513 testing `translate' inside the loop. */
5514 if (RE_TRANSLATE_P (translate))
5517 PREFETCH ();
5518 if (RE_TRANSLATE (translate, *d) != *p++)
5520 d = dfail;
5521 goto fail;
5523 d++;
5525 while (--mcnt);
5526 else
5529 PREFETCH ();
5530 if (*d++ != *p++)
5532 d = dfail;
5533 goto fail;
5536 while (--mcnt);
5537 #else /* emacs */
5538 /* The cost of testing `translate' is comparatively small. */
5539 if (target_multibyte)
5542 int pat_charlen, buf_charlen;
5543 int pat_ch, buf_ch;
5545 PREFETCH ();
5546 if (multibyte)
5547 pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen);
5548 else
5550 pat_ch = RE_CHAR_TO_MULTIBYTE (*p);
5551 pat_charlen = 1;
5553 buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen);
5555 if (TRANSLATE (buf_ch) != pat_ch)
5557 d = dfail;
5558 goto fail;
5561 p += pat_charlen;
5562 d += buf_charlen;
5563 mcnt -= pat_charlen;
5565 while (mcnt > 0);
5566 else
5569 int pat_charlen, buf_charlen;
5570 int pat_ch, buf_ch;
5572 PREFETCH ();
5573 if (multibyte)
5575 pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen);
5576 pat_ch = RE_CHAR_TO_UNIBYTE (pat_ch);
5578 else
5580 pat_ch = *p;
5581 pat_charlen = 1;
5583 buf_ch = RE_CHAR_TO_MULTIBYTE (*d);
5584 if (! CHAR_BYTE8_P (buf_ch))
5586 buf_ch = TRANSLATE (buf_ch);
5587 buf_ch = RE_CHAR_TO_UNIBYTE (buf_ch);
5588 if (buf_ch < 0)
5589 buf_ch = *d;
5591 else
5592 buf_ch = *d;
5593 if (buf_ch != pat_ch)
5595 d = dfail;
5596 goto fail;
5598 p += pat_charlen;
5599 d++;
5601 while (--mcnt);
5602 #endif
5603 break;
5606 /* Match any character except possibly a newline or a null. */
5607 case anychar:
5609 int buf_charlen;
5610 re_wchar_t buf_ch;
5612 DEBUG_PRINT1 ("EXECUTING anychar.\n");
5614 PREFETCH ();
5615 buf_ch = RE_STRING_CHAR_AND_LENGTH (d, buf_charlen,
5616 target_multibyte);
5617 buf_ch = TRANSLATE (buf_ch);
5619 if ((!(bufp->syntax & RE_DOT_NEWLINE)
5620 && buf_ch == '\n')
5621 || ((bufp->syntax & RE_DOT_NOT_NULL)
5622 && buf_ch == '\000'))
5623 goto fail;
5625 DEBUG_PRINT2 (" Matched `%d'.\n", *d);
5626 d += buf_charlen;
5628 break;
5631 case charset:
5632 case charset_not:
5634 register unsigned int c;
5635 boolean not = (re_opcode_t) *(p - 1) == charset_not;
5636 int len;
5638 /* Start of actual range_table, or end of bitmap if there is no
5639 range table. */
5640 re_char *range_table;
5642 /* Nonzero if there is a range table. */
5643 int range_table_exists;
5645 /* Number of ranges of range table. This is not included
5646 in the initial byte-length of the command. */
5647 int count = 0;
5649 /* Whether matching against a unibyte character. */
5650 boolean unibyte_char = false;
5652 DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : "");
5654 range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]);
5656 if (range_table_exists)
5658 range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap. */
5659 EXTRACT_NUMBER_AND_INCR (count, range_table);
5662 PREFETCH ();
5663 c = RE_STRING_CHAR_AND_LENGTH (d, len, target_multibyte);
5664 if (target_multibyte)
5666 int c1;
5668 c = TRANSLATE (c);
5669 c1 = RE_CHAR_TO_UNIBYTE (c);
5670 if (c1 >= 0)
5672 unibyte_char = true;
5673 c = c1;
5676 else
5678 int c1 = RE_CHAR_TO_MULTIBYTE (c);
5680 if (! CHAR_BYTE8_P (c1))
5682 c1 = TRANSLATE (c1);
5683 c1 = RE_CHAR_TO_UNIBYTE (c1);
5684 if (c1 >= 0)
5686 unibyte_char = true;
5687 c = c1;
5690 else
5691 unibyte_char = true;
5694 if (unibyte_char && c < (1 << BYTEWIDTH))
5695 { /* Lookup bitmap. */
5696 /* Cast to `unsigned' instead of `unsigned char' in
5697 case the bit list is a full 32 bytes long. */
5698 if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH)
5699 && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
5700 not = !not;
5702 #ifdef emacs
5703 else if (range_table_exists)
5705 int class_bits = CHARSET_RANGE_TABLE_BITS (&p[-1]);
5707 if ( (class_bits & BIT_LOWER && ISLOWER (c))
5708 | (class_bits & BIT_MULTIBYTE)
5709 | (class_bits & BIT_PUNCT && ISPUNCT (c))
5710 | (class_bits & BIT_SPACE && ISSPACE (c))
5711 | (class_bits & BIT_UPPER && ISUPPER (c))
5712 | (class_bits & BIT_WORD && ISWORD (c)))
5713 not = !not;
5714 else
5715 CHARSET_LOOKUP_RANGE_TABLE_RAW (not, c, range_table, count);
5717 #endif /* emacs */
5719 if (range_table_exists)
5720 p = CHARSET_RANGE_TABLE_END (range_table, count);
5721 else
5722 p += CHARSET_BITMAP_SIZE (&p[-1]) + 1;
5724 if (!not) goto fail;
5726 d += len;
5727 break;
5731 /* The beginning of a group is represented by start_memory.
5732 The argument is the register number. The text
5733 matched within the group is recorded (in the internal
5734 registers data structure) under the register number. */
5735 case start_memory:
5736 DEBUG_PRINT2 ("EXECUTING start_memory %d:\n", *p);
5738 /* In case we need to undo this operation (via backtracking). */
5739 PUSH_FAILURE_REG ((unsigned int)*p);
5741 regstart[*p] = d;
5742 regend[*p] = NULL; /* probably unnecessary. -sm */
5743 DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p]));
5745 /* Move past the register number and inner group count. */
5746 p += 1;
5747 break;
5750 /* The stop_memory opcode represents the end of a group. Its
5751 argument is the same as start_memory's: the register number. */
5752 case stop_memory:
5753 DEBUG_PRINT2 ("EXECUTING stop_memory %d:\n", *p);
5755 assert (!REG_UNSET (regstart[*p]));
5756 /* Strictly speaking, there should be code such as:
5758 assert (REG_UNSET (regend[*p]));
5759 PUSH_FAILURE_REGSTOP ((unsigned int)*p);
5761 But the only info to be pushed is regend[*p] and it is known to
5762 be UNSET, so there really isn't anything to push.
5763 Not pushing anything, on the other hand deprives us from the
5764 guarantee that regend[*p] is UNSET since undoing this operation
5765 will not reset its value properly. This is not important since
5766 the value will only be read on the next start_memory or at
5767 the very end and both events can only happen if this stop_memory
5768 is *not* undone. */
5770 regend[*p] = d;
5771 DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p]));
5773 /* Move past the register number and the inner group count. */
5774 p += 1;
5775 break;
5778 /* \<digit> has been turned into a `duplicate' command which is
5779 followed by the numeric value of <digit> as the register number. */
5780 case duplicate:
5782 register re_char *d2, *dend2;
5783 int regno = *p++; /* Get which register to match against. */
5784 DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno);
5786 /* Can't back reference a group which we've never matched. */
5787 if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno]))
5788 goto fail;
5790 /* Where in input to try to start matching. */
5791 d2 = regstart[regno];
5793 /* Remember the start point to rollback upon failure. */
5794 dfail = d;
5796 /* Where to stop matching; if both the place to start and
5797 the place to stop matching are in the same string, then
5798 set to the place to stop, otherwise, for now have to use
5799 the end of the first string. */
5801 dend2 = ((FIRST_STRING_P (regstart[regno])
5802 == FIRST_STRING_P (regend[regno]))
5803 ? regend[regno] : end_match_1);
5804 for (;;)
5806 /* If necessary, advance to next segment in register
5807 contents. */
5808 while (d2 == dend2)
5810 if (dend2 == end_match_2) break;
5811 if (dend2 == regend[regno]) break;
5813 /* End of string1 => advance to string2. */
5814 d2 = string2;
5815 dend2 = regend[regno];
5817 /* At end of register contents => success */
5818 if (d2 == dend2) break;
5820 /* If necessary, advance to next segment in data. */
5821 PREFETCH ();
5823 /* How many characters left in this segment to match. */
5824 mcnt = dend - d;
5826 /* Want how many consecutive characters we can match in
5827 one shot, so, if necessary, adjust the count. */
5828 if (mcnt > dend2 - d2)
5829 mcnt = dend2 - d2;
5831 /* Compare that many; failure if mismatch, else move
5832 past them. */
5833 if (RE_TRANSLATE_P (translate)
5834 ? bcmp_translate (d, d2, mcnt, translate, target_multibyte)
5835 : memcmp (d, d2, mcnt))
5837 d = dfail;
5838 goto fail;
5840 d += mcnt, d2 += mcnt;
5843 break;
5846 /* begline matches the empty string at the beginning of the string
5847 (unless `not_bol' is set in `bufp'), and after newlines. */
5848 case begline:
5849 DEBUG_PRINT1 ("EXECUTING begline.\n");
5851 if (AT_STRINGS_BEG (d))
5853 if (!bufp->not_bol) break;
5855 else
5857 unsigned c;
5858 GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2);
5859 if (c == '\n')
5860 break;
5862 /* In all other cases, we fail. */
5863 goto fail;
5866 /* endline is the dual of begline. */
5867 case endline:
5868 DEBUG_PRINT1 ("EXECUTING endline.\n");
5870 if (AT_STRINGS_END (d))
5872 if (!bufp->not_eol) break;
5874 else
5876 PREFETCH_NOLIMIT ();
5877 if (*d == '\n')
5878 break;
5880 goto fail;
5883 /* Match at the very beginning of the data. */
5884 case begbuf:
5885 DEBUG_PRINT1 ("EXECUTING begbuf.\n");
5886 if (AT_STRINGS_BEG (d))
5887 break;
5888 goto fail;
5891 /* Match at the very end of the data. */
5892 case endbuf:
5893 DEBUG_PRINT1 ("EXECUTING endbuf.\n");
5894 if (AT_STRINGS_END (d))
5895 break;
5896 goto fail;
5899 /* on_failure_keep_string_jump is used to optimize `.*\n'. It
5900 pushes NULL as the value for the string on the stack. Then
5901 `POP_FAILURE_POINT' will keep the current value for the
5902 string, instead of restoring it. To see why, consider
5903 matching `foo\nbar' against `.*\n'. The .* matches the foo;
5904 then the . fails against the \n. But the next thing we want
5905 to do is match the \n against the \n; if we restored the
5906 string value, we would be back at the foo.
5908 Because this is used only in specific cases, we don't need to
5909 check all the things that `on_failure_jump' does, to make
5910 sure the right things get saved on the stack. Hence we don't
5911 share its code. The only reason to push anything on the
5912 stack at all is that otherwise we would have to change
5913 `anychar's code to do something besides goto fail in this
5914 case; that seems worse than this. */
5915 case on_failure_keep_string_jump:
5916 EXTRACT_NUMBER_AND_INCR (mcnt, p);
5917 DEBUG_PRINT3 ("EXECUTING on_failure_keep_string_jump %d (to %p):\n",
5918 mcnt, p + mcnt);
5920 PUSH_FAILURE_POINT (p - 3, NULL);
5921 break;
5923 /* A nasty loop is introduced by the non-greedy *? and +?.
5924 With such loops, the stack only ever contains one failure point
5925 at a time, so that a plain on_failure_jump_loop kind of
5926 cycle detection cannot work. Worse yet, such a detection
5927 can not only fail to detect a cycle, but it can also wrongly
5928 detect a cycle (between different instantiations of the same
5929 loop).
5930 So the method used for those nasty loops is a little different:
5931 We use a special cycle-detection-stack-frame which is pushed
5932 when the on_failure_jump_nastyloop failure-point is *popped*.
5933 This special frame thus marks the beginning of one iteration
5934 through the loop and we can hence easily check right here
5935 whether something matched between the beginning and the end of
5936 the loop. */
5937 case on_failure_jump_nastyloop:
5938 EXTRACT_NUMBER_AND_INCR (mcnt, p);
5939 DEBUG_PRINT3 ("EXECUTING on_failure_jump_nastyloop %d (to %p):\n",
5940 mcnt, p + mcnt);
5942 assert ((re_opcode_t)p[-4] == no_op);
5944 int cycle = 0;
5945 CHECK_INFINITE_LOOP (p - 4, d);
5946 if (!cycle)
5947 /* If there's a cycle, just continue without pushing
5948 this failure point. The failure point is the "try again"
5949 option, which shouldn't be tried.
5950 We want (x?)*?y\1z to match both xxyz and xxyxz. */
5951 PUSH_FAILURE_POINT (p - 3, d);
5953 break;
5955 /* Simple loop detecting on_failure_jump: just check on the
5956 failure stack if the same spot was already hit earlier. */
5957 case on_failure_jump_loop:
5958 on_failure:
5959 EXTRACT_NUMBER_AND_INCR (mcnt, p);
5960 DEBUG_PRINT3 ("EXECUTING on_failure_jump_loop %d (to %p):\n",
5961 mcnt, p + mcnt);
5963 int cycle = 0;
5964 CHECK_INFINITE_LOOP (p - 3, d);
5965 if (cycle)
5966 /* If there's a cycle, get out of the loop, as if the matching
5967 had failed. We used to just `goto fail' here, but that was
5968 aborting the search a bit too early: we want to keep the
5969 empty-loop-match and keep matching after the loop.
5970 We want (x?)*y\1z to match both xxyz and xxyxz. */
5971 p += mcnt;
5972 else
5973 PUSH_FAILURE_POINT (p - 3, d);
5975 break;
5978 /* Uses of on_failure_jump:
5980 Each alternative starts with an on_failure_jump that points
5981 to the beginning of the next alternative. Each alternative
5982 except the last ends with a jump that in effect jumps past
5983 the rest of the alternatives. (They really jump to the
5984 ending jump of the following alternative, because tensioning
5985 these jumps is a hassle.)
5987 Repeats start with an on_failure_jump that points past both
5988 the repetition text and either the following jump or
5989 pop_failure_jump back to this on_failure_jump. */
5990 case on_failure_jump:
5991 EXTRACT_NUMBER_AND_INCR (mcnt, p);
5992 DEBUG_PRINT3 ("EXECUTING on_failure_jump %d (to %p):\n",
5993 mcnt, p + mcnt);
5995 PUSH_FAILURE_POINT (p -3, d);
5996 break;
5998 /* This operation is used for greedy *.
5999 Compare the beginning of the repeat with what in the
6000 pattern follows its end. If we can establish that there
6001 is nothing that they would both match, i.e., that we
6002 would have to backtrack because of (as in, e.g., `a*a')
6003 then we can use a non-backtracking loop based on
6004 on_failure_keep_string_jump instead of on_failure_jump. */
6005 case on_failure_jump_smart:
6006 EXTRACT_NUMBER_AND_INCR (mcnt, p);
6007 DEBUG_PRINT3 ("EXECUTING on_failure_jump_smart %d (to %p).\n",
6008 mcnt, p + mcnt);
6010 re_char *p1 = p; /* Next operation. */
6011 /* Here, we discard `const', making re_match non-reentrant. */
6012 unsigned char *p2 = (unsigned char*) p + mcnt; /* Jump dest. */
6013 unsigned char *p3 = (unsigned char*) p - 3; /* opcode location. */
6015 p -= 3; /* Reset so that we will re-execute the
6016 instruction once it's been changed. */
6018 EXTRACT_NUMBER (mcnt, p2 - 2);
6020 /* Ensure this is a indeed the trivial kind of loop
6021 we are expecting. */
6022 assert (skip_one_char (p1) == p2 - 3);
6023 assert ((re_opcode_t) p2[-3] == jump && p2 + mcnt == p);
6024 DEBUG_STATEMENT (debug += 2);
6025 if (mutually_exclusive_p (bufp, p1, p2))
6027 /* Use a fast `on_failure_keep_string_jump' loop. */
6028 DEBUG_PRINT1 (" smart exclusive => fast loop.\n");
6029 *p3 = (unsigned char) on_failure_keep_string_jump;
6030 STORE_NUMBER (p2 - 2, mcnt + 3);
6032 else
6034 /* Default to a safe `on_failure_jump' loop. */
6035 DEBUG_PRINT1 (" smart default => slow loop.\n");
6036 *p3 = (unsigned char) on_failure_jump;
6038 DEBUG_STATEMENT (debug -= 2);
6040 break;
6042 /* Unconditionally jump (without popping any failure points). */
6043 case jump:
6044 unconditional_jump:
6045 IMMEDIATE_QUIT_CHECK;
6046 EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */
6047 DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt);
6048 p += mcnt; /* Do the jump. */
6049 DEBUG_PRINT2 ("(to %p).\n", p);
6050 break;
6053 /* Have to succeed matching what follows at least n times.
6054 After that, handle like `on_failure_jump'. */
6055 case succeed_n:
6056 /* Signedness doesn't matter since we only compare MCNT to 0. */
6057 EXTRACT_NUMBER (mcnt, p + 2);
6058 DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt);
6060 /* Originally, mcnt is how many times we HAVE to succeed. */
6061 if (mcnt != 0)
6063 /* Here, we discard `const', making re_match non-reentrant. */
6064 unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */
6065 mcnt--;
6066 p += 4;
6067 PUSH_NUMBER (p2, mcnt);
6069 else
6070 /* The two bytes encoding mcnt == 0 are two no_op opcodes. */
6071 goto on_failure;
6072 break;
6074 case jump_n:
6075 /* Signedness doesn't matter since we only compare MCNT to 0. */
6076 EXTRACT_NUMBER (mcnt, p + 2);
6077 DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt);
6079 /* Originally, this is how many times we CAN jump. */
6080 if (mcnt != 0)
6082 /* Here, we discard `const', making re_match non-reentrant. */
6083 unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */
6084 mcnt--;
6085 PUSH_NUMBER (p2, mcnt);
6086 goto unconditional_jump;
6088 /* If don't have to jump any more, skip over the rest of command. */
6089 else
6090 p += 4;
6091 break;
6093 case set_number_at:
6095 unsigned char *p2; /* Location of the counter. */
6096 DEBUG_PRINT1 ("EXECUTING set_number_at.\n");
6098 EXTRACT_NUMBER_AND_INCR (mcnt, p);
6099 /* Here, we discard `const', making re_match non-reentrant. */
6100 p2 = (unsigned char*) p + mcnt;
6101 /* Signedness doesn't matter since we only copy MCNT's bits . */
6102 EXTRACT_NUMBER_AND_INCR (mcnt, p);
6103 DEBUG_PRINT3 (" Setting %p to %d.\n", p2, mcnt);
6104 PUSH_NUMBER (p2, mcnt);
6105 break;
6108 case wordbound:
6109 case notwordbound:
6110 not = (re_opcode_t) *(p - 1) == notwordbound;
6111 DEBUG_PRINT2 ("EXECUTING %swordbound.\n", not?"not":"");
6113 /* We SUCCEED (or FAIL) in one of the following cases: */
6115 /* Case 1: D is at the beginning or the end of string. */
6116 if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d))
6117 not = !not;
6118 else
6120 /* C1 is the character before D, S1 is the syntax of C1, C2
6121 is the character at D, and S2 is the syntax of C2. */
6122 re_wchar_t c1, c2;
6123 int s1, s2;
6124 int dummy;
6125 #ifdef emacs
6126 int offset = PTR_TO_OFFSET (d - 1);
6127 int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
6128 UPDATE_SYNTAX_TABLE (charpos);
6129 #endif
6130 GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
6131 s1 = SYNTAX (c1);
6132 #ifdef emacs
6133 UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1);
6134 #endif
6135 PREFETCH_NOLIMIT ();
6136 GET_CHAR_AFTER (c2, d, dummy);
6137 s2 = SYNTAX (c2);
6139 if (/* Case 2: Only one of S1 and S2 is Sword. */
6140 ((s1 == Sword) != (s2 == Sword))
6141 /* Case 3: Both of S1 and S2 are Sword, and macro
6142 WORD_BOUNDARY_P (C1, C2) returns nonzero. */
6143 || ((s1 == Sword) && WORD_BOUNDARY_P (c1, c2)))
6144 not = !not;
6146 if (not)
6147 break;
6148 else
6149 goto fail;
6151 case wordbeg:
6152 DEBUG_PRINT1 ("EXECUTING wordbeg.\n");
6154 /* We FAIL in one of the following cases: */
6156 /* Case 1: D is at the end of string. */
6157 if (AT_STRINGS_END (d))
6158 goto fail;
6159 else
6161 /* C1 is the character before D, S1 is the syntax of C1, C2
6162 is the character at D, and S2 is the syntax of C2. */
6163 re_wchar_t c1, c2;
6164 int s1, s2;
6165 int dummy;
6166 #ifdef emacs
6167 int offset = PTR_TO_OFFSET (d);
6168 int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
6169 UPDATE_SYNTAX_TABLE (charpos);
6170 #endif
6171 PREFETCH ();
6172 GET_CHAR_AFTER (c2, d, dummy);
6173 s2 = SYNTAX (c2);
6175 /* Case 2: S2 is not Sword. */
6176 if (s2 != Sword)
6177 goto fail;
6179 /* Case 3: D is not at the beginning of string ... */
6180 if (!AT_STRINGS_BEG (d))
6182 GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
6183 #ifdef emacs
6184 UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1);
6185 #endif
6186 s1 = SYNTAX (c1);
6188 /* ... and S1 is Sword, and WORD_BOUNDARY_P (C1, C2)
6189 returns 0. */
6190 if ((s1 == Sword) && !WORD_BOUNDARY_P (c1, c2))
6191 goto fail;
6194 break;
6196 case wordend:
6197 DEBUG_PRINT1 ("EXECUTING wordend.\n");
6199 /* We FAIL in one of the following cases: */
6201 /* Case 1: D is at the beginning of string. */
6202 if (AT_STRINGS_BEG (d))
6203 goto fail;
6204 else
6206 /* C1 is the character before D, S1 is the syntax of C1, C2
6207 is the character at D, and S2 is the syntax of C2. */
6208 re_wchar_t c1, c2;
6209 int s1, s2;
6210 int dummy;
6211 #ifdef emacs
6212 int offset = PTR_TO_OFFSET (d) - 1;
6213 int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
6214 UPDATE_SYNTAX_TABLE (charpos);
6215 #endif
6216 GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
6217 s1 = SYNTAX (c1);
6219 /* Case 2: S1 is not Sword. */
6220 if (s1 != Sword)
6221 goto fail;
6223 /* Case 3: D is not at the end of string ... */
6224 if (!AT_STRINGS_END (d))
6226 PREFETCH_NOLIMIT ();
6227 GET_CHAR_AFTER (c2, d, dummy);
6228 #ifdef emacs
6229 UPDATE_SYNTAX_TABLE_FORWARD (charpos);
6230 #endif
6231 s2 = SYNTAX (c2);
6233 /* ... and S2 is Sword, and WORD_BOUNDARY_P (C1, C2)
6234 returns 0. */
6235 if ((s2 == Sword) && !WORD_BOUNDARY_P (c1, c2))
6236 goto fail;
6239 break;
6241 case symbeg:
6242 DEBUG_PRINT1 ("EXECUTING symbeg.\n");
6244 /* We FAIL in one of the following cases: */
6246 /* Case 1: D is at the end of string. */
6247 if (AT_STRINGS_END (d))
6248 goto fail;
6249 else
6251 /* C1 is the character before D, S1 is the syntax of C1, C2
6252 is the character at D, and S2 is the syntax of C2. */
6253 re_wchar_t c1, c2;
6254 int s1, s2;
6255 #ifdef emacs
6256 int offset = PTR_TO_OFFSET (d);
6257 int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
6258 UPDATE_SYNTAX_TABLE (charpos);
6259 #endif
6260 PREFETCH ();
6261 c2 = RE_STRING_CHAR (d, target_multibyte);
6262 s2 = SYNTAX (c2);
6264 /* Case 2: S2 is neither Sword nor Ssymbol. */
6265 if (s2 != Sword && s2 != Ssymbol)
6266 goto fail;
6268 /* Case 3: D is not at the beginning of string ... */
6269 if (!AT_STRINGS_BEG (d))
6271 GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
6272 #ifdef emacs
6273 UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1);
6274 #endif
6275 s1 = SYNTAX (c1);
6277 /* ... and S1 is Sword or Ssymbol. */
6278 if (s1 == Sword || s1 == Ssymbol)
6279 goto fail;
6282 break;
6284 case symend:
6285 DEBUG_PRINT1 ("EXECUTING symend.\n");
6287 /* We FAIL in one of the following cases: */
6289 /* Case 1: D is at the beginning of string. */
6290 if (AT_STRINGS_BEG (d))
6291 goto fail;
6292 else
6294 /* C1 is the character before D, S1 is the syntax of C1, C2
6295 is the character at D, and S2 is the syntax of C2. */
6296 re_wchar_t c1, c2;
6297 int s1, s2;
6298 #ifdef emacs
6299 int offset = PTR_TO_OFFSET (d) - 1;
6300 int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
6301 UPDATE_SYNTAX_TABLE (charpos);
6302 #endif
6303 GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
6304 s1 = SYNTAX (c1);
6306 /* Case 2: S1 is neither Ssymbol nor Sword. */
6307 if (s1 != Sword && s1 != Ssymbol)
6308 goto fail;
6310 /* Case 3: D is not at the end of string ... */
6311 if (!AT_STRINGS_END (d))
6313 PREFETCH_NOLIMIT ();
6314 c2 = RE_STRING_CHAR (d, target_multibyte);
6315 #ifdef emacs
6316 UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1);
6317 #endif
6318 s2 = SYNTAX (c2);
6320 /* ... and S2 is Sword or Ssymbol. */
6321 if (s2 == Sword || s2 == Ssymbol)
6322 goto fail;
6325 break;
6327 case syntaxspec:
6328 case notsyntaxspec:
6329 not = (re_opcode_t) *(p - 1) == notsyntaxspec;
6330 mcnt = *p++;
6331 DEBUG_PRINT3 ("EXECUTING %ssyntaxspec %d.\n", not?"not":"", mcnt);
6332 PREFETCH ();
6333 #ifdef emacs
6335 int offset = PTR_TO_OFFSET (d);
6336 int pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
6337 UPDATE_SYNTAX_TABLE (pos1);
6339 #endif
6341 int len;
6342 re_wchar_t c;
6344 GET_CHAR_AFTER (c, d, len);
6345 if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not)
6346 goto fail;
6347 d += len;
6349 break;
6351 #ifdef emacs
6352 case before_dot:
6353 DEBUG_PRINT1 ("EXECUTING before_dot.\n");
6354 if (PTR_BYTE_POS (d) >= PT_BYTE)
6355 goto fail;
6356 break;
6358 case at_dot:
6359 DEBUG_PRINT1 ("EXECUTING at_dot.\n");
6360 if (PTR_BYTE_POS (d) != PT_BYTE)
6361 goto fail;
6362 break;
6364 case after_dot:
6365 DEBUG_PRINT1 ("EXECUTING after_dot.\n");
6366 if (PTR_BYTE_POS (d) <= PT_BYTE)
6367 goto fail;
6368 break;
6370 case categoryspec:
6371 case notcategoryspec:
6372 not = (re_opcode_t) *(p - 1) == notcategoryspec;
6373 mcnt = *p++;
6374 DEBUG_PRINT3 ("EXECUTING %scategoryspec %d.\n", not?"not":"", mcnt);
6375 PREFETCH ();
6377 int len;
6378 re_wchar_t c;
6380 GET_CHAR_AFTER (c, d, len);
6381 if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not)
6382 goto fail;
6383 d += len;
6385 break;
6387 #endif /* emacs */
6389 default:
6390 abort ();
6392 continue; /* Successfully executed one pattern command; keep going. */
6395 /* We goto here if a matching operation fails. */
6396 fail:
6397 IMMEDIATE_QUIT_CHECK;
6398 if (!FAIL_STACK_EMPTY ())
6400 re_char *str, *pat;
6401 /* A restart point is known. Restore to that state. */
6402 DEBUG_PRINT1 ("\nFAIL:\n");
6403 POP_FAILURE_POINT (str, pat);
6404 switch (SWITCH_ENUM_CAST ((re_opcode_t) *pat++))
6406 case on_failure_keep_string_jump:
6407 assert (str == NULL);
6408 goto continue_failure_jump;
6410 case on_failure_jump_nastyloop:
6411 assert ((re_opcode_t)pat[-2] == no_op);
6412 PUSH_FAILURE_POINT (pat - 2, str);
6413 /* Fallthrough */
6415 case on_failure_jump_loop:
6416 case on_failure_jump:
6417 case succeed_n:
6418 d = str;
6419 continue_failure_jump:
6420 EXTRACT_NUMBER_AND_INCR (mcnt, pat);
6421 p = pat + mcnt;
6422 break;
6424 case no_op:
6425 /* A special frame used for nastyloops. */
6426 goto fail;
6428 default:
6429 abort();
6432 assert (p >= bufp->buffer && p <= pend);
6434 if (d >= string1 && d <= end1)
6435 dend = end_match_1;
6437 else
6438 break; /* Matching at this starting point really fails. */
6439 } /* for (;;) */
6441 if (best_regs_set)
6442 goto restore_best_regs;
6444 FREE_VARIABLES ();
6446 return -1; /* Failure to match. */
6447 } /* re_match_2 */
6449 /* Subroutine definitions for re_match_2. */
6451 /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN
6452 bytes; nonzero otherwise. */
6454 static int
6455 bcmp_translate (s1, s2, len, translate, target_multibyte)
6456 re_char *s1, *s2;
6457 register int len;
6458 RE_TRANSLATE_TYPE translate;
6459 const int target_multibyte;
6461 register re_char *p1 = s1, *p2 = s2;
6462 re_char *p1_end = s1 + len;
6463 re_char *p2_end = s2 + len;
6465 /* FIXME: Checking both p1 and p2 presumes that the two strings might have
6466 different lengths, but relying on a single `len' would break this. -sm */
6467 while (p1 < p1_end && p2 < p2_end)
6469 int p1_charlen, p2_charlen;
6470 re_wchar_t p1_ch, p2_ch;
6472 GET_CHAR_AFTER (p1_ch, p1, p1_charlen);
6473 GET_CHAR_AFTER (p2_ch, p2, p2_charlen);
6475 if (RE_TRANSLATE (translate, p1_ch)
6476 != RE_TRANSLATE (translate, p2_ch))
6477 return 1;
6479 p1 += p1_charlen, p2 += p2_charlen;
6482 if (p1 != p1_end || p2 != p2_end)
6483 return 1;
6485 return 0;
6488 /* Entry points for GNU code. */
6490 /* re_compile_pattern is the GNU regular expression compiler: it
6491 compiles PATTERN (of length SIZE) and puts the result in BUFP.
6492 Returns 0 if the pattern was valid, otherwise an error string.
6494 Assumes the `allocated' (and perhaps `buffer') and `translate' fields
6495 are set in BUFP on entry.
6497 We call regex_compile to do the actual compilation. */
6499 const char *
6500 re_compile_pattern (pattern, length, bufp)
6501 const char *pattern;
6502 size_t length;
6503 struct re_pattern_buffer *bufp;
6505 reg_errcode_t ret;
6507 #ifdef emacs
6508 gl_state.current_syntax_table = current_buffer->syntax_table;
6509 #endif
6511 /* GNU code is written to assume at least RE_NREGS registers will be set
6512 (and at least one extra will be -1). */
6513 bufp->regs_allocated = REGS_UNALLOCATED;
6515 /* And GNU code determines whether or not to get register information
6516 by passing null for the REGS argument to re_match, etc., not by
6517 setting no_sub. */
6518 bufp->no_sub = 0;
6520 ret = regex_compile ((re_char*) pattern, length, re_syntax_options, bufp);
6522 if (!ret)
6523 return NULL;
6524 return gettext (re_error_msgid[(int) ret]);
6526 WEAK_ALIAS (__re_compile_pattern, re_compile_pattern)
6528 /* Entry points compatible with 4.2 BSD regex library. We don't define
6529 them unless specifically requested. */
6531 #if defined _REGEX_RE_COMP || defined _LIBC
6533 /* BSD has one and only one pattern buffer. */
6534 static struct re_pattern_buffer re_comp_buf;
6536 char *
6537 # ifdef _LIBC
6538 /* Make these definitions weak in libc, so POSIX programs can redefine
6539 these names if they don't use our functions, and still use
6540 regcomp/regexec below without link errors. */
6541 weak_function
6542 # endif
6543 re_comp (s)
6544 const char *s;
6546 reg_errcode_t ret;
6548 if (!s)
6550 if (!re_comp_buf.buffer)
6551 /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
6552 return (char *) gettext ("No previous regular expression");
6553 return 0;
6556 if (!re_comp_buf.buffer)
6558 re_comp_buf.buffer = (unsigned char *) malloc (200);
6559 if (re_comp_buf.buffer == NULL)
6560 /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
6561 return (char *) gettext (re_error_msgid[(int) REG_ESPACE]);
6562 re_comp_buf.allocated = 200;
6564 re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH);
6565 if (re_comp_buf.fastmap == NULL)
6566 /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
6567 return (char *) gettext (re_error_msgid[(int) REG_ESPACE]);
6570 /* Since `re_exec' always passes NULL for the `regs' argument, we
6571 don't need to initialize the pattern buffer fields which affect it. */
6573 ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf);
6575 if (!ret)
6576 return NULL;
6578 /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
6579 return (char *) gettext (re_error_msgid[(int) ret]);
6584 # ifdef _LIBC
6585 weak_function
6586 # endif
6587 re_exec (s)
6588 const char *s;
6590 const int len = strlen (s);
6591 return
6592 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0);
6594 #endif /* _REGEX_RE_COMP */
6596 /* POSIX.2 functions. Don't define these for Emacs. */
6598 #ifndef emacs
6600 /* regcomp takes a regular expression as a string and compiles it.
6602 PREG is a regex_t *. We do not expect any fields to be initialized,
6603 since POSIX says we shouldn't. Thus, we set
6605 `buffer' to the compiled pattern;
6606 `used' to the length of the compiled pattern;
6607 `syntax' to RE_SYNTAX_POSIX_EXTENDED if the
6608 REG_EXTENDED bit in CFLAGS is set; otherwise, to
6609 RE_SYNTAX_POSIX_BASIC;
6610 `fastmap' to an allocated space for the fastmap;
6611 `fastmap_accurate' to zero;
6612 `re_nsub' to the number of subexpressions in PATTERN.
6614 PATTERN is the address of the pattern string.
6616 CFLAGS is a series of bits which affect compilation.
6618 If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we
6619 use POSIX basic syntax.
6621 If REG_NEWLINE is set, then . and [^...] don't match newline.
6622 Also, regexec will try a match beginning after every newline.
6624 If REG_ICASE is set, then we considers upper- and lowercase
6625 versions of letters to be equivalent when matching.
6627 If REG_NOSUB is set, then when PREG is passed to regexec, that
6628 routine will report only success or failure, and nothing about the
6629 registers.
6631 It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for
6632 the return codes and their meanings.) */
6635 regcomp (preg, pattern, cflags)
6636 regex_t *__restrict preg;
6637 const char *__restrict pattern;
6638 int cflags;
6640 reg_errcode_t ret;
6641 reg_syntax_t syntax
6642 = (cflags & REG_EXTENDED) ?
6643 RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC;
6645 /* regex_compile will allocate the space for the compiled pattern. */
6646 preg->buffer = 0;
6647 preg->allocated = 0;
6648 preg->used = 0;
6650 /* Try to allocate space for the fastmap. */
6651 preg->fastmap = (char *) malloc (1 << BYTEWIDTH);
6653 if (cflags & REG_ICASE)
6655 unsigned i;
6657 preg->translate
6658 = (RE_TRANSLATE_TYPE) malloc (CHAR_SET_SIZE
6659 * sizeof (*(RE_TRANSLATE_TYPE)0));
6660 if (preg->translate == NULL)
6661 return (int) REG_ESPACE;
6663 /* Map uppercase characters to corresponding lowercase ones. */
6664 for (i = 0; i < CHAR_SET_SIZE; i++)
6665 preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i;
6667 else
6668 preg->translate = NULL;
6670 /* If REG_NEWLINE is set, newlines are treated differently. */
6671 if (cflags & REG_NEWLINE)
6672 { /* REG_NEWLINE implies neither . nor [^...] match newline. */
6673 syntax &= ~RE_DOT_NEWLINE;
6674 syntax |= RE_HAT_LISTS_NOT_NEWLINE;
6676 else
6677 syntax |= RE_NO_NEWLINE_ANCHOR;
6679 preg->no_sub = !!(cflags & REG_NOSUB);
6681 /* POSIX says a null character in the pattern terminates it, so we
6682 can use strlen here in compiling the pattern. */
6683 ret = regex_compile ((re_char*) pattern, strlen (pattern), syntax, preg);
6685 /* POSIX doesn't distinguish between an unmatched open-group and an
6686 unmatched close-group: both are REG_EPAREN. */
6687 if (ret == REG_ERPAREN)
6688 ret = REG_EPAREN;
6690 if (ret == REG_NOERROR && preg->fastmap)
6691 { /* Compute the fastmap now, since regexec cannot modify the pattern
6692 buffer. */
6693 re_compile_fastmap (preg);
6694 if (preg->can_be_null)
6695 { /* The fastmap can't be used anyway. */
6696 free (preg->fastmap);
6697 preg->fastmap = NULL;
6700 return (int) ret;
6702 WEAK_ALIAS (__regcomp, regcomp)
6705 /* regexec searches for a given pattern, specified by PREG, in the
6706 string STRING.
6708 If NMATCH is zero or REG_NOSUB was set in the cflags argument to
6709 `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at
6710 least NMATCH elements, and we set them to the offsets of the
6711 corresponding matched substrings.
6713 EFLAGS specifies `execution flags' which affect matching: if
6714 REG_NOTBOL is set, then ^ does not match at the beginning of the
6715 string; if REG_NOTEOL is set, then $ does not match at the end.
6717 We return 0 if we find a match and REG_NOMATCH if not. */
6720 regexec (preg, string, nmatch, pmatch, eflags)
6721 const regex_t *__restrict preg;
6722 const char *__restrict string;
6723 size_t nmatch;
6724 regmatch_t pmatch[__restrict_arr];
6725 int eflags;
6727 int ret;
6728 struct re_registers regs;
6729 regex_t private_preg;
6730 int len = strlen (string);
6731 boolean want_reg_info = !preg->no_sub && nmatch > 0 && pmatch;
6733 private_preg = *preg;
6735 private_preg.not_bol = !!(eflags & REG_NOTBOL);
6736 private_preg.not_eol = !!(eflags & REG_NOTEOL);
6738 /* The user has told us exactly how many registers to return
6739 information about, via `nmatch'. We have to pass that on to the
6740 matching routines. */
6741 private_preg.regs_allocated = REGS_FIXED;
6743 if (want_reg_info)
6745 regs.num_regs = nmatch;
6746 regs.start = TALLOC (nmatch * 2, regoff_t);
6747 if (regs.start == NULL)
6748 return (int) REG_NOMATCH;
6749 regs.end = regs.start + nmatch;
6752 /* Instead of using not_eol to implement REG_NOTEOL, we could simply
6753 pass (&private_preg, string, len + 1, 0, len, ...) pretending the string
6754 was a little bit longer but still only matching the real part.
6755 This works because the `endline' will check for a '\n' and will find a
6756 '\0', correctly deciding that this is not the end of a line.
6757 But it doesn't work out so nicely for REG_NOTBOL, since we don't have
6758 a convenient '\0' there. For all we know, the string could be preceded
6759 by '\n' which would throw things off. */
6761 /* Perform the searching operation. */
6762 ret = re_search (&private_preg, string, len,
6763 /* start: */ 0, /* range: */ len,
6764 want_reg_info ? &regs : (struct re_registers *) 0);
6766 /* Copy the register information to the POSIX structure. */
6767 if (want_reg_info)
6769 if (ret >= 0)
6771 unsigned r;
6773 for (r = 0; r < nmatch; r++)
6775 pmatch[r].rm_so = regs.start[r];
6776 pmatch[r].rm_eo = regs.end[r];
6780 /* If we needed the temporary register info, free the space now. */
6781 free (regs.start);
6784 /* We want zero return to mean success, unlike `re_search'. */
6785 return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH;
6787 WEAK_ALIAS (__regexec, regexec)
6790 /* Returns a message corresponding to an error code, ERR_CODE, returned
6791 from either regcomp or regexec. We don't use PREG here.
6793 ERR_CODE was previously called ERRCODE, but that name causes an
6794 error with msvc8 compiler. */
6796 size_t
6797 regerror (err_code, preg, errbuf, errbuf_size)
6798 int err_code;
6799 const regex_t *preg;
6800 char *errbuf;
6801 size_t errbuf_size;
6803 const char *msg;
6804 size_t msg_size;
6806 if (err_code < 0
6807 || err_code >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0])))
6808 /* Only error codes returned by the rest of the code should be passed
6809 to this routine. If we are given anything else, or if other regex
6810 code generates an invalid error code, then the program has a bug.
6811 Dump core so we can fix it. */
6812 abort ();
6814 msg = gettext (re_error_msgid[err_code]);
6816 msg_size = strlen (msg) + 1; /* Includes the null. */
6818 if (errbuf_size != 0)
6820 if (msg_size > errbuf_size)
6822 strncpy (errbuf, msg, errbuf_size - 1);
6823 errbuf[errbuf_size - 1] = 0;
6825 else
6826 strcpy (errbuf, msg);
6829 return msg_size;
6831 WEAK_ALIAS (__regerror, regerror)
6834 /* Free dynamically allocated space used by PREG. */
6836 void
6837 regfree (preg)
6838 regex_t *preg;
6840 free (preg->buffer);
6841 preg->buffer = NULL;
6843 preg->allocated = 0;
6844 preg->used = 0;
6846 free (preg->fastmap);
6847 preg->fastmap = NULL;
6848 preg->fastmap_accurate = 0;
6850 free (preg->translate);
6851 preg->translate = NULL;
6853 WEAK_ALIAS (__regfree, regfree)
6855 #endif /* not emacs */
6857 /* arch-tag: 4ffd68ba-2a9e-435b-a21a-018990f9eeb2
6858 (do not change this comment) */