Commit | Line | Data |
---|---|---|
e318085a | 1 | /* Extended regular expression matching and search library, version |
0b32bf0e | 2 | 0.12. (Implements POSIX draft P1003.2/D11.2, except for some of the |
bc78d348 KB |
3 | internationalization features.) |
4 | ||
0b5538bd | 5 | Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, |
114f9c96 | 6 | 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010 |
e468b87f | 7 | Free Software Foundation, Inc. |
bc78d348 | 8 | |
fa9a63c5 RM |
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 | |
e468b87f | 11 | the Free Software Foundation; either version 3, or (at your option) |
fa9a63c5 RM |
12 | any later version. |
13 | ||
14 | This program is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
7814e705 | 16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
fa9a63c5 RM |
17 | GNU General Public License for more details. |
18 | ||
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 | |
4fc5845f | 21 | Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, |
7814e705 | 22 | USA. */ |
fa9a63c5 | 23 | |
6df42991 | 24 | /* TODO: |
505bde11 | 25 | - structure the opcode space into opcode+flag. |
dc1e502d | 26 | - merge with glibc's regex.[ch]. |
01618498 | 27 | - replace (succeed_n + jump_n + set_number_at) with something that doesn't |
6dcf2d0e SM |
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. | |
01618498 | 31 | */ |
505bde11 | 32 | |
fa9a63c5 | 33 | /* AIX requires this to be the first thing in the file. */ |
0b32bf0e | 34 | #if defined _AIX && !defined REGEX_MALLOC |
fa9a63c5 RM |
35 | #pragma alloca |
36 | #endif | |
37 | ||
fa9a63c5 | 38 | #ifdef HAVE_CONFIG_H |
0b32bf0e | 39 | # include <config.h> |
fa9a63c5 RM |
40 | #endif |
41 | ||
4bb91c68 SM |
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 | |
fa9a63c5 | 48 | |
14473664 SM |
49 | /* Whether to use ISO C Amendment 1 wide char functions. |
50 | Those should not be used for Emacs since it uses its own. */ | |
5e5388f6 GM |
51 | #if defined _LIBC |
52 | #define WIDE_CHAR_SUPPORT 1 | |
53 | #else | |
14473664 | 54 | #define WIDE_CHAR_SUPPORT \ |
5e5388f6 GM |
55 | (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC && !emacs) |
56 | #endif | |
14473664 SM |
57 | |
58 | /* For platform which support the ISO C amendement 1 functionality we | |
59 | support user defined character classes. */ | |
a0ad02f7 | 60 | #if WIDE_CHAR_SUPPORT |
14473664 SM |
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 | |
65 | ||
c0f9ea08 SM |
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) | |
ec869672 JR |
71 | # define regerror(err_code, preg, errbuf, errbuf_size) \ |
72 | __regerror(err_code, preg, errbuf, errbuf_size) | |
c0f9ea08 SM |
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) | |
87 | ||
14473664 SM |
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 | |
92 | ||
c0f9ea08 SM |
93 | # define WEAK_ALIAS(a,b) weak_alias (a, b) |
94 | ||
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 | |
102 | ||
4bb91c68 | 103 | /* This is for other GNU distributions with internationalized messages. */ |
0b32bf0e | 104 | #if HAVE_LIBINTL_H || defined _LIBC |
fa9a63c5 RM |
105 | # include <libintl.h> |
106 | #else | |
107 | # define gettext(msgid) (msgid) | |
108 | #endif | |
109 | ||
5e69f11e RM |
110 | #ifndef gettext_noop |
111 | /* This define is so xgettext can find the internationalizable | |
112 | strings. */ | |
0b32bf0e | 113 | # define gettext_noop(String) String |
5e69f11e RM |
114 | #endif |
115 | ||
fa9a63c5 RM |
116 | /* The `emacs' switch turns on certain matching commands |
117 | that make sense only in Emacs. */ | |
118 | #ifdef emacs | |
119 | ||
d7306fe6 | 120 | # include <setjmp.h> |
0b32bf0e SM |
121 | # include "lisp.h" |
122 | # include "buffer.h" | |
b18215fc RS |
123 | |
124 | /* Make syntax table lookup grant data in gl_state. */ | |
0b32bf0e | 125 | # define SYNTAX_ENTRY_VIA_PROPERTY |
b18215fc | 126 | |
0b32bf0e | 127 | # include "syntax.h" |
9117d724 | 128 | # include "character.h" |
0b32bf0e | 129 | # include "category.h" |
fa9a63c5 | 130 | |
7689ef0b EZ |
131 | # ifdef malloc |
132 | # undef malloc | |
133 | # endif | |
0b32bf0e | 134 | # define malloc xmalloc |
7689ef0b EZ |
135 | # ifdef realloc |
136 | # undef realloc | |
137 | # endif | |
0b32bf0e | 138 | # define realloc xrealloc |
7689ef0b EZ |
139 | # ifdef free |
140 | # undef free | |
141 | # endif | |
0b32bf0e | 142 | # define free xfree |
9abbd165 | 143 | |
7814e705 | 144 | /* Converts the pointer to the char to BEG-based offset from the start. */ |
0b32bf0e SM |
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))) | |
147 | ||
148 | # define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte) | |
bf216479 | 149 | # define RE_TARGET_MULTIBYTE_P(bufp) ((bufp)->target_multibyte) |
62a6e103 AS |
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))) | |
2d1675e4 | 154 | |
4c0354d7 | 155 | # define RE_CHAR_TO_MULTIBYTE(c) UNIBYTE_TO_CHAR (c) |
cf9c99bc | 156 | |
2afc21f5 | 157 | # define RE_CHAR_TO_UNIBYTE(c) CHAR_TO_BYTE_SAFE (c) |
cf9c99bc | 158 | |
6fdd04b0 KH |
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). */ | |
bf216479 KH |
162 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
163 | do { \ | |
02cb78b5 | 164 | if (target_multibyte) \ |
bf216479 KH |
165 | { \ |
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)); \ | |
62a6e103 | 169 | c = STRING_CHAR (dtemp); \ |
bf216479 KH |
170 | } \ |
171 | else \ | |
172 | { \ | |
173 | (c = ((p) == (str2) ? (end1) : (p))[-1]); \ | |
cf9c99bc | 174 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
bf216479 | 175 | } \ |
2d1675e4 SM |
176 | } while (0) |
177 | ||
6fdd04b0 KH |
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 { \ | |
02cb78b5 | 182 | if (target_multibyte) \ |
62a6e103 | 183 | (c) = STRING_CHAR_AND_LENGTH (p, len); \ |
6fdd04b0 KH |
184 | else \ |
185 | { \ | |
cf9c99bc | 186 | (c) = *p; \ |
6fdd04b0 | 187 | len = 1; \ |
cf9c99bc | 188 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
6fdd04b0 | 189 | } \ |
8f924df7 | 190 | } while (0) |
4e8a9132 | 191 | |
fa9a63c5 RM |
192 | #else /* not emacs */ |
193 | ||
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. */ | |
0b32bf0e | 197 | # undef REL_ALLOC |
fa9a63c5 | 198 | |
0b32bf0e SM |
199 | # if defined STDC_HEADERS || defined _LIBC |
200 | # include <stdlib.h> | |
201 | # else | |
fa9a63c5 RM |
202 | char *malloc (); |
203 | char *realloc (); | |
0b32bf0e | 204 | # endif |
fa9a63c5 | 205 | |
a77f947b CY |
206 | /* When used in Emacs's lib-src, we need xmalloc and xrealloc. */ |
207 | ||
208 | void * | |
209 | xmalloc (size) | |
210 | size_t size; | |
211 | { | |
212 | register void *val; | |
213 | val = (void *) malloc (size); | |
214 | if (!val && size) | |
215 | { | |
216 | write (2, "virtual memory exhausted\n", 25); | |
217 | exit (1); | |
218 | } | |
219 | return val; | |
220 | } | |
221 | ||
222 | void * | |
223 | xrealloc (block, size) | |
224 | void *block; | |
225 | size_t size; | |
226 | { | |
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) | |
235 | { | |
236 | write (2, "virtual memory exhausted\n", 25); | |
237 | exit (1); | |
238 | } | |
239 | return val; | |
240 | } | |
241 | ||
a073faa6 CY |
242 | # ifdef malloc |
243 | # undef malloc | |
244 | # endif | |
245 | # define malloc xmalloc | |
246 | # ifdef realloc | |
247 | # undef realloc | |
248 | # endif | |
249 | # define realloc xrealloc | |
250 | ||
9e4ecb26 | 251 | /* When used in Emacs's lib-src, we need to get bzero and bcopy somehow. |
4bb91c68 | 252 | If nothing else has been done, use the method below. */ |
0b32bf0e SM |
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 | |
9e4ecb26 | 260 | |
4bb91c68 | 261 | /* This is the normal way of making sure we have memcpy, memcmp and bzero. |
9e4ecb26 KH |
262 | This is used in most programs--a few other programs avoid this |
263 | by defining INHIBIT_STRING_HEADER. */ | |
0b32bf0e SM |
264 | # ifndef INHIBIT_STRING_HEADER |
265 | # if defined HAVE_STRING_H || defined STDC_HEADERS || defined _LIBC | |
266 | # include <string.h> | |
0b32bf0e | 267 | # ifndef bzero |
4bb91c68 SM |
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 | |
0b32bf0e SM |
273 | # endif |
274 | # else | |
275 | # include <strings.h> | |
4bb91c68 SM |
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 | |
0b32bf0e SM |
282 | # endif |
283 | # endif | |
fa9a63c5 RM |
284 | |
285 | /* Define the syntax stuff for \<, \>, etc. */ | |
286 | ||
990b2375 | 287 | /* Sword must be nonzero for the wordchar pattern commands in re_match_2. */ |
669fa600 | 288 | enum syntaxcode { Swhitespace = 0, Sword = 1, Ssymbol = 2 }; |
fa9a63c5 | 289 | |
0b32bf0e | 290 | # define SWITCH_ENUM_CAST(x) (x) |
fa9a63c5 | 291 | |
e934739e | 292 | /* Dummy macros for non-Emacs environments. */ |
0b32bf0e SM |
293 | # define CHAR_CHARSET(c) 0 |
294 | # define CHARSET_LEADING_CODE_BASE(c) 0 | |
295 | # define MAX_MULTIBYTE_LENGTH 1 | |
296 | # define RE_MULTIBYTE_P(x) 0 | |
bf216479 | 297 | # define RE_TARGET_MULTIBYTE_P(x) 0 |
0b32bf0e SM |
298 | # define WORD_BOUNDARY_P(c1, c2) (0) |
299 | # define CHAR_HEAD_P(p) (1) | |
300 | # define SINGLE_BYTE_CHAR_P(c) (1) | |
301 | # define SAME_CHARSET_P(c1, c2) (1) | |
aa3830c4 | 302 | # define BYTES_BY_CHAR_HEAD(p) (1) |
70806df6 | 303 | # define PREV_CHAR_BOUNDARY(p, limit) ((p)--) |
62a6e103 AS |
304 | # define STRING_CHAR(p) (*(p)) |
305 | # define RE_STRING_CHAR(p, multibyte) STRING_CHAR (p) | |
0b32bf0e | 306 | # define CHAR_STRING(c, s) (*(s) = (c), 1) |
62a6e103 AS |
307 | # define STRING_CHAR_AND_LENGTH(p, actual_len) ((actual_len) = 1, *(p)) |
308 | # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) STRING_CHAR_AND_LENGTH (p, len) | |
cf9c99bc KH |
309 | # define RE_CHAR_TO_MULTIBYTE(c) (c) |
310 | # define RE_CHAR_TO_UNIBYTE(c) (c) | |
0b32bf0e | 311 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
b18215fc | 312 | (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1))) |
6fdd04b0 KH |
313 | # define GET_CHAR_AFTER(c, p, len) \ |
314 | (c = *p, len = 1) | |
0b32bf0e | 315 | # define MAKE_CHAR(charset, c1, c2) (c1) |
9117d724 KH |
316 | # define BYTE8_TO_CHAR(c) (c) |
317 | # define CHAR_BYTE8_P(c) (0) | |
bf216479 | 318 | # define CHAR_LEADING_CODE(c) (c) |
8f924df7 | 319 | |
fa9a63c5 | 320 | #endif /* not emacs */ |
4e8a9132 SM |
321 | |
322 | #ifndef RE_TRANSLATE | |
0b32bf0e SM |
323 | # define RE_TRANSLATE(TBL, C) ((unsigned char)(TBL)[C]) |
324 | # define RE_TRANSLATE_P(TBL) (TBL) | |
4e8a9132 | 325 | #endif |
fa9a63c5 RM |
326 | \f |
327 | /* Get the interface, including the syntax bits. */ | |
328 | #include "regex.h" | |
329 | ||
f71b19b6 DL |
330 | /* isalpha etc. are used for the character classes. */ |
331 | #include <ctype.h> | |
fa9a63c5 | 332 | |
f71b19b6 | 333 | #ifdef emacs |
fa9a63c5 | 334 | |
f71b19b6 | 335 | /* 1 if C is an ASCII character. */ |
0b32bf0e | 336 | # define IS_REAL_ASCII(c) ((c) < 0200) |
fa9a63c5 | 337 | |
f71b19b6 | 338 | /* 1 if C is a unibyte character. */ |
0b32bf0e | 339 | # define ISUNIBYTE(c) (SINGLE_BYTE_CHAR_P ((c))) |
96cc36cc | 340 | |
f71b19b6 | 341 | /* The Emacs definitions should not be directly affected by locales. */ |
96cc36cc | 342 | |
f71b19b6 | 343 | /* In Emacs, these are only used for single-byte characters. */ |
0b32bf0e SM |
344 | # define ISDIGIT(c) ((c) >= '0' && (c) <= '9') |
345 | # define ISCNTRL(c) ((c) < ' ') | |
346 | # define ISXDIGIT(c) (((c) >= '0' && (c) <= '9') \ | |
f71b19b6 DL |
347 | || ((c) >= 'a' && (c) <= 'f') \ |
348 | || ((c) >= 'A' && (c) <= 'F')) | |
96cc36cc RS |
349 | |
350 | /* This is only used for single-byte characters. */ | |
0b32bf0e | 351 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') |
96cc36cc RS |
352 | |
353 | /* The rest must handle multibyte characters. */ | |
354 | ||
0b32bf0e | 355 | # define ISGRAPH(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 356 | ? (c) > ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
357 | : 1) |
358 | ||
14473664 | 359 | # define ISPRINT(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 360 | ? (c) >= ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
361 | : 1) |
362 | ||
0b32bf0e | 363 | # define ISALNUM(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
364 | ? (((c) >= 'a' && (c) <= 'z') \ |
365 | || ((c) >= 'A' && (c) <= 'Z') \ | |
366 | || ((c) >= '0' && (c) <= '9')) \ | |
96cc36cc RS |
367 | : SYNTAX (c) == Sword) |
368 | ||
0b32bf0e | 369 | # define ISALPHA(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
370 | ? (((c) >= 'a' && (c) <= 'z') \ |
371 | || ((c) >= 'A' && (c) <= 'Z')) \ | |
96cc36cc RS |
372 | : SYNTAX (c) == Sword) |
373 | ||
0b32bf0e | 374 | # define ISLOWER(c) (LOWERCASEP (c)) |
96cc36cc | 375 | |
0b32bf0e | 376 | # define ISPUNCT(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
377 | ? ((c) > ' ' && (c) < 0177 \ |
378 | && !(((c) >= 'a' && (c) <= 'z') \ | |
4bb91c68 SM |
379 | || ((c) >= 'A' && (c) <= 'Z') \ |
380 | || ((c) >= '0' && (c) <= '9'))) \ | |
96cc36cc RS |
381 | : SYNTAX (c) != Sword) |
382 | ||
0b32bf0e | 383 | # define ISSPACE(c) (SYNTAX (c) == Swhitespace) |
96cc36cc | 384 | |
0b32bf0e | 385 | # define ISUPPER(c) (UPPERCASEP (c)) |
96cc36cc | 386 | |
0b32bf0e | 387 | # define ISWORD(c) (SYNTAX (c) == Sword) |
96cc36cc RS |
388 | |
389 | #else /* not emacs */ | |
390 | ||
f71b19b6 DL |
391 | /* Jim Meyering writes: |
392 | ||
393 | "... Some ctype macros are valid only for character codes that | |
394 | isascii says are ASCII (SGI's IRIX-4.0.5 is one such system --when | |
395 | using /bin/cc or gcc but without giving an ansi option). So, all | |
4bb91c68 | 396 | ctype uses should be through macros like ISPRINT... If |
f71b19b6 DL |
397 | STDC_HEADERS is defined, then autoconf has verified that the ctype |
398 | macros don't need to be guarded with references to isascii. ... | |
399 | Defining isascii to 1 should let any compiler worth its salt | |
4bb91c68 SM |
400 | eliminate the && through constant folding." |
401 | Solaris defines some of these symbols so we must undefine them first. */ | |
f71b19b6 | 402 | |
4bb91c68 | 403 | # undef ISASCII |
0b32bf0e SM |
404 | # if defined STDC_HEADERS || (!defined isascii && !defined HAVE_ISASCII) |
405 | # define ISASCII(c) 1 | |
406 | # else | |
407 | # define ISASCII(c) isascii(c) | |
408 | # endif | |
f71b19b6 DL |
409 | |
410 | /* 1 if C is an ASCII character. */ | |
0b32bf0e | 411 | # define IS_REAL_ASCII(c) ((c) < 0200) |
f71b19b6 DL |
412 | |
413 | /* This distinction is not meaningful, except in Emacs. */ | |
0b32bf0e SM |
414 | # define ISUNIBYTE(c) 1 |
415 | ||
416 | # ifdef isblank | |
417 | # define ISBLANK(c) (ISASCII (c) && isblank (c)) | |
418 | # else | |
419 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') | |
420 | # endif | |
421 | # ifdef isgraph | |
422 | # define ISGRAPH(c) (ISASCII (c) && isgraph (c)) | |
423 | # else | |
424 | # define ISGRAPH(c) (ISASCII (c) && isprint (c) && !isspace (c)) | |
425 | # endif | |
426 | ||
4bb91c68 | 427 | # undef ISPRINT |
0b32bf0e SM |
428 | # define ISPRINT(c) (ISASCII (c) && isprint (c)) |
429 | # define ISDIGIT(c) (ISASCII (c) && isdigit (c)) | |
430 | # define ISALNUM(c) (ISASCII (c) && isalnum (c)) | |
431 | # define ISALPHA(c) (ISASCII (c) && isalpha (c)) | |
432 | # define ISCNTRL(c) (ISASCII (c) && iscntrl (c)) | |
433 | # define ISLOWER(c) (ISASCII (c) && islower (c)) | |
434 | # define ISPUNCT(c) (ISASCII (c) && ispunct (c)) | |
435 | # define ISSPACE(c) (ISASCII (c) && isspace (c)) | |
436 | # define ISUPPER(c) (ISASCII (c) && isupper (c)) | |
437 | # define ISXDIGIT(c) (ISASCII (c) && isxdigit (c)) | |
438 | ||
439 | # define ISWORD(c) ISALPHA(c) | |
440 | ||
4bb91c68 SM |
441 | # ifdef _tolower |
442 | # define TOLOWER(c) _tolower(c) | |
443 | # else | |
444 | # define TOLOWER(c) tolower(c) | |
445 | # endif | |
446 | ||
447 | /* How many characters in the character set. */ | |
448 | # define CHAR_SET_SIZE 256 | |
449 | ||
0b32bf0e | 450 | # ifdef SYNTAX_TABLE |
f71b19b6 | 451 | |
0b32bf0e | 452 | extern char *re_syntax_table; |
f71b19b6 | 453 | |
0b32bf0e SM |
454 | # else /* not SYNTAX_TABLE */ |
455 | ||
0b32bf0e SM |
456 | static char re_syntax_table[CHAR_SET_SIZE]; |
457 | ||
458 | static void | |
459 | init_syntax_once () | |
460 | { | |
461 | register int c; | |
462 | static int done = 0; | |
463 | ||
464 | if (done) | |
465 | return; | |
466 | ||
467 | bzero (re_syntax_table, sizeof re_syntax_table); | |
468 | ||
4bb91c68 SM |
469 | for (c = 0; c < CHAR_SET_SIZE; ++c) |
470 | if (ISALNUM (c)) | |
471 | re_syntax_table[c] = Sword; | |
fa9a63c5 | 472 | |
669fa600 | 473 | re_syntax_table['_'] = Ssymbol; |
fa9a63c5 | 474 | |
0b32bf0e SM |
475 | done = 1; |
476 | } | |
477 | ||
478 | # endif /* not SYNTAX_TABLE */ | |
96cc36cc | 479 | |
4bb91c68 SM |
480 | # define SYNTAX(c) re_syntax_table[(c)] |
481 | ||
96cc36cc RS |
482 | #endif /* not emacs */ |
483 | \f | |
fa9a63c5 | 484 | #ifndef NULL |
0b32bf0e | 485 | # define NULL (void *)0 |
fa9a63c5 RM |
486 | #endif |
487 | ||
488 | /* We remove any previous definition of `SIGN_EXTEND_CHAR', | |
489 | since ours (we hope) works properly with all combinations of | |
490 | machines, compilers, `char' and `unsigned char' argument types. | |
4bb91c68 | 491 | (Per Bothner suggested the basic approach.) */ |
fa9a63c5 RM |
492 | #undef SIGN_EXTEND_CHAR |
493 | #if __STDC__ | |
0b32bf0e | 494 | # define SIGN_EXTEND_CHAR(c) ((signed char) (c)) |
fa9a63c5 RM |
495 | #else /* not __STDC__ */ |
496 | /* As in Harbison and Steele. */ | |
0b32bf0e | 497 | # define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128) |
fa9a63c5 RM |
498 | #endif |
499 | \f | |
500 | /* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we | |
501 | use `alloca' instead of `malloc'. This is because using malloc in | |
502 | re_search* or re_match* could cause memory leaks when C-g is used in | |
503 | Emacs; also, malloc is slower and causes storage fragmentation. On | |
5e69f11e RM |
504 | the other hand, malloc is more portable, and easier to debug. |
505 | ||
fa9a63c5 RM |
506 | Because we sometimes use alloca, some routines have to be macros, |
507 | not functions -- `alloca'-allocated space disappears at the end of the | |
508 | function it is called in. */ | |
509 | ||
510 | #ifdef REGEX_MALLOC | |
511 | ||
0b32bf0e SM |
512 | # define REGEX_ALLOCATE malloc |
513 | # define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize) | |
514 | # define REGEX_FREE free | |
fa9a63c5 RM |
515 | |
516 | #else /* not REGEX_MALLOC */ | |
517 | ||
518 | /* Emacs already defines alloca, sometimes. */ | |
0b32bf0e | 519 | # ifndef alloca |
fa9a63c5 RM |
520 | |
521 | /* Make alloca work the best possible way. */ | |
0b32bf0e SM |
522 | # ifdef __GNUC__ |
523 | # define alloca __builtin_alloca | |
524 | # else /* not __GNUC__ */ | |
7f585e7a | 525 | # ifdef HAVE_ALLOCA_H |
0b32bf0e SM |
526 | # include <alloca.h> |
527 | # endif /* HAVE_ALLOCA_H */ | |
528 | # endif /* not __GNUC__ */ | |
fa9a63c5 | 529 | |
0b32bf0e | 530 | # endif /* not alloca */ |
fa9a63c5 | 531 | |
0b32bf0e | 532 | # define REGEX_ALLOCATE alloca |
fa9a63c5 RM |
533 | |
534 | /* Assumes a `char *destination' variable. */ | |
0b32bf0e | 535 | # define REGEX_REALLOCATE(source, osize, nsize) \ |
fa9a63c5 | 536 | (destination = (char *) alloca (nsize), \ |
4bb91c68 | 537 | memcpy (destination, source, osize)) |
fa9a63c5 RM |
538 | |
539 | /* No need to do anything to free, after alloca. */ | |
0b32bf0e | 540 | # define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
541 | |
542 | #endif /* not REGEX_MALLOC */ | |
543 | ||
544 | /* Define how to allocate the failure stack. */ | |
545 | ||
0b32bf0e | 546 | #if defined REL_ALLOC && defined REGEX_MALLOC |
4297555e | 547 | |
0b32bf0e | 548 | # define REGEX_ALLOCATE_STACK(size) \ |
fa9a63c5 | 549 | r_alloc (&failure_stack_ptr, (size)) |
0b32bf0e | 550 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 551 | r_re_alloc (&failure_stack_ptr, (nsize)) |
0b32bf0e | 552 | # define REGEX_FREE_STACK(ptr) \ |
fa9a63c5 RM |
553 | r_alloc_free (&failure_stack_ptr) |
554 | ||
4297555e | 555 | #else /* not using relocating allocator */ |
fa9a63c5 | 556 | |
0b32bf0e | 557 | # ifdef REGEX_MALLOC |
fa9a63c5 | 558 | |
0b32bf0e SM |
559 | # define REGEX_ALLOCATE_STACK malloc |
560 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize) | |
561 | # define REGEX_FREE_STACK free | |
fa9a63c5 | 562 | |
0b32bf0e | 563 | # else /* not REGEX_MALLOC */ |
fa9a63c5 | 564 | |
0b32bf0e | 565 | # define REGEX_ALLOCATE_STACK alloca |
fa9a63c5 | 566 | |
0b32bf0e | 567 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 568 | REGEX_REALLOCATE (source, osize, nsize) |
7814e705 | 569 | /* No need to explicitly free anything. */ |
0b32bf0e | 570 | # define REGEX_FREE_STACK(arg) ((void)0) |
fa9a63c5 | 571 | |
0b32bf0e | 572 | # endif /* not REGEX_MALLOC */ |
4297555e | 573 | #endif /* not using relocating allocator */ |
fa9a63c5 RM |
574 | |
575 | ||
576 | /* True if `size1' is non-NULL and PTR is pointing anywhere inside | |
577 | `string1' or just past its end. This works if PTR is NULL, which is | |
578 | a good thing. */ | |
25fe55af | 579 | #define FIRST_STRING_P(ptr) \ |
fa9a63c5 RM |
580 | (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) |
581 | ||
582 | /* (Re)Allocate N items of type T using malloc, or fail. */ | |
583 | #define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t))) | |
584 | #define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t))) | |
585 | #define RETALLOC_IF(addr, n, t) \ | |
586 | if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t) | |
587 | #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) | |
588 | ||
4bb91c68 | 589 | #define BYTEWIDTH 8 /* In bits. */ |
fa9a63c5 RM |
590 | |
591 | #define STREQ(s1, s2) ((strcmp (s1, s2) == 0)) | |
592 | ||
593 | #undef MAX | |
594 | #undef MIN | |
595 | #define MAX(a, b) ((a) > (b) ? (a) : (b)) | |
596 | #define MIN(a, b) ((a) < (b) ? (a) : (b)) | |
597 | ||
66f0296e SM |
598 | /* Type of source-pattern and string chars. */ |
599 | typedef const unsigned char re_char; | |
600 | ||
fa9a63c5 RM |
601 | typedef char boolean; |
602 | #define false 0 | |
603 | #define true 1 | |
604 | ||
4bb91c68 SM |
605 | static int re_match_2_internal _RE_ARGS ((struct re_pattern_buffer *bufp, |
606 | re_char *string1, int size1, | |
607 | re_char *string2, int size2, | |
608 | int pos, | |
609 | struct re_registers *regs, | |
610 | int stop)); | |
fa9a63c5 RM |
611 | \f |
612 | /* These are the command codes that appear in compiled regular | |
4bb91c68 | 613 | expressions. Some opcodes are followed by argument bytes. A |
fa9a63c5 RM |
614 | command code can specify any interpretation whatsoever for its |
615 | arguments. Zero bytes may appear in the compiled regular expression. */ | |
616 | ||
617 | typedef enum | |
618 | { | |
619 | no_op = 0, | |
620 | ||
4bb91c68 | 621 | /* Succeed right away--no more backtracking. */ |
fa9a63c5 RM |
622 | succeed, |
623 | ||
25fe55af | 624 | /* Followed by one byte giving n, then by n literal bytes. */ |
fa9a63c5 RM |
625 | exactn, |
626 | ||
25fe55af | 627 | /* Matches any (more or less) character. */ |
fa9a63c5 RM |
628 | anychar, |
629 | ||
25fe55af RS |
630 | /* Matches any one char belonging to specified set. First |
631 | following byte is number of bitmap bytes. Then come bytes | |
632 | for a bitmap saying which chars are in. Bits in each byte | |
633 | are ordered low-bit-first. A character is in the set if its | |
634 | bit is 1. A character too large to have a bit in the map is | |
96cc36cc RS |
635 | automatically not in the set. |
636 | ||
637 | If the length byte has the 0x80 bit set, then that stuff | |
638 | is followed by a range table: | |
639 | 2 bytes of flags for character sets (low 8 bits, high 8 bits) | |
0b32bf0e | 640 | See RANGE_TABLE_WORK_BITS below. |
01618498 | 641 | 2 bytes, the number of pairs that follow (upto 32767) |
96cc36cc | 642 | pairs, each 2 multibyte characters, |
0b32bf0e | 643 | each multibyte character represented as 3 bytes. */ |
fa9a63c5 RM |
644 | charset, |
645 | ||
25fe55af | 646 | /* Same parameters as charset, but match any character that is |
4bb91c68 | 647 | not one of those specified. */ |
fa9a63c5 RM |
648 | charset_not, |
649 | ||
25fe55af RS |
650 | /* Start remembering the text that is matched, for storing in a |
651 | register. Followed by one byte with the register number, in | |
652 | the range 0 to one less than the pattern buffer's re_nsub | |
505bde11 | 653 | field. */ |
fa9a63c5 RM |
654 | start_memory, |
655 | ||
25fe55af RS |
656 | /* Stop remembering the text that is matched and store it in a |
657 | memory register. Followed by one byte with the register | |
658 | number, in the range 0 to one less than `re_nsub' in the | |
505bde11 | 659 | pattern buffer. */ |
fa9a63c5 RM |
660 | stop_memory, |
661 | ||
25fe55af | 662 | /* Match a duplicate of something remembered. Followed by one |
4bb91c68 | 663 | byte containing the register number. */ |
fa9a63c5 RM |
664 | duplicate, |
665 | ||
25fe55af | 666 | /* Fail unless at beginning of line. */ |
fa9a63c5 RM |
667 | begline, |
668 | ||
4bb91c68 | 669 | /* Fail unless at end of line. */ |
fa9a63c5 RM |
670 | endline, |
671 | ||
25fe55af RS |
672 | /* Succeeds if at beginning of buffer (if emacs) or at beginning |
673 | of string to be matched (if not). */ | |
fa9a63c5 RM |
674 | begbuf, |
675 | ||
25fe55af | 676 | /* Analogously, for end of buffer/string. */ |
fa9a63c5 | 677 | endbuf, |
5e69f11e | 678 | |
25fe55af | 679 | /* Followed by two byte relative address to which to jump. */ |
5e69f11e | 680 | jump, |
fa9a63c5 | 681 | |
25fe55af | 682 | /* Followed by two-byte relative address of place to resume at |
7814e705 | 683 | in case of failure. */ |
fa9a63c5 | 684 | on_failure_jump, |
5e69f11e | 685 | |
25fe55af RS |
686 | /* Like on_failure_jump, but pushes a placeholder instead of the |
687 | current string position when executed. */ | |
fa9a63c5 | 688 | on_failure_keep_string_jump, |
5e69f11e | 689 | |
505bde11 SM |
690 | /* Just like `on_failure_jump', except that it checks that we |
691 | don't get stuck in an infinite loop (matching an empty string | |
692 | indefinitely). */ | |
693 | on_failure_jump_loop, | |
694 | ||
0683b6fa SM |
695 | /* Just like `on_failure_jump_loop', except that it checks for |
696 | a different kind of loop (the kind that shows up with non-greedy | |
697 | operators). This operation has to be immediately preceded | |
698 | by a `no_op'. */ | |
699 | on_failure_jump_nastyloop, | |
700 | ||
0b32bf0e | 701 | /* A smart `on_failure_jump' used for greedy * and + operators. |
505bde11 SM |
702 | It analyses the loop before which it is put and if the |
703 | loop does not require backtracking, it changes itself to | |
4e8a9132 SM |
704 | `on_failure_keep_string_jump' and short-circuits the loop, |
705 | else it just defaults to changing itself into `on_failure_jump'. | |
706 | It assumes that it is pointing to just past a `jump'. */ | |
505bde11 | 707 | on_failure_jump_smart, |
fa9a63c5 | 708 | |
25fe55af | 709 | /* Followed by two-byte relative address and two-byte number n. |
ed0767d8 SM |
710 | After matching N times, jump to the address upon failure. |
711 | Does not work if N starts at 0: use on_failure_jump_loop | |
712 | instead. */ | |
fa9a63c5 RM |
713 | succeed_n, |
714 | ||
25fe55af RS |
715 | /* Followed by two-byte relative address, and two-byte number n. |
716 | Jump to the address N times, then fail. */ | |
fa9a63c5 RM |
717 | jump_n, |
718 | ||
25fe55af | 719 | /* Set the following two-byte relative address to the |
7814e705 | 720 | subsequent two-byte number. The address *includes* the two |
25fe55af | 721 | bytes of number. */ |
fa9a63c5 RM |
722 | set_number_at, |
723 | ||
fa9a63c5 RM |
724 | wordbeg, /* Succeeds if at word beginning. */ |
725 | wordend, /* Succeeds if at word end. */ | |
726 | ||
727 | wordbound, /* Succeeds if at a word boundary. */ | |
7814e705 | 728 | notwordbound, /* Succeeds if not at a word boundary. */ |
fa9a63c5 | 729 | |
669fa600 SM |
730 | symbeg, /* Succeeds if at symbol beginning. */ |
731 | symend, /* Succeeds if at symbol end. */ | |
732 | ||
fa9a63c5 | 733 | /* Matches any character whose syntax is specified. Followed by |
25fe55af | 734 | a byte which contains a syntax code, e.g., Sword. */ |
fa9a63c5 RM |
735 | syntaxspec, |
736 | ||
737 | /* Matches any character whose syntax is not that specified. */ | |
1fb352e0 SM |
738 | notsyntaxspec |
739 | ||
740 | #ifdef emacs | |
741 | ,before_dot, /* Succeeds if before point. */ | |
742 | at_dot, /* Succeeds if at point. */ | |
743 | after_dot, /* Succeeds if after point. */ | |
b18215fc RS |
744 | |
745 | /* Matches any character whose category-set contains the specified | |
7814e705 JB |
746 | category. The operator is followed by a byte which contains a |
747 | category code (mnemonic ASCII character). */ | |
b18215fc RS |
748 | categoryspec, |
749 | ||
750 | /* Matches any character whose category-set does not contain the | |
751 | specified category. The operator is followed by a byte which | |
752 | contains the category code (mnemonic ASCII character). */ | |
753 | notcategoryspec | |
fa9a63c5 RM |
754 | #endif /* emacs */ |
755 | } re_opcode_t; | |
756 | \f | |
757 | /* Common operations on the compiled pattern. */ | |
758 | ||
759 | /* Store NUMBER in two contiguous bytes starting at DESTINATION. */ | |
760 | ||
761 | #define STORE_NUMBER(destination, number) \ | |
762 | do { \ | |
763 | (destination)[0] = (number) & 0377; \ | |
764 | (destination)[1] = (number) >> 8; \ | |
765 | } while (0) | |
766 | ||
767 | /* Same as STORE_NUMBER, except increment DESTINATION to | |
768 | the byte after where the number is stored. Therefore, DESTINATION | |
769 | must be an lvalue. */ | |
770 | ||
771 | #define STORE_NUMBER_AND_INCR(destination, number) \ | |
772 | do { \ | |
773 | STORE_NUMBER (destination, number); \ | |
774 | (destination) += 2; \ | |
775 | } while (0) | |
776 | ||
777 | /* Put into DESTINATION a number stored in two contiguous bytes starting | |
778 | at SOURCE. */ | |
779 | ||
780 | #define EXTRACT_NUMBER(destination, source) \ | |
781 | do { \ | |
782 | (destination) = *(source) & 0377; \ | |
783 | (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \ | |
784 | } while (0) | |
785 | ||
786 | #ifdef DEBUG | |
4bb91c68 | 787 | static void extract_number _RE_ARGS ((int *dest, re_char *source)); |
fa9a63c5 RM |
788 | static void |
789 | extract_number (dest, source) | |
790 | int *dest; | |
01618498 | 791 | re_char *source; |
fa9a63c5 | 792 | { |
5e69f11e | 793 | int temp = SIGN_EXTEND_CHAR (*(source + 1)); |
fa9a63c5 RM |
794 | *dest = *source & 0377; |
795 | *dest += temp << 8; | |
796 | } | |
797 | ||
4bb91c68 | 798 | # ifndef EXTRACT_MACROS /* To debug the macros. */ |
0b32bf0e SM |
799 | # undef EXTRACT_NUMBER |
800 | # define EXTRACT_NUMBER(dest, src) extract_number (&dest, src) | |
801 | # endif /* not EXTRACT_MACROS */ | |
fa9a63c5 RM |
802 | |
803 | #endif /* DEBUG */ | |
804 | ||
805 | /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. | |
806 | SOURCE must be an lvalue. */ | |
807 | ||
808 | #define EXTRACT_NUMBER_AND_INCR(destination, source) \ | |
809 | do { \ | |
810 | EXTRACT_NUMBER (destination, source); \ | |
25fe55af | 811 | (source) += 2; \ |
fa9a63c5 RM |
812 | } while (0) |
813 | ||
814 | #ifdef DEBUG | |
4bb91c68 SM |
815 | static void extract_number_and_incr _RE_ARGS ((int *destination, |
816 | re_char **source)); | |
fa9a63c5 RM |
817 | static void |
818 | extract_number_and_incr (destination, source) | |
819 | int *destination; | |
01618498 | 820 | re_char **source; |
5e69f11e | 821 | { |
fa9a63c5 RM |
822 | extract_number (destination, *source); |
823 | *source += 2; | |
824 | } | |
825 | ||
0b32bf0e SM |
826 | # ifndef EXTRACT_MACROS |
827 | # undef EXTRACT_NUMBER_AND_INCR | |
828 | # define EXTRACT_NUMBER_AND_INCR(dest, src) \ | |
fa9a63c5 | 829 | extract_number_and_incr (&dest, &src) |
0b32bf0e | 830 | # endif /* not EXTRACT_MACROS */ |
fa9a63c5 RM |
831 | |
832 | #endif /* DEBUG */ | |
833 | \f | |
b18215fc RS |
834 | /* Store a multibyte character in three contiguous bytes starting |
835 | DESTINATION, and increment DESTINATION to the byte after where the | |
7814e705 | 836 | character is stored. Therefore, DESTINATION must be an lvalue. */ |
b18215fc RS |
837 | |
838 | #define STORE_CHARACTER_AND_INCR(destination, character) \ | |
839 | do { \ | |
840 | (destination)[0] = (character) & 0377; \ | |
841 | (destination)[1] = ((character) >> 8) & 0377; \ | |
842 | (destination)[2] = (character) >> 16; \ | |
843 | (destination) += 3; \ | |
844 | } while (0) | |
845 | ||
846 | /* Put into DESTINATION a character stored in three contiguous bytes | |
7814e705 | 847 | starting at SOURCE. */ |
b18215fc RS |
848 | |
849 | #define EXTRACT_CHARACTER(destination, source) \ | |
850 | do { \ | |
851 | (destination) = ((source)[0] \ | |
852 | | ((source)[1] << 8) \ | |
853 | | ((source)[2] << 16)); \ | |
854 | } while (0) | |
855 | ||
856 | ||
857 | /* Macros for charset. */ | |
858 | ||
859 | /* Size of bitmap of charset P in bytes. P is a start of charset, | |
860 | i.e. *P is (re_opcode_t) charset or (re_opcode_t) charset_not. */ | |
861 | #define CHARSET_BITMAP_SIZE(p) ((p)[1] & 0x7F) | |
862 | ||
863 | /* Nonzero if charset P has range table. */ | |
25fe55af | 864 | #define CHARSET_RANGE_TABLE_EXISTS_P(p) ((p)[1] & 0x80) |
b18215fc RS |
865 | |
866 | /* Return the address of range table of charset P. But not the start | |
867 | of table itself, but the before where the number of ranges is | |
96cc36cc RS |
868 | stored. `2 +' means to skip re_opcode_t and size of bitmap, |
869 | and the 2 bytes of flags at the start of the range table. */ | |
870 | #define CHARSET_RANGE_TABLE(p) (&(p)[4 + CHARSET_BITMAP_SIZE (p)]) | |
871 | ||
872 | /* Extract the bit flags that start a range table. */ | |
873 | #define CHARSET_RANGE_TABLE_BITS(p) \ | |
874 | ((p)[2 + CHARSET_BITMAP_SIZE (p)] \ | |
875 | + (p)[3 + CHARSET_BITMAP_SIZE (p)] * 0x100) | |
b18215fc RS |
876 | |
877 | /* Test if C is listed in the bitmap of charset P. */ | |
878 | #define CHARSET_LOOKUP_BITMAP(p, c) \ | |
879 | ((c) < CHARSET_BITMAP_SIZE (p) * BYTEWIDTH \ | |
880 | && (p)[2 + (c) / BYTEWIDTH] & (1 << ((c) % BYTEWIDTH))) | |
881 | ||
882 | /* Return the address of end of RANGE_TABLE. COUNT is number of | |
7814e705 JB |
883 | ranges (which is a pair of (start, end)) in the RANGE_TABLE. `* 2' |
884 | is start of range and end of range. `* 3' is size of each start | |
b18215fc RS |
885 | and end. */ |
886 | #define CHARSET_RANGE_TABLE_END(range_table, count) \ | |
887 | ((range_table) + (count) * 2 * 3) | |
888 | ||
7814e705 | 889 | /* Test if C is in RANGE_TABLE. A flag NOT is negated if C is in. |
b18215fc RS |
890 | COUNT is number of ranges in RANGE_TABLE. */ |
891 | #define CHARSET_LOOKUP_RANGE_TABLE_RAW(not, c, range_table, count) \ | |
892 | do \ | |
893 | { \ | |
01618498 SM |
894 | re_wchar_t range_start, range_end; \ |
895 | re_char *p; \ | |
896 | re_char *range_table_end \ | |
b18215fc RS |
897 | = CHARSET_RANGE_TABLE_END ((range_table), (count)); \ |
898 | \ | |
899 | for (p = (range_table); p < range_table_end; p += 2 * 3) \ | |
900 | { \ | |
901 | EXTRACT_CHARACTER (range_start, p); \ | |
902 | EXTRACT_CHARACTER (range_end, p + 3); \ | |
903 | \ | |
904 | if (range_start <= (c) && (c) <= range_end) \ | |
905 | { \ | |
906 | (not) = !(not); \ | |
907 | break; \ | |
908 | } \ | |
909 | } \ | |
910 | } \ | |
911 | while (0) | |
912 | ||
913 | /* Test if C is in range table of CHARSET. The flag NOT is negated if | |
914 | C is listed in it. */ | |
915 | #define CHARSET_LOOKUP_RANGE_TABLE(not, c, charset) \ | |
916 | do \ | |
917 | { \ | |
918 | /* Number of ranges in range table. */ \ | |
919 | int count; \ | |
01618498 SM |
920 | re_char *range_table = CHARSET_RANGE_TABLE (charset); \ |
921 | \ | |
b18215fc RS |
922 | EXTRACT_NUMBER_AND_INCR (count, range_table); \ |
923 | CHARSET_LOOKUP_RANGE_TABLE_RAW ((not), (c), range_table, count); \ | |
924 | } \ | |
925 | while (0) | |
926 | \f | |
fa9a63c5 RM |
927 | /* If DEBUG is defined, Regex prints many voluminous messages about what |
928 | it is doing (if the variable `debug' is nonzero). If linked with the | |
929 | main program in `iregex.c', you can enter patterns and strings | |
930 | interactively. And if linked with the main program in `main.c' and | |
4bb91c68 | 931 | the other test files, you can run the already-written tests. */ |
fa9a63c5 RM |
932 | |
933 | #ifdef DEBUG | |
934 | ||
935 | /* We use standard I/O for debugging. */ | |
0b32bf0e | 936 | # include <stdio.h> |
fa9a63c5 RM |
937 | |
938 | /* It is useful to test things that ``must'' be true when debugging. */ | |
0b32bf0e | 939 | # include <assert.h> |
fa9a63c5 | 940 | |
99633e97 | 941 | static int debug = -100000; |
fa9a63c5 | 942 | |
0b32bf0e SM |
943 | # define DEBUG_STATEMENT(e) e |
944 | # define DEBUG_PRINT1(x) if (debug > 0) printf (x) | |
945 | # define DEBUG_PRINT2(x1, x2) if (debug > 0) printf (x1, x2) | |
946 | # define DEBUG_PRINT3(x1, x2, x3) if (debug > 0) printf (x1, x2, x3) | |
947 | # define DEBUG_PRINT4(x1, x2, x3, x4) if (debug > 0) printf (x1, x2, x3, x4) | |
948 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ | |
99633e97 | 949 | if (debug > 0) print_partial_compiled_pattern (s, e) |
0b32bf0e | 950 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ |
99633e97 | 951 | if (debug > 0) print_double_string (w, s1, sz1, s2, sz2) |
fa9a63c5 RM |
952 | |
953 | ||
954 | /* Print the fastmap in human-readable form. */ | |
955 | ||
956 | void | |
957 | print_fastmap (fastmap) | |
958 | char *fastmap; | |
959 | { | |
960 | unsigned was_a_range = 0; | |
5e69f11e RM |
961 | unsigned i = 0; |
962 | ||
fa9a63c5 RM |
963 | while (i < (1 << BYTEWIDTH)) |
964 | { | |
965 | if (fastmap[i++]) | |
966 | { | |
967 | was_a_range = 0; | |
25fe55af RS |
968 | putchar (i - 1); |
969 | while (i < (1 << BYTEWIDTH) && fastmap[i]) | |
970 | { | |
971 | was_a_range = 1; | |
972 | i++; | |
973 | } | |
fa9a63c5 | 974 | if (was_a_range) |
25fe55af RS |
975 | { |
976 | printf ("-"); | |
977 | putchar (i - 1); | |
978 | } | |
979 | } | |
fa9a63c5 | 980 | } |
5e69f11e | 981 | putchar ('\n'); |
fa9a63c5 RM |
982 | } |
983 | ||
984 | ||
985 | /* Print a compiled pattern string in human-readable form, starting at | |
986 | the START pointer into it and ending just before the pointer END. */ | |
987 | ||
988 | void | |
989 | print_partial_compiled_pattern (start, end) | |
01618498 SM |
990 | re_char *start; |
991 | re_char *end; | |
fa9a63c5 RM |
992 | { |
993 | int mcnt, mcnt2; | |
01618498 SM |
994 | re_char *p = start; |
995 | re_char *pend = end; | |
fa9a63c5 RM |
996 | |
997 | if (start == NULL) | |
998 | { | |
a1a052df | 999 | fprintf (stderr, "(null)\n"); |
fa9a63c5 RM |
1000 | return; |
1001 | } | |
5e69f11e | 1002 | |
fa9a63c5 RM |
1003 | /* Loop over pattern commands. */ |
1004 | while (p < pend) | |
1005 | { | |
a1a052df | 1006 | fprintf (stderr, "%d:\t", p - start); |
fa9a63c5 RM |
1007 | |
1008 | switch ((re_opcode_t) *p++) | |
1009 | { | |
25fe55af | 1010 | case no_op: |
a1a052df | 1011 | fprintf (stderr, "/no_op"); |
25fe55af | 1012 | break; |
fa9a63c5 | 1013 | |
99633e97 | 1014 | case succeed: |
a1a052df | 1015 | fprintf (stderr, "/succeed"); |
99633e97 SM |
1016 | break; |
1017 | ||
fa9a63c5 RM |
1018 | case exactn: |
1019 | mcnt = *p++; | |
a1a052df | 1020 | fprintf (stderr, "/exactn/%d", mcnt); |
25fe55af | 1021 | do |
fa9a63c5 | 1022 | { |
a1a052df | 1023 | fprintf (stderr, "/%c", *p++); |
25fe55af RS |
1024 | } |
1025 | while (--mcnt); | |
1026 | break; | |
fa9a63c5 RM |
1027 | |
1028 | case start_memory: | |
a1a052df | 1029 | fprintf (stderr, "/start_memory/%d", *p++); |
25fe55af | 1030 | break; |
fa9a63c5 RM |
1031 | |
1032 | case stop_memory: | |
a1a052df | 1033 | fprintf (stderr, "/stop_memory/%d", *p++); |
25fe55af | 1034 | break; |
fa9a63c5 RM |
1035 | |
1036 | case duplicate: | |
a1a052df | 1037 | fprintf (stderr, "/duplicate/%d", *p++); |
fa9a63c5 RM |
1038 | break; |
1039 | ||
1040 | case anychar: | |
a1a052df | 1041 | fprintf (stderr, "/anychar"); |
fa9a63c5 RM |
1042 | break; |
1043 | ||
1044 | case charset: | |
25fe55af RS |
1045 | case charset_not: |
1046 | { | |
1047 | register int c, last = -100; | |
fa9a63c5 | 1048 | register int in_range = 0; |
99633e97 SM |
1049 | int length = CHARSET_BITMAP_SIZE (p - 1); |
1050 | int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1); | |
fa9a63c5 | 1051 | |
a1a052df | 1052 | fprintf (stderr, "/charset [%s", |
839966f3 | 1053 | (re_opcode_t) *(p - 1) == charset_not ? "^" : ""); |
5e69f11e | 1054 | |
839966f3 KH |
1055 | if (p + *p >= pend) |
1056 | fprintf (stderr, " !extends past end of pattern! "); | |
fa9a63c5 | 1057 | |
25fe55af | 1058 | for (c = 0; c < 256; c++) |
96cc36cc | 1059 | if (c / 8 < length |
fa9a63c5 RM |
1060 | && (p[1 + (c/8)] & (1 << (c % 8)))) |
1061 | { | |
1062 | /* Are we starting a range? */ | |
1063 | if (last + 1 == c && ! in_range) | |
1064 | { | |
a1a052df | 1065 | fprintf (stderr, "-"); |
fa9a63c5 RM |
1066 | in_range = 1; |
1067 | } | |
1068 | /* Have we broken a range? */ | |
1069 | else if (last + 1 != c && in_range) | |
96cc36cc | 1070 | { |
a1a052df | 1071 | fprintf (stderr, "%c", last); |
fa9a63c5 RM |
1072 | in_range = 0; |
1073 | } | |
5e69f11e | 1074 | |
fa9a63c5 | 1075 | if (! in_range) |
a1a052df | 1076 | fprintf (stderr, "%c", c); |
fa9a63c5 RM |
1077 | |
1078 | last = c; | |
25fe55af | 1079 | } |
fa9a63c5 RM |
1080 | |
1081 | if (in_range) | |
a1a052df | 1082 | fprintf (stderr, "%c", last); |
fa9a63c5 | 1083 | |
a1a052df | 1084 | fprintf (stderr, "]"); |
fa9a63c5 | 1085 | |
99633e97 | 1086 | p += 1 + length; |
96cc36cc | 1087 | |
96cc36cc | 1088 | if (has_range_table) |
99633e97 SM |
1089 | { |
1090 | int count; | |
a1a052df | 1091 | fprintf (stderr, "has-range-table"); |
99633e97 SM |
1092 | |
1093 | /* ??? Should print the range table; for now, just skip it. */ | |
1094 | p += 2; /* skip range table bits */ | |
1095 | EXTRACT_NUMBER_AND_INCR (count, p); | |
1096 | p = CHARSET_RANGE_TABLE_END (p, count); | |
1097 | } | |
fa9a63c5 RM |
1098 | } |
1099 | break; | |
1100 | ||
1101 | case begline: | |
a1a052df | 1102 | fprintf (stderr, "/begline"); |
25fe55af | 1103 | break; |
fa9a63c5 RM |
1104 | |
1105 | case endline: | |
a1a052df | 1106 | fprintf (stderr, "/endline"); |
25fe55af | 1107 | break; |
fa9a63c5 RM |
1108 | |
1109 | case on_failure_jump: | |
25fe55af | 1110 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1111 | fprintf (stderr, "/on_failure_jump to %d", p + mcnt - start); |
25fe55af | 1112 | break; |
fa9a63c5 RM |
1113 | |
1114 | case on_failure_keep_string_jump: | |
25fe55af | 1115 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1116 | fprintf (stderr, "/on_failure_keep_string_jump to %d", p + mcnt - start); |
25fe55af | 1117 | break; |
fa9a63c5 | 1118 | |
0683b6fa SM |
1119 | case on_failure_jump_nastyloop: |
1120 | extract_number_and_incr (&mcnt, &p); | |
a1a052df | 1121 | fprintf (stderr, "/on_failure_jump_nastyloop to %d", p + mcnt - start); |
0683b6fa SM |
1122 | break; |
1123 | ||
505bde11 | 1124 | case on_failure_jump_loop: |
fa9a63c5 | 1125 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1126 | fprintf (stderr, "/on_failure_jump_loop to %d", p + mcnt - start); |
5e69f11e RM |
1127 | break; |
1128 | ||
505bde11 | 1129 | case on_failure_jump_smart: |
fa9a63c5 | 1130 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1131 | fprintf (stderr, "/on_failure_jump_smart to %d", p + mcnt - start); |
5e69f11e RM |
1132 | break; |
1133 | ||
25fe55af | 1134 | case jump: |
fa9a63c5 | 1135 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1136 | fprintf (stderr, "/jump to %d", p + mcnt - start); |
fa9a63c5 RM |
1137 | break; |
1138 | ||
25fe55af RS |
1139 | case succeed_n: |
1140 | extract_number_and_incr (&mcnt, &p); | |
1141 | extract_number_and_incr (&mcnt2, &p); | |
a1a052df | 1142 | fprintf (stderr, "/succeed_n to %d, %d times", p - 2 + mcnt - start, mcnt2); |
25fe55af | 1143 | break; |
5e69f11e | 1144 | |
25fe55af RS |
1145 | case jump_n: |
1146 | extract_number_and_incr (&mcnt, &p); | |
1147 | extract_number_and_incr (&mcnt2, &p); | |
a1a052df | 1148 | fprintf (stderr, "/jump_n to %d, %d times", p - 2 + mcnt - start, mcnt2); |
25fe55af | 1149 | break; |
5e69f11e | 1150 | |
25fe55af RS |
1151 | case set_number_at: |
1152 | extract_number_and_incr (&mcnt, &p); | |
1153 | extract_number_and_incr (&mcnt2, &p); | |
a1a052df | 1154 | fprintf (stderr, "/set_number_at location %d to %d", p - 2 + mcnt - start, mcnt2); |
25fe55af | 1155 | break; |
5e69f11e | 1156 | |
25fe55af | 1157 | case wordbound: |
a1a052df | 1158 | fprintf (stderr, "/wordbound"); |
fa9a63c5 RM |
1159 | break; |
1160 | ||
1161 | case notwordbound: | |
a1a052df | 1162 | fprintf (stderr, "/notwordbound"); |
25fe55af | 1163 | break; |
fa9a63c5 RM |
1164 | |
1165 | case wordbeg: | |
a1a052df | 1166 | fprintf (stderr, "/wordbeg"); |
fa9a63c5 | 1167 | break; |
5e69f11e | 1168 | |
fa9a63c5 | 1169 | case wordend: |
a1a052df | 1170 | fprintf (stderr, "/wordend"); |
e2543b02 | 1171 | break; |
5e69f11e | 1172 | |
669fa600 | 1173 | case symbeg: |
e2543b02 | 1174 | fprintf (stderr, "/symbeg"); |
669fa600 SM |
1175 | break; |
1176 | ||
1177 | case symend: | |
e2543b02 | 1178 | fprintf (stderr, "/symend"); |
669fa600 | 1179 | break; |
5e69f11e | 1180 | |
1fb352e0 | 1181 | case syntaxspec: |
a1a052df | 1182 | fprintf (stderr, "/syntaxspec"); |
1fb352e0 | 1183 | mcnt = *p++; |
a1a052df | 1184 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1185 | break; |
1186 | ||
1187 | case notsyntaxspec: | |
a1a052df | 1188 | fprintf (stderr, "/notsyntaxspec"); |
1fb352e0 | 1189 | mcnt = *p++; |
a1a052df | 1190 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1191 | break; |
1192 | ||
0b32bf0e | 1193 | # ifdef emacs |
fa9a63c5 | 1194 | case before_dot: |
a1a052df | 1195 | fprintf (stderr, "/before_dot"); |
25fe55af | 1196 | break; |
fa9a63c5 RM |
1197 | |
1198 | case at_dot: | |
a1a052df | 1199 | fprintf (stderr, "/at_dot"); |
25fe55af | 1200 | break; |
fa9a63c5 RM |
1201 | |
1202 | case after_dot: | |
a1a052df | 1203 | fprintf (stderr, "/after_dot"); |
25fe55af | 1204 | break; |
fa9a63c5 | 1205 | |
1fb352e0 | 1206 | case categoryspec: |
a1a052df | 1207 | fprintf (stderr, "/categoryspec"); |
fa9a63c5 | 1208 | mcnt = *p++; |
a1a052df | 1209 | fprintf (stderr, "/%d", mcnt); |
25fe55af | 1210 | break; |
5e69f11e | 1211 | |
1fb352e0 | 1212 | case notcategoryspec: |
a1a052df | 1213 | fprintf (stderr, "/notcategoryspec"); |
fa9a63c5 | 1214 | mcnt = *p++; |
a1a052df | 1215 | fprintf (stderr, "/%d", mcnt); |
fa9a63c5 | 1216 | break; |
0b32bf0e | 1217 | # endif /* emacs */ |
fa9a63c5 | 1218 | |
fa9a63c5 | 1219 | case begbuf: |
a1a052df | 1220 | fprintf (stderr, "/begbuf"); |
25fe55af | 1221 | break; |
fa9a63c5 RM |
1222 | |
1223 | case endbuf: | |
a1a052df | 1224 | fprintf (stderr, "/endbuf"); |
25fe55af | 1225 | break; |
fa9a63c5 | 1226 | |
25fe55af | 1227 | default: |
a1a052df | 1228 | fprintf (stderr, "?%d", *(p-1)); |
fa9a63c5 RM |
1229 | } |
1230 | ||
a1a052df | 1231 | fprintf (stderr, "\n"); |
fa9a63c5 RM |
1232 | } |
1233 | ||
a1a052df | 1234 | fprintf (stderr, "%d:\tend of pattern.\n", p - start); |
fa9a63c5 RM |
1235 | } |
1236 | ||
1237 | ||
1238 | void | |
1239 | print_compiled_pattern (bufp) | |
1240 | struct re_pattern_buffer *bufp; | |
1241 | { | |
01618498 | 1242 | re_char *buffer = bufp->buffer; |
fa9a63c5 RM |
1243 | |
1244 | print_partial_compiled_pattern (buffer, buffer + bufp->used); | |
4bb91c68 SM |
1245 | printf ("%ld bytes used/%ld bytes allocated.\n", |
1246 | bufp->used, bufp->allocated); | |
fa9a63c5 RM |
1247 | |
1248 | if (bufp->fastmap_accurate && bufp->fastmap) | |
1249 | { | |
1250 | printf ("fastmap: "); | |
1251 | print_fastmap (bufp->fastmap); | |
1252 | } | |
1253 | ||
1254 | printf ("re_nsub: %d\t", bufp->re_nsub); | |
1255 | printf ("regs_alloc: %d\t", bufp->regs_allocated); | |
1256 | printf ("can_be_null: %d\t", bufp->can_be_null); | |
fa9a63c5 RM |
1257 | printf ("no_sub: %d\t", bufp->no_sub); |
1258 | printf ("not_bol: %d\t", bufp->not_bol); | |
1259 | printf ("not_eol: %d\t", bufp->not_eol); | |
4bb91c68 | 1260 | printf ("syntax: %lx\n", bufp->syntax); |
505bde11 | 1261 | fflush (stdout); |
fa9a63c5 RM |
1262 | /* Perhaps we should print the translate table? */ |
1263 | } | |
1264 | ||
1265 | ||
1266 | void | |
1267 | print_double_string (where, string1, size1, string2, size2) | |
66f0296e SM |
1268 | re_char *where; |
1269 | re_char *string1; | |
1270 | re_char *string2; | |
fa9a63c5 RM |
1271 | int size1; |
1272 | int size2; | |
1273 | { | |
4bb91c68 | 1274 | int this_char; |
5e69f11e | 1275 | |
fa9a63c5 RM |
1276 | if (where == NULL) |
1277 | printf ("(null)"); | |
1278 | else | |
1279 | { | |
1280 | if (FIRST_STRING_P (where)) | |
25fe55af RS |
1281 | { |
1282 | for (this_char = where - string1; this_char < size1; this_char++) | |
1283 | putchar (string1[this_char]); | |
fa9a63c5 | 1284 | |
25fe55af RS |
1285 | where = string2; |
1286 | } | |
fa9a63c5 RM |
1287 | |
1288 | for (this_char = where - string2; this_char < size2; this_char++) | |
25fe55af | 1289 | putchar (string2[this_char]); |
fa9a63c5 RM |
1290 | } |
1291 | } | |
1292 | ||
1293 | #else /* not DEBUG */ | |
1294 | ||
0b32bf0e SM |
1295 | # undef assert |
1296 | # define assert(e) | |
fa9a63c5 | 1297 | |
0b32bf0e SM |
1298 | # define DEBUG_STATEMENT(e) |
1299 | # define DEBUG_PRINT1(x) | |
1300 | # define DEBUG_PRINT2(x1, x2) | |
1301 | # define DEBUG_PRINT3(x1, x2, x3) | |
1302 | # define DEBUG_PRINT4(x1, x2, x3, x4) | |
1303 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) | |
1304 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
fa9a63c5 RM |
1305 | |
1306 | #endif /* not DEBUG */ | |
1307 | \f | |
1308 | /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can | |
1309 | also be assigned to arbitrarily: each pattern buffer stores its own | |
1310 | syntax, so it can be changed between regex compilations. */ | |
1311 | /* This has no initializer because initialized variables in Emacs | |
1312 | become read-only after dumping. */ | |
1313 | reg_syntax_t re_syntax_options; | |
1314 | ||
1315 | ||
1316 | /* Specify the precise syntax of regexps for compilation. This provides | |
1317 | for compatibility for various utilities which historically have | |
1318 | different, incompatible syntaxes. | |
1319 | ||
1320 | The argument SYNTAX is a bit mask comprised of the various bits | |
4bb91c68 | 1321 | defined in regex.h. We return the old syntax. */ |
fa9a63c5 RM |
1322 | |
1323 | reg_syntax_t | |
1324 | re_set_syntax (syntax) | |
f9b0fd99 | 1325 | reg_syntax_t syntax; |
fa9a63c5 RM |
1326 | { |
1327 | reg_syntax_t ret = re_syntax_options; | |
5e69f11e | 1328 | |
fa9a63c5 RM |
1329 | re_syntax_options = syntax; |
1330 | return ret; | |
1331 | } | |
c0f9ea08 | 1332 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) |
f9b0fd99 RS |
1333 | |
1334 | /* Regexp to use to replace spaces, or NULL meaning don't. */ | |
1335 | static re_char *whitespace_regexp; | |
1336 | ||
1337 | void | |
1338 | re_set_whitespace_regexp (regexp) | |
6470ea05 | 1339 | const char *regexp; |
f9b0fd99 | 1340 | { |
6470ea05 | 1341 | whitespace_regexp = (re_char *) regexp; |
f9b0fd99 RS |
1342 | } |
1343 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) | |
fa9a63c5 RM |
1344 | \f |
1345 | /* This table gives an error message for each of the error codes listed | |
4bb91c68 | 1346 | in regex.h. Obviously the order here has to be same as there. |
fa9a63c5 | 1347 | POSIX doesn't require that we do anything for REG_NOERROR, |
4bb91c68 | 1348 | but why not be nice? */ |
fa9a63c5 RM |
1349 | |
1350 | static const char *re_error_msgid[] = | |
5e69f11e RM |
1351 | { |
1352 | gettext_noop ("Success"), /* REG_NOERROR */ | |
1353 | gettext_noop ("No match"), /* REG_NOMATCH */ | |
1354 | gettext_noop ("Invalid regular expression"), /* REG_BADPAT */ | |
1355 | gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */ | |
1356 | gettext_noop ("Invalid character class name"), /* REG_ECTYPE */ | |
1357 | gettext_noop ("Trailing backslash"), /* REG_EESCAPE */ | |
1358 | gettext_noop ("Invalid back reference"), /* REG_ESUBREG */ | |
1359 | gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */ | |
1360 | gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */ | |
1361 | gettext_noop ("Unmatched \\{"), /* REG_EBRACE */ | |
1362 | gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */ | |
1363 | gettext_noop ("Invalid range end"), /* REG_ERANGE */ | |
1364 | gettext_noop ("Memory exhausted"), /* REG_ESPACE */ | |
1365 | gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */ | |
1366 | gettext_noop ("Premature end of regular expression"), /* REG_EEND */ | |
1367 | gettext_noop ("Regular expression too big"), /* REG_ESIZE */ | |
1368 | gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */ | |
b3e4c897 | 1369 | gettext_noop ("Range striding over charsets") /* REG_ERANGEX */ |
fa9a63c5 RM |
1370 | }; |
1371 | \f | |
4bb91c68 | 1372 | /* Avoiding alloca during matching, to placate r_alloc. */ |
fa9a63c5 RM |
1373 | |
1374 | /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the | |
1375 | searching and matching functions should not call alloca. On some | |
1376 | systems, alloca is implemented in terms of malloc, and if we're | |
1377 | using the relocating allocator routines, then malloc could cause a | |
1378 | relocation, which might (if the strings being searched are in the | |
1379 | ralloc heap) shift the data out from underneath the regexp | |
1380 | routines. | |
1381 | ||
5e69f11e | 1382 | Here's another reason to avoid allocation: Emacs |
fa9a63c5 RM |
1383 | processes input from X in a signal handler; processing X input may |
1384 | call malloc; if input arrives while a matching routine is calling | |
1385 | malloc, then we're scrod. But Emacs can't just block input while | |
1386 | calling matching routines; then we don't notice interrupts when | |
1387 | they come in. So, Emacs blocks input around all regexp calls | |
1388 | except the matching calls, which it leaves unprotected, in the | |
1389 | faith that they will not malloc. */ | |
1390 | ||
1391 | /* Normally, this is fine. */ | |
1392 | #define MATCH_MAY_ALLOCATE | |
1393 | ||
fa9a63c5 RM |
1394 | /* The match routines may not allocate if (1) they would do it with malloc |
1395 | and (2) it's not safe for them to use malloc. | |
1396 | Note that if REL_ALLOC is defined, matching would not use malloc for the | |
1397 | failure stack, but we would still use it for the register vectors; | |
4bb91c68 | 1398 | so REL_ALLOC should not affect this. */ |
b588157e | 1399 | #if defined REGEX_MALLOC && defined emacs |
0b32bf0e | 1400 | # undef MATCH_MAY_ALLOCATE |
fa9a63c5 RM |
1401 | #endif |
1402 | ||
1403 | \f | |
1404 | /* Failure stack declarations and macros; both re_compile_fastmap and | |
1405 | re_match_2 use a failure stack. These have to be macros because of | |
1406 | REGEX_ALLOCATE_STACK. */ | |
5e69f11e | 1407 | |
fa9a63c5 | 1408 | |
320a2a73 | 1409 | /* Approximate number of failure points for which to initially allocate space |
fa9a63c5 RM |
1410 | when matching. If this number is exceeded, we allocate more |
1411 | space, so it is not a hard limit. */ | |
1412 | #ifndef INIT_FAILURE_ALLOC | |
0b32bf0e | 1413 | # define INIT_FAILURE_ALLOC 20 |
fa9a63c5 RM |
1414 | #endif |
1415 | ||
1416 | /* Roughly the maximum number of failure points on the stack. Would be | |
320a2a73 | 1417 | exactly that if always used TYPICAL_FAILURE_SIZE items each time we failed. |
fa9a63c5 | 1418 | This is a variable only so users of regex can assign to it; we never |
ada30c0e SM |
1419 | change it ourselves. We always multiply it by TYPICAL_FAILURE_SIZE |
1420 | before using it, so it should probably be a byte-count instead. */ | |
c0f9ea08 SM |
1421 | # if defined MATCH_MAY_ALLOCATE |
1422 | /* Note that 4400 was enough to cause a crash on Alpha OSF/1, | |
320a2a73 KH |
1423 | whose default stack limit is 2mb. In order for a larger |
1424 | value to work reliably, you have to try to make it accord | |
1425 | with the process stack limit. */ | |
c0f9ea08 SM |
1426 | size_t re_max_failures = 40000; |
1427 | # else | |
1428 | size_t re_max_failures = 4000; | |
1429 | # endif | |
fa9a63c5 RM |
1430 | |
1431 | union fail_stack_elt | |
1432 | { | |
01618498 | 1433 | re_char *pointer; |
c0f9ea08 SM |
1434 | /* This should be the biggest `int' that's no bigger than a pointer. */ |
1435 | long integer; | |
fa9a63c5 RM |
1436 | }; |
1437 | ||
1438 | typedef union fail_stack_elt fail_stack_elt_t; | |
1439 | ||
1440 | typedef struct | |
1441 | { | |
1442 | fail_stack_elt_t *stack; | |
c0f9ea08 SM |
1443 | size_t size; |
1444 | size_t avail; /* Offset of next open position. */ | |
1445 | size_t frame; /* Offset of the cur constructed frame. */ | |
fa9a63c5 RM |
1446 | } fail_stack_type; |
1447 | ||
505bde11 | 1448 | #define FAIL_STACK_EMPTY() (fail_stack.frame == 0) |
fa9a63c5 RM |
1449 | #define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size) |
1450 | ||
1451 | ||
1452 | /* Define macros to initialize and free the failure stack. | |
1453 | Do `return -2' if the alloc fails. */ | |
1454 | ||
1455 | #ifdef MATCH_MAY_ALLOCATE | |
0b32bf0e | 1456 | # define INIT_FAIL_STACK() \ |
fa9a63c5 RM |
1457 | do { \ |
1458 | fail_stack.stack = (fail_stack_elt_t *) \ | |
320a2a73 KH |
1459 | REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * TYPICAL_FAILURE_SIZE \ |
1460 | * sizeof (fail_stack_elt_t)); \ | |
fa9a63c5 RM |
1461 | \ |
1462 | if (fail_stack.stack == NULL) \ | |
1463 | return -2; \ | |
1464 | \ | |
1465 | fail_stack.size = INIT_FAILURE_ALLOC; \ | |
1466 | fail_stack.avail = 0; \ | |
505bde11 | 1467 | fail_stack.frame = 0; \ |
fa9a63c5 RM |
1468 | } while (0) |
1469 | ||
0b32bf0e | 1470 | # define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack) |
fa9a63c5 | 1471 | #else |
0b32bf0e | 1472 | # define INIT_FAIL_STACK() \ |
fa9a63c5 RM |
1473 | do { \ |
1474 | fail_stack.avail = 0; \ | |
505bde11 | 1475 | fail_stack.frame = 0; \ |
fa9a63c5 RM |
1476 | } while (0) |
1477 | ||
0b32bf0e | 1478 | # define RESET_FAIL_STACK() ((void)0) |
fa9a63c5 RM |
1479 | #endif |
1480 | ||
1481 | ||
320a2a73 KH |
1482 | /* Double the size of FAIL_STACK, up to a limit |
1483 | which allows approximately `re_max_failures' items. | |
fa9a63c5 RM |
1484 | |
1485 | Return 1 if succeeds, and 0 if either ran out of memory | |
5e69f11e RM |
1486 | allocating space for it or it was already too large. |
1487 | ||
4bb91c68 | 1488 | REGEX_REALLOCATE_STACK requires `destination' be declared. */ |
fa9a63c5 | 1489 | |
320a2a73 KH |
1490 | /* Factor to increase the failure stack size by |
1491 | when we increase it. | |
1492 | This used to be 2, but 2 was too wasteful | |
1493 | because the old discarded stacks added up to as much space | |
1494 | were as ultimate, maximum-size stack. */ | |
1495 | #define FAIL_STACK_GROWTH_FACTOR 4 | |
1496 | ||
1497 | #define GROW_FAIL_STACK(fail_stack) \ | |
eead07d6 KH |
1498 | (((fail_stack).size * sizeof (fail_stack_elt_t) \ |
1499 | >= re_max_failures * TYPICAL_FAILURE_SIZE) \ | |
fa9a63c5 | 1500 | ? 0 \ |
320a2a73 KH |
1501 | : ((fail_stack).stack \ |
1502 | = (fail_stack_elt_t *) \ | |
25fe55af RS |
1503 | REGEX_REALLOCATE_STACK ((fail_stack).stack, \ |
1504 | (fail_stack).size * sizeof (fail_stack_elt_t), \ | |
320a2a73 KH |
1505 | MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ |
1506 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1507 | * FAIL_STACK_GROWTH_FACTOR))), \ | |
fa9a63c5 RM |
1508 | \ |
1509 | (fail_stack).stack == NULL \ | |
1510 | ? 0 \ | |
6453db45 KH |
1511 | : ((fail_stack).size \ |
1512 | = (MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ | |
1513 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1514 | * FAIL_STACK_GROWTH_FACTOR)) \ | |
1515 | / sizeof (fail_stack_elt_t)), \ | |
25fe55af | 1516 | 1))) |
fa9a63c5 RM |
1517 | |
1518 | ||
fa9a63c5 RM |
1519 | /* Push a pointer value onto the failure stack. |
1520 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1521 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 | 1522 | #define PUSH_FAILURE_POINTER(item) \ |
01618498 | 1523 | fail_stack.stack[fail_stack.avail++].pointer = (item) |
fa9a63c5 RM |
1524 | |
1525 | /* This pushes an integer-valued item onto the failure stack. | |
1526 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1527 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 RM |
1528 | #define PUSH_FAILURE_INT(item) \ |
1529 | fail_stack.stack[fail_stack.avail++].integer = (item) | |
1530 | ||
1531 | /* Push a fail_stack_elt_t value onto the failure stack. | |
1532 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1533 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 RM |
1534 | #define PUSH_FAILURE_ELT(item) \ |
1535 | fail_stack.stack[fail_stack.avail++] = (item) | |
1536 | ||
1537 | /* These three POP... operations complement the three PUSH... operations. | |
1538 | All assume that `fail_stack' is nonempty. */ | |
1539 | #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer | |
1540 | #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer | |
1541 | #define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail] | |
1542 | ||
505bde11 SM |
1543 | /* Individual items aside from the registers. */ |
1544 | #define NUM_NONREG_ITEMS 3 | |
1545 | ||
1546 | /* Used to examine the stack (to detect infinite loops). */ | |
1547 | #define FAILURE_PAT(h) fail_stack.stack[(h) - 1].pointer | |
66f0296e | 1548 | #define FAILURE_STR(h) (fail_stack.stack[(h) - 2].pointer) |
505bde11 SM |
1549 | #define NEXT_FAILURE_HANDLE(h) fail_stack.stack[(h) - 3].integer |
1550 | #define TOP_FAILURE_HANDLE() fail_stack.frame | |
fa9a63c5 RM |
1551 | |
1552 | ||
505bde11 SM |
1553 | #define ENSURE_FAIL_STACK(space) \ |
1554 | while (REMAINING_AVAIL_SLOTS <= space) { \ | |
1555 | if (!GROW_FAIL_STACK (fail_stack)) \ | |
1556 | return -2; \ | |
1557 | DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", (fail_stack).size);\ | |
1558 | DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\ | |
1559 | } | |
1560 | ||
1561 | /* Push register NUM onto the stack. */ | |
1562 | #define PUSH_FAILURE_REG(num) \ | |
1563 | do { \ | |
1564 | char *destination; \ | |
1565 | ENSURE_FAIL_STACK(3); \ | |
1566 | DEBUG_PRINT4 (" Push reg %d (spanning %p -> %p)\n", \ | |
1567 | num, regstart[num], regend[num]); \ | |
1568 | PUSH_FAILURE_POINTER (regstart[num]); \ | |
1569 | PUSH_FAILURE_POINTER (regend[num]); \ | |
1570 | PUSH_FAILURE_INT (num); \ | |
1571 | } while (0) | |
1572 | ||
01618498 SM |
1573 | /* Change the counter's value to VAL, but make sure that it will |
1574 | be reset when backtracking. */ | |
1575 | #define PUSH_NUMBER(ptr,val) \ | |
dc1e502d SM |
1576 | do { \ |
1577 | char *destination; \ | |
1578 | int c; \ | |
1579 | ENSURE_FAIL_STACK(3); \ | |
1580 | EXTRACT_NUMBER (c, ptr); \ | |
01618498 | 1581 | DEBUG_PRINT4 (" Push number %p = %d -> %d\n", ptr, c, val); \ |
dc1e502d SM |
1582 | PUSH_FAILURE_INT (c); \ |
1583 | PUSH_FAILURE_POINTER (ptr); \ | |
1584 | PUSH_FAILURE_INT (-1); \ | |
01618498 | 1585 | STORE_NUMBER (ptr, val); \ |
dc1e502d SM |
1586 | } while (0) |
1587 | ||
505bde11 | 1588 | /* Pop a saved register off the stack. */ |
dc1e502d | 1589 | #define POP_FAILURE_REG_OR_COUNT() \ |
505bde11 SM |
1590 | do { \ |
1591 | int reg = POP_FAILURE_INT (); \ | |
dc1e502d SM |
1592 | if (reg == -1) \ |
1593 | { \ | |
1594 | /* It's a counter. */ \ | |
6dcf2d0e SM |
1595 | /* Here, we discard `const', making re_match non-reentrant. */ \ |
1596 | unsigned char *ptr = (unsigned char*) POP_FAILURE_POINTER (); \ | |
dc1e502d SM |
1597 | reg = POP_FAILURE_INT (); \ |
1598 | STORE_NUMBER (ptr, reg); \ | |
1599 | DEBUG_PRINT3 (" Pop counter %p = %d\n", ptr, reg); \ | |
1600 | } \ | |
1601 | else \ | |
1602 | { \ | |
1603 | regend[reg] = POP_FAILURE_POINTER (); \ | |
1604 | regstart[reg] = POP_FAILURE_POINTER (); \ | |
1605 | DEBUG_PRINT4 (" Pop reg %d (spanning %p -> %p)\n", \ | |
1606 | reg, regstart[reg], regend[reg]); \ | |
1607 | } \ | |
505bde11 SM |
1608 | } while (0) |
1609 | ||
1610 | /* Check that we are not stuck in an infinite loop. */ | |
1611 | #define CHECK_INFINITE_LOOP(pat_cur, string_place) \ | |
1612 | do { \ | |
f6df485f | 1613 | int failure = TOP_FAILURE_HANDLE (); \ |
505bde11 | 1614 | /* Check for infinite matching loops */ \ |
f6df485f RS |
1615 | while (failure > 0 \ |
1616 | && (FAILURE_STR (failure) == string_place \ | |
1617 | || FAILURE_STR (failure) == NULL)) \ | |
505bde11 SM |
1618 | { \ |
1619 | assert (FAILURE_PAT (failure) >= bufp->buffer \ | |
66f0296e | 1620 | && FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \ |
505bde11 | 1621 | if (FAILURE_PAT (failure) == pat_cur) \ |
f6df485f | 1622 | { \ |
6df42991 SM |
1623 | cycle = 1; \ |
1624 | break; \ | |
f6df485f | 1625 | } \ |
66f0296e | 1626 | DEBUG_PRINT2 (" Other pattern: %p\n", FAILURE_PAT (failure)); \ |
505bde11 SM |
1627 | failure = NEXT_FAILURE_HANDLE(failure); \ |
1628 | } \ | |
1629 | DEBUG_PRINT2 (" Other string: %p\n", FAILURE_STR (failure)); \ | |
1630 | } while (0) | |
6df42991 | 1631 | |
fa9a63c5 | 1632 | /* Push the information about the state we will need |
5e69f11e RM |
1633 | if we ever fail back to it. |
1634 | ||
505bde11 | 1635 | Requires variables fail_stack, regstart, regend and |
320a2a73 | 1636 | num_regs be declared. GROW_FAIL_STACK requires `destination' be |
fa9a63c5 | 1637 | declared. |
5e69f11e | 1638 | |
fa9a63c5 RM |
1639 | Does `return FAILURE_CODE' if runs out of memory. */ |
1640 | ||
505bde11 SM |
1641 | #define PUSH_FAILURE_POINT(pattern, string_place) \ |
1642 | do { \ | |
1643 | char *destination; \ | |
1644 | /* Must be int, so when we don't save any registers, the arithmetic \ | |
1645 | of 0 + -1 isn't done as unsigned. */ \ | |
1646 | \ | |
505bde11 | 1647 | DEBUG_STATEMENT (nfailure_points_pushed++); \ |
4bb91c68 | 1648 | DEBUG_PRINT1 ("\nPUSH_FAILURE_POINT:\n"); \ |
505bde11 SM |
1649 | DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail); \ |
1650 | DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\ | |
1651 | \ | |
1652 | ENSURE_FAIL_STACK (NUM_NONREG_ITEMS); \ | |
1653 | \ | |
1654 | DEBUG_PRINT1 ("\n"); \ | |
1655 | \ | |
1656 | DEBUG_PRINT2 (" Push frame index: %d\n", fail_stack.frame); \ | |
1657 | PUSH_FAILURE_INT (fail_stack.frame); \ | |
1658 | \ | |
1659 | DEBUG_PRINT2 (" Push string %p: `", string_place); \ | |
1660 | DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, size2);\ | |
1661 | DEBUG_PRINT1 ("'\n"); \ | |
1662 | PUSH_FAILURE_POINTER (string_place); \ | |
1663 | \ | |
1664 | DEBUG_PRINT2 (" Push pattern %p: ", pattern); \ | |
1665 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern, pend); \ | |
1666 | PUSH_FAILURE_POINTER (pattern); \ | |
1667 | \ | |
1668 | /* Close the frame by moving the frame pointer past it. */ \ | |
1669 | fail_stack.frame = fail_stack.avail; \ | |
1670 | } while (0) | |
fa9a63c5 | 1671 | |
320a2a73 KH |
1672 | /* Estimate the size of data pushed by a typical failure stack entry. |
1673 | An estimate is all we need, because all we use this for | |
1674 | is to choose a limit for how big to make the failure stack. */ | |
ada30c0e | 1675 | /* BEWARE, the value `20' is hard-coded in emacs.c:main(). */ |
320a2a73 | 1676 | #define TYPICAL_FAILURE_SIZE 20 |
fa9a63c5 | 1677 | |
fa9a63c5 RM |
1678 | /* How many items can still be added to the stack without overflowing it. */ |
1679 | #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) | |
1680 | ||
1681 | ||
1682 | /* Pops what PUSH_FAIL_STACK pushes. | |
1683 | ||
1684 | We restore into the parameters, all of which should be lvalues: | |
1685 | STR -- the saved data position. | |
1686 | PAT -- the saved pattern position. | |
fa9a63c5 | 1687 | REGSTART, REGEND -- arrays of string positions. |
5e69f11e | 1688 | |
fa9a63c5 | 1689 | Also assumes the variables `fail_stack' and (if debugging), `bufp', |
7814e705 | 1690 | `pend', `string1', `size1', `string2', and `size2'. */ |
fa9a63c5 | 1691 | |
505bde11 SM |
1692 | #define POP_FAILURE_POINT(str, pat) \ |
1693 | do { \ | |
fa9a63c5 RM |
1694 | assert (!FAIL_STACK_EMPTY ()); \ |
1695 | \ | |
1696 | /* Remove failure points and point to how many regs pushed. */ \ | |
1697 | DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \ | |
1698 | DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \ | |
25fe55af | 1699 | DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \ |
fa9a63c5 | 1700 | \ |
505bde11 SM |
1701 | /* Pop the saved registers. */ \ |
1702 | while (fail_stack.frame < fail_stack.avail) \ | |
dc1e502d | 1703 | POP_FAILURE_REG_OR_COUNT (); \ |
fa9a63c5 | 1704 | \ |
01618498 | 1705 | pat = POP_FAILURE_POINTER (); \ |
505bde11 SM |
1706 | DEBUG_PRINT2 (" Popping pattern %p: ", pat); \ |
1707 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ | |
fa9a63c5 RM |
1708 | \ |
1709 | /* If the saved string location is NULL, it came from an \ | |
1710 | on_failure_keep_string_jump opcode, and we want to throw away the \ | |
1711 | saved NULL, thus retaining our current position in the string. */ \ | |
01618498 | 1712 | str = POP_FAILURE_POINTER (); \ |
505bde11 | 1713 | DEBUG_PRINT2 (" Popping string %p: `", str); \ |
fa9a63c5 RM |
1714 | DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ |
1715 | DEBUG_PRINT1 ("'\n"); \ | |
1716 | \ | |
505bde11 SM |
1717 | fail_stack.frame = POP_FAILURE_INT (); \ |
1718 | DEBUG_PRINT2 (" Popping frame index: %d\n", fail_stack.frame); \ | |
fa9a63c5 | 1719 | \ |
505bde11 SM |
1720 | assert (fail_stack.avail >= 0); \ |
1721 | assert (fail_stack.frame <= fail_stack.avail); \ | |
fa9a63c5 | 1722 | \ |
fa9a63c5 | 1723 | DEBUG_STATEMENT (nfailure_points_popped++); \ |
505bde11 | 1724 | } while (0) /* POP_FAILURE_POINT */ |
fa9a63c5 RM |
1725 | |
1726 | ||
1727 | \f | |
fa9a63c5 | 1728 | /* Registers are set to a sentinel when they haven't yet matched. */ |
4bb91c68 | 1729 | #define REG_UNSET(e) ((e) == NULL) |
fa9a63c5 RM |
1730 | \f |
1731 | /* Subroutine declarations and macros for regex_compile. */ | |
1732 | ||
4bb91c68 SM |
1733 | static reg_errcode_t regex_compile _RE_ARGS ((re_char *pattern, size_t size, |
1734 | reg_syntax_t syntax, | |
1735 | struct re_pattern_buffer *bufp)); | |
1736 | static void store_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc, int arg)); | |
1737 | static void store_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc, | |
1738 | int arg1, int arg2)); | |
1739 | static void insert_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc, | |
1740 | int arg, unsigned char *end)); | |
1741 | static void insert_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc, | |
1742 | int arg1, int arg2, unsigned char *end)); | |
01618498 SM |
1743 | static boolean at_begline_loc_p _RE_ARGS ((re_char *pattern, |
1744 | re_char *p, | |
4bb91c68 | 1745 | reg_syntax_t syntax)); |
01618498 SM |
1746 | static boolean at_endline_loc_p _RE_ARGS ((re_char *p, |
1747 | re_char *pend, | |
4bb91c68 | 1748 | reg_syntax_t syntax)); |
01618498 SM |
1749 | static re_char *skip_one_char _RE_ARGS ((re_char *p)); |
1750 | static int analyse_first _RE_ARGS ((re_char *p, re_char *pend, | |
4bb91c68 | 1751 | char *fastmap, const int multibyte)); |
fa9a63c5 | 1752 | |
fa9a63c5 | 1753 | /* Fetch the next character in the uncompiled pattern, with no |
4bb91c68 | 1754 | translation. */ |
36595814 | 1755 | #define PATFETCH(c) \ |
2d1675e4 SM |
1756 | do { \ |
1757 | int len; \ | |
1758 | if (p == pend) return REG_EEND; \ | |
62a6e103 | 1759 | c = RE_STRING_CHAR_AND_LENGTH (p, len, multibyte); \ |
2d1675e4 | 1760 | p += len; \ |
fa9a63c5 RM |
1761 | } while (0) |
1762 | ||
fa9a63c5 RM |
1763 | |
1764 | /* If `translate' is non-null, return translate[D], else just D. We | |
1765 | cast the subscript to translate because some data is declared as | |
1766 | `char *', to avoid warnings when a string constant is passed. But | |
1767 | when we use a character as a subscript we must make it unsigned. */ | |
6676cb1c | 1768 | #ifndef TRANSLATE |
0b32bf0e | 1769 | # define TRANSLATE(d) \ |
66f0296e | 1770 | (RE_TRANSLATE_P (translate) ? RE_TRANSLATE (translate, (d)) : (d)) |
6676cb1c | 1771 | #endif |
fa9a63c5 RM |
1772 | |
1773 | ||
1774 | /* Macros for outputting the compiled pattern into `buffer'. */ | |
1775 | ||
1776 | /* If the buffer isn't allocated when it comes in, use this. */ | |
1777 | #define INIT_BUF_SIZE 32 | |
1778 | ||
4bb91c68 | 1779 | /* Make sure we have at least N more bytes of space in buffer. */ |
fa9a63c5 | 1780 | #define GET_BUFFER_SPACE(n) \ |
01618498 | 1781 | while ((size_t) (b - bufp->buffer + (n)) > bufp->allocated) \ |
fa9a63c5 RM |
1782 | EXTEND_BUFFER () |
1783 | ||
1784 | /* Make sure we have one more byte of buffer space and then add C to it. */ | |
1785 | #define BUF_PUSH(c) \ | |
1786 | do { \ | |
1787 | GET_BUFFER_SPACE (1); \ | |
1788 | *b++ = (unsigned char) (c); \ | |
1789 | } while (0) | |
1790 | ||
1791 | ||
1792 | /* Ensure we have two more bytes of buffer space and then append C1 and C2. */ | |
1793 | #define BUF_PUSH_2(c1, c2) \ | |
1794 | do { \ | |
1795 | GET_BUFFER_SPACE (2); \ | |
1796 | *b++ = (unsigned char) (c1); \ | |
1797 | *b++ = (unsigned char) (c2); \ | |
1798 | } while (0) | |
1799 | ||
1800 | ||
4bb91c68 | 1801 | /* As with BUF_PUSH_2, except for three bytes. */ |
fa9a63c5 RM |
1802 | #define BUF_PUSH_3(c1, c2, c3) \ |
1803 | do { \ | |
1804 | GET_BUFFER_SPACE (3); \ | |
1805 | *b++ = (unsigned char) (c1); \ | |
1806 | *b++ = (unsigned char) (c2); \ | |
1807 | *b++ = (unsigned char) (c3); \ | |
1808 | } while (0) | |
1809 | ||
1810 | ||
1811 | /* Store a jump with opcode OP at LOC to location TO. We store a | |
4bb91c68 | 1812 | relative address offset by the three bytes the jump itself occupies. */ |
fa9a63c5 RM |
1813 | #define STORE_JUMP(op, loc, to) \ |
1814 | store_op1 (op, loc, (to) - (loc) - 3) | |
1815 | ||
1816 | /* Likewise, for a two-argument jump. */ | |
1817 | #define STORE_JUMP2(op, loc, to, arg) \ | |
1818 | store_op2 (op, loc, (to) - (loc) - 3, arg) | |
1819 | ||
4bb91c68 | 1820 | /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ |
fa9a63c5 RM |
1821 | #define INSERT_JUMP(op, loc, to) \ |
1822 | insert_op1 (op, loc, (to) - (loc) - 3, b) | |
1823 | ||
1824 | /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ | |
1825 | #define INSERT_JUMP2(op, loc, to, arg) \ | |
1826 | insert_op2 (op, loc, (to) - (loc) - 3, arg, b) | |
1827 | ||
1828 | ||
1829 | /* This is not an arbitrary limit: the arguments which represent offsets | |
839966f3 | 1830 | into the pattern are two bytes long. So if 2^15 bytes turns out to |
fa9a63c5 | 1831 | be too small, many things would have to change. */ |
839966f3 KH |
1832 | # define MAX_BUF_SIZE (1L << 15) |
1833 | ||
1834 | #if 0 /* This is when we thought it could be 2^16 bytes. */ | |
4bb91c68 SM |
1835 | /* Any other compiler which, like MSC, has allocation limit below 2^16 |
1836 | bytes will have to use approach similar to what was done below for | |
1837 | MSC and drop MAX_BUF_SIZE a bit. Otherwise you may end up | |
1838 | reallocating to 0 bytes. Such thing is not going to work too well. | |
1839 | You have been warned!! */ | |
1840 | #if defined _MSC_VER && !defined WIN32 | |
1841 | /* Microsoft C 16-bit versions limit malloc to approx 65512 bytes. */ | |
1842 | # define MAX_BUF_SIZE 65500L | |
1843 | #else | |
1844 | # define MAX_BUF_SIZE (1L << 16) | |
1845 | #endif | |
839966f3 | 1846 | #endif /* 0 */ |
fa9a63c5 RM |
1847 | |
1848 | /* Extend the buffer by twice its current size via realloc and | |
1849 | reset the pointers that pointed into the old block to point to the | |
1850 | correct places in the new one. If extending the buffer results in it | |
4bb91c68 SM |
1851 | being larger than MAX_BUF_SIZE, then flag memory exhausted. */ |
1852 | #if __BOUNDED_POINTERS__ | |
1853 | # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) | |
381880b0 CY |
1854 | # define MOVE_BUFFER_POINTER(P) \ |
1855 | (__ptrlow (P) = new_buffer + (__ptrlow (P) - old_buffer), \ | |
1856 | SET_HIGH_BOUND (P), \ | |
1857 | __ptrvalue (P) = new_buffer + (__ptrvalue (P) - old_buffer)) | |
4bb91c68 SM |
1858 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
1859 | else \ | |
1860 | { \ | |
1861 | SET_HIGH_BOUND (b); \ | |
1862 | SET_HIGH_BOUND (begalt); \ | |
1863 | if (fixup_alt_jump) \ | |
1864 | SET_HIGH_BOUND (fixup_alt_jump); \ | |
1865 | if (laststart) \ | |
1866 | SET_HIGH_BOUND (laststart); \ | |
1867 | if (pending_exact) \ | |
1868 | SET_HIGH_BOUND (pending_exact); \ | |
1869 | } | |
1870 | #else | |
381880b0 | 1871 | # define MOVE_BUFFER_POINTER(P) ((P) = new_buffer + ((P) - old_buffer)) |
4bb91c68 SM |
1872 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND |
1873 | #endif | |
fa9a63c5 | 1874 | #define EXTEND_BUFFER() \ |
25fe55af | 1875 | do { \ |
381880b0 | 1876 | unsigned char *old_buffer = bufp->buffer; \ |
25fe55af | 1877 | if (bufp->allocated == MAX_BUF_SIZE) \ |
fa9a63c5 RM |
1878 | return REG_ESIZE; \ |
1879 | bufp->allocated <<= 1; \ | |
1880 | if (bufp->allocated > MAX_BUF_SIZE) \ | |
25fe55af | 1881 | bufp->allocated = MAX_BUF_SIZE; \ |
01618498 | 1882 | RETALLOC (bufp->buffer, bufp->allocated, unsigned char); \ |
fa9a63c5 RM |
1883 | if (bufp->buffer == NULL) \ |
1884 | return REG_ESPACE; \ | |
1885 | /* If the buffer moved, move all the pointers into it. */ \ | |
1886 | if (old_buffer != bufp->buffer) \ | |
1887 | { \ | |
381880b0 | 1888 | unsigned char *new_buffer = bufp->buffer; \ |
4bb91c68 SM |
1889 | MOVE_BUFFER_POINTER (b); \ |
1890 | MOVE_BUFFER_POINTER (begalt); \ | |
25fe55af | 1891 | if (fixup_alt_jump) \ |
4bb91c68 | 1892 | MOVE_BUFFER_POINTER (fixup_alt_jump); \ |
25fe55af | 1893 | if (laststart) \ |
4bb91c68 | 1894 | MOVE_BUFFER_POINTER (laststart); \ |
25fe55af | 1895 | if (pending_exact) \ |
4bb91c68 | 1896 | MOVE_BUFFER_POINTER (pending_exact); \ |
fa9a63c5 | 1897 | } \ |
4bb91c68 | 1898 | ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
fa9a63c5 RM |
1899 | } while (0) |
1900 | ||
1901 | ||
1902 | /* Since we have one byte reserved for the register number argument to | |
1903 | {start,stop}_memory, the maximum number of groups we can report | |
1904 | things about is what fits in that byte. */ | |
1905 | #define MAX_REGNUM 255 | |
1906 | ||
1907 | /* But patterns can have more than `MAX_REGNUM' registers. We just | |
1908 | ignore the excess. */ | |
098d42af | 1909 | typedef int regnum_t; |
fa9a63c5 RM |
1910 | |
1911 | ||
1912 | /* Macros for the compile stack. */ | |
1913 | ||
1914 | /* Since offsets can go either forwards or backwards, this type needs to | |
4bb91c68 SM |
1915 | be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ |
1916 | /* int may be not enough when sizeof(int) == 2. */ | |
1917 | typedef long pattern_offset_t; | |
fa9a63c5 RM |
1918 | |
1919 | typedef struct | |
1920 | { | |
1921 | pattern_offset_t begalt_offset; | |
1922 | pattern_offset_t fixup_alt_jump; | |
5e69f11e | 1923 | pattern_offset_t laststart_offset; |
fa9a63c5 RM |
1924 | regnum_t regnum; |
1925 | } compile_stack_elt_t; | |
1926 | ||
1927 | ||
1928 | typedef struct | |
1929 | { | |
1930 | compile_stack_elt_t *stack; | |
1931 | unsigned size; | |
1932 | unsigned avail; /* Offset of next open position. */ | |
1933 | } compile_stack_type; | |
1934 | ||
1935 | ||
1936 | #define INIT_COMPILE_STACK_SIZE 32 | |
1937 | ||
1938 | #define COMPILE_STACK_EMPTY (compile_stack.avail == 0) | |
1939 | #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) | |
1940 | ||
4bb91c68 | 1941 | /* The next available element. */ |
fa9a63c5 RM |
1942 | #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) |
1943 | ||
1cee1e27 SM |
1944 | /* Explicit quit checking is only used on NTemacs and whenever we |
1945 | use polling to process input events. */ | |
1946 | #if defined emacs && (defined WINDOWSNT || defined SYNC_INPUT) && defined QUIT | |
77d11aec RS |
1947 | extern int immediate_quit; |
1948 | # define IMMEDIATE_QUIT_CHECK \ | |
1949 | do { \ | |
1950 | if (immediate_quit) QUIT; \ | |
1951 | } while (0) | |
1952 | #else | |
1953 | # define IMMEDIATE_QUIT_CHECK ((void)0) | |
1954 | #endif | |
1955 | \f | |
b18215fc RS |
1956 | /* Structure to manage work area for range table. */ |
1957 | struct range_table_work_area | |
1958 | { | |
1959 | int *table; /* actual work area. */ | |
1960 | int allocated; /* allocated size for work area in bytes. */ | |
7814e705 | 1961 | int used; /* actually used size in words. */ |
96cc36cc | 1962 | int bits; /* flag to record character classes */ |
b18215fc RS |
1963 | }; |
1964 | ||
77d11aec RS |
1965 | /* Make sure that WORK_AREA can hold more N multibyte characters. |
1966 | This is used only in set_image_of_range and set_image_of_range_1. | |
1967 | It expects WORK_AREA to be a pointer. | |
1968 | If it can't get the space, it returns from the surrounding function. */ | |
1969 | ||
1970 | #define EXTEND_RANGE_TABLE(work_area, n) \ | |
1971 | do { \ | |
8f924df7 | 1972 | if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \ |
77d11aec | 1973 | { \ |
8f924df7 KH |
1974 | extend_range_table_work_area (&work_area); \ |
1975 | if ((work_area).table == 0) \ | |
77d11aec RS |
1976 | return (REG_ESPACE); \ |
1977 | } \ | |
b18215fc RS |
1978 | } while (0) |
1979 | ||
96cc36cc RS |
1980 | #define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \ |
1981 | (work_area).bits |= (bit) | |
1982 | ||
14473664 SM |
1983 | /* Bits used to implement the multibyte-part of the various character classes |
1984 | such as [:alnum:] in a charset's range table. */ | |
1985 | #define BIT_WORD 0x1 | |
1986 | #define BIT_LOWER 0x2 | |
1987 | #define BIT_PUNCT 0x4 | |
1988 | #define BIT_SPACE 0x8 | |
1989 | #define BIT_UPPER 0x10 | |
1990 | #define BIT_MULTIBYTE 0x20 | |
96cc36cc | 1991 | |
b18215fc RS |
1992 | /* Set a range (RANGE_START, RANGE_END) to WORK_AREA. */ |
1993 | #define SET_RANGE_TABLE_WORK_AREA(work_area, range_start, range_end) \ | |
77d11aec | 1994 | do { \ |
8f924df7 | 1995 | EXTEND_RANGE_TABLE ((work_area), 2); \ |
b18215fc RS |
1996 | (work_area).table[(work_area).used++] = (range_start); \ |
1997 | (work_area).table[(work_area).used++] = (range_end); \ | |
1998 | } while (0) | |
1999 | ||
7814e705 | 2000 | /* Free allocated memory for WORK_AREA. */ |
b18215fc RS |
2001 | #define FREE_RANGE_TABLE_WORK_AREA(work_area) \ |
2002 | do { \ | |
2003 | if ((work_area).table) \ | |
2004 | free ((work_area).table); \ | |
2005 | } while (0) | |
2006 | ||
96cc36cc | 2007 | #define CLEAR_RANGE_TABLE_WORK_USED(work_area) ((work_area).used = 0, (work_area).bits = 0) |
b18215fc | 2008 | #define RANGE_TABLE_WORK_USED(work_area) ((work_area).used) |
96cc36cc | 2009 | #define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits) |
b18215fc | 2010 | #define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i]) |
77d11aec | 2011 | \f |
b18215fc | 2012 | |
fa9a63c5 | 2013 | /* Set the bit for character C in a list. */ |
01618498 | 2014 | #define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH)) |
fa9a63c5 RM |
2015 | |
2016 | ||
bf216479 KH |
2017 | #ifdef emacs |
2018 | ||
cf9c99bc KH |
2019 | /* Store characters in the range FROM to TO in the bitmap at B (for |
2020 | ASCII and unibyte characters) and WORK_AREA (for multibyte | |
2021 | characters) while translating them and paying attention to the | |
2022 | continuity of translated characters. | |
8f924df7 | 2023 | |
cf9c99bc KH |
2024 | Implementation note: It is better to implement these fairly big |
2025 | macros by a function, but it's not that easy because macros called | |
8f924df7 | 2026 | in this macro assume various local variables already declared. */ |
bf216479 | 2027 | |
cf9c99bc KH |
2028 | /* Both FROM and TO are ASCII characters. */ |
2029 | ||
2030 | #define SETUP_ASCII_RANGE(work_area, FROM, TO) \ | |
2031 | do { \ | |
2032 | int C0, C1; \ | |
2033 | \ | |
2034 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
2035 | { \ | |
2036 | C1 = TRANSLATE (C0); \ | |
2037 | if (! ASCII_CHAR_P (C1)) \ | |
2038 | { \ | |
2039 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
2040 | if ((C1 = RE_CHAR_TO_UNIBYTE (C1)) < 0) \ | |
2041 | C1 = C0; \ | |
2042 | } \ | |
2043 | SET_LIST_BIT (C1); \ | |
2044 | } \ | |
2045 | } while (0) | |
2046 | ||
2047 | ||
2048 | /* Both FROM and TO are unibyte characters (0x80..0xFF). */ | |
2049 | ||
2050 | #define SETUP_UNIBYTE_RANGE(work_area, FROM, TO) \ | |
2051 | do { \ | |
2052 | int C0, C1, C2, I; \ | |
2053 | int USED = RANGE_TABLE_WORK_USED (work_area); \ | |
2054 | \ | |
2055 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
2056 | { \ | |
2057 | C1 = RE_CHAR_TO_MULTIBYTE (C0); \ | |
2058 | if (CHAR_BYTE8_P (C1)) \ | |
2059 | SET_LIST_BIT (C0); \ | |
2060 | else \ | |
2061 | { \ | |
2062 | C2 = TRANSLATE (C1); \ | |
2063 | if (C2 == C1 \ | |
2064 | || (C1 = RE_CHAR_TO_UNIBYTE (C2)) < 0) \ | |
2065 | C1 = C0; \ | |
2066 | SET_LIST_BIT (C1); \ | |
2067 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
2068 | { \ | |
2069 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
2070 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
2071 | \ | |
2072 | if (C2 >= from - 1 && C2 <= to + 1) \ | |
2073 | { \ | |
2074 | if (C2 == from - 1) \ | |
2075 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
2076 | else if (C2 == to + 1) \ | |
2077 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
2078 | break; \ | |
2079 | } \ | |
2080 | } \ | |
2081 | if (I < USED) \ | |
2082 | SET_RANGE_TABLE_WORK_AREA ((work_area), C2, C2); \ | |
2083 | } \ | |
2084 | } \ | |
2085 | } while (0) | |
2086 | ||
2087 | ||
78edd3b7 | 2088 | /* Both FROM and TO are multibyte characters. */ |
cf9c99bc KH |
2089 | |
2090 | #define SETUP_MULTIBYTE_RANGE(work_area, FROM, TO) \ | |
2091 | do { \ | |
2092 | int C0, C1, C2, I, USED = RANGE_TABLE_WORK_USED (work_area); \ | |
2093 | \ | |
2094 | SET_RANGE_TABLE_WORK_AREA ((work_area), (FROM), (TO)); \ | |
2095 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
2096 | { \ | |
2097 | C1 = TRANSLATE (C0); \ | |
2098 | if ((C2 = RE_CHAR_TO_UNIBYTE (C1)) >= 0 \ | |
2099 | || (C1 != C0 && (C2 = RE_CHAR_TO_UNIBYTE (C0)) >= 0)) \ | |
2100 | SET_LIST_BIT (C2); \ | |
2101 | if (C1 >= (FROM) && C1 <= (TO)) \ | |
2102 | continue; \ | |
2103 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
2104 | { \ | |
2105 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
2106 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
2107 | \ | |
2108 | if (C1 >= from - 1 && C1 <= to + 1) \ | |
2109 | { \ | |
2110 | if (C1 == from - 1) \ | |
2111 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
2112 | else if (C1 == to + 1) \ | |
2113 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
2114 | break; \ | |
2115 | } \ | |
2116 | } \ | |
2117 | if (I < USED) \ | |
2118 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
2119 | } \ | |
bf216479 KH |
2120 | } while (0) |
2121 | ||
2122 | #endif /* emacs */ | |
2123 | ||
fa9a63c5 | 2124 | /* Get the next unsigned number in the uncompiled pattern. */ |
25fe55af | 2125 | #define GET_UNSIGNED_NUMBER(num) \ |
c72b0edd SM |
2126 | do { \ |
2127 | if (p == pend) \ | |
2128 | FREE_STACK_RETURN (REG_EBRACE); \ | |
2129 | else \ | |
2130 | { \ | |
2131 | PATFETCH (c); \ | |
2132 | while ('0' <= c && c <= '9') \ | |
2133 | { \ | |
2134 | int prev; \ | |
2135 | if (num < 0) \ | |
2136 | num = 0; \ | |
2137 | prev = num; \ | |
2138 | num = num * 10 + c - '0'; \ | |
2139 | if (num / 10 != prev) \ | |
2140 | FREE_STACK_RETURN (REG_BADBR); \ | |
2141 | if (p == pend) \ | |
2142 | FREE_STACK_RETURN (REG_EBRACE); \ | |
2143 | PATFETCH (c); \ | |
2144 | } \ | |
2145 | } \ | |
2146 | } while (0) | |
77d11aec | 2147 | \f |
1fdab503 | 2148 | #if ! WIDE_CHAR_SUPPORT |
01618498 | 2149 | |
14473664 | 2150 | /* Map a string to the char class it names (if any). */ |
1fdab503 | 2151 | re_wctype_t |
ada30c0e SM |
2152 | re_wctype (str) |
2153 | re_char *str; | |
14473664 | 2154 | { |
ada30c0e | 2155 | const char *string = str; |
14473664 SM |
2156 | if (STREQ (string, "alnum")) return RECC_ALNUM; |
2157 | else if (STREQ (string, "alpha")) return RECC_ALPHA; | |
2158 | else if (STREQ (string, "word")) return RECC_WORD; | |
2159 | else if (STREQ (string, "ascii")) return RECC_ASCII; | |
2160 | else if (STREQ (string, "nonascii")) return RECC_NONASCII; | |
2161 | else if (STREQ (string, "graph")) return RECC_GRAPH; | |
2162 | else if (STREQ (string, "lower")) return RECC_LOWER; | |
2163 | else if (STREQ (string, "print")) return RECC_PRINT; | |
2164 | else if (STREQ (string, "punct")) return RECC_PUNCT; | |
2165 | else if (STREQ (string, "space")) return RECC_SPACE; | |
2166 | else if (STREQ (string, "upper")) return RECC_UPPER; | |
2167 | else if (STREQ (string, "unibyte")) return RECC_UNIBYTE; | |
2168 | else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE; | |
2169 | else if (STREQ (string, "digit")) return RECC_DIGIT; | |
2170 | else if (STREQ (string, "xdigit")) return RECC_XDIGIT; | |
2171 | else if (STREQ (string, "cntrl")) return RECC_CNTRL; | |
2172 | else if (STREQ (string, "blank")) return RECC_BLANK; | |
2173 | else return 0; | |
2174 | } | |
2175 | ||
e0f24100 | 2176 | /* True if CH is in the char class CC. */ |
1fdab503 | 2177 | boolean |
14473664 SM |
2178 | re_iswctype (ch, cc) |
2179 | int ch; | |
2180 | re_wctype_t cc; | |
2181 | { | |
2182 | switch (cc) | |
2183 | { | |
0cdd06f8 SM |
2184 | case RECC_ALNUM: return ISALNUM (ch); |
2185 | case RECC_ALPHA: return ISALPHA (ch); | |
2186 | case RECC_BLANK: return ISBLANK (ch); | |
2187 | case RECC_CNTRL: return ISCNTRL (ch); | |
2188 | case RECC_DIGIT: return ISDIGIT (ch); | |
2189 | case RECC_GRAPH: return ISGRAPH (ch); | |
2190 | case RECC_LOWER: return ISLOWER (ch); | |
2191 | case RECC_PRINT: return ISPRINT (ch); | |
2192 | case RECC_PUNCT: return ISPUNCT (ch); | |
2193 | case RECC_SPACE: return ISSPACE (ch); | |
2194 | case RECC_UPPER: return ISUPPER (ch); | |
2195 | case RECC_XDIGIT: return ISXDIGIT (ch); | |
2196 | case RECC_ASCII: return IS_REAL_ASCII (ch); | |
2197 | case RECC_NONASCII: return !IS_REAL_ASCII (ch); | |
2198 | case RECC_UNIBYTE: return ISUNIBYTE (ch); | |
2199 | case RECC_MULTIBYTE: return !ISUNIBYTE (ch); | |
2200 | case RECC_WORD: return ISWORD (ch); | |
2201 | case RECC_ERROR: return false; | |
2202 | default: | |
2203 | abort(); | |
14473664 SM |
2204 | } |
2205 | } | |
fa9a63c5 | 2206 | |
14473664 SM |
2207 | /* Return a bit-pattern to use in the range-table bits to match multibyte |
2208 | chars of class CC. */ | |
2209 | static int | |
2210 | re_wctype_to_bit (cc) | |
2211 | re_wctype_t cc; | |
2212 | { | |
2213 | switch (cc) | |
2214 | { | |
2215 | case RECC_NONASCII: case RECC_PRINT: case RECC_GRAPH: | |
0cdd06f8 SM |
2216 | case RECC_MULTIBYTE: return BIT_MULTIBYTE; |
2217 | case RECC_ALPHA: case RECC_ALNUM: case RECC_WORD: return BIT_WORD; | |
2218 | case RECC_LOWER: return BIT_LOWER; | |
2219 | case RECC_UPPER: return BIT_UPPER; | |
2220 | case RECC_PUNCT: return BIT_PUNCT; | |
2221 | case RECC_SPACE: return BIT_SPACE; | |
14473664 | 2222 | case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL: |
0cdd06f8 SM |
2223 | case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0; |
2224 | default: | |
2225 | abort(); | |
14473664 SM |
2226 | } |
2227 | } | |
2228 | #endif | |
77d11aec RS |
2229 | \f |
2230 | /* Filling in the work area of a range. */ | |
2231 | ||
2232 | /* Actually extend the space in WORK_AREA. */ | |
2233 | ||
2234 | static void | |
2235 | extend_range_table_work_area (work_area) | |
2236 | struct range_table_work_area *work_area; | |
177c0ea7 | 2237 | { |
77d11aec RS |
2238 | work_area->allocated += 16 * sizeof (int); |
2239 | if (work_area->table) | |
2240 | work_area->table | |
2241 | = (int *) realloc (work_area->table, work_area->allocated); | |
2242 | else | |
2243 | work_area->table | |
2244 | = (int *) malloc (work_area->allocated); | |
2245 | } | |
2246 | ||
8f924df7 | 2247 | #if 0 |
77d11aec RS |
2248 | #ifdef emacs |
2249 | ||
2250 | /* Carefully find the ranges of codes that are equivalent | |
2251 | under case conversion to the range start..end when passed through | |
2252 | TRANSLATE. Handle the case where non-letters can come in between | |
2253 | two upper-case letters (which happens in Latin-1). | |
2254 | Also handle the case of groups of more than 2 case-equivalent chars. | |
2255 | ||
2256 | The basic method is to look at consecutive characters and see | |
2257 | if they can form a run that can be handled as one. | |
2258 | ||
2259 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2260 | ||
2261 | static int | |
2262 | set_image_of_range_1 (work_area, start, end, translate) | |
2263 | RE_TRANSLATE_TYPE translate; | |
2264 | struct range_table_work_area *work_area; | |
2265 | re_wchar_t start, end; | |
2266 | { | |
2267 | /* `one_case' indicates a character, or a run of characters, | |
2268 | each of which is an isolate (no case-equivalents). | |
2269 | This includes all ASCII non-letters. | |
2270 | ||
2271 | `two_case' indicates a character, or a run of characters, | |
2272 | each of which has two case-equivalent forms. | |
2273 | This includes all ASCII letters. | |
2274 | ||
2275 | `strange' indicates a character that has more than one | |
2276 | case-equivalent. */ | |
177c0ea7 | 2277 | |
77d11aec RS |
2278 | enum case_type {one_case, two_case, strange}; |
2279 | ||
2280 | /* Describe the run that is in progress, | |
2281 | which the next character can try to extend. | |
2282 | If run_type is strange, that means there really is no run. | |
2283 | If run_type is one_case, then run_start...run_end is the run. | |
2284 | If run_type is two_case, then the run is run_start...run_end, | |
2285 | and the case-equivalents end at run_eqv_end. */ | |
2286 | ||
2287 | enum case_type run_type = strange; | |
2288 | int run_start, run_end, run_eqv_end; | |
2289 | ||
2290 | Lisp_Object eqv_table; | |
2291 | ||
2292 | if (!RE_TRANSLATE_P (translate)) | |
2293 | { | |
b7c12565 | 2294 | EXTEND_RANGE_TABLE (work_area, 2); |
77d11aec RS |
2295 | work_area->table[work_area->used++] = (start); |
2296 | work_area->table[work_area->used++] = (end); | |
b7c12565 | 2297 | return -1; |
77d11aec RS |
2298 | } |
2299 | ||
2300 | eqv_table = XCHAR_TABLE (translate)->extras[2]; | |
99633e97 | 2301 | |
77d11aec RS |
2302 | for (; start <= end; start++) |
2303 | { | |
2304 | enum case_type this_type; | |
2305 | int eqv = RE_TRANSLATE (eqv_table, start); | |
2306 | int minchar, maxchar; | |
2307 | ||
2308 | /* Classify this character */ | |
2309 | if (eqv == start) | |
2310 | this_type = one_case; | |
2311 | else if (RE_TRANSLATE (eqv_table, eqv) == start) | |
2312 | this_type = two_case; | |
2313 | else | |
2314 | this_type = strange; | |
2315 | ||
2316 | if (start < eqv) | |
2317 | minchar = start, maxchar = eqv; | |
2318 | else | |
2319 | minchar = eqv, maxchar = start; | |
2320 | ||
2321 | /* Can this character extend the run in progress? */ | |
2322 | if (this_type == strange || this_type != run_type | |
2323 | || !(minchar == run_end + 1 | |
2324 | && (run_type == two_case | |
2325 | ? maxchar == run_eqv_end + 1 : 1))) | |
2326 | { | |
2327 | /* No, end the run. | |
2328 | Record each of its equivalent ranges. */ | |
2329 | if (run_type == one_case) | |
2330 | { | |
2331 | EXTEND_RANGE_TABLE (work_area, 2); | |
2332 | work_area->table[work_area->used++] = run_start; | |
2333 | work_area->table[work_area->used++] = run_end; | |
2334 | } | |
2335 | else if (run_type == two_case) | |
2336 | { | |
2337 | EXTEND_RANGE_TABLE (work_area, 4); | |
2338 | work_area->table[work_area->used++] = run_start; | |
2339 | work_area->table[work_area->used++] = run_end; | |
2340 | work_area->table[work_area->used++] | |
2341 | = RE_TRANSLATE (eqv_table, run_start); | |
2342 | work_area->table[work_area->used++] | |
2343 | = RE_TRANSLATE (eqv_table, run_end); | |
2344 | } | |
2345 | run_type = strange; | |
2346 | } | |
177c0ea7 | 2347 | |
77d11aec RS |
2348 | if (this_type == strange) |
2349 | { | |
2350 | /* For a strange character, add each of its equivalents, one | |
2351 | by one. Don't start a range. */ | |
2352 | do | |
2353 | { | |
2354 | EXTEND_RANGE_TABLE (work_area, 2); | |
2355 | work_area->table[work_area->used++] = eqv; | |
2356 | work_area->table[work_area->used++] = eqv; | |
2357 | eqv = RE_TRANSLATE (eqv_table, eqv); | |
2358 | } | |
2359 | while (eqv != start); | |
2360 | } | |
2361 | ||
2362 | /* Add this char to the run, or start a new run. */ | |
2363 | else if (run_type == strange) | |
2364 | { | |
2365 | /* Initialize a new range. */ | |
2366 | run_type = this_type; | |
2367 | run_start = start; | |
2368 | run_end = start; | |
2369 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2370 | } | |
2371 | else | |
2372 | { | |
2373 | /* Extend a running range. */ | |
2374 | run_end = minchar; | |
2375 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2376 | } | |
2377 | } | |
2378 | ||
2379 | /* If a run is still in progress at the end, finish it now | |
2380 | by recording its equivalent ranges. */ | |
2381 | if (run_type == one_case) | |
2382 | { | |
2383 | EXTEND_RANGE_TABLE (work_area, 2); | |
2384 | work_area->table[work_area->used++] = run_start; | |
2385 | work_area->table[work_area->used++] = run_end; | |
2386 | } | |
2387 | else if (run_type == two_case) | |
2388 | { | |
2389 | EXTEND_RANGE_TABLE (work_area, 4); | |
2390 | work_area->table[work_area->used++] = run_start; | |
2391 | work_area->table[work_area->used++] = run_end; | |
2392 | work_area->table[work_area->used++] | |
2393 | = RE_TRANSLATE (eqv_table, run_start); | |
2394 | work_area->table[work_area->used++] | |
2395 | = RE_TRANSLATE (eqv_table, run_end); | |
2396 | } | |
2397 | ||
2398 | return -1; | |
2399 | } | |
36595814 | 2400 | |
77d11aec | 2401 | #endif /* emacs */ |
36595814 | 2402 | |
2b34df4e | 2403 | /* Record the image of the range start..end when passed through |
36595814 SM |
2404 | TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end) |
2405 | and is not even necessarily contiguous. | |
b7c12565 RS |
2406 | Normally we approximate it with the smallest contiguous range that contains |
2407 | all the chars we need. However, for Latin-1 we go to extra effort | |
2408 | to do a better job. | |
2409 | ||
2410 | This function is not called for ASCII ranges. | |
77d11aec RS |
2411 | |
2412 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2413 | ||
2414 | static int | |
36595814 SM |
2415 | set_image_of_range (work_area, start, end, translate) |
2416 | RE_TRANSLATE_TYPE translate; | |
2417 | struct range_table_work_area *work_area; | |
2418 | re_wchar_t start, end; | |
2419 | { | |
77d11aec RS |
2420 | re_wchar_t cmin, cmax; |
2421 | ||
2422 | #ifdef emacs | |
2423 | /* For Latin-1 ranges, use set_image_of_range_1 | |
2424 | to get proper handling of ranges that include letters and nonletters. | |
b7c12565 | 2425 | For a range that includes the whole of Latin-1, this is not necessary. |
77d11aec | 2426 | For other character sets, we don't bother to get this right. */ |
b7c12565 RS |
2427 | if (RE_TRANSLATE_P (translate) && start < 04400 |
2428 | && !(start < 04200 && end >= 04377)) | |
77d11aec | 2429 | { |
b7c12565 | 2430 | int newend; |
77d11aec | 2431 | int tem; |
b7c12565 RS |
2432 | newend = end; |
2433 | if (newend > 04377) | |
2434 | newend = 04377; | |
2435 | tem = set_image_of_range_1 (work_area, start, newend, translate); | |
77d11aec RS |
2436 | if (tem > 0) |
2437 | return tem; | |
2438 | ||
2439 | start = 04400; | |
2440 | if (end < 04400) | |
2441 | return -1; | |
2442 | } | |
2443 | #endif | |
2444 | ||
b7c12565 RS |
2445 | EXTEND_RANGE_TABLE (work_area, 2); |
2446 | work_area->table[work_area->used++] = (start); | |
2447 | work_area->table[work_area->used++] = (end); | |
2448 | ||
2449 | cmin = -1, cmax = -1; | |
77d11aec | 2450 | |
36595814 | 2451 | if (RE_TRANSLATE_P (translate)) |
b7c12565 RS |
2452 | { |
2453 | int ch; | |
77d11aec | 2454 | |
b7c12565 RS |
2455 | for (ch = start; ch <= end; ch++) |
2456 | { | |
2457 | re_wchar_t c = TRANSLATE (ch); | |
2458 | if (! (start <= c && c <= end)) | |
2459 | { | |
2460 | if (cmin == -1) | |
2461 | cmin = c, cmax = c; | |
2462 | else | |
2463 | { | |
2464 | cmin = MIN (cmin, c); | |
2465 | cmax = MAX (cmax, c); | |
2466 | } | |
2467 | } | |
2468 | } | |
2469 | ||
2470 | if (cmin != -1) | |
2471 | { | |
2472 | EXTEND_RANGE_TABLE (work_area, 2); | |
2473 | work_area->table[work_area->used++] = (cmin); | |
2474 | work_area->table[work_area->used++] = (cmax); | |
2475 | } | |
2476 | } | |
36595814 | 2477 | |
77d11aec RS |
2478 | return -1; |
2479 | } | |
8f924df7 | 2480 | #endif /* 0 */ |
fa9a63c5 RM |
2481 | \f |
2482 | #ifndef MATCH_MAY_ALLOCATE | |
2483 | ||
2484 | /* If we cannot allocate large objects within re_match_2_internal, | |
2485 | we make the fail stack and register vectors global. | |
2486 | The fail stack, we grow to the maximum size when a regexp | |
2487 | is compiled. | |
2488 | The register vectors, we adjust in size each time we | |
2489 | compile a regexp, according to the number of registers it needs. */ | |
2490 | ||
2491 | static fail_stack_type fail_stack; | |
2492 | ||
2493 | /* Size with which the following vectors are currently allocated. | |
2494 | That is so we can make them bigger as needed, | |
4bb91c68 | 2495 | but never make them smaller. */ |
fa9a63c5 RM |
2496 | static int regs_allocated_size; |
2497 | ||
66f0296e SM |
2498 | static re_char ** regstart, ** regend; |
2499 | static re_char **best_regstart, **best_regend; | |
fa9a63c5 RM |
2500 | |
2501 | /* Make the register vectors big enough for NUM_REGS registers, | |
4bb91c68 | 2502 | but don't make them smaller. */ |
fa9a63c5 RM |
2503 | |
2504 | static | |
2505 | regex_grow_registers (num_regs) | |
2506 | int num_regs; | |
2507 | { | |
2508 | if (num_regs > regs_allocated_size) | |
2509 | { | |
66f0296e SM |
2510 | RETALLOC_IF (regstart, num_regs, re_char *); |
2511 | RETALLOC_IF (regend, num_regs, re_char *); | |
2512 | RETALLOC_IF (best_regstart, num_regs, re_char *); | |
2513 | RETALLOC_IF (best_regend, num_regs, re_char *); | |
fa9a63c5 RM |
2514 | |
2515 | regs_allocated_size = num_regs; | |
2516 | } | |
2517 | } | |
2518 | ||
2519 | #endif /* not MATCH_MAY_ALLOCATE */ | |
2520 | \f | |
99633e97 SM |
2521 | static boolean group_in_compile_stack _RE_ARGS ((compile_stack_type |
2522 | compile_stack, | |
2523 | regnum_t regnum)); | |
2524 | ||
fa9a63c5 RM |
2525 | /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. |
2526 | Returns one of error codes defined in `regex.h', or zero for success. | |
2527 | ||
2528 | Assumes the `allocated' (and perhaps `buffer') and `translate' | |
2529 | fields are set in BUFP on entry. | |
2530 | ||
2531 | If it succeeds, results are put in BUFP (if it returns an error, the | |
2532 | contents of BUFP are undefined): | |
2533 | `buffer' is the compiled pattern; | |
2534 | `syntax' is set to SYNTAX; | |
2535 | `used' is set to the length of the compiled pattern; | |
2536 | `fastmap_accurate' is zero; | |
2537 | `re_nsub' is the number of subexpressions in PATTERN; | |
2538 | `not_bol' and `not_eol' are zero; | |
5e69f11e | 2539 | |
c0f9ea08 | 2540 | The `fastmap' field is neither examined nor set. */ |
fa9a63c5 | 2541 | |
505bde11 SM |
2542 | /* Insert the `jump' from the end of last alternative to "here". |
2543 | The space for the jump has already been allocated. */ | |
2544 | #define FIXUP_ALT_JUMP() \ | |
2545 | do { \ | |
2546 | if (fixup_alt_jump) \ | |
2547 | STORE_JUMP (jump, fixup_alt_jump, b); \ | |
2548 | } while (0) | |
2549 | ||
2550 | ||
fa9a63c5 RM |
2551 | /* Return, freeing storage we allocated. */ |
2552 | #define FREE_STACK_RETURN(value) \ | |
b18215fc RS |
2553 | do { \ |
2554 | FREE_RANGE_TABLE_WORK_AREA (range_table_work); \ | |
2555 | free (compile_stack.stack); \ | |
2556 | return value; \ | |
2557 | } while (0) | |
fa9a63c5 RM |
2558 | |
2559 | static reg_errcode_t | |
2560 | regex_compile (pattern, size, syntax, bufp) | |
66f0296e | 2561 | re_char *pattern; |
4bb91c68 | 2562 | size_t size; |
fa9a63c5 RM |
2563 | reg_syntax_t syntax; |
2564 | struct re_pattern_buffer *bufp; | |
2565 | { | |
01618498 SM |
2566 | /* We fetch characters from PATTERN here. */ |
2567 | register re_wchar_t c, c1; | |
5e69f11e | 2568 | |
fa9a63c5 | 2569 | /* A random temporary spot in PATTERN. */ |
66f0296e | 2570 | re_char *p1; |
fa9a63c5 RM |
2571 | |
2572 | /* Points to the end of the buffer, where we should append. */ | |
2573 | register unsigned char *b; | |
5e69f11e | 2574 | |
fa9a63c5 RM |
2575 | /* Keeps track of unclosed groups. */ |
2576 | compile_stack_type compile_stack; | |
2577 | ||
2578 | /* Points to the current (ending) position in the pattern. */ | |
22336245 RS |
2579 | #ifdef AIX |
2580 | /* `const' makes AIX compiler fail. */ | |
66f0296e | 2581 | unsigned char *p = pattern; |
22336245 | 2582 | #else |
66f0296e | 2583 | re_char *p = pattern; |
22336245 | 2584 | #endif |
66f0296e | 2585 | re_char *pend = pattern + size; |
5e69f11e | 2586 | |
fa9a63c5 | 2587 | /* How to translate the characters in the pattern. */ |
6676cb1c | 2588 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 RM |
2589 | |
2590 | /* Address of the count-byte of the most recently inserted `exactn' | |
2591 | command. This makes it possible to tell if a new exact-match | |
2592 | character can be added to that command or if the character requires | |
2593 | a new `exactn' command. */ | |
2594 | unsigned char *pending_exact = 0; | |
2595 | ||
2596 | /* Address of start of the most recently finished expression. | |
2597 | This tells, e.g., postfix * where to find the start of its | |
2598 | operand. Reset at the beginning of groups and alternatives. */ | |
2599 | unsigned char *laststart = 0; | |
2600 | ||
2601 | /* Address of beginning of regexp, or inside of last group. */ | |
2602 | unsigned char *begalt; | |
2603 | ||
2604 | /* Place in the uncompiled pattern (i.e., the {) to | |
2605 | which to go back if the interval is invalid. */ | |
66f0296e | 2606 | re_char *beg_interval; |
5e69f11e | 2607 | |
fa9a63c5 | 2608 | /* Address of the place where a forward jump should go to the end of |
7814e705 | 2609 | the containing expression. Each alternative of an `or' -- except the |
fa9a63c5 RM |
2610 | last -- ends with a forward jump of this sort. */ |
2611 | unsigned char *fixup_alt_jump = 0; | |
2612 | ||
b18215fc RS |
2613 | /* Work area for range table of charset. */ |
2614 | struct range_table_work_area range_table_work; | |
2615 | ||
2d1675e4 SM |
2616 | /* If the object matched can contain multibyte characters. */ |
2617 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | |
2618 | ||
8f924df7 | 2619 | /* If a target of matching can contain multibyte characters. */ |
6fdd04b0 KH |
2620 | const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp); |
2621 | ||
f9b0fd99 RS |
2622 | /* Nonzero if we have pushed down into a subpattern. */ |
2623 | int in_subpattern = 0; | |
2624 | ||
2625 | /* These hold the values of p, pattern, and pend from the main | |
2626 | pattern when we have pushed into a subpattern. */ | |
2627 | re_char *main_p; | |
2628 | re_char *main_pattern; | |
2629 | re_char *main_pend; | |
2630 | ||
fa9a63c5 | 2631 | #ifdef DEBUG |
99633e97 | 2632 | debug++; |
fa9a63c5 | 2633 | DEBUG_PRINT1 ("\nCompiling pattern: "); |
99633e97 | 2634 | if (debug > 0) |
fa9a63c5 RM |
2635 | { |
2636 | unsigned debug_count; | |
5e69f11e | 2637 | |
fa9a63c5 | 2638 | for (debug_count = 0; debug_count < size; debug_count++) |
25fe55af | 2639 | putchar (pattern[debug_count]); |
fa9a63c5 RM |
2640 | putchar ('\n'); |
2641 | } | |
2642 | #endif /* DEBUG */ | |
2643 | ||
2644 | /* Initialize the compile stack. */ | |
2645 | compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); | |
2646 | if (compile_stack.stack == NULL) | |
2647 | return REG_ESPACE; | |
2648 | ||
2649 | compile_stack.size = INIT_COMPILE_STACK_SIZE; | |
2650 | compile_stack.avail = 0; | |
2651 | ||
b18215fc RS |
2652 | range_table_work.table = 0; |
2653 | range_table_work.allocated = 0; | |
2654 | ||
fa9a63c5 RM |
2655 | /* Initialize the pattern buffer. */ |
2656 | bufp->syntax = syntax; | |
2657 | bufp->fastmap_accurate = 0; | |
2658 | bufp->not_bol = bufp->not_eol = 0; | |
6224b623 | 2659 | bufp->used_syntax = 0; |
fa9a63c5 RM |
2660 | |
2661 | /* Set `used' to zero, so that if we return an error, the pattern | |
2662 | printer (for debugging) will think there's no pattern. We reset it | |
2663 | at the end. */ | |
2664 | bufp->used = 0; | |
5e69f11e | 2665 | |
fa9a63c5 | 2666 | /* Always count groups, whether or not bufp->no_sub is set. */ |
5e69f11e | 2667 | bufp->re_nsub = 0; |
fa9a63c5 | 2668 | |
0b32bf0e | 2669 | #if !defined emacs && !defined SYNTAX_TABLE |
fa9a63c5 RM |
2670 | /* Initialize the syntax table. */ |
2671 | init_syntax_once (); | |
2672 | #endif | |
2673 | ||
2674 | if (bufp->allocated == 0) | |
2675 | { | |
2676 | if (bufp->buffer) | |
2677 | { /* If zero allocated, but buffer is non-null, try to realloc | |
25fe55af | 2678 | enough space. This loses if buffer's address is bogus, but |
7814e705 | 2679 | that is the user's responsibility. */ |
25fe55af RS |
2680 | RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); |
2681 | } | |
fa9a63c5 | 2682 | else |
7814e705 | 2683 | { /* Caller did not allocate a buffer. Do it for them. */ |
25fe55af RS |
2684 | bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); |
2685 | } | |
fa9a63c5 RM |
2686 | if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE); |
2687 | ||
2688 | bufp->allocated = INIT_BUF_SIZE; | |
2689 | } | |
2690 | ||
2691 | begalt = b = bufp->buffer; | |
2692 | ||
2693 | /* Loop through the uncompiled pattern until we're at the end. */ | |
f9b0fd99 | 2694 | while (1) |
fa9a63c5 | 2695 | { |
f9b0fd99 RS |
2696 | if (p == pend) |
2697 | { | |
2698 | /* If this is the end of an included regexp, | |
2699 | pop back to the main regexp and try again. */ | |
2700 | if (in_subpattern) | |
2701 | { | |
2702 | in_subpattern = 0; | |
2703 | pattern = main_pattern; | |
2704 | p = main_p; | |
2705 | pend = main_pend; | |
2706 | continue; | |
2707 | } | |
2708 | /* If this is the end of the main regexp, we are done. */ | |
2709 | break; | |
2710 | } | |
2711 | ||
fa9a63c5 RM |
2712 | PATFETCH (c); |
2713 | ||
2714 | switch (c) | |
25fe55af | 2715 | { |
f9b0fd99 RS |
2716 | case ' ': |
2717 | { | |
2718 | re_char *p1 = p; | |
2719 | ||
2720 | /* If there's no special whitespace regexp, treat | |
4fb680cd RS |
2721 | spaces normally. And don't try to do this recursively. */ |
2722 | if (!whitespace_regexp || in_subpattern) | |
f9b0fd99 RS |
2723 | goto normal_char; |
2724 | ||
2725 | /* Peek past following spaces. */ | |
2726 | while (p1 != pend) | |
2727 | { | |
2728 | if (*p1 != ' ') | |
2729 | break; | |
2730 | p1++; | |
2731 | } | |
2732 | /* If the spaces are followed by a repetition op, | |
2733 | treat them normally. */ | |
c721eee5 RS |
2734 | if (p1 != pend |
2735 | && (*p1 == '*' || *p1 == '+' || *p1 == '?' | |
f9b0fd99 RS |
2736 | || (*p1 == '\\' && p1 + 1 != pend && p1[1] == '{'))) |
2737 | goto normal_char; | |
2738 | ||
2739 | /* Replace the spaces with the whitespace regexp. */ | |
2740 | in_subpattern = 1; | |
2741 | main_p = p1; | |
2742 | main_pend = pend; | |
2743 | main_pattern = pattern; | |
2744 | p = pattern = whitespace_regexp; | |
2745 | pend = p + strlen (p); | |
2746 | break; | |
7814e705 | 2747 | } |
f9b0fd99 | 2748 | |
25fe55af RS |
2749 | case '^': |
2750 | { | |
7814e705 | 2751 | if ( /* If at start of pattern, it's an operator. */ |
25fe55af | 2752 | p == pattern + 1 |
7814e705 | 2753 | /* If context independent, it's an operator. */ |
25fe55af | 2754 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
7814e705 | 2755 | /* Otherwise, depends on what's come before. */ |
25fe55af | 2756 | || at_begline_loc_p (pattern, p, syntax)) |
c0f9ea08 | 2757 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? begbuf : begline); |
25fe55af RS |
2758 | else |
2759 | goto normal_char; | |
2760 | } | |
2761 | break; | |
2762 | ||
2763 | ||
2764 | case '$': | |
2765 | { | |
2766 | if ( /* If at end of pattern, it's an operator. */ | |
2767 | p == pend | |
7814e705 | 2768 | /* If context independent, it's an operator. */ |
25fe55af RS |
2769 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
2770 | /* Otherwise, depends on what's next. */ | |
2771 | || at_endline_loc_p (p, pend, syntax)) | |
c0f9ea08 | 2772 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? endbuf : endline); |
25fe55af RS |
2773 | else |
2774 | goto normal_char; | |
2775 | } | |
2776 | break; | |
fa9a63c5 RM |
2777 | |
2778 | ||
2779 | case '+': | |
25fe55af RS |
2780 | case '?': |
2781 | if ((syntax & RE_BK_PLUS_QM) | |
2782 | || (syntax & RE_LIMITED_OPS)) | |
2783 | goto normal_char; | |
2784 | handle_plus: | |
2785 | case '*': | |
2786 | /* If there is no previous pattern... */ | |
2787 | if (!laststart) | |
2788 | { | |
2789 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
2790 | FREE_STACK_RETURN (REG_BADRPT); | |
2791 | else if (!(syntax & RE_CONTEXT_INDEP_OPS)) | |
2792 | goto normal_char; | |
2793 | } | |
2794 | ||
2795 | { | |
7814e705 | 2796 | /* 1 means zero (many) matches is allowed. */ |
66f0296e SM |
2797 | boolean zero_times_ok = 0, many_times_ok = 0; |
2798 | boolean greedy = 1; | |
25fe55af RS |
2799 | |
2800 | /* If there is a sequence of repetition chars, collapse it | |
2801 | down to just one (the right one). We can't combine | |
2802 | interval operators with these because of, e.g., `a{2}*', | |
7814e705 | 2803 | which should only match an even number of `a's. */ |
25fe55af RS |
2804 | |
2805 | for (;;) | |
2806 | { | |
0b32bf0e | 2807 | if ((syntax & RE_FRUGAL) |
1c8c6d39 DL |
2808 | && c == '?' && (zero_times_ok || many_times_ok)) |
2809 | greedy = 0; | |
2810 | else | |
2811 | { | |
2812 | zero_times_ok |= c != '+'; | |
2813 | many_times_ok |= c != '?'; | |
2814 | } | |
25fe55af RS |
2815 | |
2816 | if (p == pend) | |
2817 | break; | |
ed0767d8 SM |
2818 | else if (*p == '*' |
2819 | || (!(syntax & RE_BK_PLUS_QM) | |
2820 | && (*p == '+' || *p == '?'))) | |
25fe55af | 2821 | ; |
ed0767d8 | 2822 | else if (syntax & RE_BK_PLUS_QM && *p == '\\') |
25fe55af | 2823 | { |
ed0767d8 SM |
2824 | if (p+1 == pend) |
2825 | FREE_STACK_RETURN (REG_EESCAPE); | |
2826 | if (p[1] == '+' || p[1] == '?') | |
2827 | PATFETCH (c); /* Gobble up the backslash. */ | |
2828 | else | |
2829 | break; | |
25fe55af RS |
2830 | } |
2831 | else | |
ed0767d8 | 2832 | break; |
25fe55af | 2833 | /* If we get here, we found another repeat character. */ |
ed0767d8 SM |
2834 | PATFETCH (c); |
2835 | } | |
25fe55af RS |
2836 | |
2837 | /* Star, etc. applied to an empty pattern is equivalent | |
2838 | to an empty pattern. */ | |
4e8a9132 | 2839 | if (!laststart || laststart == b) |
25fe55af RS |
2840 | break; |
2841 | ||
2842 | /* Now we know whether or not zero matches is allowed | |
7814e705 | 2843 | and also whether or not two or more matches is allowed. */ |
1c8c6d39 DL |
2844 | if (greedy) |
2845 | { | |
99633e97 | 2846 | if (many_times_ok) |
4e8a9132 SM |
2847 | { |
2848 | boolean simple = skip_one_char (laststart) == b; | |
2849 | unsigned int startoffset = 0; | |
f6a3f532 | 2850 | re_opcode_t ofj = |
01618498 | 2851 | /* Check if the loop can match the empty string. */ |
6df42991 SM |
2852 | (simple || !analyse_first (laststart, b, NULL, 0)) |
2853 | ? on_failure_jump : on_failure_jump_loop; | |
4e8a9132 | 2854 | assert (skip_one_char (laststart) <= b); |
177c0ea7 | 2855 | |
4e8a9132 SM |
2856 | if (!zero_times_ok && simple) |
2857 | { /* Since simple * loops can be made faster by using | |
2858 | on_failure_keep_string_jump, we turn simple P+ | |
2859 | into PP* if P is simple. */ | |
2860 | unsigned char *p1, *p2; | |
2861 | startoffset = b - laststart; | |
2862 | GET_BUFFER_SPACE (startoffset); | |
2863 | p1 = b; p2 = laststart; | |
2864 | while (p2 < p1) | |
2865 | *b++ = *p2++; | |
2866 | zero_times_ok = 1; | |
99633e97 | 2867 | } |
4e8a9132 SM |
2868 | |
2869 | GET_BUFFER_SPACE (6); | |
2870 | if (!zero_times_ok) | |
2871 | /* A + loop. */ | |
f6a3f532 | 2872 | STORE_JUMP (ofj, b, b + 6); |
99633e97 | 2873 | else |
4e8a9132 SM |
2874 | /* Simple * loops can use on_failure_keep_string_jump |
2875 | depending on what follows. But since we don't know | |
2876 | that yet, we leave the decision up to | |
2877 | on_failure_jump_smart. */ | |
f6a3f532 | 2878 | INSERT_JUMP (simple ? on_failure_jump_smart : ofj, |
4e8a9132 | 2879 | laststart + startoffset, b + 6); |
99633e97 | 2880 | b += 3; |
4e8a9132 | 2881 | STORE_JUMP (jump, b, laststart + startoffset); |
99633e97 SM |
2882 | b += 3; |
2883 | } | |
2884 | else | |
2885 | { | |
4e8a9132 SM |
2886 | /* A simple ? pattern. */ |
2887 | assert (zero_times_ok); | |
2888 | GET_BUFFER_SPACE (3); | |
2889 | INSERT_JUMP (on_failure_jump, laststart, b + 3); | |
99633e97 SM |
2890 | b += 3; |
2891 | } | |
1c8c6d39 DL |
2892 | } |
2893 | else /* not greedy */ | |
2894 | { /* I wish the greedy and non-greedy cases could be merged. */ | |
2895 | ||
0683b6fa | 2896 | GET_BUFFER_SPACE (7); /* We might use less. */ |
1c8c6d39 DL |
2897 | if (many_times_ok) |
2898 | { | |
f6a3f532 SM |
2899 | boolean emptyp = analyse_first (laststart, b, NULL, 0); |
2900 | ||
6df42991 SM |
2901 | /* The non-greedy multiple match looks like |
2902 | a repeat..until: we only need a conditional jump | |
2903 | at the end of the loop. */ | |
f6a3f532 SM |
2904 | if (emptyp) BUF_PUSH (no_op); |
2905 | STORE_JUMP (emptyp ? on_failure_jump_nastyloop | |
2906 | : on_failure_jump, b, laststart); | |
1c8c6d39 DL |
2907 | b += 3; |
2908 | if (zero_times_ok) | |
2909 | { | |
2910 | /* The repeat...until naturally matches one or more. | |
2911 | To also match zero times, we need to first jump to | |
6df42991 | 2912 | the end of the loop (its conditional jump). */ |
1c8c6d39 DL |
2913 | INSERT_JUMP (jump, laststart, b); |
2914 | b += 3; | |
2915 | } | |
2916 | } | |
2917 | else | |
2918 | { | |
2919 | /* non-greedy a?? */ | |
1c8c6d39 DL |
2920 | INSERT_JUMP (jump, laststart, b + 3); |
2921 | b += 3; | |
2922 | INSERT_JUMP (on_failure_jump, laststart, laststart + 6); | |
2923 | b += 3; | |
2924 | } | |
2925 | } | |
2926 | } | |
4e8a9132 | 2927 | pending_exact = 0; |
fa9a63c5 RM |
2928 | break; |
2929 | ||
2930 | ||
2931 | case '.': | |
25fe55af RS |
2932 | laststart = b; |
2933 | BUF_PUSH (anychar); | |
2934 | break; | |
fa9a63c5 RM |
2935 | |
2936 | ||
25fe55af RS |
2937 | case '[': |
2938 | { | |
b18215fc | 2939 | CLEAR_RANGE_TABLE_WORK_USED (range_table_work); |
fa9a63c5 | 2940 | |
25fe55af | 2941 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2942 | |
25fe55af RS |
2943 | /* Ensure that we have enough space to push a charset: the |
2944 | opcode, the length count, and the bitset; 34 bytes in all. */ | |
fa9a63c5 RM |
2945 | GET_BUFFER_SPACE (34); |
2946 | ||
25fe55af | 2947 | laststart = b; |
e318085a | 2948 | |
25fe55af | 2949 | /* We test `*p == '^' twice, instead of using an if |
7814e705 | 2950 | statement, so we only need one BUF_PUSH. */ |
25fe55af RS |
2951 | BUF_PUSH (*p == '^' ? charset_not : charset); |
2952 | if (*p == '^') | |
2953 | p++; | |
e318085a | 2954 | |
25fe55af RS |
2955 | /* Remember the first position in the bracket expression. */ |
2956 | p1 = p; | |
e318085a | 2957 | |
7814e705 | 2958 | /* Push the number of bytes in the bitmap. */ |
25fe55af | 2959 | BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2960 | |
25fe55af RS |
2961 | /* Clear the whole map. */ |
2962 | bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH); | |
e318085a | 2963 | |
25fe55af RS |
2964 | /* charset_not matches newline according to a syntax bit. */ |
2965 | if ((re_opcode_t) b[-2] == charset_not | |
2966 | && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) | |
2967 | SET_LIST_BIT ('\n'); | |
fa9a63c5 | 2968 | |
7814e705 | 2969 | /* Read in characters and ranges, setting map bits. */ |
25fe55af RS |
2970 | for (;;) |
2971 | { | |
b18215fc | 2972 | boolean escaped_char = false; |
2d1675e4 | 2973 | const unsigned char *p2 = p; |
cf9c99bc | 2974 | re_wchar_t ch, c2; |
e318085a | 2975 | |
25fe55af | 2976 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
e318085a | 2977 | |
36595814 SM |
2978 | /* Don't translate yet. The range TRANSLATE(X..Y) cannot |
2979 | always be determined from TRANSLATE(X) and TRANSLATE(Y) | |
2980 | So the translation is done later in a loop. Example: | |
2981 | (let ((case-fold-search t)) (string-match "[A-_]" "A")) */ | |
25fe55af | 2982 | PATFETCH (c); |
e318085a | 2983 | |
25fe55af RS |
2984 | /* \ might escape characters inside [...] and [^...]. */ |
2985 | if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') | |
2986 | { | |
2987 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
e318085a RS |
2988 | |
2989 | PATFETCH (c); | |
b18215fc | 2990 | escaped_char = true; |
25fe55af | 2991 | } |
b18215fc RS |
2992 | else |
2993 | { | |
7814e705 | 2994 | /* Could be the end of the bracket expression. If it's |
657fcfbd RS |
2995 | not (i.e., when the bracket expression is `[]' so |
2996 | far), the ']' character bit gets set way below. */ | |
2d1675e4 | 2997 | if (c == ']' && p2 != p1) |
657fcfbd | 2998 | break; |
25fe55af | 2999 | } |
b18215fc | 3000 | |
25fe55af RS |
3001 | /* See if we're at the beginning of a possible character |
3002 | class. */ | |
b18215fc | 3003 | |
2d1675e4 SM |
3004 | if (!escaped_char && |
3005 | syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') | |
657fcfbd | 3006 | { |
7814e705 | 3007 | /* Leave room for the null. */ |
14473664 | 3008 | unsigned char str[CHAR_CLASS_MAX_LENGTH + 1]; |
ed0767d8 | 3009 | const unsigned char *class_beg; |
b18215fc | 3010 | |
25fe55af RS |
3011 | PATFETCH (c); |
3012 | c1 = 0; | |
ed0767d8 | 3013 | class_beg = p; |
b18215fc | 3014 | |
25fe55af RS |
3015 | /* If pattern is `[[:'. */ |
3016 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
b18215fc | 3017 | |
25fe55af RS |
3018 | for (;;) |
3019 | { | |
14473664 SM |
3020 | PATFETCH (c); |
3021 | if ((c == ':' && *p == ']') || p == pend) | |
3022 | break; | |
3023 | if (c1 < CHAR_CLASS_MAX_LENGTH) | |
3024 | str[c1++] = c; | |
3025 | else | |
3026 | /* This is in any case an invalid class name. */ | |
3027 | str[0] = '\0'; | |
25fe55af RS |
3028 | } |
3029 | str[c1] = '\0'; | |
b18215fc RS |
3030 | |
3031 | /* If isn't a word bracketed by `[:' and `:]': | |
3032 | undo the ending character, the letters, and | |
3033 | leave the leading `:' and `[' (but set bits for | |
3034 | them). */ | |
25fe55af RS |
3035 | if (c == ':' && *p == ']') |
3036 | { | |
14473664 | 3037 | re_wctype_t cc; |
8f924df7 | 3038 | int limit; |
14473664 SM |
3039 | |
3040 | cc = re_wctype (str); | |
3041 | ||
3042 | if (cc == 0) | |
fa9a63c5 RM |
3043 | FREE_STACK_RETURN (REG_ECTYPE); |
3044 | ||
14473664 SM |
3045 | /* Throw away the ] at the end of the character |
3046 | class. */ | |
3047 | PATFETCH (c); | |
fa9a63c5 | 3048 | |
14473664 | 3049 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 3050 | |
cf9c99bc KH |
3051 | #ifndef emacs |
3052 | for (ch = 0; ch < (1 << BYTEWIDTH); ++ch) | |
8f924df7 KH |
3053 | if (re_iswctype (btowc (ch), cc)) |
3054 | { | |
3055 | c = TRANSLATE (ch); | |
ed00c2ac KH |
3056 | if (c < (1 << BYTEWIDTH)) |
3057 | SET_LIST_BIT (c); | |
8f924df7 | 3058 | } |
cf9c99bc KH |
3059 | #else /* emacs */ |
3060 | /* Most character classes in a multibyte match | |
3061 | just set a flag. Exceptions are is_blank, | |
3062 | is_digit, is_cntrl, and is_xdigit, since | |
3063 | they can only match ASCII characters. We | |
3064 | don't need to handle them for multibyte. | |
3065 | They are distinguished by a negative wctype. */ | |
96cc36cc | 3066 | |
254c06a8 SM |
3067 | /* Setup the gl_state object to its buffer-defined |
3068 | value. This hardcodes the buffer-global | |
3069 | syntax-table for ASCII chars, while the other chars | |
3070 | will obey syntax-table properties. It's not ideal, | |
3071 | but it's the way it's been done until now. */ | |
d48cd3f4 | 3072 | SETUP_BUFFER_SYNTAX_TABLE (); |
254c06a8 | 3073 | |
cf9c99bc | 3074 | for (ch = 0; ch < 256; ++ch) |
25fe55af | 3075 | { |
cf9c99bc KH |
3076 | c = RE_CHAR_TO_MULTIBYTE (ch); |
3077 | if (! CHAR_BYTE8_P (c) | |
3078 | && re_iswctype (c, cc)) | |
8f924df7 | 3079 | { |
cf9c99bc KH |
3080 | SET_LIST_BIT (ch); |
3081 | c1 = TRANSLATE (c); | |
3082 | if (c1 == c) | |
3083 | continue; | |
3084 | if (ASCII_CHAR_P (c1)) | |
3085 | SET_LIST_BIT (c1); | |
3086 | else if ((c1 = RE_CHAR_TO_UNIBYTE (c1)) >= 0) | |
3087 | SET_LIST_BIT (c1); | |
8f924df7 | 3088 | } |
25fe55af | 3089 | } |
cf9c99bc KH |
3090 | SET_RANGE_TABLE_WORK_AREA_BIT |
3091 | (range_table_work, re_wctype_to_bit (cc)); | |
3092 | #endif /* emacs */ | |
6224b623 SM |
3093 | /* In most cases the matching rule for char classes |
3094 | only uses the syntax table for multibyte chars, | |
3095 | so that the content of the syntax-table it is not | |
3096 | hardcoded in the range_table. SPACE and WORD are | |
3097 | the two exceptions. */ | |
3098 | if ((1 << cc) & ((1 << RECC_SPACE) | (1 << RECC_WORD))) | |
3099 | bufp->used_syntax = 1; | |
3100 | ||
b18215fc RS |
3101 | /* Repeat the loop. */ |
3102 | continue; | |
25fe55af RS |
3103 | } |
3104 | else | |
3105 | { | |
ed0767d8 SM |
3106 | /* Go back to right after the "[:". */ |
3107 | p = class_beg; | |
25fe55af | 3108 | SET_LIST_BIT ('['); |
b18215fc RS |
3109 | |
3110 | /* Because the `:' may starts the range, we | |
3111 | can't simply set bit and repeat the loop. | |
7814e705 | 3112 | Instead, just set it to C and handle below. */ |
b18215fc | 3113 | c = ':'; |
25fe55af RS |
3114 | } |
3115 | } | |
b18215fc RS |
3116 | |
3117 | if (p < pend && p[0] == '-' && p[1] != ']') | |
3118 | { | |
3119 | ||
3120 | /* Discard the `-'. */ | |
3121 | PATFETCH (c1); | |
3122 | ||
3123 | /* Fetch the character which ends the range. */ | |
3124 | PATFETCH (c1); | |
cf9c99bc KH |
3125 | #ifdef emacs |
3126 | if (CHAR_BYTE8_P (c1) | |
3127 | && ! ASCII_CHAR_P (c) && ! CHAR_BYTE8_P (c)) | |
3128 | /* Treat the range from a multibyte character to | |
3129 | raw-byte character as empty. */ | |
3130 | c = c1 + 1; | |
3131 | #endif /* emacs */ | |
e318085a | 3132 | } |
25fe55af | 3133 | else |
b18215fc RS |
3134 | /* Range from C to C. */ |
3135 | c1 = c; | |
3136 | ||
cf9c99bc | 3137 | if (c > c1) |
25fe55af | 3138 | { |
cf9c99bc KH |
3139 | if (syntax & RE_NO_EMPTY_RANGES) |
3140 | FREE_STACK_RETURN (REG_ERANGEX); | |
3141 | /* Else, repeat the loop. */ | |
bf216479 | 3142 | } |
6fdd04b0 | 3143 | else |
25fe55af | 3144 | { |
cf9c99bc KH |
3145 | #ifndef emacs |
3146 | /* Set the range into bitmap */ | |
8f924df7 | 3147 | for (; c <= c1; c++) |
b18215fc | 3148 | { |
cf9c99bc KH |
3149 | ch = TRANSLATE (c); |
3150 | if (ch < (1 << BYTEWIDTH)) | |
3151 | SET_LIST_BIT (ch); | |
3152 | } | |
3153 | #else /* emacs */ | |
3154 | if (c < 128) | |
3155 | { | |
3156 | ch = MIN (127, c1); | |
3157 | SETUP_ASCII_RANGE (range_table_work, c, ch); | |
3158 | c = ch + 1; | |
3159 | if (CHAR_BYTE8_P (c1)) | |
3160 | c = BYTE8_TO_CHAR (128); | |
3161 | } | |
3162 | if (c <= c1) | |
3163 | { | |
3164 | if (CHAR_BYTE8_P (c)) | |
3165 | { | |
3166 | c = CHAR_TO_BYTE8 (c); | |
3167 | c1 = CHAR_TO_BYTE8 (c1); | |
3168 | for (; c <= c1; c++) | |
3169 | SET_LIST_BIT (c); | |
3170 | } | |
3171 | else if (multibyte) | |
3172 | { | |
3173 | SETUP_MULTIBYTE_RANGE (range_table_work, c, c1); | |
3174 | } | |
3175 | else | |
3176 | { | |
3177 | SETUP_UNIBYTE_RANGE (range_table_work, c, c1); | |
3178 | } | |
e934739e | 3179 | } |
cf9c99bc | 3180 | #endif /* emacs */ |
25fe55af | 3181 | } |
e318085a RS |
3182 | } |
3183 | ||
25fe55af | 3184 | /* Discard any (non)matching list bytes that are all 0 at the |
7814e705 | 3185 | end of the map. Decrease the map-length byte too. */ |
25fe55af RS |
3186 | while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) |
3187 | b[-1]--; | |
3188 | b += b[-1]; | |
fa9a63c5 | 3189 | |
96cc36cc RS |
3190 | /* Build real range table from work area. */ |
3191 | if (RANGE_TABLE_WORK_USED (range_table_work) | |
3192 | || RANGE_TABLE_WORK_BITS (range_table_work)) | |
b18215fc RS |
3193 | { |
3194 | int i; | |
3195 | int used = RANGE_TABLE_WORK_USED (range_table_work); | |
fa9a63c5 | 3196 | |
b18215fc | 3197 | /* Allocate space for COUNT + RANGE_TABLE. Needs two |
96cc36cc RS |
3198 | bytes for flags, two for COUNT, and three bytes for |
3199 | each character. */ | |
3200 | GET_BUFFER_SPACE (4 + used * 3); | |
fa9a63c5 | 3201 | |
b18215fc RS |
3202 | /* Indicate the existence of range table. */ |
3203 | laststart[1] |= 0x80; | |
fa9a63c5 | 3204 | |
96cc36cc RS |
3205 | /* Store the character class flag bits into the range table. |
3206 | If not in emacs, these flag bits are always 0. */ | |
3207 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) & 0xff; | |
3208 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) >> 8; | |
3209 | ||
b18215fc RS |
3210 | STORE_NUMBER_AND_INCR (b, used / 2); |
3211 | for (i = 0; i < used; i++) | |
3212 | STORE_CHARACTER_AND_INCR | |
3213 | (b, RANGE_TABLE_WORK_ELT (range_table_work, i)); | |
3214 | } | |
25fe55af RS |
3215 | } |
3216 | break; | |
fa9a63c5 RM |
3217 | |
3218 | ||
b18215fc | 3219 | case '(': |
25fe55af RS |
3220 | if (syntax & RE_NO_BK_PARENS) |
3221 | goto handle_open; | |
3222 | else | |
3223 | goto normal_char; | |
fa9a63c5 RM |
3224 | |
3225 | ||
25fe55af RS |
3226 | case ')': |
3227 | if (syntax & RE_NO_BK_PARENS) | |
3228 | goto handle_close; | |
3229 | else | |
3230 | goto normal_char; | |
e318085a RS |
3231 | |
3232 | ||
25fe55af RS |
3233 | case '\n': |
3234 | if (syntax & RE_NEWLINE_ALT) | |
3235 | goto handle_alt; | |
3236 | else | |
3237 | goto normal_char; | |
e318085a RS |
3238 | |
3239 | ||
b18215fc | 3240 | case '|': |
25fe55af RS |
3241 | if (syntax & RE_NO_BK_VBAR) |
3242 | goto handle_alt; | |
3243 | else | |
3244 | goto normal_char; | |
3245 | ||
3246 | ||
3247 | case '{': | |
3248 | if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) | |
3249 | goto handle_interval; | |
3250 | else | |
3251 | goto normal_char; | |
3252 | ||
3253 | ||
3254 | case '\\': | |
3255 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
3256 | ||
3257 | /* Do not translate the character after the \, so that we can | |
3258 | distinguish, e.g., \B from \b, even if we normally would | |
3259 | translate, e.g., B to b. */ | |
36595814 | 3260 | PATFETCH (c); |
25fe55af RS |
3261 | |
3262 | switch (c) | |
3263 | { | |
3264 | case '(': | |
3265 | if (syntax & RE_NO_BK_PARENS) | |
3266 | goto normal_backslash; | |
3267 | ||
3268 | handle_open: | |
505bde11 SM |
3269 | { |
3270 | int shy = 0; | |
c69b0314 | 3271 | regnum_t regnum = 0; |
505bde11 SM |
3272 | if (p+1 < pend) |
3273 | { | |
3274 | /* Look for a special (?...) construct */ | |
ed0767d8 | 3275 | if ((syntax & RE_SHY_GROUPS) && *p == '?') |
505bde11 | 3276 | { |
ed0767d8 | 3277 | PATFETCH (c); /* Gobble up the '?'. */ |
c69b0314 | 3278 | while (!shy) |
505bde11 | 3279 | { |
c69b0314 SM |
3280 | PATFETCH (c); |
3281 | switch (c) | |
3282 | { | |
3283 | case ':': shy = 1; break; | |
3284 | case '0': | |
3285 | /* An explicitly specified regnum must start | |
3286 | with non-0. */ | |
3287 | if (regnum == 0) | |
3288 | FREE_STACK_RETURN (REG_BADPAT); | |
3289 | case '1': case '2': case '3': case '4': | |
3290 | case '5': case '6': case '7': case '8': case '9': | |
3291 | regnum = 10*regnum + (c - '0'); break; | |
3292 | default: | |
3293 | /* Only (?:...) is supported right now. */ | |
3294 | FREE_STACK_RETURN (REG_BADPAT); | |
3295 | } | |
505bde11 SM |
3296 | } |
3297 | } | |
505bde11 SM |
3298 | } |
3299 | ||
3300 | if (!shy) | |
c69b0314 SM |
3301 | regnum = ++bufp->re_nsub; |
3302 | else if (regnum) | |
3303 | { /* It's actually not shy, but explicitly numbered. */ | |
3304 | shy = 0; | |
3305 | if (regnum > bufp->re_nsub) | |
3306 | bufp->re_nsub = regnum; | |
3307 | else if (regnum > bufp->re_nsub | |
3308 | /* Ideally, we'd want to check that the specified | |
3309 | group can't have matched (i.e. all subgroups | |
3310 | using the same regnum are in other branches of | |
3311 | OR patterns), but we don't currently keep track | |
3312 | of enough info to do that easily. */ | |
3313 | || group_in_compile_stack (compile_stack, regnum)) | |
3314 | FREE_STACK_RETURN (REG_BADPAT); | |
505bde11 | 3315 | } |
c69b0314 SM |
3316 | else |
3317 | /* It's really shy. */ | |
3318 | regnum = - bufp->re_nsub; | |
25fe55af | 3319 | |
99633e97 SM |
3320 | if (COMPILE_STACK_FULL) |
3321 | { | |
3322 | RETALLOC (compile_stack.stack, compile_stack.size << 1, | |
3323 | compile_stack_elt_t); | |
3324 | if (compile_stack.stack == NULL) return REG_ESPACE; | |
25fe55af | 3325 | |
99633e97 SM |
3326 | compile_stack.size <<= 1; |
3327 | } | |
25fe55af | 3328 | |
99633e97 | 3329 | /* These are the values to restore when we hit end of this |
7814e705 | 3330 | group. They are all relative offsets, so that if the |
99633e97 SM |
3331 | whole pattern moves because of realloc, they will still |
3332 | be valid. */ | |
3333 | COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; | |
3334 | COMPILE_STACK_TOP.fixup_alt_jump | |
3335 | = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; | |
3336 | COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer; | |
c69b0314 | 3337 | COMPILE_STACK_TOP.regnum = regnum; |
99633e97 | 3338 | |
c69b0314 SM |
3339 | /* Do not push a start_memory for groups beyond the last one |
3340 | we can represent in the compiled pattern. */ | |
3341 | if (regnum <= MAX_REGNUM && regnum > 0) | |
99633e97 SM |
3342 | BUF_PUSH_2 (start_memory, regnum); |
3343 | ||
3344 | compile_stack.avail++; | |
3345 | ||
3346 | fixup_alt_jump = 0; | |
3347 | laststart = 0; | |
3348 | begalt = b; | |
3349 | /* If we've reached MAX_REGNUM groups, then this open | |
3350 | won't actually generate any code, so we'll have to | |
3351 | clear pending_exact explicitly. */ | |
3352 | pending_exact = 0; | |
3353 | break; | |
505bde11 | 3354 | } |
25fe55af RS |
3355 | |
3356 | case ')': | |
3357 | if (syntax & RE_NO_BK_PARENS) goto normal_backslash; | |
3358 | ||
3359 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3360 | { |
3361 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3362 | goto normal_backslash; | |
3363 | else | |
3364 | FREE_STACK_RETURN (REG_ERPAREN); | |
3365 | } | |
25fe55af RS |
3366 | |
3367 | handle_close: | |
505bde11 | 3368 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3369 | |
3370 | /* See similar code for backslashed left paren above. */ | |
3371 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3372 | { |
3373 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3374 | goto normal_char; | |
3375 | else | |
3376 | FREE_STACK_RETURN (REG_ERPAREN); | |
3377 | } | |
25fe55af RS |
3378 | |
3379 | /* Since we just checked for an empty stack above, this | |
3380 | ``can't happen''. */ | |
3381 | assert (compile_stack.avail != 0); | |
3382 | { | |
3383 | /* We don't just want to restore into `regnum', because | |
3384 | later groups should continue to be numbered higher, | |
7814e705 | 3385 | as in `(ab)c(de)' -- the second group is #2. */ |
c69b0314 | 3386 | regnum_t regnum; |
25fe55af RS |
3387 | |
3388 | compile_stack.avail--; | |
3389 | begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; | |
3390 | fixup_alt_jump | |
3391 | = COMPILE_STACK_TOP.fixup_alt_jump | |
3392 | ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 | |
3393 | : 0; | |
3394 | laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; | |
c69b0314 | 3395 | regnum = COMPILE_STACK_TOP.regnum; |
b18215fc RS |
3396 | /* If we've reached MAX_REGNUM groups, then this open |
3397 | won't actually generate any code, so we'll have to | |
3398 | clear pending_exact explicitly. */ | |
3399 | pending_exact = 0; | |
e318085a | 3400 | |
25fe55af | 3401 | /* We're at the end of the group, so now we know how many |
7814e705 | 3402 | groups were inside this one. */ |
c69b0314 SM |
3403 | if (regnum <= MAX_REGNUM && regnum > 0) |
3404 | BUF_PUSH_2 (stop_memory, regnum); | |
25fe55af RS |
3405 | } |
3406 | break; | |
3407 | ||
3408 | ||
3409 | case '|': /* `\|'. */ | |
3410 | if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) | |
3411 | goto normal_backslash; | |
3412 | handle_alt: | |
3413 | if (syntax & RE_LIMITED_OPS) | |
3414 | goto normal_char; | |
3415 | ||
3416 | /* Insert before the previous alternative a jump which | |
7814e705 | 3417 | jumps to this alternative if the former fails. */ |
25fe55af RS |
3418 | GET_BUFFER_SPACE (3); |
3419 | INSERT_JUMP (on_failure_jump, begalt, b + 6); | |
3420 | pending_exact = 0; | |
3421 | b += 3; | |
3422 | ||
3423 | /* The alternative before this one has a jump after it | |
3424 | which gets executed if it gets matched. Adjust that | |
3425 | jump so it will jump to this alternative's analogous | |
3426 | jump (put in below, which in turn will jump to the next | |
3427 | (if any) alternative's such jump, etc.). The last such | |
3428 | jump jumps to the correct final destination. A picture: | |
3429 | _____ _____ | |
3430 | | | | | | |
3431 | | v | v | |
3432 | a | b | c | |
3433 | ||
3434 | If we are at `b', then fixup_alt_jump right now points to a | |
3435 | three-byte space after `a'. We'll put in the jump, set | |
3436 | fixup_alt_jump to right after `b', and leave behind three | |
3437 | bytes which we'll fill in when we get to after `c'. */ | |
3438 | ||
505bde11 | 3439 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3440 | |
3441 | /* Mark and leave space for a jump after this alternative, | |
3442 | to be filled in later either by next alternative or | |
3443 | when know we're at the end of a series of alternatives. */ | |
3444 | fixup_alt_jump = b; | |
3445 | GET_BUFFER_SPACE (3); | |
3446 | b += 3; | |
3447 | ||
3448 | laststart = 0; | |
3449 | begalt = b; | |
3450 | break; | |
3451 | ||
3452 | ||
3453 | case '{': | |
3454 | /* If \{ is a literal. */ | |
3455 | if (!(syntax & RE_INTERVALS) | |
3456 | /* If we're at `\{' and it's not the open-interval | |
3457 | operator. */ | |
4bb91c68 | 3458 | || (syntax & RE_NO_BK_BRACES)) |
25fe55af RS |
3459 | goto normal_backslash; |
3460 | ||
3461 | handle_interval: | |
3462 | { | |
3463 | /* If got here, then the syntax allows intervals. */ | |
3464 | ||
3465 | /* At least (most) this many matches must be made. */ | |
99633e97 | 3466 | int lower_bound = 0, upper_bound = -1; |
25fe55af | 3467 | |
ed0767d8 | 3468 | beg_interval = p; |
25fe55af | 3469 | |
25fe55af RS |
3470 | GET_UNSIGNED_NUMBER (lower_bound); |
3471 | ||
3472 | if (c == ',') | |
ed0767d8 | 3473 | GET_UNSIGNED_NUMBER (upper_bound); |
25fe55af RS |
3474 | else |
3475 | /* Interval such as `{1}' => match exactly once. */ | |
3476 | upper_bound = lower_bound; | |
3477 | ||
3478 | if (lower_bound < 0 || upper_bound > RE_DUP_MAX | |
ed0767d8 | 3479 | || (upper_bound >= 0 && lower_bound > upper_bound)) |
4bb91c68 | 3480 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3481 | |
3482 | if (!(syntax & RE_NO_BK_BRACES)) | |
3483 | { | |
4bb91c68 SM |
3484 | if (c != '\\') |
3485 | FREE_STACK_RETURN (REG_BADBR); | |
c72b0edd SM |
3486 | if (p == pend) |
3487 | FREE_STACK_RETURN (REG_EESCAPE); | |
25fe55af RS |
3488 | PATFETCH (c); |
3489 | } | |
3490 | ||
3491 | if (c != '}') | |
4bb91c68 | 3492 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3493 | |
3494 | /* We just parsed a valid interval. */ | |
3495 | ||
3496 | /* If it's invalid to have no preceding re. */ | |
3497 | if (!laststart) | |
3498 | { | |
3499 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
3500 | FREE_STACK_RETURN (REG_BADRPT); | |
3501 | else if (syntax & RE_CONTEXT_INDEP_OPS) | |
3502 | laststart = b; | |
3503 | else | |
3504 | goto unfetch_interval; | |
3505 | } | |
3506 | ||
6df42991 SM |
3507 | if (upper_bound == 0) |
3508 | /* If the upper bound is zero, just drop the sub pattern | |
3509 | altogether. */ | |
3510 | b = laststart; | |
3511 | else if (lower_bound == 1 && upper_bound == 1) | |
3512 | /* Just match it once: nothing to do here. */ | |
3513 | ; | |
3514 | ||
3515 | /* Otherwise, we have a nontrivial interval. When | |
3516 | we're all done, the pattern will look like: | |
3517 | set_number_at <jump count> <upper bound> | |
3518 | set_number_at <succeed_n count> <lower bound> | |
3519 | succeed_n <after jump addr> <succeed_n count> | |
3520 | <body of loop> | |
3521 | jump_n <succeed_n addr> <jump count> | |
3522 | (The upper bound and `jump_n' are omitted if | |
3523 | `upper_bound' is 1, though.) */ | |
3524 | else | |
3525 | { /* If the upper bound is > 1, we need to insert | |
3526 | more at the end of the loop. */ | |
3527 | unsigned int nbytes = (upper_bound < 0 ? 3 | |
3528 | : upper_bound > 1 ? 5 : 0); | |
3529 | unsigned int startoffset = 0; | |
3530 | ||
3531 | GET_BUFFER_SPACE (20); /* We might use less. */ | |
3532 | ||
3533 | if (lower_bound == 0) | |
3534 | { | |
3535 | /* A succeed_n that starts with 0 is really a | |
3536 | a simple on_failure_jump_loop. */ | |
3537 | INSERT_JUMP (on_failure_jump_loop, laststart, | |
3538 | b + 3 + nbytes); | |
3539 | b += 3; | |
3540 | } | |
3541 | else | |
3542 | { | |
3543 | /* Initialize lower bound of the `succeed_n', even | |
3544 | though it will be set during matching by its | |
3545 | attendant `set_number_at' (inserted next), | |
3546 | because `re_compile_fastmap' needs to know. | |
3547 | Jump to the `jump_n' we might insert below. */ | |
3548 | INSERT_JUMP2 (succeed_n, laststart, | |
3549 | b + 5 + nbytes, | |
3550 | lower_bound); | |
3551 | b += 5; | |
3552 | ||
3553 | /* Code to initialize the lower bound. Insert | |
7814e705 | 3554 | before the `succeed_n'. The `5' is the last two |
6df42991 SM |
3555 | bytes of this `set_number_at', plus 3 bytes of |
3556 | the following `succeed_n'. */ | |
3557 | insert_op2 (set_number_at, laststart, 5, lower_bound, b); | |
3558 | b += 5; | |
3559 | startoffset += 5; | |
3560 | } | |
3561 | ||
3562 | if (upper_bound < 0) | |
3563 | { | |
3564 | /* A negative upper bound stands for infinity, | |
3565 | in which case it degenerates to a plain jump. */ | |
3566 | STORE_JUMP (jump, b, laststart + startoffset); | |
3567 | b += 3; | |
3568 | } | |
3569 | else if (upper_bound > 1) | |
3570 | { /* More than one repetition is allowed, so | |
3571 | append a backward jump to the `succeed_n' | |
3572 | that starts this interval. | |
3573 | ||
3574 | When we've reached this during matching, | |
3575 | we'll have matched the interval once, so | |
3576 | jump back only `upper_bound - 1' times. */ | |
3577 | STORE_JUMP2 (jump_n, b, laststart + startoffset, | |
3578 | upper_bound - 1); | |
3579 | b += 5; | |
3580 | ||
3581 | /* The location we want to set is the second | |
3582 | parameter of the `jump_n'; that is `b-2' as | |
3583 | an absolute address. `laststart' will be | |
3584 | the `set_number_at' we're about to insert; | |
3585 | `laststart+3' the number to set, the source | |
3586 | for the relative address. But we are | |
3587 | inserting into the middle of the pattern -- | |
3588 | so everything is getting moved up by 5. | |
3589 | Conclusion: (b - 2) - (laststart + 3) + 5, | |
3590 | i.e., b - laststart. | |
3591 | ||
3592 | We insert this at the beginning of the loop | |
3593 | so that if we fail during matching, we'll | |
3594 | reinitialize the bounds. */ | |
3595 | insert_op2 (set_number_at, laststart, b - laststart, | |
3596 | upper_bound - 1, b); | |
3597 | b += 5; | |
3598 | } | |
3599 | } | |
25fe55af RS |
3600 | pending_exact = 0; |
3601 | beg_interval = NULL; | |
3602 | } | |
3603 | break; | |
3604 | ||
3605 | unfetch_interval: | |
3606 | /* If an invalid interval, match the characters as literals. */ | |
3607 | assert (beg_interval); | |
3608 | p = beg_interval; | |
3609 | beg_interval = NULL; | |
3610 | ||
3611 | /* normal_char and normal_backslash need `c'. */ | |
ed0767d8 | 3612 | c = '{'; |
25fe55af RS |
3613 | |
3614 | if (!(syntax & RE_NO_BK_BRACES)) | |
3615 | { | |
ed0767d8 SM |
3616 | assert (p > pattern && p[-1] == '\\'); |
3617 | goto normal_backslash; | |
25fe55af | 3618 | } |
ed0767d8 SM |
3619 | else |
3620 | goto normal_char; | |
e318085a | 3621 | |
b18215fc | 3622 | #ifdef emacs |
25fe55af | 3623 | /* There is no way to specify the before_dot and after_dot |
7814e705 | 3624 | operators. rms says this is ok. --karl */ |
25fe55af RS |
3625 | case '=': |
3626 | BUF_PUSH (at_dot); | |
3627 | break; | |
3628 | ||
3629 | case 's': | |
3630 | laststart = b; | |
3631 | PATFETCH (c); | |
3632 | BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); | |
3633 | break; | |
3634 | ||
3635 | case 'S': | |
3636 | laststart = b; | |
3637 | PATFETCH (c); | |
3638 | BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); | |
3639 | break; | |
b18215fc RS |
3640 | |
3641 | case 'c': | |
3642 | laststart = b; | |
36595814 | 3643 | PATFETCH (c); |
b18215fc RS |
3644 | BUF_PUSH_2 (categoryspec, c); |
3645 | break; | |
e318085a | 3646 | |
b18215fc RS |
3647 | case 'C': |
3648 | laststart = b; | |
36595814 | 3649 | PATFETCH (c); |
b18215fc RS |
3650 | BUF_PUSH_2 (notcategoryspec, c); |
3651 | break; | |
3652 | #endif /* emacs */ | |
e318085a | 3653 | |
e318085a | 3654 | |
25fe55af | 3655 | case 'w': |
4bb91c68 SM |
3656 | if (syntax & RE_NO_GNU_OPS) |
3657 | goto normal_char; | |
25fe55af | 3658 | laststart = b; |
1fb352e0 | 3659 | BUF_PUSH_2 (syntaxspec, Sword); |
25fe55af | 3660 | break; |
e318085a | 3661 | |
e318085a | 3662 | |
25fe55af | 3663 | case 'W': |
4bb91c68 SM |
3664 | if (syntax & RE_NO_GNU_OPS) |
3665 | goto normal_char; | |
25fe55af | 3666 | laststart = b; |
1fb352e0 | 3667 | BUF_PUSH_2 (notsyntaxspec, Sword); |
25fe55af | 3668 | break; |
e318085a RS |
3669 | |
3670 | ||
25fe55af | 3671 | case '<': |
4bb91c68 SM |
3672 | if (syntax & RE_NO_GNU_OPS) |
3673 | goto normal_char; | |
25fe55af RS |
3674 | BUF_PUSH (wordbeg); |
3675 | break; | |
e318085a | 3676 | |
25fe55af | 3677 | case '>': |
4bb91c68 SM |
3678 | if (syntax & RE_NO_GNU_OPS) |
3679 | goto normal_char; | |
25fe55af RS |
3680 | BUF_PUSH (wordend); |
3681 | break; | |
e318085a | 3682 | |
669fa600 SM |
3683 | case '_': |
3684 | if (syntax & RE_NO_GNU_OPS) | |
3685 | goto normal_char; | |
3686 | laststart = b; | |
3687 | PATFETCH (c); | |
3688 | if (c == '<') | |
3689 | BUF_PUSH (symbeg); | |
3690 | else if (c == '>') | |
3691 | BUF_PUSH (symend); | |
3692 | else | |
3693 | FREE_STACK_RETURN (REG_BADPAT); | |
3694 | break; | |
3695 | ||
25fe55af | 3696 | case 'b': |
4bb91c68 SM |
3697 | if (syntax & RE_NO_GNU_OPS) |
3698 | goto normal_char; | |
25fe55af RS |
3699 | BUF_PUSH (wordbound); |
3700 | break; | |
e318085a | 3701 | |
25fe55af | 3702 | case 'B': |
4bb91c68 SM |
3703 | if (syntax & RE_NO_GNU_OPS) |
3704 | goto normal_char; | |
25fe55af RS |
3705 | BUF_PUSH (notwordbound); |
3706 | break; | |
fa9a63c5 | 3707 | |
25fe55af | 3708 | case '`': |
4bb91c68 SM |
3709 | if (syntax & RE_NO_GNU_OPS) |
3710 | goto normal_char; | |
25fe55af RS |
3711 | BUF_PUSH (begbuf); |
3712 | break; | |
e318085a | 3713 | |
25fe55af | 3714 | case '\'': |
4bb91c68 SM |
3715 | if (syntax & RE_NO_GNU_OPS) |
3716 | goto normal_char; | |
25fe55af RS |
3717 | BUF_PUSH (endbuf); |
3718 | break; | |
e318085a | 3719 | |
25fe55af RS |
3720 | case '1': case '2': case '3': case '4': case '5': |
3721 | case '6': case '7': case '8': case '9': | |
0cdd06f8 SM |
3722 | { |
3723 | regnum_t reg; | |
e318085a | 3724 | |
0cdd06f8 SM |
3725 | if (syntax & RE_NO_BK_REFS) |
3726 | goto normal_backslash; | |
e318085a | 3727 | |
0cdd06f8 | 3728 | reg = c - '0'; |
e318085a | 3729 | |
c69b0314 SM |
3730 | if (reg > bufp->re_nsub || reg < 1 |
3731 | /* Can't back reference to a subexp before its end. */ | |
3732 | || group_in_compile_stack (compile_stack, reg)) | |
0cdd06f8 | 3733 | FREE_STACK_RETURN (REG_ESUBREG); |
e318085a | 3734 | |
0cdd06f8 SM |
3735 | laststart = b; |
3736 | BUF_PUSH_2 (duplicate, reg); | |
3737 | } | |
25fe55af | 3738 | break; |
e318085a | 3739 | |
e318085a | 3740 | |
25fe55af RS |
3741 | case '+': |
3742 | case '?': | |
3743 | if (syntax & RE_BK_PLUS_QM) | |
3744 | goto handle_plus; | |
3745 | else | |
3746 | goto normal_backslash; | |
3747 | ||
3748 | default: | |
3749 | normal_backslash: | |
3750 | /* You might think it would be useful for \ to mean | |
3751 | not to translate; but if we don't translate it | |
4bb91c68 | 3752 | it will never match anything. */ |
25fe55af RS |
3753 | goto normal_char; |
3754 | } | |
3755 | break; | |
fa9a63c5 RM |
3756 | |
3757 | ||
3758 | default: | |
25fe55af | 3759 | /* Expects the character in `c'. */ |
fa9a63c5 | 3760 | normal_char: |
36595814 | 3761 | /* If no exactn currently being built. */ |
25fe55af | 3762 | if (!pending_exact |
fa9a63c5 | 3763 | |
25fe55af RS |
3764 | /* If last exactn not at current position. */ |
3765 | || pending_exact + *pending_exact + 1 != b | |
5e69f11e | 3766 | |
25fe55af | 3767 | /* We have only one byte following the exactn for the count. */ |
2d1675e4 | 3768 | || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH |
fa9a63c5 | 3769 | |
7814e705 | 3770 | /* If followed by a repetition operator. */ |
9d99031f | 3771 | || (p != pend && (*p == '*' || *p == '^')) |
fa9a63c5 | 3772 | || ((syntax & RE_BK_PLUS_QM) |
9d99031f RS |
3773 | ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?') |
3774 | : p != pend && (*p == '+' || *p == '?')) | |
fa9a63c5 | 3775 | || ((syntax & RE_INTERVALS) |
25fe55af | 3776 | && ((syntax & RE_NO_BK_BRACES) |
9d99031f RS |
3777 | ? p != pend && *p == '{' |
3778 | : p + 1 < pend && p[0] == '\\' && p[1] == '{'))) | |
fa9a63c5 RM |
3779 | { |
3780 | /* Start building a new exactn. */ | |
5e69f11e | 3781 | |
25fe55af | 3782 | laststart = b; |
fa9a63c5 RM |
3783 | |
3784 | BUF_PUSH_2 (exactn, 0); | |
3785 | pending_exact = b - 1; | |
25fe55af | 3786 | } |
5e69f11e | 3787 | |
2d1675e4 SM |
3788 | GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH); |
3789 | { | |
e0277a47 KH |
3790 | int len; |
3791 | ||
cf9c99bc | 3792 | if (multibyte) |
6fdd04b0 | 3793 | { |
cf9c99bc | 3794 | c = TRANSLATE (c); |
6fdd04b0 KH |
3795 | len = CHAR_STRING (c, b); |
3796 | b += len; | |
3797 | } | |
e0277a47 | 3798 | else |
6fdd04b0 | 3799 | { |
cf9c99bc KH |
3800 | c1 = RE_CHAR_TO_MULTIBYTE (c); |
3801 | if (! CHAR_BYTE8_P (c1)) | |
3802 | { | |
3803 | re_wchar_t c2 = TRANSLATE (c1); | |
3804 | ||
3805 | if (c1 != c2 && (c1 = RE_CHAR_TO_UNIBYTE (c2)) >= 0) | |
3806 | c = c1; | |
409f2919 | 3807 | } |
6fdd04b0 KH |
3808 | *b++ = c; |
3809 | len = 1; | |
3810 | } | |
2d1675e4 SM |
3811 | (*pending_exact) += len; |
3812 | } | |
3813 | ||
fa9a63c5 | 3814 | break; |
25fe55af | 3815 | } /* switch (c) */ |
fa9a63c5 RM |
3816 | } /* while p != pend */ |
3817 | ||
5e69f11e | 3818 | |
fa9a63c5 | 3819 | /* Through the pattern now. */ |
5e69f11e | 3820 | |
505bde11 | 3821 | FIXUP_ALT_JUMP (); |
fa9a63c5 | 3822 | |
5e69f11e | 3823 | if (!COMPILE_STACK_EMPTY) |
fa9a63c5 RM |
3824 | FREE_STACK_RETURN (REG_EPAREN); |
3825 | ||
3826 | /* If we don't want backtracking, force success | |
3827 | the first time we reach the end of the compiled pattern. */ | |
3828 | if (syntax & RE_NO_POSIX_BACKTRACKING) | |
3829 | BUF_PUSH (succeed); | |
3830 | ||
fa9a63c5 RM |
3831 | /* We have succeeded; set the length of the buffer. */ |
3832 | bufp->used = b - bufp->buffer; | |
3833 | ||
3834 | #ifdef DEBUG | |
99633e97 | 3835 | if (debug > 0) |
fa9a63c5 | 3836 | { |
505bde11 | 3837 | re_compile_fastmap (bufp); |
fa9a63c5 RM |
3838 | DEBUG_PRINT1 ("\nCompiled pattern: \n"); |
3839 | print_compiled_pattern (bufp); | |
3840 | } | |
99633e97 | 3841 | debug--; |
fa9a63c5 RM |
3842 | #endif /* DEBUG */ |
3843 | ||
3844 | #ifndef MATCH_MAY_ALLOCATE | |
3845 | /* Initialize the failure stack to the largest possible stack. This | |
3846 | isn't necessary unless we're trying to avoid calling alloca in | |
3847 | the search and match routines. */ | |
3848 | { | |
3849 | int num_regs = bufp->re_nsub + 1; | |
3850 | ||
320a2a73 | 3851 | if (fail_stack.size < re_max_failures * TYPICAL_FAILURE_SIZE) |
fa9a63c5 | 3852 | { |
a26f4ccd | 3853 | fail_stack.size = re_max_failures * TYPICAL_FAILURE_SIZE; |
fa9a63c5 | 3854 | |
fa9a63c5 RM |
3855 | if (! fail_stack.stack) |
3856 | fail_stack.stack | |
5e69f11e | 3857 | = (fail_stack_elt_t *) malloc (fail_stack.size |
fa9a63c5 RM |
3858 | * sizeof (fail_stack_elt_t)); |
3859 | else | |
3860 | fail_stack.stack | |
3861 | = (fail_stack_elt_t *) realloc (fail_stack.stack, | |
3862 | (fail_stack.size | |
3863 | * sizeof (fail_stack_elt_t))); | |
fa9a63c5 RM |
3864 | } |
3865 | ||
3866 | regex_grow_registers (num_regs); | |
3867 | } | |
3868 | #endif /* not MATCH_MAY_ALLOCATE */ | |
3869 | ||
839966f3 | 3870 | FREE_STACK_RETURN (REG_NOERROR); |
fa9a63c5 RM |
3871 | } /* regex_compile */ |
3872 | \f | |
3873 | /* Subroutines for `regex_compile'. */ | |
3874 | ||
7814e705 | 3875 | /* Store OP at LOC followed by two-byte integer parameter ARG. */ |
fa9a63c5 RM |
3876 | |
3877 | static void | |
3878 | store_op1 (op, loc, arg) | |
3879 | re_opcode_t op; | |
3880 | unsigned char *loc; | |
3881 | int arg; | |
3882 | { | |
3883 | *loc = (unsigned char) op; | |
3884 | STORE_NUMBER (loc + 1, arg); | |
3885 | } | |
3886 | ||
3887 | ||
3888 | /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3889 | ||
3890 | static void | |
3891 | store_op2 (op, loc, arg1, arg2) | |
3892 | re_opcode_t op; | |
3893 | unsigned char *loc; | |
3894 | int arg1, arg2; | |
3895 | { | |
3896 | *loc = (unsigned char) op; | |
3897 | STORE_NUMBER (loc + 1, arg1); | |
3898 | STORE_NUMBER (loc + 3, arg2); | |
3899 | } | |
3900 | ||
3901 | ||
3902 | /* Copy the bytes from LOC to END to open up three bytes of space at LOC | |
3903 | for OP followed by two-byte integer parameter ARG. */ | |
3904 | ||
3905 | static void | |
3906 | insert_op1 (op, loc, arg, end) | |
3907 | re_opcode_t op; | |
3908 | unsigned char *loc; | |
3909 | int arg; | |
5e69f11e | 3910 | unsigned char *end; |
fa9a63c5 RM |
3911 | { |
3912 | register unsigned char *pfrom = end; | |
3913 | register unsigned char *pto = end + 3; | |
3914 | ||
3915 | while (pfrom != loc) | |
3916 | *--pto = *--pfrom; | |
5e69f11e | 3917 | |
fa9a63c5 RM |
3918 | store_op1 (op, loc, arg); |
3919 | } | |
3920 | ||
3921 | ||
3922 | /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3923 | ||
3924 | static void | |
3925 | insert_op2 (op, loc, arg1, arg2, end) | |
3926 | re_opcode_t op; | |
3927 | unsigned char *loc; | |
3928 | int arg1, arg2; | |
5e69f11e | 3929 | unsigned char *end; |
fa9a63c5 RM |
3930 | { |
3931 | register unsigned char *pfrom = end; | |
3932 | register unsigned char *pto = end + 5; | |
3933 | ||
3934 | while (pfrom != loc) | |
3935 | *--pto = *--pfrom; | |
5e69f11e | 3936 | |
fa9a63c5 RM |
3937 | store_op2 (op, loc, arg1, arg2); |
3938 | } | |
3939 | ||
3940 | ||
3941 | /* P points to just after a ^ in PATTERN. Return true if that ^ comes | |
3942 | after an alternative or a begin-subexpression. We assume there is at | |
3943 | least one character before the ^. */ | |
3944 | ||
3945 | static boolean | |
3946 | at_begline_loc_p (pattern, p, syntax) | |
01618498 | 3947 | re_char *pattern, *p; |
fa9a63c5 RM |
3948 | reg_syntax_t syntax; |
3949 | { | |
01618498 | 3950 | re_char *prev = p - 2; |
fa9a63c5 | 3951 | boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\'; |
5e69f11e | 3952 | |
fa9a63c5 RM |
3953 | return |
3954 | /* After a subexpression? */ | |
3955 | (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash)) | |
25fe55af | 3956 | /* After an alternative? */ |
d2af47df SM |
3957 | || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash)) |
3958 | /* After a shy subexpression? */ | |
3959 | || ((syntax & RE_SHY_GROUPS) && prev - 2 >= pattern | |
3960 | && prev[-1] == '?' && prev[-2] == '(' | |
3961 | && (syntax & RE_NO_BK_PARENS | |
3962 | || (prev - 3 >= pattern && prev[-3] == '\\'))); | |
fa9a63c5 RM |
3963 | } |
3964 | ||
3965 | ||
3966 | /* The dual of at_begline_loc_p. This one is for $. We assume there is | |
3967 | at least one character after the $, i.e., `P < PEND'. */ | |
3968 | ||
3969 | static boolean | |
3970 | at_endline_loc_p (p, pend, syntax) | |
01618498 | 3971 | re_char *p, *pend; |
99633e97 | 3972 | reg_syntax_t syntax; |
fa9a63c5 | 3973 | { |
01618498 | 3974 | re_char *next = p; |
fa9a63c5 | 3975 | boolean next_backslash = *next == '\\'; |
01618498 | 3976 | re_char *next_next = p + 1 < pend ? p + 1 : 0; |
5e69f11e | 3977 | |
fa9a63c5 RM |
3978 | return |
3979 | /* Before a subexpression? */ | |
3980 | (syntax & RE_NO_BK_PARENS ? *next == ')' | |
25fe55af | 3981 | : next_backslash && next_next && *next_next == ')') |
fa9a63c5 RM |
3982 | /* Before an alternative? */ |
3983 | || (syntax & RE_NO_BK_VBAR ? *next == '|' | |
25fe55af | 3984 | : next_backslash && next_next && *next_next == '|'); |
fa9a63c5 RM |
3985 | } |
3986 | ||
3987 | ||
5e69f11e | 3988 | /* Returns true if REGNUM is in one of COMPILE_STACK's elements and |
fa9a63c5 RM |
3989 | false if it's not. */ |
3990 | ||
3991 | static boolean | |
3992 | group_in_compile_stack (compile_stack, regnum) | |
3993 | compile_stack_type compile_stack; | |
3994 | regnum_t regnum; | |
3995 | { | |
3996 | int this_element; | |
3997 | ||
5e69f11e RM |
3998 | for (this_element = compile_stack.avail - 1; |
3999 | this_element >= 0; | |
fa9a63c5 RM |
4000 | this_element--) |
4001 | if (compile_stack.stack[this_element].regnum == regnum) | |
4002 | return true; | |
4003 | ||
4004 | return false; | |
4005 | } | |
fa9a63c5 | 4006 | \f |
f6a3f532 SM |
4007 | /* analyse_first. |
4008 | If fastmap is non-NULL, go through the pattern and fill fastmap | |
4009 | with all the possible leading chars. If fastmap is NULL, don't | |
4010 | bother filling it up (obviously) and only return whether the | |
4011 | pattern could potentially match the empty string. | |
4012 | ||
4013 | Return 1 if p..pend might match the empty string. | |
4014 | Return 0 if p..pend matches at least one char. | |
01618498 | 4015 | Return -1 if fastmap was not updated accurately. */ |
f6a3f532 SM |
4016 | |
4017 | static int | |
4018 | analyse_first (p, pend, fastmap, multibyte) | |
01618498 | 4019 | re_char *p, *pend; |
f6a3f532 SM |
4020 | char *fastmap; |
4021 | const int multibyte; | |
fa9a63c5 | 4022 | { |
505bde11 | 4023 | int j, k; |
1fb352e0 | 4024 | boolean not; |
fa9a63c5 | 4025 | |
b18215fc | 4026 | /* If all elements for base leading-codes in fastmap is set, this |
7814e705 | 4027 | flag is set true. */ |
b18215fc RS |
4028 | boolean match_any_multibyte_characters = false; |
4029 | ||
f6a3f532 | 4030 | assert (p); |
5e69f11e | 4031 | |
505bde11 SM |
4032 | /* The loop below works as follows: |
4033 | - It has a working-list kept in the PATTERN_STACK and which basically | |
4034 | starts by only containing a pointer to the first operation. | |
4035 | - If the opcode we're looking at is a match against some set of | |
4036 | chars, then we add those chars to the fastmap and go on to the | |
4037 | next work element from the worklist (done via `break'). | |
4038 | - If the opcode is a control operator on the other hand, we either | |
4039 | ignore it (if it's meaningless at this point, such as `start_memory') | |
4040 | or execute it (if it's a jump). If the jump has several destinations | |
4041 | (i.e. `on_failure_jump'), then we push the other destination onto the | |
4042 | worklist. | |
4043 | We guarantee termination by ignoring backward jumps (more or less), | |
4044 | so that `p' is monotonically increasing. More to the point, we | |
4045 | never set `p' (or push) anything `<= p1'. */ | |
4046 | ||
01618498 | 4047 | while (p < pend) |
fa9a63c5 | 4048 | { |
505bde11 SM |
4049 | /* `p1' is used as a marker of how far back a `on_failure_jump' |
4050 | can go without being ignored. It is normally equal to `p' | |
4051 | (which prevents any backward `on_failure_jump') except right | |
4052 | after a plain `jump', to allow patterns such as: | |
4053 | 0: jump 10 | |
4054 | 3..9: <body> | |
4055 | 10: on_failure_jump 3 | |
4056 | as used for the *? operator. */ | |
01618498 | 4057 | re_char *p1 = p; |
5e69f11e | 4058 | |
fa9a63c5 RM |
4059 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++)) |
4060 | { | |
f6a3f532 | 4061 | case succeed: |
01618498 | 4062 | return 1; |
f6a3f532 | 4063 | continue; |
fa9a63c5 | 4064 | |
fa9a63c5 | 4065 | case duplicate: |
505bde11 SM |
4066 | /* If the first character has to match a backreference, that means |
4067 | that the group was empty (since it already matched). Since this | |
4068 | is the only case that interests us here, we can assume that the | |
4069 | backreference must match the empty string. */ | |
4070 | p++; | |
4071 | continue; | |
fa9a63c5 RM |
4072 | |
4073 | ||
4074 | /* Following are the cases which match a character. These end | |
7814e705 | 4075 | with `break'. */ |
fa9a63c5 RM |
4076 | |
4077 | case exactn: | |
e0277a47 | 4078 | if (fastmap) |
cf9c99bc KH |
4079 | { |
4080 | /* If multibyte is nonzero, the first byte of each | |
4081 | character is an ASCII or a leading code. Otherwise, | |
4082 | each byte is a character. Thus, this works in both | |
4083 | cases. */ | |
4084 | fastmap[p[1]] = 1; | |
4085 | if (! multibyte) | |
4086 | { | |
4087 | /* For the case of matching this unibyte regex | |
4088 | against multibyte, we must set a leading code of | |
4089 | the corresponding multibyte character. */ | |
4090 | int c = RE_CHAR_TO_MULTIBYTE (p[1]); | |
4091 | ||
86e893e3 | 4092 | fastmap[CHAR_LEADING_CODE (c)] = 1; |
cf9c99bc KH |
4093 | } |
4094 | } | |
fa9a63c5 RM |
4095 | break; |
4096 | ||
4097 | ||
1fb352e0 SM |
4098 | case anychar: |
4099 | /* We could put all the chars except for \n (and maybe \0) | |
4100 | but we don't bother since it is generally not worth it. */ | |
f6a3f532 | 4101 | if (!fastmap) break; |
01618498 | 4102 | return -1; |
fa9a63c5 RM |
4103 | |
4104 | ||
b18215fc | 4105 | case charset_not: |
1fb352e0 | 4106 | if (!fastmap) break; |
bf216479 KH |
4107 | { |
4108 | /* Chars beyond end of bitmap are possible matches. */ | |
bf216479 | 4109 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH; |
cf9c99bc | 4110 | j < (1 << BYTEWIDTH); j++) |
bf216479 KH |
4111 | fastmap[j] = 1; |
4112 | } | |
4113 | ||
1fb352e0 SM |
4114 | /* Fallthrough */ |
4115 | case charset: | |
4116 | if (!fastmap) break; | |
4117 | not = (re_opcode_t) *(p - 1) == charset_not; | |
b18215fc RS |
4118 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH - 1, p++; |
4119 | j >= 0; j--) | |
1fb352e0 | 4120 | if (!!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) ^ not) |
49da453b | 4121 | fastmap[j] = 1; |
b18215fc | 4122 | |
6482db2e KH |
4123 | #ifdef emacs |
4124 | if (/* Any leading code can possibly start a character | |
1fb352e0 | 4125 | which doesn't match the specified set of characters. */ |
6482db2e | 4126 | not |
409f2919 | 4127 | || |
6482db2e KH |
4128 | /* If we can match a character class, we can match any |
4129 | multibyte characters. */ | |
4130 | (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) | |
4131 | && CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0)) | |
4132 | ||
b18215fc | 4133 | { |
b18215fc RS |
4134 | if (match_any_multibyte_characters == false) |
4135 | { | |
6482db2e KH |
4136 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
4137 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
6fdd04b0 | 4138 | fastmap[j] = 1; |
b18215fc RS |
4139 | match_any_multibyte_characters = true; |
4140 | } | |
4141 | } | |
b18215fc | 4142 | |
1fb352e0 SM |
4143 | else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) |
4144 | && match_any_multibyte_characters == false) | |
4145 | { | |
bf216479 | 4146 | /* Set fastmap[I] to 1 where I is a leading code of each |
9117d724 | 4147 | multibyte characer in the range table. */ |
1fb352e0 | 4148 | int c, count; |
bf216479 | 4149 | unsigned char lc1, lc2; |
b18215fc | 4150 | |
1fb352e0 | 4151 | /* Make P points the range table. `+ 2' is to skip flag |
0b32bf0e | 4152 | bits for a character class. */ |
1fb352e0 | 4153 | p += CHARSET_BITMAP_SIZE (&p[-2]) + 2; |
b18215fc | 4154 | |
1fb352e0 SM |
4155 | /* Extract the number of ranges in range table into COUNT. */ |
4156 | EXTRACT_NUMBER_AND_INCR (count, p); | |
cf9c99bc | 4157 | for (; count > 0; count--, p += 3) |
1fb352e0 | 4158 | { |
9117d724 KH |
4159 | /* Extract the start and end of each range. */ |
4160 | EXTRACT_CHARACTER (c, p); | |
bf216479 | 4161 | lc1 = CHAR_LEADING_CODE (c); |
9117d724 | 4162 | p += 3; |
1fb352e0 | 4163 | EXTRACT_CHARACTER (c, p); |
bf216479 KH |
4164 | lc2 = CHAR_LEADING_CODE (c); |
4165 | for (j = lc1; j <= lc2; j++) | |
9117d724 | 4166 | fastmap[j] = 1; |
1fb352e0 SM |
4167 | } |
4168 | } | |
6482db2e | 4169 | #endif |
b18215fc RS |
4170 | break; |
4171 | ||
1fb352e0 SM |
4172 | case syntaxspec: |
4173 | case notsyntaxspec: | |
4174 | if (!fastmap) break; | |
4175 | #ifndef emacs | |
4176 | not = (re_opcode_t)p[-1] == notsyntaxspec; | |
4177 | k = *p++; | |
4178 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
990b2375 | 4179 | if ((SYNTAX (j) == (enum syntaxcode) k) ^ not) |
b18215fc | 4180 | fastmap[j] = 1; |
b18215fc | 4181 | break; |
1fb352e0 | 4182 | #else /* emacs */ |
b18215fc RS |
4183 | /* This match depends on text properties. These end with |
4184 | aborting optimizations. */ | |
01618498 | 4185 | return -1; |
b18215fc RS |
4186 | |
4187 | case categoryspec: | |
b18215fc | 4188 | case notcategoryspec: |
1fb352e0 SM |
4189 | if (!fastmap) break; |
4190 | not = (re_opcode_t)p[-1] == notcategoryspec; | |
b18215fc | 4191 | k = *p++; |
6482db2e | 4192 | for (j = (1 << BYTEWIDTH); j >= 0; j--) |
1fb352e0 | 4193 | if ((CHAR_HAS_CATEGORY (j, k)) ^ not) |
b18215fc RS |
4194 | fastmap[j] = 1; |
4195 | ||
6482db2e KH |
4196 | /* Any leading code can possibly start a character which |
4197 | has or doesn't has the specified category. */ | |
4198 | if (match_any_multibyte_characters == false) | |
6fdd04b0 | 4199 | { |
6482db2e KH |
4200 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
4201 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
4202 | fastmap[j] = 1; | |
4203 | match_any_multibyte_characters = true; | |
6fdd04b0 | 4204 | } |
b18215fc RS |
4205 | break; |
4206 | ||
fa9a63c5 | 4207 | /* All cases after this match the empty string. These end with |
25fe55af | 4208 | `continue'. */ |
fa9a63c5 | 4209 | |
fa9a63c5 RM |
4210 | case before_dot: |
4211 | case at_dot: | |
4212 | case after_dot: | |
1fb352e0 | 4213 | #endif /* !emacs */ |
25fe55af RS |
4214 | case no_op: |
4215 | case begline: | |
4216 | case endline: | |
fa9a63c5 RM |
4217 | case begbuf: |
4218 | case endbuf: | |
4219 | case wordbound: | |
4220 | case notwordbound: | |
4221 | case wordbeg: | |
4222 | case wordend: | |
669fa600 SM |
4223 | case symbeg: |
4224 | case symend: | |
25fe55af | 4225 | continue; |
fa9a63c5 RM |
4226 | |
4227 | ||
fa9a63c5 | 4228 | case jump: |
25fe55af | 4229 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 SM |
4230 | if (j < 0) |
4231 | /* Backward jumps can only go back to code that we've already | |
4232 | visited. `re_compile' should make sure this is true. */ | |
4233 | break; | |
25fe55af | 4234 | p += j; |
505bde11 SM |
4235 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p)) |
4236 | { | |
4237 | case on_failure_jump: | |
4238 | case on_failure_keep_string_jump: | |
505bde11 | 4239 | case on_failure_jump_loop: |
0683b6fa | 4240 | case on_failure_jump_nastyloop: |
505bde11 SM |
4241 | case on_failure_jump_smart: |
4242 | p++; | |
4243 | break; | |
4244 | default: | |
4245 | continue; | |
4246 | }; | |
4247 | /* Keep `p1' to allow the `on_failure_jump' we are jumping to | |
4248 | to jump back to "just after here". */ | |
4249 | /* Fallthrough */ | |
fa9a63c5 | 4250 | |
25fe55af RS |
4251 | case on_failure_jump: |
4252 | case on_failure_keep_string_jump: | |
0683b6fa | 4253 | case on_failure_jump_nastyloop: |
505bde11 SM |
4254 | case on_failure_jump_loop: |
4255 | case on_failure_jump_smart: | |
25fe55af | 4256 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 | 4257 | if (p + j <= p1) |
ed0767d8 | 4258 | ; /* Backward jump to be ignored. */ |
01618498 SM |
4259 | else |
4260 | { /* We have to look down both arms. | |
4261 | We first go down the "straight" path so as to minimize | |
4262 | stack usage when going through alternatives. */ | |
4263 | int r = analyse_first (p, pend, fastmap, multibyte); | |
4264 | if (r) return r; | |
4265 | p += j; | |
4266 | } | |
25fe55af | 4267 | continue; |
fa9a63c5 RM |
4268 | |
4269 | ||
ed0767d8 SM |
4270 | case jump_n: |
4271 | /* This code simply does not properly handle forward jump_n. */ | |
4272 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p); assert (j < 0)); | |
4273 | p += 4; | |
4274 | /* jump_n can either jump or fall through. The (backward) jump | |
4275 | case has already been handled, so we only need to look at the | |
4276 | fallthrough case. */ | |
4277 | continue; | |
177c0ea7 | 4278 | |
fa9a63c5 | 4279 | case succeed_n: |
ed0767d8 SM |
4280 | /* If N == 0, it should be an on_failure_jump_loop instead. */ |
4281 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0)); | |
4282 | p += 4; | |
4283 | /* We only care about one iteration of the loop, so we don't | |
4284 | need to consider the case where this behaves like an | |
4285 | on_failure_jump. */ | |
25fe55af | 4286 | continue; |
fa9a63c5 RM |
4287 | |
4288 | ||
4289 | case set_number_at: | |
25fe55af RS |
4290 | p += 4; |
4291 | continue; | |
fa9a63c5 RM |
4292 | |
4293 | ||
4294 | case start_memory: | |
25fe55af | 4295 | case stop_memory: |
505bde11 | 4296 | p += 1; |
fa9a63c5 RM |
4297 | continue; |
4298 | ||
4299 | ||
4300 | default: | |
25fe55af RS |
4301 | abort (); /* We have listed all the cases. */ |
4302 | } /* switch *p++ */ | |
fa9a63c5 RM |
4303 | |
4304 | /* Getting here means we have found the possible starting | |
25fe55af | 4305 | characters for one path of the pattern -- and that the empty |
7814e705 | 4306 | string does not match. We need not follow this path further. */ |
01618498 | 4307 | return 0; |
fa9a63c5 RM |
4308 | } /* while p */ |
4309 | ||
01618498 SM |
4310 | /* We reached the end without matching anything. */ |
4311 | return 1; | |
4312 | ||
f6a3f532 SM |
4313 | } /* analyse_first */ |
4314 | \f | |
4315 | /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in | |
4316 | BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible | |
4317 | characters can start a string that matches the pattern. This fastmap | |
4318 | is used by re_search to skip quickly over impossible starting points. | |
4319 | ||
4320 | Character codes above (1 << BYTEWIDTH) are not represented in the | |
4321 | fastmap, but the leading codes are represented. Thus, the fastmap | |
4322 | indicates which character sets could start a match. | |
4323 | ||
4324 | The caller must supply the address of a (1 << BYTEWIDTH)-byte data | |
4325 | area as BUFP->fastmap. | |
4326 | ||
4327 | We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in | |
4328 | the pattern buffer. | |
4329 | ||
4330 | Returns 0 if we succeed, -2 if an internal error. */ | |
4331 | ||
4332 | int | |
4333 | re_compile_fastmap (bufp) | |
4334 | struct re_pattern_buffer *bufp; | |
4335 | { | |
4336 | char *fastmap = bufp->fastmap; | |
4337 | int analysis; | |
4338 | ||
4339 | assert (fastmap && bufp->buffer); | |
4340 | ||
7814e705 | 4341 | bzero (fastmap, 1 << BYTEWIDTH); /* Assume nothing's valid. */ |
f6a3f532 SM |
4342 | bufp->fastmap_accurate = 1; /* It will be when we're done. */ |
4343 | ||
4344 | analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used, | |
2d1675e4 | 4345 | fastmap, RE_MULTIBYTE_P (bufp)); |
c0f9ea08 | 4346 | bufp->can_be_null = (analysis != 0); |
fa9a63c5 RM |
4347 | return 0; |
4348 | } /* re_compile_fastmap */ | |
4349 | \f | |
4350 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | |
4351 | ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | |
4352 | this memory for recording register information. STARTS and ENDS | |
4353 | must be allocated using the malloc library routine, and must each | |
4354 | be at least NUM_REGS * sizeof (regoff_t) bytes long. | |
4355 | ||
4356 | If NUM_REGS == 0, then subsequent matches should allocate their own | |
4357 | register data. | |
4358 | ||
4359 | Unless this function is called, the first search or match using | |
4360 | PATTERN_BUFFER will allocate its own register data, without | |
4361 | freeing the old data. */ | |
4362 | ||
4363 | void | |
4364 | re_set_registers (bufp, regs, num_regs, starts, ends) | |
4365 | struct re_pattern_buffer *bufp; | |
4366 | struct re_registers *regs; | |
4367 | unsigned num_regs; | |
4368 | regoff_t *starts, *ends; | |
4369 | { | |
4370 | if (num_regs) | |
4371 | { | |
4372 | bufp->regs_allocated = REGS_REALLOCATE; | |
4373 | regs->num_regs = num_regs; | |
4374 | regs->start = starts; | |
4375 | regs->end = ends; | |
4376 | } | |
4377 | else | |
4378 | { | |
4379 | bufp->regs_allocated = REGS_UNALLOCATED; | |
4380 | regs->num_regs = 0; | |
4381 | regs->start = regs->end = (regoff_t *) 0; | |
4382 | } | |
4383 | } | |
c0f9ea08 | 4384 | WEAK_ALIAS (__re_set_registers, re_set_registers) |
fa9a63c5 | 4385 | \f |
7814e705 | 4386 | /* Searching routines. */ |
fa9a63c5 RM |
4387 | |
4388 | /* Like re_search_2, below, but only one string is specified, and | |
4389 | doesn't let you say where to stop matching. */ | |
4390 | ||
4391 | int | |
4392 | re_search (bufp, string, size, startpos, range, regs) | |
4393 | struct re_pattern_buffer *bufp; | |
4394 | const char *string; | |
4395 | int size, startpos, range; | |
4396 | struct re_registers *regs; | |
4397 | { | |
5e69f11e | 4398 | return re_search_2 (bufp, NULL, 0, string, size, startpos, range, |
fa9a63c5 RM |
4399 | regs, size); |
4400 | } | |
c0f9ea08 | 4401 | WEAK_ALIAS (__re_search, re_search) |
fa9a63c5 | 4402 | |
70806df6 KH |
4403 | /* Head address of virtual concatenation of string. */ |
4404 | #define HEAD_ADDR_VSTRING(P) \ | |
4405 | (((P) >= size1 ? string2 : string1)) | |
4406 | ||
b18215fc RS |
4407 | /* End address of virtual concatenation of string. */ |
4408 | #define STOP_ADDR_VSTRING(P) \ | |
4409 | (((P) >= size1 ? string2 + size2 : string1 + size1)) | |
4410 | ||
4411 | /* Address of POS in the concatenation of virtual string. */ | |
4412 | #define POS_ADDR_VSTRING(POS) \ | |
4413 | (((POS) >= size1 ? string2 - size1 : string1) + (POS)) | |
fa9a63c5 RM |
4414 | |
4415 | /* Using the compiled pattern in BUFP->buffer, first tries to match the | |
4416 | virtual concatenation of STRING1 and STRING2, starting first at index | |
4417 | STARTPOS, then at STARTPOS + 1, and so on. | |
5e69f11e | 4418 | |
fa9a63c5 | 4419 | STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. |
5e69f11e | 4420 | |
fa9a63c5 RM |
4421 | RANGE is how far to scan while trying to match. RANGE = 0 means try |
4422 | only at STARTPOS; in general, the last start tried is STARTPOS + | |
4423 | RANGE. | |
5e69f11e | 4424 | |
fa9a63c5 RM |
4425 | In REGS, return the indices of the virtual concatenation of STRING1 |
4426 | and STRING2 that matched the entire BUFP->buffer and its contained | |
4427 | subexpressions. | |
5e69f11e | 4428 | |
fa9a63c5 RM |
4429 | Do not consider matching one past the index STOP in the virtual |
4430 | concatenation of STRING1 and STRING2. | |
4431 | ||
4432 | We return either the position in the strings at which the match was | |
4433 | found, -1 if no match, or -2 if error (such as failure | |
4434 | stack overflow). */ | |
4435 | ||
4436 | int | |
66f0296e | 4437 | re_search_2 (bufp, str1, size1, str2, size2, startpos, range, regs, stop) |
fa9a63c5 | 4438 | struct re_pattern_buffer *bufp; |
66f0296e | 4439 | const char *str1, *str2; |
fa9a63c5 RM |
4440 | int size1, size2; |
4441 | int startpos; | |
4442 | int range; | |
4443 | struct re_registers *regs; | |
4444 | int stop; | |
4445 | { | |
4446 | int val; | |
66f0296e SM |
4447 | re_char *string1 = (re_char*) str1; |
4448 | re_char *string2 = (re_char*) str2; | |
fa9a63c5 | 4449 | register char *fastmap = bufp->fastmap; |
6676cb1c | 4450 | register RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 RM |
4451 | int total_size = size1 + size2; |
4452 | int endpos = startpos + range; | |
c0f9ea08 | 4453 | boolean anchored_start; |
cf9c99bc KH |
4454 | /* Nonzero if we are searching multibyte string. */ |
4455 | const boolean multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
b18215fc | 4456 | |
fa9a63c5 RM |
4457 | /* Check for out-of-range STARTPOS. */ |
4458 | if (startpos < 0 || startpos > total_size) | |
4459 | return -1; | |
5e69f11e | 4460 | |
fa9a63c5 | 4461 | /* Fix up RANGE if it might eventually take us outside |
34597fa9 | 4462 | the virtual concatenation of STRING1 and STRING2. |
5e69f11e | 4463 | Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ |
34597fa9 RS |
4464 | if (endpos < 0) |
4465 | range = 0 - startpos; | |
fa9a63c5 RM |
4466 | else if (endpos > total_size) |
4467 | range = total_size - startpos; | |
4468 | ||
4469 | /* If the search isn't to be a backwards one, don't waste time in a | |
7b140fd7 | 4470 | search for a pattern anchored at beginning of buffer. */ |
fa9a63c5 RM |
4471 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) |
4472 | { | |
4473 | if (startpos > 0) | |
4474 | return -1; | |
4475 | else | |
7b140fd7 | 4476 | range = 0; |
fa9a63c5 RM |
4477 | } |
4478 | ||
ae4788a8 RS |
4479 | #ifdef emacs |
4480 | /* In a forward search for something that starts with \=. | |
4481 | don't keep searching past point. */ | |
4482 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) | |
4483 | { | |
7b140fd7 RS |
4484 | range = PT_BYTE - BEGV_BYTE - startpos; |
4485 | if (range < 0) | |
ae4788a8 RS |
4486 | return -1; |
4487 | } | |
4488 | #endif /* emacs */ | |
4489 | ||
fa9a63c5 RM |
4490 | /* Update the fastmap now if not correct already. */ |
4491 | if (fastmap && !bufp->fastmap_accurate) | |
01618498 | 4492 | re_compile_fastmap (bufp); |
5e69f11e | 4493 | |
c8499ba5 | 4494 | /* See whether the pattern is anchored. */ |
c0f9ea08 | 4495 | anchored_start = (bufp->buffer[0] == begline); |
c8499ba5 | 4496 | |
b18215fc | 4497 | #ifdef emacs |
d48cd3f4 | 4498 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
cc9b4df2 | 4499 | { |
99633e97 | 4500 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (startpos)); |
cc9b4df2 KH |
4501 | |
4502 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); | |
4503 | } | |
b18215fc RS |
4504 | #endif |
4505 | ||
fa9a63c5 RM |
4506 | /* Loop through the string, looking for a place to start matching. */ |
4507 | for (;;) | |
5e69f11e | 4508 | { |
c8499ba5 RS |
4509 | /* If the pattern is anchored, |
4510 | skip quickly past places we cannot match. | |
4511 | We don't bother to treat startpos == 0 specially | |
4512 | because that case doesn't repeat. */ | |
4513 | if (anchored_start && startpos > 0) | |
4514 | { | |
c0f9ea08 SM |
4515 | if (! ((startpos <= size1 ? string1[startpos - 1] |
4516 | : string2[startpos - size1 - 1]) | |
4517 | == '\n')) | |
c8499ba5 RS |
4518 | goto advance; |
4519 | } | |
4520 | ||
fa9a63c5 | 4521 | /* If a fastmap is supplied, skip quickly over characters that |
25fe55af RS |
4522 | cannot be the start of a match. If the pattern can match the |
4523 | null string, however, we don't need to skip characters; we want | |
7814e705 | 4524 | the first null string. */ |
fa9a63c5 RM |
4525 | if (fastmap && startpos < total_size && !bufp->can_be_null) |
4526 | { | |
66f0296e | 4527 | register re_char *d; |
01618498 | 4528 | register re_wchar_t buf_ch; |
e934739e RS |
4529 | |
4530 | d = POS_ADDR_VSTRING (startpos); | |
4531 | ||
7814e705 | 4532 | if (range > 0) /* Searching forwards. */ |
fa9a63c5 | 4533 | { |
fa9a63c5 RM |
4534 | register int lim = 0; |
4535 | int irange = range; | |
4536 | ||
25fe55af RS |
4537 | if (startpos < size1 && startpos + range >= size1) |
4538 | lim = range - (size1 - startpos); | |
fa9a63c5 | 4539 | |
25fe55af RS |
4540 | /* Written out as an if-else to avoid testing `translate' |
4541 | inside the loop. */ | |
28ae27ae AS |
4542 | if (RE_TRANSLATE_P (translate)) |
4543 | { | |
e934739e RS |
4544 | if (multibyte) |
4545 | while (range > lim) | |
4546 | { | |
4547 | int buf_charlen; | |
4548 | ||
62a6e103 | 4549 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 4550 | buf_ch = RE_TRANSLATE (translate, buf_ch); |
bf216479 | 4551 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
e934739e RS |
4552 | break; |
4553 | ||
4554 | range -= buf_charlen; | |
4555 | d += buf_charlen; | |
4556 | } | |
4557 | else | |
bf216479 | 4558 | while (range > lim) |
33c46939 | 4559 | { |
cf9c99bc KH |
4560 | register re_wchar_t ch, translated; |
4561 | ||
bf216479 | 4562 | buf_ch = *d; |
cf9c99bc KH |
4563 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); |
4564 | translated = RE_TRANSLATE (translate, ch); | |
4565 | if (translated != ch | |
4566 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4567 | buf_ch = ch; | |
6fdd04b0 | 4568 | if (fastmap[buf_ch]) |
bf216479 | 4569 | break; |
33c46939 RS |
4570 | d++; |
4571 | range--; | |
4572 | } | |
e934739e | 4573 | } |
fa9a63c5 | 4574 | else |
6fdd04b0 KH |
4575 | { |
4576 | if (multibyte) | |
4577 | while (range > lim) | |
4578 | { | |
4579 | int buf_charlen; | |
fa9a63c5 | 4580 | |
62a6e103 | 4581 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
6fdd04b0 KH |
4582 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
4583 | break; | |
4584 | range -= buf_charlen; | |
4585 | d += buf_charlen; | |
4586 | } | |
e934739e | 4587 | else |
6fdd04b0 | 4588 | while (range > lim && !fastmap[*d]) |
33c46939 RS |
4589 | { |
4590 | d++; | |
4591 | range--; | |
4592 | } | |
e934739e | 4593 | } |
fa9a63c5 RM |
4594 | startpos += irange - range; |
4595 | } | |
7814e705 | 4596 | else /* Searching backwards. */ |
fa9a63c5 | 4597 | { |
ba5e343c KH |
4598 | if (multibyte) |
4599 | { | |
62a6e103 | 4600 | buf_ch = STRING_CHAR (d); |
ba5e343c KH |
4601 | buf_ch = TRANSLATE (buf_ch); |
4602 | if (! fastmap[CHAR_LEADING_CODE (buf_ch)]) | |
4603 | goto advance; | |
4604 | } | |
4605 | else | |
4606 | { | |
cf9c99bc KH |
4607 | register re_wchar_t ch, translated; |
4608 | ||
4609 | buf_ch = *d; | |
4610 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); | |
4611 | translated = TRANSLATE (ch); | |
4612 | if (translated != ch | |
4613 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4614 | buf_ch = ch; | |
4615 | if (! fastmap[TRANSLATE (buf_ch)]) | |
ba5e343c KH |
4616 | goto advance; |
4617 | } | |
fa9a63c5 RM |
4618 | } |
4619 | } | |
4620 | ||
4621 | /* If can't match the null string, and that's all we have left, fail. */ | |
4622 | if (range >= 0 && startpos == total_size && fastmap | |
25fe55af | 4623 | && !bufp->can_be_null) |
fa9a63c5 RM |
4624 | return -1; |
4625 | ||
4626 | val = re_match_2_internal (bufp, string1, size1, string2, size2, | |
4627 | startpos, regs, stop); | |
fa9a63c5 RM |
4628 | |
4629 | if (val >= 0) | |
4630 | return startpos; | |
5e69f11e | 4631 | |
fa9a63c5 RM |
4632 | if (val == -2) |
4633 | return -2; | |
4634 | ||
4635 | advance: | |
5e69f11e | 4636 | if (!range) |
25fe55af | 4637 | break; |
5e69f11e | 4638 | else if (range > 0) |
25fe55af | 4639 | { |
b18215fc RS |
4640 | /* Update STARTPOS to the next character boundary. */ |
4641 | if (multibyte) | |
4642 | { | |
66f0296e SM |
4643 | re_char *p = POS_ADDR_VSTRING (startpos); |
4644 | re_char *pend = STOP_ADDR_VSTRING (startpos); | |
aa3830c4 | 4645 | int len = BYTES_BY_CHAR_HEAD (*p); |
b18215fc RS |
4646 | |
4647 | range -= len; | |
4648 | if (range < 0) | |
4649 | break; | |
4650 | startpos += len; | |
4651 | } | |
4652 | else | |
4653 | { | |
b560c397 RS |
4654 | range--; |
4655 | startpos++; | |
4656 | } | |
e318085a | 4657 | } |
fa9a63c5 | 4658 | else |
25fe55af RS |
4659 | { |
4660 | range++; | |
4661 | startpos--; | |
b18215fc RS |
4662 | |
4663 | /* Update STARTPOS to the previous character boundary. */ | |
4664 | if (multibyte) | |
4665 | { | |
70806df6 KH |
4666 | re_char *p = POS_ADDR_VSTRING (startpos) + 1; |
4667 | re_char *p0 = p; | |
4668 | re_char *phead = HEAD_ADDR_VSTRING (startpos); | |
b18215fc RS |
4669 | |
4670 | /* Find the head of multibyte form. */ | |
70806df6 KH |
4671 | PREV_CHAR_BOUNDARY (p, phead); |
4672 | range += p0 - 1 - p; | |
4673 | if (range > 0) | |
4674 | break; | |
b18215fc | 4675 | |
70806df6 | 4676 | startpos -= p0 - 1 - p; |
b18215fc | 4677 | } |
25fe55af | 4678 | } |
fa9a63c5 RM |
4679 | } |
4680 | return -1; | |
4681 | } /* re_search_2 */ | |
c0f9ea08 | 4682 | WEAK_ALIAS (__re_search_2, re_search_2) |
fa9a63c5 RM |
4683 | \f |
4684 | /* Declarations and macros for re_match_2. */ | |
4685 | ||
2d1675e4 SM |
4686 | static int bcmp_translate _RE_ARGS((re_char *s1, re_char *s2, |
4687 | register int len, | |
4688 | RE_TRANSLATE_TYPE translate, | |
4689 | const int multibyte)); | |
fa9a63c5 RM |
4690 | |
4691 | /* This converts PTR, a pointer into one of the search strings `string1' | |
4692 | and `string2' into an offset from the beginning of that string. */ | |
4693 | #define POINTER_TO_OFFSET(ptr) \ | |
4694 | (FIRST_STRING_P (ptr) \ | |
4695 | ? ((regoff_t) ((ptr) - string1)) \ | |
4696 | : ((regoff_t) ((ptr) - string2 + size1))) | |
4697 | ||
fa9a63c5 | 4698 | /* Call before fetching a character with *d. This switches over to |
419d1c74 SM |
4699 | string2 if necessary. |
4700 | Check re_match_2_internal for a discussion of why end_match_2 might | |
4701 | not be within string2 (but be equal to end_match_1 instead). */ | |
fa9a63c5 | 4702 | #define PREFETCH() \ |
25fe55af | 4703 | while (d == dend) \ |
fa9a63c5 RM |
4704 | { \ |
4705 | /* End of string2 => fail. */ \ | |
25fe55af RS |
4706 | if (dend == end_match_2) \ |
4707 | goto fail; \ | |
4bb91c68 | 4708 | /* End of string1 => advance to string2. */ \ |
25fe55af | 4709 | d = string2; \ |
fa9a63c5 RM |
4710 | dend = end_match_2; \ |
4711 | } | |
4712 | ||
f1ad044f SM |
4713 | /* Call before fetching a char with *d if you already checked other limits. |
4714 | This is meant for use in lookahead operations like wordend, etc.. | |
4715 | where we might need to look at parts of the string that might be | |
4716 | outside of the LIMITs (i.e past `stop'). */ | |
4717 | #define PREFETCH_NOLIMIT() \ | |
4718 | if (d == end1) \ | |
4719 | { \ | |
4720 | d = string2; \ | |
4721 | dend = end_match_2; \ | |
4722 | } \ | |
fa9a63c5 RM |
4723 | |
4724 | /* Test if at very beginning or at very end of the virtual concatenation | |
7814e705 | 4725 | of `string1' and `string2'. If only one string, it's `string2'. */ |
fa9a63c5 | 4726 | #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) |
5e69f11e | 4727 | #define AT_STRINGS_END(d) ((d) == end2) |
fa9a63c5 RM |
4728 | |
4729 | ||
4730 | /* Test if D points to a character which is word-constituent. We have | |
4731 | two special cases to check for: if past the end of string1, look at | |
4732 | the first character in string2; and if before the beginning of | |
4733 | string2, look at the last character in string1. */ | |
4734 | #define WORDCHAR_P(d) \ | |
4735 | (SYNTAX ((d) == end1 ? *string2 \ | |
25fe55af | 4736 | : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ |
fa9a63c5 RM |
4737 | == Sword) |
4738 | ||
9121ca40 | 4739 | /* Disabled due to a compiler bug -- see comment at case wordbound */ |
b18215fc RS |
4740 | |
4741 | /* The comment at case wordbound is following one, but we don't use | |
4742 | AT_WORD_BOUNDARY anymore to support multibyte form. | |
4743 | ||
4744 | The DEC Alpha C compiler 3.x generates incorrect code for the | |
25fe55af | 4745 | test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of |
7814e705 | 4746 | AT_WORD_BOUNDARY, so this code is disabled. Expanding the |
b18215fc RS |
4747 | macro and introducing temporary variables works around the bug. */ |
4748 | ||
9121ca40 | 4749 | #if 0 |
fa9a63c5 RM |
4750 | /* Test if the character before D and the one at D differ with respect |
4751 | to being word-constituent. */ | |
4752 | #define AT_WORD_BOUNDARY(d) \ | |
4753 | (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ | |
4754 | || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) | |
9121ca40 | 4755 | #endif |
fa9a63c5 RM |
4756 | |
4757 | /* Free everything we malloc. */ | |
4758 | #ifdef MATCH_MAY_ALLOCATE | |
0b32bf0e SM |
4759 | # define FREE_VAR(var) if (var) { REGEX_FREE (var); var = NULL; } else |
4760 | # define FREE_VARIABLES() \ | |
fa9a63c5 RM |
4761 | do { \ |
4762 | REGEX_FREE_STACK (fail_stack.stack); \ | |
4763 | FREE_VAR (regstart); \ | |
4764 | FREE_VAR (regend); \ | |
fa9a63c5 RM |
4765 | FREE_VAR (best_regstart); \ |
4766 | FREE_VAR (best_regend); \ | |
fa9a63c5 RM |
4767 | } while (0) |
4768 | #else | |
0b32bf0e | 4769 | # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
4770 | #endif /* not MATCH_MAY_ALLOCATE */ |
4771 | ||
505bde11 SM |
4772 | \f |
4773 | /* Optimization routines. */ | |
4774 | ||
4e8a9132 SM |
4775 | /* If the operation is a match against one or more chars, |
4776 | return a pointer to the next operation, else return NULL. */ | |
01618498 | 4777 | static re_char * |
4e8a9132 | 4778 | skip_one_char (p) |
01618498 | 4779 | re_char *p; |
4e8a9132 SM |
4780 | { |
4781 | switch (SWITCH_ENUM_CAST (*p++)) | |
4782 | { | |
4783 | case anychar: | |
4784 | break; | |
177c0ea7 | 4785 | |
4e8a9132 SM |
4786 | case exactn: |
4787 | p += *p + 1; | |
4788 | break; | |
4789 | ||
4790 | case charset_not: | |
4791 | case charset: | |
4792 | if (CHARSET_RANGE_TABLE_EXISTS_P (p - 1)) | |
4793 | { | |
4794 | int mcnt; | |
4795 | p = CHARSET_RANGE_TABLE (p - 1); | |
4796 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4797 | p = CHARSET_RANGE_TABLE_END (p, mcnt); | |
4798 | } | |
4799 | else | |
4800 | p += 1 + CHARSET_BITMAP_SIZE (p - 1); | |
4801 | break; | |
177c0ea7 | 4802 | |
4e8a9132 SM |
4803 | case syntaxspec: |
4804 | case notsyntaxspec: | |
1fb352e0 | 4805 | #ifdef emacs |
4e8a9132 SM |
4806 | case categoryspec: |
4807 | case notcategoryspec: | |
4808 | #endif /* emacs */ | |
4809 | p++; | |
4810 | break; | |
4811 | ||
4812 | default: | |
4813 | p = NULL; | |
4814 | } | |
4815 | return p; | |
4816 | } | |
4817 | ||
4818 | ||
505bde11 | 4819 | /* Jump over non-matching operations. */ |
839966f3 | 4820 | static re_char * |
4e8a9132 | 4821 | skip_noops (p, pend) |
839966f3 | 4822 | re_char *p, *pend; |
505bde11 SM |
4823 | { |
4824 | int mcnt; | |
4825 | while (p < pend) | |
4826 | { | |
4827 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p)) | |
4828 | { | |
4829 | case start_memory: | |
505bde11 SM |
4830 | case stop_memory: |
4831 | p += 2; break; | |
4832 | case no_op: | |
4833 | p += 1; break; | |
4834 | case jump: | |
4835 | p += 1; | |
4836 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4837 | p += mcnt; | |
4838 | break; | |
4839 | default: | |
4840 | return p; | |
4841 | } | |
4842 | } | |
4843 | assert (p == pend); | |
4844 | return p; | |
4845 | } | |
4846 | ||
4847 | /* Non-zero if "p1 matches something" implies "p2 fails". */ | |
4848 | static int | |
4849 | mutually_exclusive_p (bufp, p1, p2) | |
4850 | struct re_pattern_buffer *bufp; | |
839966f3 | 4851 | re_char *p1, *p2; |
505bde11 | 4852 | { |
4e8a9132 | 4853 | re_opcode_t op2; |
2d1675e4 | 4854 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
505bde11 SM |
4855 | unsigned char *pend = bufp->buffer + bufp->used; |
4856 | ||
4e8a9132 | 4857 | assert (p1 >= bufp->buffer && p1 < pend |
505bde11 SM |
4858 | && p2 >= bufp->buffer && p2 <= pend); |
4859 | ||
4860 | /* Skip over open/close-group commands. | |
4861 | If what follows this loop is a ...+ construct, | |
4862 | look at what begins its body, since we will have to | |
4863 | match at least one of that. */ | |
4e8a9132 SM |
4864 | p2 = skip_noops (p2, pend); |
4865 | /* The same skip can be done for p1, except that this function | |
4866 | is only used in the case where p1 is a simple match operator. */ | |
4867 | /* p1 = skip_noops (p1, pend); */ | |
4868 | ||
4869 | assert (p1 >= bufp->buffer && p1 < pend | |
4870 | && p2 >= bufp->buffer && p2 <= pend); | |
4871 | ||
4872 | op2 = p2 == pend ? succeed : *p2; | |
4873 | ||
4874 | switch (SWITCH_ENUM_CAST (op2)) | |
505bde11 | 4875 | { |
4e8a9132 SM |
4876 | case succeed: |
4877 | case endbuf: | |
4878 | /* If we're at the end of the pattern, we can change. */ | |
4879 | if (skip_one_char (p1)) | |
505bde11 | 4880 | { |
505bde11 SM |
4881 | DEBUG_PRINT1 (" End of pattern: fast loop.\n"); |
4882 | return 1; | |
505bde11 | 4883 | } |
4e8a9132 | 4884 | break; |
177c0ea7 | 4885 | |
4e8a9132 | 4886 | case endline: |
4e8a9132 SM |
4887 | case exactn: |
4888 | { | |
01618498 | 4889 | register re_wchar_t c |
4e8a9132 | 4890 | = (re_opcode_t) *p2 == endline ? '\n' |
62a6e103 | 4891 | : RE_STRING_CHAR (p2 + 2, multibyte); |
505bde11 | 4892 | |
4e8a9132 SM |
4893 | if ((re_opcode_t) *p1 == exactn) |
4894 | { | |
62a6e103 | 4895 | if (c != RE_STRING_CHAR (p1 + 2, multibyte)) |
4e8a9132 SM |
4896 | { |
4897 | DEBUG_PRINT3 (" '%c' != '%c' => fast loop.\n", c, p1[2]); | |
4898 | return 1; | |
4899 | } | |
4900 | } | |
505bde11 | 4901 | |
4e8a9132 SM |
4902 | else if ((re_opcode_t) *p1 == charset |
4903 | || (re_opcode_t) *p1 == charset_not) | |
4904 | { | |
4905 | int not = (re_opcode_t) *p1 == charset_not; | |
505bde11 | 4906 | |
4e8a9132 SM |
4907 | /* Test if C is listed in charset (or charset_not) |
4908 | at `p1'. */ | |
6fdd04b0 | 4909 | if (! multibyte || IS_REAL_ASCII (c)) |
4e8a9132 SM |
4910 | { |
4911 | if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH | |
4912 | && p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
4913 | not = !not; | |
4914 | } | |
4915 | else if (CHARSET_RANGE_TABLE_EXISTS_P (p1)) | |
4916 | CHARSET_LOOKUP_RANGE_TABLE (not, c, p1); | |
505bde11 | 4917 | |
4e8a9132 SM |
4918 | /* `not' is equal to 1 if c would match, which means |
4919 | that we can't change to pop_failure_jump. */ | |
4920 | if (!not) | |
4921 | { | |
4922 | DEBUG_PRINT1 (" No match => fast loop.\n"); | |
4923 | return 1; | |
4924 | } | |
4925 | } | |
4926 | else if ((re_opcode_t) *p1 == anychar | |
4927 | && c == '\n') | |
4928 | { | |
4929 | DEBUG_PRINT1 (" . != \\n => fast loop.\n"); | |
4930 | return 1; | |
4931 | } | |
4932 | } | |
4933 | break; | |
505bde11 | 4934 | |
4e8a9132 | 4935 | case charset: |
4e8a9132 SM |
4936 | { |
4937 | if ((re_opcode_t) *p1 == exactn) | |
4938 | /* Reuse the code above. */ | |
4939 | return mutually_exclusive_p (bufp, p2, p1); | |
505bde11 | 4940 | |
505bde11 SM |
4941 | /* It is hard to list up all the character in charset |
4942 | P2 if it includes multibyte character. Give up in | |
4943 | such case. */ | |
4944 | else if (!multibyte || !CHARSET_RANGE_TABLE_EXISTS_P (p2)) | |
4945 | { | |
4946 | /* Now, we are sure that P2 has no range table. | |
4947 | So, for the size of bitmap in P2, `p2[1]' is | |
7814e705 | 4948 | enough. But P1 may have range table, so the |
505bde11 SM |
4949 | size of bitmap table of P1 is extracted by |
4950 | using macro `CHARSET_BITMAP_SIZE'. | |
4951 | ||
6fdd04b0 KH |
4952 | In a multibyte case, we know that all the character |
4953 | listed in P2 is ASCII. In a unibyte case, P1 has only a | |
4954 | bitmap table. So, in both cases, it is enough to test | |
4955 | only the bitmap table of P1. */ | |
505bde11 | 4956 | |
411e4203 | 4957 | if ((re_opcode_t) *p1 == charset) |
505bde11 SM |
4958 | { |
4959 | int idx; | |
4960 | /* We win if the charset inside the loop | |
4961 | has no overlap with the one after the loop. */ | |
4962 | for (idx = 0; | |
4963 | (idx < (int) p2[1] | |
4964 | && idx < CHARSET_BITMAP_SIZE (p1)); | |
4965 | idx++) | |
4966 | if ((p2[2 + idx] & p1[2 + idx]) != 0) | |
4967 | break; | |
4968 | ||
4969 | if (idx == p2[1] | |
4970 | || idx == CHARSET_BITMAP_SIZE (p1)) | |
4971 | { | |
4972 | DEBUG_PRINT1 (" No match => fast loop.\n"); | |
4973 | return 1; | |
4974 | } | |
4975 | } | |
411e4203 | 4976 | else if ((re_opcode_t) *p1 == charset_not) |
505bde11 SM |
4977 | { |
4978 | int idx; | |
4979 | /* We win if the charset_not inside the loop lists | |
7814e705 | 4980 | every character listed in the charset after. */ |
505bde11 SM |
4981 | for (idx = 0; idx < (int) p2[1]; idx++) |
4982 | if (! (p2[2 + idx] == 0 | |
4983 | || (idx < CHARSET_BITMAP_SIZE (p1) | |
4984 | && ((p2[2 + idx] & ~ p1[2 + idx]) == 0)))) | |
4985 | break; | |
4986 | ||
4e8a9132 SM |
4987 | if (idx == p2[1]) |
4988 | { | |
4989 | DEBUG_PRINT1 (" No match => fast loop.\n"); | |
4990 | return 1; | |
4991 | } | |
4992 | } | |
4993 | } | |
4994 | } | |
609b757a | 4995 | break; |
177c0ea7 | 4996 | |
411e4203 SM |
4997 | case charset_not: |
4998 | switch (SWITCH_ENUM_CAST (*p1)) | |
4999 | { | |
5000 | case exactn: | |
5001 | case charset: | |
5002 | /* Reuse the code above. */ | |
5003 | return mutually_exclusive_p (bufp, p2, p1); | |
5004 | case charset_not: | |
5005 | /* When we have two charset_not, it's very unlikely that | |
5006 | they don't overlap. The union of the two sets of excluded | |
5007 | chars should cover all possible chars, which, as a matter of | |
5008 | fact, is virtually impossible in multibyte buffers. */ | |
36595814 | 5009 | break; |
411e4203 SM |
5010 | } |
5011 | break; | |
5012 | ||
4e8a9132 | 5013 | case wordend: |
669fa600 SM |
5014 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == Sword); |
5015 | case symend: | |
4e8a9132 | 5016 | return ((re_opcode_t) *p1 == syntaxspec |
669fa600 SM |
5017 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
5018 | case notsyntaxspec: | |
5019 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
5020 | |
5021 | case wordbeg: | |
669fa600 SM |
5022 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == Sword); |
5023 | case symbeg: | |
4e8a9132 | 5024 | return ((re_opcode_t) *p1 == notsyntaxspec |
669fa600 SM |
5025 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
5026 | case syntaxspec: | |
5027 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
5028 | |
5029 | case wordbound: | |
5030 | return (((re_opcode_t) *p1 == notsyntaxspec | |
5031 | || (re_opcode_t) *p1 == syntaxspec) | |
5032 | && p1[1] == Sword); | |
5033 | ||
1fb352e0 | 5034 | #ifdef emacs |
4e8a9132 SM |
5035 | case categoryspec: |
5036 | return ((re_opcode_t) *p1 == notcategoryspec && p1[1] == p2[1]); | |
5037 | case notcategoryspec: | |
5038 | return ((re_opcode_t) *p1 == categoryspec && p1[1] == p2[1]); | |
5039 | #endif /* emacs */ | |
5040 | ||
5041 | default: | |
5042 | ; | |
505bde11 SM |
5043 | } |
5044 | ||
5045 | /* Safe default. */ | |
5046 | return 0; | |
5047 | } | |
5048 | ||
fa9a63c5 RM |
5049 | \f |
5050 | /* Matching routines. */ | |
5051 | ||
25fe55af | 5052 | #ifndef emacs /* Emacs never uses this. */ |
fa9a63c5 RM |
5053 | /* re_match is like re_match_2 except it takes only a single string. */ |
5054 | ||
5055 | int | |
5056 | re_match (bufp, string, size, pos, regs) | |
5057 | struct re_pattern_buffer *bufp; | |
5058 | const char *string; | |
5059 | int size, pos; | |
5060 | struct re_registers *regs; | |
5061 | { | |
4bb91c68 | 5062 | int result = re_match_2_internal (bufp, NULL, 0, (re_char*) string, size, |
fa9a63c5 | 5063 | pos, regs, size); |
fa9a63c5 RM |
5064 | return result; |
5065 | } | |
c0f9ea08 | 5066 | WEAK_ALIAS (__re_match, re_match) |
fa9a63c5 RM |
5067 | #endif /* not emacs */ |
5068 | ||
b18215fc RS |
5069 | #ifdef emacs |
5070 | /* In Emacs, this is the string or buffer in which we | |
7814e705 | 5071 | are matching. It is used for looking up syntax properties. */ |
b18215fc RS |
5072 | Lisp_Object re_match_object; |
5073 | #endif | |
fa9a63c5 RM |
5074 | |
5075 | /* re_match_2 matches the compiled pattern in BUFP against the | |
5076 | the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 | |
5077 | and SIZE2, respectively). We start matching at POS, and stop | |
5078 | matching at STOP. | |
5e69f11e | 5079 | |
fa9a63c5 | 5080 | If REGS is non-null and the `no_sub' field of BUFP is nonzero, we |
7814e705 | 5081 | store offsets for the substring each group matched in REGS. See the |
fa9a63c5 RM |
5082 | documentation for exactly how many groups we fill. |
5083 | ||
5084 | We return -1 if no match, -2 if an internal error (such as the | |
7814e705 | 5085 | failure stack overflowing). Otherwise, we return the length of the |
fa9a63c5 RM |
5086 | matched substring. */ |
5087 | ||
5088 | int | |
5089 | re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) | |
5090 | struct re_pattern_buffer *bufp; | |
5091 | const char *string1, *string2; | |
5092 | int size1, size2; | |
5093 | int pos; | |
5094 | struct re_registers *regs; | |
5095 | int stop; | |
5096 | { | |
b18215fc | 5097 | int result; |
25fe55af | 5098 | |
b18215fc | 5099 | #ifdef emacs |
cc9b4df2 | 5100 | int charpos; |
d48cd3f4 | 5101 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
99633e97 | 5102 | charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (pos)); |
cc9b4df2 | 5103 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); |
b18215fc RS |
5104 | #endif |
5105 | ||
4bb91c68 SM |
5106 | result = re_match_2_internal (bufp, (re_char*) string1, size1, |
5107 | (re_char*) string2, size2, | |
cc9b4df2 | 5108 | pos, regs, stop); |
fa9a63c5 RM |
5109 | return result; |
5110 | } | |
c0f9ea08 | 5111 | WEAK_ALIAS (__re_match_2, re_match_2) |
fa9a63c5 | 5112 | |
bf216479 | 5113 | |
fa9a63c5 | 5114 | /* This is a separate function so that we can force an alloca cleanup |
7814e705 | 5115 | afterwards. */ |
fa9a63c5 RM |
5116 | static int |
5117 | re_match_2_internal (bufp, string1, size1, string2, size2, pos, regs, stop) | |
5118 | struct re_pattern_buffer *bufp; | |
66f0296e | 5119 | re_char *string1, *string2; |
fa9a63c5 RM |
5120 | int size1, size2; |
5121 | int pos; | |
5122 | struct re_registers *regs; | |
5123 | int stop; | |
5124 | { | |
5125 | /* General temporaries. */ | |
5126 | int mcnt; | |
01618498 | 5127 | size_t reg; |
66f0296e | 5128 | boolean not; |
fa9a63c5 RM |
5129 | |
5130 | /* Just past the end of the corresponding string. */ | |
66f0296e | 5131 | re_char *end1, *end2; |
fa9a63c5 RM |
5132 | |
5133 | /* Pointers into string1 and string2, just past the last characters in | |
7814e705 | 5134 | each to consider matching. */ |
66f0296e | 5135 | re_char *end_match_1, *end_match_2; |
fa9a63c5 RM |
5136 | |
5137 | /* Where we are in the data, and the end of the current string. */ | |
66f0296e | 5138 | re_char *d, *dend; |
5e69f11e | 5139 | |
99633e97 SM |
5140 | /* Used sometimes to remember where we were before starting matching |
5141 | an operator so that we can go back in case of failure. This "atomic" | |
5142 | behavior of matching opcodes is indispensable to the correctness | |
5143 | of the on_failure_keep_string_jump optimization. */ | |
5144 | re_char *dfail; | |
5145 | ||
fa9a63c5 | 5146 | /* Where we are in the pattern, and the end of the pattern. */ |
01618498 SM |
5147 | re_char *p = bufp->buffer; |
5148 | re_char *pend = p + bufp->used; | |
fa9a63c5 | 5149 | |
25fe55af | 5150 | /* We use this to map every character in the string. */ |
6676cb1c | 5151 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 | 5152 | |
cf9c99bc | 5153 | /* Nonzero if BUFP is setup from a multibyte regex. */ |
2d1675e4 | 5154 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
b18215fc | 5155 | |
cf9c99bc KH |
5156 | /* Nonzero if STRING1/STRING2 are multibyte. */ |
5157 | const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
5158 | ||
fa9a63c5 RM |
5159 | /* Failure point stack. Each place that can handle a failure further |
5160 | down the line pushes a failure point on this stack. It consists of | |
505bde11 | 5161 | regstart, and regend for all registers corresponding to |
fa9a63c5 RM |
5162 | the subexpressions we're currently inside, plus the number of such |
5163 | registers, and, finally, two char *'s. The first char * is where | |
5164 | to resume scanning the pattern; the second one is where to resume | |
7814e705 JB |
5165 | scanning the strings. */ |
5166 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | |
fa9a63c5 RM |
5167 | fail_stack_type fail_stack; |
5168 | #endif | |
5169 | #ifdef DEBUG | |
fa9a63c5 RM |
5170 | unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; |
5171 | #endif | |
5172 | ||
0b32bf0e | 5173 | #if defined REL_ALLOC && defined REGEX_MALLOC |
fa9a63c5 RM |
5174 | /* This holds the pointer to the failure stack, when |
5175 | it is allocated relocatably. */ | |
5176 | fail_stack_elt_t *failure_stack_ptr; | |
99633e97 | 5177 | #endif |
fa9a63c5 RM |
5178 | |
5179 | /* We fill all the registers internally, independent of what we | |
7814e705 | 5180 | return, for use in backreferences. The number here includes |
fa9a63c5 | 5181 | an element for register zero. */ |
4bb91c68 | 5182 | size_t num_regs = bufp->re_nsub + 1; |
5e69f11e | 5183 | |
fa9a63c5 RM |
5184 | /* Information on the contents of registers. These are pointers into |
5185 | the input strings; they record just what was matched (on this | |
5186 | attempt) by a subexpression part of the pattern, that is, the | |
5187 | regnum-th regstart pointer points to where in the pattern we began | |
5188 | matching and the regnum-th regend points to right after where we | |
5189 | stopped matching the regnum-th subexpression. (The zeroth register | |
5190 | keeps track of what the whole pattern matches.) */ | |
5191 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5192 | re_char **regstart, **regend; |
fa9a63c5 RM |
5193 | #endif |
5194 | ||
fa9a63c5 | 5195 | /* The following record the register info as found in the above |
5e69f11e | 5196 | variables when we find a match better than any we've seen before. |
fa9a63c5 RM |
5197 | This happens as we backtrack through the failure points, which in |
5198 | turn happens only if we have not yet matched the entire string. */ | |
5199 | unsigned best_regs_set = false; | |
5200 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5201 | re_char **best_regstart, **best_regend; |
fa9a63c5 | 5202 | #endif |
5e69f11e | 5203 | |
fa9a63c5 RM |
5204 | /* Logically, this is `best_regend[0]'. But we don't want to have to |
5205 | allocate space for that if we're not allocating space for anything | |
7814e705 | 5206 | else (see below). Also, we never need info about register 0 for |
fa9a63c5 RM |
5207 | any of the other register vectors, and it seems rather a kludge to |
5208 | treat `best_regend' differently than the rest. So we keep track of | |
5209 | the end of the best match so far in a separate variable. We | |
5210 | initialize this to NULL so that when we backtrack the first time | |
5211 | and need to test it, it's not garbage. */ | |
66f0296e | 5212 | re_char *match_end = NULL; |
fa9a63c5 | 5213 | |
fa9a63c5 RM |
5214 | #ifdef DEBUG |
5215 | /* Counts the total number of registers pushed. */ | |
5e69f11e | 5216 | unsigned num_regs_pushed = 0; |
fa9a63c5 RM |
5217 | #endif |
5218 | ||
5219 | DEBUG_PRINT1 ("\n\nEntering re_match_2.\n"); | |
5e69f11e | 5220 | |
fa9a63c5 | 5221 | INIT_FAIL_STACK (); |
5e69f11e | 5222 | |
fa9a63c5 RM |
5223 | #ifdef MATCH_MAY_ALLOCATE |
5224 | /* Do not bother to initialize all the register variables if there are | |
5225 | no groups in the pattern, as it takes a fair amount of time. If | |
5226 | there are groups, we include space for register 0 (the whole | |
5227 | pattern), even though we never use it, since it simplifies the | |
5228 | array indexing. We should fix this. */ | |
5229 | if (bufp->re_nsub) | |
5230 | { | |
66f0296e SM |
5231 | regstart = REGEX_TALLOC (num_regs, re_char *); |
5232 | regend = REGEX_TALLOC (num_regs, re_char *); | |
5233 | best_regstart = REGEX_TALLOC (num_regs, re_char *); | |
5234 | best_regend = REGEX_TALLOC (num_regs, re_char *); | |
fa9a63c5 | 5235 | |
505bde11 | 5236 | if (!(regstart && regend && best_regstart && best_regend)) |
25fe55af RS |
5237 | { |
5238 | FREE_VARIABLES (); | |
5239 | return -2; | |
5240 | } | |
fa9a63c5 RM |
5241 | } |
5242 | else | |
5243 | { | |
5244 | /* We must initialize all our variables to NULL, so that | |
25fe55af | 5245 | `FREE_VARIABLES' doesn't try to free them. */ |
505bde11 | 5246 | regstart = regend = best_regstart = best_regend = NULL; |
fa9a63c5 RM |
5247 | } |
5248 | #endif /* MATCH_MAY_ALLOCATE */ | |
5249 | ||
5250 | /* The starting position is bogus. */ | |
5251 | if (pos < 0 || pos > size1 + size2) | |
5252 | { | |
5253 | FREE_VARIABLES (); | |
5254 | return -1; | |
5255 | } | |
5e69f11e | 5256 | |
fa9a63c5 RM |
5257 | /* Initialize subexpression text positions to -1 to mark ones that no |
5258 | start_memory/stop_memory has been seen for. Also initialize the | |
5259 | register information struct. */ | |
01618498 SM |
5260 | for (reg = 1; reg < num_regs; reg++) |
5261 | regstart[reg] = regend[reg] = NULL; | |
99633e97 | 5262 | |
fa9a63c5 | 5263 | /* We move `string1' into `string2' if the latter's empty -- but not if |
7814e705 | 5264 | `string1' is null. */ |
fa9a63c5 RM |
5265 | if (size2 == 0 && string1 != NULL) |
5266 | { | |
5267 | string2 = string1; | |
5268 | size2 = size1; | |
5269 | string1 = 0; | |
5270 | size1 = 0; | |
5271 | } | |
5272 | end1 = string1 + size1; | |
5273 | end2 = string2 + size2; | |
5274 | ||
5e69f11e | 5275 | /* `p' scans through the pattern as `d' scans through the data. |
fa9a63c5 RM |
5276 | `dend' is the end of the input string that `d' points within. `d' |
5277 | is advanced into the following input string whenever necessary, but | |
5278 | this happens before fetching; therefore, at the beginning of the | |
5279 | loop, `d' can be pointing at the end of a string, but it cannot | |
5280 | equal `string2'. */ | |
419d1c74 | 5281 | if (pos >= size1) |
fa9a63c5 | 5282 | { |
419d1c74 SM |
5283 | /* Only match within string2. */ |
5284 | d = string2 + pos - size1; | |
5285 | dend = end_match_2 = string2 + stop - size1; | |
5286 | end_match_1 = end1; /* Just to give it a value. */ | |
fa9a63c5 RM |
5287 | } |
5288 | else | |
5289 | { | |
f1ad044f | 5290 | if (stop < size1) |
419d1c74 SM |
5291 | { |
5292 | /* Only match within string1. */ | |
5293 | end_match_1 = string1 + stop; | |
5294 | /* BEWARE! | |
5295 | When we reach end_match_1, PREFETCH normally switches to string2. | |
5296 | But in the present case, this means that just doing a PREFETCH | |
5297 | makes us jump from `stop' to `gap' within the string. | |
5298 | What we really want here is for the search to stop as | |
5299 | soon as we hit end_match_1. That's why we set end_match_2 | |
5300 | to end_match_1 (since PREFETCH fails as soon as we hit | |
5301 | end_match_2). */ | |
5302 | end_match_2 = end_match_1; | |
5303 | } | |
5304 | else | |
f1ad044f SM |
5305 | { /* It's important to use this code when stop == size so that |
5306 | moving `d' from end1 to string2 will not prevent the d == dend | |
5307 | check from catching the end of string. */ | |
419d1c74 SM |
5308 | end_match_1 = end1; |
5309 | end_match_2 = string2 + stop - size1; | |
5310 | } | |
5311 | d = string1 + pos; | |
5312 | dend = end_match_1; | |
fa9a63c5 RM |
5313 | } |
5314 | ||
5315 | DEBUG_PRINT1 ("The compiled pattern is: "); | |
5316 | DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); | |
5317 | DEBUG_PRINT1 ("The string to match is: `"); | |
5318 | DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); | |
5319 | DEBUG_PRINT1 ("'\n"); | |
5e69f11e | 5320 | |
7814e705 | 5321 | /* This loops over pattern commands. It exits by returning from the |
fa9a63c5 RM |
5322 | function if the match is complete, or it drops through if the match |
5323 | fails at this starting point in the input data. */ | |
5324 | for (;;) | |
5325 | { | |
505bde11 | 5326 | DEBUG_PRINT2 ("\n%p: ", p); |
fa9a63c5 RM |
5327 | |
5328 | if (p == pend) | |
5329 | { /* End of pattern means we might have succeeded. */ | |
25fe55af | 5330 | DEBUG_PRINT1 ("end of pattern ... "); |
5e69f11e | 5331 | |
fa9a63c5 | 5332 | /* If we haven't matched the entire string, and we want the |
25fe55af RS |
5333 | longest match, try backtracking. */ |
5334 | if (d != end_match_2) | |
fa9a63c5 RM |
5335 | { |
5336 | /* 1 if this match ends in the same string (string1 or string2) | |
5337 | as the best previous match. */ | |
5e69f11e | 5338 | boolean same_str_p = (FIRST_STRING_P (match_end) |
99633e97 | 5339 | == FIRST_STRING_P (d)); |
fa9a63c5 RM |
5340 | /* 1 if this match is the best seen so far. */ |
5341 | boolean best_match_p; | |
5342 | ||
5343 | /* AIX compiler got confused when this was combined | |
7814e705 | 5344 | with the previous declaration. */ |
fa9a63c5 RM |
5345 | if (same_str_p) |
5346 | best_match_p = d > match_end; | |
5347 | else | |
99633e97 | 5348 | best_match_p = !FIRST_STRING_P (d); |
fa9a63c5 | 5349 | |
25fe55af RS |
5350 | DEBUG_PRINT1 ("backtracking.\n"); |
5351 | ||
5352 | if (!FAIL_STACK_EMPTY ()) | |
5353 | { /* More failure points to try. */ | |
5354 | ||
5355 | /* If exceeds best match so far, save it. */ | |
5356 | if (!best_regs_set || best_match_p) | |
5357 | { | |
5358 | best_regs_set = true; | |
5359 | match_end = d; | |
5360 | ||
5361 | DEBUG_PRINT1 ("\nSAVING match as best so far.\n"); | |
5362 | ||
01618498 | 5363 | for (reg = 1; reg < num_regs; reg++) |
25fe55af | 5364 | { |
01618498 SM |
5365 | best_regstart[reg] = regstart[reg]; |
5366 | best_regend[reg] = regend[reg]; | |
25fe55af RS |
5367 | } |
5368 | } | |
5369 | goto fail; | |
5370 | } | |
5371 | ||
5372 | /* If no failure points, don't restore garbage. And if | |
5373 | last match is real best match, don't restore second | |
5374 | best one. */ | |
5375 | else if (best_regs_set && !best_match_p) | |
5376 | { | |
5377 | restore_best_regs: | |
5378 | /* Restore best match. It may happen that `dend == | |
5379 | end_match_1' while the restored d is in string2. | |
5380 | For example, the pattern `x.*y.*z' against the | |
5381 | strings `x-' and `y-z-', if the two strings are | |
7814e705 | 5382 | not consecutive in memory. */ |
25fe55af RS |
5383 | DEBUG_PRINT1 ("Restoring best registers.\n"); |
5384 | ||
5385 | d = match_end; | |
5386 | dend = ((d >= string1 && d <= end1) | |
5387 | ? end_match_1 : end_match_2); | |
fa9a63c5 | 5388 | |
01618498 | 5389 | for (reg = 1; reg < num_regs; reg++) |
fa9a63c5 | 5390 | { |
01618498 SM |
5391 | regstart[reg] = best_regstart[reg]; |
5392 | regend[reg] = best_regend[reg]; | |
fa9a63c5 | 5393 | } |
25fe55af RS |
5394 | } |
5395 | } /* d != end_match_2 */ | |
fa9a63c5 RM |
5396 | |
5397 | succeed_label: | |
25fe55af | 5398 | DEBUG_PRINT1 ("Accepting match.\n"); |
fa9a63c5 | 5399 | |
25fe55af RS |
5400 | /* If caller wants register contents data back, do it. */ |
5401 | if (regs && !bufp->no_sub) | |
fa9a63c5 | 5402 | { |
25fe55af RS |
5403 | /* Have the register data arrays been allocated? */ |
5404 | if (bufp->regs_allocated == REGS_UNALLOCATED) | |
7814e705 | 5405 | { /* No. So allocate them with malloc. We need one |
25fe55af RS |
5406 | extra element beyond `num_regs' for the `-1' marker |
5407 | GNU code uses. */ | |
5408 | regs->num_regs = MAX (RE_NREGS, num_regs + 1); | |
5409 | regs->start = TALLOC (regs->num_regs, regoff_t); | |
5410 | regs->end = TALLOC (regs->num_regs, regoff_t); | |
5411 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5412 | { |
5413 | FREE_VARIABLES (); | |
5414 | return -2; | |
5415 | } | |
25fe55af RS |
5416 | bufp->regs_allocated = REGS_REALLOCATE; |
5417 | } | |
5418 | else if (bufp->regs_allocated == REGS_REALLOCATE) | |
5419 | { /* Yes. If we need more elements than were already | |
5420 | allocated, reallocate them. If we need fewer, just | |
5421 | leave it alone. */ | |
5422 | if (regs->num_regs < num_regs + 1) | |
5423 | { | |
5424 | regs->num_regs = num_regs + 1; | |
5425 | RETALLOC (regs->start, regs->num_regs, regoff_t); | |
5426 | RETALLOC (regs->end, regs->num_regs, regoff_t); | |
5427 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5428 | { |
5429 | FREE_VARIABLES (); | |
5430 | return -2; | |
5431 | } | |
25fe55af RS |
5432 | } |
5433 | } | |
5434 | else | |
fa9a63c5 RM |
5435 | { |
5436 | /* These braces fend off a "empty body in an else-statement" | |
7814e705 | 5437 | warning under GCC when assert expands to nothing. */ |
fa9a63c5 RM |
5438 | assert (bufp->regs_allocated == REGS_FIXED); |
5439 | } | |
5440 | ||
25fe55af RS |
5441 | /* Convert the pointer data in `regstart' and `regend' to |
5442 | indices. Register zero has to be set differently, | |
5443 | since we haven't kept track of any info for it. */ | |
5444 | if (regs->num_regs > 0) | |
5445 | { | |
5446 | regs->start[0] = pos; | |
99633e97 | 5447 | regs->end[0] = POINTER_TO_OFFSET (d); |
25fe55af | 5448 | } |
5e69f11e | 5449 | |
25fe55af RS |
5450 | /* Go through the first `min (num_regs, regs->num_regs)' |
5451 | registers, since that is all we initialized. */ | |
01618498 | 5452 | for (reg = 1; reg < MIN (num_regs, regs->num_regs); reg++) |
fa9a63c5 | 5453 | { |
01618498 SM |
5454 | if (REG_UNSET (regstart[reg]) || REG_UNSET (regend[reg])) |
5455 | regs->start[reg] = regs->end[reg] = -1; | |
25fe55af RS |
5456 | else |
5457 | { | |
01618498 SM |
5458 | regs->start[reg] |
5459 | = (regoff_t) POINTER_TO_OFFSET (regstart[reg]); | |
5460 | regs->end[reg] | |
5461 | = (regoff_t) POINTER_TO_OFFSET (regend[reg]); | |
25fe55af | 5462 | } |
fa9a63c5 | 5463 | } |
5e69f11e | 5464 | |
25fe55af RS |
5465 | /* If the regs structure we return has more elements than |
5466 | were in the pattern, set the extra elements to -1. If | |
5467 | we (re)allocated the registers, this is the case, | |
5468 | because we always allocate enough to have at least one | |
7814e705 | 5469 | -1 at the end. */ |
01618498 SM |
5470 | for (reg = num_regs; reg < regs->num_regs; reg++) |
5471 | regs->start[reg] = regs->end[reg] = -1; | |
fa9a63c5 RM |
5472 | } /* regs && !bufp->no_sub */ |
5473 | ||
25fe55af RS |
5474 | DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n", |
5475 | nfailure_points_pushed, nfailure_points_popped, | |
5476 | nfailure_points_pushed - nfailure_points_popped); | |
5477 | DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed); | |
fa9a63c5 | 5478 | |
99633e97 | 5479 | mcnt = POINTER_TO_OFFSET (d) - pos; |
fa9a63c5 | 5480 | |
25fe55af | 5481 | DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt); |
fa9a63c5 | 5482 | |
25fe55af RS |
5483 | FREE_VARIABLES (); |
5484 | return mcnt; | |
5485 | } | |
fa9a63c5 | 5486 | |
7814e705 | 5487 | /* Otherwise match next pattern command. */ |
fa9a63c5 RM |
5488 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++)) |
5489 | { | |
25fe55af RS |
5490 | /* Ignore these. Used to ignore the n of succeed_n's which |
5491 | currently have n == 0. */ | |
5492 | case no_op: | |
5493 | DEBUG_PRINT1 ("EXECUTING no_op.\n"); | |
5494 | break; | |
fa9a63c5 RM |
5495 | |
5496 | case succeed: | |
25fe55af | 5497 | DEBUG_PRINT1 ("EXECUTING succeed.\n"); |
fa9a63c5 RM |
5498 | goto succeed_label; |
5499 | ||
7814e705 | 5500 | /* Match the next n pattern characters exactly. The following |
25fe55af | 5501 | byte in the pattern defines n, and the n bytes after that |
7814e705 | 5502 | are the characters to match. */ |
fa9a63c5 RM |
5503 | case exactn: |
5504 | mcnt = *p++; | |
25fe55af | 5505 | DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt); |
fa9a63c5 | 5506 | |
99633e97 SM |
5507 | /* Remember the start point to rollback upon failure. */ |
5508 | dfail = d; | |
5509 | ||
6fdd04b0 | 5510 | #ifndef emacs |
25fe55af RS |
5511 | /* This is written out as an if-else so we don't waste time |
5512 | testing `translate' inside the loop. */ | |
28703c16 | 5513 | if (RE_TRANSLATE_P (translate)) |
6fdd04b0 KH |
5514 | do |
5515 | { | |
5516 | PREFETCH (); | |
5517 | if (RE_TRANSLATE (translate, *d) != *p++) | |
e934739e | 5518 | { |
6fdd04b0 KH |
5519 | d = dfail; |
5520 | goto fail; | |
e934739e | 5521 | } |
6fdd04b0 KH |
5522 | d++; |
5523 | } | |
5524 | while (--mcnt); | |
fa9a63c5 | 5525 | else |
6fdd04b0 KH |
5526 | do |
5527 | { | |
5528 | PREFETCH (); | |
5529 | if (*d++ != *p++) | |
bf216479 | 5530 | { |
6fdd04b0 KH |
5531 | d = dfail; |
5532 | goto fail; | |
bf216479 | 5533 | } |
6fdd04b0 KH |
5534 | } |
5535 | while (--mcnt); | |
5536 | #else /* emacs */ | |
5537 | /* The cost of testing `translate' is comparatively small. */ | |
cf9c99bc | 5538 | if (target_multibyte) |
6fdd04b0 KH |
5539 | do |
5540 | { | |
5541 | int pat_charlen, buf_charlen; | |
cf9c99bc | 5542 | int pat_ch, buf_ch; |
e934739e | 5543 | |
6fdd04b0 | 5544 | PREFETCH (); |
cf9c99bc | 5545 | if (multibyte) |
62a6e103 | 5546 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
cf9c99bc KH |
5547 | else |
5548 | { | |
5549 | pat_ch = RE_CHAR_TO_MULTIBYTE (*p); | |
5550 | pat_charlen = 1; | |
5551 | } | |
62a6e103 | 5552 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 5553 | |
6fdd04b0 | 5554 | if (TRANSLATE (buf_ch) != pat_ch) |
e934739e | 5555 | { |
6fdd04b0 KH |
5556 | d = dfail; |
5557 | goto fail; | |
e934739e | 5558 | } |
bf216479 | 5559 | |
6fdd04b0 KH |
5560 | p += pat_charlen; |
5561 | d += buf_charlen; | |
5562 | mcnt -= pat_charlen; | |
5563 | } | |
5564 | while (mcnt > 0); | |
fa9a63c5 | 5565 | else |
6fdd04b0 KH |
5566 | do |
5567 | { | |
cf9c99bc KH |
5568 | int pat_charlen, buf_charlen; |
5569 | int pat_ch, buf_ch; | |
bf216479 | 5570 | |
6fdd04b0 | 5571 | PREFETCH (); |
cf9c99bc KH |
5572 | if (multibyte) |
5573 | { | |
62a6e103 | 5574 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
2afc21f5 | 5575 | pat_ch = RE_CHAR_TO_UNIBYTE (pat_ch); |
cf9c99bc KH |
5576 | } |
5577 | else | |
5578 | { | |
5579 | pat_ch = *p; | |
5580 | pat_charlen = 1; | |
5581 | } | |
5582 | buf_ch = RE_CHAR_TO_MULTIBYTE (*d); | |
5583 | if (! CHAR_BYTE8_P (buf_ch)) | |
5584 | { | |
5585 | buf_ch = TRANSLATE (buf_ch); | |
5586 | buf_ch = RE_CHAR_TO_UNIBYTE (buf_ch); | |
5587 | if (buf_ch < 0) | |
5588 | buf_ch = *d; | |
5589 | } | |
0e2501ed AS |
5590 | else |
5591 | buf_ch = *d; | |
cf9c99bc | 5592 | if (buf_ch != pat_ch) |
6fdd04b0 KH |
5593 | { |
5594 | d = dfail; | |
5595 | goto fail; | |
bf216479 | 5596 | } |
cf9c99bc KH |
5597 | p += pat_charlen; |
5598 | d++; | |
6fdd04b0 KH |
5599 | } |
5600 | while (--mcnt); | |
5601 | #endif | |
25fe55af | 5602 | break; |
fa9a63c5 RM |
5603 | |
5604 | ||
25fe55af | 5605 | /* Match any character except possibly a newline or a null. */ |
fa9a63c5 | 5606 | case anychar: |
e934739e RS |
5607 | { |
5608 | int buf_charlen; | |
01618498 | 5609 | re_wchar_t buf_ch; |
fa9a63c5 | 5610 | |
e934739e | 5611 | DEBUG_PRINT1 ("EXECUTING anychar.\n"); |
fa9a63c5 | 5612 | |
e934739e | 5613 | PREFETCH (); |
62a6e103 | 5614 | buf_ch = RE_STRING_CHAR_AND_LENGTH (d, buf_charlen, |
cf9c99bc | 5615 | target_multibyte); |
e934739e RS |
5616 | buf_ch = TRANSLATE (buf_ch); |
5617 | ||
5618 | if ((!(bufp->syntax & RE_DOT_NEWLINE) | |
5619 | && buf_ch == '\n') | |
5620 | || ((bufp->syntax & RE_DOT_NOT_NULL) | |
5621 | && buf_ch == '\000')) | |
5622 | goto fail; | |
5623 | ||
e934739e RS |
5624 | DEBUG_PRINT2 (" Matched `%d'.\n", *d); |
5625 | d += buf_charlen; | |
5626 | } | |
fa9a63c5 RM |
5627 | break; |
5628 | ||
5629 | ||
5630 | case charset: | |
5631 | case charset_not: | |
5632 | { | |
b18215fc | 5633 | register unsigned int c; |
fa9a63c5 | 5634 | boolean not = (re_opcode_t) *(p - 1) == charset_not; |
b18215fc RS |
5635 | int len; |
5636 | ||
5637 | /* Start of actual range_table, or end of bitmap if there is no | |
5638 | range table. */ | |
01618498 | 5639 | re_char *range_table; |
b18215fc | 5640 | |
96cc36cc | 5641 | /* Nonzero if there is a range table. */ |
b18215fc RS |
5642 | int range_table_exists; |
5643 | ||
96cc36cc RS |
5644 | /* Number of ranges of range table. This is not included |
5645 | in the initial byte-length of the command. */ | |
5646 | int count = 0; | |
fa9a63c5 | 5647 | |
f5020181 AS |
5648 | /* Whether matching against a unibyte character. */ |
5649 | boolean unibyte_char = false; | |
5650 | ||
25fe55af | 5651 | DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : ""); |
fa9a63c5 | 5652 | |
b18215fc | 5653 | range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]); |
96cc36cc | 5654 | |
b18215fc | 5655 | if (range_table_exists) |
96cc36cc RS |
5656 | { |
5657 | range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap. */ | |
5658 | EXTRACT_NUMBER_AND_INCR (count, range_table); | |
5659 | } | |
b18215fc | 5660 | |
2d1675e4 | 5661 | PREFETCH (); |
62a6e103 | 5662 | c = RE_STRING_CHAR_AND_LENGTH (d, len, target_multibyte); |
cf9c99bc KH |
5663 | if (target_multibyte) |
5664 | { | |
5665 | int c1; | |
b18215fc | 5666 | |
cf9c99bc KH |
5667 | c = TRANSLATE (c); |
5668 | c1 = RE_CHAR_TO_UNIBYTE (c); | |
5669 | if (c1 >= 0) | |
f5020181 AS |
5670 | { |
5671 | unibyte_char = true; | |
5672 | c = c1; | |
5673 | } | |
cf9c99bc KH |
5674 | } |
5675 | else | |
5676 | { | |
5677 | int c1 = RE_CHAR_TO_MULTIBYTE (c); | |
5678 | ||
5679 | if (! CHAR_BYTE8_P (c1)) | |
5680 | { | |
5681 | c1 = TRANSLATE (c1); | |
5682 | c1 = RE_CHAR_TO_UNIBYTE (c1); | |
5683 | if (c1 >= 0) | |
f5020181 AS |
5684 | { |
5685 | unibyte_char = true; | |
5686 | c = c1; | |
5687 | } | |
cf9c99bc | 5688 | } |
0b8be006 AS |
5689 | else |
5690 | unibyte_char = true; | |
cf9c99bc KH |
5691 | } |
5692 | ||
f5020181 | 5693 | if (unibyte_char && c < (1 << BYTEWIDTH)) |
b18215fc | 5694 | { /* Lookup bitmap. */ |
b18215fc RS |
5695 | /* Cast to `unsigned' instead of `unsigned char' in |
5696 | case the bit list is a full 32 bytes long. */ | |
5697 | if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH) | |
96cc36cc RS |
5698 | && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) |
5699 | not = !not; | |
b18215fc | 5700 | } |
96cc36cc | 5701 | #ifdef emacs |
b18215fc | 5702 | else if (range_table_exists) |
96cc36cc RS |
5703 | { |
5704 | int class_bits = CHARSET_RANGE_TABLE_BITS (&p[-1]); | |
5705 | ||
14473664 SM |
5706 | if ( (class_bits & BIT_LOWER && ISLOWER (c)) |
5707 | | (class_bits & BIT_MULTIBYTE) | |
96cc36cc RS |
5708 | | (class_bits & BIT_PUNCT && ISPUNCT (c)) |
5709 | | (class_bits & BIT_SPACE && ISSPACE (c)) | |
5710 | | (class_bits & BIT_UPPER && ISUPPER (c)) | |
5711 | | (class_bits & BIT_WORD && ISWORD (c))) | |
5712 | not = !not; | |
5713 | else | |
5714 | CHARSET_LOOKUP_RANGE_TABLE_RAW (not, c, range_table, count); | |
5715 | } | |
5716 | #endif /* emacs */ | |
fa9a63c5 | 5717 | |
96cc36cc RS |
5718 | if (range_table_exists) |
5719 | p = CHARSET_RANGE_TABLE_END (range_table, count); | |
5720 | else | |
5721 | p += CHARSET_BITMAP_SIZE (&p[-1]) + 1; | |
fa9a63c5 RM |
5722 | |
5723 | if (!not) goto fail; | |
5e69f11e | 5724 | |
b18215fc | 5725 | d += len; |
fa9a63c5 RM |
5726 | break; |
5727 | } | |
5728 | ||
5729 | ||
25fe55af | 5730 | /* The beginning of a group is represented by start_memory. |
505bde11 | 5731 | The argument is the register number. The text |
25fe55af | 5732 | matched within the group is recorded (in the internal |
7814e705 | 5733 | registers data structure) under the register number. */ |
25fe55af | 5734 | case start_memory: |
505bde11 SM |
5735 | DEBUG_PRINT2 ("EXECUTING start_memory %d:\n", *p); |
5736 | ||
5737 | /* In case we need to undo this operation (via backtracking). */ | |
5738 | PUSH_FAILURE_REG ((unsigned int)*p); | |
fa9a63c5 | 5739 | |
25fe55af | 5740 | regstart[*p] = d; |
4bb91c68 | 5741 | regend[*p] = NULL; /* probably unnecessary. -sm */ |
fa9a63c5 RM |
5742 | DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p])); |
5743 | ||
25fe55af | 5744 | /* Move past the register number and inner group count. */ |
505bde11 | 5745 | p += 1; |
25fe55af | 5746 | break; |
fa9a63c5 RM |
5747 | |
5748 | ||
25fe55af | 5749 | /* The stop_memory opcode represents the end of a group. Its |
505bde11 | 5750 | argument is the same as start_memory's: the register number. */ |
fa9a63c5 | 5751 | case stop_memory: |
505bde11 SM |
5752 | DEBUG_PRINT2 ("EXECUTING stop_memory %d:\n", *p); |
5753 | ||
5754 | assert (!REG_UNSET (regstart[*p])); | |
5755 | /* Strictly speaking, there should be code such as: | |
177c0ea7 | 5756 | |
0b32bf0e | 5757 | assert (REG_UNSET (regend[*p])); |
505bde11 SM |
5758 | PUSH_FAILURE_REGSTOP ((unsigned int)*p); |
5759 | ||
5760 | But the only info to be pushed is regend[*p] and it is known to | |
5761 | be UNSET, so there really isn't anything to push. | |
5762 | Not pushing anything, on the other hand deprives us from the | |
5763 | guarantee that regend[*p] is UNSET since undoing this operation | |
5764 | will not reset its value properly. This is not important since | |
5765 | the value will only be read on the next start_memory or at | |
5766 | the very end and both events can only happen if this stop_memory | |
5767 | is *not* undone. */ | |
fa9a63c5 | 5768 | |
25fe55af | 5769 | regend[*p] = d; |
fa9a63c5 RM |
5770 | DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p])); |
5771 | ||
25fe55af | 5772 | /* Move past the register number and the inner group count. */ |
505bde11 | 5773 | p += 1; |
25fe55af | 5774 | break; |
fa9a63c5 RM |
5775 | |
5776 | ||
5777 | /* \<digit> has been turned into a `duplicate' command which is | |
25fe55af RS |
5778 | followed by the numeric value of <digit> as the register number. */ |
5779 | case duplicate: | |
fa9a63c5 | 5780 | { |
66f0296e | 5781 | register re_char *d2, *dend2; |
7814e705 | 5782 | int regno = *p++; /* Get which register to match against. */ |
fa9a63c5 RM |
5783 | DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno); |
5784 | ||
7814e705 | 5785 | /* Can't back reference a group which we've never matched. */ |
25fe55af RS |
5786 | if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) |
5787 | goto fail; | |
5e69f11e | 5788 | |
7814e705 | 5789 | /* Where in input to try to start matching. */ |
25fe55af | 5790 | d2 = regstart[regno]; |
5e69f11e | 5791 | |
99633e97 SM |
5792 | /* Remember the start point to rollback upon failure. */ |
5793 | dfail = d; | |
5794 | ||
25fe55af RS |
5795 | /* Where to stop matching; if both the place to start and |
5796 | the place to stop matching are in the same string, then | |
5797 | set to the place to stop, otherwise, for now have to use | |
5798 | the end of the first string. */ | |
fa9a63c5 | 5799 | |
25fe55af | 5800 | dend2 = ((FIRST_STRING_P (regstart[regno]) |
fa9a63c5 RM |
5801 | == FIRST_STRING_P (regend[regno])) |
5802 | ? regend[regno] : end_match_1); | |
5803 | for (;;) | |
5804 | { | |
5805 | /* If necessary, advance to next segment in register | |
25fe55af | 5806 | contents. */ |
fa9a63c5 RM |
5807 | while (d2 == dend2) |
5808 | { | |
5809 | if (dend2 == end_match_2) break; | |
5810 | if (dend2 == regend[regno]) break; | |
5811 | ||
25fe55af RS |
5812 | /* End of string1 => advance to string2. */ |
5813 | d2 = string2; | |
5814 | dend2 = regend[regno]; | |
fa9a63c5 RM |
5815 | } |
5816 | /* At end of register contents => success */ | |
5817 | if (d2 == dend2) break; | |
5818 | ||
5819 | /* If necessary, advance to next segment in data. */ | |
5820 | PREFETCH (); | |
5821 | ||
5822 | /* How many characters left in this segment to match. */ | |
5823 | mcnt = dend - d; | |
5e69f11e | 5824 | |
fa9a63c5 | 5825 | /* Want how many consecutive characters we can match in |
25fe55af RS |
5826 | one shot, so, if necessary, adjust the count. */ |
5827 | if (mcnt > dend2 - d2) | |
fa9a63c5 | 5828 | mcnt = dend2 - d2; |
5e69f11e | 5829 | |
fa9a63c5 | 5830 | /* Compare that many; failure if mismatch, else move |
25fe55af | 5831 | past them. */ |
28703c16 | 5832 | if (RE_TRANSLATE_P (translate) |
02cb78b5 | 5833 | ? bcmp_translate (d, d2, mcnt, translate, target_multibyte) |
4bb91c68 | 5834 | : memcmp (d, d2, mcnt)) |
99633e97 SM |
5835 | { |
5836 | d = dfail; | |
5837 | goto fail; | |
5838 | } | |
fa9a63c5 | 5839 | d += mcnt, d2 += mcnt; |
fa9a63c5 RM |
5840 | } |
5841 | } | |
5842 | break; | |
5843 | ||
5844 | ||
25fe55af | 5845 | /* begline matches the empty string at the beginning of the string |
c0f9ea08 | 5846 | (unless `not_bol' is set in `bufp'), and after newlines. */ |
fa9a63c5 | 5847 | case begline: |
25fe55af | 5848 | DEBUG_PRINT1 ("EXECUTING begline.\n"); |
5e69f11e | 5849 | |
25fe55af RS |
5850 | if (AT_STRINGS_BEG (d)) |
5851 | { | |
5852 | if (!bufp->not_bol) break; | |
5853 | } | |
419d1c74 | 5854 | else |
25fe55af | 5855 | { |
bf216479 | 5856 | unsigned c; |
419d1c74 | 5857 | GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2); |
c0f9ea08 | 5858 | if (c == '\n') |
419d1c74 | 5859 | break; |
25fe55af RS |
5860 | } |
5861 | /* In all other cases, we fail. */ | |
5862 | goto fail; | |
fa9a63c5 RM |
5863 | |
5864 | ||
25fe55af | 5865 | /* endline is the dual of begline. */ |
fa9a63c5 | 5866 | case endline: |
25fe55af | 5867 | DEBUG_PRINT1 ("EXECUTING endline.\n"); |
fa9a63c5 | 5868 | |
25fe55af RS |
5869 | if (AT_STRINGS_END (d)) |
5870 | { | |
5871 | if (!bufp->not_eol) break; | |
5872 | } | |
f1ad044f | 5873 | else |
25fe55af | 5874 | { |
f1ad044f | 5875 | PREFETCH_NOLIMIT (); |
c0f9ea08 | 5876 | if (*d == '\n') |
f1ad044f | 5877 | break; |
25fe55af RS |
5878 | } |
5879 | goto fail; | |
fa9a63c5 RM |
5880 | |
5881 | ||
5882 | /* Match at the very beginning of the data. */ | |
25fe55af RS |
5883 | case begbuf: |
5884 | DEBUG_PRINT1 ("EXECUTING begbuf.\n"); | |
5885 | if (AT_STRINGS_BEG (d)) | |
5886 | break; | |
5887 | goto fail; | |
fa9a63c5 RM |
5888 | |
5889 | ||
5890 | /* Match at the very end of the data. */ | |
25fe55af RS |
5891 | case endbuf: |
5892 | DEBUG_PRINT1 ("EXECUTING endbuf.\n"); | |
fa9a63c5 RM |
5893 | if (AT_STRINGS_END (d)) |
5894 | break; | |
25fe55af | 5895 | goto fail; |
5e69f11e | 5896 | |
5e69f11e | 5897 | |
25fe55af RS |
5898 | /* on_failure_keep_string_jump is used to optimize `.*\n'. It |
5899 | pushes NULL as the value for the string on the stack. Then | |
505bde11 | 5900 | `POP_FAILURE_POINT' will keep the current value for the |
25fe55af | 5901 | string, instead of restoring it. To see why, consider |
7814e705 | 5902 | matching `foo\nbar' against `.*\n'. The .* matches the foo; |
25fe55af RS |
5903 | then the . fails against the \n. But the next thing we want |
5904 | to do is match the \n against the \n; if we restored the | |
5905 | string value, we would be back at the foo. | |
5906 | ||
5907 | Because this is used only in specific cases, we don't need to | |
5908 | check all the things that `on_failure_jump' does, to make | |
5909 | sure the right things get saved on the stack. Hence we don't | |
5910 | share its code. The only reason to push anything on the | |
5911 | stack at all is that otherwise we would have to change | |
5912 | `anychar's code to do something besides goto fail in this | |
5913 | case; that seems worse than this. */ | |
5914 | case on_failure_keep_string_jump: | |
505bde11 SM |
5915 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
5916 | DEBUG_PRINT3 ("EXECUTING on_failure_keep_string_jump %d (to %p):\n", | |
5917 | mcnt, p + mcnt); | |
fa9a63c5 | 5918 | |
505bde11 SM |
5919 | PUSH_FAILURE_POINT (p - 3, NULL); |
5920 | break; | |
5921 | ||
0683b6fa SM |
5922 | /* A nasty loop is introduced by the non-greedy *? and +?. |
5923 | With such loops, the stack only ever contains one failure point | |
5924 | at a time, so that a plain on_failure_jump_loop kind of | |
5925 | cycle detection cannot work. Worse yet, such a detection | |
5926 | can not only fail to detect a cycle, but it can also wrongly | |
5927 | detect a cycle (between different instantiations of the same | |
6df42991 | 5928 | loop). |
0683b6fa SM |
5929 | So the method used for those nasty loops is a little different: |
5930 | We use a special cycle-detection-stack-frame which is pushed | |
5931 | when the on_failure_jump_nastyloop failure-point is *popped*. | |
5932 | This special frame thus marks the beginning of one iteration | |
5933 | through the loop and we can hence easily check right here | |
5934 | whether something matched between the beginning and the end of | |
5935 | the loop. */ | |
5936 | case on_failure_jump_nastyloop: | |
5937 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
5938 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_nastyloop %d (to %p):\n", | |
5939 | mcnt, p + mcnt); | |
5940 | ||
5941 | assert ((re_opcode_t)p[-4] == no_op); | |
6df42991 SM |
5942 | { |
5943 | int cycle = 0; | |
5944 | CHECK_INFINITE_LOOP (p - 4, d); | |
5945 | if (!cycle) | |
5946 | /* If there's a cycle, just continue without pushing | |
5947 | this failure point. The failure point is the "try again" | |
5948 | option, which shouldn't be tried. | |
5949 | We want (x?)*?y\1z to match both xxyz and xxyxz. */ | |
5950 | PUSH_FAILURE_POINT (p - 3, d); | |
5951 | } | |
0683b6fa SM |
5952 | break; |
5953 | ||
4e8a9132 SM |
5954 | /* Simple loop detecting on_failure_jump: just check on the |
5955 | failure stack if the same spot was already hit earlier. */ | |
505bde11 SM |
5956 | case on_failure_jump_loop: |
5957 | on_failure: | |
5958 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
5959 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_loop %d (to %p):\n", | |
5960 | mcnt, p + mcnt); | |
6df42991 SM |
5961 | { |
5962 | int cycle = 0; | |
5963 | CHECK_INFINITE_LOOP (p - 3, d); | |
5964 | if (cycle) | |
5965 | /* If there's a cycle, get out of the loop, as if the matching | |
5966 | had failed. We used to just `goto fail' here, but that was | |
5967 | aborting the search a bit too early: we want to keep the | |
5968 | empty-loop-match and keep matching after the loop. | |
5969 | We want (x?)*y\1z to match both xxyz and xxyxz. */ | |
5970 | p += mcnt; | |
5971 | else | |
5972 | PUSH_FAILURE_POINT (p - 3, d); | |
5973 | } | |
25fe55af | 5974 | break; |
fa9a63c5 RM |
5975 | |
5976 | ||
5977 | /* Uses of on_failure_jump: | |
5e69f11e | 5978 | |
25fe55af RS |
5979 | Each alternative starts with an on_failure_jump that points |
5980 | to the beginning of the next alternative. Each alternative | |
5981 | except the last ends with a jump that in effect jumps past | |
5982 | the rest of the alternatives. (They really jump to the | |
5983 | ending jump of the following alternative, because tensioning | |
5984 | these jumps is a hassle.) | |
fa9a63c5 | 5985 | |
25fe55af RS |
5986 | Repeats start with an on_failure_jump that points past both |
5987 | the repetition text and either the following jump or | |
5988 | pop_failure_jump back to this on_failure_jump. */ | |
fa9a63c5 | 5989 | case on_failure_jump: |
25fe55af | 5990 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
505bde11 SM |
5991 | DEBUG_PRINT3 ("EXECUTING on_failure_jump %d (to %p):\n", |
5992 | mcnt, p + mcnt); | |
25fe55af | 5993 | |
505bde11 | 5994 | PUSH_FAILURE_POINT (p -3, d); |
25fe55af RS |
5995 | break; |
5996 | ||
4e8a9132 | 5997 | /* This operation is used for greedy *. |
505bde11 SM |
5998 | Compare the beginning of the repeat with what in the |
5999 | pattern follows its end. If we can establish that there | |
6000 | is nothing that they would both match, i.e., that we | |
6001 | would have to backtrack because of (as in, e.g., `a*a') | |
6002 | then we can use a non-backtracking loop based on | |
4e8a9132 | 6003 | on_failure_keep_string_jump instead of on_failure_jump. */ |
505bde11 | 6004 | case on_failure_jump_smart: |
25fe55af | 6005 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
505bde11 SM |
6006 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_smart %d (to %p).\n", |
6007 | mcnt, p + mcnt); | |
25fe55af | 6008 | { |
01618498 | 6009 | re_char *p1 = p; /* Next operation. */ |
6dcf2d0e SM |
6010 | /* Here, we discard `const', making re_match non-reentrant. */ |
6011 | unsigned char *p2 = (unsigned char*) p + mcnt; /* Jump dest. */ | |
6012 | unsigned char *p3 = (unsigned char*) p - 3; /* opcode location. */ | |
fa9a63c5 | 6013 | |
505bde11 SM |
6014 | p -= 3; /* Reset so that we will re-execute the |
6015 | instruction once it's been changed. */ | |
fa9a63c5 | 6016 | |
4e8a9132 SM |
6017 | EXTRACT_NUMBER (mcnt, p2 - 2); |
6018 | ||
6019 | /* Ensure this is a indeed the trivial kind of loop | |
6020 | we are expecting. */ | |
6021 | assert (skip_one_char (p1) == p2 - 3); | |
6022 | assert ((re_opcode_t) p2[-3] == jump && p2 + mcnt == p); | |
99633e97 | 6023 | DEBUG_STATEMENT (debug += 2); |
505bde11 | 6024 | if (mutually_exclusive_p (bufp, p1, p2)) |
fa9a63c5 | 6025 | { |
505bde11 | 6026 | /* Use a fast `on_failure_keep_string_jump' loop. */ |
4e8a9132 | 6027 | DEBUG_PRINT1 (" smart exclusive => fast loop.\n"); |
01618498 | 6028 | *p3 = (unsigned char) on_failure_keep_string_jump; |
4e8a9132 | 6029 | STORE_NUMBER (p2 - 2, mcnt + 3); |
25fe55af | 6030 | } |
505bde11 | 6031 | else |
fa9a63c5 | 6032 | { |
505bde11 SM |
6033 | /* Default to a safe `on_failure_jump' loop. */ |
6034 | DEBUG_PRINT1 (" smart default => slow loop.\n"); | |
01618498 | 6035 | *p3 = (unsigned char) on_failure_jump; |
fa9a63c5 | 6036 | } |
99633e97 | 6037 | DEBUG_STATEMENT (debug -= 2); |
25fe55af | 6038 | } |
505bde11 | 6039 | break; |
25fe55af RS |
6040 | |
6041 | /* Unconditionally jump (without popping any failure points). */ | |
6042 | case jump: | |
fa9a63c5 | 6043 | unconditional_jump: |
5b370c2b | 6044 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 6045 | EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ |
25fe55af | 6046 | DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt); |
7814e705 | 6047 | p += mcnt; /* Do the jump. */ |
505bde11 | 6048 | DEBUG_PRINT2 ("(to %p).\n", p); |
25fe55af RS |
6049 | break; |
6050 | ||
6051 | ||
25fe55af RS |
6052 | /* Have to succeed matching what follows at least n times. |
6053 | After that, handle like `on_failure_jump'. */ | |
6054 | case succeed_n: | |
01618498 | 6055 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af RS |
6056 | EXTRACT_NUMBER (mcnt, p + 2); |
6057 | DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt); | |
5e69f11e | 6058 | |
dc1e502d SM |
6059 | /* Originally, mcnt is how many times we HAVE to succeed. */ |
6060 | if (mcnt != 0) | |
25fe55af | 6061 | { |
6dcf2d0e SM |
6062 | /* Here, we discard `const', making re_match non-reentrant. */ |
6063 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 6064 | mcnt--; |
01618498 SM |
6065 | p += 4; |
6066 | PUSH_NUMBER (p2, mcnt); | |
25fe55af | 6067 | } |
dc1e502d SM |
6068 | else |
6069 | /* The two bytes encoding mcnt == 0 are two no_op opcodes. */ | |
6070 | goto on_failure; | |
25fe55af RS |
6071 | break; |
6072 | ||
6073 | case jump_n: | |
01618498 | 6074 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af RS |
6075 | EXTRACT_NUMBER (mcnt, p + 2); |
6076 | DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt); | |
6077 | ||
6078 | /* Originally, this is how many times we CAN jump. */ | |
dc1e502d | 6079 | if (mcnt != 0) |
25fe55af | 6080 | { |
6dcf2d0e SM |
6081 | /* Here, we discard `const', making re_match non-reentrant. */ |
6082 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 6083 | mcnt--; |
01618498 | 6084 | PUSH_NUMBER (p2, mcnt); |
dc1e502d | 6085 | goto unconditional_jump; |
25fe55af RS |
6086 | } |
6087 | /* If don't have to jump any more, skip over the rest of command. */ | |
5e69f11e RM |
6088 | else |
6089 | p += 4; | |
25fe55af | 6090 | break; |
5e69f11e | 6091 | |
fa9a63c5 RM |
6092 | case set_number_at: |
6093 | { | |
01618498 | 6094 | unsigned char *p2; /* Location of the counter. */ |
25fe55af | 6095 | DEBUG_PRINT1 ("EXECUTING set_number_at.\n"); |
fa9a63c5 | 6096 | |
25fe55af | 6097 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
6dcf2d0e SM |
6098 | /* Here, we discard `const', making re_match non-reentrant. */ |
6099 | p2 = (unsigned char*) p + mcnt; | |
01618498 | 6100 | /* Signedness doesn't matter since we only copy MCNT's bits . */ |
25fe55af | 6101 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
01618498 SM |
6102 | DEBUG_PRINT3 (" Setting %p to %d.\n", p2, mcnt); |
6103 | PUSH_NUMBER (p2, mcnt); | |
25fe55af RS |
6104 | break; |
6105 | } | |
9121ca40 KH |
6106 | |
6107 | case wordbound: | |
66f0296e SM |
6108 | case notwordbound: |
6109 | not = (re_opcode_t) *(p - 1) == notwordbound; | |
6110 | DEBUG_PRINT2 ("EXECUTING %swordbound.\n", not?"not":""); | |
fa9a63c5 | 6111 | |
99633e97 | 6112 | /* We SUCCEED (or FAIL) in one of the following cases: */ |
9121ca40 | 6113 | |
b18215fc | 6114 | /* Case 1: D is at the beginning or the end of string. */ |
9121ca40 | 6115 | if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) |
66f0296e | 6116 | not = !not; |
b18215fc RS |
6117 | else |
6118 | { | |
6119 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6120 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
6121 | re_wchar_t c1, c2; |
6122 | int s1, s2; | |
bf216479 | 6123 | int dummy; |
b18215fc | 6124 | #ifdef emacs |
2d1675e4 SM |
6125 | int offset = PTR_TO_OFFSET (d - 1); |
6126 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
5d967c7a | 6127 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 6128 | #endif |
66f0296e | 6129 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
b18215fc RS |
6130 | s1 = SYNTAX (c1); |
6131 | #ifdef emacs | |
5d967c7a | 6132 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
25fe55af | 6133 | #endif |
f1ad044f | 6134 | PREFETCH_NOLIMIT (); |
6fdd04b0 | 6135 | GET_CHAR_AFTER (c2, d, dummy); |
b18215fc RS |
6136 | s2 = SYNTAX (c2); |
6137 | ||
6138 | if (/* Case 2: Only one of S1 and S2 is Sword. */ | |
6139 | ((s1 == Sword) != (s2 == Sword)) | |
6140 | /* Case 3: Both of S1 and S2 are Sword, and macro | |
7814e705 | 6141 | WORD_BOUNDARY_P (C1, C2) returns nonzero. */ |
b18215fc | 6142 | || ((s1 == Sword) && WORD_BOUNDARY_P (c1, c2))) |
66f0296e SM |
6143 | not = !not; |
6144 | } | |
6145 | if (not) | |
9121ca40 | 6146 | break; |
b18215fc | 6147 | else |
9121ca40 | 6148 | goto fail; |
fa9a63c5 RM |
6149 | |
6150 | case wordbeg: | |
25fe55af | 6151 | DEBUG_PRINT1 ("EXECUTING wordbeg.\n"); |
fa9a63c5 | 6152 | |
b18215fc RS |
6153 | /* We FAIL in one of the following cases: */ |
6154 | ||
7814e705 | 6155 | /* Case 1: D is at the end of string. */ |
b18215fc | 6156 | if (AT_STRINGS_END (d)) |
99633e97 | 6157 | goto fail; |
b18215fc RS |
6158 | else |
6159 | { | |
6160 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6161 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
6162 | re_wchar_t c1, c2; |
6163 | int s1, s2; | |
bf216479 | 6164 | int dummy; |
fa9a63c5 | 6165 | #ifdef emacs |
2d1675e4 SM |
6166 | int offset = PTR_TO_OFFSET (d); |
6167 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 6168 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 6169 | #endif |
99633e97 | 6170 | PREFETCH (); |
6fdd04b0 | 6171 | GET_CHAR_AFTER (c2, d, dummy); |
b18215fc | 6172 | s2 = SYNTAX (c2); |
177c0ea7 | 6173 | |
b18215fc RS |
6174 | /* Case 2: S2 is not Sword. */ |
6175 | if (s2 != Sword) | |
6176 | goto fail; | |
6177 | ||
6178 | /* Case 3: D is not at the beginning of string ... */ | |
6179 | if (!AT_STRINGS_BEG (d)) | |
6180 | { | |
6181 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6182 | #ifdef emacs | |
5d967c7a | 6183 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); |
25fe55af | 6184 | #endif |
b18215fc RS |
6185 | s1 = SYNTAX (c1); |
6186 | ||
6187 | /* ... and S1 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6188 | returns 0. */ |
b18215fc RS |
6189 | if ((s1 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
6190 | goto fail; | |
6191 | } | |
6192 | } | |
e318085a RS |
6193 | break; |
6194 | ||
b18215fc | 6195 | case wordend: |
25fe55af | 6196 | DEBUG_PRINT1 ("EXECUTING wordend.\n"); |
b18215fc RS |
6197 | |
6198 | /* We FAIL in one of the following cases: */ | |
6199 | ||
6200 | /* Case 1: D is at the beginning of string. */ | |
6201 | if (AT_STRINGS_BEG (d)) | |
e318085a | 6202 | goto fail; |
b18215fc RS |
6203 | else |
6204 | { | |
6205 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6206 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
6207 | re_wchar_t c1, c2; |
6208 | int s1, s2; | |
bf216479 | 6209 | int dummy; |
5d967c7a | 6210 | #ifdef emacs |
2d1675e4 SM |
6211 | int offset = PTR_TO_OFFSET (d) - 1; |
6212 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 6213 | UPDATE_SYNTAX_TABLE (charpos); |
5d967c7a | 6214 | #endif |
99633e97 | 6215 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
b18215fc RS |
6216 | s1 = SYNTAX (c1); |
6217 | ||
6218 | /* Case 2: S1 is not Sword. */ | |
6219 | if (s1 != Sword) | |
6220 | goto fail; | |
6221 | ||
6222 | /* Case 3: D is not at the end of string ... */ | |
6223 | if (!AT_STRINGS_END (d)) | |
6224 | { | |
f1ad044f | 6225 | PREFETCH_NOLIMIT (); |
6fdd04b0 | 6226 | GET_CHAR_AFTER (c2, d, dummy); |
5d967c7a RS |
6227 | #ifdef emacs |
6228 | UPDATE_SYNTAX_TABLE_FORWARD (charpos); | |
6229 | #endif | |
b18215fc RS |
6230 | s2 = SYNTAX (c2); |
6231 | ||
6232 | /* ... and S2 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6233 | returns 0. */ |
b18215fc | 6234 | if ((s2 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
25fe55af | 6235 | goto fail; |
b18215fc RS |
6236 | } |
6237 | } | |
e318085a RS |
6238 | break; |
6239 | ||
669fa600 SM |
6240 | case symbeg: |
6241 | DEBUG_PRINT1 ("EXECUTING symbeg.\n"); | |
6242 | ||
6243 | /* We FAIL in one of the following cases: */ | |
6244 | ||
7814e705 | 6245 | /* Case 1: D is at the end of string. */ |
669fa600 SM |
6246 | if (AT_STRINGS_END (d)) |
6247 | goto fail; | |
6248 | else | |
6249 | { | |
6250 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6251 | is the character at D, and S2 is the syntax of C2. */ | |
6252 | re_wchar_t c1, c2; | |
6253 | int s1, s2; | |
6254 | #ifdef emacs | |
6255 | int offset = PTR_TO_OFFSET (d); | |
6256 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
6257 | UPDATE_SYNTAX_TABLE (charpos); | |
6258 | #endif | |
6259 | PREFETCH (); | |
62a6e103 | 6260 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6261 | s2 = SYNTAX (c2); |
7814e705 | 6262 | |
669fa600 SM |
6263 | /* Case 2: S2 is neither Sword nor Ssymbol. */ |
6264 | if (s2 != Sword && s2 != Ssymbol) | |
6265 | goto fail; | |
6266 | ||
6267 | /* Case 3: D is not at the beginning of string ... */ | |
6268 | if (!AT_STRINGS_BEG (d)) | |
6269 | { | |
6270 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6271 | #ifdef emacs | |
6272 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); | |
6273 | #endif | |
6274 | s1 = SYNTAX (c1); | |
6275 | ||
6276 | /* ... and S1 is Sword or Ssymbol. */ | |
6277 | if (s1 == Sword || s1 == Ssymbol) | |
6278 | goto fail; | |
6279 | } | |
6280 | } | |
6281 | break; | |
6282 | ||
6283 | case symend: | |
6284 | DEBUG_PRINT1 ("EXECUTING symend.\n"); | |
6285 | ||
6286 | /* We FAIL in one of the following cases: */ | |
6287 | ||
6288 | /* Case 1: D is at the beginning of string. */ | |
6289 | if (AT_STRINGS_BEG (d)) | |
6290 | goto fail; | |
6291 | else | |
6292 | { | |
6293 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6294 | is the character at D, and S2 is the syntax of C2. */ | |
6295 | re_wchar_t c1, c2; | |
6296 | int s1, s2; | |
6297 | #ifdef emacs | |
6298 | int offset = PTR_TO_OFFSET (d) - 1; | |
6299 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
6300 | UPDATE_SYNTAX_TABLE (charpos); | |
6301 | #endif | |
6302 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6303 | s1 = SYNTAX (c1); | |
6304 | ||
6305 | /* Case 2: S1 is neither Ssymbol nor Sword. */ | |
6306 | if (s1 != Sword && s1 != Ssymbol) | |
6307 | goto fail; | |
6308 | ||
6309 | /* Case 3: D is not at the end of string ... */ | |
6310 | if (!AT_STRINGS_END (d)) | |
6311 | { | |
6312 | PREFETCH_NOLIMIT (); | |
62a6e103 | 6313 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6314 | #ifdef emacs |
134579f2 | 6315 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
669fa600 SM |
6316 | #endif |
6317 | s2 = SYNTAX (c2); | |
6318 | ||
6319 | /* ... and S2 is Sword or Ssymbol. */ | |
6320 | if (s2 == Sword || s2 == Ssymbol) | |
6321 | goto fail; | |
b18215fc RS |
6322 | } |
6323 | } | |
e318085a RS |
6324 | break; |
6325 | ||
fa9a63c5 | 6326 | case syntaxspec: |
1fb352e0 SM |
6327 | case notsyntaxspec: |
6328 | not = (re_opcode_t) *(p - 1) == notsyntaxspec; | |
fa9a63c5 | 6329 | mcnt = *p++; |
1fb352e0 | 6330 | DEBUG_PRINT3 ("EXECUTING %ssyntaxspec %d.\n", not?"not":"", mcnt); |
fa9a63c5 | 6331 | PREFETCH (); |
b18215fc RS |
6332 | #ifdef emacs |
6333 | { | |
2d1675e4 SM |
6334 | int offset = PTR_TO_OFFSET (d); |
6335 | int pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
b18215fc RS |
6336 | UPDATE_SYNTAX_TABLE (pos1); |
6337 | } | |
25fe55af | 6338 | #endif |
b18215fc | 6339 | { |
01618498 SM |
6340 | int len; |
6341 | re_wchar_t c; | |
b18215fc | 6342 | |
6fdd04b0 | 6343 | GET_CHAR_AFTER (c, d, len); |
990b2375 | 6344 | if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not) |
1fb352e0 | 6345 | goto fail; |
b18215fc RS |
6346 | d += len; |
6347 | } | |
fa9a63c5 RM |
6348 | break; |
6349 | ||
b18215fc | 6350 | #ifdef emacs |
1fb352e0 SM |
6351 | case before_dot: |
6352 | DEBUG_PRINT1 ("EXECUTING before_dot.\n"); | |
6353 | if (PTR_BYTE_POS (d) >= PT_BYTE) | |
fa9a63c5 | 6354 | goto fail; |
b18215fc RS |
6355 | break; |
6356 | ||
1fb352e0 SM |
6357 | case at_dot: |
6358 | DEBUG_PRINT1 ("EXECUTING at_dot.\n"); | |
6359 | if (PTR_BYTE_POS (d) != PT_BYTE) | |
6360 | goto fail; | |
6361 | break; | |
b18215fc | 6362 | |
1fb352e0 SM |
6363 | case after_dot: |
6364 | DEBUG_PRINT1 ("EXECUTING after_dot.\n"); | |
6365 | if (PTR_BYTE_POS (d) <= PT_BYTE) | |
6366 | goto fail; | |
e318085a | 6367 | break; |
fa9a63c5 | 6368 | |
1fb352e0 | 6369 | case categoryspec: |
b18215fc | 6370 | case notcategoryspec: |
1fb352e0 | 6371 | not = (re_opcode_t) *(p - 1) == notcategoryspec; |
b18215fc | 6372 | mcnt = *p++; |
1fb352e0 | 6373 | DEBUG_PRINT3 ("EXECUTING %scategoryspec %d.\n", not?"not":"", mcnt); |
b18215fc RS |
6374 | PREFETCH (); |
6375 | { | |
01618498 SM |
6376 | int len; |
6377 | re_wchar_t c; | |
6378 | ||
6fdd04b0 | 6379 | GET_CHAR_AFTER (c, d, len); |
1fb352e0 | 6380 | if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not) |
b18215fc RS |
6381 | goto fail; |
6382 | d += len; | |
6383 | } | |
fa9a63c5 | 6384 | break; |
5e69f11e | 6385 | |
1fb352e0 | 6386 | #endif /* emacs */ |
5e69f11e | 6387 | |
0b32bf0e SM |
6388 | default: |
6389 | abort (); | |
fa9a63c5 | 6390 | } |
b18215fc | 6391 | continue; /* Successfully executed one pattern command; keep going. */ |
fa9a63c5 RM |
6392 | |
6393 | ||
6394 | /* We goto here if a matching operation fails. */ | |
6395 | fail: | |
5b370c2b | 6396 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 6397 | if (!FAIL_STACK_EMPTY ()) |
505bde11 | 6398 | { |
01618498 | 6399 | re_char *str, *pat; |
505bde11 | 6400 | /* A restart point is known. Restore to that state. */ |
0b32bf0e SM |
6401 | DEBUG_PRINT1 ("\nFAIL:\n"); |
6402 | POP_FAILURE_POINT (str, pat); | |
505bde11 SM |
6403 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *pat++)) |
6404 | { | |
6405 | case on_failure_keep_string_jump: | |
6406 | assert (str == NULL); | |
6407 | goto continue_failure_jump; | |
6408 | ||
0683b6fa SM |
6409 | case on_failure_jump_nastyloop: |
6410 | assert ((re_opcode_t)pat[-2] == no_op); | |
6411 | PUSH_FAILURE_POINT (pat - 2, str); | |
6412 | /* Fallthrough */ | |
6413 | ||
505bde11 SM |
6414 | case on_failure_jump_loop: |
6415 | case on_failure_jump: | |
6416 | case succeed_n: | |
6417 | d = str; | |
6418 | continue_failure_jump: | |
6419 | EXTRACT_NUMBER_AND_INCR (mcnt, pat); | |
6420 | p = pat + mcnt; | |
6421 | break; | |
b18215fc | 6422 | |
0683b6fa SM |
6423 | case no_op: |
6424 | /* A special frame used for nastyloops. */ | |
6425 | goto fail; | |
6426 | ||
505bde11 SM |
6427 | default: |
6428 | abort(); | |
6429 | } | |
fa9a63c5 | 6430 | |
505bde11 | 6431 | assert (p >= bufp->buffer && p <= pend); |
b18215fc | 6432 | |
0b32bf0e | 6433 | if (d >= string1 && d <= end1) |
fa9a63c5 | 6434 | dend = end_match_1; |
0b32bf0e | 6435 | } |
fa9a63c5 | 6436 | else |
0b32bf0e | 6437 | break; /* Matching at this starting point really fails. */ |
fa9a63c5 RM |
6438 | } /* for (;;) */ |
6439 | ||
6440 | if (best_regs_set) | |
6441 | goto restore_best_regs; | |
6442 | ||
6443 | FREE_VARIABLES (); | |
6444 | ||
b18215fc | 6445 | return -1; /* Failure to match. */ |
fa9a63c5 RM |
6446 | } /* re_match_2 */ |
6447 | \f | |
6448 | /* Subroutine definitions for re_match_2. */ | |
6449 | ||
fa9a63c5 RM |
6450 | /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN |
6451 | bytes; nonzero otherwise. */ | |
5e69f11e | 6452 | |
fa9a63c5 | 6453 | static int |
02cb78b5 | 6454 | bcmp_translate (s1, s2, len, translate, target_multibyte) |
2d1675e4 | 6455 | re_char *s1, *s2; |
fa9a63c5 | 6456 | register int len; |
6676cb1c | 6457 | RE_TRANSLATE_TYPE translate; |
02cb78b5 | 6458 | const int target_multibyte; |
fa9a63c5 | 6459 | { |
2d1675e4 SM |
6460 | register re_char *p1 = s1, *p2 = s2; |
6461 | re_char *p1_end = s1 + len; | |
6462 | re_char *p2_end = s2 + len; | |
e934739e | 6463 | |
4bb91c68 SM |
6464 | /* FIXME: Checking both p1 and p2 presumes that the two strings might have |
6465 | different lengths, but relying on a single `len' would break this. -sm */ | |
6466 | while (p1 < p1_end && p2 < p2_end) | |
fa9a63c5 | 6467 | { |
e934739e | 6468 | int p1_charlen, p2_charlen; |
01618498 | 6469 | re_wchar_t p1_ch, p2_ch; |
e934739e | 6470 | |
6fdd04b0 KH |
6471 | GET_CHAR_AFTER (p1_ch, p1, p1_charlen); |
6472 | GET_CHAR_AFTER (p2_ch, p2, p2_charlen); | |
e934739e RS |
6473 | |
6474 | if (RE_TRANSLATE (translate, p1_ch) | |
6475 | != RE_TRANSLATE (translate, p2_ch)) | |
bc192b5b | 6476 | return 1; |
e934739e RS |
6477 | |
6478 | p1 += p1_charlen, p2 += p2_charlen; | |
fa9a63c5 | 6479 | } |
e934739e RS |
6480 | |
6481 | if (p1 != p1_end || p2 != p2_end) | |
6482 | return 1; | |
6483 | ||
fa9a63c5 RM |
6484 | return 0; |
6485 | } | |
6486 | \f | |
6487 | /* Entry points for GNU code. */ | |
6488 | ||
6489 | /* re_compile_pattern is the GNU regular expression compiler: it | |
6490 | compiles PATTERN (of length SIZE) and puts the result in BUFP. | |
6491 | Returns 0 if the pattern was valid, otherwise an error string. | |
5e69f11e | 6492 | |
fa9a63c5 RM |
6493 | Assumes the `allocated' (and perhaps `buffer') and `translate' fields |
6494 | are set in BUFP on entry. | |
5e69f11e | 6495 | |
b18215fc | 6496 | We call regex_compile to do the actual compilation. */ |
fa9a63c5 RM |
6497 | |
6498 | const char * | |
6499 | re_compile_pattern (pattern, length, bufp) | |
6500 | const char *pattern; | |
0b32bf0e | 6501 | size_t length; |
fa9a63c5 RM |
6502 | struct re_pattern_buffer *bufp; |
6503 | { | |
6504 | reg_errcode_t ret; | |
5e69f11e | 6505 | |
fa9a63c5 RM |
6506 | /* GNU code is written to assume at least RE_NREGS registers will be set |
6507 | (and at least one extra will be -1). */ | |
6508 | bufp->regs_allocated = REGS_UNALLOCATED; | |
5e69f11e | 6509 | |
fa9a63c5 RM |
6510 | /* And GNU code determines whether or not to get register information |
6511 | by passing null for the REGS argument to re_match, etc., not by | |
6512 | setting no_sub. */ | |
6513 | bufp->no_sub = 0; | |
5e69f11e | 6514 | |
4bb91c68 | 6515 | ret = regex_compile ((re_char*) pattern, length, re_syntax_options, bufp); |
fa9a63c5 RM |
6516 | |
6517 | if (!ret) | |
6518 | return NULL; | |
6519 | return gettext (re_error_msgid[(int) ret]); | |
5e69f11e | 6520 | } |
c0f9ea08 | 6521 | WEAK_ALIAS (__re_compile_pattern, re_compile_pattern) |
fa9a63c5 | 6522 | \f |
b18215fc RS |
6523 | /* Entry points compatible with 4.2 BSD regex library. We don't define |
6524 | them unless specifically requested. */ | |
fa9a63c5 | 6525 | |
0b32bf0e | 6526 | #if defined _REGEX_RE_COMP || defined _LIBC |
fa9a63c5 RM |
6527 | |
6528 | /* BSD has one and only one pattern buffer. */ | |
6529 | static struct re_pattern_buffer re_comp_buf; | |
6530 | ||
6531 | char * | |
0b32bf0e | 6532 | # ifdef _LIBC |
48afdd44 RM |
6533 | /* Make these definitions weak in libc, so POSIX programs can redefine |
6534 | these names if they don't use our functions, and still use | |
6535 | regcomp/regexec below without link errors. */ | |
6536 | weak_function | |
0b32bf0e | 6537 | # endif |
fa9a63c5 RM |
6538 | re_comp (s) |
6539 | const char *s; | |
6540 | { | |
6541 | reg_errcode_t ret; | |
5e69f11e | 6542 | |
fa9a63c5 RM |
6543 | if (!s) |
6544 | { | |
6545 | if (!re_comp_buf.buffer) | |
0b32bf0e | 6546 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
a60198e5 | 6547 | return (char *) gettext ("No previous regular expression"); |
fa9a63c5 RM |
6548 | return 0; |
6549 | } | |
6550 | ||
6551 | if (!re_comp_buf.buffer) | |
6552 | { | |
6553 | re_comp_buf.buffer = (unsigned char *) malloc (200); | |
6554 | if (re_comp_buf.buffer == NULL) | |
0b32bf0e SM |
6555 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6556 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6557 | re_comp_buf.allocated = 200; |
6558 | ||
6559 | re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH); | |
6560 | if (re_comp_buf.fastmap == NULL) | |
a60198e5 SM |
6561 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6562 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6563 | } |
6564 | ||
6565 | /* Since `re_exec' always passes NULL for the `regs' argument, we | |
6566 | don't need to initialize the pattern buffer fields which affect it. */ | |
6567 | ||
fa9a63c5 | 6568 | ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); |
5e69f11e | 6569 | |
fa9a63c5 RM |
6570 | if (!ret) |
6571 | return NULL; | |
6572 | ||
6573 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | |
6574 | return (char *) gettext (re_error_msgid[(int) ret]); | |
6575 | } | |
6576 | ||
6577 | ||
6578 | int | |
0b32bf0e | 6579 | # ifdef _LIBC |
48afdd44 | 6580 | weak_function |
0b32bf0e | 6581 | # endif |
fa9a63c5 RM |
6582 | re_exec (s) |
6583 | const char *s; | |
6584 | { | |
6585 | const int len = strlen (s); | |
6586 | return | |
6587 | 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0); | |
6588 | } | |
6589 | #endif /* _REGEX_RE_COMP */ | |
6590 | \f | |
6591 | /* POSIX.2 functions. Don't define these for Emacs. */ | |
6592 | ||
6593 | #ifndef emacs | |
6594 | ||
6595 | /* regcomp takes a regular expression as a string and compiles it. | |
6596 | ||
b18215fc | 6597 | PREG is a regex_t *. We do not expect any fields to be initialized, |
fa9a63c5 RM |
6598 | since POSIX says we shouldn't. Thus, we set |
6599 | ||
6600 | `buffer' to the compiled pattern; | |
6601 | `used' to the length of the compiled pattern; | |
6602 | `syntax' to RE_SYNTAX_POSIX_EXTENDED if the | |
6603 | REG_EXTENDED bit in CFLAGS is set; otherwise, to | |
6604 | RE_SYNTAX_POSIX_BASIC; | |
c0f9ea08 SM |
6605 | `fastmap' to an allocated space for the fastmap; |
6606 | `fastmap_accurate' to zero; | |
fa9a63c5 RM |
6607 | `re_nsub' to the number of subexpressions in PATTERN. |
6608 | ||
6609 | PATTERN is the address of the pattern string. | |
6610 | ||
6611 | CFLAGS is a series of bits which affect compilation. | |
6612 | ||
6613 | If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we | |
6614 | use POSIX basic syntax. | |
6615 | ||
6616 | If REG_NEWLINE is set, then . and [^...] don't match newline. | |
6617 | Also, regexec will try a match beginning after every newline. | |
6618 | ||
6619 | If REG_ICASE is set, then we considers upper- and lowercase | |
6620 | versions of letters to be equivalent when matching. | |
6621 | ||
6622 | If REG_NOSUB is set, then when PREG is passed to regexec, that | |
6623 | routine will report only success or failure, and nothing about the | |
6624 | registers. | |
6625 | ||
b18215fc | 6626 | It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for |
fa9a63c5 RM |
6627 | the return codes and their meanings.) */ |
6628 | ||
6629 | int | |
6630 | regcomp (preg, pattern, cflags) | |
ada30c0e SM |
6631 | regex_t *__restrict preg; |
6632 | const char *__restrict pattern; | |
fa9a63c5 RM |
6633 | int cflags; |
6634 | { | |
6635 | reg_errcode_t ret; | |
4bb91c68 | 6636 | reg_syntax_t syntax |
fa9a63c5 RM |
6637 | = (cflags & REG_EXTENDED) ? |
6638 | RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; | |
6639 | ||
6640 | /* regex_compile will allocate the space for the compiled pattern. */ | |
6641 | preg->buffer = 0; | |
6642 | preg->allocated = 0; | |
6643 | preg->used = 0; | |
5e69f11e | 6644 | |
c0f9ea08 SM |
6645 | /* Try to allocate space for the fastmap. */ |
6646 | preg->fastmap = (char *) malloc (1 << BYTEWIDTH); | |
5e69f11e | 6647 | |
fa9a63c5 RM |
6648 | if (cflags & REG_ICASE) |
6649 | { | |
6650 | unsigned i; | |
5e69f11e | 6651 | |
6676cb1c RS |
6652 | preg->translate |
6653 | = (RE_TRANSLATE_TYPE) malloc (CHAR_SET_SIZE | |
6654 | * sizeof (*(RE_TRANSLATE_TYPE)0)); | |
fa9a63c5 | 6655 | if (preg->translate == NULL) |
0b32bf0e | 6656 | return (int) REG_ESPACE; |
fa9a63c5 RM |
6657 | |
6658 | /* Map uppercase characters to corresponding lowercase ones. */ | |
6659 | for (i = 0; i < CHAR_SET_SIZE; i++) | |
4bb91c68 | 6660 | preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i; |
fa9a63c5 RM |
6661 | } |
6662 | else | |
6663 | preg->translate = NULL; | |
6664 | ||
6665 | /* If REG_NEWLINE is set, newlines are treated differently. */ | |
6666 | if (cflags & REG_NEWLINE) | |
6667 | { /* REG_NEWLINE implies neither . nor [^...] match newline. */ | |
6668 | syntax &= ~RE_DOT_NEWLINE; | |
6669 | syntax |= RE_HAT_LISTS_NOT_NEWLINE; | |
fa9a63c5 RM |
6670 | } |
6671 | else | |
c0f9ea08 | 6672 | syntax |= RE_NO_NEWLINE_ANCHOR; |
fa9a63c5 RM |
6673 | |
6674 | preg->no_sub = !!(cflags & REG_NOSUB); | |
6675 | ||
5e69f11e | 6676 | /* POSIX says a null character in the pattern terminates it, so we |
fa9a63c5 | 6677 | can use strlen here in compiling the pattern. */ |
4bb91c68 | 6678 | ret = regex_compile ((re_char*) pattern, strlen (pattern), syntax, preg); |
5e69f11e | 6679 | |
fa9a63c5 RM |
6680 | /* POSIX doesn't distinguish between an unmatched open-group and an |
6681 | unmatched close-group: both are REG_EPAREN. */ | |
c0f9ea08 SM |
6682 | if (ret == REG_ERPAREN) |
6683 | ret = REG_EPAREN; | |
6684 | ||
6685 | if (ret == REG_NOERROR && preg->fastmap) | |
6686 | { /* Compute the fastmap now, since regexec cannot modify the pattern | |
6687 | buffer. */ | |
6688 | re_compile_fastmap (preg); | |
6689 | if (preg->can_be_null) | |
6690 | { /* The fastmap can't be used anyway. */ | |
6691 | free (preg->fastmap); | |
6692 | preg->fastmap = NULL; | |
6693 | } | |
6694 | } | |
fa9a63c5 RM |
6695 | return (int) ret; |
6696 | } | |
c0f9ea08 | 6697 | WEAK_ALIAS (__regcomp, regcomp) |
fa9a63c5 RM |
6698 | |
6699 | ||
6700 | /* regexec searches for a given pattern, specified by PREG, in the | |
6701 | string STRING. | |
5e69f11e | 6702 | |
fa9a63c5 | 6703 | If NMATCH is zero or REG_NOSUB was set in the cflags argument to |
b18215fc | 6704 | `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at |
fa9a63c5 RM |
6705 | least NMATCH elements, and we set them to the offsets of the |
6706 | corresponding matched substrings. | |
5e69f11e | 6707 | |
fa9a63c5 RM |
6708 | EFLAGS specifies `execution flags' which affect matching: if |
6709 | REG_NOTBOL is set, then ^ does not match at the beginning of the | |
6710 | string; if REG_NOTEOL is set, then $ does not match at the end. | |
5e69f11e | 6711 | |
fa9a63c5 RM |
6712 | We return 0 if we find a match and REG_NOMATCH if not. */ |
6713 | ||
6714 | int | |
6715 | regexec (preg, string, nmatch, pmatch, eflags) | |
ada30c0e SM |
6716 | const regex_t *__restrict preg; |
6717 | const char *__restrict string; | |
5e69f11e | 6718 | size_t nmatch; |
9f2dbe01 | 6719 | regmatch_t pmatch[__restrict_arr]; |
fa9a63c5 RM |
6720 | int eflags; |
6721 | { | |
6722 | int ret; | |
6723 | struct re_registers regs; | |
6724 | regex_t private_preg; | |
6725 | int len = strlen (string); | |
c0f9ea08 | 6726 | boolean want_reg_info = !preg->no_sub && nmatch > 0 && pmatch; |
fa9a63c5 RM |
6727 | |
6728 | private_preg = *preg; | |
5e69f11e | 6729 | |
fa9a63c5 RM |
6730 | private_preg.not_bol = !!(eflags & REG_NOTBOL); |
6731 | private_preg.not_eol = !!(eflags & REG_NOTEOL); | |
5e69f11e | 6732 | |
fa9a63c5 RM |
6733 | /* The user has told us exactly how many registers to return |
6734 | information about, via `nmatch'. We have to pass that on to the | |
b18215fc | 6735 | matching routines. */ |
fa9a63c5 | 6736 | private_preg.regs_allocated = REGS_FIXED; |
5e69f11e | 6737 | |
fa9a63c5 RM |
6738 | if (want_reg_info) |
6739 | { | |
6740 | regs.num_regs = nmatch; | |
4bb91c68 SM |
6741 | regs.start = TALLOC (nmatch * 2, regoff_t); |
6742 | if (regs.start == NULL) | |
0b32bf0e | 6743 | return (int) REG_NOMATCH; |
4bb91c68 | 6744 | regs.end = regs.start + nmatch; |
fa9a63c5 RM |
6745 | } |
6746 | ||
c0f9ea08 SM |
6747 | /* Instead of using not_eol to implement REG_NOTEOL, we could simply |
6748 | pass (&private_preg, string, len + 1, 0, len, ...) pretending the string | |
6749 | was a little bit longer but still only matching the real part. | |
6750 | This works because the `endline' will check for a '\n' and will find a | |
6751 | '\0', correctly deciding that this is not the end of a line. | |
6752 | But it doesn't work out so nicely for REG_NOTBOL, since we don't have | |
6753 | a convenient '\0' there. For all we know, the string could be preceded | |
6754 | by '\n' which would throw things off. */ | |
6755 | ||
fa9a63c5 RM |
6756 | /* Perform the searching operation. */ |
6757 | ret = re_search (&private_preg, string, len, | |
0b32bf0e SM |
6758 | /* start: */ 0, /* range: */ len, |
6759 | want_reg_info ? ®s : (struct re_registers *) 0); | |
5e69f11e | 6760 | |
fa9a63c5 RM |
6761 | /* Copy the register information to the POSIX structure. */ |
6762 | if (want_reg_info) | |
6763 | { | |
6764 | if (ret >= 0) | |
0b32bf0e SM |
6765 | { |
6766 | unsigned r; | |
fa9a63c5 | 6767 | |
0b32bf0e SM |
6768 | for (r = 0; r < nmatch; r++) |
6769 | { | |
6770 | pmatch[r].rm_so = regs.start[r]; | |
6771 | pmatch[r].rm_eo = regs.end[r]; | |
6772 | } | |
6773 | } | |
fa9a63c5 | 6774 | |
b18215fc | 6775 | /* If we needed the temporary register info, free the space now. */ |
fa9a63c5 | 6776 | free (regs.start); |
fa9a63c5 RM |
6777 | } |
6778 | ||
6779 | /* We want zero return to mean success, unlike `re_search'. */ | |
6780 | return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH; | |
6781 | } | |
c0f9ea08 | 6782 | WEAK_ALIAS (__regexec, regexec) |
fa9a63c5 RM |
6783 | |
6784 | ||
ec869672 JR |
6785 | /* Returns a message corresponding to an error code, ERR_CODE, returned |
6786 | from either regcomp or regexec. We don't use PREG here. | |
6787 | ||
6788 | ERR_CODE was previously called ERRCODE, but that name causes an | |
6789 | error with msvc8 compiler. */ | |
fa9a63c5 RM |
6790 | |
6791 | size_t | |
ec869672 JR |
6792 | regerror (err_code, preg, errbuf, errbuf_size) |
6793 | int err_code; | |
fa9a63c5 RM |
6794 | const regex_t *preg; |
6795 | char *errbuf; | |
6796 | size_t errbuf_size; | |
6797 | { | |
6798 | const char *msg; | |
6799 | size_t msg_size; | |
6800 | ||
ec869672 JR |
6801 | if (err_code < 0 |
6802 | || err_code >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0]))) | |
5e69f11e | 6803 | /* Only error codes returned by the rest of the code should be passed |
b18215fc | 6804 | to this routine. If we are given anything else, or if other regex |
fa9a63c5 RM |
6805 | code generates an invalid error code, then the program has a bug. |
6806 | Dump core so we can fix it. */ | |
6807 | abort (); | |
6808 | ||
ec869672 | 6809 | msg = gettext (re_error_msgid[err_code]); |
fa9a63c5 RM |
6810 | |
6811 | msg_size = strlen (msg) + 1; /* Includes the null. */ | |
5e69f11e | 6812 | |
fa9a63c5 RM |
6813 | if (errbuf_size != 0) |
6814 | { | |
6815 | if (msg_size > errbuf_size) | |
0b32bf0e SM |
6816 | { |
6817 | strncpy (errbuf, msg, errbuf_size - 1); | |
6818 | errbuf[errbuf_size - 1] = 0; | |
6819 | } | |
fa9a63c5 | 6820 | else |
0b32bf0e | 6821 | strcpy (errbuf, msg); |
fa9a63c5 RM |
6822 | } |
6823 | ||
6824 | return msg_size; | |
6825 | } | |
c0f9ea08 | 6826 | WEAK_ALIAS (__regerror, regerror) |
fa9a63c5 RM |
6827 | |
6828 | ||
6829 | /* Free dynamically allocated space used by PREG. */ | |
6830 | ||
6831 | void | |
6832 | regfree (preg) | |
6833 | regex_t *preg; | |
6834 | { | |
c2cd06e6 | 6835 | free (preg->buffer); |
fa9a63c5 | 6836 | preg->buffer = NULL; |
5e69f11e | 6837 | |
fa9a63c5 RM |
6838 | preg->allocated = 0; |
6839 | preg->used = 0; | |
6840 | ||
c2cd06e6 | 6841 | free (preg->fastmap); |
fa9a63c5 RM |
6842 | preg->fastmap = NULL; |
6843 | preg->fastmap_accurate = 0; | |
6844 | ||
c2cd06e6 | 6845 | free (preg->translate); |
fa9a63c5 RM |
6846 | preg->translate = NULL; |
6847 | } | |
c0f9ea08 | 6848 | WEAK_ALIAS (__regfree, regfree) |
fa9a63c5 RM |
6849 | |
6850 | #endif /* not emacs */ | |
839966f3 KH |
6851 | |
6852 | /* arch-tag: 4ffd68ba-2a9e-435b-a21a-018990f9eeb2 | |
6853 | (do not change this comment) */ |