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 | ||
ab422c4d | 5 | Copyright (C) 1993-2013 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. */ | |
908589fd | 36 | #if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 5) |
b8df54ff PE |
37 | # pragma GCC diagnostic ignored "-Wstrict-overflow" |
38 | # ifndef emacs | |
39 | # pragma GCC diagnostic ignored "-Wunused-but-set-variable" | |
40 | # pragma GCC diagnostic ignored "-Wunused-function" | |
41 | # pragma GCC diagnostic ignored "-Wunused-macros" | |
42 | # pragma GCC diagnostic ignored "-Wunused-result" | |
43 | # pragma GCC diagnostic ignored "-Wunused-variable" | |
44 | # endif | |
45 | #endif | |
46 | ||
cf38a720 | 47 | #include <config.h> |
fa9a63c5 | 48 | |
0e926e56 PE |
49 | #include <stddef.h> |
50 | ||
51 | #ifdef emacs | |
4bb91c68 SM |
52 | /* We need this for `regex.h', and perhaps for the Emacs include files. */ |
53 | # include <sys/types.h> | |
54 | #endif | |
fa9a63c5 | 55 | |
14473664 SM |
56 | /* Whether to use ISO C Amendment 1 wide char functions. |
57 | Those should not be used for Emacs since it uses its own. */ | |
5e5388f6 GM |
58 | #if defined _LIBC |
59 | #define WIDE_CHAR_SUPPORT 1 | |
60 | #else | |
14473664 | 61 | #define WIDE_CHAR_SUPPORT \ |
5e5388f6 GM |
62 | (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC && !emacs) |
63 | #endif | |
14473664 | 64 | |
fa463103 | 65 | /* For platform which support the ISO C amendment 1 functionality we |
14473664 | 66 | support user defined character classes. */ |
a0ad02f7 | 67 | #if WIDE_CHAR_SUPPORT |
14473664 SM |
68 | /* Solaris 2.5 has a bug: <wchar.h> must be included before <wctype.h>. */ |
69 | # include <wchar.h> | |
70 | # include <wctype.h> | |
71 | #endif | |
72 | ||
c0f9ea08 SM |
73 | #ifdef _LIBC |
74 | /* We have to keep the namespace clean. */ | |
75 | # define regfree(preg) __regfree (preg) | |
76 | # define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef) | |
77 | # define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags) | |
ec869672 | 78 | # define regerror(err_code, preg, errbuf, errbuf_size) \ |
5e617bc2 | 79 | __regerror (err_code, preg, errbuf, errbuf_size) |
c0f9ea08 SM |
80 | # define re_set_registers(bu, re, nu, st, en) \ |
81 | __re_set_registers (bu, re, nu, st, en) | |
82 | # define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \ | |
83 | __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) | |
84 | # define re_match(bufp, string, size, pos, regs) \ | |
85 | __re_match (bufp, string, size, pos, regs) | |
86 | # define re_search(bufp, string, size, startpos, range, regs) \ | |
87 | __re_search (bufp, string, size, startpos, range, regs) | |
88 | # define re_compile_pattern(pattern, length, bufp) \ | |
89 | __re_compile_pattern (pattern, length, bufp) | |
90 | # define re_set_syntax(syntax) __re_set_syntax (syntax) | |
91 | # define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \ | |
92 | __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop) | |
93 | # define re_compile_fastmap(bufp) __re_compile_fastmap (bufp) | |
94 | ||
14473664 SM |
95 | /* Make sure we call libc's function even if the user overrides them. */ |
96 | # define btowc __btowc | |
97 | # define iswctype __iswctype | |
98 | # define wctype __wctype | |
99 | ||
c0f9ea08 SM |
100 | # define WEAK_ALIAS(a,b) weak_alias (a, b) |
101 | ||
102 | /* We are also using some library internals. */ | |
103 | # include <locale/localeinfo.h> | |
104 | # include <locale/elem-hash.h> | |
105 | # include <langinfo.h> | |
106 | #else | |
107 | # define WEAK_ALIAS(a,b) | |
108 | #endif | |
109 | ||
4bb91c68 | 110 | /* This is for other GNU distributions with internationalized messages. */ |
0b32bf0e | 111 | #if HAVE_LIBINTL_H || defined _LIBC |
fa9a63c5 RM |
112 | # include <libintl.h> |
113 | #else | |
114 | # define gettext(msgid) (msgid) | |
115 | #endif | |
116 | ||
5e69f11e RM |
117 | #ifndef gettext_noop |
118 | /* This define is so xgettext can find the internationalizable | |
119 | strings. */ | |
0b32bf0e | 120 | # define gettext_noop(String) String |
5e69f11e RM |
121 | #endif |
122 | ||
fa9a63c5 RM |
123 | /* The `emacs' switch turns on certain matching commands |
124 | that make sense only in Emacs. */ | |
125 | #ifdef emacs | |
126 | ||
0b32bf0e | 127 | # include "lisp.h" |
e5560ff7 | 128 | # include "character.h" |
0b32bf0e | 129 | # include "buffer.h" |
b18215fc RS |
130 | |
131 | /* Make syntax table lookup grant data in gl_state. */ | |
0b32bf0e | 132 | # define SYNTAX_ENTRY_VIA_PROPERTY |
b18215fc | 133 | |
0b32bf0e | 134 | # include "syntax.h" |
0b32bf0e | 135 | # include "category.h" |
fa9a63c5 | 136 | |
7689ef0b EZ |
137 | # ifdef malloc |
138 | # undef malloc | |
139 | # endif | |
0b32bf0e | 140 | # define malloc xmalloc |
7689ef0b EZ |
141 | # ifdef realloc |
142 | # undef realloc | |
143 | # endif | |
0b32bf0e | 144 | # define realloc xrealloc |
7689ef0b EZ |
145 | # ifdef free |
146 | # undef free | |
147 | # endif | |
0b32bf0e | 148 | # define free xfree |
9abbd165 | 149 | |
7814e705 | 150 | /* Converts the pointer to the char to BEG-based offset from the start. */ |
0b32bf0e SM |
151 | # define PTR_TO_OFFSET(d) POS_AS_IN_BUFFER (POINTER_TO_OFFSET (d)) |
152 | # define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object))) | |
153 | ||
154 | # define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte) | |
bf216479 | 155 | # define RE_TARGET_MULTIBYTE_P(bufp) ((bufp)->target_multibyte) |
62a6e103 AS |
156 | # define RE_STRING_CHAR(p, multibyte) \ |
157 | (multibyte ? (STRING_CHAR (p)) : (*(p))) | |
158 | # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) \ | |
159 | (multibyte ? (STRING_CHAR_AND_LENGTH (p, len)) : ((len) = 1, *(p))) | |
2d1675e4 | 160 | |
4c0354d7 | 161 | # define RE_CHAR_TO_MULTIBYTE(c) UNIBYTE_TO_CHAR (c) |
cf9c99bc | 162 | |
2afc21f5 | 163 | # define RE_CHAR_TO_UNIBYTE(c) CHAR_TO_BYTE_SAFE (c) |
cf9c99bc | 164 | |
6fdd04b0 KH |
165 | /* Set C a (possibly converted to multibyte) character before P. P |
166 | points into a string which is the virtual concatenation of STR1 | |
167 | (which ends at END1) or STR2 (which ends at END2). */ | |
bf216479 KH |
168 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
169 | do { \ | |
02cb78b5 | 170 | if (target_multibyte) \ |
bf216479 KH |
171 | { \ |
172 | re_char *dtemp = (p) == (str2) ? (end1) : (p); \ | |
173 | re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \ | |
174 | while (dtemp-- > dlimit && !CHAR_HEAD_P (*dtemp)); \ | |
62a6e103 | 175 | c = STRING_CHAR (dtemp); \ |
bf216479 KH |
176 | } \ |
177 | else \ | |
178 | { \ | |
179 | (c = ((p) == (str2) ? (end1) : (p))[-1]); \ | |
cf9c99bc | 180 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
bf216479 | 181 | } \ |
2d1675e4 SM |
182 | } while (0) |
183 | ||
6fdd04b0 KH |
184 | /* Set C a (possibly converted to multibyte) character at P, and set |
185 | LEN to the byte length of that character. */ | |
186 | # define GET_CHAR_AFTER(c, p, len) \ | |
187 | do { \ | |
02cb78b5 | 188 | if (target_multibyte) \ |
62a6e103 | 189 | (c) = STRING_CHAR_AND_LENGTH (p, len); \ |
6fdd04b0 KH |
190 | else \ |
191 | { \ | |
cf9c99bc | 192 | (c) = *p; \ |
6fdd04b0 | 193 | len = 1; \ |
cf9c99bc | 194 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
6fdd04b0 | 195 | } \ |
8f924df7 | 196 | } while (0) |
4e8a9132 | 197 | |
fa9a63c5 RM |
198 | #else /* not emacs */ |
199 | ||
200 | /* If we are not linking with Emacs proper, | |
201 | we can't use the relocating allocator | |
202 | even if config.h says that we can. */ | |
0b32bf0e | 203 | # undef REL_ALLOC |
fa9a63c5 | 204 | |
4004364e | 205 | # include <unistd.h> |
fa9a63c5 | 206 | |
a77f947b CY |
207 | /* When used in Emacs's lib-src, we need xmalloc and xrealloc. */ |
208 | ||
b8df54ff | 209 | static void * |
d2762c86 | 210 | xmalloc (size_t size) |
a77f947b | 211 | { |
38182d90 | 212 | void *val = malloc (size); |
a77f947b CY |
213 | if (!val && size) |
214 | { | |
215 | write (2, "virtual memory exhausted\n", 25); | |
216 | exit (1); | |
217 | } | |
218 | return val; | |
219 | } | |
220 | ||
b8df54ff | 221 | static void * |
d2762c86 | 222 | xrealloc (void *block, size_t size) |
a77f947b | 223 | { |
38182d90 | 224 | void *val; |
a77f947b CY |
225 | /* We must call malloc explicitly when BLOCK is 0, since some |
226 | reallocs don't do this. */ | |
227 | if (! block) | |
38182d90 | 228 | val = malloc (size); |
a77f947b | 229 | else |
38182d90 | 230 | val = realloc (block, size); |
a77f947b CY |
231 | if (!val && size) |
232 | { | |
233 | write (2, "virtual memory exhausted\n", 25); | |
234 | exit (1); | |
235 | } | |
236 | return val; | |
237 | } | |
238 | ||
a073faa6 CY |
239 | # ifdef malloc |
240 | # undef malloc | |
241 | # endif | |
242 | # define malloc xmalloc | |
243 | # ifdef realloc | |
244 | # undef realloc | |
245 | # endif | |
246 | # define realloc xrealloc | |
247 | ||
f5d9e83a | 248 | # include <stdbool.h> |
9cfdb3ec | 249 | # include <string.h> |
fa9a63c5 RM |
250 | |
251 | /* Define the syntax stuff for \<, \>, etc. */ | |
252 | ||
990b2375 | 253 | /* Sword must be nonzero for the wordchar pattern commands in re_match_2. */ |
669fa600 | 254 | enum syntaxcode { Swhitespace = 0, Sword = 1, Ssymbol = 2 }; |
fa9a63c5 | 255 | |
e934739e | 256 | /* Dummy macros for non-Emacs environments. */ |
0b32bf0e SM |
257 | # define CHAR_CHARSET(c) 0 |
258 | # define CHARSET_LEADING_CODE_BASE(c) 0 | |
259 | # define MAX_MULTIBYTE_LENGTH 1 | |
260 | # define RE_MULTIBYTE_P(x) 0 | |
bf216479 | 261 | # define RE_TARGET_MULTIBYTE_P(x) 0 |
0b32bf0e SM |
262 | # define WORD_BOUNDARY_P(c1, c2) (0) |
263 | # define CHAR_HEAD_P(p) (1) | |
264 | # define SINGLE_BYTE_CHAR_P(c) (1) | |
265 | # define SAME_CHARSET_P(c1, c2) (1) | |
aa3830c4 | 266 | # define BYTES_BY_CHAR_HEAD(p) (1) |
70806df6 | 267 | # define PREV_CHAR_BOUNDARY(p, limit) ((p)--) |
62a6e103 AS |
268 | # define STRING_CHAR(p) (*(p)) |
269 | # define RE_STRING_CHAR(p, multibyte) STRING_CHAR (p) | |
0b32bf0e | 270 | # define CHAR_STRING(c, s) (*(s) = (c), 1) |
62a6e103 AS |
271 | # define STRING_CHAR_AND_LENGTH(p, actual_len) ((actual_len) = 1, *(p)) |
272 | # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) STRING_CHAR_AND_LENGTH (p, len) | |
cf9c99bc KH |
273 | # define RE_CHAR_TO_MULTIBYTE(c) (c) |
274 | # define RE_CHAR_TO_UNIBYTE(c) (c) | |
0b32bf0e | 275 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
b18215fc | 276 | (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1))) |
6fdd04b0 KH |
277 | # define GET_CHAR_AFTER(c, p, len) \ |
278 | (c = *p, len = 1) | |
0b32bf0e | 279 | # define MAKE_CHAR(charset, c1, c2) (c1) |
9117d724 KH |
280 | # define BYTE8_TO_CHAR(c) (c) |
281 | # define CHAR_BYTE8_P(c) (0) | |
bf216479 | 282 | # define CHAR_LEADING_CODE(c) (c) |
8f924df7 | 283 | |
fa9a63c5 | 284 | #endif /* not emacs */ |
4e8a9132 SM |
285 | |
286 | #ifndef RE_TRANSLATE | |
0b32bf0e SM |
287 | # define RE_TRANSLATE(TBL, C) ((unsigned char)(TBL)[C]) |
288 | # define RE_TRANSLATE_P(TBL) (TBL) | |
4e8a9132 | 289 | #endif |
fa9a63c5 RM |
290 | \f |
291 | /* Get the interface, including the syntax bits. */ | |
292 | #include "regex.h" | |
293 | ||
f71b19b6 DL |
294 | /* isalpha etc. are used for the character classes. */ |
295 | #include <ctype.h> | |
fa9a63c5 | 296 | |
f71b19b6 | 297 | #ifdef emacs |
fa9a63c5 | 298 | |
f71b19b6 | 299 | /* 1 if C is an ASCII character. */ |
0b32bf0e | 300 | # define IS_REAL_ASCII(c) ((c) < 0200) |
fa9a63c5 | 301 | |
f71b19b6 | 302 | /* 1 if C is a unibyte character. */ |
0b32bf0e | 303 | # define ISUNIBYTE(c) (SINGLE_BYTE_CHAR_P ((c))) |
96cc36cc | 304 | |
f71b19b6 | 305 | /* The Emacs definitions should not be directly affected by locales. */ |
96cc36cc | 306 | |
f71b19b6 | 307 | /* In Emacs, these are only used for single-byte characters. */ |
0b32bf0e SM |
308 | # define ISDIGIT(c) ((c) >= '0' && (c) <= '9') |
309 | # define ISCNTRL(c) ((c) < ' ') | |
310 | # define ISXDIGIT(c) (((c) >= '0' && (c) <= '9') \ | |
f71b19b6 DL |
311 | || ((c) >= 'a' && (c) <= 'f') \ |
312 | || ((c) >= 'A' && (c) <= 'F')) | |
96cc36cc RS |
313 | |
314 | /* This is only used for single-byte characters. */ | |
0b32bf0e | 315 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') |
96cc36cc RS |
316 | |
317 | /* The rest must handle multibyte characters. */ | |
318 | ||
0b32bf0e | 319 | # define ISGRAPH(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 320 | ? (c) > ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
321 | : 1) |
322 | ||
14473664 | 323 | # define ISPRINT(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 324 | ? (c) >= ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
325 | : 1) |
326 | ||
0b32bf0e | 327 | # define ISALNUM(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
328 | ? (((c) >= 'a' && (c) <= 'z') \ |
329 | || ((c) >= 'A' && (c) <= 'Z') \ | |
330 | || ((c) >= '0' && (c) <= '9')) \ | |
96cc36cc RS |
331 | : SYNTAX (c) == Sword) |
332 | ||
0b32bf0e | 333 | # define ISALPHA(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
334 | ? (((c) >= 'a' && (c) <= 'z') \ |
335 | || ((c) >= 'A' && (c) <= 'Z')) \ | |
96cc36cc RS |
336 | : SYNTAX (c) == Sword) |
337 | ||
5da9919f | 338 | # define ISLOWER(c) lowercasep (c) |
96cc36cc | 339 | |
0b32bf0e | 340 | # define ISPUNCT(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
341 | ? ((c) > ' ' && (c) < 0177 \ |
342 | && !(((c) >= 'a' && (c) <= 'z') \ | |
4bb91c68 SM |
343 | || ((c) >= 'A' && (c) <= 'Z') \ |
344 | || ((c) >= '0' && (c) <= '9'))) \ | |
96cc36cc RS |
345 | : SYNTAX (c) != Sword) |
346 | ||
0b32bf0e | 347 | # define ISSPACE(c) (SYNTAX (c) == Swhitespace) |
96cc36cc | 348 | |
5da9919f | 349 | # define ISUPPER(c) uppercasep (c) |
96cc36cc | 350 | |
0b32bf0e | 351 | # define ISWORD(c) (SYNTAX (c) == Sword) |
96cc36cc RS |
352 | |
353 | #else /* not emacs */ | |
354 | ||
f71b19b6 | 355 | /* 1 if C is an ASCII character. */ |
0b32bf0e | 356 | # define IS_REAL_ASCII(c) ((c) < 0200) |
f71b19b6 DL |
357 | |
358 | /* This distinction is not meaningful, except in Emacs. */ | |
0b32bf0e SM |
359 | # define ISUNIBYTE(c) 1 |
360 | ||
361 | # ifdef isblank | |
0e926e56 | 362 | # define ISBLANK(c) isblank (c) |
0b32bf0e SM |
363 | # else |
364 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') | |
365 | # endif | |
366 | # ifdef isgraph | |
0e926e56 | 367 | # define ISGRAPH(c) isgraph (c) |
0b32bf0e | 368 | # else |
0e926e56 | 369 | # define ISGRAPH(c) (isprint (c) && !isspace (c)) |
0b32bf0e SM |
370 | # endif |
371 | ||
0e926e56 | 372 | /* Solaris defines ISPRINT so we must undefine it first. */ |
4bb91c68 | 373 | # undef ISPRINT |
0e926e56 PE |
374 | # define ISPRINT(c) isprint (c) |
375 | # define ISDIGIT(c) isdigit (c) | |
376 | # define ISALNUM(c) isalnum (c) | |
377 | # define ISALPHA(c) isalpha (c) | |
378 | # define ISCNTRL(c) iscntrl (c) | |
379 | # define ISLOWER(c) islower (c) | |
380 | # define ISPUNCT(c) ispunct (c) | |
381 | # define ISSPACE(c) isspace (c) | |
382 | # define ISUPPER(c) isupper (c) | |
383 | # define ISXDIGIT(c) isxdigit (c) | |
0b32bf0e | 384 | |
5e617bc2 | 385 | # define ISWORD(c) ISALPHA (c) |
0b32bf0e | 386 | |
4bb91c68 | 387 | # ifdef _tolower |
5e617bc2 | 388 | # define TOLOWER(c) _tolower (c) |
4bb91c68 | 389 | # else |
5e617bc2 | 390 | # define TOLOWER(c) tolower (c) |
4bb91c68 SM |
391 | # endif |
392 | ||
393 | /* How many characters in the character set. */ | |
394 | # define CHAR_SET_SIZE 256 | |
395 | ||
0b32bf0e | 396 | # ifdef SYNTAX_TABLE |
f71b19b6 | 397 | |
0b32bf0e | 398 | extern char *re_syntax_table; |
f71b19b6 | 399 | |
0b32bf0e SM |
400 | # else /* not SYNTAX_TABLE */ |
401 | ||
0b32bf0e SM |
402 | static char re_syntax_table[CHAR_SET_SIZE]; |
403 | ||
404 | static void | |
d2762c86 | 405 | init_syntax_once (void) |
0b32bf0e SM |
406 | { |
407 | register int c; | |
408 | static int done = 0; | |
409 | ||
410 | if (done) | |
411 | return; | |
412 | ||
72af86bd | 413 | memset (re_syntax_table, 0, sizeof re_syntax_table); |
0b32bf0e | 414 | |
4bb91c68 SM |
415 | for (c = 0; c < CHAR_SET_SIZE; ++c) |
416 | if (ISALNUM (c)) | |
417 | re_syntax_table[c] = Sword; | |
fa9a63c5 | 418 | |
669fa600 | 419 | re_syntax_table['_'] = Ssymbol; |
fa9a63c5 | 420 | |
0b32bf0e SM |
421 | done = 1; |
422 | } | |
423 | ||
424 | # endif /* not SYNTAX_TABLE */ | |
96cc36cc | 425 | |
4bb91c68 SM |
426 | # define SYNTAX(c) re_syntax_table[(c)] |
427 | ||
96cc36cc RS |
428 | #endif /* not emacs */ |
429 | \f | |
261cb4bb | 430 | #define SIGN_EXTEND_CHAR(c) ((signed char) (c)) |
fa9a63c5 RM |
431 | \f |
432 | /* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we | |
433 | use `alloca' instead of `malloc'. This is because using malloc in | |
434 | re_search* or re_match* could cause memory leaks when C-g is used in | |
435 | Emacs; also, malloc is slower and causes storage fragmentation. On | |
5e69f11e RM |
436 | the other hand, malloc is more portable, and easier to debug. |
437 | ||
fa9a63c5 RM |
438 | Because we sometimes use alloca, some routines have to be macros, |
439 | not functions -- `alloca'-allocated space disappears at the end of the | |
440 | function it is called in. */ | |
441 | ||
442 | #ifdef REGEX_MALLOC | |
443 | ||
0b32bf0e SM |
444 | # define REGEX_ALLOCATE malloc |
445 | # define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize) | |
446 | # define REGEX_FREE free | |
fa9a63c5 RM |
447 | |
448 | #else /* not REGEX_MALLOC */ | |
449 | ||
450 | /* Emacs already defines alloca, sometimes. */ | |
0b32bf0e | 451 | # ifndef alloca |
fa9a63c5 RM |
452 | |
453 | /* Make alloca work the best possible way. */ | |
0b32bf0e SM |
454 | # ifdef __GNUC__ |
455 | # define alloca __builtin_alloca | |
456 | # else /* not __GNUC__ */ | |
7f585e7a | 457 | # ifdef HAVE_ALLOCA_H |
0b32bf0e SM |
458 | # include <alloca.h> |
459 | # endif /* HAVE_ALLOCA_H */ | |
460 | # endif /* not __GNUC__ */ | |
fa9a63c5 | 461 | |
0b32bf0e | 462 | # endif /* not alloca */ |
fa9a63c5 | 463 | |
0b32bf0e | 464 | # define REGEX_ALLOCATE alloca |
fa9a63c5 RM |
465 | |
466 | /* Assumes a `char *destination' variable. */ | |
0b32bf0e | 467 | # define REGEX_REALLOCATE(source, osize, nsize) \ |
fa9a63c5 | 468 | (destination = (char *) alloca (nsize), \ |
4bb91c68 | 469 | memcpy (destination, source, osize)) |
fa9a63c5 RM |
470 | |
471 | /* No need to do anything to free, after alloca. */ | |
0b32bf0e | 472 | # define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
473 | |
474 | #endif /* not REGEX_MALLOC */ | |
475 | ||
476 | /* Define how to allocate the failure stack. */ | |
477 | ||
0b32bf0e | 478 | #if defined REL_ALLOC && defined REGEX_MALLOC |
4297555e | 479 | |
0b32bf0e | 480 | # define REGEX_ALLOCATE_STACK(size) \ |
fa9a63c5 | 481 | r_alloc (&failure_stack_ptr, (size)) |
0b32bf0e | 482 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 483 | r_re_alloc (&failure_stack_ptr, (nsize)) |
0b32bf0e | 484 | # define REGEX_FREE_STACK(ptr) \ |
fa9a63c5 RM |
485 | r_alloc_free (&failure_stack_ptr) |
486 | ||
4297555e | 487 | #else /* not using relocating allocator */ |
fa9a63c5 | 488 | |
0b32bf0e | 489 | # ifdef REGEX_MALLOC |
fa9a63c5 | 490 | |
0b32bf0e SM |
491 | # define REGEX_ALLOCATE_STACK malloc |
492 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize) | |
493 | # define REGEX_FREE_STACK free | |
fa9a63c5 | 494 | |
0b32bf0e | 495 | # else /* not REGEX_MALLOC */ |
fa9a63c5 | 496 | |
0b32bf0e | 497 | # define REGEX_ALLOCATE_STACK alloca |
fa9a63c5 | 498 | |
0b32bf0e | 499 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 500 | REGEX_REALLOCATE (source, osize, nsize) |
7814e705 | 501 | /* No need to explicitly free anything. */ |
0b32bf0e | 502 | # define REGEX_FREE_STACK(arg) ((void)0) |
fa9a63c5 | 503 | |
0b32bf0e | 504 | # endif /* not REGEX_MALLOC */ |
4297555e | 505 | #endif /* not using relocating allocator */ |
fa9a63c5 RM |
506 | |
507 | ||
508 | /* True if `size1' is non-NULL and PTR is pointing anywhere inside | |
509 | `string1' or just past its end. This works if PTR is NULL, which is | |
510 | a good thing. */ | |
25fe55af | 511 | #define FIRST_STRING_P(ptr) \ |
fa9a63c5 RM |
512 | (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) |
513 | ||
514 | /* (Re)Allocate N items of type T using malloc, or fail. */ | |
515 | #define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t))) | |
516 | #define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t))) | |
fa9a63c5 RM |
517 | #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) |
518 | ||
4bb91c68 | 519 | #define BYTEWIDTH 8 /* In bits. */ |
fa9a63c5 RM |
520 | |
521 | #define STREQ(s1, s2) ((strcmp (s1, s2) == 0)) | |
522 | ||
523 | #undef MAX | |
524 | #undef MIN | |
525 | #define MAX(a, b) ((a) > (b) ? (a) : (b)) | |
526 | #define MIN(a, b) ((a) < (b) ? (a) : (b)) | |
527 | ||
66f0296e | 528 | /* Type of source-pattern and string chars. */ |
a6fc3b5c EZ |
529 | #ifdef _MSC_VER |
530 | typedef unsigned char re_char; | |
531 | #else | |
66f0296e | 532 | typedef const unsigned char re_char; |
a6fc3b5c | 533 | #endif |
66f0296e | 534 | |
fa9a63c5 | 535 | typedef char boolean; |
fa9a63c5 | 536 | |
261cb4bb PE |
537 | static regoff_t re_match_2_internal (struct re_pattern_buffer *bufp, |
538 | re_char *string1, size_t size1, | |
539 | re_char *string2, size_t size2, | |
540 | ssize_t pos, | |
541 | struct re_registers *regs, | |
542 | ssize_t stop); | |
fa9a63c5 RM |
543 | \f |
544 | /* These are the command codes that appear in compiled regular | |
4bb91c68 | 545 | expressions. Some opcodes are followed by argument bytes. A |
fa9a63c5 RM |
546 | command code can specify any interpretation whatsoever for its |
547 | arguments. Zero bytes may appear in the compiled regular expression. */ | |
548 | ||
549 | typedef enum | |
550 | { | |
551 | no_op = 0, | |
552 | ||
4bb91c68 | 553 | /* Succeed right away--no more backtracking. */ |
fa9a63c5 RM |
554 | succeed, |
555 | ||
25fe55af | 556 | /* Followed by one byte giving n, then by n literal bytes. */ |
fa9a63c5 RM |
557 | exactn, |
558 | ||
25fe55af | 559 | /* Matches any (more or less) character. */ |
fa9a63c5 RM |
560 | anychar, |
561 | ||
25fe55af RS |
562 | /* Matches any one char belonging to specified set. First |
563 | following byte is number of bitmap bytes. Then come bytes | |
564 | for a bitmap saying which chars are in. Bits in each byte | |
565 | are ordered low-bit-first. A character is in the set if its | |
566 | bit is 1. A character too large to have a bit in the map is | |
96cc36cc RS |
567 | automatically not in the set. |
568 | ||
569 | If the length byte has the 0x80 bit set, then that stuff | |
570 | is followed by a range table: | |
571 | 2 bytes of flags for character sets (low 8 bits, high 8 bits) | |
0b32bf0e | 572 | See RANGE_TABLE_WORK_BITS below. |
01618498 | 573 | 2 bytes, the number of pairs that follow (upto 32767) |
96cc36cc | 574 | pairs, each 2 multibyte characters, |
0b32bf0e | 575 | each multibyte character represented as 3 bytes. */ |
fa9a63c5 RM |
576 | charset, |
577 | ||
25fe55af | 578 | /* Same parameters as charset, but match any character that is |
4bb91c68 | 579 | not one of those specified. */ |
fa9a63c5 RM |
580 | charset_not, |
581 | ||
25fe55af RS |
582 | /* Start remembering the text that is matched, for storing in a |
583 | register. Followed by one byte with the register number, in | |
584 | the range 0 to one less than the pattern buffer's re_nsub | |
505bde11 | 585 | field. */ |
fa9a63c5 RM |
586 | start_memory, |
587 | ||
25fe55af RS |
588 | /* Stop remembering the text that is matched and store it in a |
589 | memory register. Followed by one byte with the register | |
590 | number, in the range 0 to one less than `re_nsub' in the | |
505bde11 | 591 | pattern buffer. */ |
fa9a63c5 RM |
592 | stop_memory, |
593 | ||
25fe55af | 594 | /* Match a duplicate of something remembered. Followed by one |
4bb91c68 | 595 | byte containing the register number. */ |
fa9a63c5 RM |
596 | duplicate, |
597 | ||
25fe55af | 598 | /* Fail unless at beginning of line. */ |
fa9a63c5 RM |
599 | begline, |
600 | ||
4bb91c68 | 601 | /* Fail unless at end of line. */ |
fa9a63c5 RM |
602 | endline, |
603 | ||
25fe55af RS |
604 | /* Succeeds if at beginning of buffer (if emacs) or at beginning |
605 | of string to be matched (if not). */ | |
fa9a63c5 RM |
606 | begbuf, |
607 | ||
25fe55af | 608 | /* Analogously, for end of buffer/string. */ |
fa9a63c5 | 609 | endbuf, |
5e69f11e | 610 | |
25fe55af | 611 | /* Followed by two byte relative address to which to jump. */ |
5e69f11e | 612 | jump, |
fa9a63c5 | 613 | |
25fe55af | 614 | /* Followed by two-byte relative address of place to resume at |
7814e705 | 615 | in case of failure. */ |
fa9a63c5 | 616 | on_failure_jump, |
5e69f11e | 617 | |
25fe55af RS |
618 | /* Like on_failure_jump, but pushes a placeholder instead of the |
619 | current string position when executed. */ | |
fa9a63c5 | 620 | on_failure_keep_string_jump, |
5e69f11e | 621 | |
505bde11 SM |
622 | /* Just like `on_failure_jump', except that it checks that we |
623 | don't get stuck in an infinite loop (matching an empty string | |
624 | indefinitely). */ | |
625 | on_failure_jump_loop, | |
626 | ||
0683b6fa SM |
627 | /* Just like `on_failure_jump_loop', except that it checks for |
628 | a different kind of loop (the kind that shows up with non-greedy | |
629 | operators). This operation has to be immediately preceded | |
630 | by a `no_op'. */ | |
631 | on_failure_jump_nastyloop, | |
632 | ||
0b32bf0e | 633 | /* A smart `on_failure_jump' used for greedy * and + operators. |
c7015153 | 634 | It analyzes the loop before which it is put and if the |
505bde11 | 635 | loop does not require backtracking, it changes itself to |
4e8a9132 SM |
636 | `on_failure_keep_string_jump' and short-circuits the loop, |
637 | else it just defaults to changing itself into `on_failure_jump'. | |
638 | It assumes that it is pointing to just past a `jump'. */ | |
505bde11 | 639 | on_failure_jump_smart, |
fa9a63c5 | 640 | |
25fe55af | 641 | /* Followed by two-byte relative address and two-byte number n. |
ed0767d8 SM |
642 | After matching N times, jump to the address upon failure. |
643 | Does not work if N starts at 0: use on_failure_jump_loop | |
644 | instead. */ | |
fa9a63c5 RM |
645 | succeed_n, |
646 | ||
25fe55af RS |
647 | /* Followed by two-byte relative address, and two-byte number n. |
648 | Jump to the address N times, then fail. */ | |
fa9a63c5 RM |
649 | jump_n, |
650 | ||
25fe55af | 651 | /* Set the following two-byte relative address to the |
7814e705 | 652 | subsequent two-byte number. The address *includes* the two |
25fe55af | 653 | bytes of number. */ |
fa9a63c5 RM |
654 | set_number_at, |
655 | ||
fa9a63c5 RM |
656 | wordbeg, /* Succeeds if at word beginning. */ |
657 | wordend, /* Succeeds if at word end. */ | |
658 | ||
659 | wordbound, /* Succeeds if at a word boundary. */ | |
7814e705 | 660 | notwordbound, /* Succeeds if not at a word boundary. */ |
fa9a63c5 | 661 | |
669fa600 SM |
662 | symbeg, /* Succeeds if at symbol beginning. */ |
663 | symend, /* Succeeds if at symbol end. */ | |
664 | ||
fa9a63c5 | 665 | /* Matches any character whose syntax is specified. Followed by |
25fe55af | 666 | a byte which contains a syntax code, e.g., Sword. */ |
fa9a63c5 RM |
667 | syntaxspec, |
668 | ||
669 | /* Matches any character whose syntax is not that specified. */ | |
1fb352e0 SM |
670 | notsyntaxspec |
671 | ||
672 | #ifdef emacs | |
673 | ,before_dot, /* Succeeds if before point. */ | |
674 | at_dot, /* Succeeds if at point. */ | |
675 | after_dot, /* Succeeds if after point. */ | |
b18215fc RS |
676 | |
677 | /* Matches any character whose category-set contains the specified | |
7814e705 JB |
678 | category. The operator is followed by a byte which contains a |
679 | category code (mnemonic ASCII character). */ | |
b18215fc RS |
680 | categoryspec, |
681 | ||
682 | /* Matches any character whose category-set does not contain the | |
683 | specified category. The operator is followed by a byte which | |
684 | contains the category code (mnemonic ASCII character). */ | |
685 | notcategoryspec | |
fa9a63c5 RM |
686 | #endif /* emacs */ |
687 | } re_opcode_t; | |
688 | \f | |
689 | /* Common operations on the compiled pattern. */ | |
690 | ||
691 | /* Store NUMBER in two contiguous bytes starting at DESTINATION. */ | |
692 | ||
693 | #define STORE_NUMBER(destination, number) \ | |
694 | do { \ | |
695 | (destination)[0] = (number) & 0377; \ | |
696 | (destination)[1] = (number) >> 8; \ | |
697 | } while (0) | |
698 | ||
699 | /* Same as STORE_NUMBER, except increment DESTINATION to | |
700 | the byte after where the number is stored. Therefore, DESTINATION | |
701 | must be an lvalue. */ | |
702 | ||
703 | #define STORE_NUMBER_AND_INCR(destination, number) \ | |
704 | do { \ | |
705 | STORE_NUMBER (destination, number); \ | |
706 | (destination) += 2; \ | |
707 | } while (0) | |
708 | ||
709 | /* Put into DESTINATION a number stored in two contiguous bytes starting | |
710 | at SOURCE. */ | |
711 | ||
712 | #define EXTRACT_NUMBER(destination, source) \ | |
713 | do { \ | |
714 | (destination) = *(source) & 0377; \ | |
715 | (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \ | |
716 | } while (0) | |
717 | ||
718 | #ifdef DEBUG | |
719 | static void | |
261cb4bb | 720 | extract_number (int *dest, re_char *source) |
fa9a63c5 | 721 | { |
5e69f11e | 722 | int temp = SIGN_EXTEND_CHAR (*(source + 1)); |
fa9a63c5 RM |
723 | *dest = *source & 0377; |
724 | *dest += temp << 8; | |
725 | } | |
726 | ||
4bb91c68 | 727 | # ifndef EXTRACT_MACROS /* To debug the macros. */ |
0b32bf0e SM |
728 | # undef EXTRACT_NUMBER |
729 | # define EXTRACT_NUMBER(dest, src) extract_number (&dest, src) | |
730 | # endif /* not EXTRACT_MACROS */ | |
fa9a63c5 RM |
731 | |
732 | #endif /* DEBUG */ | |
733 | ||
734 | /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. | |
735 | SOURCE must be an lvalue. */ | |
736 | ||
737 | #define EXTRACT_NUMBER_AND_INCR(destination, source) \ | |
738 | do { \ | |
739 | EXTRACT_NUMBER (destination, source); \ | |
25fe55af | 740 | (source) += 2; \ |
fa9a63c5 RM |
741 | } while (0) |
742 | ||
743 | #ifdef DEBUG | |
744 | static void | |
261cb4bb | 745 | extract_number_and_incr (int *destination, re_char **source) |
5e69f11e | 746 | { |
fa9a63c5 RM |
747 | extract_number (destination, *source); |
748 | *source += 2; | |
749 | } | |
750 | ||
0b32bf0e SM |
751 | # ifndef EXTRACT_MACROS |
752 | # undef EXTRACT_NUMBER_AND_INCR | |
753 | # define EXTRACT_NUMBER_AND_INCR(dest, src) \ | |
fa9a63c5 | 754 | extract_number_and_incr (&dest, &src) |
0b32bf0e | 755 | # endif /* not EXTRACT_MACROS */ |
fa9a63c5 RM |
756 | |
757 | #endif /* DEBUG */ | |
758 | \f | |
b18215fc RS |
759 | /* Store a multibyte character in three contiguous bytes starting |
760 | DESTINATION, and increment DESTINATION to the byte after where the | |
7814e705 | 761 | character is stored. Therefore, DESTINATION must be an lvalue. */ |
b18215fc RS |
762 | |
763 | #define STORE_CHARACTER_AND_INCR(destination, character) \ | |
764 | do { \ | |
765 | (destination)[0] = (character) & 0377; \ | |
766 | (destination)[1] = ((character) >> 8) & 0377; \ | |
767 | (destination)[2] = (character) >> 16; \ | |
768 | (destination) += 3; \ | |
769 | } while (0) | |
770 | ||
771 | /* Put into DESTINATION a character stored in three contiguous bytes | |
7814e705 | 772 | starting at SOURCE. */ |
b18215fc RS |
773 | |
774 | #define EXTRACT_CHARACTER(destination, source) \ | |
775 | do { \ | |
776 | (destination) = ((source)[0] \ | |
777 | | ((source)[1] << 8) \ | |
778 | | ((source)[2] << 16)); \ | |
779 | } while (0) | |
780 | ||
781 | ||
782 | /* Macros for charset. */ | |
783 | ||
784 | /* Size of bitmap of charset P in bytes. P is a start of charset, | |
785 | i.e. *P is (re_opcode_t) charset or (re_opcode_t) charset_not. */ | |
786 | #define CHARSET_BITMAP_SIZE(p) ((p)[1] & 0x7F) | |
787 | ||
788 | /* Nonzero if charset P has range table. */ | |
25fe55af | 789 | #define CHARSET_RANGE_TABLE_EXISTS_P(p) ((p)[1] & 0x80) |
b18215fc RS |
790 | |
791 | /* Return the address of range table of charset P. But not the start | |
792 | of table itself, but the before where the number of ranges is | |
96cc36cc RS |
793 | stored. `2 +' means to skip re_opcode_t and size of bitmap, |
794 | and the 2 bytes of flags at the start of the range table. */ | |
795 | #define CHARSET_RANGE_TABLE(p) (&(p)[4 + CHARSET_BITMAP_SIZE (p)]) | |
796 | ||
797 | /* Extract the bit flags that start a range table. */ | |
798 | #define CHARSET_RANGE_TABLE_BITS(p) \ | |
799 | ((p)[2 + CHARSET_BITMAP_SIZE (p)] \ | |
800 | + (p)[3 + CHARSET_BITMAP_SIZE (p)] * 0x100) | |
b18215fc | 801 | |
b18215fc | 802 | /* Return the address of end of RANGE_TABLE. COUNT is number of |
7814e705 JB |
803 | ranges (which is a pair of (start, end)) in the RANGE_TABLE. `* 2' |
804 | is start of range and end of range. `* 3' is size of each start | |
b18215fc RS |
805 | and end. */ |
806 | #define CHARSET_RANGE_TABLE_END(range_table, count) \ | |
807 | ((range_table) + (count) * 2 * 3) | |
808 | ||
7814e705 | 809 | /* Test if C is in RANGE_TABLE. A flag NOT is negated if C is in. |
b18215fc RS |
810 | COUNT is number of ranges in RANGE_TABLE. */ |
811 | #define CHARSET_LOOKUP_RANGE_TABLE_RAW(not, c, range_table, count) \ | |
812 | do \ | |
813 | { \ | |
01618498 | 814 | re_wchar_t range_start, range_end; \ |
19ed5445 | 815 | re_char *rtp; \ |
01618498 | 816 | re_char *range_table_end \ |
b18215fc RS |
817 | = CHARSET_RANGE_TABLE_END ((range_table), (count)); \ |
818 | \ | |
19ed5445 | 819 | for (rtp = (range_table); rtp < range_table_end; rtp += 2 * 3) \ |
b18215fc | 820 | { \ |
19ed5445 PE |
821 | EXTRACT_CHARACTER (range_start, rtp); \ |
822 | EXTRACT_CHARACTER (range_end, rtp + 3); \ | |
b18215fc RS |
823 | \ |
824 | if (range_start <= (c) && (c) <= range_end) \ | |
825 | { \ | |
826 | (not) = !(not); \ | |
827 | break; \ | |
828 | } \ | |
829 | } \ | |
830 | } \ | |
831 | while (0) | |
832 | ||
833 | /* Test if C is in range table of CHARSET. The flag NOT is negated if | |
834 | C is listed in it. */ | |
835 | #define CHARSET_LOOKUP_RANGE_TABLE(not, c, charset) \ | |
836 | do \ | |
837 | { \ | |
838 | /* Number of ranges in range table. */ \ | |
839 | int count; \ | |
01618498 SM |
840 | re_char *range_table = CHARSET_RANGE_TABLE (charset); \ |
841 | \ | |
b18215fc RS |
842 | EXTRACT_NUMBER_AND_INCR (count, range_table); \ |
843 | CHARSET_LOOKUP_RANGE_TABLE_RAW ((not), (c), range_table, count); \ | |
844 | } \ | |
845 | while (0) | |
846 | \f | |
fa9a63c5 RM |
847 | /* If DEBUG is defined, Regex prints many voluminous messages about what |
848 | it is doing (if the variable `debug' is nonzero). If linked with the | |
849 | main program in `iregex.c', you can enter patterns and strings | |
850 | interactively. And if linked with the main program in `main.c' and | |
4bb91c68 | 851 | the other test files, you can run the already-written tests. */ |
fa9a63c5 RM |
852 | |
853 | #ifdef DEBUG | |
854 | ||
855 | /* We use standard I/O for debugging. */ | |
0b32bf0e | 856 | # include <stdio.h> |
fa9a63c5 RM |
857 | |
858 | /* It is useful to test things that ``must'' be true when debugging. */ | |
0b32bf0e | 859 | # include <assert.h> |
fa9a63c5 | 860 | |
99633e97 | 861 | static int debug = -100000; |
fa9a63c5 | 862 | |
0b32bf0e SM |
863 | # define DEBUG_STATEMENT(e) e |
864 | # define DEBUG_PRINT1(x) if (debug > 0) printf (x) | |
865 | # define DEBUG_PRINT2(x1, x2) if (debug > 0) printf (x1, x2) | |
866 | # define DEBUG_PRINT3(x1, x2, x3) if (debug > 0) printf (x1, x2, x3) | |
867 | # define DEBUG_PRINT4(x1, x2, x3, x4) if (debug > 0) printf (x1, x2, x3, x4) | |
868 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ | |
99633e97 | 869 | if (debug > 0) print_partial_compiled_pattern (s, e) |
0b32bf0e | 870 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ |
99633e97 | 871 | if (debug > 0) print_double_string (w, s1, sz1, s2, sz2) |
fa9a63c5 RM |
872 | |
873 | ||
874 | /* Print the fastmap in human-readable form. */ | |
875 | ||
876 | void | |
877 | print_fastmap (fastmap) | |
878 | char *fastmap; | |
879 | { | |
880 | unsigned was_a_range = 0; | |
5e69f11e RM |
881 | unsigned i = 0; |
882 | ||
fa9a63c5 RM |
883 | while (i < (1 << BYTEWIDTH)) |
884 | { | |
885 | if (fastmap[i++]) | |
886 | { | |
887 | was_a_range = 0; | |
25fe55af RS |
888 | putchar (i - 1); |
889 | while (i < (1 << BYTEWIDTH) && fastmap[i]) | |
890 | { | |
891 | was_a_range = 1; | |
892 | i++; | |
893 | } | |
fa9a63c5 | 894 | if (was_a_range) |
25fe55af RS |
895 | { |
896 | printf ("-"); | |
897 | putchar (i - 1); | |
898 | } | |
899 | } | |
fa9a63c5 | 900 | } |
5e69f11e | 901 | putchar ('\n'); |
fa9a63c5 RM |
902 | } |
903 | ||
904 | ||
905 | /* Print a compiled pattern string in human-readable form, starting at | |
906 | the START pointer into it and ending just before the pointer END. */ | |
907 | ||
908 | void | |
909 | print_partial_compiled_pattern (start, end) | |
01618498 SM |
910 | re_char *start; |
911 | re_char *end; | |
fa9a63c5 RM |
912 | { |
913 | int mcnt, mcnt2; | |
01618498 SM |
914 | re_char *p = start; |
915 | re_char *pend = end; | |
fa9a63c5 RM |
916 | |
917 | if (start == NULL) | |
918 | { | |
a1a052df | 919 | fprintf (stderr, "(null)\n"); |
fa9a63c5 RM |
920 | return; |
921 | } | |
5e69f11e | 922 | |
fa9a63c5 RM |
923 | /* Loop over pattern commands. */ |
924 | while (p < pend) | |
925 | { | |
a1a052df | 926 | fprintf (stderr, "%d:\t", p - start); |
fa9a63c5 RM |
927 | |
928 | switch ((re_opcode_t) *p++) | |
929 | { | |
25fe55af | 930 | case no_op: |
a1a052df | 931 | fprintf (stderr, "/no_op"); |
25fe55af | 932 | break; |
fa9a63c5 | 933 | |
99633e97 | 934 | case succeed: |
a1a052df | 935 | fprintf (stderr, "/succeed"); |
99633e97 SM |
936 | break; |
937 | ||
fa9a63c5 RM |
938 | case exactn: |
939 | mcnt = *p++; | |
a1a052df | 940 | fprintf (stderr, "/exactn/%d", mcnt); |
25fe55af | 941 | do |
fa9a63c5 | 942 | { |
a1a052df | 943 | fprintf (stderr, "/%c", *p++); |
25fe55af RS |
944 | } |
945 | while (--mcnt); | |
946 | break; | |
fa9a63c5 RM |
947 | |
948 | case start_memory: | |
a1a052df | 949 | fprintf (stderr, "/start_memory/%d", *p++); |
25fe55af | 950 | break; |
fa9a63c5 RM |
951 | |
952 | case stop_memory: | |
a1a052df | 953 | fprintf (stderr, "/stop_memory/%d", *p++); |
25fe55af | 954 | break; |
fa9a63c5 RM |
955 | |
956 | case duplicate: | |
a1a052df | 957 | fprintf (stderr, "/duplicate/%d", *p++); |
fa9a63c5 RM |
958 | break; |
959 | ||
960 | case anychar: | |
a1a052df | 961 | fprintf (stderr, "/anychar"); |
fa9a63c5 RM |
962 | break; |
963 | ||
964 | case charset: | |
25fe55af RS |
965 | case charset_not: |
966 | { | |
967 | register int c, last = -100; | |
fa9a63c5 | 968 | register int in_range = 0; |
99633e97 SM |
969 | int length = CHARSET_BITMAP_SIZE (p - 1); |
970 | int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1); | |
fa9a63c5 | 971 | |
a1a052df | 972 | fprintf (stderr, "/charset [%s", |
839966f3 | 973 | (re_opcode_t) *(p - 1) == charset_not ? "^" : ""); |
5e69f11e | 974 | |
839966f3 KH |
975 | if (p + *p >= pend) |
976 | fprintf (stderr, " !extends past end of pattern! "); | |
fa9a63c5 | 977 | |
25fe55af | 978 | for (c = 0; c < 256; c++) |
96cc36cc | 979 | if (c / 8 < length |
fa9a63c5 RM |
980 | && (p[1 + (c/8)] & (1 << (c % 8)))) |
981 | { | |
982 | /* Are we starting a range? */ | |
983 | if (last + 1 == c && ! in_range) | |
984 | { | |
a1a052df | 985 | fprintf (stderr, "-"); |
fa9a63c5 RM |
986 | in_range = 1; |
987 | } | |
988 | /* Have we broken a range? */ | |
989 | else if (last + 1 != c && in_range) | |
96cc36cc | 990 | { |
a1a052df | 991 | fprintf (stderr, "%c", last); |
fa9a63c5 RM |
992 | in_range = 0; |
993 | } | |
5e69f11e | 994 | |
fa9a63c5 | 995 | if (! in_range) |
a1a052df | 996 | fprintf (stderr, "%c", c); |
fa9a63c5 RM |
997 | |
998 | last = c; | |
25fe55af | 999 | } |
fa9a63c5 RM |
1000 | |
1001 | if (in_range) | |
a1a052df | 1002 | fprintf (stderr, "%c", last); |
fa9a63c5 | 1003 | |
a1a052df | 1004 | fprintf (stderr, "]"); |
fa9a63c5 | 1005 | |
99633e97 | 1006 | p += 1 + length; |
96cc36cc | 1007 | |
96cc36cc | 1008 | if (has_range_table) |
99633e97 SM |
1009 | { |
1010 | int count; | |
a1a052df | 1011 | fprintf (stderr, "has-range-table"); |
99633e97 SM |
1012 | |
1013 | /* ??? Should print the range table; for now, just skip it. */ | |
1014 | p += 2; /* skip range table bits */ | |
1015 | EXTRACT_NUMBER_AND_INCR (count, p); | |
1016 | p = CHARSET_RANGE_TABLE_END (p, count); | |
1017 | } | |
fa9a63c5 RM |
1018 | } |
1019 | break; | |
1020 | ||
1021 | case begline: | |
a1a052df | 1022 | fprintf (stderr, "/begline"); |
25fe55af | 1023 | break; |
fa9a63c5 RM |
1024 | |
1025 | case endline: | |
a1a052df | 1026 | fprintf (stderr, "/endline"); |
25fe55af | 1027 | break; |
fa9a63c5 RM |
1028 | |
1029 | case on_failure_jump: | |
25fe55af | 1030 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1031 | fprintf (stderr, "/on_failure_jump to %d", p + mcnt - start); |
25fe55af | 1032 | break; |
fa9a63c5 RM |
1033 | |
1034 | case on_failure_keep_string_jump: | |
25fe55af | 1035 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1036 | fprintf (stderr, "/on_failure_keep_string_jump to %d", p + mcnt - start); |
25fe55af | 1037 | break; |
fa9a63c5 | 1038 | |
0683b6fa SM |
1039 | case on_failure_jump_nastyloop: |
1040 | extract_number_and_incr (&mcnt, &p); | |
a1a052df | 1041 | fprintf (stderr, "/on_failure_jump_nastyloop to %d", p + mcnt - start); |
0683b6fa SM |
1042 | break; |
1043 | ||
505bde11 | 1044 | case on_failure_jump_loop: |
fa9a63c5 | 1045 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1046 | fprintf (stderr, "/on_failure_jump_loop to %d", p + mcnt - start); |
5e69f11e RM |
1047 | break; |
1048 | ||
505bde11 | 1049 | case on_failure_jump_smart: |
fa9a63c5 | 1050 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1051 | fprintf (stderr, "/on_failure_jump_smart to %d", p + mcnt - start); |
5e69f11e RM |
1052 | break; |
1053 | ||
25fe55af | 1054 | case jump: |
fa9a63c5 | 1055 | extract_number_and_incr (&mcnt, &p); |
a1a052df | 1056 | fprintf (stderr, "/jump to %d", p + mcnt - start); |
fa9a63c5 RM |
1057 | break; |
1058 | ||
25fe55af RS |
1059 | case succeed_n: |
1060 | extract_number_and_incr (&mcnt, &p); | |
1061 | extract_number_and_incr (&mcnt2, &p); | |
a1a052df | 1062 | fprintf (stderr, "/succeed_n to %d, %d times", p - 2 + mcnt - start, mcnt2); |
25fe55af | 1063 | break; |
5e69f11e | 1064 | |
25fe55af RS |
1065 | case jump_n: |
1066 | extract_number_and_incr (&mcnt, &p); | |
1067 | extract_number_and_incr (&mcnt2, &p); | |
a1a052df | 1068 | fprintf (stderr, "/jump_n to %d, %d times", p - 2 + mcnt - start, mcnt2); |
25fe55af | 1069 | break; |
5e69f11e | 1070 | |
25fe55af RS |
1071 | case set_number_at: |
1072 | extract_number_and_incr (&mcnt, &p); | |
1073 | extract_number_and_incr (&mcnt2, &p); | |
a1a052df | 1074 | fprintf (stderr, "/set_number_at location %d to %d", p - 2 + mcnt - start, mcnt2); |
25fe55af | 1075 | break; |
5e69f11e | 1076 | |
25fe55af | 1077 | case wordbound: |
a1a052df | 1078 | fprintf (stderr, "/wordbound"); |
fa9a63c5 RM |
1079 | break; |
1080 | ||
1081 | case notwordbound: | |
a1a052df | 1082 | fprintf (stderr, "/notwordbound"); |
25fe55af | 1083 | break; |
fa9a63c5 RM |
1084 | |
1085 | case wordbeg: | |
a1a052df | 1086 | fprintf (stderr, "/wordbeg"); |
fa9a63c5 | 1087 | break; |
5e69f11e | 1088 | |
fa9a63c5 | 1089 | case wordend: |
a1a052df | 1090 | fprintf (stderr, "/wordend"); |
e2543b02 | 1091 | break; |
5e69f11e | 1092 | |
669fa600 | 1093 | case symbeg: |
e2543b02 | 1094 | fprintf (stderr, "/symbeg"); |
669fa600 SM |
1095 | break; |
1096 | ||
1097 | case symend: | |
e2543b02 | 1098 | fprintf (stderr, "/symend"); |
669fa600 | 1099 | break; |
5e69f11e | 1100 | |
1fb352e0 | 1101 | case syntaxspec: |
a1a052df | 1102 | fprintf (stderr, "/syntaxspec"); |
1fb352e0 | 1103 | mcnt = *p++; |
a1a052df | 1104 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1105 | break; |
1106 | ||
1107 | case notsyntaxspec: | |
a1a052df | 1108 | fprintf (stderr, "/notsyntaxspec"); |
1fb352e0 | 1109 | mcnt = *p++; |
a1a052df | 1110 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1111 | break; |
1112 | ||
0b32bf0e | 1113 | # ifdef emacs |
fa9a63c5 | 1114 | case before_dot: |
a1a052df | 1115 | fprintf (stderr, "/before_dot"); |
25fe55af | 1116 | break; |
fa9a63c5 RM |
1117 | |
1118 | case at_dot: | |
a1a052df | 1119 | fprintf (stderr, "/at_dot"); |
25fe55af | 1120 | break; |
fa9a63c5 RM |
1121 | |
1122 | case after_dot: | |
a1a052df | 1123 | fprintf (stderr, "/after_dot"); |
25fe55af | 1124 | break; |
fa9a63c5 | 1125 | |
1fb352e0 | 1126 | case categoryspec: |
a1a052df | 1127 | fprintf (stderr, "/categoryspec"); |
fa9a63c5 | 1128 | mcnt = *p++; |
a1a052df | 1129 | fprintf (stderr, "/%d", mcnt); |
25fe55af | 1130 | break; |
5e69f11e | 1131 | |
1fb352e0 | 1132 | case notcategoryspec: |
a1a052df | 1133 | fprintf (stderr, "/notcategoryspec"); |
fa9a63c5 | 1134 | mcnt = *p++; |
a1a052df | 1135 | fprintf (stderr, "/%d", mcnt); |
fa9a63c5 | 1136 | break; |
0b32bf0e | 1137 | # endif /* emacs */ |
fa9a63c5 | 1138 | |
fa9a63c5 | 1139 | case begbuf: |
a1a052df | 1140 | fprintf (stderr, "/begbuf"); |
25fe55af | 1141 | break; |
fa9a63c5 RM |
1142 | |
1143 | case endbuf: | |
a1a052df | 1144 | fprintf (stderr, "/endbuf"); |
25fe55af | 1145 | break; |
fa9a63c5 | 1146 | |
25fe55af | 1147 | default: |
a1a052df | 1148 | fprintf (stderr, "?%d", *(p-1)); |
fa9a63c5 RM |
1149 | } |
1150 | ||
a1a052df | 1151 | fprintf (stderr, "\n"); |
fa9a63c5 RM |
1152 | } |
1153 | ||
a1a052df | 1154 | fprintf (stderr, "%d:\tend of pattern.\n", p - start); |
fa9a63c5 RM |
1155 | } |
1156 | ||
1157 | ||
1158 | void | |
1159 | print_compiled_pattern (bufp) | |
1160 | struct re_pattern_buffer *bufp; | |
1161 | { | |
01618498 | 1162 | re_char *buffer = bufp->buffer; |
fa9a63c5 RM |
1163 | |
1164 | print_partial_compiled_pattern (buffer, buffer + bufp->used); | |
4bb91c68 SM |
1165 | printf ("%ld bytes used/%ld bytes allocated.\n", |
1166 | bufp->used, bufp->allocated); | |
fa9a63c5 RM |
1167 | |
1168 | if (bufp->fastmap_accurate && bufp->fastmap) | |
1169 | { | |
1170 | printf ("fastmap: "); | |
1171 | print_fastmap (bufp->fastmap); | |
1172 | } | |
1173 | ||
1174 | printf ("re_nsub: %d\t", bufp->re_nsub); | |
1175 | printf ("regs_alloc: %d\t", bufp->regs_allocated); | |
1176 | printf ("can_be_null: %d\t", bufp->can_be_null); | |
fa9a63c5 RM |
1177 | printf ("no_sub: %d\t", bufp->no_sub); |
1178 | printf ("not_bol: %d\t", bufp->not_bol); | |
1179 | printf ("not_eol: %d\t", bufp->not_eol); | |
4bb91c68 | 1180 | printf ("syntax: %lx\n", bufp->syntax); |
505bde11 | 1181 | fflush (stdout); |
fa9a63c5 RM |
1182 | /* Perhaps we should print the translate table? */ |
1183 | } | |
1184 | ||
1185 | ||
1186 | void | |
1187 | print_double_string (where, string1, size1, string2, size2) | |
66f0296e SM |
1188 | re_char *where; |
1189 | re_char *string1; | |
1190 | re_char *string2; | |
d1dfb56c EZ |
1191 | ssize_t size1; |
1192 | ssize_t size2; | |
fa9a63c5 | 1193 | { |
d1dfb56c | 1194 | ssize_t this_char; |
5e69f11e | 1195 | |
fa9a63c5 RM |
1196 | if (where == NULL) |
1197 | printf ("(null)"); | |
1198 | else | |
1199 | { | |
1200 | if (FIRST_STRING_P (where)) | |
25fe55af RS |
1201 | { |
1202 | for (this_char = where - string1; this_char < size1; this_char++) | |
1203 | putchar (string1[this_char]); | |
fa9a63c5 | 1204 | |
25fe55af RS |
1205 | where = string2; |
1206 | } | |
fa9a63c5 RM |
1207 | |
1208 | for (this_char = where - string2; this_char < size2; this_char++) | |
25fe55af | 1209 | putchar (string2[this_char]); |
fa9a63c5 RM |
1210 | } |
1211 | } | |
1212 | ||
1213 | #else /* not DEBUG */ | |
1214 | ||
0b32bf0e SM |
1215 | # undef assert |
1216 | # define assert(e) | |
fa9a63c5 | 1217 | |
0b32bf0e SM |
1218 | # define DEBUG_STATEMENT(e) |
1219 | # define DEBUG_PRINT1(x) | |
1220 | # define DEBUG_PRINT2(x1, x2) | |
1221 | # define DEBUG_PRINT3(x1, x2, x3) | |
1222 | # define DEBUG_PRINT4(x1, x2, x3, x4) | |
1223 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) | |
1224 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
fa9a63c5 RM |
1225 | |
1226 | #endif /* not DEBUG */ | |
1227 | \f | |
4da60324 PE |
1228 | /* Use this to suppress gcc's `...may be used before initialized' warnings. */ |
1229 | #ifdef lint | |
1230 | # define IF_LINT(Code) Code | |
1231 | #else | |
1232 | # define IF_LINT(Code) /* empty */ | |
1233 | #endif | |
1234 | \f | |
fa9a63c5 RM |
1235 | /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can |
1236 | also be assigned to arbitrarily: each pattern buffer stores its own | |
1237 | syntax, so it can be changed between regex compilations. */ | |
1238 | /* This has no initializer because initialized variables in Emacs | |
1239 | become read-only after dumping. */ | |
1240 | reg_syntax_t re_syntax_options; | |
1241 | ||
1242 | ||
1243 | /* Specify the precise syntax of regexps for compilation. This provides | |
1244 | for compatibility for various utilities which historically have | |
1245 | different, incompatible syntaxes. | |
1246 | ||
1247 | The argument SYNTAX is a bit mask comprised of the various bits | |
4bb91c68 | 1248 | defined in regex.h. We return the old syntax. */ |
fa9a63c5 RM |
1249 | |
1250 | reg_syntax_t | |
971de7fb | 1251 | re_set_syntax (reg_syntax_t syntax) |
fa9a63c5 RM |
1252 | { |
1253 | reg_syntax_t ret = re_syntax_options; | |
5e69f11e | 1254 | |
fa9a63c5 RM |
1255 | re_syntax_options = syntax; |
1256 | return ret; | |
1257 | } | |
c0f9ea08 | 1258 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) |
f9b0fd99 RS |
1259 | |
1260 | /* Regexp to use to replace spaces, or NULL meaning don't. */ | |
1261 | static re_char *whitespace_regexp; | |
1262 | ||
1263 | void | |
971de7fb | 1264 | re_set_whitespace_regexp (const char *regexp) |
f9b0fd99 | 1265 | { |
6470ea05 | 1266 | whitespace_regexp = (re_char *) regexp; |
f9b0fd99 RS |
1267 | } |
1268 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) | |
fa9a63c5 RM |
1269 | \f |
1270 | /* This table gives an error message for each of the error codes listed | |
4bb91c68 | 1271 | in regex.h. Obviously the order here has to be same as there. |
fa9a63c5 | 1272 | POSIX doesn't require that we do anything for REG_NOERROR, |
4bb91c68 | 1273 | but why not be nice? */ |
fa9a63c5 RM |
1274 | |
1275 | static const char *re_error_msgid[] = | |
5e69f11e RM |
1276 | { |
1277 | gettext_noop ("Success"), /* REG_NOERROR */ | |
1278 | gettext_noop ("No match"), /* REG_NOMATCH */ | |
1279 | gettext_noop ("Invalid regular expression"), /* REG_BADPAT */ | |
1280 | gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */ | |
1281 | gettext_noop ("Invalid character class name"), /* REG_ECTYPE */ | |
1282 | gettext_noop ("Trailing backslash"), /* REG_EESCAPE */ | |
1283 | gettext_noop ("Invalid back reference"), /* REG_ESUBREG */ | |
1284 | gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */ | |
1285 | gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */ | |
1286 | gettext_noop ("Unmatched \\{"), /* REG_EBRACE */ | |
1287 | gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */ | |
1288 | gettext_noop ("Invalid range end"), /* REG_ERANGE */ | |
1289 | gettext_noop ("Memory exhausted"), /* REG_ESPACE */ | |
1290 | gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */ | |
1291 | gettext_noop ("Premature end of regular expression"), /* REG_EEND */ | |
1292 | gettext_noop ("Regular expression too big"), /* REG_ESIZE */ | |
1293 | gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */ | |
b3e4c897 | 1294 | gettext_noop ("Range striding over charsets") /* REG_ERANGEX */ |
fa9a63c5 RM |
1295 | }; |
1296 | \f | |
4bb91c68 | 1297 | /* Avoiding alloca during matching, to placate r_alloc. */ |
fa9a63c5 RM |
1298 | |
1299 | /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the | |
1300 | searching and matching functions should not call alloca. On some | |
1301 | systems, alloca is implemented in terms of malloc, and if we're | |
1302 | using the relocating allocator routines, then malloc could cause a | |
1303 | relocation, which might (if the strings being searched are in the | |
1304 | ralloc heap) shift the data out from underneath the regexp | |
1305 | routines. | |
1306 | ||
5e69f11e | 1307 | Here's another reason to avoid allocation: Emacs |
fa9a63c5 RM |
1308 | processes input from X in a signal handler; processing X input may |
1309 | call malloc; if input arrives while a matching routine is calling | |
1310 | malloc, then we're scrod. But Emacs can't just block input while | |
1311 | calling matching routines; then we don't notice interrupts when | |
1312 | they come in. So, Emacs blocks input around all regexp calls | |
1313 | except the matching calls, which it leaves unprotected, in the | |
1314 | faith that they will not malloc. */ | |
1315 | ||
1316 | /* Normally, this is fine. */ | |
1317 | #define MATCH_MAY_ALLOCATE | |
1318 | ||
fa9a63c5 RM |
1319 | /* The match routines may not allocate if (1) they would do it with malloc |
1320 | and (2) it's not safe for them to use malloc. | |
1321 | Note that if REL_ALLOC is defined, matching would not use malloc for the | |
1322 | failure stack, but we would still use it for the register vectors; | |
4bb91c68 | 1323 | so REL_ALLOC should not affect this. */ |
b588157e | 1324 | #if defined REGEX_MALLOC && defined emacs |
0b32bf0e | 1325 | # undef MATCH_MAY_ALLOCATE |
fa9a63c5 RM |
1326 | #endif |
1327 | ||
1328 | \f | |
1329 | /* Failure stack declarations and macros; both re_compile_fastmap and | |
1330 | re_match_2 use a failure stack. These have to be macros because of | |
1331 | REGEX_ALLOCATE_STACK. */ | |
5e69f11e | 1332 | |
fa9a63c5 | 1333 | |
320a2a73 | 1334 | /* Approximate number of failure points for which to initially allocate space |
fa9a63c5 RM |
1335 | when matching. If this number is exceeded, we allocate more |
1336 | space, so it is not a hard limit. */ | |
1337 | #ifndef INIT_FAILURE_ALLOC | |
0b32bf0e | 1338 | # define INIT_FAILURE_ALLOC 20 |
fa9a63c5 RM |
1339 | #endif |
1340 | ||
1341 | /* Roughly the maximum number of failure points on the stack. Would be | |
320a2a73 | 1342 | exactly that if always used TYPICAL_FAILURE_SIZE items each time we failed. |
fa9a63c5 | 1343 | This is a variable only so users of regex can assign to it; we never |
ada30c0e SM |
1344 | change it ourselves. We always multiply it by TYPICAL_FAILURE_SIZE |
1345 | before using it, so it should probably be a byte-count instead. */ | |
c0f9ea08 SM |
1346 | # if defined MATCH_MAY_ALLOCATE |
1347 | /* Note that 4400 was enough to cause a crash on Alpha OSF/1, | |
320a2a73 KH |
1348 | whose default stack limit is 2mb. In order for a larger |
1349 | value to work reliably, you have to try to make it accord | |
1350 | with the process stack limit. */ | |
c0f9ea08 SM |
1351 | size_t re_max_failures = 40000; |
1352 | # else | |
1353 | size_t re_max_failures = 4000; | |
1354 | # endif | |
fa9a63c5 RM |
1355 | |
1356 | union fail_stack_elt | |
1357 | { | |
01618498 | 1358 | re_char *pointer; |
c0f9ea08 SM |
1359 | /* This should be the biggest `int' that's no bigger than a pointer. */ |
1360 | long integer; | |
fa9a63c5 RM |
1361 | }; |
1362 | ||
1363 | typedef union fail_stack_elt fail_stack_elt_t; | |
1364 | ||
1365 | typedef struct | |
1366 | { | |
1367 | fail_stack_elt_t *stack; | |
c0f9ea08 SM |
1368 | size_t size; |
1369 | size_t avail; /* Offset of next open position. */ | |
1370 | size_t frame; /* Offset of the cur constructed frame. */ | |
fa9a63c5 RM |
1371 | } fail_stack_type; |
1372 | ||
505bde11 | 1373 | #define FAIL_STACK_EMPTY() (fail_stack.frame == 0) |
fa9a63c5 RM |
1374 | |
1375 | ||
1376 | /* Define macros to initialize and free the failure stack. | |
1377 | Do `return -2' if the alloc fails. */ | |
1378 | ||
1379 | #ifdef MATCH_MAY_ALLOCATE | |
0b32bf0e | 1380 | # define INIT_FAIL_STACK() \ |
fa9a63c5 | 1381 | do { \ |
38182d90 | 1382 | fail_stack.stack = \ |
320a2a73 KH |
1383 | REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * TYPICAL_FAILURE_SIZE \ |
1384 | * sizeof (fail_stack_elt_t)); \ | |
fa9a63c5 RM |
1385 | \ |
1386 | if (fail_stack.stack == NULL) \ | |
1387 | return -2; \ | |
1388 | \ | |
1389 | fail_stack.size = INIT_FAILURE_ALLOC; \ | |
1390 | fail_stack.avail = 0; \ | |
505bde11 | 1391 | fail_stack.frame = 0; \ |
fa9a63c5 | 1392 | } while (0) |
fa9a63c5 | 1393 | #else |
0b32bf0e | 1394 | # define INIT_FAIL_STACK() \ |
fa9a63c5 RM |
1395 | do { \ |
1396 | fail_stack.avail = 0; \ | |
505bde11 | 1397 | fail_stack.frame = 0; \ |
fa9a63c5 RM |
1398 | } while (0) |
1399 | ||
b313f9d8 PE |
1400 | # define RETALLOC_IF(addr, n, t) \ |
1401 | if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t) | |
fa9a63c5 RM |
1402 | #endif |
1403 | ||
1404 | ||
320a2a73 KH |
1405 | /* Double the size of FAIL_STACK, up to a limit |
1406 | which allows approximately `re_max_failures' items. | |
fa9a63c5 RM |
1407 | |
1408 | Return 1 if succeeds, and 0 if either ran out of memory | |
5e69f11e RM |
1409 | allocating space for it or it was already too large. |
1410 | ||
4bb91c68 | 1411 | REGEX_REALLOCATE_STACK requires `destination' be declared. */ |
fa9a63c5 | 1412 | |
320a2a73 KH |
1413 | /* Factor to increase the failure stack size by |
1414 | when we increase it. | |
1415 | This used to be 2, but 2 was too wasteful | |
1416 | because the old discarded stacks added up to as much space | |
1417 | were as ultimate, maximum-size stack. */ | |
1418 | #define FAIL_STACK_GROWTH_FACTOR 4 | |
1419 | ||
1420 | #define GROW_FAIL_STACK(fail_stack) \ | |
eead07d6 KH |
1421 | (((fail_stack).size * sizeof (fail_stack_elt_t) \ |
1422 | >= re_max_failures * TYPICAL_FAILURE_SIZE) \ | |
fa9a63c5 | 1423 | ? 0 \ |
320a2a73 | 1424 | : ((fail_stack).stack \ |
38182d90 | 1425 | = REGEX_REALLOCATE_STACK ((fail_stack).stack, \ |
25fe55af | 1426 | (fail_stack).size * sizeof (fail_stack_elt_t), \ |
320a2a73 KH |
1427 | MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ |
1428 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1429 | * FAIL_STACK_GROWTH_FACTOR))), \ | |
fa9a63c5 RM |
1430 | \ |
1431 | (fail_stack).stack == NULL \ | |
1432 | ? 0 \ | |
6453db45 KH |
1433 | : ((fail_stack).size \ |
1434 | = (MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ | |
1435 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1436 | * FAIL_STACK_GROWTH_FACTOR)) \ | |
1437 | / sizeof (fail_stack_elt_t)), \ | |
25fe55af | 1438 | 1))) |
fa9a63c5 RM |
1439 | |
1440 | ||
fa9a63c5 RM |
1441 | /* Push a pointer value onto the failure stack. |
1442 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1443 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 | 1444 | #define PUSH_FAILURE_POINTER(item) \ |
01618498 | 1445 | fail_stack.stack[fail_stack.avail++].pointer = (item) |
fa9a63c5 RM |
1446 | |
1447 | /* This pushes an integer-valued item onto the failure stack. | |
1448 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1449 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 RM |
1450 | #define PUSH_FAILURE_INT(item) \ |
1451 | fail_stack.stack[fail_stack.avail++].integer = (item) | |
1452 | ||
b313f9d8 | 1453 | /* These POP... operations complement the PUSH... operations. |
fa9a63c5 RM |
1454 | All assume that `fail_stack' is nonempty. */ |
1455 | #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer | |
1456 | #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer | |
fa9a63c5 | 1457 | |
505bde11 SM |
1458 | /* Individual items aside from the registers. */ |
1459 | #define NUM_NONREG_ITEMS 3 | |
1460 | ||
1461 | /* Used to examine the stack (to detect infinite loops). */ | |
1462 | #define FAILURE_PAT(h) fail_stack.stack[(h) - 1].pointer | |
66f0296e | 1463 | #define FAILURE_STR(h) (fail_stack.stack[(h) - 2].pointer) |
505bde11 SM |
1464 | #define NEXT_FAILURE_HANDLE(h) fail_stack.stack[(h) - 3].integer |
1465 | #define TOP_FAILURE_HANDLE() fail_stack.frame | |
fa9a63c5 RM |
1466 | |
1467 | ||
505bde11 SM |
1468 | #define ENSURE_FAIL_STACK(space) \ |
1469 | while (REMAINING_AVAIL_SLOTS <= space) { \ | |
1470 | if (!GROW_FAIL_STACK (fail_stack)) \ | |
1471 | return -2; \ | |
1472 | DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", (fail_stack).size);\ | |
1473 | DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\ | |
1474 | } | |
1475 | ||
1476 | /* Push register NUM onto the stack. */ | |
1477 | #define PUSH_FAILURE_REG(num) \ | |
1478 | do { \ | |
1479 | char *destination; \ | |
1480 | ENSURE_FAIL_STACK(3); \ | |
1481 | DEBUG_PRINT4 (" Push reg %d (spanning %p -> %p)\n", \ | |
1482 | num, regstart[num], regend[num]); \ | |
1483 | PUSH_FAILURE_POINTER (regstart[num]); \ | |
1484 | PUSH_FAILURE_POINTER (regend[num]); \ | |
1485 | PUSH_FAILURE_INT (num); \ | |
1486 | } while (0) | |
1487 | ||
01618498 SM |
1488 | /* Change the counter's value to VAL, but make sure that it will |
1489 | be reset when backtracking. */ | |
1490 | #define PUSH_NUMBER(ptr,val) \ | |
dc1e502d SM |
1491 | do { \ |
1492 | char *destination; \ | |
1493 | int c; \ | |
1494 | ENSURE_FAIL_STACK(3); \ | |
1495 | EXTRACT_NUMBER (c, ptr); \ | |
01618498 | 1496 | DEBUG_PRINT4 (" Push number %p = %d -> %d\n", ptr, c, val); \ |
dc1e502d SM |
1497 | PUSH_FAILURE_INT (c); \ |
1498 | PUSH_FAILURE_POINTER (ptr); \ | |
1499 | PUSH_FAILURE_INT (-1); \ | |
01618498 | 1500 | STORE_NUMBER (ptr, val); \ |
dc1e502d SM |
1501 | } while (0) |
1502 | ||
505bde11 | 1503 | /* Pop a saved register off the stack. */ |
dc1e502d | 1504 | #define POP_FAILURE_REG_OR_COUNT() \ |
505bde11 | 1505 | do { \ |
d1dfb56c | 1506 | long pfreg = POP_FAILURE_INT (); \ |
19ed5445 | 1507 | if (pfreg == -1) \ |
dc1e502d SM |
1508 | { \ |
1509 | /* It's a counter. */ \ | |
6dcf2d0e SM |
1510 | /* Here, we discard `const', making re_match non-reentrant. */ \ |
1511 | unsigned char *ptr = (unsigned char*) POP_FAILURE_POINTER (); \ | |
19ed5445 PE |
1512 | pfreg = POP_FAILURE_INT (); \ |
1513 | STORE_NUMBER (ptr, pfreg); \ | |
1514 | DEBUG_PRINT3 (" Pop counter %p = %d\n", ptr, pfreg); \ | |
dc1e502d SM |
1515 | } \ |
1516 | else \ | |
1517 | { \ | |
19ed5445 PE |
1518 | regend[pfreg] = POP_FAILURE_POINTER (); \ |
1519 | regstart[pfreg] = POP_FAILURE_POINTER (); \ | |
dc1e502d | 1520 | DEBUG_PRINT4 (" Pop reg %d (spanning %p -> %p)\n", \ |
19ed5445 | 1521 | pfreg, regstart[pfreg], regend[pfreg]); \ |
dc1e502d | 1522 | } \ |
505bde11 SM |
1523 | } while (0) |
1524 | ||
1525 | /* Check that we are not stuck in an infinite loop. */ | |
1526 | #define CHECK_INFINITE_LOOP(pat_cur, string_place) \ | |
1527 | do { \ | |
d1dfb56c | 1528 | ssize_t failure = TOP_FAILURE_HANDLE (); \ |
505bde11 | 1529 | /* Check for infinite matching loops */ \ |
f6df485f RS |
1530 | while (failure > 0 \ |
1531 | && (FAILURE_STR (failure) == string_place \ | |
1532 | || FAILURE_STR (failure) == NULL)) \ | |
505bde11 SM |
1533 | { \ |
1534 | assert (FAILURE_PAT (failure) >= bufp->buffer \ | |
66f0296e | 1535 | && FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \ |
505bde11 | 1536 | if (FAILURE_PAT (failure) == pat_cur) \ |
f6df485f | 1537 | { \ |
6df42991 SM |
1538 | cycle = 1; \ |
1539 | break; \ | |
f6df485f | 1540 | } \ |
66f0296e | 1541 | DEBUG_PRINT2 (" Other pattern: %p\n", FAILURE_PAT (failure)); \ |
505bde11 SM |
1542 | failure = NEXT_FAILURE_HANDLE(failure); \ |
1543 | } \ | |
1544 | DEBUG_PRINT2 (" Other string: %p\n", FAILURE_STR (failure)); \ | |
1545 | } while (0) | |
6df42991 | 1546 | |
fa9a63c5 | 1547 | /* Push the information about the state we will need |
5e69f11e RM |
1548 | if we ever fail back to it. |
1549 | ||
505bde11 | 1550 | Requires variables fail_stack, regstart, regend and |
320a2a73 | 1551 | num_regs be declared. GROW_FAIL_STACK requires `destination' be |
fa9a63c5 | 1552 | declared. |
5e69f11e | 1553 | |
fa9a63c5 RM |
1554 | Does `return FAILURE_CODE' if runs out of memory. */ |
1555 | ||
505bde11 SM |
1556 | #define PUSH_FAILURE_POINT(pattern, string_place) \ |
1557 | do { \ | |
1558 | char *destination; \ | |
1559 | /* Must be int, so when we don't save any registers, the arithmetic \ | |
1560 | of 0 + -1 isn't done as unsigned. */ \ | |
1561 | \ | |
505bde11 | 1562 | DEBUG_STATEMENT (nfailure_points_pushed++); \ |
4bb91c68 | 1563 | DEBUG_PRINT1 ("\nPUSH_FAILURE_POINT:\n"); \ |
505bde11 SM |
1564 | DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail); \ |
1565 | DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\ | |
1566 | \ | |
1567 | ENSURE_FAIL_STACK (NUM_NONREG_ITEMS); \ | |
1568 | \ | |
1569 | DEBUG_PRINT1 ("\n"); \ | |
1570 | \ | |
1571 | DEBUG_PRINT2 (" Push frame index: %d\n", fail_stack.frame); \ | |
1572 | PUSH_FAILURE_INT (fail_stack.frame); \ | |
1573 | \ | |
1574 | DEBUG_PRINT2 (" Push string %p: `", string_place); \ | |
1575 | DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, size2);\ | |
1576 | DEBUG_PRINT1 ("'\n"); \ | |
1577 | PUSH_FAILURE_POINTER (string_place); \ | |
1578 | \ | |
1579 | DEBUG_PRINT2 (" Push pattern %p: ", pattern); \ | |
1580 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern, pend); \ | |
1581 | PUSH_FAILURE_POINTER (pattern); \ | |
1582 | \ | |
1583 | /* Close the frame by moving the frame pointer past it. */ \ | |
1584 | fail_stack.frame = fail_stack.avail; \ | |
1585 | } while (0) | |
fa9a63c5 | 1586 | |
320a2a73 KH |
1587 | /* Estimate the size of data pushed by a typical failure stack entry. |
1588 | An estimate is all we need, because all we use this for | |
1589 | is to choose a limit for how big to make the failure stack. */ | |
ada30c0e | 1590 | /* BEWARE, the value `20' is hard-coded in emacs.c:main(). */ |
320a2a73 | 1591 | #define TYPICAL_FAILURE_SIZE 20 |
fa9a63c5 | 1592 | |
fa9a63c5 RM |
1593 | /* How many items can still be added to the stack without overflowing it. */ |
1594 | #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) | |
1595 | ||
1596 | ||
1597 | /* Pops what PUSH_FAIL_STACK pushes. | |
1598 | ||
1599 | We restore into the parameters, all of which should be lvalues: | |
1600 | STR -- the saved data position. | |
1601 | PAT -- the saved pattern position. | |
fa9a63c5 | 1602 | REGSTART, REGEND -- arrays of string positions. |
5e69f11e | 1603 | |
fa9a63c5 | 1604 | Also assumes the variables `fail_stack' and (if debugging), `bufp', |
7814e705 | 1605 | `pend', `string1', `size1', `string2', and `size2'. */ |
fa9a63c5 | 1606 | |
505bde11 SM |
1607 | #define POP_FAILURE_POINT(str, pat) \ |
1608 | do { \ | |
fa9a63c5 RM |
1609 | assert (!FAIL_STACK_EMPTY ()); \ |
1610 | \ | |
1611 | /* Remove failure points and point to how many regs pushed. */ \ | |
1612 | DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \ | |
1613 | DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \ | |
25fe55af | 1614 | DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \ |
fa9a63c5 | 1615 | \ |
505bde11 SM |
1616 | /* Pop the saved registers. */ \ |
1617 | while (fail_stack.frame < fail_stack.avail) \ | |
dc1e502d | 1618 | POP_FAILURE_REG_OR_COUNT (); \ |
fa9a63c5 | 1619 | \ |
01618498 | 1620 | pat = POP_FAILURE_POINTER (); \ |
505bde11 SM |
1621 | DEBUG_PRINT2 (" Popping pattern %p: ", pat); \ |
1622 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ | |
fa9a63c5 RM |
1623 | \ |
1624 | /* If the saved string location is NULL, it came from an \ | |
1625 | on_failure_keep_string_jump opcode, and we want to throw away the \ | |
1626 | saved NULL, thus retaining our current position in the string. */ \ | |
01618498 | 1627 | str = POP_FAILURE_POINTER (); \ |
505bde11 | 1628 | DEBUG_PRINT2 (" Popping string %p: `", str); \ |
fa9a63c5 RM |
1629 | DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ |
1630 | DEBUG_PRINT1 ("'\n"); \ | |
1631 | \ | |
505bde11 SM |
1632 | fail_stack.frame = POP_FAILURE_INT (); \ |
1633 | DEBUG_PRINT2 (" Popping frame index: %d\n", fail_stack.frame); \ | |
fa9a63c5 | 1634 | \ |
505bde11 SM |
1635 | assert (fail_stack.avail >= 0); \ |
1636 | assert (fail_stack.frame <= fail_stack.avail); \ | |
fa9a63c5 | 1637 | \ |
fa9a63c5 | 1638 | DEBUG_STATEMENT (nfailure_points_popped++); \ |
505bde11 | 1639 | } while (0) /* POP_FAILURE_POINT */ |
fa9a63c5 RM |
1640 | |
1641 | ||
1642 | \f | |
fa9a63c5 | 1643 | /* Registers are set to a sentinel when they haven't yet matched. */ |
4bb91c68 | 1644 | #define REG_UNSET(e) ((e) == NULL) |
fa9a63c5 RM |
1645 | \f |
1646 | /* Subroutine declarations and macros for regex_compile. */ | |
1647 | ||
261cb4bb PE |
1648 | static reg_errcode_t regex_compile (re_char *pattern, size_t size, |
1649 | reg_syntax_t syntax, | |
1650 | struct re_pattern_buffer *bufp); | |
1651 | static void store_op1 (re_opcode_t op, unsigned char *loc, int arg); | |
1652 | static void store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2); | |
1653 | static void insert_op1 (re_opcode_t op, unsigned char *loc, | |
1654 | int arg, unsigned char *end); | |
1655 | static void insert_op2 (re_opcode_t op, unsigned char *loc, | |
1656 | int arg1, int arg2, unsigned char *end); | |
1657 | static boolean at_begline_loc_p (re_char *pattern, re_char *p, | |
1658 | reg_syntax_t syntax); | |
1659 | static boolean at_endline_loc_p (re_char *p, re_char *pend, | |
1660 | reg_syntax_t syntax); | |
1661 | static re_char *skip_one_char (re_char *p); | |
1662 | static int analyse_first (re_char *p, re_char *pend, | |
1663 | char *fastmap, const int multibyte); | |
fa9a63c5 | 1664 | |
fa9a63c5 | 1665 | /* Fetch the next character in the uncompiled pattern, with no |
4bb91c68 | 1666 | translation. */ |
36595814 | 1667 | #define PATFETCH(c) \ |
2d1675e4 SM |
1668 | do { \ |
1669 | int len; \ | |
1670 | if (p == pend) return REG_EEND; \ | |
62a6e103 | 1671 | c = RE_STRING_CHAR_AND_LENGTH (p, len, multibyte); \ |
2d1675e4 | 1672 | p += len; \ |
fa9a63c5 RM |
1673 | } while (0) |
1674 | ||
fa9a63c5 RM |
1675 | |
1676 | /* If `translate' is non-null, return translate[D], else just D. We | |
1677 | cast the subscript to translate because some data is declared as | |
1678 | `char *', to avoid warnings when a string constant is passed. But | |
1679 | when we use a character as a subscript we must make it unsigned. */ | |
6676cb1c | 1680 | #ifndef TRANSLATE |
0b32bf0e | 1681 | # define TRANSLATE(d) \ |
66f0296e | 1682 | (RE_TRANSLATE_P (translate) ? RE_TRANSLATE (translate, (d)) : (d)) |
6676cb1c | 1683 | #endif |
fa9a63c5 RM |
1684 | |
1685 | ||
1686 | /* Macros for outputting the compiled pattern into `buffer'. */ | |
1687 | ||
1688 | /* If the buffer isn't allocated when it comes in, use this. */ | |
1689 | #define INIT_BUF_SIZE 32 | |
1690 | ||
4bb91c68 | 1691 | /* Make sure we have at least N more bytes of space in buffer. */ |
fa9a63c5 | 1692 | #define GET_BUFFER_SPACE(n) \ |
01618498 | 1693 | while ((size_t) (b - bufp->buffer + (n)) > bufp->allocated) \ |
fa9a63c5 RM |
1694 | EXTEND_BUFFER () |
1695 | ||
1696 | /* Make sure we have one more byte of buffer space and then add C to it. */ | |
1697 | #define BUF_PUSH(c) \ | |
1698 | do { \ | |
1699 | GET_BUFFER_SPACE (1); \ | |
1700 | *b++ = (unsigned char) (c); \ | |
1701 | } while (0) | |
1702 | ||
1703 | ||
1704 | /* Ensure we have two more bytes of buffer space and then append C1 and C2. */ | |
1705 | #define BUF_PUSH_2(c1, c2) \ | |
1706 | do { \ | |
1707 | GET_BUFFER_SPACE (2); \ | |
1708 | *b++ = (unsigned char) (c1); \ | |
1709 | *b++ = (unsigned char) (c2); \ | |
1710 | } while (0) | |
1711 | ||
1712 | ||
fa9a63c5 | 1713 | /* Store a jump with opcode OP at LOC to location TO. We store a |
4bb91c68 | 1714 | relative address offset by the three bytes the jump itself occupies. */ |
fa9a63c5 RM |
1715 | #define STORE_JUMP(op, loc, to) \ |
1716 | store_op1 (op, loc, (to) - (loc) - 3) | |
1717 | ||
1718 | /* Likewise, for a two-argument jump. */ | |
1719 | #define STORE_JUMP2(op, loc, to, arg) \ | |
1720 | store_op2 (op, loc, (to) - (loc) - 3, arg) | |
1721 | ||
4bb91c68 | 1722 | /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ |
fa9a63c5 RM |
1723 | #define INSERT_JUMP(op, loc, to) \ |
1724 | insert_op1 (op, loc, (to) - (loc) - 3, b) | |
1725 | ||
1726 | /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ | |
1727 | #define INSERT_JUMP2(op, loc, to, arg) \ | |
1728 | insert_op2 (op, loc, (to) - (loc) - 3, arg, b) | |
1729 | ||
1730 | ||
1731 | /* This is not an arbitrary limit: the arguments which represent offsets | |
839966f3 | 1732 | into the pattern are two bytes long. So if 2^15 bytes turns out to |
fa9a63c5 | 1733 | be too small, many things would have to change. */ |
839966f3 KH |
1734 | # define MAX_BUF_SIZE (1L << 15) |
1735 | ||
fa9a63c5 RM |
1736 | /* Extend the buffer by twice its current size via realloc and |
1737 | reset the pointers that pointed into the old block to point to the | |
1738 | correct places in the new one. If extending the buffer results in it | |
4bb91c68 SM |
1739 | being larger than MAX_BUF_SIZE, then flag memory exhausted. */ |
1740 | #if __BOUNDED_POINTERS__ | |
1741 | # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) | |
381880b0 CY |
1742 | # define MOVE_BUFFER_POINTER(P) \ |
1743 | (__ptrlow (P) = new_buffer + (__ptrlow (P) - old_buffer), \ | |
1744 | SET_HIGH_BOUND (P), \ | |
1745 | __ptrvalue (P) = new_buffer + (__ptrvalue (P) - old_buffer)) | |
4bb91c68 SM |
1746 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
1747 | else \ | |
1748 | { \ | |
1749 | SET_HIGH_BOUND (b); \ | |
1750 | SET_HIGH_BOUND (begalt); \ | |
1751 | if (fixup_alt_jump) \ | |
1752 | SET_HIGH_BOUND (fixup_alt_jump); \ | |
1753 | if (laststart) \ | |
1754 | SET_HIGH_BOUND (laststart); \ | |
1755 | if (pending_exact) \ | |
1756 | SET_HIGH_BOUND (pending_exact); \ | |
1757 | } | |
1758 | #else | |
381880b0 | 1759 | # define MOVE_BUFFER_POINTER(P) ((P) = new_buffer + ((P) - old_buffer)) |
4bb91c68 SM |
1760 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND |
1761 | #endif | |
fa9a63c5 | 1762 | #define EXTEND_BUFFER() \ |
25fe55af | 1763 | do { \ |
381880b0 | 1764 | unsigned char *old_buffer = bufp->buffer; \ |
25fe55af | 1765 | if (bufp->allocated == MAX_BUF_SIZE) \ |
fa9a63c5 RM |
1766 | return REG_ESIZE; \ |
1767 | bufp->allocated <<= 1; \ | |
1768 | if (bufp->allocated > MAX_BUF_SIZE) \ | |
25fe55af | 1769 | bufp->allocated = MAX_BUF_SIZE; \ |
01618498 | 1770 | RETALLOC (bufp->buffer, bufp->allocated, unsigned char); \ |
fa9a63c5 RM |
1771 | if (bufp->buffer == NULL) \ |
1772 | return REG_ESPACE; \ | |
1773 | /* If the buffer moved, move all the pointers into it. */ \ | |
1774 | if (old_buffer != bufp->buffer) \ | |
1775 | { \ | |
381880b0 | 1776 | unsigned char *new_buffer = bufp->buffer; \ |
4bb91c68 SM |
1777 | MOVE_BUFFER_POINTER (b); \ |
1778 | MOVE_BUFFER_POINTER (begalt); \ | |
25fe55af | 1779 | if (fixup_alt_jump) \ |
4bb91c68 | 1780 | MOVE_BUFFER_POINTER (fixup_alt_jump); \ |
25fe55af | 1781 | if (laststart) \ |
4bb91c68 | 1782 | MOVE_BUFFER_POINTER (laststart); \ |
25fe55af | 1783 | if (pending_exact) \ |
4bb91c68 | 1784 | MOVE_BUFFER_POINTER (pending_exact); \ |
fa9a63c5 | 1785 | } \ |
4bb91c68 | 1786 | ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
fa9a63c5 RM |
1787 | } while (0) |
1788 | ||
1789 | ||
1790 | /* Since we have one byte reserved for the register number argument to | |
1791 | {start,stop}_memory, the maximum number of groups we can report | |
1792 | things about is what fits in that byte. */ | |
1793 | #define MAX_REGNUM 255 | |
1794 | ||
1795 | /* But patterns can have more than `MAX_REGNUM' registers. We just | |
1796 | ignore the excess. */ | |
098d42af | 1797 | typedef int regnum_t; |
fa9a63c5 RM |
1798 | |
1799 | ||
1800 | /* Macros for the compile stack. */ | |
1801 | ||
1802 | /* Since offsets can go either forwards or backwards, this type needs to | |
4bb91c68 SM |
1803 | be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ |
1804 | /* int may be not enough when sizeof(int) == 2. */ | |
1805 | typedef long pattern_offset_t; | |
fa9a63c5 RM |
1806 | |
1807 | typedef struct | |
1808 | { | |
1809 | pattern_offset_t begalt_offset; | |
1810 | pattern_offset_t fixup_alt_jump; | |
5e69f11e | 1811 | pattern_offset_t laststart_offset; |
fa9a63c5 RM |
1812 | regnum_t regnum; |
1813 | } compile_stack_elt_t; | |
1814 | ||
1815 | ||
1816 | typedef struct | |
1817 | { | |
1818 | compile_stack_elt_t *stack; | |
d1dfb56c EZ |
1819 | size_t size; |
1820 | size_t avail; /* Offset of next open position. */ | |
fa9a63c5 RM |
1821 | } compile_stack_type; |
1822 | ||
1823 | ||
1824 | #define INIT_COMPILE_STACK_SIZE 32 | |
1825 | ||
1826 | #define COMPILE_STACK_EMPTY (compile_stack.avail == 0) | |
1827 | #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) | |
1828 | ||
4bb91c68 | 1829 | /* The next available element. */ |
fa9a63c5 RM |
1830 | #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) |
1831 | ||
0caaedb1 PE |
1832 | /* Explicit quit checking is needed for Emacs, which uses polling to |
1833 | process input events. */ | |
1834 | #ifdef emacs | |
77d11aec RS |
1835 | # define IMMEDIATE_QUIT_CHECK \ |
1836 | do { \ | |
1837 | if (immediate_quit) QUIT; \ | |
1838 | } while (0) | |
1839 | #else | |
1840 | # define IMMEDIATE_QUIT_CHECK ((void)0) | |
1841 | #endif | |
1842 | \f | |
b18215fc RS |
1843 | /* Structure to manage work area for range table. */ |
1844 | struct range_table_work_area | |
1845 | { | |
1846 | int *table; /* actual work area. */ | |
1847 | int allocated; /* allocated size for work area in bytes. */ | |
7814e705 | 1848 | int used; /* actually used size in words. */ |
96cc36cc | 1849 | int bits; /* flag to record character classes */ |
b18215fc RS |
1850 | }; |
1851 | ||
77d11aec RS |
1852 | /* Make sure that WORK_AREA can hold more N multibyte characters. |
1853 | This is used only in set_image_of_range and set_image_of_range_1. | |
1854 | It expects WORK_AREA to be a pointer. | |
1855 | If it can't get the space, it returns from the surrounding function. */ | |
1856 | ||
1857 | #define EXTEND_RANGE_TABLE(work_area, n) \ | |
1858 | do { \ | |
8f924df7 | 1859 | if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \ |
77d11aec | 1860 | { \ |
8f924df7 KH |
1861 | extend_range_table_work_area (&work_area); \ |
1862 | if ((work_area).table == 0) \ | |
77d11aec RS |
1863 | return (REG_ESPACE); \ |
1864 | } \ | |
b18215fc RS |
1865 | } while (0) |
1866 | ||
96cc36cc RS |
1867 | #define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \ |
1868 | (work_area).bits |= (bit) | |
1869 | ||
14473664 SM |
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 | |
96cc36cc | 1878 | |
b18215fc RS |
1879 | /* Set a range (RANGE_START, RANGE_END) to WORK_AREA. */ |
1880 | #define SET_RANGE_TABLE_WORK_AREA(work_area, range_start, range_end) \ | |
77d11aec | 1881 | do { \ |
8f924df7 | 1882 | EXTEND_RANGE_TABLE ((work_area), 2); \ |
b18215fc RS |
1883 | (work_area).table[(work_area).used++] = (range_start); \ |
1884 | (work_area).table[(work_area).used++] = (range_end); \ | |
1885 | } while (0) | |
1886 | ||
7814e705 | 1887 | /* Free allocated memory for WORK_AREA. */ |
b18215fc RS |
1888 | #define FREE_RANGE_TABLE_WORK_AREA(work_area) \ |
1889 | do { \ | |
1890 | if ((work_area).table) \ | |
1891 | free ((work_area).table); \ | |
1892 | } while (0) | |
1893 | ||
96cc36cc | 1894 | #define CLEAR_RANGE_TABLE_WORK_USED(work_area) ((work_area).used = 0, (work_area).bits = 0) |
b18215fc | 1895 | #define RANGE_TABLE_WORK_USED(work_area) ((work_area).used) |
96cc36cc | 1896 | #define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits) |
b18215fc | 1897 | #define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i]) |
77d11aec | 1898 | \f |
b18215fc | 1899 | |
fa9a63c5 | 1900 | /* Set the bit for character C in a list. */ |
01618498 | 1901 | #define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH)) |
fa9a63c5 RM |
1902 | |
1903 | ||
bf216479 KH |
1904 | #ifdef emacs |
1905 | ||
cf9c99bc KH |
1906 | /* Store characters in the range FROM to TO in the bitmap at B (for |
1907 | ASCII and unibyte characters) and WORK_AREA (for multibyte | |
1908 | characters) while translating them and paying attention to the | |
1909 | continuity of translated characters. | |
8f924df7 | 1910 | |
cf9c99bc KH |
1911 | Implementation note: It is better to implement these fairly big |
1912 | macros by a function, but it's not that easy because macros called | |
8f924df7 | 1913 | in this macro assume various local variables already declared. */ |
bf216479 | 1914 | |
cf9c99bc KH |
1915 | /* Both FROM and TO are ASCII characters. */ |
1916 | ||
1917 | #define SETUP_ASCII_RANGE(work_area, FROM, TO) \ | |
1918 | do { \ | |
1919 | int C0, C1; \ | |
1920 | \ | |
1921 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1922 | { \ | |
1923 | C1 = TRANSLATE (C0); \ | |
1924 | if (! ASCII_CHAR_P (C1)) \ | |
1925 | { \ | |
1926 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
1927 | if ((C1 = RE_CHAR_TO_UNIBYTE (C1)) < 0) \ | |
1928 | C1 = C0; \ | |
1929 | } \ | |
1930 | SET_LIST_BIT (C1); \ | |
1931 | } \ | |
1932 | } while (0) | |
1933 | ||
1934 | ||
1935 | /* Both FROM and TO are unibyte characters (0x80..0xFF). */ | |
1936 | ||
1937 | #define SETUP_UNIBYTE_RANGE(work_area, FROM, TO) \ | |
1938 | do { \ | |
1939 | int C0, C1, C2, I; \ | |
1940 | int USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1941 | \ | |
1942 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1943 | { \ | |
1944 | C1 = RE_CHAR_TO_MULTIBYTE (C0); \ | |
1945 | if (CHAR_BYTE8_P (C1)) \ | |
1946 | SET_LIST_BIT (C0); \ | |
1947 | else \ | |
1948 | { \ | |
1949 | C2 = TRANSLATE (C1); \ | |
1950 | if (C2 == C1 \ | |
1951 | || (C1 = RE_CHAR_TO_UNIBYTE (C2)) < 0) \ | |
1952 | C1 = C0; \ | |
1953 | SET_LIST_BIT (C1); \ | |
1954 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1955 | { \ | |
1956 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1957 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1958 | \ | |
1959 | if (C2 >= from - 1 && C2 <= to + 1) \ | |
1960 | { \ | |
1961 | if (C2 == from - 1) \ | |
1962 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1963 | else if (C2 == to + 1) \ | |
1964 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
1965 | break; \ | |
1966 | } \ | |
1967 | } \ | |
1968 | if (I < USED) \ | |
1969 | SET_RANGE_TABLE_WORK_AREA ((work_area), C2, C2); \ | |
1970 | } \ | |
1971 | } \ | |
1972 | } while (0) | |
1973 | ||
1974 | ||
78edd3b7 | 1975 | /* Both FROM and TO are multibyte characters. */ |
cf9c99bc KH |
1976 | |
1977 | #define SETUP_MULTIBYTE_RANGE(work_area, FROM, TO) \ | |
1978 | do { \ | |
1979 | int C0, C1, C2, I, USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1980 | \ | |
1981 | SET_RANGE_TABLE_WORK_AREA ((work_area), (FROM), (TO)); \ | |
1982 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1983 | { \ | |
1984 | C1 = TRANSLATE (C0); \ | |
1985 | if ((C2 = RE_CHAR_TO_UNIBYTE (C1)) >= 0 \ | |
1986 | || (C1 != C0 && (C2 = RE_CHAR_TO_UNIBYTE (C0)) >= 0)) \ | |
1987 | SET_LIST_BIT (C2); \ | |
1988 | if (C1 >= (FROM) && C1 <= (TO)) \ | |
1989 | continue; \ | |
1990 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1991 | { \ | |
1992 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1993 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1994 | \ | |
1995 | if (C1 >= from - 1 && C1 <= to + 1) \ | |
1996 | { \ | |
1997 | if (C1 == from - 1) \ | |
1998 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1999 | else if (C1 == to + 1) \ | |
2000 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
2001 | break; \ | |
2002 | } \ | |
2003 | } \ | |
2004 | if (I < USED) \ | |
2005 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
2006 | } \ | |
bf216479 KH |
2007 | } while (0) |
2008 | ||
2009 | #endif /* emacs */ | |
2010 | ||
fa9a63c5 | 2011 | /* Get the next unsigned number in the uncompiled pattern. */ |
25fe55af | 2012 | #define GET_UNSIGNED_NUMBER(num) \ |
c72b0edd SM |
2013 | do { \ |
2014 | if (p == pend) \ | |
2015 | FREE_STACK_RETURN (REG_EBRACE); \ | |
2016 | else \ | |
2017 | { \ | |
2018 | PATFETCH (c); \ | |
2019 | while ('0' <= c && c <= '9') \ | |
2020 | { \ | |
2021 | int prev; \ | |
2022 | if (num < 0) \ | |
2023 | num = 0; \ | |
2024 | prev = num; \ | |
2025 | num = num * 10 + c - '0'; \ | |
2026 | if (num / 10 != prev) \ | |
2027 | FREE_STACK_RETURN (REG_BADBR); \ | |
2028 | if (p == pend) \ | |
2029 | FREE_STACK_RETURN (REG_EBRACE); \ | |
2030 | PATFETCH (c); \ | |
2031 | } \ | |
2032 | } \ | |
2033 | } while (0) | |
77d11aec | 2034 | \f |
1fdab503 | 2035 | #if ! WIDE_CHAR_SUPPORT |
01618498 | 2036 | |
14473664 | 2037 | /* Map a string to the char class it names (if any). */ |
1fdab503 | 2038 | re_wctype_t |
971de7fb | 2039 | re_wctype (const re_char *str) |
14473664 | 2040 | { |
5b0534c8 | 2041 | const char *string = (const char *) str; |
14473664 SM |
2042 | if (STREQ (string, "alnum")) return RECC_ALNUM; |
2043 | else if (STREQ (string, "alpha")) return RECC_ALPHA; | |
2044 | else if (STREQ (string, "word")) return RECC_WORD; | |
2045 | else if (STREQ (string, "ascii")) return RECC_ASCII; | |
2046 | else if (STREQ (string, "nonascii")) return RECC_NONASCII; | |
2047 | else if (STREQ (string, "graph")) return RECC_GRAPH; | |
2048 | else if (STREQ (string, "lower")) return RECC_LOWER; | |
2049 | else if (STREQ (string, "print")) return RECC_PRINT; | |
2050 | else if (STREQ (string, "punct")) return RECC_PUNCT; | |
2051 | else if (STREQ (string, "space")) return RECC_SPACE; | |
2052 | else if (STREQ (string, "upper")) return RECC_UPPER; | |
2053 | else if (STREQ (string, "unibyte")) return RECC_UNIBYTE; | |
2054 | else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE; | |
2055 | else if (STREQ (string, "digit")) return RECC_DIGIT; | |
2056 | else if (STREQ (string, "xdigit")) return RECC_XDIGIT; | |
2057 | else if (STREQ (string, "cntrl")) return RECC_CNTRL; | |
2058 | else if (STREQ (string, "blank")) return RECC_BLANK; | |
2059 | else return 0; | |
2060 | } | |
2061 | ||
e0f24100 | 2062 | /* True if CH is in the char class CC. */ |
1fdab503 | 2063 | boolean |
971de7fb | 2064 | re_iswctype (int ch, re_wctype_t cc) |
14473664 SM |
2065 | { |
2066 | switch (cc) | |
2067 | { | |
f3fcc40d AS |
2068 | case RECC_ALNUM: return ISALNUM (ch) != 0; |
2069 | case RECC_ALPHA: return ISALPHA (ch) != 0; | |
2070 | case RECC_BLANK: return ISBLANK (ch) != 0; | |
2071 | case RECC_CNTRL: return ISCNTRL (ch) != 0; | |
2072 | case RECC_DIGIT: return ISDIGIT (ch) != 0; | |
2073 | case RECC_GRAPH: return ISGRAPH (ch) != 0; | |
2074 | case RECC_LOWER: return ISLOWER (ch) != 0; | |
2075 | case RECC_PRINT: return ISPRINT (ch) != 0; | |
2076 | case RECC_PUNCT: return ISPUNCT (ch) != 0; | |
2077 | case RECC_SPACE: return ISSPACE (ch) != 0; | |
2078 | case RECC_UPPER: return ISUPPER (ch) != 0; | |
2079 | case RECC_XDIGIT: return ISXDIGIT (ch) != 0; | |
2080 | case RECC_ASCII: return IS_REAL_ASCII (ch) != 0; | |
213bd7f2 | 2081 | case RECC_NONASCII: return !IS_REAL_ASCII (ch); |
f3fcc40d | 2082 | case RECC_UNIBYTE: return ISUNIBYTE (ch) != 0; |
213bd7f2 | 2083 | case RECC_MULTIBYTE: return !ISUNIBYTE (ch); |
f3fcc40d | 2084 | case RECC_WORD: return ISWORD (ch) != 0; |
0cdd06f8 SM |
2085 | case RECC_ERROR: return false; |
2086 | default: | |
5e617bc2 | 2087 | abort (); |
14473664 SM |
2088 | } |
2089 | } | |
fa9a63c5 | 2090 | |
14473664 SM |
2091 | /* Return a bit-pattern to use in the range-table bits to match multibyte |
2092 | chars of class CC. */ | |
2093 | static int | |
971de7fb | 2094 | re_wctype_to_bit (re_wctype_t cc) |
14473664 SM |
2095 | { |
2096 | switch (cc) | |
2097 | { | |
2098 | case RECC_NONASCII: case RECC_PRINT: case RECC_GRAPH: | |
0cdd06f8 SM |
2099 | case RECC_MULTIBYTE: return BIT_MULTIBYTE; |
2100 | case RECC_ALPHA: case RECC_ALNUM: case RECC_WORD: return BIT_WORD; | |
2101 | case RECC_LOWER: return BIT_LOWER; | |
2102 | case RECC_UPPER: return BIT_UPPER; | |
2103 | case RECC_PUNCT: return BIT_PUNCT; | |
2104 | case RECC_SPACE: return BIT_SPACE; | |
14473664 | 2105 | case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL: |
0cdd06f8 SM |
2106 | case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0; |
2107 | default: | |
5e617bc2 | 2108 | abort (); |
14473664 SM |
2109 | } |
2110 | } | |
2111 | #endif | |
77d11aec RS |
2112 | \f |
2113 | /* Filling in the work area of a range. */ | |
2114 | ||
2115 | /* Actually extend the space in WORK_AREA. */ | |
2116 | ||
2117 | static void | |
971de7fb | 2118 | extend_range_table_work_area (struct range_table_work_area *work_area) |
177c0ea7 | 2119 | { |
77d11aec | 2120 | work_area->allocated += 16 * sizeof (int); |
38182d90 | 2121 | work_area->table = realloc (work_area->table, work_area->allocated); |
77d11aec RS |
2122 | } |
2123 | ||
8f924df7 | 2124 | #if 0 |
77d11aec RS |
2125 | #ifdef emacs |
2126 | ||
2127 | /* Carefully find the ranges of codes that are equivalent | |
2128 | under case conversion to the range start..end when passed through | |
2129 | TRANSLATE. Handle the case where non-letters can come in between | |
2130 | two upper-case letters (which happens in Latin-1). | |
2131 | Also handle the case of groups of more than 2 case-equivalent chars. | |
2132 | ||
2133 | The basic method is to look at consecutive characters and see | |
2134 | if they can form a run that can be handled as one. | |
2135 | ||
2136 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2137 | ||
2138 | static int | |
1dae0f0a AS |
2139 | set_image_of_range_1 (struct range_table_work_area *work_area, |
2140 | re_wchar_t start, re_wchar_t end, | |
2141 | RE_TRANSLATE_TYPE translate) | |
77d11aec RS |
2142 | { |
2143 | /* `one_case' indicates a character, or a run of characters, | |
2144 | each of which is an isolate (no case-equivalents). | |
2145 | This includes all ASCII non-letters. | |
2146 | ||
2147 | `two_case' indicates a character, or a run of characters, | |
2148 | each of which has two case-equivalent forms. | |
2149 | This includes all ASCII letters. | |
2150 | ||
2151 | `strange' indicates a character that has more than one | |
2152 | case-equivalent. */ | |
177c0ea7 | 2153 | |
77d11aec RS |
2154 | enum case_type {one_case, two_case, strange}; |
2155 | ||
2156 | /* Describe the run that is in progress, | |
2157 | which the next character can try to extend. | |
2158 | If run_type is strange, that means there really is no run. | |
2159 | If run_type is one_case, then run_start...run_end is the run. | |
2160 | If run_type is two_case, then the run is run_start...run_end, | |
2161 | and the case-equivalents end at run_eqv_end. */ | |
2162 | ||
2163 | enum case_type run_type = strange; | |
2164 | int run_start, run_end, run_eqv_end; | |
2165 | ||
2166 | Lisp_Object eqv_table; | |
2167 | ||
2168 | if (!RE_TRANSLATE_P (translate)) | |
2169 | { | |
b7c12565 | 2170 | EXTEND_RANGE_TABLE (work_area, 2); |
77d11aec RS |
2171 | work_area->table[work_area->used++] = (start); |
2172 | work_area->table[work_area->used++] = (end); | |
b7c12565 | 2173 | return -1; |
77d11aec RS |
2174 | } |
2175 | ||
2176 | eqv_table = XCHAR_TABLE (translate)->extras[2]; | |
99633e97 | 2177 | |
77d11aec RS |
2178 | for (; start <= end; start++) |
2179 | { | |
2180 | enum case_type this_type; | |
2181 | int eqv = RE_TRANSLATE (eqv_table, start); | |
2182 | int minchar, maxchar; | |
2183 | ||
2184 | /* Classify this character */ | |
2185 | if (eqv == start) | |
2186 | this_type = one_case; | |
2187 | else if (RE_TRANSLATE (eqv_table, eqv) == start) | |
2188 | this_type = two_case; | |
2189 | else | |
2190 | this_type = strange; | |
2191 | ||
2192 | if (start < eqv) | |
2193 | minchar = start, maxchar = eqv; | |
2194 | else | |
2195 | minchar = eqv, maxchar = start; | |
2196 | ||
2197 | /* Can this character extend the run in progress? */ | |
2198 | if (this_type == strange || this_type != run_type | |
2199 | || !(minchar == run_end + 1 | |
2200 | && (run_type == two_case | |
2201 | ? maxchar == run_eqv_end + 1 : 1))) | |
2202 | { | |
2203 | /* No, end the run. | |
2204 | Record each of its equivalent ranges. */ | |
2205 | if (run_type == one_case) | |
2206 | { | |
2207 | EXTEND_RANGE_TABLE (work_area, 2); | |
2208 | work_area->table[work_area->used++] = run_start; | |
2209 | work_area->table[work_area->used++] = run_end; | |
2210 | } | |
2211 | else if (run_type == two_case) | |
2212 | { | |
2213 | EXTEND_RANGE_TABLE (work_area, 4); | |
2214 | work_area->table[work_area->used++] = run_start; | |
2215 | work_area->table[work_area->used++] = run_end; | |
2216 | work_area->table[work_area->used++] | |
2217 | = RE_TRANSLATE (eqv_table, run_start); | |
2218 | work_area->table[work_area->used++] | |
2219 | = RE_TRANSLATE (eqv_table, run_end); | |
2220 | } | |
2221 | run_type = strange; | |
2222 | } | |
177c0ea7 | 2223 | |
77d11aec RS |
2224 | if (this_type == strange) |
2225 | { | |
2226 | /* For a strange character, add each of its equivalents, one | |
2227 | by one. Don't start a range. */ | |
2228 | do | |
2229 | { | |
2230 | EXTEND_RANGE_TABLE (work_area, 2); | |
2231 | work_area->table[work_area->used++] = eqv; | |
2232 | work_area->table[work_area->used++] = eqv; | |
2233 | eqv = RE_TRANSLATE (eqv_table, eqv); | |
2234 | } | |
2235 | while (eqv != start); | |
2236 | } | |
2237 | ||
2238 | /* Add this char to the run, or start a new run. */ | |
2239 | else if (run_type == strange) | |
2240 | { | |
2241 | /* Initialize a new range. */ | |
2242 | run_type = this_type; | |
2243 | run_start = start; | |
2244 | run_end = start; | |
2245 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2246 | } | |
2247 | else | |
2248 | { | |
2249 | /* Extend a running range. */ | |
2250 | run_end = minchar; | |
2251 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2252 | } | |
2253 | } | |
2254 | ||
2255 | /* If a run is still in progress at the end, finish it now | |
2256 | by recording its equivalent ranges. */ | |
2257 | if (run_type == one_case) | |
2258 | { | |
2259 | EXTEND_RANGE_TABLE (work_area, 2); | |
2260 | work_area->table[work_area->used++] = run_start; | |
2261 | work_area->table[work_area->used++] = run_end; | |
2262 | } | |
2263 | else if (run_type == two_case) | |
2264 | { | |
2265 | EXTEND_RANGE_TABLE (work_area, 4); | |
2266 | work_area->table[work_area->used++] = run_start; | |
2267 | work_area->table[work_area->used++] = run_end; | |
2268 | work_area->table[work_area->used++] | |
2269 | = RE_TRANSLATE (eqv_table, run_start); | |
2270 | work_area->table[work_area->used++] | |
2271 | = RE_TRANSLATE (eqv_table, run_end); | |
2272 | } | |
2273 | ||
2274 | return -1; | |
2275 | } | |
36595814 | 2276 | |
77d11aec | 2277 | #endif /* emacs */ |
36595814 | 2278 | |
2b34df4e | 2279 | /* Record the image of the range start..end when passed through |
36595814 SM |
2280 | TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end) |
2281 | and is not even necessarily contiguous. | |
b7c12565 RS |
2282 | Normally we approximate it with the smallest contiguous range that contains |
2283 | all the chars we need. However, for Latin-1 we go to extra effort | |
2284 | to do a better job. | |
2285 | ||
2286 | This function is not called for ASCII ranges. | |
77d11aec RS |
2287 | |
2288 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2289 | ||
2290 | static int | |
1dae0f0a AS |
2291 | set_image_of_range (struct range_table_work_area *work_area, |
2292 | re_wchar_t start, re_wchar_t end, | |
2293 | RE_TRANSLATE_TYPE translate) | |
36595814 | 2294 | { |
77d11aec RS |
2295 | re_wchar_t cmin, cmax; |
2296 | ||
2297 | #ifdef emacs | |
2298 | /* For Latin-1 ranges, use set_image_of_range_1 | |
2299 | to get proper handling of ranges that include letters and nonletters. | |
b7c12565 | 2300 | For a range that includes the whole of Latin-1, this is not necessary. |
77d11aec | 2301 | For other character sets, we don't bother to get this right. */ |
b7c12565 RS |
2302 | if (RE_TRANSLATE_P (translate) && start < 04400 |
2303 | && !(start < 04200 && end >= 04377)) | |
77d11aec | 2304 | { |
b7c12565 | 2305 | int newend; |
77d11aec | 2306 | int tem; |
b7c12565 RS |
2307 | newend = end; |
2308 | if (newend > 04377) | |
2309 | newend = 04377; | |
2310 | tem = set_image_of_range_1 (work_area, start, newend, translate); | |
77d11aec RS |
2311 | if (tem > 0) |
2312 | return tem; | |
2313 | ||
2314 | start = 04400; | |
2315 | if (end < 04400) | |
2316 | return -1; | |
2317 | } | |
2318 | #endif | |
2319 | ||
b7c12565 RS |
2320 | EXTEND_RANGE_TABLE (work_area, 2); |
2321 | work_area->table[work_area->used++] = (start); | |
2322 | work_area->table[work_area->used++] = (end); | |
2323 | ||
2324 | cmin = -1, cmax = -1; | |
77d11aec | 2325 | |
36595814 | 2326 | if (RE_TRANSLATE_P (translate)) |
b7c12565 RS |
2327 | { |
2328 | int ch; | |
77d11aec | 2329 | |
b7c12565 RS |
2330 | for (ch = start; ch <= end; ch++) |
2331 | { | |
2332 | re_wchar_t c = TRANSLATE (ch); | |
2333 | if (! (start <= c && c <= end)) | |
2334 | { | |
2335 | if (cmin == -1) | |
2336 | cmin = c, cmax = c; | |
2337 | else | |
2338 | { | |
2339 | cmin = MIN (cmin, c); | |
2340 | cmax = MAX (cmax, c); | |
2341 | } | |
2342 | } | |
2343 | } | |
2344 | ||
2345 | if (cmin != -1) | |
2346 | { | |
2347 | EXTEND_RANGE_TABLE (work_area, 2); | |
2348 | work_area->table[work_area->used++] = (cmin); | |
2349 | work_area->table[work_area->used++] = (cmax); | |
2350 | } | |
2351 | } | |
36595814 | 2352 | |
77d11aec RS |
2353 | return -1; |
2354 | } | |
8f924df7 | 2355 | #endif /* 0 */ |
fa9a63c5 RM |
2356 | \f |
2357 | #ifndef MATCH_MAY_ALLOCATE | |
2358 | ||
2359 | /* If we cannot allocate large objects within re_match_2_internal, | |
2360 | we make the fail stack and register vectors global. | |
2361 | The fail stack, we grow to the maximum size when a regexp | |
2362 | is compiled. | |
2363 | The register vectors, we adjust in size each time we | |
2364 | compile a regexp, according to the number of registers it needs. */ | |
2365 | ||
2366 | static fail_stack_type fail_stack; | |
2367 | ||
2368 | /* Size with which the following vectors are currently allocated. | |
2369 | That is so we can make them bigger as needed, | |
4bb91c68 | 2370 | but never make them smaller. */ |
fa9a63c5 RM |
2371 | static int regs_allocated_size; |
2372 | ||
66f0296e SM |
2373 | static re_char ** regstart, ** regend; |
2374 | static re_char **best_regstart, **best_regend; | |
fa9a63c5 RM |
2375 | |
2376 | /* Make the register vectors big enough for NUM_REGS registers, | |
4bb91c68 | 2377 | but don't make them smaller. */ |
fa9a63c5 RM |
2378 | |
2379 | static | |
1dae0f0a | 2380 | regex_grow_registers (int num_regs) |
fa9a63c5 RM |
2381 | { |
2382 | if (num_regs > regs_allocated_size) | |
2383 | { | |
66f0296e SM |
2384 | RETALLOC_IF (regstart, num_regs, re_char *); |
2385 | RETALLOC_IF (regend, num_regs, re_char *); | |
2386 | RETALLOC_IF (best_regstart, num_regs, re_char *); | |
2387 | RETALLOC_IF (best_regend, num_regs, re_char *); | |
fa9a63c5 RM |
2388 | |
2389 | regs_allocated_size = num_regs; | |
2390 | } | |
2391 | } | |
2392 | ||
2393 | #endif /* not MATCH_MAY_ALLOCATE */ | |
2394 | \f | |
261cb4bb PE |
2395 | static boolean group_in_compile_stack (compile_stack_type compile_stack, |
2396 | regnum_t regnum); | |
99633e97 | 2397 | |
fa9a63c5 RM |
2398 | /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. |
2399 | Returns one of error codes defined in `regex.h', or zero for success. | |
2400 | ||
2401 | Assumes the `allocated' (and perhaps `buffer') and `translate' | |
2402 | fields are set in BUFP on entry. | |
2403 | ||
2404 | If it succeeds, results are put in BUFP (if it returns an error, the | |
2405 | contents of BUFP are undefined): | |
2406 | `buffer' is the compiled pattern; | |
2407 | `syntax' is set to SYNTAX; | |
2408 | `used' is set to the length of the compiled pattern; | |
2409 | `fastmap_accurate' is zero; | |
2410 | `re_nsub' is the number of subexpressions in PATTERN; | |
2411 | `not_bol' and `not_eol' are zero; | |
5e69f11e | 2412 | |
c0f9ea08 | 2413 | The `fastmap' field is neither examined nor set. */ |
fa9a63c5 | 2414 | |
505bde11 SM |
2415 | /* Insert the `jump' from the end of last alternative to "here". |
2416 | The space for the jump has already been allocated. */ | |
2417 | #define FIXUP_ALT_JUMP() \ | |
2418 | do { \ | |
2419 | if (fixup_alt_jump) \ | |
2420 | STORE_JUMP (jump, fixup_alt_jump, b); \ | |
2421 | } while (0) | |
2422 | ||
2423 | ||
fa9a63c5 RM |
2424 | /* Return, freeing storage we allocated. */ |
2425 | #define FREE_STACK_RETURN(value) \ | |
b18215fc RS |
2426 | do { \ |
2427 | FREE_RANGE_TABLE_WORK_AREA (range_table_work); \ | |
2428 | free (compile_stack.stack); \ | |
2429 | return value; \ | |
2430 | } while (0) | |
fa9a63c5 RM |
2431 | |
2432 | static reg_errcode_t | |
971de7fb | 2433 | regex_compile (const re_char *pattern, size_t size, reg_syntax_t syntax, struct re_pattern_buffer *bufp) |
fa9a63c5 | 2434 | { |
01618498 SM |
2435 | /* We fetch characters from PATTERN here. */ |
2436 | register re_wchar_t c, c1; | |
5e69f11e | 2437 | |
fa9a63c5 RM |
2438 | /* Points to the end of the buffer, where we should append. */ |
2439 | register unsigned char *b; | |
5e69f11e | 2440 | |
fa9a63c5 RM |
2441 | /* Keeps track of unclosed groups. */ |
2442 | compile_stack_type compile_stack; | |
2443 | ||
2444 | /* Points to the current (ending) position in the pattern. */ | |
22336245 RS |
2445 | #ifdef AIX |
2446 | /* `const' makes AIX compiler fail. */ | |
66f0296e | 2447 | unsigned char *p = pattern; |
22336245 | 2448 | #else |
66f0296e | 2449 | re_char *p = pattern; |
22336245 | 2450 | #endif |
66f0296e | 2451 | re_char *pend = pattern + size; |
5e69f11e | 2452 | |
fa9a63c5 | 2453 | /* How to translate the characters in the pattern. */ |
6676cb1c | 2454 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 RM |
2455 | |
2456 | /* Address of the count-byte of the most recently inserted `exactn' | |
2457 | command. This makes it possible to tell if a new exact-match | |
2458 | character can be added to that command or if the character requires | |
2459 | a new `exactn' command. */ | |
2460 | unsigned char *pending_exact = 0; | |
2461 | ||
2462 | /* Address of start of the most recently finished expression. | |
2463 | This tells, e.g., postfix * where to find the start of its | |
2464 | operand. Reset at the beginning of groups and alternatives. */ | |
2465 | unsigned char *laststart = 0; | |
2466 | ||
2467 | /* Address of beginning of regexp, or inside of last group. */ | |
2468 | unsigned char *begalt; | |
2469 | ||
2470 | /* Place in the uncompiled pattern (i.e., the {) to | |
2471 | which to go back if the interval is invalid. */ | |
66f0296e | 2472 | re_char *beg_interval; |
5e69f11e | 2473 | |
fa9a63c5 | 2474 | /* Address of the place where a forward jump should go to the end of |
7814e705 | 2475 | the containing expression. Each alternative of an `or' -- except the |
fa9a63c5 RM |
2476 | last -- ends with a forward jump of this sort. */ |
2477 | unsigned char *fixup_alt_jump = 0; | |
2478 | ||
b18215fc RS |
2479 | /* Work area for range table of charset. */ |
2480 | struct range_table_work_area range_table_work; | |
2481 | ||
2d1675e4 SM |
2482 | /* If the object matched can contain multibyte characters. */ |
2483 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | |
2484 | ||
f9b0fd99 RS |
2485 | /* Nonzero if we have pushed down into a subpattern. */ |
2486 | int in_subpattern = 0; | |
2487 | ||
2488 | /* These hold the values of p, pattern, and pend from the main | |
2489 | pattern when we have pushed into a subpattern. */ | |
da053e48 PE |
2490 | re_char *main_p IF_LINT (= NULL); |
2491 | re_char *main_pattern IF_LINT (= NULL); | |
2492 | re_char *main_pend IF_LINT (= NULL); | |
f9b0fd99 | 2493 | |
fa9a63c5 | 2494 | #ifdef DEBUG |
99633e97 | 2495 | debug++; |
fa9a63c5 | 2496 | DEBUG_PRINT1 ("\nCompiling pattern: "); |
99633e97 | 2497 | if (debug > 0) |
fa9a63c5 RM |
2498 | { |
2499 | unsigned debug_count; | |
5e69f11e | 2500 | |
fa9a63c5 | 2501 | for (debug_count = 0; debug_count < size; debug_count++) |
25fe55af | 2502 | putchar (pattern[debug_count]); |
fa9a63c5 RM |
2503 | putchar ('\n'); |
2504 | } | |
2505 | #endif /* DEBUG */ | |
2506 | ||
2507 | /* Initialize the compile stack. */ | |
2508 | compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); | |
2509 | if (compile_stack.stack == NULL) | |
2510 | return REG_ESPACE; | |
2511 | ||
2512 | compile_stack.size = INIT_COMPILE_STACK_SIZE; | |
2513 | compile_stack.avail = 0; | |
2514 | ||
b18215fc RS |
2515 | range_table_work.table = 0; |
2516 | range_table_work.allocated = 0; | |
2517 | ||
fa9a63c5 RM |
2518 | /* Initialize the pattern buffer. */ |
2519 | bufp->syntax = syntax; | |
2520 | bufp->fastmap_accurate = 0; | |
2521 | bufp->not_bol = bufp->not_eol = 0; | |
6224b623 | 2522 | bufp->used_syntax = 0; |
fa9a63c5 RM |
2523 | |
2524 | /* Set `used' to zero, so that if we return an error, the pattern | |
2525 | printer (for debugging) will think there's no pattern. We reset it | |
2526 | at the end. */ | |
2527 | bufp->used = 0; | |
5e69f11e | 2528 | |
fa9a63c5 | 2529 | /* Always count groups, whether or not bufp->no_sub is set. */ |
5e69f11e | 2530 | bufp->re_nsub = 0; |
fa9a63c5 | 2531 | |
0b32bf0e | 2532 | #if !defined emacs && !defined SYNTAX_TABLE |
fa9a63c5 RM |
2533 | /* Initialize the syntax table. */ |
2534 | init_syntax_once (); | |
2535 | #endif | |
2536 | ||
2537 | if (bufp->allocated == 0) | |
2538 | { | |
2539 | if (bufp->buffer) | |
2540 | { /* If zero allocated, but buffer is non-null, try to realloc | |
25fe55af | 2541 | enough space. This loses if buffer's address is bogus, but |
7814e705 | 2542 | that is the user's responsibility. */ |
25fe55af RS |
2543 | RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); |
2544 | } | |
fa9a63c5 | 2545 | else |
7814e705 | 2546 | { /* Caller did not allocate a buffer. Do it for them. */ |
25fe55af RS |
2547 | bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); |
2548 | } | |
fa9a63c5 RM |
2549 | if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE); |
2550 | ||
2551 | bufp->allocated = INIT_BUF_SIZE; | |
2552 | } | |
2553 | ||
2554 | begalt = b = bufp->buffer; | |
2555 | ||
2556 | /* Loop through the uncompiled pattern until we're at the end. */ | |
f9b0fd99 | 2557 | while (1) |
fa9a63c5 | 2558 | { |
f9b0fd99 RS |
2559 | if (p == pend) |
2560 | { | |
2561 | /* If this is the end of an included regexp, | |
2562 | pop back to the main regexp and try again. */ | |
2563 | if (in_subpattern) | |
2564 | { | |
2565 | in_subpattern = 0; | |
2566 | pattern = main_pattern; | |
2567 | p = main_p; | |
2568 | pend = main_pend; | |
2569 | continue; | |
2570 | } | |
2571 | /* If this is the end of the main regexp, we are done. */ | |
2572 | break; | |
2573 | } | |
2574 | ||
fa9a63c5 RM |
2575 | PATFETCH (c); |
2576 | ||
2577 | switch (c) | |
25fe55af | 2578 | { |
f9b0fd99 RS |
2579 | case ' ': |
2580 | { | |
2581 | re_char *p1 = p; | |
2582 | ||
2583 | /* If there's no special whitespace regexp, treat | |
4fb680cd RS |
2584 | spaces normally. And don't try to do this recursively. */ |
2585 | if (!whitespace_regexp || in_subpattern) | |
f9b0fd99 RS |
2586 | goto normal_char; |
2587 | ||
2588 | /* Peek past following spaces. */ | |
2589 | while (p1 != pend) | |
2590 | { | |
2591 | if (*p1 != ' ') | |
2592 | break; | |
2593 | p1++; | |
2594 | } | |
2595 | /* If the spaces are followed by a repetition op, | |
2596 | treat them normally. */ | |
c721eee5 RS |
2597 | if (p1 != pend |
2598 | && (*p1 == '*' || *p1 == '+' || *p1 == '?' | |
f9b0fd99 RS |
2599 | || (*p1 == '\\' && p1 + 1 != pend && p1[1] == '{'))) |
2600 | goto normal_char; | |
2601 | ||
2602 | /* Replace the spaces with the whitespace regexp. */ | |
2603 | in_subpattern = 1; | |
2604 | main_p = p1; | |
2605 | main_pend = pend; | |
2606 | main_pattern = pattern; | |
2607 | p = pattern = whitespace_regexp; | |
5b0534c8 | 2608 | pend = p + strlen ((const char *) p); |
f9b0fd99 | 2609 | break; |
7814e705 | 2610 | } |
f9b0fd99 | 2611 | |
25fe55af RS |
2612 | case '^': |
2613 | { | |
7814e705 | 2614 | if ( /* If at start of pattern, it's an operator. */ |
25fe55af | 2615 | p == pattern + 1 |
7814e705 | 2616 | /* If context independent, it's an operator. */ |
25fe55af | 2617 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
7814e705 | 2618 | /* Otherwise, depends on what's come before. */ |
25fe55af | 2619 | || at_begline_loc_p (pattern, p, syntax)) |
c0f9ea08 | 2620 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? begbuf : begline); |
25fe55af RS |
2621 | else |
2622 | goto normal_char; | |
2623 | } | |
2624 | break; | |
2625 | ||
2626 | ||
2627 | case '$': | |
2628 | { | |
2629 | if ( /* If at end of pattern, it's an operator. */ | |
2630 | p == pend | |
7814e705 | 2631 | /* If context independent, it's an operator. */ |
25fe55af RS |
2632 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
2633 | /* Otherwise, depends on what's next. */ | |
2634 | || at_endline_loc_p (p, pend, syntax)) | |
c0f9ea08 | 2635 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? endbuf : endline); |
25fe55af RS |
2636 | else |
2637 | goto normal_char; | |
2638 | } | |
2639 | break; | |
fa9a63c5 RM |
2640 | |
2641 | ||
2642 | case '+': | |
25fe55af RS |
2643 | case '?': |
2644 | if ((syntax & RE_BK_PLUS_QM) | |
2645 | || (syntax & RE_LIMITED_OPS)) | |
2646 | goto normal_char; | |
2647 | handle_plus: | |
2648 | case '*': | |
2649 | /* If there is no previous pattern... */ | |
2650 | if (!laststart) | |
2651 | { | |
2652 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
2653 | FREE_STACK_RETURN (REG_BADRPT); | |
2654 | else if (!(syntax & RE_CONTEXT_INDEP_OPS)) | |
2655 | goto normal_char; | |
2656 | } | |
2657 | ||
2658 | { | |
7814e705 | 2659 | /* 1 means zero (many) matches is allowed. */ |
66f0296e SM |
2660 | boolean zero_times_ok = 0, many_times_ok = 0; |
2661 | boolean greedy = 1; | |
25fe55af RS |
2662 | |
2663 | /* If there is a sequence of repetition chars, collapse it | |
2664 | down to just one (the right one). We can't combine | |
2665 | interval operators with these because of, e.g., `a{2}*', | |
7814e705 | 2666 | which should only match an even number of `a's. */ |
25fe55af RS |
2667 | |
2668 | for (;;) | |
2669 | { | |
0b32bf0e | 2670 | if ((syntax & RE_FRUGAL) |
1c8c6d39 DL |
2671 | && c == '?' && (zero_times_ok || many_times_ok)) |
2672 | greedy = 0; | |
2673 | else | |
2674 | { | |
2675 | zero_times_ok |= c != '+'; | |
2676 | many_times_ok |= c != '?'; | |
2677 | } | |
25fe55af RS |
2678 | |
2679 | if (p == pend) | |
2680 | break; | |
ed0767d8 SM |
2681 | else if (*p == '*' |
2682 | || (!(syntax & RE_BK_PLUS_QM) | |
2683 | && (*p == '+' || *p == '?'))) | |
25fe55af | 2684 | ; |
ed0767d8 | 2685 | else if (syntax & RE_BK_PLUS_QM && *p == '\\') |
25fe55af | 2686 | { |
ed0767d8 SM |
2687 | if (p+1 == pend) |
2688 | FREE_STACK_RETURN (REG_EESCAPE); | |
2689 | if (p[1] == '+' || p[1] == '?') | |
2690 | PATFETCH (c); /* Gobble up the backslash. */ | |
2691 | else | |
2692 | break; | |
25fe55af RS |
2693 | } |
2694 | else | |
ed0767d8 | 2695 | break; |
25fe55af | 2696 | /* If we get here, we found another repeat character. */ |
ed0767d8 SM |
2697 | PATFETCH (c); |
2698 | } | |
25fe55af RS |
2699 | |
2700 | /* Star, etc. applied to an empty pattern is equivalent | |
2701 | to an empty pattern. */ | |
4e8a9132 | 2702 | if (!laststart || laststart == b) |
25fe55af RS |
2703 | break; |
2704 | ||
2705 | /* Now we know whether or not zero matches is allowed | |
7814e705 | 2706 | and also whether or not two or more matches is allowed. */ |
1c8c6d39 DL |
2707 | if (greedy) |
2708 | { | |
99633e97 | 2709 | if (many_times_ok) |
4e8a9132 SM |
2710 | { |
2711 | boolean simple = skip_one_char (laststart) == b; | |
d1dfb56c | 2712 | size_t startoffset = 0; |
f6a3f532 | 2713 | re_opcode_t ofj = |
01618498 | 2714 | /* Check if the loop can match the empty string. */ |
6df42991 SM |
2715 | (simple || !analyse_first (laststart, b, NULL, 0)) |
2716 | ? on_failure_jump : on_failure_jump_loop; | |
4e8a9132 | 2717 | assert (skip_one_char (laststart) <= b); |
177c0ea7 | 2718 | |
4e8a9132 SM |
2719 | if (!zero_times_ok && simple) |
2720 | { /* Since simple * loops can be made faster by using | |
2721 | on_failure_keep_string_jump, we turn simple P+ | |
2722 | into PP* if P is simple. */ | |
2723 | unsigned char *p1, *p2; | |
2724 | startoffset = b - laststart; | |
2725 | GET_BUFFER_SPACE (startoffset); | |
2726 | p1 = b; p2 = laststart; | |
2727 | while (p2 < p1) | |
2728 | *b++ = *p2++; | |
2729 | zero_times_ok = 1; | |
99633e97 | 2730 | } |
4e8a9132 SM |
2731 | |
2732 | GET_BUFFER_SPACE (6); | |
2733 | if (!zero_times_ok) | |
2734 | /* A + loop. */ | |
f6a3f532 | 2735 | STORE_JUMP (ofj, b, b + 6); |
99633e97 | 2736 | else |
4e8a9132 SM |
2737 | /* Simple * loops can use on_failure_keep_string_jump |
2738 | depending on what follows. But since we don't know | |
2739 | that yet, we leave the decision up to | |
2740 | on_failure_jump_smart. */ | |
f6a3f532 | 2741 | INSERT_JUMP (simple ? on_failure_jump_smart : ofj, |
4e8a9132 | 2742 | laststart + startoffset, b + 6); |
99633e97 | 2743 | b += 3; |
4e8a9132 | 2744 | STORE_JUMP (jump, b, laststart + startoffset); |
99633e97 SM |
2745 | b += 3; |
2746 | } | |
2747 | else | |
2748 | { | |
4e8a9132 SM |
2749 | /* A simple ? pattern. */ |
2750 | assert (zero_times_ok); | |
2751 | GET_BUFFER_SPACE (3); | |
2752 | INSERT_JUMP (on_failure_jump, laststart, b + 3); | |
99633e97 SM |
2753 | b += 3; |
2754 | } | |
1c8c6d39 DL |
2755 | } |
2756 | else /* not greedy */ | |
2757 | { /* I wish the greedy and non-greedy cases could be merged. */ | |
2758 | ||
0683b6fa | 2759 | GET_BUFFER_SPACE (7); /* We might use less. */ |
1c8c6d39 DL |
2760 | if (many_times_ok) |
2761 | { | |
f6a3f532 SM |
2762 | boolean emptyp = analyse_first (laststart, b, NULL, 0); |
2763 | ||
6df42991 SM |
2764 | /* The non-greedy multiple match looks like |
2765 | a repeat..until: we only need a conditional jump | |
2766 | at the end of the loop. */ | |
f6a3f532 SM |
2767 | if (emptyp) BUF_PUSH (no_op); |
2768 | STORE_JUMP (emptyp ? on_failure_jump_nastyloop | |
2769 | : on_failure_jump, b, laststart); | |
1c8c6d39 DL |
2770 | b += 3; |
2771 | if (zero_times_ok) | |
2772 | { | |
2773 | /* The repeat...until naturally matches one or more. | |
2774 | To also match zero times, we need to first jump to | |
6df42991 | 2775 | the end of the loop (its conditional jump). */ |
1c8c6d39 DL |
2776 | INSERT_JUMP (jump, laststart, b); |
2777 | b += 3; | |
2778 | } | |
2779 | } | |
2780 | else | |
2781 | { | |
2782 | /* non-greedy a?? */ | |
1c8c6d39 DL |
2783 | INSERT_JUMP (jump, laststart, b + 3); |
2784 | b += 3; | |
2785 | INSERT_JUMP (on_failure_jump, laststart, laststart + 6); | |
2786 | b += 3; | |
2787 | } | |
2788 | } | |
2789 | } | |
4e8a9132 | 2790 | pending_exact = 0; |
fa9a63c5 RM |
2791 | break; |
2792 | ||
2793 | ||
2794 | case '.': | |
25fe55af RS |
2795 | laststart = b; |
2796 | BUF_PUSH (anychar); | |
2797 | break; | |
fa9a63c5 RM |
2798 | |
2799 | ||
25fe55af RS |
2800 | case '[': |
2801 | { | |
19ed5445 PE |
2802 | re_char *p1; |
2803 | ||
b18215fc | 2804 | CLEAR_RANGE_TABLE_WORK_USED (range_table_work); |
fa9a63c5 | 2805 | |
25fe55af | 2806 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2807 | |
25fe55af RS |
2808 | /* Ensure that we have enough space to push a charset: the |
2809 | opcode, the length count, and the bitset; 34 bytes in all. */ | |
fa9a63c5 RM |
2810 | GET_BUFFER_SPACE (34); |
2811 | ||
25fe55af | 2812 | laststart = b; |
e318085a | 2813 | |
25fe55af | 2814 | /* We test `*p == '^' twice, instead of using an if |
7814e705 | 2815 | statement, so we only need one BUF_PUSH. */ |
25fe55af RS |
2816 | BUF_PUSH (*p == '^' ? charset_not : charset); |
2817 | if (*p == '^') | |
2818 | p++; | |
e318085a | 2819 | |
25fe55af RS |
2820 | /* Remember the first position in the bracket expression. */ |
2821 | p1 = p; | |
e318085a | 2822 | |
7814e705 | 2823 | /* Push the number of bytes in the bitmap. */ |
25fe55af | 2824 | BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2825 | |
25fe55af | 2826 | /* Clear the whole map. */ |
72af86bd | 2827 | memset (b, 0, (1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2828 | |
25fe55af RS |
2829 | /* charset_not matches newline according to a syntax bit. */ |
2830 | if ((re_opcode_t) b[-2] == charset_not | |
2831 | && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) | |
2832 | SET_LIST_BIT ('\n'); | |
fa9a63c5 | 2833 | |
7814e705 | 2834 | /* Read in characters and ranges, setting map bits. */ |
25fe55af RS |
2835 | for (;;) |
2836 | { | |
b18215fc | 2837 | boolean escaped_char = false; |
2d1675e4 | 2838 | const unsigned char *p2 = p; |
abbd1bcf | 2839 | re_wchar_t ch; |
e318085a | 2840 | |
25fe55af | 2841 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
e318085a | 2842 | |
36595814 SM |
2843 | /* Don't translate yet. The range TRANSLATE(X..Y) cannot |
2844 | always be determined from TRANSLATE(X) and TRANSLATE(Y) | |
2845 | So the translation is done later in a loop. Example: | |
2846 | (let ((case-fold-search t)) (string-match "[A-_]" "A")) */ | |
25fe55af | 2847 | PATFETCH (c); |
e318085a | 2848 | |
25fe55af RS |
2849 | /* \ might escape characters inside [...] and [^...]. */ |
2850 | if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') | |
2851 | { | |
2852 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
e318085a RS |
2853 | |
2854 | PATFETCH (c); | |
b18215fc | 2855 | escaped_char = true; |
25fe55af | 2856 | } |
b18215fc RS |
2857 | else |
2858 | { | |
7814e705 | 2859 | /* Could be the end of the bracket expression. If it's |
657fcfbd RS |
2860 | not (i.e., when the bracket expression is `[]' so |
2861 | far), the ']' character bit gets set way below. */ | |
2d1675e4 | 2862 | if (c == ']' && p2 != p1) |
657fcfbd | 2863 | break; |
25fe55af | 2864 | } |
b18215fc | 2865 | |
25fe55af RS |
2866 | /* See if we're at the beginning of a possible character |
2867 | class. */ | |
b18215fc | 2868 | |
2d1675e4 SM |
2869 | if (!escaped_char && |
2870 | syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') | |
657fcfbd | 2871 | { |
7814e705 | 2872 | /* Leave room for the null. */ |
14473664 | 2873 | unsigned char str[CHAR_CLASS_MAX_LENGTH + 1]; |
ed0767d8 | 2874 | const unsigned char *class_beg; |
b18215fc | 2875 | |
25fe55af RS |
2876 | PATFETCH (c); |
2877 | c1 = 0; | |
ed0767d8 | 2878 | class_beg = p; |
b18215fc | 2879 | |
25fe55af RS |
2880 | /* If pattern is `[[:'. */ |
2881 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
b18215fc | 2882 | |
25fe55af RS |
2883 | for (;;) |
2884 | { | |
14473664 SM |
2885 | PATFETCH (c); |
2886 | if ((c == ':' && *p == ']') || p == pend) | |
2887 | break; | |
2888 | if (c1 < CHAR_CLASS_MAX_LENGTH) | |
2889 | str[c1++] = c; | |
2890 | else | |
2891 | /* This is in any case an invalid class name. */ | |
2892 | str[0] = '\0'; | |
25fe55af RS |
2893 | } |
2894 | str[c1] = '\0'; | |
b18215fc RS |
2895 | |
2896 | /* If isn't a word bracketed by `[:' and `:]': | |
2897 | undo the ending character, the letters, and | |
2898 | leave the leading `:' and `[' (but set bits for | |
2899 | them). */ | |
25fe55af RS |
2900 | if (c == ':' && *p == ']') |
2901 | { | |
abbd1bcf | 2902 | re_wctype_t cc = re_wctype (str); |
14473664 SM |
2903 | |
2904 | if (cc == 0) | |
fa9a63c5 RM |
2905 | FREE_STACK_RETURN (REG_ECTYPE); |
2906 | ||
14473664 SM |
2907 | /* Throw away the ] at the end of the character |
2908 | class. */ | |
2909 | PATFETCH (c); | |
fa9a63c5 | 2910 | |
14473664 | 2911 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2912 | |
cf9c99bc KH |
2913 | #ifndef emacs |
2914 | for (ch = 0; ch < (1 << BYTEWIDTH); ++ch) | |
8f924df7 KH |
2915 | if (re_iswctype (btowc (ch), cc)) |
2916 | { | |
2917 | c = TRANSLATE (ch); | |
ed00c2ac KH |
2918 | if (c < (1 << BYTEWIDTH)) |
2919 | SET_LIST_BIT (c); | |
8f924df7 | 2920 | } |
cf9c99bc KH |
2921 | #else /* emacs */ |
2922 | /* Most character classes in a multibyte match | |
2923 | just set a flag. Exceptions are is_blank, | |
2924 | is_digit, is_cntrl, and is_xdigit, since | |
2925 | they can only match ASCII characters. We | |
2926 | don't need to handle them for multibyte. | |
2927 | They are distinguished by a negative wctype. */ | |
96cc36cc | 2928 | |
254c06a8 SM |
2929 | /* Setup the gl_state object to its buffer-defined |
2930 | value. This hardcodes the buffer-global | |
2931 | syntax-table for ASCII chars, while the other chars | |
2932 | will obey syntax-table properties. It's not ideal, | |
2933 | but it's the way it's been done until now. */ | |
d48cd3f4 | 2934 | SETUP_BUFFER_SYNTAX_TABLE (); |
254c06a8 | 2935 | |
cf9c99bc | 2936 | for (ch = 0; ch < 256; ++ch) |
25fe55af | 2937 | { |
cf9c99bc KH |
2938 | c = RE_CHAR_TO_MULTIBYTE (ch); |
2939 | if (! CHAR_BYTE8_P (c) | |
2940 | && re_iswctype (c, cc)) | |
8f924df7 | 2941 | { |
cf9c99bc KH |
2942 | SET_LIST_BIT (ch); |
2943 | c1 = TRANSLATE (c); | |
2944 | if (c1 == c) | |
2945 | continue; | |
2946 | if (ASCII_CHAR_P (c1)) | |
2947 | SET_LIST_BIT (c1); | |
2948 | else if ((c1 = RE_CHAR_TO_UNIBYTE (c1)) >= 0) | |
2949 | SET_LIST_BIT (c1); | |
8f924df7 | 2950 | } |
25fe55af | 2951 | } |
cf9c99bc KH |
2952 | SET_RANGE_TABLE_WORK_AREA_BIT |
2953 | (range_table_work, re_wctype_to_bit (cc)); | |
2954 | #endif /* emacs */ | |
6224b623 SM |
2955 | /* In most cases the matching rule for char classes |
2956 | only uses the syntax table for multibyte chars, | |
2957 | so that the content of the syntax-table it is not | |
2958 | hardcoded in the range_table. SPACE and WORD are | |
2959 | the two exceptions. */ | |
2960 | if ((1 << cc) & ((1 << RECC_SPACE) | (1 << RECC_WORD))) | |
2961 | bufp->used_syntax = 1; | |
2962 | ||
b18215fc RS |
2963 | /* Repeat the loop. */ |
2964 | continue; | |
25fe55af RS |
2965 | } |
2966 | else | |
2967 | { | |
ed0767d8 SM |
2968 | /* Go back to right after the "[:". */ |
2969 | p = class_beg; | |
25fe55af | 2970 | SET_LIST_BIT ('['); |
b18215fc RS |
2971 | |
2972 | /* Because the `:' may starts the range, we | |
2973 | can't simply set bit and repeat the loop. | |
7814e705 | 2974 | Instead, just set it to C and handle below. */ |
b18215fc | 2975 | c = ':'; |
25fe55af RS |
2976 | } |
2977 | } | |
b18215fc RS |
2978 | |
2979 | if (p < pend && p[0] == '-' && p[1] != ']') | |
2980 | { | |
2981 | ||
2982 | /* Discard the `-'. */ | |
2983 | PATFETCH (c1); | |
2984 | ||
2985 | /* Fetch the character which ends the range. */ | |
2986 | PATFETCH (c1); | |
cf9c99bc KH |
2987 | #ifdef emacs |
2988 | if (CHAR_BYTE8_P (c1) | |
2989 | && ! ASCII_CHAR_P (c) && ! CHAR_BYTE8_P (c)) | |
2990 | /* Treat the range from a multibyte character to | |
2991 | raw-byte character as empty. */ | |
2992 | c = c1 + 1; | |
2993 | #endif /* emacs */ | |
e318085a | 2994 | } |
25fe55af | 2995 | else |
b18215fc RS |
2996 | /* Range from C to C. */ |
2997 | c1 = c; | |
2998 | ||
cf9c99bc | 2999 | if (c > c1) |
25fe55af | 3000 | { |
cf9c99bc KH |
3001 | if (syntax & RE_NO_EMPTY_RANGES) |
3002 | FREE_STACK_RETURN (REG_ERANGEX); | |
3003 | /* Else, repeat the loop. */ | |
bf216479 | 3004 | } |
6fdd04b0 | 3005 | else |
25fe55af | 3006 | { |
cf9c99bc KH |
3007 | #ifndef emacs |
3008 | /* Set the range into bitmap */ | |
8f924df7 | 3009 | for (; c <= c1; c++) |
b18215fc | 3010 | { |
cf9c99bc KH |
3011 | ch = TRANSLATE (c); |
3012 | if (ch < (1 << BYTEWIDTH)) | |
3013 | SET_LIST_BIT (ch); | |
3014 | } | |
3015 | #else /* emacs */ | |
3016 | if (c < 128) | |
3017 | { | |
3018 | ch = MIN (127, c1); | |
3019 | SETUP_ASCII_RANGE (range_table_work, c, ch); | |
3020 | c = ch + 1; | |
3021 | if (CHAR_BYTE8_P (c1)) | |
3022 | c = BYTE8_TO_CHAR (128); | |
3023 | } | |
3024 | if (c <= c1) | |
3025 | { | |
3026 | if (CHAR_BYTE8_P (c)) | |
3027 | { | |
3028 | c = CHAR_TO_BYTE8 (c); | |
3029 | c1 = CHAR_TO_BYTE8 (c1); | |
3030 | for (; c <= c1; c++) | |
3031 | SET_LIST_BIT (c); | |
3032 | } | |
3033 | else if (multibyte) | |
3034 | { | |
3035 | SETUP_MULTIBYTE_RANGE (range_table_work, c, c1); | |
3036 | } | |
3037 | else | |
3038 | { | |
3039 | SETUP_UNIBYTE_RANGE (range_table_work, c, c1); | |
3040 | } | |
e934739e | 3041 | } |
cf9c99bc | 3042 | #endif /* emacs */ |
25fe55af | 3043 | } |
e318085a RS |
3044 | } |
3045 | ||
25fe55af | 3046 | /* Discard any (non)matching list bytes that are all 0 at the |
7814e705 | 3047 | end of the map. Decrease the map-length byte too. */ |
25fe55af RS |
3048 | while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) |
3049 | b[-1]--; | |
3050 | b += b[-1]; | |
fa9a63c5 | 3051 | |
96cc36cc RS |
3052 | /* Build real range table from work area. */ |
3053 | if (RANGE_TABLE_WORK_USED (range_table_work) | |
3054 | || RANGE_TABLE_WORK_BITS (range_table_work)) | |
b18215fc RS |
3055 | { |
3056 | int i; | |
3057 | int used = RANGE_TABLE_WORK_USED (range_table_work); | |
fa9a63c5 | 3058 | |
b18215fc | 3059 | /* Allocate space for COUNT + RANGE_TABLE. Needs two |
96cc36cc RS |
3060 | bytes for flags, two for COUNT, and three bytes for |
3061 | each character. */ | |
3062 | GET_BUFFER_SPACE (4 + used * 3); | |
fa9a63c5 | 3063 | |
b18215fc RS |
3064 | /* Indicate the existence of range table. */ |
3065 | laststart[1] |= 0x80; | |
fa9a63c5 | 3066 | |
96cc36cc RS |
3067 | /* Store the character class flag bits into the range table. |
3068 | If not in emacs, these flag bits are always 0. */ | |
3069 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) & 0xff; | |
3070 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) >> 8; | |
3071 | ||
b18215fc RS |
3072 | STORE_NUMBER_AND_INCR (b, used / 2); |
3073 | for (i = 0; i < used; i++) | |
3074 | STORE_CHARACTER_AND_INCR | |
3075 | (b, RANGE_TABLE_WORK_ELT (range_table_work, i)); | |
3076 | } | |
25fe55af RS |
3077 | } |
3078 | break; | |
fa9a63c5 RM |
3079 | |
3080 | ||
b18215fc | 3081 | case '(': |
25fe55af RS |
3082 | if (syntax & RE_NO_BK_PARENS) |
3083 | goto handle_open; | |
3084 | else | |
3085 | goto normal_char; | |
fa9a63c5 RM |
3086 | |
3087 | ||
25fe55af RS |
3088 | case ')': |
3089 | if (syntax & RE_NO_BK_PARENS) | |
3090 | goto handle_close; | |
3091 | else | |
3092 | goto normal_char; | |
e318085a RS |
3093 | |
3094 | ||
25fe55af RS |
3095 | case '\n': |
3096 | if (syntax & RE_NEWLINE_ALT) | |
3097 | goto handle_alt; | |
3098 | else | |
3099 | goto normal_char; | |
e318085a RS |
3100 | |
3101 | ||
b18215fc | 3102 | case '|': |
25fe55af RS |
3103 | if (syntax & RE_NO_BK_VBAR) |
3104 | goto handle_alt; | |
3105 | else | |
3106 | goto normal_char; | |
3107 | ||
3108 | ||
3109 | case '{': | |
3110 | if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) | |
3111 | goto handle_interval; | |
3112 | else | |
3113 | goto normal_char; | |
3114 | ||
3115 | ||
3116 | case '\\': | |
3117 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
3118 | ||
3119 | /* Do not translate the character after the \, so that we can | |
3120 | distinguish, e.g., \B from \b, even if we normally would | |
3121 | translate, e.g., B to b. */ | |
36595814 | 3122 | PATFETCH (c); |
25fe55af RS |
3123 | |
3124 | switch (c) | |
3125 | { | |
3126 | case '(': | |
3127 | if (syntax & RE_NO_BK_PARENS) | |
3128 | goto normal_backslash; | |
3129 | ||
3130 | handle_open: | |
505bde11 SM |
3131 | { |
3132 | int shy = 0; | |
c69b0314 | 3133 | regnum_t regnum = 0; |
505bde11 SM |
3134 | if (p+1 < pend) |
3135 | { | |
3136 | /* Look for a special (?...) construct */ | |
ed0767d8 | 3137 | if ((syntax & RE_SHY_GROUPS) && *p == '?') |
505bde11 | 3138 | { |
ed0767d8 | 3139 | PATFETCH (c); /* Gobble up the '?'. */ |
c69b0314 | 3140 | while (!shy) |
505bde11 | 3141 | { |
c69b0314 SM |
3142 | PATFETCH (c); |
3143 | switch (c) | |
3144 | { | |
3145 | case ':': shy = 1; break; | |
3146 | case '0': | |
3147 | /* An explicitly specified regnum must start | |
3148 | with non-0. */ | |
3149 | if (regnum == 0) | |
3150 | FREE_STACK_RETURN (REG_BADPAT); | |
3151 | case '1': case '2': case '3': case '4': | |
3152 | case '5': case '6': case '7': case '8': case '9': | |
3153 | regnum = 10*regnum + (c - '0'); break; | |
3154 | default: | |
3155 | /* Only (?:...) is supported right now. */ | |
3156 | FREE_STACK_RETURN (REG_BADPAT); | |
3157 | } | |
505bde11 SM |
3158 | } |
3159 | } | |
505bde11 SM |
3160 | } |
3161 | ||
3162 | if (!shy) | |
c69b0314 SM |
3163 | regnum = ++bufp->re_nsub; |
3164 | else if (regnum) | |
3165 | { /* It's actually not shy, but explicitly numbered. */ | |
3166 | shy = 0; | |
3167 | if (regnum > bufp->re_nsub) | |
3168 | bufp->re_nsub = regnum; | |
3169 | else if (regnum > bufp->re_nsub | |
3170 | /* Ideally, we'd want to check that the specified | |
3171 | group can't have matched (i.e. all subgroups | |
3172 | using the same regnum are in other branches of | |
3173 | OR patterns), but we don't currently keep track | |
3174 | of enough info to do that easily. */ | |
3175 | || group_in_compile_stack (compile_stack, regnum)) | |
3176 | FREE_STACK_RETURN (REG_BADPAT); | |
505bde11 | 3177 | } |
c69b0314 SM |
3178 | else |
3179 | /* It's really shy. */ | |
3180 | regnum = - bufp->re_nsub; | |
25fe55af | 3181 | |
99633e97 SM |
3182 | if (COMPILE_STACK_FULL) |
3183 | { | |
3184 | RETALLOC (compile_stack.stack, compile_stack.size << 1, | |
3185 | compile_stack_elt_t); | |
3186 | if (compile_stack.stack == NULL) return REG_ESPACE; | |
25fe55af | 3187 | |
99633e97 SM |
3188 | compile_stack.size <<= 1; |
3189 | } | |
25fe55af | 3190 | |
99633e97 | 3191 | /* These are the values to restore when we hit end of this |
7814e705 | 3192 | group. They are all relative offsets, so that if the |
99633e97 SM |
3193 | whole pattern moves because of realloc, they will still |
3194 | be valid. */ | |
3195 | COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; | |
3196 | COMPILE_STACK_TOP.fixup_alt_jump | |
3197 | = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; | |
3198 | COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer; | |
c69b0314 | 3199 | COMPILE_STACK_TOP.regnum = regnum; |
99633e97 | 3200 | |
c69b0314 SM |
3201 | /* Do not push a start_memory for groups beyond the last one |
3202 | we can represent in the compiled pattern. */ | |
3203 | if (regnum <= MAX_REGNUM && regnum > 0) | |
99633e97 SM |
3204 | BUF_PUSH_2 (start_memory, regnum); |
3205 | ||
3206 | compile_stack.avail++; | |
3207 | ||
3208 | fixup_alt_jump = 0; | |
3209 | laststart = 0; | |
3210 | begalt = b; | |
3211 | /* If we've reached MAX_REGNUM groups, then this open | |
3212 | won't actually generate any code, so we'll have to | |
3213 | clear pending_exact explicitly. */ | |
3214 | pending_exact = 0; | |
3215 | break; | |
505bde11 | 3216 | } |
25fe55af RS |
3217 | |
3218 | case ')': | |
3219 | if (syntax & RE_NO_BK_PARENS) goto normal_backslash; | |
3220 | ||
3221 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3222 | { |
3223 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3224 | goto normal_backslash; | |
3225 | else | |
3226 | FREE_STACK_RETURN (REG_ERPAREN); | |
3227 | } | |
25fe55af RS |
3228 | |
3229 | handle_close: | |
505bde11 | 3230 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3231 | |
3232 | /* See similar code for backslashed left paren above. */ | |
3233 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3234 | { |
3235 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3236 | goto normal_char; | |
3237 | else | |
3238 | FREE_STACK_RETURN (REG_ERPAREN); | |
3239 | } | |
25fe55af RS |
3240 | |
3241 | /* Since we just checked for an empty stack above, this | |
3242 | ``can't happen''. */ | |
3243 | assert (compile_stack.avail != 0); | |
3244 | { | |
3245 | /* We don't just want to restore into `regnum', because | |
3246 | later groups should continue to be numbered higher, | |
7814e705 | 3247 | as in `(ab)c(de)' -- the second group is #2. */ |
c69b0314 | 3248 | regnum_t regnum; |
25fe55af RS |
3249 | |
3250 | compile_stack.avail--; | |
3251 | begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; | |
3252 | fixup_alt_jump | |
3253 | = COMPILE_STACK_TOP.fixup_alt_jump | |
3254 | ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 | |
3255 | : 0; | |
3256 | laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; | |
c69b0314 | 3257 | regnum = COMPILE_STACK_TOP.regnum; |
b18215fc RS |
3258 | /* If we've reached MAX_REGNUM groups, then this open |
3259 | won't actually generate any code, so we'll have to | |
3260 | clear pending_exact explicitly. */ | |
3261 | pending_exact = 0; | |
e318085a | 3262 | |
25fe55af | 3263 | /* We're at the end of the group, so now we know how many |
7814e705 | 3264 | groups were inside this one. */ |
c69b0314 SM |
3265 | if (regnum <= MAX_REGNUM && regnum > 0) |
3266 | BUF_PUSH_2 (stop_memory, regnum); | |
25fe55af RS |
3267 | } |
3268 | break; | |
3269 | ||
3270 | ||
3271 | case '|': /* `\|'. */ | |
3272 | if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) | |
3273 | goto normal_backslash; | |
3274 | handle_alt: | |
3275 | if (syntax & RE_LIMITED_OPS) | |
3276 | goto normal_char; | |
3277 | ||
3278 | /* Insert before the previous alternative a jump which | |
7814e705 | 3279 | jumps to this alternative if the former fails. */ |
25fe55af RS |
3280 | GET_BUFFER_SPACE (3); |
3281 | INSERT_JUMP (on_failure_jump, begalt, b + 6); | |
3282 | pending_exact = 0; | |
3283 | b += 3; | |
3284 | ||
3285 | /* The alternative before this one has a jump after it | |
3286 | which gets executed if it gets matched. Adjust that | |
3287 | jump so it will jump to this alternative's analogous | |
3288 | jump (put in below, which in turn will jump to the next | |
3289 | (if any) alternative's such jump, etc.). The last such | |
3290 | jump jumps to the correct final destination. A picture: | |
3291 | _____ _____ | |
3292 | | | | | | |
3293 | | v | v | |
d1dfb56c | 3294 | a | b | c |
25fe55af RS |
3295 | |
3296 | If we are at `b', then fixup_alt_jump right now points to a | |
3297 | three-byte space after `a'. We'll put in the jump, set | |
3298 | fixup_alt_jump to right after `b', and leave behind three | |
3299 | bytes which we'll fill in when we get to after `c'. */ | |
3300 | ||
505bde11 | 3301 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3302 | |
3303 | /* Mark and leave space for a jump after this alternative, | |
3304 | to be filled in later either by next alternative or | |
3305 | when know we're at the end of a series of alternatives. */ | |
3306 | fixup_alt_jump = b; | |
3307 | GET_BUFFER_SPACE (3); | |
3308 | b += 3; | |
3309 | ||
3310 | laststart = 0; | |
3311 | begalt = b; | |
3312 | break; | |
3313 | ||
3314 | ||
3315 | case '{': | |
3316 | /* If \{ is a literal. */ | |
3317 | if (!(syntax & RE_INTERVALS) | |
3318 | /* If we're at `\{' and it's not the open-interval | |
3319 | operator. */ | |
4bb91c68 | 3320 | || (syntax & RE_NO_BK_BRACES)) |
25fe55af RS |
3321 | goto normal_backslash; |
3322 | ||
3323 | handle_interval: | |
3324 | { | |
3325 | /* If got here, then the syntax allows intervals. */ | |
3326 | ||
3327 | /* At least (most) this many matches must be made. */ | |
99633e97 | 3328 | int lower_bound = 0, upper_bound = -1; |
25fe55af | 3329 | |
ed0767d8 | 3330 | beg_interval = p; |
25fe55af | 3331 | |
25fe55af RS |
3332 | GET_UNSIGNED_NUMBER (lower_bound); |
3333 | ||
3334 | if (c == ',') | |
ed0767d8 | 3335 | GET_UNSIGNED_NUMBER (upper_bound); |
25fe55af RS |
3336 | else |
3337 | /* Interval such as `{1}' => match exactly once. */ | |
3338 | upper_bound = lower_bound; | |
3339 | ||
3340 | if (lower_bound < 0 || upper_bound > RE_DUP_MAX | |
ed0767d8 | 3341 | || (upper_bound >= 0 && lower_bound > upper_bound)) |
4bb91c68 | 3342 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3343 | |
3344 | if (!(syntax & RE_NO_BK_BRACES)) | |
3345 | { | |
4bb91c68 SM |
3346 | if (c != '\\') |
3347 | FREE_STACK_RETURN (REG_BADBR); | |
c72b0edd SM |
3348 | if (p == pend) |
3349 | FREE_STACK_RETURN (REG_EESCAPE); | |
25fe55af RS |
3350 | PATFETCH (c); |
3351 | } | |
3352 | ||
3353 | if (c != '}') | |
4bb91c68 | 3354 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3355 | |
3356 | /* We just parsed a valid interval. */ | |
3357 | ||
3358 | /* If it's invalid to have no preceding re. */ | |
3359 | if (!laststart) | |
3360 | { | |
3361 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
3362 | FREE_STACK_RETURN (REG_BADRPT); | |
3363 | else if (syntax & RE_CONTEXT_INDEP_OPS) | |
3364 | laststart = b; | |
3365 | else | |
3366 | goto unfetch_interval; | |
3367 | } | |
3368 | ||
6df42991 SM |
3369 | if (upper_bound == 0) |
3370 | /* If the upper bound is zero, just drop the sub pattern | |
3371 | altogether. */ | |
3372 | b = laststart; | |
3373 | else if (lower_bound == 1 && upper_bound == 1) | |
3374 | /* Just match it once: nothing to do here. */ | |
3375 | ; | |
3376 | ||
3377 | /* Otherwise, we have a nontrivial interval. When | |
3378 | we're all done, the pattern will look like: | |
3379 | set_number_at <jump count> <upper bound> | |
3380 | set_number_at <succeed_n count> <lower bound> | |
3381 | succeed_n <after jump addr> <succeed_n count> | |
3382 | <body of loop> | |
3383 | jump_n <succeed_n addr> <jump count> | |
3384 | (The upper bound and `jump_n' are omitted if | |
3385 | `upper_bound' is 1, though.) */ | |
3386 | else | |
3387 | { /* If the upper bound is > 1, we need to insert | |
3388 | more at the end of the loop. */ | |
3389 | unsigned int nbytes = (upper_bound < 0 ? 3 | |
3390 | : upper_bound > 1 ? 5 : 0); | |
3391 | unsigned int startoffset = 0; | |
3392 | ||
3393 | GET_BUFFER_SPACE (20); /* We might use less. */ | |
3394 | ||
3395 | if (lower_bound == 0) | |
3396 | { | |
3397 | /* A succeed_n that starts with 0 is really a | |
3398 | a simple on_failure_jump_loop. */ | |
3399 | INSERT_JUMP (on_failure_jump_loop, laststart, | |
3400 | b + 3 + nbytes); | |
3401 | b += 3; | |
3402 | } | |
3403 | else | |
3404 | { | |
3405 | /* Initialize lower bound of the `succeed_n', even | |
3406 | though it will be set during matching by its | |
3407 | attendant `set_number_at' (inserted next), | |
3408 | because `re_compile_fastmap' needs to know. | |
3409 | Jump to the `jump_n' we might insert below. */ | |
3410 | INSERT_JUMP2 (succeed_n, laststart, | |
3411 | b + 5 + nbytes, | |
3412 | lower_bound); | |
3413 | b += 5; | |
3414 | ||
3415 | /* Code to initialize the lower bound. Insert | |
7814e705 | 3416 | before the `succeed_n'. The `5' is the last two |
6df42991 SM |
3417 | bytes of this `set_number_at', plus 3 bytes of |
3418 | the following `succeed_n'. */ | |
3419 | insert_op2 (set_number_at, laststart, 5, lower_bound, b); | |
3420 | b += 5; | |
3421 | startoffset += 5; | |
3422 | } | |
3423 | ||
3424 | if (upper_bound < 0) | |
3425 | { | |
3426 | /* A negative upper bound stands for infinity, | |
3427 | in which case it degenerates to a plain jump. */ | |
3428 | STORE_JUMP (jump, b, laststart + startoffset); | |
3429 | b += 3; | |
3430 | } | |
3431 | else if (upper_bound > 1) | |
3432 | { /* More than one repetition is allowed, so | |
3433 | append a backward jump to the `succeed_n' | |
3434 | that starts this interval. | |
3435 | ||
3436 | When we've reached this during matching, | |
3437 | we'll have matched the interval once, so | |
3438 | jump back only `upper_bound - 1' times. */ | |
3439 | STORE_JUMP2 (jump_n, b, laststart + startoffset, | |
3440 | upper_bound - 1); | |
3441 | b += 5; | |
3442 | ||
3443 | /* The location we want to set is the second | |
3444 | parameter of the `jump_n'; that is `b-2' as | |
3445 | an absolute address. `laststart' will be | |
3446 | the `set_number_at' we're about to insert; | |
3447 | `laststart+3' the number to set, the source | |
3448 | for the relative address. But we are | |
3449 | inserting into the middle of the pattern -- | |
3450 | so everything is getting moved up by 5. | |
3451 | Conclusion: (b - 2) - (laststart + 3) + 5, | |
3452 | i.e., b - laststart. | |
3453 | ||
3454 | We insert this at the beginning of the loop | |
3455 | so that if we fail during matching, we'll | |
3456 | reinitialize the bounds. */ | |
3457 | insert_op2 (set_number_at, laststart, b - laststart, | |
3458 | upper_bound - 1, b); | |
3459 | b += 5; | |
3460 | } | |
3461 | } | |
25fe55af RS |
3462 | pending_exact = 0; |
3463 | beg_interval = NULL; | |
3464 | } | |
3465 | break; | |
3466 | ||
3467 | unfetch_interval: | |
3468 | /* If an invalid interval, match the characters as literals. */ | |
3469 | assert (beg_interval); | |
3470 | p = beg_interval; | |
3471 | beg_interval = NULL; | |
3472 | ||
3473 | /* normal_char and normal_backslash need `c'. */ | |
ed0767d8 | 3474 | c = '{'; |
25fe55af RS |
3475 | |
3476 | if (!(syntax & RE_NO_BK_BRACES)) | |
3477 | { | |
ed0767d8 SM |
3478 | assert (p > pattern && p[-1] == '\\'); |
3479 | goto normal_backslash; | |
25fe55af | 3480 | } |
ed0767d8 SM |
3481 | else |
3482 | goto normal_char; | |
e318085a | 3483 | |
b18215fc | 3484 | #ifdef emacs |
25fe55af | 3485 | /* There is no way to specify the before_dot and after_dot |
7814e705 | 3486 | operators. rms says this is ok. --karl */ |
25fe55af RS |
3487 | case '=': |
3488 | BUF_PUSH (at_dot); | |
3489 | break; | |
3490 | ||
3491 | case 's': | |
3492 | laststart = b; | |
3493 | PATFETCH (c); | |
3494 | BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); | |
3495 | break; | |
3496 | ||
3497 | case 'S': | |
3498 | laststart = b; | |
3499 | PATFETCH (c); | |
3500 | BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); | |
3501 | break; | |
b18215fc RS |
3502 | |
3503 | case 'c': | |
3504 | laststart = b; | |
36595814 | 3505 | PATFETCH (c); |
b18215fc RS |
3506 | BUF_PUSH_2 (categoryspec, c); |
3507 | break; | |
e318085a | 3508 | |
b18215fc RS |
3509 | case 'C': |
3510 | laststart = b; | |
36595814 | 3511 | PATFETCH (c); |
b18215fc RS |
3512 | BUF_PUSH_2 (notcategoryspec, c); |
3513 | break; | |
3514 | #endif /* emacs */ | |
e318085a | 3515 | |
e318085a | 3516 | |
25fe55af | 3517 | case 'w': |
4bb91c68 SM |
3518 | if (syntax & RE_NO_GNU_OPS) |
3519 | goto normal_char; | |
25fe55af | 3520 | laststart = b; |
1fb352e0 | 3521 | BUF_PUSH_2 (syntaxspec, Sword); |
25fe55af | 3522 | break; |
e318085a | 3523 | |
e318085a | 3524 | |
25fe55af | 3525 | case 'W': |
4bb91c68 SM |
3526 | if (syntax & RE_NO_GNU_OPS) |
3527 | goto normal_char; | |
25fe55af | 3528 | laststart = b; |
1fb352e0 | 3529 | BUF_PUSH_2 (notsyntaxspec, Sword); |
25fe55af | 3530 | break; |
e318085a RS |
3531 | |
3532 | ||
25fe55af | 3533 | case '<': |
4bb91c68 SM |
3534 | if (syntax & RE_NO_GNU_OPS) |
3535 | goto normal_char; | |
25fe55af RS |
3536 | BUF_PUSH (wordbeg); |
3537 | break; | |
e318085a | 3538 | |
25fe55af | 3539 | case '>': |
4bb91c68 SM |
3540 | if (syntax & RE_NO_GNU_OPS) |
3541 | goto normal_char; | |
25fe55af RS |
3542 | BUF_PUSH (wordend); |
3543 | break; | |
e318085a | 3544 | |
669fa600 SM |
3545 | case '_': |
3546 | if (syntax & RE_NO_GNU_OPS) | |
3547 | goto normal_char; | |
3548 | laststart = b; | |
3549 | PATFETCH (c); | |
3550 | if (c == '<') | |
3551 | BUF_PUSH (symbeg); | |
3552 | else if (c == '>') | |
3553 | BUF_PUSH (symend); | |
3554 | else | |
3555 | FREE_STACK_RETURN (REG_BADPAT); | |
3556 | break; | |
3557 | ||
25fe55af | 3558 | case 'b': |
4bb91c68 SM |
3559 | if (syntax & RE_NO_GNU_OPS) |
3560 | goto normal_char; | |
25fe55af RS |
3561 | BUF_PUSH (wordbound); |
3562 | break; | |
e318085a | 3563 | |
25fe55af | 3564 | case 'B': |
4bb91c68 SM |
3565 | if (syntax & RE_NO_GNU_OPS) |
3566 | goto normal_char; | |
25fe55af RS |
3567 | BUF_PUSH (notwordbound); |
3568 | break; | |
fa9a63c5 | 3569 | |
25fe55af | 3570 | case '`': |
4bb91c68 SM |
3571 | if (syntax & RE_NO_GNU_OPS) |
3572 | goto normal_char; | |
25fe55af RS |
3573 | BUF_PUSH (begbuf); |
3574 | break; | |
e318085a | 3575 | |
25fe55af | 3576 | case '\'': |
4bb91c68 SM |
3577 | if (syntax & RE_NO_GNU_OPS) |
3578 | goto normal_char; | |
25fe55af RS |
3579 | BUF_PUSH (endbuf); |
3580 | break; | |
e318085a | 3581 | |
25fe55af RS |
3582 | case '1': case '2': case '3': case '4': case '5': |
3583 | case '6': case '7': case '8': case '9': | |
0cdd06f8 SM |
3584 | { |
3585 | regnum_t reg; | |
e318085a | 3586 | |
0cdd06f8 SM |
3587 | if (syntax & RE_NO_BK_REFS) |
3588 | goto normal_backslash; | |
e318085a | 3589 | |
0cdd06f8 | 3590 | reg = c - '0'; |
e318085a | 3591 | |
c69b0314 SM |
3592 | if (reg > bufp->re_nsub || reg < 1 |
3593 | /* Can't back reference to a subexp before its end. */ | |
3594 | || group_in_compile_stack (compile_stack, reg)) | |
0cdd06f8 | 3595 | FREE_STACK_RETURN (REG_ESUBREG); |
e318085a | 3596 | |
0cdd06f8 SM |
3597 | laststart = b; |
3598 | BUF_PUSH_2 (duplicate, reg); | |
3599 | } | |
25fe55af | 3600 | break; |
e318085a | 3601 | |
e318085a | 3602 | |
25fe55af RS |
3603 | case '+': |
3604 | case '?': | |
3605 | if (syntax & RE_BK_PLUS_QM) | |
3606 | goto handle_plus; | |
3607 | else | |
3608 | goto normal_backslash; | |
3609 | ||
3610 | default: | |
3611 | normal_backslash: | |
3612 | /* You might think it would be useful for \ to mean | |
3613 | not to translate; but if we don't translate it | |
4bb91c68 | 3614 | it will never match anything. */ |
25fe55af RS |
3615 | goto normal_char; |
3616 | } | |
3617 | break; | |
fa9a63c5 RM |
3618 | |
3619 | ||
3620 | default: | |
25fe55af | 3621 | /* Expects the character in `c'. */ |
fa9a63c5 | 3622 | normal_char: |
36595814 | 3623 | /* If no exactn currently being built. */ |
25fe55af | 3624 | if (!pending_exact |
fa9a63c5 | 3625 | |
25fe55af RS |
3626 | /* If last exactn not at current position. */ |
3627 | || pending_exact + *pending_exact + 1 != b | |
5e69f11e | 3628 | |
25fe55af | 3629 | /* We have only one byte following the exactn for the count. */ |
2d1675e4 | 3630 | || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH |
fa9a63c5 | 3631 | |
7814e705 | 3632 | /* If followed by a repetition operator. */ |
9d99031f | 3633 | || (p != pend && (*p == '*' || *p == '^')) |
fa9a63c5 | 3634 | || ((syntax & RE_BK_PLUS_QM) |
9d99031f RS |
3635 | ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?') |
3636 | : p != pend && (*p == '+' || *p == '?')) | |
fa9a63c5 | 3637 | || ((syntax & RE_INTERVALS) |
25fe55af | 3638 | && ((syntax & RE_NO_BK_BRACES) |
9d99031f RS |
3639 | ? p != pend && *p == '{' |
3640 | : p + 1 < pend && p[0] == '\\' && p[1] == '{'))) | |
fa9a63c5 RM |
3641 | { |
3642 | /* Start building a new exactn. */ | |
5e69f11e | 3643 | |
25fe55af | 3644 | laststart = b; |
fa9a63c5 RM |
3645 | |
3646 | BUF_PUSH_2 (exactn, 0); | |
3647 | pending_exact = b - 1; | |
25fe55af | 3648 | } |
5e69f11e | 3649 | |
2d1675e4 SM |
3650 | GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH); |
3651 | { | |
e0277a47 KH |
3652 | int len; |
3653 | ||
cf9c99bc | 3654 | if (multibyte) |
6fdd04b0 | 3655 | { |
cf9c99bc | 3656 | c = TRANSLATE (c); |
6fdd04b0 KH |
3657 | len = CHAR_STRING (c, b); |
3658 | b += len; | |
3659 | } | |
e0277a47 | 3660 | else |
6fdd04b0 | 3661 | { |
cf9c99bc KH |
3662 | c1 = RE_CHAR_TO_MULTIBYTE (c); |
3663 | if (! CHAR_BYTE8_P (c1)) | |
3664 | { | |
3665 | re_wchar_t c2 = TRANSLATE (c1); | |
3666 | ||
3667 | if (c1 != c2 && (c1 = RE_CHAR_TO_UNIBYTE (c2)) >= 0) | |
3668 | c = c1; | |
409f2919 | 3669 | } |
6fdd04b0 KH |
3670 | *b++ = c; |
3671 | len = 1; | |
3672 | } | |
2d1675e4 SM |
3673 | (*pending_exact) += len; |
3674 | } | |
3675 | ||
fa9a63c5 | 3676 | break; |
25fe55af | 3677 | } /* switch (c) */ |
fa9a63c5 RM |
3678 | } /* while p != pend */ |
3679 | ||
5e69f11e | 3680 | |
fa9a63c5 | 3681 | /* Through the pattern now. */ |
5e69f11e | 3682 | |
505bde11 | 3683 | FIXUP_ALT_JUMP (); |
fa9a63c5 | 3684 | |
5e69f11e | 3685 | if (!COMPILE_STACK_EMPTY) |
fa9a63c5 RM |
3686 | FREE_STACK_RETURN (REG_EPAREN); |
3687 | ||
3688 | /* If we don't want backtracking, force success | |
3689 | the first time we reach the end of the compiled pattern. */ | |
3690 | if (syntax & RE_NO_POSIX_BACKTRACKING) | |
3691 | BUF_PUSH (succeed); | |
3692 | ||
fa9a63c5 RM |
3693 | /* We have succeeded; set the length of the buffer. */ |
3694 | bufp->used = b - bufp->buffer; | |
3695 | ||
3696 | #ifdef DEBUG | |
99633e97 | 3697 | if (debug > 0) |
fa9a63c5 | 3698 | { |
505bde11 | 3699 | re_compile_fastmap (bufp); |
fa9a63c5 RM |
3700 | DEBUG_PRINT1 ("\nCompiled pattern: \n"); |
3701 | print_compiled_pattern (bufp); | |
3702 | } | |
99633e97 | 3703 | debug--; |
fa9a63c5 RM |
3704 | #endif /* DEBUG */ |
3705 | ||
3706 | #ifndef MATCH_MAY_ALLOCATE | |
3707 | /* Initialize the failure stack to the largest possible stack. This | |
3708 | isn't necessary unless we're trying to avoid calling alloca in | |
3709 | the search and match routines. */ | |
3710 | { | |
3711 | int num_regs = bufp->re_nsub + 1; | |
3712 | ||
320a2a73 | 3713 | if (fail_stack.size < re_max_failures * TYPICAL_FAILURE_SIZE) |
fa9a63c5 | 3714 | { |
a26f4ccd | 3715 | fail_stack.size = re_max_failures * TYPICAL_FAILURE_SIZE; |
38182d90 PE |
3716 | falk_stack.stack = realloc (fail_stack.stack, |
3717 | fail_stack.size * sizeof *falk_stack.stack); | |
fa9a63c5 RM |
3718 | } |
3719 | ||
3720 | regex_grow_registers (num_regs); | |
3721 | } | |
3722 | #endif /* not MATCH_MAY_ALLOCATE */ | |
3723 | ||
839966f3 | 3724 | FREE_STACK_RETURN (REG_NOERROR); |
fa9a63c5 RM |
3725 | } /* regex_compile */ |
3726 | \f | |
3727 | /* Subroutines for `regex_compile'. */ | |
3728 | ||
7814e705 | 3729 | /* Store OP at LOC followed by two-byte integer parameter ARG. */ |
fa9a63c5 RM |
3730 | |
3731 | static void | |
971de7fb | 3732 | store_op1 (re_opcode_t op, unsigned char *loc, int arg) |
fa9a63c5 RM |
3733 | { |
3734 | *loc = (unsigned char) op; | |
3735 | STORE_NUMBER (loc + 1, arg); | |
3736 | } | |
3737 | ||
3738 | ||
3739 | /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3740 | ||
3741 | static void | |
971de7fb | 3742 | store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2) |
fa9a63c5 RM |
3743 | { |
3744 | *loc = (unsigned char) op; | |
3745 | STORE_NUMBER (loc + 1, arg1); | |
3746 | STORE_NUMBER (loc + 3, arg2); | |
3747 | } | |
3748 | ||
3749 | ||
3750 | /* Copy the bytes from LOC to END to open up three bytes of space at LOC | |
3751 | for OP followed by two-byte integer parameter ARG. */ | |
3752 | ||
3753 | static void | |
971de7fb | 3754 | insert_op1 (re_opcode_t op, unsigned char *loc, int arg, unsigned char *end) |
fa9a63c5 RM |
3755 | { |
3756 | register unsigned char *pfrom = end; | |
3757 | register unsigned char *pto = end + 3; | |
3758 | ||
3759 | while (pfrom != loc) | |
3760 | *--pto = *--pfrom; | |
5e69f11e | 3761 | |
fa9a63c5 RM |
3762 | store_op1 (op, loc, arg); |
3763 | } | |
3764 | ||
3765 | ||
3766 | /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3767 | ||
3768 | static void | |
971de7fb | 3769 | insert_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2, unsigned char *end) |
fa9a63c5 RM |
3770 | { |
3771 | register unsigned char *pfrom = end; | |
3772 | register unsigned char *pto = end + 5; | |
3773 | ||
3774 | while (pfrom != loc) | |
3775 | *--pto = *--pfrom; | |
5e69f11e | 3776 | |
fa9a63c5 RM |
3777 | store_op2 (op, loc, arg1, arg2); |
3778 | } | |
3779 | ||
3780 | ||
3781 | /* P points to just after a ^ in PATTERN. Return true if that ^ comes | |
3782 | after an alternative or a begin-subexpression. We assume there is at | |
3783 | least one character before the ^. */ | |
3784 | ||
3785 | static boolean | |
971de7fb | 3786 | at_begline_loc_p (const re_char *pattern, const re_char *p, reg_syntax_t syntax) |
fa9a63c5 | 3787 | { |
01618498 | 3788 | re_char *prev = p - 2; |
95988fcf | 3789 | boolean odd_backslashes; |
5e69f11e | 3790 | |
95988fcf AS |
3791 | /* After a subexpression? */ |
3792 | if (*prev == '(') | |
3793 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3794 | ||
3795 | /* After an alternative? */ | |
3796 | else if (*prev == '|') | |
3797 | odd_backslashes = (syntax & RE_NO_BK_VBAR) == 0; | |
3798 | ||
3799 | /* After a shy subexpression? */ | |
3800 | else if (*prev == ':' && (syntax & RE_SHY_GROUPS)) | |
3801 | { | |
3802 | /* Skip over optional regnum. */ | |
3803 | while (prev - 1 >= pattern && prev[-1] >= '0' && prev[-1] <= '9') | |
3804 | --prev; | |
3805 | ||
3806 | if (!(prev - 2 >= pattern | |
3807 | && prev[-1] == '?' && prev[-2] == '(')) | |
3808 | return false; | |
3809 | prev -= 2; | |
3810 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3811 | } | |
3812 | else | |
3813 | return false; | |
3814 | ||
3815 | /* Count the number of preceding backslashes. */ | |
3816 | p = prev; | |
3817 | while (prev - 1 >= pattern && prev[-1] == '\\') | |
3818 | --prev; | |
3819 | return (p - prev) & odd_backslashes; | |
fa9a63c5 RM |
3820 | } |
3821 | ||
3822 | ||
3823 | /* The dual of at_begline_loc_p. This one is for $. We assume there is | |
3824 | at least one character after the $, i.e., `P < PEND'. */ | |
3825 | ||
3826 | static boolean | |
971de7fb | 3827 | at_endline_loc_p (const re_char *p, const re_char *pend, reg_syntax_t syntax) |
fa9a63c5 | 3828 | { |
01618498 | 3829 | re_char *next = p; |
fa9a63c5 | 3830 | boolean next_backslash = *next == '\\'; |
01618498 | 3831 | re_char *next_next = p + 1 < pend ? p + 1 : 0; |
5e69f11e | 3832 | |
fa9a63c5 RM |
3833 | return |
3834 | /* Before a subexpression? */ | |
3835 | (syntax & RE_NO_BK_PARENS ? *next == ')' | |
25fe55af | 3836 | : next_backslash && next_next && *next_next == ')') |
fa9a63c5 RM |
3837 | /* Before an alternative? */ |
3838 | || (syntax & RE_NO_BK_VBAR ? *next == '|' | |
25fe55af | 3839 | : next_backslash && next_next && *next_next == '|'); |
fa9a63c5 RM |
3840 | } |
3841 | ||
3842 | ||
5e69f11e | 3843 | /* Returns true if REGNUM is in one of COMPILE_STACK's elements and |
fa9a63c5 RM |
3844 | false if it's not. */ |
3845 | ||
3846 | static boolean | |
971de7fb | 3847 | group_in_compile_stack (compile_stack_type compile_stack, regnum_t regnum) |
fa9a63c5 | 3848 | { |
d1dfb56c | 3849 | ssize_t this_element; |
fa9a63c5 | 3850 | |
5e69f11e RM |
3851 | for (this_element = compile_stack.avail - 1; |
3852 | this_element >= 0; | |
fa9a63c5 RM |
3853 | this_element--) |
3854 | if (compile_stack.stack[this_element].regnum == regnum) | |
3855 | return true; | |
3856 | ||
3857 | return false; | |
3858 | } | |
fa9a63c5 | 3859 | \f |
f6a3f532 SM |
3860 | /* analyse_first. |
3861 | If fastmap is non-NULL, go through the pattern and fill fastmap | |
3862 | with all the possible leading chars. If fastmap is NULL, don't | |
3863 | bother filling it up (obviously) and only return whether the | |
3864 | pattern could potentially match the empty string. | |
3865 | ||
3866 | Return 1 if p..pend might match the empty string. | |
3867 | Return 0 if p..pend matches at least one char. | |
01618498 | 3868 | Return -1 if fastmap was not updated accurately. */ |
f6a3f532 SM |
3869 | |
3870 | static int | |
438105ed | 3871 | analyse_first (const re_char *p, const re_char *pend, char *fastmap, const int multibyte) |
fa9a63c5 | 3872 | { |
505bde11 | 3873 | int j, k; |
1fb352e0 | 3874 | boolean not; |
fa9a63c5 | 3875 | |
b18215fc | 3876 | /* If all elements for base leading-codes in fastmap is set, this |
7814e705 | 3877 | flag is set true. */ |
b18215fc RS |
3878 | boolean match_any_multibyte_characters = false; |
3879 | ||
f6a3f532 | 3880 | assert (p); |
5e69f11e | 3881 | |
505bde11 SM |
3882 | /* The loop below works as follows: |
3883 | - It has a working-list kept in the PATTERN_STACK and which basically | |
3884 | starts by only containing a pointer to the first operation. | |
3885 | - If the opcode we're looking at is a match against some set of | |
3886 | chars, then we add those chars to the fastmap and go on to the | |
3887 | next work element from the worklist (done via `break'). | |
3888 | - If the opcode is a control operator on the other hand, we either | |
3889 | ignore it (if it's meaningless at this point, such as `start_memory') | |
3890 | or execute it (if it's a jump). If the jump has several destinations | |
3891 | (i.e. `on_failure_jump'), then we push the other destination onto the | |
3892 | worklist. | |
3893 | We guarantee termination by ignoring backward jumps (more or less), | |
3894 | so that `p' is monotonically increasing. More to the point, we | |
3895 | never set `p' (or push) anything `<= p1'. */ | |
3896 | ||
01618498 | 3897 | while (p < pend) |
fa9a63c5 | 3898 | { |
505bde11 SM |
3899 | /* `p1' is used as a marker of how far back a `on_failure_jump' |
3900 | can go without being ignored. It is normally equal to `p' | |
3901 | (which prevents any backward `on_failure_jump') except right | |
3902 | after a plain `jump', to allow patterns such as: | |
3903 | 0: jump 10 | |
3904 | 3..9: <body> | |
3905 | 10: on_failure_jump 3 | |
3906 | as used for the *? operator. */ | |
01618498 | 3907 | re_char *p1 = p; |
5e69f11e | 3908 | |
7393bcbb | 3909 | switch (*p++) |
fa9a63c5 | 3910 | { |
f6a3f532 | 3911 | case succeed: |
01618498 | 3912 | return 1; |
fa9a63c5 | 3913 | |
fa9a63c5 | 3914 | case duplicate: |
505bde11 SM |
3915 | /* If the first character has to match a backreference, that means |
3916 | that the group was empty (since it already matched). Since this | |
3917 | is the only case that interests us here, we can assume that the | |
3918 | backreference must match the empty string. */ | |
3919 | p++; | |
3920 | continue; | |
fa9a63c5 RM |
3921 | |
3922 | ||
3923 | /* Following are the cases which match a character. These end | |
7814e705 | 3924 | with `break'. */ |
fa9a63c5 RM |
3925 | |
3926 | case exactn: | |
e0277a47 | 3927 | if (fastmap) |
cf9c99bc KH |
3928 | { |
3929 | /* If multibyte is nonzero, the first byte of each | |
3930 | character is an ASCII or a leading code. Otherwise, | |
3931 | each byte is a character. Thus, this works in both | |
3932 | cases. */ | |
3933 | fastmap[p[1]] = 1; | |
3934 | if (! multibyte) | |
3935 | { | |
3936 | /* For the case of matching this unibyte regex | |
3937 | against multibyte, we must set a leading code of | |
3938 | the corresponding multibyte character. */ | |
3939 | int c = RE_CHAR_TO_MULTIBYTE (p[1]); | |
3940 | ||
86e893e3 | 3941 | fastmap[CHAR_LEADING_CODE (c)] = 1; |
cf9c99bc KH |
3942 | } |
3943 | } | |
fa9a63c5 RM |
3944 | break; |
3945 | ||
3946 | ||
1fb352e0 SM |
3947 | case anychar: |
3948 | /* We could put all the chars except for \n (and maybe \0) | |
3949 | but we don't bother since it is generally not worth it. */ | |
f6a3f532 | 3950 | if (!fastmap) break; |
01618498 | 3951 | return -1; |
fa9a63c5 RM |
3952 | |
3953 | ||
b18215fc | 3954 | case charset_not: |
1fb352e0 | 3955 | if (!fastmap) break; |
bf216479 KH |
3956 | { |
3957 | /* Chars beyond end of bitmap are possible matches. */ | |
bf216479 | 3958 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH; |
cf9c99bc | 3959 | j < (1 << BYTEWIDTH); j++) |
bf216479 KH |
3960 | fastmap[j] = 1; |
3961 | } | |
3962 | ||
1fb352e0 SM |
3963 | /* Fallthrough */ |
3964 | case charset: | |
3965 | if (!fastmap) break; | |
3966 | not = (re_opcode_t) *(p - 1) == charset_not; | |
b18215fc RS |
3967 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH - 1, p++; |
3968 | j >= 0; j--) | |
1fb352e0 | 3969 | if (!!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) ^ not) |
49da453b | 3970 | fastmap[j] = 1; |
b18215fc | 3971 | |
6482db2e KH |
3972 | #ifdef emacs |
3973 | if (/* Any leading code can possibly start a character | |
1fb352e0 | 3974 | which doesn't match the specified set of characters. */ |
6482db2e | 3975 | not |
409f2919 | 3976 | || |
6482db2e KH |
3977 | /* If we can match a character class, we can match any |
3978 | multibyte characters. */ | |
3979 | (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) | |
3980 | && CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0)) | |
3981 | ||
b18215fc | 3982 | { |
b18215fc RS |
3983 | if (match_any_multibyte_characters == false) |
3984 | { | |
6482db2e KH |
3985 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
3986 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
6fdd04b0 | 3987 | fastmap[j] = 1; |
b18215fc RS |
3988 | match_any_multibyte_characters = true; |
3989 | } | |
3990 | } | |
b18215fc | 3991 | |
1fb352e0 SM |
3992 | else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) |
3993 | && match_any_multibyte_characters == false) | |
3994 | { | |
bf216479 | 3995 | /* Set fastmap[I] to 1 where I is a leading code of each |
51e4f4a8 | 3996 | multibyte character in the range table. */ |
1fb352e0 | 3997 | int c, count; |
bf216479 | 3998 | unsigned char lc1, lc2; |
b18215fc | 3999 | |
1fb352e0 | 4000 | /* Make P points the range table. `+ 2' is to skip flag |
0b32bf0e | 4001 | bits for a character class. */ |
1fb352e0 | 4002 | p += CHARSET_BITMAP_SIZE (&p[-2]) + 2; |
b18215fc | 4003 | |
1fb352e0 SM |
4004 | /* Extract the number of ranges in range table into COUNT. */ |
4005 | EXTRACT_NUMBER_AND_INCR (count, p); | |
cf9c99bc | 4006 | for (; count > 0; count--, p += 3) |
1fb352e0 | 4007 | { |
9117d724 KH |
4008 | /* Extract the start and end of each range. */ |
4009 | EXTRACT_CHARACTER (c, p); | |
bf216479 | 4010 | lc1 = CHAR_LEADING_CODE (c); |
9117d724 | 4011 | p += 3; |
1fb352e0 | 4012 | EXTRACT_CHARACTER (c, p); |
bf216479 KH |
4013 | lc2 = CHAR_LEADING_CODE (c); |
4014 | for (j = lc1; j <= lc2; j++) | |
9117d724 | 4015 | fastmap[j] = 1; |
1fb352e0 SM |
4016 | } |
4017 | } | |
6482db2e | 4018 | #endif |
b18215fc RS |
4019 | break; |
4020 | ||
1fb352e0 SM |
4021 | case syntaxspec: |
4022 | case notsyntaxspec: | |
4023 | if (!fastmap) break; | |
4024 | #ifndef emacs | |
4025 | not = (re_opcode_t)p[-1] == notsyntaxspec; | |
4026 | k = *p++; | |
4027 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
990b2375 | 4028 | if ((SYNTAX (j) == (enum syntaxcode) k) ^ not) |
b18215fc | 4029 | fastmap[j] = 1; |
b18215fc | 4030 | break; |
1fb352e0 | 4031 | #else /* emacs */ |
b18215fc RS |
4032 | /* This match depends on text properties. These end with |
4033 | aborting optimizations. */ | |
01618498 | 4034 | return -1; |
b18215fc RS |
4035 | |
4036 | case categoryspec: | |
b18215fc | 4037 | case notcategoryspec: |
1fb352e0 SM |
4038 | if (!fastmap) break; |
4039 | not = (re_opcode_t)p[-1] == notcategoryspec; | |
b18215fc | 4040 | k = *p++; |
6482db2e | 4041 | for (j = (1 << BYTEWIDTH); j >= 0; j--) |
1fb352e0 | 4042 | if ((CHAR_HAS_CATEGORY (j, k)) ^ not) |
b18215fc RS |
4043 | fastmap[j] = 1; |
4044 | ||
6482db2e KH |
4045 | /* Any leading code can possibly start a character which |
4046 | has or doesn't has the specified category. */ | |
4047 | if (match_any_multibyte_characters == false) | |
6fdd04b0 | 4048 | { |
6482db2e KH |
4049 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
4050 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
4051 | fastmap[j] = 1; | |
4052 | match_any_multibyte_characters = true; | |
6fdd04b0 | 4053 | } |
b18215fc RS |
4054 | break; |
4055 | ||
fa9a63c5 | 4056 | /* All cases after this match the empty string. These end with |
25fe55af | 4057 | `continue'. */ |
fa9a63c5 | 4058 | |
fa9a63c5 RM |
4059 | case before_dot: |
4060 | case at_dot: | |
4061 | case after_dot: | |
1fb352e0 | 4062 | #endif /* !emacs */ |
25fe55af RS |
4063 | case no_op: |
4064 | case begline: | |
4065 | case endline: | |
fa9a63c5 RM |
4066 | case begbuf: |
4067 | case endbuf: | |
4068 | case wordbound: | |
4069 | case notwordbound: | |
4070 | case wordbeg: | |
4071 | case wordend: | |
669fa600 SM |
4072 | case symbeg: |
4073 | case symend: | |
25fe55af | 4074 | continue; |
fa9a63c5 RM |
4075 | |
4076 | ||
fa9a63c5 | 4077 | case jump: |
25fe55af | 4078 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 SM |
4079 | if (j < 0) |
4080 | /* Backward jumps can only go back to code that we've already | |
4081 | visited. `re_compile' should make sure this is true. */ | |
4082 | break; | |
25fe55af | 4083 | p += j; |
7393bcbb | 4084 | switch (*p) |
505bde11 SM |
4085 | { |
4086 | case on_failure_jump: | |
4087 | case on_failure_keep_string_jump: | |
505bde11 | 4088 | case on_failure_jump_loop: |
0683b6fa | 4089 | case on_failure_jump_nastyloop: |
505bde11 SM |
4090 | case on_failure_jump_smart: |
4091 | p++; | |
4092 | break; | |
4093 | default: | |
4094 | continue; | |
4095 | }; | |
4096 | /* Keep `p1' to allow the `on_failure_jump' we are jumping to | |
4097 | to jump back to "just after here". */ | |
4098 | /* Fallthrough */ | |
fa9a63c5 | 4099 | |
25fe55af RS |
4100 | case on_failure_jump: |
4101 | case on_failure_keep_string_jump: | |
0683b6fa | 4102 | case on_failure_jump_nastyloop: |
505bde11 SM |
4103 | case on_failure_jump_loop: |
4104 | case on_failure_jump_smart: | |
25fe55af | 4105 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 | 4106 | if (p + j <= p1) |
ed0767d8 | 4107 | ; /* Backward jump to be ignored. */ |
01618498 SM |
4108 | else |
4109 | { /* We have to look down both arms. | |
4110 | We first go down the "straight" path so as to minimize | |
4111 | stack usage when going through alternatives. */ | |
4112 | int r = analyse_first (p, pend, fastmap, multibyte); | |
4113 | if (r) return r; | |
4114 | p += j; | |
4115 | } | |
25fe55af | 4116 | continue; |
fa9a63c5 RM |
4117 | |
4118 | ||
ed0767d8 SM |
4119 | case jump_n: |
4120 | /* This code simply does not properly handle forward jump_n. */ | |
4121 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p); assert (j < 0)); | |
4122 | p += 4; | |
4123 | /* jump_n can either jump or fall through. The (backward) jump | |
4124 | case has already been handled, so we only need to look at the | |
4125 | fallthrough case. */ | |
4126 | continue; | |
177c0ea7 | 4127 | |
fa9a63c5 | 4128 | case succeed_n: |
ed0767d8 SM |
4129 | /* If N == 0, it should be an on_failure_jump_loop instead. */ |
4130 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0)); | |
4131 | p += 4; | |
4132 | /* We only care about one iteration of the loop, so we don't | |
4133 | need to consider the case where this behaves like an | |
4134 | on_failure_jump. */ | |
25fe55af | 4135 | continue; |
fa9a63c5 RM |
4136 | |
4137 | ||
4138 | case set_number_at: | |
25fe55af RS |
4139 | p += 4; |
4140 | continue; | |
fa9a63c5 RM |
4141 | |
4142 | ||
4143 | case start_memory: | |
25fe55af | 4144 | case stop_memory: |
505bde11 | 4145 | p += 1; |
fa9a63c5 RM |
4146 | continue; |
4147 | ||
4148 | ||
4149 | default: | |
25fe55af RS |
4150 | abort (); /* We have listed all the cases. */ |
4151 | } /* switch *p++ */ | |
fa9a63c5 RM |
4152 | |
4153 | /* Getting here means we have found the possible starting | |
25fe55af | 4154 | characters for one path of the pattern -- and that the empty |
7814e705 | 4155 | string does not match. We need not follow this path further. */ |
01618498 | 4156 | return 0; |
fa9a63c5 RM |
4157 | } /* while p */ |
4158 | ||
01618498 SM |
4159 | /* We reached the end without matching anything. */ |
4160 | return 1; | |
4161 | ||
f6a3f532 SM |
4162 | } /* analyse_first */ |
4163 | \f | |
4164 | /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in | |
4165 | BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible | |
4166 | characters can start a string that matches the pattern. This fastmap | |
4167 | is used by re_search to skip quickly over impossible starting points. | |
4168 | ||
4169 | Character codes above (1 << BYTEWIDTH) are not represented in the | |
4170 | fastmap, but the leading codes are represented. Thus, the fastmap | |
4171 | indicates which character sets could start a match. | |
4172 | ||
4173 | The caller must supply the address of a (1 << BYTEWIDTH)-byte data | |
4174 | area as BUFP->fastmap. | |
4175 | ||
4176 | We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in | |
4177 | the pattern buffer. | |
4178 | ||
4179 | Returns 0 if we succeed, -2 if an internal error. */ | |
4180 | ||
4181 | int | |
971de7fb | 4182 | re_compile_fastmap (struct re_pattern_buffer *bufp) |
f6a3f532 SM |
4183 | { |
4184 | char *fastmap = bufp->fastmap; | |
4185 | int analysis; | |
4186 | ||
4187 | assert (fastmap && bufp->buffer); | |
4188 | ||
72af86bd | 4189 | memset (fastmap, 0, 1 << BYTEWIDTH); /* Assume nothing's valid. */ |
f6a3f532 SM |
4190 | bufp->fastmap_accurate = 1; /* It will be when we're done. */ |
4191 | ||
4192 | analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used, | |
2d1675e4 | 4193 | fastmap, RE_MULTIBYTE_P (bufp)); |
c0f9ea08 | 4194 | bufp->can_be_null = (analysis != 0); |
fa9a63c5 RM |
4195 | return 0; |
4196 | } /* re_compile_fastmap */ | |
4197 | \f | |
4198 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | |
4199 | ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | |
4200 | this memory for recording register information. STARTS and ENDS | |
4201 | must be allocated using the malloc library routine, and must each | |
4202 | be at least NUM_REGS * sizeof (regoff_t) bytes long. | |
4203 | ||
4204 | If NUM_REGS == 0, then subsequent matches should allocate their own | |
4205 | register data. | |
4206 | ||
4207 | Unless this function is called, the first search or match using | |
4208 | PATTERN_BUFFER will allocate its own register data, without | |
4209 | freeing the old data. */ | |
4210 | ||
4211 | void | |
971de7fb | 4212 | re_set_registers (struct re_pattern_buffer *bufp, struct re_registers *regs, unsigned int num_regs, regoff_t *starts, regoff_t *ends) |
fa9a63c5 RM |
4213 | { |
4214 | if (num_regs) | |
4215 | { | |
4216 | bufp->regs_allocated = REGS_REALLOCATE; | |
4217 | regs->num_regs = num_regs; | |
4218 | regs->start = starts; | |
4219 | regs->end = ends; | |
4220 | } | |
4221 | else | |
4222 | { | |
4223 | bufp->regs_allocated = REGS_UNALLOCATED; | |
4224 | regs->num_regs = 0; | |
4225 | regs->start = regs->end = (regoff_t *) 0; | |
4226 | } | |
4227 | } | |
c0f9ea08 | 4228 | WEAK_ALIAS (__re_set_registers, re_set_registers) |
fa9a63c5 | 4229 | \f |
7814e705 | 4230 | /* Searching routines. */ |
fa9a63c5 RM |
4231 | |
4232 | /* Like re_search_2, below, but only one string is specified, and | |
4233 | doesn't let you say where to stop matching. */ | |
4234 | ||
d1dfb56c EZ |
4235 | regoff_t |
4236 | re_search (struct re_pattern_buffer *bufp, const char *string, size_t size, | |
4237 | ssize_t startpos, ssize_t range, struct re_registers *regs) | |
fa9a63c5 | 4238 | { |
5e69f11e | 4239 | return re_search_2 (bufp, NULL, 0, string, size, startpos, range, |
fa9a63c5 RM |
4240 | regs, size); |
4241 | } | |
c0f9ea08 | 4242 | WEAK_ALIAS (__re_search, re_search) |
fa9a63c5 | 4243 | |
70806df6 KH |
4244 | /* Head address of virtual concatenation of string. */ |
4245 | #define HEAD_ADDR_VSTRING(P) \ | |
4246 | (((P) >= size1 ? string2 : string1)) | |
4247 | ||
b18215fc RS |
4248 | /* Address of POS in the concatenation of virtual string. */ |
4249 | #define POS_ADDR_VSTRING(POS) \ | |
4250 | (((POS) >= size1 ? string2 - size1 : string1) + (POS)) | |
fa9a63c5 RM |
4251 | |
4252 | /* Using the compiled pattern in BUFP->buffer, first tries to match the | |
4253 | virtual concatenation of STRING1 and STRING2, starting first at index | |
4254 | STARTPOS, then at STARTPOS + 1, and so on. | |
5e69f11e | 4255 | |
fa9a63c5 | 4256 | STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. |
5e69f11e | 4257 | |
fa9a63c5 RM |
4258 | RANGE is how far to scan while trying to match. RANGE = 0 means try |
4259 | only at STARTPOS; in general, the last start tried is STARTPOS + | |
4260 | RANGE. | |
5e69f11e | 4261 | |
fa9a63c5 RM |
4262 | In REGS, return the indices of the virtual concatenation of STRING1 |
4263 | and STRING2 that matched the entire BUFP->buffer and its contained | |
4264 | subexpressions. | |
5e69f11e | 4265 | |
fa9a63c5 RM |
4266 | Do not consider matching one past the index STOP in the virtual |
4267 | concatenation of STRING1 and STRING2. | |
4268 | ||
4269 | We return either the position in the strings at which the match was | |
4270 | found, -1 if no match, or -2 if error (such as failure | |
4271 | stack overflow). */ | |
4272 | ||
d1dfb56c EZ |
4273 | regoff_t |
4274 | re_search_2 (struct re_pattern_buffer *bufp, const char *str1, size_t size1, | |
4275 | const char *str2, size_t size2, ssize_t startpos, ssize_t range, | |
4276 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4277 | { |
d1dfb56c | 4278 | regoff_t val; |
66f0296e SM |
4279 | re_char *string1 = (re_char*) str1; |
4280 | re_char *string2 = (re_char*) str2; | |
fa9a63c5 | 4281 | register char *fastmap = bufp->fastmap; |
6676cb1c | 4282 | register RE_TRANSLATE_TYPE translate = bufp->translate; |
d1dfb56c EZ |
4283 | size_t total_size = size1 + size2; |
4284 | ssize_t endpos = startpos + range; | |
c0f9ea08 | 4285 | boolean anchored_start; |
cf9c99bc KH |
4286 | /* Nonzero if we are searching multibyte string. */ |
4287 | const boolean multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
b18215fc | 4288 | |
fa9a63c5 RM |
4289 | /* Check for out-of-range STARTPOS. */ |
4290 | if (startpos < 0 || startpos > total_size) | |
4291 | return -1; | |
5e69f11e | 4292 | |
fa9a63c5 | 4293 | /* Fix up RANGE if it might eventually take us outside |
34597fa9 | 4294 | the virtual concatenation of STRING1 and STRING2. |
5e69f11e | 4295 | Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ |
34597fa9 RS |
4296 | if (endpos < 0) |
4297 | range = 0 - startpos; | |
fa9a63c5 RM |
4298 | else if (endpos > total_size) |
4299 | range = total_size - startpos; | |
4300 | ||
4301 | /* If the search isn't to be a backwards one, don't waste time in a | |
7b140fd7 | 4302 | search for a pattern anchored at beginning of buffer. */ |
fa9a63c5 RM |
4303 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) |
4304 | { | |
4305 | if (startpos > 0) | |
4306 | return -1; | |
4307 | else | |
7b140fd7 | 4308 | range = 0; |
fa9a63c5 RM |
4309 | } |
4310 | ||
ae4788a8 RS |
4311 | #ifdef emacs |
4312 | /* In a forward search for something that starts with \=. | |
4313 | don't keep searching past point. */ | |
4314 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) | |
4315 | { | |
7b140fd7 RS |
4316 | range = PT_BYTE - BEGV_BYTE - startpos; |
4317 | if (range < 0) | |
ae4788a8 RS |
4318 | return -1; |
4319 | } | |
4320 | #endif /* emacs */ | |
4321 | ||
fa9a63c5 RM |
4322 | /* Update the fastmap now if not correct already. */ |
4323 | if (fastmap && !bufp->fastmap_accurate) | |
01618498 | 4324 | re_compile_fastmap (bufp); |
5e69f11e | 4325 | |
c8499ba5 | 4326 | /* See whether the pattern is anchored. */ |
c0f9ea08 | 4327 | anchored_start = (bufp->buffer[0] == begline); |
c8499ba5 | 4328 | |
b18215fc | 4329 | #ifdef emacs |
d48cd3f4 | 4330 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
cc9b4df2 | 4331 | { |
d1dfb56c | 4332 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (startpos)); |
cc9b4df2 KH |
4333 | |
4334 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); | |
4335 | } | |
b18215fc RS |
4336 | #endif |
4337 | ||
fa9a63c5 RM |
4338 | /* Loop through the string, looking for a place to start matching. */ |
4339 | for (;;) | |
5e69f11e | 4340 | { |
c8499ba5 RS |
4341 | /* If the pattern is anchored, |
4342 | skip quickly past places we cannot match. | |
4343 | We don't bother to treat startpos == 0 specially | |
4344 | because that case doesn't repeat. */ | |
4345 | if (anchored_start && startpos > 0) | |
4346 | { | |
c0f9ea08 SM |
4347 | if (! ((startpos <= size1 ? string1[startpos - 1] |
4348 | : string2[startpos - size1 - 1]) | |
4349 | == '\n')) | |
c8499ba5 RS |
4350 | goto advance; |
4351 | } | |
4352 | ||
fa9a63c5 | 4353 | /* If a fastmap is supplied, skip quickly over characters that |
25fe55af RS |
4354 | cannot be the start of a match. If the pattern can match the |
4355 | null string, however, we don't need to skip characters; we want | |
7814e705 | 4356 | the first null string. */ |
fa9a63c5 RM |
4357 | if (fastmap && startpos < total_size && !bufp->can_be_null) |
4358 | { | |
66f0296e | 4359 | register re_char *d; |
01618498 | 4360 | register re_wchar_t buf_ch; |
e934739e RS |
4361 | |
4362 | d = POS_ADDR_VSTRING (startpos); | |
4363 | ||
7814e705 | 4364 | if (range > 0) /* Searching forwards. */ |
fa9a63c5 | 4365 | { |
fa9a63c5 | 4366 | register int lim = 0; |
d1dfb56c | 4367 | ssize_t irange = range; |
fa9a63c5 | 4368 | |
25fe55af RS |
4369 | if (startpos < size1 && startpos + range >= size1) |
4370 | lim = range - (size1 - startpos); | |
fa9a63c5 | 4371 | |
25fe55af RS |
4372 | /* Written out as an if-else to avoid testing `translate' |
4373 | inside the loop. */ | |
28ae27ae AS |
4374 | if (RE_TRANSLATE_P (translate)) |
4375 | { | |
e934739e RS |
4376 | if (multibyte) |
4377 | while (range > lim) | |
4378 | { | |
4379 | int buf_charlen; | |
4380 | ||
62a6e103 | 4381 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 4382 | buf_ch = RE_TRANSLATE (translate, buf_ch); |
bf216479 | 4383 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
e934739e RS |
4384 | break; |
4385 | ||
4386 | range -= buf_charlen; | |
4387 | d += buf_charlen; | |
4388 | } | |
4389 | else | |
bf216479 | 4390 | while (range > lim) |
33c46939 | 4391 | { |
cf9c99bc KH |
4392 | register re_wchar_t ch, translated; |
4393 | ||
bf216479 | 4394 | buf_ch = *d; |
cf9c99bc KH |
4395 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); |
4396 | translated = RE_TRANSLATE (translate, ch); | |
4397 | if (translated != ch | |
4398 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4399 | buf_ch = ch; | |
6fdd04b0 | 4400 | if (fastmap[buf_ch]) |
bf216479 | 4401 | break; |
33c46939 RS |
4402 | d++; |
4403 | range--; | |
4404 | } | |
e934739e | 4405 | } |
fa9a63c5 | 4406 | else |
6fdd04b0 KH |
4407 | { |
4408 | if (multibyte) | |
4409 | while (range > lim) | |
4410 | { | |
4411 | int buf_charlen; | |
fa9a63c5 | 4412 | |
62a6e103 | 4413 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
6fdd04b0 KH |
4414 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
4415 | break; | |
4416 | range -= buf_charlen; | |
4417 | d += buf_charlen; | |
4418 | } | |
e934739e | 4419 | else |
6fdd04b0 | 4420 | while (range > lim && !fastmap[*d]) |
33c46939 RS |
4421 | { |
4422 | d++; | |
4423 | range--; | |
4424 | } | |
e934739e | 4425 | } |
fa9a63c5 RM |
4426 | startpos += irange - range; |
4427 | } | |
7814e705 | 4428 | else /* Searching backwards. */ |
fa9a63c5 | 4429 | { |
ba5e343c KH |
4430 | if (multibyte) |
4431 | { | |
62a6e103 | 4432 | buf_ch = STRING_CHAR (d); |
ba5e343c KH |
4433 | buf_ch = TRANSLATE (buf_ch); |
4434 | if (! fastmap[CHAR_LEADING_CODE (buf_ch)]) | |
4435 | goto advance; | |
4436 | } | |
4437 | else | |
4438 | { | |
cf9c99bc KH |
4439 | register re_wchar_t ch, translated; |
4440 | ||
4441 | buf_ch = *d; | |
4442 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); | |
4443 | translated = TRANSLATE (ch); | |
4444 | if (translated != ch | |
4445 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4446 | buf_ch = ch; | |
4447 | if (! fastmap[TRANSLATE (buf_ch)]) | |
ba5e343c KH |
4448 | goto advance; |
4449 | } | |
fa9a63c5 RM |
4450 | } |
4451 | } | |
4452 | ||
4453 | /* If can't match the null string, and that's all we have left, fail. */ | |
4454 | if (range >= 0 && startpos == total_size && fastmap | |
25fe55af | 4455 | && !bufp->can_be_null) |
fa9a63c5 RM |
4456 | return -1; |
4457 | ||
4458 | val = re_match_2_internal (bufp, string1, size1, string2, size2, | |
4459 | startpos, regs, stop); | |
fa9a63c5 RM |
4460 | |
4461 | if (val >= 0) | |
4462 | return startpos; | |
5e69f11e | 4463 | |
fa9a63c5 RM |
4464 | if (val == -2) |
4465 | return -2; | |
4466 | ||
4467 | advance: | |
5e69f11e | 4468 | if (!range) |
25fe55af | 4469 | break; |
5e69f11e | 4470 | else if (range > 0) |
25fe55af | 4471 | { |
b18215fc RS |
4472 | /* Update STARTPOS to the next character boundary. */ |
4473 | if (multibyte) | |
4474 | { | |
66f0296e | 4475 | re_char *p = POS_ADDR_VSTRING (startpos); |
aa3830c4 | 4476 | int len = BYTES_BY_CHAR_HEAD (*p); |
b18215fc RS |
4477 | |
4478 | range -= len; | |
4479 | if (range < 0) | |
4480 | break; | |
4481 | startpos += len; | |
4482 | } | |
4483 | else | |
4484 | { | |
b560c397 RS |
4485 | range--; |
4486 | startpos++; | |
4487 | } | |
e318085a | 4488 | } |
fa9a63c5 | 4489 | else |
25fe55af RS |
4490 | { |
4491 | range++; | |
4492 | startpos--; | |
b18215fc RS |
4493 | |
4494 | /* Update STARTPOS to the previous character boundary. */ | |
4495 | if (multibyte) | |
4496 | { | |
70806df6 KH |
4497 | re_char *p = POS_ADDR_VSTRING (startpos) + 1; |
4498 | re_char *p0 = p; | |
4499 | re_char *phead = HEAD_ADDR_VSTRING (startpos); | |
b18215fc RS |
4500 | |
4501 | /* Find the head of multibyte form. */ | |
70806df6 KH |
4502 | PREV_CHAR_BOUNDARY (p, phead); |
4503 | range += p0 - 1 - p; | |
4504 | if (range > 0) | |
4505 | break; | |
b18215fc | 4506 | |
70806df6 | 4507 | startpos -= p0 - 1 - p; |
b18215fc | 4508 | } |
25fe55af | 4509 | } |
fa9a63c5 RM |
4510 | } |
4511 | return -1; | |
4512 | } /* re_search_2 */ | |
c0f9ea08 | 4513 | WEAK_ALIAS (__re_search_2, re_search_2) |
fa9a63c5 RM |
4514 | \f |
4515 | /* Declarations and macros for re_match_2. */ | |
4516 | ||
261cb4bb PE |
4517 | static int bcmp_translate (re_char *s1, re_char *s2, |
4518 | register ssize_t len, | |
4519 | RE_TRANSLATE_TYPE translate, | |
4520 | const int multibyte); | |
fa9a63c5 RM |
4521 | |
4522 | /* This converts PTR, a pointer into one of the search strings `string1' | |
4523 | and `string2' into an offset from the beginning of that string. */ | |
4524 | #define POINTER_TO_OFFSET(ptr) \ | |
4525 | (FIRST_STRING_P (ptr) \ | |
4526 | ? ((regoff_t) ((ptr) - string1)) \ | |
4527 | : ((regoff_t) ((ptr) - string2 + size1))) | |
4528 | ||
fa9a63c5 | 4529 | /* Call before fetching a character with *d. This switches over to |
419d1c74 SM |
4530 | string2 if necessary. |
4531 | Check re_match_2_internal for a discussion of why end_match_2 might | |
4532 | not be within string2 (but be equal to end_match_1 instead). */ | |
fa9a63c5 | 4533 | #define PREFETCH() \ |
25fe55af | 4534 | while (d == dend) \ |
fa9a63c5 RM |
4535 | { \ |
4536 | /* End of string2 => fail. */ \ | |
25fe55af RS |
4537 | if (dend == end_match_2) \ |
4538 | goto fail; \ | |
4bb91c68 | 4539 | /* End of string1 => advance to string2. */ \ |
25fe55af | 4540 | d = string2; \ |
fa9a63c5 RM |
4541 | dend = end_match_2; \ |
4542 | } | |
4543 | ||
f1ad044f SM |
4544 | /* Call before fetching a char with *d if you already checked other limits. |
4545 | This is meant for use in lookahead operations like wordend, etc.. | |
4546 | where we might need to look at parts of the string that might be | |
4547 | outside of the LIMITs (i.e past `stop'). */ | |
4548 | #define PREFETCH_NOLIMIT() \ | |
4549 | if (d == end1) \ | |
4550 | { \ | |
4551 | d = string2; \ | |
4552 | dend = end_match_2; \ | |
4553 | } \ | |
fa9a63c5 RM |
4554 | |
4555 | /* Test if at very beginning or at very end of the virtual concatenation | |
7814e705 | 4556 | of `string1' and `string2'. If only one string, it's `string2'. */ |
fa9a63c5 | 4557 | #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) |
5e69f11e | 4558 | #define AT_STRINGS_END(d) ((d) == end2) |
fa9a63c5 | 4559 | |
9121ca40 | 4560 | /* Disabled due to a compiler bug -- see comment at case wordbound */ |
b18215fc RS |
4561 | |
4562 | /* The comment at case wordbound is following one, but we don't use | |
4563 | AT_WORD_BOUNDARY anymore to support multibyte form. | |
4564 | ||
4565 | The DEC Alpha C compiler 3.x generates incorrect code for the | |
25fe55af | 4566 | test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of |
7814e705 | 4567 | AT_WORD_BOUNDARY, so this code is disabled. Expanding the |
b18215fc RS |
4568 | macro and introducing temporary variables works around the bug. */ |
4569 | ||
9121ca40 | 4570 | #if 0 |
b313f9d8 PE |
4571 | /* Test if D points to a character which is word-constituent. We have |
4572 | two special cases to check for: if past the end of string1, look at | |
4573 | the first character in string2; and if before the beginning of | |
4574 | string2, look at the last character in string1. */ | |
4575 | #define WORDCHAR_P(d) \ | |
4576 | (SYNTAX ((d) == end1 ? *string2 \ | |
4577 | : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ | |
4578 | == Sword) | |
4579 | ||
fa9a63c5 RM |
4580 | /* Test if the character before D and the one at D differ with respect |
4581 | to being word-constituent. */ | |
4582 | #define AT_WORD_BOUNDARY(d) \ | |
4583 | (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ | |
4584 | || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) | |
9121ca40 | 4585 | #endif |
fa9a63c5 RM |
4586 | |
4587 | /* Free everything we malloc. */ | |
4588 | #ifdef MATCH_MAY_ALLOCATE | |
952db0d7 PE |
4589 | # define FREE_VAR(var) \ |
4590 | do { \ | |
4591 | if (var) \ | |
4592 | { \ | |
4593 | REGEX_FREE (var); \ | |
4594 | var = NULL; \ | |
4595 | } \ | |
4596 | } while (0) | |
0b32bf0e | 4597 | # define FREE_VARIABLES() \ |
fa9a63c5 RM |
4598 | do { \ |
4599 | REGEX_FREE_STACK (fail_stack.stack); \ | |
4600 | FREE_VAR (regstart); \ | |
4601 | FREE_VAR (regend); \ | |
fa9a63c5 RM |
4602 | FREE_VAR (best_regstart); \ |
4603 | FREE_VAR (best_regend); \ | |
fa9a63c5 RM |
4604 | } while (0) |
4605 | #else | |
0b32bf0e | 4606 | # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
4607 | #endif /* not MATCH_MAY_ALLOCATE */ |
4608 | ||
505bde11 SM |
4609 | \f |
4610 | /* Optimization routines. */ | |
4611 | ||
4e8a9132 SM |
4612 | /* If the operation is a match against one or more chars, |
4613 | return a pointer to the next operation, else return NULL. */ | |
01618498 | 4614 | static re_char * |
971de7fb | 4615 | skip_one_char (const re_char *p) |
4e8a9132 | 4616 | { |
7393bcbb | 4617 | switch (*p++) |
4e8a9132 SM |
4618 | { |
4619 | case anychar: | |
4620 | break; | |
177c0ea7 | 4621 | |
4e8a9132 SM |
4622 | case exactn: |
4623 | p += *p + 1; | |
4624 | break; | |
4625 | ||
4626 | case charset_not: | |
4627 | case charset: | |
4628 | if (CHARSET_RANGE_TABLE_EXISTS_P (p - 1)) | |
4629 | { | |
4630 | int mcnt; | |
4631 | p = CHARSET_RANGE_TABLE (p - 1); | |
4632 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4633 | p = CHARSET_RANGE_TABLE_END (p, mcnt); | |
4634 | } | |
4635 | else | |
4636 | p += 1 + CHARSET_BITMAP_SIZE (p - 1); | |
4637 | break; | |
177c0ea7 | 4638 | |
4e8a9132 SM |
4639 | case syntaxspec: |
4640 | case notsyntaxspec: | |
1fb352e0 | 4641 | #ifdef emacs |
4e8a9132 SM |
4642 | case categoryspec: |
4643 | case notcategoryspec: | |
4644 | #endif /* emacs */ | |
4645 | p++; | |
4646 | break; | |
4647 | ||
4648 | default: | |
4649 | p = NULL; | |
4650 | } | |
4651 | return p; | |
4652 | } | |
4653 | ||
4654 | ||
505bde11 | 4655 | /* Jump over non-matching operations. */ |
839966f3 | 4656 | static re_char * |
971de7fb | 4657 | skip_noops (const re_char *p, const re_char *pend) |
505bde11 SM |
4658 | { |
4659 | int mcnt; | |
4660 | while (p < pend) | |
4661 | { | |
7393bcbb | 4662 | switch (*p) |
505bde11 SM |
4663 | { |
4664 | case start_memory: | |
505bde11 SM |
4665 | case stop_memory: |
4666 | p += 2; break; | |
4667 | case no_op: | |
4668 | p += 1; break; | |
4669 | case jump: | |
4670 | p += 1; | |
4671 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4672 | p += mcnt; | |
4673 | break; | |
4674 | default: | |
4675 | return p; | |
4676 | } | |
4677 | } | |
4678 | assert (p == pend); | |
4679 | return p; | |
4680 | } | |
4681 | ||
4682 | /* Non-zero if "p1 matches something" implies "p2 fails". */ | |
4683 | static int | |
971de7fb | 4684 | mutually_exclusive_p (struct re_pattern_buffer *bufp, const re_char *p1, const re_char *p2) |
505bde11 | 4685 | { |
4e8a9132 | 4686 | re_opcode_t op2; |
2d1675e4 | 4687 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
505bde11 SM |
4688 | unsigned char *pend = bufp->buffer + bufp->used; |
4689 | ||
4e8a9132 | 4690 | assert (p1 >= bufp->buffer && p1 < pend |
505bde11 SM |
4691 | && p2 >= bufp->buffer && p2 <= pend); |
4692 | ||
4693 | /* Skip over open/close-group commands. | |
4694 | If what follows this loop is a ...+ construct, | |
4695 | look at what begins its body, since we will have to | |
4696 | match at least one of that. */ | |
4e8a9132 SM |
4697 | p2 = skip_noops (p2, pend); |
4698 | /* The same skip can be done for p1, except that this function | |
4699 | is only used in the case where p1 is a simple match operator. */ | |
4700 | /* p1 = skip_noops (p1, pend); */ | |
4701 | ||
4702 | assert (p1 >= bufp->buffer && p1 < pend | |
4703 | && p2 >= bufp->buffer && p2 <= pend); | |
4704 | ||
4705 | op2 = p2 == pend ? succeed : *p2; | |
4706 | ||
7393bcbb | 4707 | switch (op2) |
505bde11 | 4708 | { |
4e8a9132 SM |
4709 | case succeed: |
4710 | case endbuf: | |
4711 | /* If we're at the end of the pattern, we can change. */ | |
4712 | if (skip_one_char (p1)) | |
505bde11 | 4713 | { |
505bde11 SM |
4714 | DEBUG_PRINT1 (" End of pattern: fast loop.\n"); |
4715 | return 1; | |
505bde11 | 4716 | } |
4e8a9132 | 4717 | break; |
177c0ea7 | 4718 | |
4e8a9132 | 4719 | case endline: |
4e8a9132 SM |
4720 | case exactn: |
4721 | { | |
01618498 | 4722 | register re_wchar_t c |
4e8a9132 | 4723 | = (re_opcode_t) *p2 == endline ? '\n' |
62a6e103 | 4724 | : RE_STRING_CHAR (p2 + 2, multibyte); |
505bde11 | 4725 | |
4e8a9132 SM |
4726 | if ((re_opcode_t) *p1 == exactn) |
4727 | { | |
62a6e103 | 4728 | if (c != RE_STRING_CHAR (p1 + 2, multibyte)) |
4e8a9132 SM |
4729 | { |
4730 | DEBUG_PRINT3 (" '%c' != '%c' => fast loop.\n", c, p1[2]); | |
4731 | return 1; | |
4732 | } | |
4733 | } | |
505bde11 | 4734 | |
4e8a9132 SM |
4735 | else if ((re_opcode_t) *p1 == charset |
4736 | || (re_opcode_t) *p1 == charset_not) | |
4737 | { | |
4738 | int not = (re_opcode_t) *p1 == charset_not; | |
505bde11 | 4739 | |
4e8a9132 SM |
4740 | /* Test if C is listed in charset (or charset_not) |
4741 | at `p1'. */ | |
6fdd04b0 | 4742 | if (! multibyte || IS_REAL_ASCII (c)) |
4e8a9132 SM |
4743 | { |
4744 | if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH | |
4745 | && p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
4746 | not = !not; | |
4747 | } | |
4748 | else if (CHARSET_RANGE_TABLE_EXISTS_P (p1)) | |
4749 | CHARSET_LOOKUP_RANGE_TABLE (not, c, p1); | |
505bde11 | 4750 | |
4e8a9132 SM |
4751 | /* `not' is equal to 1 if c would match, which means |
4752 | that we can't change to pop_failure_jump. */ | |
4753 | if (!not) | |
4754 | { | |
4755 | DEBUG_PRINT1 (" No match => fast loop.\n"); | |
4756 | return 1; | |
4757 | } | |
4758 | } | |
4759 | else if ((re_opcode_t) *p1 == anychar | |
4760 | && c == '\n') | |
4761 | { | |
4762 | DEBUG_PRINT1 (" . != \\n => fast loop.\n"); | |
4763 | return 1; | |
4764 | } | |
4765 | } | |
4766 | break; | |
505bde11 | 4767 | |
4e8a9132 | 4768 | case charset: |
4e8a9132 SM |
4769 | { |
4770 | if ((re_opcode_t) *p1 == exactn) | |
4771 | /* Reuse the code above. */ | |
4772 | return mutually_exclusive_p (bufp, p2, p1); | |
505bde11 | 4773 | |
505bde11 SM |
4774 | /* It is hard to list up all the character in charset |
4775 | P2 if it includes multibyte character. Give up in | |
4776 | such case. */ | |
4777 | else if (!multibyte || !CHARSET_RANGE_TABLE_EXISTS_P (p2)) | |
4778 | { | |
4779 | /* Now, we are sure that P2 has no range table. | |
4780 | So, for the size of bitmap in P2, `p2[1]' is | |
7814e705 | 4781 | enough. But P1 may have range table, so the |
505bde11 SM |
4782 | size of bitmap table of P1 is extracted by |
4783 | using macro `CHARSET_BITMAP_SIZE'. | |
4784 | ||
6fdd04b0 KH |
4785 | In a multibyte case, we know that all the character |
4786 | listed in P2 is ASCII. In a unibyte case, P1 has only a | |
4787 | bitmap table. So, in both cases, it is enough to test | |
4788 | only the bitmap table of P1. */ | |
505bde11 | 4789 | |
411e4203 | 4790 | if ((re_opcode_t) *p1 == charset) |
505bde11 SM |
4791 | { |
4792 | int idx; | |
4793 | /* We win if the charset inside the loop | |
4794 | has no overlap with the one after the loop. */ | |
4795 | for (idx = 0; | |
4796 | (idx < (int) p2[1] | |
4797 | && idx < CHARSET_BITMAP_SIZE (p1)); | |
4798 | idx++) | |
4799 | if ((p2[2 + idx] & p1[2 + idx]) != 0) | |
4800 | break; | |
4801 | ||
4802 | if (idx == p2[1] | |
4803 | || idx == CHARSET_BITMAP_SIZE (p1)) | |
4804 | { | |
4805 | DEBUG_PRINT1 (" No match => fast loop.\n"); | |
4806 | return 1; | |
4807 | } | |
4808 | } | |
411e4203 | 4809 | else if ((re_opcode_t) *p1 == charset_not) |
505bde11 SM |
4810 | { |
4811 | int idx; | |
4812 | /* We win if the charset_not inside the loop lists | |
7814e705 | 4813 | every character listed in the charset after. */ |
505bde11 SM |
4814 | for (idx = 0; idx < (int) p2[1]; idx++) |
4815 | if (! (p2[2 + idx] == 0 | |
4816 | || (idx < CHARSET_BITMAP_SIZE (p1) | |
4817 | && ((p2[2 + idx] & ~ p1[2 + idx]) == 0)))) | |
4818 | break; | |
4819 | ||
d1dfb56c EZ |
4820 | if (idx == p2[1]) |
4821 | { | |
4822 | DEBUG_PRINT1 (" No match => fast loop.\n"); | |
4823 | return 1; | |
4824 | } | |
4e8a9132 SM |
4825 | } |
4826 | } | |
4827 | } | |
609b757a | 4828 | break; |
177c0ea7 | 4829 | |
411e4203 | 4830 | case charset_not: |
7393bcbb | 4831 | switch (*p1) |
411e4203 SM |
4832 | { |
4833 | case exactn: | |
4834 | case charset: | |
4835 | /* Reuse the code above. */ | |
4836 | return mutually_exclusive_p (bufp, p2, p1); | |
4837 | case charset_not: | |
4838 | /* When we have two charset_not, it's very unlikely that | |
4839 | they don't overlap. The union of the two sets of excluded | |
4840 | chars should cover all possible chars, which, as a matter of | |
4841 | fact, is virtually impossible in multibyte buffers. */ | |
36595814 | 4842 | break; |
411e4203 SM |
4843 | } |
4844 | break; | |
4845 | ||
4e8a9132 | 4846 | case wordend: |
669fa600 SM |
4847 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == Sword); |
4848 | case symend: | |
4e8a9132 | 4849 | return ((re_opcode_t) *p1 == syntaxspec |
669fa600 SM |
4850 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4851 | case notsyntaxspec: | |
4852 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4853 | |
4854 | case wordbeg: | |
669fa600 SM |
4855 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == Sword); |
4856 | case symbeg: | |
4e8a9132 | 4857 | return ((re_opcode_t) *p1 == notsyntaxspec |
669fa600 SM |
4858 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4859 | case syntaxspec: | |
4860 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4861 | |
4862 | case wordbound: | |
4863 | return (((re_opcode_t) *p1 == notsyntaxspec | |
4864 | || (re_opcode_t) *p1 == syntaxspec) | |
4865 | && p1[1] == Sword); | |
4866 | ||
1fb352e0 | 4867 | #ifdef emacs |
4e8a9132 SM |
4868 | case categoryspec: |
4869 | return ((re_opcode_t) *p1 == notcategoryspec && p1[1] == p2[1]); | |
4870 | case notcategoryspec: | |
4871 | return ((re_opcode_t) *p1 == categoryspec && p1[1] == p2[1]); | |
4872 | #endif /* emacs */ | |
4873 | ||
4874 | default: | |
4875 | ; | |
505bde11 SM |
4876 | } |
4877 | ||
4878 | /* Safe default. */ | |
4879 | return 0; | |
4880 | } | |
4881 | ||
fa9a63c5 RM |
4882 | \f |
4883 | /* Matching routines. */ | |
4884 | ||
25fe55af | 4885 | #ifndef emacs /* Emacs never uses this. */ |
fa9a63c5 RM |
4886 | /* re_match is like re_match_2 except it takes only a single string. */ |
4887 | ||
d1dfb56c | 4888 | regoff_t |
d2762c86 | 4889 | re_match (struct re_pattern_buffer *bufp, const char *string, |
d1dfb56c | 4890 | size_t size, ssize_t pos, struct re_registers *regs) |
fa9a63c5 | 4891 | { |
d1dfb56c EZ |
4892 | regoff_t result = re_match_2_internal (bufp, NULL, 0, (re_char*) string, |
4893 | size, pos, regs, size); | |
fa9a63c5 RM |
4894 | return result; |
4895 | } | |
c0f9ea08 | 4896 | WEAK_ALIAS (__re_match, re_match) |
fa9a63c5 RM |
4897 | #endif /* not emacs */ |
4898 | ||
b18215fc RS |
4899 | #ifdef emacs |
4900 | /* In Emacs, this is the string or buffer in which we | |
7814e705 | 4901 | are matching. It is used for looking up syntax properties. */ |
b18215fc RS |
4902 | Lisp_Object re_match_object; |
4903 | #endif | |
fa9a63c5 RM |
4904 | |
4905 | /* re_match_2 matches the compiled pattern in BUFP against the | |
4906 | the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 | |
4907 | and SIZE2, respectively). We start matching at POS, and stop | |
4908 | matching at STOP. | |
5e69f11e | 4909 | |
fa9a63c5 | 4910 | If REGS is non-null and the `no_sub' field of BUFP is nonzero, we |
7814e705 | 4911 | store offsets for the substring each group matched in REGS. See the |
fa9a63c5 RM |
4912 | documentation for exactly how many groups we fill. |
4913 | ||
4914 | We return -1 if no match, -2 if an internal error (such as the | |
7814e705 | 4915 | failure stack overflowing). Otherwise, we return the length of the |
fa9a63c5 RM |
4916 | matched substring. */ |
4917 | ||
d1dfb56c EZ |
4918 | regoff_t |
4919 | re_match_2 (struct re_pattern_buffer *bufp, const char *string1, | |
4920 | size_t size1, const char *string2, size_t size2, ssize_t pos, | |
4921 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4922 | { |
d1dfb56c | 4923 | regoff_t result; |
25fe55af | 4924 | |
b18215fc | 4925 | #ifdef emacs |
d1dfb56c | 4926 | ssize_t charpos; |
d48cd3f4 | 4927 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
99633e97 | 4928 | charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (pos)); |
cc9b4df2 | 4929 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); |
b18215fc RS |
4930 | #endif |
4931 | ||
4bb91c68 SM |
4932 | result = re_match_2_internal (bufp, (re_char*) string1, size1, |
4933 | (re_char*) string2, size2, | |
cc9b4df2 | 4934 | pos, regs, stop); |
fa9a63c5 RM |
4935 | return result; |
4936 | } | |
c0f9ea08 | 4937 | WEAK_ALIAS (__re_match_2, re_match_2) |
fa9a63c5 | 4938 | |
bf216479 | 4939 | |
fa9a63c5 | 4940 | /* This is a separate function so that we can force an alloca cleanup |
7814e705 | 4941 | afterwards. */ |
d1dfb56c EZ |
4942 | static regoff_t |
4943 | re_match_2_internal (struct re_pattern_buffer *bufp, const re_char *string1, | |
4944 | size_t size1, const re_char *string2, size_t size2, | |
4945 | ssize_t pos, struct re_registers *regs, ssize_t stop) | |
fa9a63c5 RM |
4946 | { |
4947 | /* General temporaries. */ | |
d1dfb56c | 4948 | ssize_t mcnt; |
01618498 | 4949 | size_t reg; |
fa9a63c5 RM |
4950 | |
4951 | /* Just past the end of the corresponding string. */ | |
66f0296e | 4952 | re_char *end1, *end2; |
fa9a63c5 RM |
4953 | |
4954 | /* Pointers into string1 and string2, just past the last characters in | |
7814e705 | 4955 | each to consider matching. */ |
66f0296e | 4956 | re_char *end_match_1, *end_match_2; |
fa9a63c5 RM |
4957 | |
4958 | /* Where we are in the data, and the end of the current string. */ | |
66f0296e | 4959 | re_char *d, *dend; |
5e69f11e | 4960 | |
99633e97 SM |
4961 | /* Used sometimes to remember where we were before starting matching |
4962 | an operator so that we can go back in case of failure. This "atomic" | |
4963 | behavior of matching opcodes is indispensable to the correctness | |
4964 | of the on_failure_keep_string_jump optimization. */ | |
4965 | re_char *dfail; | |
4966 | ||
fa9a63c5 | 4967 | /* Where we are in the pattern, and the end of the pattern. */ |
01618498 SM |
4968 | re_char *p = bufp->buffer; |
4969 | re_char *pend = p + bufp->used; | |
fa9a63c5 | 4970 | |
25fe55af | 4971 | /* We use this to map every character in the string. */ |
6676cb1c | 4972 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 | 4973 | |
cf9c99bc | 4974 | /* Nonzero if BUFP is setup from a multibyte regex. */ |
2d1675e4 | 4975 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
b18215fc | 4976 | |
cf9c99bc KH |
4977 | /* Nonzero if STRING1/STRING2 are multibyte. */ |
4978 | const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
4979 | ||
fa9a63c5 RM |
4980 | /* Failure point stack. Each place that can handle a failure further |
4981 | down the line pushes a failure point on this stack. It consists of | |
505bde11 | 4982 | regstart, and regend for all registers corresponding to |
fa9a63c5 RM |
4983 | the subexpressions we're currently inside, plus the number of such |
4984 | registers, and, finally, two char *'s. The first char * is where | |
4985 | to resume scanning the pattern; the second one is where to resume | |
7814e705 JB |
4986 | scanning the strings. */ |
4987 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | |
fa9a63c5 RM |
4988 | fail_stack_type fail_stack; |
4989 | #endif | |
4990 | #ifdef DEBUG | |
fa9a63c5 RM |
4991 | unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; |
4992 | #endif | |
4993 | ||
0b32bf0e | 4994 | #if defined REL_ALLOC && defined REGEX_MALLOC |
fa9a63c5 RM |
4995 | /* This holds the pointer to the failure stack, when |
4996 | it is allocated relocatably. */ | |
4997 | fail_stack_elt_t *failure_stack_ptr; | |
99633e97 | 4998 | #endif |
fa9a63c5 RM |
4999 | |
5000 | /* We fill all the registers internally, independent of what we | |
7814e705 | 5001 | return, for use in backreferences. The number here includes |
fa9a63c5 | 5002 | an element for register zero. */ |
4bb91c68 | 5003 | size_t num_regs = bufp->re_nsub + 1; |
5e69f11e | 5004 | |
fa9a63c5 RM |
5005 | /* Information on the contents of registers. These are pointers into |
5006 | the input strings; they record just what was matched (on this | |
5007 | attempt) by a subexpression part of the pattern, that is, the | |
5008 | regnum-th regstart pointer points to where in the pattern we began | |
5009 | matching and the regnum-th regend points to right after where we | |
5010 | stopped matching the regnum-th subexpression. (The zeroth register | |
5011 | keeps track of what the whole pattern matches.) */ | |
5012 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5013 | re_char **regstart, **regend; |
fa9a63c5 RM |
5014 | #endif |
5015 | ||
fa9a63c5 | 5016 | /* The following record the register info as found in the above |
5e69f11e | 5017 | variables when we find a match better than any we've seen before. |
fa9a63c5 RM |
5018 | This happens as we backtrack through the failure points, which in |
5019 | turn happens only if we have not yet matched the entire string. */ | |
5020 | unsigned best_regs_set = false; | |
5021 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5022 | re_char **best_regstart, **best_regend; |
fa9a63c5 | 5023 | #endif |
5e69f11e | 5024 | |
fa9a63c5 RM |
5025 | /* Logically, this is `best_regend[0]'. But we don't want to have to |
5026 | allocate space for that if we're not allocating space for anything | |
7814e705 | 5027 | else (see below). Also, we never need info about register 0 for |
fa9a63c5 RM |
5028 | any of the other register vectors, and it seems rather a kludge to |
5029 | treat `best_regend' differently than the rest. So we keep track of | |
5030 | the end of the best match so far in a separate variable. We | |
5031 | initialize this to NULL so that when we backtrack the first time | |
5032 | and need to test it, it's not garbage. */ | |
66f0296e | 5033 | re_char *match_end = NULL; |
fa9a63c5 | 5034 | |
fa9a63c5 RM |
5035 | #ifdef DEBUG |
5036 | /* Counts the total number of registers pushed. */ | |
5e69f11e | 5037 | unsigned num_regs_pushed = 0; |
fa9a63c5 RM |
5038 | #endif |
5039 | ||
5040 | DEBUG_PRINT1 ("\n\nEntering re_match_2.\n"); | |
5e69f11e | 5041 | |
fa9a63c5 | 5042 | INIT_FAIL_STACK (); |
5e69f11e | 5043 | |
fa9a63c5 RM |
5044 | #ifdef MATCH_MAY_ALLOCATE |
5045 | /* Do not bother to initialize all the register variables if there are | |
5046 | no groups in the pattern, as it takes a fair amount of time. If | |
5047 | there are groups, we include space for register 0 (the whole | |
5048 | pattern), even though we never use it, since it simplifies the | |
5049 | array indexing. We should fix this. */ | |
5050 | if (bufp->re_nsub) | |
5051 | { | |
66f0296e SM |
5052 | regstart = REGEX_TALLOC (num_regs, re_char *); |
5053 | regend = REGEX_TALLOC (num_regs, re_char *); | |
5054 | best_regstart = REGEX_TALLOC (num_regs, re_char *); | |
5055 | best_regend = REGEX_TALLOC (num_regs, re_char *); | |
fa9a63c5 | 5056 | |
505bde11 | 5057 | if (!(regstart && regend && best_regstart && best_regend)) |
25fe55af RS |
5058 | { |
5059 | FREE_VARIABLES (); | |
5060 | return -2; | |
5061 | } | |
fa9a63c5 RM |
5062 | } |
5063 | else | |
5064 | { | |
5065 | /* We must initialize all our variables to NULL, so that | |
25fe55af | 5066 | `FREE_VARIABLES' doesn't try to free them. */ |
505bde11 | 5067 | regstart = regend = best_regstart = best_regend = NULL; |
fa9a63c5 RM |
5068 | } |
5069 | #endif /* MATCH_MAY_ALLOCATE */ | |
5070 | ||
5071 | /* The starting position is bogus. */ | |
5072 | if (pos < 0 || pos > size1 + size2) | |
5073 | { | |
5074 | FREE_VARIABLES (); | |
5075 | return -1; | |
5076 | } | |
5e69f11e | 5077 | |
fa9a63c5 RM |
5078 | /* Initialize subexpression text positions to -1 to mark ones that no |
5079 | start_memory/stop_memory has been seen for. Also initialize the | |
5080 | register information struct. */ | |
01618498 SM |
5081 | for (reg = 1; reg < num_regs; reg++) |
5082 | regstart[reg] = regend[reg] = NULL; | |
99633e97 | 5083 | |
fa9a63c5 | 5084 | /* We move `string1' into `string2' if the latter's empty -- but not if |
7814e705 | 5085 | `string1' is null. */ |
fa9a63c5 RM |
5086 | if (size2 == 0 && string1 != NULL) |
5087 | { | |
5088 | string2 = string1; | |
5089 | size2 = size1; | |
5090 | string1 = 0; | |
5091 | size1 = 0; | |
5092 | } | |
5093 | end1 = string1 + size1; | |
5094 | end2 = string2 + size2; | |
5095 | ||
5e69f11e | 5096 | /* `p' scans through the pattern as `d' scans through the data. |
fa9a63c5 RM |
5097 | `dend' is the end of the input string that `d' points within. `d' |
5098 | is advanced into the following input string whenever necessary, but | |
5099 | this happens before fetching; therefore, at the beginning of the | |
5100 | loop, `d' can be pointing at the end of a string, but it cannot | |
5101 | equal `string2'. */ | |
419d1c74 | 5102 | if (pos >= size1) |
fa9a63c5 | 5103 | { |
419d1c74 SM |
5104 | /* Only match within string2. */ |
5105 | d = string2 + pos - size1; | |
5106 | dend = end_match_2 = string2 + stop - size1; | |
5107 | end_match_1 = end1; /* Just to give it a value. */ | |
fa9a63c5 RM |
5108 | } |
5109 | else | |
5110 | { | |
f1ad044f | 5111 | if (stop < size1) |
419d1c74 SM |
5112 | { |
5113 | /* Only match within string1. */ | |
5114 | end_match_1 = string1 + stop; | |
5115 | /* BEWARE! | |
5116 | When we reach end_match_1, PREFETCH normally switches to string2. | |
5117 | But in the present case, this means that just doing a PREFETCH | |
5118 | makes us jump from `stop' to `gap' within the string. | |
5119 | What we really want here is for the search to stop as | |
5120 | soon as we hit end_match_1. That's why we set end_match_2 | |
5121 | to end_match_1 (since PREFETCH fails as soon as we hit | |
5122 | end_match_2). */ | |
5123 | end_match_2 = end_match_1; | |
5124 | } | |
5125 | else | |
f1ad044f SM |
5126 | { /* It's important to use this code when stop == size so that |
5127 | moving `d' from end1 to string2 will not prevent the d == dend | |
5128 | check from catching the end of string. */ | |
419d1c74 SM |
5129 | end_match_1 = end1; |
5130 | end_match_2 = string2 + stop - size1; | |
5131 | } | |
5132 | d = string1 + pos; | |
5133 | dend = end_match_1; | |
fa9a63c5 RM |
5134 | } |
5135 | ||
5136 | DEBUG_PRINT1 ("The compiled pattern is: "); | |
5137 | DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); | |
5138 | DEBUG_PRINT1 ("The string to match is: `"); | |
5139 | DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); | |
5140 | DEBUG_PRINT1 ("'\n"); | |
5e69f11e | 5141 | |
7814e705 | 5142 | /* This loops over pattern commands. It exits by returning from the |
fa9a63c5 RM |
5143 | function if the match is complete, or it drops through if the match |
5144 | fails at this starting point in the input data. */ | |
5145 | for (;;) | |
5146 | { | |
505bde11 | 5147 | DEBUG_PRINT2 ("\n%p: ", p); |
fa9a63c5 RM |
5148 | |
5149 | if (p == pend) | |
5150 | { /* End of pattern means we might have succeeded. */ | |
25fe55af | 5151 | DEBUG_PRINT1 ("end of pattern ... "); |
5e69f11e | 5152 | |
fa9a63c5 | 5153 | /* If we haven't matched the entire string, and we want the |
25fe55af RS |
5154 | longest match, try backtracking. */ |
5155 | if (d != end_match_2) | |
fa9a63c5 RM |
5156 | { |
5157 | /* 1 if this match ends in the same string (string1 or string2) | |
5158 | as the best previous match. */ | |
d42f4f0f PE |
5159 | boolean same_str_p = (FIRST_STRING_P (match_end) |
5160 | == FIRST_STRING_P (d)); | |
fa9a63c5 RM |
5161 | /* 1 if this match is the best seen so far. */ |
5162 | boolean best_match_p; | |
5163 | ||
5164 | /* AIX compiler got confused when this was combined | |
7814e705 | 5165 | with the previous declaration. */ |
fa9a63c5 RM |
5166 | if (same_str_p) |
5167 | best_match_p = d > match_end; | |
5168 | else | |
99633e97 | 5169 | best_match_p = !FIRST_STRING_P (d); |
fa9a63c5 | 5170 | |
25fe55af RS |
5171 | DEBUG_PRINT1 ("backtracking.\n"); |
5172 | ||
5173 | if (!FAIL_STACK_EMPTY ()) | |
5174 | { /* More failure points to try. */ | |
5175 | ||
5176 | /* If exceeds best match so far, save it. */ | |
5177 | if (!best_regs_set || best_match_p) | |
5178 | { | |
5179 | best_regs_set = true; | |
5180 | match_end = d; | |
5181 | ||
5182 | DEBUG_PRINT1 ("\nSAVING match as best so far.\n"); | |
5183 | ||
01618498 | 5184 | for (reg = 1; reg < num_regs; reg++) |
25fe55af | 5185 | { |
01618498 SM |
5186 | best_regstart[reg] = regstart[reg]; |
5187 | best_regend[reg] = regend[reg]; | |
25fe55af RS |
5188 | } |
5189 | } | |
5190 | goto fail; | |
5191 | } | |
5192 | ||
5193 | /* If no failure points, don't restore garbage. And if | |
5194 | last match is real best match, don't restore second | |
5195 | best one. */ | |
5196 | else if (best_regs_set && !best_match_p) | |
5197 | { | |
5198 | restore_best_regs: | |
5199 | /* Restore best match. It may happen that `dend == | |
5200 | end_match_1' while the restored d is in string2. | |
5201 | For example, the pattern `x.*y.*z' against the | |
5202 | strings `x-' and `y-z-', if the two strings are | |
7814e705 | 5203 | not consecutive in memory. */ |
25fe55af RS |
5204 | DEBUG_PRINT1 ("Restoring best registers.\n"); |
5205 | ||
5206 | d = match_end; | |
5207 | dend = ((d >= string1 && d <= end1) | |
5208 | ? end_match_1 : end_match_2); | |
fa9a63c5 | 5209 | |
01618498 | 5210 | for (reg = 1; reg < num_regs; reg++) |
fa9a63c5 | 5211 | { |
01618498 SM |
5212 | regstart[reg] = best_regstart[reg]; |
5213 | regend[reg] = best_regend[reg]; | |
fa9a63c5 | 5214 | } |
25fe55af RS |
5215 | } |
5216 | } /* d != end_match_2 */ | |
fa9a63c5 RM |
5217 | |
5218 | succeed_label: | |
25fe55af | 5219 | DEBUG_PRINT1 ("Accepting match.\n"); |
fa9a63c5 | 5220 | |
25fe55af RS |
5221 | /* If caller wants register contents data back, do it. */ |
5222 | if (regs && !bufp->no_sub) | |
fa9a63c5 | 5223 | { |
25fe55af RS |
5224 | /* Have the register data arrays been allocated? */ |
5225 | if (bufp->regs_allocated == REGS_UNALLOCATED) | |
7814e705 | 5226 | { /* No. So allocate them with malloc. We need one |
25fe55af RS |
5227 | extra element beyond `num_regs' for the `-1' marker |
5228 | GNU code uses. */ | |
5229 | regs->num_regs = MAX (RE_NREGS, num_regs + 1); | |
5230 | regs->start = TALLOC (regs->num_regs, regoff_t); | |
5231 | regs->end = TALLOC (regs->num_regs, regoff_t); | |
5232 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5233 | { |
5234 | FREE_VARIABLES (); | |
5235 | return -2; | |
5236 | } | |
25fe55af RS |
5237 | bufp->regs_allocated = REGS_REALLOCATE; |
5238 | } | |
5239 | else if (bufp->regs_allocated == REGS_REALLOCATE) | |
5240 | { /* Yes. If we need more elements than were already | |
5241 | allocated, reallocate them. If we need fewer, just | |
5242 | leave it alone. */ | |
5243 | if (regs->num_regs < num_regs + 1) | |
5244 | { | |
5245 | regs->num_regs = num_regs + 1; | |
5246 | RETALLOC (regs->start, regs->num_regs, regoff_t); | |
5247 | RETALLOC (regs->end, regs->num_regs, regoff_t); | |
5248 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5249 | { |
5250 | FREE_VARIABLES (); | |
5251 | return -2; | |
5252 | } | |
25fe55af RS |
5253 | } |
5254 | } | |
5255 | else | |
fa9a63c5 RM |
5256 | { |
5257 | /* These braces fend off a "empty body in an else-statement" | |
7814e705 | 5258 | warning under GCC when assert expands to nothing. */ |
fa9a63c5 RM |
5259 | assert (bufp->regs_allocated == REGS_FIXED); |
5260 | } | |
5261 | ||
25fe55af RS |
5262 | /* Convert the pointer data in `regstart' and `regend' to |
5263 | indices. Register zero has to be set differently, | |
5264 | since we haven't kept track of any info for it. */ | |
5265 | if (regs->num_regs > 0) | |
5266 | { | |
5267 | regs->start[0] = pos; | |
99633e97 | 5268 | regs->end[0] = POINTER_TO_OFFSET (d); |
25fe55af | 5269 | } |
5e69f11e | 5270 | |
25fe55af RS |
5271 | /* Go through the first `min (num_regs, regs->num_regs)' |
5272 | registers, since that is all we initialized. */ | |
01618498 | 5273 | for (reg = 1; reg < MIN (num_regs, regs->num_regs); reg++) |
fa9a63c5 | 5274 | { |
01618498 SM |
5275 | if (REG_UNSET (regstart[reg]) || REG_UNSET (regend[reg])) |
5276 | regs->start[reg] = regs->end[reg] = -1; | |
25fe55af RS |
5277 | else |
5278 | { | |
01618498 SM |
5279 | regs->start[reg] |
5280 | = (regoff_t) POINTER_TO_OFFSET (regstart[reg]); | |
5281 | regs->end[reg] | |
5282 | = (regoff_t) POINTER_TO_OFFSET (regend[reg]); | |
25fe55af | 5283 | } |
fa9a63c5 | 5284 | } |
5e69f11e | 5285 | |
25fe55af RS |
5286 | /* If the regs structure we return has more elements than |
5287 | were in the pattern, set the extra elements to -1. If | |
5288 | we (re)allocated the registers, this is the case, | |
5289 | because we always allocate enough to have at least one | |
7814e705 | 5290 | -1 at the end. */ |
01618498 SM |
5291 | for (reg = num_regs; reg < regs->num_regs; reg++) |
5292 | regs->start[reg] = regs->end[reg] = -1; | |
fa9a63c5 RM |
5293 | } /* regs && !bufp->no_sub */ |
5294 | ||
25fe55af RS |
5295 | DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n", |
5296 | nfailure_points_pushed, nfailure_points_popped, | |
5297 | nfailure_points_pushed - nfailure_points_popped); | |
5298 | DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed); | |
fa9a63c5 | 5299 | |
99633e97 | 5300 | mcnt = POINTER_TO_OFFSET (d) - pos; |
fa9a63c5 | 5301 | |
25fe55af | 5302 | DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt); |
fa9a63c5 | 5303 | |
25fe55af RS |
5304 | FREE_VARIABLES (); |
5305 | return mcnt; | |
5306 | } | |
fa9a63c5 | 5307 | |
7814e705 | 5308 | /* Otherwise match next pattern command. */ |
7393bcbb | 5309 | switch (*p++) |
fa9a63c5 | 5310 | { |
25fe55af RS |
5311 | /* Ignore these. Used to ignore the n of succeed_n's which |
5312 | currently have n == 0. */ | |
5313 | case no_op: | |
5314 | DEBUG_PRINT1 ("EXECUTING no_op.\n"); | |
5315 | break; | |
fa9a63c5 RM |
5316 | |
5317 | case succeed: | |
25fe55af | 5318 | DEBUG_PRINT1 ("EXECUTING succeed.\n"); |
fa9a63c5 RM |
5319 | goto succeed_label; |
5320 | ||
7814e705 | 5321 | /* Match the next n pattern characters exactly. The following |
25fe55af | 5322 | byte in the pattern defines n, and the n bytes after that |
7814e705 | 5323 | are the characters to match. */ |
fa9a63c5 RM |
5324 | case exactn: |
5325 | mcnt = *p++; | |
25fe55af | 5326 | DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt); |
fa9a63c5 | 5327 | |
99633e97 SM |
5328 | /* Remember the start point to rollback upon failure. */ |
5329 | dfail = d; | |
5330 | ||
6fdd04b0 | 5331 | #ifndef emacs |
25fe55af RS |
5332 | /* This is written out as an if-else so we don't waste time |
5333 | testing `translate' inside the loop. */ | |
28703c16 | 5334 | if (RE_TRANSLATE_P (translate)) |
6fdd04b0 KH |
5335 | do |
5336 | { | |
5337 | PREFETCH (); | |
5338 | if (RE_TRANSLATE (translate, *d) != *p++) | |
e934739e | 5339 | { |
6fdd04b0 KH |
5340 | d = dfail; |
5341 | goto fail; | |
e934739e | 5342 | } |
6fdd04b0 KH |
5343 | d++; |
5344 | } | |
5345 | while (--mcnt); | |
fa9a63c5 | 5346 | else |
6fdd04b0 KH |
5347 | do |
5348 | { | |
5349 | PREFETCH (); | |
5350 | if (*d++ != *p++) | |
bf216479 | 5351 | { |
6fdd04b0 KH |
5352 | d = dfail; |
5353 | goto fail; | |
bf216479 | 5354 | } |
6fdd04b0 KH |
5355 | } |
5356 | while (--mcnt); | |
5357 | #else /* emacs */ | |
5358 | /* The cost of testing `translate' is comparatively small. */ | |
cf9c99bc | 5359 | if (target_multibyte) |
6fdd04b0 KH |
5360 | do |
5361 | { | |
5362 | int pat_charlen, buf_charlen; | |
cf9c99bc | 5363 | int pat_ch, buf_ch; |
e934739e | 5364 | |
6fdd04b0 | 5365 | PREFETCH (); |
cf9c99bc | 5366 | if (multibyte) |
62a6e103 | 5367 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
cf9c99bc KH |
5368 | else |
5369 | { | |
5370 | pat_ch = RE_CHAR_TO_MULTIBYTE (*p); | |
5371 | pat_charlen = 1; | |
5372 | } | |
62a6e103 | 5373 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 5374 | |
6fdd04b0 | 5375 | if (TRANSLATE (buf_ch) != pat_ch) |
e934739e | 5376 | { |
6fdd04b0 KH |
5377 | d = dfail; |
5378 | goto fail; | |
e934739e | 5379 | } |
bf216479 | 5380 | |
6fdd04b0 KH |
5381 | p += pat_charlen; |
5382 | d += buf_charlen; | |
5383 | mcnt -= pat_charlen; | |
5384 | } | |
5385 | while (mcnt > 0); | |
fa9a63c5 | 5386 | else |
6fdd04b0 KH |
5387 | do |
5388 | { | |
abbd1bcf | 5389 | int pat_charlen; |
cf9c99bc | 5390 | int pat_ch, buf_ch; |
bf216479 | 5391 | |
6fdd04b0 | 5392 | PREFETCH (); |
cf9c99bc KH |
5393 | if (multibyte) |
5394 | { | |
62a6e103 | 5395 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
2afc21f5 | 5396 | pat_ch = RE_CHAR_TO_UNIBYTE (pat_ch); |
cf9c99bc KH |
5397 | } |
5398 | else | |
5399 | { | |
5400 | pat_ch = *p; | |
5401 | pat_charlen = 1; | |
5402 | } | |
5403 | buf_ch = RE_CHAR_TO_MULTIBYTE (*d); | |
5404 | if (! CHAR_BYTE8_P (buf_ch)) | |
5405 | { | |
5406 | buf_ch = TRANSLATE (buf_ch); | |
5407 | buf_ch = RE_CHAR_TO_UNIBYTE (buf_ch); | |
5408 | if (buf_ch < 0) | |
5409 | buf_ch = *d; | |
5410 | } | |
0e2501ed AS |
5411 | else |
5412 | buf_ch = *d; | |
cf9c99bc | 5413 | if (buf_ch != pat_ch) |
6fdd04b0 KH |
5414 | { |
5415 | d = dfail; | |
5416 | goto fail; | |
bf216479 | 5417 | } |
cf9c99bc KH |
5418 | p += pat_charlen; |
5419 | d++; | |
6fdd04b0 KH |
5420 | } |
5421 | while (--mcnt); | |
5422 | #endif | |
25fe55af | 5423 | break; |
fa9a63c5 RM |
5424 | |
5425 | ||
25fe55af | 5426 | /* Match any character except possibly a newline or a null. */ |
fa9a63c5 | 5427 | case anychar: |
e934739e RS |
5428 | { |
5429 | int buf_charlen; | |
01618498 | 5430 | re_wchar_t buf_ch; |
fa9a63c5 | 5431 | |
e934739e | 5432 | DEBUG_PRINT1 ("EXECUTING anychar.\n"); |
fa9a63c5 | 5433 | |
e934739e | 5434 | PREFETCH (); |
62a6e103 | 5435 | buf_ch = RE_STRING_CHAR_AND_LENGTH (d, buf_charlen, |
cf9c99bc | 5436 | target_multibyte); |
e934739e RS |
5437 | buf_ch = TRANSLATE (buf_ch); |
5438 | ||
5439 | if ((!(bufp->syntax & RE_DOT_NEWLINE) | |
5440 | && buf_ch == '\n') | |
5441 | || ((bufp->syntax & RE_DOT_NOT_NULL) | |
5442 | && buf_ch == '\000')) | |
5443 | goto fail; | |
5444 | ||
e934739e RS |
5445 | DEBUG_PRINT2 (" Matched `%d'.\n", *d); |
5446 | d += buf_charlen; | |
5447 | } | |
fa9a63c5 RM |
5448 | break; |
5449 | ||
5450 | ||
5451 | case charset: | |
5452 | case charset_not: | |
5453 | { | |
b18215fc | 5454 | register unsigned int c; |
fa9a63c5 | 5455 | boolean not = (re_opcode_t) *(p - 1) == charset_not; |
b18215fc RS |
5456 | int len; |
5457 | ||
5458 | /* Start of actual range_table, or end of bitmap if there is no | |
5459 | range table. */ | |
da053e48 | 5460 | re_char *range_table IF_LINT (= NULL); |
b18215fc | 5461 | |
96cc36cc | 5462 | /* Nonzero if there is a range table. */ |
b18215fc RS |
5463 | int range_table_exists; |
5464 | ||
96cc36cc RS |
5465 | /* Number of ranges of range table. This is not included |
5466 | in the initial byte-length of the command. */ | |
5467 | int count = 0; | |
fa9a63c5 | 5468 | |
f5020181 AS |
5469 | /* Whether matching against a unibyte character. */ |
5470 | boolean unibyte_char = false; | |
5471 | ||
25fe55af | 5472 | DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : ""); |
fa9a63c5 | 5473 | |
b18215fc | 5474 | range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]); |
96cc36cc | 5475 | |
b18215fc | 5476 | if (range_table_exists) |
96cc36cc RS |
5477 | { |
5478 | range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap. */ | |
5479 | EXTRACT_NUMBER_AND_INCR (count, range_table); | |
5480 | } | |
b18215fc | 5481 | |
2d1675e4 | 5482 | PREFETCH (); |
62a6e103 | 5483 | c = RE_STRING_CHAR_AND_LENGTH (d, len, target_multibyte); |
cf9c99bc KH |
5484 | if (target_multibyte) |
5485 | { | |
5486 | int c1; | |
b18215fc | 5487 | |
cf9c99bc KH |
5488 | c = TRANSLATE (c); |
5489 | c1 = RE_CHAR_TO_UNIBYTE (c); | |
5490 | if (c1 >= 0) | |
f5020181 AS |
5491 | { |
5492 | unibyte_char = true; | |
5493 | c = c1; | |
5494 | } | |
cf9c99bc KH |
5495 | } |
5496 | else | |
5497 | { | |
5498 | int c1 = RE_CHAR_TO_MULTIBYTE (c); | |
5499 | ||
5500 | if (! CHAR_BYTE8_P (c1)) | |
5501 | { | |
5502 | c1 = TRANSLATE (c1); | |
5503 | c1 = RE_CHAR_TO_UNIBYTE (c1); | |
5504 | if (c1 >= 0) | |
f5020181 AS |
5505 | { |
5506 | unibyte_char = true; | |
5507 | c = c1; | |
5508 | } | |
cf9c99bc | 5509 | } |
0b8be006 AS |
5510 | else |
5511 | unibyte_char = true; | |
cf9c99bc KH |
5512 | } |
5513 | ||
f5020181 | 5514 | if (unibyte_char && c < (1 << BYTEWIDTH)) |
b18215fc | 5515 | { /* Lookup bitmap. */ |
b18215fc RS |
5516 | /* Cast to `unsigned' instead of `unsigned char' in |
5517 | case the bit list is a full 32 bytes long. */ | |
5518 | if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH) | |
96cc36cc RS |
5519 | && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) |
5520 | not = !not; | |
b18215fc | 5521 | } |
96cc36cc | 5522 | #ifdef emacs |
b18215fc | 5523 | else if (range_table_exists) |
96cc36cc RS |
5524 | { |
5525 | int class_bits = CHARSET_RANGE_TABLE_BITS (&p[-1]); | |
5526 | ||
14473664 SM |
5527 | if ( (class_bits & BIT_LOWER && ISLOWER (c)) |
5528 | | (class_bits & BIT_MULTIBYTE) | |
96cc36cc RS |
5529 | | (class_bits & BIT_PUNCT && ISPUNCT (c)) |
5530 | | (class_bits & BIT_SPACE && ISSPACE (c)) | |
5531 | | (class_bits & BIT_UPPER && ISUPPER (c)) | |
5532 | | (class_bits & BIT_WORD && ISWORD (c))) | |
5533 | not = !not; | |
5534 | else | |
5535 | CHARSET_LOOKUP_RANGE_TABLE_RAW (not, c, range_table, count); | |
5536 | } | |
5537 | #endif /* emacs */ | |
fa9a63c5 | 5538 | |
96cc36cc RS |
5539 | if (range_table_exists) |
5540 | p = CHARSET_RANGE_TABLE_END (range_table, count); | |
5541 | else | |
5542 | p += CHARSET_BITMAP_SIZE (&p[-1]) + 1; | |
fa9a63c5 RM |
5543 | |
5544 | if (!not) goto fail; | |
5e69f11e | 5545 | |
b18215fc | 5546 | d += len; |
fa9a63c5 | 5547 | } |
8fb31792 | 5548 | break; |
fa9a63c5 RM |
5549 | |
5550 | ||
25fe55af | 5551 | /* The beginning of a group is represented by start_memory. |
505bde11 | 5552 | The argument is the register number. The text |
25fe55af | 5553 | matched within the group is recorded (in the internal |
7814e705 | 5554 | registers data structure) under the register number. */ |
25fe55af | 5555 | case start_memory: |
505bde11 SM |
5556 | DEBUG_PRINT2 ("EXECUTING start_memory %d:\n", *p); |
5557 | ||
5558 | /* In case we need to undo this operation (via backtracking). */ | |
5559 | PUSH_FAILURE_REG ((unsigned int)*p); | |
fa9a63c5 | 5560 | |
25fe55af | 5561 | regstart[*p] = d; |
4bb91c68 | 5562 | regend[*p] = NULL; /* probably unnecessary. -sm */ |
fa9a63c5 RM |
5563 | DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p])); |
5564 | ||
25fe55af | 5565 | /* Move past the register number and inner group count. */ |
505bde11 | 5566 | p += 1; |
25fe55af | 5567 | break; |
fa9a63c5 RM |
5568 | |
5569 | ||
25fe55af | 5570 | /* The stop_memory opcode represents the end of a group. Its |
505bde11 | 5571 | argument is the same as start_memory's: the register number. */ |
fa9a63c5 | 5572 | case stop_memory: |
505bde11 SM |
5573 | DEBUG_PRINT2 ("EXECUTING stop_memory %d:\n", *p); |
5574 | ||
5575 | assert (!REG_UNSET (regstart[*p])); | |
5576 | /* Strictly speaking, there should be code such as: | |
177c0ea7 | 5577 | |
0b32bf0e | 5578 | assert (REG_UNSET (regend[*p])); |
505bde11 SM |
5579 | PUSH_FAILURE_REGSTOP ((unsigned int)*p); |
5580 | ||
5581 | But the only info to be pushed is regend[*p] and it is known to | |
5582 | be UNSET, so there really isn't anything to push. | |
5583 | Not pushing anything, on the other hand deprives us from the | |
5584 | guarantee that regend[*p] is UNSET since undoing this operation | |
5585 | will not reset its value properly. This is not important since | |
5586 | the value will only be read on the next start_memory or at | |
5587 | the very end and both events can only happen if this stop_memory | |
5588 | is *not* undone. */ | |
fa9a63c5 | 5589 | |
25fe55af | 5590 | regend[*p] = d; |
fa9a63c5 RM |
5591 | DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p])); |
5592 | ||
25fe55af | 5593 | /* Move past the register number and the inner group count. */ |
505bde11 | 5594 | p += 1; |
25fe55af | 5595 | break; |
fa9a63c5 RM |
5596 | |
5597 | ||
5598 | /* \<digit> has been turned into a `duplicate' command which is | |
25fe55af RS |
5599 | followed by the numeric value of <digit> as the register number. */ |
5600 | case duplicate: | |
fa9a63c5 | 5601 | { |
66f0296e | 5602 | register re_char *d2, *dend2; |
7814e705 | 5603 | int regno = *p++; /* Get which register to match against. */ |
fa9a63c5 RM |
5604 | DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno); |
5605 | ||
7814e705 | 5606 | /* Can't back reference a group which we've never matched. */ |
25fe55af RS |
5607 | if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) |
5608 | goto fail; | |
5e69f11e | 5609 | |
7814e705 | 5610 | /* Where in input to try to start matching. */ |
25fe55af | 5611 | d2 = regstart[regno]; |
5e69f11e | 5612 | |
99633e97 SM |
5613 | /* Remember the start point to rollback upon failure. */ |
5614 | dfail = d; | |
5615 | ||
25fe55af RS |
5616 | /* Where to stop matching; if both the place to start and |
5617 | the place to stop matching are in the same string, then | |
5618 | set to the place to stop, otherwise, for now have to use | |
5619 | the end of the first string. */ | |
fa9a63c5 | 5620 | |
25fe55af | 5621 | dend2 = ((FIRST_STRING_P (regstart[regno]) |
fa9a63c5 RM |
5622 | == FIRST_STRING_P (regend[regno])) |
5623 | ? regend[regno] : end_match_1); | |
5624 | for (;;) | |
5625 | { | |
5626 | /* If necessary, advance to next segment in register | |
25fe55af | 5627 | contents. */ |
fa9a63c5 RM |
5628 | while (d2 == dend2) |
5629 | { | |
5630 | if (dend2 == end_match_2) break; | |
5631 | if (dend2 == regend[regno]) break; | |
5632 | ||
25fe55af RS |
5633 | /* End of string1 => advance to string2. */ |
5634 | d2 = string2; | |
5635 | dend2 = regend[regno]; | |
fa9a63c5 RM |
5636 | } |
5637 | /* At end of register contents => success */ | |
5638 | if (d2 == dend2) break; | |
5639 | ||
5640 | /* If necessary, advance to next segment in data. */ | |
5641 | PREFETCH (); | |
5642 | ||
5643 | /* How many characters left in this segment to match. */ | |
5644 | mcnt = dend - d; | |
5e69f11e | 5645 | |
fa9a63c5 | 5646 | /* Want how many consecutive characters we can match in |
25fe55af RS |
5647 | one shot, so, if necessary, adjust the count. */ |
5648 | if (mcnt > dend2 - d2) | |
fa9a63c5 | 5649 | mcnt = dend2 - d2; |
5e69f11e | 5650 | |
fa9a63c5 | 5651 | /* Compare that many; failure if mismatch, else move |
25fe55af | 5652 | past them. */ |
28703c16 | 5653 | if (RE_TRANSLATE_P (translate) |
02cb78b5 | 5654 | ? bcmp_translate (d, d2, mcnt, translate, target_multibyte) |
4bb91c68 | 5655 | : memcmp (d, d2, mcnt)) |
99633e97 SM |
5656 | { |
5657 | d = dfail; | |
5658 | goto fail; | |
5659 | } | |
fa9a63c5 | 5660 | d += mcnt, d2 += mcnt; |
fa9a63c5 RM |
5661 | } |
5662 | } | |
5663 | break; | |
5664 | ||
5665 | ||
25fe55af | 5666 | /* begline matches the empty string at the beginning of the string |
c0f9ea08 | 5667 | (unless `not_bol' is set in `bufp'), and after newlines. */ |
fa9a63c5 | 5668 | case begline: |
25fe55af | 5669 | DEBUG_PRINT1 ("EXECUTING begline.\n"); |
5e69f11e | 5670 | |
25fe55af RS |
5671 | if (AT_STRINGS_BEG (d)) |
5672 | { | |
5673 | if (!bufp->not_bol) break; | |
5674 | } | |
419d1c74 | 5675 | else |
25fe55af | 5676 | { |
bf216479 | 5677 | unsigned c; |
419d1c74 | 5678 | GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2); |
c0f9ea08 | 5679 | if (c == '\n') |
419d1c74 | 5680 | break; |
25fe55af RS |
5681 | } |
5682 | /* In all other cases, we fail. */ | |
5683 | goto fail; | |
fa9a63c5 RM |
5684 | |
5685 | ||
25fe55af | 5686 | /* endline is the dual of begline. */ |
fa9a63c5 | 5687 | case endline: |
25fe55af | 5688 | DEBUG_PRINT1 ("EXECUTING endline.\n"); |
fa9a63c5 | 5689 | |
25fe55af RS |
5690 | if (AT_STRINGS_END (d)) |
5691 | { | |
5692 | if (!bufp->not_eol) break; | |
5693 | } | |
f1ad044f | 5694 | else |
25fe55af | 5695 | { |
f1ad044f | 5696 | PREFETCH_NOLIMIT (); |
c0f9ea08 | 5697 | if (*d == '\n') |
f1ad044f | 5698 | break; |
25fe55af RS |
5699 | } |
5700 | goto fail; | |
fa9a63c5 RM |
5701 | |
5702 | ||
5703 | /* Match at the very beginning of the data. */ | |
25fe55af RS |
5704 | case begbuf: |
5705 | DEBUG_PRINT1 ("EXECUTING begbuf.\n"); | |
5706 | if (AT_STRINGS_BEG (d)) | |
5707 | break; | |
5708 | goto fail; | |
fa9a63c5 RM |
5709 | |
5710 | ||
5711 | /* Match at the very end of the data. */ | |
25fe55af RS |
5712 | case endbuf: |
5713 | DEBUG_PRINT1 ("EXECUTING endbuf.\n"); | |
fa9a63c5 RM |
5714 | if (AT_STRINGS_END (d)) |
5715 | break; | |
25fe55af | 5716 | goto fail; |
5e69f11e | 5717 | |
5e69f11e | 5718 | |
25fe55af RS |
5719 | /* on_failure_keep_string_jump is used to optimize `.*\n'. It |
5720 | pushes NULL as the value for the string on the stack. Then | |
505bde11 | 5721 | `POP_FAILURE_POINT' will keep the current value for the |
25fe55af | 5722 | string, instead of restoring it. To see why, consider |
7814e705 | 5723 | matching `foo\nbar' against `.*\n'. The .* matches the foo; |
25fe55af RS |
5724 | then the . fails against the \n. But the next thing we want |
5725 | to do is match the \n against the \n; if we restored the | |
5726 | string value, we would be back at the foo. | |
5727 | ||
5728 | Because this is used only in specific cases, we don't need to | |
5729 | check all the things that `on_failure_jump' does, to make | |
5730 | sure the right things get saved on the stack. Hence we don't | |
5731 | share its code. The only reason to push anything on the | |
5732 | stack at all is that otherwise we would have to change | |
5733 | `anychar's code to do something besides goto fail in this | |
5734 | case; that seems worse than this. */ | |
5735 | case on_failure_keep_string_jump: | |
505bde11 SM |
5736 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
5737 | DEBUG_PRINT3 ("EXECUTING on_failure_keep_string_jump %d (to %p):\n", | |
5738 | mcnt, p + mcnt); | |
fa9a63c5 | 5739 | |
505bde11 SM |
5740 | PUSH_FAILURE_POINT (p - 3, NULL); |
5741 | break; | |
5742 | ||
0683b6fa SM |
5743 | /* A nasty loop is introduced by the non-greedy *? and +?. |
5744 | With such loops, the stack only ever contains one failure point | |
5745 | at a time, so that a plain on_failure_jump_loop kind of | |
5746 | cycle detection cannot work. Worse yet, such a detection | |
5747 | can not only fail to detect a cycle, but it can also wrongly | |
5748 | detect a cycle (between different instantiations of the same | |
6df42991 | 5749 | loop). |
0683b6fa SM |
5750 | So the method used for those nasty loops is a little different: |
5751 | We use a special cycle-detection-stack-frame which is pushed | |
5752 | when the on_failure_jump_nastyloop failure-point is *popped*. | |
5753 | This special frame thus marks the beginning of one iteration | |
5754 | through the loop and we can hence easily check right here | |
5755 | whether something matched between the beginning and the end of | |
5756 | the loop. */ | |
5757 | case on_failure_jump_nastyloop: | |
5758 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
5759 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_nastyloop %d (to %p):\n", | |
5760 | mcnt, p + mcnt); | |
5761 | ||
5762 | assert ((re_opcode_t)p[-4] == no_op); | |
6df42991 SM |
5763 | { |
5764 | int cycle = 0; | |
5765 | CHECK_INFINITE_LOOP (p - 4, d); | |
5766 | if (!cycle) | |
5767 | /* If there's a cycle, just continue without pushing | |
5768 | this failure point. The failure point is the "try again" | |
5769 | option, which shouldn't be tried. | |
5770 | We want (x?)*?y\1z to match both xxyz and xxyxz. */ | |
5771 | PUSH_FAILURE_POINT (p - 3, d); | |
5772 | } | |
0683b6fa SM |
5773 | break; |
5774 | ||
4e8a9132 SM |
5775 | /* Simple loop detecting on_failure_jump: just check on the |
5776 | failure stack if the same spot was already hit earlier. */ | |
505bde11 SM |
5777 | case on_failure_jump_loop: |
5778 | on_failure: | |
5779 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
5780 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_loop %d (to %p):\n", | |
5781 | mcnt, p + mcnt); | |
6df42991 SM |
5782 | { |
5783 | int cycle = 0; | |
5784 | CHECK_INFINITE_LOOP (p - 3, d); | |
5785 | if (cycle) | |
5786 | /* If there's a cycle, get out of the loop, as if the matching | |
5787 | had failed. We used to just `goto fail' here, but that was | |
5788 | aborting the search a bit too early: we want to keep the | |
5789 | empty-loop-match and keep matching after the loop. | |
5790 | We want (x?)*y\1z to match both xxyz and xxyxz. */ | |
5791 | p += mcnt; | |
5792 | else | |
5793 | PUSH_FAILURE_POINT (p - 3, d); | |
5794 | } | |
25fe55af | 5795 | break; |
fa9a63c5 RM |
5796 | |
5797 | ||
5798 | /* Uses of on_failure_jump: | |
5e69f11e | 5799 | |
25fe55af RS |
5800 | Each alternative starts with an on_failure_jump that points |
5801 | to the beginning of the next alternative. Each alternative | |
5802 | except the last ends with a jump that in effect jumps past | |
5803 | the rest of the alternatives. (They really jump to the | |
5804 | ending jump of the following alternative, because tensioning | |
5805 | these jumps is a hassle.) | |
fa9a63c5 | 5806 | |
25fe55af RS |
5807 | Repeats start with an on_failure_jump that points past both |
5808 | the repetition text and either the following jump or | |
5809 | pop_failure_jump back to this on_failure_jump. */ | |
fa9a63c5 | 5810 | case on_failure_jump: |
25fe55af | 5811 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
505bde11 SM |
5812 | DEBUG_PRINT3 ("EXECUTING on_failure_jump %d (to %p):\n", |
5813 | mcnt, p + mcnt); | |
25fe55af | 5814 | |
505bde11 | 5815 | PUSH_FAILURE_POINT (p -3, d); |
25fe55af RS |
5816 | break; |
5817 | ||
4e8a9132 | 5818 | /* This operation is used for greedy *. |
505bde11 SM |
5819 | Compare the beginning of the repeat with what in the |
5820 | pattern follows its end. If we can establish that there | |
5821 | is nothing that they would both match, i.e., that we | |
5822 | would have to backtrack because of (as in, e.g., `a*a') | |
5823 | then we can use a non-backtracking loop based on | |
4e8a9132 | 5824 | on_failure_keep_string_jump instead of on_failure_jump. */ |
505bde11 | 5825 | case on_failure_jump_smart: |
25fe55af | 5826 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
505bde11 SM |
5827 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_smart %d (to %p).\n", |
5828 | mcnt, p + mcnt); | |
25fe55af | 5829 | { |
01618498 | 5830 | re_char *p1 = p; /* Next operation. */ |
6dcf2d0e SM |
5831 | /* Here, we discard `const', making re_match non-reentrant. */ |
5832 | unsigned char *p2 = (unsigned char*) p + mcnt; /* Jump dest. */ | |
5833 | unsigned char *p3 = (unsigned char*) p - 3; /* opcode location. */ | |
fa9a63c5 | 5834 | |
505bde11 SM |
5835 | p -= 3; /* Reset so that we will re-execute the |
5836 | instruction once it's been changed. */ | |
fa9a63c5 | 5837 | |
4e8a9132 SM |
5838 | EXTRACT_NUMBER (mcnt, p2 - 2); |
5839 | ||
5840 | /* Ensure this is a indeed the trivial kind of loop | |
5841 | we are expecting. */ | |
5842 | assert (skip_one_char (p1) == p2 - 3); | |
5843 | assert ((re_opcode_t) p2[-3] == jump && p2 + mcnt == p); | |
99633e97 | 5844 | DEBUG_STATEMENT (debug += 2); |
505bde11 | 5845 | if (mutually_exclusive_p (bufp, p1, p2)) |
fa9a63c5 | 5846 | { |
505bde11 | 5847 | /* Use a fast `on_failure_keep_string_jump' loop. */ |
4e8a9132 | 5848 | DEBUG_PRINT1 (" smart exclusive => fast loop.\n"); |
01618498 | 5849 | *p3 = (unsigned char) on_failure_keep_string_jump; |
4e8a9132 | 5850 | STORE_NUMBER (p2 - 2, mcnt + 3); |
25fe55af | 5851 | } |
505bde11 | 5852 | else |
fa9a63c5 | 5853 | { |
505bde11 SM |
5854 | /* Default to a safe `on_failure_jump' loop. */ |
5855 | DEBUG_PRINT1 (" smart default => slow loop.\n"); | |
01618498 | 5856 | *p3 = (unsigned char) on_failure_jump; |
fa9a63c5 | 5857 | } |
99633e97 | 5858 | DEBUG_STATEMENT (debug -= 2); |
25fe55af | 5859 | } |
505bde11 | 5860 | break; |
25fe55af RS |
5861 | |
5862 | /* Unconditionally jump (without popping any failure points). */ | |
5863 | case jump: | |
fa9a63c5 | 5864 | unconditional_jump: |
5b370c2b | 5865 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 5866 | EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ |
25fe55af | 5867 | DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt); |
7814e705 | 5868 | p += mcnt; /* Do the jump. */ |
505bde11 | 5869 | DEBUG_PRINT2 ("(to %p).\n", p); |
25fe55af RS |
5870 | break; |
5871 | ||
5872 | ||
25fe55af RS |
5873 | /* Have to succeed matching what follows at least n times. |
5874 | After that, handle like `on_failure_jump'. */ | |
5875 | case succeed_n: | |
01618498 | 5876 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af RS |
5877 | EXTRACT_NUMBER (mcnt, p + 2); |
5878 | DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt); | |
5e69f11e | 5879 | |
dc1e502d SM |
5880 | /* Originally, mcnt is how many times we HAVE to succeed. */ |
5881 | if (mcnt != 0) | |
25fe55af | 5882 | { |
6dcf2d0e SM |
5883 | /* Here, we discard `const', making re_match non-reentrant. */ |
5884 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5885 | mcnt--; |
01618498 SM |
5886 | p += 4; |
5887 | PUSH_NUMBER (p2, mcnt); | |
25fe55af | 5888 | } |
dc1e502d SM |
5889 | else |
5890 | /* The two bytes encoding mcnt == 0 are two no_op opcodes. */ | |
5891 | goto on_failure; | |
25fe55af RS |
5892 | break; |
5893 | ||
5894 | case jump_n: | |
01618498 | 5895 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af RS |
5896 | EXTRACT_NUMBER (mcnt, p + 2); |
5897 | DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt); | |
5898 | ||
5899 | /* Originally, this is how many times we CAN jump. */ | |
dc1e502d | 5900 | if (mcnt != 0) |
25fe55af | 5901 | { |
6dcf2d0e SM |
5902 | /* Here, we discard `const', making re_match non-reentrant. */ |
5903 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5904 | mcnt--; |
01618498 | 5905 | PUSH_NUMBER (p2, mcnt); |
dc1e502d | 5906 | goto unconditional_jump; |
25fe55af RS |
5907 | } |
5908 | /* If don't have to jump any more, skip over the rest of command. */ | |
5e69f11e RM |
5909 | else |
5910 | p += 4; | |
25fe55af | 5911 | break; |
5e69f11e | 5912 | |
fa9a63c5 RM |
5913 | case set_number_at: |
5914 | { | |
01618498 | 5915 | unsigned char *p2; /* Location of the counter. */ |
25fe55af | 5916 | DEBUG_PRINT1 ("EXECUTING set_number_at.\n"); |
fa9a63c5 | 5917 | |
25fe55af | 5918 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
6dcf2d0e SM |
5919 | /* Here, we discard `const', making re_match non-reentrant. */ |
5920 | p2 = (unsigned char*) p + mcnt; | |
01618498 | 5921 | /* Signedness doesn't matter since we only copy MCNT's bits . */ |
25fe55af | 5922 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
01618498 SM |
5923 | DEBUG_PRINT3 (" Setting %p to %d.\n", p2, mcnt); |
5924 | PUSH_NUMBER (p2, mcnt); | |
25fe55af RS |
5925 | break; |
5926 | } | |
9121ca40 KH |
5927 | |
5928 | case wordbound: | |
66f0296e | 5929 | case notwordbound: |
19ed5445 PE |
5930 | { |
5931 | boolean not = (re_opcode_t) *(p - 1) == notwordbound; | |
5932 | DEBUG_PRINT2 ("EXECUTING %swordbound.\n", not?"not":""); | |
fa9a63c5 | 5933 | |
19ed5445 | 5934 | /* We SUCCEED (or FAIL) in one of the following cases: */ |
9121ca40 | 5935 | |
19ed5445 PE |
5936 | /* Case 1: D is at the beginning or the end of string. */ |
5937 | if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) | |
5938 | not = !not; | |
5939 | else | |
5940 | { | |
5941 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5942 | is the character at D, and S2 is the syntax of C2. */ | |
5943 | re_wchar_t c1, c2; | |
5944 | int s1, s2; | |
5945 | int dummy; | |
b18215fc | 5946 | #ifdef emacs |
d1dfb56c EZ |
5947 | ssize_t offset = PTR_TO_OFFSET (d - 1); |
5948 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 | 5949 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5950 | #endif |
19ed5445 PE |
5951 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
5952 | s1 = SYNTAX (c1); | |
b18215fc | 5953 | #ifdef emacs |
19ed5445 | 5954 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
25fe55af | 5955 | #endif |
19ed5445 PE |
5956 | PREFETCH_NOLIMIT (); |
5957 | GET_CHAR_AFTER (c2, d, dummy); | |
5958 | s2 = SYNTAX (c2); | |
5959 | ||
5960 | if (/* Case 2: Only one of S1 and S2 is Sword. */ | |
5961 | ((s1 == Sword) != (s2 == Sword)) | |
5962 | /* Case 3: Both of S1 and S2 are Sword, and macro | |
5963 | WORD_BOUNDARY_P (C1, C2) returns nonzero. */ | |
5964 | || ((s1 == Sword) && WORD_BOUNDARY_P (c1, c2))) | |
5965 | not = !not; | |
5966 | } | |
5967 | if (not) | |
5968 | break; | |
5969 | else | |
5970 | goto fail; | |
5971 | } | |
fa9a63c5 RM |
5972 | |
5973 | case wordbeg: | |
25fe55af | 5974 | DEBUG_PRINT1 ("EXECUTING wordbeg.\n"); |
fa9a63c5 | 5975 | |
b18215fc RS |
5976 | /* We FAIL in one of the following cases: */ |
5977 | ||
7814e705 | 5978 | /* Case 1: D is at the end of string. */ |
b18215fc | 5979 | if (AT_STRINGS_END (d)) |
99633e97 | 5980 | goto fail; |
b18215fc RS |
5981 | else |
5982 | { | |
5983 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5984 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
5985 | re_wchar_t c1, c2; |
5986 | int s1, s2; | |
bf216479 | 5987 | int dummy; |
fa9a63c5 | 5988 | #ifdef emacs |
d1dfb56c EZ |
5989 | ssize_t offset = PTR_TO_OFFSET (d); |
5990 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 5991 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5992 | #endif |
99633e97 | 5993 | PREFETCH (); |
6fdd04b0 | 5994 | GET_CHAR_AFTER (c2, d, dummy); |
b18215fc | 5995 | s2 = SYNTAX (c2); |
177c0ea7 | 5996 | |
b18215fc RS |
5997 | /* Case 2: S2 is not Sword. */ |
5998 | if (s2 != Sword) | |
5999 | goto fail; | |
6000 | ||
6001 | /* Case 3: D is not at the beginning of string ... */ | |
6002 | if (!AT_STRINGS_BEG (d)) | |
6003 | { | |
6004 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6005 | #ifdef emacs | |
5d967c7a | 6006 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); |
25fe55af | 6007 | #endif |
b18215fc RS |
6008 | s1 = SYNTAX (c1); |
6009 | ||
6010 | /* ... and S1 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6011 | returns 0. */ |
b18215fc RS |
6012 | if ((s1 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
6013 | goto fail; | |
6014 | } | |
6015 | } | |
e318085a RS |
6016 | break; |
6017 | ||
b18215fc | 6018 | case wordend: |
25fe55af | 6019 | DEBUG_PRINT1 ("EXECUTING wordend.\n"); |
b18215fc RS |
6020 | |
6021 | /* We FAIL in one of the following cases: */ | |
6022 | ||
6023 | /* Case 1: D is at the beginning of string. */ | |
6024 | if (AT_STRINGS_BEG (d)) | |
e318085a | 6025 | goto fail; |
b18215fc RS |
6026 | else |
6027 | { | |
6028 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6029 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
6030 | re_wchar_t c1, c2; |
6031 | int s1, s2; | |
bf216479 | 6032 | int dummy; |
5d967c7a | 6033 | #ifdef emacs |
d1dfb56c EZ |
6034 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6035 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 6036 | UPDATE_SYNTAX_TABLE (charpos); |
5d967c7a | 6037 | #endif |
99633e97 | 6038 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
b18215fc RS |
6039 | s1 = SYNTAX (c1); |
6040 | ||
6041 | /* Case 2: S1 is not Sword. */ | |
6042 | if (s1 != Sword) | |
6043 | goto fail; | |
6044 | ||
6045 | /* Case 3: D is not at the end of string ... */ | |
6046 | if (!AT_STRINGS_END (d)) | |
6047 | { | |
f1ad044f | 6048 | PREFETCH_NOLIMIT (); |
6fdd04b0 | 6049 | GET_CHAR_AFTER (c2, d, dummy); |
5d967c7a RS |
6050 | #ifdef emacs |
6051 | UPDATE_SYNTAX_TABLE_FORWARD (charpos); | |
6052 | #endif | |
b18215fc RS |
6053 | s2 = SYNTAX (c2); |
6054 | ||
6055 | /* ... and S2 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6056 | returns 0. */ |
b18215fc | 6057 | if ((s2 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
25fe55af | 6058 | goto fail; |
b18215fc RS |
6059 | } |
6060 | } | |
e318085a RS |
6061 | break; |
6062 | ||
669fa600 SM |
6063 | case symbeg: |
6064 | DEBUG_PRINT1 ("EXECUTING symbeg.\n"); | |
6065 | ||
6066 | /* We FAIL in one of the following cases: */ | |
6067 | ||
7814e705 | 6068 | /* Case 1: D is at the end of string. */ |
669fa600 SM |
6069 | if (AT_STRINGS_END (d)) |
6070 | goto fail; | |
6071 | else | |
6072 | { | |
6073 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6074 | is the character at D, and S2 is the syntax of C2. */ | |
6075 | re_wchar_t c1, c2; | |
6076 | int s1, s2; | |
6077 | #ifdef emacs | |
d1dfb56c EZ |
6078 | ssize_t offset = PTR_TO_OFFSET (d); |
6079 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6080 | UPDATE_SYNTAX_TABLE (charpos); |
6081 | #endif | |
6082 | PREFETCH (); | |
62a6e103 | 6083 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6084 | s2 = SYNTAX (c2); |
7814e705 | 6085 | |
669fa600 SM |
6086 | /* Case 2: S2 is neither Sword nor Ssymbol. */ |
6087 | if (s2 != Sword && s2 != Ssymbol) | |
6088 | goto fail; | |
6089 | ||
6090 | /* Case 3: D is not at the beginning of string ... */ | |
6091 | if (!AT_STRINGS_BEG (d)) | |
6092 | { | |
6093 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6094 | #ifdef emacs | |
6095 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); | |
6096 | #endif | |
6097 | s1 = SYNTAX (c1); | |
6098 | ||
6099 | /* ... and S1 is Sword or Ssymbol. */ | |
6100 | if (s1 == Sword || s1 == Ssymbol) | |
6101 | goto fail; | |
6102 | } | |
6103 | } | |
6104 | break; | |
6105 | ||
6106 | case symend: | |
6107 | DEBUG_PRINT1 ("EXECUTING symend.\n"); | |
6108 | ||
6109 | /* We FAIL in one of the following cases: */ | |
6110 | ||
6111 | /* Case 1: D is at the beginning of string. */ | |
6112 | if (AT_STRINGS_BEG (d)) | |
6113 | goto fail; | |
6114 | else | |
6115 | { | |
6116 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6117 | is the character at D, and S2 is the syntax of C2. */ | |
6118 | re_wchar_t c1, c2; | |
6119 | int s1, s2; | |
6120 | #ifdef emacs | |
d1dfb56c EZ |
6121 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6122 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6123 | UPDATE_SYNTAX_TABLE (charpos); |
6124 | #endif | |
6125 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6126 | s1 = SYNTAX (c1); | |
6127 | ||
6128 | /* Case 2: S1 is neither Ssymbol nor Sword. */ | |
6129 | if (s1 != Sword && s1 != Ssymbol) | |
6130 | goto fail; | |
6131 | ||
6132 | /* Case 3: D is not at the end of string ... */ | |
6133 | if (!AT_STRINGS_END (d)) | |
6134 | { | |
6135 | PREFETCH_NOLIMIT (); | |
62a6e103 | 6136 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6137 | #ifdef emacs |
134579f2 | 6138 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
669fa600 SM |
6139 | #endif |
6140 | s2 = SYNTAX (c2); | |
6141 | ||
6142 | /* ... and S2 is Sword or Ssymbol. */ | |
6143 | if (s2 == Sword || s2 == Ssymbol) | |
6144 | goto fail; | |
b18215fc RS |
6145 | } |
6146 | } | |
e318085a RS |
6147 | break; |
6148 | ||
fa9a63c5 | 6149 | case syntaxspec: |
1fb352e0 | 6150 | case notsyntaxspec: |
b18215fc | 6151 | { |
19ed5445 PE |
6152 | boolean not = (re_opcode_t) *(p - 1) == notsyntaxspec; |
6153 | mcnt = *p++; | |
6154 | DEBUG_PRINT3 ("EXECUTING %ssyntaxspec %d.\n", not?"not":"", mcnt); | |
6155 | PREFETCH (); | |
6156 | #ifdef emacs | |
6157 | { | |
d1dfb56c EZ |
6158 | ssize_t offset = PTR_TO_OFFSET (d); |
6159 | ssize_t pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 PE |
6160 | UPDATE_SYNTAX_TABLE (pos1); |
6161 | } | |
25fe55af | 6162 | #endif |
19ed5445 PE |
6163 | { |
6164 | int len; | |
6165 | re_wchar_t c; | |
b18215fc | 6166 | |
19ed5445 PE |
6167 | GET_CHAR_AFTER (c, d, len); |
6168 | if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not) | |
6169 | goto fail; | |
6170 | d += len; | |
6171 | } | |
b18215fc | 6172 | } |
8fb31792 | 6173 | break; |
fa9a63c5 | 6174 | |
b18215fc | 6175 | #ifdef emacs |
1fb352e0 SM |
6176 | case before_dot: |
6177 | DEBUG_PRINT1 ("EXECUTING before_dot.\n"); | |
6178 | if (PTR_BYTE_POS (d) >= PT_BYTE) | |
fa9a63c5 | 6179 | goto fail; |
b18215fc RS |
6180 | break; |
6181 | ||
1fb352e0 SM |
6182 | case at_dot: |
6183 | DEBUG_PRINT1 ("EXECUTING at_dot.\n"); | |
6184 | if (PTR_BYTE_POS (d) != PT_BYTE) | |
6185 | goto fail; | |
6186 | break; | |
b18215fc | 6187 | |
1fb352e0 SM |
6188 | case after_dot: |
6189 | DEBUG_PRINT1 ("EXECUTING after_dot.\n"); | |
6190 | if (PTR_BYTE_POS (d) <= PT_BYTE) | |
6191 | goto fail; | |
e318085a | 6192 | break; |
fa9a63c5 | 6193 | |
1fb352e0 | 6194 | case categoryspec: |
b18215fc | 6195 | case notcategoryspec: |
b18215fc | 6196 | { |
8fb31792 PE |
6197 | boolean not = (re_opcode_t) *(p - 1) == notcategoryspec; |
6198 | mcnt = *p++; | |
6199 | DEBUG_PRINT3 ("EXECUTING %scategoryspec %d.\n", | |
6200 | not?"not":"", mcnt); | |
6201 | PREFETCH (); | |
01618498 | 6202 | |
8fb31792 PE |
6203 | { |
6204 | int len; | |
6205 | re_wchar_t c; | |
6206 | GET_CHAR_AFTER (c, d, len); | |
6207 | if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not) | |
6208 | goto fail; | |
6209 | d += len; | |
6210 | } | |
b18215fc | 6211 | } |
fa9a63c5 | 6212 | break; |
5e69f11e | 6213 | |
1fb352e0 | 6214 | #endif /* emacs */ |
5e69f11e | 6215 | |
0b32bf0e SM |
6216 | default: |
6217 | abort (); | |
fa9a63c5 | 6218 | } |
b18215fc | 6219 | continue; /* Successfully executed one pattern command; keep going. */ |
fa9a63c5 RM |
6220 | |
6221 | ||
6222 | /* We goto here if a matching operation fails. */ | |
6223 | fail: | |
5b370c2b | 6224 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 6225 | if (!FAIL_STACK_EMPTY ()) |
505bde11 | 6226 | { |
01618498 | 6227 | re_char *str, *pat; |
505bde11 | 6228 | /* A restart point is known. Restore to that state. */ |
0b32bf0e SM |
6229 | DEBUG_PRINT1 ("\nFAIL:\n"); |
6230 | POP_FAILURE_POINT (str, pat); | |
7393bcbb | 6231 | switch (*pat++) |
505bde11 SM |
6232 | { |
6233 | case on_failure_keep_string_jump: | |
6234 | assert (str == NULL); | |
6235 | goto continue_failure_jump; | |
6236 | ||
0683b6fa SM |
6237 | case on_failure_jump_nastyloop: |
6238 | assert ((re_opcode_t)pat[-2] == no_op); | |
6239 | PUSH_FAILURE_POINT (pat - 2, str); | |
6240 | /* Fallthrough */ | |
6241 | ||
505bde11 SM |
6242 | case on_failure_jump_loop: |
6243 | case on_failure_jump: | |
6244 | case succeed_n: | |
6245 | d = str; | |
6246 | continue_failure_jump: | |
6247 | EXTRACT_NUMBER_AND_INCR (mcnt, pat); | |
6248 | p = pat + mcnt; | |
6249 | break; | |
b18215fc | 6250 | |
0683b6fa SM |
6251 | case no_op: |
6252 | /* A special frame used for nastyloops. */ | |
6253 | goto fail; | |
6254 | ||
505bde11 | 6255 | default: |
5e617bc2 | 6256 | abort (); |
505bde11 | 6257 | } |
fa9a63c5 | 6258 | |
505bde11 | 6259 | assert (p >= bufp->buffer && p <= pend); |
b18215fc | 6260 | |
0b32bf0e | 6261 | if (d >= string1 && d <= end1) |
fa9a63c5 | 6262 | dend = end_match_1; |
0b32bf0e | 6263 | } |
fa9a63c5 | 6264 | else |
0b32bf0e | 6265 | break; /* Matching at this starting point really fails. */ |
fa9a63c5 RM |
6266 | } /* for (;;) */ |
6267 | ||
6268 | if (best_regs_set) | |
6269 | goto restore_best_regs; | |
6270 | ||
6271 | FREE_VARIABLES (); | |
6272 | ||
b18215fc | 6273 | return -1; /* Failure to match. */ |
fa9a63c5 RM |
6274 | } /* re_match_2 */ |
6275 | \f | |
6276 | /* Subroutine definitions for re_match_2. */ | |
6277 | ||
fa9a63c5 RM |
6278 | /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN |
6279 | bytes; nonzero otherwise. */ | |
5e69f11e | 6280 | |
fa9a63c5 | 6281 | static int |
d1dfb56c | 6282 | bcmp_translate (const re_char *s1, const re_char *s2, register ssize_t len, |
438105ed | 6283 | RE_TRANSLATE_TYPE translate, const int target_multibyte) |
fa9a63c5 | 6284 | { |
2d1675e4 SM |
6285 | register re_char *p1 = s1, *p2 = s2; |
6286 | re_char *p1_end = s1 + len; | |
6287 | re_char *p2_end = s2 + len; | |
e934739e | 6288 | |
4bb91c68 SM |
6289 | /* FIXME: Checking both p1 and p2 presumes that the two strings might have |
6290 | different lengths, but relying on a single `len' would break this. -sm */ | |
6291 | while (p1 < p1_end && p2 < p2_end) | |
fa9a63c5 | 6292 | { |
e934739e | 6293 | int p1_charlen, p2_charlen; |
01618498 | 6294 | re_wchar_t p1_ch, p2_ch; |
e934739e | 6295 | |
6fdd04b0 KH |
6296 | GET_CHAR_AFTER (p1_ch, p1, p1_charlen); |
6297 | GET_CHAR_AFTER (p2_ch, p2, p2_charlen); | |
e934739e RS |
6298 | |
6299 | if (RE_TRANSLATE (translate, p1_ch) | |
6300 | != RE_TRANSLATE (translate, p2_ch)) | |
bc192b5b | 6301 | return 1; |
e934739e RS |
6302 | |
6303 | p1 += p1_charlen, p2 += p2_charlen; | |
fa9a63c5 | 6304 | } |
e934739e RS |
6305 | |
6306 | if (p1 != p1_end || p2 != p2_end) | |
6307 | return 1; | |
6308 | ||
fa9a63c5 RM |
6309 | return 0; |
6310 | } | |
6311 | \f | |
6312 | /* Entry points for GNU code. */ | |
6313 | ||
6314 | /* re_compile_pattern is the GNU regular expression compiler: it | |
6315 | compiles PATTERN (of length SIZE) and puts the result in BUFP. | |
6316 | Returns 0 if the pattern was valid, otherwise an error string. | |
5e69f11e | 6317 | |
fa9a63c5 RM |
6318 | Assumes the `allocated' (and perhaps `buffer') and `translate' fields |
6319 | are set in BUFP on entry. | |
5e69f11e | 6320 | |
b18215fc | 6321 | We call regex_compile to do the actual compilation. */ |
fa9a63c5 RM |
6322 | |
6323 | const char * | |
d1dfb56c EZ |
6324 | re_compile_pattern (const char *pattern, size_t length, |
6325 | struct re_pattern_buffer *bufp) | |
fa9a63c5 RM |
6326 | { |
6327 | reg_errcode_t ret; | |
5e69f11e | 6328 | |
fa9a63c5 RM |
6329 | /* GNU code is written to assume at least RE_NREGS registers will be set |
6330 | (and at least one extra will be -1). */ | |
6331 | bufp->regs_allocated = REGS_UNALLOCATED; | |
5e69f11e | 6332 | |
fa9a63c5 RM |
6333 | /* And GNU code determines whether or not to get register information |
6334 | by passing null for the REGS argument to re_match, etc., not by | |
6335 | setting no_sub. */ | |
6336 | bufp->no_sub = 0; | |
5e69f11e | 6337 | |
4bb91c68 | 6338 | ret = regex_compile ((re_char*) pattern, length, re_syntax_options, bufp); |
fa9a63c5 RM |
6339 | |
6340 | if (!ret) | |
6341 | return NULL; | |
6342 | return gettext (re_error_msgid[(int) ret]); | |
5e69f11e | 6343 | } |
c0f9ea08 | 6344 | WEAK_ALIAS (__re_compile_pattern, re_compile_pattern) |
fa9a63c5 | 6345 | \f |
b18215fc RS |
6346 | /* Entry points compatible with 4.2 BSD regex library. We don't define |
6347 | them unless specifically requested. */ | |
fa9a63c5 | 6348 | |
0b32bf0e | 6349 | #if defined _REGEX_RE_COMP || defined _LIBC |
fa9a63c5 RM |
6350 | |
6351 | /* BSD has one and only one pattern buffer. */ | |
6352 | static struct re_pattern_buffer re_comp_buf; | |
6353 | ||
6354 | char * | |
0b32bf0e | 6355 | # ifdef _LIBC |
48afdd44 RM |
6356 | /* Make these definitions weak in libc, so POSIX programs can redefine |
6357 | these names if they don't use our functions, and still use | |
6358 | regcomp/regexec below without link errors. */ | |
6359 | weak_function | |
0b32bf0e | 6360 | # endif |
31011111 | 6361 | re_comp (const char *s) |
fa9a63c5 RM |
6362 | { |
6363 | reg_errcode_t ret; | |
5e69f11e | 6364 | |
fa9a63c5 RM |
6365 | if (!s) |
6366 | { | |
6367 | if (!re_comp_buf.buffer) | |
0b32bf0e | 6368 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
a60198e5 | 6369 | return (char *) gettext ("No previous regular expression"); |
fa9a63c5 RM |
6370 | return 0; |
6371 | } | |
6372 | ||
6373 | if (!re_comp_buf.buffer) | |
6374 | { | |
38182d90 | 6375 | re_comp_buf.buffer = malloc (200); |
fa9a63c5 | 6376 | if (re_comp_buf.buffer == NULL) |
0b32bf0e SM |
6377 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6378 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6379 | re_comp_buf.allocated = 200; |
6380 | ||
38182d90 | 6381 | re_comp_buf.fastmap = malloc (1 << BYTEWIDTH); |
fa9a63c5 | 6382 | if (re_comp_buf.fastmap == NULL) |
a60198e5 SM |
6383 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6384 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6385 | } |
6386 | ||
6387 | /* Since `re_exec' always passes NULL for the `regs' argument, we | |
6388 | don't need to initialize the pattern buffer fields which affect it. */ | |
6389 | ||
fa9a63c5 | 6390 | ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); |
5e69f11e | 6391 | |
fa9a63c5 RM |
6392 | if (!ret) |
6393 | return NULL; | |
6394 | ||
6395 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | |
6396 | return (char *) gettext (re_error_msgid[(int) ret]); | |
6397 | } | |
6398 | ||
6399 | ||
31011111 | 6400 | int |
0b32bf0e | 6401 | # ifdef _LIBC |
48afdd44 | 6402 | weak_function |
0b32bf0e | 6403 | # endif |
d1dfb56c | 6404 | re_exec (const char *s) |
fa9a63c5 | 6405 | { |
d1dfb56c | 6406 | const size_t len = strlen (s); |
908589fd AS |
6407 | return (re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0) |
6408 | >= 0); | |
fa9a63c5 RM |
6409 | } |
6410 | #endif /* _REGEX_RE_COMP */ | |
6411 | \f | |
6412 | /* POSIX.2 functions. Don't define these for Emacs. */ | |
6413 | ||
6414 | #ifndef emacs | |
6415 | ||
6416 | /* regcomp takes a regular expression as a string and compiles it. | |
6417 | ||
b18215fc | 6418 | PREG is a regex_t *. We do not expect any fields to be initialized, |
fa9a63c5 RM |
6419 | since POSIX says we shouldn't. Thus, we set |
6420 | ||
6421 | `buffer' to the compiled pattern; | |
6422 | `used' to the length of the compiled pattern; | |
6423 | `syntax' to RE_SYNTAX_POSIX_EXTENDED if the | |
6424 | REG_EXTENDED bit in CFLAGS is set; otherwise, to | |
6425 | RE_SYNTAX_POSIX_BASIC; | |
c0f9ea08 SM |
6426 | `fastmap' to an allocated space for the fastmap; |
6427 | `fastmap_accurate' to zero; | |
fa9a63c5 RM |
6428 | `re_nsub' to the number of subexpressions in PATTERN. |
6429 | ||
6430 | PATTERN is the address of the pattern string. | |
6431 | ||
6432 | CFLAGS is a series of bits which affect compilation. | |
6433 | ||
6434 | If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we | |
6435 | use POSIX basic syntax. | |
6436 | ||
6437 | If REG_NEWLINE is set, then . and [^...] don't match newline. | |
6438 | Also, regexec will try a match beginning after every newline. | |
6439 | ||
6440 | If REG_ICASE is set, then we considers upper- and lowercase | |
6441 | versions of letters to be equivalent when matching. | |
6442 | ||
6443 | If REG_NOSUB is set, then when PREG is passed to regexec, that | |
6444 | routine will report only success or failure, and nothing about the | |
6445 | registers. | |
6446 | ||
b18215fc | 6447 | It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for |
fa9a63c5 RM |
6448 | the return codes and their meanings.) */ |
6449 | ||
d1dfb56c | 6450 | reg_errcode_t |
d2762c86 DN |
6451 | regcomp (regex_t *__restrict preg, const char *__restrict pattern, |
6452 | int cflags) | |
fa9a63c5 RM |
6453 | { |
6454 | reg_errcode_t ret; | |
4bb91c68 | 6455 | reg_syntax_t syntax |
fa9a63c5 RM |
6456 | = (cflags & REG_EXTENDED) ? |
6457 | RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; | |
6458 | ||
6459 | /* regex_compile will allocate the space for the compiled pattern. */ | |
6460 | preg->buffer = 0; | |
6461 | preg->allocated = 0; | |
6462 | preg->used = 0; | |
5e69f11e | 6463 | |
c0f9ea08 | 6464 | /* Try to allocate space for the fastmap. */ |
38182d90 | 6465 | preg->fastmap = malloc (1 << BYTEWIDTH); |
5e69f11e | 6466 | |
fa9a63c5 RM |
6467 | if (cflags & REG_ICASE) |
6468 | { | |
6469 | unsigned i; | |
5e69f11e | 6470 | |
38182d90 | 6471 | preg->translate = malloc (CHAR_SET_SIZE * sizeof *preg->translate); |
fa9a63c5 | 6472 | if (preg->translate == NULL) |
0b32bf0e | 6473 | return (int) REG_ESPACE; |
fa9a63c5 RM |
6474 | |
6475 | /* Map uppercase characters to corresponding lowercase ones. */ | |
6476 | for (i = 0; i < CHAR_SET_SIZE; i++) | |
4bb91c68 | 6477 | preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i; |
fa9a63c5 RM |
6478 | } |
6479 | else | |
6480 | preg->translate = NULL; | |
6481 | ||
6482 | /* If REG_NEWLINE is set, newlines are treated differently. */ | |
6483 | if (cflags & REG_NEWLINE) | |
6484 | { /* REG_NEWLINE implies neither . nor [^...] match newline. */ | |
6485 | syntax &= ~RE_DOT_NEWLINE; | |
6486 | syntax |= RE_HAT_LISTS_NOT_NEWLINE; | |
fa9a63c5 RM |
6487 | } |
6488 | else | |
c0f9ea08 | 6489 | syntax |= RE_NO_NEWLINE_ANCHOR; |
fa9a63c5 RM |
6490 | |
6491 | preg->no_sub = !!(cflags & REG_NOSUB); | |
6492 | ||
5e69f11e | 6493 | /* POSIX says a null character in the pattern terminates it, so we |
fa9a63c5 | 6494 | can use strlen here in compiling the pattern. */ |
4bb91c68 | 6495 | ret = regex_compile ((re_char*) pattern, strlen (pattern), syntax, preg); |
5e69f11e | 6496 | |
fa9a63c5 RM |
6497 | /* POSIX doesn't distinguish between an unmatched open-group and an |
6498 | unmatched close-group: both are REG_EPAREN. */ | |
c0f9ea08 SM |
6499 | if (ret == REG_ERPAREN) |
6500 | ret = REG_EPAREN; | |
6501 | ||
6502 | if (ret == REG_NOERROR && preg->fastmap) | |
6503 | { /* Compute the fastmap now, since regexec cannot modify the pattern | |
6504 | buffer. */ | |
6505 | re_compile_fastmap (preg); | |
6506 | if (preg->can_be_null) | |
6507 | { /* The fastmap can't be used anyway. */ | |
6508 | free (preg->fastmap); | |
6509 | preg->fastmap = NULL; | |
6510 | } | |
6511 | } | |
d1dfb56c | 6512 | return ret; |
fa9a63c5 | 6513 | } |
c0f9ea08 | 6514 | WEAK_ALIAS (__regcomp, regcomp) |
fa9a63c5 RM |
6515 | |
6516 | ||
6517 | /* regexec searches for a given pattern, specified by PREG, in the | |
6518 | string STRING. | |
5e69f11e | 6519 | |
fa9a63c5 | 6520 | If NMATCH is zero or REG_NOSUB was set in the cflags argument to |
b18215fc | 6521 | `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at |
fa9a63c5 RM |
6522 | least NMATCH elements, and we set them to the offsets of the |
6523 | corresponding matched substrings. | |
5e69f11e | 6524 | |
fa9a63c5 RM |
6525 | EFLAGS specifies `execution flags' which affect matching: if |
6526 | REG_NOTBOL is set, then ^ does not match at the beginning of the | |
6527 | string; if REG_NOTEOL is set, then $ does not match at the end. | |
5e69f11e | 6528 | |
fa9a63c5 RM |
6529 | We return 0 if we find a match and REG_NOMATCH if not. */ |
6530 | ||
d1dfb56c | 6531 | reg_errcode_t |
d2762c86 DN |
6532 | regexec (const regex_t *__restrict preg, const char *__restrict string, |
6533 | size_t nmatch, regmatch_t pmatch[__restrict_arr], int eflags) | |
fa9a63c5 | 6534 | { |
31011111 | 6535 | regoff_t ret; |
fa9a63c5 RM |
6536 | struct re_registers regs; |
6537 | regex_t private_preg; | |
d1dfb56c | 6538 | size_t len = strlen (string); |
c0f9ea08 | 6539 | boolean want_reg_info = !preg->no_sub && nmatch > 0 && pmatch; |
fa9a63c5 RM |
6540 | |
6541 | private_preg = *preg; | |
5e69f11e | 6542 | |
fa9a63c5 RM |
6543 | private_preg.not_bol = !!(eflags & REG_NOTBOL); |
6544 | private_preg.not_eol = !!(eflags & REG_NOTEOL); | |
5e69f11e | 6545 | |
fa9a63c5 RM |
6546 | /* The user has told us exactly how many registers to return |
6547 | information about, via `nmatch'. We have to pass that on to the | |
b18215fc | 6548 | matching routines. */ |
fa9a63c5 | 6549 | private_preg.regs_allocated = REGS_FIXED; |
5e69f11e | 6550 | |
fa9a63c5 RM |
6551 | if (want_reg_info) |
6552 | { | |
6553 | regs.num_regs = nmatch; | |
4bb91c68 SM |
6554 | regs.start = TALLOC (nmatch * 2, regoff_t); |
6555 | if (regs.start == NULL) | |
d1dfb56c | 6556 | return REG_NOMATCH; |
4bb91c68 | 6557 | regs.end = regs.start + nmatch; |
fa9a63c5 RM |
6558 | } |
6559 | ||
c0f9ea08 SM |
6560 | /* Instead of using not_eol to implement REG_NOTEOL, we could simply |
6561 | pass (&private_preg, string, len + 1, 0, len, ...) pretending the string | |
6562 | was a little bit longer but still only matching the real part. | |
6563 | This works because the `endline' will check for a '\n' and will find a | |
6564 | '\0', correctly deciding that this is not the end of a line. | |
6565 | But it doesn't work out so nicely for REG_NOTBOL, since we don't have | |
6566 | a convenient '\0' there. For all we know, the string could be preceded | |
6567 | by '\n' which would throw things off. */ | |
6568 | ||
fa9a63c5 RM |
6569 | /* Perform the searching operation. */ |
6570 | ret = re_search (&private_preg, string, len, | |
0b32bf0e SM |
6571 | /* start: */ 0, /* range: */ len, |
6572 | want_reg_info ? ®s : (struct re_registers *) 0); | |
5e69f11e | 6573 | |
fa9a63c5 RM |
6574 | /* Copy the register information to the POSIX structure. */ |
6575 | if (want_reg_info) | |
6576 | { | |
6577 | if (ret >= 0) | |
0b32bf0e SM |
6578 | { |
6579 | unsigned r; | |
fa9a63c5 | 6580 | |
0b32bf0e SM |
6581 | for (r = 0; r < nmatch; r++) |
6582 | { | |
6583 | pmatch[r].rm_so = regs.start[r]; | |
6584 | pmatch[r].rm_eo = regs.end[r]; | |
6585 | } | |
6586 | } | |
fa9a63c5 | 6587 | |
b18215fc | 6588 | /* If we needed the temporary register info, free the space now. */ |
fa9a63c5 | 6589 | free (regs.start); |
fa9a63c5 RM |
6590 | } |
6591 | ||
6592 | /* We want zero return to mean success, unlike `re_search'. */ | |
d1dfb56c | 6593 | return ret >= 0 ? REG_NOERROR : REG_NOMATCH; |
fa9a63c5 | 6594 | } |
c0f9ea08 | 6595 | WEAK_ALIAS (__regexec, regexec) |
fa9a63c5 RM |
6596 | |
6597 | ||
ec869672 JR |
6598 | /* Returns a message corresponding to an error code, ERR_CODE, returned |
6599 | from either regcomp or regexec. We don't use PREG here. | |
6600 | ||
6601 | ERR_CODE was previously called ERRCODE, but that name causes an | |
6602 | error with msvc8 compiler. */ | |
fa9a63c5 RM |
6603 | |
6604 | size_t | |
d2762c86 | 6605 | regerror (int err_code, const regex_t *preg, char *errbuf, size_t errbuf_size) |
fa9a63c5 RM |
6606 | { |
6607 | const char *msg; | |
6608 | size_t msg_size; | |
6609 | ||
ec869672 JR |
6610 | if (err_code < 0 |
6611 | || err_code >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0]))) | |
5e69f11e | 6612 | /* Only error codes returned by the rest of the code should be passed |
b18215fc | 6613 | to this routine. If we are given anything else, or if other regex |
fa9a63c5 RM |
6614 | code generates an invalid error code, then the program has a bug. |
6615 | Dump core so we can fix it. */ | |
6616 | abort (); | |
6617 | ||
ec869672 | 6618 | msg = gettext (re_error_msgid[err_code]); |
fa9a63c5 RM |
6619 | |
6620 | msg_size = strlen (msg) + 1; /* Includes the null. */ | |
5e69f11e | 6621 | |
fa9a63c5 RM |
6622 | if (errbuf_size != 0) |
6623 | { | |
6624 | if (msg_size > errbuf_size) | |
0b32bf0e | 6625 | { |
e99a530f | 6626 | memcpy (errbuf, msg, errbuf_size - 1); |
0b32bf0e SM |
6627 | errbuf[errbuf_size - 1] = 0; |
6628 | } | |
fa9a63c5 | 6629 | else |
0b32bf0e | 6630 | strcpy (errbuf, msg); |
fa9a63c5 RM |
6631 | } |
6632 | ||
6633 | return msg_size; | |
6634 | } | |
c0f9ea08 | 6635 | WEAK_ALIAS (__regerror, regerror) |
fa9a63c5 RM |
6636 | |
6637 | ||
6638 | /* Free dynamically allocated space used by PREG. */ | |
6639 | ||
6640 | void | |
d2762c86 | 6641 | regfree (regex_t *preg) |
fa9a63c5 | 6642 | { |
c2cd06e6 | 6643 | free (preg->buffer); |
fa9a63c5 | 6644 | preg->buffer = NULL; |
5e69f11e | 6645 | |
fa9a63c5 RM |
6646 | preg->allocated = 0; |
6647 | preg->used = 0; | |
6648 | ||
c2cd06e6 | 6649 | free (preg->fastmap); |
fa9a63c5 RM |
6650 | preg->fastmap = NULL; |
6651 | preg->fastmap_accurate = 0; | |
6652 | ||
c2cd06e6 | 6653 | free (preg->translate); |
fa9a63c5 RM |
6654 | preg->translate = NULL; |
6655 | } | |
c0f9ea08 | 6656 | WEAK_ALIAS (__regfree, regfree) |
fa9a63c5 RM |
6657 | |
6658 | #endif /* not emacs */ |