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 | { |
dc4a2ee0 | 716 | return (SIGN_EXTEND_CHAR (source[1]) << 8) + source[0]; |
fa9a63c5 RM |
717 | } |
718 | ||
fa9a63c5 RM |
719 | /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. |
720 | SOURCE must be an lvalue. */ | |
721 | ||
722 | #define EXTRACT_NUMBER_AND_INCR(destination, source) \ | |
dc4a2ee0 | 723 | ((destination) = extract_number_and_incr (&source)) |
fa9a63c5 | 724 | |
dc4a2ee0 PE |
725 | static int |
726 | extract_number_and_incr (re_char **source) | |
5e69f11e | 727 | { |
dc4a2ee0 | 728 | int num = extract_number (*source); |
fa9a63c5 | 729 | *source += 2; |
dc4a2ee0 | 730 | return num; |
fa9a63c5 | 731 | } |
fa9a63c5 | 732 | \f |
b18215fc RS |
733 | /* Store a multibyte character in three contiguous bytes starting |
734 | DESTINATION, and increment DESTINATION to the byte after where the | |
7814e705 | 735 | character is stored. Therefore, DESTINATION must be an lvalue. */ |
b18215fc RS |
736 | |
737 | #define STORE_CHARACTER_AND_INCR(destination, character) \ | |
738 | do { \ | |
739 | (destination)[0] = (character) & 0377; \ | |
740 | (destination)[1] = ((character) >> 8) & 0377; \ | |
741 | (destination)[2] = (character) >> 16; \ | |
742 | (destination) += 3; \ | |
743 | } while (0) | |
744 | ||
745 | /* Put into DESTINATION a character stored in three contiguous bytes | |
7814e705 | 746 | starting at SOURCE. */ |
b18215fc RS |
747 | |
748 | #define EXTRACT_CHARACTER(destination, source) \ | |
749 | do { \ | |
750 | (destination) = ((source)[0] \ | |
751 | | ((source)[1] << 8) \ | |
752 | | ((source)[2] << 16)); \ | |
753 | } while (0) | |
754 | ||
755 | ||
756 | /* Macros for charset. */ | |
757 | ||
758 | /* Size of bitmap of charset P in bytes. P is a start of charset, | |
759 | i.e. *P is (re_opcode_t) charset or (re_opcode_t) charset_not. */ | |
760 | #define CHARSET_BITMAP_SIZE(p) ((p)[1] & 0x7F) | |
761 | ||
762 | /* Nonzero if charset P has range table. */ | |
25fe55af | 763 | #define CHARSET_RANGE_TABLE_EXISTS_P(p) ((p)[1] & 0x80) |
b18215fc RS |
764 | |
765 | /* Return the address of range table of charset P. But not the start | |
766 | of table itself, but the before where the number of ranges is | |
96cc36cc RS |
767 | stored. `2 +' means to skip re_opcode_t and size of bitmap, |
768 | and the 2 bytes of flags at the start of the range table. */ | |
769 | #define CHARSET_RANGE_TABLE(p) (&(p)[4 + CHARSET_BITMAP_SIZE (p)]) | |
770 | ||
78779650 | 771 | #ifdef emacs |
96cc36cc RS |
772 | /* Extract the bit flags that start a range table. */ |
773 | #define CHARSET_RANGE_TABLE_BITS(p) \ | |
774 | ((p)[2 + CHARSET_BITMAP_SIZE (p)] \ | |
775 | + (p)[3 + CHARSET_BITMAP_SIZE (p)] * 0x100) | |
78779650 | 776 | #endif |
b18215fc | 777 | |
b18215fc | 778 | /* Return the address of end of RANGE_TABLE. COUNT is number of |
7814e705 JB |
779 | ranges (which is a pair of (start, end)) in the RANGE_TABLE. `* 2' |
780 | is start of range and end of range. `* 3' is size of each start | |
b18215fc RS |
781 | and end. */ |
782 | #define CHARSET_RANGE_TABLE_END(range_table, count) \ | |
783 | ((range_table) + (count) * 2 * 3) | |
784 | ||
7814e705 | 785 | /* Test if C is in RANGE_TABLE. A flag NOT is negated if C is in. |
b18215fc RS |
786 | COUNT is number of ranges in RANGE_TABLE. */ |
787 | #define CHARSET_LOOKUP_RANGE_TABLE_RAW(not, c, range_table, count) \ | |
788 | do \ | |
789 | { \ | |
01618498 | 790 | re_wchar_t range_start, range_end; \ |
19ed5445 | 791 | re_char *rtp; \ |
01618498 | 792 | re_char *range_table_end \ |
b18215fc RS |
793 | = CHARSET_RANGE_TABLE_END ((range_table), (count)); \ |
794 | \ | |
19ed5445 | 795 | for (rtp = (range_table); rtp < range_table_end; rtp += 2 * 3) \ |
b18215fc | 796 | { \ |
19ed5445 PE |
797 | EXTRACT_CHARACTER (range_start, rtp); \ |
798 | EXTRACT_CHARACTER (range_end, rtp + 3); \ | |
b18215fc RS |
799 | \ |
800 | if (range_start <= (c) && (c) <= range_end) \ | |
801 | { \ | |
802 | (not) = !(not); \ | |
803 | break; \ | |
804 | } \ | |
805 | } \ | |
806 | } \ | |
807 | while (0) | |
808 | ||
809 | /* Test if C is in range table of CHARSET. The flag NOT is negated if | |
810 | C is listed in it. */ | |
811 | #define CHARSET_LOOKUP_RANGE_TABLE(not, c, charset) \ | |
812 | do \ | |
813 | { \ | |
814 | /* Number of ranges in range table. */ \ | |
815 | int count; \ | |
01618498 SM |
816 | re_char *range_table = CHARSET_RANGE_TABLE (charset); \ |
817 | \ | |
b18215fc RS |
818 | EXTRACT_NUMBER_AND_INCR (count, range_table); \ |
819 | CHARSET_LOOKUP_RANGE_TABLE_RAW ((not), (c), range_table, count); \ | |
820 | } \ | |
821 | while (0) | |
822 | \f | |
fa9a63c5 RM |
823 | /* If DEBUG is defined, Regex prints many voluminous messages about what |
824 | it is doing (if the variable `debug' is nonzero). If linked with the | |
825 | main program in `iregex.c', you can enter patterns and strings | |
826 | interactively. And if linked with the main program in `main.c' and | |
4bb91c68 | 827 | the other test files, you can run the already-written tests. */ |
fa9a63c5 RM |
828 | |
829 | #ifdef DEBUG | |
830 | ||
831 | /* We use standard I/O for debugging. */ | |
0b32bf0e | 832 | # include <stdio.h> |
fa9a63c5 RM |
833 | |
834 | /* It is useful to test things that ``must'' be true when debugging. */ | |
0b32bf0e | 835 | # include <assert.h> |
fa9a63c5 | 836 | |
99633e97 | 837 | static int debug = -100000; |
fa9a63c5 | 838 | |
0b32bf0e | 839 | # define DEBUG_STATEMENT(e) e |
dc4a2ee0 PE |
840 | # define DEBUG_PRINT(...) if (debug > 0) printf (__VA_ARGS__) |
841 | # define DEBUG_COMPILES_ARGUMENTS | |
0b32bf0e | 842 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ |
99633e97 | 843 | if (debug > 0) print_partial_compiled_pattern (s, e) |
0b32bf0e | 844 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ |
99633e97 | 845 | if (debug > 0) print_double_string (w, s1, sz1, s2, sz2) |
fa9a63c5 RM |
846 | |
847 | ||
848 | /* Print the fastmap in human-readable form. */ | |
849 | ||
dc4a2ee0 PE |
850 | static void |
851 | print_fastmap (char *fastmap) | |
fa9a63c5 RM |
852 | { |
853 | unsigned was_a_range = 0; | |
5e69f11e RM |
854 | unsigned i = 0; |
855 | ||
fa9a63c5 RM |
856 | while (i < (1 << BYTEWIDTH)) |
857 | { | |
858 | if (fastmap[i++]) | |
859 | { | |
860 | was_a_range = 0; | |
25fe55af RS |
861 | putchar (i - 1); |
862 | while (i < (1 << BYTEWIDTH) && fastmap[i]) | |
863 | { | |
864 | was_a_range = 1; | |
865 | i++; | |
866 | } | |
fa9a63c5 | 867 | if (was_a_range) |
25fe55af RS |
868 | { |
869 | printf ("-"); | |
870 | putchar (i - 1); | |
871 | } | |
872 | } | |
fa9a63c5 | 873 | } |
5e69f11e | 874 | putchar ('\n'); |
fa9a63c5 RM |
875 | } |
876 | ||
877 | ||
878 | /* Print a compiled pattern string in human-readable form, starting at | |
879 | the START pointer into it and ending just before the pointer END. */ | |
880 | ||
dc4a2ee0 PE |
881 | static void |
882 | print_partial_compiled_pattern (re_char *start, re_char *end) | |
fa9a63c5 RM |
883 | { |
884 | int mcnt, mcnt2; | |
01618498 SM |
885 | re_char *p = start; |
886 | re_char *pend = end; | |
fa9a63c5 RM |
887 | |
888 | if (start == NULL) | |
889 | { | |
a1a052df | 890 | fprintf (stderr, "(null)\n"); |
fa9a63c5 RM |
891 | return; |
892 | } | |
5e69f11e | 893 | |
fa9a63c5 RM |
894 | /* Loop over pattern commands. */ |
895 | while (p < pend) | |
896 | { | |
dc4a2ee0 | 897 | fprintf (stderr, "%td:\t", p - start); |
fa9a63c5 RM |
898 | |
899 | switch ((re_opcode_t) *p++) | |
900 | { | |
25fe55af | 901 | case no_op: |
a1a052df | 902 | fprintf (stderr, "/no_op"); |
25fe55af | 903 | break; |
fa9a63c5 | 904 | |
99633e97 | 905 | case succeed: |
a1a052df | 906 | fprintf (stderr, "/succeed"); |
99633e97 SM |
907 | break; |
908 | ||
fa9a63c5 RM |
909 | case exactn: |
910 | mcnt = *p++; | |
a1a052df | 911 | fprintf (stderr, "/exactn/%d", mcnt); |
25fe55af | 912 | do |
fa9a63c5 | 913 | { |
a1a052df | 914 | fprintf (stderr, "/%c", *p++); |
25fe55af RS |
915 | } |
916 | while (--mcnt); | |
917 | break; | |
fa9a63c5 RM |
918 | |
919 | case start_memory: | |
a1a052df | 920 | fprintf (stderr, "/start_memory/%d", *p++); |
25fe55af | 921 | break; |
fa9a63c5 RM |
922 | |
923 | case stop_memory: | |
a1a052df | 924 | fprintf (stderr, "/stop_memory/%d", *p++); |
25fe55af | 925 | break; |
fa9a63c5 RM |
926 | |
927 | case duplicate: | |
a1a052df | 928 | fprintf (stderr, "/duplicate/%d", *p++); |
fa9a63c5 RM |
929 | break; |
930 | ||
931 | case anychar: | |
a1a052df | 932 | fprintf (stderr, "/anychar"); |
fa9a63c5 RM |
933 | break; |
934 | ||
935 | case charset: | |
25fe55af RS |
936 | case charset_not: |
937 | { | |
938 | register int c, last = -100; | |
fa9a63c5 | 939 | register int in_range = 0; |
99633e97 SM |
940 | int length = CHARSET_BITMAP_SIZE (p - 1); |
941 | int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1); | |
fa9a63c5 | 942 | |
a1a052df | 943 | fprintf (stderr, "/charset [%s", |
839966f3 | 944 | (re_opcode_t) *(p - 1) == charset_not ? "^" : ""); |
5e69f11e | 945 | |
839966f3 KH |
946 | if (p + *p >= pend) |
947 | fprintf (stderr, " !extends past end of pattern! "); | |
fa9a63c5 | 948 | |
25fe55af | 949 | for (c = 0; c < 256; c++) |
96cc36cc | 950 | if (c / 8 < length |
fa9a63c5 RM |
951 | && (p[1 + (c/8)] & (1 << (c % 8)))) |
952 | { | |
953 | /* Are we starting a range? */ | |
954 | if (last + 1 == c && ! in_range) | |
955 | { | |
a1a052df | 956 | fprintf (stderr, "-"); |
fa9a63c5 RM |
957 | in_range = 1; |
958 | } | |
959 | /* Have we broken a range? */ | |
960 | else if (last + 1 != c && in_range) | |
96cc36cc | 961 | { |
a1a052df | 962 | fprintf (stderr, "%c", last); |
fa9a63c5 RM |
963 | in_range = 0; |
964 | } | |
5e69f11e | 965 | |
fa9a63c5 | 966 | if (! in_range) |
a1a052df | 967 | fprintf (stderr, "%c", c); |
fa9a63c5 RM |
968 | |
969 | last = c; | |
25fe55af | 970 | } |
fa9a63c5 RM |
971 | |
972 | if (in_range) | |
a1a052df | 973 | fprintf (stderr, "%c", last); |
fa9a63c5 | 974 | |
a1a052df | 975 | fprintf (stderr, "]"); |
fa9a63c5 | 976 | |
99633e97 | 977 | p += 1 + length; |
96cc36cc | 978 | |
96cc36cc | 979 | if (has_range_table) |
99633e97 SM |
980 | { |
981 | int count; | |
a1a052df | 982 | fprintf (stderr, "has-range-table"); |
99633e97 SM |
983 | |
984 | /* ??? Should print the range table; for now, just skip it. */ | |
985 | p += 2; /* skip range table bits */ | |
986 | EXTRACT_NUMBER_AND_INCR (count, p); | |
987 | p = CHARSET_RANGE_TABLE_END (p, count); | |
988 | } | |
fa9a63c5 RM |
989 | } |
990 | break; | |
991 | ||
992 | case begline: | |
a1a052df | 993 | fprintf (stderr, "/begline"); |
25fe55af | 994 | break; |
fa9a63c5 RM |
995 | |
996 | case endline: | |
a1a052df | 997 | fprintf (stderr, "/endline"); |
25fe55af | 998 | break; |
fa9a63c5 RM |
999 | |
1000 | case on_failure_jump: | |
dc4a2ee0 PE |
1001 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1002 | fprintf (stderr, "/on_failure_jump to %td", p + mcnt - start); | |
25fe55af | 1003 | break; |
fa9a63c5 RM |
1004 | |
1005 | case on_failure_keep_string_jump: | |
dc4a2ee0 PE |
1006 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1007 | fprintf (stderr, "/on_failure_keep_string_jump to %td", | |
1008 | p + mcnt - start); | |
25fe55af | 1009 | break; |
fa9a63c5 | 1010 | |
0683b6fa | 1011 | case on_failure_jump_nastyloop: |
dc4a2ee0 PE |
1012 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1013 | fprintf (stderr, "/on_failure_jump_nastyloop to %td", | |
1014 | p + mcnt - start); | |
0683b6fa SM |
1015 | break; |
1016 | ||
505bde11 | 1017 | case on_failure_jump_loop: |
dc4a2ee0 PE |
1018 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1019 | fprintf (stderr, "/on_failure_jump_loop to %td", | |
1020 | p + mcnt - start); | |
5e69f11e RM |
1021 | break; |
1022 | ||
505bde11 | 1023 | case on_failure_jump_smart: |
dc4a2ee0 PE |
1024 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1025 | fprintf (stderr, "/on_failure_jump_smart to %td", | |
1026 | p + mcnt - start); | |
5e69f11e RM |
1027 | break; |
1028 | ||
25fe55af | 1029 | case jump: |
dc4a2ee0 PE |
1030 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1031 | fprintf (stderr, "/jump to %td", p + mcnt - start); | |
fa9a63c5 RM |
1032 | break; |
1033 | ||
25fe55af | 1034 | case succeed_n: |
dc4a2ee0 PE |
1035 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1036 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1037 | fprintf (stderr, "/succeed_n to %td, %d times", | |
1038 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1039 | break; |
5e69f11e | 1040 | |
25fe55af | 1041 | case jump_n: |
dc4a2ee0 PE |
1042 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1043 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1044 | fprintf (stderr, "/jump_n to %td, %d times", | |
1045 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1046 | break; |
5e69f11e | 1047 | |
25fe55af | 1048 | case set_number_at: |
dc4a2ee0 PE |
1049 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1050 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1051 | fprintf (stderr, "/set_number_at location %td to %d", | |
1052 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1053 | break; |
5e69f11e | 1054 | |
25fe55af | 1055 | case wordbound: |
a1a052df | 1056 | fprintf (stderr, "/wordbound"); |
fa9a63c5 RM |
1057 | break; |
1058 | ||
1059 | case notwordbound: | |
a1a052df | 1060 | fprintf (stderr, "/notwordbound"); |
25fe55af | 1061 | break; |
fa9a63c5 RM |
1062 | |
1063 | case wordbeg: | |
a1a052df | 1064 | fprintf (stderr, "/wordbeg"); |
fa9a63c5 | 1065 | break; |
5e69f11e | 1066 | |
fa9a63c5 | 1067 | case wordend: |
a1a052df | 1068 | fprintf (stderr, "/wordend"); |
e2543b02 | 1069 | break; |
5e69f11e | 1070 | |
669fa600 | 1071 | case symbeg: |
e2543b02 | 1072 | fprintf (stderr, "/symbeg"); |
669fa600 SM |
1073 | break; |
1074 | ||
1075 | case symend: | |
e2543b02 | 1076 | fprintf (stderr, "/symend"); |
669fa600 | 1077 | break; |
5e69f11e | 1078 | |
1fb352e0 | 1079 | case syntaxspec: |
a1a052df | 1080 | fprintf (stderr, "/syntaxspec"); |
1fb352e0 | 1081 | mcnt = *p++; |
a1a052df | 1082 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1083 | break; |
1084 | ||
1085 | case notsyntaxspec: | |
a1a052df | 1086 | fprintf (stderr, "/notsyntaxspec"); |
1fb352e0 | 1087 | mcnt = *p++; |
a1a052df | 1088 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1089 | break; |
1090 | ||
0b32bf0e | 1091 | # ifdef emacs |
fa9a63c5 | 1092 | case before_dot: |
a1a052df | 1093 | fprintf (stderr, "/before_dot"); |
25fe55af | 1094 | break; |
fa9a63c5 RM |
1095 | |
1096 | case at_dot: | |
a1a052df | 1097 | fprintf (stderr, "/at_dot"); |
25fe55af | 1098 | break; |
fa9a63c5 RM |
1099 | |
1100 | case after_dot: | |
a1a052df | 1101 | fprintf (stderr, "/after_dot"); |
25fe55af | 1102 | break; |
fa9a63c5 | 1103 | |
1fb352e0 | 1104 | case categoryspec: |
a1a052df | 1105 | fprintf (stderr, "/categoryspec"); |
fa9a63c5 | 1106 | mcnt = *p++; |
a1a052df | 1107 | fprintf (stderr, "/%d", mcnt); |
25fe55af | 1108 | break; |
5e69f11e | 1109 | |
1fb352e0 | 1110 | case notcategoryspec: |
a1a052df | 1111 | fprintf (stderr, "/notcategoryspec"); |
fa9a63c5 | 1112 | mcnt = *p++; |
a1a052df | 1113 | fprintf (stderr, "/%d", mcnt); |
fa9a63c5 | 1114 | break; |
0b32bf0e | 1115 | # endif /* emacs */ |
fa9a63c5 | 1116 | |
fa9a63c5 | 1117 | case begbuf: |
a1a052df | 1118 | fprintf (stderr, "/begbuf"); |
25fe55af | 1119 | break; |
fa9a63c5 RM |
1120 | |
1121 | case endbuf: | |
a1a052df | 1122 | fprintf (stderr, "/endbuf"); |
25fe55af | 1123 | break; |
fa9a63c5 | 1124 | |
25fe55af | 1125 | default: |
a1a052df | 1126 | fprintf (stderr, "?%d", *(p-1)); |
fa9a63c5 RM |
1127 | } |
1128 | ||
a1a052df | 1129 | fprintf (stderr, "\n"); |
fa9a63c5 RM |
1130 | } |
1131 | ||
dc4a2ee0 | 1132 | fprintf (stderr, "%td:\tend of pattern.\n", p - start); |
fa9a63c5 RM |
1133 | } |
1134 | ||
1135 | ||
dc4a2ee0 PE |
1136 | static void |
1137 | print_compiled_pattern (struct re_pattern_buffer *bufp) | |
fa9a63c5 | 1138 | { |
01618498 | 1139 | re_char *buffer = bufp->buffer; |
fa9a63c5 RM |
1140 | |
1141 | print_partial_compiled_pattern (buffer, buffer + bufp->used); | |
4bb91c68 SM |
1142 | printf ("%ld bytes used/%ld bytes allocated.\n", |
1143 | bufp->used, bufp->allocated); | |
fa9a63c5 RM |
1144 | |
1145 | if (bufp->fastmap_accurate && bufp->fastmap) | |
1146 | { | |
1147 | printf ("fastmap: "); | |
1148 | print_fastmap (bufp->fastmap); | |
1149 | } | |
1150 | ||
dc4a2ee0 | 1151 | printf ("re_nsub: %zu\t", bufp->re_nsub); |
fa9a63c5 RM |
1152 | printf ("regs_alloc: %d\t", bufp->regs_allocated); |
1153 | printf ("can_be_null: %d\t", bufp->can_be_null); | |
fa9a63c5 RM |
1154 | printf ("no_sub: %d\t", bufp->no_sub); |
1155 | printf ("not_bol: %d\t", bufp->not_bol); | |
1156 | printf ("not_eol: %d\t", bufp->not_eol); | |
4bb91c68 | 1157 | printf ("syntax: %lx\n", bufp->syntax); |
505bde11 | 1158 | fflush (stdout); |
fa9a63c5 RM |
1159 | /* Perhaps we should print the translate table? */ |
1160 | } | |
1161 | ||
1162 | ||
dc4a2ee0 PE |
1163 | static void |
1164 | print_double_string (re_char *where, re_char *string1, ssize_t size1, | |
1165 | re_char *string2, ssize_t size2) | |
fa9a63c5 | 1166 | { |
d1dfb56c | 1167 | ssize_t this_char; |
5e69f11e | 1168 | |
fa9a63c5 RM |
1169 | if (where == NULL) |
1170 | printf ("(null)"); | |
1171 | else | |
1172 | { | |
1173 | if (FIRST_STRING_P (where)) | |
25fe55af RS |
1174 | { |
1175 | for (this_char = where - string1; this_char < size1; this_char++) | |
1176 | putchar (string1[this_char]); | |
fa9a63c5 | 1177 | |
25fe55af RS |
1178 | where = string2; |
1179 | } | |
fa9a63c5 RM |
1180 | |
1181 | for (this_char = where - string2; this_char < size2; this_char++) | |
25fe55af | 1182 | putchar (string2[this_char]); |
fa9a63c5 RM |
1183 | } |
1184 | } | |
1185 | ||
1186 | #else /* not DEBUG */ | |
1187 | ||
0b32bf0e SM |
1188 | # undef assert |
1189 | # define assert(e) | |
fa9a63c5 | 1190 | |
0b32bf0e | 1191 | # define DEBUG_STATEMENT(e) |
dc4a2ee0 PE |
1192 | # if __STDC_VERSION__ < 199901L |
1193 | # define DEBUG_COMPILES_ARGUMENTS | |
1194 | # define DEBUG_PRINT /* 'DEBUG_PRINT (x, y)' discards X and Y. */ (void) | |
1195 | # else | |
1196 | # define DEBUG_PRINT(...) | |
1197 | # endif | |
0b32bf0e SM |
1198 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) |
1199 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
fa9a63c5 RM |
1200 | |
1201 | #endif /* not DEBUG */ | |
1202 | \f | |
4da60324 PE |
1203 | /* Use this to suppress gcc's `...may be used before initialized' warnings. */ |
1204 | #ifdef lint | |
1205 | # define IF_LINT(Code) Code | |
1206 | #else | |
1207 | # define IF_LINT(Code) /* empty */ | |
1208 | #endif | |
1209 | \f | |
fa9a63c5 RM |
1210 | /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can |
1211 | also be assigned to arbitrarily: each pattern buffer stores its own | |
1212 | syntax, so it can be changed between regex compilations. */ | |
1213 | /* This has no initializer because initialized variables in Emacs | |
1214 | become read-only after dumping. */ | |
1215 | reg_syntax_t re_syntax_options; | |
1216 | ||
1217 | ||
1218 | /* Specify the precise syntax of regexps for compilation. This provides | |
1219 | for compatibility for various utilities which historically have | |
1220 | different, incompatible syntaxes. | |
1221 | ||
1222 | The argument SYNTAX is a bit mask comprised of the various bits | |
4bb91c68 | 1223 | defined in regex.h. We return the old syntax. */ |
fa9a63c5 RM |
1224 | |
1225 | reg_syntax_t | |
971de7fb | 1226 | re_set_syntax (reg_syntax_t syntax) |
fa9a63c5 RM |
1227 | { |
1228 | reg_syntax_t ret = re_syntax_options; | |
5e69f11e | 1229 | |
fa9a63c5 RM |
1230 | re_syntax_options = syntax; |
1231 | return ret; | |
1232 | } | |
c0f9ea08 | 1233 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) |
f9b0fd99 RS |
1234 | |
1235 | /* Regexp to use to replace spaces, or NULL meaning don't. */ | |
f462f075 | 1236 | static const_re_char *whitespace_regexp; |
f9b0fd99 RS |
1237 | |
1238 | void | |
971de7fb | 1239 | re_set_whitespace_regexp (const char *regexp) |
f9b0fd99 | 1240 | { |
f462f075 | 1241 | whitespace_regexp = (const_re_char *) regexp; |
f9b0fd99 RS |
1242 | } |
1243 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) | |
fa9a63c5 RM |
1244 | \f |
1245 | /* This table gives an error message for each of the error codes listed | |
4bb91c68 | 1246 | in regex.h. Obviously the order here has to be same as there. |
fa9a63c5 | 1247 | POSIX doesn't require that we do anything for REG_NOERROR, |
4bb91c68 | 1248 | but why not be nice? */ |
fa9a63c5 RM |
1249 | |
1250 | static const char *re_error_msgid[] = | |
5e69f11e RM |
1251 | { |
1252 | gettext_noop ("Success"), /* REG_NOERROR */ | |
1253 | gettext_noop ("No match"), /* REG_NOMATCH */ | |
1254 | gettext_noop ("Invalid regular expression"), /* REG_BADPAT */ | |
1255 | gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */ | |
1256 | gettext_noop ("Invalid character class name"), /* REG_ECTYPE */ | |
1257 | gettext_noop ("Trailing backslash"), /* REG_EESCAPE */ | |
1258 | gettext_noop ("Invalid back reference"), /* REG_ESUBREG */ | |
1259 | gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */ | |
1260 | gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */ | |
1261 | gettext_noop ("Unmatched \\{"), /* REG_EBRACE */ | |
1262 | gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */ | |
1263 | gettext_noop ("Invalid range end"), /* REG_ERANGE */ | |
1264 | gettext_noop ("Memory exhausted"), /* REG_ESPACE */ | |
1265 | gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */ | |
1266 | gettext_noop ("Premature end of regular expression"), /* REG_EEND */ | |
1267 | gettext_noop ("Regular expression too big"), /* REG_ESIZE */ | |
1268 | gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */ | |
b3e4c897 | 1269 | gettext_noop ("Range striding over charsets") /* REG_ERANGEX */ |
fa9a63c5 RM |
1270 | }; |
1271 | \f | |
4bb91c68 | 1272 | /* Avoiding alloca during matching, to placate r_alloc. */ |
fa9a63c5 RM |
1273 | |
1274 | /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the | |
1275 | searching and matching functions should not call alloca. On some | |
1276 | systems, alloca is implemented in terms of malloc, and if we're | |
1277 | using the relocating allocator routines, then malloc could cause a | |
1278 | relocation, which might (if the strings being searched are in the | |
1279 | ralloc heap) shift the data out from underneath the regexp | |
1280 | routines. | |
1281 | ||
5e69f11e | 1282 | Here's another reason to avoid allocation: Emacs |
fa9a63c5 RM |
1283 | processes input from X in a signal handler; processing X input may |
1284 | call malloc; if input arrives while a matching routine is calling | |
1285 | malloc, then we're scrod. But Emacs can't just block input while | |
1286 | calling matching routines; then we don't notice interrupts when | |
1287 | they come in. So, Emacs blocks input around all regexp calls | |
1288 | except the matching calls, which it leaves unprotected, in the | |
1289 | faith that they will not malloc. */ | |
1290 | ||
1291 | /* Normally, this is fine. */ | |
1292 | #define MATCH_MAY_ALLOCATE | |
1293 | ||
fa9a63c5 RM |
1294 | /* The match routines may not allocate if (1) they would do it with malloc |
1295 | and (2) it's not safe for them to use malloc. | |
1296 | Note that if REL_ALLOC is defined, matching would not use malloc for the | |
1297 | failure stack, but we would still use it for the register vectors; | |
4bb91c68 | 1298 | so REL_ALLOC should not affect this. */ |
b588157e | 1299 | #if defined REGEX_MALLOC && defined emacs |
0b32bf0e | 1300 | # undef MATCH_MAY_ALLOCATE |
fa9a63c5 RM |
1301 | #endif |
1302 | ||
1303 | \f | |
1304 | /* Failure stack declarations and macros; both re_compile_fastmap and | |
1305 | re_match_2 use a failure stack. These have to be macros because of | |
1306 | REGEX_ALLOCATE_STACK. */ | |
5e69f11e | 1307 | |
fa9a63c5 | 1308 | |
320a2a73 | 1309 | /* Approximate number of failure points for which to initially allocate space |
fa9a63c5 RM |
1310 | when matching. If this number is exceeded, we allocate more |
1311 | space, so it is not a hard limit. */ | |
1312 | #ifndef INIT_FAILURE_ALLOC | |
0b32bf0e | 1313 | # define INIT_FAILURE_ALLOC 20 |
fa9a63c5 RM |
1314 | #endif |
1315 | ||
1316 | /* Roughly the maximum number of failure points on the stack. Would be | |
320a2a73 | 1317 | exactly that if always used TYPICAL_FAILURE_SIZE items each time we failed. |
fa9a63c5 | 1318 | This is a variable only so users of regex can assign to it; we never |
ada30c0e SM |
1319 | change it ourselves. We always multiply it by TYPICAL_FAILURE_SIZE |
1320 | before using it, so it should probably be a byte-count instead. */ | |
c0f9ea08 SM |
1321 | # if defined MATCH_MAY_ALLOCATE |
1322 | /* Note that 4400 was enough to cause a crash on Alpha OSF/1, | |
320a2a73 KH |
1323 | whose default stack limit is 2mb. In order for a larger |
1324 | value to work reliably, you have to try to make it accord | |
1325 | with the process stack limit. */ | |
c0f9ea08 SM |
1326 | size_t re_max_failures = 40000; |
1327 | # else | |
1328 | size_t re_max_failures = 4000; | |
1329 | # endif | |
fa9a63c5 RM |
1330 | |
1331 | union fail_stack_elt | |
1332 | { | |
01618498 | 1333 | re_char *pointer; |
c0f9ea08 SM |
1334 | /* This should be the biggest `int' that's no bigger than a pointer. */ |
1335 | long integer; | |
fa9a63c5 RM |
1336 | }; |
1337 | ||
1338 | typedef union fail_stack_elt fail_stack_elt_t; | |
1339 | ||
1340 | typedef struct | |
1341 | { | |
1342 | fail_stack_elt_t *stack; | |
c0f9ea08 SM |
1343 | size_t size; |
1344 | size_t avail; /* Offset of next open position. */ | |
1345 | size_t frame; /* Offset of the cur constructed frame. */ | |
fa9a63c5 RM |
1346 | } fail_stack_type; |
1347 | ||
505bde11 | 1348 | #define FAIL_STACK_EMPTY() (fail_stack.frame == 0) |
fa9a63c5 RM |
1349 | |
1350 | ||
1351 | /* Define macros to initialize and free the failure stack. | |
1352 | Do `return -2' if the alloc fails. */ | |
1353 | ||
1354 | #ifdef MATCH_MAY_ALLOCATE | |
0b32bf0e | 1355 | # define INIT_FAIL_STACK() \ |
fa9a63c5 | 1356 | do { \ |
38182d90 | 1357 | fail_stack.stack = \ |
320a2a73 KH |
1358 | REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * TYPICAL_FAILURE_SIZE \ |
1359 | * sizeof (fail_stack_elt_t)); \ | |
fa9a63c5 RM |
1360 | \ |
1361 | if (fail_stack.stack == NULL) \ | |
1362 | return -2; \ | |
1363 | \ | |
1364 | fail_stack.size = INIT_FAILURE_ALLOC; \ | |
1365 | fail_stack.avail = 0; \ | |
505bde11 | 1366 | fail_stack.frame = 0; \ |
fa9a63c5 | 1367 | } while (0) |
fa9a63c5 | 1368 | #else |
0b32bf0e | 1369 | # define INIT_FAIL_STACK() \ |
fa9a63c5 RM |
1370 | do { \ |
1371 | fail_stack.avail = 0; \ | |
505bde11 | 1372 | fail_stack.frame = 0; \ |
fa9a63c5 RM |
1373 | } while (0) |
1374 | ||
b313f9d8 PE |
1375 | # define RETALLOC_IF(addr, n, t) \ |
1376 | if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t) | |
fa9a63c5 RM |
1377 | #endif |
1378 | ||
1379 | ||
320a2a73 KH |
1380 | /* Double the size of FAIL_STACK, up to a limit |
1381 | which allows approximately `re_max_failures' items. | |
fa9a63c5 RM |
1382 | |
1383 | Return 1 if succeeds, and 0 if either ran out of memory | |
5e69f11e RM |
1384 | allocating space for it or it was already too large. |
1385 | ||
4bb91c68 | 1386 | REGEX_REALLOCATE_STACK requires `destination' be declared. */ |
fa9a63c5 | 1387 | |
320a2a73 KH |
1388 | /* Factor to increase the failure stack size by |
1389 | when we increase it. | |
1390 | This used to be 2, but 2 was too wasteful | |
1391 | because the old discarded stacks added up to as much space | |
1392 | were as ultimate, maximum-size stack. */ | |
1393 | #define FAIL_STACK_GROWTH_FACTOR 4 | |
1394 | ||
1395 | #define GROW_FAIL_STACK(fail_stack) \ | |
eead07d6 KH |
1396 | (((fail_stack).size * sizeof (fail_stack_elt_t) \ |
1397 | >= re_max_failures * TYPICAL_FAILURE_SIZE) \ | |
fa9a63c5 | 1398 | ? 0 \ |
320a2a73 | 1399 | : ((fail_stack).stack \ |
38182d90 | 1400 | = REGEX_REALLOCATE_STACK ((fail_stack).stack, \ |
25fe55af | 1401 | (fail_stack).size * sizeof (fail_stack_elt_t), \ |
320a2a73 KH |
1402 | MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ |
1403 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1404 | * FAIL_STACK_GROWTH_FACTOR))), \ | |
fa9a63c5 RM |
1405 | \ |
1406 | (fail_stack).stack == NULL \ | |
1407 | ? 0 \ | |
6453db45 KH |
1408 | : ((fail_stack).size \ |
1409 | = (MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ | |
1410 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1411 | * FAIL_STACK_GROWTH_FACTOR)) \ | |
1412 | / sizeof (fail_stack_elt_t)), \ | |
25fe55af | 1413 | 1))) |
fa9a63c5 RM |
1414 | |
1415 | ||
fa9a63c5 RM |
1416 | /* Push a pointer value onto the failure stack. |
1417 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1418 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 | 1419 | #define PUSH_FAILURE_POINTER(item) \ |
01618498 | 1420 | fail_stack.stack[fail_stack.avail++].pointer = (item) |
fa9a63c5 RM |
1421 | |
1422 | /* This pushes an integer-valued item onto the failure stack. | |
1423 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1424 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 RM |
1425 | #define PUSH_FAILURE_INT(item) \ |
1426 | fail_stack.stack[fail_stack.avail++].integer = (item) | |
1427 | ||
b313f9d8 | 1428 | /* These POP... operations complement the PUSH... operations. |
fa9a63c5 RM |
1429 | All assume that `fail_stack' is nonempty. */ |
1430 | #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer | |
1431 | #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer | |
fa9a63c5 | 1432 | |
505bde11 SM |
1433 | /* Individual items aside from the registers. */ |
1434 | #define NUM_NONREG_ITEMS 3 | |
1435 | ||
1436 | /* Used to examine the stack (to detect infinite loops). */ | |
1437 | #define FAILURE_PAT(h) fail_stack.stack[(h) - 1].pointer | |
66f0296e | 1438 | #define FAILURE_STR(h) (fail_stack.stack[(h) - 2].pointer) |
505bde11 SM |
1439 | #define NEXT_FAILURE_HANDLE(h) fail_stack.stack[(h) - 3].integer |
1440 | #define TOP_FAILURE_HANDLE() fail_stack.frame | |
fa9a63c5 RM |
1441 | |
1442 | ||
505bde11 SM |
1443 | #define ENSURE_FAIL_STACK(space) \ |
1444 | while (REMAINING_AVAIL_SLOTS <= space) { \ | |
1445 | if (!GROW_FAIL_STACK (fail_stack)) \ | |
1446 | return -2; \ | |
dc4a2ee0 PE |
1447 | DEBUG_PRINT ("\n Doubled stack; size now: %zd\n", (fail_stack).size);\ |
1448 | DEBUG_PRINT (" slots available: %zd\n", REMAINING_AVAIL_SLOTS);\ | |
505bde11 SM |
1449 | } |
1450 | ||
1451 | /* Push register NUM onto the stack. */ | |
1452 | #define PUSH_FAILURE_REG(num) \ | |
1453 | do { \ | |
1454 | char *destination; \ | |
dc4a2ee0 | 1455 | long n = num; \ |
505bde11 | 1456 | ENSURE_FAIL_STACK(3); \ |
dc4a2ee0 PE |
1457 | DEBUG_PRINT (" Push reg %ld (spanning %p -> %p)\n", \ |
1458 | n, regstart[n], regend[n]); \ | |
1459 | PUSH_FAILURE_POINTER (regstart[n]); \ | |
1460 | PUSH_FAILURE_POINTER (regend[n]); \ | |
1461 | PUSH_FAILURE_INT (n); \ | |
505bde11 SM |
1462 | } while (0) |
1463 | ||
01618498 SM |
1464 | /* Change the counter's value to VAL, but make sure that it will |
1465 | be reset when backtracking. */ | |
1466 | #define PUSH_NUMBER(ptr,val) \ | |
dc1e502d SM |
1467 | do { \ |
1468 | char *destination; \ | |
1469 | int c; \ | |
1470 | ENSURE_FAIL_STACK(3); \ | |
1471 | EXTRACT_NUMBER (c, ptr); \ | |
dc4a2ee0 | 1472 | DEBUG_PRINT (" Push number %p = %d -> %d\n", ptr, c, val); \ |
dc1e502d SM |
1473 | PUSH_FAILURE_INT (c); \ |
1474 | PUSH_FAILURE_POINTER (ptr); \ | |
1475 | PUSH_FAILURE_INT (-1); \ | |
01618498 | 1476 | STORE_NUMBER (ptr, val); \ |
dc1e502d SM |
1477 | } while (0) |
1478 | ||
505bde11 | 1479 | /* Pop a saved register off the stack. */ |
dc1e502d | 1480 | #define POP_FAILURE_REG_OR_COUNT() \ |
505bde11 | 1481 | do { \ |
d1dfb56c | 1482 | long pfreg = POP_FAILURE_INT (); \ |
19ed5445 | 1483 | if (pfreg == -1) \ |
dc1e502d SM |
1484 | { \ |
1485 | /* It's a counter. */ \ | |
6dcf2d0e SM |
1486 | /* Here, we discard `const', making re_match non-reentrant. */ \ |
1487 | unsigned char *ptr = (unsigned char*) POP_FAILURE_POINTER (); \ | |
19ed5445 PE |
1488 | pfreg = POP_FAILURE_INT (); \ |
1489 | STORE_NUMBER (ptr, pfreg); \ | |
dc4a2ee0 | 1490 | DEBUG_PRINT (" Pop counter %p = %ld\n", ptr, pfreg); \ |
dc1e502d SM |
1491 | } \ |
1492 | else \ | |
1493 | { \ | |
19ed5445 PE |
1494 | regend[pfreg] = POP_FAILURE_POINTER (); \ |
1495 | regstart[pfreg] = POP_FAILURE_POINTER (); \ | |
dc4a2ee0 PE |
1496 | DEBUG_PRINT (" Pop reg %ld (spanning %p -> %p)\n", \ |
1497 | pfreg, regstart[pfreg], regend[pfreg]); \ | |
dc1e502d | 1498 | } \ |
505bde11 SM |
1499 | } while (0) |
1500 | ||
1501 | /* Check that we are not stuck in an infinite loop. */ | |
1502 | #define CHECK_INFINITE_LOOP(pat_cur, string_place) \ | |
1503 | do { \ | |
d1dfb56c | 1504 | ssize_t failure = TOP_FAILURE_HANDLE (); \ |
505bde11 | 1505 | /* Check for infinite matching loops */ \ |
f6df485f RS |
1506 | while (failure > 0 \ |
1507 | && (FAILURE_STR (failure) == string_place \ | |
1508 | || FAILURE_STR (failure) == NULL)) \ | |
505bde11 SM |
1509 | { \ |
1510 | assert (FAILURE_PAT (failure) >= bufp->buffer \ | |
66f0296e | 1511 | && FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \ |
505bde11 | 1512 | if (FAILURE_PAT (failure) == pat_cur) \ |
f6df485f | 1513 | { \ |
6df42991 SM |
1514 | cycle = 1; \ |
1515 | break; \ | |
f6df485f | 1516 | } \ |
dc4a2ee0 | 1517 | DEBUG_PRINT (" Other pattern: %p\n", FAILURE_PAT (failure)); \ |
505bde11 SM |
1518 | failure = NEXT_FAILURE_HANDLE(failure); \ |
1519 | } \ | |
dc4a2ee0 | 1520 | DEBUG_PRINT (" Other string: %p\n", FAILURE_STR (failure)); \ |
505bde11 | 1521 | } while (0) |
6df42991 | 1522 | |
fa9a63c5 | 1523 | /* Push the information about the state we will need |
5e69f11e RM |
1524 | if we ever fail back to it. |
1525 | ||
505bde11 | 1526 | Requires variables fail_stack, regstart, regend and |
320a2a73 | 1527 | num_regs be declared. GROW_FAIL_STACK requires `destination' be |
fa9a63c5 | 1528 | declared. |
5e69f11e | 1529 | |
fa9a63c5 RM |
1530 | Does `return FAILURE_CODE' if runs out of memory. */ |
1531 | ||
505bde11 SM |
1532 | #define PUSH_FAILURE_POINT(pattern, string_place) \ |
1533 | do { \ | |
1534 | char *destination; \ | |
1535 | /* Must be int, so when we don't save any registers, the arithmetic \ | |
1536 | of 0 + -1 isn't done as unsigned. */ \ | |
1537 | \ | |
505bde11 | 1538 | DEBUG_STATEMENT (nfailure_points_pushed++); \ |
dc4a2ee0 PE |
1539 | DEBUG_PRINT ("\nPUSH_FAILURE_POINT:\n"); \ |
1540 | DEBUG_PRINT (" Before push, next avail: %zd\n", (fail_stack).avail); \ | |
1541 | DEBUG_PRINT (" size: %zd\n", (fail_stack).size);\ | |
505bde11 SM |
1542 | \ |
1543 | ENSURE_FAIL_STACK (NUM_NONREG_ITEMS); \ | |
1544 | \ | |
dc4a2ee0 | 1545 | DEBUG_PRINT ("\n"); \ |
505bde11 | 1546 | \ |
dc4a2ee0 | 1547 | DEBUG_PRINT (" Push frame index: %zd\n", fail_stack.frame); \ |
505bde11 SM |
1548 | PUSH_FAILURE_INT (fail_stack.frame); \ |
1549 | \ | |
dc4a2ee0 | 1550 | DEBUG_PRINT (" Push string %p: `", string_place); \ |
505bde11 | 1551 | DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, size2);\ |
dc4a2ee0 | 1552 | DEBUG_PRINT ("'\n"); \ |
505bde11 SM |
1553 | PUSH_FAILURE_POINTER (string_place); \ |
1554 | \ | |
dc4a2ee0 | 1555 | DEBUG_PRINT (" Push pattern %p: ", pattern); \ |
505bde11 SM |
1556 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern, pend); \ |
1557 | PUSH_FAILURE_POINTER (pattern); \ | |
1558 | \ | |
1559 | /* Close the frame by moving the frame pointer past it. */ \ | |
1560 | fail_stack.frame = fail_stack.avail; \ | |
1561 | } while (0) | |
fa9a63c5 | 1562 | |
320a2a73 KH |
1563 | /* Estimate the size of data pushed by a typical failure stack entry. |
1564 | An estimate is all we need, because all we use this for | |
1565 | is to choose a limit for how big to make the failure stack. */ | |
ada30c0e | 1566 | /* BEWARE, the value `20' is hard-coded in emacs.c:main(). */ |
320a2a73 | 1567 | #define TYPICAL_FAILURE_SIZE 20 |
fa9a63c5 | 1568 | |
fa9a63c5 RM |
1569 | /* How many items can still be added to the stack without overflowing it. */ |
1570 | #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) | |
1571 | ||
1572 | ||
1573 | /* Pops what PUSH_FAIL_STACK pushes. | |
1574 | ||
1575 | We restore into the parameters, all of which should be lvalues: | |
1576 | STR -- the saved data position. | |
1577 | PAT -- the saved pattern position. | |
fa9a63c5 | 1578 | REGSTART, REGEND -- arrays of string positions. |
5e69f11e | 1579 | |
fa9a63c5 | 1580 | Also assumes the variables `fail_stack' and (if debugging), `bufp', |
7814e705 | 1581 | `pend', `string1', `size1', `string2', and `size2'. */ |
fa9a63c5 | 1582 | |
505bde11 SM |
1583 | #define POP_FAILURE_POINT(str, pat) \ |
1584 | do { \ | |
fa9a63c5 RM |
1585 | assert (!FAIL_STACK_EMPTY ()); \ |
1586 | \ | |
1587 | /* Remove failure points and point to how many regs pushed. */ \ | |
dc4a2ee0 PE |
1588 | DEBUG_PRINT ("POP_FAILURE_POINT:\n"); \ |
1589 | DEBUG_PRINT (" Before pop, next avail: %zd\n", fail_stack.avail); \ | |
1590 | DEBUG_PRINT (" size: %zd\n", fail_stack.size); \ | |
fa9a63c5 | 1591 | \ |
505bde11 SM |
1592 | /* Pop the saved registers. */ \ |
1593 | while (fail_stack.frame < fail_stack.avail) \ | |
dc1e502d | 1594 | POP_FAILURE_REG_OR_COUNT (); \ |
fa9a63c5 | 1595 | \ |
dc4a2ee0 PE |
1596 | pat = POP_FAILURE_POINTER (); \ |
1597 | DEBUG_PRINT (" Popping pattern %p: ", pat); \ | |
505bde11 | 1598 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ |
fa9a63c5 RM |
1599 | \ |
1600 | /* If the saved string location is NULL, it came from an \ | |
1601 | on_failure_keep_string_jump opcode, and we want to throw away the \ | |
1602 | saved NULL, thus retaining our current position in the string. */ \ | |
01618498 | 1603 | str = POP_FAILURE_POINTER (); \ |
dc4a2ee0 | 1604 | DEBUG_PRINT (" Popping string %p: `", str); \ |
fa9a63c5 | 1605 | DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ |
dc4a2ee0 | 1606 | DEBUG_PRINT ("'\n"); \ |
fa9a63c5 | 1607 | \ |
505bde11 | 1608 | fail_stack.frame = POP_FAILURE_INT (); \ |
dc4a2ee0 | 1609 | DEBUG_PRINT (" Popping frame index: %zd\n", fail_stack.frame); \ |
fa9a63c5 | 1610 | \ |
505bde11 SM |
1611 | assert (fail_stack.avail >= 0); \ |
1612 | assert (fail_stack.frame <= fail_stack.avail); \ | |
fa9a63c5 | 1613 | \ |
fa9a63c5 | 1614 | DEBUG_STATEMENT (nfailure_points_popped++); \ |
505bde11 | 1615 | } while (0) /* POP_FAILURE_POINT */ |
fa9a63c5 RM |
1616 | |
1617 | ||
1618 | \f | |
fa9a63c5 | 1619 | /* Registers are set to a sentinel when they haven't yet matched. */ |
4bb91c68 | 1620 | #define REG_UNSET(e) ((e) == NULL) |
fa9a63c5 RM |
1621 | \f |
1622 | /* Subroutine declarations and macros for regex_compile. */ | |
1623 | ||
261cb4bb PE |
1624 | static reg_errcode_t regex_compile (re_char *pattern, size_t size, |
1625 | reg_syntax_t syntax, | |
1626 | struct re_pattern_buffer *bufp); | |
1627 | static void store_op1 (re_opcode_t op, unsigned char *loc, int arg); | |
1628 | static void store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2); | |
1629 | static void insert_op1 (re_opcode_t op, unsigned char *loc, | |
1630 | int arg, unsigned char *end); | |
1631 | static void insert_op2 (re_opcode_t op, unsigned char *loc, | |
1632 | int arg1, int arg2, unsigned char *end); | |
1633 | static boolean at_begline_loc_p (re_char *pattern, re_char *p, | |
1634 | reg_syntax_t syntax); | |
1635 | static boolean at_endline_loc_p (re_char *p, re_char *pend, | |
1636 | reg_syntax_t syntax); | |
1637 | static re_char *skip_one_char (re_char *p); | |
1638 | static int analyse_first (re_char *p, re_char *pend, | |
1639 | char *fastmap, const int multibyte); | |
fa9a63c5 | 1640 | |
fa9a63c5 | 1641 | /* Fetch the next character in the uncompiled pattern, with no |
4bb91c68 | 1642 | translation. */ |
36595814 | 1643 | #define PATFETCH(c) \ |
2d1675e4 SM |
1644 | do { \ |
1645 | int len; \ | |
1646 | if (p == pend) return REG_EEND; \ | |
62a6e103 | 1647 | c = RE_STRING_CHAR_AND_LENGTH (p, len, multibyte); \ |
2d1675e4 | 1648 | p += len; \ |
fa9a63c5 RM |
1649 | } while (0) |
1650 | ||
fa9a63c5 RM |
1651 | |
1652 | /* If `translate' is non-null, return translate[D], else just D. We | |
1653 | cast the subscript to translate because some data is declared as | |
1654 | `char *', to avoid warnings when a string constant is passed. But | |
1655 | when we use a character as a subscript we must make it unsigned. */ | |
6676cb1c | 1656 | #ifndef TRANSLATE |
0b32bf0e | 1657 | # define TRANSLATE(d) \ |
66f0296e | 1658 | (RE_TRANSLATE_P (translate) ? RE_TRANSLATE (translate, (d)) : (d)) |
6676cb1c | 1659 | #endif |
fa9a63c5 RM |
1660 | |
1661 | ||
1662 | /* Macros for outputting the compiled pattern into `buffer'. */ | |
1663 | ||
1664 | /* If the buffer isn't allocated when it comes in, use this. */ | |
1665 | #define INIT_BUF_SIZE 32 | |
1666 | ||
4bb91c68 | 1667 | /* Make sure we have at least N more bytes of space in buffer. */ |
fa9a63c5 | 1668 | #define GET_BUFFER_SPACE(n) \ |
01618498 | 1669 | while ((size_t) (b - bufp->buffer + (n)) > bufp->allocated) \ |
fa9a63c5 RM |
1670 | EXTEND_BUFFER () |
1671 | ||
1672 | /* Make sure we have one more byte of buffer space and then add C to it. */ | |
1673 | #define BUF_PUSH(c) \ | |
1674 | do { \ | |
1675 | GET_BUFFER_SPACE (1); \ | |
1676 | *b++ = (unsigned char) (c); \ | |
1677 | } while (0) | |
1678 | ||
1679 | ||
1680 | /* Ensure we have two more bytes of buffer space and then append C1 and C2. */ | |
1681 | #define BUF_PUSH_2(c1, c2) \ | |
1682 | do { \ | |
1683 | GET_BUFFER_SPACE (2); \ | |
1684 | *b++ = (unsigned char) (c1); \ | |
1685 | *b++ = (unsigned char) (c2); \ | |
1686 | } while (0) | |
1687 | ||
1688 | ||
fa9a63c5 | 1689 | /* Store a jump with opcode OP at LOC to location TO. We store a |
4bb91c68 | 1690 | relative address offset by the three bytes the jump itself occupies. */ |
fa9a63c5 RM |
1691 | #define STORE_JUMP(op, loc, to) \ |
1692 | store_op1 (op, loc, (to) - (loc) - 3) | |
1693 | ||
1694 | /* Likewise, for a two-argument jump. */ | |
1695 | #define STORE_JUMP2(op, loc, to, arg) \ | |
1696 | store_op2 (op, loc, (to) - (loc) - 3, arg) | |
1697 | ||
4bb91c68 | 1698 | /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ |
fa9a63c5 RM |
1699 | #define INSERT_JUMP(op, loc, to) \ |
1700 | insert_op1 (op, loc, (to) - (loc) - 3, b) | |
1701 | ||
1702 | /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ | |
1703 | #define INSERT_JUMP2(op, loc, to, arg) \ | |
1704 | insert_op2 (op, loc, (to) - (loc) - 3, arg, b) | |
1705 | ||
1706 | ||
1707 | /* This is not an arbitrary limit: the arguments which represent offsets | |
839966f3 | 1708 | into the pattern are two bytes long. So if 2^15 bytes turns out to |
fa9a63c5 | 1709 | be too small, many things would have to change. */ |
839966f3 KH |
1710 | # define MAX_BUF_SIZE (1L << 15) |
1711 | ||
fa9a63c5 RM |
1712 | /* Extend the buffer by twice its current size via realloc and |
1713 | reset the pointers that pointed into the old block to point to the | |
1714 | correct places in the new one. If extending the buffer results in it | |
4bb91c68 SM |
1715 | being larger than MAX_BUF_SIZE, then flag memory exhausted. */ |
1716 | #if __BOUNDED_POINTERS__ | |
1717 | # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) | |
381880b0 CY |
1718 | # define MOVE_BUFFER_POINTER(P) \ |
1719 | (__ptrlow (P) = new_buffer + (__ptrlow (P) - old_buffer), \ | |
1720 | SET_HIGH_BOUND (P), \ | |
1721 | __ptrvalue (P) = new_buffer + (__ptrvalue (P) - old_buffer)) | |
4bb91c68 SM |
1722 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
1723 | else \ | |
1724 | { \ | |
1725 | SET_HIGH_BOUND (b); \ | |
1726 | SET_HIGH_BOUND (begalt); \ | |
1727 | if (fixup_alt_jump) \ | |
1728 | SET_HIGH_BOUND (fixup_alt_jump); \ | |
1729 | if (laststart) \ | |
1730 | SET_HIGH_BOUND (laststart); \ | |
1731 | if (pending_exact) \ | |
1732 | SET_HIGH_BOUND (pending_exact); \ | |
1733 | } | |
1734 | #else | |
381880b0 | 1735 | # define MOVE_BUFFER_POINTER(P) ((P) = new_buffer + ((P) - old_buffer)) |
4bb91c68 SM |
1736 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND |
1737 | #endif | |
fa9a63c5 | 1738 | #define EXTEND_BUFFER() \ |
25fe55af | 1739 | do { \ |
381880b0 | 1740 | unsigned char *old_buffer = bufp->buffer; \ |
25fe55af | 1741 | if (bufp->allocated == MAX_BUF_SIZE) \ |
fa9a63c5 RM |
1742 | return REG_ESIZE; \ |
1743 | bufp->allocated <<= 1; \ | |
1744 | if (bufp->allocated > MAX_BUF_SIZE) \ | |
25fe55af | 1745 | bufp->allocated = MAX_BUF_SIZE; \ |
01618498 | 1746 | RETALLOC (bufp->buffer, bufp->allocated, unsigned char); \ |
fa9a63c5 RM |
1747 | if (bufp->buffer == NULL) \ |
1748 | return REG_ESPACE; \ | |
1749 | /* If the buffer moved, move all the pointers into it. */ \ | |
1750 | if (old_buffer != bufp->buffer) \ | |
1751 | { \ | |
381880b0 | 1752 | unsigned char *new_buffer = bufp->buffer; \ |
4bb91c68 SM |
1753 | MOVE_BUFFER_POINTER (b); \ |
1754 | MOVE_BUFFER_POINTER (begalt); \ | |
25fe55af | 1755 | if (fixup_alt_jump) \ |
4bb91c68 | 1756 | MOVE_BUFFER_POINTER (fixup_alt_jump); \ |
25fe55af | 1757 | if (laststart) \ |
4bb91c68 | 1758 | MOVE_BUFFER_POINTER (laststart); \ |
25fe55af | 1759 | if (pending_exact) \ |
4bb91c68 | 1760 | MOVE_BUFFER_POINTER (pending_exact); \ |
fa9a63c5 | 1761 | } \ |
4bb91c68 | 1762 | ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
fa9a63c5 RM |
1763 | } while (0) |
1764 | ||
1765 | ||
1766 | /* Since we have one byte reserved for the register number argument to | |
1767 | {start,stop}_memory, the maximum number of groups we can report | |
1768 | things about is what fits in that byte. */ | |
1769 | #define MAX_REGNUM 255 | |
1770 | ||
1771 | /* But patterns can have more than `MAX_REGNUM' registers. We just | |
1772 | ignore the excess. */ | |
098d42af | 1773 | typedef int regnum_t; |
fa9a63c5 RM |
1774 | |
1775 | ||
1776 | /* Macros for the compile stack. */ | |
1777 | ||
1778 | /* Since offsets can go either forwards or backwards, this type needs to | |
4bb91c68 SM |
1779 | be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ |
1780 | /* int may be not enough when sizeof(int) == 2. */ | |
1781 | typedef long pattern_offset_t; | |
fa9a63c5 RM |
1782 | |
1783 | typedef struct | |
1784 | { | |
1785 | pattern_offset_t begalt_offset; | |
1786 | pattern_offset_t fixup_alt_jump; | |
5e69f11e | 1787 | pattern_offset_t laststart_offset; |
fa9a63c5 RM |
1788 | regnum_t regnum; |
1789 | } compile_stack_elt_t; | |
1790 | ||
1791 | ||
1792 | typedef struct | |
1793 | { | |
1794 | compile_stack_elt_t *stack; | |
d1dfb56c EZ |
1795 | size_t size; |
1796 | size_t avail; /* Offset of next open position. */ | |
fa9a63c5 RM |
1797 | } compile_stack_type; |
1798 | ||
1799 | ||
1800 | #define INIT_COMPILE_STACK_SIZE 32 | |
1801 | ||
1802 | #define COMPILE_STACK_EMPTY (compile_stack.avail == 0) | |
1803 | #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) | |
1804 | ||
4bb91c68 | 1805 | /* The next available element. */ |
fa9a63c5 RM |
1806 | #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) |
1807 | ||
0caaedb1 PE |
1808 | /* Explicit quit checking is needed for Emacs, which uses polling to |
1809 | process input events. */ | |
1810 | #ifdef emacs | |
77d11aec RS |
1811 | # define IMMEDIATE_QUIT_CHECK \ |
1812 | do { \ | |
1813 | if (immediate_quit) QUIT; \ | |
1814 | } while (0) | |
1815 | #else | |
1816 | # define IMMEDIATE_QUIT_CHECK ((void)0) | |
1817 | #endif | |
1818 | \f | |
b18215fc RS |
1819 | /* Structure to manage work area for range table. */ |
1820 | struct range_table_work_area | |
1821 | { | |
1822 | int *table; /* actual work area. */ | |
1823 | int allocated; /* allocated size for work area in bytes. */ | |
7814e705 | 1824 | int used; /* actually used size in words. */ |
96cc36cc | 1825 | int bits; /* flag to record character classes */ |
b18215fc RS |
1826 | }; |
1827 | ||
78779650 AS |
1828 | #ifdef emacs |
1829 | ||
77d11aec RS |
1830 | /* Make sure that WORK_AREA can hold more N multibyte characters. |
1831 | This is used only in set_image_of_range and set_image_of_range_1. | |
1832 | It expects WORK_AREA to be a pointer. | |
1833 | If it can't get the space, it returns from the surrounding function. */ | |
1834 | ||
1835 | #define EXTEND_RANGE_TABLE(work_area, n) \ | |
1836 | do { \ | |
8f924df7 | 1837 | if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \ |
77d11aec | 1838 | { \ |
8f924df7 KH |
1839 | extend_range_table_work_area (&work_area); \ |
1840 | if ((work_area).table == 0) \ | |
77d11aec RS |
1841 | return (REG_ESPACE); \ |
1842 | } \ | |
b18215fc RS |
1843 | } while (0) |
1844 | ||
96cc36cc RS |
1845 | #define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \ |
1846 | (work_area).bits |= (bit) | |
1847 | ||
b18215fc RS |
1848 | /* Set a range (RANGE_START, RANGE_END) to WORK_AREA. */ |
1849 | #define SET_RANGE_TABLE_WORK_AREA(work_area, range_start, range_end) \ | |
77d11aec | 1850 | do { \ |
8f924df7 | 1851 | EXTEND_RANGE_TABLE ((work_area), 2); \ |
b18215fc RS |
1852 | (work_area).table[(work_area).used++] = (range_start); \ |
1853 | (work_area).table[(work_area).used++] = (range_end); \ | |
1854 | } while (0) | |
1855 | ||
78779650 AS |
1856 | #endif /* emacs */ |
1857 | ||
7814e705 | 1858 | /* Free allocated memory for WORK_AREA. */ |
b18215fc RS |
1859 | #define FREE_RANGE_TABLE_WORK_AREA(work_area) \ |
1860 | do { \ | |
1861 | if ((work_area).table) \ | |
1862 | free ((work_area).table); \ | |
1863 | } while (0) | |
1864 | ||
96cc36cc | 1865 | #define CLEAR_RANGE_TABLE_WORK_USED(work_area) ((work_area).used = 0, (work_area).bits = 0) |
b18215fc | 1866 | #define RANGE_TABLE_WORK_USED(work_area) ((work_area).used) |
96cc36cc | 1867 | #define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits) |
b18215fc | 1868 | #define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i]) |
78779650 AS |
1869 | |
1870 | /* Bits used to implement the multibyte-part of the various character classes | |
1871 | such as [:alnum:] in a charset's range table. */ | |
1872 | #define BIT_WORD 0x1 | |
1873 | #define BIT_LOWER 0x2 | |
1874 | #define BIT_PUNCT 0x4 | |
1875 | #define BIT_SPACE 0x8 | |
1876 | #define BIT_UPPER 0x10 | |
1877 | #define BIT_MULTIBYTE 0x20 | |
77d11aec | 1878 | \f |
b18215fc | 1879 | |
fa9a63c5 | 1880 | /* Set the bit for character C in a list. */ |
01618498 | 1881 | #define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH)) |
fa9a63c5 RM |
1882 | |
1883 | ||
bf216479 KH |
1884 | #ifdef emacs |
1885 | ||
cf9c99bc KH |
1886 | /* Store characters in the range FROM to TO in the bitmap at B (for |
1887 | ASCII and unibyte characters) and WORK_AREA (for multibyte | |
1888 | characters) while translating them and paying attention to the | |
1889 | continuity of translated characters. | |
8f924df7 | 1890 | |
cf9c99bc KH |
1891 | Implementation note: It is better to implement these fairly big |
1892 | macros by a function, but it's not that easy because macros called | |
8f924df7 | 1893 | in this macro assume various local variables already declared. */ |
bf216479 | 1894 | |
cf9c99bc KH |
1895 | /* Both FROM and TO are ASCII characters. */ |
1896 | ||
1897 | #define SETUP_ASCII_RANGE(work_area, FROM, TO) \ | |
1898 | do { \ | |
1899 | int C0, C1; \ | |
1900 | \ | |
1901 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1902 | { \ | |
1903 | C1 = TRANSLATE (C0); \ | |
1904 | if (! ASCII_CHAR_P (C1)) \ | |
1905 | { \ | |
1906 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
1907 | if ((C1 = RE_CHAR_TO_UNIBYTE (C1)) < 0) \ | |
1908 | C1 = C0; \ | |
1909 | } \ | |
1910 | SET_LIST_BIT (C1); \ | |
1911 | } \ | |
1912 | } while (0) | |
1913 | ||
1914 | ||
1915 | /* Both FROM and TO are unibyte characters (0x80..0xFF). */ | |
1916 | ||
1917 | #define SETUP_UNIBYTE_RANGE(work_area, FROM, TO) \ | |
1918 | do { \ | |
1919 | int C0, C1, C2, I; \ | |
1920 | int USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1921 | \ | |
1922 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1923 | { \ | |
1924 | C1 = RE_CHAR_TO_MULTIBYTE (C0); \ | |
1925 | if (CHAR_BYTE8_P (C1)) \ | |
1926 | SET_LIST_BIT (C0); \ | |
1927 | else \ | |
1928 | { \ | |
1929 | C2 = TRANSLATE (C1); \ | |
1930 | if (C2 == C1 \ | |
1931 | || (C1 = RE_CHAR_TO_UNIBYTE (C2)) < 0) \ | |
1932 | C1 = C0; \ | |
1933 | SET_LIST_BIT (C1); \ | |
1934 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1935 | { \ | |
1936 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1937 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1938 | \ | |
1939 | if (C2 >= from - 1 && C2 <= to + 1) \ | |
1940 | { \ | |
1941 | if (C2 == from - 1) \ | |
1942 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1943 | else if (C2 == to + 1) \ | |
1944 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
1945 | break; \ | |
1946 | } \ | |
1947 | } \ | |
1948 | if (I < USED) \ | |
1949 | SET_RANGE_TABLE_WORK_AREA ((work_area), C2, C2); \ | |
1950 | } \ | |
1951 | } \ | |
1952 | } while (0) | |
1953 | ||
1954 | ||
78edd3b7 | 1955 | /* Both FROM and TO are multibyte characters. */ |
cf9c99bc KH |
1956 | |
1957 | #define SETUP_MULTIBYTE_RANGE(work_area, FROM, TO) \ | |
1958 | do { \ | |
1959 | int C0, C1, C2, I, USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1960 | \ | |
1961 | SET_RANGE_TABLE_WORK_AREA ((work_area), (FROM), (TO)); \ | |
1962 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1963 | { \ | |
1964 | C1 = TRANSLATE (C0); \ | |
1965 | if ((C2 = RE_CHAR_TO_UNIBYTE (C1)) >= 0 \ | |
1966 | || (C1 != C0 && (C2 = RE_CHAR_TO_UNIBYTE (C0)) >= 0)) \ | |
1967 | SET_LIST_BIT (C2); \ | |
1968 | if (C1 >= (FROM) && C1 <= (TO)) \ | |
1969 | continue; \ | |
1970 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1971 | { \ | |
1972 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1973 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1974 | \ | |
1975 | if (C1 >= from - 1 && C1 <= to + 1) \ | |
1976 | { \ | |
1977 | if (C1 == from - 1) \ | |
1978 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1979 | else if (C1 == to + 1) \ | |
1980 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
1981 | break; \ | |
1982 | } \ | |
1983 | } \ | |
1984 | if (I < USED) \ | |
1985 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
1986 | } \ | |
bf216479 KH |
1987 | } while (0) |
1988 | ||
1989 | #endif /* emacs */ | |
1990 | ||
fa9a63c5 | 1991 | /* Get the next unsigned number in the uncompiled pattern. */ |
4618713a | 1992 | #define GET_INTERVAL_COUNT(num) \ |
c72b0edd SM |
1993 | do { \ |
1994 | if (p == pend) \ | |
1995 | FREE_STACK_RETURN (REG_EBRACE); \ | |
1996 | else \ | |
1997 | { \ | |
1998 | PATFETCH (c); \ | |
1999 | while ('0' <= c && c <= '9') \ | |
2000 | { \ | |
c72b0edd SM |
2001 | if (num < 0) \ |
2002 | num = 0; \ | |
4618713a | 2003 | if (RE_DUP_MAX / 10 - (RE_DUP_MAX % 10 < c - '0') < num) \ |
c72b0edd | 2004 | FREE_STACK_RETURN (REG_BADBR); \ |
4618713a | 2005 | num = num * 10 + c - '0'; \ |
c72b0edd SM |
2006 | if (p == pend) \ |
2007 | FREE_STACK_RETURN (REG_EBRACE); \ | |
2008 | PATFETCH (c); \ | |
2009 | } \ | |
2010 | } \ | |
2011 | } while (0) | |
77d11aec | 2012 | \f |
1fdab503 | 2013 | #if ! WIDE_CHAR_SUPPORT |
01618498 | 2014 | |
14473664 | 2015 | /* Map a string to the char class it names (if any). */ |
1fdab503 | 2016 | re_wctype_t |
29abe551 | 2017 | re_wctype (const_re_char *str) |
14473664 | 2018 | { |
5b0534c8 | 2019 | const char *string = (const char *) str; |
14473664 SM |
2020 | if (STREQ (string, "alnum")) return RECC_ALNUM; |
2021 | else if (STREQ (string, "alpha")) return RECC_ALPHA; | |
2022 | else if (STREQ (string, "word")) return RECC_WORD; | |
2023 | else if (STREQ (string, "ascii")) return RECC_ASCII; | |
2024 | else if (STREQ (string, "nonascii")) return RECC_NONASCII; | |
2025 | else if (STREQ (string, "graph")) return RECC_GRAPH; | |
2026 | else if (STREQ (string, "lower")) return RECC_LOWER; | |
2027 | else if (STREQ (string, "print")) return RECC_PRINT; | |
2028 | else if (STREQ (string, "punct")) return RECC_PUNCT; | |
2029 | else if (STREQ (string, "space")) return RECC_SPACE; | |
2030 | else if (STREQ (string, "upper")) return RECC_UPPER; | |
2031 | else if (STREQ (string, "unibyte")) return RECC_UNIBYTE; | |
2032 | else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE; | |
2033 | else if (STREQ (string, "digit")) return RECC_DIGIT; | |
2034 | else if (STREQ (string, "xdigit")) return RECC_XDIGIT; | |
2035 | else if (STREQ (string, "cntrl")) return RECC_CNTRL; | |
2036 | else if (STREQ (string, "blank")) return RECC_BLANK; | |
2037 | else return 0; | |
2038 | } | |
2039 | ||
e0f24100 | 2040 | /* True if CH is in the char class CC. */ |
1fdab503 | 2041 | boolean |
971de7fb | 2042 | re_iswctype (int ch, re_wctype_t cc) |
14473664 SM |
2043 | { |
2044 | switch (cc) | |
2045 | { | |
f3fcc40d AS |
2046 | case RECC_ALNUM: return ISALNUM (ch) != 0; |
2047 | case RECC_ALPHA: return ISALPHA (ch) != 0; | |
2048 | case RECC_BLANK: return ISBLANK (ch) != 0; | |
2049 | case RECC_CNTRL: return ISCNTRL (ch) != 0; | |
2050 | case RECC_DIGIT: return ISDIGIT (ch) != 0; | |
2051 | case RECC_GRAPH: return ISGRAPH (ch) != 0; | |
2052 | case RECC_LOWER: return ISLOWER (ch) != 0; | |
2053 | case RECC_PRINT: return ISPRINT (ch) != 0; | |
2054 | case RECC_PUNCT: return ISPUNCT (ch) != 0; | |
2055 | case RECC_SPACE: return ISSPACE (ch) != 0; | |
2056 | case RECC_UPPER: return ISUPPER (ch) != 0; | |
2057 | case RECC_XDIGIT: return ISXDIGIT (ch) != 0; | |
2058 | case RECC_ASCII: return IS_REAL_ASCII (ch) != 0; | |
213bd7f2 | 2059 | case RECC_NONASCII: return !IS_REAL_ASCII (ch); |
f3fcc40d | 2060 | case RECC_UNIBYTE: return ISUNIBYTE (ch) != 0; |
213bd7f2 | 2061 | case RECC_MULTIBYTE: return !ISUNIBYTE (ch); |
f3fcc40d | 2062 | case RECC_WORD: return ISWORD (ch) != 0; |
0cdd06f8 SM |
2063 | case RECC_ERROR: return false; |
2064 | default: | |
5e617bc2 | 2065 | abort (); |
14473664 SM |
2066 | } |
2067 | } | |
fa9a63c5 | 2068 | |
14473664 SM |
2069 | /* Return a bit-pattern to use in the range-table bits to match multibyte |
2070 | chars of class CC. */ | |
2071 | static int | |
971de7fb | 2072 | re_wctype_to_bit (re_wctype_t cc) |
14473664 SM |
2073 | { |
2074 | switch (cc) | |
2075 | { | |
2076 | case RECC_NONASCII: case RECC_PRINT: case RECC_GRAPH: | |
0cdd06f8 SM |
2077 | case RECC_MULTIBYTE: return BIT_MULTIBYTE; |
2078 | case RECC_ALPHA: case RECC_ALNUM: case RECC_WORD: return BIT_WORD; | |
2079 | case RECC_LOWER: return BIT_LOWER; | |
2080 | case RECC_UPPER: return BIT_UPPER; | |
2081 | case RECC_PUNCT: return BIT_PUNCT; | |
2082 | case RECC_SPACE: return BIT_SPACE; | |
14473664 | 2083 | case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL: |
0cdd06f8 SM |
2084 | case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0; |
2085 | default: | |
5e617bc2 | 2086 | abort (); |
14473664 SM |
2087 | } |
2088 | } | |
2089 | #endif | |
77d11aec RS |
2090 | \f |
2091 | /* Filling in the work area of a range. */ | |
2092 | ||
2093 | /* Actually extend the space in WORK_AREA. */ | |
2094 | ||
2095 | static void | |
971de7fb | 2096 | extend_range_table_work_area (struct range_table_work_area *work_area) |
177c0ea7 | 2097 | { |
77d11aec | 2098 | work_area->allocated += 16 * sizeof (int); |
38182d90 | 2099 | work_area->table = realloc (work_area->table, work_area->allocated); |
77d11aec RS |
2100 | } |
2101 | ||
8f924df7 | 2102 | #if 0 |
77d11aec RS |
2103 | #ifdef emacs |
2104 | ||
2105 | /* Carefully find the ranges of codes that are equivalent | |
2106 | under case conversion to the range start..end when passed through | |
2107 | TRANSLATE. Handle the case where non-letters can come in between | |
2108 | two upper-case letters (which happens in Latin-1). | |
2109 | Also handle the case of groups of more than 2 case-equivalent chars. | |
2110 | ||
2111 | The basic method is to look at consecutive characters and see | |
2112 | if they can form a run that can be handled as one. | |
2113 | ||
2114 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2115 | ||
2116 | static int | |
1dae0f0a AS |
2117 | set_image_of_range_1 (struct range_table_work_area *work_area, |
2118 | re_wchar_t start, re_wchar_t end, | |
2119 | RE_TRANSLATE_TYPE translate) | |
77d11aec RS |
2120 | { |
2121 | /* `one_case' indicates a character, or a run of characters, | |
2122 | each of which is an isolate (no case-equivalents). | |
2123 | This includes all ASCII non-letters. | |
2124 | ||
2125 | `two_case' indicates a character, or a run of characters, | |
2126 | each of which has two case-equivalent forms. | |
2127 | This includes all ASCII letters. | |
2128 | ||
2129 | `strange' indicates a character that has more than one | |
2130 | case-equivalent. */ | |
177c0ea7 | 2131 | |
77d11aec RS |
2132 | enum case_type {one_case, two_case, strange}; |
2133 | ||
2134 | /* Describe the run that is in progress, | |
2135 | which the next character can try to extend. | |
2136 | If run_type is strange, that means there really is no run. | |
2137 | If run_type is one_case, then run_start...run_end is the run. | |
2138 | If run_type is two_case, then the run is run_start...run_end, | |
2139 | and the case-equivalents end at run_eqv_end. */ | |
2140 | ||
2141 | enum case_type run_type = strange; | |
2142 | int run_start, run_end, run_eqv_end; | |
2143 | ||
2144 | Lisp_Object eqv_table; | |
2145 | ||
2146 | if (!RE_TRANSLATE_P (translate)) | |
2147 | { | |
b7c12565 | 2148 | EXTEND_RANGE_TABLE (work_area, 2); |
77d11aec RS |
2149 | work_area->table[work_area->used++] = (start); |
2150 | work_area->table[work_area->used++] = (end); | |
b7c12565 | 2151 | return -1; |
77d11aec RS |
2152 | } |
2153 | ||
2154 | eqv_table = XCHAR_TABLE (translate)->extras[2]; | |
99633e97 | 2155 | |
77d11aec RS |
2156 | for (; start <= end; start++) |
2157 | { | |
2158 | enum case_type this_type; | |
2159 | int eqv = RE_TRANSLATE (eqv_table, start); | |
2160 | int minchar, maxchar; | |
2161 | ||
2162 | /* Classify this character */ | |
2163 | if (eqv == start) | |
2164 | this_type = one_case; | |
2165 | else if (RE_TRANSLATE (eqv_table, eqv) == start) | |
2166 | this_type = two_case; | |
2167 | else | |
2168 | this_type = strange; | |
2169 | ||
2170 | if (start < eqv) | |
2171 | minchar = start, maxchar = eqv; | |
2172 | else | |
2173 | minchar = eqv, maxchar = start; | |
2174 | ||
2175 | /* Can this character extend the run in progress? */ | |
2176 | if (this_type == strange || this_type != run_type | |
2177 | || !(minchar == run_end + 1 | |
2178 | && (run_type == two_case | |
2179 | ? maxchar == run_eqv_end + 1 : 1))) | |
2180 | { | |
2181 | /* No, end the run. | |
2182 | Record each of its equivalent ranges. */ | |
2183 | if (run_type == one_case) | |
2184 | { | |
2185 | EXTEND_RANGE_TABLE (work_area, 2); | |
2186 | work_area->table[work_area->used++] = run_start; | |
2187 | work_area->table[work_area->used++] = run_end; | |
2188 | } | |
2189 | else if (run_type == two_case) | |
2190 | { | |
2191 | EXTEND_RANGE_TABLE (work_area, 4); | |
2192 | work_area->table[work_area->used++] = run_start; | |
2193 | work_area->table[work_area->used++] = run_end; | |
2194 | work_area->table[work_area->used++] | |
2195 | = RE_TRANSLATE (eqv_table, run_start); | |
2196 | work_area->table[work_area->used++] | |
2197 | = RE_TRANSLATE (eqv_table, run_end); | |
2198 | } | |
2199 | run_type = strange; | |
2200 | } | |
177c0ea7 | 2201 | |
77d11aec RS |
2202 | if (this_type == strange) |
2203 | { | |
2204 | /* For a strange character, add each of its equivalents, one | |
2205 | by one. Don't start a range. */ | |
2206 | do | |
2207 | { | |
2208 | EXTEND_RANGE_TABLE (work_area, 2); | |
2209 | work_area->table[work_area->used++] = eqv; | |
2210 | work_area->table[work_area->used++] = eqv; | |
2211 | eqv = RE_TRANSLATE (eqv_table, eqv); | |
2212 | } | |
2213 | while (eqv != start); | |
2214 | } | |
2215 | ||
2216 | /* Add this char to the run, or start a new run. */ | |
2217 | else if (run_type == strange) | |
2218 | { | |
2219 | /* Initialize a new range. */ | |
2220 | run_type = this_type; | |
2221 | run_start = start; | |
2222 | run_end = start; | |
2223 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2224 | } | |
2225 | else | |
2226 | { | |
2227 | /* Extend a running range. */ | |
2228 | run_end = minchar; | |
2229 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2230 | } | |
2231 | } | |
2232 | ||
2233 | /* If a run is still in progress at the end, finish it now | |
2234 | by recording its equivalent ranges. */ | |
2235 | if (run_type == one_case) | |
2236 | { | |
2237 | EXTEND_RANGE_TABLE (work_area, 2); | |
2238 | work_area->table[work_area->used++] = run_start; | |
2239 | work_area->table[work_area->used++] = run_end; | |
2240 | } | |
2241 | else if (run_type == two_case) | |
2242 | { | |
2243 | EXTEND_RANGE_TABLE (work_area, 4); | |
2244 | work_area->table[work_area->used++] = run_start; | |
2245 | work_area->table[work_area->used++] = run_end; | |
2246 | work_area->table[work_area->used++] | |
2247 | = RE_TRANSLATE (eqv_table, run_start); | |
2248 | work_area->table[work_area->used++] | |
2249 | = RE_TRANSLATE (eqv_table, run_end); | |
2250 | } | |
2251 | ||
2252 | return -1; | |
2253 | } | |
36595814 | 2254 | |
77d11aec | 2255 | #endif /* emacs */ |
36595814 | 2256 | |
2b34df4e | 2257 | /* Record the image of the range start..end when passed through |
36595814 SM |
2258 | TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end) |
2259 | and is not even necessarily contiguous. | |
b7c12565 RS |
2260 | Normally we approximate it with the smallest contiguous range that contains |
2261 | all the chars we need. However, for Latin-1 we go to extra effort | |
2262 | to do a better job. | |
2263 | ||
2264 | This function is not called for ASCII ranges. | |
77d11aec RS |
2265 | |
2266 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2267 | ||
2268 | static int | |
1dae0f0a AS |
2269 | set_image_of_range (struct range_table_work_area *work_area, |
2270 | re_wchar_t start, re_wchar_t end, | |
2271 | RE_TRANSLATE_TYPE translate) | |
36595814 | 2272 | { |
77d11aec RS |
2273 | re_wchar_t cmin, cmax; |
2274 | ||
2275 | #ifdef emacs | |
2276 | /* For Latin-1 ranges, use set_image_of_range_1 | |
2277 | to get proper handling of ranges that include letters and nonletters. | |
b7c12565 | 2278 | For a range that includes the whole of Latin-1, this is not necessary. |
77d11aec | 2279 | For other character sets, we don't bother to get this right. */ |
b7c12565 RS |
2280 | if (RE_TRANSLATE_P (translate) && start < 04400 |
2281 | && !(start < 04200 && end >= 04377)) | |
77d11aec | 2282 | { |
b7c12565 | 2283 | int newend; |
77d11aec | 2284 | int tem; |
b7c12565 RS |
2285 | newend = end; |
2286 | if (newend > 04377) | |
2287 | newend = 04377; | |
2288 | tem = set_image_of_range_1 (work_area, start, newend, translate); | |
77d11aec RS |
2289 | if (tem > 0) |
2290 | return tem; | |
2291 | ||
2292 | start = 04400; | |
2293 | if (end < 04400) | |
2294 | return -1; | |
2295 | } | |
2296 | #endif | |
2297 | ||
b7c12565 RS |
2298 | EXTEND_RANGE_TABLE (work_area, 2); |
2299 | work_area->table[work_area->used++] = (start); | |
2300 | work_area->table[work_area->used++] = (end); | |
2301 | ||
2302 | cmin = -1, cmax = -1; | |
77d11aec | 2303 | |
36595814 | 2304 | if (RE_TRANSLATE_P (translate)) |
b7c12565 RS |
2305 | { |
2306 | int ch; | |
77d11aec | 2307 | |
b7c12565 RS |
2308 | for (ch = start; ch <= end; ch++) |
2309 | { | |
2310 | re_wchar_t c = TRANSLATE (ch); | |
2311 | if (! (start <= c && c <= end)) | |
2312 | { | |
2313 | if (cmin == -1) | |
2314 | cmin = c, cmax = c; | |
2315 | else | |
2316 | { | |
2317 | cmin = MIN (cmin, c); | |
2318 | cmax = MAX (cmax, c); | |
2319 | } | |
2320 | } | |
2321 | } | |
2322 | ||
2323 | if (cmin != -1) | |
2324 | { | |
2325 | EXTEND_RANGE_TABLE (work_area, 2); | |
2326 | work_area->table[work_area->used++] = (cmin); | |
2327 | work_area->table[work_area->used++] = (cmax); | |
2328 | } | |
2329 | } | |
36595814 | 2330 | |
77d11aec RS |
2331 | return -1; |
2332 | } | |
8f924df7 | 2333 | #endif /* 0 */ |
fa9a63c5 RM |
2334 | \f |
2335 | #ifndef MATCH_MAY_ALLOCATE | |
2336 | ||
2337 | /* If we cannot allocate large objects within re_match_2_internal, | |
2338 | we make the fail stack and register vectors global. | |
2339 | The fail stack, we grow to the maximum size when a regexp | |
2340 | is compiled. | |
2341 | The register vectors, we adjust in size each time we | |
2342 | compile a regexp, according to the number of registers it needs. */ | |
2343 | ||
2344 | static fail_stack_type fail_stack; | |
2345 | ||
2346 | /* Size with which the following vectors are currently allocated. | |
2347 | That is so we can make them bigger as needed, | |
4bb91c68 | 2348 | but never make them smaller. */ |
fa9a63c5 RM |
2349 | static int regs_allocated_size; |
2350 | ||
66f0296e SM |
2351 | static re_char ** regstart, ** regend; |
2352 | static re_char **best_regstart, **best_regend; | |
fa9a63c5 RM |
2353 | |
2354 | /* Make the register vectors big enough for NUM_REGS registers, | |
4bb91c68 | 2355 | but don't make them smaller. */ |
fa9a63c5 RM |
2356 | |
2357 | static | |
1dae0f0a | 2358 | regex_grow_registers (int num_regs) |
fa9a63c5 RM |
2359 | { |
2360 | if (num_regs > regs_allocated_size) | |
2361 | { | |
66f0296e SM |
2362 | RETALLOC_IF (regstart, num_regs, re_char *); |
2363 | RETALLOC_IF (regend, num_regs, re_char *); | |
2364 | RETALLOC_IF (best_regstart, num_regs, re_char *); | |
2365 | RETALLOC_IF (best_regend, num_regs, re_char *); | |
fa9a63c5 RM |
2366 | |
2367 | regs_allocated_size = num_regs; | |
2368 | } | |
2369 | } | |
2370 | ||
2371 | #endif /* not MATCH_MAY_ALLOCATE */ | |
2372 | \f | |
261cb4bb PE |
2373 | static boolean group_in_compile_stack (compile_stack_type compile_stack, |
2374 | regnum_t regnum); | |
99633e97 | 2375 | |
fa9a63c5 RM |
2376 | /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. |
2377 | Returns one of error codes defined in `regex.h', or zero for success. | |
2378 | ||
2379 | Assumes the `allocated' (and perhaps `buffer') and `translate' | |
2380 | fields are set in BUFP on entry. | |
2381 | ||
2382 | If it succeeds, results are put in BUFP (if it returns an error, the | |
2383 | contents of BUFP are undefined): | |
2384 | `buffer' is the compiled pattern; | |
2385 | `syntax' is set to SYNTAX; | |
2386 | `used' is set to the length of the compiled pattern; | |
2387 | `fastmap_accurate' is zero; | |
2388 | `re_nsub' is the number of subexpressions in PATTERN; | |
2389 | `not_bol' and `not_eol' are zero; | |
5e69f11e | 2390 | |
c0f9ea08 | 2391 | The `fastmap' field is neither examined nor set. */ |
fa9a63c5 | 2392 | |
505bde11 SM |
2393 | /* Insert the `jump' from the end of last alternative to "here". |
2394 | The space for the jump has already been allocated. */ | |
2395 | #define FIXUP_ALT_JUMP() \ | |
2396 | do { \ | |
2397 | if (fixup_alt_jump) \ | |
2398 | STORE_JUMP (jump, fixup_alt_jump, b); \ | |
2399 | } while (0) | |
2400 | ||
2401 | ||
fa9a63c5 RM |
2402 | /* Return, freeing storage we allocated. */ |
2403 | #define FREE_STACK_RETURN(value) \ | |
b18215fc RS |
2404 | do { \ |
2405 | FREE_RANGE_TABLE_WORK_AREA (range_table_work); \ | |
2406 | free (compile_stack.stack); \ | |
2407 | return value; \ | |
2408 | } while (0) | |
fa9a63c5 RM |
2409 | |
2410 | static reg_errcode_t | |
29abe551 PE |
2411 | regex_compile (const_re_char *pattern, size_t size, reg_syntax_t syntax, |
2412 | struct re_pattern_buffer *bufp) | |
fa9a63c5 | 2413 | { |
01618498 SM |
2414 | /* We fetch characters from PATTERN here. */ |
2415 | register re_wchar_t c, c1; | |
5e69f11e | 2416 | |
fa9a63c5 RM |
2417 | /* Points to the end of the buffer, where we should append. */ |
2418 | register unsigned char *b; | |
5e69f11e | 2419 | |
fa9a63c5 RM |
2420 | /* Keeps track of unclosed groups. */ |
2421 | compile_stack_type compile_stack; | |
2422 | ||
2423 | /* Points to the current (ending) position in the pattern. */ | |
22336245 RS |
2424 | #ifdef AIX |
2425 | /* `const' makes AIX compiler fail. */ | |
66f0296e | 2426 | unsigned char *p = pattern; |
22336245 | 2427 | #else |
66f0296e | 2428 | re_char *p = pattern; |
22336245 | 2429 | #endif |
66f0296e | 2430 | re_char *pend = pattern + size; |
5e69f11e | 2431 | |
fa9a63c5 | 2432 | /* How to translate the characters in the pattern. */ |
6676cb1c | 2433 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 RM |
2434 | |
2435 | /* Address of the count-byte of the most recently inserted `exactn' | |
2436 | command. This makes it possible to tell if a new exact-match | |
2437 | character can be added to that command or if the character requires | |
2438 | a new `exactn' command. */ | |
2439 | unsigned char *pending_exact = 0; | |
2440 | ||
2441 | /* Address of start of the most recently finished expression. | |
2442 | This tells, e.g., postfix * where to find the start of its | |
2443 | operand. Reset at the beginning of groups and alternatives. */ | |
2444 | unsigned char *laststart = 0; | |
2445 | ||
2446 | /* Address of beginning of regexp, or inside of last group. */ | |
2447 | unsigned char *begalt; | |
2448 | ||
2449 | /* Place in the uncompiled pattern (i.e., the {) to | |
2450 | which to go back if the interval is invalid. */ | |
66f0296e | 2451 | re_char *beg_interval; |
5e69f11e | 2452 | |
fa9a63c5 | 2453 | /* Address of the place where a forward jump should go to the end of |
7814e705 | 2454 | the containing expression. Each alternative of an `or' -- except the |
fa9a63c5 RM |
2455 | last -- ends with a forward jump of this sort. */ |
2456 | unsigned char *fixup_alt_jump = 0; | |
2457 | ||
b18215fc RS |
2458 | /* Work area for range table of charset. */ |
2459 | struct range_table_work_area range_table_work; | |
2460 | ||
2d1675e4 SM |
2461 | /* If the object matched can contain multibyte characters. */ |
2462 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | |
2463 | ||
f9b0fd99 RS |
2464 | /* Nonzero if we have pushed down into a subpattern. */ |
2465 | int in_subpattern = 0; | |
2466 | ||
2467 | /* These hold the values of p, pattern, and pend from the main | |
2468 | pattern when we have pushed into a subpattern. */ | |
da053e48 PE |
2469 | re_char *main_p IF_LINT (= NULL); |
2470 | re_char *main_pattern IF_LINT (= NULL); | |
2471 | re_char *main_pend IF_LINT (= NULL); | |
f9b0fd99 | 2472 | |
fa9a63c5 | 2473 | #ifdef DEBUG |
99633e97 | 2474 | debug++; |
dc4a2ee0 | 2475 | DEBUG_PRINT ("\nCompiling pattern: "); |
99633e97 | 2476 | if (debug > 0) |
fa9a63c5 RM |
2477 | { |
2478 | unsigned debug_count; | |
5e69f11e | 2479 | |
fa9a63c5 | 2480 | for (debug_count = 0; debug_count < size; debug_count++) |
25fe55af | 2481 | putchar (pattern[debug_count]); |
fa9a63c5 RM |
2482 | putchar ('\n'); |
2483 | } | |
2484 | #endif /* DEBUG */ | |
2485 | ||
2486 | /* Initialize the compile stack. */ | |
2487 | compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); | |
2488 | if (compile_stack.stack == NULL) | |
2489 | return REG_ESPACE; | |
2490 | ||
2491 | compile_stack.size = INIT_COMPILE_STACK_SIZE; | |
2492 | compile_stack.avail = 0; | |
2493 | ||
b18215fc RS |
2494 | range_table_work.table = 0; |
2495 | range_table_work.allocated = 0; | |
2496 | ||
fa9a63c5 RM |
2497 | /* Initialize the pattern buffer. */ |
2498 | bufp->syntax = syntax; | |
2499 | bufp->fastmap_accurate = 0; | |
2500 | bufp->not_bol = bufp->not_eol = 0; | |
6224b623 | 2501 | bufp->used_syntax = 0; |
fa9a63c5 RM |
2502 | |
2503 | /* Set `used' to zero, so that if we return an error, the pattern | |
2504 | printer (for debugging) will think there's no pattern. We reset it | |
2505 | at the end. */ | |
2506 | bufp->used = 0; | |
5e69f11e | 2507 | |
fa9a63c5 | 2508 | /* Always count groups, whether or not bufp->no_sub is set. */ |
5e69f11e | 2509 | bufp->re_nsub = 0; |
fa9a63c5 | 2510 | |
0b32bf0e | 2511 | #if !defined emacs && !defined SYNTAX_TABLE |
fa9a63c5 RM |
2512 | /* Initialize the syntax table. */ |
2513 | init_syntax_once (); | |
2514 | #endif | |
2515 | ||
2516 | if (bufp->allocated == 0) | |
2517 | { | |
2518 | if (bufp->buffer) | |
2519 | { /* If zero allocated, but buffer is non-null, try to realloc | |
25fe55af | 2520 | enough space. This loses if buffer's address is bogus, but |
7814e705 | 2521 | that is the user's responsibility. */ |
25fe55af RS |
2522 | RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); |
2523 | } | |
fa9a63c5 | 2524 | else |
7814e705 | 2525 | { /* Caller did not allocate a buffer. Do it for them. */ |
25fe55af RS |
2526 | bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); |
2527 | } | |
fa9a63c5 RM |
2528 | if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE); |
2529 | ||
2530 | bufp->allocated = INIT_BUF_SIZE; | |
2531 | } | |
2532 | ||
2533 | begalt = b = bufp->buffer; | |
2534 | ||
2535 | /* Loop through the uncompiled pattern until we're at the end. */ | |
f9b0fd99 | 2536 | while (1) |
fa9a63c5 | 2537 | { |
f9b0fd99 RS |
2538 | if (p == pend) |
2539 | { | |
2540 | /* If this is the end of an included regexp, | |
2541 | pop back to the main regexp and try again. */ | |
2542 | if (in_subpattern) | |
2543 | { | |
2544 | in_subpattern = 0; | |
2545 | pattern = main_pattern; | |
2546 | p = main_p; | |
2547 | pend = main_pend; | |
2548 | continue; | |
2549 | } | |
2550 | /* If this is the end of the main regexp, we are done. */ | |
2551 | break; | |
2552 | } | |
2553 | ||
fa9a63c5 RM |
2554 | PATFETCH (c); |
2555 | ||
2556 | switch (c) | |
25fe55af | 2557 | { |
f9b0fd99 RS |
2558 | case ' ': |
2559 | { | |
2560 | re_char *p1 = p; | |
2561 | ||
2562 | /* If there's no special whitespace regexp, treat | |
4fb680cd RS |
2563 | spaces normally. And don't try to do this recursively. */ |
2564 | if (!whitespace_regexp || in_subpattern) | |
f9b0fd99 RS |
2565 | goto normal_char; |
2566 | ||
2567 | /* Peek past following spaces. */ | |
2568 | while (p1 != pend) | |
2569 | { | |
2570 | if (*p1 != ' ') | |
2571 | break; | |
2572 | p1++; | |
2573 | } | |
2574 | /* If the spaces are followed by a repetition op, | |
2575 | treat them normally. */ | |
c721eee5 RS |
2576 | if (p1 != pend |
2577 | && (*p1 == '*' || *p1 == '+' || *p1 == '?' | |
f9b0fd99 RS |
2578 | || (*p1 == '\\' && p1 + 1 != pend && p1[1] == '{'))) |
2579 | goto normal_char; | |
2580 | ||
2581 | /* Replace the spaces with the whitespace regexp. */ | |
2582 | in_subpattern = 1; | |
2583 | main_p = p1; | |
2584 | main_pend = pend; | |
2585 | main_pattern = pattern; | |
2586 | p = pattern = whitespace_regexp; | |
5b0534c8 | 2587 | pend = p + strlen ((const char *) p); |
f9b0fd99 | 2588 | break; |
7814e705 | 2589 | } |
f9b0fd99 | 2590 | |
25fe55af RS |
2591 | case '^': |
2592 | { | |
7814e705 | 2593 | if ( /* If at start of pattern, it's an operator. */ |
25fe55af | 2594 | p == pattern + 1 |
7814e705 | 2595 | /* If context independent, it's an operator. */ |
25fe55af | 2596 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
7814e705 | 2597 | /* Otherwise, depends on what's come before. */ |
25fe55af | 2598 | || at_begline_loc_p (pattern, p, syntax)) |
c0f9ea08 | 2599 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? begbuf : begline); |
25fe55af RS |
2600 | else |
2601 | goto normal_char; | |
2602 | } | |
2603 | break; | |
2604 | ||
2605 | ||
2606 | case '$': | |
2607 | { | |
2608 | if ( /* If at end of pattern, it's an operator. */ | |
2609 | p == pend | |
7814e705 | 2610 | /* If context independent, it's an operator. */ |
25fe55af RS |
2611 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
2612 | /* Otherwise, depends on what's next. */ | |
2613 | || at_endline_loc_p (p, pend, syntax)) | |
c0f9ea08 | 2614 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? endbuf : endline); |
25fe55af RS |
2615 | else |
2616 | goto normal_char; | |
2617 | } | |
2618 | break; | |
fa9a63c5 RM |
2619 | |
2620 | ||
2621 | case '+': | |
25fe55af RS |
2622 | case '?': |
2623 | if ((syntax & RE_BK_PLUS_QM) | |
2624 | || (syntax & RE_LIMITED_OPS)) | |
2625 | goto normal_char; | |
2626 | handle_plus: | |
2627 | case '*': | |
5ac2eb34 | 2628 | /* If there is no previous pattern... */ |
25fe55af RS |
2629 | if (!laststart) |
2630 | { | |
2631 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
2632 | FREE_STACK_RETURN (REG_BADRPT); | |
2633 | else if (!(syntax & RE_CONTEXT_INDEP_OPS)) | |
2634 | goto normal_char; | |
2635 | } | |
2636 | ||
2637 | { | |
7814e705 | 2638 | /* 1 means zero (many) matches is allowed. */ |
66f0296e SM |
2639 | boolean zero_times_ok = 0, many_times_ok = 0; |
2640 | boolean greedy = 1; | |
25fe55af RS |
2641 | |
2642 | /* If there is a sequence of repetition chars, collapse it | |
2643 | down to just one (the right one). We can't combine | |
2644 | interval operators with these because of, e.g., `a{2}*', | |
7814e705 | 2645 | which should only match an even number of `a's. */ |
25fe55af RS |
2646 | |
2647 | for (;;) | |
2648 | { | |
0b32bf0e | 2649 | if ((syntax & RE_FRUGAL) |
1c8c6d39 DL |
2650 | && c == '?' && (zero_times_ok || many_times_ok)) |
2651 | greedy = 0; | |
2652 | else | |
2653 | { | |
2654 | zero_times_ok |= c != '+'; | |
2655 | many_times_ok |= c != '?'; | |
2656 | } | |
25fe55af RS |
2657 | |
2658 | if (p == pend) | |
2659 | break; | |
ed0767d8 SM |
2660 | else if (*p == '*' |
2661 | || (!(syntax & RE_BK_PLUS_QM) | |
2662 | && (*p == '+' || *p == '?'))) | |
25fe55af | 2663 | ; |
ed0767d8 | 2664 | else if (syntax & RE_BK_PLUS_QM && *p == '\\') |
25fe55af | 2665 | { |
ed0767d8 SM |
2666 | if (p+1 == pend) |
2667 | FREE_STACK_RETURN (REG_EESCAPE); | |
2668 | if (p[1] == '+' || p[1] == '?') | |
2669 | PATFETCH (c); /* Gobble up the backslash. */ | |
2670 | else | |
2671 | break; | |
25fe55af RS |
2672 | } |
2673 | else | |
ed0767d8 | 2674 | break; |
25fe55af | 2675 | /* If we get here, we found another repeat character. */ |
ed0767d8 SM |
2676 | PATFETCH (c); |
2677 | } | |
25fe55af RS |
2678 | |
2679 | /* Star, etc. applied to an empty pattern is equivalent | |
2680 | to an empty pattern. */ | |
4e8a9132 | 2681 | if (!laststart || laststart == b) |
25fe55af RS |
2682 | break; |
2683 | ||
2684 | /* Now we know whether or not zero matches is allowed | |
7814e705 | 2685 | and also whether or not two or more matches is allowed. */ |
1c8c6d39 DL |
2686 | if (greedy) |
2687 | { | |
99633e97 | 2688 | if (many_times_ok) |
4e8a9132 SM |
2689 | { |
2690 | boolean simple = skip_one_char (laststart) == b; | |
d1dfb56c | 2691 | size_t startoffset = 0; |
f6a3f532 | 2692 | re_opcode_t ofj = |
01618498 | 2693 | /* Check if the loop can match the empty string. */ |
6df42991 SM |
2694 | (simple || !analyse_first (laststart, b, NULL, 0)) |
2695 | ? on_failure_jump : on_failure_jump_loop; | |
4e8a9132 | 2696 | assert (skip_one_char (laststart) <= b); |
177c0ea7 | 2697 | |
4e8a9132 SM |
2698 | if (!zero_times_ok && simple) |
2699 | { /* Since simple * loops can be made faster by using | |
2700 | on_failure_keep_string_jump, we turn simple P+ | |
2701 | into PP* if P is simple. */ | |
2702 | unsigned char *p1, *p2; | |
2703 | startoffset = b - laststart; | |
2704 | GET_BUFFER_SPACE (startoffset); | |
2705 | p1 = b; p2 = laststart; | |
2706 | while (p2 < p1) | |
2707 | *b++ = *p2++; | |
2708 | zero_times_ok = 1; | |
99633e97 | 2709 | } |
4e8a9132 SM |
2710 | |
2711 | GET_BUFFER_SPACE (6); | |
2712 | if (!zero_times_ok) | |
2713 | /* A + loop. */ | |
f6a3f532 | 2714 | STORE_JUMP (ofj, b, b + 6); |
99633e97 | 2715 | else |
4e8a9132 SM |
2716 | /* Simple * loops can use on_failure_keep_string_jump |
2717 | depending on what follows. But since we don't know | |
2718 | that yet, we leave the decision up to | |
2719 | on_failure_jump_smart. */ | |
f6a3f532 | 2720 | INSERT_JUMP (simple ? on_failure_jump_smart : ofj, |
4e8a9132 | 2721 | laststart + startoffset, b + 6); |
99633e97 | 2722 | b += 3; |
4e8a9132 | 2723 | STORE_JUMP (jump, b, laststart + startoffset); |
99633e97 SM |
2724 | b += 3; |
2725 | } | |
2726 | else | |
2727 | { | |
4e8a9132 SM |
2728 | /* A simple ? pattern. */ |
2729 | assert (zero_times_ok); | |
2730 | GET_BUFFER_SPACE (3); | |
2731 | INSERT_JUMP (on_failure_jump, laststart, b + 3); | |
99633e97 SM |
2732 | b += 3; |
2733 | } | |
1c8c6d39 DL |
2734 | } |
2735 | else /* not greedy */ | |
5ac2eb34 | 2736 | { /* I wish the greedy and non-greedy cases could be merged. */ |
1c8c6d39 | 2737 | |
0683b6fa | 2738 | GET_BUFFER_SPACE (7); /* We might use less. */ |
1c8c6d39 DL |
2739 | if (many_times_ok) |
2740 | { | |
f6a3f532 SM |
2741 | boolean emptyp = analyse_first (laststart, b, NULL, 0); |
2742 | ||
6df42991 SM |
2743 | /* The non-greedy multiple match looks like |
2744 | a repeat..until: we only need a conditional jump | |
2745 | at the end of the loop. */ | |
f6a3f532 SM |
2746 | if (emptyp) BUF_PUSH (no_op); |
2747 | STORE_JUMP (emptyp ? on_failure_jump_nastyloop | |
2748 | : on_failure_jump, b, laststart); | |
1c8c6d39 DL |
2749 | b += 3; |
2750 | if (zero_times_ok) | |
2751 | { | |
2752 | /* The repeat...until naturally matches one or more. | |
2753 | To also match zero times, we need to first jump to | |
6df42991 | 2754 | the end of the loop (its conditional jump). */ |
1c8c6d39 DL |
2755 | INSERT_JUMP (jump, laststart, b); |
2756 | b += 3; | |
2757 | } | |
2758 | } | |
2759 | else | |
2760 | { | |
2761 | /* non-greedy a?? */ | |
1c8c6d39 DL |
2762 | INSERT_JUMP (jump, laststart, b + 3); |
2763 | b += 3; | |
2764 | INSERT_JUMP (on_failure_jump, laststart, laststart + 6); | |
2765 | b += 3; | |
2766 | } | |
2767 | } | |
2768 | } | |
4e8a9132 | 2769 | pending_exact = 0; |
fa9a63c5 RM |
2770 | break; |
2771 | ||
2772 | ||
2773 | case '.': | |
25fe55af RS |
2774 | laststart = b; |
2775 | BUF_PUSH (anychar); | |
2776 | break; | |
fa9a63c5 RM |
2777 | |
2778 | ||
25fe55af RS |
2779 | case '[': |
2780 | { | |
19ed5445 PE |
2781 | re_char *p1; |
2782 | ||
b18215fc | 2783 | CLEAR_RANGE_TABLE_WORK_USED (range_table_work); |
fa9a63c5 | 2784 | |
25fe55af | 2785 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2786 | |
25fe55af RS |
2787 | /* Ensure that we have enough space to push a charset: the |
2788 | opcode, the length count, and the bitset; 34 bytes in all. */ | |
fa9a63c5 RM |
2789 | GET_BUFFER_SPACE (34); |
2790 | ||
25fe55af | 2791 | laststart = b; |
e318085a | 2792 | |
25fe55af | 2793 | /* We test `*p == '^' twice, instead of using an if |
7814e705 | 2794 | statement, so we only need one BUF_PUSH. */ |
25fe55af RS |
2795 | BUF_PUSH (*p == '^' ? charset_not : charset); |
2796 | if (*p == '^') | |
2797 | p++; | |
e318085a | 2798 | |
25fe55af RS |
2799 | /* Remember the first position in the bracket expression. */ |
2800 | p1 = p; | |
e318085a | 2801 | |
7814e705 | 2802 | /* Push the number of bytes in the bitmap. */ |
25fe55af | 2803 | BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2804 | |
25fe55af | 2805 | /* Clear the whole map. */ |
72af86bd | 2806 | memset (b, 0, (1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2807 | |
25fe55af RS |
2808 | /* charset_not matches newline according to a syntax bit. */ |
2809 | if ((re_opcode_t) b[-2] == charset_not | |
2810 | && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) | |
2811 | SET_LIST_BIT ('\n'); | |
fa9a63c5 | 2812 | |
7814e705 | 2813 | /* Read in characters and ranges, setting map bits. */ |
25fe55af RS |
2814 | for (;;) |
2815 | { | |
b18215fc | 2816 | boolean escaped_char = false; |
2d1675e4 | 2817 | const unsigned char *p2 = p; |
abbd1bcf | 2818 | re_wchar_t ch; |
e318085a | 2819 | |
25fe55af | 2820 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
e318085a | 2821 | |
36595814 SM |
2822 | /* Don't translate yet. The range TRANSLATE(X..Y) cannot |
2823 | always be determined from TRANSLATE(X) and TRANSLATE(Y) | |
2824 | So the translation is done later in a loop. Example: | |
2825 | (let ((case-fold-search t)) (string-match "[A-_]" "A")) */ | |
25fe55af | 2826 | PATFETCH (c); |
e318085a | 2827 | |
25fe55af RS |
2828 | /* \ might escape characters inside [...] and [^...]. */ |
2829 | if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') | |
2830 | { | |
2831 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
e318085a RS |
2832 | |
2833 | PATFETCH (c); | |
b18215fc | 2834 | escaped_char = true; |
25fe55af | 2835 | } |
b18215fc RS |
2836 | else |
2837 | { | |
7814e705 | 2838 | /* Could be the end of the bracket expression. If it's |
657fcfbd RS |
2839 | not (i.e., when the bracket expression is `[]' so |
2840 | far), the ']' character bit gets set way below. */ | |
2d1675e4 | 2841 | if (c == ']' && p2 != p1) |
657fcfbd | 2842 | break; |
25fe55af | 2843 | } |
b18215fc | 2844 | |
25fe55af RS |
2845 | /* See if we're at the beginning of a possible character |
2846 | class. */ | |
b18215fc | 2847 | |
2d1675e4 SM |
2848 | if (!escaped_char && |
2849 | syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') | |
657fcfbd | 2850 | { |
7814e705 | 2851 | /* Leave room for the null. */ |
14473664 | 2852 | unsigned char str[CHAR_CLASS_MAX_LENGTH + 1]; |
ed0767d8 | 2853 | const unsigned char *class_beg; |
b18215fc | 2854 | |
25fe55af RS |
2855 | PATFETCH (c); |
2856 | c1 = 0; | |
ed0767d8 | 2857 | class_beg = p; |
b18215fc | 2858 | |
25fe55af RS |
2859 | /* If pattern is `[[:'. */ |
2860 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
b18215fc | 2861 | |
25fe55af RS |
2862 | for (;;) |
2863 | { | |
14473664 SM |
2864 | PATFETCH (c); |
2865 | if ((c == ':' && *p == ']') || p == pend) | |
2866 | break; | |
2867 | if (c1 < CHAR_CLASS_MAX_LENGTH) | |
2868 | str[c1++] = c; | |
2869 | else | |
2870 | /* This is in any case an invalid class name. */ | |
2871 | str[0] = '\0'; | |
25fe55af RS |
2872 | } |
2873 | str[c1] = '\0'; | |
b18215fc RS |
2874 | |
2875 | /* If isn't a word bracketed by `[:' and `:]': | |
2876 | undo the ending character, the letters, and | |
2877 | leave the leading `:' and `[' (but set bits for | |
2878 | them). */ | |
25fe55af RS |
2879 | if (c == ':' && *p == ']') |
2880 | { | |
abbd1bcf | 2881 | re_wctype_t cc = re_wctype (str); |
14473664 SM |
2882 | |
2883 | if (cc == 0) | |
fa9a63c5 RM |
2884 | FREE_STACK_RETURN (REG_ECTYPE); |
2885 | ||
14473664 SM |
2886 | /* Throw away the ] at the end of the character |
2887 | class. */ | |
2888 | PATFETCH (c); | |
fa9a63c5 | 2889 | |
14473664 | 2890 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2891 | |
cf9c99bc KH |
2892 | #ifndef emacs |
2893 | for (ch = 0; ch < (1 << BYTEWIDTH); ++ch) | |
8f924df7 KH |
2894 | if (re_iswctype (btowc (ch), cc)) |
2895 | { | |
2896 | c = TRANSLATE (ch); | |
ed00c2ac KH |
2897 | if (c < (1 << BYTEWIDTH)) |
2898 | SET_LIST_BIT (c); | |
8f924df7 | 2899 | } |
cf9c99bc KH |
2900 | #else /* emacs */ |
2901 | /* Most character classes in a multibyte match | |
2902 | just set a flag. Exceptions are is_blank, | |
2903 | is_digit, is_cntrl, and is_xdigit, since | |
2904 | they can only match ASCII characters. We | |
2905 | don't need to handle them for multibyte. | |
2906 | They are distinguished by a negative wctype. */ | |
96cc36cc | 2907 | |
254c06a8 SM |
2908 | /* Setup the gl_state object to its buffer-defined |
2909 | value. This hardcodes the buffer-global | |
2910 | syntax-table for ASCII chars, while the other chars | |
2911 | will obey syntax-table properties. It's not ideal, | |
2912 | but it's the way it's been done until now. */ | |
d48cd3f4 | 2913 | SETUP_BUFFER_SYNTAX_TABLE (); |
254c06a8 | 2914 | |
cf9c99bc | 2915 | for (ch = 0; ch < 256; ++ch) |
25fe55af | 2916 | { |
cf9c99bc KH |
2917 | c = RE_CHAR_TO_MULTIBYTE (ch); |
2918 | if (! CHAR_BYTE8_P (c) | |
2919 | && re_iswctype (c, cc)) | |
8f924df7 | 2920 | { |
cf9c99bc KH |
2921 | SET_LIST_BIT (ch); |
2922 | c1 = TRANSLATE (c); | |
2923 | if (c1 == c) | |
2924 | continue; | |
2925 | if (ASCII_CHAR_P (c1)) | |
2926 | SET_LIST_BIT (c1); | |
2927 | else if ((c1 = RE_CHAR_TO_UNIBYTE (c1)) >= 0) | |
2928 | SET_LIST_BIT (c1); | |
8f924df7 | 2929 | } |
25fe55af | 2930 | } |
cf9c99bc KH |
2931 | SET_RANGE_TABLE_WORK_AREA_BIT |
2932 | (range_table_work, re_wctype_to_bit (cc)); | |
2933 | #endif /* emacs */ | |
6224b623 SM |
2934 | /* In most cases the matching rule for char classes |
2935 | only uses the syntax table for multibyte chars, | |
2936 | so that the content of the syntax-table it is not | |
2937 | hardcoded in the range_table. SPACE and WORD are | |
2938 | the two exceptions. */ | |
2939 | if ((1 << cc) & ((1 << RECC_SPACE) | (1 << RECC_WORD))) | |
2940 | bufp->used_syntax = 1; | |
2941 | ||
b18215fc RS |
2942 | /* Repeat the loop. */ |
2943 | continue; | |
25fe55af RS |
2944 | } |
2945 | else | |
2946 | { | |
ed0767d8 SM |
2947 | /* Go back to right after the "[:". */ |
2948 | p = class_beg; | |
25fe55af | 2949 | SET_LIST_BIT ('['); |
b18215fc RS |
2950 | |
2951 | /* Because the `:' may starts the range, we | |
2952 | can't simply set bit and repeat the loop. | |
7814e705 | 2953 | Instead, just set it to C and handle below. */ |
b18215fc | 2954 | c = ':'; |
25fe55af RS |
2955 | } |
2956 | } | |
b18215fc RS |
2957 | |
2958 | if (p < pend && p[0] == '-' && p[1] != ']') | |
2959 | { | |
2960 | ||
2961 | /* Discard the `-'. */ | |
2962 | PATFETCH (c1); | |
2963 | ||
2964 | /* Fetch the character which ends the range. */ | |
2965 | PATFETCH (c1); | |
cf9c99bc KH |
2966 | #ifdef emacs |
2967 | if (CHAR_BYTE8_P (c1) | |
2968 | && ! ASCII_CHAR_P (c) && ! CHAR_BYTE8_P (c)) | |
2969 | /* Treat the range from a multibyte character to | |
2970 | raw-byte character as empty. */ | |
2971 | c = c1 + 1; | |
2972 | #endif /* emacs */ | |
e318085a | 2973 | } |
25fe55af | 2974 | else |
b18215fc RS |
2975 | /* Range from C to C. */ |
2976 | c1 = c; | |
2977 | ||
cf9c99bc | 2978 | if (c > c1) |
25fe55af | 2979 | { |
cf9c99bc KH |
2980 | if (syntax & RE_NO_EMPTY_RANGES) |
2981 | FREE_STACK_RETURN (REG_ERANGEX); | |
2982 | /* Else, repeat the loop. */ | |
bf216479 | 2983 | } |
6fdd04b0 | 2984 | else |
25fe55af | 2985 | { |
cf9c99bc KH |
2986 | #ifndef emacs |
2987 | /* Set the range into bitmap */ | |
8f924df7 | 2988 | for (; c <= c1; c++) |
b18215fc | 2989 | { |
cf9c99bc KH |
2990 | ch = TRANSLATE (c); |
2991 | if (ch < (1 << BYTEWIDTH)) | |
2992 | SET_LIST_BIT (ch); | |
2993 | } | |
2994 | #else /* emacs */ | |
2995 | if (c < 128) | |
2996 | { | |
2997 | ch = MIN (127, c1); | |
2998 | SETUP_ASCII_RANGE (range_table_work, c, ch); | |
2999 | c = ch + 1; | |
3000 | if (CHAR_BYTE8_P (c1)) | |
3001 | c = BYTE8_TO_CHAR (128); | |
3002 | } | |
3003 | if (c <= c1) | |
3004 | { | |
3005 | if (CHAR_BYTE8_P (c)) | |
3006 | { | |
3007 | c = CHAR_TO_BYTE8 (c); | |
3008 | c1 = CHAR_TO_BYTE8 (c1); | |
3009 | for (; c <= c1; c++) | |
3010 | SET_LIST_BIT (c); | |
3011 | } | |
3012 | else if (multibyte) | |
3013 | { | |
3014 | SETUP_MULTIBYTE_RANGE (range_table_work, c, c1); | |
3015 | } | |
3016 | else | |
3017 | { | |
3018 | SETUP_UNIBYTE_RANGE (range_table_work, c, c1); | |
3019 | } | |
e934739e | 3020 | } |
cf9c99bc | 3021 | #endif /* emacs */ |
25fe55af | 3022 | } |
e318085a RS |
3023 | } |
3024 | ||
25fe55af | 3025 | /* Discard any (non)matching list bytes that are all 0 at the |
7814e705 | 3026 | end of the map. Decrease the map-length byte too. */ |
25fe55af RS |
3027 | while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) |
3028 | b[-1]--; | |
3029 | b += b[-1]; | |
fa9a63c5 | 3030 | |
96cc36cc RS |
3031 | /* Build real range table from work area. */ |
3032 | if (RANGE_TABLE_WORK_USED (range_table_work) | |
3033 | || RANGE_TABLE_WORK_BITS (range_table_work)) | |
b18215fc RS |
3034 | { |
3035 | int i; | |
3036 | int used = RANGE_TABLE_WORK_USED (range_table_work); | |
fa9a63c5 | 3037 | |
b18215fc | 3038 | /* Allocate space for COUNT + RANGE_TABLE. Needs two |
96cc36cc | 3039 | bytes for flags, two for COUNT, and three bytes for |
5ac2eb34 | 3040 | each character. */ |
96cc36cc | 3041 | GET_BUFFER_SPACE (4 + used * 3); |
fa9a63c5 | 3042 | |
b18215fc RS |
3043 | /* Indicate the existence of range table. */ |
3044 | laststart[1] |= 0x80; | |
fa9a63c5 | 3045 | |
96cc36cc RS |
3046 | /* Store the character class flag bits into the range table. |
3047 | If not in emacs, these flag bits are always 0. */ | |
3048 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) & 0xff; | |
3049 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) >> 8; | |
3050 | ||
b18215fc RS |
3051 | STORE_NUMBER_AND_INCR (b, used / 2); |
3052 | for (i = 0; i < used; i++) | |
3053 | STORE_CHARACTER_AND_INCR | |
3054 | (b, RANGE_TABLE_WORK_ELT (range_table_work, i)); | |
3055 | } | |
25fe55af RS |
3056 | } |
3057 | break; | |
fa9a63c5 RM |
3058 | |
3059 | ||
b18215fc | 3060 | case '(': |
25fe55af RS |
3061 | if (syntax & RE_NO_BK_PARENS) |
3062 | goto handle_open; | |
3063 | else | |
3064 | goto normal_char; | |
fa9a63c5 RM |
3065 | |
3066 | ||
25fe55af RS |
3067 | case ')': |
3068 | if (syntax & RE_NO_BK_PARENS) | |
3069 | goto handle_close; | |
3070 | else | |
3071 | goto normal_char; | |
e318085a RS |
3072 | |
3073 | ||
25fe55af RS |
3074 | case '\n': |
3075 | if (syntax & RE_NEWLINE_ALT) | |
3076 | goto handle_alt; | |
3077 | else | |
3078 | goto normal_char; | |
e318085a RS |
3079 | |
3080 | ||
b18215fc | 3081 | case '|': |
25fe55af RS |
3082 | if (syntax & RE_NO_BK_VBAR) |
3083 | goto handle_alt; | |
3084 | else | |
3085 | goto normal_char; | |
3086 | ||
3087 | ||
3088 | case '{': | |
3089 | if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) | |
3090 | goto handle_interval; | |
3091 | else | |
3092 | goto normal_char; | |
3093 | ||
3094 | ||
3095 | case '\\': | |
3096 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
3097 | ||
3098 | /* Do not translate the character after the \, so that we can | |
3099 | distinguish, e.g., \B from \b, even if we normally would | |
3100 | translate, e.g., B to b. */ | |
36595814 | 3101 | PATFETCH (c); |
25fe55af RS |
3102 | |
3103 | switch (c) | |
3104 | { | |
3105 | case '(': | |
3106 | if (syntax & RE_NO_BK_PARENS) | |
3107 | goto normal_backslash; | |
3108 | ||
3109 | handle_open: | |
505bde11 SM |
3110 | { |
3111 | int shy = 0; | |
c69b0314 | 3112 | regnum_t regnum = 0; |
505bde11 SM |
3113 | if (p+1 < pend) |
3114 | { | |
3115 | /* Look for a special (?...) construct */ | |
ed0767d8 | 3116 | if ((syntax & RE_SHY_GROUPS) && *p == '?') |
505bde11 | 3117 | { |
ed0767d8 | 3118 | PATFETCH (c); /* Gobble up the '?'. */ |
c69b0314 | 3119 | while (!shy) |
505bde11 | 3120 | { |
c69b0314 SM |
3121 | PATFETCH (c); |
3122 | switch (c) | |
3123 | { | |
3124 | case ':': shy = 1; break; | |
3125 | case '0': | |
3126 | /* An explicitly specified regnum must start | |
3127 | with non-0. */ | |
3128 | if (regnum == 0) | |
3129 | FREE_STACK_RETURN (REG_BADPAT); | |
3130 | case '1': case '2': case '3': case '4': | |
3131 | case '5': case '6': case '7': case '8': case '9': | |
3132 | regnum = 10*regnum + (c - '0'); break; | |
3133 | default: | |
3134 | /* Only (?:...) is supported right now. */ | |
3135 | FREE_STACK_RETURN (REG_BADPAT); | |
3136 | } | |
505bde11 SM |
3137 | } |
3138 | } | |
505bde11 SM |
3139 | } |
3140 | ||
3141 | if (!shy) | |
c69b0314 SM |
3142 | regnum = ++bufp->re_nsub; |
3143 | else if (regnum) | |
3144 | { /* It's actually not shy, but explicitly numbered. */ | |
3145 | shy = 0; | |
3146 | if (regnum > bufp->re_nsub) | |
3147 | bufp->re_nsub = regnum; | |
3148 | else if (regnum > bufp->re_nsub | |
3149 | /* Ideally, we'd want to check that the specified | |
3150 | group can't have matched (i.e. all subgroups | |
3151 | using the same regnum are in other branches of | |
3152 | OR patterns), but we don't currently keep track | |
3153 | of enough info to do that easily. */ | |
3154 | || group_in_compile_stack (compile_stack, regnum)) | |
3155 | FREE_STACK_RETURN (REG_BADPAT); | |
505bde11 | 3156 | } |
c69b0314 SM |
3157 | else |
3158 | /* It's really shy. */ | |
3159 | regnum = - bufp->re_nsub; | |
25fe55af | 3160 | |
99633e97 SM |
3161 | if (COMPILE_STACK_FULL) |
3162 | { | |
3163 | RETALLOC (compile_stack.stack, compile_stack.size << 1, | |
3164 | compile_stack_elt_t); | |
3165 | if (compile_stack.stack == NULL) return REG_ESPACE; | |
25fe55af | 3166 | |
99633e97 SM |
3167 | compile_stack.size <<= 1; |
3168 | } | |
25fe55af | 3169 | |
99633e97 | 3170 | /* These are the values to restore when we hit end of this |
7814e705 | 3171 | group. They are all relative offsets, so that if the |
99633e97 SM |
3172 | whole pattern moves because of realloc, they will still |
3173 | be valid. */ | |
3174 | COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; | |
3175 | COMPILE_STACK_TOP.fixup_alt_jump | |
3176 | = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; | |
3177 | COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer; | |
c69b0314 | 3178 | COMPILE_STACK_TOP.regnum = regnum; |
99633e97 | 3179 | |
c69b0314 SM |
3180 | /* Do not push a start_memory for groups beyond the last one |
3181 | we can represent in the compiled pattern. */ | |
3182 | if (regnum <= MAX_REGNUM && regnum > 0) | |
99633e97 SM |
3183 | BUF_PUSH_2 (start_memory, regnum); |
3184 | ||
3185 | compile_stack.avail++; | |
3186 | ||
3187 | fixup_alt_jump = 0; | |
3188 | laststart = 0; | |
3189 | begalt = b; | |
3190 | /* If we've reached MAX_REGNUM groups, then this open | |
3191 | won't actually generate any code, so we'll have to | |
3192 | clear pending_exact explicitly. */ | |
3193 | pending_exact = 0; | |
3194 | break; | |
505bde11 | 3195 | } |
25fe55af RS |
3196 | |
3197 | case ')': | |
3198 | if (syntax & RE_NO_BK_PARENS) goto normal_backslash; | |
3199 | ||
3200 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3201 | { |
3202 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3203 | goto normal_backslash; | |
3204 | else | |
3205 | FREE_STACK_RETURN (REG_ERPAREN); | |
3206 | } | |
25fe55af RS |
3207 | |
3208 | handle_close: | |
505bde11 | 3209 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3210 | |
3211 | /* See similar code for backslashed left paren above. */ | |
3212 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3213 | { |
3214 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3215 | goto normal_char; | |
3216 | else | |
3217 | FREE_STACK_RETURN (REG_ERPAREN); | |
3218 | } | |
25fe55af RS |
3219 | |
3220 | /* Since we just checked for an empty stack above, this | |
3221 | ``can't happen''. */ | |
3222 | assert (compile_stack.avail != 0); | |
3223 | { | |
3224 | /* We don't just want to restore into `regnum', because | |
3225 | later groups should continue to be numbered higher, | |
7814e705 | 3226 | as in `(ab)c(de)' -- the second group is #2. */ |
c69b0314 | 3227 | regnum_t regnum; |
25fe55af RS |
3228 | |
3229 | compile_stack.avail--; | |
3230 | begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; | |
3231 | fixup_alt_jump | |
3232 | = COMPILE_STACK_TOP.fixup_alt_jump | |
3233 | ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 | |
3234 | : 0; | |
3235 | laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; | |
c69b0314 | 3236 | regnum = COMPILE_STACK_TOP.regnum; |
b18215fc RS |
3237 | /* If we've reached MAX_REGNUM groups, then this open |
3238 | won't actually generate any code, so we'll have to | |
3239 | clear pending_exact explicitly. */ | |
3240 | pending_exact = 0; | |
e318085a | 3241 | |
25fe55af | 3242 | /* We're at the end of the group, so now we know how many |
7814e705 | 3243 | groups were inside this one. */ |
c69b0314 SM |
3244 | if (regnum <= MAX_REGNUM && regnum > 0) |
3245 | BUF_PUSH_2 (stop_memory, regnum); | |
25fe55af RS |
3246 | } |
3247 | break; | |
3248 | ||
3249 | ||
3250 | case '|': /* `\|'. */ | |
3251 | if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) | |
3252 | goto normal_backslash; | |
3253 | handle_alt: | |
3254 | if (syntax & RE_LIMITED_OPS) | |
3255 | goto normal_char; | |
3256 | ||
3257 | /* Insert before the previous alternative a jump which | |
7814e705 | 3258 | jumps to this alternative if the former fails. */ |
25fe55af RS |
3259 | GET_BUFFER_SPACE (3); |
3260 | INSERT_JUMP (on_failure_jump, begalt, b + 6); | |
3261 | pending_exact = 0; | |
3262 | b += 3; | |
3263 | ||
3264 | /* The alternative before this one has a jump after it | |
3265 | which gets executed if it gets matched. Adjust that | |
3266 | jump so it will jump to this alternative's analogous | |
3267 | jump (put in below, which in turn will jump to the next | |
3268 | (if any) alternative's such jump, etc.). The last such | |
3269 | jump jumps to the correct final destination. A picture: | |
3270 | _____ _____ | |
3271 | | | | | | |
3272 | | v | v | |
d1dfb56c | 3273 | a | b | c |
25fe55af RS |
3274 | |
3275 | If we are at `b', then fixup_alt_jump right now points to a | |
3276 | three-byte space after `a'. We'll put in the jump, set | |
3277 | fixup_alt_jump to right after `b', and leave behind three | |
3278 | bytes which we'll fill in when we get to after `c'. */ | |
3279 | ||
505bde11 | 3280 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3281 | |
3282 | /* Mark and leave space for a jump after this alternative, | |
3283 | to be filled in later either by next alternative or | |
3284 | when know we're at the end of a series of alternatives. */ | |
3285 | fixup_alt_jump = b; | |
3286 | GET_BUFFER_SPACE (3); | |
3287 | b += 3; | |
3288 | ||
3289 | laststart = 0; | |
3290 | begalt = b; | |
3291 | break; | |
3292 | ||
3293 | ||
3294 | case '{': | |
3295 | /* If \{ is a literal. */ | |
3296 | if (!(syntax & RE_INTERVALS) | |
3297 | /* If we're at `\{' and it's not the open-interval | |
3298 | operator. */ | |
4bb91c68 | 3299 | || (syntax & RE_NO_BK_BRACES)) |
25fe55af RS |
3300 | goto normal_backslash; |
3301 | ||
3302 | handle_interval: | |
3303 | { | |
3304 | /* If got here, then the syntax allows intervals. */ | |
3305 | ||
3306 | /* At least (most) this many matches must be made. */ | |
99633e97 | 3307 | int lower_bound = 0, upper_bound = -1; |
25fe55af | 3308 | |
ed0767d8 | 3309 | beg_interval = p; |
25fe55af | 3310 | |
4618713a | 3311 | GET_INTERVAL_COUNT (lower_bound); |
25fe55af RS |
3312 | |
3313 | if (c == ',') | |
a3fab1ec | 3314 | GET_INTERVAL_COUNT (upper_bound); |
25fe55af RS |
3315 | else |
3316 | /* Interval such as `{1}' => match exactly once. */ | |
3317 | upper_bound = lower_bound; | |
3318 | ||
a3fab1ec PE |
3319 | if (lower_bound < 0 |
3320 | || (0 <= upper_bound && upper_bound < lower_bound)) | |
3321 | FREE_STACK_RETURN (REG_BADBR); | |
3322 | ||
25fe55af RS |
3323 | if (!(syntax & RE_NO_BK_BRACES)) |
3324 | { | |
4bb91c68 SM |
3325 | if (c != '\\') |
3326 | FREE_STACK_RETURN (REG_BADBR); | |
c72b0edd SM |
3327 | if (p == pend) |
3328 | FREE_STACK_RETURN (REG_EESCAPE); | |
25fe55af RS |
3329 | PATFETCH (c); |
3330 | } | |
3331 | ||
3332 | if (c != '}') | |
4bb91c68 | 3333 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3334 | |
3335 | /* We just parsed a valid interval. */ | |
3336 | ||
3337 | /* If it's invalid to have no preceding re. */ | |
3338 | if (!laststart) | |
3339 | { | |
3340 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
3341 | FREE_STACK_RETURN (REG_BADRPT); | |
3342 | else if (syntax & RE_CONTEXT_INDEP_OPS) | |
3343 | laststart = b; | |
3344 | else | |
3345 | goto unfetch_interval; | |
3346 | } | |
3347 | ||
6df42991 SM |
3348 | if (upper_bound == 0) |
3349 | /* If the upper bound is zero, just drop the sub pattern | |
3350 | altogether. */ | |
3351 | b = laststart; | |
3352 | else if (lower_bound == 1 && upper_bound == 1) | |
3353 | /* Just match it once: nothing to do here. */ | |
3354 | ; | |
3355 | ||
3356 | /* Otherwise, we have a nontrivial interval. When | |
3357 | we're all done, the pattern will look like: | |
3358 | set_number_at <jump count> <upper bound> | |
3359 | set_number_at <succeed_n count> <lower bound> | |
3360 | succeed_n <after jump addr> <succeed_n count> | |
3361 | <body of loop> | |
3362 | jump_n <succeed_n addr> <jump count> | |
3363 | (The upper bound and `jump_n' are omitted if | |
3364 | `upper_bound' is 1, though.) */ | |
3365 | else | |
3366 | { /* If the upper bound is > 1, we need to insert | |
3367 | more at the end of the loop. */ | |
3368 | unsigned int nbytes = (upper_bound < 0 ? 3 | |
3369 | : upper_bound > 1 ? 5 : 0); | |
3370 | unsigned int startoffset = 0; | |
3371 | ||
3372 | GET_BUFFER_SPACE (20); /* We might use less. */ | |
3373 | ||
3374 | if (lower_bound == 0) | |
3375 | { | |
3376 | /* A succeed_n that starts with 0 is really a | |
3377 | a simple on_failure_jump_loop. */ | |
3378 | INSERT_JUMP (on_failure_jump_loop, laststart, | |
3379 | b + 3 + nbytes); | |
3380 | b += 3; | |
3381 | } | |
3382 | else | |
3383 | { | |
3384 | /* Initialize lower bound of the `succeed_n', even | |
3385 | though it will be set during matching by its | |
3386 | attendant `set_number_at' (inserted next), | |
3387 | because `re_compile_fastmap' needs to know. | |
3388 | Jump to the `jump_n' we might insert below. */ | |
3389 | INSERT_JUMP2 (succeed_n, laststart, | |
3390 | b + 5 + nbytes, | |
3391 | lower_bound); | |
3392 | b += 5; | |
3393 | ||
3394 | /* Code to initialize the lower bound. Insert | |
7814e705 | 3395 | before the `succeed_n'. The `5' is the last two |
6df42991 SM |
3396 | bytes of this `set_number_at', plus 3 bytes of |
3397 | the following `succeed_n'. */ | |
3398 | insert_op2 (set_number_at, laststart, 5, lower_bound, b); | |
3399 | b += 5; | |
3400 | startoffset += 5; | |
3401 | } | |
3402 | ||
3403 | if (upper_bound < 0) | |
3404 | { | |
3405 | /* A negative upper bound stands for infinity, | |
3406 | in which case it degenerates to a plain jump. */ | |
3407 | STORE_JUMP (jump, b, laststart + startoffset); | |
3408 | b += 3; | |
3409 | } | |
3410 | else if (upper_bound > 1) | |
3411 | { /* More than one repetition is allowed, so | |
3412 | append a backward jump to the `succeed_n' | |
3413 | that starts this interval. | |
3414 | ||
3415 | When we've reached this during matching, | |
3416 | we'll have matched the interval once, so | |
3417 | jump back only `upper_bound - 1' times. */ | |
3418 | STORE_JUMP2 (jump_n, b, laststart + startoffset, | |
3419 | upper_bound - 1); | |
3420 | b += 5; | |
3421 | ||
3422 | /* The location we want to set is the second | |
3423 | parameter of the `jump_n'; that is `b-2' as | |
3424 | an absolute address. `laststart' will be | |
3425 | the `set_number_at' we're about to insert; | |
3426 | `laststart+3' the number to set, the source | |
3427 | for the relative address. But we are | |
3428 | inserting into the middle of the pattern -- | |
3429 | so everything is getting moved up by 5. | |
3430 | Conclusion: (b - 2) - (laststart + 3) + 5, | |
3431 | i.e., b - laststart. | |
3432 | ||
3433 | We insert this at the beginning of the loop | |
3434 | so that if we fail during matching, we'll | |
3435 | reinitialize the bounds. */ | |
3436 | insert_op2 (set_number_at, laststart, b - laststart, | |
3437 | upper_bound - 1, b); | |
3438 | b += 5; | |
3439 | } | |
3440 | } | |
25fe55af RS |
3441 | pending_exact = 0; |
3442 | beg_interval = NULL; | |
3443 | } | |
3444 | break; | |
3445 | ||
3446 | unfetch_interval: | |
3447 | /* If an invalid interval, match the characters as literals. */ | |
3448 | assert (beg_interval); | |
3449 | p = beg_interval; | |
3450 | beg_interval = NULL; | |
3451 | ||
3452 | /* normal_char and normal_backslash need `c'. */ | |
ed0767d8 | 3453 | c = '{'; |
25fe55af RS |
3454 | |
3455 | if (!(syntax & RE_NO_BK_BRACES)) | |
3456 | { | |
ed0767d8 SM |
3457 | assert (p > pattern && p[-1] == '\\'); |
3458 | goto normal_backslash; | |
25fe55af | 3459 | } |
ed0767d8 SM |
3460 | else |
3461 | goto normal_char; | |
e318085a | 3462 | |
b18215fc | 3463 | #ifdef emacs |
25fe55af | 3464 | /* There is no way to specify the before_dot and after_dot |
7814e705 | 3465 | operators. rms says this is ok. --karl */ |
25fe55af | 3466 | case '=': |
5ac2eb34 | 3467 | laststart = b; |
25fe55af RS |
3468 | BUF_PUSH (at_dot); |
3469 | break; | |
3470 | ||
3471 | case 's': | |
3472 | laststart = b; | |
3473 | PATFETCH (c); | |
3474 | BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); | |
3475 | break; | |
3476 | ||
3477 | case 'S': | |
3478 | laststart = b; | |
3479 | PATFETCH (c); | |
3480 | BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); | |
3481 | break; | |
b18215fc RS |
3482 | |
3483 | case 'c': | |
3484 | laststart = b; | |
36595814 | 3485 | PATFETCH (c); |
b18215fc RS |
3486 | BUF_PUSH_2 (categoryspec, c); |
3487 | break; | |
e318085a | 3488 | |
b18215fc RS |
3489 | case 'C': |
3490 | laststart = b; | |
36595814 | 3491 | PATFETCH (c); |
b18215fc RS |
3492 | BUF_PUSH_2 (notcategoryspec, c); |
3493 | break; | |
3494 | #endif /* emacs */ | |
e318085a | 3495 | |
e318085a | 3496 | |
25fe55af | 3497 | case 'w': |
4bb91c68 SM |
3498 | if (syntax & RE_NO_GNU_OPS) |
3499 | goto normal_char; | |
25fe55af | 3500 | laststart = b; |
1fb352e0 | 3501 | BUF_PUSH_2 (syntaxspec, Sword); |
25fe55af | 3502 | break; |
e318085a | 3503 | |
e318085a | 3504 | |
25fe55af | 3505 | case 'W': |
4bb91c68 SM |
3506 | if (syntax & RE_NO_GNU_OPS) |
3507 | goto normal_char; | |
25fe55af | 3508 | laststart = b; |
1fb352e0 | 3509 | BUF_PUSH_2 (notsyntaxspec, Sword); |
25fe55af | 3510 | break; |
e318085a RS |
3511 | |
3512 | ||
25fe55af | 3513 | case '<': |
4bb91c68 SM |
3514 | if (syntax & RE_NO_GNU_OPS) |
3515 | goto normal_char; | |
5ac2eb34 | 3516 | laststart = b; |
25fe55af RS |
3517 | BUF_PUSH (wordbeg); |
3518 | break; | |
e318085a | 3519 | |
25fe55af | 3520 | case '>': |
4bb91c68 SM |
3521 | if (syntax & RE_NO_GNU_OPS) |
3522 | goto normal_char; | |
5ac2eb34 | 3523 | laststart = b; |
25fe55af RS |
3524 | BUF_PUSH (wordend); |
3525 | break; | |
e318085a | 3526 | |
669fa600 SM |
3527 | case '_': |
3528 | if (syntax & RE_NO_GNU_OPS) | |
3529 | goto normal_char; | |
3530 | laststart = b; | |
3531 | PATFETCH (c); | |
3532 | if (c == '<') | |
3533 | BUF_PUSH (symbeg); | |
3534 | else if (c == '>') | |
3535 | BUF_PUSH (symend); | |
3536 | else | |
3537 | FREE_STACK_RETURN (REG_BADPAT); | |
3538 | break; | |
3539 | ||
25fe55af | 3540 | case 'b': |
4bb91c68 SM |
3541 | if (syntax & RE_NO_GNU_OPS) |
3542 | goto normal_char; | |
25fe55af RS |
3543 | BUF_PUSH (wordbound); |
3544 | break; | |
e318085a | 3545 | |
25fe55af | 3546 | case 'B': |
4bb91c68 SM |
3547 | if (syntax & RE_NO_GNU_OPS) |
3548 | goto normal_char; | |
25fe55af RS |
3549 | BUF_PUSH (notwordbound); |
3550 | break; | |
fa9a63c5 | 3551 | |
25fe55af | 3552 | case '`': |
4bb91c68 SM |
3553 | if (syntax & RE_NO_GNU_OPS) |
3554 | goto normal_char; | |
25fe55af RS |
3555 | BUF_PUSH (begbuf); |
3556 | break; | |
e318085a | 3557 | |
25fe55af | 3558 | case '\'': |
4bb91c68 SM |
3559 | if (syntax & RE_NO_GNU_OPS) |
3560 | goto normal_char; | |
25fe55af RS |
3561 | BUF_PUSH (endbuf); |
3562 | break; | |
e318085a | 3563 | |
25fe55af RS |
3564 | case '1': case '2': case '3': case '4': case '5': |
3565 | case '6': case '7': case '8': case '9': | |
0cdd06f8 SM |
3566 | { |
3567 | regnum_t reg; | |
e318085a | 3568 | |
0cdd06f8 SM |
3569 | if (syntax & RE_NO_BK_REFS) |
3570 | goto normal_backslash; | |
e318085a | 3571 | |
0cdd06f8 | 3572 | reg = c - '0'; |
e318085a | 3573 | |
c69b0314 SM |
3574 | if (reg > bufp->re_nsub || reg < 1 |
3575 | /* Can't back reference to a subexp before its end. */ | |
3576 | || group_in_compile_stack (compile_stack, reg)) | |
0cdd06f8 | 3577 | FREE_STACK_RETURN (REG_ESUBREG); |
e318085a | 3578 | |
0cdd06f8 SM |
3579 | laststart = b; |
3580 | BUF_PUSH_2 (duplicate, reg); | |
3581 | } | |
25fe55af | 3582 | break; |
e318085a | 3583 | |
e318085a | 3584 | |
25fe55af RS |
3585 | case '+': |
3586 | case '?': | |
3587 | if (syntax & RE_BK_PLUS_QM) | |
3588 | goto handle_plus; | |
3589 | else | |
3590 | goto normal_backslash; | |
3591 | ||
3592 | default: | |
3593 | normal_backslash: | |
3594 | /* You might think it would be useful for \ to mean | |
3595 | not to translate; but if we don't translate it | |
4bb91c68 | 3596 | it will never match anything. */ |
25fe55af RS |
3597 | goto normal_char; |
3598 | } | |
3599 | break; | |
fa9a63c5 RM |
3600 | |
3601 | ||
3602 | default: | |
25fe55af | 3603 | /* Expects the character in `c'. */ |
fa9a63c5 | 3604 | normal_char: |
36595814 | 3605 | /* If no exactn currently being built. */ |
25fe55af | 3606 | if (!pending_exact |
fa9a63c5 | 3607 | |
25fe55af RS |
3608 | /* If last exactn not at current position. */ |
3609 | || pending_exact + *pending_exact + 1 != b | |
5e69f11e | 3610 | |
25fe55af | 3611 | /* We have only one byte following the exactn for the count. */ |
2d1675e4 | 3612 | || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH |
fa9a63c5 | 3613 | |
7814e705 | 3614 | /* If followed by a repetition operator. */ |
9d99031f | 3615 | || (p != pend && (*p == '*' || *p == '^')) |
fa9a63c5 | 3616 | || ((syntax & RE_BK_PLUS_QM) |
9d99031f RS |
3617 | ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?') |
3618 | : p != pend && (*p == '+' || *p == '?')) | |
fa9a63c5 | 3619 | || ((syntax & RE_INTERVALS) |
25fe55af | 3620 | && ((syntax & RE_NO_BK_BRACES) |
9d99031f RS |
3621 | ? p != pend && *p == '{' |
3622 | : p + 1 < pend && p[0] == '\\' && p[1] == '{'))) | |
fa9a63c5 RM |
3623 | { |
3624 | /* Start building a new exactn. */ | |
5e69f11e | 3625 | |
25fe55af | 3626 | laststart = b; |
fa9a63c5 RM |
3627 | |
3628 | BUF_PUSH_2 (exactn, 0); | |
3629 | pending_exact = b - 1; | |
25fe55af | 3630 | } |
5e69f11e | 3631 | |
2d1675e4 SM |
3632 | GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH); |
3633 | { | |
e0277a47 KH |
3634 | int len; |
3635 | ||
cf9c99bc | 3636 | if (multibyte) |
6fdd04b0 | 3637 | { |
cf9c99bc | 3638 | c = TRANSLATE (c); |
6fdd04b0 KH |
3639 | len = CHAR_STRING (c, b); |
3640 | b += len; | |
3641 | } | |
e0277a47 | 3642 | else |
6fdd04b0 | 3643 | { |
cf9c99bc KH |
3644 | c1 = RE_CHAR_TO_MULTIBYTE (c); |
3645 | if (! CHAR_BYTE8_P (c1)) | |
3646 | { | |
3647 | re_wchar_t c2 = TRANSLATE (c1); | |
3648 | ||
3649 | if (c1 != c2 && (c1 = RE_CHAR_TO_UNIBYTE (c2)) >= 0) | |
3650 | c = c1; | |
409f2919 | 3651 | } |
6fdd04b0 KH |
3652 | *b++ = c; |
3653 | len = 1; | |
3654 | } | |
2d1675e4 SM |
3655 | (*pending_exact) += len; |
3656 | } | |
3657 | ||
fa9a63c5 | 3658 | break; |
25fe55af | 3659 | } /* switch (c) */ |
fa9a63c5 RM |
3660 | } /* while p != pend */ |
3661 | ||
5e69f11e | 3662 | |
fa9a63c5 | 3663 | /* Through the pattern now. */ |
5e69f11e | 3664 | |
505bde11 | 3665 | FIXUP_ALT_JUMP (); |
fa9a63c5 | 3666 | |
5e69f11e | 3667 | if (!COMPILE_STACK_EMPTY) |
fa9a63c5 RM |
3668 | FREE_STACK_RETURN (REG_EPAREN); |
3669 | ||
3670 | /* If we don't want backtracking, force success | |
3671 | the first time we reach the end of the compiled pattern. */ | |
3672 | if (syntax & RE_NO_POSIX_BACKTRACKING) | |
3673 | BUF_PUSH (succeed); | |
3674 | ||
fa9a63c5 RM |
3675 | /* We have succeeded; set the length of the buffer. */ |
3676 | bufp->used = b - bufp->buffer; | |
3677 | ||
3678 | #ifdef DEBUG | |
99633e97 | 3679 | if (debug > 0) |
fa9a63c5 | 3680 | { |
505bde11 | 3681 | re_compile_fastmap (bufp); |
dc4a2ee0 | 3682 | DEBUG_PRINT ("\nCompiled pattern: \n"); |
fa9a63c5 RM |
3683 | print_compiled_pattern (bufp); |
3684 | } | |
99633e97 | 3685 | debug--; |
fa9a63c5 RM |
3686 | #endif /* DEBUG */ |
3687 | ||
3688 | #ifndef MATCH_MAY_ALLOCATE | |
3689 | /* Initialize the failure stack to the largest possible stack. This | |
3690 | isn't necessary unless we're trying to avoid calling alloca in | |
3691 | the search and match routines. */ | |
3692 | { | |
3693 | int num_regs = bufp->re_nsub + 1; | |
3694 | ||
320a2a73 | 3695 | if (fail_stack.size < re_max_failures * TYPICAL_FAILURE_SIZE) |
fa9a63c5 | 3696 | { |
a26f4ccd | 3697 | fail_stack.size = re_max_failures * TYPICAL_FAILURE_SIZE; |
38182d90 PE |
3698 | falk_stack.stack = realloc (fail_stack.stack, |
3699 | fail_stack.size * sizeof *falk_stack.stack); | |
fa9a63c5 RM |
3700 | } |
3701 | ||
3702 | regex_grow_registers (num_regs); | |
3703 | } | |
3704 | #endif /* not MATCH_MAY_ALLOCATE */ | |
3705 | ||
839966f3 | 3706 | FREE_STACK_RETURN (REG_NOERROR); |
fa9a63c5 RM |
3707 | } /* regex_compile */ |
3708 | \f | |
3709 | /* Subroutines for `regex_compile'. */ | |
3710 | ||
7814e705 | 3711 | /* Store OP at LOC followed by two-byte integer parameter ARG. */ |
fa9a63c5 RM |
3712 | |
3713 | static void | |
971de7fb | 3714 | store_op1 (re_opcode_t op, unsigned char *loc, int arg) |
fa9a63c5 RM |
3715 | { |
3716 | *loc = (unsigned char) op; | |
3717 | STORE_NUMBER (loc + 1, arg); | |
3718 | } | |
3719 | ||
3720 | ||
3721 | /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3722 | ||
3723 | static void | |
971de7fb | 3724 | store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2) |
fa9a63c5 RM |
3725 | { |
3726 | *loc = (unsigned char) op; | |
3727 | STORE_NUMBER (loc + 1, arg1); | |
3728 | STORE_NUMBER (loc + 3, arg2); | |
3729 | } | |
3730 | ||
3731 | ||
3732 | /* Copy the bytes from LOC to END to open up three bytes of space at LOC | |
3733 | for OP followed by two-byte integer parameter ARG. */ | |
3734 | ||
3735 | static void | |
971de7fb | 3736 | insert_op1 (re_opcode_t op, unsigned char *loc, int arg, unsigned char *end) |
fa9a63c5 RM |
3737 | { |
3738 | register unsigned char *pfrom = end; | |
3739 | register unsigned char *pto = end + 3; | |
3740 | ||
3741 | while (pfrom != loc) | |
3742 | *--pto = *--pfrom; | |
5e69f11e | 3743 | |
fa9a63c5 RM |
3744 | store_op1 (op, loc, arg); |
3745 | } | |
3746 | ||
3747 | ||
3748 | /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3749 | ||
3750 | static void | |
971de7fb | 3751 | insert_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2, unsigned char *end) |
fa9a63c5 RM |
3752 | { |
3753 | register unsigned char *pfrom = end; | |
3754 | register unsigned char *pto = end + 5; | |
3755 | ||
3756 | while (pfrom != loc) | |
3757 | *--pto = *--pfrom; | |
5e69f11e | 3758 | |
fa9a63c5 RM |
3759 | store_op2 (op, loc, arg1, arg2); |
3760 | } | |
3761 | ||
3762 | ||
3763 | /* P points to just after a ^ in PATTERN. Return true if that ^ comes | |
3764 | after an alternative or a begin-subexpression. We assume there is at | |
3765 | least one character before the ^. */ | |
3766 | ||
3767 | static boolean | |
29abe551 | 3768 | at_begline_loc_p (const_re_char *pattern, const_re_char *p, reg_syntax_t syntax) |
fa9a63c5 | 3769 | { |
01618498 | 3770 | re_char *prev = p - 2; |
95988fcf | 3771 | boolean odd_backslashes; |
5e69f11e | 3772 | |
95988fcf AS |
3773 | /* After a subexpression? */ |
3774 | if (*prev == '(') | |
3775 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3776 | ||
3777 | /* After an alternative? */ | |
3778 | else if (*prev == '|') | |
3779 | odd_backslashes = (syntax & RE_NO_BK_VBAR) == 0; | |
3780 | ||
3781 | /* After a shy subexpression? */ | |
3782 | else if (*prev == ':' && (syntax & RE_SHY_GROUPS)) | |
3783 | { | |
3784 | /* Skip over optional regnum. */ | |
3785 | while (prev - 1 >= pattern && prev[-1] >= '0' && prev[-1] <= '9') | |
3786 | --prev; | |
3787 | ||
3788 | if (!(prev - 2 >= pattern | |
3789 | && prev[-1] == '?' && prev[-2] == '(')) | |
3790 | return false; | |
3791 | prev -= 2; | |
3792 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3793 | } | |
3794 | else | |
3795 | return false; | |
3796 | ||
3797 | /* Count the number of preceding backslashes. */ | |
3798 | p = prev; | |
3799 | while (prev - 1 >= pattern && prev[-1] == '\\') | |
3800 | --prev; | |
3801 | return (p - prev) & odd_backslashes; | |
fa9a63c5 RM |
3802 | } |
3803 | ||
3804 | ||
3805 | /* The dual of at_begline_loc_p. This one is for $. We assume there is | |
3806 | at least one character after the $, i.e., `P < PEND'. */ | |
3807 | ||
3808 | static boolean | |
29abe551 | 3809 | at_endline_loc_p (const_re_char *p, const_re_char *pend, reg_syntax_t syntax) |
fa9a63c5 | 3810 | { |
01618498 | 3811 | re_char *next = p; |
fa9a63c5 | 3812 | boolean next_backslash = *next == '\\'; |
01618498 | 3813 | re_char *next_next = p + 1 < pend ? p + 1 : 0; |
5e69f11e | 3814 | |
fa9a63c5 RM |
3815 | return |
3816 | /* Before a subexpression? */ | |
3817 | (syntax & RE_NO_BK_PARENS ? *next == ')' | |
25fe55af | 3818 | : next_backslash && next_next && *next_next == ')') |
fa9a63c5 RM |
3819 | /* Before an alternative? */ |
3820 | || (syntax & RE_NO_BK_VBAR ? *next == '|' | |
25fe55af | 3821 | : next_backslash && next_next && *next_next == '|'); |
fa9a63c5 RM |
3822 | } |
3823 | ||
3824 | ||
5e69f11e | 3825 | /* Returns true if REGNUM is in one of COMPILE_STACK's elements and |
fa9a63c5 RM |
3826 | false if it's not. */ |
3827 | ||
3828 | static boolean | |
971de7fb | 3829 | group_in_compile_stack (compile_stack_type compile_stack, regnum_t regnum) |
fa9a63c5 | 3830 | { |
d1dfb56c | 3831 | ssize_t this_element; |
fa9a63c5 | 3832 | |
5e69f11e RM |
3833 | for (this_element = compile_stack.avail - 1; |
3834 | this_element >= 0; | |
fa9a63c5 RM |
3835 | this_element--) |
3836 | if (compile_stack.stack[this_element].regnum == regnum) | |
3837 | return true; | |
3838 | ||
3839 | return false; | |
3840 | } | |
fa9a63c5 | 3841 | \f |
f6a3f532 SM |
3842 | /* analyse_first. |
3843 | If fastmap is non-NULL, go through the pattern and fill fastmap | |
3844 | with all the possible leading chars. If fastmap is NULL, don't | |
3845 | bother filling it up (obviously) and only return whether the | |
3846 | pattern could potentially match the empty string. | |
3847 | ||
3848 | Return 1 if p..pend might match the empty string. | |
3849 | Return 0 if p..pend matches at least one char. | |
01618498 | 3850 | Return -1 if fastmap was not updated accurately. */ |
f6a3f532 SM |
3851 | |
3852 | static int | |
29abe551 PE |
3853 | analyse_first (const_re_char *p, const_re_char *pend, char *fastmap, |
3854 | const int multibyte) | |
fa9a63c5 | 3855 | { |
505bde11 | 3856 | int j, k; |
1fb352e0 | 3857 | boolean not; |
fa9a63c5 | 3858 | |
b18215fc | 3859 | /* If all elements for base leading-codes in fastmap is set, this |
7814e705 | 3860 | flag is set true. */ |
b18215fc RS |
3861 | boolean match_any_multibyte_characters = false; |
3862 | ||
f6a3f532 | 3863 | assert (p); |
5e69f11e | 3864 | |
505bde11 SM |
3865 | /* The loop below works as follows: |
3866 | - It has a working-list kept in the PATTERN_STACK and which basically | |
3867 | starts by only containing a pointer to the first operation. | |
3868 | - If the opcode we're looking at is a match against some set of | |
3869 | chars, then we add those chars to the fastmap and go on to the | |
3870 | next work element from the worklist (done via `break'). | |
3871 | - If the opcode is a control operator on the other hand, we either | |
3872 | ignore it (if it's meaningless at this point, such as `start_memory') | |
3873 | or execute it (if it's a jump). If the jump has several destinations | |
3874 | (i.e. `on_failure_jump'), then we push the other destination onto the | |
3875 | worklist. | |
3876 | We guarantee termination by ignoring backward jumps (more or less), | |
3877 | so that `p' is monotonically increasing. More to the point, we | |
3878 | never set `p' (or push) anything `<= p1'. */ | |
3879 | ||
01618498 | 3880 | while (p < pend) |
fa9a63c5 | 3881 | { |
505bde11 SM |
3882 | /* `p1' is used as a marker of how far back a `on_failure_jump' |
3883 | can go without being ignored. It is normally equal to `p' | |
3884 | (which prevents any backward `on_failure_jump') except right | |
3885 | after a plain `jump', to allow patterns such as: | |
3886 | 0: jump 10 | |
3887 | 3..9: <body> | |
3888 | 10: on_failure_jump 3 | |
3889 | as used for the *? operator. */ | |
01618498 | 3890 | re_char *p1 = p; |
5e69f11e | 3891 | |
7393bcbb | 3892 | switch (*p++) |
fa9a63c5 | 3893 | { |
f6a3f532 | 3894 | case succeed: |
01618498 | 3895 | return 1; |
fa9a63c5 | 3896 | |
fa9a63c5 | 3897 | case duplicate: |
505bde11 SM |
3898 | /* If the first character has to match a backreference, that means |
3899 | that the group was empty (since it already matched). Since this | |
3900 | is the only case that interests us here, we can assume that the | |
3901 | backreference must match the empty string. */ | |
3902 | p++; | |
3903 | continue; | |
fa9a63c5 RM |
3904 | |
3905 | ||
3906 | /* Following are the cases which match a character. These end | |
7814e705 | 3907 | with `break'. */ |
fa9a63c5 RM |
3908 | |
3909 | case exactn: | |
e0277a47 | 3910 | if (fastmap) |
cf9c99bc KH |
3911 | { |
3912 | /* If multibyte is nonzero, the first byte of each | |
3913 | character is an ASCII or a leading code. Otherwise, | |
3914 | each byte is a character. Thus, this works in both | |
3915 | cases. */ | |
3916 | fastmap[p[1]] = 1; | |
3917 | if (! multibyte) | |
3918 | { | |
3919 | /* For the case of matching this unibyte regex | |
3920 | against multibyte, we must set a leading code of | |
3921 | the corresponding multibyte character. */ | |
3922 | int c = RE_CHAR_TO_MULTIBYTE (p[1]); | |
3923 | ||
86e893e3 | 3924 | fastmap[CHAR_LEADING_CODE (c)] = 1; |
cf9c99bc KH |
3925 | } |
3926 | } | |
fa9a63c5 RM |
3927 | break; |
3928 | ||
3929 | ||
1fb352e0 SM |
3930 | case anychar: |
3931 | /* We could put all the chars except for \n (and maybe \0) | |
3932 | but we don't bother since it is generally not worth it. */ | |
f6a3f532 | 3933 | if (!fastmap) break; |
01618498 | 3934 | return -1; |
fa9a63c5 RM |
3935 | |
3936 | ||
b18215fc | 3937 | case charset_not: |
1fb352e0 | 3938 | if (!fastmap) break; |
bf216479 KH |
3939 | { |
3940 | /* Chars beyond end of bitmap are possible matches. */ | |
bf216479 | 3941 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH; |
cf9c99bc | 3942 | j < (1 << BYTEWIDTH); j++) |
bf216479 KH |
3943 | fastmap[j] = 1; |
3944 | } | |
3945 | ||
1fb352e0 SM |
3946 | /* Fallthrough */ |
3947 | case charset: | |
3948 | if (!fastmap) break; | |
3949 | not = (re_opcode_t) *(p - 1) == charset_not; | |
b18215fc RS |
3950 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH - 1, p++; |
3951 | j >= 0; j--) | |
1fb352e0 | 3952 | if (!!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) ^ not) |
49da453b | 3953 | fastmap[j] = 1; |
b18215fc | 3954 | |
6482db2e KH |
3955 | #ifdef emacs |
3956 | if (/* Any leading code can possibly start a character | |
1fb352e0 | 3957 | which doesn't match the specified set of characters. */ |
6482db2e | 3958 | not |
409f2919 | 3959 | || |
6482db2e KH |
3960 | /* If we can match a character class, we can match any |
3961 | multibyte characters. */ | |
3962 | (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) | |
3963 | && CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0)) | |
3964 | ||
b18215fc | 3965 | { |
b18215fc RS |
3966 | if (match_any_multibyte_characters == false) |
3967 | { | |
6482db2e KH |
3968 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
3969 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
6fdd04b0 | 3970 | fastmap[j] = 1; |
b18215fc RS |
3971 | match_any_multibyte_characters = true; |
3972 | } | |
3973 | } | |
b18215fc | 3974 | |
1fb352e0 SM |
3975 | else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) |
3976 | && match_any_multibyte_characters == false) | |
3977 | { | |
bf216479 | 3978 | /* Set fastmap[I] to 1 where I is a leading code of each |
51e4f4a8 | 3979 | multibyte character in the range table. */ |
1fb352e0 | 3980 | int c, count; |
bf216479 | 3981 | unsigned char lc1, lc2; |
b18215fc | 3982 | |
1fb352e0 | 3983 | /* Make P points the range table. `+ 2' is to skip flag |
0b32bf0e | 3984 | bits for a character class. */ |
1fb352e0 | 3985 | p += CHARSET_BITMAP_SIZE (&p[-2]) + 2; |
b18215fc | 3986 | |
1fb352e0 SM |
3987 | /* Extract the number of ranges in range table into COUNT. */ |
3988 | EXTRACT_NUMBER_AND_INCR (count, p); | |
cf9c99bc | 3989 | for (; count > 0; count--, p += 3) |
1fb352e0 | 3990 | { |
9117d724 KH |
3991 | /* Extract the start and end of each range. */ |
3992 | EXTRACT_CHARACTER (c, p); | |
bf216479 | 3993 | lc1 = CHAR_LEADING_CODE (c); |
9117d724 | 3994 | p += 3; |
1fb352e0 | 3995 | EXTRACT_CHARACTER (c, p); |
bf216479 KH |
3996 | lc2 = CHAR_LEADING_CODE (c); |
3997 | for (j = lc1; j <= lc2; j++) | |
9117d724 | 3998 | fastmap[j] = 1; |
1fb352e0 SM |
3999 | } |
4000 | } | |
6482db2e | 4001 | #endif |
b18215fc RS |
4002 | break; |
4003 | ||
1fb352e0 SM |
4004 | case syntaxspec: |
4005 | case notsyntaxspec: | |
4006 | if (!fastmap) break; | |
4007 | #ifndef emacs | |
4008 | not = (re_opcode_t)p[-1] == notsyntaxspec; | |
4009 | k = *p++; | |
4010 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
990b2375 | 4011 | if ((SYNTAX (j) == (enum syntaxcode) k) ^ not) |
b18215fc | 4012 | fastmap[j] = 1; |
b18215fc | 4013 | break; |
1fb352e0 | 4014 | #else /* emacs */ |
b18215fc RS |
4015 | /* This match depends on text properties. These end with |
4016 | aborting optimizations. */ | |
01618498 | 4017 | return -1; |
b18215fc RS |
4018 | |
4019 | case categoryspec: | |
b18215fc | 4020 | case notcategoryspec: |
1fb352e0 SM |
4021 | if (!fastmap) break; |
4022 | not = (re_opcode_t)p[-1] == notcategoryspec; | |
b18215fc | 4023 | k = *p++; |
6482db2e | 4024 | for (j = (1 << BYTEWIDTH); j >= 0; j--) |
1fb352e0 | 4025 | if ((CHAR_HAS_CATEGORY (j, k)) ^ not) |
b18215fc RS |
4026 | fastmap[j] = 1; |
4027 | ||
6482db2e KH |
4028 | /* Any leading code can possibly start a character which |
4029 | has or doesn't has the specified category. */ | |
4030 | if (match_any_multibyte_characters == false) | |
6fdd04b0 | 4031 | { |
6482db2e KH |
4032 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
4033 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
4034 | fastmap[j] = 1; | |
4035 | match_any_multibyte_characters = true; | |
6fdd04b0 | 4036 | } |
b18215fc RS |
4037 | break; |
4038 | ||
fa9a63c5 | 4039 | /* All cases after this match the empty string. These end with |
25fe55af | 4040 | `continue'. */ |
fa9a63c5 | 4041 | |
fa9a63c5 RM |
4042 | case before_dot: |
4043 | case at_dot: | |
4044 | case after_dot: | |
1fb352e0 | 4045 | #endif /* !emacs */ |
25fe55af RS |
4046 | case no_op: |
4047 | case begline: | |
4048 | case endline: | |
fa9a63c5 RM |
4049 | case begbuf: |
4050 | case endbuf: | |
4051 | case wordbound: | |
4052 | case notwordbound: | |
4053 | case wordbeg: | |
4054 | case wordend: | |
669fa600 SM |
4055 | case symbeg: |
4056 | case symend: | |
25fe55af | 4057 | continue; |
fa9a63c5 RM |
4058 | |
4059 | ||
fa9a63c5 | 4060 | case jump: |
25fe55af | 4061 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 SM |
4062 | if (j < 0) |
4063 | /* Backward jumps can only go back to code that we've already | |
4064 | visited. `re_compile' should make sure this is true. */ | |
4065 | break; | |
25fe55af | 4066 | p += j; |
7393bcbb | 4067 | switch (*p) |
505bde11 SM |
4068 | { |
4069 | case on_failure_jump: | |
4070 | case on_failure_keep_string_jump: | |
505bde11 | 4071 | case on_failure_jump_loop: |
0683b6fa | 4072 | case on_failure_jump_nastyloop: |
505bde11 SM |
4073 | case on_failure_jump_smart: |
4074 | p++; | |
4075 | break; | |
4076 | default: | |
4077 | continue; | |
4078 | }; | |
4079 | /* Keep `p1' to allow the `on_failure_jump' we are jumping to | |
4080 | to jump back to "just after here". */ | |
4081 | /* Fallthrough */ | |
fa9a63c5 | 4082 | |
25fe55af RS |
4083 | case on_failure_jump: |
4084 | case on_failure_keep_string_jump: | |
0683b6fa | 4085 | case on_failure_jump_nastyloop: |
505bde11 SM |
4086 | case on_failure_jump_loop: |
4087 | case on_failure_jump_smart: | |
25fe55af | 4088 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 | 4089 | if (p + j <= p1) |
ed0767d8 | 4090 | ; /* Backward jump to be ignored. */ |
01618498 SM |
4091 | else |
4092 | { /* We have to look down both arms. | |
4093 | We first go down the "straight" path so as to minimize | |
4094 | stack usage when going through alternatives. */ | |
4095 | int r = analyse_first (p, pend, fastmap, multibyte); | |
4096 | if (r) return r; | |
4097 | p += j; | |
4098 | } | |
25fe55af | 4099 | continue; |
fa9a63c5 RM |
4100 | |
4101 | ||
ed0767d8 SM |
4102 | case jump_n: |
4103 | /* This code simply does not properly handle forward jump_n. */ | |
4104 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p); assert (j < 0)); | |
4105 | p += 4; | |
4106 | /* jump_n can either jump or fall through. The (backward) jump | |
4107 | case has already been handled, so we only need to look at the | |
4108 | fallthrough case. */ | |
4109 | continue; | |
177c0ea7 | 4110 | |
fa9a63c5 | 4111 | case succeed_n: |
ed0767d8 SM |
4112 | /* If N == 0, it should be an on_failure_jump_loop instead. */ |
4113 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0)); | |
4114 | p += 4; | |
4115 | /* We only care about one iteration of the loop, so we don't | |
4116 | need to consider the case where this behaves like an | |
4117 | on_failure_jump. */ | |
25fe55af | 4118 | continue; |
fa9a63c5 RM |
4119 | |
4120 | ||
4121 | case set_number_at: | |
25fe55af RS |
4122 | p += 4; |
4123 | continue; | |
fa9a63c5 RM |
4124 | |
4125 | ||
4126 | case start_memory: | |
25fe55af | 4127 | case stop_memory: |
505bde11 | 4128 | p += 1; |
fa9a63c5 RM |
4129 | continue; |
4130 | ||
4131 | ||
4132 | default: | |
25fe55af RS |
4133 | abort (); /* We have listed all the cases. */ |
4134 | } /* switch *p++ */ | |
fa9a63c5 RM |
4135 | |
4136 | /* Getting here means we have found the possible starting | |
25fe55af | 4137 | characters for one path of the pattern -- and that the empty |
7814e705 | 4138 | string does not match. We need not follow this path further. */ |
01618498 | 4139 | return 0; |
fa9a63c5 RM |
4140 | } /* while p */ |
4141 | ||
01618498 SM |
4142 | /* We reached the end without matching anything. */ |
4143 | return 1; | |
4144 | ||
f6a3f532 SM |
4145 | } /* analyse_first */ |
4146 | \f | |
4147 | /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in | |
4148 | BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible | |
4149 | characters can start a string that matches the pattern. This fastmap | |
4150 | is used by re_search to skip quickly over impossible starting points. | |
4151 | ||
4152 | Character codes above (1 << BYTEWIDTH) are not represented in the | |
4153 | fastmap, but the leading codes are represented. Thus, the fastmap | |
4154 | indicates which character sets could start a match. | |
4155 | ||
4156 | The caller must supply the address of a (1 << BYTEWIDTH)-byte data | |
4157 | area as BUFP->fastmap. | |
4158 | ||
4159 | We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in | |
4160 | the pattern buffer. | |
4161 | ||
4162 | Returns 0 if we succeed, -2 if an internal error. */ | |
4163 | ||
4164 | int | |
971de7fb | 4165 | re_compile_fastmap (struct re_pattern_buffer *bufp) |
f6a3f532 SM |
4166 | { |
4167 | char *fastmap = bufp->fastmap; | |
4168 | int analysis; | |
4169 | ||
4170 | assert (fastmap && bufp->buffer); | |
4171 | ||
72af86bd | 4172 | memset (fastmap, 0, 1 << BYTEWIDTH); /* Assume nothing's valid. */ |
f6a3f532 SM |
4173 | bufp->fastmap_accurate = 1; /* It will be when we're done. */ |
4174 | ||
4175 | analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used, | |
2d1675e4 | 4176 | fastmap, RE_MULTIBYTE_P (bufp)); |
c0f9ea08 | 4177 | bufp->can_be_null = (analysis != 0); |
fa9a63c5 RM |
4178 | return 0; |
4179 | } /* re_compile_fastmap */ | |
4180 | \f | |
4181 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | |
4182 | ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | |
4183 | this memory for recording register information. STARTS and ENDS | |
4184 | must be allocated using the malloc library routine, and must each | |
4185 | be at least NUM_REGS * sizeof (regoff_t) bytes long. | |
4186 | ||
4187 | If NUM_REGS == 0, then subsequent matches should allocate their own | |
4188 | register data. | |
4189 | ||
4190 | Unless this function is called, the first search or match using | |
4191 | PATTERN_BUFFER will allocate its own register data, without | |
4192 | freeing the old data. */ | |
4193 | ||
4194 | void | |
971de7fb | 4195 | re_set_registers (struct re_pattern_buffer *bufp, struct re_registers *regs, unsigned int num_regs, regoff_t *starts, regoff_t *ends) |
fa9a63c5 RM |
4196 | { |
4197 | if (num_regs) | |
4198 | { | |
4199 | bufp->regs_allocated = REGS_REALLOCATE; | |
4200 | regs->num_regs = num_regs; | |
4201 | regs->start = starts; | |
4202 | regs->end = ends; | |
4203 | } | |
4204 | else | |
4205 | { | |
4206 | bufp->regs_allocated = REGS_UNALLOCATED; | |
4207 | regs->num_regs = 0; | |
7d652d97 | 4208 | regs->start = regs->end = 0; |
fa9a63c5 RM |
4209 | } |
4210 | } | |
c0f9ea08 | 4211 | WEAK_ALIAS (__re_set_registers, re_set_registers) |
fa9a63c5 | 4212 | \f |
7814e705 | 4213 | /* Searching routines. */ |
fa9a63c5 RM |
4214 | |
4215 | /* Like re_search_2, below, but only one string is specified, and | |
4216 | doesn't let you say where to stop matching. */ | |
4217 | ||
d1dfb56c EZ |
4218 | regoff_t |
4219 | re_search (struct re_pattern_buffer *bufp, const char *string, size_t size, | |
4220 | ssize_t startpos, ssize_t range, struct re_registers *regs) | |
fa9a63c5 | 4221 | { |
5e69f11e | 4222 | return re_search_2 (bufp, NULL, 0, string, size, startpos, range, |
fa9a63c5 RM |
4223 | regs, size); |
4224 | } | |
c0f9ea08 | 4225 | WEAK_ALIAS (__re_search, re_search) |
fa9a63c5 | 4226 | |
70806df6 KH |
4227 | /* Head address of virtual concatenation of string. */ |
4228 | #define HEAD_ADDR_VSTRING(P) \ | |
4229 | (((P) >= size1 ? string2 : string1)) | |
4230 | ||
b18215fc RS |
4231 | /* Address of POS in the concatenation of virtual string. */ |
4232 | #define POS_ADDR_VSTRING(POS) \ | |
4233 | (((POS) >= size1 ? string2 - size1 : string1) + (POS)) | |
fa9a63c5 RM |
4234 | |
4235 | /* Using the compiled pattern in BUFP->buffer, first tries to match the | |
4236 | virtual concatenation of STRING1 and STRING2, starting first at index | |
4237 | STARTPOS, then at STARTPOS + 1, and so on. | |
5e69f11e | 4238 | |
fa9a63c5 | 4239 | STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. |
5e69f11e | 4240 | |
fa9a63c5 RM |
4241 | RANGE is how far to scan while trying to match. RANGE = 0 means try |
4242 | only at STARTPOS; in general, the last start tried is STARTPOS + | |
4243 | RANGE. | |
5e69f11e | 4244 | |
fa9a63c5 RM |
4245 | In REGS, return the indices of the virtual concatenation of STRING1 |
4246 | and STRING2 that matched the entire BUFP->buffer and its contained | |
4247 | subexpressions. | |
5e69f11e | 4248 | |
fa9a63c5 RM |
4249 | Do not consider matching one past the index STOP in the virtual |
4250 | concatenation of STRING1 and STRING2. | |
4251 | ||
4252 | We return either the position in the strings at which the match was | |
4253 | found, -1 if no match, or -2 if error (such as failure | |
4254 | stack overflow). */ | |
4255 | ||
d1dfb56c EZ |
4256 | regoff_t |
4257 | re_search_2 (struct re_pattern_buffer *bufp, const char *str1, size_t size1, | |
4258 | const char *str2, size_t size2, ssize_t startpos, ssize_t range, | |
4259 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4260 | { |
d1dfb56c | 4261 | regoff_t val; |
66f0296e SM |
4262 | re_char *string1 = (re_char*) str1; |
4263 | re_char *string2 = (re_char*) str2; | |
fa9a63c5 | 4264 | register char *fastmap = bufp->fastmap; |
6676cb1c | 4265 | register RE_TRANSLATE_TYPE translate = bufp->translate; |
d1dfb56c EZ |
4266 | size_t total_size = size1 + size2; |
4267 | ssize_t endpos = startpos + range; | |
c0f9ea08 | 4268 | boolean anchored_start; |
cf9c99bc KH |
4269 | /* Nonzero if we are searching multibyte string. */ |
4270 | const boolean multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
b18215fc | 4271 | |
fa9a63c5 RM |
4272 | /* Check for out-of-range STARTPOS. */ |
4273 | if (startpos < 0 || startpos > total_size) | |
4274 | return -1; | |
5e69f11e | 4275 | |
fa9a63c5 | 4276 | /* Fix up RANGE if it might eventually take us outside |
34597fa9 | 4277 | the virtual concatenation of STRING1 and STRING2. |
5e69f11e | 4278 | Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ |
34597fa9 RS |
4279 | if (endpos < 0) |
4280 | range = 0 - startpos; | |
fa9a63c5 RM |
4281 | else if (endpos > total_size) |
4282 | range = total_size - startpos; | |
4283 | ||
4284 | /* If the search isn't to be a backwards one, don't waste time in a | |
7b140fd7 | 4285 | search for a pattern anchored at beginning of buffer. */ |
fa9a63c5 RM |
4286 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) |
4287 | { | |
4288 | if (startpos > 0) | |
4289 | return -1; | |
4290 | else | |
7b140fd7 | 4291 | range = 0; |
fa9a63c5 RM |
4292 | } |
4293 | ||
ae4788a8 RS |
4294 | #ifdef emacs |
4295 | /* In a forward search for something that starts with \=. | |
4296 | don't keep searching past point. */ | |
4297 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) | |
4298 | { | |
7b140fd7 RS |
4299 | range = PT_BYTE - BEGV_BYTE - startpos; |
4300 | if (range < 0) | |
ae4788a8 RS |
4301 | return -1; |
4302 | } | |
4303 | #endif /* emacs */ | |
4304 | ||
fa9a63c5 RM |
4305 | /* Update the fastmap now if not correct already. */ |
4306 | if (fastmap && !bufp->fastmap_accurate) | |
01618498 | 4307 | re_compile_fastmap (bufp); |
5e69f11e | 4308 | |
c8499ba5 | 4309 | /* See whether the pattern is anchored. */ |
c0f9ea08 | 4310 | anchored_start = (bufp->buffer[0] == begline); |
c8499ba5 | 4311 | |
b18215fc | 4312 | #ifdef emacs |
d48cd3f4 | 4313 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
cc9b4df2 | 4314 | { |
d1dfb56c | 4315 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (startpos)); |
cc9b4df2 KH |
4316 | |
4317 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); | |
4318 | } | |
b18215fc RS |
4319 | #endif |
4320 | ||
fa9a63c5 RM |
4321 | /* Loop through the string, looking for a place to start matching. */ |
4322 | for (;;) | |
5e69f11e | 4323 | { |
c8499ba5 RS |
4324 | /* If the pattern is anchored, |
4325 | skip quickly past places we cannot match. | |
4326 | We don't bother to treat startpos == 0 specially | |
4327 | because that case doesn't repeat. */ | |
4328 | if (anchored_start && startpos > 0) | |
4329 | { | |
c0f9ea08 SM |
4330 | if (! ((startpos <= size1 ? string1[startpos - 1] |
4331 | : string2[startpos - size1 - 1]) | |
4332 | == '\n')) | |
c8499ba5 RS |
4333 | goto advance; |
4334 | } | |
4335 | ||
fa9a63c5 | 4336 | /* If a fastmap is supplied, skip quickly over characters that |
25fe55af RS |
4337 | cannot be the start of a match. If the pattern can match the |
4338 | null string, however, we don't need to skip characters; we want | |
7814e705 | 4339 | the first null string. */ |
fa9a63c5 RM |
4340 | if (fastmap && startpos < total_size && !bufp->can_be_null) |
4341 | { | |
66f0296e | 4342 | register re_char *d; |
01618498 | 4343 | register re_wchar_t buf_ch; |
e934739e RS |
4344 | |
4345 | d = POS_ADDR_VSTRING (startpos); | |
4346 | ||
7814e705 | 4347 | if (range > 0) /* Searching forwards. */ |
fa9a63c5 | 4348 | { |
fa9a63c5 | 4349 | register int lim = 0; |
d1dfb56c | 4350 | ssize_t irange = range; |
fa9a63c5 | 4351 | |
25fe55af RS |
4352 | if (startpos < size1 && startpos + range >= size1) |
4353 | lim = range - (size1 - startpos); | |
fa9a63c5 | 4354 | |
25fe55af RS |
4355 | /* Written out as an if-else to avoid testing `translate' |
4356 | inside the loop. */ | |
28ae27ae AS |
4357 | if (RE_TRANSLATE_P (translate)) |
4358 | { | |
e934739e RS |
4359 | if (multibyte) |
4360 | while (range > lim) | |
4361 | { | |
4362 | int buf_charlen; | |
4363 | ||
62a6e103 | 4364 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 4365 | buf_ch = RE_TRANSLATE (translate, buf_ch); |
bf216479 | 4366 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
e934739e RS |
4367 | break; |
4368 | ||
4369 | range -= buf_charlen; | |
4370 | d += buf_charlen; | |
4371 | } | |
4372 | else | |
bf216479 | 4373 | while (range > lim) |
33c46939 | 4374 | { |
cf9c99bc KH |
4375 | register re_wchar_t ch, translated; |
4376 | ||
bf216479 | 4377 | buf_ch = *d; |
cf9c99bc KH |
4378 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); |
4379 | translated = RE_TRANSLATE (translate, ch); | |
4380 | if (translated != ch | |
4381 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4382 | buf_ch = ch; | |
6fdd04b0 | 4383 | if (fastmap[buf_ch]) |
bf216479 | 4384 | break; |
33c46939 RS |
4385 | d++; |
4386 | range--; | |
4387 | } | |
e934739e | 4388 | } |
fa9a63c5 | 4389 | else |
6fdd04b0 KH |
4390 | { |
4391 | if (multibyte) | |
4392 | while (range > lim) | |
4393 | { | |
4394 | int buf_charlen; | |
fa9a63c5 | 4395 | |
62a6e103 | 4396 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
6fdd04b0 KH |
4397 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
4398 | break; | |
4399 | range -= buf_charlen; | |
4400 | d += buf_charlen; | |
4401 | } | |
e934739e | 4402 | else |
6fdd04b0 | 4403 | while (range > lim && !fastmap[*d]) |
33c46939 RS |
4404 | { |
4405 | d++; | |
4406 | range--; | |
4407 | } | |
e934739e | 4408 | } |
fa9a63c5 RM |
4409 | startpos += irange - range; |
4410 | } | |
7814e705 | 4411 | else /* Searching backwards. */ |
fa9a63c5 | 4412 | { |
ba5e343c KH |
4413 | if (multibyte) |
4414 | { | |
62a6e103 | 4415 | buf_ch = STRING_CHAR (d); |
ba5e343c KH |
4416 | buf_ch = TRANSLATE (buf_ch); |
4417 | if (! fastmap[CHAR_LEADING_CODE (buf_ch)]) | |
4418 | goto advance; | |
4419 | } | |
4420 | else | |
4421 | { | |
cf9c99bc KH |
4422 | register re_wchar_t ch, translated; |
4423 | ||
4424 | buf_ch = *d; | |
4425 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); | |
4426 | translated = TRANSLATE (ch); | |
4427 | if (translated != ch | |
4428 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4429 | buf_ch = ch; | |
4430 | if (! fastmap[TRANSLATE (buf_ch)]) | |
ba5e343c KH |
4431 | goto advance; |
4432 | } | |
fa9a63c5 RM |
4433 | } |
4434 | } | |
4435 | ||
4436 | /* If can't match the null string, and that's all we have left, fail. */ | |
4437 | if (range >= 0 && startpos == total_size && fastmap | |
25fe55af | 4438 | && !bufp->can_be_null) |
fa9a63c5 RM |
4439 | return -1; |
4440 | ||
4441 | val = re_match_2_internal (bufp, string1, size1, string2, size2, | |
4442 | startpos, regs, stop); | |
fa9a63c5 RM |
4443 | |
4444 | if (val >= 0) | |
4445 | return startpos; | |
5e69f11e | 4446 | |
fa9a63c5 RM |
4447 | if (val == -2) |
4448 | return -2; | |
4449 | ||
4450 | advance: | |
5e69f11e | 4451 | if (!range) |
25fe55af | 4452 | break; |
5e69f11e | 4453 | else if (range > 0) |
25fe55af | 4454 | { |
b18215fc RS |
4455 | /* Update STARTPOS to the next character boundary. */ |
4456 | if (multibyte) | |
4457 | { | |
66f0296e | 4458 | re_char *p = POS_ADDR_VSTRING (startpos); |
aa3830c4 | 4459 | int len = BYTES_BY_CHAR_HEAD (*p); |
b18215fc RS |
4460 | |
4461 | range -= len; | |
4462 | if (range < 0) | |
4463 | break; | |
4464 | startpos += len; | |
4465 | } | |
4466 | else | |
4467 | { | |
b560c397 RS |
4468 | range--; |
4469 | startpos++; | |
4470 | } | |
e318085a | 4471 | } |
fa9a63c5 | 4472 | else |
25fe55af RS |
4473 | { |
4474 | range++; | |
4475 | startpos--; | |
b18215fc RS |
4476 | |
4477 | /* Update STARTPOS to the previous character boundary. */ | |
4478 | if (multibyte) | |
4479 | { | |
70806df6 KH |
4480 | re_char *p = POS_ADDR_VSTRING (startpos) + 1; |
4481 | re_char *p0 = p; | |
4482 | re_char *phead = HEAD_ADDR_VSTRING (startpos); | |
b18215fc RS |
4483 | |
4484 | /* Find the head of multibyte form. */ | |
70806df6 KH |
4485 | PREV_CHAR_BOUNDARY (p, phead); |
4486 | range += p0 - 1 - p; | |
4487 | if (range > 0) | |
4488 | break; | |
b18215fc | 4489 | |
70806df6 | 4490 | startpos -= p0 - 1 - p; |
b18215fc | 4491 | } |
25fe55af | 4492 | } |
fa9a63c5 RM |
4493 | } |
4494 | return -1; | |
4495 | } /* re_search_2 */ | |
c0f9ea08 | 4496 | WEAK_ALIAS (__re_search_2, re_search_2) |
fa9a63c5 RM |
4497 | \f |
4498 | /* Declarations and macros for re_match_2. */ | |
4499 | ||
261cb4bb PE |
4500 | static int bcmp_translate (re_char *s1, re_char *s2, |
4501 | register ssize_t len, | |
4502 | RE_TRANSLATE_TYPE translate, | |
4503 | const int multibyte); | |
fa9a63c5 RM |
4504 | |
4505 | /* This converts PTR, a pointer into one of the search strings `string1' | |
4506 | and `string2' into an offset from the beginning of that string. */ | |
4507 | #define POINTER_TO_OFFSET(ptr) \ | |
4508 | (FIRST_STRING_P (ptr) \ | |
dc4a2ee0 PE |
4509 | ? (ptr) - string1 \ |
4510 | : (ptr) - string2 + (ptrdiff_t) size1) | |
fa9a63c5 | 4511 | |
fa9a63c5 | 4512 | /* Call before fetching a character with *d. This switches over to |
419d1c74 SM |
4513 | string2 if necessary. |
4514 | Check re_match_2_internal for a discussion of why end_match_2 might | |
4515 | not be within string2 (but be equal to end_match_1 instead). */ | |
fa9a63c5 | 4516 | #define PREFETCH() \ |
25fe55af | 4517 | while (d == dend) \ |
fa9a63c5 RM |
4518 | { \ |
4519 | /* End of string2 => fail. */ \ | |
25fe55af RS |
4520 | if (dend == end_match_2) \ |
4521 | goto fail; \ | |
4bb91c68 | 4522 | /* End of string1 => advance to string2. */ \ |
25fe55af | 4523 | d = string2; \ |
fa9a63c5 RM |
4524 | dend = end_match_2; \ |
4525 | } | |
4526 | ||
f1ad044f SM |
4527 | /* Call before fetching a char with *d if you already checked other limits. |
4528 | This is meant for use in lookahead operations like wordend, etc.. | |
4529 | where we might need to look at parts of the string that might be | |
4530 | outside of the LIMITs (i.e past `stop'). */ | |
4531 | #define PREFETCH_NOLIMIT() \ | |
4532 | if (d == end1) \ | |
4533 | { \ | |
4534 | d = string2; \ | |
4535 | dend = end_match_2; \ | |
4536 | } \ | |
fa9a63c5 RM |
4537 | |
4538 | /* Test if at very beginning or at very end of the virtual concatenation | |
7814e705 | 4539 | of `string1' and `string2'. If only one string, it's `string2'. */ |
fa9a63c5 | 4540 | #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) |
5e69f11e | 4541 | #define AT_STRINGS_END(d) ((d) == end2) |
fa9a63c5 | 4542 | |
9121ca40 | 4543 | /* Disabled due to a compiler bug -- see comment at case wordbound */ |
b18215fc RS |
4544 | |
4545 | /* The comment at case wordbound is following one, but we don't use | |
4546 | AT_WORD_BOUNDARY anymore to support multibyte form. | |
4547 | ||
4548 | The DEC Alpha C compiler 3.x generates incorrect code for the | |
25fe55af | 4549 | test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of |
7814e705 | 4550 | AT_WORD_BOUNDARY, so this code is disabled. Expanding the |
b18215fc RS |
4551 | macro and introducing temporary variables works around the bug. */ |
4552 | ||
9121ca40 | 4553 | #if 0 |
b313f9d8 PE |
4554 | /* Test if D points to a character which is word-constituent. We have |
4555 | two special cases to check for: if past the end of string1, look at | |
4556 | the first character in string2; and if before the beginning of | |
4557 | string2, look at the last character in string1. */ | |
4558 | #define WORDCHAR_P(d) \ | |
4559 | (SYNTAX ((d) == end1 ? *string2 \ | |
4560 | : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ | |
4561 | == Sword) | |
4562 | ||
fa9a63c5 RM |
4563 | /* Test if the character before D and the one at D differ with respect |
4564 | to being word-constituent. */ | |
4565 | #define AT_WORD_BOUNDARY(d) \ | |
4566 | (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ | |
4567 | || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) | |
9121ca40 | 4568 | #endif |
fa9a63c5 RM |
4569 | |
4570 | /* Free everything we malloc. */ | |
4571 | #ifdef MATCH_MAY_ALLOCATE | |
952db0d7 PE |
4572 | # define FREE_VAR(var) \ |
4573 | do { \ | |
4574 | if (var) \ | |
4575 | { \ | |
4576 | REGEX_FREE (var); \ | |
4577 | var = NULL; \ | |
4578 | } \ | |
4579 | } while (0) | |
0b32bf0e | 4580 | # define FREE_VARIABLES() \ |
fa9a63c5 RM |
4581 | do { \ |
4582 | REGEX_FREE_STACK (fail_stack.stack); \ | |
4583 | FREE_VAR (regstart); \ | |
4584 | FREE_VAR (regend); \ | |
fa9a63c5 RM |
4585 | FREE_VAR (best_regstart); \ |
4586 | FREE_VAR (best_regend); \ | |
fa9a63c5 RM |
4587 | } while (0) |
4588 | #else | |
0b32bf0e | 4589 | # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
4590 | #endif /* not MATCH_MAY_ALLOCATE */ |
4591 | ||
505bde11 SM |
4592 | \f |
4593 | /* Optimization routines. */ | |
4594 | ||
4e8a9132 SM |
4595 | /* If the operation is a match against one or more chars, |
4596 | return a pointer to the next operation, else return NULL. */ | |
01618498 | 4597 | static re_char * |
29abe551 | 4598 | skip_one_char (const_re_char *p) |
4e8a9132 | 4599 | { |
7393bcbb | 4600 | switch (*p++) |
4e8a9132 SM |
4601 | { |
4602 | case anychar: | |
4603 | break; | |
177c0ea7 | 4604 | |
4e8a9132 SM |
4605 | case exactn: |
4606 | p += *p + 1; | |
4607 | break; | |
4608 | ||
4609 | case charset_not: | |
4610 | case charset: | |
4611 | if (CHARSET_RANGE_TABLE_EXISTS_P (p - 1)) | |
4612 | { | |
4613 | int mcnt; | |
4614 | p = CHARSET_RANGE_TABLE (p - 1); | |
4615 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4616 | p = CHARSET_RANGE_TABLE_END (p, mcnt); | |
4617 | } | |
4618 | else | |
4619 | p += 1 + CHARSET_BITMAP_SIZE (p - 1); | |
4620 | break; | |
177c0ea7 | 4621 | |
4e8a9132 SM |
4622 | case syntaxspec: |
4623 | case notsyntaxspec: | |
1fb352e0 | 4624 | #ifdef emacs |
4e8a9132 SM |
4625 | case categoryspec: |
4626 | case notcategoryspec: | |
4627 | #endif /* emacs */ | |
4628 | p++; | |
4629 | break; | |
4630 | ||
4631 | default: | |
4632 | p = NULL; | |
4633 | } | |
4634 | return p; | |
4635 | } | |
4636 | ||
4637 | ||
505bde11 | 4638 | /* Jump over non-matching operations. */ |
839966f3 | 4639 | static re_char * |
29abe551 | 4640 | skip_noops (const_re_char *p, const_re_char *pend) |
505bde11 SM |
4641 | { |
4642 | int mcnt; | |
4643 | while (p < pend) | |
4644 | { | |
7393bcbb | 4645 | switch (*p) |
505bde11 SM |
4646 | { |
4647 | case start_memory: | |
505bde11 SM |
4648 | case stop_memory: |
4649 | p += 2; break; | |
4650 | case no_op: | |
4651 | p += 1; break; | |
4652 | case jump: | |
4653 | p += 1; | |
4654 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4655 | p += mcnt; | |
4656 | break; | |
4657 | default: | |
4658 | return p; | |
4659 | } | |
4660 | } | |
4661 | assert (p == pend); | |
4662 | return p; | |
4663 | } | |
4664 | ||
4665 | /* Non-zero if "p1 matches something" implies "p2 fails". */ | |
4666 | static int | |
29abe551 PE |
4667 | mutually_exclusive_p (struct re_pattern_buffer *bufp, const_re_char *p1, |
4668 | const_re_char *p2) | |
505bde11 | 4669 | { |
4e8a9132 | 4670 | re_opcode_t op2; |
2d1675e4 | 4671 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
505bde11 SM |
4672 | unsigned char *pend = bufp->buffer + bufp->used; |
4673 | ||
4e8a9132 | 4674 | assert (p1 >= bufp->buffer && p1 < pend |
505bde11 SM |
4675 | && p2 >= bufp->buffer && p2 <= pend); |
4676 | ||
4677 | /* Skip over open/close-group commands. | |
4678 | If what follows this loop is a ...+ construct, | |
4679 | look at what begins its body, since we will have to | |
4680 | match at least one of that. */ | |
4e8a9132 SM |
4681 | p2 = skip_noops (p2, pend); |
4682 | /* The same skip can be done for p1, except that this function | |
4683 | is only used in the case where p1 is a simple match operator. */ | |
4684 | /* p1 = skip_noops (p1, pend); */ | |
4685 | ||
4686 | assert (p1 >= bufp->buffer && p1 < pend | |
4687 | && p2 >= bufp->buffer && p2 <= pend); | |
4688 | ||
4689 | op2 = p2 == pend ? succeed : *p2; | |
4690 | ||
7393bcbb | 4691 | switch (op2) |
505bde11 | 4692 | { |
4e8a9132 SM |
4693 | case succeed: |
4694 | case endbuf: | |
4695 | /* If we're at the end of the pattern, we can change. */ | |
4696 | if (skip_one_char (p1)) | |
505bde11 | 4697 | { |
dc4a2ee0 | 4698 | DEBUG_PRINT (" End of pattern: fast loop.\n"); |
505bde11 | 4699 | return 1; |
505bde11 | 4700 | } |
4e8a9132 | 4701 | break; |
177c0ea7 | 4702 | |
4e8a9132 | 4703 | case endline: |
4e8a9132 SM |
4704 | case exactn: |
4705 | { | |
01618498 | 4706 | register re_wchar_t c |
4e8a9132 | 4707 | = (re_opcode_t) *p2 == endline ? '\n' |
62a6e103 | 4708 | : RE_STRING_CHAR (p2 + 2, multibyte); |
505bde11 | 4709 | |
4e8a9132 SM |
4710 | if ((re_opcode_t) *p1 == exactn) |
4711 | { | |
62a6e103 | 4712 | if (c != RE_STRING_CHAR (p1 + 2, multibyte)) |
4e8a9132 | 4713 | { |
dc4a2ee0 | 4714 | DEBUG_PRINT (" '%c' != '%c' => fast loop.\n", c, p1[2]); |
4e8a9132 SM |
4715 | return 1; |
4716 | } | |
4717 | } | |
505bde11 | 4718 | |
4e8a9132 SM |
4719 | else if ((re_opcode_t) *p1 == charset |
4720 | || (re_opcode_t) *p1 == charset_not) | |
4721 | { | |
4722 | int not = (re_opcode_t) *p1 == charset_not; | |
505bde11 | 4723 | |
4e8a9132 SM |
4724 | /* Test if C is listed in charset (or charset_not) |
4725 | at `p1'. */ | |
6fdd04b0 | 4726 | if (! multibyte || IS_REAL_ASCII (c)) |
4e8a9132 SM |
4727 | { |
4728 | if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH | |
4729 | && p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
4730 | not = !not; | |
4731 | } | |
4732 | else if (CHARSET_RANGE_TABLE_EXISTS_P (p1)) | |
4733 | CHARSET_LOOKUP_RANGE_TABLE (not, c, p1); | |
505bde11 | 4734 | |
4e8a9132 SM |
4735 | /* `not' is equal to 1 if c would match, which means |
4736 | that we can't change to pop_failure_jump. */ | |
4737 | if (!not) | |
4738 | { | |
dc4a2ee0 | 4739 | DEBUG_PRINT (" No match => fast loop.\n"); |
4e8a9132 SM |
4740 | return 1; |
4741 | } | |
4742 | } | |
4743 | else if ((re_opcode_t) *p1 == anychar | |
4744 | && c == '\n') | |
4745 | { | |
dc4a2ee0 | 4746 | DEBUG_PRINT (" . != \\n => fast loop.\n"); |
4e8a9132 SM |
4747 | return 1; |
4748 | } | |
4749 | } | |
4750 | break; | |
505bde11 | 4751 | |
4e8a9132 | 4752 | case charset: |
4e8a9132 SM |
4753 | { |
4754 | if ((re_opcode_t) *p1 == exactn) | |
4755 | /* Reuse the code above. */ | |
4756 | return mutually_exclusive_p (bufp, p2, p1); | |
505bde11 | 4757 | |
505bde11 SM |
4758 | /* It is hard to list up all the character in charset |
4759 | P2 if it includes multibyte character. Give up in | |
4760 | such case. */ | |
4761 | else if (!multibyte || !CHARSET_RANGE_TABLE_EXISTS_P (p2)) | |
4762 | { | |
4763 | /* Now, we are sure that P2 has no range table. | |
4764 | So, for the size of bitmap in P2, `p2[1]' is | |
7814e705 | 4765 | enough. But P1 may have range table, so the |
505bde11 SM |
4766 | size of bitmap table of P1 is extracted by |
4767 | using macro `CHARSET_BITMAP_SIZE'. | |
4768 | ||
6fdd04b0 KH |
4769 | In a multibyte case, we know that all the character |
4770 | listed in P2 is ASCII. In a unibyte case, P1 has only a | |
4771 | bitmap table. So, in both cases, it is enough to test | |
4772 | only the bitmap table of P1. */ | |
505bde11 | 4773 | |
411e4203 | 4774 | if ((re_opcode_t) *p1 == charset) |
505bde11 SM |
4775 | { |
4776 | int idx; | |
4777 | /* We win if the charset inside the loop | |
4778 | has no overlap with the one after the loop. */ | |
4779 | for (idx = 0; | |
4780 | (idx < (int) p2[1] | |
4781 | && idx < CHARSET_BITMAP_SIZE (p1)); | |
4782 | idx++) | |
4783 | if ((p2[2 + idx] & p1[2 + idx]) != 0) | |
4784 | break; | |
4785 | ||
4786 | if (idx == p2[1] | |
4787 | || idx == CHARSET_BITMAP_SIZE (p1)) | |
4788 | { | |
dc4a2ee0 | 4789 | DEBUG_PRINT (" No match => fast loop.\n"); |
505bde11 SM |
4790 | return 1; |
4791 | } | |
4792 | } | |
411e4203 | 4793 | else if ((re_opcode_t) *p1 == charset_not) |
505bde11 SM |
4794 | { |
4795 | int idx; | |
4796 | /* We win if the charset_not inside the loop lists | |
7814e705 | 4797 | every character listed in the charset after. */ |
505bde11 SM |
4798 | for (idx = 0; idx < (int) p2[1]; idx++) |
4799 | if (! (p2[2 + idx] == 0 | |
4800 | || (idx < CHARSET_BITMAP_SIZE (p1) | |
4801 | && ((p2[2 + idx] & ~ p1[2 + idx]) == 0)))) | |
4802 | break; | |
4803 | ||
d1dfb56c EZ |
4804 | if (idx == p2[1]) |
4805 | { | |
dc4a2ee0 | 4806 | DEBUG_PRINT (" No match => fast loop.\n"); |
d1dfb56c EZ |
4807 | return 1; |
4808 | } | |
4e8a9132 SM |
4809 | } |
4810 | } | |
4811 | } | |
609b757a | 4812 | break; |
177c0ea7 | 4813 | |
411e4203 | 4814 | case charset_not: |
7393bcbb | 4815 | switch (*p1) |
411e4203 SM |
4816 | { |
4817 | case exactn: | |
4818 | case charset: | |
4819 | /* Reuse the code above. */ | |
4820 | return mutually_exclusive_p (bufp, p2, p1); | |
4821 | case charset_not: | |
4822 | /* When we have two charset_not, it's very unlikely that | |
4823 | they don't overlap. The union of the two sets of excluded | |
4824 | chars should cover all possible chars, which, as a matter of | |
4825 | fact, is virtually impossible in multibyte buffers. */ | |
36595814 | 4826 | break; |
411e4203 SM |
4827 | } |
4828 | break; | |
4829 | ||
4e8a9132 | 4830 | case wordend: |
669fa600 SM |
4831 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == Sword); |
4832 | case symend: | |
4e8a9132 | 4833 | return ((re_opcode_t) *p1 == syntaxspec |
669fa600 SM |
4834 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4835 | case notsyntaxspec: | |
4836 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4837 | |
4838 | case wordbeg: | |
669fa600 SM |
4839 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == Sword); |
4840 | case symbeg: | |
4e8a9132 | 4841 | return ((re_opcode_t) *p1 == notsyntaxspec |
669fa600 SM |
4842 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4843 | case syntaxspec: | |
4844 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4845 | |
4846 | case wordbound: | |
4847 | return (((re_opcode_t) *p1 == notsyntaxspec | |
4848 | || (re_opcode_t) *p1 == syntaxspec) | |
4849 | && p1[1] == Sword); | |
4850 | ||
1fb352e0 | 4851 | #ifdef emacs |
4e8a9132 SM |
4852 | case categoryspec: |
4853 | return ((re_opcode_t) *p1 == notcategoryspec && p1[1] == p2[1]); | |
4854 | case notcategoryspec: | |
4855 | return ((re_opcode_t) *p1 == categoryspec && p1[1] == p2[1]); | |
4856 | #endif /* emacs */ | |
4857 | ||
4858 | default: | |
4859 | ; | |
505bde11 SM |
4860 | } |
4861 | ||
4862 | /* Safe default. */ | |
4863 | return 0; | |
4864 | } | |
4865 | ||
fa9a63c5 RM |
4866 | \f |
4867 | /* Matching routines. */ | |
4868 | ||
25fe55af | 4869 | #ifndef emacs /* Emacs never uses this. */ |
fa9a63c5 RM |
4870 | /* re_match is like re_match_2 except it takes only a single string. */ |
4871 | ||
d1dfb56c | 4872 | regoff_t |
d2762c86 | 4873 | re_match (struct re_pattern_buffer *bufp, const char *string, |
d1dfb56c | 4874 | size_t size, ssize_t pos, struct re_registers *regs) |
fa9a63c5 | 4875 | { |
d1dfb56c EZ |
4876 | regoff_t result = re_match_2_internal (bufp, NULL, 0, (re_char*) string, |
4877 | size, pos, regs, size); | |
fa9a63c5 RM |
4878 | return result; |
4879 | } | |
c0f9ea08 | 4880 | WEAK_ALIAS (__re_match, re_match) |
fa9a63c5 RM |
4881 | #endif /* not emacs */ |
4882 | ||
b18215fc RS |
4883 | #ifdef emacs |
4884 | /* In Emacs, this is the string or buffer in which we | |
7814e705 | 4885 | are matching. It is used for looking up syntax properties. */ |
b18215fc RS |
4886 | Lisp_Object re_match_object; |
4887 | #endif | |
fa9a63c5 RM |
4888 | |
4889 | /* re_match_2 matches the compiled pattern in BUFP against the | |
4890 | the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 | |
4891 | and SIZE2, respectively). We start matching at POS, and stop | |
4892 | matching at STOP. | |
5e69f11e | 4893 | |
fa9a63c5 | 4894 | If REGS is non-null and the `no_sub' field of BUFP is nonzero, we |
7814e705 | 4895 | store offsets for the substring each group matched in REGS. See the |
fa9a63c5 RM |
4896 | documentation for exactly how many groups we fill. |
4897 | ||
4898 | We return -1 if no match, -2 if an internal error (such as the | |
7814e705 | 4899 | failure stack overflowing). Otherwise, we return the length of the |
fa9a63c5 RM |
4900 | matched substring. */ |
4901 | ||
d1dfb56c EZ |
4902 | regoff_t |
4903 | re_match_2 (struct re_pattern_buffer *bufp, const char *string1, | |
4904 | size_t size1, const char *string2, size_t size2, ssize_t pos, | |
4905 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4906 | { |
d1dfb56c | 4907 | regoff_t result; |
25fe55af | 4908 | |
b18215fc | 4909 | #ifdef emacs |
d1dfb56c | 4910 | ssize_t charpos; |
d48cd3f4 | 4911 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
99633e97 | 4912 | charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (pos)); |
cc9b4df2 | 4913 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); |
b18215fc RS |
4914 | #endif |
4915 | ||
4bb91c68 SM |
4916 | result = re_match_2_internal (bufp, (re_char*) string1, size1, |
4917 | (re_char*) string2, size2, | |
cc9b4df2 | 4918 | pos, regs, stop); |
fa9a63c5 RM |
4919 | return result; |
4920 | } | |
c0f9ea08 | 4921 | WEAK_ALIAS (__re_match_2, re_match_2) |
fa9a63c5 | 4922 | |
bf216479 | 4923 | |
fa9a63c5 | 4924 | /* This is a separate function so that we can force an alloca cleanup |
7814e705 | 4925 | afterwards. */ |
d1dfb56c | 4926 | static regoff_t |
29abe551 PE |
4927 | re_match_2_internal (struct re_pattern_buffer *bufp, const_re_char *string1, |
4928 | size_t size1, const_re_char *string2, size_t size2, | |
d1dfb56c | 4929 | ssize_t pos, struct re_registers *regs, ssize_t stop) |
fa9a63c5 RM |
4930 | { |
4931 | /* General temporaries. */ | |
dc4a2ee0 | 4932 | int mcnt; |
01618498 | 4933 | size_t reg; |
fa9a63c5 RM |
4934 | |
4935 | /* Just past the end of the corresponding string. */ | |
66f0296e | 4936 | re_char *end1, *end2; |
fa9a63c5 RM |
4937 | |
4938 | /* Pointers into string1 and string2, just past the last characters in | |
7814e705 | 4939 | each to consider matching. */ |
66f0296e | 4940 | re_char *end_match_1, *end_match_2; |
fa9a63c5 RM |
4941 | |
4942 | /* Where we are in the data, and the end of the current string. */ | |
66f0296e | 4943 | re_char *d, *dend; |
5e69f11e | 4944 | |
99633e97 SM |
4945 | /* Used sometimes to remember where we were before starting matching |
4946 | an operator so that we can go back in case of failure. This "atomic" | |
4947 | behavior of matching opcodes is indispensable to the correctness | |
4948 | of the on_failure_keep_string_jump optimization. */ | |
4949 | re_char *dfail; | |
4950 | ||
fa9a63c5 | 4951 | /* Where we are in the pattern, and the end of the pattern. */ |
01618498 SM |
4952 | re_char *p = bufp->buffer; |
4953 | re_char *pend = p + bufp->used; | |
fa9a63c5 | 4954 | |
25fe55af | 4955 | /* We use this to map every character in the string. */ |
6676cb1c | 4956 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 | 4957 | |
cf9c99bc | 4958 | /* Nonzero if BUFP is setup from a multibyte regex. */ |
2d1675e4 | 4959 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
b18215fc | 4960 | |
cf9c99bc KH |
4961 | /* Nonzero if STRING1/STRING2 are multibyte. */ |
4962 | const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
4963 | ||
fa9a63c5 RM |
4964 | /* Failure point stack. Each place that can handle a failure further |
4965 | down the line pushes a failure point on this stack. It consists of | |
505bde11 | 4966 | regstart, and regend for all registers corresponding to |
fa9a63c5 RM |
4967 | the subexpressions we're currently inside, plus the number of such |
4968 | registers, and, finally, two char *'s. The first char * is where | |
4969 | to resume scanning the pattern; the second one is where to resume | |
7814e705 JB |
4970 | scanning the strings. */ |
4971 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | |
fa9a63c5 RM |
4972 | fail_stack_type fail_stack; |
4973 | #endif | |
dc4a2ee0 | 4974 | #ifdef DEBUG_COMPILES_ARGUMENTS |
fa9a63c5 RM |
4975 | unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; |
4976 | #endif | |
4977 | ||
0b32bf0e | 4978 | #if defined REL_ALLOC && defined REGEX_MALLOC |
fa9a63c5 RM |
4979 | /* This holds the pointer to the failure stack, when |
4980 | it is allocated relocatably. */ | |
4981 | fail_stack_elt_t *failure_stack_ptr; | |
99633e97 | 4982 | #endif |
fa9a63c5 RM |
4983 | |
4984 | /* We fill all the registers internally, independent of what we | |
7814e705 | 4985 | return, for use in backreferences. The number here includes |
fa9a63c5 | 4986 | an element for register zero. */ |
4bb91c68 | 4987 | size_t num_regs = bufp->re_nsub + 1; |
5e69f11e | 4988 | |
fa9a63c5 RM |
4989 | /* Information on the contents of registers. These are pointers into |
4990 | the input strings; they record just what was matched (on this | |
4991 | attempt) by a subexpression part of the pattern, that is, the | |
4992 | regnum-th regstart pointer points to where in the pattern we began | |
4993 | matching and the regnum-th regend points to right after where we | |
4994 | stopped matching the regnum-th subexpression. (The zeroth register | |
4995 | keeps track of what the whole pattern matches.) */ | |
4996 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 4997 | re_char **regstart, **regend; |
fa9a63c5 RM |
4998 | #endif |
4999 | ||
fa9a63c5 | 5000 | /* The following record the register info as found in the above |
5e69f11e | 5001 | variables when we find a match better than any we've seen before. |
fa9a63c5 RM |
5002 | This happens as we backtrack through the failure points, which in |
5003 | turn happens only if we have not yet matched the entire string. */ | |
5004 | unsigned best_regs_set = false; | |
5005 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5006 | re_char **best_regstart, **best_regend; |
fa9a63c5 | 5007 | #endif |
5e69f11e | 5008 | |
fa9a63c5 RM |
5009 | /* Logically, this is `best_regend[0]'. But we don't want to have to |
5010 | allocate space for that if we're not allocating space for anything | |
7814e705 | 5011 | else (see below). Also, we never need info about register 0 for |
fa9a63c5 RM |
5012 | any of the other register vectors, and it seems rather a kludge to |
5013 | treat `best_regend' differently than the rest. So we keep track of | |
5014 | the end of the best match so far in a separate variable. We | |
5015 | initialize this to NULL so that when we backtrack the first time | |
5016 | and need to test it, it's not garbage. */ | |
66f0296e | 5017 | re_char *match_end = NULL; |
fa9a63c5 | 5018 | |
dc4a2ee0 | 5019 | #ifdef DEBUG_COMPILES_ARGUMENTS |
fa9a63c5 | 5020 | /* Counts the total number of registers pushed. */ |
5e69f11e | 5021 | unsigned num_regs_pushed = 0; |
fa9a63c5 RM |
5022 | #endif |
5023 | ||
dc4a2ee0 | 5024 | DEBUG_PRINT ("\n\nEntering re_match_2.\n"); |
5e69f11e | 5025 | |
fa9a63c5 | 5026 | INIT_FAIL_STACK (); |
5e69f11e | 5027 | |
fa9a63c5 RM |
5028 | #ifdef MATCH_MAY_ALLOCATE |
5029 | /* Do not bother to initialize all the register variables if there are | |
5030 | no groups in the pattern, as it takes a fair amount of time. If | |
5031 | there are groups, we include space for register 0 (the whole | |
5032 | pattern), even though we never use it, since it simplifies the | |
5033 | array indexing. We should fix this. */ | |
5034 | if (bufp->re_nsub) | |
5035 | { | |
66f0296e SM |
5036 | regstart = REGEX_TALLOC (num_regs, re_char *); |
5037 | regend = REGEX_TALLOC (num_regs, re_char *); | |
5038 | best_regstart = REGEX_TALLOC (num_regs, re_char *); | |
5039 | best_regend = REGEX_TALLOC (num_regs, re_char *); | |
fa9a63c5 | 5040 | |
505bde11 | 5041 | if (!(regstart && regend && best_regstart && best_regend)) |
25fe55af RS |
5042 | { |
5043 | FREE_VARIABLES (); | |
5044 | return -2; | |
5045 | } | |
fa9a63c5 RM |
5046 | } |
5047 | else | |
5048 | { | |
5049 | /* We must initialize all our variables to NULL, so that | |
25fe55af | 5050 | `FREE_VARIABLES' doesn't try to free them. */ |
505bde11 | 5051 | regstart = regend = best_regstart = best_regend = NULL; |
fa9a63c5 RM |
5052 | } |
5053 | #endif /* MATCH_MAY_ALLOCATE */ | |
5054 | ||
5055 | /* The starting position is bogus. */ | |
5056 | if (pos < 0 || pos > size1 + size2) | |
5057 | { | |
5058 | FREE_VARIABLES (); | |
5059 | return -1; | |
5060 | } | |
5e69f11e | 5061 | |
fa9a63c5 RM |
5062 | /* Initialize subexpression text positions to -1 to mark ones that no |
5063 | start_memory/stop_memory has been seen for. Also initialize the | |
5064 | register information struct. */ | |
01618498 SM |
5065 | for (reg = 1; reg < num_regs; reg++) |
5066 | regstart[reg] = regend[reg] = NULL; | |
99633e97 | 5067 | |
fa9a63c5 | 5068 | /* We move `string1' into `string2' if the latter's empty -- but not if |
7814e705 | 5069 | `string1' is null. */ |
fa9a63c5 RM |
5070 | if (size2 == 0 && string1 != NULL) |
5071 | { | |
5072 | string2 = string1; | |
5073 | size2 = size1; | |
5074 | string1 = 0; | |
5075 | size1 = 0; | |
5076 | } | |
5077 | end1 = string1 + size1; | |
5078 | end2 = string2 + size2; | |
5079 | ||
5e69f11e | 5080 | /* `p' scans through the pattern as `d' scans through the data. |
fa9a63c5 RM |
5081 | `dend' is the end of the input string that `d' points within. `d' |
5082 | is advanced into the following input string whenever necessary, but | |
5083 | this happens before fetching; therefore, at the beginning of the | |
5084 | loop, `d' can be pointing at the end of a string, but it cannot | |
5085 | equal `string2'. */ | |
419d1c74 | 5086 | if (pos >= size1) |
fa9a63c5 | 5087 | { |
419d1c74 SM |
5088 | /* Only match within string2. */ |
5089 | d = string2 + pos - size1; | |
5090 | dend = end_match_2 = string2 + stop - size1; | |
5091 | end_match_1 = end1; /* Just to give it a value. */ | |
fa9a63c5 RM |
5092 | } |
5093 | else | |
5094 | { | |
f1ad044f | 5095 | if (stop < size1) |
419d1c74 SM |
5096 | { |
5097 | /* Only match within string1. */ | |
5098 | end_match_1 = string1 + stop; | |
5099 | /* BEWARE! | |
5100 | When we reach end_match_1, PREFETCH normally switches to string2. | |
5101 | But in the present case, this means that just doing a PREFETCH | |
5102 | makes us jump from `stop' to `gap' within the string. | |
5103 | What we really want here is for the search to stop as | |
5104 | soon as we hit end_match_1. That's why we set end_match_2 | |
5105 | to end_match_1 (since PREFETCH fails as soon as we hit | |
5106 | end_match_2). */ | |
5107 | end_match_2 = end_match_1; | |
5108 | } | |
5109 | else | |
f1ad044f SM |
5110 | { /* It's important to use this code when stop == size so that |
5111 | moving `d' from end1 to string2 will not prevent the d == dend | |
5112 | check from catching the end of string. */ | |
419d1c74 SM |
5113 | end_match_1 = end1; |
5114 | end_match_2 = string2 + stop - size1; | |
5115 | } | |
5116 | d = string1 + pos; | |
5117 | dend = end_match_1; | |
fa9a63c5 RM |
5118 | } |
5119 | ||
dc4a2ee0 | 5120 | DEBUG_PRINT ("The compiled pattern is: "); |
fa9a63c5 | 5121 | DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); |
dc4a2ee0 | 5122 | DEBUG_PRINT ("The string to match is: `"); |
fa9a63c5 | 5123 | DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); |
dc4a2ee0 | 5124 | DEBUG_PRINT ("'\n"); |
5e69f11e | 5125 | |
7814e705 | 5126 | /* This loops over pattern commands. It exits by returning from the |
fa9a63c5 RM |
5127 | function if the match is complete, or it drops through if the match |
5128 | fails at this starting point in the input data. */ | |
5129 | for (;;) | |
5130 | { | |
dc4a2ee0 | 5131 | DEBUG_PRINT ("\n%p: ", p); |
fa9a63c5 RM |
5132 | |
5133 | if (p == pend) | |
dc4a2ee0 PE |
5134 | { |
5135 | ptrdiff_t dcnt; | |
5136 | ||
5137 | /* End of pattern means we might have succeeded. */ | |
5138 | DEBUG_PRINT ("end of pattern ... "); | |
5e69f11e | 5139 | |
fa9a63c5 | 5140 | /* If we haven't matched the entire string, and we want the |
25fe55af RS |
5141 | longest match, try backtracking. */ |
5142 | if (d != end_match_2) | |
fa9a63c5 RM |
5143 | { |
5144 | /* 1 if this match ends in the same string (string1 or string2) | |
5145 | as the best previous match. */ | |
d42f4f0f PE |
5146 | boolean same_str_p = (FIRST_STRING_P (match_end) |
5147 | == FIRST_STRING_P (d)); | |
fa9a63c5 RM |
5148 | /* 1 if this match is the best seen so far. */ |
5149 | boolean best_match_p; | |
5150 | ||
5151 | /* AIX compiler got confused when this was combined | |
7814e705 | 5152 | with the previous declaration. */ |
fa9a63c5 RM |
5153 | if (same_str_p) |
5154 | best_match_p = d > match_end; | |
5155 | else | |
99633e97 | 5156 | best_match_p = !FIRST_STRING_P (d); |
fa9a63c5 | 5157 | |
dc4a2ee0 | 5158 | DEBUG_PRINT ("backtracking.\n"); |
25fe55af RS |
5159 | |
5160 | if (!FAIL_STACK_EMPTY ()) | |
5161 | { /* More failure points to try. */ | |
5162 | ||
5163 | /* If exceeds best match so far, save it. */ | |
5164 | if (!best_regs_set || best_match_p) | |
5165 | { | |
5166 | best_regs_set = true; | |
5167 | match_end = d; | |
5168 | ||
dc4a2ee0 | 5169 | DEBUG_PRINT ("\nSAVING match as best so far.\n"); |
25fe55af | 5170 | |
01618498 | 5171 | for (reg = 1; reg < num_regs; reg++) |
25fe55af | 5172 | { |
01618498 SM |
5173 | best_regstart[reg] = regstart[reg]; |
5174 | best_regend[reg] = regend[reg]; | |
25fe55af RS |
5175 | } |
5176 | } | |
5177 | goto fail; | |
5178 | } | |
5179 | ||
5180 | /* If no failure points, don't restore garbage. And if | |
5181 | last match is real best match, don't restore second | |
5182 | best one. */ | |
5183 | else if (best_regs_set && !best_match_p) | |
5184 | { | |
5185 | restore_best_regs: | |
5186 | /* Restore best match. It may happen that `dend == | |
5187 | end_match_1' while the restored d is in string2. | |
5188 | For example, the pattern `x.*y.*z' against the | |
5189 | strings `x-' and `y-z-', if the two strings are | |
7814e705 | 5190 | not consecutive in memory. */ |
dc4a2ee0 | 5191 | DEBUG_PRINT ("Restoring best registers.\n"); |
25fe55af RS |
5192 | |
5193 | d = match_end; | |
5194 | dend = ((d >= string1 && d <= end1) | |
5195 | ? end_match_1 : end_match_2); | |
fa9a63c5 | 5196 | |
01618498 | 5197 | for (reg = 1; reg < num_regs; reg++) |
fa9a63c5 | 5198 | { |
01618498 SM |
5199 | regstart[reg] = best_regstart[reg]; |
5200 | regend[reg] = best_regend[reg]; | |
fa9a63c5 | 5201 | } |
25fe55af RS |
5202 | } |
5203 | } /* d != end_match_2 */ | |
fa9a63c5 RM |
5204 | |
5205 | succeed_label: | |
dc4a2ee0 | 5206 | DEBUG_PRINT ("Accepting match.\n"); |
fa9a63c5 | 5207 | |
25fe55af RS |
5208 | /* If caller wants register contents data back, do it. */ |
5209 | if (regs && !bufp->no_sub) | |
fa9a63c5 | 5210 | { |
25fe55af RS |
5211 | /* Have the register data arrays been allocated? */ |
5212 | if (bufp->regs_allocated == REGS_UNALLOCATED) | |
7814e705 | 5213 | { /* No. So allocate them with malloc. We need one |
25fe55af RS |
5214 | extra element beyond `num_regs' for the `-1' marker |
5215 | GNU code uses. */ | |
5216 | regs->num_regs = MAX (RE_NREGS, num_regs + 1); | |
5217 | regs->start = TALLOC (regs->num_regs, regoff_t); | |
5218 | regs->end = TALLOC (regs->num_regs, regoff_t); | |
5219 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5220 | { |
5221 | FREE_VARIABLES (); | |
5222 | return -2; | |
5223 | } | |
25fe55af RS |
5224 | bufp->regs_allocated = REGS_REALLOCATE; |
5225 | } | |
5226 | else if (bufp->regs_allocated == REGS_REALLOCATE) | |
5227 | { /* Yes. If we need more elements than were already | |
5228 | allocated, reallocate them. If we need fewer, just | |
5229 | leave it alone. */ | |
5230 | if (regs->num_regs < num_regs + 1) | |
5231 | { | |
5232 | regs->num_regs = num_regs + 1; | |
5233 | RETALLOC (regs->start, regs->num_regs, regoff_t); | |
5234 | RETALLOC (regs->end, regs->num_regs, regoff_t); | |
5235 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5236 | { |
5237 | FREE_VARIABLES (); | |
5238 | return -2; | |
5239 | } | |
25fe55af RS |
5240 | } |
5241 | } | |
5242 | else | |
fa9a63c5 RM |
5243 | { |
5244 | /* These braces fend off a "empty body in an else-statement" | |
7814e705 | 5245 | warning under GCC when assert expands to nothing. */ |
fa9a63c5 RM |
5246 | assert (bufp->regs_allocated == REGS_FIXED); |
5247 | } | |
5248 | ||
25fe55af RS |
5249 | /* Convert the pointer data in `regstart' and `regend' to |
5250 | indices. Register zero has to be set differently, | |
5251 | since we haven't kept track of any info for it. */ | |
5252 | if (regs->num_regs > 0) | |
5253 | { | |
5254 | regs->start[0] = pos; | |
99633e97 | 5255 | regs->end[0] = POINTER_TO_OFFSET (d); |
25fe55af | 5256 | } |
5e69f11e | 5257 | |
25fe55af RS |
5258 | /* Go through the first `min (num_regs, regs->num_regs)' |
5259 | registers, since that is all we initialized. */ | |
01618498 | 5260 | for (reg = 1; reg < MIN (num_regs, regs->num_regs); reg++) |
fa9a63c5 | 5261 | { |
01618498 SM |
5262 | if (REG_UNSET (regstart[reg]) || REG_UNSET (regend[reg])) |
5263 | regs->start[reg] = regs->end[reg] = -1; | |
25fe55af RS |
5264 | else |
5265 | { | |
dc4a2ee0 PE |
5266 | regs->start[reg] = POINTER_TO_OFFSET (regstart[reg]); |
5267 | regs->end[reg] = POINTER_TO_OFFSET (regend[reg]); | |
25fe55af | 5268 | } |
fa9a63c5 | 5269 | } |
5e69f11e | 5270 | |
25fe55af RS |
5271 | /* If the regs structure we return has more elements than |
5272 | were in the pattern, set the extra elements to -1. If | |
5273 | we (re)allocated the registers, this is the case, | |
5274 | because we always allocate enough to have at least one | |
7814e705 | 5275 | -1 at the end. */ |
01618498 SM |
5276 | for (reg = num_regs; reg < regs->num_regs; reg++) |
5277 | regs->start[reg] = regs->end[reg] = -1; | |
fa9a63c5 RM |
5278 | } /* regs && !bufp->no_sub */ |
5279 | ||
dc4a2ee0 PE |
5280 | DEBUG_PRINT ("%u failure points pushed, %u popped (%u remain).\n", |
5281 | nfailure_points_pushed, nfailure_points_popped, | |
5282 | nfailure_points_pushed - nfailure_points_popped); | |
5283 | DEBUG_PRINT ("%u registers pushed.\n", num_regs_pushed); | |
fa9a63c5 | 5284 | |
dc4a2ee0 | 5285 | dcnt = POINTER_TO_OFFSET (d) - pos; |
fa9a63c5 | 5286 | |
dc4a2ee0 | 5287 | DEBUG_PRINT ("Returning %td from re_match_2.\n", dcnt); |
fa9a63c5 | 5288 | |
25fe55af | 5289 | FREE_VARIABLES (); |
dc4a2ee0 | 5290 | return dcnt; |
25fe55af | 5291 | } |
fa9a63c5 | 5292 | |
7814e705 | 5293 | /* Otherwise match next pattern command. */ |
7393bcbb | 5294 | switch (*p++) |
fa9a63c5 | 5295 | { |
25fe55af RS |
5296 | /* Ignore these. Used to ignore the n of succeed_n's which |
5297 | currently have n == 0. */ | |
5298 | case no_op: | |
dc4a2ee0 | 5299 | DEBUG_PRINT ("EXECUTING no_op.\n"); |
25fe55af | 5300 | break; |
fa9a63c5 RM |
5301 | |
5302 | case succeed: | |
dc4a2ee0 | 5303 | DEBUG_PRINT ("EXECUTING succeed.\n"); |
fa9a63c5 RM |
5304 | goto succeed_label; |
5305 | ||
7814e705 | 5306 | /* Match the next n pattern characters exactly. The following |
25fe55af | 5307 | byte in the pattern defines n, and the n bytes after that |
7814e705 | 5308 | are the characters to match. */ |
fa9a63c5 RM |
5309 | case exactn: |
5310 | mcnt = *p++; | |
dc4a2ee0 | 5311 | DEBUG_PRINT ("EXECUTING exactn %d.\n", mcnt); |
fa9a63c5 | 5312 | |
99633e97 SM |
5313 | /* Remember the start point to rollback upon failure. */ |
5314 | dfail = d; | |
5315 | ||
6fdd04b0 | 5316 | #ifndef emacs |
25fe55af RS |
5317 | /* This is written out as an if-else so we don't waste time |
5318 | testing `translate' inside the loop. */ | |
28703c16 | 5319 | if (RE_TRANSLATE_P (translate)) |
6fdd04b0 KH |
5320 | do |
5321 | { | |
5322 | PREFETCH (); | |
5323 | if (RE_TRANSLATE (translate, *d) != *p++) | |
e934739e | 5324 | { |
6fdd04b0 KH |
5325 | d = dfail; |
5326 | goto fail; | |
e934739e | 5327 | } |
6fdd04b0 KH |
5328 | d++; |
5329 | } | |
5330 | while (--mcnt); | |
fa9a63c5 | 5331 | else |
6fdd04b0 KH |
5332 | do |
5333 | { | |
5334 | PREFETCH (); | |
5335 | if (*d++ != *p++) | |
bf216479 | 5336 | { |
6fdd04b0 KH |
5337 | d = dfail; |
5338 | goto fail; | |
bf216479 | 5339 | } |
6fdd04b0 KH |
5340 | } |
5341 | while (--mcnt); | |
5342 | #else /* emacs */ | |
5343 | /* The cost of testing `translate' is comparatively small. */ | |
cf9c99bc | 5344 | if (target_multibyte) |
6fdd04b0 KH |
5345 | do |
5346 | { | |
5347 | int pat_charlen, buf_charlen; | |
cf9c99bc | 5348 | int pat_ch, buf_ch; |
e934739e | 5349 | |
6fdd04b0 | 5350 | PREFETCH (); |
cf9c99bc | 5351 | if (multibyte) |
62a6e103 | 5352 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
cf9c99bc KH |
5353 | else |
5354 | { | |
5355 | pat_ch = RE_CHAR_TO_MULTIBYTE (*p); | |
5356 | pat_charlen = 1; | |
5357 | } | |
62a6e103 | 5358 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 5359 | |
6fdd04b0 | 5360 | if (TRANSLATE (buf_ch) != pat_ch) |
e934739e | 5361 | { |
6fdd04b0 KH |
5362 | d = dfail; |
5363 | goto fail; | |
e934739e | 5364 | } |
bf216479 | 5365 | |
6fdd04b0 KH |
5366 | p += pat_charlen; |
5367 | d += buf_charlen; | |
5368 | mcnt -= pat_charlen; | |
5369 | } | |
5370 | while (mcnt > 0); | |
fa9a63c5 | 5371 | else |
6fdd04b0 KH |
5372 | do |
5373 | { | |
abbd1bcf | 5374 | int pat_charlen; |
cf9c99bc | 5375 | int pat_ch, buf_ch; |
bf216479 | 5376 | |
6fdd04b0 | 5377 | PREFETCH (); |
cf9c99bc KH |
5378 | if (multibyte) |
5379 | { | |
62a6e103 | 5380 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
2afc21f5 | 5381 | pat_ch = RE_CHAR_TO_UNIBYTE (pat_ch); |
cf9c99bc KH |
5382 | } |
5383 | else | |
5384 | { | |
5385 | pat_ch = *p; | |
5386 | pat_charlen = 1; | |
5387 | } | |
5388 | buf_ch = RE_CHAR_TO_MULTIBYTE (*d); | |
5389 | if (! CHAR_BYTE8_P (buf_ch)) | |
5390 | { | |
5391 | buf_ch = TRANSLATE (buf_ch); | |
5392 | buf_ch = RE_CHAR_TO_UNIBYTE (buf_ch); | |
5393 | if (buf_ch < 0) | |
5394 | buf_ch = *d; | |
5395 | } | |
0e2501ed AS |
5396 | else |
5397 | buf_ch = *d; | |
cf9c99bc | 5398 | if (buf_ch != pat_ch) |
6fdd04b0 KH |
5399 | { |
5400 | d = dfail; | |
5401 | goto fail; | |
bf216479 | 5402 | } |
cf9c99bc KH |
5403 | p += pat_charlen; |
5404 | d++; | |
6fdd04b0 KH |
5405 | } |
5406 | while (--mcnt); | |
5407 | #endif | |
25fe55af | 5408 | break; |
fa9a63c5 RM |
5409 | |
5410 | ||
25fe55af | 5411 | /* Match any character except possibly a newline or a null. */ |
fa9a63c5 | 5412 | case anychar: |
e934739e RS |
5413 | { |
5414 | int buf_charlen; | |
01618498 | 5415 | re_wchar_t buf_ch; |
fa9a63c5 | 5416 | |
dc4a2ee0 | 5417 | DEBUG_PRINT ("EXECUTING anychar.\n"); |
fa9a63c5 | 5418 | |
e934739e | 5419 | PREFETCH (); |
62a6e103 | 5420 | buf_ch = RE_STRING_CHAR_AND_LENGTH (d, buf_charlen, |
cf9c99bc | 5421 | target_multibyte); |
e934739e RS |
5422 | buf_ch = TRANSLATE (buf_ch); |
5423 | ||
5424 | if ((!(bufp->syntax & RE_DOT_NEWLINE) | |
5425 | && buf_ch == '\n') | |
5426 | || ((bufp->syntax & RE_DOT_NOT_NULL) | |
5427 | && buf_ch == '\000')) | |
5428 | goto fail; | |
5429 | ||
dc4a2ee0 | 5430 | DEBUG_PRINT (" Matched `%d'.\n", *d); |
e934739e RS |
5431 | d += buf_charlen; |
5432 | } | |
fa9a63c5 RM |
5433 | break; |
5434 | ||
5435 | ||
5436 | case charset: | |
5437 | case charset_not: | |
5438 | { | |
b18215fc | 5439 | register unsigned int c; |
fa9a63c5 | 5440 | boolean not = (re_opcode_t) *(p - 1) == charset_not; |
b18215fc RS |
5441 | int len; |
5442 | ||
5443 | /* Start of actual range_table, or end of bitmap if there is no | |
5444 | range table. */ | |
da053e48 | 5445 | re_char *range_table IF_LINT (= NULL); |
b18215fc | 5446 | |
96cc36cc | 5447 | /* Nonzero if there is a range table. */ |
b18215fc RS |
5448 | int range_table_exists; |
5449 | ||
96cc36cc RS |
5450 | /* Number of ranges of range table. This is not included |
5451 | in the initial byte-length of the command. */ | |
5452 | int count = 0; | |
fa9a63c5 | 5453 | |
f5020181 AS |
5454 | /* Whether matching against a unibyte character. */ |
5455 | boolean unibyte_char = false; | |
5456 | ||
dc4a2ee0 | 5457 | DEBUG_PRINT ("EXECUTING charset%s.\n", not ? "_not" : ""); |
fa9a63c5 | 5458 | |
b18215fc | 5459 | range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]); |
96cc36cc | 5460 | |
b18215fc | 5461 | if (range_table_exists) |
96cc36cc RS |
5462 | { |
5463 | range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap. */ | |
5464 | EXTRACT_NUMBER_AND_INCR (count, range_table); | |
5465 | } | |
b18215fc | 5466 | |
2d1675e4 | 5467 | PREFETCH (); |
62a6e103 | 5468 | c = RE_STRING_CHAR_AND_LENGTH (d, len, target_multibyte); |
cf9c99bc KH |
5469 | if (target_multibyte) |
5470 | { | |
5471 | int c1; | |
b18215fc | 5472 | |
cf9c99bc KH |
5473 | c = TRANSLATE (c); |
5474 | c1 = RE_CHAR_TO_UNIBYTE (c); | |
5475 | if (c1 >= 0) | |
f5020181 AS |
5476 | { |
5477 | unibyte_char = true; | |
5478 | c = c1; | |
5479 | } | |
cf9c99bc KH |
5480 | } |
5481 | else | |
5482 | { | |
5483 | int c1 = RE_CHAR_TO_MULTIBYTE (c); | |
5484 | ||
5485 | if (! CHAR_BYTE8_P (c1)) | |
5486 | { | |
5487 | c1 = TRANSLATE (c1); | |
5488 | c1 = RE_CHAR_TO_UNIBYTE (c1); | |
5489 | if (c1 >= 0) | |
f5020181 AS |
5490 | { |
5491 | unibyte_char = true; | |
5492 | c = c1; | |
5493 | } | |
cf9c99bc | 5494 | } |
0b8be006 AS |
5495 | else |
5496 | unibyte_char = true; | |
cf9c99bc KH |
5497 | } |
5498 | ||
f5020181 | 5499 | if (unibyte_char && c < (1 << BYTEWIDTH)) |
b18215fc | 5500 | { /* Lookup bitmap. */ |
b18215fc RS |
5501 | /* Cast to `unsigned' instead of `unsigned char' in |
5502 | case the bit list is a full 32 bytes long. */ | |
5503 | if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH) | |
96cc36cc RS |
5504 | && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) |
5505 | not = !not; | |
b18215fc | 5506 | } |
96cc36cc | 5507 | #ifdef emacs |
b18215fc | 5508 | else if (range_table_exists) |
96cc36cc RS |
5509 | { |
5510 | int class_bits = CHARSET_RANGE_TABLE_BITS (&p[-1]); | |
5511 | ||
14473664 SM |
5512 | if ( (class_bits & BIT_LOWER && ISLOWER (c)) |
5513 | | (class_bits & BIT_MULTIBYTE) | |
96cc36cc RS |
5514 | | (class_bits & BIT_PUNCT && ISPUNCT (c)) |
5515 | | (class_bits & BIT_SPACE && ISSPACE (c)) | |
5516 | | (class_bits & BIT_UPPER && ISUPPER (c)) | |
5517 | | (class_bits & BIT_WORD && ISWORD (c))) | |
5518 | not = !not; | |
5519 | else | |
5520 | CHARSET_LOOKUP_RANGE_TABLE_RAW (not, c, range_table, count); | |
5521 | } | |
5522 | #endif /* emacs */ | |
fa9a63c5 | 5523 | |
96cc36cc RS |
5524 | if (range_table_exists) |
5525 | p = CHARSET_RANGE_TABLE_END (range_table, count); | |
5526 | else | |
5527 | p += CHARSET_BITMAP_SIZE (&p[-1]) + 1; | |
fa9a63c5 RM |
5528 | |
5529 | if (!not) goto fail; | |
5e69f11e | 5530 | |
b18215fc | 5531 | d += len; |
fa9a63c5 | 5532 | } |
8fb31792 | 5533 | break; |
fa9a63c5 RM |
5534 | |
5535 | ||
25fe55af | 5536 | /* The beginning of a group is represented by start_memory. |
505bde11 | 5537 | The argument is the register number. The text |
25fe55af | 5538 | matched within the group is recorded (in the internal |
7814e705 | 5539 | registers data structure) under the register number. */ |
25fe55af | 5540 | case start_memory: |
dc4a2ee0 | 5541 | DEBUG_PRINT ("EXECUTING start_memory %d:\n", *p); |
505bde11 SM |
5542 | |
5543 | /* In case we need to undo this operation (via backtracking). */ | |
dc4a2ee0 | 5544 | PUSH_FAILURE_REG (*p); |
fa9a63c5 | 5545 | |
25fe55af | 5546 | regstart[*p] = d; |
4bb91c68 | 5547 | regend[*p] = NULL; /* probably unnecessary. -sm */ |
dc4a2ee0 | 5548 | DEBUG_PRINT (" regstart: %td\n", POINTER_TO_OFFSET (regstart[*p])); |
fa9a63c5 | 5549 | |
25fe55af | 5550 | /* Move past the register number and inner group count. */ |
505bde11 | 5551 | p += 1; |
25fe55af | 5552 | break; |
fa9a63c5 RM |
5553 | |
5554 | ||
25fe55af | 5555 | /* The stop_memory opcode represents the end of a group. Its |
505bde11 | 5556 | argument is the same as start_memory's: the register number. */ |
fa9a63c5 | 5557 | case stop_memory: |
dc4a2ee0 | 5558 | DEBUG_PRINT ("EXECUTING stop_memory %d:\n", *p); |
505bde11 SM |
5559 | |
5560 | assert (!REG_UNSET (regstart[*p])); | |
5561 | /* Strictly speaking, there should be code such as: | |
177c0ea7 | 5562 | |
0b32bf0e | 5563 | assert (REG_UNSET (regend[*p])); |
505bde11 SM |
5564 | PUSH_FAILURE_REGSTOP ((unsigned int)*p); |
5565 | ||
5566 | But the only info to be pushed is regend[*p] and it is known to | |
5567 | be UNSET, so there really isn't anything to push. | |
5568 | Not pushing anything, on the other hand deprives us from the | |
5569 | guarantee that regend[*p] is UNSET since undoing this operation | |
5570 | will not reset its value properly. This is not important since | |
5571 | the value will only be read on the next start_memory or at | |
5572 | the very end and both events can only happen if this stop_memory | |
5573 | is *not* undone. */ | |
fa9a63c5 | 5574 | |
25fe55af | 5575 | regend[*p] = d; |
dc4a2ee0 | 5576 | DEBUG_PRINT (" regend: %td\n", POINTER_TO_OFFSET (regend[*p])); |
fa9a63c5 | 5577 | |
25fe55af | 5578 | /* Move past the register number and the inner group count. */ |
505bde11 | 5579 | p += 1; |
25fe55af | 5580 | break; |
fa9a63c5 RM |
5581 | |
5582 | ||
5583 | /* \<digit> has been turned into a `duplicate' command which is | |
25fe55af RS |
5584 | followed by the numeric value of <digit> as the register number. */ |
5585 | case duplicate: | |
fa9a63c5 | 5586 | { |
66f0296e | 5587 | register re_char *d2, *dend2; |
7814e705 | 5588 | int regno = *p++; /* Get which register to match against. */ |
dc4a2ee0 | 5589 | DEBUG_PRINT ("EXECUTING duplicate %d.\n", regno); |
fa9a63c5 | 5590 | |
7814e705 | 5591 | /* Can't back reference a group which we've never matched. */ |
25fe55af RS |
5592 | if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) |
5593 | goto fail; | |
5e69f11e | 5594 | |
7814e705 | 5595 | /* Where in input to try to start matching. */ |
25fe55af | 5596 | d2 = regstart[regno]; |
5e69f11e | 5597 | |
99633e97 SM |
5598 | /* Remember the start point to rollback upon failure. */ |
5599 | dfail = d; | |
5600 | ||
25fe55af RS |
5601 | /* Where to stop matching; if both the place to start and |
5602 | the place to stop matching are in the same string, then | |
5603 | set to the place to stop, otherwise, for now have to use | |
5604 | the end of the first string. */ | |
fa9a63c5 | 5605 | |
25fe55af | 5606 | dend2 = ((FIRST_STRING_P (regstart[regno]) |
fa9a63c5 RM |
5607 | == FIRST_STRING_P (regend[regno])) |
5608 | ? regend[regno] : end_match_1); | |
5609 | for (;;) | |
5610 | { | |
dc4a2ee0 PE |
5611 | ptrdiff_t dcnt; |
5612 | ||
fa9a63c5 | 5613 | /* If necessary, advance to next segment in register |
25fe55af | 5614 | contents. */ |
fa9a63c5 RM |
5615 | while (d2 == dend2) |
5616 | { | |
5617 | if (dend2 == end_match_2) break; | |
5618 | if (dend2 == regend[regno]) break; | |
5619 | ||
25fe55af RS |
5620 | /* End of string1 => advance to string2. */ |
5621 | d2 = string2; | |
5622 | dend2 = regend[regno]; | |
fa9a63c5 RM |
5623 | } |
5624 | /* At end of register contents => success */ | |
5625 | if (d2 == dend2) break; | |
5626 | ||
5627 | /* If necessary, advance to next segment in data. */ | |
5628 | PREFETCH (); | |
5629 | ||
5630 | /* How many characters left in this segment to match. */ | |
dc4a2ee0 | 5631 | dcnt = dend - d; |
5e69f11e | 5632 | |
fa9a63c5 | 5633 | /* Want how many consecutive characters we can match in |
25fe55af | 5634 | one shot, so, if necessary, adjust the count. */ |
dc4a2ee0 PE |
5635 | if (dcnt > dend2 - d2) |
5636 | dcnt = dend2 - d2; | |
5e69f11e | 5637 | |
fa9a63c5 | 5638 | /* Compare that many; failure if mismatch, else move |
25fe55af | 5639 | past them. */ |
28703c16 | 5640 | if (RE_TRANSLATE_P (translate) |
dc4a2ee0 PE |
5641 | ? bcmp_translate (d, d2, dcnt, translate, target_multibyte) |
5642 | : memcmp (d, d2, dcnt)) | |
99633e97 SM |
5643 | { |
5644 | d = dfail; | |
5645 | goto fail; | |
5646 | } | |
dc4a2ee0 | 5647 | d += dcnt, d2 += dcnt; |
fa9a63c5 RM |
5648 | } |
5649 | } | |
5650 | break; | |
5651 | ||
5652 | ||
25fe55af | 5653 | /* begline matches the empty string at the beginning of the string |
c0f9ea08 | 5654 | (unless `not_bol' is set in `bufp'), and after newlines. */ |
fa9a63c5 | 5655 | case begline: |
dc4a2ee0 | 5656 | DEBUG_PRINT ("EXECUTING begline.\n"); |
5e69f11e | 5657 | |
25fe55af RS |
5658 | if (AT_STRINGS_BEG (d)) |
5659 | { | |
5660 | if (!bufp->not_bol) break; | |
5661 | } | |
419d1c74 | 5662 | else |
25fe55af | 5663 | { |
bf216479 | 5664 | unsigned c; |
419d1c74 | 5665 | GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2); |
c0f9ea08 | 5666 | if (c == '\n') |
419d1c74 | 5667 | break; |
25fe55af RS |
5668 | } |
5669 | /* In all other cases, we fail. */ | |
5670 | goto fail; | |
fa9a63c5 RM |
5671 | |
5672 | ||
25fe55af | 5673 | /* endline is the dual of begline. */ |
fa9a63c5 | 5674 | case endline: |
dc4a2ee0 | 5675 | DEBUG_PRINT ("EXECUTING endline.\n"); |
fa9a63c5 | 5676 | |
25fe55af RS |
5677 | if (AT_STRINGS_END (d)) |
5678 | { | |
5679 | if (!bufp->not_eol) break; | |
5680 | } | |
f1ad044f | 5681 | else |
25fe55af | 5682 | { |
f1ad044f | 5683 | PREFETCH_NOLIMIT (); |
c0f9ea08 | 5684 | if (*d == '\n') |
f1ad044f | 5685 | break; |
25fe55af RS |
5686 | } |
5687 | goto fail; | |
fa9a63c5 RM |
5688 | |
5689 | ||
5690 | /* Match at the very beginning of the data. */ | |
25fe55af | 5691 | case begbuf: |
dc4a2ee0 | 5692 | DEBUG_PRINT ("EXECUTING begbuf.\n"); |
25fe55af RS |
5693 | if (AT_STRINGS_BEG (d)) |
5694 | break; | |
5695 | goto fail; | |
fa9a63c5 RM |
5696 | |
5697 | ||
5698 | /* Match at the very end of the data. */ | |
25fe55af | 5699 | case endbuf: |
dc4a2ee0 | 5700 | DEBUG_PRINT ("EXECUTING endbuf.\n"); |
fa9a63c5 RM |
5701 | if (AT_STRINGS_END (d)) |
5702 | break; | |
25fe55af | 5703 | goto fail; |
5e69f11e | 5704 | |
5e69f11e | 5705 | |
25fe55af RS |
5706 | /* on_failure_keep_string_jump is used to optimize `.*\n'. It |
5707 | pushes NULL as the value for the string on the stack. Then | |
505bde11 | 5708 | `POP_FAILURE_POINT' will keep the current value for the |
25fe55af | 5709 | string, instead of restoring it. To see why, consider |
7814e705 | 5710 | matching `foo\nbar' against `.*\n'. The .* matches the foo; |
25fe55af RS |
5711 | then the . fails against the \n. But the next thing we want |
5712 | to do is match the \n against the \n; if we restored the | |
5713 | string value, we would be back at the foo. | |
5714 | ||
5715 | Because this is used only in specific cases, we don't need to | |
5716 | check all the things that `on_failure_jump' does, to make | |
5717 | sure the right things get saved on the stack. Hence we don't | |
5718 | share its code. The only reason to push anything on the | |
5719 | stack at all is that otherwise we would have to change | |
5720 | `anychar's code to do something besides goto fail in this | |
5721 | case; that seems worse than this. */ | |
5722 | case on_failure_keep_string_jump: | |
505bde11 | 5723 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5724 | DEBUG_PRINT ("EXECUTING on_failure_keep_string_jump %d (to %p):\n", |
5725 | mcnt, p + mcnt); | |
fa9a63c5 | 5726 | |
505bde11 SM |
5727 | PUSH_FAILURE_POINT (p - 3, NULL); |
5728 | break; | |
5729 | ||
0683b6fa SM |
5730 | /* A nasty loop is introduced by the non-greedy *? and +?. |
5731 | With such loops, the stack only ever contains one failure point | |
5732 | at a time, so that a plain on_failure_jump_loop kind of | |
5733 | cycle detection cannot work. Worse yet, such a detection | |
5734 | can not only fail to detect a cycle, but it can also wrongly | |
5735 | detect a cycle (between different instantiations of the same | |
6df42991 | 5736 | loop). |
0683b6fa SM |
5737 | So the method used for those nasty loops is a little different: |
5738 | We use a special cycle-detection-stack-frame which is pushed | |
5739 | when the on_failure_jump_nastyloop failure-point is *popped*. | |
5740 | This special frame thus marks the beginning of one iteration | |
5741 | through the loop and we can hence easily check right here | |
5742 | whether something matched between the beginning and the end of | |
5743 | the loop. */ | |
5744 | case on_failure_jump_nastyloop: | |
5745 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
dc4a2ee0 PE |
5746 | DEBUG_PRINT ("EXECUTING on_failure_jump_nastyloop %d (to %p):\n", |
5747 | mcnt, p + mcnt); | |
0683b6fa SM |
5748 | |
5749 | assert ((re_opcode_t)p[-4] == no_op); | |
6df42991 SM |
5750 | { |
5751 | int cycle = 0; | |
5752 | CHECK_INFINITE_LOOP (p - 4, d); | |
5753 | if (!cycle) | |
5754 | /* If there's a cycle, just continue without pushing | |
5755 | this failure point. The failure point is the "try again" | |
5756 | option, which shouldn't be tried. | |
5757 | We want (x?)*?y\1z to match both xxyz and xxyxz. */ | |
5758 | PUSH_FAILURE_POINT (p - 3, d); | |
5759 | } | |
0683b6fa SM |
5760 | break; |
5761 | ||
4e8a9132 SM |
5762 | /* Simple loop detecting on_failure_jump: just check on the |
5763 | failure stack if the same spot was already hit earlier. */ | |
505bde11 SM |
5764 | case on_failure_jump_loop: |
5765 | on_failure: | |
5766 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
dc4a2ee0 PE |
5767 | DEBUG_PRINT ("EXECUTING on_failure_jump_loop %d (to %p):\n", |
5768 | mcnt, p + mcnt); | |
6df42991 SM |
5769 | { |
5770 | int cycle = 0; | |
5771 | CHECK_INFINITE_LOOP (p - 3, d); | |
5772 | if (cycle) | |
5773 | /* If there's a cycle, get out of the loop, as if the matching | |
5774 | had failed. We used to just `goto fail' here, but that was | |
5775 | aborting the search a bit too early: we want to keep the | |
5776 | empty-loop-match and keep matching after the loop. | |
5777 | We want (x?)*y\1z to match both xxyz and xxyxz. */ | |
5778 | p += mcnt; | |
5779 | else | |
5780 | PUSH_FAILURE_POINT (p - 3, d); | |
5781 | } | |
25fe55af | 5782 | break; |
fa9a63c5 RM |
5783 | |
5784 | ||
5785 | /* Uses of on_failure_jump: | |
5e69f11e | 5786 | |
25fe55af RS |
5787 | Each alternative starts with an on_failure_jump that points |
5788 | to the beginning of the next alternative. Each alternative | |
5789 | except the last ends with a jump that in effect jumps past | |
5790 | the rest of the alternatives. (They really jump to the | |
5791 | ending jump of the following alternative, because tensioning | |
5792 | these jumps is a hassle.) | |
fa9a63c5 | 5793 | |
25fe55af RS |
5794 | Repeats start with an on_failure_jump that points past both |
5795 | the repetition text and either the following jump or | |
5796 | pop_failure_jump back to this on_failure_jump. */ | |
fa9a63c5 | 5797 | case on_failure_jump: |
25fe55af | 5798 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5799 | DEBUG_PRINT ("EXECUTING on_failure_jump %d (to %p):\n", |
5800 | mcnt, p + mcnt); | |
25fe55af | 5801 | |
505bde11 | 5802 | PUSH_FAILURE_POINT (p -3, d); |
25fe55af RS |
5803 | break; |
5804 | ||
4e8a9132 | 5805 | /* This operation is used for greedy *. |
505bde11 SM |
5806 | Compare the beginning of the repeat with what in the |
5807 | pattern follows its end. If we can establish that there | |
5808 | is nothing that they would both match, i.e., that we | |
5809 | would have to backtrack because of (as in, e.g., `a*a') | |
5810 | then we can use a non-backtracking loop based on | |
4e8a9132 | 5811 | on_failure_keep_string_jump instead of on_failure_jump. */ |
505bde11 | 5812 | case on_failure_jump_smart: |
25fe55af | 5813 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5814 | DEBUG_PRINT ("EXECUTING on_failure_jump_smart %d (to %p).\n", |
5815 | mcnt, p + mcnt); | |
25fe55af | 5816 | { |
01618498 | 5817 | re_char *p1 = p; /* Next operation. */ |
6dcf2d0e SM |
5818 | /* Here, we discard `const', making re_match non-reentrant. */ |
5819 | unsigned char *p2 = (unsigned char*) p + mcnt; /* Jump dest. */ | |
5820 | unsigned char *p3 = (unsigned char*) p - 3; /* opcode location. */ | |
fa9a63c5 | 5821 | |
505bde11 SM |
5822 | p -= 3; /* Reset so that we will re-execute the |
5823 | instruction once it's been changed. */ | |
fa9a63c5 | 5824 | |
4e8a9132 SM |
5825 | EXTRACT_NUMBER (mcnt, p2 - 2); |
5826 | ||
5827 | /* Ensure this is a indeed the trivial kind of loop | |
5828 | we are expecting. */ | |
5829 | assert (skip_one_char (p1) == p2 - 3); | |
5830 | assert ((re_opcode_t) p2[-3] == jump && p2 + mcnt == p); | |
99633e97 | 5831 | DEBUG_STATEMENT (debug += 2); |
505bde11 | 5832 | if (mutually_exclusive_p (bufp, p1, p2)) |
fa9a63c5 | 5833 | { |
505bde11 | 5834 | /* Use a fast `on_failure_keep_string_jump' loop. */ |
dc4a2ee0 | 5835 | DEBUG_PRINT (" smart exclusive => fast loop.\n"); |
01618498 | 5836 | *p3 = (unsigned char) on_failure_keep_string_jump; |
4e8a9132 | 5837 | STORE_NUMBER (p2 - 2, mcnt + 3); |
25fe55af | 5838 | } |
505bde11 | 5839 | else |
fa9a63c5 | 5840 | { |
505bde11 | 5841 | /* Default to a safe `on_failure_jump' loop. */ |
dc4a2ee0 | 5842 | DEBUG_PRINT (" smart default => slow loop.\n"); |
01618498 | 5843 | *p3 = (unsigned char) on_failure_jump; |
fa9a63c5 | 5844 | } |
99633e97 | 5845 | DEBUG_STATEMENT (debug -= 2); |
25fe55af | 5846 | } |
505bde11 | 5847 | break; |
25fe55af RS |
5848 | |
5849 | /* Unconditionally jump (without popping any failure points). */ | |
5850 | case jump: | |
fa9a63c5 | 5851 | unconditional_jump: |
5b370c2b | 5852 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 5853 | EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ |
dc4a2ee0 | 5854 | DEBUG_PRINT ("EXECUTING jump %d ", mcnt); |
7814e705 | 5855 | p += mcnt; /* Do the jump. */ |
dc4a2ee0 | 5856 | DEBUG_PRINT ("(to %p).\n", p); |
25fe55af RS |
5857 | break; |
5858 | ||
5859 | ||
25fe55af RS |
5860 | /* Have to succeed matching what follows at least n times. |
5861 | After that, handle like `on_failure_jump'. */ | |
5862 | case succeed_n: | |
01618498 | 5863 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af | 5864 | EXTRACT_NUMBER (mcnt, p + 2); |
dc4a2ee0 | 5865 | DEBUG_PRINT ("EXECUTING succeed_n %d.\n", mcnt); |
5e69f11e | 5866 | |
dc1e502d SM |
5867 | /* Originally, mcnt is how many times we HAVE to succeed. */ |
5868 | if (mcnt != 0) | |
25fe55af | 5869 | { |
6dcf2d0e SM |
5870 | /* Here, we discard `const', making re_match non-reentrant. */ |
5871 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5872 | mcnt--; |
01618498 SM |
5873 | p += 4; |
5874 | PUSH_NUMBER (p2, mcnt); | |
25fe55af | 5875 | } |
dc1e502d SM |
5876 | else |
5877 | /* The two bytes encoding mcnt == 0 are two no_op opcodes. */ | |
5878 | goto on_failure; | |
25fe55af RS |
5879 | break; |
5880 | ||
5881 | case jump_n: | |
01618498 | 5882 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af | 5883 | EXTRACT_NUMBER (mcnt, p + 2); |
dc4a2ee0 | 5884 | DEBUG_PRINT ("EXECUTING jump_n %d.\n", mcnt); |
25fe55af RS |
5885 | |
5886 | /* Originally, this is how many times we CAN jump. */ | |
dc1e502d | 5887 | if (mcnt != 0) |
25fe55af | 5888 | { |
6dcf2d0e SM |
5889 | /* Here, we discard `const', making re_match non-reentrant. */ |
5890 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5891 | mcnt--; |
01618498 | 5892 | PUSH_NUMBER (p2, mcnt); |
dc1e502d | 5893 | goto unconditional_jump; |
25fe55af RS |
5894 | } |
5895 | /* If don't have to jump any more, skip over the rest of command. */ | |
5e69f11e RM |
5896 | else |
5897 | p += 4; | |
25fe55af | 5898 | break; |
5e69f11e | 5899 | |
fa9a63c5 RM |
5900 | case set_number_at: |
5901 | { | |
01618498 | 5902 | unsigned char *p2; /* Location of the counter. */ |
dc4a2ee0 | 5903 | DEBUG_PRINT ("EXECUTING set_number_at.\n"); |
fa9a63c5 | 5904 | |
25fe55af | 5905 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
6dcf2d0e SM |
5906 | /* Here, we discard `const', making re_match non-reentrant. */ |
5907 | p2 = (unsigned char*) p + mcnt; | |
f224e500 | 5908 | /* Signedness doesn't matter since we only copy MCNT's bits. */ |
25fe55af | 5909 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 | 5910 | DEBUG_PRINT (" Setting %p to %d.\n", p2, mcnt); |
01618498 | 5911 | PUSH_NUMBER (p2, mcnt); |
25fe55af RS |
5912 | break; |
5913 | } | |
9121ca40 KH |
5914 | |
5915 | case wordbound: | |
66f0296e | 5916 | case notwordbound: |
19ed5445 PE |
5917 | { |
5918 | boolean not = (re_opcode_t) *(p - 1) == notwordbound; | |
dc4a2ee0 | 5919 | DEBUG_PRINT ("EXECUTING %swordbound.\n", not ? "not" : ""); |
fa9a63c5 | 5920 | |
19ed5445 | 5921 | /* We SUCCEED (or FAIL) in one of the following cases: */ |
9121ca40 | 5922 | |
19ed5445 PE |
5923 | /* Case 1: D is at the beginning or the end of string. */ |
5924 | if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) | |
5925 | not = !not; | |
5926 | else | |
5927 | { | |
5928 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5929 | is the character at D, and S2 is the syntax of C2. */ | |
5930 | re_wchar_t c1, c2; | |
5931 | int s1, s2; | |
5932 | int dummy; | |
b18215fc | 5933 | #ifdef emacs |
d1dfb56c EZ |
5934 | ssize_t offset = PTR_TO_OFFSET (d - 1); |
5935 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 | 5936 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5937 | #endif |
19ed5445 PE |
5938 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
5939 | s1 = SYNTAX (c1); | |
b18215fc | 5940 | #ifdef emacs |
19ed5445 | 5941 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
25fe55af | 5942 | #endif |
19ed5445 PE |
5943 | PREFETCH_NOLIMIT (); |
5944 | GET_CHAR_AFTER (c2, d, dummy); | |
5945 | s2 = SYNTAX (c2); | |
5946 | ||
5947 | if (/* Case 2: Only one of S1 and S2 is Sword. */ | |
5948 | ((s1 == Sword) != (s2 == Sword)) | |
5949 | /* Case 3: Both of S1 and S2 are Sword, and macro | |
5950 | WORD_BOUNDARY_P (C1, C2) returns nonzero. */ | |
5951 | || ((s1 == Sword) && WORD_BOUNDARY_P (c1, c2))) | |
5952 | not = !not; | |
5953 | } | |
5954 | if (not) | |
5955 | break; | |
5956 | else | |
5957 | goto fail; | |
5958 | } | |
fa9a63c5 RM |
5959 | |
5960 | case wordbeg: | |
dc4a2ee0 | 5961 | DEBUG_PRINT ("EXECUTING wordbeg.\n"); |
fa9a63c5 | 5962 | |
b18215fc RS |
5963 | /* We FAIL in one of the following cases: */ |
5964 | ||
7814e705 | 5965 | /* Case 1: D is at the end of string. */ |
b18215fc | 5966 | if (AT_STRINGS_END (d)) |
99633e97 | 5967 | goto fail; |
b18215fc RS |
5968 | else |
5969 | { | |
5970 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5971 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
5972 | re_wchar_t c1, c2; |
5973 | int s1, s2; | |
bf216479 | 5974 | int dummy; |
fa9a63c5 | 5975 | #ifdef emacs |
d1dfb56c EZ |
5976 | ssize_t offset = PTR_TO_OFFSET (d); |
5977 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 5978 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5979 | #endif |
99633e97 | 5980 | PREFETCH (); |
6fdd04b0 | 5981 | GET_CHAR_AFTER (c2, d, dummy); |
b18215fc | 5982 | s2 = SYNTAX (c2); |
177c0ea7 | 5983 | |
b18215fc RS |
5984 | /* Case 2: S2 is not Sword. */ |
5985 | if (s2 != Sword) | |
5986 | goto fail; | |
5987 | ||
5988 | /* Case 3: D is not at the beginning of string ... */ | |
5989 | if (!AT_STRINGS_BEG (d)) | |
5990 | { | |
5991 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
5992 | #ifdef emacs | |
5d967c7a | 5993 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); |
25fe55af | 5994 | #endif |
b18215fc RS |
5995 | s1 = SYNTAX (c1); |
5996 | ||
5997 | /* ... and S1 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 5998 | returns 0. */ |
b18215fc RS |
5999 | if ((s1 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
6000 | goto fail; | |
6001 | } | |
6002 | } | |
e318085a RS |
6003 | break; |
6004 | ||
b18215fc | 6005 | case wordend: |
dc4a2ee0 | 6006 | DEBUG_PRINT ("EXECUTING wordend.\n"); |
b18215fc RS |
6007 | |
6008 | /* We FAIL in one of the following cases: */ | |
6009 | ||
6010 | /* Case 1: D is at the beginning of string. */ | |
6011 | if (AT_STRINGS_BEG (d)) | |
e318085a | 6012 | goto fail; |
b18215fc RS |
6013 | else |
6014 | { | |
6015 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6016 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
6017 | re_wchar_t c1, c2; |
6018 | int s1, s2; | |
bf216479 | 6019 | int dummy; |
5d967c7a | 6020 | #ifdef emacs |
d1dfb56c EZ |
6021 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6022 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 6023 | UPDATE_SYNTAX_TABLE (charpos); |
5d967c7a | 6024 | #endif |
99633e97 | 6025 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
b18215fc RS |
6026 | s1 = SYNTAX (c1); |
6027 | ||
6028 | /* Case 2: S1 is not Sword. */ | |
6029 | if (s1 != Sword) | |
6030 | goto fail; | |
6031 | ||
6032 | /* Case 3: D is not at the end of string ... */ | |
6033 | if (!AT_STRINGS_END (d)) | |
6034 | { | |
f1ad044f | 6035 | PREFETCH_NOLIMIT (); |
6fdd04b0 | 6036 | GET_CHAR_AFTER (c2, d, dummy); |
5d967c7a RS |
6037 | #ifdef emacs |
6038 | UPDATE_SYNTAX_TABLE_FORWARD (charpos); | |
6039 | #endif | |
b18215fc RS |
6040 | s2 = SYNTAX (c2); |
6041 | ||
6042 | /* ... and S2 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6043 | returns 0. */ |
b18215fc | 6044 | if ((s2 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
25fe55af | 6045 | goto fail; |
b18215fc RS |
6046 | } |
6047 | } | |
e318085a RS |
6048 | break; |
6049 | ||
669fa600 | 6050 | case symbeg: |
dc4a2ee0 | 6051 | DEBUG_PRINT ("EXECUTING symbeg.\n"); |
669fa600 SM |
6052 | |
6053 | /* We FAIL in one of the following cases: */ | |
6054 | ||
7814e705 | 6055 | /* Case 1: D is at the end of string. */ |
669fa600 SM |
6056 | if (AT_STRINGS_END (d)) |
6057 | goto fail; | |
6058 | else | |
6059 | { | |
6060 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6061 | is the character at D, and S2 is the syntax of C2. */ | |
6062 | re_wchar_t c1, c2; | |
6063 | int s1, s2; | |
6064 | #ifdef emacs | |
d1dfb56c EZ |
6065 | ssize_t offset = PTR_TO_OFFSET (d); |
6066 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6067 | UPDATE_SYNTAX_TABLE (charpos); |
6068 | #endif | |
6069 | PREFETCH (); | |
62a6e103 | 6070 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6071 | s2 = SYNTAX (c2); |
7814e705 | 6072 | |
669fa600 SM |
6073 | /* Case 2: S2 is neither Sword nor Ssymbol. */ |
6074 | if (s2 != Sword && s2 != Ssymbol) | |
6075 | goto fail; | |
6076 | ||
6077 | /* Case 3: D is not at the beginning of string ... */ | |
6078 | if (!AT_STRINGS_BEG (d)) | |
6079 | { | |
6080 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6081 | #ifdef emacs | |
6082 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); | |
6083 | #endif | |
6084 | s1 = SYNTAX (c1); | |
6085 | ||
6086 | /* ... and S1 is Sword or Ssymbol. */ | |
6087 | if (s1 == Sword || s1 == Ssymbol) | |
6088 | goto fail; | |
6089 | } | |
6090 | } | |
6091 | break; | |
6092 | ||
6093 | case symend: | |
dc4a2ee0 | 6094 | DEBUG_PRINT ("EXECUTING symend.\n"); |
669fa600 SM |
6095 | |
6096 | /* We FAIL in one of the following cases: */ | |
6097 | ||
6098 | /* Case 1: D is at the beginning of string. */ | |
6099 | if (AT_STRINGS_BEG (d)) | |
6100 | goto fail; | |
6101 | else | |
6102 | { | |
6103 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6104 | is the character at D, and S2 is the syntax of C2. */ | |
6105 | re_wchar_t c1, c2; | |
6106 | int s1, s2; | |
6107 | #ifdef emacs | |
d1dfb56c EZ |
6108 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6109 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6110 | UPDATE_SYNTAX_TABLE (charpos); |
6111 | #endif | |
6112 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6113 | s1 = SYNTAX (c1); | |
6114 | ||
6115 | /* Case 2: S1 is neither Ssymbol nor Sword. */ | |
6116 | if (s1 != Sword && s1 != Ssymbol) | |
6117 | goto fail; | |
6118 | ||
6119 | /* Case 3: D is not at the end of string ... */ | |
6120 | if (!AT_STRINGS_END (d)) | |
6121 | { | |
6122 | PREFETCH_NOLIMIT (); | |
62a6e103 | 6123 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6124 | #ifdef emacs |
134579f2 | 6125 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
669fa600 SM |
6126 | #endif |
6127 | s2 = SYNTAX (c2); | |
6128 | ||
6129 | /* ... and S2 is Sword or Ssymbol. */ | |
6130 | if (s2 == Sword || s2 == Ssymbol) | |
6131 | goto fail; | |
b18215fc RS |
6132 | } |
6133 | } | |
e318085a RS |
6134 | break; |
6135 | ||
fa9a63c5 | 6136 | case syntaxspec: |
1fb352e0 | 6137 | case notsyntaxspec: |
b18215fc | 6138 | { |
19ed5445 PE |
6139 | boolean not = (re_opcode_t) *(p - 1) == notsyntaxspec; |
6140 | mcnt = *p++; | |
dc4a2ee0 PE |
6141 | DEBUG_PRINT ("EXECUTING %ssyntaxspec %d.\n", not ? "not" : "", |
6142 | mcnt); | |
19ed5445 PE |
6143 | PREFETCH (); |
6144 | #ifdef emacs | |
6145 | { | |
d1dfb56c EZ |
6146 | ssize_t offset = PTR_TO_OFFSET (d); |
6147 | ssize_t pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 PE |
6148 | UPDATE_SYNTAX_TABLE (pos1); |
6149 | } | |
25fe55af | 6150 | #endif |
19ed5445 PE |
6151 | { |
6152 | int len; | |
6153 | re_wchar_t c; | |
b18215fc | 6154 | |
19ed5445 PE |
6155 | GET_CHAR_AFTER (c, d, len); |
6156 | if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not) | |
6157 | goto fail; | |
6158 | d += len; | |
6159 | } | |
b18215fc | 6160 | } |
8fb31792 | 6161 | break; |
fa9a63c5 | 6162 | |
b18215fc | 6163 | #ifdef emacs |
1fb352e0 | 6164 | case before_dot: |
dc4a2ee0 | 6165 | DEBUG_PRINT ("EXECUTING before_dot.\n"); |
1fb352e0 | 6166 | if (PTR_BYTE_POS (d) >= PT_BYTE) |
fa9a63c5 | 6167 | goto fail; |
b18215fc RS |
6168 | break; |
6169 | ||
1fb352e0 | 6170 | case at_dot: |
dc4a2ee0 | 6171 | DEBUG_PRINT ("EXECUTING at_dot.\n"); |
1fb352e0 SM |
6172 | if (PTR_BYTE_POS (d) != PT_BYTE) |
6173 | goto fail; | |
6174 | break; | |
b18215fc | 6175 | |
1fb352e0 | 6176 | case after_dot: |
dc4a2ee0 | 6177 | DEBUG_PRINT ("EXECUTING after_dot.\n"); |
1fb352e0 SM |
6178 | if (PTR_BYTE_POS (d) <= PT_BYTE) |
6179 | goto fail; | |
e318085a | 6180 | break; |
fa9a63c5 | 6181 | |
1fb352e0 | 6182 | case categoryspec: |
b18215fc | 6183 | case notcategoryspec: |
b18215fc | 6184 | { |
8fb31792 PE |
6185 | boolean not = (re_opcode_t) *(p - 1) == notcategoryspec; |
6186 | mcnt = *p++; | |
dc4a2ee0 PE |
6187 | DEBUG_PRINT ("EXECUTING %scategoryspec %d.\n", |
6188 | not ? "not" : "", mcnt); | |
8fb31792 | 6189 | PREFETCH (); |
01618498 | 6190 | |
8fb31792 PE |
6191 | { |
6192 | int len; | |
6193 | re_wchar_t c; | |
6194 | GET_CHAR_AFTER (c, d, len); | |
6195 | if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not) | |
6196 | goto fail; | |
6197 | d += len; | |
6198 | } | |
b18215fc | 6199 | } |
fa9a63c5 | 6200 | break; |
5e69f11e | 6201 | |
1fb352e0 | 6202 | #endif /* emacs */ |
5e69f11e | 6203 | |
0b32bf0e SM |
6204 | default: |
6205 | abort (); | |
fa9a63c5 | 6206 | } |
b18215fc | 6207 | continue; /* Successfully executed one pattern command; keep going. */ |
fa9a63c5 RM |
6208 | |
6209 | ||
6210 | /* We goto here if a matching operation fails. */ | |
6211 | fail: | |
5b370c2b | 6212 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 6213 | if (!FAIL_STACK_EMPTY ()) |
505bde11 | 6214 | { |
01618498 | 6215 | re_char *str, *pat; |
505bde11 | 6216 | /* A restart point is known. Restore to that state. */ |
dc4a2ee0 | 6217 | DEBUG_PRINT ("\nFAIL:\n"); |
0b32bf0e | 6218 | POP_FAILURE_POINT (str, pat); |
7393bcbb | 6219 | switch (*pat++) |
505bde11 SM |
6220 | { |
6221 | case on_failure_keep_string_jump: | |
6222 | assert (str == NULL); | |
6223 | goto continue_failure_jump; | |
6224 | ||
0683b6fa SM |
6225 | case on_failure_jump_nastyloop: |
6226 | assert ((re_opcode_t)pat[-2] == no_op); | |
6227 | PUSH_FAILURE_POINT (pat - 2, str); | |
6228 | /* Fallthrough */ | |
6229 | ||
505bde11 SM |
6230 | case on_failure_jump_loop: |
6231 | case on_failure_jump: | |
6232 | case succeed_n: | |
6233 | d = str; | |
6234 | continue_failure_jump: | |
6235 | EXTRACT_NUMBER_AND_INCR (mcnt, pat); | |
6236 | p = pat + mcnt; | |
6237 | break; | |
b18215fc | 6238 | |
0683b6fa SM |
6239 | case no_op: |
6240 | /* A special frame used for nastyloops. */ | |
6241 | goto fail; | |
6242 | ||
505bde11 | 6243 | default: |
5e617bc2 | 6244 | abort (); |
505bde11 | 6245 | } |
fa9a63c5 | 6246 | |
505bde11 | 6247 | assert (p >= bufp->buffer && p <= pend); |
b18215fc | 6248 | |
0b32bf0e | 6249 | if (d >= string1 && d <= end1) |
fa9a63c5 | 6250 | dend = end_match_1; |
0b32bf0e | 6251 | } |
fa9a63c5 | 6252 | else |
0b32bf0e | 6253 | break; /* Matching at this starting point really fails. */ |
fa9a63c5 RM |
6254 | } /* for (;;) */ |
6255 | ||
6256 | if (best_regs_set) | |
6257 | goto restore_best_regs; | |
6258 | ||
6259 | FREE_VARIABLES (); | |
6260 | ||
b18215fc | 6261 | return -1; /* Failure to match. */ |
dc4a2ee0 | 6262 | } |
fa9a63c5 RM |
6263 | \f |
6264 | /* Subroutine definitions for re_match_2. */ | |
6265 | ||
fa9a63c5 RM |
6266 | /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN |
6267 | bytes; nonzero otherwise. */ | |
5e69f11e | 6268 | |
fa9a63c5 | 6269 | static int |
29abe551 | 6270 | bcmp_translate (const_re_char *s1, const_re_char *s2, register ssize_t len, |
438105ed | 6271 | RE_TRANSLATE_TYPE translate, const int target_multibyte) |
fa9a63c5 | 6272 | { |
2d1675e4 SM |
6273 | register re_char *p1 = s1, *p2 = s2; |
6274 | re_char *p1_end = s1 + len; | |
6275 | re_char *p2_end = s2 + len; | |
e934739e | 6276 | |
4bb91c68 SM |
6277 | /* FIXME: Checking both p1 and p2 presumes that the two strings might have |
6278 | different lengths, but relying on a single `len' would break this. -sm */ | |
6279 | while (p1 < p1_end && p2 < p2_end) | |
fa9a63c5 | 6280 | { |
e934739e | 6281 | int p1_charlen, p2_charlen; |
01618498 | 6282 | re_wchar_t p1_ch, p2_ch; |
e934739e | 6283 | |
6fdd04b0 KH |
6284 | GET_CHAR_AFTER (p1_ch, p1, p1_charlen); |
6285 | GET_CHAR_AFTER (p2_ch, p2, p2_charlen); | |
e934739e RS |
6286 | |
6287 | if (RE_TRANSLATE (translate, p1_ch) | |
6288 | != RE_TRANSLATE (translate, p2_ch)) | |
bc192b5b | 6289 | return 1; |
e934739e RS |
6290 | |
6291 | p1 += p1_charlen, p2 += p2_charlen; | |
fa9a63c5 | 6292 | } |
e934739e RS |
6293 | |
6294 | if (p1 != p1_end || p2 != p2_end) | |
6295 | return 1; | |
6296 | ||
fa9a63c5 RM |
6297 | return 0; |
6298 | } | |
6299 | \f | |
6300 | /* Entry points for GNU code. */ | |
6301 | ||
6302 | /* re_compile_pattern is the GNU regular expression compiler: it | |
6303 | compiles PATTERN (of length SIZE) and puts the result in BUFP. | |
6304 | Returns 0 if the pattern was valid, otherwise an error string. | |
5e69f11e | 6305 | |
fa9a63c5 RM |
6306 | Assumes the `allocated' (and perhaps `buffer') and `translate' fields |
6307 | are set in BUFP on entry. | |
5e69f11e | 6308 | |
b18215fc | 6309 | We call regex_compile to do the actual compilation. */ |
fa9a63c5 RM |
6310 | |
6311 | const char * | |
d1dfb56c EZ |
6312 | re_compile_pattern (const char *pattern, size_t length, |
6313 | struct re_pattern_buffer *bufp) | |
fa9a63c5 RM |
6314 | { |
6315 | reg_errcode_t ret; | |
5e69f11e | 6316 | |
fa9a63c5 RM |
6317 | /* GNU code is written to assume at least RE_NREGS registers will be set |
6318 | (and at least one extra will be -1). */ | |
6319 | bufp->regs_allocated = REGS_UNALLOCATED; | |
5e69f11e | 6320 | |
fa9a63c5 RM |
6321 | /* And GNU code determines whether or not to get register information |
6322 | by passing null for the REGS argument to re_match, etc., not by | |
6323 | setting no_sub. */ | |
6324 | bufp->no_sub = 0; | |
5e69f11e | 6325 | |
4bb91c68 | 6326 | ret = regex_compile ((re_char*) pattern, length, re_syntax_options, bufp); |
fa9a63c5 RM |
6327 | |
6328 | if (!ret) | |
6329 | return NULL; | |
6330 | return gettext (re_error_msgid[(int) ret]); | |
5e69f11e | 6331 | } |
c0f9ea08 | 6332 | WEAK_ALIAS (__re_compile_pattern, re_compile_pattern) |
fa9a63c5 | 6333 | \f |
b18215fc RS |
6334 | /* Entry points compatible with 4.2 BSD regex library. We don't define |
6335 | them unless specifically requested. */ | |
fa9a63c5 | 6336 | |
0b32bf0e | 6337 | #if defined _REGEX_RE_COMP || defined _LIBC |
fa9a63c5 RM |
6338 | |
6339 | /* BSD has one and only one pattern buffer. */ | |
6340 | static struct re_pattern_buffer re_comp_buf; | |
6341 | ||
6342 | char * | |
0b32bf0e | 6343 | # ifdef _LIBC |
48afdd44 RM |
6344 | /* Make these definitions weak in libc, so POSIX programs can redefine |
6345 | these names if they don't use our functions, and still use | |
6346 | regcomp/regexec below without link errors. */ | |
6347 | weak_function | |
0b32bf0e | 6348 | # endif |
31011111 | 6349 | re_comp (const char *s) |
fa9a63c5 RM |
6350 | { |
6351 | reg_errcode_t ret; | |
5e69f11e | 6352 | |
fa9a63c5 RM |
6353 | if (!s) |
6354 | { | |
6355 | if (!re_comp_buf.buffer) | |
0b32bf0e | 6356 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
a60198e5 | 6357 | return (char *) gettext ("No previous regular expression"); |
fa9a63c5 RM |
6358 | return 0; |
6359 | } | |
6360 | ||
6361 | if (!re_comp_buf.buffer) | |
6362 | { | |
38182d90 | 6363 | re_comp_buf.buffer = malloc (200); |
fa9a63c5 | 6364 | if (re_comp_buf.buffer == NULL) |
0b32bf0e SM |
6365 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6366 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6367 | re_comp_buf.allocated = 200; |
6368 | ||
38182d90 | 6369 | re_comp_buf.fastmap = malloc (1 << BYTEWIDTH); |
fa9a63c5 | 6370 | if (re_comp_buf.fastmap == NULL) |
a60198e5 SM |
6371 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6372 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6373 | } |
6374 | ||
6375 | /* Since `re_exec' always passes NULL for the `regs' argument, we | |
6376 | don't need to initialize the pattern buffer fields which affect it. */ | |
6377 | ||
fa9a63c5 | 6378 | ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); |
5e69f11e | 6379 | |
fa9a63c5 RM |
6380 | if (!ret) |
6381 | return NULL; | |
6382 | ||
6383 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | |
6384 | return (char *) gettext (re_error_msgid[(int) ret]); | |
6385 | } | |
6386 | ||
6387 | ||
31011111 | 6388 | int |
0b32bf0e | 6389 | # ifdef _LIBC |
48afdd44 | 6390 | weak_function |
0b32bf0e | 6391 | # endif |
d1dfb56c | 6392 | re_exec (const char *s) |
fa9a63c5 | 6393 | { |
d1dfb56c | 6394 | const size_t len = strlen (s); |
7d652d97 | 6395 | return re_search (&re_comp_buf, s, len, 0, len, 0) >= 0; |
fa9a63c5 RM |
6396 | } |
6397 | #endif /* _REGEX_RE_COMP */ | |
6398 | \f | |
6399 | /* POSIX.2 functions. Don't define these for Emacs. */ | |
6400 | ||
6401 | #ifndef emacs | |
6402 | ||
6403 | /* regcomp takes a regular expression as a string and compiles it. | |
6404 | ||
b18215fc | 6405 | PREG is a regex_t *. We do not expect any fields to be initialized, |
fa9a63c5 RM |
6406 | since POSIX says we shouldn't. Thus, we set |
6407 | ||
6408 | `buffer' to the compiled pattern; | |
6409 | `used' to the length of the compiled pattern; | |
6410 | `syntax' to RE_SYNTAX_POSIX_EXTENDED if the | |
6411 | REG_EXTENDED bit in CFLAGS is set; otherwise, to | |
6412 | RE_SYNTAX_POSIX_BASIC; | |
c0f9ea08 SM |
6413 | `fastmap' to an allocated space for the fastmap; |
6414 | `fastmap_accurate' to zero; | |
fa9a63c5 RM |
6415 | `re_nsub' to the number of subexpressions in PATTERN. |
6416 | ||
6417 | PATTERN is the address of the pattern string. | |
6418 | ||
6419 | CFLAGS is a series of bits which affect compilation. | |
6420 | ||
6421 | If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we | |
6422 | use POSIX basic syntax. | |
6423 | ||
6424 | If REG_NEWLINE is set, then . and [^...] don't match newline. | |
6425 | Also, regexec will try a match beginning after every newline. | |
6426 | ||
6427 | If REG_ICASE is set, then we considers upper- and lowercase | |
6428 | versions of letters to be equivalent when matching. | |
6429 | ||
6430 | If REG_NOSUB is set, then when PREG is passed to regexec, that | |
6431 | routine will report only success or failure, and nothing about the | |
6432 | registers. | |
6433 | ||
b18215fc | 6434 | It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for |
fa9a63c5 RM |
6435 | the return codes and their meanings.) */ |
6436 | ||
d1dfb56c | 6437 | reg_errcode_t |
29abe551 | 6438 | regcomp (regex_t *_Restrict_ preg, const char *_Restrict_ pattern, |
d2762c86 | 6439 | int cflags) |
fa9a63c5 RM |
6440 | { |
6441 | reg_errcode_t ret; | |
4bb91c68 | 6442 | reg_syntax_t syntax |
fa9a63c5 RM |
6443 | = (cflags & REG_EXTENDED) ? |
6444 | RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; | |
6445 | ||
6446 | /* regex_compile will allocate the space for the compiled pattern. */ | |
6447 | preg->buffer = 0; | |
6448 | preg->allocated = 0; | |
6449 | preg->used = 0; | |
5e69f11e | 6450 | |
c0f9ea08 | 6451 | /* Try to allocate space for the fastmap. */ |
38182d90 | 6452 | preg->fastmap = malloc (1 << BYTEWIDTH); |
5e69f11e | 6453 | |
fa9a63c5 RM |
6454 | if (cflags & REG_ICASE) |
6455 | { | |
6456 | unsigned i; | |
5e69f11e | 6457 | |
38182d90 | 6458 | preg->translate = malloc (CHAR_SET_SIZE * sizeof *preg->translate); |
fa9a63c5 | 6459 | if (preg->translate == NULL) |
0b32bf0e | 6460 | return (int) REG_ESPACE; |
fa9a63c5 RM |
6461 | |
6462 | /* Map uppercase characters to corresponding lowercase ones. */ | |
6463 | for (i = 0; i < CHAR_SET_SIZE; i++) | |
4bb91c68 | 6464 | preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i; |
fa9a63c5 RM |
6465 | } |
6466 | else | |
6467 | preg->translate = NULL; | |
6468 | ||
6469 | /* If REG_NEWLINE is set, newlines are treated differently. */ | |
6470 | if (cflags & REG_NEWLINE) | |
6471 | { /* REG_NEWLINE implies neither . nor [^...] match newline. */ | |
6472 | syntax &= ~RE_DOT_NEWLINE; | |
6473 | syntax |= RE_HAT_LISTS_NOT_NEWLINE; | |
fa9a63c5 RM |
6474 | } |
6475 | else | |
c0f9ea08 | 6476 | syntax |= RE_NO_NEWLINE_ANCHOR; |
fa9a63c5 RM |
6477 | |
6478 | preg->no_sub = !!(cflags & REG_NOSUB); | |
6479 | ||
5e69f11e | 6480 | /* POSIX says a null character in the pattern terminates it, so we |
fa9a63c5 | 6481 | can use strlen here in compiling the pattern. */ |
4bb91c68 | 6482 | ret = regex_compile ((re_char*) pattern, strlen (pattern), syntax, preg); |
5e69f11e | 6483 | |
fa9a63c5 RM |
6484 | /* POSIX doesn't distinguish between an unmatched open-group and an |
6485 | unmatched close-group: both are REG_EPAREN. */ | |
c0f9ea08 SM |
6486 | if (ret == REG_ERPAREN) |
6487 | ret = REG_EPAREN; | |
6488 | ||
6489 | if (ret == REG_NOERROR && preg->fastmap) | |
6490 | { /* Compute the fastmap now, since regexec cannot modify the pattern | |
6491 | buffer. */ | |
6492 | re_compile_fastmap (preg); | |
6493 | if (preg->can_be_null) | |
6494 | { /* The fastmap can't be used anyway. */ | |
6495 | free (preg->fastmap); | |
6496 | preg->fastmap = NULL; | |
6497 | } | |
6498 | } | |
d1dfb56c | 6499 | return ret; |
fa9a63c5 | 6500 | } |
c0f9ea08 | 6501 | WEAK_ALIAS (__regcomp, regcomp) |
fa9a63c5 RM |
6502 | |
6503 | ||
6504 | /* regexec searches for a given pattern, specified by PREG, in the | |
6505 | string STRING. | |
5e69f11e | 6506 | |
fa9a63c5 | 6507 | If NMATCH is zero or REG_NOSUB was set in the cflags argument to |
b18215fc | 6508 | `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at |
fa9a63c5 RM |
6509 | least NMATCH elements, and we set them to the offsets of the |
6510 | corresponding matched substrings. | |
5e69f11e | 6511 | |
fa9a63c5 RM |
6512 | EFLAGS specifies `execution flags' which affect matching: if |
6513 | REG_NOTBOL is set, then ^ does not match at the beginning of the | |
6514 | string; if REG_NOTEOL is set, then $ does not match at the end. | |
5e69f11e | 6515 | |
fa9a63c5 RM |
6516 | We return 0 if we find a match and REG_NOMATCH if not. */ |
6517 | ||
d1dfb56c | 6518 | reg_errcode_t |
29abe551 PE |
6519 | regexec (const regex_t *_Restrict_ preg, const char *_Restrict_ string, |
6520 | size_t nmatch, regmatch_t pmatch[_Restrict_arr_], int eflags) | |
fa9a63c5 | 6521 | { |
31011111 | 6522 | regoff_t ret; |
fa9a63c5 RM |
6523 | struct re_registers regs; |
6524 | regex_t private_preg; | |
d1dfb56c | 6525 | size_t len = strlen (string); |
c0f9ea08 | 6526 | boolean want_reg_info = !preg->no_sub && nmatch > 0 && pmatch; |
fa9a63c5 RM |
6527 | |
6528 | private_preg = *preg; | |
5e69f11e | 6529 | |
fa9a63c5 RM |
6530 | private_preg.not_bol = !!(eflags & REG_NOTBOL); |
6531 | private_preg.not_eol = !!(eflags & REG_NOTEOL); | |
5e69f11e | 6532 | |
fa9a63c5 RM |
6533 | /* The user has told us exactly how many registers to return |
6534 | information about, via `nmatch'. We have to pass that on to the | |
b18215fc | 6535 | matching routines. */ |
fa9a63c5 | 6536 | private_preg.regs_allocated = REGS_FIXED; |
5e69f11e | 6537 | |
fa9a63c5 RM |
6538 | if (want_reg_info) |
6539 | { | |
6540 | regs.num_regs = nmatch; | |
4bb91c68 SM |
6541 | regs.start = TALLOC (nmatch * 2, regoff_t); |
6542 | if (regs.start == NULL) | |
d1dfb56c | 6543 | return REG_NOMATCH; |
4bb91c68 | 6544 | regs.end = regs.start + nmatch; |
fa9a63c5 RM |
6545 | } |
6546 | ||
c0f9ea08 SM |
6547 | /* Instead of using not_eol to implement REG_NOTEOL, we could simply |
6548 | pass (&private_preg, string, len + 1, 0, len, ...) pretending the string | |
6549 | was a little bit longer but still only matching the real part. | |
6550 | This works because the `endline' will check for a '\n' and will find a | |
6551 | '\0', correctly deciding that this is not the end of a line. | |
6552 | But it doesn't work out so nicely for REG_NOTBOL, since we don't have | |
6553 | a convenient '\0' there. For all we know, the string could be preceded | |
6554 | by '\n' which would throw things off. */ | |
6555 | ||
fa9a63c5 RM |
6556 | /* Perform the searching operation. */ |
6557 | ret = re_search (&private_preg, string, len, | |
0b32bf0e | 6558 | /* start: */ 0, /* range: */ len, |
7d652d97 | 6559 | want_reg_info ? ®s : 0); |
5e69f11e | 6560 | |
fa9a63c5 RM |
6561 | /* Copy the register information to the POSIX structure. */ |
6562 | if (want_reg_info) | |
6563 | { | |
6564 | if (ret >= 0) | |
0b32bf0e SM |
6565 | { |
6566 | unsigned r; | |
fa9a63c5 | 6567 | |
0b32bf0e SM |
6568 | for (r = 0; r < nmatch; r++) |
6569 | { | |
6570 | pmatch[r].rm_so = regs.start[r]; | |
6571 | pmatch[r].rm_eo = regs.end[r]; | |
6572 | } | |
6573 | } | |
fa9a63c5 | 6574 | |
b18215fc | 6575 | /* If we needed the temporary register info, free the space now. */ |
fa9a63c5 | 6576 | free (regs.start); |
fa9a63c5 RM |
6577 | } |
6578 | ||
6579 | /* We want zero return to mean success, unlike `re_search'. */ | |
d1dfb56c | 6580 | return ret >= 0 ? REG_NOERROR : REG_NOMATCH; |
fa9a63c5 | 6581 | } |
c0f9ea08 | 6582 | WEAK_ALIAS (__regexec, regexec) |
fa9a63c5 RM |
6583 | |
6584 | ||
ec869672 JR |
6585 | /* Returns a message corresponding to an error code, ERR_CODE, returned |
6586 | from either regcomp or regexec. We don't use PREG here. | |
6587 | ||
6588 | ERR_CODE was previously called ERRCODE, but that name causes an | |
6589 | error with msvc8 compiler. */ | |
fa9a63c5 RM |
6590 | |
6591 | size_t | |
d2762c86 | 6592 | regerror (int err_code, const regex_t *preg, char *errbuf, size_t errbuf_size) |
fa9a63c5 RM |
6593 | { |
6594 | const char *msg; | |
6595 | size_t msg_size; | |
6596 | ||
ec869672 JR |
6597 | if (err_code < 0 |
6598 | || err_code >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0]))) | |
5e69f11e | 6599 | /* Only error codes returned by the rest of the code should be passed |
b18215fc | 6600 | to this routine. If we are given anything else, or if other regex |
fa9a63c5 RM |
6601 | code generates an invalid error code, then the program has a bug. |
6602 | Dump core so we can fix it. */ | |
6603 | abort (); | |
6604 | ||
ec869672 | 6605 | msg = gettext (re_error_msgid[err_code]); |
fa9a63c5 RM |
6606 | |
6607 | msg_size = strlen (msg) + 1; /* Includes the null. */ | |
5e69f11e | 6608 | |
fa9a63c5 RM |
6609 | if (errbuf_size != 0) |
6610 | { | |
6611 | if (msg_size > errbuf_size) | |
0b32bf0e | 6612 | { |
e99a530f | 6613 | memcpy (errbuf, msg, errbuf_size - 1); |
0b32bf0e SM |
6614 | errbuf[errbuf_size - 1] = 0; |
6615 | } | |
fa9a63c5 | 6616 | else |
0b32bf0e | 6617 | strcpy (errbuf, msg); |
fa9a63c5 RM |
6618 | } |
6619 | ||
6620 | return msg_size; | |
6621 | } | |
c0f9ea08 | 6622 | WEAK_ALIAS (__regerror, regerror) |
fa9a63c5 RM |
6623 | |
6624 | ||
6625 | /* Free dynamically allocated space used by PREG. */ | |
6626 | ||
6627 | void | |
d2762c86 | 6628 | regfree (regex_t *preg) |
fa9a63c5 | 6629 | { |
c2cd06e6 | 6630 | free (preg->buffer); |
fa9a63c5 | 6631 | preg->buffer = NULL; |
5e69f11e | 6632 | |
fa9a63c5 RM |
6633 | preg->allocated = 0; |
6634 | preg->used = 0; | |
6635 | ||
c2cd06e6 | 6636 | free (preg->fastmap); |
fa9a63c5 RM |
6637 | preg->fastmap = NULL; |
6638 | preg->fastmap_accurate = 0; | |
6639 | ||
c2cd06e6 | 6640 | free (preg->translate); |
fa9a63c5 RM |
6641 | preg->translate = NULL; |
6642 | } | |
c0f9ea08 | 6643 | WEAK_ALIAS (__regfree, regfree) |
fa9a63c5 RM |
6644 | |
6645 | #endif /* not emacs */ |