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 | ||
ba318903 | 5 | Copyright (C) 1993-2014 Free Software Foundation, Inc. |
bc78d348 | 6 | |
fa9a63c5 RM |
7 | This program is free software; you can redistribute it and/or modify |
8 | it under the terms of the GNU General Public License as published by | |
e468b87f | 9 | the Free Software Foundation; either version 3, or (at your option) |
fa9a63c5 RM |
10 | any later version. |
11 | ||
12 | This program is distributed in the hope that it will be useful, | |
13 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
7814e705 | 14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
fa9a63c5 RM |
15 | GNU General Public License for more details. |
16 | ||
17 | You should have received a copy of the GNU General Public License | |
fee0bd5f | 18 | along with this program. If not, see <http://www.gnu.org/licenses/>. */ |
fa9a63c5 | 19 | |
6df42991 | 20 | /* TODO: |
505bde11 | 21 | - structure the opcode space into opcode+flag. |
dc1e502d | 22 | - merge with glibc's regex.[ch]. |
01618498 | 23 | - replace (succeed_n + jump_n + set_number_at) with something that doesn't |
6dcf2d0e SM |
24 | need to modify the compiled regexp so that re_match can be reentrant. |
25 | - get rid of on_failure_jump_smart by doing the optimization in re_comp | |
26 | rather than at run-time, so that re_match can be reentrant. | |
01618498 | 27 | */ |
505bde11 | 28 | |
b7432bb2 | 29 | /* AIX requires this to be the first thing in the file. */ |
0b32bf0e | 30 | #if defined _AIX && !defined REGEX_MALLOC |
fa9a63c5 RM |
31 | #pragma alloca |
32 | #endif | |
33 | ||
b8df54ff PE |
34 | /* Ignore some GCC warnings for now. This section should go away |
35 | once the Emacs and Gnulib regex code is merged. */ | |
31ff141c | 36 | #if 4 < __GNUC__ + (5 <= __GNUC_MINOR__) || defined __clang__ |
b8df54ff PE |
37 | # pragma GCC diagnostic ignored "-Wstrict-overflow" |
38 | # ifndef emacs | |
b8df54ff PE |
39 | # pragma GCC diagnostic ignored "-Wunused-function" |
40 | # pragma GCC diagnostic ignored "-Wunused-macros" | |
41 | # pragma GCC diagnostic ignored "-Wunused-result" | |
42 | # pragma GCC diagnostic ignored "-Wunused-variable" | |
43 | # endif | |
44 | #endif | |
45 | ||
a5d376b0 | 46 | #if 4 < __GNUC__ + (6 <= __GNUC_MINOR__) && ! defined __clang__ |
31ff141c PE |
47 | # pragma GCC diagnostic ignored "-Wunused-but-set-variable" |
48 | #endif | |
49 | ||
cf38a720 | 50 | #include <config.h> |
fa9a63c5 | 51 | |
0e926e56 PE |
52 | #include <stddef.h> |
53 | ||
54 | #ifdef emacs | |
4bb91c68 SM |
55 | /* We need this for `regex.h', and perhaps for the Emacs include files. */ |
56 | # include <sys/types.h> | |
57 | #endif | |
fa9a63c5 | 58 | |
14473664 SM |
59 | /* Whether to use ISO C Amendment 1 wide char functions. |
60 | Those should not be used for Emacs since it uses its own. */ | |
5e5388f6 GM |
61 | #if defined _LIBC |
62 | #define WIDE_CHAR_SUPPORT 1 | |
63 | #else | |
14473664 | 64 | #define WIDE_CHAR_SUPPORT \ |
5e5388f6 GM |
65 | (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC && !emacs) |
66 | #endif | |
14473664 | 67 | |
fa463103 | 68 | /* For platform which support the ISO C amendment 1 functionality we |
14473664 | 69 | support user defined character classes. */ |
a0ad02f7 | 70 | #if WIDE_CHAR_SUPPORT |
14473664 SM |
71 | /* Solaris 2.5 has a bug: <wchar.h> must be included before <wctype.h>. */ |
72 | # include <wchar.h> | |
73 | # include <wctype.h> | |
74 | #endif | |
75 | ||
c0f9ea08 SM |
76 | #ifdef _LIBC |
77 | /* We have to keep the namespace clean. */ | |
78 | # define regfree(preg) __regfree (preg) | |
79 | # define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef) | |
80 | # define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags) | |
ec869672 | 81 | # define regerror(err_code, preg, errbuf, errbuf_size) \ |
5e617bc2 | 82 | __regerror (err_code, preg, errbuf, errbuf_size) |
c0f9ea08 SM |
83 | # define re_set_registers(bu, re, nu, st, en) \ |
84 | __re_set_registers (bu, re, nu, st, en) | |
85 | # define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \ | |
86 | __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) | |
87 | # define re_match(bufp, string, size, pos, regs) \ | |
88 | __re_match (bufp, string, size, pos, regs) | |
89 | # define re_search(bufp, string, size, startpos, range, regs) \ | |
90 | __re_search (bufp, string, size, startpos, range, regs) | |
91 | # define re_compile_pattern(pattern, length, bufp) \ | |
92 | __re_compile_pattern (pattern, length, bufp) | |
93 | # define re_set_syntax(syntax) __re_set_syntax (syntax) | |
94 | # define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \ | |
95 | __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop) | |
96 | # define re_compile_fastmap(bufp) __re_compile_fastmap (bufp) | |
97 | ||
14473664 SM |
98 | /* Make sure we call libc's function even if the user overrides them. */ |
99 | # define btowc __btowc | |
100 | # define iswctype __iswctype | |
101 | # define wctype __wctype | |
102 | ||
c0f9ea08 SM |
103 | # define WEAK_ALIAS(a,b) weak_alias (a, b) |
104 | ||
105 | /* We are also using some library internals. */ | |
106 | # include <locale/localeinfo.h> | |
107 | # include <locale/elem-hash.h> | |
108 | # include <langinfo.h> | |
109 | #else | |
110 | # define WEAK_ALIAS(a,b) | |
111 | #endif | |
112 | ||
4bb91c68 | 113 | /* This is for other GNU distributions with internationalized messages. */ |
0b32bf0e | 114 | #if HAVE_LIBINTL_H || defined _LIBC |
fa9a63c5 RM |
115 | # include <libintl.h> |
116 | #else | |
117 | # define gettext(msgid) (msgid) | |
118 | #endif | |
119 | ||
5e69f11e RM |
120 | #ifndef gettext_noop |
121 | /* This define is so xgettext can find the internationalizable | |
122 | strings. */ | |
0b32bf0e | 123 | # define gettext_noop(String) String |
5e69f11e RM |
124 | #endif |
125 | ||
fa9a63c5 RM |
126 | /* The `emacs' switch turns on certain matching commands |
127 | that make sense only in Emacs. */ | |
128 | #ifdef emacs | |
129 | ||
0b32bf0e | 130 | # include "lisp.h" |
e5560ff7 | 131 | # include "character.h" |
0b32bf0e | 132 | # include "buffer.h" |
b18215fc | 133 | |
0b32bf0e | 134 | # include "syntax.h" |
0b32bf0e | 135 | # include "category.h" |
fa9a63c5 | 136 | |
c6cfd910 PE |
137 | /* Make syntax table lookup grant data in gl_state. */ |
138 | # define SYNTAX(c) syntax_property (c, 1) | |
139 | ||
7689ef0b EZ |
140 | # ifdef malloc |
141 | # undef malloc | |
142 | # endif | |
0b32bf0e | 143 | # define malloc xmalloc |
7689ef0b EZ |
144 | # ifdef realloc |
145 | # undef realloc | |
146 | # endif | |
0b32bf0e | 147 | # define realloc xrealloc |
7689ef0b EZ |
148 | # ifdef free |
149 | # undef free | |
150 | # endif | |
0b32bf0e | 151 | # define free xfree |
9abbd165 | 152 | |
7814e705 | 153 | /* Converts the pointer to the char to BEG-based offset from the start. */ |
0b32bf0e SM |
154 | # define PTR_TO_OFFSET(d) POS_AS_IN_BUFFER (POINTER_TO_OFFSET (d)) |
155 | # define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object))) | |
156 | ||
157 | # define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte) | |
bf216479 | 158 | # define RE_TARGET_MULTIBYTE_P(bufp) ((bufp)->target_multibyte) |
62a6e103 AS |
159 | # define RE_STRING_CHAR(p, multibyte) \ |
160 | (multibyte ? (STRING_CHAR (p)) : (*(p))) | |
161 | # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) \ | |
162 | (multibyte ? (STRING_CHAR_AND_LENGTH (p, len)) : ((len) = 1, *(p))) | |
2d1675e4 | 163 | |
4c0354d7 | 164 | # define RE_CHAR_TO_MULTIBYTE(c) UNIBYTE_TO_CHAR (c) |
cf9c99bc | 165 | |
2afc21f5 | 166 | # define RE_CHAR_TO_UNIBYTE(c) CHAR_TO_BYTE_SAFE (c) |
cf9c99bc | 167 | |
6fdd04b0 KH |
168 | /* Set C a (possibly converted to multibyte) character before P. P |
169 | points into a string which is the virtual concatenation of STR1 | |
170 | (which ends at END1) or STR2 (which ends at END2). */ | |
bf216479 KH |
171 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
172 | do { \ | |
02cb78b5 | 173 | if (target_multibyte) \ |
bf216479 KH |
174 | { \ |
175 | re_char *dtemp = (p) == (str2) ? (end1) : (p); \ | |
176 | re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \ | |
177 | while (dtemp-- > dlimit && !CHAR_HEAD_P (*dtemp)); \ | |
62a6e103 | 178 | c = STRING_CHAR (dtemp); \ |
bf216479 KH |
179 | } \ |
180 | else \ | |
181 | { \ | |
182 | (c = ((p) == (str2) ? (end1) : (p))[-1]); \ | |
cf9c99bc | 183 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
bf216479 | 184 | } \ |
2d1675e4 SM |
185 | } while (0) |
186 | ||
6fdd04b0 KH |
187 | /* Set C a (possibly converted to multibyte) character at P, and set |
188 | LEN to the byte length of that character. */ | |
189 | # define GET_CHAR_AFTER(c, p, len) \ | |
190 | do { \ | |
02cb78b5 | 191 | if (target_multibyte) \ |
62a6e103 | 192 | (c) = STRING_CHAR_AND_LENGTH (p, len); \ |
6fdd04b0 KH |
193 | else \ |
194 | { \ | |
cf9c99bc | 195 | (c) = *p; \ |
6fdd04b0 | 196 | len = 1; \ |
cf9c99bc | 197 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
6fdd04b0 | 198 | } \ |
8f924df7 | 199 | } while (0) |
4e8a9132 | 200 | |
fa9a63c5 RM |
201 | #else /* not emacs */ |
202 | ||
203 | /* If we are not linking with Emacs proper, | |
204 | we can't use the relocating allocator | |
205 | even if config.h says that we can. */ | |
0b32bf0e | 206 | # undef REL_ALLOC |
fa9a63c5 | 207 | |
4004364e | 208 | # include <unistd.h> |
fa9a63c5 | 209 | |
a77f947b CY |
210 | /* When used in Emacs's lib-src, we need xmalloc and xrealloc. */ |
211 | ||
b8df54ff | 212 | static void * |
d2762c86 | 213 | xmalloc (size_t size) |
a77f947b | 214 | { |
38182d90 | 215 | void *val = malloc (size); |
a77f947b CY |
216 | if (!val && size) |
217 | { | |
218 | write (2, "virtual memory exhausted\n", 25); | |
219 | exit (1); | |
220 | } | |
221 | return val; | |
222 | } | |
223 | ||
b8df54ff | 224 | static void * |
d2762c86 | 225 | xrealloc (void *block, size_t size) |
a77f947b | 226 | { |
38182d90 | 227 | void *val; |
a77f947b CY |
228 | /* We must call malloc explicitly when BLOCK is 0, since some |
229 | reallocs don't do this. */ | |
230 | if (! block) | |
38182d90 | 231 | val = malloc (size); |
a77f947b | 232 | else |
38182d90 | 233 | val = realloc (block, size); |
a77f947b CY |
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 | ||
f5d9e83a | 251 | # include <stdbool.h> |
9cfdb3ec | 252 | # include <string.h> |
fa9a63c5 RM |
253 | |
254 | /* Define the syntax stuff for \<, \>, etc. */ | |
255 | ||
990b2375 | 256 | /* Sword must be nonzero for the wordchar pattern commands in re_match_2. */ |
669fa600 | 257 | enum syntaxcode { Swhitespace = 0, Sword = 1, Ssymbol = 2 }; |
fa9a63c5 | 258 | |
e934739e | 259 | /* Dummy macros for non-Emacs environments. */ |
0b32bf0e SM |
260 | # define MAX_MULTIBYTE_LENGTH 1 |
261 | # define RE_MULTIBYTE_P(x) 0 | |
bf216479 | 262 | # define RE_TARGET_MULTIBYTE_P(x) 0 |
0b32bf0e | 263 | # define WORD_BOUNDARY_P(c1, c2) (0) |
aa3830c4 | 264 | # define BYTES_BY_CHAR_HEAD(p) (1) |
70806df6 | 265 | # define PREV_CHAR_BOUNDARY(p, limit) ((p)--) |
62a6e103 AS |
266 | # define STRING_CHAR(p) (*(p)) |
267 | # define RE_STRING_CHAR(p, multibyte) STRING_CHAR (p) | |
0b32bf0e | 268 | # define CHAR_STRING(c, s) (*(s) = (c), 1) |
62a6e103 AS |
269 | # define STRING_CHAR_AND_LENGTH(p, actual_len) ((actual_len) = 1, *(p)) |
270 | # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) STRING_CHAR_AND_LENGTH (p, len) | |
cf9c99bc KH |
271 | # define RE_CHAR_TO_MULTIBYTE(c) (c) |
272 | # define RE_CHAR_TO_UNIBYTE(c) (c) | |
0b32bf0e | 273 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
b18215fc | 274 | (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1))) |
6fdd04b0 KH |
275 | # define GET_CHAR_AFTER(c, p, len) \ |
276 | (c = *p, len = 1) | |
9117d724 | 277 | # define CHAR_BYTE8_P(c) (0) |
bf216479 | 278 | # define CHAR_LEADING_CODE(c) (c) |
8f924df7 | 279 | |
fa9a63c5 | 280 | #endif /* not emacs */ |
4e8a9132 SM |
281 | |
282 | #ifndef RE_TRANSLATE | |
0b32bf0e SM |
283 | # define RE_TRANSLATE(TBL, C) ((unsigned char)(TBL)[C]) |
284 | # define RE_TRANSLATE_P(TBL) (TBL) | |
4e8a9132 | 285 | #endif |
fa9a63c5 RM |
286 | \f |
287 | /* Get the interface, including the syntax bits. */ | |
288 | #include "regex.h" | |
289 | ||
f71b19b6 DL |
290 | /* isalpha etc. are used for the character classes. */ |
291 | #include <ctype.h> | |
fa9a63c5 | 292 | |
f71b19b6 | 293 | #ifdef emacs |
fa9a63c5 | 294 | |
f71b19b6 | 295 | /* 1 if C is an ASCII character. */ |
0b32bf0e | 296 | # define IS_REAL_ASCII(c) ((c) < 0200) |
fa9a63c5 | 297 | |
f71b19b6 | 298 | /* 1 if C is a unibyte character. */ |
0b32bf0e | 299 | # define ISUNIBYTE(c) (SINGLE_BYTE_CHAR_P ((c))) |
96cc36cc | 300 | |
f71b19b6 | 301 | /* The Emacs definitions should not be directly affected by locales. */ |
96cc36cc | 302 | |
f71b19b6 | 303 | /* In Emacs, these are only used for single-byte characters. */ |
0b32bf0e SM |
304 | # define ISDIGIT(c) ((c) >= '0' && (c) <= '9') |
305 | # define ISCNTRL(c) ((c) < ' ') | |
306 | # define ISXDIGIT(c) (((c) >= '0' && (c) <= '9') \ | |
f71b19b6 DL |
307 | || ((c) >= 'a' && (c) <= 'f') \ |
308 | || ((c) >= 'A' && (c) <= 'F')) | |
96cc36cc RS |
309 | |
310 | /* This is only used for single-byte characters. */ | |
0b32bf0e | 311 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') |
96cc36cc RS |
312 | |
313 | /* The rest must handle multibyte characters. */ | |
314 | ||
0b32bf0e | 315 | # define ISGRAPH(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 316 | ? (c) > ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
317 | : 1) |
318 | ||
14473664 | 319 | # define ISPRINT(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 320 | ? (c) >= ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
321 | : 1) |
322 | ||
0b32bf0e | 323 | # define ISALNUM(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
324 | ? (((c) >= 'a' && (c) <= 'z') \ |
325 | || ((c) >= 'A' && (c) <= 'Z') \ | |
326 | || ((c) >= '0' && (c) <= '9')) \ | |
96cc36cc RS |
327 | : SYNTAX (c) == Sword) |
328 | ||
0b32bf0e | 329 | # define ISALPHA(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
330 | ? (((c) >= 'a' && (c) <= 'z') \ |
331 | || ((c) >= 'A' && (c) <= 'Z')) \ | |
96cc36cc RS |
332 | : SYNTAX (c) == Sword) |
333 | ||
5da9919f | 334 | # define ISLOWER(c) lowercasep (c) |
96cc36cc | 335 | |
0b32bf0e | 336 | # define ISPUNCT(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
337 | ? ((c) > ' ' && (c) < 0177 \ |
338 | && !(((c) >= 'a' && (c) <= 'z') \ | |
4bb91c68 SM |
339 | || ((c) >= 'A' && (c) <= 'Z') \ |
340 | || ((c) >= '0' && (c) <= '9'))) \ | |
96cc36cc RS |
341 | : SYNTAX (c) != Sword) |
342 | ||
0b32bf0e | 343 | # define ISSPACE(c) (SYNTAX (c) == Swhitespace) |
96cc36cc | 344 | |
5da9919f | 345 | # define ISUPPER(c) uppercasep (c) |
96cc36cc | 346 | |
0b32bf0e | 347 | # define ISWORD(c) (SYNTAX (c) == Sword) |
96cc36cc RS |
348 | |
349 | #else /* not emacs */ | |
350 | ||
f71b19b6 | 351 | /* 1 if C is an ASCII character. */ |
0b32bf0e | 352 | # define IS_REAL_ASCII(c) ((c) < 0200) |
f71b19b6 DL |
353 | |
354 | /* This distinction is not meaningful, except in Emacs. */ | |
0b32bf0e SM |
355 | # define ISUNIBYTE(c) 1 |
356 | ||
357 | # ifdef isblank | |
0e926e56 | 358 | # define ISBLANK(c) isblank (c) |
0b32bf0e SM |
359 | # else |
360 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') | |
361 | # endif | |
362 | # ifdef isgraph | |
0e926e56 | 363 | # define ISGRAPH(c) isgraph (c) |
0b32bf0e | 364 | # else |
0e926e56 | 365 | # define ISGRAPH(c) (isprint (c) && !isspace (c)) |
0b32bf0e SM |
366 | # endif |
367 | ||
0e926e56 | 368 | /* Solaris defines ISPRINT so we must undefine it first. */ |
4bb91c68 | 369 | # undef ISPRINT |
0e926e56 PE |
370 | # define ISPRINT(c) isprint (c) |
371 | # define ISDIGIT(c) isdigit (c) | |
372 | # define ISALNUM(c) isalnum (c) | |
373 | # define ISALPHA(c) isalpha (c) | |
374 | # define ISCNTRL(c) iscntrl (c) | |
375 | # define ISLOWER(c) islower (c) | |
376 | # define ISPUNCT(c) ispunct (c) | |
377 | # define ISSPACE(c) isspace (c) | |
378 | # define ISUPPER(c) isupper (c) | |
379 | # define ISXDIGIT(c) isxdigit (c) | |
0b32bf0e | 380 | |
5e617bc2 | 381 | # define ISWORD(c) ISALPHA (c) |
0b32bf0e | 382 | |
4bb91c68 | 383 | # ifdef _tolower |
5e617bc2 | 384 | # define TOLOWER(c) _tolower (c) |
4bb91c68 | 385 | # else |
5e617bc2 | 386 | # define TOLOWER(c) tolower (c) |
4bb91c68 SM |
387 | # endif |
388 | ||
389 | /* How many characters in the character set. */ | |
390 | # define CHAR_SET_SIZE 256 | |
391 | ||
0b32bf0e | 392 | # ifdef SYNTAX_TABLE |
f71b19b6 | 393 | |
0b32bf0e | 394 | extern char *re_syntax_table; |
f71b19b6 | 395 | |
0b32bf0e SM |
396 | # else /* not SYNTAX_TABLE */ |
397 | ||
0b32bf0e SM |
398 | static char re_syntax_table[CHAR_SET_SIZE]; |
399 | ||
400 | static void | |
d2762c86 | 401 | init_syntax_once (void) |
0b32bf0e SM |
402 | { |
403 | register int c; | |
404 | static int done = 0; | |
405 | ||
406 | if (done) | |
407 | return; | |
408 | ||
72af86bd | 409 | memset (re_syntax_table, 0, sizeof re_syntax_table); |
0b32bf0e | 410 | |
4bb91c68 SM |
411 | for (c = 0; c < CHAR_SET_SIZE; ++c) |
412 | if (ISALNUM (c)) | |
413 | re_syntax_table[c] = Sword; | |
fa9a63c5 | 414 | |
669fa600 | 415 | re_syntax_table['_'] = Ssymbol; |
fa9a63c5 | 416 | |
0b32bf0e SM |
417 | done = 1; |
418 | } | |
419 | ||
420 | # endif /* not SYNTAX_TABLE */ | |
96cc36cc | 421 | |
4bb91c68 SM |
422 | # define SYNTAX(c) re_syntax_table[(c)] |
423 | ||
96cc36cc RS |
424 | #endif /* not emacs */ |
425 | \f | |
261cb4bb | 426 | #define SIGN_EXTEND_CHAR(c) ((signed char) (c)) |
fa9a63c5 RM |
427 | \f |
428 | /* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we | |
429 | use `alloca' instead of `malloc'. This is because using malloc in | |
430 | re_search* or re_match* could cause memory leaks when C-g is used in | |
431 | Emacs; also, malloc is slower and causes storage fragmentation. On | |
5e69f11e RM |
432 | the other hand, malloc is more portable, and easier to debug. |
433 | ||
fa9a63c5 RM |
434 | Because we sometimes use alloca, some routines have to be macros, |
435 | not functions -- `alloca'-allocated space disappears at the end of the | |
436 | function it is called in. */ | |
437 | ||
438 | #ifdef REGEX_MALLOC | |
439 | ||
0b32bf0e SM |
440 | # define REGEX_ALLOCATE malloc |
441 | # define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize) | |
442 | # define REGEX_FREE free | |
fa9a63c5 RM |
443 | |
444 | #else /* not REGEX_MALLOC */ | |
445 | ||
446 | /* Emacs already defines alloca, sometimes. */ | |
0b32bf0e | 447 | # ifndef alloca |
fa9a63c5 RM |
448 | |
449 | /* Make alloca work the best possible way. */ | |
0b32bf0e SM |
450 | # ifdef __GNUC__ |
451 | # define alloca __builtin_alloca | |
452 | # else /* not __GNUC__ */ | |
7f585e7a | 453 | # ifdef HAVE_ALLOCA_H |
0b32bf0e SM |
454 | # include <alloca.h> |
455 | # endif /* HAVE_ALLOCA_H */ | |
456 | # endif /* not __GNUC__ */ | |
fa9a63c5 | 457 | |
0b32bf0e | 458 | # endif /* not alloca */ |
fa9a63c5 | 459 | |
0b32bf0e | 460 | # define REGEX_ALLOCATE alloca |
fa9a63c5 RM |
461 | |
462 | /* Assumes a `char *destination' variable. */ | |
0b32bf0e | 463 | # define REGEX_REALLOCATE(source, osize, nsize) \ |
7d652d97 | 464 | (destination = alloca (nsize), \ |
4bb91c68 | 465 | memcpy (destination, source, osize)) |
fa9a63c5 RM |
466 | |
467 | /* No need to do anything to free, after alloca. */ | |
0b32bf0e | 468 | # define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
469 | |
470 | #endif /* not REGEX_MALLOC */ | |
471 | ||
472 | /* Define how to allocate the failure stack. */ | |
473 | ||
0b32bf0e | 474 | #if defined REL_ALLOC && defined REGEX_MALLOC |
4297555e | 475 | |
0b32bf0e | 476 | # define REGEX_ALLOCATE_STACK(size) \ |
fa9a63c5 | 477 | r_alloc (&failure_stack_ptr, (size)) |
0b32bf0e | 478 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 479 | r_re_alloc (&failure_stack_ptr, (nsize)) |
0b32bf0e | 480 | # define REGEX_FREE_STACK(ptr) \ |
fa9a63c5 RM |
481 | r_alloc_free (&failure_stack_ptr) |
482 | ||
4297555e | 483 | #else /* not using relocating allocator */ |
fa9a63c5 | 484 | |
0b32bf0e | 485 | # ifdef REGEX_MALLOC |
fa9a63c5 | 486 | |
0b32bf0e SM |
487 | # define REGEX_ALLOCATE_STACK malloc |
488 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize) | |
489 | # define REGEX_FREE_STACK free | |
fa9a63c5 | 490 | |
0b32bf0e | 491 | # else /* not REGEX_MALLOC */ |
fa9a63c5 | 492 | |
0b32bf0e | 493 | # define REGEX_ALLOCATE_STACK alloca |
fa9a63c5 | 494 | |
0b32bf0e | 495 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 496 | REGEX_REALLOCATE (source, osize, nsize) |
7814e705 | 497 | /* No need to explicitly free anything. */ |
0b32bf0e | 498 | # define REGEX_FREE_STACK(arg) ((void)0) |
fa9a63c5 | 499 | |
0b32bf0e | 500 | # endif /* not REGEX_MALLOC */ |
4297555e | 501 | #endif /* not using relocating allocator */ |
fa9a63c5 RM |
502 | |
503 | ||
504 | /* True if `size1' is non-NULL and PTR is pointing anywhere inside | |
505 | `string1' or just past its end. This works if PTR is NULL, which is | |
506 | a good thing. */ | |
25fe55af | 507 | #define FIRST_STRING_P(ptr) \ |
fa9a63c5 RM |
508 | (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) |
509 | ||
510 | /* (Re)Allocate N items of type T using malloc, or fail. */ | |
511 | #define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t))) | |
512 | #define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t))) | |
fa9a63c5 RM |
513 | #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) |
514 | ||
4bb91c68 | 515 | #define BYTEWIDTH 8 /* In bits. */ |
fa9a63c5 RM |
516 | |
517 | #define STREQ(s1, s2) ((strcmp (s1, s2) == 0)) | |
518 | ||
519 | #undef MAX | |
520 | #undef MIN | |
521 | #define MAX(a, b) ((a) > (b) ? (a) : (b)) | |
522 | #define MIN(a, b) ((a) < (b) ? (a) : (b)) | |
523 | ||
66f0296e | 524 | /* Type of source-pattern and string chars. */ |
a6fc3b5c EZ |
525 | #ifdef _MSC_VER |
526 | typedef unsigned char re_char; | |
29abe551 | 527 | typedef const re_char const_re_char; |
a6fc3b5c | 528 | #else |
66f0296e | 529 | typedef const unsigned char re_char; |
29abe551 | 530 | typedef re_char const_re_char; |
a6fc3b5c | 531 | #endif |
66f0296e | 532 | |
fa9a63c5 | 533 | typedef char boolean; |
fa9a63c5 | 534 | |
261cb4bb PE |
535 | static regoff_t re_match_2_internal (struct re_pattern_buffer *bufp, |
536 | re_char *string1, size_t size1, | |
537 | re_char *string2, size_t size2, | |
538 | ssize_t pos, | |
539 | struct re_registers *regs, | |
540 | ssize_t stop); | |
fa9a63c5 RM |
541 | \f |
542 | /* These are the command codes that appear in compiled regular | |
4bb91c68 | 543 | expressions. Some opcodes are followed by argument bytes. A |
fa9a63c5 RM |
544 | command code can specify any interpretation whatsoever for its |
545 | arguments. Zero bytes may appear in the compiled regular expression. */ | |
546 | ||
547 | typedef enum | |
548 | { | |
549 | no_op = 0, | |
550 | ||
4bb91c68 | 551 | /* Succeed right away--no more backtracking. */ |
fa9a63c5 RM |
552 | succeed, |
553 | ||
25fe55af | 554 | /* Followed by one byte giving n, then by n literal bytes. */ |
fa9a63c5 RM |
555 | exactn, |
556 | ||
25fe55af | 557 | /* Matches any (more or less) character. */ |
fa9a63c5 RM |
558 | anychar, |
559 | ||
25fe55af RS |
560 | /* Matches any one char belonging to specified set. First |
561 | following byte is number of bitmap bytes. Then come bytes | |
562 | for a bitmap saying which chars are in. Bits in each byte | |
563 | are ordered low-bit-first. A character is in the set if its | |
564 | bit is 1. A character too large to have a bit in the map is | |
96cc36cc RS |
565 | automatically not in the set. |
566 | ||
567 | If the length byte has the 0x80 bit set, then that stuff | |
568 | is followed by a range table: | |
569 | 2 bytes of flags for character sets (low 8 bits, high 8 bits) | |
0b32bf0e | 570 | See RANGE_TABLE_WORK_BITS below. |
01618498 | 571 | 2 bytes, the number of pairs that follow (upto 32767) |
96cc36cc | 572 | pairs, each 2 multibyte characters, |
0b32bf0e | 573 | each multibyte character represented as 3 bytes. */ |
fa9a63c5 RM |
574 | charset, |
575 | ||
25fe55af | 576 | /* Same parameters as charset, but match any character that is |
4bb91c68 | 577 | not one of those specified. */ |
fa9a63c5 RM |
578 | charset_not, |
579 | ||
25fe55af RS |
580 | /* Start remembering the text that is matched, for storing in a |
581 | register. Followed by one byte with the register number, in | |
582 | the range 0 to one less than the pattern buffer's re_nsub | |
505bde11 | 583 | field. */ |
fa9a63c5 RM |
584 | start_memory, |
585 | ||
25fe55af RS |
586 | /* Stop remembering the text that is matched and store it in a |
587 | memory register. Followed by one byte with the register | |
588 | number, in the range 0 to one less than `re_nsub' in the | |
505bde11 | 589 | pattern buffer. */ |
fa9a63c5 RM |
590 | stop_memory, |
591 | ||
25fe55af | 592 | /* Match a duplicate of something remembered. Followed by one |
4bb91c68 | 593 | byte containing the register number. */ |
fa9a63c5 RM |
594 | duplicate, |
595 | ||
25fe55af | 596 | /* Fail unless at beginning of line. */ |
fa9a63c5 RM |
597 | begline, |
598 | ||
4bb91c68 | 599 | /* Fail unless at end of line. */ |
fa9a63c5 RM |
600 | endline, |
601 | ||
25fe55af RS |
602 | /* Succeeds if at beginning of buffer (if emacs) or at beginning |
603 | of string to be matched (if not). */ | |
fa9a63c5 RM |
604 | begbuf, |
605 | ||
25fe55af | 606 | /* Analogously, for end of buffer/string. */ |
fa9a63c5 | 607 | endbuf, |
5e69f11e | 608 | |
25fe55af | 609 | /* Followed by two byte relative address to which to jump. */ |
5e69f11e | 610 | jump, |
fa9a63c5 | 611 | |
25fe55af | 612 | /* Followed by two-byte relative address of place to resume at |
7814e705 | 613 | in case of failure. */ |
fa9a63c5 | 614 | on_failure_jump, |
5e69f11e | 615 | |
25fe55af RS |
616 | /* Like on_failure_jump, but pushes a placeholder instead of the |
617 | current string position when executed. */ | |
fa9a63c5 | 618 | on_failure_keep_string_jump, |
5e69f11e | 619 | |
505bde11 SM |
620 | /* Just like `on_failure_jump', except that it checks that we |
621 | don't get stuck in an infinite loop (matching an empty string | |
622 | indefinitely). */ | |
623 | on_failure_jump_loop, | |
624 | ||
0683b6fa SM |
625 | /* Just like `on_failure_jump_loop', except that it checks for |
626 | a different kind of loop (the kind that shows up with non-greedy | |
627 | operators). This operation has to be immediately preceded | |
628 | by a `no_op'. */ | |
629 | on_failure_jump_nastyloop, | |
630 | ||
0b32bf0e | 631 | /* A smart `on_failure_jump' used for greedy * and + operators. |
c7015153 | 632 | It analyzes the loop before which it is put and if the |
505bde11 | 633 | loop does not require backtracking, it changes itself to |
4e8a9132 SM |
634 | `on_failure_keep_string_jump' and short-circuits the loop, |
635 | else it just defaults to changing itself into `on_failure_jump'. | |
636 | It assumes that it is pointing to just past a `jump'. */ | |
505bde11 | 637 | on_failure_jump_smart, |
fa9a63c5 | 638 | |
25fe55af | 639 | /* Followed by two-byte relative address and two-byte number n. |
ed0767d8 SM |
640 | After matching N times, jump to the address upon failure. |
641 | Does not work if N starts at 0: use on_failure_jump_loop | |
642 | instead. */ | |
fa9a63c5 RM |
643 | succeed_n, |
644 | ||
25fe55af RS |
645 | /* Followed by two-byte relative address, and two-byte number n. |
646 | Jump to the address N times, then fail. */ | |
fa9a63c5 RM |
647 | jump_n, |
648 | ||
25fe55af | 649 | /* Set the following two-byte relative address to the |
7814e705 | 650 | subsequent two-byte number. The address *includes* the two |
25fe55af | 651 | bytes of number. */ |
fa9a63c5 RM |
652 | set_number_at, |
653 | ||
fa9a63c5 RM |
654 | wordbeg, /* Succeeds if at word beginning. */ |
655 | wordend, /* Succeeds if at word end. */ | |
656 | ||
657 | wordbound, /* Succeeds if at a word boundary. */ | |
7814e705 | 658 | notwordbound, /* Succeeds if not at a word boundary. */ |
fa9a63c5 | 659 | |
669fa600 SM |
660 | symbeg, /* Succeeds if at symbol beginning. */ |
661 | symend, /* Succeeds if at symbol end. */ | |
662 | ||
fa9a63c5 | 663 | /* Matches any character whose syntax is specified. Followed by |
25fe55af | 664 | a byte which contains a syntax code, e.g., Sword. */ |
fa9a63c5 RM |
665 | syntaxspec, |
666 | ||
667 | /* Matches any character whose syntax is not that specified. */ | |
1fb352e0 SM |
668 | notsyntaxspec |
669 | ||
670 | #ifdef emacs | |
671 | ,before_dot, /* Succeeds if before point. */ | |
672 | at_dot, /* Succeeds if at point. */ | |
673 | after_dot, /* Succeeds if after point. */ | |
b18215fc RS |
674 | |
675 | /* Matches any character whose category-set contains the specified | |
7814e705 JB |
676 | category. The operator is followed by a byte which contains a |
677 | category code (mnemonic ASCII character). */ | |
b18215fc RS |
678 | categoryspec, |
679 | ||
680 | /* Matches any character whose category-set does not contain the | |
681 | specified category. The operator is followed by a byte which | |
682 | contains the category code (mnemonic ASCII character). */ | |
683 | notcategoryspec | |
fa9a63c5 RM |
684 | #endif /* emacs */ |
685 | } re_opcode_t; | |
686 | \f | |
687 | /* Common operations on the compiled pattern. */ | |
688 | ||
689 | /* Store NUMBER in two contiguous bytes starting at DESTINATION. */ | |
690 | ||
691 | #define STORE_NUMBER(destination, number) \ | |
692 | do { \ | |
693 | (destination)[0] = (number) & 0377; \ | |
694 | (destination)[1] = (number) >> 8; \ | |
695 | } while (0) | |
696 | ||
697 | /* Same as STORE_NUMBER, except increment DESTINATION to | |
698 | the byte after where the number is stored. Therefore, DESTINATION | |
699 | must be an lvalue. */ | |
700 | ||
701 | #define STORE_NUMBER_AND_INCR(destination, number) \ | |
702 | do { \ | |
703 | STORE_NUMBER (destination, number); \ | |
704 | (destination) += 2; \ | |
705 | } while (0) | |
706 | ||
707 | /* Put into DESTINATION a number stored in two contiguous bytes starting | |
708 | at SOURCE. */ | |
709 | ||
710 | #define EXTRACT_NUMBER(destination, source) \ | |
dc4a2ee0 | 711 | ((destination) = extract_number (source)) |
fa9a63c5 | 712 | |
dc4a2ee0 PE |
713 | static int |
714 | extract_number (re_char *source) | |
fa9a63c5 | 715 | { |
6ab1b16c PE |
716 | unsigned leading_byte = SIGN_EXTEND_CHAR (source[1]); |
717 | return (leading_byte << 8) + source[0]; | |
fa9a63c5 RM |
718 | } |
719 | ||
fa9a63c5 RM |
720 | /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. |
721 | SOURCE must be an lvalue. */ | |
722 | ||
723 | #define EXTRACT_NUMBER_AND_INCR(destination, source) \ | |
dc4a2ee0 | 724 | ((destination) = extract_number_and_incr (&source)) |
fa9a63c5 | 725 | |
dc4a2ee0 PE |
726 | static int |
727 | extract_number_and_incr (re_char **source) | |
5e69f11e | 728 | { |
dc4a2ee0 | 729 | int num = extract_number (*source); |
fa9a63c5 | 730 | *source += 2; |
dc4a2ee0 | 731 | return num; |
fa9a63c5 | 732 | } |
fa9a63c5 | 733 | \f |
b18215fc RS |
734 | /* Store a multibyte character in three contiguous bytes starting |
735 | DESTINATION, and increment DESTINATION to the byte after where the | |
7814e705 | 736 | character is stored. Therefore, DESTINATION must be an lvalue. */ |
b18215fc RS |
737 | |
738 | #define STORE_CHARACTER_AND_INCR(destination, character) \ | |
739 | do { \ | |
740 | (destination)[0] = (character) & 0377; \ | |
741 | (destination)[1] = ((character) >> 8) & 0377; \ | |
742 | (destination)[2] = (character) >> 16; \ | |
743 | (destination) += 3; \ | |
744 | } while (0) | |
745 | ||
746 | /* Put into DESTINATION a character stored in three contiguous bytes | |
7814e705 | 747 | starting at SOURCE. */ |
b18215fc RS |
748 | |
749 | #define EXTRACT_CHARACTER(destination, source) \ | |
750 | do { \ | |
751 | (destination) = ((source)[0] \ | |
752 | | ((source)[1] << 8) \ | |
753 | | ((source)[2] << 16)); \ | |
754 | } while (0) | |
755 | ||
756 | ||
757 | /* Macros for charset. */ | |
758 | ||
759 | /* Size of bitmap of charset P in bytes. P is a start of charset, | |
760 | i.e. *P is (re_opcode_t) charset or (re_opcode_t) charset_not. */ | |
761 | #define CHARSET_BITMAP_SIZE(p) ((p)[1] & 0x7F) | |
762 | ||
763 | /* Nonzero if charset P has range table. */ | |
25fe55af | 764 | #define CHARSET_RANGE_TABLE_EXISTS_P(p) ((p)[1] & 0x80) |
b18215fc RS |
765 | |
766 | /* Return the address of range table of charset P. But not the start | |
767 | of table itself, but the before where the number of ranges is | |
96cc36cc RS |
768 | stored. `2 +' means to skip re_opcode_t and size of bitmap, |
769 | and the 2 bytes of flags at the start of the range table. */ | |
770 | #define CHARSET_RANGE_TABLE(p) (&(p)[4 + CHARSET_BITMAP_SIZE (p)]) | |
771 | ||
78779650 | 772 | #ifdef emacs |
96cc36cc RS |
773 | /* Extract the bit flags that start a range table. */ |
774 | #define CHARSET_RANGE_TABLE_BITS(p) \ | |
775 | ((p)[2 + CHARSET_BITMAP_SIZE (p)] \ | |
776 | + (p)[3 + CHARSET_BITMAP_SIZE (p)] * 0x100) | |
78779650 | 777 | #endif |
b18215fc | 778 | |
b18215fc | 779 | /* Return the address of end of RANGE_TABLE. COUNT is number of |
7814e705 JB |
780 | ranges (which is a pair of (start, end)) in the RANGE_TABLE. `* 2' |
781 | is start of range and end of range. `* 3' is size of each start | |
b18215fc RS |
782 | and end. */ |
783 | #define CHARSET_RANGE_TABLE_END(range_table, count) \ | |
784 | ((range_table) + (count) * 2 * 3) | |
785 | ||
7814e705 | 786 | /* Test if C is in RANGE_TABLE. A flag NOT is negated if C is in. |
b18215fc RS |
787 | COUNT is number of ranges in RANGE_TABLE. */ |
788 | #define CHARSET_LOOKUP_RANGE_TABLE_RAW(not, c, range_table, count) \ | |
789 | do \ | |
790 | { \ | |
01618498 | 791 | re_wchar_t range_start, range_end; \ |
19ed5445 | 792 | re_char *rtp; \ |
01618498 | 793 | re_char *range_table_end \ |
b18215fc RS |
794 | = CHARSET_RANGE_TABLE_END ((range_table), (count)); \ |
795 | \ | |
19ed5445 | 796 | for (rtp = (range_table); rtp < range_table_end; rtp += 2 * 3) \ |
b18215fc | 797 | { \ |
19ed5445 PE |
798 | EXTRACT_CHARACTER (range_start, rtp); \ |
799 | EXTRACT_CHARACTER (range_end, rtp + 3); \ | |
b18215fc RS |
800 | \ |
801 | if (range_start <= (c) && (c) <= range_end) \ | |
802 | { \ | |
803 | (not) = !(not); \ | |
804 | break; \ | |
805 | } \ | |
806 | } \ | |
807 | } \ | |
808 | while (0) | |
809 | ||
810 | /* Test if C is in range table of CHARSET. The flag NOT is negated if | |
811 | C is listed in it. */ | |
812 | #define CHARSET_LOOKUP_RANGE_TABLE(not, c, charset) \ | |
813 | do \ | |
814 | { \ | |
815 | /* Number of ranges in range table. */ \ | |
816 | int count; \ | |
01618498 SM |
817 | re_char *range_table = CHARSET_RANGE_TABLE (charset); \ |
818 | \ | |
b18215fc RS |
819 | EXTRACT_NUMBER_AND_INCR (count, range_table); \ |
820 | CHARSET_LOOKUP_RANGE_TABLE_RAW ((not), (c), range_table, count); \ | |
821 | } \ | |
822 | while (0) | |
823 | \f | |
fa9a63c5 RM |
824 | /* If DEBUG is defined, Regex prints many voluminous messages about what |
825 | it is doing (if the variable `debug' is nonzero). If linked with the | |
826 | main program in `iregex.c', you can enter patterns and strings | |
827 | interactively. And if linked with the main program in `main.c' and | |
4bb91c68 | 828 | the other test files, you can run the already-written tests. */ |
fa9a63c5 RM |
829 | |
830 | #ifdef DEBUG | |
831 | ||
832 | /* We use standard I/O for debugging. */ | |
0b32bf0e | 833 | # include <stdio.h> |
fa9a63c5 RM |
834 | |
835 | /* It is useful to test things that ``must'' be true when debugging. */ | |
0b32bf0e | 836 | # include <assert.h> |
fa9a63c5 | 837 | |
99633e97 | 838 | static int debug = -100000; |
fa9a63c5 | 839 | |
0b32bf0e | 840 | # define DEBUG_STATEMENT(e) e |
dc4a2ee0 PE |
841 | # define DEBUG_PRINT(...) if (debug > 0) printf (__VA_ARGS__) |
842 | # define DEBUG_COMPILES_ARGUMENTS | |
0b32bf0e | 843 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ |
99633e97 | 844 | if (debug > 0) print_partial_compiled_pattern (s, e) |
0b32bf0e | 845 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ |
99633e97 | 846 | if (debug > 0) print_double_string (w, s1, sz1, s2, sz2) |
fa9a63c5 RM |
847 | |
848 | ||
849 | /* Print the fastmap in human-readable form. */ | |
850 | ||
dc4a2ee0 PE |
851 | static void |
852 | print_fastmap (char *fastmap) | |
fa9a63c5 RM |
853 | { |
854 | unsigned was_a_range = 0; | |
5e69f11e RM |
855 | unsigned i = 0; |
856 | ||
fa9a63c5 RM |
857 | while (i < (1 << BYTEWIDTH)) |
858 | { | |
859 | if (fastmap[i++]) | |
860 | { | |
861 | was_a_range = 0; | |
25fe55af RS |
862 | putchar (i - 1); |
863 | while (i < (1 << BYTEWIDTH) && fastmap[i]) | |
864 | { | |
865 | was_a_range = 1; | |
866 | i++; | |
867 | } | |
fa9a63c5 | 868 | if (was_a_range) |
25fe55af RS |
869 | { |
870 | printf ("-"); | |
871 | putchar (i - 1); | |
872 | } | |
873 | } | |
fa9a63c5 | 874 | } |
5e69f11e | 875 | putchar ('\n'); |
fa9a63c5 RM |
876 | } |
877 | ||
878 | ||
879 | /* Print a compiled pattern string in human-readable form, starting at | |
880 | the START pointer into it and ending just before the pointer END. */ | |
881 | ||
dc4a2ee0 PE |
882 | static void |
883 | print_partial_compiled_pattern (re_char *start, re_char *end) | |
fa9a63c5 RM |
884 | { |
885 | int mcnt, mcnt2; | |
01618498 SM |
886 | re_char *p = start; |
887 | re_char *pend = end; | |
fa9a63c5 RM |
888 | |
889 | if (start == NULL) | |
890 | { | |
a1a052df | 891 | fprintf (stderr, "(null)\n"); |
fa9a63c5 RM |
892 | return; |
893 | } | |
5e69f11e | 894 | |
fa9a63c5 RM |
895 | /* Loop over pattern commands. */ |
896 | while (p < pend) | |
897 | { | |
dc4a2ee0 | 898 | fprintf (stderr, "%td:\t", p - start); |
fa9a63c5 RM |
899 | |
900 | switch ((re_opcode_t) *p++) | |
901 | { | |
25fe55af | 902 | case no_op: |
a1a052df | 903 | fprintf (stderr, "/no_op"); |
25fe55af | 904 | break; |
fa9a63c5 | 905 | |
99633e97 | 906 | case succeed: |
a1a052df | 907 | fprintf (stderr, "/succeed"); |
99633e97 SM |
908 | break; |
909 | ||
fa9a63c5 RM |
910 | case exactn: |
911 | mcnt = *p++; | |
a1a052df | 912 | fprintf (stderr, "/exactn/%d", mcnt); |
25fe55af | 913 | do |
fa9a63c5 | 914 | { |
a1a052df | 915 | fprintf (stderr, "/%c", *p++); |
25fe55af RS |
916 | } |
917 | while (--mcnt); | |
918 | break; | |
fa9a63c5 RM |
919 | |
920 | case start_memory: | |
a1a052df | 921 | fprintf (stderr, "/start_memory/%d", *p++); |
25fe55af | 922 | break; |
fa9a63c5 RM |
923 | |
924 | case stop_memory: | |
a1a052df | 925 | fprintf (stderr, "/stop_memory/%d", *p++); |
25fe55af | 926 | break; |
fa9a63c5 RM |
927 | |
928 | case duplicate: | |
a1a052df | 929 | fprintf (stderr, "/duplicate/%d", *p++); |
fa9a63c5 RM |
930 | break; |
931 | ||
932 | case anychar: | |
a1a052df | 933 | fprintf (stderr, "/anychar"); |
fa9a63c5 RM |
934 | break; |
935 | ||
936 | case charset: | |
25fe55af RS |
937 | case charset_not: |
938 | { | |
939 | register int c, last = -100; | |
fa9a63c5 | 940 | register int in_range = 0; |
99633e97 SM |
941 | int length = CHARSET_BITMAP_SIZE (p - 1); |
942 | int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1); | |
fa9a63c5 | 943 | |
a1a052df | 944 | fprintf (stderr, "/charset [%s", |
839966f3 | 945 | (re_opcode_t) *(p - 1) == charset_not ? "^" : ""); |
5e69f11e | 946 | |
839966f3 KH |
947 | if (p + *p >= pend) |
948 | fprintf (stderr, " !extends past end of pattern! "); | |
fa9a63c5 | 949 | |
25fe55af | 950 | for (c = 0; c < 256; c++) |
96cc36cc | 951 | if (c / 8 < length |
fa9a63c5 RM |
952 | && (p[1 + (c/8)] & (1 << (c % 8)))) |
953 | { | |
954 | /* Are we starting a range? */ | |
955 | if (last + 1 == c && ! in_range) | |
956 | { | |
a1a052df | 957 | fprintf (stderr, "-"); |
fa9a63c5 RM |
958 | in_range = 1; |
959 | } | |
960 | /* Have we broken a range? */ | |
961 | else if (last + 1 != c && in_range) | |
96cc36cc | 962 | { |
a1a052df | 963 | fprintf (stderr, "%c", last); |
fa9a63c5 RM |
964 | in_range = 0; |
965 | } | |
5e69f11e | 966 | |
fa9a63c5 | 967 | if (! in_range) |
a1a052df | 968 | fprintf (stderr, "%c", c); |
fa9a63c5 RM |
969 | |
970 | last = c; | |
25fe55af | 971 | } |
fa9a63c5 RM |
972 | |
973 | if (in_range) | |
a1a052df | 974 | fprintf (stderr, "%c", last); |
fa9a63c5 | 975 | |
a1a052df | 976 | fprintf (stderr, "]"); |
fa9a63c5 | 977 | |
99633e97 | 978 | p += 1 + length; |
96cc36cc | 979 | |
96cc36cc | 980 | if (has_range_table) |
99633e97 SM |
981 | { |
982 | int count; | |
a1a052df | 983 | fprintf (stderr, "has-range-table"); |
99633e97 SM |
984 | |
985 | /* ??? Should print the range table; for now, just skip it. */ | |
986 | p += 2; /* skip range table bits */ | |
987 | EXTRACT_NUMBER_AND_INCR (count, p); | |
988 | p = CHARSET_RANGE_TABLE_END (p, count); | |
989 | } | |
fa9a63c5 RM |
990 | } |
991 | break; | |
992 | ||
993 | case begline: | |
a1a052df | 994 | fprintf (stderr, "/begline"); |
25fe55af | 995 | break; |
fa9a63c5 RM |
996 | |
997 | case endline: | |
a1a052df | 998 | fprintf (stderr, "/endline"); |
25fe55af | 999 | break; |
fa9a63c5 RM |
1000 | |
1001 | case on_failure_jump: | |
dc4a2ee0 PE |
1002 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1003 | fprintf (stderr, "/on_failure_jump to %td", p + mcnt - start); | |
25fe55af | 1004 | break; |
fa9a63c5 RM |
1005 | |
1006 | case on_failure_keep_string_jump: | |
dc4a2ee0 PE |
1007 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1008 | fprintf (stderr, "/on_failure_keep_string_jump to %td", | |
1009 | p + mcnt - start); | |
25fe55af | 1010 | break; |
fa9a63c5 | 1011 | |
0683b6fa | 1012 | case on_failure_jump_nastyloop: |
dc4a2ee0 PE |
1013 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1014 | fprintf (stderr, "/on_failure_jump_nastyloop to %td", | |
1015 | p + mcnt - start); | |
0683b6fa SM |
1016 | break; |
1017 | ||
505bde11 | 1018 | case on_failure_jump_loop: |
dc4a2ee0 PE |
1019 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1020 | fprintf (stderr, "/on_failure_jump_loop to %td", | |
1021 | p + mcnt - start); | |
5e69f11e RM |
1022 | break; |
1023 | ||
505bde11 | 1024 | case on_failure_jump_smart: |
dc4a2ee0 PE |
1025 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1026 | fprintf (stderr, "/on_failure_jump_smart to %td", | |
1027 | p + mcnt - start); | |
5e69f11e RM |
1028 | break; |
1029 | ||
25fe55af | 1030 | case jump: |
dc4a2ee0 PE |
1031 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1032 | fprintf (stderr, "/jump to %td", p + mcnt - start); | |
fa9a63c5 RM |
1033 | break; |
1034 | ||
25fe55af | 1035 | case succeed_n: |
dc4a2ee0 PE |
1036 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1037 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1038 | fprintf (stderr, "/succeed_n to %td, %d times", | |
1039 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1040 | break; |
5e69f11e | 1041 | |
25fe55af | 1042 | case jump_n: |
dc4a2ee0 PE |
1043 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1044 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1045 | fprintf (stderr, "/jump_n to %td, %d times", | |
1046 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1047 | break; |
5e69f11e | 1048 | |
25fe55af | 1049 | case set_number_at: |
dc4a2ee0 PE |
1050 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1051 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1052 | fprintf (stderr, "/set_number_at location %td to %d", | |
1053 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1054 | break; |
5e69f11e | 1055 | |
25fe55af | 1056 | case wordbound: |
a1a052df | 1057 | fprintf (stderr, "/wordbound"); |
fa9a63c5 RM |
1058 | break; |
1059 | ||
1060 | case notwordbound: | |
a1a052df | 1061 | fprintf (stderr, "/notwordbound"); |
25fe55af | 1062 | break; |
fa9a63c5 RM |
1063 | |
1064 | case wordbeg: | |
a1a052df | 1065 | fprintf (stderr, "/wordbeg"); |
fa9a63c5 | 1066 | break; |
5e69f11e | 1067 | |
fa9a63c5 | 1068 | case wordend: |
a1a052df | 1069 | fprintf (stderr, "/wordend"); |
e2543b02 | 1070 | break; |
5e69f11e | 1071 | |
669fa600 | 1072 | case symbeg: |
e2543b02 | 1073 | fprintf (stderr, "/symbeg"); |
669fa600 SM |
1074 | break; |
1075 | ||
1076 | case symend: | |
e2543b02 | 1077 | fprintf (stderr, "/symend"); |
669fa600 | 1078 | break; |
5e69f11e | 1079 | |
1fb352e0 | 1080 | case syntaxspec: |
a1a052df | 1081 | fprintf (stderr, "/syntaxspec"); |
1fb352e0 | 1082 | mcnt = *p++; |
a1a052df | 1083 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1084 | break; |
1085 | ||
1086 | case notsyntaxspec: | |
a1a052df | 1087 | fprintf (stderr, "/notsyntaxspec"); |
1fb352e0 | 1088 | mcnt = *p++; |
a1a052df | 1089 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1090 | break; |
1091 | ||
0b32bf0e | 1092 | # ifdef emacs |
fa9a63c5 | 1093 | case before_dot: |
a1a052df | 1094 | fprintf (stderr, "/before_dot"); |
25fe55af | 1095 | break; |
fa9a63c5 RM |
1096 | |
1097 | case at_dot: | |
a1a052df | 1098 | fprintf (stderr, "/at_dot"); |
25fe55af | 1099 | break; |
fa9a63c5 RM |
1100 | |
1101 | case after_dot: | |
a1a052df | 1102 | fprintf (stderr, "/after_dot"); |
25fe55af | 1103 | break; |
fa9a63c5 | 1104 | |
1fb352e0 | 1105 | case categoryspec: |
a1a052df | 1106 | fprintf (stderr, "/categoryspec"); |
fa9a63c5 | 1107 | mcnt = *p++; |
a1a052df | 1108 | fprintf (stderr, "/%d", mcnt); |
25fe55af | 1109 | break; |
5e69f11e | 1110 | |
1fb352e0 | 1111 | case notcategoryspec: |
a1a052df | 1112 | fprintf (stderr, "/notcategoryspec"); |
fa9a63c5 | 1113 | mcnt = *p++; |
a1a052df | 1114 | fprintf (stderr, "/%d", mcnt); |
fa9a63c5 | 1115 | break; |
0b32bf0e | 1116 | # endif /* emacs */ |
fa9a63c5 | 1117 | |
fa9a63c5 | 1118 | case begbuf: |
a1a052df | 1119 | fprintf (stderr, "/begbuf"); |
25fe55af | 1120 | break; |
fa9a63c5 RM |
1121 | |
1122 | case endbuf: | |
a1a052df | 1123 | fprintf (stderr, "/endbuf"); |
25fe55af | 1124 | break; |
fa9a63c5 | 1125 | |
25fe55af | 1126 | default: |
a1a052df | 1127 | fprintf (stderr, "?%d", *(p-1)); |
fa9a63c5 RM |
1128 | } |
1129 | ||
a1a052df | 1130 | fprintf (stderr, "\n"); |
fa9a63c5 RM |
1131 | } |
1132 | ||
dc4a2ee0 | 1133 | fprintf (stderr, "%td:\tend of pattern.\n", p - start); |
fa9a63c5 RM |
1134 | } |
1135 | ||
1136 | ||
dc4a2ee0 PE |
1137 | static void |
1138 | print_compiled_pattern (struct re_pattern_buffer *bufp) | |
fa9a63c5 | 1139 | { |
01618498 | 1140 | re_char *buffer = bufp->buffer; |
fa9a63c5 RM |
1141 | |
1142 | print_partial_compiled_pattern (buffer, buffer + bufp->used); | |
4bb91c68 SM |
1143 | printf ("%ld bytes used/%ld bytes allocated.\n", |
1144 | bufp->used, bufp->allocated); | |
fa9a63c5 RM |
1145 | |
1146 | if (bufp->fastmap_accurate && bufp->fastmap) | |
1147 | { | |
1148 | printf ("fastmap: "); | |
1149 | print_fastmap (bufp->fastmap); | |
1150 | } | |
1151 | ||
dc4a2ee0 | 1152 | printf ("re_nsub: %zu\t", bufp->re_nsub); |
fa9a63c5 RM |
1153 | printf ("regs_alloc: %d\t", bufp->regs_allocated); |
1154 | printf ("can_be_null: %d\t", bufp->can_be_null); | |
fa9a63c5 RM |
1155 | printf ("no_sub: %d\t", bufp->no_sub); |
1156 | printf ("not_bol: %d\t", bufp->not_bol); | |
1157 | printf ("not_eol: %d\t", bufp->not_eol); | |
4bb91c68 | 1158 | printf ("syntax: %lx\n", bufp->syntax); |
505bde11 | 1159 | fflush (stdout); |
fa9a63c5 RM |
1160 | /* Perhaps we should print the translate table? */ |
1161 | } | |
1162 | ||
1163 | ||
dc4a2ee0 PE |
1164 | static void |
1165 | print_double_string (re_char *where, re_char *string1, ssize_t size1, | |
1166 | re_char *string2, ssize_t size2) | |
fa9a63c5 | 1167 | { |
d1dfb56c | 1168 | ssize_t this_char; |
5e69f11e | 1169 | |
fa9a63c5 RM |
1170 | if (where == NULL) |
1171 | printf ("(null)"); | |
1172 | else | |
1173 | { | |
1174 | if (FIRST_STRING_P (where)) | |
25fe55af RS |
1175 | { |
1176 | for (this_char = where - string1; this_char < size1; this_char++) | |
1177 | putchar (string1[this_char]); | |
fa9a63c5 | 1178 | |
25fe55af RS |
1179 | where = string2; |
1180 | } | |
fa9a63c5 RM |
1181 | |
1182 | for (this_char = where - string2; this_char < size2; this_char++) | |
25fe55af | 1183 | putchar (string2[this_char]); |
fa9a63c5 RM |
1184 | } |
1185 | } | |
1186 | ||
1187 | #else /* not DEBUG */ | |
1188 | ||
0b32bf0e SM |
1189 | # undef assert |
1190 | # define assert(e) | |
fa9a63c5 | 1191 | |
0b32bf0e | 1192 | # define DEBUG_STATEMENT(e) |
8208d2bf | 1193 | # define DEBUG_PRINT(...) |
0b32bf0e SM |
1194 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) |
1195 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
fa9a63c5 RM |
1196 | |
1197 | #endif /* not DEBUG */ | |
1198 | \f | |
4da60324 PE |
1199 | /* Use this to suppress gcc's `...may be used before initialized' warnings. */ |
1200 | #ifdef lint | |
1201 | # define IF_LINT(Code) Code | |
1202 | #else | |
1203 | # define IF_LINT(Code) /* empty */ | |
1204 | #endif | |
1205 | \f | |
fa9a63c5 RM |
1206 | /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can |
1207 | also be assigned to arbitrarily: each pattern buffer stores its own | |
1208 | syntax, so it can be changed between regex compilations. */ | |
1209 | /* This has no initializer because initialized variables in Emacs | |
1210 | become read-only after dumping. */ | |
1211 | reg_syntax_t re_syntax_options; | |
1212 | ||
1213 | ||
1214 | /* Specify the precise syntax of regexps for compilation. This provides | |
1215 | for compatibility for various utilities which historically have | |
1216 | different, incompatible syntaxes. | |
1217 | ||
1218 | The argument SYNTAX is a bit mask comprised of the various bits | |
4bb91c68 | 1219 | defined in regex.h. We return the old syntax. */ |
fa9a63c5 RM |
1220 | |
1221 | reg_syntax_t | |
971de7fb | 1222 | re_set_syntax (reg_syntax_t syntax) |
fa9a63c5 RM |
1223 | { |
1224 | reg_syntax_t ret = re_syntax_options; | |
5e69f11e | 1225 | |
fa9a63c5 RM |
1226 | re_syntax_options = syntax; |
1227 | return ret; | |
1228 | } | |
c0f9ea08 | 1229 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) |
f9b0fd99 RS |
1230 | |
1231 | /* Regexp to use to replace spaces, or NULL meaning don't. */ | |
f462f075 | 1232 | static const_re_char *whitespace_regexp; |
f9b0fd99 RS |
1233 | |
1234 | void | |
971de7fb | 1235 | re_set_whitespace_regexp (const char *regexp) |
f9b0fd99 | 1236 | { |
f462f075 | 1237 | whitespace_regexp = (const_re_char *) regexp; |
f9b0fd99 RS |
1238 | } |
1239 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) | |
fa9a63c5 RM |
1240 | \f |
1241 | /* This table gives an error message for each of the error codes listed | |
4bb91c68 | 1242 | in regex.h. Obviously the order here has to be same as there. |
fa9a63c5 | 1243 | POSIX doesn't require that we do anything for REG_NOERROR, |
4bb91c68 | 1244 | but why not be nice? */ |
fa9a63c5 RM |
1245 | |
1246 | static const char *re_error_msgid[] = | |
5e69f11e RM |
1247 | { |
1248 | gettext_noop ("Success"), /* REG_NOERROR */ | |
1249 | gettext_noop ("No match"), /* REG_NOMATCH */ | |
1250 | gettext_noop ("Invalid regular expression"), /* REG_BADPAT */ | |
1251 | gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */ | |
1252 | gettext_noop ("Invalid character class name"), /* REG_ECTYPE */ | |
1253 | gettext_noop ("Trailing backslash"), /* REG_EESCAPE */ | |
1254 | gettext_noop ("Invalid back reference"), /* REG_ESUBREG */ | |
1255 | gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */ | |
1256 | gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */ | |
1257 | gettext_noop ("Unmatched \\{"), /* REG_EBRACE */ | |
1258 | gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */ | |
1259 | gettext_noop ("Invalid range end"), /* REG_ERANGE */ | |
1260 | gettext_noop ("Memory exhausted"), /* REG_ESPACE */ | |
1261 | gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */ | |
1262 | gettext_noop ("Premature end of regular expression"), /* REG_EEND */ | |
1263 | gettext_noop ("Regular expression too big"), /* REG_ESIZE */ | |
1264 | gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */ | |
b3e4c897 | 1265 | gettext_noop ("Range striding over charsets") /* REG_ERANGEX */ |
fa9a63c5 RM |
1266 | }; |
1267 | \f | |
4bb91c68 | 1268 | /* Avoiding alloca during matching, to placate r_alloc. */ |
fa9a63c5 RM |
1269 | |
1270 | /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the | |
1271 | searching and matching functions should not call alloca. On some | |
1272 | systems, alloca is implemented in terms of malloc, and if we're | |
1273 | using the relocating allocator routines, then malloc could cause a | |
1274 | relocation, which might (if the strings being searched are in the | |
1275 | ralloc heap) shift the data out from underneath the regexp | |
1276 | routines. | |
1277 | ||
5e69f11e | 1278 | Here's another reason to avoid allocation: Emacs |
fa9a63c5 RM |
1279 | processes input from X in a signal handler; processing X input may |
1280 | call malloc; if input arrives while a matching routine is calling | |
1281 | malloc, then we're scrod. But Emacs can't just block input while | |
1282 | calling matching routines; then we don't notice interrupts when | |
1283 | they come in. So, Emacs blocks input around all regexp calls | |
1284 | except the matching calls, which it leaves unprotected, in the | |
1285 | faith that they will not malloc. */ | |
1286 | ||
1287 | /* Normally, this is fine. */ | |
1288 | #define MATCH_MAY_ALLOCATE | |
1289 | ||
fa9a63c5 RM |
1290 | /* The match routines may not allocate if (1) they would do it with malloc |
1291 | and (2) it's not safe for them to use malloc. | |
1292 | Note that if REL_ALLOC is defined, matching would not use malloc for the | |
1293 | failure stack, but we would still use it for the register vectors; | |
4bb91c68 | 1294 | so REL_ALLOC should not affect this. */ |
b588157e | 1295 | #if defined REGEX_MALLOC && defined emacs |
0b32bf0e | 1296 | # undef MATCH_MAY_ALLOCATE |
fa9a63c5 RM |
1297 | #endif |
1298 | ||
1299 | \f | |
1300 | /* Failure stack declarations and macros; both re_compile_fastmap and | |
1301 | re_match_2 use a failure stack. These have to be macros because of | |
1302 | REGEX_ALLOCATE_STACK. */ | |
5e69f11e | 1303 | |
fa9a63c5 | 1304 | |
320a2a73 | 1305 | /* Approximate number of failure points for which to initially allocate space |
fa9a63c5 RM |
1306 | when matching. If this number is exceeded, we allocate more |
1307 | space, so it is not a hard limit. */ | |
1308 | #ifndef INIT_FAILURE_ALLOC | |
0b32bf0e | 1309 | # define INIT_FAILURE_ALLOC 20 |
fa9a63c5 RM |
1310 | #endif |
1311 | ||
1312 | /* Roughly the maximum number of failure points on the stack. Would be | |
320a2a73 | 1313 | exactly that if always used TYPICAL_FAILURE_SIZE items each time we failed. |
fa9a63c5 | 1314 | This is a variable only so users of regex can assign to it; we never |
ada30c0e SM |
1315 | change it ourselves. We always multiply it by TYPICAL_FAILURE_SIZE |
1316 | before using it, so it should probably be a byte-count instead. */ | |
c0f9ea08 SM |
1317 | # if defined MATCH_MAY_ALLOCATE |
1318 | /* Note that 4400 was enough to cause a crash on Alpha OSF/1, | |
320a2a73 KH |
1319 | whose default stack limit is 2mb. In order for a larger |
1320 | value to work reliably, you have to try to make it accord | |
1321 | with the process stack limit. */ | |
c0f9ea08 SM |
1322 | size_t re_max_failures = 40000; |
1323 | # else | |
1324 | size_t re_max_failures = 4000; | |
1325 | # endif | |
fa9a63c5 RM |
1326 | |
1327 | union fail_stack_elt | |
1328 | { | |
01618498 | 1329 | re_char *pointer; |
c0f9ea08 SM |
1330 | /* This should be the biggest `int' that's no bigger than a pointer. */ |
1331 | long integer; | |
fa9a63c5 RM |
1332 | }; |
1333 | ||
1334 | typedef union fail_stack_elt fail_stack_elt_t; | |
1335 | ||
1336 | typedef struct | |
1337 | { | |
1338 | fail_stack_elt_t *stack; | |
c0f9ea08 SM |
1339 | size_t size; |
1340 | size_t avail; /* Offset of next open position. */ | |
1341 | size_t frame; /* Offset of the cur constructed frame. */ | |
fa9a63c5 RM |
1342 | } fail_stack_type; |
1343 | ||
505bde11 | 1344 | #define FAIL_STACK_EMPTY() (fail_stack.frame == 0) |
fa9a63c5 RM |
1345 | |
1346 | ||
1347 | /* Define macros to initialize and free the failure stack. | |
1348 | Do `return -2' if the alloc fails. */ | |
1349 | ||
1350 | #ifdef MATCH_MAY_ALLOCATE | |
0b32bf0e | 1351 | # define INIT_FAIL_STACK() \ |
fa9a63c5 | 1352 | do { \ |
38182d90 | 1353 | fail_stack.stack = \ |
320a2a73 KH |
1354 | REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * TYPICAL_FAILURE_SIZE \ |
1355 | * sizeof (fail_stack_elt_t)); \ | |
fa9a63c5 RM |
1356 | \ |
1357 | if (fail_stack.stack == NULL) \ | |
1358 | return -2; \ | |
1359 | \ | |
1360 | fail_stack.size = INIT_FAILURE_ALLOC; \ | |
1361 | fail_stack.avail = 0; \ | |
505bde11 | 1362 | fail_stack.frame = 0; \ |
fa9a63c5 | 1363 | } while (0) |
fa9a63c5 | 1364 | #else |
0b32bf0e | 1365 | # define INIT_FAIL_STACK() \ |
fa9a63c5 RM |
1366 | do { \ |
1367 | fail_stack.avail = 0; \ | |
505bde11 | 1368 | fail_stack.frame = 0; \ |
fa9a63c5 RM |
1369 | } while (0) |
1370 | ||
b313f9d8 PE |
1371 | # define RETALLOC_IF(addr, n, t) \ |
1372 | if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t) | |
fa9a63c5 RM |
1373 | #endif |
1374 | ||
1375 | ||
320a2a73 KH |
1376 | /* Double the size of FAIL_STACK, up to a limit |
1377 | which allows approximately `re_max_failures' items. | |
fa9a63c5 RM |
1378 | |
1379 | Return 1 if succeeds, and 0 if either ran out of memory | |
5e69f11e RM |
1380 | allocating space for it or it was already too large. |
1381 | ||
4bb91c68 | 1382 | REGEX_REALLOCATE_STACK requires `destination' be declared. */ |
fa9a63c5 | 1383 | |
320a2a73 KH |
1384 | /* Factor to increase the failure stack size by |
1385 | when we increase it. | |
1386 | This used to be 2, but 2 was too wasteful | |
1387 | because the old discarded stacks added up to as much space | |
1388 | were as ultimate, maximum-size stack. */ | |
1389 | #define FAIL_STACK_GROWTH_FACTOR 4 | |
1390 | ||
1391 | #define GROW_FAIL_STACK(fail_stack) \ | |
eead07d6 KH |
1392 | (((fail_stack).size * sizeof (fail_stack_elt_t) \ |
1393 | >= re_max_failures * TYPICAL_FAILURE_SIZE) \ | |
fa9a63c5 | 1394 | ? 0 \ |
320a2a73 | 1395 | : ((fail_stack).stack \ |
38182d90 | 1396 | = REGEX_REALLOCATE_STACK ((fail_stack).stack, \ |
25fe55af | 1397 | (fail_stack).size * sizeof (fail_stack_elt_t), \ |
320a2a73 KH |
1398 | MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ |
1399 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1400 | * FAIL_STACK_GROWTH_FACTOR))), \ | |
fa9a63c5 RM |
1401 | \ |
1402 | (fail_stack).stack == NULL \ | |
1403 | ? 0 \ | |
6453db45 KH |
1404 | : ((fail_stack).size \ |
1405 | = (MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ | |
1406 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1407 | * FAIL_STACK_GROWTH_FACTOR)) \ | |
1408 | / sizeof (fail_stack_elt_t)), \ | |
25fe55af | 1409 | 1))) |
fa9a63c5 RM |
1410 | |
1411 | ||
fa9a63c5 RM |
1412 | /* Push a pointer value onto the failure stack. |
1413 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1414 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 | 1415 | #define PUSH_FAILURE_POINTER(item) \ |
01618498 | 1416 | fail_stack.stack[fail_stack.avail++].pointer = (item) |
fa9a63c5 RM |
1417 | |
1418 | /* This pushes an integer-valued item onto the failure stack. | |
1419 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1420 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 RM |
1421 | #define PUSH_FAILURE_INT(item) \ |
1422 | fail_stack.stack[fail_stack.avail++].integer = (item) | |
1423 | ||
b313f9d8 | 1424 | /* These POP... operations complement the PUSH... operations. |
fa9a63c5 RM |
1425 | All assume that `fail_stack' is nonempty. */ |
1426 | #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer | |
1427 | #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer | |
fa9a63c5 | 1428 | |
505bde11 SM |
1429 | /* Individual items aside from the registers. */ |
1430 | #define NUM_NONREG_ITEMS 3 | |
1431 | ||
1432 | /* Used to examine the stack (to detect infinite loops). */ | |
1433 | #define FAILURE_PAT(h) fail_stack.stack[(h) - 1].pointer | |
66f0296e | 1434 | #define FAILURE_STR(h) (fail_stack.stack[(h) - 2].pointer) |
505bde11 SM |
1435 | #define NEXT_FAILURE_HANDLE(h) fail_stack.stack[(h) - 3].integer |
1436 | #define TOP_FAILURE_HANDLE() fail_stack.frame | |
fa9a63c5 RM |
1437 | |
1438 | ||
505bde11 SM |
1439 | #define ENSURE_FAIL_STACK(space) \ |
1440 | while (REMAINING_AVAIL_SLOTS <= space) { \ | |
1441 | if (!GROW_FAIL_STACK (fail_stack)) \ | |
1442 | return -2; \ | |
dc4a2ee0 PE |
1443 | DEBUG_PRINT ("\n Doubled stack; size now: %zd\n", (fail_stack).size);\ |
1444 | DEBUG_PRINT (" slots available: %zd\n", REMAINING_AVAIL_SLOTS);\ | |
505bde11 SM |
1445 | } |
1446 | ||
1447 | /* Push register NUM onto the stack. */ | |
1448 | #define PUSH_FAILURE_REG(num) \ | |
1449 | do { \ | |
1450 | char *destination; \ | |
dc4a2ee0 | 1451 | long n = num; \ |
505bde11 | 1452 | ENSURE_FAIL_STACK(3); \ |
dc4a2ee0 PE |
1453 | DEBUG_PRINT (" Push reg %ld (spanning %p -> %p)\n", \ |
1454 | n, regstart[n], regend[n]); \ | |
1455 | PUSH_FAILURE_POINTER (regstart[n]); \ | |
1456 | PUSH_FAILURE_POINTER (regend[n]); \ | |
1457 | PUSH_FAILURE_INT (n); \ | |
505bde11 SM |
1458 | } while (0) |
1459 | ||
01618498 SM |
1460 | /* Change the counter's value to VAL, but make sure that it will |
1461 | be reset when backtracking. */ | |
1462 | #define PUSH_NUMBER(ptr,val) \ | |
dc1e502d SM |
1463 | do { \ |
1464 | char *destination; \ | |
1465 | int c; \ | |
1466 | ENSURE_FAIL_STACK(3); \ | |
1467 | EXTRACT_NUMBER (c, ptr); \ | |
dc4a2ee0 | 1468 | DEBUG_PRINT (" Push number %p = %d -> %d\n", ptr, c, val); \ |
dc1e502d SM |
1469 | PUSH_FAILURE_INT (c); \ |
1470 | PUSH_FAILURE_POINTER (ptr); \ | |
1471 | PUSH_FAILURE_INT (-1); \ | |
01618498 | 1472 | STORE_NUMBER (ptr, val); \ |
dc1e502d SM |
1473 | } while (0) |
1474 | ||
505bde11 | 1475 | /* Pop a saved register off the stack. */ |
dc1e502d | 1476 | #define POP_FAILURE_REG_OR_COUNT() \ |
505bde11 | 1477 | do { \ |
d1dfb56c | 1478 | long pfreg = POP_FAILURE_INT (); \ |
19ed5445 | 1479 | if (pfreg == -1) \ |
dc1e502d SM |
1480 | { \ |
1481 | /* It's a counter. */ \ | |
6dcf2d0e SM |
1482 | /* Here, we discard `const', making re_match non-reentrant. */ \ |
1483 | unsigned char *ptr = (unsigned char*) POP_FAILURE_POINTER (); \ | |
19ed5445 PE |
1484 | pfreg = POP_FAILURE_INT (); \ |
1485 | STORE_NUMBER (ptr, pfreg); \ | |
dc4a2ee0 | 1486 | DEBUG_PRINT (" Pop counter %p = %ld\n", ptr, pfreg); \ |
dc1e502d SM |
1487 | } \ |
1488 | else \ | |
1489 | { \ | |
19ed5445 PE |
1490 | regend[pfreg] = POP_FAILURE_POINTER (); \ |
1491 | regstart[pfreg] = POP_FAILURE_POINTER (); \ | |
dc4a2ee0 PE |
1492 | DEBUG_PRINT (" Pop reg %ld (spanning %p -> %p)\n", \ |
1493 | pfreg, regstart[pfreg], regend[pfreg]); \ | |
dc1e502d | 1494 | } \ |
505bde11 SM |
1495 | } while (0) |
1496 | ||
1497 | /* Check that we are not stuck in an infinite loop. */ | |
1498 | #define CHECK_INFINITE_LOOP(pat_cur, string_place) \ | |
1499 | do { \ | |
d1dfb56c | 1500 | ssize_t failure = TOP_FAILURE_HANDLE (); \ |
505bde11 | 1501 | /* Check for infinite matching loops */ \ |
f6df485f RS |
1502 | while (failure > 0 \ |
1503 | && (FAILURE_STR (failure) == string_place \ | |
1504 | || FAILURE_STR (failure) == NULL)) \ | |
505bde11 SM |
1505 | { \ |
1506 | assert (FAILURE_PAT (failure) >= bufp->buffer \ | |
66f0296e | 1507 | && FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \ |
505bde11 | 1508 | if (FAILURE_PAT (failure) == pat_cur) \ |
f6df485f | 1509 | { \ |
6df42991 SM |
1510 | cycle = 1; \ |
1511 | break; \ | |
f6df485f | 1512 | } \ |
dc4a2ee0 | 1513 | DEBUG_PRINT (" Other pattern: %p\n", FAILURE_PAT (failure)); \ |
505bde11 SM |
1514 | failure = NEXT_FAILURE_HANDLE(failure); \ |
1515 | } \ | |
dc4a2ee0 | 1516 | DEBUG_PRINT (" Other string: %p\n", FAILURE_STR (failure)); \ |
505bde11 | 1517 | } while (0) |
6df42991 | 1518 | |
fa9a63c5 | 1519 | /* Push the information about the state we will need |
5e69f11e RM |
1520 | if we ever fail back to it. |
1521 | ||
505bde11 | 1522 | Requires variables fail_stack, regstart, regend and |
320a2a73 | 1523 | num_regs be declared. GROW_FAIL_STACK requires `destination' be |
fa9a63c5 | 1524 | declared. |
5e69f11e | 1525 | |
fa9a63c5 RM |
1526 | Does `return FAILURE_CODE' if runs out of memory. */ |
1527 | ||
505bde11 SM |
1528 | #define PUSH_FAILURE_POINT(pattern, string_place) \ |
1529 | do { \ | |
1530 | char *destination; \ | |
1531 | /* Must be int, so when we don't save any registers, the arithmetic \ | |
1532 | of 0 + -1 isn't done as unsigned. */ \ | |
1533 | \ | |
505bde11 | 1534 | DEBUG_STATEMENT (nfailure_points_pushed++); \ |
dc4a2ee0 PE |
1535 | DEBUG_PRINT ("\nPUSH_FAILURE_POINT:\n"); \ |
1536 | DEBUG_PRINT (" Before push, next avail: %zd\n", (fail_stack).avail); \ | |
1537 | DEBUG_PRINT (" size: %zd\n", (fail_stack).size);\ | |
505bde11 SM |
1538 | \ |
1539 | ENSURE_FAIL_STACK (NUM_NONREG_ITEMS); \ | |
1540 | \ | |
dc4a2ee0 | 1541 | DEBUG_PRINT ("\n"); \ |
505bde11 | 1542 | \ |
dc4a2ee0 | 1543 | DEBUG_PRINT (" Push frame index: %zd\n", fail_stack.frame); \ |
505bde11 SM |
1544 | PUSH_FAILURE_INT (fail_stack.frame); \ |
1545 | \ | |
dc4a2ee0 | 1546 | DEBUG_PRINT (" Push string %p: `", string_place); \ |
505bde11 | 1547 | DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, size2);\ |
dc4a2ee0 | 1548 | DEBUG_PRINT ("'\n"); \ |
505bde11 SM |
1549 | PUSH_FAILURE_POINTER (string_place); \ |
1550 | \ | |
dc4a2ee0 | 1551 | DEBUG_PRINT (" Push pattern %p: ", pattern); \ |
505bde11 SM |
1552 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern, pend); \ |
1553 | PUSH_FAILURE_POINTER (pattern); \ | |
1554 | \ | |
1555 | /* Close the frame by moving the frame pointer past it. */ \ | |
1556 | fail_stack.frame = fail_stack.avail; \ | |
1557 | } while (0) | |
fa9a63c5 | 1558 | |
320a2a73 KH |
1559 | /* Estimate the size of data pushed by a typical failure stack entry. |
1560 | An estimate is all we need, because all we use this for | |
1561 | is to choose a limit for how big to make the failure stack. */ | |
ada30c0e | 1562 | /* BEWARE, the value `20' is hard-coded in emacs.c:main(). */ |
320a2a73 | 1563 | #define TYPICAL_FAILURE_SIZE 20 |
fa9a63c5 | 1564 | |
fa9a63c5 RM |
1565 | /* How many items can still be added to the stack without overflowing it. */ |
1566 | #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) | |
1567 | ||
1568 | ||
1569 | /* Pops what PUSH_FAIL_STACK pushes. | |
1570 | ||
1571 | We restore into the parameters, all of which should be lvalues: | |
1572 | STR -- the saved data position. | |
1573 | PAT -- the saved pattern position. | |
fa9a63c5 | 1574 | REGSTART, REGEND -- arrays of string positions. |
5e69f11e | 1575 | |
fa9a63c5 | 1576 | Also assumes the variables `fail_stack' and (if debugging), `bufp', |
7814e705 | 1577 | `pend', `string1', `size1', `string2', and `size2'. */ |
fa9a63c5 | 1578 | |
505bde11 SM |
1579 | #define POP_FAILURE_POINT(str, pat) \ |
1580 | do { \ | |
fa9a63c5 RM |
1581 | assert (!FAIL_STACK_EMPTY ()); \ |
1582 | \ | |
1583 | /* Remove failure points and point to how many regs pushed. */ \ | |
dc4a2ee0 PE |
1584 | DEBUG_PRINT ("POP_FAILURE_POINT:\n"); \ |
1585 | DEBUG_PRINT (" Before pop, next avail: %zd\n", fail_stack.avail); \ | |
1586 | DEBUG_PRINT (" size: %zd\n", fail_stack.size); \ | |
fa9a63c5 | 1587 | \ |
505bde11 SM |
1588 | /* Pop the saved registers. */ \ |
1589 | while (fail_stack.frame < fail_stack.avail) \ | |
dc1e502d | 1590 | POP_FAILURE_REG_OR_COUNT (); \ |
fa9a63c5 | 1591 | \ |
dc4a2ee0 PE |
1592 | pat = POP_FAILURE_POINTER (); \ |
1593 | DEBUG_PRINT (" Popping pattern %p: ", pat); \ | |
505bde11 | 1594 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ |
fa9a63c5 RM |
1595 | \ |
1596 | /* If the saved string location is NULL, it came from an \ | |
1597 | on_failure_keep_string_jump opcode, and we want to throw away the \ | |
1598 | saved NULL, thus retaining our current position in the string. */ \ | |
01618498 | 1599 | str = POP_FAILURE_POINTER (); \ |
dc4a2ee0 | 1600 | DEBUG_PRINT (" Popping string %p: `", str); \ |
fa9a63c5 | 1601 | DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ |
dc4a2ee0 | 1602 | DEBUG_PRINT ("'\n"); \ |
fa9a63c5 | 1603 | \ |
505bde11 | 1604 | fail_stack.frame = POP_FAILURE_INT (); \ |
dc4a2ee0 | 1605 | DEBUG_PRINT (" Popping frame index: %zd\n", fail_stack.frame); \ |
fa9a63c5 | 1606 | \ |
505bde11 SM |
1607 | assert (fail_stack.avail >= 0); \ |
1608 | assert (fail_stack.frame <= fail_stack.avail); \ | |
fa9a63c5 | 1609 | \ |
fa9a63c5 | 1610 | DEBUG_STATEMENT (nfailure_points_popped++); \ |
505bde11 | 1611 | } while (0) /* POP_FAILURE_POINT */ |
fa9a63c5 RM |
1612 | |
1613 | ||
1614 | \f | |
fa9a63c5 | 1615 | /* Registers are set to a sentinel when they haven't yet matched. */ |
4bb91c68 | 1616 | #define REG_UNSET(e) ((e) == NULL) |
fa9a63c5 RM |
1617 | \f |
1618 | /* Subroutine declarations and macros for regex_compile. */ | |
1619 | ||
261cb4bb PE |
1620 | static reg_errcode_t regex_compile (re_char *pattern, size_t size, |
1621 | reg_syntax_t syntax, | |
1622 | struct re_pattern_buffer *bufp); | |
1623 | static void store_op1 (re_opcode_t op, unsigned char *loc, int arg); | |
1624 | static void store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2); | |
1625 | static void insert_op1 (re_opcode_t op, unsigned char *loc, | |
1626 | int arg, unsigned char *end); | |
1627 | static void insert_op2 (re_opcode_t op, unsigned char *loc, | |
1628 | int arg1, int arg2, unsigned char *end); | |
1629 | static boolean at_begline_loc_p (re_char *pattern, re_char *p, | |
1630 | reg_syntax_t syntax); | |
1631 | static boolean at_endline_loc_p (re_char *p, re_char *pend, | |
1632 | reg_syntax_t syntax); | |
1633 | static re_char *skip_one_char (re_char *p); | |
1634 | static int analyse_first (re_char *p, re_char *pend, | |
1635 | char *fastmap, const int multibyte); | |
fa9a63c5 | 1636 | |
fa9a63c5 | 1637 | /* Fetch the next character in the uncompiled pattern, with no |
4bb91c68 | 1638 | translation. */ |
36595814 | 1639 | #define PATFETCH(c) \ |
2d1675e4 SM |
1640 | do { \ |
1641 | int len; \ | |
1642 | if (p == pend) return REG_EEND; \ | |
62a6e103 | 1643 | c = RE_STRING_CHAR_AND_LENGTH (p, len, multibyte); \ |
2d1675e4 | 1644 | p += len; \ |
fa9a63c5 RM |
1645 | } while (0) |
1646 | ||
fa9a63c5 RM |
1647 | |
1648 | /* If `translate' is non-null, return translate[D], else just D. We | |
1649 | cast the subscript to translate because some data is declared as | |
1650 | `char *', to avoid warnings when a string constant is passed. But | |
1651 | when we use a character as a subscript we must make it unsigned. */ | |
6676cb1c | 1652 | #ifndef TRANSLATE |
0b32bf0e | 1653 | # define TRANSLATE(d) \ |
66f0296e | 1654 | (RE_TRANSLATE_P (translate) ? RE_TRANSLATE (translate, (d)) : (d)) |
6676cb1c | 1655 | #endif |
fa9a63c5 RM |
1656 | |
1657 | ||
1658 | /* Macros for outputting the compiled pattern into `buffer'. */ | |
1659 | ||
1660 | /* If the buffer isn't allocated when it comes in, use this. */ | |
1661 | #define INIT_BUF_SIZE 32 | |
1662 | ||
4bb91c68 | 1663 | /* Make sure we have at least N more bytes of space in buffer. */ |
fa9a63c5 | 1664 | #define GET_BUFFER_SPACE(n) \ |
01618498 | 1665 | while ((size_t) (b - bufp->buffer + (n)) > bufp->allocated) \ |
fa9a63c5 RM |
1666 | EXTEND_BUFFER () |
1667 | ||
1668 | /* Make sure we have one more byte of buffer space and then add C to it. */ | |
1669 | #define BUF_PUSH(c) \ | |
1670 | do { \ | |
1671 | GET_BUFFER_SPACE (1); \ | |
1672 | *b++ = (unsigned char) (c); \ | |
1673 | } while (0) | |
1674 | ||
1675 | ||
1676 | /* Ensure we have two more bytes of buffer space and then append C1 and C2. */ | |
1677 | #define BUF_PUSH_2(c1, c2) \ | |
1678 | do { \ | |
1679 | GET_BUFFER_SPACE (2); \ | |
1680 | *b++ = (unsigned char) (c1); \ | |
1681 | *b++ = (unsigned char) (c2); \ | |
1682 | } while (0) | |
1683 | ||
1684 | ||
fa9a63c5 | 1685 | /* Store a jump with opcode OP at LOC to location TO. We store a |
4bb91c68 | 1686 | relative address offset by the three bytes the jump itself occupies. */ |
fa9a63c5 RM |
1687 | #define STORE_JUMP(op, loc, to) \ |
1688 | store_op1 (op, loc, (to) - (loc) - 3) | |
1689 | ||
1690 | /* Likewise, for a two-argument jump. */ | |
1691 | #define STORE_JUMP2(op, loc, to, arg) \ | |
1692 | store_op2 (op, loc, (to) - (loc) - 3, arg) | |
1693 | ||
4bb91c68 | 1694 | /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ |
fa9a63c5 RM |
1695 | #define INSERT_JUMP(op, loc, to) \ |
1696 | insert_op1 (op, loc, (to) - (loc) - 3, b) | |
1697 | ||
1698 | /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ | |
1699 | #define INSERT_JUMP2(op, loc, to, arg) \ | |
1700 | insert_op2 (op, loc, (to) - (loc) - 3, arg, b) | |
1701 | ||
1702 | ||
1703 | /* This is not an arbitrary limit: the arguments which represent offsets | |
839966f3 | 1704 | into the pattern are two bytes long. So if 2^15 bytes turns out to |
fa9a63c5 | 1705 | be too small, many things would have to change. */ |
839966f3 KH |
1706 | # define MAX_BUF_SIZE (1L << 15) |
1707 | ||
fa9a63c5 RM |
1708 | /* Extend the buffer by twice its current size via realloc and |
1709 | reset the pointers that pointed into the old block to point to the | |
1710 | correct places in the new one. If extending the buffer results in it | |
4bb91c68 SM |
1711 | being larger than MAX_BUF_SIZE, then flag memory exhausted. */ |
1712 | #if __BOUNDED_POINTERS__ | |
1713 | # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) | |
381880b0 CY |
1714 | # define MOVE_BUFFER_POINTER(P) \ |
1715 | (__ptrlow (P) = new_buffer + (__ptrlow (P) - old_buffer), \ | |
1716 | SET_HIGH_BOUND (P), \ | |
1717 | __ptrvalue (P) = new_buffer + (__ptrvalue (P) - old_buffer)) | |
4bb91c68 SM |
1718 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
1719 | else \ | |
1720 | { \ | |
1721 | SET_HIGH_BOUND (b); \ | |
1722 | SET_HIGH_BOUND (begalt); \ | |
1723 | if (fixup_alt_jump) \ | |
1724 | SET_HIGH_BOUND (fixup_alt_jump); \ | |
1725 | if (laststart) \ | |
1726 | SET_HIGH_BOUND (laststart); \ | |
1727 | if (pending_exact) \ | |
1728 | SET_HIGH_BOUND (pending_exact); \ | |
1729 | } | |
1730 | #else | |
381880b0 | 1731 | # define MOVE_BUFFER_POINTER(P) ((P) = new_buffer + ((P) - old_buffer)) |
4bb91c68 SM |
1732 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND |
1733 | #endif | |
fa9a63c5 | 1734 | #define EXTEND_BUFFER() \ |
25fe55af | 1735 | do { \ |
381880b0 | 1736 | unsigned char *old_buffer = bufp->buffer; \ |
25fe55af | 1737 | if (bufp->allocated == MAX_BUF_SIZE) \ |
fa9a63c5 RM |
1738 | return REG_ESIZE; \ |
1739 | bufp->allocated <<= 1; \ | |
1740 | if (bufp->allocated > MAX_BUF_SIZE) \ | |
25fe55af | 1741 | bufp->allocated = MAX_BUF_SIZE; \ |
01618498 | 1742 | RETALLOC (bufp->buffer, bufp->allocated, unsigned char); \ |
fa9a63c5 RM |
1743 | if (bufp->buffer == NULL) \ |
1744 | return REG_ESPACE; \ | |
1745 | /* If the buffer moved, move all the pointers into it. */ \ | |
1746 | if (old_buffer != bufp->buffer) \ | |
1747 | { \ | |
381880b0 | 1748 | unsigned char *new_buffer = bufp->buffer; \ |
4bb91c68 SM |
1749 | MOVE_BUFFER_POINTER (b); \ |
1750 | MOVE_BUFFER_POINTER (begalt); \ | |
25fe55af | 1751 | if (fixup_alt_jump) \ |
4bb91c68 | 1752 | MOVE_BUFFER_POINTER (fixup_alt_jump); \ |
25fe55af | 1753 | if (laststart) \ |
4bb91c68 | 1754 | MOVE_BUFFER_POINTER (laststart); \ |
25fe55af | 1755 | if (pending_exact) \ |
4bb91c68 | 1756 | MOVE_BUFFER_POINTER (pending_exact); \ |
fa9a63c5 | 1757 | } \ |
4bb91c68 | 1758 | ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
fa9a63c5 RM |
1759 | } while (0) |
1760 | ||
1761 | ||
1762 | /* Since we have one byte reserved for the register number argument to | |
1763 | {start,stop}_memory, the maximum number of groups we can report | |
1764 | things about is what fits in that byte. */ | |
1765 | #define MAX_REGNUM 255 | |
1766 | ||
1767 | /* But patterns can have more than `MAX_REGNUM' registers. We just | |
1768 | ignore the excess. */ | |
098d42af | 1769 | typedef int regnum_t; |
fa9a63c5 RM |
1770 | |
1771 | ||
1772 | /* Macros for the compile stack. */ | |
1773 | ||
1774 | /* Since offsets can go either forwards or backwards, this type needs to | |
4bb91c68 SM |
1775 | be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ |
1776 | /* int may be not enough when sizeof(int) == 2. */ | |
1777 | typedef long pattern_offset_t; | |
fa9a63c5 RM |
1778 | |
1779 | typedef struct | |
1780 | { | |
1781 | pattern_offset_t begalt_offset; | |
1782 | pattern_offset_t fixup_alt_jump; | |
5e69f11e | 1783 | pattern_offset_t laststart_offset; |
fa9a63c5 RM |
1784 | regnum_t regnum; |
1785 | } compile_stack_elt_t; | |
1786 | ||
1787 | ||
1788 | typedef struct | |
1789 | { | |
1790 | compile_stack_elt_t *stack; | |
d1dfb56c EZ |
1791 | size_t size; |
1792 | size_t avail; /* Offset of next open position. */ | |
fa9a63c5 RM |
1793 | } compile_stack_type; |
1794 | ||
1795 | ||
1796 | #define INIT_COMPILE_STACK_SIZE 32 | |
1797 | ||
1798 | #define COMPILE_STACK_EMPTY (compile_stack.avail == 0) | |
1799 | #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) | |
1800 | ||
4bb91c68 | 1801 | /* The next available element. */ |
fa9a63c5 RM |
1802 | #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) |
1803 | ||
0caaedb1 PE |
1804 | /* Explicit quit checking is needed for Emacs, which uses polling to |
1805 | process input events. */ | |
1806 | #ifdef emacs | |
77d11aec RS |
1807 | # define IMMEDIATE_QUIT_CHECK \ |
1808 | do { \ | |
1809 | if (immediate_quit) QUIT; \ | |
1810 | } while (0) | |
1811 | #else | |
1812 | # define IMMEDIATE_QUIT_CHECK ((void)0) | |
1813 | #endif | |
1814 | \f | |
b18215fc RS |
1815 | /* Structure to manage work area for range table. */ |
1816 | struct range_table_work_area | |
1817 | { | |
1818 | int *table; /* actual work area. */ | |
1819 | int allocated; /* allocated size for work area in bytes. */ | |
7814e705 | 1820 | int used; /* actually used size in words. */ |
96cc36cc | 1821 | int bits; /* flag to record character classes */ |
b18215fc RS |
1822 | }; |
1823 | ||
78779650 AS |
1824 | #ifdef emacs |
1825 | ||
77d11aec RS |
1826 | /* Make sure that WORK_AREA can hold more N multibyte characters. |
1827 | This is used only in set_image_of_range and set_image_of_range_1. | |
1828 | It expects WORK_AREA to be a pointer. | |
1829 | If it can't get the space, it returns from the surrounding function. */ | |
1830 | ||
1831 | #define EXTEND_RANGE_TABLE(work_area, n) \ | |
1832 | do { \ | |
8f924df7 | 1833 | if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \ |
77d11aec | 1834 | { \ |
8f924df7 KH |
1835 | extend_range_table_work_area (&work_area); \ |
1836 | if ((work_area).table == 0) \ | |
77d11aec RS |
1837 | return (REG_ESPACE); \ |
1838 | } \ | |
b18215fc RS |
1839 | } while (0) |
1840 | ||
96cc36cc RS |
1841 | #define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \ |
1842 | (work_area).bits |= (bit) | |
1843 | ||
b18215fc RS |
1844 | /* Set a range (RANGE_START, RANGE_END) to WORK_AREA. */ |
1845 | #define SET_RANGE_TABLE_WORK_AREA(work_area, range_start, range_end) \ | |
77d11aec | 1846 | do { \ |
8f924df7 | 1847 | EXTEND_RANGE_TABLE ((work_area), 2); \ |
b18215fc RS |
1848 | (work_area).table[(work_area).used++] = (range_start); \ |
1849 | (work_area).table[(work_area).used++] = (range_end); \ | |
1850 | } while (0) | |
1851 | ||
78779650 AS |
1852 | #endif /* emacs */ |
1853 | ||
7814e705 | 1854 | /* Free allocated memory for WORK_AREA. */ |
b18215fc RS |
1855 | #define FREE_RANGE_TABLE_WORK_AREA(work_area) \ |
1856 | do { \ | |
1857 | if ((work_area).table) \ | |
1858 | free ((work_area).table); \ | |
1859 | } while (0) | |
1860 | ||
96cc36cc | 1861 | #define CLEAR_RANGE_TABLE_WORK_USED(work_area) ((work_area).used = 0, (work_area).bits = 0) |
b18215fc | 1862 | #define RANGE_TABLE_WORK_USED(work_area) ((work_area).used) |
96cc36cc | 1863 | #define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits) |
b18215fc | 1864 | #define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i]) |
78779650 AS |
1865 | |
1866 | /* Bits used to implement the multibyte-part of the various character classes | |
1867 | such as [:alnum:] in a charset's range table. */ | |
1868 | #define BIT_WORD 0x1 | |
1869 | #define BIT_LOWER 0x2 | |
1870 | #define BIT_PUNCT 0x4 | |
1871 | #define BIT_SPACE 0x8 | |
1872 | #define BIT_UPPER 0x10 | |
1873 | #define BIT_MULTIBYTE 0x20 | |
77d11aec | 1874 | \f |
b18215fc | 1875 | |
fa9a63c5 | 1876 | /* Set the bit for character C in a list. */ |
01618498 | 1877 | #define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH)) |
fa9a63c5 RM |
1878 | |
1879 | ||
bf216479 KH |
1880 | #ifdef emacs |
1881 | ||
cf9c99bc KH |
1882 | /* Store characters in the range FROM to TO in the bitmap at B (for |
1883 | ASCII and unibyte characters) and WORK_AREA (for multibyte | |
1884 | characters) while translating them and paying attention to the | |
1885 | continuity of translated characters. | |
8f924df7 | 1886 | |
cf9c99bc KH |
1887 | Implementation note: It is better to implement these fairly big |
1888 | macros by a function, but it's not that easy because macros called | |
8f924df7 | 1889 | in this macro assume various local variables already declared. */ |
bf216479 | 1890 | |
cf9c99bc KH |
1891 | /* Both FROM and TO are ASCII characters. */ |
1892 | ||
1893 | #define SETUP_ASCII_RANGE(work_area, FROM, TO) \ | |
1894 | do { \ | |
1895 | int C0, C1; \ | |
1896 | \ | |
1897 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1898 | { \ | |
1899 | C1 = TRANSLATE (C0); \ | |
1900 | if (! ASCII_CHAR_P (C1)) \ | |
1901 | { \ | |
1902 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
1903 | if ((C1 = RE_CHAR_TO_UNIBYTE (C1)) < 0) \ | |
1904 | C1 = C0; \ | |
1905 | } \ | |
1906 | SET_LIST_BIT (C1); \ | |
1907 | } \ | |
1908 | } while (0) | |
1909 | ||
1910 | ||
1911 | /* Both FROM and TO are unibyte characters (0x80..0xFF). */ | |
1912 | ||
1913 | #define SETUP_UNIBYTE_RANGE(work_area, FROM, TO) \ | |
1914 | do { \ | |
1915 | int C0, C1, C2, I; \ | |
1916 | int USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1917 | \ | |
1918 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1919 | { \ | |
1920 | C1 = RE_CHAR_TO_MULTIBYTE (C0); \ | |
1921 | if (CHAR_BYTE8_P (C1)) \ | |
1922 | SET_LIST_BIT (C0); \ | |
1923 | else \ | |
1924 | { \ | |
1925 | C2 = TRANSLATE (C1); \ | |
1926 | if (C2 == C1 \ | |
1927 | || (C1 = RE_CHAR_TO_UNIBYTE (C2)) < 0) \ | |
1928 | C1 = C0; \ | |
1929 | SET_LIST_BIT (C1); \ | |
1930 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1931 | { \ | |
1932 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1933 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1934 | \ | |
1935 | if (C2 >= from - 1 && C2 <= to + 1) \ | |
1936 | { \ | |
1937 | if (C2 == from - 1) \ | |
1938 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1939 | else if (C2 == to + 1) \ | |
1940 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
1941 | break; \ | |
1942 | } \ | |
1943 | } \ | |
1944 | if (I < USED) \ | |
1945 | SET_RANGE_TABLE_WORK_AREA ((work_area), C2, C2); \ | |
1946 | } \ | |
1947 | } \ | |
1948 | } while (0) | |
1949 | ||
1950 | ||
78edd3b7 | 1951 | /* Both FROM and TO are multibyte characters. */ |
cf9c99bc KH |
1952 | |
1953 | #define SETUP_MULTIBYTE_RANGE(work_area, FROM, TO) \ | |
1954 | do { \ | |
1955 | int C0, C1, C2, I, USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1956 | \ | |
1957 | SET_RANGE_TABLE_WORK_AREA ((work_area), (FROM), (TO)); \ | |
1958 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1959 | { \ | |
1960 | C1 = TRANSLATE (C0); \ | |
1961 | if ((C2 = RE_CHAR_TO_UNIBYTE (C1)) >= 0 \ | |
1962 | || (C1 != C0 && (C2 = RE_CHAR_TO_UNIBYTE (C0)) >= 0)) \ | |
1963 | SET_LIST_BIT (C2); \ | |
1964 | if (C1 >= (FROM) && C1 <= (TO)) \ | |
1965 | continue; \ | |
1966 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1967 | { \ | |
1968 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1969 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1970 | \ | |
1971 | if (C1 >= from - 1 && C1 <= to + 1) \ | |
1972 | { \ | |
1973 | if (C1 == from - 1) \ | |
1974 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1975 | else if (C1 == to + 1) \ | |
1976 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
1977 | break; \ | |
1978 | } \ | |
1979 | } \ | |
1980 | if (I < USED) \ | |
1981 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
1982 | } \ | |
bf216479 KH |
1983 | } while (0) |
1984 | ||
1985 | #endif /* emacs */ | |
1986 | ||
fa9a63c5 | 1987 | /* Get the next unsigned number in the uncompiled pattern. */ |
4618713a | 1988 | #define GET_INTERVAL_COUNT(num) \ |
c72b0edd SM |
1989 | do { \ |
1990 | if (p == pend) \ | |
1991 | FREE_STACK_RETURN (REG_EBRACE); \ | |
1992 | else \ | |
1993 | { \ | |
1994 | PATFETCH (c); \ | |
1995 | while ('0' <= c && c <= '9') \ | |
1996 | { \ | |
c72b0edd SM |
1997 | if (num < 0) \ |
1998 | num = 0; \ | |
4618713a | 1999 | if (RE_DUP_MAX / 10 - (RE_DUP_MAX % 10 < c - '0') < num) \ |
c72b0edd | 2000 | FREE_STACK_RETURN (REG_BADBR); \ |
4618713a | 2001 | num = num * 10 + c - '0'; \ |
c72b0edd SM |
2002 | if (p == pend) \ |
2003 | FREE_STACK_RETURN (REG_EBRACE); \ | |
2004 | PATFETCH (c); \ | |
2005 | } \ | |
2006 | } \ | |
2007 | } while (0) | |
77d11aec | 2008 | \f |
1fdab503 | 2009 | #if ! WIDE_CHAR_SUPPORT |
01618498 | 2010 | |
14473664 | 2011 | /* Map a string to the char class it names (if any). */ |
1fdab503 | 2012 | re_wctype_t |
29abe551 | 2013 | re_wctype (const_re_char *str) |
14473664 | 2014 | { |
5b0534c8 | 2015 | const char *string = (const char *) str; |
14473664 SM |
2016 | if (STREQ (string, "alnum")) return RECC_ALNUM; |
2017 | else if (STREQ (string, "alpha")) return RECC_ALPHA; | |
2018 | else if (STREQ (string, "word")) return RECC_WORD; | |
2019 | else if (STREQ (string, "ascii")) return RECC_ASCII; | |
2020 | else if (STREQ (string, "nonascii")) return RECC_NONASCII; | |
2021 | else if (STREQ (string, "graph")) return RECC_GRAPH; | |
2022 | else if (STREQ (string, "lower")) return RECC_LOWER; | |
2023 | else if (STREQ (string, "print")) return RECC_PRINT; | |
2024 | else if (STREQ (string, "punct")) return RECC_PUNCT; | |
2025 | else if (STREQ (string, "space")) return RECC_SPACE; | |
2026 | else if (STREQ (string, "upper")) return RECC_UPPER; | |
2027 | else if (STREQ (string, "unibyte")) return RECC_UNIBYTE; | |
2028 | else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE; | |
2029 | else if (STREQ (string, "digit")) return RECC_DIGIT; | |
2030 | else if (STREQ (string, "xdigit")) return RECC_XDIGIT; | |
2031 | else if (STREQ (string, "cntrl")) return RECC_CNTRL; | |
2032 | else if (STREQ (string, "blank")) return RECC_BLANK; | |
2033 | else return 0; | |
2034 | } | |
2035 | ||
e0f24100 | 2036 | /* True if CH is in the char class CC. */ |
1fdab503 | 2037 | boolean |
971de7fb | 2038 | re_iswctype (int ch, re_wctype_t cc) |
14473664 SM |
2039 | { |
2040 | switch (cc) | |
2041 | { | |
f3fcc40d AS |
2042 | case RECC_ALNUM: return ISALNUM (ch) != 0; |
2043 | case RECC_ALPHA: return ISALPHA (ch) != 0; | |
2044 | case RECC_BLANK: return ISBLANK (ch) != 0; | |
2045 | case RECC_CNTRL: return ISCNTRL (ch) != 0; | |
2046 | case RECC_DIGIT: return ISDIGIT (ch) != 0; | |
2047 | case RECC_GRAPH: return ISGRAPH (ch) != 0; | |
2048 | case RECC_LOWER: return ISLOWER (ch) != 0; | |
2049 | case RECC_PRINT: return ISPRINT (ch) != 0; | |
2050 | case RECC_PUNCT: return ISPUNCT (ch) != 0; | |
2051 | case RECC_SPACE: return ISSPACE (ch) != 0; | |
2052 | case RECC_UPPER: return ISUPPER (ch) != 0; | |
2053 | case RECC_XDIGIT: return ISXDIGIT (ch) != 0; | |
2054 | case RECC_ASCII: return IS_REAL_ASCII (ch) != 0; | |
213bd7f2 | 2055 | case RECC_NONASCII: return !IS_REAL_ASCII (ch); |
f3fcc40d | 2056 | case RECC_UNIBYTE: return ISUNIBYTE (ch) != 0; |
213bd7f2 | 2057 | case RECC_MULTIBYTE: return !ISUNIBYTE (ch); |
f3fcc40d | 2058 | case RECC_WORD: return ISWORD (ch) != 0; |
0cdd06f8 SM |
2059 | case RECC_ERROR: return false; |
2060 | default: | |
5e617bc2 | 2061 | abort (); |
14473664 SM |
2062 | } |
2063 | } | |
fa9a63c5 | 2064 | |
14473664 SM |
2065 | /* Return a bit-pattern to use in the range-table bits to match multibyte |
2066 | chars of class CC. */ | |
2067 | static int | |
971de7fb | 2068 | re_wctype_to_bit (re_wctype_t cc) |
14473664 SM |
2069 | { |
2070 | switch (cc) | |
2071 | { | |
2072 | case RECC_NONASCII: case RECC_PRINT: case RECC_GRAPH: | |
0cdd06f8 SM |
2073 | case RECC_MULTIBYTE: return BIT_MULTIBYTE; |
2074 | case RECC_ALPHA: case RECC_ALNUM: case RECC_WORD: return BIT_WORD; | |
2075 | case RECC_LOWER: return BIT_LOWER; | |
2076 | case RECC_UPPER: return BIT_UPPER; | |
2077 | case RECC_PUNCT: return BIT_PUNCT; | |
2078 | case RECC_SPACE: return BIT_SPACE; | |
14473664 | 2079 | case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL: |
0cdd06f8 SM |
2080 | case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0; |
2081 | default: | |
5e617bc2 | 2082 | abort (); |
14473664 SM |
2083 | } |
2084 | } | |
2085 | #endif | |
77d11aec RS |
2086 | \f |
2087 | /* Filling in the work area of a range. */ | |
2088 | ||
2089 | /* Actually extend the space in WORK_AREA. */ | |
2090 | ||
2091 | static void | |
971de7fb | 2092 | extend_range_table_work_area (struct range_table_work_area *work_area) |
177c0ea7 | 2093 | { |
77d11aec | 2094 | work_area->allocated += 16 * sizeof (int); |
38182d90 | 2095 | work_area->table = realloc (work_area->table, work_area->allocated); |
77d11aec RS |
2096 | } |
2097 | ||
8f924df7 | 2098 | #if 0 |
77d11aec RS |
2099 | #ifdef emacs |
2100 | ||
2101 | /* Carefully find the ranges of codes that are equivalent | |
2102 | under case conversion to the range start..end when passed through | |
2103 | TRANSLATE. Handle the case where non-letters can come in between | |
2104 | two upper-case letters (which happens in Latin-1). | |
2105 | Also handle the case of groups of more than 2 case-equivalent chars. | |
2106 | ||
2107 | The basic method is to look at consecutive characters and see | |
2108 | if they can form a run that can be handled as one. | |
2109 | ||
2110 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2111 | ||
2112 | static int | |
1dae0f0a AS |
2113 | set_image_of_range_1 (struct range_table_work_area *work_area, |
2114 | re_wchar_t start, re_wchar_t end, | |
2115 | RE_TRANSLATE_TYPE translate) | |
77d11aec RS |
2116 | { |
2117 | /* `one_case' indicates a character, or a run of characters, | |
2118 | each of which is an isolate (no case-equivalents). | |
2119 | This includes all ASCII non-letters. | |
2120 | ||
2121 | `two_case' indicates a character, or a run of characters, | |
2122 | each of which has two case-equivalent forms. | |
2123 | This includes all ASCII letters. | |
2124 | ||
2125 | `strange' indicates a character that has more than one | |
2126 | case-equivalent. */ | |
177c0ea7 | 2127 | |
77d11aec RS |
2128 | enum case_type {one_case, two_case, strange}; |
2129 | ||
2130 | /* Describe the run that is in progress, | |
2131 | which the next character can try to extend. | |
2132 | If run_type is strange, that means there really is no run. | |
2133 | If run_type is one_case, then run_start...run_end is the run. | |
2134 | If run_type is two_case, then the run is run_start...run_end, | |
2135 | and the case-equivalents end at run_eqv_end. */ | |
2136 | ||
2137 | enum case_type run_type = strange; | |
2138 | int run_start, run_end, run_eqv_end; | |
2139 | ||
2140 | Lisp_Object eqv_table; | |
2141 | ||
2142 | if (!RE_TRANSLATE_P (translate)) | |
2143 | { | |
b7c12565 | 2144 | EXTEND_RANGE_TABLE (work_area, 2); |
77d11aec RS |
2145 | work_area->table[work_area->used++] = (start); |
2146 | work_area->table[work_area->used++] = (end); | |
b7c12565 | 2147 | return -1; |
77d11aec RS |
2148 | } |
2149 | ||
2150 | eqv_table = XCHAR_TABLE (translate)->extras[2]; | |
99633e97 | 2151 | |
77d11aec RS |
2152 | for (; start <= end; start++) |
2153 | { | |
2154 | enum case_type this_type; | |
2155 | int eqv = RE_TRANSLATE (eqv_table, start); | |
2156 | int minchar, maxchar; | |
2157 | ||
2158 | /* Classify this character */ | |
2159 | if (eqv == start) | |
2160 | this_type = one_case; | |
2161 | else if (RE_TRANSLATE (eqv_table, eqv) == start) | |
2162 | this_type = two_case; | |
2163 | else | |
2164 | this_type = strange; | |
2165 | ||
2166 | if (start < eqv) | |
2167 | minchar = start, maxchar = eqv; | |
2168 | else | |
2169 | minchar = eqv, maxchar = start; | |
2170 | ||
2171 | /* Can this character extend the run in progress? */ | |
2172 | if (this_type == strange || this_type != run_type | |
2173 | || !(minchar == run_end + 1 | |
2174 | && (run_type == two_case | |
2175 | ? maxchar == run_eqv_end + 1 : 1))) | |
2176 | { | |
2177 | /* No, end the run. | |
2178 | Record each of its equivalent ranges. */ | |
2179 | if (run_type == one_case) | |
2180 | { | |
2181 | EXTEND_RANGE_TABLE (work_area, 2); | |
2182 | work_area->table[work_area->used++] = run_start; | |
2183 | work_area->table[work_area->used++] = run_end; | |
2184 | } | |
2185 | else if (run_type == two_case) | |
2186 | { | |
2187 | EXTEND_RANGE_TABLE (work_area, 4); | |
2188 | work_area->table[work_area->used++] = run_start; | |
2189 | work_area->table[work_area->used++] = run_end; | |
2190 | work_area->table[work_area->used++] | |
2191 | = RE_TRANSLATE (eqv_table, run_start); | |
2192 | work_area->table[work_area->used++] | |
2193 | = RE_TRANSLATE (eqv_table, run_end); | |
2194 | } | |
2195 | run_type = strange; | |
2196 | } | |
177c0ea7 | 2197 | |
77d11aec RS |
2198 | if (this_type == strange) |
2199 | { | |
2200 | /* For a strange character, add each of its equivalents, one | |
2201 | by one. Don't start a range. */ | |
2202 | do | |
2203 | { | |
2204 | EXTEND_RANGE_TABLE (work_area, 2); | |
2205 | work_area->table[work_area->used++] = eqv; | |
2206 | work_area->table[work_area->used++] = eqv; | |
2207 | eqv = RE_TRANSLATE (eqv_table, eqv); | |
2208 | } | |
2209 | while (eqv != start); | |
2210 | } | |
2211 | ||
2212 | /* Add this char to the run, or start a new run. */ | |
2213 | else if (run_type == strange) | |
2214 | { | |
2215 | /* Initialize a new range. */ | |
2216 | run_type = this_type; | |
2217 | run_start = start; | |
2218 | run_end = start; | |
2219 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2220 | } | |
2221 | else | |
2222 | { | |
2223 | /* Extend a running range. */ | |
2224 | run_end = minchar; | |
2225 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2226 | } | |
2227 | } | |
2228 | ||
2229 | /* If a run is still in progress at the end, finish it now | |
2230 | by recording its equivalent ranges. */ | |
2231 | if (run_type == one_case) | |
2232 | { | |
2233 | EXTEND_RANGE_TABLE (work_area, 2); | |
2234 | work_area->table[work_area->used++] = run_start; | |
2235 | work_area->table[work_area->used++] = run_end; | |
2236 | } | |
2237 | else if (run_type == two_case) | |
2238 | { | |
2239 | EXTEND_RANGE_TABLE (work_area, 4); | |
2240 | work_area->table[work_area->used++] = run_start; | |
2241 | work_area->table[work_area->used++] = run_end; | |
2242 | work_area->table[work_area->used++] | |
2243 | = RE_TRANSLATE (eqv_table, run_start); | |
2244 | work_area->table[work_area->used++] | |
2245 | = RE_TRANSLATE (eqv_table, run_end); | |
2246 | } | |
2247 | ||
2248 | return -1; | |
2249 | } | |
36595814 | 2250 | |
77d11aec | 2251 | #endif /* emacs */ |
36595814 | 2252 | |
2b34df4e | 2253 | /* Record the image of the range start..end when passed through |
36595814 SM |
2254 | TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end) |
2255 | and is not even necessarily contiguous. | |
b7c12565 RS |
2256 | Normally we approximate it with the smallest contiguous range that contains |
2257 | all the chars we need. However, for Latin-1 we go to extra effort | |
2258 | to do a better job. | |
2259 | ||
2260 | This function is not called for ASCII ranges. | |
77d11aec RS |
2261 | |
2262 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2263 | ||
2264 | static int | |
1dae0f0a AS |
2265 | set_image_of_range (struct range_table_work_area *work_area, |
2266 | re_wchar_t start, re_wchar_t end, | |
2267 | RE_TRANSLATE_TYPE translate) | |
36595814 | 2268 | { |
77d11aec RS |
2269 | re_wchar_t cmin, cmax; |
2270 | ||
2271 | #ifdef emacs | |
2272 | /* For Latin-1 ranges, use set_image_of_range_1 | |
2273 | to get proper handling of ranges that include letters and nonletters. | |
b7c12565 | 2274 | For a range that includes the whole of Latin-1, this is not necessary. |
77d11aec | 2275 | For other character sets, we don't bother to get this right. */ |
b7c12565 RS |
2276 | if (RE_TRANSLATE_P (translate) && start < 04400 |
2277 | && !(start < 04200 && end >= 04377)) | |
77d11aec | 2278 | { |
b7c12565 | 2279 | int newend; |
77d11aec | 2280 | int tem; |
b7c12565 RS |
2281 | newend = end; |
2282 | if (newend > 04377) | |
2283 | newend = 04377; | |
2284 | tem = set_image_of_range_1 (work_area, start, newend, translate); | |
77d11aec RS |
2285 | if (tem > 0) |
2286 | return tem; | |
2287 | ||
2288 | start = 04400; | |
2289 | if (end < 04400) | |
2290 | return -1; | |
2291 | } | |
2292 | #endif | |
2293 | ||
b7c12565 RS |
2294 | EXTEND_RANGE_TABLE (work_area, 2); |
2295 | work_area->table[work_area->used++] = (start); | |
2296 | work_area->table[work_area->used++] = (end); | |
2297 | ||
2298 | cmin = -1, cmax = -1; | |
77d11aec | 2299 | |
36595814 | 2300 | if (RE_TRANSLATE_P (translate)) |
b7c12565 RS |
2301 | { |
2302 | int ch; | |
77d11aec | 2303 | |
b7c12565 RS |
2304 | for (ch = start; ch <= end; ch++) |
2305 | { | |
2306 | re_wchar_t c = TRANSLATE (ch); | |
2307 | if (! (start <= c && c <= end)) | |
2308 | { | |
2309 | if (cmin == -1) | |
2310 | cmin = c, cmax = c; | |
2311 | else | |
2312 | { | |
2313 | cmin = MIN (cmin, c); | |
2314 | cmax = MAX (cmax, c); | |
2315 | } | |
2316 | } | |
2317 | } | |
2318 | ||
2319 | if (cmin != -1) | |
2320 | { | |
2321 | EXTEND_RANGE_TABLE (work_area, 2); | |
2322 | work_area->table[work_area->used++] = (cmin); | |
2323 | work_area->table[work_area->used++] = (cmax); | |
2324 | } | |
2325 | } | |
36595814 | 2326 | |
77d11aec RS |
2327 | return -1; |
2328 | } | |
8f924df7 | 2329 | #endif /* 0 */ |
fa9a63c5 RM |
2330 | \f |
2331 | #ifndef MATCH_MAY_ALLOCATE | |
2332 | ||
2333 | /* If we cannot allocate large objects within re_match_2_internal, | |
2334 | we make the fail stack and register vectors global. | |
2335 | The fail stack, we grow to the maximum size when a regexp | |
2336 | is compiled. | |
2337 | The register vectors, we adjust in size each time we | |
2338 | compile a regexp, according to the number of registers it needs. */ | |
2339 | ||
2340 | static fail_stack_type fail_stack; | |
2341 | ||
2342 | /* Size with which the following vectors are currently allocated. | |
2343 | That is so we can make them bigger as needed, | |
4bb91c68 | 2344 | but never make them smaller. */ |
fa9a63c5 RM |
2345 | static int regs_allocated_size; |
2346 | ||
66f0296e SM |
2347 | static re_char ** regstart, ** regend; |
2348 | static re_char **best_regstart, **best_regend; | |
fa9a63c5 RM |
2349 | |
2350 | /* Make the register vectors big enough for NUM_REGS registers, | |
4bb91c68 | 2351 | but don't make them smaller. */ |
fa9a63c5 RM |
2352 | |
2353 | static | |
1dae0f0a | 2354 | regex_grow_registers (int num_regs) |
fa9a63c5 RM |
2355 | { |
2356 | if (num_regs > regs_allocated_size) | |
2357 | { | |
66f0296e SM |
2358 | RETALLOC_IF (regstart, num_regs, re_char *); |
2359 | RETALLOC_IF (regend, num_regs, re_char *); | |
2360 | RETALLOC_IF (best_regstart, num_regs, re_char *); | |
2361 | RETALLOC_IF (best_regend, num_regs, re_char *); | |
fa9a63c5 RM |
2362 | |
2363 | regs_allocated_size = num_regs; | |
2364 | } | |
2365 | } | |
2366 | ||
2367 | #endif /* not MATCH_MAY_ALLOCATE */ | |
2368 | \f | |
261cb4bb PE |
2369 | static boolean group_in_compile_stack (compile_stack_type compile_stack, |
2370 | regnum_t regnum); | |
99633e97 | 2371 | |
fa9a63c5 RM |
2372 | /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. |
2373 | Returns one of error codes defined in `regex.h', or zero for success. | |
2374 | ||
2375 | Assumes the `allocated' (and perhaps `buffer') and `translate' | |
2376 | fields are set in BUFP on entry. | |
2377 | ||
2378 | If it succeeds, results are put in BUFP (if it returns an error, the | |
2379 | contents of BUFP are undefined): | |
2380 | `buffer' is the compiled pattern; | |
2381 | `syntax' is set to SYNTAX; | |
2382 | `used' is set to the length of the compiled pattern; | |
2383 | `fastmap_accurate' is zero; | |
2384 | `re_nsub' is the number of subexpressions in PATTERN; | |
2385 | `not_bol' and `not_eol' are zero; | |
5e69f11e | 2386 | |
c0f9ea08 | 2387 | The `fastmap' field is neither examined nor set. */ |
fa9a63c5 | 2388 | |
505bde11 SM |
2389 | /* Insert the `jump' from the end of last alternative to "here". |
2390 | The space for the jump has already been allocated. */ | |
2391 | #define FIXUP_ALT_JUMP() \ | |
2392 | do { \ | |
2393 | if (fixup_alt_jump) \ | |
2394 | STORE_JUMP (jump, fixup_alt_jump, b); \ | |
2395 | } while (0) | |
2396 | ||
2397 | ||
fa9a63c5 RM |
2398 | /* Return, freeing storage we allocated. */ |
2399 | #define FREE_STACK_RETURN(value) \ | |
b18215fc RS |
2400 | do { \ |
2401 | FREE_RANGE_TABLE_WORK_AREA (range_table_work); \ | |
2402 | free (compile_stack.stack); \ | |
2403 | return value; \ | |
2404 | } while (0) | |
fa9a63c5 RM |
2405 | |
2406 | static reg_errcode_t | |
29abe551 PE |
2407 | regex_compile (const_re_char *pattern, size_t size, reg_syntax_t syntax, |
2408 | struct re_pattern_buffer *bufp) | |
fa9a63c5 | 2409 | { |
01618498 SM |
2410 | /* We fetch characters from PATTERN here. */ |
2411 | register re_wchar_t c, c1; | |
5e69f11e | 2412 | |
fa9a63c5 RM |
2413 | /* Points to the end of the buffer, where we should append. */ |
2414 | register unsigned char *b; | |
5e69f11e | 2415 | |
fa9a63c5 RM |
2416 | /* Keeps track of unclosed groups. */ |
2417 | compile_stack_type compile_stack; | |
2418 | ||
2419 | /* Points to the current (ending) position in the pattern. */ | |
22336245 RS |
2420 | #ifdef AIX |
2421 | /* `const' makes AIX compiler fail. */ | |
66f0296e | 2422 | unsigned char *p = pattern; |
22336245 | 2423 | #else |
66f0296e | 2424 | re_char *p = pattern; |
22336245 | 2425 | #endif |
66f0296e | 2426 | re_char *pend = pattern + size; |
5e69f11e | 2427 | |
fa9a63c5 | 2428 | /* How to translate the characters in the pattern. */ |
6676cb1c | 2429 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 RM |
2430 | |
2431 | /* Address of the count-byte of the most recently inserted `exactn' | |
2432 | command. This makes it possible to tell if a new exact-match | |
2433 | character can be added to that command or if the character requires | |
2434 | a new `exactn' command. */ | |
2435 | unsigned char *pending_exact = 0; | |
2436 | ||
2437 | /* Address of start of the most recently finished expression. | |
2438 | This tells, e.g., postfix * where to find the start of its | |
2439 | operand. Reset at the beginning of groups and alternatives. */ | |
2440 | unsigned char *laststart = 0; | |
2441 | ||
2442 | /* Address of beginning of regexp, or inside of last group. */ | |
2443 | unsigned char *begalt; | |
2444 | ||
2445 | /* Place in the uncompiled pattern (i.e., the {) to | |
2446 | which to go back if the interval is invalid. */ | |
66f0296e | 2447 | re_char *beg_interval; |
5e69f11e | 2448 | |
fa9a63c5 | 2449 | /* Address of the place where a forward jump should go to the end of |
7814e705 | 2450 | the containing expression. Each alternative of an `or' -- except the |
fa9a63c5 RM |
2451 | last -- ends with a forward jump of this sort. */ |
2452 | unsigned char *fixup_alt_jump = 0; | |
2453 | ||
b18215fc RS |
2454 | /* Work area for range table of charset. */ |
2455 | struct range_table_work_area range_table_work; | |
2456 | ||
2d1675e4 SM |
2457 | /* If the object matched can contain multibyte characters. */ |
2458 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | |
2459 | ||
f9b0fd99 RS |
2460 | /* Nonzero if we have pushed down into a subpattern. */ |
2461 | int in_subpattern = 0; | |
2462 | ||
2463 | /* These hold the values of p, pattern, and pend from the main | |
2464 | pattern when we have pushed into a subpattern. */ | |
da053e48 PE |
2465 | re_char *main_p IF_LINT (= NULL); |
2466 | re_char *main_pattern IF_LINT (= NULL); | |
2467 | re_char *main_pend IF_LINT (= NULL); | |
f9b0fd99 | 2468 | |
fa9a63c5 | 2469 | #ifdef DEBUG |
99633e97 | 2470 | debug++; |
dc4a2ee0 | 2471 | DEBUG_PRINT ("\nCompiling pattern: "); |
99633e97 | 2472 | if (debug > 0) |
fa9a63c5 RM |
2473 | { |
2474 | unsigned debug_count; | |
5e69f11e | 2475 | |
fa9a63c5 | 2476 | for (debug_count = 0; debug_count < size; debug_count++) |
25fe55af | 2477 | putchar (pattern[debug_count]); |
fa9a63c5 RM |
2478 | putchar ('\n'); |
2479 | } | |
2480 | #endif /* DEBUG */ | |
2481 | ||
2482 | /* Initialize the compile stack. */ | |
2483 | compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); | |
2484 | if (compile_stack.stack == NULL) | |
2485 | return REG_ESPACE; | |
2486 | ||
2487 | compile_stack.size = INIT_COMPILE_STACK_SIZE; | |
2488 | compile_stack.avail = 0; | |
2489 | ||
b18215fc RS |
2490 | range_table_work.table = 0; |
2491 | range_table_work.allocated = 0; | |
2492 | ||
fa9a63c5 RM |
2493 | /* Initialize the pattern buffer. */ |
2494 | bufp->syntax = syntax; | |
2495 | bufp->fastmap_accurate = 0; | |
2496 | bufp->not_bol = bufp->not_eol = 0; | |
6224b623 | 2497 | bufp->used_syntax = 0; |
fa9a63c5 RM |
2498 | |
2499 | /* Set `used' to zero, so that if we return an error, the pattern | |
2500 | printer (for debugging) will think there's no pattern. We reset it | |
2501 | at the end. */ | |
2502 | bufp->used = 0; | |
5e69f11e | 2503 | |
fa9a63c5 | 2504 | /* Always count groups, whether or not bufp->no_sub is set. */ |
5e69f11e | 2505 | bufp->re_nsub = 0; |
fa9a63c5 | 2506 | |
0b32bf0e | 2507 | #if !defined emacs && !defined SYNTAX_TABLE |
fa9a63c5 RM |
2508 | /* Initialize the syntax table. */ |
2509 | init_syntax_once (); | |
2510 | #endif | |
2511 | ||
2512 | if (bufp->allocated == 0) | |
2513 | { | |
2514 | if (bufp->buffer) | |
2515 | { /* If zero allocated, but buffer is non-null, try to realloc | |
25fe55af | 2516 | enough space. This loses if buffer's address is bogus, but |
7814e705 | 2517 | that is the user's responsibility. */ |
25fe55af RS |
2518 | RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); |
2519 | } | |
fa9a63c5 | 2520 | else |
7814e705 | 2521 | { /* Caller did not allocate a buffer. Do it for them. */ |
25fe55af RS |
2522 | bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); |
2523 | } | |
fa9a63c5 RM |
2524 | if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE); |
2525 | ||
2526 | bufp->allocated = INIT_BUF_SIZE; | |
2527 | } | |
2528 | ||
2529 | begalt = b = bufp->buffer; | |
2530 | ||
2531 | /* Loop through the uncompiled pattern until we're at the end. */ | |
f9b0fd99 | 2532 | while (1) |
fa9a63c5 | 2533 | { |
f9b0fd99 RS |
2534 | if (p == pend) |
2535 | { | |
2536 | /* If this is the end of an included regexp, | |
2537 | pop back to the main regexp and try again. */ | |
2538 | if (in_subpattern) | |
2539 | { | |
2540 | in_subpattern = 0; | |
2541 | pattern = main_pattern; | |
2542 | p = main_p; | |
2543 | pend = main_pend; | |
2544 | continue; | |
2545 | } | |
2546 | /* If this is the end of the main regexp, we are done. */ | |
2547 | break; | |
2548 | } | |
2549 | ||
fa9a63c5 RM |
2550 | PATFETCH (c); |
2551 | ||
2552 | switch (c) | |
25fe55af | 2553 | { |
f9b0fd99 RS |
2554 | case ' ': |
2555 | { | |
2556 | re_char *p1 = p; | |
2557 | ||
2558 | /* If there's no special whitespace regexp, treat | |
4fb680cd RS |
2559 | spaces normally. And don't try to do this recursively. */ |
2560 | if (!whitespace_regexp || in_subpattern) | |
f9b0fd99 RS |
2561 | goto normal_char; |
2562 | ||
2563 | /* Peek past following spaces. */ | |
2564 | while (p1 != pend) | |
2565 | { | |
2566 | if (*p1 != ' ') | |
2567 | break; | |
2568 | p1++; | |
2569 | } | |
2570 | /* If the spaces are followed by a repetition op, | |
2571 | treat them normally. */ | |
c721eee5 RS |
2572 | if (p1 != pend |
2573 | && (*p1 == '*' || *p1 == '+' || *p1 == '?' | |
f9b0fd99 RS |
2574 | || (*p1 == '\\' && p1 + 1 != pend && p1[1] == '{'))) |
2575 | goto normal_char; | |
2576 | ||
2577 | /* Replace the spaces with the whitespace regexp. */ | |
2578 | in_subpattern = 1; | |
2579 | main_p = p1; | |
2580 | main_pend = pend; | |
2581 | main_pattern = pattern; | |
2582 | p = pattern = whitespace_regexp; | |
5b0534c8 | 2583 | pend = p + strlen ((const char *) p); |
f9b0fd99 | 2584 | break; |
7814e705 | 2585 | } |
f9b0fd99 | 2586 | |
25fe55af RS |
2587 | case '^': |
2588 | { | |
7814e705 | 2589 | if ( /* If at start of pattern, it's an operator. */ |
25fe55af | 2590 | p == pattern + 1 |
7814e705 | 2591 | /* If context independent, it's an operator. */ |
25fe55af | 2592 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
7814e705 | 2593 | /* Otherwise, depends on what's come before. */ |
25fe55af | 2594 | || at_begline_loc_p (pattern, p, syntax)) |
c0f9ea08 | 2595 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? begbuf : begline); |
25fe55af RS |
2596 | else |
2597 | goto normal_char; | |
2598 | } | |
2599 | break; | |
2600 | ||
2601 | ||
2602 | case '$': | |
2603 | { | |
2604 | if ( /* If at end of pattern, it's an operator. */ | |
2605 | p == pend | |
7814e705 | 2606 | /* If context independent, it's an operator. */ |
25fe55af RS |
2607 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
2608 | /* Otherwise, depends on what's next. */ | |
2609 | || at_endline_loc_p (p, pend, syntax)) | |
c0f9ea08 | 2610 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? endbuf : endline); |
25fe55af RS |
2611 | else |
2612 | goto normal_char; | |
2613 | } | |
2614 | break; | |
fa9a63c5 RM |
2615 | |
2616 | ||
2617 | case '+': | |
25fe55af RS |
2618 | case '?': |
2619 | if ((syntax & RE_BK_PLUS_QM) | |
2620 | || (syntax & RE_LIMITED_OPS)) | |
2621 | goto normal_char; | |
2622 | handle_plus: | |
2623 | case '*': | |
5ac2eb34 | 2624 | /* If there is no previous pattern... */ |
25fe55af RS |
2625 | if (!laststart) |
2626 | { | |
2627 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
2628 | FREE_STACK_RETURN (REG_BADRPT); | |
2629 | else if (!(syntax & RE_CONTEXT_INDEP_OPS)) | |
2630 | goto normal_char; | |
2631 | } | |
2632 | ||
2633 | { | |
7814e705 | 2634 | /* 1 means zero (many) matches is allowed. */ |
66f0296e SM |
2635 | boolean zero_times_ok = 0, many_times_ok = 0; |
2636 | boolean greedy = 1; | |
25fe55af RS |
2637 | |
2638 | /* If there is a sequence of repetition chars, collapse it | |
2639 | down to just one (the right one). We can't combine | |
2640 | interval operators with these because of, e.g., `a{2}*', | |
7814e705 | 2641 | which should only match an even number of `a's. */ |
25fe55af RS |
2642 | |
2643 | for (;;) | |
2644 | { | |
0b32bf0e | 2645 | if ((syntax & RE_FRUGAL) |
1c8c6d39 DL |
2646 | && c == '?' && (zero_times_ok || many_times_ok)) |
2647 | greedy = 0; | |
2648 | else | |
2649 | { | |
2650 | zero_times_ok |= c != '+'; | |
2651 | many_times_ok |= c != '?'; | |
2652 | } | |
25fe55af RS |
2653 | |
2654 | if (p == pend) | |
2655 | break; | |
ed0767d8 SM |
2656 | else if (*p == '*' |
2657 | || (!(syntax & RE_BK_PLUS_QM) | |
2658 | && (*p == '+' || *p == '?'))) | |
25fe55af | 2659 | ; |
ed0767d8 | 2660 | else if (syntax & RE_BK_PLUS_QM && *p == '\\') |
25fe55af | 2661 | { |
ed0767d8 SM |
2662 | if (p+1 == pend) |
2663 | FREE_STACK_RETURN (REG_EESCAPE); | |
2664 | if (p[1] == '+' || p[1] == '?') | |
2665 | PATFETCH (c); /* Gobble up the backslash. */ | |
2666 | else | |
2667 | break; | |
25fe55af RS |
2668 | } |
2669 | else | |
ed0767d8 | 2670 | break; |
25fe55af | 2671 | /* If we get here, we found another repeat character. */ |
ed0767d8 SM |
2672 | PATFETCH (c); |
2673 | } | |
25fe55af RS |
2674 | |
2675 | /* Star, etc. applied to an empty pattern is equivalent | |
2676 | to an empty pattern. */ | |
4e8a9132 | 2677 | if (!laststart || laststart == b) |
25fe55af RS |
2678 | break; |
2679 | ||
2680 | /* Now we know whether or not zero matches is allowed | |
7814e705 | 2681 | and also whether or not two or more matches is allowed. */ |
1c8c6d39 DL |
2682 | if (greedy) |
2683 | { | |
99633e97 | 2684 | if (many_times_ok) |
4e8a9132 SM |
2685 | { |
2686 | boolean simple = skip_one_char (laststart) == b; | |
d1dfb56c | 2687 | size_t startoffset = 0; |
f6a3f532 | 2688 | re_opcode_t ofj = |
01618498 | 2689 | /* Check if the loop can match the empty string. */ |
6df42991 SM |
2690 | (simple || !analyse_first (laststart, b, NULL, 0)) |
2691 | ? on_failure_jump : on_failure_jump_loop; | |
4e8a9132 | 2692 | assert (skip_one_char (laststart) <= b); |
177c0ea7 | 2693 | |
4e8a9132 SM |
2694 | if (!zero_times_ok && simple) |
2695 | { /* Since simple * loops can be made faster by using | |
2696 | on_failure_keep_string_jump, we turn simple P+ | |
2697 | into PP* if P is simple. */ | |
2698 | unsigned char *p1, *p2; | |
2699 | startoffset = b - laststart; | |
2700 | GET_BUFFER_SPACE (startoffset); | |
2701 | p1 = b; p2 = laststart; | |
2702 | while (p2 < p1) | |
2703 | *b++ = *p2++; | |
2704 | zero_times_ok = 1; | |
99633e97 | 2705 | } |
4e8a9132 SM |
2706 | |
2707 | GET_BUFFER_SPACE (6); | |
2708 | if (!zero_times_ok) | |
2709 | /* A + loop. */ | |
f6a3f532 | 2710 | STORE_JUMP (ofj, b, b + 6); |
99633e97 | 2711 | else |
4e8a9132 SM |
2712 | /* Simple * loops can use on_failure_keep_string_jump |
2713 | depending on what follows. But since we don't know | |
2714 | that yet, we leave the decision up to | |
2715 | on_failure_jump_smart. */ | |
f6a3f532 | 2716 | INSERT_JUMP (simple ? on_failure_jump_smart : ofj, |
4e8a9132 | 2717 | laststart + startoffset, b + 6); |
99633e97 | 2718 | b += 3; |
4e8a9132 | 2719 | STORE_JUMP (jump, b, laststart + startoffset); |
99633e97 SM |
2720 | b += 3; |
2721 | } | |
2722 | else | |
2723 | { | |
4e8a9132 SM |
2724 | /* A simple ? pattern. */ |
2725 | assert (zero_times_ok); | |
2726 | GET_BUFFER_SPACE (3); | |
2727 | INSERT_JUMP (on_failure_jump, laststart, b + 3); | |
99633e97 SM |
2728 | b += 3; |
2729 | } | |
1c8c6d39 DL |
2730 | } |
2731 | else /* not greedy */ | |
5ac2eb34 | 2732 | { /* I wish the greedy and non-greedy cases could be merged. */ |
1c8c6d39 | 2733 | |
0683b6fa | 2734 | GET_BUFFER_SPACE (7); /* We might use less. */ |
1c8c6d39 DL |
2735 | if (many_times_ok) |
2736 | { | |
f6a3f532 SM |
2737 | boolean emptyp = analyse_first (laststart, b, NULL, 0); |
2738 | ||
6df42991 SM |
2739 | /* The non-greedy multiple match looks like |
2740 | a repeat..until: we only need a conditional jump | |
2741 | at the end of the loop. */ | |
f6a3f532 SM |
2742 | if (emptyp) BUF_PUSH (no_op); |
2743 | STORE_JUMP (emptyp ? on_failure_jump_nastyloop | |
2744 | : on_failure_jump, b, laststart); | |
1c8c6d39 DL |
2745 | b += 3; |
2746 | if (zero_times_ok) | |
2747 | { | |
2748 | /* The repeat...until naturally matches one or more. | |
2749 | To also match zero times, we need to first jump to | |
6df42991 | 2750 | the end of the loop (its conditional jump). */ |
1c8c6d39 DL |
2751 | INSERT_JUMP (jump, laststart, b); |
2752 | b += 3; | |
2753 | } | |
2754 | } | |
2755 | else | |
2756 | { | |
2757 | /* non-greedy a?? */ | |
1c8c6d39 DL |
2758 | INSERT_JUMP (jump, laststart, b + 3); |
2759 | b += 3; | |
2760 | INSERT_JUMP (on_failure_jump, laststart, laststart + 6); | |
2761 | b += 3; | |
2762 | } | |
2763 | } | |
2764 | } | |
4e8a9132 | 2765 | pending_exact = 0; |
fa9a63c5 RM |
2766 | break; |
2767 | ||
2768 | ||
2769 | case '.': | |
25fe55af RS |
2770 | laststart = b; |
2771 | BUF_PUSH (anychar); | |
2772 | break; | |
fa9a63c5 RM |
2773 | |
2774 | ||
25fe55af RS |
2775 | case '[': |
2776 | { | |
19ed5445 PE |
2777 | re_char *p1; |
2778 | ||
b18215fc | 2779 | CLEAR_RANGE_TABLE_WORK_USED (range_table_work); |
fa9a63c5 | 2780 | |
25fe55af | 2781 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2782 | |
25fe55af RS |
2783 | /* Ensure that we have enough space to push a charset: the |
2784 | opcode, the length count, and the bitset; 34 bytes in all. */ | |
fa9a63c5 RM |
2785 | GET_BUFFER_SPACE (34); |
2786 | ||
25fe55af | 2787 | laststart = b; |
e318085a | 2788 | |
25fe55af | 2789 | /* We test `*p == '^' twice, instead of using an if |
7814e705 | 2790 | statement, so we only need one BUF_PUSH. */ |
25fe55af RS |
2791 | BUF_PUSH (*p == '^' ? charset_not : charset); |
2792 | if (*p == '^') | |
2793 | p++; | |
e318085a | 2794 | |
25fe55af RS |
2795 | /* Remember the first position in the bracket expression. */ |
2796 | p1 = p; | |
e318085a | 2797 | |
7814e705 | 2798 | /* Push the number of bytes in the bitmap. */ |
25fe55af | 2799 | BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2800 | |
25fe55af | 2801 | /* Clear the whole map. */ |
72af86bd | 2802 | memset (b, 0, (1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2803 | |
25fe55af RS |
2804 | /* charset_not matches newline according to a syntax bit. */ |
2805 | if ((re_opcode_t) b[-2] == charset_not | |
2806 | && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) | |
2807 | SET_LIST_BIT ('\n'); | |
fa9a63c5 | 2808 | |
7814e705 | 2809 | /* Read in characters and ranges, setting map bits. */ |
25fe55af RS |
2810 | for (;;) |
2811 | { | |
b18215fc | 2812 | boolean escaped_char = false; |
2d1675e4 | 2813 | const unsigned char *p2 = p; |
abbd1bcf | 2814 | re_wchar_t ch; |
e318085a | 2815 | |
25fe55af | 2816 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
e318085a | 2817 | |
36595814 SM |
2818 | /* Don't translate yet. The range TRANSLATE(X..Y) cannot |
2819 | always be determined from TRANSLATE(X) and TRANSLATE(Y) | |
2820 | So the translation is done later in a loop. Example: | |
2821 | (let ((case-fold-search t)) (string-match "[A-_]" "A")) */ | |
25fe55af | 2822 | PATFETCH (c); |
e318085a | 2823 | |
25fe55af RS |
2824 | /* \ might escape characters inside [...] and [^...]. */ |
2825 | if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') | |
2826 | { | |
2827 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
e318085a RS |
2828 | |
2829 | PATFETCH (c); | |
b18215fc | 2830 | escaped_char = true; |
25fe55af | 2831 | } |
b18215fc RS |
2832 | else |
2833 | { | |
7814e705 | 2834 | /* Could be the end of the bracket expression. If it's |
657fcfbd RS |
2835 | not (i.e., when the bracket expression is `[]' so |
2836 | far), the ']' character bit gets set way below. */ | |
2d1675e4 | 2837 | if (c == ']' && p2 != p1) |
657fcfbd | 2838 | break; |
25fe55af | 2839 | } |
b18215fc | 2840 | |
25fe55af RS |
2841 | /* See if we're at the beginning of a possible character |
2842 | class. */ | |
b18215fc | 2843 | |
2d1675e4 SM |
2844 | if (!escaped_char && |
2845 | syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') | |
657fcfbd | 2846 | { |
7814e705 | 2847 | /* Leave room for the null. */ |
14473664 | 2848 | unsigned char str[CHAR_CLASS_MAX_LENGTH + 1]; |
ed0767d8 | 2849 | const unsigned char *class_beg; |
b18215fc | 2850 | |
25fe55af RS |
2851 | PATFETCH (c); |
2852 | c1 = 0; | |
ed0767d8 | 2853 | class_beg = p; |
b18215fc | 2854 | |
25fe55af RS |
2855 | /* If pattern is `[[:'. */ |
2856 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
b18215fc | 2857 | |
25fe55af RS |
2858 | for (;;) |
2859 | { | |
14473664 SM |
2860 | PATFETCH (c); |
2861 | if ((c == ':' && *p == ']') || p == pend) | |
2862 | break; | |
2863 | if (c1 < CHAR_CLASS_MAX_LENGTH) | |
2864 | str[c1++] = c; | |
2865 | else | |
2866 | /* This is in any case an invalid class name. */ | |
2867 | str[0] = '\0'; | |
25fe55af RS |
2868 | } |
2869 | str[c1] = '\0'; | |
b18215fc RS |
2870 | |
2871 | /* If isn't a word bracketed by `[:' and `:]': | |
2872 | undo the ending character, the letters, and | |
2873 | leave the leading `:' and `[' (but set bits for | |
2874 | them). */ | |
25fe55af RS |
2875 | if (c == ':' && *p == ']') |
2876 | { | |
abbd1bcf | 2877 | re_wctype_t cc = re_wctype (str); |
14473664 SM |
2878 | |
2879 | if (cc == 0) | |
fa9a63c5 RM |
2880 | FREE_STACK_RETURN (REG_ECTYPE); |
2881 | ||
14473664 SM |
2882 | /* Throw away the ] at the end of the character |
2883 | class. */ | |
2884 | PATFETCH (c); | |
fa9a63c5 | 2885 | |
14473664 | 2886 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2887 | |
cf9c99bc KH |
2888 | #ifndef emacs |
2889 | for (ch = 0; ch < (1 << BYTEWIDTH); ++ch) | |
8f924df7 KH |
2890 | if (re_iswctype (btowc (ch), cc)) |
2891 | { | |
2892 | c = TRANSLATE (ch); | |
ed00c2ac KH |
2893 | if (c < (1 << BYTEWIDTH)) |
2894 | SET_LIST_BIT (c); | |
8f924df7 | 2895 | } |
cf9c99bc KH |
2896 | #else /* emacs */ |
2897 | /* Most character classes in a multibyte match | |
2898 | just set a flag. Exceptions are is_blank, | |
2899 | is_digit, is_cntrl, and is_xdigit, since | |
2900 | they can only match ASCII characters. We | |
2901 | don't need to handle them for multibyte. | |
2902 | They are distinguished by a negative wctype. */ | |
96cc36cc | 2903 | |
254c06a8 SM |
2904 | /* Setup the gl_state object to its buffer-defined |
2905 | value. This hardcodes the buffer-global | |
2906 | syntax-table for ASCII chars, while the other chars | |
2907 | will obey syntax-table properties. It's not ideal, | |
2908 | but it's the way it's been done until now. */ | |
d48cd3f4 | 2909 | SETUP_BUFFER_SYNTAX_TABLE (); |
254c06a8 | 2910 | |
cf9c99bc | 2911 | for (ch = 0; ch < 256; ++ch) |
25fe55af | 2912 | { |
cf9c99bc KH |
2913 | c = RE_CHAR_TO_MULTIBYTE (ch); |
2914 | if (! CHAR_BYTE8_P (c) | |
2915 | && re_iswctype (c, cc)) | |
8f924df7 | 2916 | { |
cf9c99bc KH |
2917 | SET_LIST_BIT (ch); |
2918 | c1 = TRANSLATE (c); | |
2919 | if (c1 == c) | |
2920 | continue; | |
2921 | if (ASCII_CHAR_P (c1)) | |
2922 | SET_LIST_BIT (c1); | |
2923 | else if ((c1 = RE_CHAR_TO_UNIBYTE (c1)) >= 0) | |
2924 | SET_LIST_BIT (c1); | |
8f924df7 | 2925 | } |
25fe55af | 2926 | } |
cf9c99bc KH |
2927 | SET_RANGE_TABLE_WORK_AREA_BIT |
2928 | (range_table_work, re_wctype_to_bit (cc)); | |
2929 | #endif /* emacs */ | |
6224b623 SM |
2930 | /* In most cases the matching rule for char classes |
2931 | only uses the syntax table for multibyte chars, | |
2932 | so that the content of the syntax-table it is not | |
2933 | hardcoded in the range_table. SPACE and WORD are | |
2934 | the two exceptions. */ | |
2935 | if ((1 << cc) & ((1 << RECC_SPACE) | (1 << RECC_WORD))) | |
2936 | bufp->used_syntax = 1; | |
2937 | ||
b18215fc RS |
2938 | /* Repeat the loop. */ |
2939 | continue; | |
25fe55af RS |
2940 | } |
2941 | else | |
2942 | { | |
ed0767d8 SM |
2943 | /* Go back to right after the "[:". */ |
2944 | p = class_beg; | |
25fe55af | 2945 | SET_LIST_BIT ('['); |
b18215fc RS |
2946 | |
2947 | /* Because the `:' may starts the range, we | |
2948 | can't simply set bit and repeat the loop. | |
7814e705 | 2949 | Instead, just set it to C and handle below. */ |
b18215fc | 2950 | c = ':'; |
25fe55af RS |
2951 | } |
2952 | } | |
b18215fc RS |
2953 | |
2954 | if (p < pend && p[0] == '-' && p[1] != ']') | |
2955 | { | |
2956 | ||
2957 | /* Discard the `-'. */ | |
2958 | PATFETCH (c1); | |
2959 | ||
2960 | /* Fetch the character which ends the range. */ | |
2961 | PATFETCH (c1); | |
cf9c99bc KH |
2962 | #ifdef emacs |
2963 | if (CHAR_BYTE8_P (c1) | |
2964 | && ! ASCII_CHAR_P (c) && ! CHAR_BYTE8_P (c)) | |
2965 | /* Treat the range from a multibyte character to | |
2966 | raw-byte character as empty. */ | |
2967 | c = c1 + 1; | |
2968 | #endif /* emacs */ | |
e318085a | 2969 | } |
25fe55af | 2970 | else |
b18215fc RS |
2971 | /* Range from C to C. */ |
2972 | c1 = c; | |
2973 | ||
cf9c99bc | 2974 | if (c > c1) |
25fe55af | 2975 | { |
cf9c99bc KH |
2976 | if (syntax & RE_NO_EMPTY_RANGES) |
2977 | FREE_STACK_RETURN (REG_ERANGEX); | |
2978 | /* Else, repeat the loop. */ | |
bf216479 | 2979 | } |
6fdd04b0 | 2980 | else |
25fe55af | 2981 | { |
cf9c99bc KH |
2982 | #ifndef emacs |
2983 | /* Set the range into bitmap */ | |
8f924df7 | 2984 | for (; c <= c1; c++) |
b18215fc | 2985 | { |
cf9c99bc KH |
2986 | ch = TRANSLATE (c); |
2987 | if (ch < (1 << BYTEWIDTH)) | |
2988 | SET_LIST_BIT (ch); | |
2989 | } | |
2990 | #else /* emacs */ | |
2991 | if (c < 128) | |
2992 | { | |
2993 | ch = MIN (127, c1); | |
2994 | SETUP_ASCII_RANGE (range_table_work, c, ch); | |
2995 | c = ch + 1; | |
2996 | if (CHAR_BYTE8_P (c1)) | |
2997 | c = BYTE8_TO_CHAR (128); | |
2998 | } | |
2999 | if (c <= c1) | |
3000 | { | |
3001 | if (CHAR_BYTE8_P (c)) | |
3002 | { | |
3003 | c = CHAR_TO_BYTE8 (c); | |
3004 | c1 = CHAR_TO_BYTE8 (c1); | |
3005 | for (; c <= c1; c++) | |
3006 | SET_LIST_BIT (c); | |
3007 | } | |
3008 | else if (multibyte) | |
3009 | { | |
3010 | SETUP_MULTIBYTE_RANGE (range_table_work, c, c1); | |
3011 | } | |
3012 | else | |
3013 | { | |
3014 | SETUP_UNIBYTE_RANGE (range_table_work, c, c1); | |
3015 | } | |
e934739e | 3016 | } |
cf9c99bc | 3017 | #endif /* emacs */ |
25fe55af | 3018 | } |
e318085a RS |
3019 | } |
3020 | ||
25fe55af | 3021 | /* Discard any (non)matching list bytes that are all 0 at the |
7814e705 | 3022 | end of the map. Decrease the map-length byte too. */ |
25fe55af RS |
3023 | while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) |
3024 | b[-1]--; | |
3025 | b += b[-1]; | |
fa9a63c5 | 3026 | |
96cc36cc RS |
3027 | /* Build real range table from work area. */ |
3028 | if (RANGE_TABLE_WORK_USED (range_table_work) | |
3029 | || RANGE_TABLE_WORK_BITS (range_table_work)) | |
b18215fc RS |
3030 | { |
3031 | int i; | |
3032 | int used = RANGE_TABLE_WORK_USED (range_table_work); | |
fa9a63c5 | 3033 | |
b18215fc | 3034 | /* Allocate space for COUNT + RANGE_TABLE. Needs two |
96cc36cc | 3035 | bytes for flags, two for COUNT, and three bytes for |
5ac2eb34 | 3036 | each character. */ |
96cc36cc | 3037 | GET_BUFFER_SPACE (4 + used * 3); |
fa9a63c5 | 3038 | |
b18215fc RS |
3039 | /* Indicate the existence of range table. */ |
3040 | laststart[1] |= 0x80; | |
fa9a63c5 | 3041 | |
96cc36cc RS |
3042 | /* Store the character class flag bits into the range table. |
3043 | If not in emacs, these flag bits are always 0. */ | |
3044 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) & 0xff; | |
3045 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) >> 8; | |
3046 | ||
b18215fc RS |
3047 | STORE_NUMBER_AND_INCR (b, used / 2); |
3048 | for (i = 0; i < used; i++) | |
3049 | STORE_CHARACTER_AND_INCR | |
3050 | (b, RANGE_TABLE_WORK_ELT (range_table_work, i)); | |
3051 | } | |
25fe55af RS |
3052 | } |
3053 | break; | |
fa9a63c5 RM |
3054 | |
3055 | ||
b18215fc | 3056 | case '(': |
25fe55af RS |
3057 | if (syntax & RE_NO_BK_PARENS) |
3058 | goto handle_open; | |
3059 | else | |
3060 | goto normal_char; | |
fa9a63c5 RM |
3061 | |
3062 | ||
25fe55af RS |
3063 | case ')': |
3064 | if (syntax & RE_NO_BK_PARENS) | |
3065 | goto handle_close; | |
3066 | else | |
3067 | goto normal_char; | |
e318085a RS |
3068 | |
3069 | ||
25fe55af RS |
3070 | case '\n': |
3071 | if (syntax & RE_NEWLINE_ALT) | |
3072 | goto handle_alt; | |
3073 | else | |
3074 | goto normal_char; | |
e318085a RS |
3075 | |
3076 | ||
b18215fc | 3077 | case '|': |
25fe55af RS |
3078 | if (syntax & RE_NO_BK_VBAR) |
3079 | goto handle_alt; | |
3080 | else | |
3081 | goto normal_char; | |
3082 | ||
3083 | ||
3084 | case '{': | |
3085 | if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) | |
3086 | goto handle_interval; | |
3087 | else | |
3088 | goto normal_char; | |
3089 | ||
3090 | ||
3091 | case '\\': | |
3092 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
3093 | ||
3094 | /* Do not translate the character after the \, so that we can | |
3095 | distinguish, e.g., \B from \b, even if we normally would | |
3096 | translate, e.g., B to b. */ | |
36595814 | 3097 | PATFETCH (c); |
25fe55af RS |
3098 | |
3099 | switch (c) | |
3100 | { | |
3101 | case '(': | |
3102 | if (syntax & RE_NO_BK_PARENS) | |
3103 | goto normal_backslash; | |
3104 | ||
3105 | handle_open: | |
505bde11 SM |
3106 | { |
3107 | int shy = 0; | |
c69b0314 | 3108 | regnum_t regnum = 0; |
505bde11 SM |
3109 | if (p+1 < pend) |
3110 | { | |
3111 | /* Look for a special (?...) construct */ | |
ed0767d8 | 3112 | if ((syntax & RE_SHY_GROUPS) && *p == '?') |
505bde11 | 3113 | { |
ed0767d8 | 3114 | PATFETCH (c); /* Gobble up the '?'. */ |
c69b0314 | 3115 | while (!shy) |
505bde11 | 3116 | { |
c69b0314 SM |
3117 | PATFETCH (c); |
3118 | switch (c) | |
3119 | { | |
3120 | case ':': shy = 1; break; | |
3121 | case '0': | |
3122 | /* An explicitly specified regnum must start | |
3123 | with non-0. */ | |
3124 | if (regnum == 0) | |
3125 | FREE_STACK_RETURN (REG_BADPAT); | |
3126 | case '1': case '2': case '3': case '4': | |
3127 | case '5': case '6': case '7': case '8': case '9': | |
3128 | regnum = 10*regnum + (c - '0'); break; | |
3129 | default: | |
3130 | /* Only (?:...) is supported right now. */ | |
3131 | FREE_STACK_RETURN (REG_BADPAT); | |
3132 | } | |
505bde11 SM |
3133 | } |
3134 | } | |
505bde11 SM |
3135 | } |
3136 | ||
3137 | if (!shy) | |
c69b0314 SM |
3138 | regnum = ++bufp->re_nsub; |
3139 | else if (regnum) | |
3140 | { /* It's actually not shy, but explicitly numbered. */ | |
3141 | shy = 0; | |
3142 | if (regnum > bufp->re_nsub) | |
3143 | bufp->re_nsub = regnum; | |
3144 | else if (regnum > bufp->re_nsub | |
3145 | /* Ideally, we'd want to check that the specified | |
3146 | group can't have matched (i.e. all subgroups | |
3147 | using the same regnum are in other branches of | |
3148 | OR patterns), but we don't currently keep track | |
3149 | of enough info to do that easily. */ | |
3150 | || group_in_compile_stack (compile_stack, regnum)) | |
3151 | FREE_STACK_RETURN (REG_BADPAT); | |
505bde11 | 3152 | } |
c69b0314 SM |
3153 | else |
3154 | /* It's really shy. */ | |
3155 | regnum = - bufp->re_nsub; | |
25fe55af | 3156 | |
99633e97 SM |
3157 | if (COMPILE_STACK_FULL) |
3158 | { | |
3159 | RETALLOC (compile_stack.stack, compile_stack.size << 1, | |
3160 | compile_stack_elt_t); | |
3161 | if (compile_stack.stack == NULL) return REG_ESPACE; | |
25fe55af | 3162 | |
99633e97 SM |
3163 | compile_stack.size <<= 1; |
3164 | } | |
25fe55af | 3165 | |
99633e97 | 3166 | /* These are the values to restore when we hit end of this |
7814e705 | 3167 | group. They are all relative offsets, so that if the |
99633e97 SM |
3168 | whole pattern moves because of realloc, they will still |
3169 | be valid. */ | |
3170 | COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; | |
3171 | COMPILE_STACK_TOP.fixup_alt_jump | |
3172 | = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; | |
3173 | COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer; | |
c69b0314 | 3174 | COMPILE_STACK_TOP.regnum = regnum; |
99633e97 | 3175 | |
c69b0314 SM |
3176 | /* Do not push a start_memory for groups beyond the last one |
3177 | we can represent in the compiled pattern. */ | |
3178 | if (regnum <= MAX_REGNUM && regnum > 0) | |
99633e97 SM |
3179 | BUF_PUSH_2 (start_memory, regnum); |
3180 | ||
3181 | compile_stack.avail++; | |
3182 | ||
3183 | fixup_alt_jump = 0; | |
3184 | laststart = 0; | |
3185 | begalt = b; | |
3186 | /* If we've reached MAX_REGNUM groups, then this open | |
3187 | won't actually generate any code, so we'll have to | |
3188 | clear pending_exact explicitly. */ | |
3189 | pending_exact = 0; | |
3190 | break; | |
505bde11 | 3191 | } |
25fe55af RS |
3192 | |
3193 | case ')': | |
3194 | if (syntax & RE_NO_BK_PARENS) goto normal_backslash; | |
3195 | ||
3196 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3197 | { |
3198 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3199 | goto normal_backslash; | |
3200 | else | |
3201 | FREE_STACK_RETURN (REG_ERPAREN); | |
3202 | } | |
25fe55af RS |
3203 | |
3204 | handle_close: | |
505bde11 | 3205 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3206 | |
3207 | /* See similar code for backslashed left paren above. */ | |
3208 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3209 | { |
3210 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3211 | goto normal_char; | |
3212 | else | |
3213 | FREE_STACK_RETURN (REG_ERPAREN); | |
3214 | } | |
25fe55af RS |
3215 | |
3216 | /* Since we just checked for an empty stack above, this | |
3217 | ``can't happen''. */ | |
3218 | assert (compile_stack.avail != 0); | |
3219 | { | |
3220 | /* We don't just want to restore into `regnum', because | |
3221 | later groups should continue to be numbered higher, | |
7814e705 | 3222 | as in `(ab)c(de)' -- the second group is #2. */ |
c69b0314 | 3223 | regnum_t regnum; |
25fe55af RS |
3224 | |
3225 | compile_stack.avail--; | |
3226 | begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; | |
3227 | fixup_alt_jump | |
3228 | = COMPILE_STACK_TOP.fixup_alt_jump | |
3229 | ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 | |
3230 | : 0; | |
3231 | laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; | |
c69b0314 | 3232 | regnum = COMPILE_STACK_TOP.regnum; |
b18215fc RS |
3233 | /* If we've reached MAX_REGNUM groups, then this open |
3234 | won't actually generate any code, so we'll have to | |
3235 | clear pending_exact explicitly. */ | |
3236 | pending_exact = 0; | |
e318085a | 3237 | |
25fe55af | 3238 | /* We're at the end of the group, so now we know how many |
7814e705 | 3239 | groups were inside this one. */ |
c69b0314 SM |
3240 | if (regnum <= MAX_REGNUM && regnum > 0) |
3241 | BUF_PUSH_2 (stop_memory, regnum); | |
25fe55af RS |
3242 | } |
3243 | break; | |
3244 | ||
3245 | ||
3246 | case '|': /* `\|'. */ | |
3247 | if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) | |
3248 | goto normal_backslash; | |
3249 | handle_alt: | |
3250 | if (syntax & RE_LIMITED_OPS) | |
3251 | goto normal_char; | |
3252 | ||
3253 | /* Insert before the previous alternative a jump which | |
7814e705 | 3254 | jumps to this alternative if the former fails. */ |
25fe55af RS |
3255 | GET_BUFFER_SPACE (3); |
3256 | INSERT_JUMP (on_failure_jump, begalt, b + 6); | |
3257 | pending_exact = 0; | |
3258 | b += 3; | |
3259 | ||
3260 | /* The alternative before this one has a jump after it | |
3261 | which gets executed if it gets matched. Adjust that | |
3262 | jump so it will jump to this alternative's analogous | |
3263 | jump (put in below, which in turn will jump to the next | |
3264 | (if any) alternative's such jump, etc.). The last such | |
3265 | jump jumps to the correct final destination. A picture: | |
3266 | _____ _____ | |
3267 | | | | | | |
3268 | | v | v | |
d1dfb56c | 3269 | a | b | c |
25fe55af RS |
3270 | |
3271 | If we are at `b', then fixup_alt_jump right now points to a | |
3272 | three-byte space after `a'. We'll put in the jump, set | |
3273 | fixup_alt_jump to right after `b', and leave behind three | |
3274 | bytes which we'll fill in when we get to after `c'. */ | |
3275 | ||
505bde11 | 3276 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3277 | |
3278 | /* Mark and leave space for a jump after this alternative, | |
3279 | to be filled in later either by next alternative or | |
3280 | when know we're at the end of a series of alternatives. */ | |
3281 | fixup_alt_jump = b; | |
3282 | GET_BUFFER_SPACE (3); | |
3283 | b += 3; | |
3284 | ||
3285 | laststart = 0; | |
3286 | begalt = b; | |
3287 | break; | |
3288 | ||
3289 | ||
3290 | case '{': | |
3291 | /* If \{ is a literal. */ | |
3292 | if (!(syntax & RE_INTERVALS) | |
3293 | /* If we're at `\{' and it's not the open-interval | |
3294 | operator. */ | |
4bb91c68 | 3295 | || (syntax & RE_NO_BK_BRACES)) |
25fe55af RS |
3296 | goto normal_backslash; |
3297 | ||
3298 | handle_interval: | |
3299 | { | |
3300 | /* If got here, then the syntax allows intervals. */ | |
3301 | ||
3302 | /* At least (most) this many matches must be made. */ | |
99633e97 | 3303 | int lower_bound = 0, upper_bound = -1; |
25fe55af | 3304 | |
ed0767d8 | 3305 | beg_interval = p; |
25fe55af | 3306 | |
4618713a | 3307 | GET_INTERVAL_COUNT (lower_bound); |
25fe55af RS |
3308 | |
3309 | if (c == ',') | |
a3fab1ec | 3310 | GET_INTERVAL_COUNT (upper_bound); |
25fe55af RS |
3311 | else |
3312 | /* Interval such as `{1}' => match exactly once. */ | |
3313 | upper_bound = lower_bound; | |
3314 | ||
a3fab1ec PE |
3315 | if (lower_bound < 0 |
3316 | || (0 <= upper_bound && upper_bound < lower_bound)) | |
3317 | FREE_STACK_RETURN (REG_BADBR); | |
3318 | ||
25fe55af RS |
3319 | if (!(syntax & RE_NO_BK_BRACES)) |
3320 | { | |
4bb91c68 SM |
3321 | if (c != '\\') |
3322 | FREE_STACK_RETURN (REG_BADBR); | |
c72b0edd SM |
3323 | if (p == pend) |
3324 | FREE_STACK_RETURN (REG_EESCAPE); | |
25fe55af RS |
3325 | PATFETCH (c); |
3326 | } | |
3327 | ||
3328 | if (c != '}') | |
4bb91c68 | 3329 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3330 | |
3331 | /* We just parsed a valid interval. */ | |
3332 | ||
3333 | /* If it's invalid to have no preceding re. */ | |
3334 | if (!laststart) | |
3335 | { | |
3336 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
3337 | FREE_STACK_RETURN (REG_BADRPT); | |
3338 | else if (syntax & RE_CONTEXT_INDEP_OPS) | |
3339 | laststart = b; | |
3340 | else | |
3341 | goto unfetch_interval; | |
3342 | } | |
3343 | ||
6df42991 SM |
3344 | if (upper_bound == 0) |
3345 | /* If the upper bound is zero, just drop the sub pattern | |
3346 | altogether. */ | |
3347 | b = laststart; | |
3348 | else if (lower_bound == 1 && upper_bound == 1) | |
3349 | /* Just match it once: nothing to do here. */ | |
3350 | ; | |
3351 | ||
3352 | /* Otherwise, we have a nontrivial interval. When | |
3353 | we're all done, the pattern will look like: | |
3354 | set_number_at <jump count> <upper bound> | |
3355 | set_number_at <succeed_n count> <lower bound> | |
3356 | succeed_n <after jump addr> <succeed_n count> | |
3357 | <body of loop> | |
3358 | jump_n <succeed_n addr> <jump count> | |
3359 | (The upper bound and `jump_n' are omitted if | |
3360 | `upper_bound' is 1, though.) */ | |
3361 | else | |
3362 | { /* If the upper bound is > 1, we need to insert | |
3363 | more at the end of the loop. */ | |
3364 | unsigned int nbytes = (upper_bound < 0 ? 3 | |
3365 | : upper_bound > 1 ? 5 : 0); | |
3366 | unsigned int startoffset = 0; | |
3367 | ||
3368 | GET_BUFFER_SPACE (20); /* We might use less. */ | |
3369 | ||
3370 | if (lower_bound == 0) | |
3371 | { | |
3372 | /* A succeed_n that starts with 0 is really a | |
3373 | a simple on_failure_jump_loop. */ | |
3374 | INSERT_JUMP (on_failure_jump_loop, laststart, | |
3375 | b + 3 + nbytes); | |
3376 | b += 3; | |
3377 | } | |
3378 | else | |
3379 | { | |
3380 | /* Initialize lower bound of the `succeed_n', even | |
3381 | though it will be set during matching by its | |
3382 | attendant `set_number_at' (inserted next), | |
3383 | because `re_compile_fastmap' needs to know. | |
3384 | Jump to the `jump_n' we might insert below. */ | |
3385 | INSERT_JUMP2 (succeed_n, laststart, | |
3386 | b + 5 + nbytes, | |
3387 | lower_bound); | |
3388 | b += 5; | |
3389 | ||
3390 | /* Code to initialize the lower bound. Insert | |
7814e705 | 3391 | before the `succeed_n'. The `5' is the last two |
6df42991 SM |
3392 | bytes of this `set_number_at', plus 3 bytes of |
3393 | the following `succeed_n'. */ | |
3394 | insert_op2 (set_number_at, laststart, 5, lower_bound, b); | |
3395 | b += 5; | |
3396 | startoffset += 5; | |
3397 | } | |
3398 | ||
3399 | if (upper_bound < 0) | |
3400 | { | |
3401 | /* A negative upper bound stands for infinity, | |
3402 | in which case it degenerates to a plain jump. */ | |
3403 | STORE_JUMP (jump, b, laststart + startoffset); | |
3404 | b += 3; | |
3405 | } | |
3406 | else if (upper_bound > 1) | |
3407 | { /* More than one repetition is allowed, so | |
3408 | append a backward jump to the `succeed_n' | |
3409 | that starts this interval. | |
3410 | ||
3411 | When we've reached this during matching, | |
3412 | we'll have matched the interval once, so | |
3413 | jump back only `upper_bound - 1' times. */ | |
3414 | STORE_JUMP2 (jump_n, b, laststart + startoffset, | |
3415 | upper_bound - 1); | |
3416 | b += 5; | |
3417 | ||
3418 | /* The location we want to set is the second | |
3419 | parameter of the `jump_n'; that is `b-2' as | |
3420 | an absolute address. `laststart' will be | |
3421 | the `set_number_at' we're about to insert; | |
3422 | `laststart+3' the number to set, the source | |
3423 | for the relative address. But we are | |
3424 | inserting into the middle of the pattern -- | |
3425 | so everything is getting moved up by 5. | |
3426 | Conclusion: (b - 2) - (laststart + 3) + 5, | |
3427 | i.e., b - laststart. | |
3428 | ||
3429 | We insert this at the beginning of the loop | |
3430 | so that if we fail during matching, we'll | |
3431 | reinitialize the bounds. */ | |
3432 | insert_op2 (set_number_at, laststart, b - laststart, | |
3433 | upper_bound - 1, b); | |
3434 | b += 5; | |
3435 | } | |
3436 | } | |
25fe55af RS |
3437 | pending_exact = 0; |
3438 | beg_interval = NULL; | |
3439 | } | |
3440 | break; | |
3441 | ||
3442 | unfetch_interval: | |
3443 | /* If an invalid interval, match the characters as literals. */ | |
3444 | assert (beg_interval); | |
3445 | p = beg_interval; | |
3446 | beg_interval = NULL; | |
3447 | ||
3448 | /* normal_char and normal_backslash need `c'. */ | |
ed0767d8 | 3449 | c = '{'; |
25fe55af RS |
3450 | |
3451 | if (!(syntax & RE_NO_BK_BRACES)) | |
3452 | { | |
ed0767d8 SM |
3453 | assert (p > pattern && p[-1] == '\\'); |
3454 | goto normal_backslash; | |
25fe55af | 3455 | } |
ed0767d8 SM |
3456 | else |
3457 | goto normal_char; | |
e318085a | 3458 | |
b18215fc | 3459 | #ifdef emacs |
25fe55af | 3460 | /* There is no way to specify the before_dot and after_dot |
7814e705 | 3461 | operators. rms says this is ok. --karl */ |
25fe55af | 3462 | case '=': |
5ac2eb34 | 3463 | laststart = b; |
25fe55af RS |
3464 | BUF_PUSH (at_dot); |
3465 | break; | |
3466 | ||
3467 | case 's': | |
3468 | laststart = b; | |
3469 | PATFETCH (c); | |
3470 | BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); | |
3471 | break; | |
3472 | ||
3473 | case 'S': | |
3474 | laststart = b; | |
3475 | PATFETCH (c); | |
3476 | BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); | |
3477 | break; | |
b18215fc RS |
3478 | |
3479 | case 'c': | |
3480 | laststart = b; | |
36595814 | 3481 | PATFETCH (c); |
b18215fc RS |
3482 | BUF_PUSH_2 (categoryspec, c); |
3483 | break; | |
e318085a | 3484 | |
b18215fc RS |
3485 | case 'C': |
3486 | laststart = b; | |
36595814 | 3487 | PATFETCH (c); |
b18215fc RS |
3488 | BUF_PUSH_2 (notcategoryspec, c); |
3489 | break; | |
3490 | #endif /* emacs */ | |
e318085a | 3491 | |
e318085a | 3492 | |
25fe55af | 3493 | case 'w': |
4bb91c68 SM |
3494 | if (syntax & RE_NO_GNU_OPS) |
3495 | goto normal_char; | |
25fe55af | 3496 | laststart = b; |
1fb352e0 | 3497 | BUF_PUSH_2 (syntaxspec, Sword); |
25fe55af | 3498 | break; |
e318085a | 3499 | |
e318085a | 3500 | |
25fe55af | 3501 | case 'W': |
4bb91c68 SM |
3502 | if (syntax & RE_NO_GNU_OPS) |
3503 | goto normal_char; | |
25fe55af | 3504 | laststart = b; |
1fb352e0 | 3505 | BUF_PUSH_2 (notsyntaxspec, Sword); |
25fe55af | 3506 | break; |
e318085a RS |
3507 | |
3508 | ||
25fe55af | 3509 | case '<': |
4bb91c68 SM |
3510 | if (syntax & RE_NO_GNU_OPS) |
3511 | goto normal_char; | |
5ac2eb34 | 3512 | laststart = b; |
25fe55af RS |
3513 | BUF_PUSH (wordbeg); |
3514 | break; | |
e318085a | 3515 | |
25fe55af | 3516 | case '>': |
4bb91c68 SM |
3517 | if (syntax & RE_NO_GNU_OPS) |
3518 | goto normal_char; | |
5ac2eb34 | 3519 | laststart = b; |
25fe55af RS |
3520 | BUF_PUSH (wordend); |
3521 | break; | |
e318085a | 3522 | |
669fa600 SM |
3523 | case '_': |
3524 | if (syntax & RE_NO_GNU_OPS) | |
3525 | goto normal_char; | |
3526 | laststart = b; | |
3527 | PATFETCH (c); | |
3528 | if (c == '<') | |
3529 | BUF_PUSH (symbeg); | |
3530 | else if (c == '>') | |
3531 | BUF_PUSH (symend); | |
3532 | else | |
3533 | FREE_STACK_RETURN (REG_BADPAT); | |
3534 | break; | |
3535 | ||
25fe55af | 3536 | case 'b': |
4bb91c68 SM |
3537 | if (syntax & RE_NO_GNU_OPS) |
3538 | goto normal_char; | |
25fe55af RS |
3539 | BUF_PUSH (wordbound); |
3540 | break; | |
e318085a | 3541 | |
25fe55af | 3542 | case 'B': |
4bb91c68 SM |
3543 | if (syntax & RE_NO_GNU_OPS) |
3544 | goto normal_char; | |
25fe55af RS |
3545 | BUF_PUSH (notwordbound); |
3546 | break; | |
fa9a63c5 | 3547 | |
25fe55af | 3548 | case '`': |
4bb91c68 SM |
3549 | if (syntax & RE_NO_GNU_OPS) |
3550 | goto normal_char; | |
25fe55af RS |
3551 | BUF_PUSH (begbuf); |
3552 | break; | |
e318085a | 3553 | |
25fe55af | 3554 | case '\'': |
4bb91c68 SM |
3555 | if (syntax & RE_NO_GNU_OPS) |
3556 | goto normal_char; | |
25fe55af RS |
3557 | BUF_PUSH (endbuf); |
3558 | break; | |
e318085a | 3559 | |
25fe55af RS |
3560 | case '1': case '2': case '3': case '4': case '5': |
3561 | case '6': case '7': case '8': case '9': | |
0cdd06f8 SM |
3562 | { |
3563 | regnum_t reg; | |
e318085a | 3564 | |
0cdd06f8 SM |
3565 | if (syntax & RE_NO_BK_REFS) |
3566 | goto normal_backslash; | |
e318085a | 3567 | |
0cdd06f8 | 3568 | reg = c - '0'; |
e318085a | 3569 | |
c69b0314 SM |
3570 | if (reg > bufp->re_nsub || reg < 1 |
3571 | /* Can't back reference to a subexp before its end. */ | |
3572 | || group_in_compile_stack (compile_stack, reg)) | |
0cdd06f8 | 3573 | FREE_STACK_RETURN (REG_ESUBREG); |
e318085a | 3574 | |
0cdd06f8 SM |
3575 | laststart = b; |
3576 | BUF_PUSH_2 (duplicate, reg); | |
3577 | } | |
25fe55af | 3578 | break; |
e318085a | 3579 | |
e318085a | 3580 | |
25fe55af RS |
3581 | case '+': |
3582 | case '?': | |
3583 | if (syntax & RE_BK_PLUS_QM) | |
3584 | goto handle_plus; | |
3585 | else | |
3586 | goto normal_backslash; | |
3587 | ||
3588 | default: | |
3589 | normal_backslash: | |
3590 | /* You might think it would be useful for \ to mean | |
3591 | not to translate; but if we don't translate it | |
4bb91c68 | 3592 | it will never match anything. */ |
25fe55af RS |
3593 | goto normal_char; |
3594 | } | |
3595 | break; | |
fa9a63c5 RM |
3596 | |
3597 | ||
3598 | default: | |
25fe55af | 3599 | /* Expects the character in `c'. */ |
fa9a63c5 | 3600 | normal_char: |
36595814 | 3601 | /* If no exactn currently being built. */ |
25fe55af | 3602 | if (!pending_exact |
fa9a63c5 | 3603 | |
25fe55af RS |
3604 | /* If last exactn not at current position. */ |
3605 | || pending_exact + *pending_exact + 1 != b | |
5e69f11e | 3606 | |
25fe55af | 3607 | /* We have only one byte following the exactn for the count. */ |
2d1675e4 | 3608 | || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH |
fa9a63c5 | 3609 | |
7814e705 | 3610 | /* If followed by a repetition operator. */ |
9d99031f | 3611 | || (p != pend && (*p == '*' || *p == '^')) |
fa9a63c5 | 3612 | || ((syntax & RE_BK_PLUS_QM) |
9d99031f RS |
3613 | ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?') |
3614 | : p != pend && (*p == '+' || *p == '?')) | |
fa9a63c5 | 3615 | || ((syntax & RE_INTERVALS) |
25fe55af | 3616 | && ((syntax & RE_NO_BK_BRACES) |
9d99031f RS |
3617 | ? p != pend && *p == '{' |
3618 | : p + 1 < pend && p[0] == '\\' && p[1] == '{'))) | |
fa9a63c5 RM |
3619 | { |
3620 | /* Start building a new exactn. */ | |
5e69f11e | 3621 | |
25fe55af | 3622 | laststart = b; |
fa9a63c5 RM |
3623 | |
3624 | BUF_PUSH_2 (exactn, 0); | |
3625 | pending_exact = b - 1; | |
25fe55af | 3626 | } |
5e69f11e | 3627 | |
2d1675e4 SM |
3628 | GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH); |
3629 | { | |
e0277a47 KH |
3630 | int len; |
3631 | ||
cf9c99bc | 3632 | if (multibyte) |
6fdd04b0 | 3633 | { |
cf9c99bc | 3634 | c = TRANSLATE (c); |
6fdd04b0 KH |
3635 | len = CHAR_STRING (c, b); |
3636 | b += len; | |
3637 | } | |
e0277a47 | 3638 | else |
6fdd04b0 | 3639 | { |
cf9c99bc KH |
3640 | c1 = RE_CHAR_TO_MULTIBYTE (c); |
3641 | if (! CHAR_BYTE8_P (c1)) | |
3642 | { | |
3643 | re_wchar_t c2 = TRANSLATE (c1); | |
3644 | ||
3645 | if (c1 != c2 && (c1 = RE_CHAR_TO_UNIBYTE (c2)) >= 0) | |
3646 | c = c1; | |
409f2919 | 3647 | } |
6fdd04b0 KH |
3648 | *b++ = c; |
3649 | len = 1; | |
3650 | } | |
2d1675e4 SM |
3651 | (*pending_exact) += len; |
3652 | } | |
3653 | ||
fa9a63c5 | 3654 | break; |
25fe55af | 3655 | } /* switch (c) */ |
fa9a63c5 RM |
3656 | } /* while p != pend */ |
3657 | ||
5e69f11e | 3658 | |
fa9a63c5 | 3659 | /* Through the pattern now. */ |
5e69f11e | 3660 | |
505bde11 | 3661 | FIXUP_ALT_JUMP (); |
fa9a63c5 | 3662 | |
5e69f11e | 3663 | if (!COMPILE_STACK_EMPTY) |
fa9a63c5 RM |
3664 | FREE_STACK_RETURN (REG_EPAREN); |
3665 | ||
3666 | /* If we don't want backtracking, force success | |
3667 | the first time we reach the end of the compiled pattern. */ | |
3668 | if (syntax & RE_NO_POSIX_BACKTRACKING) | |
3669 | BUF_PUSH (succeed); | |
3670 | ||
fa9a63c5 RM |
3671 | /* We have succeeded; set the length of the buffer. */ |
3672 | bufp->used = b - bufp->buffer; | |
3673 | ||
3674 | #ifdef DEBUG | |
99633e97 | 3675 | if (debug > 0) |
fa9a63c5 | 3676 | { |
505bde11 | 3677 | re_compile_fastmap (bufp); |
dc4a2ee0 | 3678 | DEBUG_PRINT ("\nCompiled pattern: \n"); |
fa9a63c5 RM |
3679 | print_compiled_pattern (bufp); |
3680 | } | |
99633e97 | 3681 | debug--; |
fa9a63c5 RM |
3682 | #endif /* DEBUG */ |
3683 | ||
3684 | #ifndef MATCH_MAY_ALLOCATE | |
3685 | /* Initialize the failure stack to the largest possible stack. This | |
3686 | isn't necessary unless we're trying to avoid calling alloca in | |
3687 | the search and match routines. */ | |
3688 | { | |
3689 | int num_regs = bufp->re_nsub + 1; | |
3690 | ||
320a2a73 | 3691 | if (fail_stack.size < re_max_failures * TYPICAL_FAILURE_SIZE) |
fa9a63c5 | 3692 | { |
a26f4ccd | 3693 | fail_stack.size = re_max_failures * TYPICAL_FAILURE_SIZE; |
38182d90 PE |
3694 | falk_stack.stack = realloc (fail_stack.stack, |
3695 | fail_stack.size * sizeof *falk_stack.stack); | |
fa9a63c5 RM |
3696 | } |
3697 | ||
3698 | regex_grow_registers (num_regs); | |
3699 | } | |
3700 | #endif /* not MATCH_MAY_ALLOCATE */ | |
3701 | ||
839966f3 | 3702 | FREE_STACK_RETURN (REG_NOERROR); |
fa9a63c5 RM |
3703 | } /* regex_compile */ |
3704 | \f | |
3705 | /* Subroutines for `regex_compile'. */ | |
3706 | ||
7814e705 | 3707 | /* Store OP at LOC followed by two-byte integer parameter ARG. */ |
fa9a63c5 RM |
3708 | |
3709 | static void | |
971de7fb | 3710 | store_op1 (re_opcode_t op, unsigned char *loc, int arg) |
fa9a63c5 RM |
3711 | { |
3712 | *loc = (unsigned char) op; | |
3713 | STORE_NUMBER (loc + 1, arg); | |
3714 | } | |
3715 | ||
3716 | ||
3717 | /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3718 | ||
3719 | static void | |
971de7fb | 3720 | store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2) |
fa9a63c5 RM |
3721 | { |
3722 | *loc = (unsigned char) op; | |
3723 | STORE_NUMBER (loc + 1, arg1); | |
3724 | STORE_NUMBER (loc + 3, arg2); | |
3725 | } | |
3726 | ||
3727 | ||
3728 | /* Copy the bytes from LOC to END to open up three bytes of space at LOC | |
3729 | for OP followed by two-byte integer parameter ARG. */ | |
3730 | ||
3731 | static void | |
971de7fb | 3732 | insert_op1 (re_opcode_t op, unsigned char *loc, int arg, unsigned char *end) |
fa9a63c5 RM |
3733 | { |
3734 | register unsigned char *pfrom = end; | |
3735 | register unsigned char *pto = end + 3; | |
3736 | ||
3737 | while (pfrom != loc) | |
3738 | *--pto = *--pfrom; | |
5e69f11e | 3739 | |
fa9a63c5 RM |
3740 | store_op1 (op, loc, arg); |
3741 | } | |
3742 | ||
3743 | ||
3744 | /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3745 | ||
3746 | static void | |
971de7fb | 3747 | insert_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2, unsigned char *end) |
fa9a63c5 RM |
3748 | { |
3749 | register unsigned char *pfrom = end; | |
3750 | register unsigned char *pto = end + 5; | |
3751 | ||
3752 | while (pfrom != loc) | |
3753 | *--pto = *--pfrom; | |
5e69f11e | 3754 | |
fa9a63c5 RM |
3755 | store_op2 (op, loc, arg1, arg2); |
3756 | } | |
3757 | ||
3758 | ||
3759 | /* P points to just after a ^ in PATTERN. Return true if that ^ comes | |
3760 | after an alternative or a begin-subexpression. We assume there is at | |
3761 | least one character before the ^. */ | |
3762 | ||
3763 | static boolean | |
29abe551 | 3764 | at_begline_loc_p (const_re_char *pattern, const_re_char *p, reg_syntax_t syntax) |
fa9a63c5 | 3765 | { |
01618498 | 3766 | re_char *prev = p - 2; |
95988fcf | 3767 | boolean odd_backslashes; |
5e69f11e | 3768 | |
95988fcf AS |
3769 | /* After a subexpression? */ |
3770 | if (*prev == '(') | |
3771 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3772 | ||
3773 | /* After an alternative? */ | |
3774 | else if (*prev == '|') | |
3775 | odd_backslashes = (syntax & RE_NO_BK_VBAR) == 0; | |
3776 | ||
3777 | /* After a shy subexpression? */ | |
3778 | else if (*prev == ':' && (syntax & RE_SHY_GROUPS)) | |
3779 | { | |
3780 | /* Skip over optional regnum. */ | |
3781 | while (prev - 1 >= pattern && prev[-1] >= '0' && prev[-1] <= '9') | |
3782 | --prev; | |
3783 | ||
3784 | if (!(prev - 2 >= pattern | |
3785 | && prev[-1] == '?' && prev[-2] == '(')) | |
3786 | return false; | |
3787 | prev -= 2; | |
3788 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3789 | } | |
3790 | else | |
3791 | return false; | |
3792 | ||
3793 | /* Count the number of preceding backslashes. */ | |
3794 | p = prev; | |
3795 | while (prev - 1 >= pattern && prev[-1] == '\\') | |
3796 | --prev; | |
3797 | return (p - prev) & odd_backslashes; | |
fa9a63c5 RM |
3798 | } |
3799 | ||
3800 | ||
3801 | /* The dual of at_begline_loc_p. This one is for $. We assume there is | |
3802 | at least one character after the $, i.e., `P < PEND'. */ | |
3803 | ||
3804 | static boolean | |
29abe551 | 3805 | at_endline_loc_p (const_re_char *p, const_re_char *pend, reg_syntax_t syntax) |
fa9a63c5 | 3806 | { |
01618498 | 3807 | re_char *next = p; |
fa9a63c5 | 3808 | boolean next_backslash = *next == '\\'; |
01618498 | 3809 | re_char *next_next = p + 1 < pend ? p + 1 : 0; |
5e69f11e | 3810 | |
fa9a63c5 RM |
3811 | return |
3812 | /* Before a subexpression? */ | |
3813 | (syntax & RE_NO_BK_PARENS ? *next == ')' | |
25fe55af | 3814 | : next_backslash && next_next && *next_next == ')') |
fa9a63c5 RM |
3815 | /* Before an alternative? */ |
3816 | || (syntax & RE_NO_BK_VBAR ? *next == '|' | |
25fe55af | 3817 | : next_backslash && next_next && *next_next == '|'); |
fa9a63c5 RM |
3818 | } |
3819 | ||
3820 | ||
5e69f11e | 3821 | /* Returns true if REGNUM is in one of COMPILE_STACK's elements and |
fa9a63c5 RM |
3822 | false if it's not. */ |
3823 | ||
3824 | static boolean | |
971de7fb | 3825 | group_in_compile_stack (compile_stack_type compile_stack, regnum_t regnum) |
fa9a63c5 | 3826 | { |
d1dfb56c | 3827 | ssize_t this_element; |
fa9a63c5 | 3828 | |
5e69f11e RM |
3829 | for (this_element = compile_stack.avail - 1; |
3830 | this_element >= 0; | |
fa9a63c5 RM |
3831 | this_element--) |
3832 | if (compile_stack.stack[this_element].regnum == regnum) | |
3833 | return true; | |
3834 | ||
3835 | return false; | |
3836 | } | |
fa9a63c5 | 3837 | \f |
f6a3f532 SM |
3838 | /* analyse_first. |
3839 | If fastmap is non-NULL, go through the pattern and fill fastmap | |
3840 | with all the possible leading chars. If fastmap is NULL, don't | |
3841 | bother filling it up (obviously) and only return whether the | |
3842 | pattern could potentially match the empty string. | |
3843 | ||
3844 | Return 1 if p..pend might match the empty string. | |
3845 | Return 0 if p..pend matches at least one char. | |
01618498 | 3846 | Return -1 if fastmap was not updated accurately. */ |
f6a3f532 SM |
3847 | |
3848 | static int | |
29abe551 PE |
3849 | analyse_first (const_re_char *p, const_re_char *pend, char *fastmap, |
3850 | const int multibyte) | |
fa9a63c5 | 3851 | { |
505bde11 | 3852 | int j, k; |
1fb352e0 | 3853 | boolean not; |
fa9a63c5 | 3854 | |
b18215fc | 3855 | /* If all elements for base leading-codes in fastmap is set, this |
7814e705 | 3856 | flag is set true. */ |
b18215fc RS |
3857 | boolean match_any_multibyte_characters = false; |
3858 | ||
f6a3f532 | 3859 | assert (p); |
5e69f11e | 3860 | |
505bde11 SM |
3861 | /* The loop below works as follows: |
3862 | - It has a working-list kept in the PATTERN_STACK and which basically | |
3863 | starts by only containing a pointer to the first operation. | |
3864 | - If the opcode we're looking at is a match against some set of | |
3865 | chars, then we add those chars to the fastmap and go on to the | |
3866 | next work element from the worklist (done via `break'). | |
3867 | - If the opcode is a control operator on the other hand, we either | |
3868 | ignore it (if it's meaningless at this point, such as `start_memory') | |
3869 | or execute it (if it's a jump). If the jump has several destinations | |
3870 | (i.e. `on_failure_jump'), then we push the other destination onto the | |
3871 | worklist. | |
3872 | We guarantee termination by ignoring backward jumps (more or less), | |
3873 | so that `p' is monotonically increasing. More to the point, we | |
3874 | never set `p' (or push) anything `<= p1'. */ | |
3875 | ||
01618498 | 3876 | while (p < pend) |
fa9a63c5 | 3877 | { |
505bde11 SM |
3878 | /* `p1' is used as a marker of how far back a `on_failure_jump' |
3879 | can go without being ignored. It is normally equal to `p' | |
3880 | (which prevents any backward `on_failure_jump') except right | |
3881 | after a plain `jump', to allow patterns such as: | |
3882 | 0: jump 10 | |
3883 | 3..9: <body> | |
3884 | 10: on_failure_jump 3 | |
3885 | as used for the *? operator. */ | |
01618498 | 3886 | re_char *p1 = p; |
5e69f11e | 3887 | |
7393bcbb | 3888 | switch (*p++) |
fa9a63c5 | 3889 | { |
f6a3f532 | 3890 | case succeed: |
01618498 | 3891 | return 1; |
fa9a63c5 | 3892 | |
fa9a63c5 | 3893 | case duplicate: |
505bde11 SM |
3894 | /* If the first character has to match a backreference, that means |
3895 | that the group was empty (since it already matched). Since this | |
3896 | is the only case that interests us here, we can assume that the | |
3897 | backreference must match the empty string. */ | |
3898 | p++; | |
3899 | continue; | |
fa9a63c5 RM |
3900 | |
3901 | ||
3902 | /* Following are the cases which match a character. These end | |
7814e705 | 3903 | with `break'. */ |
fa9a63c5 RM |
3904 | |
3905 | case exactn: | |
e0277a47 | 3906 | if (fastmap) |
cf9c99bc KH |
3907 | { |
3908 | /* If multibyte is nonzero, the first byte of each | |
3909 | character is an ASCII or a leading code. Otherwise, | |
3910 | each byte is a character. Thus, this works in both | |
3911 | cases. */ | |
3912 | fastmap[p[1]] = 1; | |
3913 | if (! multibyte) | |
3914 | { | |
3915 | /* For the case of matching this unibyte regex | |
3916 | against multibyte, we must set a leading code of | |
3917 | the corresponding multibyte character. */ | |
3918 | int c = RE_CHAR_TO_MULTIBYTE (p[1]); | |
3919 | ||
86e893e3 | 3920 | fastmap[CHAR_LEADING_CODE (c)] = 1; |
cf9c99bc KH |
3921 | } |
3922 | } | |
fa9a63c5 RM |
3923 | break; |
3924 | ||
3925 | ||
1fb352e0 SM |
3926 | case anychar: |
3927 | /* We could put all the chars except for \n (and maybe \0) | |
3928 | but we don't bother since it is generally not worth it. */ | |
f6a3f532 | 3929 | if (!fastmap) break; |
01618498 | 3930 | return -1; |
fa9a63c5 RM |
3931 | |
3932 | ||
b18215fc | 3933 | case charset_not: |
1fb352e0 | 3934 | if (!fastmap) break; |
bf216479 KH |
3935 | { |
3936 | /* Chars beyond end of bitmap are possible matches. */ | |
bf216479 | 3937 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH; |
cf9c99bc | 3938 | j < (1 << BYTEWIDTH); j++) |
bf216479 KH |
3939 | fastmap[j] = 1; |
3940 | } | |
3941 | ||
1fb352e0 SM |
3942 | /* Fallthrough */ |
3943 | case charset: | |
3944 | if (!fastmap) break; | |
3945 | not = (re_opcode_t) *(p - 1) == charset_not; | |
b18215fc RS |
3946 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH - 1, p++; |
3947 | j >= 0; j--) | |
1fb352e0 | 3948 | if (!!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) ^ not) |
49da453b | 3949 | fastmap[j] = 1; |
b18215fc | 3950 | |
6482db2e KH |
3951 | #ifdef emacs |
3952 | if (/* Any leading code can possibly start a character | |
1fb352e0 | 3953 | which doesn't match the specified set of characters. */ |
6482db2e | 3954 | not |
409f2919 | 3955 | || |
6482db2e KH |
3956 | /* If we can match a character class, we can match any |
3957 | multibyte characters. */ | |
3958 | (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) | |
3959 | && CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0)) | |
3960 | ||
b18215fc | 3961 | { |
b18215fc RS |
3962 | if (match_any_multibyte_characters == false) |
3963 | { | |
6482db2e KH |
3964 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
3965 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
6fdd04b0 | 3966 | fastmap[j] = 1; |
b18215fc RS |
3967 | match_any_multibyte_characters = true; |
3968 | } | |
3969 | } | |
b18215fc | 3970 | |
1fb352e0 SM |
3971 | else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) |
3972 | && match_any_multibyte_characters == false) | |
3973 | { | |
bf216479 | 3974 | /* Set fastmap[I] to 1 where I is a leading code of each |
51e4f4a8 | 3975 | multibyte character in the range table. */ |
1fb352e0 | 3976 | int c, count; |
bf216479 | 3977 | unsigned char lc1, lc2; |
b18215fc | 3978 | |
1fb352e0 | 3979 | /* Make P points the range table. `+ 2' is to skip flag |
0b32bf0e | 3980 | bits for a character class. */ |
1fb352e0 | 3981 | p += CHARSET_BITMAP_SIZE (&p[-2]) + 2; |
b18215fc | 3982 | |
1fb352e0 SM |
3983 | /* Extract the number of ranges in range table into COUNT. */ |
3984 | EXTRACT_NUMBER_AND_INCR (count, p); | |
cf9c99bc | 3985 | for (; count > 0; count--, p += 3) |
1fb352e0 | 3986 | { |
9117d724 KH |
3987 | /* Extract the start and end of each range. */ |
3988 | EXTRACT_CHARACTER (c, p); | |
bf216479 | 3989 | lc1 = CHAR_LEADING_CODE (c); |
9117d724 | 3990 | p += 3; |
1fb352e0 | 3991 | EXTRACT_CHARACTER (c, p); |
bf216479 KH |
3992 | lc2 = CHAR_LEADING_CODE (c); |
3993 | for (j = lc1; j <= lc2; j++) | |
9117d724 | 3994 | fastmap[j] = 1; |
1fb352e0 SM |
3995 | } |
3996 | } | |
6482db2e | 3997 | #endif |
b18215fc RS |
3998 | break; |
3999 | ||
1fb352e0 SM |
4000 | case syntaxspec: |
4001 | case notsyntaxspec: | |
4002 | if (!fastmap) break; | |
4003 | #ifndef emacs | |
4004 | not = (re_opcode_t)p[-1] == notsyntaxspec; | |
4005 | k = *p++; | |
4006 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
990b2375 | 4007 | if ((SYNTAX (j) == (enum syntaxcode) k) ^ not) |
b18215fc | 4008 | fastmap[j] = 1; |
b18215fc | 4009 | break; |
1fb352e0 | 4010 | #else /* emacs */ |
b18215fc RS |
4011 | /* This match depends on text properties. These end with |
4012 | aborting optimizations. */ | |
01618498 | 4013 | return -1; |
b18215fc RS |
4014 | |
4015 | case categoryspec: | |
b18215fc | 4016 | case notcategoryspec: |
1fb352e0 SM |
4017 | if (!fastmap) break; |
4018 | not = (re_opcode_t)p[-1] == notcategoryspec; | |
b18215fc | 4019 | k = *p++; |
6482db2e | 4020 | for (j = (1 << BYTEWIDTH); j >= 0; j--) |
1fb352e0 | 4021 | if ((CHAR_HAS_CATEGORY (j, k)) ^ not) |
b18215fc RS |
4022 | fastmap[j] = 1; |
4023 | ||
6482db2e KH |
4024 | /* Any leading code can possibly start a character which |
4025 | has or doesn't has the specified category. */ | |
4026 | if (match_any_multibyte_characters == false) | |
6fdd04b0 | 4027 | { |
6482db2e KH |
4028 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
4029 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
4030 | fastmap[j] = 1; | |
4031 | match_any_multibyte_characters = true; | |
6fdd04b0 | 4032 | } |
b18215fc RS |
4033 | break; |
4034 | ||
fa9a63c5 | 4035 | /* All cases after this match the empty string. These end with |
25fe55af | 4036 | `continue'. */ |
fa9a63c5 | 4037 | |
fa9a63c5 RM |
4038 | case before_dot: |
4039 | case at_dot: | |
4040 | case after_dot: | |
1fb352e0 | 4041 | #endif /* !emacs */ |
25fe55af RS |
4042 | case no_op: |
4043 | case begline: | |
4044 | case endline: | |
fa9a63c5 RM |
4045 | case begbuf: |
4046 | case endbuf: | |
4047 | case wordbound: | |
4048 | case notwordbound: | |
4049 | case wordbeg: | |
4050 | case wordend: | |
669fa600 SM |
4051 | case symbeg: |
4052 | case symend: | |
25fe55af | 4053 | continue; |
fa9a63c5 RM |
4054 | |
4055 | ||
fa9a63c5 | 4056 | case jump: |
25fe55af | 4057 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 SM |
4058 | if (j < 0) |
4059 | /* Backward jumps can only go back to code that we've already | |
4060 | visited. `re_compile' should make sure this is true. */ | |
4061 | break; | |
25fe55af | 4062 | p += j; |
7393bcbb | 4063 | switch (*p) |
505bde11 SM |
4064 | { |
4065 | case on_failure_jump: | |
4066 | case on_failure_keep_string_jump: | |
505bde11 | 4067 | case on_failure_jump_loop: |
0683b6fa | 4068 | case on_failure_jump_nastyloop: |
505bde11 SM |
4069 | case on_failure_jump_smart: |
4070 | p++; | |
4071 | break; | |
4072 | default: | |
4073 | continue; | |
4074 | }; | |
4075 | /* Keep `p1' to allow the `on_failure_jump' we are jumping to | |
4076 | to jump back to "just after here". */ | |
4077 | /* Fallthrough */ | |
fa9a63c5 | 4078 | |
25fe55af RS |
4079 | case on_failure_jump: |
4080 | case on_failure_keep_string_jump: | |
0683b6fa | 4081 | case on_failure_jump_nastyloop: |
505bde11 SM |
4082 | case on_failure_jump_loop: |
4083 | case on_failure_jump_smart: | |
25fe55af | 4084 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 | 4085 | if (p + j <= p1) |
ed0767d8 | 4086 | ; /* Backward jump to be ignored. */ |
01618498 SM |
4087 | else |
4088 | { /* We have to look down both arms. | |
4089 | We first go down the "straight" path so as to minimize | |
4090 | stack usage when going through alternatives. */ | |
4091 | int r = analyse_first (p, pend, fastmap, multibyte); | |
4092 | if (r) return r; | |
4093 | p += j; | |
4094 | } | |
25fe55af | 4095 | continue; |
fa9a63c5 RM |
4096 | |
4097 | ||
ed0767d8 SM |
4098 | case jump_n: |
4099 | /* This code simply does not properly handle forward jump_n. */ | |
4100 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p); assert (j < 0)); | |
4101 | p += 4; | |
4102 | /* jump_n can either jump or fall through. The (backward) jump | |
4103 | case has already been handled, so we only need to look at the | |
4104 | fallthrough case. */ | |
4105 | continue; | |
177c0ea7 | 4106 | |
fa9a63c5 | 4107 | case succeed_n: |
ed0767d8 SM |
4108 | /* If N == 0, it should be an on_failure_jump_loop instead. */ |
4109 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0)); | |
4110 | p += 4; | |
4111 | /* We only care about one iteration of the loop, so we don't | |
4112 | need to consider the case where this behaves like an | |
4113 | on_failure_jump. */ | |
25fe55af | 4114 | continue; |
fa9a63c5 RM |
4115 | |
4116 | ||
4117 | case set_number_at: | |
25fe55af RS |
4118 | p += 4; |
4119 | continue; | |
fa9a63c5 RM |
4120 | |
4121 | ||
4122 | case start_memory: | |
25fe55af | 4123 | case stop_memory: |
505bde11 | 4124 | p += 1; |
fa9a63c5 RM |
4125 | continue; |
4126 | ||
4127 | ||
4128 | default: | |
25fe55af RS |
4129 | abort (); /* We have listed all the cases. */ |
4130 | } /* switch *p++ */ | |
fa9a63c5 RM |
4131 | |
4132 | /* Getting here means we have found the possible starting | |
25fe55af | 4133 | characters for one path of the pattern -- and that the empty |
7814e705 | 4134 | string does not match. We need not follow this path further. */ |
01618498 | 4135 | return 0; |
fa9a63c5 RM |
4136 | } /* while p */ |
4137 | ||
01618498 SM |
4138 | /* We reached the end without matching anything. */ |
4139 | return 1; | |
4140 | ||
f6a3f532 SM |
4141 | } /* analyse_first */ |
4142 | \f | |
4143 | /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in | |
4144 | BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible | |
4145 | characters can start a string that matches the pattern. This fastmap | |
4146 | is used by re_search to skip quickly over impossible starting points. | |
4147 | ||
4148 | Character codes above (1 << BYTEWIDTH) are not represented in the | |
4149 | fastmap, but the leading codes are represented. Thus, the fastmap | |
4150 | indicates which character sets could start a match. | |
4151 | ||
4152 | The caller must supply the address of a (1 << BYTEWIDTH)-byte data | |
4153 | area as BUFP->fastmap. | |
4154 | ||
4155 | We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in | |
4156 | the pattern buffer. | |
4157 | ||
4158 | Returns 0 if we succeed, -2 if an internal error. */ | |
4159 | ||
4160 | int | |
971de7fb | 4161 | re_compile_fastmap (struct re_pattern_buffer *bufp) |
f6a3f532 SM |
4162 | { |
4163 | char *fastmap = bufp->fastmap; | |
4164 | int analysis; | |
4165 | ||
4166 | assert (fastmap && bufp->buffer); | |
4167 | ||
72af86bd | 4168 | memset (fastmap, 0, 1 << BYTEWIDTH); /* Assume nothing's valid. */ |
f6a3f532 SM |
4169 | bufp->fastmap_accurate = 1; /* It will be when we're done. */ |
4170 | ||
4171 | analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used, | |
2d1675e4 | 4172 | fastmap, RE_MULTIBYTE_P (bufp)); |
c0f9ea08 | 4173 | bufp->can_be_null = (analysis != 0); |
fa9a63c5 RM |
4174 | return 0; |
4175 | } /* re_compile_fastmap */ | |
4176 | \f | |
4177 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | |
4178 | ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | |
4179 | this memory for recording register information. STARTS and ENDS | |
4180 | must be allocated using the malloc library routine, and must each | |
4181 | be at least NUM_REGS * sizeof (regoff_t) bytes long. | |
4182 | ||
4183 | If NUM_REGS == 0, then subsequent matches should allocate their own | |
4184 | register data. | |
4185 | ||
4186 | Unless this function is called, the first search or match using | |
4187 | PATTERN_BUFFER will allocate its own register data, without | |
4188 | freeing the old data. */ | |
4189 | ||
4190 | void | |
971de7fb | 4191 | re_set_registers (struct re_pattern_buffer *bufp, struct re_registers *regs, unsigned int num_regs, regoff_t *starts, regoff_t *ends) |
fa9a63c5 RM |
4192 | { |
4193 | if (num_regs) | |
4194 | { | |
4195 | bufp->regs_allocated = REGS_REALLOCATE; | |
4196 | regs->num_regs = num_regs; | |
4197 | regs->start = starts; | |
4198 | regs->end = ends; | |
4199 | } | |
4200 | else | |
4201 | { | |
4202 | bufp->regs_allocated = REGS_UNALLOCATED; | |
4203 | regs->num_regs = 0; | |
7d652d97 | 4204 | regs->start = regs->end = 0; |
fa9a63c5 RM |
4205 | } |
4206 | } | |
c0f9ea08 | 4207 | WEAK_ALIAS (__re_set_registers, re_set_registers) |
fa9a63c5 | 4208 | \f |
7814e705 | 4209 | /* Searching routines. */ |
fa9a63c5 RM |
4210 | |
4211 | /* Like re_search_2, below, but only one string is specified, and | |
4212 | doesn't let you say where to stop matching. */ | |
4213 | ||
d1dfb56c EZ |
4214 | regoff_t |
4215 | re_search (struct re_pattern_buffer *bufp, const char *string, size_t size, | |
4216 | ssize_t startpos, ssize_t range, struct re_registers *regs) | |
fa9a63c5 | 4217 | { |
5e69f11e | 4218 | return re_search_2 (bufp, NULL, 0, string, size, startpos, range, |
fa9a63c5 RM |
4219 | regs, size); |
4220 | } | |
c0f9ea08 | 4221 | WEAK_ALIAS (__re_search, re_search) |
fa9a63c5 | 4222 | |
70806df6 KH |
4223 | /* Head address of virtual concatenation of string. */ |
4224 | #define HEAD_ADDR_VSTRING(P) \ | |
4225 | (((P) >= size1 ? string2 : string1)) | |
4226 | ||
b18215fc RS |
4227 | /* Address of POS in the concatenation of virtual string. */ |
4228 | #define POS_ADDR_VSTRING(POS) \ | |
4229 | (((POS) >= size1 ? string2 - size1 : string1) + (POS)) | |
fa9a63c5 RM |
4230 | |
4231 | /* Using the compiled pattern in BUFP->buffer, first tries to match the | |
4232 | virtual concatenation of STRING1 and STRING2, starting first at index | |
4233 | STARTPOS, then at STARTPOS + 1, and so on. | |
5e69f11e | 4234 | |
fa9a63c5 | 4235 | STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. |
5e69f11e | 4236 | |
fa9a63c5 RM |
4237 | RANGE is how far to scan while trying to match. RANGE = 0 means try |
4238 | only at STARTPOS; in general, the last start tried is STARTPOS + | |
4239 | RANGE. | |
5e69f11e | 4240 | |
fa9a63c5 RM |
4241 | In REGS, return the indices of the virtual concatenation of STRING1 |
4242 | and STRING2 that matched the entire BUFP->buffer and its contained | |
4243 | subexpressions. | |
5e69f11e | 4244 | |
fa9a63c5 RM |
4245 | Do not consider matching one past the index STOP in the virtual |
4246 | concatenation of STRING1 and STRING2. | |
4247 | ||
4248 | We return either the position in the strings at which the match was | |
4249 | found, -1 if no match, or -2 if error (such as failure | |
4250 | stack overflow). */ | |
4251 | ||
d1dfb56c EZ |
4252 | regoff_t |
4253 | re_search_2 (struct re_pattern_buffer *bufp, const char *str1, size_t size1, | |
4254 | const char *str2, size_t size2, ssize_t startpos, ssize_t range, | |
4255 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4256 | { |
d1dfb56c | 4257 | regoff_t val; |
66f0296e SM |
4258 | re_char *string1 = (re_char*) str1; |
4259 | re_char *string2 = (re_char*) str2; | |
fa9a63c5 | 4260 | register char *fastmap = bufp->fastmap; |
6676cb1c | 4261 | register RE_TRANSLATE_TYPE translate = bufp->translate; |
d1dfb56c EZ |
4262 | size_t total_size = size1 + size2; |
4263 | ssize_t endpos = startpos + range; | |
c0f9ea08 | 4264 | boolean anchored_start; |
cf9c99bc KH |
4265 | /* Nonzero if we are searching multibyte string. */ |
4266 | const boolean multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
b18215fc | 4267 | |
fa9a63c5 RM |
4268 | /* Check for out-of-range STARTPOS. */ |
4269 | if (startpos < 0 || startpos > total_size) | |
4270 | return -1; | |
5e69f11e | 4271 | |
fa9a63c5 | 4272 | /* Fix up RANGE if it might eventually take us outside |
34597fa9 | 4273 | the virtual concatenation of STRING1 and STRING2. |
5e69f11e | 4274 | Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ |
34597fa9 RS |
4275 | if (endpos < 0) |
4276 | range = 0 - startpos; | |
fa9a63c5 RM |
4277 | else if (endpos > total_size) |
4278 | range = total_size - startpos; | |
4279 | ||
4280 | /* If the search isn't to be a backwards one, don't waste time in a | |
7b140fd7 | 4281 | search for a pattern anchored at beginning of buffer. */ |
fa9a63c5 RM |
4282 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) |
4283 | { | |
4284 | if (startpos > 0) | |
4285 | return -1; | |
4286 | else | |
7b140fd7 | 4287 | range = 0; |
fa9a63c5 RM |
4288 | } |
4289 | ||
ae4788a8 RS |
4290 | #ifdef emacs |
4291 | /* In a forward search for something that starts with \=. | |
4292 | don't keep searching past point. */ | |
4293 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) | |
4294 | { | |
7b140fd7 RS |
4295 | range = PT_BYTE - BEGV_BYTE - startpos; |
4296 | if (range < 0) | |
ae4788a8 RS |
4297 | return -1; |
4298 | } | |
4299 | #endif /* emacs */ | |
4300 | ||
fa9a63c5 RM |
4301 | /* Update the fastmap now if not correct already. */ |
4302 | if (fastmap && !bufp->fastmap_accurate) | |
01618498 | 4303 | re_compile_fastmap (bufp); |
5e69f11e | 4304 | |
c8499ba5 | 4305 | /* See whether the pattern is anchored. */ |
c0f9ea08 | 4306 | anchored_start = (bufp->buffer[0] == begline); |
c8499ba5 | 4307 | |
b18215fc | 4308 | #ifdef emacs |
d48cd3f4 | 4309 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
cc9b4df2 | 4310 | { |
d1dfb56c | 4311 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (startpos)); |
cc9b4df2 KH |
4312 | |
4313 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); | |
4314 | } | |
b18215fc RS |
4315 | #endif |
4316 | ||
fa9a63c5 RM |
4317 | /* Loop through the string, looking for a place to start matching. */ |
4318 | for (;;) | |
5e69f11e | 4319 | { |
c8499ba5 RS |
4320 | /* If the pattern is anchored, |
4321 | skip quickly past places we cannot match. | |
4322 | We don't bother to treat startpos == 0 specially | |
4323 | because that case doesn't repeat. */ | |
4324 | if (anchored_start && startpos > 0) | |
4325 | { | |
c0f9ea08 SM |
4326 | if (! ((startpos <= size1 ? string1[startpos - 1] |
4327 | : string2[startpos - size1 - 1]) | |
4328 | == '\n')) | |
c8499ba5 RS |
4329 | goto advance; |
4330 | } | |
4331 | ||
fa9a63c5 | 4332 | /* If a fastmap is supplied, skip quickly over characters that |
25fe55af RS |
4333 | cannot be the start of a match. If the pattern can match the |
4334 | null string, however, we don't need to skip characters; we want | |
7814e705 | 4335 | the first null string. */ |
fa9a63c5 RM |
4336 | if (fastmap && startpos < total_size && !bufp->can_be_null) |
4337 | { | |
66f0296e | 4338 | register re_char *d; |
01618498 | 4339 | register re_wchar_t buf_ch; |
e934739e RS |
4340 | |
4341 | d = POS_ADDR_VSTRING (startpos); | |
4342 | ||
7814e705 | 4343 | if (range > 0) /* Searching forwards. */ |
fa9a63c5 | 4344 | { |
fa9a63c5 | 4345 | register int lim = 0; |
d1dfb56c | 4346 | ssize_t irange = range; |
fa9a63c5 | 4347 | |
25fe55af RS |
4348 | if (startpos < size1 && startpos + range >= size1) |
4349 | lim = range - (size1 - startpos); | |
fa9a63c5 | 4350 | |
25fe55af RS |
4351 | /* Written out as an if-else to avoid testing `translate' |
4352 | inside the loop. */ | |
28ae27ae AS |
4353 | if (RE_TRANSLATE_P (translate)) |
4354 | { | |
e934739e RS |
4355 | if (multibyte) |
4356 | while (range > lim) | |
4357 | { | |
4358 | int buf_charlen; | |
4359 | ||
62a6e103 | 4360 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 4361 | buf_ch = RE_TRANSLATE (translate, buf_ch); |
bf216479 | 4362 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
e934739e RS |
4363 | break; |
4364 | ||
4365 | range -= buf_charlen; | |
4366 | d += buf_charlen; | |
4367 | } | |
4368 | else | |
bf216479 | 4369 | while (range > lim) |
33c46939 | 4370 | { |
cf9c99bc KH |
4371 | register re_wchar_t ch, translated; |
4372 | ||
bf216479 | 4373 | buf_ch = *d; |
cf9c99bc KH |
4374 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); |
4375 | translated = RE_TRANSLATE (translate, ch); | |
4376 | if (translated != ch | |
4377 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4378 | buf_ch = ch; | |
6fdd04b0 | 4379 | if (fastmap[buf_ch]) |
bf216479 | 4380 | break; |
33c46939 RS |
4381 | d++; |
4382 | range--; | |
4383 | } | |
e934739e | 4384 | } |
fa9a63c5 | 4385 | else |
6fdd04b0 KH |
4386 | { |
4387 | if (multibyte) | |
4388 | while (range > lim) | |
4389 | { | |
4390 | int buf_charlen; | |
fa9a63c5 | 4391 | |
62a6e103 | 4392 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
6fdd04b0 KH |
4393 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
4394 | break; | |
4395 | range -= buf_charlen; | |
4396 | d += buf_charlen; | |
4397 | } | |
e934739e | 4398 | else |
6fdd04b0 | 4399 | while (range > lim && !fastmap[*d]) |
33c46939 RS |
4400 | { |
4401 | d++; | |
4402 | range--; | |
4403 | } | |
e934739e | 4404 | } |
fa9a63c5 RM |
4405 | startpos += irange - range; |
4406 | } | |
7814e705 | 4407 | else /* Searching backwards. */ |
fa9a63c5 | 4408 | { |
ba5e343c KH |
4409 | if (multibyte) |
4410 | { | |
62a6e103 | 4411 | buf_ch = STRING_CHAR (d); |
ba5e343c KH |
4412 | buf_ch = TRANSLATE (buf_ch); |
4413 | if (! fastmap[CHAR_LEADING_CODE (buf_ch)]) | |
4414 | goto advance; | |
4415 | } | |
4416 | else | |
4417 | { | |
cf9c99bc KH |
4418 | register re_wchar_t ch, translated; |
4419 | ||
4420 | buf_ch = *d; | |
4421 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); | |
4422 | translated = TRANSLATE (ch); | |
4423 | if (translated != ch | |
4424 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4425 | buf_ch = ch; | |
4426 | if (! fastmap[TRANSLATE (buf_ch)]) | |
ba5e343c KH |
4427 | goto advance; |
4428 | } | |
fa9a63c5 RM |
4429 | } |
4430 | } | |
4431 | ||
4432 | /* If can't match the null string, and that's all we have left, fail. */ | |
4433 | if (range >= 0 && startpos == total_size && fastmap | |
25fe55af | 4434 | && !bufp->can_be_null) |
fa9a63c5 RM |
4435 | return -1; |
4436 | ||
4437 | val = re_match_2_internal (bufp, string1, size1, string2, size2, | |
4438 | startpos, regs, stop); | |
fa9a63c5 RM |
4439 | |
4440 | if (val >= 0) | |
4441 | return startpos; | |
5e69f11e | 4442 | |
fa9a63c5 RM |
4443 | if (val == -2) |
4444 | return -2; | |
4445 | ||
4446 | advance: | |
5e69f11e | 4447 | if (!range) |
25fe55af | 4448 | break; |
5e69f11e | 4449 | else if (range > 0) |
25fe55af | 4450 | { |
b18215fc RS |
4451 | /* Update STARTPOS to the next character boundary. */ |
4452 | if (multibyte) | |
4453 | { | |
66f0296e | 4454 | re_char *p = POS_ADDR_VSTRING (startpos); |
aa3830c4 | 4455 | int len = BYTES_BY_CHAR_HEAD (*p); |
b18215fc RS |
4456 | |
4457 | range -= len; | |
4458 | if (range < 0) | |
4459 | break; | |
4460 | startpos += len; | |
4461 | } | |
4462 | else | |
4463 | { | |
b560c397 RS |
4464 | range--; |
4465 | startpos++; | |
4466 | } | |
e318085a | 4467 | } |
fa9a63c5 | 4468 | else |
25fe55af RS |
4469 | { |
4470 | range++; | |
4471 | startpos--; | |
b18215fc RS |
4472 | |
4473 | /* Update STARTPOS to the previous character boundary. */ | |
4474 | if (multibyte) | |
4475 | { | |
70806df6 KH |
4476 | re_char *p = POS_ADDR_VSTRING (startpos) + 1; |
4477 | re_char *p0 = p; | |
4478 | re_char *phead = HEAD_ADDR_VSTRING (startpos); | |
b18215fc RS |
4479 | |
4480 | /* Find the head of multibyte form. */ | |
70806df6 KH |
4481 | PREV_CHAR_BOUNDARY (p, phead); |
4482 | range += p0 - 1 - p; | |
4483 | if (range > 0) | |
4484 | break; | |
b18215fc | 4485 | |
70806df6 | 4486 | startpos -= p0 - 1 - p; |
b18215fc | 4487 | } |
25fe55af | 4488 | } |
fa9a63c5 RM |
4489 | } |
4490 | return -1; | |
4491 | } /* re_search_2 */ | |
c0f9ea08 | 4492 | WEAK_ALIAS (__re_search_2, re_search_2) |
fa9a63c5 RM |
4493 | \f |
4494 | /* Declarations and macros for re_match_2. */ | |
4495 | ||
261cb4bb PE |
4496 | static int bcmp_translate (re_char *s1, re_char *s2, |
4497 | register ssize_t len, | |
4498 | RE_TRANSLATE_TYPE translate, | |
4499 | const int multibyte); | |
fa9a63c5 RM |
4500 | |
4501 | /* This converts PTR, a pointer into one of the search strings `string1' | |
4502 | and `string2' into an offset from the beginning of that string. */ | |
4503 | #define POINTER_TO_OFFSET(ptr) \ | |
4504 | (FIRST_STRING_P (ptr) \ | |
dc4a2ee0 PE |
4505 | ? (ptr) - string1 \ |
4506 | : (ptr) - string2 + (ptrdiff_t) size1) | |
fa9a63c5 | 4507 | |
fa9a63c5 | 4508 | /* Call before fetching a character with *d. This switches over to |
419d1c74 SM |
4509 | string2 if necessary. |
4510 | Check re_match_2_internal for a discussion of why end_match_2 might | |
4511 | not be within string2 (but be equal to end_match_1 instead). */ | |
fa9a63c5 | 4512 | #define PREFETCH() \ |
25fe55af | 4513 | while (d == dend) \ |
fa9a63c5 RM |
4514 | { \ |
4515 | /* End of string2 => fail. */ \ | |
25fe55af RS |
4516 | if (dend == end_match_2) \ |
4517 | goto fail; \ | |
4bb91c68 | 4518 | /* End of string1 => advance to string2. */ \ |
25fe55af | 4519 | d = string2; \ |
fa9a63c5 RM |
4520 | dend = end_match_2; \ |
4521 | } | |
4522 | ||
f1ad044f SM |
4523 | /* Call before fetching a char with *d if you already checked other limits. |
4524 | This is meant for use in lookahead operations like wordend, etc.. | |
4525 | where we might need to look at parts of the string that might be | |
4526 | outside of the LIMITs (i.e past `stop'). */ | |
4527 | #define PREFETCH_NOLIMIT() \ | |
4528 | if (d == end1) \ | |
4529 | { \ | |
4530 | d = string2; \ | |
4531 | dend = end_match_2; \ | |
4532 | } \ | |
fa9a63c5 RM |
4533 | |
4534 | /* Test if at very beginning or at very end of the virtual concatenation | |
7814e705 | 4535 | of `string1' and `string2'. If only one string, it's `string2'. */ |
fa9a63c5 | 4536 | #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) |
5e69f11e | 4537 | #define AT_STRINGS_END(d) ((d) == end2) |
fa9a63c5 | 4538 | |
9121ca40 | 4539 | /* Disabled due to a compiler bug -- see comment at case wordbound */ |
b18215fc RS |
4540 | |
4541 | /* The comment at case wordbound is following one, but we don't use | |
4542 | AT_WORD_BOUNDARY anymore to support multibyte form. | |
4543 | ||
4544 | The DEC Alpha C compiler 3.x generates incorrect code for the | |
25fe55af | 4545 | test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of |
7814e705 | 4546 | AT_WORD_BOUNDARY, so this code is disabled. Expanding the |
b18215fc RS |
4547 | macro and introducing temporary variables works around the bug. */ |
4548 | ||
9121ca40 | 4549 | #if 0 |
b313f9d8 PE |
4550 | /* Test if D points to a character which is word-constituent. We have |
4551 | two special cases to check for: if past the end of string1, look at | |
4552 | the first character in string2; and if before the beginning of | |
4553 | string2, look at the last character in string1. */ | |
4554 | #define WORDCHAR_P(d) \ | |
4555 | (SYNTAX ((d) == end1 ? *string2 \ | |
4556 | : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ | |
4557 | == Sword) | |
4558 | ||
fa9a63c5 RM |
4559 | /* Test if the character before D and the one at D differ with respect |
4560 | to being word-constituent. */ | |
4561 | #define AT_WORD_BOUNDARY(d) \ | |
4562 | (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ | |
4563 | || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) | |
9121ca40 | 4564 | #endif |
fa9a63c5 RM |
4565 | |
4566 | /* Free everything we malloc. */ | |
4567 | #ifdef MATCH_MAY_ALLOCATE | |
952db0d7 PE |
4568 | # define FREE_VAR(var) \ |
4569 | do { \ | |
4570 | if (var) \ | |
4571 | { \ | |
4572 | REGEX_FREE (var); \ | |
4573 | var = NULL; \ | |
4574 | } \ | |
4575 | } while (0) | |
0b32bf0e | 4576 | # define FREE_VARIABLES() \ |
fa9a63c5 RM |
4577 | do { \ |
4578 | REGEX_FREE_STACK (fail_stack.stack); \ | |
4579 | FREE_VAR (regstart); \ | |
4580 | FREE_VAR (regend); \ | |
fa9a63c5 RM |
4581 | FREE_VAR (best_regstart); \ |
4582 | FREE_VAR (best_regend); \ | |
fa9a63c5 RM |
4583 | } while (0) |
4584 | #else | |
0b32bf0e | 4585 | # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
4586 | #endif /* not MATCH_MAY_ALLOCATE */ |
4587 | ||
505bde11 SM |
4588 | \f |
4589 | /* Optimization routines. */ | |
4590 | ||
4e8a9132 SM |
4591 | /* If the operation is a match against one or more chars, |
4592 | return a pointer to the next operation, else return NULL. */ | |
01618498 | 4593 | static re_char * |
29abe551 | 4594 | skip_one_char (const_re_char *p) |
4e8a9132 | 4595 | { |
7393bcbb | 4596 | switch (*p++) |
4e8a9132 SM |
4597 | { |
4598 | case anychar: | |
4599 | break; | |
177c0ea7 | 4600 | |
4e8a9132 SM |
4601 | case exactn: |
4602 | p += *p + 1; | |
4603 | break; | |
4604 | ||
4605 | case charset_not: | |
4606 | case charset: | |
4607 | if (CHARSET_RANGE_TABLE_EXISTS_P (p - 1)) | |
4608 | { | |
4609 | int mcnt; | |
4610 | p = CHARSET_RANGE_TABLE (p - 1); | |
4611 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4612 | p = CHARSET_RANGE_TABLE_END (p, mcnt); | |
4613 | } | |
4614 | else | |
4615 | p += 1 + CHARSET_BITMAP_SIZE (p - 1); | |
4616 | break; | |
177c0ea7 | 4617 | |
4e8a9132 SM |
4618 | case syntaxspec: |
4619 | case notsyntaxspec: | |
1fb352e0 | 4620 | #ifdef emacs |
4e8a9132 SM |
4621 | case categoryspec: |
4622 | case notcategoryspec: | |
4623 | #endif /* emacs */ | |
4624 | p++; | |
4625 | break; | |
4626 | ||
4627 | default: | |
4628 | p = NULL; | |
4629 | } | |
4630 | return p; | |
4631 | } | |
4632 | ||
4633 | ||
505bde11 | 4634 | /* Jump over non-matching operations. */ |
839966f3 | 4635 | static re_char * |
29abe551 | 4636 | skip_noops (const_re_char *p, const_re_char *pend) |
505bde11 SM |
4637 | { |
4638 | int mcnt; | |
4639 | while (p < pend) | |
4640 | { | |
7393bcbb | 4641 | switch (*p) |
505bde11 SM |
4642 | { |
4643 | case start_memory: | |
505bde11 SM |
4644 | case stop_memory: |
4645 | p += 2; break; | |
4646 | case no_op: | |
4647 | p += 1; break; | |
4648 | case jump: | |
4649 | p += 1; | |
4650 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4651 | p += mcnt; | |
4652 | break; | |
4653 | default: | |
4654 | return p; | |
4655 | } | |
4656 | } | |
4657 | assert (p == pend); | |
4658 | return p; | |
4659 | } | |
4660 | ||
4661 | /* Non-zero if "p1 matches something" implies "p2 fails". */ | |
4662 | static int | |
29abe551 PE |
4663 | mutually_exclusive_p (struct re_pattern_buffer *bufp, const_re_char *p1, |
4664 | const_re_char *p2) | |
505bde11 | 4665 | { |
4e8a9132 | 4666 | re_opcode_t op2; |
2d1675e4 | 4667 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
505bde11 SM |
4668 | unsigned char *pend = bufp->buffer + bufp->used; |
4669 | ||
4e8a9132 | 4670 | assert (p1 >= bufp->buffer && p1 < pend |
505bde11 SM |
4671 | && p2 >= bufp->buffer && p2 <= pend); |
4672 | ||
4673 | /* Skip over open/close-group commands. | |
4674 | If what follows this loop is a ...+ construct, | |
4675 | look at what begins its body, since we will have to | |
4676 | match at least one of that. */ | |
4e8a9132 SM |
4677 | p2 = skip_noops (p2, pend); |
4678 | /* The same skip can be done for p1, except that this function | |
4679 | is only used in the case where p1 is a simple match operator. */ | |
4680 | /* p1 = skip_noops (p1, pend); */ | |
4681 | ||
4682 | assert (p1 >= bufp->buffer && p1 < pend | |
4683 | && p2 >= bufp->buffer && p2 <= pend); | |
4684 | ||
4685 | op2 = p2 == pend ? succeed : *p2; | |
4686 | ||
7393bcbb | 4687 | switch (op2) |
505bde11 | 4688 | { |
4e8a9132 SM |
4689 | case succeed: |
4690 | case endbuf: | |
4691 | /* If we're at the end of the pattern, we can change. */ | |
4692 | if (skip_one_char (p1)) | |
505bde11 | 4693 | { |
dc4a2ee0 | 4694 | DEBUG_PRINT (" End of pattern: fast loop.\n"); |
505bde11 | 4695 | return 1; |
505bde11 | 4696 | } |
4e8a9132 | 4697 | break; |
177c0ea7 | 4698 | |
4e8a9132 | 4699 | case endline: |
4e8a9132 SM |
4700 | case exactn: |
4701 | { | |
01618498 | 4702 | register re_wchar_t c |
4e8a9132 | 4703 | = (re_opcode_t) *p2 == endline ? '\n' |
62a6e103 | 4704 | : RE_STRING_CHAR (p2 + 2, multibyte); |
505bde11 | 4705 | |
4e8a9132 SM |
4706 | if ((re_opcode_t) *p1 == exactn) |
4707 | { | |
62a6e103 | 4708 | if (c != RE_STRING_CHAR (p1 + 2, multibyte)) |
4e8a9132 | 4709 | { |
dc4a2ee0 | 4710 | DEBUG_PRINT (" '%c' != '%c' => fast loop.\n", c, p1[2]); |
4e8a9132 SM |
4711 | return 1; |
4712 | } | |
4713 | } | |
505bde11 | 4714 | |
4e8a9132 SM |
4715 | else if ((re_opcode_t) *p1 == charset |
4716 | || (re_opcode_t) *p1 == charset_not) | |
4717 | { | |
4718 | int not = (re_opcode_t) *p1 == charset_not; | |
505bde11 | 4719 | |
4e8a9132 SM |
4720 | /* Test if C is listed in charset (or charset_not) |
4721 | at `p1'. */ | |
6fdd04b0 | 4722 | if (! multibyte || IS_REAL_ASCII (c)) |
4e8a9132 SM |
4723 | { |
4724 | if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH | |
4725 | && p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
4726 | not = !not; | |
4727 | } | |
4728 | else if (CHARSET_RANGE_TABLE_EXISTS_P (p1)) | |
4729 | CHARSET_LOOKUP_RANGE_TABLE (not, c, p1); | |
505bde11 | 4730 | |
4e8a9132 SM |
4731 | /* `not' is equal to 1 if c would match, which means |
4732 | that we can't change to pop_failure_jump. */ | |
4733 | if (!not) | |
4734 | { | |
dc4a2ee0 | 4735 | DEBUG_PRINT (" No match => fast loop.\n"); |
4e8a9132 SM |
4736 | return 1; |
4737 | } | |
4738 | } | |
4739 | else if ((re_opcode_t) *p1 == anychar | |
4740 | && c == '\n') | |
4741 | { | |
dc4a2ee0 | 4742 | DEBUG_PRINT (" . != \\n => fast loop.\n"); |
4e8a9132 SM |
4743 | return 1; |
4744 | } | |
4745 | } | |
4746 | break; | |
505bde11 | 4747 | |
4e8a9132 | 4748 | case charset: |
4e8a9132 SM |
4749 | { |
4750 | if ((re_opcode_t) *p1 == exactn) | |
4751 | /* Reuse the code above. */ | |
4752 | return mutually_exclusive_p (bufp, p2, p1); | |
505bde11 | 4753 | |
505bde11 SM |
4754 | /* It is hard to list up all the character in charset |
4755 | P2 if it includes multibyte character. Give up in | |
4756 | such case. */ | |
4757 | else if (!multibyte || !CHARSET_RANGE_TABLE_EXISTS_P (p2)) | |
4758 | { | |
4759 | /* Now, we are sure that P2 has no range table. | |
4760 | So, for the size of bitmap in P2, `p2[1]' is | |
7814e705 | 4761 | enough. But P1 may have range table, so the |
505bde11 SM |
4762 | size of bitmap table of P1 is extracted by |
4763 | using macro `CHARSET_BITMAP_SIZE'. | |
4764 | ||
6fdd04b0 KH |
4765 | In a multibyte case, we know that all the character |
4766 | listed in P2 is ASCII. In a unibyte case, P1 has only a | |
4767 | bitmap table. So, in both cases, it is enough to test | |
4768 | only the bitmap table of P1. */ | |
505bde11 | 4769 | |
411e4203 | 4770 | if ((re_opcode_t) *p1 == charset) |
505bde11 SM |
4771 | { |
4772 | int idx; | |
4773 | /* We win if the charset inside the loop | |
4774 | has no overlap with the one after the loop. */ | |
4775 | for (idx = 0; | |
4776 | (idx < (int) p2[1] | |
4777 | && idx < CHARSET_BITMAP_SIZE (p1)); | |
4778 | idx++) | |
4779 | if ((p2[2 + idx] & p1[2 + idx]) != 0) | |
4780 | break; | |
4781 | ||
4782 | if (idx == p2[1] | |
4783 | || idx == CHARSET_BITMAP_SIZE (p1)) | |
4784 | { | |
dc4a2ee0 | 4785 | DEBUG_PRINT (" No match => fast loop.\n"); |
505bde11 SM |
4786 | return 1; |
4787 | } | |
4788 | } | |
411e4203 | 4789 | else if ((re_opcode_t) *p1 == charset_not) |
505bde11 SM |
4790 | { |
4791 | int idx; | |
4792 | /* We win if the charset_not inside the loop lists | |
7814e705 | 4793 | every character listed in the charset after. */ |
505bde11 SM |
4794 | for (idx = 0; idx < (int) p2[1]; idx++) |
4795 | if (! (p2[2 + idx] == 0 | |
4796 | || (idx < CHARSET_BITMAP_SIZE (p1) | |
4797 | && ((p2[2 + idx] & ~ p1[2 + idx]) == 0)))) | |
4798 | break; | |
4799 | ||
d1dfb56c EZ |
4800 | if (idx == p2[1]) |
4801 | { | |
dc4a2ee0 | 4802 | DEBUG_PRINT (" No match => fast loop.\n"); |
d1dfb56c EZ |
4803 | return 1; |
4804 | } | |
4e8a9132 SM |
4805 | } |
4806 | } | |
4807 | } | |
609b757a | 4808 | break; |
177c0ea7 | 4809 | |
411e4203 | 4810 | case charset_not: |
7393bcbb | 4811 | switch (*p1) |
411e4203 SM |
4812 | { |
4813 | case exactn: | |
4814 | case charset: | |
4815 | /* Reuse the code above. */ | |
4816 | return mutually_exclusive_p (bufp, p2, p1); | |
4817 | case charset_not: | |
4818 | /* When we have two charset_not, it's very unlikely that | |
4819 | they don't overlap. The union of the two sets of excluded | |
4820 | chars should cover all possible chars, which, as a matter of | |
4821 | fact, is virtually impossible in multibyte buffers. */ | |
36595814 | 4822 | break; |
411e4203 SM |
4823 | } |
4824 | break; | |
4825 | ||
4e8a9132 | 4826 | case wordend: |
669fa600 SM |
4827 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == Sword); |
4828 | case symend: | |
4e8a9132 | 4829 | return ((re_opcode_t) *p1 == syntaxspec |
669fa600 SM |
4830 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4831 | case notsyntaxspec: | |
4832 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4833 | |
4834 | case wordbeg: | |
669fa600 SM |
4835 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == Sword); |
4836 | case symbeg: | |
4e8a9132 | 4837 | return ((re_opcode_t) *p1 == notsyntaxspec |
669fa600 SM |
4838 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4839 | case syntaxspec: | |
4840 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4841 | |
4842 | case wordbound: | |
4843 | return (((re_opcode_t) *p1 == notsyntaxspec | |
4844 | || (re_opcode_t) *p1 == syntaxspec) | |
4845 | && p1[1] == Sword); | |
4846 | ||
1fb352e0 | 4847 | #ifdef emacs |
4e8a9132 SM |
4848 | case categoryspec: |
4849 | return ((re_opcode_t) *p1 == notcategoryspec && p1[1] == p2[1]); | |
4850 | case notcategoryspec: | |
4851 | return ((re_opcode_t) *p1 == categoryspec && p1[1] == p2[1]); | |
4852 | #endif /* emacs */ | |
4853 | ||
4854 | default: | |
4855 | ; | |
505bde11 SM |
4856 | } |
4857 | ||
4858 | /* Safe default. */ | |
4859 | return 0; | |
4860 | } | |
4861 | ||
fa9a63c5 RM |
4862 | \f |
4863 | /* Matching routines. */ | |
4864 | ||
25fe55af | 4865 | #ifndef emacs /* Emacs never uses this. */ |
fa9a63c5 RM |
4866 | /* re_match is like re_match_2 except it takes only a single string. */ |
4867 | ||
d1dfb56c | 4868 | regoff_t |
d2762c86 | 4869 | re_match (struct re_pattern_buffer *bufp, const char *string, |
d1dfb56c | 4870 | size_t size, ssize_t pos, struct re_registers *regs) |
fa9a63c5 | 4871 | { |
d1dfb56c EZ |
4872 | regoff_t result = re_match_2_internal (bufp, NULL, 0, (re_char*) string, |
4873 | size, pos, regs, size); | |
fa9a63c5 RM |
4874 | return result; |
4875 | } | |
c0f9ea08 | 4876 | WEAK_ALIAS (__re_match, re_match) |
fa9a63c5 RM |
4877 | #endif /* not emacs */ |
4878 | ||
b18215fc RS |
4879 | #ifdef emacs |
4880 | /* In Emacs, this is the string or buffer in which we | |
7814e705 | 4881 | are matching. It is used for looking up syntax properties. */ |
b18215fc RS |
4882 | Lisp_Object re_match_object; |
4883 | #endif | |
fa9a63c5 RM |
4884 | |
4885 | /* re_match_2 matches the compiled pattern in BUFP against the | |
4886 | the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 | |
4887 | and SIZE2, respectively). We start matching at POS, and stop | |
4888 | matching at STOP. | |
5e69f11e | 4889 | |
fa9a63c5 | 4890 | If REGS is non-null and the `no_sub' field of BUFP is nonzero, we |
7814e705 | 4891 | store offsets for the substring each group matched in REGS. See the |
fa9a63c5 RM |
4892 | documentation for exactly how many groups we fill. |
4893 | ||
4894 | We return -1 if no match, -2 if an internal error (such as the | |
7814e705 | 4895 | failure stack overflowing). Otherwise, we return the length of the |
fa9a63c5 RM |
4896 | matched substring. */ |
4897 | ||
d1dfb56c EZ |
4898 | regoff_t |
4899 | re_match_2 (struct re_pattern_buffer *bufp, const char *string1, | |
4900 | size_t size1, const char *string2, size_t size2, ssize_t pos, | |
4901 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4902 | { |
d1dfb56c | 4903 | regoff_t result; |
25fe55af | 4904 | |
b18215fc | 4905 | #ifdef emacs |
d1dfb56c | 4906 | ssize_t charpos; |
d48cd3f4 | 4907 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
99633e97 | 4908 | charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (pos)); |
cc9b4df2 | 4909 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); |
b18215fc RS |
4910 | #endif |
4911 | ||
4bb91c68 SM |
4912 | result = re_match_2_internal (bufp, (re_char*) string1, size1, |
4913 | (re_char*) string2, size2, | |
cc9b4df2 | 4914 | pos, regs, stop); |
fa9a63c5 RM |
4915 | return result; |
4916 | } | |
c0f9ea08 | 4917 | WEAK_ALIAS (__re_match_2, re_match_2) |
fa9a63c5 | 4918 | |
bf216479 | 4919 | |
fa9a63c5 | 4920 | /* This is a separate function so that we can force an alloca cleanup |
7814e705 | 4921 | afterwards. */ |
d1dfb56c | 4922 | static regoff_t |
29abe551 PE |
4923 | re_match_2_internal (struct re_pattern_buffer *bufp, const_re_char *string1, |
4924 | size_t size1, const_re_char *string2, size_t size2, | |
d1dfb56c | 4925 | ssize_t pos, struct re_registers *regs, ssize_t stop) |
fa9a63c5 RM |
4926 | { |
4927 | /* General temporaries. */ | |
dc4a2ee0 | 4928 | int mcnt; |
01618498 | 4929 | size_t reg; |
fa9a63c5 RM |
4930 | |
4931 | /* Just past the end of the corresponding string. */ | |
66f0296e | 4932 | re_char *end1, *end2; |
fa9a63c5 RM |
4933 | |
4934 | /* Pointers into string1 and string2, just past the last characters in | |
7814e705 | 4935 | each to consider matching. */ |
66f0296e | 4936 | re_char *end_match_1, *end_match_2; |
fa9a63c5 RM |
4937 | |
4938 | /* Where we are in the data, and the end of the current string. */ | |
66f0296e | 4939 | re_char *d, *dend; |
5e69f11e | 4940 | |
99633e97 SM |
4941 | /* Used sometimes to remember where we were before starting matching |
4942 | an operator so that we can go back in case of failure. This "atomic" | |
4943 | behavior of matching opcodes is indispensable to the correctness | |
4944 | of the on_failure_keep_string_jump optimization. */ | |
4945 | re_char *dfail; | |
4946 | ||
fa9a63c5 | 4947 | /* Where we are in the pattern, and the end of the pattern. */ |
01618498 SM |
4948 | re_char *p = bufp->buffer; |
4949 | re_char *pend = p + bufp->used; | |
fa9a63c5 | 4950 | |
25fe55af | 4951 | /* We use this to map every character in the string. */ |
6676cb1c | 4952 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 | 4953 | |
cf9c99bc | 4954 | /* Nonzero if BUFP is setup from a multibyte regex. */ |
2d1675e4 | 4955 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
b18215fc | 4956 | |
cf9c99bc KH |
4957 | /* Nonzero if STRING1/STRING2 are multibyte. */ |
4958 | const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
4959 | ||
fa9a63c5 RM |
4960 | /* Failure point stack. Each place that can handle a failure further |
4961 | down the line pushes a failure point on this stack. It consists of | |
505bde11 | 4962 | regstart, and regend for all registers corresponding to |
fa9a63c5 RM |
4963 | the subexpressions we're currently inside, plus the number of such |
4964 | registers, and, finally, two char *'s. The first char * is where | |
4965 | to resume scanning the pattern; the second one is where to resume | |
7814e705 JB |
4966 | scanning the strings. */ |
4967 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | |
fa9a63c5 RM |
4968 | fail_stack_type fail_stack; |
4969 | #endif | |
dc4a2ee0 | 4970 | #ifdef DEBUG_COMPILES_ARGUMENTS |
fa9a63c5 RM |
4971 | unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; |
4972 | #endif | |
4973 | ||
0b32bf0e | 4974 | #if defined REL_ALLOC && defined REGEX_MALLOC |
fa9a63c5 RM |
4975 | /* This holds the pointer to the failure stack, when |
4976 | it is allocated relocatably. */ | |
4977 | fail_stack_elt_t *failure_stack_ptr; | |
99633e97 | 4978 | #endif |
fa9a63c5 RM |
4979 | |
4980 | /* We fill all the registers internally, independent of what we | |
7814e705 | 4981 | return, for use in backreferences. The number here includes |
fa9a63c5 | 4982 | an element for register zero. */ |
4bb91c68 | 4983 | size_t num_regs = bufp->re_nsub + 1; |
5e69f11e | 4984 | |
fa9a63c5 RM |
4985 | /* Information on the contents of registers. These are pointers into |
4986 | the input strings; they record just what was matched (on this | |
4987 | attempt) by a subexpression part of the pattern, that is, the | |
4988 | regnum-th regstart pointer points to where in the pattern we began | |
4989 | matching and the regnum-th regend points to right after where we | |
4990 | stopped matching the regnum-th subexpression. (The zeroth register | |
4991 | keeps track of what the whole pattern matches.) */ | |
4992 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 4993 | re_char **regstart, **regend; |
fa9a63c5 RM |
4994 | #endif |
4995 | ||
fa9a63c5 | 4996 | /* The following record the register info as found in the above |
5e69f11e | 4997 | variables when we find a match better than any we've seen before. |
fa9a63c5 RM |
4998 | This happens as we backtrack through the failure points, which in |
4999 | turn happens only if we have not yet matched the entire string. */ | |
5000 | unsigned best_regs_set = false; | |
5001 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5002 | re_char **best_regstart, **best_regend; |
fa9a63c5 | 5003 | #endif |
5e69f11e | 5004 | |
fa9a63c5 RM |
5005 | /* Logically, this is `best_regend[0]'. But we don't want to have to |
5006 | allocate space for that if we're not allocating space for anything | |
7814e705 | 5007 | else (see below). Also, we never need info about register 0 for |
fa9a63c5 RM |
5008 | any of the other register vectors, and it seems rather a kludge to |
5009 | treat `best_regend' differently than the rest. So we keep track of | |
5010 | the end of the best match so far in a separate variable. We | |
5011 | initialize this to NULL so that when we backtrack the first time | |
5012 | and need to test it, it's not garbage. */ | |
66f0296e | 5013 | re_char *match_end = NULL; |
fa9a63c5 | 5014 | |
dc4a2ee0 | 5015 | #ifdef DEBUG_COMPILES_ARGUMENTS |
fa9a63c5 | 5016 | /* Counts the total number of registers pushed. */ |
5e69f11e | 5017 | unsigned num_regs_pushed = 0; |
fa9a63c5 RM |
5018 | #endif |
5019 | ||
dc4a2ee0 | 5020 | DEBUG_PRINT ("\n\nEntering re_match_2.\n"); |
5e69f11e | 5021 | |
fa9a63c5 | 5022 | INIT_FAIL_STACK (); |
5e69f11e | 5023 | |
fa9a63c5 RM |
5024 | #ifdef MATCH_MAY_ALLOCATE |
5025 | /* Do not bother to initialize all the register variables if there are | |
5026 | no groups in the pattern, as it takes a fair amount of time. If | |
5027 | there are groups, we include space for register 0 (the whole | |
5028 | pattern), even though we never use it, since it simplifies the | |
5029 | array indexing. We should fix this. */ | |
5030 | if (bufp->re_nsub) | |
5031 | { | |
66f0296e SM |
5032 | regstart = REGEX_TALLOC (num_regs, re_char *); |
5033 | regend = REGEX_TALLOC (num_regs, re_char *); | |
5034 | best_regstart = REGEX_TALLOC (num_regs, re_char *); | |
5035 | best_regend = REGEX_TALLOC (num_regs, re_char *); | |
fa9a63c5 | 5036 | |
505bde11 | 5037 | if (!(regstart && regend && best_regstart && best_regend)) |
25fe55af RS |
5038 | { |
5039 | FREE_VARIABLES (); | |
5040 | return -2; | |
5041 | } | |
fa9a63c5 RM |
5042 | } |
5043 | else | |
5044 | { | |
5045 | /* We must initialize all our variables to NULL, so that | |
25fe55af | 5046 | `FREE_VARIABLES' doesn't try to free them. */ |
505bde11 | 5047 | regstart = regend = best_regstart = best_regend = NULL; |
fa9a63c5 RM |
5048 | } |
5049 | #endif /* MATCH_MAY_ALLOCATE */ | |
5050 | ||
5051 | /* The starting position is bogus. */ | |
5052 | if (pos < 0 || pos > size1 + size2) | |
5053 | { | |
5054 | FREE_VARIABLES (); | |
5055 | return -1; | |
5056 | } | |
5e69f11e | 5057 | |
fa9a63c5 RM |
5058 | /* Initialize subexpression text positions to -1 to mark ones that no |
5059 | start_memory/stop_memory has been seen for. Also initialize the | |
5060 | register information struct. */ | |
01618498 SM |
5061 | for (reg = 1; reg < num_regs; reg++) |
5062 | regstart[reg] = regend[reg] = NULL; | |
99633e97 | 5063 | |
fa9a63c5 | 5064 | /* We move `string1' into `string2' if the latter's empty -- but not if |
7814e705 | 5065 | `string1' is null. */ |
fa9a63c5 RM |
5066 | if (size2 == 0 && string1 != NULL) |
5067 | { | |
5068 | string2 = string1; | |
5069 | size2 = size1; | |
5070 | string1 = 0; | |
5071 | size1 = 0; | |
5072 | } | |
5073 | end1 = string1 + size1; | |
5074 | end2 = string2 + size2; | |
5075 | ||
5e69f11e | 5076 | /* `p' scans through the pattern as `d' scans through the data. |
fa9a63c5 RM |
5077 | `dend' is the end of the input string that `d' points within. `d' |
5078 | is advanced into the following input string whenever necessary, but | |
5079 | this happens before fetching; therefore, at the beginning of the | |
5080 | loop, `d' can be pointing at the end of a string, but it cannot | |
5081 | equal `string2'. */ | |
419d1c74 | 5082 | if (pos >= size1) |
fa9a63c5 | 5083 | { |
419d1c74 SM |
5084 | /* Only match within string2. */ |
5085 | d = string2 + pos - size1; | |
5086 | dend = end_match_2 = string2 + stop - size1; | |
5087 | end_match_1 = end1; /* Just to give it a value. */ | |
fa9a63c5 RM |
5088 | } |
5089 | else | |
5090 | { | |
f1ad044f | 5091 | if (stop < size1) |
419d1c74 SM |
5092 | { |
5093 | /* Only match within string1. */ | |
5094 | end_match_1 = string1 + stop; | |
5095 | /* BEWARE! | |
5096 | When we reach end_match_1, PREFETCH normally switches to string2. | |
5097 | But in the present case, this means that just doing a PREFETCH | |
5098 | makes us jump from `stop' to `gap' within the string. | |
5099 | What we really want here is for the search to stop as | |
5100 | soon as we hit end_match_1. That's why we set end_match_2 | |
5101 | to end_match_1 (since PREFETCH fails as soon as we hit | |
5102 | end_match_2). */ | |
5103 | end_match_2 = end_match_1; | |
5104 | } | |
5105 | else | |
f1ad044f SM |
5106 | { /* It's important to use this code when stop == size so that |
5107 | moving `d' from end1 to string2 will not prevent the d == dend | |
5108 | check from catching the end of string. */ | |
419d1c74 SM |
5109 | end_match_1 = end1; |
5110 | end_match_2 = string2 + stop - size1; | |
5111 | } | |
5112 | d = string1 + pos; | |
5113 | dend = end_match_1; | |
fa9a63c5 RM |
5114 | } |
5115 | ||
dc4a2ee0 | 5116 | DEBUG_PRINT ("The compiled pattern is: "); |
fa9a63c5 | 5117 | DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); |
dc4a2ee0 | 5118 | DEBUG_PRINT ("The string to match is: `"); |
fa9a63c5 | 5119 | DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); |
dc4a2ee0 | 5120 | DEBUG_PRINT ("'\n"); |
5e69f11e | 5121 | |
7814e705 | 5122 | /* This loops over pattern commands. It exits by returning from the |
fa9a63c5 RM |
5123 | function if the match is complete, or it drops through if the match |
5124 | fails at this starting point in the input data. */ | |
5125 | for (;;) | |
5126 | { | |
dc4a2ee0 | 5127 | DEBUG_PRINT ("\n%p: ", p); |
fa9a63c5 RM |
5128 | |
5129 | if (p == pend) | |
dc4a2ee0 PE |
5130 | { |
5131 | ptrdiff_t dcnt; | |
5132 | ||
5133 | /* End of pattern means we might have succeeded. */ | |
5134 | DEBUG_PRINT ("end of pattern ... "); | |
5e69f11e | 5135 | |
fa9a63c5 | 5136 | /* If we haven't matched the entire string, and we want the |
25fe55af RS |
5137 | longest match, try backtracking. */ |
5138 | if (d != end_match_2) | |
fa9a63c5 RM |
5139 | { |
5140 | /* 1 if this match ends in the same string (string1 or string2) | |
5141 | as the best previous match. */ | |
d42f4f0f PE |
5142 | boolean same_str_p = (FIRST_STRING_P (match_end) |
5143 | == FIRST_STRING_P (d)); | |
fa9a63c5 RM |
5144 | /* 1 if this match is the best seen so far. */ |
5145 | boolean best_match_p; | |
5146 | ||
5147 | /* AIX compiler got confused when this was combined | |
7814e705 | 5148 | with the previous declaration. */ |
fa9a63c5 RM |
5149 | if (same_str_p) |
5150 | best_match_p = d > match_end; | |
5151 | else | |
99633e97 | 5152 | best_match_p = !FIRST_STRING_P (d); |
fa9a63c5 | 5153 | |
dc4a2ee0 | 5154 | DEBUG_PRINT ("backtracking.\n"); |
25fe55af RS |
5155 | |
5156 | if (!FAIL_STACK_EMPTY ()) | |
5157 | { /* More failure points to try. */ | |
5158 | ||
5159 | /* If exceeds best match so far, save it. */ | |
5160 | if (!best_regs_set || best_match_p) | |
5161 | { | |
5162 | best_regs_set = true; | |
5163 | match_end = d; | |
5164 | ||
dc4a2ee0 | 5165 | DEBUG_PRINT ("\nSAVING match as best so far.\n"); |
25fe55af | 5166 | |
01618498 | 5167 | for (reg = 1; reg < num_regs; reg++) |
25fe55af | 5168 | { |
01618498 SM |
5169 | best_regstart[reg] = regstart[reg]; |
5170 | best_regend[reg] = regend[reg]; | |
25fe55af RS |
5171 | } |
5172 | } | |
5173 | goto fail; | |
5174 | } | |
5175 | ||
5176 | /* If no failure points, don't restore garbage. And if | |
5177 | last match is real best match, don't restore second | |
5178 | best one. */ | |
5179 | else if (best_regs_set && !best_match_p) | |
5180 | { | |
5181 | restore_best_regs: | |
5182 | /* Restore best match. It may happen that `dend == | |
5183 | end_match_1' while the restored d is in string2. | |
5184 | For example, the pattern `x.*y.*z' against the | |
5185 | strings `x-' and `y-z-', if the two strings are | |
7814e705 | 5186 | not consecutive in memory. */ |
dc4a2ee0 | 5187 | DEBUG_PRINT ("Restoring best registers.\n"); |
25fe55af RS |
5188 | |
5189 | d = match_end; | |
5190 | dend = ((d >= string1 && d <= end1) | |
5191 | ? end_match_1 : end_match_2); | |
fa9a63c5 | 5192 | |
01618498 | 5193 | for (reg = 1; reg < num_regs; reg++) |
fa9a63c5 | 5194 | { |
01618498 SM |
5195 | regstart[reg] = best_regstart[reg]; |
5196 | regend[reg] = best_regend[reg]; | |
fa9a63c5 | 5197 | } |
25fe55af RS |
5198 | } |
5199 | } /* d != end_match_2 */ | |
fa9a63c5 RM |
5200 | |
5201 | succeed_label: | |
dc4a2ee0 | 5202 | DEBUG_PRINT ("Accepting match.\n"); |
fa9a63c5 | 5203 | |
25fe55af RS |
5204 | /* If caller wants register contents data back, do it. */ |
5205 | if (regs && !bufp->no_sub) | |
fa9a63c5 | 5206 | { |
25fe55af RS |
5207 | /* Have the register data arrays been allocated? */ |
5208 | if (bufp->regs_allocated == REGS_UNALLOCATED) | |
7814e705 | 5209 | { /* No. So allocate them with malloc. We need one |
25fe55af RS |
5210 | extra element beyond `num_regs' for the `-1' marker |
5211 | GNU code uses. */ | |
5212 | regs->num_regs = MAX (RE_NREGS, num_regs + 1); | |
5213 | regs->start = TALLOC (regs->num_regs, regoff_t); | |
5214 | regs->end = TALLOC (regs->num_regs, regoff_t); | |
5215 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5216 | { |
5217 | FREE_VARIABLES (); | |
5218 | return -2; | |
5219 | } | |
25fe55af RS |
5220 | bufp->regs_allocated = REGS_REALLOCATE; |
5221 | } | |
5222 | else if (bufp->regs_allocated == REGS_REALLOCATE) | |
5223 | { /* Yes. If we need more elements than were already | |
5224 | allocated, reallocate them. If we need fewer, just | |
5225 | leave it alone. */ | |
5226 | if (regs->num_regs < num_regs + 1) | |
5227 | { | |
5228 | regs->num_regs = num_regs + 1; | |
5229 | RETALLOC (regs->start, regs->num_regs, regoff_t); | |
5230 | RETALLOC (regs->end, regs->num_regs, regoff_t); | |
5231 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5232 | { |
5233 | FREE_VARIABLES (); | |
5234 | return -2; | |
5235 | } | |
25fe55af RS |
5236 | } |
5237 | } | |
5238 | else | |
fa9a63c5 RM |
5239 | { |
5240 | /* These braces fend off a "empty body in an else-statement" | |
7814e705 | 5241 | warning under GCC when assert expands to nothing. */ |
fa9a63c5 RM |
5242 | assert (bufp->regs_allocated == REGS_FIXED); |
5243 | } | |
5244 | ||
25fe55af RS |
5245 | /* Convert the pointer data in `regstart' and `regend' to |
5246 | indices. Register zero has to be set differently, | |
5247 | since we haven't kept track of any info for it. */ | |
5248 | if (regs->num_regs > 0) | |
5249 | { | |
5250 | regs->start[0] = pos; | |
99633e97 | 5251 | regs->end[0] = POINTER_TO_OFFSET (d); |
25fe55af | 5252 | } |
5e69f11e | 5253 | |
25fe55af RS |
5254 | /* Go through the first `min (num_regs, regs->num_regs)' |
5255 | registers, since that is all we initialized. */ | |
01618498 | 5256 | for (reg = 1; reg < MIN (num_regs, regs->num_regs); reg++) |
fa9a63c5 | 5257 | { |
01618498 SM |
5258 | if (REG_UNSET (regstart[reg]) || REG_UNSET (regend[reg])) |
5259 | regs->start[reg] = regs->end[reg] = -1; | |
25fe55af RS |
5260 | else |
5261 | { | |
dc4a2ee0 PE |
5262 | regs->start[reg] = POINTER_TO_OFFSET (regstart[reg]); |
5263 | regs->end[reg] = POINTER_TO_OFFSET (regend[reg]); | |
25fe55af | 5264 | } |
fa9a63c5 | 5265 | } |
5e69f11e | 5266 | |
25fe55af RS |
5267 | /* If the regs structure we return has more elements than |
5268 | were in the pattern, set the extra elements to -1. If | |
5269 | we (re)allocated the registers, this is the case, | |
5270 | because we always allocate enough to have at least one | |
7814e705 | 5271 | -1 at the end. */ |
01618498 SM |
5272 | for (reg = num_regs; reg < regs->num_regs; reg++) |
5273 | regs->start[reg] = regs->end[reg] = -1; | |
fa9a63c5 RM |
5274 | } /* regs && !bufp->no_sub */ |
5275 | ||
dc4a2ee0 PE |
5276 | DEBUG_PRINT ("%u failure points pushed, %u popped (%u remain).\n", |
5277 | nfailure_points_pushed, nfailure_points_popped, | |
5278 | nfailure_points_pushed - nfailure_points_popped); | |
5279 | DEBUG_PRINT ("%u registers pushed.\n", num_regs_pushed); | |
fa9a63c5 | 5280 | |
dc4a2ee0 | 5281 | dcnt = POINTER_TO_OFFSET (d) - pos; |
fa9a63c5 | 5282 | |
dc4a2ee0 | 5283 | DEBUG_PRINT ("Returning %td from re_match_2.\n", dcnt); |
fa9a63c5 | 5284 | |
25fe55af | 5285 | FREE_VARIABLES (); |
dc4a2ee0 | 5286 | return dcnt; |
25fe55af | 5287 | } |
fa9a63c5 | 5288 | |
7814e705 | 5289 | /* Otherwise match next pattern command. */ |
7393bcbb | 5290 | switch (*p++) |
fa9a63c5 | 5291 | { |
25fe55af RS |
5292 | /* Ignore these. Used to ignore the n of succeed_n's which |
5293 | currently have n == 0. */ | |
5294 | case no_op: | |
dc4a2ee0 | 5295 | DEBUG_PRINT ("EXECUTING no_op.\n"); |
25fe55af | 5296 | break; |
fa9a63c5 RM |
5297 | |
5298 | case succeed: | |
dc4a2ee0 | 5299 | DEBUG_PRINT ("EXECUTING succeed.\n"); |
fa9a63c5 RM |
5300 | goto succeed_label; |
5301 | ||
7814e705 | 5302 | /* Match the next n pattern characters exactly. The following |
25fe55af | 5303 | byte in the pattern defines n, and the n bytes after that |
7814e705 | 5304 | are the characters to match. */ |
fa9a63c5 RM |
5305 | case exactn: |
5306 | mcnt = *p++; | |
dc4a2ee0 | 5307 | DEBUG_PRINT ("EXECUTING exactn %d.\n", mcnt); |
fa9a63c5 | 5308 | |
99633e97 SM |
5309 | /* Remember the start point to rollback upon failure. */ |
5310 | dfail = d; | |
5311 | ||
6fdd04b0 | 5312 | #ifndef emacs |
25fe55af RS |
5313 | /* This is written out as an if-else so we don't waste time |
5314 | testing `translate' inside the loop. */ | |
28703c16 | 5315 | if (RE_TRANSLATE_P (translate)) |
6fdd04b0 KH |
5316 | do |
5317 | { | |
5318 | PREFETCH (); | |
5319 | if (RE_TRANSLATE (translate, *d) != *p++) | |
e934739e | 5320 | { |
6fdd04b0 KH |
5321 | d = dfail; |
5322 | goto fail; | |
e934739e | 5323 | } |
6fdd04b0 KH |
5324 | d++; |
5325 | } | |
5326 | while (--mcnt); | |
fa9a63c5 | 5327 | else |
6fdd04b0 KH |
5328 | do |
5329 | { | |
5330 | PREFETCH (); | |
5331 | if (*d++ != *p++) | |
bf216479 | 5332 | { |
6fdd04b0 KH |
5333 | d = dfail; |
5334 | goto fail; | |
bf216479 | 5335 | } |
6fdd04b0 KH |
5336 | } |
5337 | while (--mcnt); | |
5338 | #else /* emacs */ | |
5339 | /* The cost of testing `translate' is comparatively small. */ | |
cf9c99bc | 5340 | if (target_multibyte) |
6fdd04b0 KH |
5341 | do |
5342 | { | |
5343 | int pat_charlen, buf_charlen; | |
cf9c99bc | 5344 | int pat_ch, buf_ch; |
e934739e | 5345 | |
6fdd04b0 | 5346 | PREFETCH (); |
cf9c99bc | 5347 | if (multibyte) |
62a6e103 | 5348 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
cf9c99bc KH |
5349 | else |
5350 | { | |
5351 | pat_ch = RE_CHAR_TO_MULTIBYTE (*p); | |
5352 | pat_charlen = 1; | |
5353 | } | |
62a6e103 | 5354 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 5355 | |
6fdd04b0 | 5356 | if (TRANSLATE (buf_ch) != pat_ch) |
e934739e | 5357 | { |
6fdd04b0 KH |
5358 | d = dfail; |
5359 | goto fail; | |
e934739e | 5360 | } |
bf216479 | 5361 | |
6fdd04b0 KH |
5362 | p += pat_charlen; |
5363 | d += buf_charlen; | |
5364 | mcnt -= pat_charlen; | |
5365 | } | |
5366 | while (mcnt > 0); | |
fa9a63c5 | 5367 | else |
6fdd04b0 KH |
5368 | do |
5369 | { | |
abbd1bcf | 5370 | int pat_charlen; |
cf9c99bc | 5371 | int pat_ch, buf_ch; |
bf216479 | 5372 | |
6fdd04b0 | 5373 | PREFETCH (); |
cf9c99bc KH |
5374 | if (multibyte) |
5375 | { | |
62a6e103 | 5376 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
2afc21f5 | 5377 | pat_ch = RE_CHAR_TO_UNIBYTE (pat_ch); |
cf9c99bc KH |
5378 | } |
5379 | else | |
5380 | { | |
5381 | pat_ch = *p; | |
5382 | pat_charlen = 1; | |
5383 | } | |
5384 | buf_ch = RE_CHAR_TO_MULTIBYTE (*d); | |
5385 | if (! CHAR_BYTE8_P (buf_ch)) | |
5386 | { | |
5387 | buf_ch = TRANSLATE (buf_ch); | |
5388 | buf_ch = RE_CHAR_TO_UNIBYTE (buf_ch); | |
5389 | if (buf_ch < 0) | |
5390 | buf_ch = *d; | |
5391 | } | |
0e2501ed AS |
5392 | else |
5393 | buf_ch = *d; | |
cf9c99bc | 5394 | if (buf_ch != pat_ch) |
6fdd04b0 KH |
5395 | { |
5396 | d = dfail; | |
5397 | goto fail; | |
bf216479 | 5398 | } |
cf9c99bc KH |
5399 | p += pat_charlen; |
5400 | d++; | |
6fdd04b0 KH |
5401 | } |
5402 | while (--mcnt); | |
5403 | #endif | |
25fe55af | 5404 | break; |
fa9a63c5 RM |
5405 | |
5406 | ||
25fe55af | 5407 | /* Match any character except possibly a newline or a null. */ |
fa9a63c5 | 5408 | case anychar: |
e934739e RS |
5409 | { |
5410 | int buf_charlen; | |
01618498 | 5411 | re_wchar_t buf_ch; |
fa9a63c5 | 5412 | |
dc4a2ee0 | 5413 | DEBUG_PRINT ("EXECUTING anychar.\n"); |
fa9a63c5 | 5414 | |
e934739e | 5415 | PREFETCH (); |
62a6e103 | 5416 | buf_ch = RE_STRING_CHAR_AND_LENGTH (d, buf_charlen, |
cf9c99bc | 5417 | target_multibyte); |
e934739e RS |
5418 | buf_ch = TRANSLATE (buf_ch); |
5419 | ||
5420 | if ((!(bufp->syntax & RE_DOT_NEWLINE) | |
5421 | && buf_ch == '\n') | |
5422 | || ((bufp->syntax & RE_DOT_NOT_NULL) | |
5423 | && buf_ch == '\000')) | |
5424 | goto fail; | |
5425 | ||
dc4a2ee0 | 5426 | DEBUG_PRINT (" Matched `%d'.\n", *d); |
e934739e RS |
5427 | d += buf_charlen; |
5428 | } | |
fa9a63c5 RM |
5429 | break; |
5430 | ||
5431 | ||
5432 | case charset: | |
5433 | case charset_not: | |
5434 | { | |
b18215fc | 5435 | register unsigned int c; |
fa9a63c5 | 5436 | boolean not = (re_opcode_t) *(p - 1) == charset_not; |
b18215fc RS |
5437 | int len; |
5438 | ||
5439 | /* Start of actual range_table, or end of bitmap if there is no | |
5440 | range table. */ | |
da053e48 | 5441 | re_char *range_table IF_LINT (= NULL); |
b18215fc | 5442 | |
96cc36cc | 5443 | /* Nonzero if there is a range table. */ |
b18215fc RS |
5444 | int range_table_exists; |
5445 | ||
96cc36cc RS |
5446 | /* Number of ranges of range table. This is not included |
5447 | in the initial byte-length of the command. */ | |
5448 | int count = 0; | |
fa9a63c5 | 5449 | |
f5020181 AS |
5450 | /* Whether matching against a unibyte character. */ |
5451 | boolean unibyte_char = false; | |
5452 | ||
dc4a2ee0 | 5453 | DEBUG_PRINT ("EXECUTING charset%s.\n", not ? "_not" : ""); |
fa9a63c5 | 5454 | |
b18215fc | 5455 | range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]); |
96cc36cc | 5456 | |
b18215fc | 5457 | if (range_table_exists) |
96cc36cc RS |
5458 | { |
5459 | range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap. */ | |
5460 | EXTRACT_NUMBER_AND_INCR (count, range_table); | |
5461 | } | |
b18215fc | 5462 | |
2d1675e4 | 5463 | PREFETCH (); |
62a6e103 | 5464 | c = RE_STRING_CHAR_AND_LENGTH (d, len, target_multibyte); |
cf9c99bc KH |
5465 | if (target_multibyte) |
5466 | { | |
5467 | int c1; | |
b18215fc | 5468 | |
cf9c99bc KH |
5469 | c = TRANSLATE (c); |
5470 | c1 = RE_CHAR_TO_UNIBYTE (c); | |
5471 | if (c1 >= 0) | |
f5020181 AS |
5472 | { |
5473 | unibyte_char = true; | |
5474 | c = c1; | |
5475 | } | |
cf9c99bc KH |
5476 | } |
5477 | else | |
5478 | { | |
5479 | int c1 = RE_CHAR_TO_MULTIBYTE (c); | |
5480 | ||
5481 | if (! CHAR_BYTE8_P (c1)) | |
5482 | { | |
5483 | c1 = TRANSLATE (c1); | |
5484 | c1 = RE_CHAR_TO_UNIBYTE (c1); | |
5485 | if (c1 >= 0) | |
f5020181 AS |
5486 | { |
5487 | unibyte_char = true; | |
5488 | c = c1; | |
5489 | } | |
cf9c99bc | 5490 | } |
0b8be006 AS |
5491 | else |
5492 | unibyte_char = true; | |
cf9c99bc KH |
5493 | } |
5494 | ||
f5020181 | 5495 | if (unibyte_char && c < (1 << BYTEWIDTH)) |
b18215fc | 5496 | { /* Lookup bitmap. */ |
b18215fc RS |
5497 | /* Cast to `unsigned' instead of `unsigned char' in |
5498 | case the bit list is a full 32 bytes long. */ | |
5499 | if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH) | |
96cc36cc RS |
5500 | && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) |
5501 | not = !not; | |
b18215fc | 5502 | } |
96cc36cc | 5503 | #ifdef emacs |
b18215fc | 5504 | else if (range_table_exists) |
96cc36cc RS |
5505 | { |
5506 | int class_bits = CHARSET_RANGE_TABLE_BITS (&p[-1]); | |
5507 | ||
14473664 SM |
5508 | if ( (class_bits & BIT_LOWER && ISLOWER (c)) |
5509 | | (class_bits & BIT_MULTIBYTE) | |
96cc36cc RS |
5510 | | (class_bits & BIT_PUNCT && ISPUNCT (c)) |
5511 | | (class_bits & BIT_SPACE && ISSPACE (c)) | |
5512 | | (class_bits & BIT_UPPER && ISUPPER (c)) | |
5513 | | (class_bits & BIT_WORD && ISWORD (c))) | |
5514 | not = !not; | |
5515 | else | |
5516 | CHARSET_LOOKUP_RANGE_TABLE_RAW (not, c, range_table, count); | |
5517 | } | |
5518 | #endif /* emacs */ | |
fa9a63c5 | 5519 | |
96cc36cc RS |
5520 | if (range_table_exists) |
5521 | p = CHARSET_RANGE_TABLE_END (range_table, count); | |
5522 | else | |
5523 | p += CHARSET_BITMAP_SIZE (&p[-1]) + 1; | |
fa9a63c5 RM |
5524 | |
5525 | if (!not) goto fail; | |
5e69f11e | 5526 | |
b18215fc | 5527 | d += len; |
fa9a63c5 | 5528 | } |
8fb31792 | 5529 | break; |
fa9a63c5 RM |
5530 | |
5531 | ||
25fe55af | 5532 | /* The beginning of a group is represented by start_memory. |
505bde11 | 5533 | The argument is the register number. The text |
25fe55af | 5534 | matched within the group is recorded (in the internal |
7814e705 | 5535 | registers data structure) under the register number. */ |
25fe55af | 5536 | case start_memory: |
dc4a2ee0 | 5537 | DEBUG_PRINT ("EXECUTING start_memory %d:\n", *p); |
505bde11 SM |
5538 | |
5539 | /* In case we need to undo this operation (via backtracking). */ | |
dc4a2ee0 | 5540 | PUSH_FAILURE_REG (*p); |
fa9a63c5 | 5541 | |
25fe55af | 5542 | regstart[*p] = d; |
4bb91c68 | 5543 | regend[*p] = NULL; /* probably unnecessary. -sm */ |
dc4a2ee0 | 5544 | DEBUG_PRINT (" regstart: %td\n", POINTER_TO_OFFSET (regstart[*p])); |
fa9a63c5 | 5545 | |
25fe55af | 5546 | /* Move past the register number and inner group count. */ |
505bde11 | 5547 | p += 1; |
25fe55af | 5548 | break; |
fa9a63c5 RM |
5549 | |
5550 | ||
25fe55af | 5551 | /* The stop_memory opcode represents the end of a group. Its |
505bde11 | 5552 | argument is the same as start_memory's: the register number. */ |
fa9a63c5 | 5553 | case stop_memory: |
dc4a2ee0 | 5554 | DEBUG_PRINT ("EXECUTING stop_memory %d:\n", *p); |
505bde11 SM |
5555 | |
5556 | assert (!REG_UNSET (regstart[*p])); | |
5557 | /* Strictly speaking, there should be code such as: | |
177c0ea7 | 5558 | |
0b32bf0e | 5559 | assert (REG_UNSET (regend[*p])); |
505bde11 SM |
5560 | PUSH_FAILURE_REGSTOP ((unsigned int)*p); |
5561 | ||
5562 | But the only info to be pushed is regend[*p] and it is known to | |
5563 | be UNSET, so there really isn't anything to push. | |
5564 | Not pushing anything, on the other hand deprives us from the | |
5565 | guarantee that regend[*p] is UNSET since undoing this operation | |
5566 | will not reset its value properly. This is not important since | |
5567 | the value will only be read on the next start_memory or at | |
5568 | the very end and both events can only happen if this stop_memory | |
5569 | is *not* undone. */ | |
fa9a63c5 | 5570 | |
25fe55af | 5571 | regend[*p] = d; |
dc4a2ee0 | 5572 | DEBUG_PRINT (" regend: %td\n", POINTER_TO_OFFSET (regend[*p])); |
fa9a63c5 | 5573 | |
25fe55af | 5574 | /* Move past the register number and the inner group count. */ |
505bde11 | 5575 | p += 1; |
25fe55af | 5576 | break; |
fa9a63c5 RM |
5577 | |
5578 | ||
5579 | /* \<digit> has been turned into a `duplicate' command which is | |
25fe55af RS |
5580 | followed by the numeric value of <digit> as the register number. */ |
5581 | case duplicate: | |
fa9a63c5 | 5582 | { |
66f0296e | 5583 | register re_char *d2, *dend2; |
7814e705 | 5584 | int regno = *p++; /* Get which register to match against. */ |
dc4a2ee0 | 5585 | DEBUG_PRINT ("EXECUTING duplicate %d.\n", regno); |
fa9a63c5 | 5586 | |
7814e705 | 5587 | /* Can't back reference a group which we've never matched. */ |
25fe55af RS |
5588 | if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) |
5589 | goto fail; | |
5e69f11e | 5590 | |
7814e705 | 5591 | /* Where in input to try to start matching. */ |
25fe55af | 5592 | d2 = regstart[regno]; |
5e69f11e | 5593 | |
99633e97 SM |
5594 | /* Remember the start point to rollback upon failure. */ |
5595 | dfail = d; | |
5596 | ||
25fe55af RS |
5597 | /* Where to stop matching; if both the place to start and |
5598 | the place to stop matching are in the same string, then | |
5599 | set to the place to stop, otherwise, for now have to use | |
5600 | the end of the first string. */ | |
fa9a63c5 | 5601 | |
25fe55af | 5602 | dend2 = ((FIRST_STRING_P (regstart[regno]) |
fa9a63c5 RM |
5603 | == FIRST_STRING_P (regend[regno])) |
5604 | ? regend[regno] : end_match_1); | |
5605 | for (;;) | |
5606 | { | |
dc4a2ee0 PE |
5607 | ptrdiff_t dcnt; |
5608 | ||
fa9a63c5 | 5609 | /* If necessary, advance to next segment in register |
25fe55af | 5610 | contents. */ |
fa9a63c5 RM |
5611 | while (d2 == dend2) |
5612 | { | |
5613 | if (dend2 == end_match_2) break; | |
5614 | if (dend2 == regend[regno]) break; | |
5615 | ||
25fe55af RS |
5616 | /* End of string1 => advance to string2. */ |
5617 | d2 = string2; | |
5618 | dend2 = regend[regno]; | |
fa9a63c5 RM |
5619 | } |
5620 | /* At end of register contents => success */ | |
5621 | if (d2 == dend2) break; | |
5622 | ||
5623 | /* If necessary, advance to next segment in data. */ | |
5624 | PREFETCH (); | |
5625 | ||
5626 | /* How many characters left in this segment to match. */ | |
dc4a2ee0 | 5627 | dcnt = dend - d; |
5e69f11e | 5628 | |
fa9a63c5 | 5629 | /* Want how many consecutive characters we can match in |
25fe55af | 5630 | one shot, so, if necessary, adjust the count. */ |
dc4a2ee0 PE |
5631 | if (dcnt > dend2 - d2) |
5632 | dcnt = dend2 - d2; | |
5e69f11e | 5633 | |
fa9a63c5 | 5634 | /* Compare that many; failure if mismatch, else move |
25fe55af | 5635 | past them. */ |
28703c16 | 5636 | if (RE_TRANSLATE_P (translate) |
dc4a2ee0 PE |
5637 | ? bcmp_translate (d, d2, dcnt, translate, target_multibyte) |
5638 | : memcmp (d, d2, dcnt)) | |
99633e97 SM |
5639 | { |
5640 | d = dfail; | |
5641 | goto fail; | |
5642 | } | |
dc4a2ee0 | 5643 | d += dcnt, d2 += dcnt; |
fa9a63c5 RM |
5644 | } |
5645 | } | |
5646 | break; | |
5647 | ||
5648 | ||
25fe55af | 5649 | /* begline matches the empty string at the beginning of the string |
c0f9ea08 | 5650 | (unless `not_bol' is set in `bufp'), and after newlines. */ |
fa9a63c5 | 5651 | case begline: |
dc4a2ee0 | 5652 | DEBUG_PRINT ("EXECUTING begline.\n"); |
5e69f11e | 5653 | |
25fe55af RS |
5654 | if (AT_STRINGS_BEG (d)) |
5655 | { | |
5656 | if (!bufp->not_bol) break; | |
5657 | } | |
419d1c74 | 5658 | else |
25fe55af | 5659 | { |
bf216479 | 5660 | unsigned c; |
419d1c74 | 5661 | GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2); |
c0f9ea08 | 5662 | if (c == '\n') |
419d1c74 | 5663 | break; |
25fe55af RS |
5664 | } |
5665 | /* In all other cases, we fail. */ | |
5666 | goto fail; | |
fa9a63c5 RM |
5667 | |
5668 | ||
25fe55af | 5669 | /* endline is the dual of begline. */ |
fa9a63c5 | 5670 | case endline: |
dc4a2ee0 | 5671 | DEBUG_PRINT ("EXECUTING endline.\n"); |
fa9a63c5 | 5672 | |
25fe55af RS |
5673 | if (AT_STRINGS_END (d)) |
5674 | { | |
5675 | if (!bufp->not_eol) break; | |
5676 | } | |
f1ad044f | 5677 | else |
25fe55af | 5678 | { |
f1ad044f | 5679 | PREFETCH_NOLIMIT (); |
c0f9ea08 | 5680 | if (*d == '\n') |
f1ad044f | 5681 | break; |
25fe55af RS |
5682 | } |
5683 | goto fail; | |
fa9a63c5 RM |
5684 | |
5685 | ||
5686 | /* Match at the very beginning of the data. */ | |
25fe55af | 5687 | case begbuf: |
dc4a2ee0 | 5688 | DEBUG_PRINT ("EXECUTING begbuf.\n"); |
25fe55af RS |
5689 | if (AT_STRINGS_BEG (d)) |
5690 | break; | |
5691 | goto fail; | |
fa9a63c5 RM |
5692 | |
5693 | ||
5694 | /* Match at the very end of the data. */ | |
25fe55af | 5695 | case endbuf: |
dc4a2ee0 | 5696 | DEBUG_PRINT ("EXECUTING endbuf.\n"); |
fa9a63c5 RM |
5697 | if (AT_STRINGS_END (d)) |
5698 | break; | |
25fe55af | 5699 | goto fail; |
5e69f11e | 5700 | |
5e69f11e | 5701 | |
25fe55af RS |
5702 | /* on_failure_keep_string_jump is used to optimize `.*\n'. It |
5703 | pushes NULL as the value for the string on the stack. Then | |
505bde11 | 5704 | `POP_FAILURE_POINT' will keep the current value for the |
25fe55af | 5705 | string, instead of restoring it. To see why, consider |
7814e705 | 5706 | matching `foo\nbar' against `.*\n'. The .* matches the foo; |
25fe55af RS |
5707 | then the . fails against the \n. But the next thing we want |
5708 | to do is match the \n against the \n; if we restored the | |
5709 | string value, we would be back at the foo. | |
5710 | ||
5711 | Because this is used only in specific cases, we don't need to | |
5712 | check all the things that `on_failure_jump' does, to make | |
5713 | sure the right things get saved on the stack. Hence we don't | |
5714 | share its code. The only reason to push anything on the | |
5715 | stack at all is that otherwise we would have to change | |
5716 | `anychar's code to do something besides goto fail in this | |
5717 | case; that seems worse than this. */ | |
5718 | case on_failure_keep_string_jump: | |
505bde11 | 5719 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5720 | DEBUG_PRINT ("EXECUTING on_failure_keep_string_jump %d (to %p):\n", |
5721 | mcnt, p + mcnt); | |
fa9a63c5 | 5722 | |
505bde11 SM |
5723 | PUSH_FAILURE_POINT (p - 3, NULL); |
5724 | break; | |
5725 | ||
0683b6fa SM |
5726 | /* A nasty loop is introduced by the non-greedy *? and +?. |
5727 | With such loops, the stack only ever contains one failure point | |
5728 | at a time, so that a plain on_failure_jump_loop kind of | |
5729 | cycle detection cannot work. Worse yet, such a detection | |
5730 | can not only fail to detect a cycle, but it can also wrongly | |
5731 | detect a cycle (between different instantiations of the same | |
6df42991 | 5732 | loop). |
0683b6fa SM |
5733 | So the method used for those nasty loops is a little different: |
5734 | We use a special cycle-detection-stack-frame which is pushed | |
5735 | when the on_failure_jump_nastyloop failure-point is *popped*. | |
5736 | This special frame thus marks the beginning of one iteration | |
5737 | through the loop and we can hence easily check right here | |
5738 | whether something matched between the beginning and the end of | |
5739 | the loop. */ | |
5740 | case on_failure_jump_nastyloop: | |
5741 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
dc4a2ee0 PE |
5742 | DEBUG_PRINT ("EXECUTING on_failure_jump_nastyloop %d (to %p):\n", |
5743 | mcnt, p + mcnt); | |
0683b6fa SM |
5744 | |
5745 | assert ((re_opcode_t)p[-4] == no_op); | |
6df42991 SM |
5746 | { |
5747 | int cycle = 0; | |
5748 | CHECK_INFINITE_LOOP (p - 4, d); | |
5749 | if (!cycle) | |
5750 | /* If there's a cycle, just continue without pushing | |
5751 | this failure point. The failure point is the "try again" | |
5752 | option, which shouldn't be tried. | |
5753 | We want (x?)*?y\1z to match both xxyz and xxyxz. */ | |
5754 | PUSH_FAILURE_POINT (p - 3, d); | |
5755 | } | |
0683b6fa SM |
5756 | break; |
5757 | ||
4e8a9132 SM |
5758 | /* Simple loop detecting on_failure_jump: just check on the |
5759 | failure stack if the same spot was already hit earlier. */ | |
505bde11 SM |
5760 | case on_failure_jump_loop: |
5761 | on_failure: | |
5762 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
dc4a2ee0 PE |
5763 | DEBUG_PRINT ("EXECUTING on_failure_jump_loop %d (to %p):\n", |
5764 | mcnt, p + mcnt); | |
6df42991 SM |
5765 | { |
5766 | int cycle = 0; | |
5767 | CHECK_INFINITE_LOOP (p - 3, d); | |
5768 | if (cycle) | |
5769 | /* If there's a cycle, get out of the loop, as if the matching | |
5770 | had failed. We used to just `goto fail' here, but that was | |
5771 | aborting the search a bit too early: we want to keep the | |
5772 | empty-loop-match and keep matching after the loop. | |
5773 | We want (x?)*y\1z to match both xxyz and xxyxz. */ | |
5774 | p += mcnt; | |
5775 | else | |
5776 | PUSH_FAILURE_POINT (p - 3, d); | |
5777 | } | |
25fe55af | 5778 | break; |
fa9a63c5 RM |
5779 | |
5780 | ||
5781 | /* Uses of on_failure_jump: | |
5e69f11e | 5782 | |
25fe55af RS |
5783 | Each alternative starts with an on_failure_jump that points |
5784 | to the beginning of the next alternative. Each alternative | |
5785 | except the last ends with a jump that in effect jumps past | |
5786 | the rest of the alternatives. (They really jump to the | |
5787 | ending jump of the following alternative, because tensioning | |
5788 | these jumps is a hassle.) | |
fa9a63c5 | 5789 | |
25fe55af RS |
5790 | Repeats start with an on_failure_jump that points past both |
5791 | the repetition text and either the following jump or | |
5792 | pop_failure_jump back to this on_failure_jump. */ | |
fa9a63c5 | 5793 | case on_failure_jump: |
25fe55af | 5794 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5795 | DEBUG_PRINT ("EXECUTING on_failure_jump %d (to %p):\n", |
5796 | mcnt, p + mcnt); | |
25fe55af | 5797 | |
505bde11 | 5798 | PUSH_FAILURE_POINT (p -3, d); |
25fe55af RS |
5799 | break; |
5800 | ||
4e8a9132 | 5801 | /* This operation is used for greedy *. |
505bde11 SM |
5802 | Compare the beginning of the repeat with what in the |
5803 | pattern follows its end. If we can establish that there | |
5804 | is nothing that they would both match, i.e., that we | |
5805 | would have to backtrack because of (as in, e.g., `a*a') | |
5806 | then we can use a non-backtracking loop based on | |
4e8a9132 | 5807 | on_failure_keep_string_jump instead of on_failure_jump. */ |
505bde11 | 5808 | case on_failure_jump_smart: |
25fe55af | 5809 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5810 | DEBUG_PRINT ("EXECUTING on_failure_jump_smart %d (to %p).\n", |
5811 | mcnt, p + mcnt); | |
25fe55af | 5812 | { |
01618498 | 5813 | re_char *p1 = p; /* Next operation. */ |
6dcf2d0e SM |
5814 | /* Here, we discard `const', making re_match non-reentrant. */ |
5815 | unsigned char *p2 = (unsigned char*) p + mcnt; /* Jump dest. */ | |
5816 | unsigned char *p3 = (unsigned char*) p - 3; /* opcode location. */ | |
fa9a63c5 | 5817 | |
505bde11 SM |
5818 | p -= 3; /* Reset so that we will re-execute the |
5819 | instruction once it's been changed. */ | |
fa9a63c5 | 5820 | |
4e8a9132 SM |
5821 | EXTRACT_NUMBER (mcnt, p2 - 2); |
5822 | ||
5823 | /* Ensure this is a indeed the trivial kind of loop | |
5824 | we are expecting. */ | |
5825 | assert (skip_one_char (p1) == p2 - 3); | |
5826 | assert ((re_opcode_t) p2[-3] == jump && p2 + mcnt == p); | |
99633e97 | 5827 | DEBUG_STATEMENT (debug += 2); |
505bde11 | 5828 | if (mutually_exclusive_p (bufp, p1, p2)) |
fa9a63c5 | 5829 | { |
505bde11 | 5830 | /* Use a fast `on_failure_keep_string_jump' loop. */ |
dc4a2ee0 | 5831 | DEBUG_PRINT (" smart exclusive => fast loop.\n"); |
01618498 | 5832 | *p3 = (unsigned char) on_failure_keep_string_jump; |
4e8a9132 | 5833 | STORE_NUMBER (p2 - 2, mcnt + 3); |
25fe55af | 5834 | } |
505bde11 | 5835 | else |
fa9a63c5 | 5836 | { |
505bde11 | 5837 | /* Default to a safe `on_failure_jump' loop. */ |
dc4a2ee0 | 5838 | DEBUG_PRINT (" smart default => slow loop.\n"); |
01618498 | 5839 | *p3 = (unsigned char) on_failure_jump; |
fa9a63c5 | 5840 | } |
99633e97 | 5841 | DEBUG_STATEMENT (debug -= 2); |
25fe55af | 5842 | } |
505bde11 | 5843 | break; |
25fe55af RS |
5844 | |
5845 | /* Unconditionally jump (without popping any failure points). */ | |
5846 | case jump: | |
fa9a63c5 | 5847 | unconditional_jump: |
5b370c2b | 5848 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 5849 | EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ |
dc4a2ee0 | 5850 | DEBUG_PRINT ("EXECUTING jump %d ", mcnt); |
7814e705 | 5851 | p += mcnt; /* Do the jump. */ |
dc4a2ee0 | 5852 | DEBUG_PRINT ("(to %p).\n", p); |
25fe55af RS |
5853 | break; |
5854 | ||
5855 | ||
25fe55af RS |
5856 | /* Have to succeed matching what follows at least n times. |
5857 | After that, handle like `on_failure_jump'. */ | |
5858 | case succeed_n: | |
01618498 | 5859 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af | 5860 | EXTRACT_NUMBER (mcnt, p + 2); |
dc4a2ee0 | 5861 | DEBUG_PRINT ("EXECUTING succeed_n %d.\n", mcnt); |
5e69f11e | 5862 | |
dc1e502d SM |
5863 | /* Originally, mcnt is how many times we HAVE to succeed. */ |
5864 | if (mcnt != 0) | |
25fe55af | 5865 | { |
6dcf2d0e SM |
5866 | /* Here, we discard `const', making re_match non-reentrant. */ |
5867 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5868 | mcnt--; |
01618498 SM |
5869 | p += 4; |
5870 | PUSH_NUMBER (p2, mcnt); | |
25fe55af | 5871 | } |
dc1e502d SM |
5872 | else |
5873 | /* The two bytes encoding mcnt == 0 are two no_op opcodes. */ | |
5874 | goto on_failure; | |
25fe55af RS |
5875 | break; |
5876 | ||
5877 | case jump_n: | |
01618498 | 5878 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af | 5879 | EXTRACT_NUMBER (mcnt, p + 2); |
dc4a2ee0 | 5880 | DEBUG_PRINT ("EXECUTING jump_n %d.\n", mcnt); |
25fe55af RS |
5881 | |
5882 | /* Originally, this is how many times we CAN jump. */ | |
dc1e502d | 5883 | if (mcnt != 0) |
25fe55af | 5884 | { |
6dcf2d0e SM |
5885 | /* Here, we discard `const', making re_match non-reentrant. */ |
5886 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5887 | mcnt--; |
01618498 | 5888 | PUSH_NUMBER (p2, mcnt); |
dc1e502d | 5889 | goto unconditional_jump; |
25fe55af RS |
5890 | } |
5891 | /* If don't have to jump any more, skip over the rest of command. */ | |
5e69f11e RM |
5892 | else |
5893 | p += 4; | |
25fe55af | 5894 | break; |
5e69f11e | 5895 | |
fa9a63c5 RM |
5896 | case set_number_at: |
5897 | { | |
01618498 | 5898 | unsigned char *p2; /* Location of the counter. */ |
dc4a2ee0 | 5899 | DEBUG_PRINT ("EXECUTING set_number_at.\n"); |
fa9a63c5 | 5900 | |
25fe55af | 5901 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
6dcf2d0e SM |
5902 | /* Here, we discard `const', making re_match non-reentrant. */ |
5903 | p2 = (unsigned char*) p + mcnt; | |
f224e500 | 5904 | /* Signedness doesn't matter since we only copy MCNT's bits. */ |
25fe55af | 5905 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 | 5906 | DEBUG_PRINT (" Setting %p to %d.\n", p2, mcnt); |
01618498 | 5907 | PUSH_NUMBER (p2, mcnt); |
25fe55af RS |
5908 | break; |
5909 | } | |
9121ca40 KH |
5910 | |
5911 | case wordbound: | |
66f0296e | 5912 | case notwordbound: |
19ed5445 PE |
5913 | { |
5914 | boolean not = (re_opcode_t) *(p - 1) == notwordbound; | |
dc4a2ee0 | 5915 | DEBUG_PRINT ("EXECUTING %swordbound.\n", not ? "not" : ""); |
fa9a63c5 | 5916 | |
19ed5445 | 5917 | /* We SUCCEED (or FAIL) in one of the following cases: */ |
9121ca40 | 5918 | |
19ed5445 PE |
5919 | /* Case 1: D is at the beginning or the end of string. */ |
5920 | if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) | |
5921 | not = !not; | |
5922 | else | |
5923 | { | |
5924 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5925 | is the character at D, and S2 is the syntax of C2. */ | |
5926 | re_wchar_t c1, c2; | |
5927 | int s1, s2; | |
5928 | int dummy; | |
b18215fc | 5929 | #ifdef emacs |
d1dfb56c EZ |
5930 | ssize_t offset = PTR_TO_OFFSET (d - 1); |
5931 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 | 5932 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5933 | #endif |
19ed5445 PE |
5934 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
5935 | s1 = SYNTAX (c1); | |
b18215fc | 5936 | #ifdef emacs |
19ed5445 | 5937 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
25fe55af | 5938 | #endif |
19ed5445 PE |
5939 | PREFETCH_NOLIMIT (); |
5940 | GET_CHAR_AFTER (c2, d, dummy); | |
5941 | s2 = SYNTAX (c2); | |
5942 | ||
5943 | if (/* Case 2: Only one of S1 and S2 is Sword. */ | |
5944 | ((s1 == Sword) != (s2 == Sword)) | |
5945 | /* Case 3: Both of S1 and S2 are Sword, and macro | |
5946 | WORD_BOUNDARY_P (C1, C2) returns nonzero. */ | |
5947 | || ((s1 == Sword) && WORD_BOUNDARY_P (c1, c2))) | |
5948 | not = !not; | |
5949 | } | |
5950 | if (not) | |
5951 | break; | |
5952 | else | |
5953 | goto fail; | |
5954 | } | |
fa9a63c5 RM |
5955 | |
5956 | case wordbeg: | |
dc4a2ee0 | 5957 | DEBUG_PRINT ("EXECUTING wordbeg.\n"); |
fa9a63c5 | 5958 | |
b18215fc RS |
5959 | /* We FAIL in one of the following cases: */ |
5960 | ||
7814e705 | 5961 | /* Case 1: D is at the end of string. */ |
b18215fc | 5962 | if (AT_STRINGS_END (d)) |
99633e97 | 5963 | goto fail; |
b18215fc RS |
5964 | else |
5965 | { | |
5966 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5967 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
5968 | re_wchar_t c1, c2; |
5969 | int s1, s2; | |
bf216479 | 5970 | int dummy; |
fa9a63c5 | 5971 | #ifdef emacs |
d1dfb56c EZ |
5972 | ssize_t offset = PTR_TO_OFFSET (d); |
5973 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 5974 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5975 | #endif |
99633e97 | 5976 | PREFETCH (); |
6fdd04b0 | 5977 | GET_CHAR_AFTER (c2, d, dummy); |
b18215fc | 5978 | s2 = SYNTAX (c2); |
177c0ea7 | 5979 | |
b18215fc RS |
5980 | /* Case 2: S2 is not Sword. */ |
5981 | if (s2 != Sword) | |
5982 | goto fail; | |
5983 | ||
5984 | /* Case 3: D is not at the beginning of string ... */ | |
5985 | if (!AT_STRINGS_BEG (d)) | |
5986 | { | |
5987 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
5988 | #ifdef emacs | |
5d967c7a | 5989 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); |
25fe55af | 5990 | #endif |
b18215fc RS |
5991 | s1 = SYNTAX (c1); |
5992 | ||
5993 | /* ... and S1 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 5994 | returns 0. */ |
b18215fc RS |
5995 | if ((s1 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
5996 | goto fail; | |
5997 | } | |
5998 | } | |
e318085a RS |
5999 | break; |
6000 | ||
b18215fc | 6001 | case wordend: |
dc4a2ee0 | 6002 | DEBUG_PRINT ("EXECUTING wordend.\n"); |
b18215fc RS |
6003 | |
6004 | /* We FAIL in one of the following cases: */ | |
6005 | ||
6006 | /* Case 1: D is at the beginning of string. */ | |
6007 | if (AT_STRINGS_BEG (d)) | |
e318085a | 6008 | goto fail; |
b18215fc RS |
6009 | else |
6010 | { | |
6011 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6012 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
6013 | re_wchar_t c1, c2; |
6014 | int s1, s2; | |
bf216479 | 6015 | int dummy; |
5d967c7a | 6016 | #ifdef emacs |
d1dfb56c EZ |
6017 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6018 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 6019 | UPDATE_SYNTAX_TABLE (charpos); |
5d967c7a | 6020 | #endif |
99633e97 | 6021 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
b18215fc RS |
6022 | s1 = SYNTAX (c1); |
6023 | ||
6024 | /* Case 2: S1 is not Sword. */ | |
6025 | if (s1 != Sword) | |
6026 | goto fail; | |
6027 | ||
6028 | /* Case 3: D is not at the end of string ... */ | |
6029 | if (!AT_STRINGS_END (d)) | |
6030 | { | |
f1ad044f | 6031 | PREFETCH_NOLIMIT (); |
6fdd04b0 | 6032 | GET_CHAR_AFTER (c2, d, dummy); |
5d967c7a RS |
6033 | #ifdef emacs |
6034 | UPDATE_SYNTAX_TABLE_FORWARD (charpos); | |
6035 | #endif | |
b18215fc RS |
6036 | s2 = SYNTAX (c2); |
6037 | ||
6038 | /* ... and S2 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6039 | returns 0. */ |
b18215fc | 6040 | if ((s2 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
25fe55af | 6041 | goto fail; |
b18215fc RS |
6042 | } |
6043 | } | |
e318085a RS |
6044 | break; |
6045 | ||
669fa600 | 6046 | case symbeg: |
dc4a2ee0 | 6047 | DEBUG_PRINT ("EXECUTING symbeg.\n"); |
669fa600 SM |
6048 | |
6049 | /* We FAIL in one of the following cases: */ | |
6050 | ||
7814e705 | 6051 | /* Case 1: D is at the end of string. */ |
669fa600 SM |
6052 | if (AT_STRINGS_END (d)) |
6053 | goto fail; | |
6054 | else | |
6055 | { | |
6056 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6057 | is the character at D, and S2 is the syntax of C2. */ | |
6058 | re_wchar_t c1, c2; | |
6059 | int s1, s2; | |
6060 | #ifdef emacs | |
d1dfb56c EZ |
6061 | ssize_t offset = PTR_TO_OFFSET (d); |
6062 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6063 | UPDATE_SYNTAX_TABLE (charpos); |
6064 | #endif | |
6065 | PREFETCH (); | |
62a6e103 | 6066 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6067 | s2 = SYNTAX (c2); |
7814e705 | 6068 | |
669fa600 SM |
6069 | /* Case 2: S2 is neither Sword nor Ssymbol. */ |
6070 | if (s2 != Sword && s2 != Ssymbol) | |
6071 | goto fail; | |
6072 | ||
6073 | /* Case 3: D is not at the beginning of string ... */ | |
6074 | if (!AT_STRINGS_BEG (d)) | |
6075 | { | |
6076 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6077 | #ifdef emacs | |
6078 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); | |
6079 | #endif | |
6080 | s1 = SYNTAX (c1); | |
6081 | ||
6082 | /* ... and S1 is Sword or Ssymbol. */ | |
6083 | if (s1 == Sword || s1 == Ssymbol) | |
6084 | goto fail; | |
6085 | } | |
6086 | } | |
6087 | break; | |
6088 | ||
6089 | case symend: | |
dc4a2ee0 | 6090 | DEBUG_PRINT ("EXECUTING symend.\n"); |
669fa600 SM |
6091 | |
6092 | /* We FAIL in one of the following cases: */ | |
6093 | ||
6094 | /* Case 1: D is at the beginning of string. */ | |
6095 | if (AT_STRINGS_BEG (d)) | |
6096 | goto fail; | |
6097 | else | |
6098 | { | |
6099 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6100 | is the character at D, and S2 is the syntax of C2. */ | |
6101 | re_wchar_t c1, c2; | |
6102 | int s1, s2; | |
6103 | #ifdef emacs | |
d1dfb56c EZ |
6104 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6105 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6106 | UPDATE_SYNTAX_TABLE (charpos); |
6107 | #endif | |
6108 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6109 | s1 = SYNTAX (c1); | |
6110 | ||
6111 | /* Case 2: S1 is neither Ssymbol nor Sword. */ | |
6112 | if (s1 != Sword && s1 != Ssymbol) | |
6113 | goto fail; | |
6114 | ||
6115 | /* Case 3: D is not at the end of string ... */ | |
6116 | if (!AT_STRINGS_END (d)) | |
6117 | { | |
6118 | PREFETCH_NOLIMIT (); | |
62a6e103 | 6119 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6120 | #ifdef emacs |
134579f2 | 6121 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
669fa600 SM |
6122 | #endif |
6123 | s2 = SYNTAX (c2); | |
6124 | ||
6125 | /* ... and S2 is Sword or Ssymbol. */ | |
6126 | if (s2 == Sword || s2 == Ssymbol) | |
6127 | goto fail; | |
b18215fc RS |
6128 | } |
6129 | } | |
e318085a RS |
6130 | break; |
6131 | ||
fa9a63c5 | 6132 | case syntaxspec: |
1fb352e0 | 6133 | case notsyntaxspec: |
b18215fc | 6134 | { |
19ed5445 PE |
6135 | boolean not = (re_opcode_t) *(p - 1) == notsyntaxspec; |
6136 | mcnt = *p++; | |
dc4a2ee0 PE |
6137 | DEBUG_PRINT ("EXECUTING %ssyntaxspec %d.\n", not ? "not" : "", |
6138 | mcnt); | |
19ed5445 PE |
6139 | PREFETCH (); |
6140 | #ifdef emacs | |
6141 | { | |
d1dfb56c EZ |
6142 | ssize_t offset = PTR_TO_OFFSET (d); |
6143 | ssize_t pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 PE |
6144 | UPDATE_SYNTAX_TABLE (pos1); |
6145 | } | |
25fe55af | 6146 | #endif |
19ed5445 PE |
6147 | { |
6148 | int len; | |
6149 | re_wchar_t c; | |
b18215fc | 6150 | |
19ed5445 PE |
6151 | GET_CHAR_AFTER (c, d, len); |
6152 | if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not) | |
6153 | goto fail; | |
6154 | d += len; | |
6155 | } | |
b18215fc | 6156 | } |
8fb31792 | 6157 | break; |
fa9a63c5 | 6158 | |
b18215fc | 6159 | #ifdef emacs |
1fb352e0 | 6160 | case before_dot: |
dc4a2ee0 | 6161 | DEBUG_PRINT ("EXECUTING before_dot.\n"); |
1fb352e0 | 6162 | if (PTR_BYTE_POS (d) >= PT_BYTE) |
fa9a63c5 | 6163 | goto fail; |
b18215fc RS |
6164 | break; |
6165 | ||
1fb352e0 | 6166 | case at_dot: |
dc4a2ee0 | 6167 | DEBUG_PRINT ("EXECUTING at_dot.\n"); |
1fb352e0 SM |
6168 | if (PTR_BYTE_POS (d) != PT_BYTE) |
6169 | goto fail; | |
6170 | break; | |
b18215fc | 6171 | |
1fb352e0 | 6172 | case after_dot: |
dc4a2ee0 | 6173 | DEBUG_PRINT ("EXECUTING after_dot.\n"); |
1fb352e0 SM |
6174 | if (PTR_BYTE_POS (d) <= PT_BYTE) |
6175 | goto fail; | |
e318085a | 6176 | break; |
fa9a63c5 | 6177 | |
1fb352e0 | 6178 | case categoryspec: |
b18215fc | 6179 | case notcategoryspec: |
b18215fc | 6180 | { |
8fb31792 PE |
6181 | boolean not = (re_opcode_t) *(p - 1) == notcategoryspec; |
6182 | mcnt = *p++; | |
dc4a2ee0 PE |
6183 | DEBUG_PRINT ("EXECUTING %scategoryspec %d.\n", |
6184 | not ? "not" : "", mcnt); | |
8fb31792 | 6185 | PREFETCH (); |
01618498 | 6186 | |
8fb31792 PE |
6187 | { |
6188 | int len; | |
6189 | re_wchar_t c; | |
6190 | GET_CHAR_AFTER (c, d, len); | |
6191 | if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not) | |
6192 | goto fail; | |
6193 | d += len; | |
6194 | } | |
b18215fc | 6195 | } |
fa9a63c5 | 6196 | break; |
5e69f11e | 6197 | |
1fb352e0 | 6198 | #endif /* emacs */ |
5e69f11e | 6199 | |
0b32bf0e SM |
6200 | default: |
6201 | abort (); | |
fa9a63c5 | 6202 | } |
b18215fc | 6203 | continue; /* Successfully executed one pattern command; keep going. */ |
fa9a63c5 RM |
6204 | |
6205 | ||
6206 | /* We goto here if a matching operation fails. */ | |
6207 | fail: | |
5b370c2b | 6208 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 6209 | if (!FAIL_STACK_EMPTY ()) |
505bde11 | 6210 | { |
01618498 | 6211 | re_char *str, *pat; |
505bde11 | 6212 | /* A restart point is known. Restore to that state. */ |
dc4a2ee0 | 6213 | DEBUG_PRINT ("\nFAIL:\n"); |
0b32bf0e | 6214 | POP_FAILURE_POINT (str, pat); |
7393bcbb | 6215 | switch (*pat++) |
505bde11 SM |
6216 | { |
6217 | case on_failure_keep_string_jump: | |
6218 | assert (str == NULL); | |
6219 | goto continue_failure_jump; | |
6220 | ||
0683b6fa SM |
6221 | case on_failure_jump_nastyloop: |
6222 | assert ((re_opcode_t)pat[-2] == no_op); | |
6223 | PUSH_FAILURE_POINT (pat - 2, str); | |
6224 | /* Fallthrough */ | |
6225 | ||
505bde11 SM |
6226 | case on_failure_jump_loop: |
6227 | case on_failure_jump: | |
6228 | case succeed_n: | |
6229 | d = str; | |
6230 | continue_failure_jump: | |
6231 | EXTRACT_NUMBER_AND_INCR (mcnt, pat); | |
6232 | p = pat + mcnt; | |
6233 | break; | |
b18215fc | 6234 | |
0683b6fa SM |
6235 | case no_op: |
6236 | /* A special frame used for nastyloops. */ | |
6237 | goto fail; | |
6238 | ||
505bde11 | 6239 | default: |
5e617bc2 | 6240 | abort (); |
505bde11 | 6241 | } |
fa9a63c5 | 6242 | |
505bde11 | 6243 | assert (p >= bufp->buffer && p <= pend); |
b18215fc | 6244 | |
0b32bf0e | 6245 | if (d >= string1 && d <= end1) |
fa9a63c5 | 6246 | dend = end_match_1; |
0b32bf0e | 6247 | } |
fa9a63c5 | 6248 | else |
0b32bf0e | 6249 | break; /* Matching at this starting point really fails. */ |
fa9a63c5 RM |
6250 | } /* for (;;) */ |
6251 | ||
6252 | if (best_regs_set) | |
6253 | goto restore_best_regs; | |
6254 | ||
6255 | FREE_VARIABLES (); | |
6256 | ||
b18215fc | 6257 | return -1; /* Failure to match. */ |
dc4a2ee0 | 6258 | } |
fa9a63c5 RM |
6259 | \f |
6260 | /* Subroutine definitions for re_match_2. */ | |
6261 | ||
fa9a63c5 RM |
6262 | /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN |
6263 | bytes; nonzero otherwise. */ | |
5e69f11e | 6264 | |
fa9a63c5 | 6265 | static int |
29abe551 | 6266 | bcmp_translate (const_re_char *s1, const_re_char *s2, register ssize_t len, |
438105ed | 6267 | RE_TRANSLATE_TYPE translate, const int target_multibyte) |
fa9a63c5 | 6268 | { |
2d1675e4 SM |
6269 | register re_char *p1 = s1, *p2 = s2; |
6270 | re_char *p1_end = s1 + len; | |
6271 | re_char *p2_end = s2 + len; | |
e934739e | 6272 | |
4bb91c68 SM |
6273 | /* FIXME: Checking both p1 and p2 presumes that the two strings might have |
6274 | different lengths, but relying on a single `len' would break this. -sm */ | |
6275 | while (p1 < p1_end && p2 < p2_end) | |
fa9a63c5 | 6276 | { |
e934739e | 6277 | int p1_charlen, p2_charlen; |
01618498 | 6278 | re_wchar_t p1_ch, p2_ch; |
e934739e | 6279 | |
6fdd04b0 KH |
6280 | GET_CHAR_AFTER (p1_ch, p1, p1_charlen); |
6281 | GET_CHAR_AFTER (p2_ch, p2, p2_charlen); | |
e934739e RS |
6282 | |
6283 | if (RE_TRANSLATE (translate, p1_ch) | |
6284 | != RE_TRANSLATE (translate, p2_ch)) | |
bc192b5b | 6285 | return 1; |
e934739e RS |
6286 | |
6287 | p1 += p1_charlen, p2 += p2_charlen; | |
fa9a63c5 | 6288 | } |
e934739e RS |
6289 | |
6290 | if (p1 != p1_end || p2 != p2_end) | |
6291 | return 1; | |
6292 | ||
fa9a63c5 RM |
6293 | return 0; |
6294 | } | |
6295 | \f | |
6296 | /* Entry points for GNU code. */ | |
6297 | ||
6298 | /* re_compile_pattern is the GNU regular expression compiler: it | |
6299 | compiles PATTERN (of length SIZE) and puts the result in BUFP. | |
6300 | Returns 0 if the pattern was valid, otherwise an error string. | |
5e69f11e | 6301 | |
fa9a63c5 RM |
6302 | Assumes the `allocated' (and perhaps `buffer') and `translate' fields |
6303 | are set in BUFP on entry. | |
5e69f11e | 6304 | |
b18215fc | 6305 | We call regex_compile to do the actual compilation. */ |
fa9a63c5 RM |
6306 | |
6307 | const char * | |
d1dfb56c EZ |
6308 | re_compile_pattern (const char *pattern, size_t length, |
6309 | struct re_pattern_buffer *bufp) | |
fa9a63c5 RM |
6310 | { |
6311 | reg_errcode_t ret; | |
5e69f11e | 6312 | |
fa9a63c5 RM |
6313 | /* GNU code is written to assume at least RE_NREGS registers will be set |
6314 | (and at least one extra will be -1). */ | |
6315 | bufp->regs_allocated = REGS_UNALLOCATED; | |
5e69f11e | 6316 | |
fa9a63c5 RM |
6317 | /* And GNU code determines whether or not to get register information |
6318 | by passing null for the REGS argument to re_match, etc., not by | |
6319 | setting no_sub. */ | |
6320 | bufp->no_sub = 0; | |
5e69f11e | 6321 | |
4bb91c68 | 6322 | ret = regex_compile ((re_char*) pattern, length, re_syntax_options, bufp); |
fa9a63c5 RM |
6323 | |
6324 | if (!ret) | |
6325 | return NULL; | |
6326 | return gettext (re_error_msgid[(int) ret]); | |
5e69f11e | 6327 | } |
c0f9ea08 | 6328 | WEAK_ALIAS (__re_compile_pattern, re_compile_pattern) |
fa9a63c5 | 6329 | \f |
b18215fc RS |
6330 | /* Entry points compatible with 4.2 BSD regex library. We don't define |
6331 | them unless specifically requested. */ | |
fa9a63c5 | 6332 | |
0b32bf0e | 6333 | #if defined _REGEX_RE_COMP || defined _LIBC |
fa9a63c5 RM |
6334 | |
6335 | /* BSD has one and only one pattern buffer. */ | |
6336 | static struct re_pattern_buffer re_comp_buf; | |
6337 | ||
6338 | char * | |
0b32bf0e | 6339 | # ifdef _LIBC |
48afdd44 RM |
6340 | /* Make these definitions weak in libc, so POSIX programs can redefine |
6341 | these names if they don't use our functions, and still use | |
6342 | regcomp/regexec below without link errors. */ | |
6343 | weak_function | |
0b32bf0e | 6344 | # endif |
31011111 | 6345 | re_comp (const char *s) |
fa9a63c5 RM |
6346 | { |
6347 | reg_errcode_t ret; | |
5e69f11e | 6348 | |
fa9a63c5 RM |
6349 | if (!s) |
6350 | { | |
6351 | if (!re_comp_buf.buffer) | |
0b32bf0e | 6352 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
a60198e5 | 6353 | return (char *) gettext ("No previous regular expression"); |
fa9a63c5 RM |
6354 | return 0; |
6355 | } | |
6356 | ||
6357 | if (!re_comp_buf.buffer) | |
6358 | { | |
38182d90 | 6359 | re_comp_buf.buffer = malloc (200); |
fa9a63c5 | 6360 | if (re_comp_buf.buffer == NULL) |
0b32bf0e SM |
6361 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6362 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6363 | re_comp_buf.allocated = 200; |
6364 | ||
38182d90 | 6365 | re_comp_buf.fastmap = malloc (1 << BYTEWIDTH); |
fa9a63c5 | 6366 | if (re_comp_buf.fastmap == NULL) |
a60198e5 SM |
6367 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6368 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6369 | } |
6370 | ||
6371 | /* Since `re_exec' always passes NULL for the `regs' argument, we | |
6372 | don't need to initialize the pattern buffer fields which affect it. */ | |
6373 | ||
fa9a63c5 | 6374 | ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); |
5e69f11e | 6375 | |
fa9a63c5 RM |
6376 | if (!ret) |
6377 | return NULL; | |
6378 | ||
6379 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | |
6380 | return (char *) gettext (re_error_msgid[(int) ret]); | |
6381 | } | |
6382 | ||
6383 | ||
31011111 | 6384 | int |
0b32bf0e | 6385 | # ifdef _LIBC |
48afdd44 | 6386 | weak_function |
0b32bf0e | 6387 | # endif |
d1dfb56c | 6388 | re_exec (const char *s) |
fa9a63c5 | 6389 | { |
d1dfb56c | 6390 | const size_t len = strlen (s); |
7d652d97 | 6391 | return re_search (&re_comp_buf, s, len, 0, len, 0) >= 0; |
fa9a63c5 RM |
6392 | } |
6393 | #endif /* _REGEX_RE_COMP */ | |
6394 | \f | |
6395 | /* POSIX.2 functions. Don't define these for Emacs. */ | |
6396 | ||
6397 | #ifndef emacs | |
6398 | ||
6399 | /* regcomp takes a regular expression as a string and compiles it. | |
6400 | ||
b18215fc | 6401 | PREG is a regex_t *. We do not expect any fields to be initialized, |
fa9a63c5 RM |
6402 | since POSIX says we shouldn't. Thus, we set |
6403 | ||
6404 | `buffer' to the compiled pattern; | |
6405 | `used' to the length of the compiled pattern; | |
6406 | `syntax' to RE_SYNTAX_POSIX_EXTENDED if the | |
6407 | REG_EXTENDED bit in CFLAGS is set; otherwise, to | |
6408 | RE_SYNTAX_POSIX_BASIC; | |
c0f9ea08 SM |
6409 | `fastmap' to an allocated space for the fastmap; |
6410 | `fastmap_accurate' to zero; | |
fa9a63c5 RM |
6411 | `re_nsub' to the number of subexpressions in PATTERN. |
6412 | ||
6413 | PATTERN is the address of the pattern string. | |
6414 | ||
6415 | CFLAGS is a series of bits which affect compilation. | |
6416 | ||
6417 | If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we | |
6418 | use POSIX basic syntax. | |
6419 | ||
6420 | If REG_NEWLINE is set, then . and [^...] don't match newline. | |
6421 | Also, regexec will try a match beginning after every newline. | |
6422 | ||
6423 | If REG_ICASE is set, then we considers upper- and lowercase | |
6424 | versions of letters to be equivalent when matching. | |
6425 | ||
6426 | If REG_NOSUB is set, then when PREG is passed to regexec, that | |
6427 | routine will report only success or failure, and nothing about the | |
6428 | registers. | |
6429 | ||
b18215fc | 6430 | It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for |
fa9a63c5 RM |
6431 | the return codes and their meanings.) */ |
6432 | ||
d1dfb56c | 6433 | reg_errcode_t |
29abe551 | 6434 | regcomp (regex_t *_Restrict_ preg, const char *_Restrict_ pattern, |
d2762c86 | 6435 | int cflags) |
fa9a63c5 RM |
6436 | { |
6437 | reg_errcode_t ret; | |
4bb91c68 | 6438 | reg_syntax_t syntax |
fa9a63c5 RM |
6439 | = (cflags & REG_EXTENDED) ? |
6440 | RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; | |
6441 | ||
6442 | /* regex_compile will allocate the space for the compiled pattern. */ | |
6443 | preg->buffer = 0; | |
6444 | preg->allocated = 0; | |
6445 | preg->used = 0; | |
5e69f11e | 6446 | |
c0f9ea08 | 6447 | /* Try to allocate space for the fastmap. */ |
38182d90 | 6448 | preg->fastmap = malloc (1 << BYTEWIDTH); |
5e69f11e | 6449 | |
fa9a63c5 RM |
6450 | if (cflags & REG_ICASE) |
6451 | { | |
6452 | unsigned i; | |
5e69f11e | 6453 | |
38182d90 | 6454 | preg->translate = malloc (CHAR_SET_SIZE * sizeof *preg->translate); |
fa9a63c5 | 6455 | if (preg->translate == NULL) |
0b32bf0e | 6456 | return (int) REG_ESPACE; |
fa9a63c5 RM |
6457 | |
6458 | /* Map uppercase characters to corresponding lowercase ones. */ | |
6459 | for (i = 0; i < CHAR_SET_SIZE; i++) | |
4bb91c68 | 6460 | preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i; |
fa9a63c5 RM |
6461 | } |
6462 | else | |
6463 | preg->translate = NULL; | |
6464 | ||
6465 | /* If REG_NEWLINE is set, newlines are treated differently. */ | |
6466 | if (cflags & REG_NEWLINE) | |
6467 | { /* REG_NEWLINE implies neither . nor [^...] match newline. */ | |
6468 | syntax &= ~RE_DOT_NEWLINE; | |
6469 | syntax |= RE_HAT_LISTS_NOT_NEWLINE; | |
fa9a63c5 RM |
6470 | } |
6471 | else | |
c0f9ea08 | 6472 | syntax |= RE_NO_NEWLINE_ANCHOR; |
fa9a63c5 RM |
6473 | |
6474 | preg->no_sub = !!(cflags & REG_NOSUB); | |
6475 | ||
5e69f11e | 6476 | /* POSIX says a null character in the pattern terminates it, so we |
fa9a63c5 | 6477 | can use strlen here in compiling the pattern. */ |
4bb91c68 | 6478 | ret = regex_compile ((re_char*) pattern, strlen (pattern), syntax, preg); |
5e69f11e | 6479 | |
fa9a63c5 RM |
6480 | /* POSIX doesn't distinguish between an unmatched open-group and an |
6481 | unmatched close-group: both are REG_EPAREN. */ | |
c0f9ea08 SM |
6482 | if (ret == REG_ERPAREN) |
6483 | ret = REG_EPAREN; | |
6484 | ||
6485 | if (ret == REG_NOERROR && preg->fastmap) | |
6486 | { /* Compute the fastmap now, since regexec cannot modify the pattern | |
6487 | buffer. */ | |
6488 | re_compile_fastmap (preg); | |
6489 | if (preg->can_be_null) | |
6490 | { /* The fastmap can't be used anyway. */ | |
6491 | free (preg->fastmap); | |
6492 | preg->fastmap = NULL; | |
6493 | } | |
6494 | } | |
d1dfb56c | 6495 | return ret; |
fa9a63c5 | 6496 | } |
c0f9ea08 | 6497 | WEAK_ALIAS (__regcomp, regcomp) |
fa9a63c5 RM |
6498 | |
6499 | ||
6500 | /* regexec searches for a given pattern, specified by PREG, in the | |
6501 | string STRING. | |
5e69f11e | 6502 | |
fa9a63c5 | 6503 | If NMATCH is zero or REG_NOSUB was set in the cflags argument to |
b18215fc | 6504 | `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at |
fa9a63c5 RM |
6505 | least NMATCH elements, and we set them to the offsets of the |
6506 | corresponding matched substrings. | |
5e69f11e | 6507 | |
fa9a63c5 RM |
6508 | EFLAGS specifies `execution flags' which affect matching: if |
6509 | REG_NOTBOL is set, then ^ does not match at the beginning of the | |
6510 | string; if REG_NOTEOL is set, then $ does not match at the end. | |
5e69f11e | 6511 | |
fa9a63c5 RM |
6512 | We return 0 if we find a match and REG_NOMATCH if not. */ |
6513 | ||
d1dfb56c | 6514 | reg_errcode_t |
29abe551 PE |
6515 | regexec (const regex_t *_Restrict_ preg, const char *_Restrict_ string, |
6516 | size_t nmatch, regmatch_t pmatch[_Restrict_arr_], int eflags) | |
fa9a63c5 | 6517 | { |
31011111 | 6518 | regoff_t ret; |
fa9a63c5 RM |
6519 | struct re_registers regs; |
6520 | regex_t private_preg; | |
d1dfb56c | 6521 | size_t len = strlen (string); |
c0f9ea08 | 6522 | boolean want_reg_info = !preg->no_sub && nmatch > 0 && pmatch; |
fa9a63c5 RM |
6523 | |
6524 | private_preg = *preg; | |
5e69f11e | 6525 | |
fa9a63c5 RM |
6526 | private_preg.not_bol = !!(eflags & REG_NOTBOL); |
6527 | private_preg.not_eol = !!(eflags & REG_NOTEOL); | |
5e69f11e | 6528 | |
fa9a63c5 RM |
6529 | /* The user has told us exactly how many registers to return |
6530 | information about, via `nmatch'. We have to pass that on to the | |
b18215fc | 6531 | matching routines. */ |
fa9a63c5 | 6532 | private_preg.regs_allocated = REGS_FIXED; |
5e69f11e | 6533 | |
fa9a63c5 RM |
6534 | if (want_reg_info) |
6535 | { | |
6536 | regs.num_regs = nmatch; | |
4bb91c68 SM |
6537 | regs.start = TALLOC (nmatch * 2, regoff_t); |
6538 | if (regs.start == NULL) | |
d1dfb56c | 6539 | return REG_NOMATCH; |
4bb91c68 | 6540 | regs.end = regs.start + nmatch; |
fa9a63c5 RM |
6541 | } |
6542 | ||
c0f9ea08 SM |
6543 | /* Instead of using not_eol to implement REG_NOTEOL, we could simply |
6544 | pass (&private_preg, string, len + 1, 0, len, ...) pretending the string | |
6545 | was a little bit longer but still only matching the real part. | |
6546 | This works because the `endline' will check for a '\n' and will find a | |
6547 | '\0', correctly deciding that this is not the end of a line. | |
6548 | But it doesn't work out so nicely for REG_NOTBOL, since we don't have | |
6549 | a convenient '\0' there. For all we know, the string could be preceded | |
6550 | by '\n' which would throw things off. */ | |
6551 | ||
fa9a63c5 RM |
6552 | /* Perform the searching operation. */ |
6553 | ret = re_search (&private_preg, string, len, | |
0b32bf0e | 6554 | /* start: */ 0, /* range: */ len, |
7d652d97 | 6555 | want_reg_info ? ®s : 0); |
5e69f11e | 6556 | |
fa9a63c5 RM |
6557 | /* Copy the register information to the POSIX structure. */ |
6558 | if (want_reg_info) | |
6559 | { | |
6560 | if (ret >= 0) | |
0b32bf0e SM |
6561 | { |
6562 | unsigned r; | |
fa9a63c5 | 6563 | |
0b32bf0e SM |
6564 | for (r = 0; r < nmatch; r++) |
6565 | { | |
6566 | pmatch[r].rm_so = regs.start[r]; | |
6567 | pmatch[r].rm_eo = regs.end[r]; | |
6568 | } | |
6569 | } | |
fa9a63c5 | 6570 | |
b18215fc | 6571 | /* If we needed the temporary register info, free the space now. */ |
fa9a63c5 | 6572 | free (regs.start); |
fa9a63c5 RM |
6573 | } |
6574 | ||
6575 | /* We want zero return to mean success, unlike `re_search'. */ | |
d1dfb56c | 6576 | return ret >= 0 ? REG_NOERROR : REG_NOMATCH; |
fa9a63c5 | 6577 | } |
c0f9ea08 | 6578 | WEAK_ALIAS (__regexec, regexec) |
fa9a63c5 RM |
6579 | |
6580 | ||
ec869672 JR |
6581 | /* Returns a message corresponding to an error code, ERR_CODE, returned |
6582 | from either regcomp or regexec. We don't use PREG here. | |
6583 | ||
6584 | ERR_CODE was previously called ERRCODE, but that name causes an | |
6585 | error with msvc8 compiler. */ | |
fa9a63c5 RM |
6586 | |
6587 | size_t | |
d2762c86 | 6588 | regerror (int err_code, const regex_t *preg, char *errbuf, size_t errbuf_size) |
fa9a63c5 RM |
6589 | { |
6590 | const char *msg; | |
6591 | size_t msg_size; | |
6592 | ||
ec869672 JR |
6593 | if (err_code < 0 |
6594 | || err_code >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0]))) | |
5e69f11e | 6595 | /* Only error codes returned by the rest of the code should be passed |
b18215fc | 6596 | to this routine. If we are given anything else, or if other regex |
fa9a63c5 RM |
6597 | code generates an invalid error code, then the program has a bug. |
6598 | Dump core so we can fix it. */ | |
6599 | abort (); | |
6600 | ||
ec869672 | 6601 | msg = gettext (re_error_msgid[err_code]); |
fa9a63c5 RM |
6602 | |
6603 | msg_size = strlen (msg) + 1; /* Includes the null. */ | |
5e69f11e | 6604 | |
fa9a63c5 RM |
6605 | if (errbuf_size != 0) |
6606 | { | |
6607 | if (msg_size > errbuf_size) | |
0b32bf0e | 6608 | { |
e99a530f | 6609 | memcpy (errbuf, msg, errbuf_size - 1); |
0b32bf0e SM |
6610 | errbuf[errbuf_size - 1] = 0; |
6611 | } | |
fa9a63c5 | 6612 | else |
0b32bf0e | 6613 | strcpy (errbuf, msg); |
fa9a63c5 RM |
6614 | } |
6615 | ||
6616 | return msg_size; | |
6617 | } | |
c0f9ea08 | 6618 | WEAK_ALIAS (__regerror, regerror) |
fa9a63c5 RM |
6619 | |
6620 | ||
6621 | /* Free dynamically allocated space used by PREG. */ | |
6622 | ||
6623 | void | |
d2762c86 | 6624 | regfree (regex_t *preg) |
fa9a63c5 | 6625 | { |
c2cd06e6 | 6626 | free (preg->buffer); |
fa9a63c5 | 6627 | preg->buffer = NULL; |
5e69f11e | 6628 | |
fa9a63c5 RM |
6629 | preg->allocated = 0; |
6630 | preg->used = 0; | |
6631 | ||
c2cd06e6 | 6632 | free (preg->fastmap); |
fa9a63c5 RM |
6633 | preg->fastmap = NULL; |
6634 | preg->fastmap_accurate = 0; | |
6635 | ||
c2cd06e6 | 6636 | free (preg->translate); |
fa9a63c5 RM |
6637 | preg->translate = NULL; |
6638 | } | |
c0f9ea08 | 6639 | WEAK_ALIAS (__regfree, regfree) |
fa9a63c5 RM |
6640 | |
6641 | #endif /* not emacs */ |