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. */ | |
31ff141c | 36 | #if 4 < __GNUC__ + (5 <= __GNUC_MINOR__) || defined __clang__ |
b8df54ff PE |
37 | # pragma GCC diagnostic ignored "-Wstrict-overflow" |
38 | # ifndef emacs | |
b8df54ff PE |
39 | # pragma GCC diagnostic ignored "-Wunused-function" |
40 | # pragma GCC diagnostic ignored "-Wunused-macros" | |
41 | # pragma GCC diagnostic ignored "-Wunused-result" | |
42 | # pragma GCC diagnostic ignored "-Wunused-variable" | |
43 | # endif | |
44 | #endif | |
45 | ||
31ff141c PE |
46 | #if 4 < __GNUC__ + (5 <= __GNUC_MINOR__) && ! defined __clang__ |
47 | # pragma GCC diagnostic ignored "-Wunused-but-set-variable" | |
48 | #endif | |
49 | ||
cf38a720 | 50 | #include <config.h> |
fa9a63c5 | 51 | |
0e926e56 PE |
52 | #include <stddef.h> |
53 | ||
54 | #ifdef emacs | |
4bb91c68 SM |
55 | /* We need this for `regex.h', and perhaps for the Emacs include files. */ |
56 | # include <sys/types.h> | |
57 | #endif | |
fa9a63c5 | 58 | |
14473664 SM |
59 | /* Whether to use ISO C Amendment 1 wide char functions. |
60 | Those should not be used for Emacs since it uses its own. */ | |
5e5388f6 GM |
61 | #if defined _LIBC |
62 | #define WIDE_CHAR_SUPPORT 1 | |
63 | #else | |
14473664 | 64 | #define WIDE_CHAR_SUPPORT \ |
5e5388f6 GM |
65 | (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC && !emacs) |
66 | #endif | |
14473664 | 67 | |
fa463103 | 68 | /* For platform which support the ISO C amendment 1 functionality we |
14473664 | 69 | support user defined character classes. */ |
a0ad02f7 | 70 | #if WIDE_CHAR_SUPPORT |
14473664 SM |
71 | /* Solaris 2.5 has a bug: <wchar.h> must be included before <wctype.h>. */ |
72 | # include <wchar.h> | |
73 | # include <wctype.h> | |
74 | #endif | |
75 | ||
c0f9ea08 SM |
76 | #ifdef _LIBC |
77 | /* We have to keep the namespace clean. */ | |
78 | # define regfree(preg) __regfree (preg) | |
79 | # define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef) | |
80 | # define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags) | |
ec869672 | 81 | # define regerror(err_code, preg, errbuf, errbuf_size) \ |
5e617bc2 | 82 | __regerror (err_code, preg, errbuf, errbuf_size) |
c0f9ea08 SM |
83 | # define re_set_registers(bu, re, nu, st, en) \ |
84 | __re_set_registers (bu, re, nu, st, en) | |
85 | # define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \ | |
86 | __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) | |
87 | # define re_match(bufp, string, size, pos, regs) \ | |
88 | __re_match (bufp, string, size, pos, regs) | |
89 | # define re_search(bufp, string, size, startpos, range, regs) \ | |
90 | __re_search (bufp, string, size, startpos, range, regs) | |
91 | # define re_compile_pattern(pattern, length, bufp) \ | |
92 | __re_compile_pattern (pattern, length, bufp) | |
93 | # define re_set_syntax(syntax) __re_set_syntax (syntax) | |
94 | # define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \ | |
95 | __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop) | |
96 | # define re_compile_fastmap(bufp) __re_compile_fastmap (bufp) | |
97 | ||
14473664 SM |
98 | /* Make sure we call libc's function even if the user overrides them. */ |
99 | # define btowc __btowc | |
100 | # define iswctype __iswctype | |
101 | # define wctype __wctype | |
102 | ||
c0f9ea08 SM |
103 | # define WEAK_ALIAS(a,b) weak_alias (a, b) |
104 | ||
105 | /* We are also using some library internals. */ | |
106 | # include <locale/localeinfo.h> | |
107 | # include <locale/elem-hash.h> | |
108 | # include <langinfo.h> | |
109 | #else | |
110 | # define WEAK_ALIAS(a,b) | |
111 | #endif | |
112 | ||
4bb91c68 | 113 | /* This is for other GNU distributions with internationalized messages. */ |
0b32bf0e | 114 | #if HAVE_LIBINTL_H || defined _LIBC |
fa9a63c5 RM |
115 | # include <libintl.h> |
116 | #else | |
117 | # define gettext(msgid) (msgid) | |
118 | #endif | |
119 | ||
5e69f11e RM |
120 | #ifndef gettext_noop |
121 | /* This define is so xgettext can find the internationalizable | |
122 | strings. */ | |
0b32bf0e | 123 | # define gettext_noop(String) String |
5e69f11e RM |
124 | #endif |
125 | ||
fa9a63c5 RM |
126 | /* The `emacs' switch turns on certain matching commands |
127 | that make sense only in Emacs. */ | |
128 | #ifdef emacs | |
129 | ||
0b32bf0e | 130 | # include "lisp.h" |
e5560ff7 | 131 | # include "character.h" |
0b32bf0e | 132 | # include "buffer.h" |
b18215fc RS |
133 | |
134 | /* Make syntax table lookup grant data in gl_state. */ | |
0b32bf0e | 135 | # define SYNTAX_ENTRY_VIA_PROPERTY |
b18215fc | 136 | |
0b32bf0e | 137 | # include "syntax.h" |
0b32bf0e | 138 | # include "category.h" |
fa9a63c5 | 139 | |
7689ef0b EZ |
140 | # ifdef malloc |
141 | # undef malloc | |
142 | # endif | |
0b32bf0e | 143 | # define malloc xmalloc |
7689ef0b EZ |
144 | # ifdef realloc |
145 | # undef realloc | |
146 | # endif | |
0b32bf0e | 147 | # define realloc xrealloc |
7689ef0b EZ |
148 | # ifdef free |
149 | # undef free | |
150 | # endif | |
0b32bf0e | 151 | # define free xfree |
9abbd165 | 152 | |
7814e705 | 153 | /* Converts the pointer to the char to BEG-based offset from the start. */ |
0b32bf0e SM |
154 | # define PTR_TO_OFFSET(d) POS_AS_IN_BUFFER (POINTER_TO_OFFSET (d)) |
155 | # define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object))) | |
156 | ||
157 | # define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte) | |
bf216479 | 158 | # define RE_TARGET_MULTIBYTE_P(bufp) ((bufp)->target_multibyte) |
62a6e103 AS |
159 | # define RE_STRING_CHAR(p, multibyte) \ |
160 | (multibyte ? (STRING_CHAR (p)) : (*(p))) | |
161 | # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) \ | |
162 | (multibyte ? (STRING_CHAR_AND_LENGTH (p, len)) : ((len) = 1, *(p))) | |
2d1675e4 | 163 | |
4c0354d7 | 164 | # define RE_CHAR_TO_MULTIBYTE(c) UNIBYTE_TO_CHAR (c) |
cf9c99bc | 165 | |
2afc21f5 | 166 | # define RE_CHAR_TO_UNIBYTE(c) CHAR_TO_BYTE_SAFE (c) |
cf9c99bc | 167 | |
6fdd04b0 KH |
168 | /* Set C a (possibly converted to multibyte) character before P. P |
169 | points into a string which is the virtual concatenation of STR1 | |
170 | (which ends at END1) or STR2 (which ends at END2). */ | |
bf216479 KH |
171 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
172 | do { \ | |
02cb78b5 | 173 | if (target_multibyte) \ |
bf216479 KH |
174 | { \ |
175 | re_char *dtemp = (p) == (str2) ? (end1) : (p); \ | |
176 | re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \ | |
177 | while (dtemp-- > dlimit && !CHAR_HEAD_P (*dtemp)); \ | |
62a6e103 | 178 | c = STRING_CHAR (dtemp); \ |
bf216479 KH |
179 | } \ |
180 | else \ | |
181 | { \ | |
182 | (c = ((p) == (str2) ? (end1) : (p))[-1]); \ | |
cf9c99bc | 183 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
bf216479 | 184 | } \ |
2d1675e4 SM |
185 | } while (0) |
186 | ||
6fdd04b0 KH |
187 | /* Set C a (possibly converted to multibyte) character at P, and set |
188 | LEN to the byte length of that character. */ | |
189 | # define GET_CHAR_AFTER(c, p, len) \ | |
190 | do { \ | |
02cb78b5 | 191 | if (target_multibyte) \ |
62a6e103 | 192 | (c) = STRING_CHAR_AND_LENGTH (p, len); \ |
6fdd04b0 KH |
193 | else \ |
194 | { \ | |
cf9c99bc | 195 | (c) = *p; \ |
6fdd04b0 | 196 | len = 1; \ |
cf9c99bc | 197 | (c) = RE_CHAR_TO_MULTIBYTE (c); \ |
6fdd04b0 | 198 | } \ |
8f924df7 | 199 | } while (0) |
4e8a9132 | 200 | |
fa9a63c5 RM |
201 | #else /* not emacs */ |
202 | ||
203 | /* If we are not linking with Emacs proper, | |
204 | we can't use the relocating allocator | |
205 | even if config.h says that we can. */ | |
0b32bf0e | 206 | # undef REL_ALLOC |
fa9a63c5 | 207 | |
4004364e | 208 | # include <unistd.h> |
fa9a63c5 | 209 | |
a77f947b CY |
210 | /* When used in Emacs's lib-src, we need xmalloc and xrealloc. */ |
211 | ||
b8df54ff | 212 | static void * |
d2762c86 | 213 | xmalloc (size_t size) |
a77f947b | 214 | { |
38182d90 | 215 | void *val = malloc (size); |
a77f947b CY |
216 | if (!val && size) |
217 | { | |
218 | write (2, "virtual memory exhausted\n", 25); | |
219 | exit (1); | |
220 | } | |
221 | return val; | |
222 | } | |
223 | ||
b8df54ff | 224 | static void * |
d2762c86 | 225 | xrealloc (void *block, size_t size) |
a77f947b | 226 | { |
38182d90 | 227 | void *val; |
a77f947b CY |
228 | /* We must call malloc explicitly when BLOCK is 0, since some |
229 | reallocs don't do this. */ | |
230 | if (! block) | |
38182d90 | 231 | val = malloc (size); |
a77f947b | 232 | else |
38182d90 | 233 | val = realloc (block, size); |
a77f947b CY |
234 | if (!val && size) |
235 | { | |
236 | write (2, "virtual memory exhausted\n", 25); | |
237 | exit (1); | |
238 | } | |
239 | return val; | |
240 | } | |
241 | ||
a073faa6 CY |
242 | # ifdef malloc |
243 | # undef malloc | |
244 | # endif | |
245 | # define malloc xmalloc | |
246 | # ifdef realloc | |
247 | # undef realloc | |
248 | # endif | |
249 | # define realloc xrealloc | |
250 | ||
f5d9e83a | 251 | # include <stdbool.h> |
9cfdb3ec | 252 | # include <string.h> |
fa9a63c5 RM |
253 | |
254 | /* Define the syntax stuff for \<, \>, etc. */ | |
255 | ||
990b2375 | 256 | /* Sword must be nonzero for the wordchar pattern commands in re_match_2. */ |
669fa600 | 257 | enum syntaxcode { Swhitespace = 0, Sword = 1, Ssymbol = 2 }; |
fa9a63c5 | 258 | |
e934739e | 259 | /* Dummy macros for non-Emacs environments. */ |
0b32bf0e SM |
260 | # define CHAR_CHARSET(c) 0 |
261 | # define CHARSET_LEADING_CODE_BASE(c) 0 | |
262 | # define MAX_MULTIBYTE_LENGTH 1 | |
263 | # define RE_MULTIBYTE_P(x) 0 | |
bf216479 | 264 | # define RE_TARGET_MULTIBYTE_P(x) 0 |
0b32bf0e SM |
265 | # define WORD_BOUNDARY_P(c1, c2) (0) |
266 | # define CHAR_HEAD_P(p) (1) | |
267 | # define SINGLE_BYTE_CHAR_P(c) (1) | |
268 | # define SAME_CHARSET_P(c1, c2) (1) | |
aa3830c4 | 269 | # define BYTES_BY_CHAR_HEAD(p) (1) |
70806df6 | 270 | # define PREV_CHAR_BOUNDARY(p, limit) ((p)--) |
62a6e103 AS |
271 | # define STRING_CHAR(p) (*(p)) |
272 | # define RE_STRING_CHAR(p, multibyte) STRING_CHAR (p) | |
0b32bf0e | 273 | # define CHAR_STRING(c, s) (*(s) = (c), 1) |
62a6e103 AS |
274 | # define STRING_CHAR_AND_LENGTH(p, actual_len) ((actual_len) = 1, *(p)) |
275 | # define RE_STRING_CHAR_AND_LENGTH(p, len, multibyte) STRING_CHAR_AND_LENGTH (p, len) | |
cf9c99bc KH |
276 | # define RE_CHAR_TO_MULTIBYTE(c) (c) |
277 | # define RE_CHAR_TO_UNIBYTE(c) (c) | |
0b32bf0e | 278 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ |
b18215fc | 279 | (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1))) |
6fdd04b0 KH |
280 | # define GET_CHAR_AFTER(c, p, len) \ |
281 | (c = *p, len = 1) | |
0b32bf0e | 282 | # define MAKE_CHAR(charset, c1, c2) (c1) |
9117d724 KH |
283 | # define BYTE8_TO_CHAR(c) (c) |
284 | # define CHAR_BYTE8_P(c) (0) | |
bf216479 | 285 | # define CHAR_LEADING_CODE(c) (c) |
8f924df7 | 286 | |
fa9a63c5 | 287 | #endif /* not emacs */ |
4e8a9132 SM |
288 | |
289 | #ifndef RE_TRANSLATE | |
0b32bf0e SM |
290 | # define RE_TRANSLATE(TBL, C) ((unsigned char)(TBL)[C]) |
291 | # define RE_TRANSLATE_P(TBL) (TBL) | |
4e8a9132 | 292 | #endif |
fa9a63c5 RM |
293 | \f |
294 | /* Get the interface, including the syntax bits. */ | |
295 | #include "regex.h" | |
296 | ||
f71b19b6 DL |
297 | /* isalpha etc. are used for the character classes. */ |
298 | #include <ctype.h> | |
fa9a63c5 | 299 | |
f71b19b6 | 300 | #ifdef emacs |
fa9a63c5 | 301 | |
f71b19b6 | 302 | /* 1 if C is an ASCII character. */ |
0b32bf0e | 303 | # define IS_REAL_ASCII(c) ((c) < 0200) |
fa9a63c5 | 304 | |
f71b19b6 | 305 | /* 1 if C is a unibyte character. */ |
0b32bf0e | 306 | # define ISUNIBYTE(c) (SINGLE_BYTE_CHAR_P ((c))) |
96cc36cc | 307 | |
f71b19b6 | 308 | /* The Emacs definitions should not be directly affected by locales. */ |
96cc36cc | 309 | |
f71b19b6 | 310 | /* In Emacs, these are only used for single-byte characters. */ |
0b32bf0e SM |
311 | # define ISDIGIT(c) ((c) >= '0' && (c) <= '9') |
312 | # define ISCNTRL(c) ((c) < ' ') | |
313 | # define ISXDIGIT(c) (((c) >= '0' && (c) <= '9') \ | |
f71b19b6 DL |
314 | || ((c) >= 'a' && (c) <= 'f') \ |
315 | || ((c) >= 'A' && (c) <= 'F')) | |
96cc36cc RS |
316 | |
317 | /* This is only used for single-byte characters. */ | |
0b32bf0e | 318 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') |
96cc36cc RS |
319 | |
320 | /* The rest must handle multibyte characters. */ | |
321 | ||
0b32bf0e | 322 | # define ISGRAPH(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 323 | ? (c) > ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
324 | : 1) |
325 | ||
14473664 | 326 | # define ISPRINT(c) (SINGLE_BYTE_CHAR_P (c) \ |
f71b19b6 | 327 | ? (c) >= ' ' && !((c) >= 0177 && (c) <= 0237) \ |
96cc36cc RS |
328 | : 1) |
329 | ||
0b32bf0e | 330 | # define ISALNUM(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
331 | ? (((c) >= 'a' && (c) <= 'z') \ |
332 | || ((c) >= 'A' && (c) <= 'Z') \ | |
333 | || ((c) >= '0' && (c) <= '9')) \ | |
96cc36cc RS |
334 | : SYNTAX (c) == Sword) |
335 | ||
0b32bf0e | 336 | # define ISALPHA(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
337 | ? (((c) >= 'a' && (c) <= 'z') \ |
338 | || ((c) >= 'A' && (c) <= 'Z')) \ | |
96cc36cc RS |
339 | : SYNTAX (c) == Sword) |
340 | ||
5da9919f | 341 | # define ISLOWER(c) lowercasep (c) |
96cc36cc | 342 | |
0b32bf0e | 343 | # define ISPUNCT(c) (IS_REAL_ASCII (c) \ |
f71b19b6 DL |
344 | ? ((c) > ' ' && (c) < 0177 \ |
345 | && !(((c) >= 'a' && (c) <= 'z') \ | |
4bb91c68 SM |
346 | || ((c) >= 'A' && (c) <= 'Z') \ |
347 | || ((c) >= '0' && (c) <= '9'))) \ | |
96cc36cc RS |
348 | : SYNTAX (c) != Sword) |
349 | ||
0b32bf0e | 350 | # define ISSPACE(c) (SYNTAX (c) == Swhitespace) |
96cc36cc | 351 | |
5da9919f | 352 | # define ISUPPER(c) uppercasep (c) |
96cc36cc | 353 | |
0b32bf0e | 354 | # define ISWORD(c) (SYNTAX (c) == Sword) |
96cc36cc RS |
355 | |
356 | #else /* not emacs */ | |
357 | ||
f71b19b6 | 358 | /* 1 if C is an ASCII character. */ |
0b32bf0e | 359 | # define IS_REAL_ASCII(c) ((c) < 0200) |
f71b19b6 DL |
360 | |
361 | /* This distinction is not meaningful, except in Emacs. */ | |
0b32bf0e SM |
362 | # define ISUNIBYTE(c) 1 |
363 | ||
364 | # ifdef isblank | |
0e926e56 | 365 | # define ISBLANK(c) isblank (c) |
0b32bf0e SM |
366 | # else |
367 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') | |
368 | # endif | |
369 | # ifdef isgraph | |
0e926e56 | 370 | # define ISGRAPH(c) isgraph (c) |
0b32bf0e | 371 | # else |
0e926e56 | 372 | # define ISGRAPH(c) (isprint (c) && !isspace (c)) |
0b32bf0e SM |
373 | # endif |
374 | ||
0e926e56 | 375 | /* Solaris defines ISPRINT so we must undefine it first. */ |
4bb91c68 | 376 | # undef ISPRINT |
0e926e56 PE |
377 | # define ISPRINT(c) isprint (c) |
378 | # define ISDIGIT(c) isdigit (c) | |
379 | # define ISALNUM(c) isalnum (c) | |
380 | # define ISALPHA(c) isalpha (c) | |
381 | # define ISCNTRL(c) iscntrl (c) | |
382 | # define ISLOWER(c) islower (c) | |
383 | # define ISPUNCT(c) ispunct (c) | |
384 | # define ISSPACE(c) isspace (c) | |
385 | # define ISUPPER(c) isupper (c) | |
386 | # define ISXDIGIT(c) isxdigit (c) | |
0b32bf0e | 387 | |
5e617bc2 | 388 | # define ISWORD(c) ISALPHA (c) |
0b32bf0e | 389 | |
4bb91c68 | 390 | # ifdef _tolower |
5e617bc2 | 391 | # define TOLOWER(c) _tolower (c) |
4bb91c68 | 392 | # else |
5e617bc2 | 393 | # define TOLOWER(c) tolower (c) |
4bb91c68 SM |
394 | # endif |
395 | ||
396 | /* How many characters in the character set. */ | |
397 | # define CHAR_SET_SIZE 256 | |
398 | ||
0b32bf0e | 399 | # ifdef SYNTAX_TABLE |
f71b19b6 | 400 | |
0b32bf0e | 401 | extern char *re_syntax_table; |
f71b19b6 | 402 | |
0b32bf0e SM |
403 | # else /* not SYNTAX_TABLE */ |
404 | ||
0b32bf0e SM |
405 | static char re_syntax_table[CHAR_SET_SIZE]; |
406 | ||
407 | static void | |
d2762c86 | 408 | init_syntax_once (void) |
0b32bf0e SM |
409 | { |
410 | register int c; | |
411 | static int done = 0; | |
412 | ||
413 | if (done) | |
414 | return; | |
415 | ||
72af86bd | 416 | memset (re_syntax_table, 0, sizeof re_syntax_table); |
0b32bf0e | 417 | |
4bb91c68 SM |
418 | for (c = 0; c < CHAR_SET_SIZE; ++c) |
419 | if (ISALNUM (c)) | |
420 | re_syntax_table[c] = Sword; | |
fa9a63c5 | 421 | |
669fa600 | 422 | re_syntax_table['_'] = Ssymbol; |
fa9a63c5 | 423 | |
0b32bf0e SM |
424 | done = 1; |
425 | } | |
426 | ||
427 | # endif /* not SYNTAX_TABLE */ | |
96cc36cc | 428 | |
4bb91c68 SM |
429 | # define SYNTAX(c) re_syntax_table[(c)] |
430 | ||
96cc36cc RS |
431 | #endif /* not emacs */ |
432 | \f | |
261cb4bb | 433 | #define SIGN_EXTEND_CHAR(c) ((signed char) (c)) |
fa9a63c5 RM |
434 | \f |
435 | /* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we | |
436 | use `alloca' instead of `malloc'. This is because using malloc in | |
437 | re_search* or re_match* could cause memory leaks when C-g is used in | |
438 | Emacs; also, malloc is slower and causes storage fragmentation. On | |
5e69f11e RM |
439 | the other hand, malloc is more portable, and easier to debug. |
440 | ||
fa9a63c5 RM |
441 | Because we sometimes use alloca, some routines have to be macros, |
442 | not functions -- `alloca'-allocated space disappears at the end of the | |
443 | function it is called in. */ | |
444 | ||
445 | #ifdef REGEX_MALLOC | |
446 | ||
0b32bf0e SM |
447 | # define REGEX_ALLOCATE malloc |
448 | # define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize) | |
449 | # define REGEX_FREE free | |
fa9a63c5 RM |
450 | |
451 | #else /* not REGEX_MALLOC */ | |
452 | ||
453 | /* Emacs already defines alloca, sometimes. */ | |
0b32bf0e | 454 | # ifndef alloca |
fa9a63c5 RM |
455 | |
456 | /* Make alloca work the best possible way. */ | |
0b32bf0e SM |
457 | # ifdef __GNUC__ |
458 | # define alloca __builtin_alloca | |
459 | # else /* not __GNUC__ */ | |
7f585e7a | 460 | # ifdef HAVE_ALLOCA_H |
0b32bf0e SM |
461 | # include <alloca.h> |
462 | # endif /* HAVE_ALLOCA_H */ | |
463 | # endif /* not __GNUC__ */ | |
fa9a63c5 | 464 | |
0b32bf0e | 465 | # endif /* not alloca */ |
fa9a63c5 | 466 | |
0b32bf0e | 467 | # define REGEX_ALLOCATE alloca |
fa9a63c5 RM |
468 | |
469 | /* Assumes a `char *destination' variable. */ | |
0b32bf0e | 470 | # define REGEX_REALLOCATE(source, osize, nsize) \ |
7d652d97 | 471 | (destination = alloca (nsize), \ |
4bb91c68 | 472 | memcpy (destination, source, osize)) |
fa9a63c5 RM |
473 | |
474 | /* No need to do anything to free, after alloca. */ | |
0b32bf0e | 475 | # define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
476 | |
477 | #endif /* not REGEX_MALLOC */ | |
478 | ||
479 | /* Define how to allocate the failure stack. */ | |
480 | ||
0b32bf0e | 481 | #if defined REL_ALLOC && defined REGEX_MALLOC |
4297555e | 482 | |
0b32bf0e | 483 | # define REGEX_ALLOCATE_STACK(size) \ |
fa9a63c5 | 484 | r_alloc (&failure_stack_ptr, (size)) |
0b32bf0e | 485 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 486 | r_re_alloc (&failure_stack_ptr, (nsize)) |
0b32bf0e | 487 | # define REGEX_FREE_STACK(ptr) \ |
fa9a63c5 RM |
488 | r_alloc_free (&failure_stack_ptr) |
489 | ||
4297555e | 490 | #else /* not using relocating allocator */ |
fa9a63c5 | 491 | |
0b32bf0e | 492 | # ifdef REGEX_MALLOC |
fa9a63c5 | 493 | |
0b32bf0e SM |
494 | # define REGEX_ALLOCATE_STACK malloc |
495 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize) | |
496 | # define REGEX_FREE_STACK free | |
fa9a63c5 | 497 | |
0b32bf0e | 498 | # else /* not REGEX_MALLOC */ |
fa9a63c5 | 499 | |
0b32bf0e | 500 | # define REGEX_ALLOCATE_STACK alloca |
fa9a63c5 | 501 | |
0b32bf0e | 502 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ |
fa9a63c5 | 503 | REGEX_REALLOCATE (source, osize, nsize) |
7814e705 | 504 | /* No need to explicitly free anything. */ |
0b32bf0e | 505 | # define REGEX_FREE_STACK(arg) ((void)0) |
fa9a63c5 | 506 | |
0b32bf0e | 507 | # endif /* not REGEX_MALLOC */ |
4297555e | 508 | #endif /* not using relocating allocator */ |
fa9a63c5 RM |
509 | |
510 | ||
511 | /* True if `size1' is non-NULL and PTR is pointing anywhere inside | |
512 | `string1' or just past its end. This works if PTR is NULL, which is | |
513 | a good thing. */ | |
25fe55af | 514 | #define FIRST_STRING_P(ptr) \ |
fa9a63c5 RM |
515 | (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) |
516 | ||
517 | /* (Re)Allocate N items of type T using malloc, or fail. */ | |
518 | #define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t))) | |
519 | #define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t))) | |
fa9a63c5 RM |
520 | #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) |
521 | ||
4bb91c68 | 522 | #define BYTEWIDTH 8 /* In bits. */ |
fa9a63c5 RM |
523 | |
524 | #define STREQ(s1, s2) ((strcmp (s1, s2) == 0)) | |
525 | ||
526 | #undef MAX | |
527 | #undef MIN | |
528 | #define MAX(a, b) ((a) > (b) ? (a) : (b)) | |
529 | #define MIN(a, b) ((a) < (b) ? (a) : (b)) | |
530 | ||
66f0296e | 531 | /* Type of source-pattern and string chars. */ |
a6fc3b5c EZ |
532 | #ifdef _MSC_VER |
533 | typedef unsigned char re_char; | |
29abe551 | 534 | typedef const re_char const_re_char; |
a6fc3b5c | 535 | #else |
66f0296e | 536 | typedef const unsigned char re_char; |
29abe551 | 537 | typedef re_char const_re_char; |
a6fc3b5c | 538 | #endif |
66f0296e | 539 | |
fa9a63c5 | 540 | typedef char boolean; |
fa9a63c5 | 541 | |
261cb4bb PE |
542 | static regoff_t re_match_2_internal (struct re_pattern_buffer *bufp, |
543 | re_char *string1, size_t size1, | |
544 | re_char *string2, size_t size2, | |
545 | ssize_t pos, | |
546 | struct re_registers *regs, | |
547 | ssize_t stop); | |
fa9a63c5 RM |
548 | \f |
549 | /* These are the command codes that appear in compiled regular | |
4bb91c68 | 550 | expressions. Some opcodes are followed by argument bytes. A |
fa9a63c5 RM |
551 | command code can specify any interpretation whatsoever for its |
552 | arguments. Zero bytes may appear in the compiled regular expression. */ | |
553 | ||
554 | typedef enum | |
555 | { | |
556 | no_op = 0, | |
557 | ||
4bb91c68 | 558 | /* Succeed right away--no more backtracking. */ |
fa9a63c5 RM |
559 | succeed, |
560 | ||
25fe55af | 561 | /* Followed by one byte giving n, then by n literal bytes. */ |
fa9a63c5 RM |
562 | exactn, |
563 | ||
25fe55af | 564 | /* Matches any (more or less) character. */ |
fa9a63c5 RM |
565 | anychar, |
566 | ||
25fe55af RS |
567 | /* Matches any one char belonging to specified set. First |
568 | following byte is number of bitmap bytes. Then come bytes | |
569 | for a bitmap saying which chars are in. Bits in each byte | |
570 | are ordered low-bit-first. A character is in the set if its | |
571 | bit is 1. A character too large to have a bit in the map is | |
96cc36cc RS |
572 | automatically not in the set. |
573 | ||
574 | If the length byte has the 0x80 bit set, then that stuff | |
575 | is followed by a range table: | |
576 | 2 bytes of flags for character sets (low 8 bits, high 8 bits) | |
0b32bf0e | 577 | See RANGE_TABLE_WORK_BITS below. |
01618498 | 578 | 2 bytes, the number of pairs that follow (upto 32767) |
96cc36cc | 579 | pairs, each 2 multibyte characters, |
0b32bf0e | 580 | each multibyte character represented as 3 bytes. */ |
fa9a63c5 RM |
581 | charset, |
582 | ||
25fe55af | 583 | /* Same parameters as charset, but match any character that is |
4bb91c68 | 584 | not one of those specified. */ |
fa9a63c5 RM |
585 | charset_not, |
586 | ||
25fe55af RS |
587 | /* Start remembering the text that is matched, for storing in a |
588 | register. Followed by one byte with the register number, in | |
589 | the range 0 to one less than the pattern buffer's re_nsub | |
505bde11 | 590 | field. */ |
fa9a63c5 RM |
591 | start_memory, |
592 | ||
25fe55af RS |
593 | /* Stop remembering the text that is matched and store it in a |
594 | memory register. Followed by one byte with the register | |
595 | number, in the range 0 to one less than `re_nsub' in the | |
505bde11 | 596 | pattern buffer. */ |
fa9a63c5 RM |
597 | stop_memory, |
598 | ||
25fe55af | 599 | /* Match a duplicate of something remembered. Followed by one |
4bb91c68 | 600 | byte containing the register number. */ |
fa9a63c5 RM |
601 | duplicate, |
602 | ||
25fe55af | 603 | /* Fail unless at beginning of line. */ |
fa9a63c5 RM |
604 | begline, |
605 | ||
4bb91c68 | 606 | /* Fail unless at end of line. */ |
fa9a63c5 RM |
607 | endline, |
608 | ||
25fe55af RS |
609 | /* Succeeds if at beginning of buffer (if emacs) or at beginning |
610 | of string to be matched (if not). */ | |
fa9a63c5 RM |
611 | begbuf, |
612 | ||
25fe55af | 613 | /* Analogously, for end of buffer/string. */ |
fa9a63c5 | 614 | endbuf, |
5e69f11e | 615 | |
25fe55af | 616 | /* Followed by two byte relative address to which to jump. */ |
5e69f11e | 617 | jump, |
fa9a63c5 | 618 | |
25fe55af | 619 | /* Followed by two-byte relative address of place to resume at |
7814e705 | 620 | in case of failure. */ |
fa9a63c5 | 621 | on_failure_jump, |
5e69f11e | 622 | |
25fe55af RS |
623 | /* Like on_failure_jump, but pushes a placeholder instead of the |
624 | current string position when executed. */ | |
fa9a63c5 | 625 | on_failure_keep_string_jump, |
5e69f11e | 626 | |
505bde11 SM |
627 | /* Just like `on_failure_jump', except that it checks that we |
628 | don't get stuck in an infinite loop (matching an empty string | |
629 | indefinitely). */ | |
630 | on_failure_jump_loop, | |
631 | ||
0683b6fa SM |
632 | /* Just like `on_failure_jump_loop', except that it checks for |
633 | a different kind of loop (the kind that shows up with non-greedy | |
634 | operators). This operation has to be immediately preceded | |
635 | by a `no_op'. */ | |
636 | on_failure_jump_nastyloop, | |
637 | ||
0b32bf0e | 638 | /* A smart `on_failure_jump' used for greedy * and + operators. |
c7015153 | 639 | It analyzes the loop before which it is put and if the |
505bde11 | 640 | loop does not require backtracking, it changes itself to |
4e8a9132 SM |
641 | `on_failure_keep_string_jump' and short-circuits the loop, |
642 | else it just defaults to changing itself into `on_failure_jump'. | |
643 | It assumes that it is pointing to just past a `jump'. */ | |
505bde11 | 644 | on_failure_jump_smart, |
fa9a63c5 | 645 | |
25fe55af | 646 | /* Followed by two-byte relative address and two-byte number n. |
ed0767d8 SM |
647 | After matching N times, jump to the address upon failure. |
648 | Does not work if N starts at 0: use on_failure_jump_loop | |
649 | instead. */ | |
fa9a63c5 RM |
650 | succeed_n, |
651 | ||
25fe55af RS |
652 | /* Followed by two-byte relative address, and two-byte number n. |
653 | Jump to the address N times, then fail. */ | |
fa9a63c5 RM |
654 | jump_n, |
655 | ||
25fe55af | 656 | /* Set the following two-byte relative address to the |
7814e705 | 657 | subsequent two-byte number. The address *includes* the two |
25fe55af | 658 | bytes of number. */ |
fa9a63c5 RM |
659 | set_number_at, |
660 | ||
fa9a63c5 RM |
661 | wordbeg, /* Succeeds if at word beginning. */ |
662 | wordend, /* Succeeds if at word end. */ | |
663 | ||
664 | wordbound, /* Succeeds if at a word boundary. */ | |
7814e705 | 665 | notwordbound, /* Succeeds if not at a word boundary. */ |
fa9a63c5 | 666 | |
669fa600 SM |
667 | symbeg, /* Succeeds if at symbol beginning. */ |
668 | symend, /* Succeeds if at symbol end. */ | |
669 | ||
fa9a63c5 | 670 | /* Matches any character whose syntax is specified. Followed by |
25fe55af | 671 | a byte which contains a syntax code, e.g., Sword. */ |
fa9a63c5 RM |
672 | syntaxspec, |
673 | ||
674 | /* Matches any character whose syntax is not that specified. */ | |
1fb352e0 SM |
675 | notsyntaxspec |
676 | ||
677 | #ifdef emacs | |
678 | ,before_dot, /* Succeeds if before point. */ | |
679 | at_dot, /* Succeeds if at point. */ | |
680 | after_dot, /* Succeeds if after point. */ | |
b18215fc RS |
681 | |
682 | /* Matches any character whose category-set contains the specified | |
7814e705 JB |
683 | category. The operator is followed by a byte which contains a |
684 | category code (mnemonic ASCII character). */ | |
b18215fc RS |
685 | categoryspec, |
686 | ||
687 | /* Matches any character whose category-set does not contain the | |
688 | specified category. The operator is followed by a byte which | |
689 | contains the category code (mnemonic ASCII character). */ | |
690 | notcategoryspec | |
fa9a63c5 RM |
691 | #endif /* emacs */ |
692 | } re_opcode_t; | |
693 | \f | |
694 | /* Common operations on the compiled pattern. */ | |
695 | ||
696 | /* Store NUMBER in two contiguous bytes starting at DESTINATION. */ | |
697 | ||
698 | #define STORE_NUMBER(destination, number) \ | |
699 | do { \ | |
700 | (destination)[0] = (number) & 0377; \ | |
701 | (destination)[1] = (number) >> 8; \ | |
702 | } while (0) | |
703 | ||
704 | /* Same as STORE_NUMBER, except increment DESTINATION to | |
705 | the byte after where the number is stored. Therefore, DESTINATION | |
706 | must be an lvalue. */ | |
707 | ||
708 | #define STORE_NUMBER_AND_INCR(destination, number) \ | |
709 | do { \ | |
710 | STORE_NUMBER (destination, number); \ | |
711 | (destination) += 2; \ | |
712 | } while (0) | |
713 | ||
714 | /* Put into DESTINATION a number stored in two contiguous bytes starting | |
715 | at SOURCE. */ | |
716 | ||
717 | #define EXTRACT_NUMBER(destination, source) \ | |
dc4a2ee0 | 718 | ((destination) = extract_number (source)) |
fa9a63c5 | 719 | |
dc4a2ee0 PE |
720 | static int |
721 | extract_number (re_char *source) | |
fa9a63c5 | 722 | { |
dc4a2ee0 | 723 | return (SIGN_EXTEND_CHAR (source[1]) << 8) + source[0]; |
fa9a63c5 RM |
724 | } |
725 | ||
fa9a63c5 RM |
726 | /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. |
727 | SOURCE must be an lvalue. */ | |
728 | ||
729 | #define EXTRACT_NUMBER_AND_INCR(destination, source) \ | |
dc4a2ee0 | 730 | ((destination) = extract_number_and_incr (&source)) |
fa9a63c5 | 731 | |
dc4a2ee0 PE |
732 | static int |
733 | extract_number_and_incr (re_char **source) | |
5e69f11e | 734 | { |
dc4a2ee0 | 735 | int num = extract_number (*source); |
fa9a63c5 | 736 | *source += 2; |
dc4a2ee0 | 737 | return num; |
fa9a63c5 | 738 | } |
fa9a63c5 | 739 | \f |
b18215fc RS |
740 | /* Store a multibyte character in three contiguous bytes starting |
741 | DESTINATION, and increment DESTINATION to the byte after where the | |
7814e705 | 742 | character is stored. Therefore, DESTINATION must be an lvalue. */ |
b18215fc RS |
743 | |
744 | #define STORE_CHARACTER_AND_INCR(destination, character) \ | |
745 | do { \ | |
746 | (destination)[0] = (character) & 0377; \ | |
747 | (destination)[1] = ((character) >> 8) & 0377; \ | |
748 | (destination)[2] = (character) >> 16; \ | |
749 | (destination) += 3; \ | |
750 | } while (0) | |
751 | ||
752 | /* Put into DESTINATION a character stored in three contiguous bytes | |
7814e705 | 753 | starting at SOURCE. */ |
b18215fc RS |
754 | |
755 | #define EXTRACT_CHARACTER(destination, source) \ | |
756 | do { \ | |
757 | (destination) = ((source)[0] \ | |
758 | | ((source)[1] << 8) \ | |
759 | | ((source)[2] << 16)); \ | |
760 | } while (0) | |
761 | ||
762 | ||
763 | /* Macros for charset. */ | |
764 | ||
765 | /* Size of bitmap of charset P in bytes. P is a start of charset, | |
766 | i.e. *P is (re_opcode_t) charset or (re_opcode_t) charset_not. */ | |
767 | #define CHARSET_BITMAP_SIZE(p) ((p)[1] & 0x7F) | |
768 | ||
769 | /* Nonzero if charset P has range table. */ | |
25fe55af | 770 | #define CHARSET_RANGE_TABLE_EXISTS_P(p) ((p)[1] & 0x80) |
b18215fc RS |
771 | |
772 | /* Return the address of range table of charset P. But not the start | |
773 | of table itself, but the before where the number of ranges is | |
96cc36cc RS |
774 | stored. `2 +' means to skip re_opcode_t and size of bitmap, |
775 | and the 2 bytes of flags at the start of the range table. */ | |
776 | #define CHARSET_RANGE_TABLE(p) (&(p)[4 + CHARSET_BITMAP_SIZE (p)]) | |
777 | ||
778 | /* Extract the bit flags that start a range table. */ | |
779 | #define CHARSET_RANGE_TABLE_BITS(p) \ | |
780 | ((p)[2 + CHARSET_BITMAP_SIZE (p)] \ | |
781 | + (p)[3 + CHARSET_BITMAP_SIZE (p)] * 0x100) | |
b18215fc | 782 | |
b18215fc | 783 | /* Return the address of end of RANGE_TABLE. COUNT is number of |
7814e705 JB |
784 | ranges (which is a pair of (start, end)) in the RANGE_TABLE. `* 2' |
785 | is start of range and end of range. `* 3' is size of each start | |
b18215fc RS |
786 | and end. */ |
787 | #define CHARSET_RANGE_TABLE_END(range_table, count) \ | |
788 | ((range_table) + (count) * 2 * 3) | |
789 | ||
7814e705 | 790 | /* Test if C is in RANGE_TABLE. A flag NOT is negated if C is in. |
b18215fc RS |
791 | COUNT is number of ranges in RANGE_TABLE. */ |
792 | #define CHARSET_LOOKUP_RANGE_TABLE_RAW(not, c, range_table, count) \ | |
793 | do \ | |
794 | { \ | |
01618498 | 795 | re_wchar_t range_start, range_end; \ |
19ed5445 | 796 | re_char *rtp; \ |
01618498 | 797 | re_char *range_table_end \ |
b18215fc RS |
798 | = CHARSET_RANGE_TABLE_END ((range_table), (count)); \ |
799 | \ | |
19ed5445 | 800 | for (rtp = (range_table); rtp < range_table_end; rtp += 2 * 3) \ |
b18215fc | 801 | { \ |
19ed5445 PE |
802 | EXTRACT_CHARACTER (range_start, rtp); \ |
803 | EXTRACT_CHARACTER (range_end, rtp + 3); \ | |
b18215fc RS |
804 | \ |
805 | if (range_start <= (c) && (c) <= range_end) \ | |
806 | { \ | |
807 | (not) = !(not); \ | |
808 | break; \ | |
809 | } \ | |
810 | } \ | |
811 | } \ | |
812 | while (0) | |
813 | ||
814 | /* Test if C is in range table of CHARSET. The flag NOT is negated if | |
815 | C is listed in it. */ | |
816 | #define CHARSET_LOOKUP_RANGE_TABLE(not, c, charset) \ | |
817 | do \ | |
818 | { \ | |
819 | /* Number of ranges in range table. */ \ | |
820 | int count; \ | |
01618498 SM |
821 | re_char *range_table = CHARSET_RANGE_TABLE (charset); \ |
822 | \ | |
b18215fc RS |
823 | EXTRACT_NUMBER_AND_INCR (count, range_table); \ |
824 | CHARSET_LOOKUP_RANGE_TABLE_RAW ((not), (c), range_table, count); \ | |
825 | } \ | |
826 | while (0) | |
827 | \f | |
fa9a63c5 RM |
828 | /* If DEBUG is defined, Regex prints many voluminous messages about what |
829 | it is doing (if the variable `debug' is nonzero). If linked with the | |
830 | main program in `iregex.c', you can enter patterns and strings | |
831 | interactively. And if linked with the main program in `main.c' and | |
4bb91c68 | 832 | the other test files, you can run the already-written tests. */ |
fa9a63c5 RM |
833 | |
834 | #ifdef DEBUG | |
835 | ||
836 | /* We use standard I/O for debugging. */ | |
0b32bf0e | 837 | # include <stdio.h> |
fa9a63c5 RM |
838 | |
839 | /* It is useful to test things that ``must'' be true when debugging. */ | |
0b32bf0e | 840 | # include <assert.h> |
fa9a63c5 | 841 | |
99633e97 | 842 | static int debug = -100000; |
fa9a63c5 | 843 | |
0b32bf0e | 844 | # define DEBUG_STATEMENT(e) e |
dc4a2ee0 PE |
845 | # define DEBUG_PRINT(...) if (debug > 0) printf (__VA_ARGS__) |
846 | # define DEBUG_COMPILES_ARGUMENTS | |
0b32bf0e | 847 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ |
99633e97 | 848 | if (debug > 0) print_partial_compiled_pattern (s, e) |
0b32bf0e | 849 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ |
99633e97 | 850 | if (debug > 0) print_double_string (w, s1, sz1, s2, sz2) |
fa9a63c5 RM |
851 | |
852 | ||
853 | /* Print the fastmap in human-readable form. */ | |
854 | ||
dc4a2ee0 PE |
855 | static void |
856 | print_fastmap (char *fastmap) | |
fa9a63c5 RM |
857 | { |
858 | unsigned was_a_range = 0; | |
5e69f11e RM |
859 | unsigned i = 0; |
860 | ||
fa9a63c5 RM |
861 | while (i < (1 << BYTEWIDTH)) |
862 | { | |
863 | if (fastmap[i++]) | |
864 | { | |
865 | was_a_range = 0; | |
25fe55af RS |
866 | putchar (i - 1); |
867 | while (i < (1 << BYTEWIDTH) && fastmap[i]) | |
868 | { | |
869 | was_a_range = 1; | |
870 | i++; | |
871 | } | |
fa9a63c5 | 872 | if (was_a_range) |
25fe55af RS |
873 | { |
874 | printf ("-"); | |
875 | putchar (i - 1); | |
876 | } | |
877 | } | |
fa9a63c5 | 878 | } |
5e69f11e | 879 | putchar ('\n'); |
fa9a63c5 RM |
880 | } |
881 | ||
882 | ||
883 | /* Print a compiled pattern string in human-readable form, starting at | |
884 | the START pointer into it and ending just before the pointer END. */ | |
885 | ||
dc4a2ee0 PE |
886 | static void |
887 | print_partial_compiled_pattern (re_char *start, re_char *end) | |
fa9a63c5 RM |
888 | { |
889 | int mcnt, mcnt2; | |
01618498 SM |
890 | re_char *p = start; |
891 | re_char *pend = end; | |
fa9a63c5 RM |
892 | |
893 | if (start == NULL) | |
894 | { | |
a1a052df | 895 | fprintf (stderr, "(null)\n"); |
fa9a63c5 RM |
896 | return; |
897 | } | |
5e69f11e | 898 | |
fa9a63c5 RM |
899 | /* Loop over pattern commands. */ |
900 | while (p < pend) | |
901 | { | |
dc4a2ee0 | 902 | fprintf (stderr, "%td:\t", p - start); |
fa9a63c5 RM |
903 | |
904 | switch ((re_opcode_t) *p++) | |
905 | { | |
25fe55af | 906 | case no_op: |
a1a052df | 907 | fprintf (stderr, "/no_op"); |
25fe55af | 908 | break; |
fa9a63c5 | 909 | |
99633e97 | 910 | case succeed: |
a1a052df | 911 | fprintf (stderr, "/succeed"); |
99633e97 SM |
912 | break; |
913 | ||
fa9a63c5 RM |
914 | case exactn: |
915 | mcnt = *p++; | |
a1a052df | 916 | fprintf (stderr, "/exactn/%d", mcnt); |
25fe55af | 917 | do |
fa9a63c5 | 918 | { |
a1a052df | 919 | fprintf (stderr, "/%c", *p++); |
25fe55af RS |
920 | } |
921 | while (--mcnt); | |
922 | break; | |
fa9a63c5 RM |
923 | |
924 | case start_memory: | |
a1a052df | 925 | fprintf (stderr, "/start_memory/%d", *p++); |
25fe55af | 926 | break; |
fa9a63c5 RM |
927 | |
928 | case stop_memory: | |
a1a052df | 929 | fprintf (stderr, "/stop_memory/%d", *p++); |
25fe55af | 930 | break; |
fa9a63c5 RM |
931 | |
932 | case duplicate: | |
a1a052df | 933 | fprintf (stderr, "/duplicate/%d", *p++); |
fa9a63c5 RM |
934 | break; |
935 | ||
936 | case anychar: | |
a1a052df | 937 | fprintf (stderr, "/anychar"); |
fa9a63c5 RM |
938 | break; |
939 | ||
940 | case charset: | |
25fe55af RS |
941 | case charset_not: |
942 | { | |
943 | register int c, last = -100; | |
fa9a63c5 | 944 | register int in_range = 0; |
99633e97 SM |
945 | int length = CHARSET_BITMAP_SIZE (p - 1); |
946 | int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1); | |
fa9a63c5 | 947 | |
a1a052df | 948 | fprintf (stderr, "/charset [%s", |
839966f3 | 949 | (re_opcode_t) *(p - 1) == charset_not ? "^" : ""); |
5e69f11e | 950 | |
839966f3 KH |
951 | if (p + *p >= pend) |
952 | fprintf (stderr, " !extends past end of pattern! "); | |
fa9a63c5 | 953 | |
25fe55af | 954 | for (c = 0; c < 256; c++) |
96cc36cc | 955 | if (c / 8 < length |
fa9a63c5 RM |
956 | && (p[1 + (c/8)] & (1 << (c % 8)))) |
957 | { | |
958 | /* Are we starting a range? */ | |
959 | if (last + 1 == c && ! in_range) | |
960 | { | |
a1a052df | 961 | fprintf (stderr, "-"); |
fa9a63c5 RM |
962 | in_range = 1; |
963 | } | |
964 | /* Have we broken a range? */ | |
965 | else if (last + 1 != c && in_range) | |
96cc36cc | 966 | { |
a1a052df | 967 | fprintf (stderr, "%c", last); |
fa9a63c5 RM |
968 | in_range = 0; |
969 | } | |
5e69f11e | 970 | |
fa9a63c5 | 971 | if (! in_range) |
a1a052df | 972 | fprintf (stderr, "%c", c); |
fa9a63c5 RM |
973 | |
974 | last = c; | |
25fe55af | 975 | } |
fa9a63c5 RM |
976 | |
977 | if (in_range) | |
a1a052df | 978 | fprintf (stderr, "%c", last); |
fa9a63c5 | 979 | |
a1a052df | 980 | fprintf (stderr, "]"); |
fa9a63c5 | 981 | |
99633e97 | 982 | p += 1 + length; |
96cc36cc | 983 | |
96cc36cc | 984 | if (has_range_table) |
99633e97 SM |
985 | { |
986 | int count; | |
a1a052df | 987 | fprintf (stderr, "has-range-table"); |
99633e97 SM |
988 | |
989 | /* ??? Should print the range table; for now, just skip it. */ | |
990 | p += 2; /* skip range table bits */ | |
991 | EXTRACT_NUMBER_AND_INCR (count, p); | |
992 | p = CHARSET_RANGE_TABLE_END (p, count); | |
993 | } | |
fa9a63c5 RM |
994 | } |
995 | break; | |
996 | ||
997 | case begline: | |
a1a052df | 998 | fprintf (stderr, "/begline"); |
25fe55af | 999 | break; |
fa9a63c5 RM |
1000 | |
1001 | case endline: | |
a1a052df | 1002 | fprintf (stderr, "/endline"); |
25fe55af | 1003 | break; |
fa9a63c5 RM |
1004 | |
1005 | case on_failure_jump: | |
dc4a2ee0 PE |
1006 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1007 | fprintf (stderr, "/on_failure_jump to %td", p + mcnt - start); | |
25fe55af | 1008 | break; |
fa9a63c5 RM |
1009 | |
1010 | case on_failure_keep_string_jump: | |
dc4a2ee0 PE |
1011 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1012 | fprintf (stderr, "/on_failure_keep_string_jump to %td", | |
1013 | p + mcnt - start); | |
25fe55af | 1014 | break; |
fa9a63c5 | 1015 | |
0683b6fa | 1016 | case on_failure_jump_nastyloop: |
dc4a2ee0 PE |
1017 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1018 | fprintf (stderr, "/on_failure_jump_nastyloop to %td", | |
1019 | p + mcnt - start); | |
0683b6fa SM |
1020 | break; |
1021 | ||
505bde11 | 1022 | case on_failure_jump_loop: |
dc4a2ee0 PE |
1023 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1024 | fprintf (stderr, "/on_failure_jump_loop to %td", | |
1025 | p + mcnt - start); | |
5e69f11e RM |
1026 | break; |
1027 | ||
505bde11 | 1028 | case on_failure_jump_smart: |
dc4a2ee0 PE |
1029 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1030 | fprintf (stderr, "/on_failure_jump_smart to %td", | |
1031 | p + mcnt - start); | |
5e69f11e RM |
1032 | break; |
1033 | ||
25fe55af | 1034 | case jump: |
dc4a2ee0 PE |
1035 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1036 | fprintf (stderr, "/jump to %td", p + mcnt - start); | |
fa9a63c5 RM |
1037 | break; |
1038 | ||
25fe55af | 1039 | case succeed_n: |
dc4a2ee0 PE |
1040 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1041 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1042 | fprintf (stderr, "/succeed_n to %td, %d times", | |
1043 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1044 | break; |
5e69f11e | 1045 | |
25fe55af | 1046 | case jump_n: |
dc4a2ee0 PE |
1047 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1048 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1049 | fprintf (stderr, "/jump_n to %td, %d times", | |
1050 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1051 | break; |
5e69f11e | 1052 | |
25fe55af | 1053 | case set_number_at: |
dc4a2ee0 PE |
1054 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
1055 | EXTRACT_NUMBER_AND_INCR (mcnt2, p); | |
1056 | fprintf (stderr, "/set_number_at location %td to %d", | |
1057 | p - 2 + mcnt - start, mcnt2); | |
25fe55af | 1058 | break; |
5e69f11e | 1059 | |
25fe55af | 1060 | case wordbound: |
a1a052df | 1061 | fprintf (stderr, "/wordbound"); |
fa9a63c5 RM |
1062 | break; |
1063 | ||
1064 | case notwordbound: | |
a1a052df | 1065 | fprintf (stderr, "/notwordbound"); |
25fe55af | 1066 | break; |
fa9a63c5 RM |
1067 | |
1068 | case wordbeg: | |
a1a052df | 1069 | fprintf (stderr, "/wordbeg"); |
fa9a63c5 | 1070 | break; |
5e69f11e | 1071 | |
fa9a63c5 | 1072 | case wordend: |
a1a052df | 1073 | fprintf (stderr, "/wordend"); |
e2543b02 | 1074 | break; |
5e69f11e | 1075 | |
669fa600 | 1076 | case symbeg: |
e2543b02 | 1077 | fprintf (stderr, "/symbeg"); |
669fa600 SM |
1078 | break; |
1079 | ||
1080 | case symend: | |
e2543b02 | 1081 | fprintf (stderr, "/symend"); |
669fa600 | 1082 | break; |
5e69f11e | 1083 | |
1fb352e0 | 1084 | case syntaxspec: |
a1a052df | 1085 | fprintf (stderr, "/syntaxspec"); |
1fb352e0 | 1086 | mcnt = *p++; |
a1a052df | 1087 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1088 | break; |
1089 | ||
1090 | case notsyntaxspec: | |
a1a052df | 1091 | fprintf (stderr, "/notsyntaxspec"); |
1fb352e0 | 1092 | mcnt = *p++; |
a1a052df | 1093 | fprintf (stderr, "/%d", mcnt); |
1fb352e0 SM |
1094 | break; |
1095 | ||
0b32bf0e | 1096 | # ifdef emacs |
fa9a63c5 | 1097 | case before_dot: |
a1a052df | 1098 | fprintf (stderr, "/before_dot"); |
25fe55af | 1099 | break; |
fa9a63c5 RM |
1100 | |
1101 | case at_dot: | |
a1a052df | 1102 | fprintf (stderr, "/at_dot"); |
25fe55af | 1103 | break; |
fa9a63c5 RM |
1104 | |
1105 | case after_dot: | |
a1a052df | 1106 | fprintf (stderr, "/after_dot"); |
25fe55af | 1107 | break; |
fa9a63c5 | 1108 | |
1fb352e0 | 1109 | case categoryspec: |
a1a052df | 1110 | fprintf (stderr, "/categoryspec"); |
fa9a63c5 | 1111 | mcnt = *p++; |
a1a052df | 1112 | fprintf (stderr, "/%d", mcnt); |
25fe55af | 1113 | break; |
5e69f11e | 1114 | |
1fb352e0 | 1115 | case notcategoryspec: |
a1a052df | 1116 | fprintf (stderr, "/notcategoryspec"); |
fa9a63c5 | 1117 | mcnt = *p++; |
a1a052df | 1118 | fprintf (stderr, "/%d", mcnt); |
fa9a63c5 | 1119 | break; |
0b32bf0e | 1120 | # endif /* emacs */ |
fa9a63c5 | 1121 | |
fa9a63c5 | 1122 | case begbuf: |
a1a052df | 1123 | fprintf (stderr, "/begbuf"); |
25fe55af | 1124 | break; |
fa9a63c5 RM |
1125 | |
1126 | case endbuf: | |
a1a052df | 1127 | fprintf (stderr, "/endbuf"); |
25fe55af | 1128 | break; |
fa9a63c5 | 1129 | |
25fe55af | 1130 | default: |
a1a052df | 1131 | fprintf (stderr, "?%d", *(p-1)); |
fa9a63c5 RM |
1132 | } |
1133 | ||
a1a052df | 1134 | fprintf (stderr, "\n"); |
fa9a63c5 RM |
1135 | } |
1136 | ||
dc4a2ee0 | 1137 | fprintf (stderr, "%td:\tend of pattern.\n", p - start); |
fa9a63c5 RM |
1138 | } |
1139 | ||
1140 | ||
dc4a2ee0 PE |
1141 | static void |
1142 | print_compiled_pattern (struct re_pattern_buffer *bufp) | |
fa9a63c5 | 1143 | { |
01618498 | 1144 | re_char *buffer = bufp->buffer; |
fa9a63c5 RM |
1145 | |
1146 | print_partial_compiled_pattern (buffer, buffer + bufp->used); | |
4bb91c68 SM |
1147 | printf ("%ld bytes used/%ld bytes allocated.\n", |
1148 | bufp->used, bufp->allocated); | |
fa9a63c5 RM |
1149 | |
1150 | if (bufp->fastmap_accurate && bufp->fastmap) | |
1151 | { | |
1152 | printf ("fastmap: "); | |
1153 | print_fastmap (bufp->fastmap); | |
1154 | } | |
1155 | ||
dc4a2ee0 | 1156 | printf ("re_nsub: %zu\t", bufp->re_nsub); |
fa9a63c5 RM |
1157 | printf ("regs_alloc: %d\t", bufp->regs_allocated); |
1158 | printf ("can_be_null: %d\t", bufp->can_be_null); | |
fa9a63c5 RM |
1159 | printf ("no_sub: %d\t", bufp->no_sub); |
1160 | printf ("not_bol: %d\t", bufp->not_bol); | |
1161 | printf ("not_eol: %d\t", bufp->not_eol); | |
4bb91c68 | 1162 | printf ("syntax: %lx\n", bufp->syntax); |
505bde11 | 1163 | fflush (stdout); |
fa9a63c5 RM |
1164 | /* Perhaps we should print the translate table? */ |
1165 | } | |
1166 | ||
1167 | ||
dc4a2ee0 PE |
1168 | static void |
1169 | print_double_string (re_char *where, re_char *string1, ssize_t size1, | |
1170 | re_char *string2, ssize_t size2) | |
fa9a63c5 | 1171 | { |
d1dfb56c | 1172 | ssize_t this_char; |
5e69f11e | 1173 | |
fa9a63c5 RM |
1174 | if (where == NULL) |
1175 | printf ("(null)"); | |
1176 | else | |
1177 | { | |
1178 | if (FIRST_STRING_P (where)) | |
25fe55af RS |
1179 | { |
1180 | for (this_char = where - string1; this_char < size1; this_char++) | |
1181 | putchar (string1[this_char]); | |
fa9a63c5 | 1182 | |
25fe55af RS |
1183 | where = string2; |
1184 | } | |
fa9a63c5 RM |
1185 | |
1186 | for (this_char = where - string2; this_char < size2; this_char++) | |
25fe55af | 1187 | putchar (string2[this_char]); |
fa9a63c5 RM |
1188 | } |
1189 | } | |
1190 | ||
1191 | #else /* not DEBUG */ | |
1192 | ||
0b32bf0e SM |
1193 | # undef assert |
1194 | # define assert(e) | |
fa9a63c5 | 1195 | |
0b32bf0e | 1196 | # define DEBUG_STATEMENT(e) |
dc4a2ee0 PE |
1197 | # if __STDC_VERSION__ < 199901L |
1198 | # define DEBUG_COMPILES_ARGUMENTS | |
1199 | # define DEBUG_PRINT /* 'DEBUG_PRINT (x, y)' discards X and Y. */ (void) | |
1200 | # else | |
1201 | # define DEBUG_PRINT(...) | |
1202 | # endif | |
0b32bf0e SM |
1203 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) |
1204 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | |
fa9a63c5 RM |
1205 | |
1206 | #endif /* not DEBUG */ | |
1207 | \f | |
4da60324 PE |
1208 | /* Use this to suppress gcc's `...may be used before initialized' warnings. */ |
1209 | #ifdef lint | |
1210 | # define IF_LINT(Code) Code | |
1211 | #else | |
1212 | # define IF_LINT(Code) /* empty */ | |
1213 | #endif | |
1214 | \f | |
fa9a63c5 RM |
1215 | /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can |
1216 | also be assigned to arbitrarily: each pattern buffer stores its own | |
1217 | syntax, so it can be changed between regex compilations. */ | |
1218 | /* This has no initializer because initialized variables in Emacs | |
1219 | become read-only after dumping. */ | |
1220 | reg_syntax_t re_syntax_options; | |
1221 | ||
1222 | ||
1223 | /* Specify the precise syntax of regexps for compilation. This provides | |
1224 | for compatibility for various utilities which historically have | |
1225 | different, incompatible syntaxes. | |
1226 | ||
1227 | The argument SYNTAX is a bit mask comprised of the various bits | |
4bb91c68 | 1228 | defined in regex.h. We return the old syntax. */ |
fa9a63c5 RM |
1229 | |
1230 | reg_syntax_t | |
971de7fb | 1231 | re_set_syntax (reg_syntax_t syntax) |
fa9a63c5 RM |
1232 | { |
1233 | reg_syntax_t ret = re_syntax_options; | |
5e69f11e | 1234 | |
fa9a63c5 RM |
1235 | re_syntax_options = syntax; |
1236 | return ret; | |
1237 | } | |
c0f9ea08 | 1238 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) |
f9b0fd99 RS |
1239 | |
1240 | /* Regexp to use to replace spaces, or NULL meaning don't. */ | |
f462f075 | 1241 | static const_re_char *whitespace_regexp; |
f9b0fd99 RS |
1242 | |
1243 | void | |
971de7fb | 1244 | re_set_whitespace_regexp (const char *regexp) |
f9b0fd99 | 1245 | { |
f462f075 | 1246 | whitespace_regexp = (const_re_char *) regexp; |
f9b0fd99 RS |
1247 | } |
1248 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) | |
fa9a63c5 RM |
1249 | \f |
1250 | /* This table gives an error message for each of the error codes listed | |
4bb91c68 | 1251 | in regex.h. Obviously the order here has to be same as there. |
fa9a63c5 | 1252 | POSIX doesn't require that we do anything for REG_NOERROR, |
4bb91c68 | 1253 | but why not be nice? */ |
fa9a63c5 RM |
1254 | |
1255 | static const char *re_error_msgid[] = | |
5e69f11e RM |
1256 | { |
1257 | gettext_noop ("Success"), /* REG_NOERROR */ | |
1258 | gettext_noop ("No match"), /* REG_NOMATCH */ | |
1259 | gettext_noop ("Invalid regular expression"), /* REG_BADPAT */ | |
1260 | gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */ | |
1261 | gettext_noop ("Invalid character class name"), /* REG_ECTYPE */ | |
1262 | gettext_noop ("Trailing backslash"), /* REG_EESCAPE */ | |
1263 | gettext_noop ("Invalid back reference"), /* REG_ESUBREG */ | |
1264 | gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */ | |
1265 | gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */ | |
1266 | gettext_noop ("Unmatched \\{"), /* REG_EBRACE */ | |
1267 | gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */ | |
1268 | gettext_noop ("Invalid range end"), /* REG_ERANGE */ | |
1269 | gettext_noop ("Memory exhausted"), /* REG_ESPACE */ | |
1270 | gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */ | |
1271 | gettext_noop ("Premature end of regular expression"), /* REG_EEND */ | |
1272 | gettext_noop ("Regular expression too big"), /* REG_ESIZE */ | |
1273 | gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */ | |
b3e4c897 | 1274 | gettext_noop ("Range striding over charsets") /* REG_ERANGEX */ |
fa9a63c5 RM |
1275 | }; |
1276 | \f | |
4bb91c68 | 1277 | /* Avoiding alloca during matching, to placate r_alloc. */ |
fa9a63c5 RM |
1278 | |
1279 | /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the | |
1280 | searching and matching functions should not call alloca. On some | |
1281 | systems, alloca is implemented in terms of malloc, and if we're | |
1282 | using the relocating allocator routines, then malloc could cause a | |
1283 | relocation, which might (if the strings being searched are in the | |
1284 | ralloc heap) shift the data out from underneath the regexp | |
1285 | routines. | |
1286 | ||
5e69f11e | 1287 | Here's another reason to avoid allocation: Emacs |
fa9a63c5 RM |
1288 | processes input from X in a signal handler; processing X input may |
1289 | call malloc; if input arrives while a matching routine is calling | |
1290 | malloc, then we're scrod. But Emacs can't just block input while | |
1291 | calling matching routines; then we don't notice interrupts when | |
1292 | they come in. So, Emacs blocks input around all regexp calls | |
1293 | except the matching calls, which it leaves unprotected, in the | |
1294 | faith that they will not malloc. */ | |
1295 | ||
1296 | /* Normally, this is fine. */ | |
1297 | #define MATCH_MAY_ALLOCATE | |
1298 | ||
fa9a63c5 RM |
1299 | /* The match routines may not allocate if (1) they would do it with malloc |
1300 | and (2) it's not safe for them to use malloc. | |
1301 | Note that if REL_ALLOC is defined, matching would not use malloc for the | |
1302 | failure stack, but we would still use it for the register vectors; | |
4bb91c68 | 1303 | so REL_ALLOC should not affect this. */ |
b588157e | 1304 | #if defined REGEX_MALLOC && defined emacs |
0b32bf0e | 1305 | # undef MATCH_MAY_ALLOCATE |
fa9a63c5 RM |
1306 | #endif |
1307 | ||
1308 | \f | |
1309 | /* Failure stack declarations and macros; both re_compile_fastmap and | |
1310 | re_match_2 use a failure stack. These have to be macros because of | |
1311 | REGEX_ALLOCATE_STACK. */ | |
5e69f11e | 1312 | |
fa9a63c5 | 1313 | |
320a2a73 | 1314 | /* Approximate number of failure points for which to initially allocate space |
fa9a63c5 RM |
1315 | when matching. If this number is exceeded, we allocate more |
1316 | space, so it is not a hard limit. */ | |
1317 | #ifndef INIT_FAILURE_ALLOC | |
0b32bf0e | 1318 | # define INIT_FAILURE_ALLOC 20 |
fa9a63c5 RM |
1319 | #endif |
1320 | ||
1321 | /* Roughly the maximum number of failure points on the stack. Would be | |
320a2a73 | 1322 | exactly that if always used TYPICAL_FAILURE_SIZE items each time we failed. |
fa9a63c5 | 1323 | This is a variable only so users of regex can assign to it; we never |
ada30c0e SM |
1324 | change it ourselves. We always multiply it by TYPICAL_FAILURE_SIZE |
1325 | before using it, so it should probably be a byte-count instead. */ | |
c0f9ea08 SM |
1326 | # if defined MATCH_MAY_ALLOCATE |
1327 | /* Note that 4400 was enough to cause a crash on Alpha OSF/1, | |
320a2a73 KH |
1328 | whose default stack limit is 2mb. In order for a larger |
1329 | value to work reliably, you have to try to make it accord | |
1330 | with the process stack limit. */ | |
c0f9ea08 SM |
1331 | size_t re_max_failures = 40000; |
1332 | # else | |
1333 | size_t re_max_failures = 4000; | |
1334 | # endif | |
fa9a63c5 RM |
1335 | |
1336 | union fail_stack_elt | |
1337 | { | |
01618498 | 1338 | re_char *pointer; |
c0f9ea08 SM |
1339 | /* This should be the biggest `int' that's no bigger than a pointer. */ |
1340 | long integer; | |
fa9a63c5 RM |
1341 | }; |
1342 | ||
1343 | typedef union fail_stack_elt fail_stack_elt_t; | |
1344 | ||
1345 | typedef struct | |
1346 | { | |
1347 | fail_stack_elt_t *stack; | |
c0f9ea08 SM |
1348 | size_t size; |
1349 | size_t avail; /* Offset of next open position. */ | |
1350 | size_t frame; /* Offset of the cur constructed frame. */ | |
fa9a63c5 RM |
1351 | } fail_stack_type; |
1352 | ||
505bde11 | 1353 | #define FAIL_STACK_EMPTY() (fail_stack.frame == 0) |
fa9a63c5 RM |
1354 | |
1355 | ||
1356 | /* Define macros to initialize and free the failure stack. | |
1357 | Do `return -2' if the alloc fails. */ | |
1358 | ||
1359 | #ifdef MATCH_MAY_ALLOCATE | |
0b32bf0e | 1360 | # define INIT_FAIL_STACK() \ |
fa9a63c5 | 1361 | do { \ |
38182d90 | 1362 | fail_stack.stack = \ |
320a2a73 KH |
1363 | REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * TYPICAL_FAILURE_SIZE \ |
1364 | * sizeof (fail_stack_elt_t)); \ | |
fa9a63c5 RM |
1365 | \ |
1366 | if (fail_stack.stack == NULL) \ | |
1367 | return -2; \ | |
1368 | \ | |
1369 | fail_stack.size = INIT_FAILURE_ALLOC; \ | |
1370 | fail_stack.avail = 0; \ | |
505bde11 | 1371 | fail_stack.frame = 0; \ |
fa9a63c5 | 1372 | } while (0) |
fa9a63c5 | 1373 | #else |
0b32bf0e | 1374 | # define INIT_FAIL_STACK() \ |
fa9a63c5 RM |
1375 | do { \ |
1376 | fail_stack.avail = 0; \ | |
505bde11 | 1377 | fail_stack.frame = 0; \ |
fa9a63c5 RM |
1378 | } while (0) |
1379 | ||
b313f9d8 PE |
1380 | # define RETALLOC_IF(addr, n, t) \ |
1381 | if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t) | |
fa9a63c5 RM |
1382 | #endif |
1383 | ||
1384 | ||
320a2a73 KH |
1385 | /* Double the size of FAIL_STACK, up to a limit |
1386 | which allows approximately `re_max_failures' items. | |
fa9a63c5 RM |
1387 | |
1388 | Return 1 if succeeds, and 0 if either ran out of memory | |
5e69f11e RM |
1389 | allocating space for it or it was already too large. |
1390 | ||
4bb91c68 | 1391 | REGEX_REALLOCATE_STACK requires `destination' be declared. */ |
fa9a63c5 | 1392 | |
320a2a73 KH |
1393 | /* Factor to increase the failure stack size by |
1394 | when we increase it. | |
1395 | This used to be 2, but 2 was too wasteful | |
1396 | because the old discarded stacks added up to as much space | |
1397 | were as ultimate, maximum-size stack. */ | |
1398 | #define FAIL_STACK_GROWTH_FACTOR 4 | |
1399 | ||
1400 | #define GROW_FAIL_STACK(fail_stack) \ | |
eead07d6 KH |
1401 | (((fail_stack).size * sizeof (fail_stack_elt_t) \ |
1402 | >= re_max_failures * TYPICAL_FAILURE_SIZE) \ | |
fa9a63c5 | 1403 | ? 0 \ |
320a2a73 | 1404 | : ((fail_stack).stack \ |
38182d90 | 1405 | = REGEX_REALLOCATE_STACK ((fail_stack).stack, \ |
25fe55af | 1406 | (fail_stack).size * sizeof (fail_stack_elt_t), \ |
320a2a73 KH |
1407 | MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ |
1408 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1409 | * FAIL_STACK_GROWTH_FACTOR))), \ | |
fa9a63c5 RM |
1410 | \ |
1411 | (fail_stack).stack == NULL \ | |
1412 | ? 0 \ | |
6453db45 KH |
1413 | : ((fail_stack).size \ |
1414 | = (MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ | |
1415 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | |
1416 | * FAIL_STACK_GROWTH_FACTOR)) \ | |
1417 | / sizeof (fail_stack_elt_t)), \ | |
25fe55af | 1418 | 1))) |
fa9a63c5 RM |
1419 | |
1420 | ||
fa9a63c5 RM |
1421 | /* Push a pointer value onto the failure stack. |
1422 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1423 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 | 1424 | #define PUSH_FAILURE_POINTER(item) \ |
01618498 | 1425 | fail_stack.stack[fail_stack.avail++].pointer = (item) |
fa9a63c5 RM |
1426 | |
1427 | /* This pushes an integer-valued item onto the failure stack. | |
1428 | Assumes the variable `fail_stack'. Probably should only | |
4bb91c68 | 1429 | be called from within `PUSH_FAILURE_POINT'. */ |
fa9a63c5 RM |
1430 | #define PUSH_FAILURE_INT(item) \ |
1431 | fail_stack.stack[fail_stack.avail++].integer = (item) | |
1432 | ||
b313f9d8 | 1433 | /* These POP... operations complement the PUSH... operations. |
fa9a63c5 RM |
1434 | All assume that `fail_stack' is nonempty. */ |
1435 | #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer | |
1436 | #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer | |
fa9a63c5 | 1437 | |
505bde11 SM |
1438 | /* Individual items aside from the registers. */ |
1439 | #define NUM_NONREG_ITEMS 3 | |
1440 | ||
1441 | /* Used to examine the stack (to detect infinite loops). */ | |
1442 | #define FAILURE_PAT(h) fail_stack.stack[(h) - 1].pointer | |
66f0296e | 1443 | #define FAILURE_STR(h) (fail_stack.stack[(h) - 2].pointer) |
505bde11 SM |
1444 | #define NEXT_FAILURE_HANDLE(h) fail_stack.stack[(h) - 3].integer |
1445 | #define TOP_FAILURE_HANDLE() fail_stack.frame | |
fa9a63c5 RM |
1446 | |
1447 | ||
505bde11 SM |
1448 | #define ENSURE_FAIL_STACK(space) \ |
1449 | while (REMAINING_AVAIL_SLOTS <= space) { \ | |
1450 | if (!GROW_FAIL_STACK (fail_stack)) \ | |
1451 | return -2; \ | |
dc4a2ee0 PE |
1452 | DEBUG_PRINT ("\n Doubled stack; size now: %zd\n", (fail_stack).size);\ |
1453 | DEBUG_PRINT (" slots available: %zd\n", REMAINING_AVAIL_SLOTS);\ | |
505bde11 SM |
1454 | } |
1455 | ||
1456 | /* Push register NUM onto the stack. */ | |
1457 | #define PUSH_FAILURE_REG(num) \ | |
1458 | do { \ | |
1459 | char *destination; \ | |
dc4a2ee0 | 1460 | long n = num; \ |
505bde11 | 1461 | ENSURE_FAIL_STACK(3); \ |
dc4a2ee0 PE |
1462 | DEBUG_PRINT (" Push reg %ld (spanning %p -> %p)\n", \ |
1463 | n, regstart[n], regend[n]); \ | |
1464 | PUSH_FAILURE_POINTER (regstart[n]); \ | |
1465 | PUSH_FAILURE_POINTER (regend[n]); \ | |
1466 | PUSH_FAILURE_INT (n); \ | |
505bde11 SM |
1467 | } while (0) |
1468 | ||
01618498 SM |
1469 | /* Change the counter's value to VAL, but make sure that it will |
1470 | be reset when backtracking. */ | |
1471 | #define PUSH_NUMBER(ptr,val) \ | |
dc1e502d SM |
1472 | do { \ |
1473 | char *destination; \ | |
1474 | int c; \ | |
1475 | ENSURE_FAIL_STACK(3); \ | |
1476 | EXTRACT_NUMBER (c, ptr); \ | |
dc4a2ee0 | 1477 | DEBUG_PRINT (" Push number %p = %d -> %d\n", ptr, c, val); \ |
dc1e502d SM |
1478 | PUSH_FAILURE_INT (c); \ |
1479 | PUSH_FAILURE_POINTER (ptr); \ | |
1480 | PUSH_FAILURE_INT (-1); \ | |
01618498 | 1481 | STORE_NUMBER (ptr, val); \ |
dc1e502d SM |
1482 | } while (0) |
1483 | ||
505bde11 | 1484 | /* Pop a saved register off the stack. */ |
dc1e502d | 1485 | #define POP_FAILURE_REG_OR_COUNT() \ |
505bde11 | 1486 | do { \ |
d1dfb56c | 1487 | long pfreg = POP_FAILURE_INT (); \ |
19ed5445 | 1488 | if (pfreg == -1) \ |
dc1e502d SM |
1489 | { \ |
1490 | /* It's a counter. */ \ | |
6dcf2d0e SM |
1491 | /* Here, we discard `const', making re_match non-reentrant. */ \ |
1492 | unsigned char *ptr = (unsigned char*) POP_FAILURE_POINTER (); \ | |
19ed5445 PE |
1493 | pfreg = POP_FAILURE_INT (); \ |
1494 | STORE_NUMBER (ptr, pfreg); \ | |
dc4a2ee0 | 1495 | DEBUG_PRINT (" Pop counter %p = %ld\n", ptr, pfreg); \ |
dc1e502d SM |
1496 | } \ |
1497 | else \ | |
1498 | { \ | |
19ed5445 PE |
1499 | regend[pfreg] = POP_FAILURE_POINTER (); \ |
1500 | regstart[pfreg] = POP_FAILURE_POINTER (); \ | |
dc4a2ee0 PE |
1501 | DEBUG_PRINT (" Pop reg %ld (spanning %p -> %p)\n", \ |
1502 | pfreg, regstart[pfreg], regend[pfreg]); \ | |
dc1e502d | 1503 | } \ |
505bde11 SM |
1504 | } while (0) |
1505 | ||
1506 | /* Check that we are not stuck in an infinite loop. */ | |
1507 | #define CHECK_INFINITE_LOOP(pat_cur, string_place) \ | |
1508 | do { \ | |
d1dfb56c | 1509 | ssize_t failure = TOP_FAILURE_HANDLE (); \ |
505bde11 | 1510 | /* Check for infinite matching loops */ \ |
f6df485f RS |
1511 | while (failure > 0 \ |
1512 | && (FAILURE_STR (failure) == string_place \ | |
1513 | || FAILURE_STR (failure) == NULL)) \ | |
505bde11 SM |
1514 | { \ |
1515 | assert (FAILURE_PAT (failure) >= bufp->buffer \ | |
66f0296e | 1516 | && FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \ |
505bde11 | 1517 | if (FAILURE_PAT (failure) == pat_cur) \ |
f6df485f | 1518 | { \ |
6df42991 SM |
1519 | cycle = 1; \ |
1520 | break; \ | |
f6df485f | 1521 | } \ |
dc4a2ee0 | 1522 | DEBUG_PRINT (" Other pattern: %p\n", FAILURE_PAT (failure)); \ |
505bde11 SM |
1523 | failure = NEXT_FAILURE_HANDLE(failure); \ |
1524 | } \ | |
dc4a2ee0 | 1525 | DEBUG_PRINT (" Other string: %p\n", FAILURE_STR (failure)); \ |
505bde11 | 1526 | } while (0) |
6df42991 | 1527 | |
fa9a63c5 | 1528 | /* Push the information about the state we will need |
5e69f11e RM |
1529 | if we ever fail back to it. |
1530 | ||
505bde11 | 1531 | Requires variables fail_stack, regstart, regend and |
320a2a73 | 1532 | num_regs be declared. GROW_FAIL_STACK requires `destination' be |
fa9a63c5 | 1533 | declared. |
5e69f11e | 1534 | |
fa9a63c5 RM |
1535 | Does `return FAILURE_CODE' if runs out of memory. */ |
1536 | ||
505bde11 SM |
1537 | #define PUSH_FAILURE_POINT(pattern, string_place) \ |
1538 | do { \ | |
1539 | char *destination; \ | |
1540 | /* Must be int, so when we don't save any registers, the arithmetic \ | |
1541 | of 0 + -1 isn't done as unsigned. */ \ | |
1542 | \ | |
505bde11 | 1543 | DEBUG_STATEMENT (nfailure_points_pushed++); \ |
dc4a2ee0 PE |
1544 | DEBUG_PRINT ("\nPUSH_FAILURE_POINT:\n"); \ |
1545 | DEBUG_PRINT (" Before push, next avail: %zd\n", (fail_stack).avail); \ | |
1546 | DEBUG_PRINT (" size: %zd\n", (fail_stack).size);\ | |
505bde11 SM |
1547 | \ |
1548 | ENSURE_FAIL_STACK (NUM_NONREG_ITEMS); \ | |
1549 | \ | |
dc4a2ee0 | 1550 | DEBUG_PRINT ("\n"); \ |
505bde11 | 1551 | \ |
dc4a2ee0 | 1552 | DEBUG_PRINT (" Push frame index: %zd\n", fail_stack.frame); \ |
505bde11 SM |
1553 | PUSH_FAILURE_INT (fail_stack.frame); \ |
1554 | \ | |
dc4a2ee0 | 1555 | DEBUG_PRINT (" Push string %p: `", string_place); \ |
505bde11 | 1556 | DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, size2);\ |
dc4a2ee0 | 1557 | DEBUG_PRINT ("'\n"); \ |
505bde11 SM |
1558 | PUSH_FAILURE_POINTER (string_place); \ |
1559 | \ | |
dc4a2ee0 | 1560 | DEBUG_PRINT (" Push pattern %p: ", pattern); \ |
505bde11 SM |
1561 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern, pend); \ |
1562 | PUSH_FAILURE_POINTER (pattern); \ | |
1563 | \ | |
1564 | /* Close the frame by moving the frame pointer past it. */ \ | |
1565 | fail_stack.frame = fail_stack.avail; \ | |
1566 | } while (0) | |
fa9a63c5 | 1567 | |
320a2a73 KH |
1568 | /* Estimate the size of data pushed by a typical failure stack entry. |
1569 | An estimate is all we need, because all we use this for | |
1570 | is to choose a limit for how big to make the failure stack. */ | |
ada30c0e | 1571 | /* BEWARE, the value `20' is hard-coded in emacs.c:main(). */ |
320a2a73 | 1572 | #define TYPICAL_FAILURE_SIZE 20 |
fa9a63c5 | 1573 | |
fa9a63c5 RM |
1574 | /* How many items can still be added to the stack without overflowing it. */ |
1575 | #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) | |
1576 | ||
1577 | ||
1578 | /* Pops what PUSH_FAIL_STACK pushes. | |
1579 | ||
1580 | We restore into the parameters, all of which should be lvalues: | |
1581 | STR -- the saved data position. | |
1582 | PAT -- the saved pattern position. | |
fa9a63c5 | 1583 | REGSTART, REGEND -- arrays of string positions. |
5e69f11e | 1584 | |
fa9a63c5 | 1585 | Also assumes the variables `fail_stack' and (if debugging), `bufp', |
7814e705 | 1586 | `pend', `string1', `size1', `string2', and `size2'. */ |
fa9a63c5 | 1587 | |
505bde11 SM |
1588 | #define POP_FAILURE_POINT(str, pat) \ |
1589 | do { \ | |
fa9a63c5 RM |
1590 | assert (!FAIL_STACK_EMPTY ()); \ |
1591 | \ | |
1592 | /* Remove failure points and point to how many regs pushed. */ \ | |
dc4a2ee0 PE |
1593 | DEBUG_PRINT ("POP_FAILURE_POINT:\n"); \ |
1594 | DEBUG_PRINT (" Before pop, next avail: %zd\n", fail_stack.avail); \ | |
1595 | DEBUG_PRINT (" size: %zd\n", fail_stack.size); \ | |
fa9a63c5 | 1596 | \ |
505bde11 SM |
1597 | /* Pop the saved registers. */ \ |
1598 | while (fail_stack.frame < fail_stack.avail) \ | |
dc1e502d | 1599 | POP_FAILURE_REG_OR_COUNT (); \ |
fa9a63c5 | 1600 | \ |
dc4a2ee0 PE |
1601 | pat = POP_FAILURE_POINTER (); \ |
1602 | DEBUG_PRINT (" Popping pattern %p: ", pat); \ | |
505bde11 | 1603 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ |
fa9a63c5 RM |
1604 | \ |
1605 | /* If the saved string location is NULL, it came from an \ | |
1606 | on_failure_keep_string_jump opcode, and we want to throw away the \ | |
1607 | saved NULL, thus retaining our current position in the string. */ \ | |
01618498 | 1608 | str = POP_FAILURE_POINTER (); \ |
dc4a2ee0 | 1609 | DEBUG_PRINT (" Popping string %p: `", str); \ |
fa9a63c5 | 1610 | DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ |
dc4a2ee0 | 1611 | DEBUG_PRINT ("'\n"); \ |
fa9a63c5 | 1612 | \ |
505bde11 | 1613 | fail_stack.frame = POP_FAILURE_INT (); \ |
dc4a2ee0 | 1614 | DEBUG_PRINT (" Popping frame index: %zd\n", fail_stack.frame); \ |
fa9a63c5 | 1615 | \ |
505bde11 SM |
1616 | assert (fail_stack.avail >= 0); \ |
1617 | assert (fail_stack.frame <= fail_stack.avail); \ | |
fa9a63c5 | 1618 | \ |
fa9a63c5 | 1619 | DEBUG_STATEMENT (nfailure_points_popped++); \ |
505bde11 | 1620 | } while (0) /* POP_FAILURE_POINT */ |
fa9a63c5 RM |
1621 | |
1622 | ||
1623 | \f | |
fa9a63c5 | 1624 | /* Registers are set to a sentinel when they haven't yet matched. */ |
4bb91c68 | 1625 | #define REG_UNSET(e) ((e) == NULL) |
fa9a63c5 RM |
1626 | \f |
1627 | /* Subroutine declarations and macros for regex_compile. */ | |
1628 | ||
261cb4bb PE |
1629 | static reg_errcode_t regex_compile (re_char *pattern, size_t size, |
1630 | reg_syntax_t syntax, | |
1631 | struct re_pattern_buffer *bufp); | |
1632 | static void store_op1 (re_opcode_t op, unsigned char *loc, int arg); | |
1633 | static void store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2); | |
1634 | static void insert_op1 (re_opcode_t op, unsigned char *loc, | |
1635 | int arg, unsigned char *end); | |
1636 | static void insert_op2 (re_opcode_t op, unsigned char *loc, | |
1637 | int arg1, int arg2, unsigned char *end); | |
1638 | static boolean at_begline_loc_p (re_char *pattern, re_char *p, | |
1639 | reg_syntax_t syntax); | |
1640 | static boolean at_endline_loc_p (re_char *p, re_char *pend, | |
1641 | reg_syntax_t syntax); | |
1642 | static re_char *skip_one_char (re_char *p); | |
1643 | static int analyse_first (re_char *p, re_char *pend, | |
1644 | char *fastmap, const int multibyte); | |
fa9a63c5 | 1645 | |
fa9a63c5 | 1646 | /* Fetch the next character in the uncompiled pattern, with no |
4bb91c68 | 1647 | translation. */ |
36595814 | 1648 | #define PATFETCH(c) \ |
2d1675e4 SM |
1649 | do { \ |
1650 | int len; \ | |
1651 | if (p == pend) return REG_EEND; \ | |
62a6e103 | 1652 | c = RE_STRING_CHAR_AND_LENGTH (p, len, multibyte); \ |
2d1675e4 | 1653 | p += len; \ |
fa9a63c5 RM |
1654 | } while (0) |
1655 | ||
fa9a63c5 RM |
1656 | |
1657 | /* If `translate' is non-null, return translate[D], else just D. We | |
1658 | cast the subscript to translate because some data is declared as | |
1659 | `char *', to avoid warnings when a string constant is passed. But | |
1660 | when we use a character as a subscript we must make it unsigned. */ | |
6676cb1c | 1661 | #ifndef TRANSLATE |
0b32bf0e | 1662 | # define TRANSLATE(d) \ |
66f0296e | 1663 | (RE_TRANSLATE_P (translate) ? RE_TRANSLATE (translate, (d)) : (d)) |
6676cb1c | 1664 | #endif |
fa9a63c5 RM |
1665 | |
1666 | ||
1667 | /* Macros for outputting the compiled pattern into `buffer'. */ | |
1668 | ||
1669 | /* If the buffer isn't allocated when it comes in, use this. */ | |
1670 | #define INIT_BUF_SIZE 32 | |
1671 | ||
4bb91c68 | 1672 | /* Make sure we have at least N more bytes of space in buffer. */ |
fa9a63c5 | 1673 | #define GET_BUFFER_SPACE(n) \ |
01618498 | 1674 | while ((size_t) (b - bufp->buffer + (n)) > bufp->allocated) \ |
fa9a63c5 RM |
1675 | EXTEND_BUFFER () |
1676 | ||
1677 | /* Make sure we have one more byte of buffer space and then add C to it. */ | |
1678 | #define BUF_PUSH(c) \ | |
1679 | do { \ | |
1680 | GET_BUFFER_SPACE (1); \ | |
1681 | *b++ = (unsigned char) (c); \ | |
1682 | } while (0) | |
1683 | ||
1684 | ||
1685 | /* Ensure we have two more bytes of buffer space and then append C1 and C2. */ | |
1686 | #define BUF_PUSH_2(c1, c2) \ | |
1687 | do { \ | |
1688 | GET_BUFFER_SPACE (2); \ | |
1689 | *b++ = (unsigned char) (c1); \ | |
1690 | *b++ = (unsigned char) (c2); \ | |
1691 | } while (0) | |
1692 | ||
1693 | ||
fa9a63c5 | 1694 | /* Store a jump with opcode OP at LOC to location TO. We store a |
4bb91c68 | 1695 | relative address offset by the three bytes the jump itself occupies. */ |
fa9a63c5 RM |
1696 | #define STORE_JUMP(op, loc, to) \ |
1697 | store_op1 (op, loc, (to) - (loc) - 3) | |
1698 | ||
1699 | /* Likewise, for a two-argument jump. */ | |
1700 | #define STORE_JUMP2(op, loc, to, arg) \ | |
1701 | store_op2 (op, loc, (to) - (loc) - 3, arg) | |
1702 | ||
4bb91c68 | 1703 | /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ |
fa9a63c5 RM |
1704 | #define INSERT_JUMP(op, loc, to) \ |
1705 | insert_op1 (op, loc, (to) - (loc) - 3, b) | |
1706 | ||
1707 | /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ | |
1708 | #define INSERT_JUMP2(op, loc, to, arg) \ | |
1709 | insert_op2 (op, loc, (to) - (loc) - 3, arg, b) | |
1710 | ||
1711 | ||
1712 | /* This is not an arbitrary limit: the arguments which represent offsets | |
839966f3 | 1713 | into the pattern are two bytes long. So if 2^15 bytes turns out to |
fa9a63c5 | 1714 | be too small, many things would have to change. */ |
839966f3 KH |
1715 | # define MAX_BUF_SIZE (1L << 15) |
1716 | ||
fa9a63c5 RM |
1717 | /* Extend the buffer by twice its current size via realloc and |
1718 | reset the pointers that pointed into the old block to point to the | |
1719 | correct places in the new one. If extending the buffer results in it | |
4bb91c68 SM |
1720 | being larger than MAX_BUF_SIZE, then flag memory exhausted. */ |
1721 | #if __BOUNDED_POINTERS__ | |
1722 | # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) | |
381880b0 CY |
1723 | # define MOVE_BUFFER_POINTER(P) \ |
1724 | (__ptrlow (P) = new_buffer + (__ptrlow (P) - old_buffer), \ | |
1725 | SET_HIGH_BOUND (P), \ | |
1726 | __ptrvalue (P) = new_buffer + (__ptrvalue (P) - old_buffer)) | |
4bb91c68 SM |
1727 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
1728 | else \ | |
1729 | { \ | |
1730 | SET_HIGH_BOUND (b); \ | |
1731 | SET_HIGH_BOUND (begalt); \ | |
1732 | if (fixup_alt_jump) \ | |
1733 | SET_HIGH_BOUND (fixup_alt_jump); \ | |
1734 | if (laststart) \ | |
1735 | SET_HIGH_BOUND (laststart); \ | |
1736 | if (pending_exact) \ | |
1737 | SET_HIGH_BOUND (pending_exact); \ | |
1738 | } | |
1739 | #else | |
381880b0 | 1740 | # define MOVE_BUFFER_POINTER(P) ((P) = new_buffer + ((P) - old_buffer)) |
4bb91c68 SM |
1741 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND |
1742 | #endif | |
fa9a63c5 | 1743 | #define EXTEND_BUFFER() \ |
25fe55af | 1744 | do { \ |
381880b0 | 1745 | unsigned char *old_buffer = bufp->buffer; \ |
25fe55af | 1746 | if (bufp->allocated == MAX_BUF_SIZE) \ |
fa9a63c5 RM |
1747 | return REG_ESIZE; \ |
1748 | bufp->allocated <<= 1; \ | |
1749 | if (bufp->allocated > MAX_BUF_SIZE) \ | |
25fe55af | 1750 | bufp->allocated = MAX_BUF_SIZE; \ |
01618498 | 1751 | RETALLOC (bufp->buffer, bufp->allocated, unsigned char); \ |
fa9a63c5 RM |
1752 | if (bufp->buffer == NULL) \ |
1753 | return REG_ESPACE; \ | |
1754 | /* If the buffer moved, move all the pointers into it. */ \ | |
1755 | if (old_buffer != bufp->buffer) \ | |
1756 | { \ | |
381880b0 | 1757 | unsigned char *new_buffer = bufp->buffer; \ |
4bb91c68 SM |
1758 | MOVE_BUFFER_POINTER (b); \ |
1759 | MOVE_BUFFER_POINTER (begalt); \ | |
25fe55af | 1760 | if (fixup_alt_jump) \ |
4bb91c68 | 1761 | MOVE_BUFFER_POINTER (fixup_alt_jump); \ |
25fe55af | 1762 | if (laststart) \ |
4bb91c68 | 1763 | MOVE_BUFFER_POINTER (laststart); \ |
25fe55af | 1764 | if (pending_exact) \ |
4bb91c68 | 1765 | MOVE_BUFFER_POINTER (pending_exact); \ |
fa9a63c5 | 1766 | } \ |
4bb91c68 | 1767 | ELSE_EXTEND_BUFFER_HIGH_BOUND \ |
fa9a63c5 RM |
1768 | } while (0) |
1769 | ||
1770 | ||
1771 | /* Since we have one byte reserved for the register number argument to | |
1772 | {start,stop}_memory, the maximum number of groups we can report | |
1773 | things about is what fits in that byte. */ | |
1774 | #define MAX_REGNUM 255 | |
1775 | ||
1776 | /* But patterns can have more than `MAX_REGNUM' registers. We just | |
1777 | ignore the excess. */ | |
098d42af | 1778 | typedef int regnum_t; |
fa9a63c5 RM |
1779 | |
1780 | ||
1781 | /* Macros for the compile stack. */ | |
1782 | ||
1783 | /* Since offsets can go either forwards or backwards, this type needs to | |
4bb91c68 SM |
1784 | be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ |
1785 | /* int may be not enough when sizeof(int) == 2. */ | |
1786 | typedef long pattern_offset_t; | |
fa9a63c5 RM |
1787 | |
1788 | typedef struct | |
1789 | { | |
1790 | pattern_offset_t begalt_offset; | |
1791 | pattern_offset_t fixup_alt_jump; | |
5e69f11e | 1792 | pattern_offset_t laststart_offset; |
fa9a63c5 RM |
1793 | regnum_t regnum; |
1794 | } compile_stack_elt_t; | |
1795 | ||
1796 | ||
1797 | typedef struct | |
1798 | { | |
1799 | compile_stack_elt_t *stack; | |
d1dfb56c EZ |
1800 | size_t size; |
1801 | size_t avail; /* Offset of next open position. */ | |
fa9a63c5 RM |
1802 | } compile_stack_type; |
1803 | ||
1804 | ||
1805 | #define INIT_COMPILE_STACK_SIZE 32 | |
1806 | ||
1807 | #define COMPILE_STACK_EMPTY (compile_stack.avail == 0) | |
1808 | #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) | |
1809 | ||
4bb91c68 | 1810 | /* The next available element. */ |
fa9a63c5 RM |
1811 | #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) |
1812 | ||
0caaedb1 PE |
1813 | /* Explicit quit checking is needed for Emacs, which uses polling to |
1814 | process input events. */ | |
1815 | #ifdef emacs | |
77d11aec RS |
1816 | # define IMMEDIATE_QUIT_CHECK \ |
1817 | do { \ | |
1818 | if (immediate_quit) QUIT; \ | |
1819 | } while (0) | |
1820 | #else | |
1821 | # define IMMEDIATE_QUIT_CHECK ((void)0) | |
1822 | #endif | |
1823 | \f | |
b18215fc RS |
1824 | /* Structure to manage work area for range table. */ |
1825 | struct range_table_work_area | |
1826 | { | |
1827 | int *table; /* actual work area. */ | |
1828 | int allocated; /* allocated size for work area in bytes. */ | |
7814e705 | 1829 | int used; /* actually used size in words. */ |
96cc36cc | 1830 | int bits; /* flag to record character classes */ |
b18215fc RS |
1831 | }; |
1832 | ||
77d11aec RS |
1833 | /* Make sure that WORK_AREA can hold more N multibyte characters. |
1834 | This is used only in set_image_of_range and set_image_of_range_1. | |
1835 | It expects WORK_AREA to be a pointer. | |
1836 | If it can't get the space, it returns from the surrounding function. */ | |
1837 | ||
1838 | #define EXTEND_RANGE_TABLE(work_area, n) \ | |
1839 | do { \ | |
8f924df7 | 1840 | if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \ |
77d11aec | 1841 | { \ |
8f924df7 KH |
1842 | extend_range_table_work_area (&work_area); \ |
1843 | if ((work_area).table == 0) \ | |
77d11aec RS |
1844 | return (REG_ESPACE); \ |
1845 | } \ | |
b18215fc RS |
1846 | } while (0) |
1847 | ||
96cc36cc RS |
1848 | #define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \ |
1849 | (work_area).bits |= (bit) | |
1850 | ||
14473664 SM |
1851 | /* Bits used to implement the multibyte-part of the various character classes |
1852 | such as [:alnum:] in a charset's range table. */ | |
1853 | #define BIT_WORD 0x1 | |
1854 | #define BIT_LOWER 0x2 | |
1855 | #define BIT_PUNCT 0x4 | |
1856 | #define BIT_SPACE 0x8 | |
1857 | #define BIT_UPPER 0x10 | |
1858 | #define BIT_MULTIBYTE 0x20 | |
96cc36cc | 1859 | |
b18215fc RS |
1860 | /* Set a range (RANGE_START, RANGE_END) to WORK_AREA. */ |
1861 | #define SET_RANGE_TABLE_WORK_AREA(work_area, range_start, range_end) \ | |
77d11aec | 1862 | do { \ |
8f924df7 | 1863 | EXTEND_RANGE_TABLE ((work_area), 2); \ |
b18215fc RS |
1864 | (work_area).table[(work_area).used++] = (range_start); \ |
1865 | (work_area).table[(work_area).used++] = (range_end); \ | |
1866 | } while (0) | |
1867 | ||
7814e705 | 1868 | /* Free allocated memory for WORK_AREA. */ |
b18215fc RS |
1869 | #define FREE_RANGE_TABLE_WORK_AREA(work_area) \ |
1870 | do { \ | |
1871 | if ((work_area).table) \ | |
1872 | free ((work_area).table); \ | |
1873 | } while (0) | |
1874 | ||
96cc36cc | 1875 | #define CLEAR_RANGE_TABLE_WORK_USED(work_area) ((work_area).used = 0, (work_area).bits = 0) |
b18215fc | 1876 | #define RANGE_TABLE_WORK_USED(work_area) ((work_area).used) |
96cc36cc | 1877 | #define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits) |
b18215fc | 1878 | #define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i]) |
77d11aec | 1879 | \f |
b18215fc | 1880 | |
fa9a63c5 | 1881 | /* Set the bit for character C in a list. */ |
01618498 | 1882 | #define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH)) |
fa9a63c5 RM |
1883 | |
1884 | ||
bf216479 KH |
1885 | #ifdef emacs |
1886 | ||
cf9c99bc KH |
1887 | /* Store characters in the range FROM to TO in the bitmap at B (for |
1888 | ASCII and unibyte characters) and WORK_AREA (for multibyte | |
1889 | characters) while translating them and paying attention to the | |
1890 | continuity of translated characters. | |
8f924df7 | 1891 | |
cf9c99bc KH |
1892 | Implementation note: It is better to implement these fairly big |
1893 | macros by a function, but it's not that easy because macros called | |
8f924df7 | 1894 | in this macro assume various local variables already declared. */ |
bf216479 | 1895 | |
cf9c99bc KH |
1896 | /* Both FROM and TO are ASCII characters. */ |
1897 | ||
1898 | #define SETUP_ASCII_RANGE(work_area, FROM, TO) \ | |
1899 | do { \ | |
1900 | int C0, C1; \ | |
1901 | \ | |
1902 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1903 | { \ | |
1904 | C1 = TRANSLATE (C0); \ | |
1905 | if (! ASCII_CHAR_P (C1)) \ | |
1906 | { \ | |
1907 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
1908 | if ((C1 = RE_CHAR_TO_UNIBYTE (C1)) < 0) \ | |
1909 | C1 = C0; \ | |
1910 | } \ | |
1911 | SET_LIST_BIT (C1); \ | |
1912 | } \ | |
1913 | } while (0) | |
1914 | ||
1915 | ||
1916 | /* Both FROM and TO are unibyte characters (0x80..0xFF). */ | |
1917 | ||
1918 | #define SETUP_UNIBYTE_RANGE(work_area, FROM, TO) \ | |
1919 | do { \ | |
1920 | int C0, C1, C2, I; \ | |
1921 | int USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1922 | \ | |
1923 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1924 | { \ | |
1925 | C1 = RE_CHAR_TO_MULTIBYTE (C0); \ | |
1926 | if (CHAR_BYTE8_P (C1)) \ | |
1927 | SET_LIST_BIT (C0); \ | |
1928 | else \ | |
1929 | { \ | |
1930 | C2 = TRANSLATE (C1); \ | |
1931 | if (C2 == C1 \ | |
1932 | || (C1 = RE_CHAR_TO_UNIBYTE (C2)) < 0) \ | |
1933 | C1 = C0; \ | |
1934 | SET_LIST_BIT (C1); \ | |
1935 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1936 | { \ | |
1937 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1938 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1939 | \ | |
1940 | if (C2 >= from - 1 && C2 <= to + 1) \ | |
1941 | { \ | |
1942 | if (C2 == from - 1) \ | |
1943 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1944 | else if (C2 == to + 1) \ | |
1945 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
1946 | break; \ | |
1947 | } \ | |
1948 | } \ | |
1949 | if (I < USED) \ | |
1950 | SET_RANGE_TABLE_WORK_AREA ((work_area), C2, C2); \ | |
1951 | } \ | |
1952 | } \ | |
1953 | } while (0) | |
1954 | ||
1955 | ||
78edd3b7 | 1956 | /* Both FROM and TO are multibyte characters. */ |
cf9c99bc KH |
1957 | |
1958 | #define SETUP_MULTIBYTE_RANGE(work_area, FROM, TO) \ | |
1959 | do { \ | |
1960 | int C0, C1, C2, I, USED = RANGE_TABLE_WORK_USED (work_area); \ | |
1961 | \ | |
1962 | SET_RANGE_TABLE_WORK_AREA ((work_area), (FROM), (TO)); \ | |
1963 | for (C0 = (FROM); C0 <= (TO); C0++) \ | |
1964 | { \ | |
1965 | C1 = TRANSLATE (C0); \ | |
1966 | if ((C2 = RE_CHAR_TO_UNIBYTE (C1)) >= 0 \ | |
1967 | || (C1 != C0 && (C2 = RE_CHAR_TO_UNIBYTE (C0)) >= 0)) \ | |
1968 | SET_LIST_BIT (C2); \ | |
1969 | if (C1 >= (FROM) && C1 <= (TO)) \ | |
1970 | continue; \ | |
1971 | for (I = RANGE_TABLE_WORK_USED (work_area) - 2; I >= USED; I -= 2) \ | |
1972 | { \ | |
1973 | int from = RANGE_TABLE_WORK_ELT (work_area, I); \ | |
1974 | int to = RANGE_TABLE_WORK_ELT (work_area, I + 1); \ | |
1975 | \ | |
1976 | if (C1 >= from - 1 && C1 <= to + 1) \ | |
1977 | { \ | |
1978 | if (C1 == from - 1) \ | |
1979 | RANGE_TABLE_WORK_ELT (work_area, I)--; \ | |
1980 | else if (C1 == to + 1) \ | |
1981 | RANGE_TABLE_WORK_ELT (work_area, I + 1)++; \ | |
1982 | break; \ | |
1983 | } \ | |
1984 | } \ | |
1985 | if (I < USED) \ | |
1986 | SET_RANGE_TABLE_WORK_AREA ((work_area), C1, C1); \ | |
1987 | } \ | |
bf216479 KH |
1988 | } while (0) |
1989 | ||
1990 | #endif /* emacs */ | |
1991 | ||
fa9a63c5 | 1992 | /* Get the next unsigned number in the uncompiled pattern. */ |
25fe55af | 1993 | #define GET_UNSIGNED_NUMBER(num) \ |
c72b0edd SM |
1994 | do { \ |
1995 | if (p == pend) \ | |
1996 | FREE_STACK_RETURN (REG_EBRACE); \ | |
1997 | else \ | |
1998 | { \ | |
1999 | PATFETCH (c); \ | |
2000 | while ('0' <= c && c <= '9') \ | |
2001 | { \ | |
2002 | int prev; \ | |
2003 | if (num < 0) \ | |
2004 | num = 0; \ | |
2005 | prev = num; \ | |
2006 | num = num * 10 + c - '0'; \ | |
2007 | if (num / 10 != prev) \ | |
2008 | FREE_STACK_RETURN (REG_BADBR); \ | |
2009 | if (p == pend) \ | |
2010 | FREE_STACK_RETURN (REG_EBRACE); \ | |
2011 | PATFETCH (c); \ | |
2012 | } \ | |
2013 | } \ | |
2014 | } while (0) | |
77d11aec | 2015 | \f |
1fdab503 | 2016 | #if ! WIDE_CHAR_SUPPORT |
01618498 | 2017 | |
14473664 | 2018 | /* Map a string to the char class it names (if any). */ |
1fdab503 | 2019 | re_wctype_t |
29abe551 | 2020 | re_wctype (const_re_char *str) |
14473664 | 2021 | { |
5b0534c8 | 2022 | const char *string = (const char *) str; |
14473664 SM |
2023 | if (STREQ (string, "alnum")) return RECC_ALNUM; |
2024 | else if (STREQ (string, "alpha")) return RECC_ALPHA; | |
2025 | else if (STREQ (string, "word")) return RECC_WORD; | |
2026 | else if (STREQ (string, "ascii")) return RECC_ASCII; | |
2027 | else if (STREQ (string, "nonascii")) return RECC_NONASCII; | |
2028 | else if (STREQ (string, "graph")) return RECC_GRAPH; | |
2029 | else if (STREQ (string, "lower")) return RECC_LOWER; | |
2030 | else if (STREQ (string, "print")) return RECC_PRINT; | |
2031 | else if (STREQ (string, "punct")) return RECC_PUNCT; | |
2032 | else if (STREQ (string, "space")) return RECC_SPACE; | |
2033 | else if (STREQ (string, "upper")) return RECC_UPPER; | |
2034 | else if (STREQ (string, "unibyte")) return RECC_UNIBYTE; | |
2035 | else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE; | |
2036 | else if (STREQ (string, "digit")) return RECC_DIGIT; | |
2037 | else if (STREQ (string, "xdigit")) return RECC_XDIGIT; | |
2038 | else if (STREQ (string, "cntrl")) return RECC_CNTRL; | |
2039 | else if (STREQ (string, "blank")) return RECC_BLANK; | |
2040 | else return 0; | |
2041 | } | |
2042 | ||
e0f24100 | 2043 | /* True if CH is in the char class CC. */ |
1fdab503 | 2044 | boolean |
971de7fb | 2045 | re_iswctype (int ch, re_wctype_t cc) |
14473664 SM |
2046 | { |
2047 | switch (cc) | |
2048 | { | |
f3fcc40d AS |
2049 | case RECC_ALNUM: return ISALNUM (ch) != 0; |
2050 | case RECC_ALPHA: return ISALPHA (ch) != 0; | |
2051 | case RECC_BLANK: return ISBLANK (ch) != 0; | |
2052 | case RECC_CNTRL: return ISCNTRL (ch) != 0; | |
2053 | case RECC_DIGIT: return ISDIGIT (ch) != 0; | |
2054 | case RECC_GRAPH: return ISGRAPH (ch) != 0; | |
2055 | case RECC_LOWER: return ISLOWER (ch) != 0; | |
2056 | case RECC_PRINT: return ISPRINT (ch) != 0; | |
2057 | case RECC_PUNCT: return ISPUNCT (ch) != 0; | |
2058 | case RECC_SPACE: return ISSPACE (ch) != 0; | |
2059 | case RECC_UPPER: return ISUPPER (ch) != 0; | |
2060 | case RECC_XDIGIT: return ISXDIGIT (ch) != 0; | |
2061 | case RECC_ASCII: return IS_REAL_ASCII (ch) != 0; | |
213bd7f2 | 2062 | case RECC_NONASCII: return !IS_REAL_ASCII (ch); |
f3fcc40d | 2063 | case RECC_UNIBYTE: return ISUNIBYTE (ch) != 0; |
213bd7f2 | 2064 | case RECC_MULTIBYTE: return !ISUNIBYTE (ch); |
f3fcc40d | 2065 | case RECC_WORD: return ISWORD (ch) != 0; |
0cdd06f8 SM |
2066 | case RECC_ERROR: return false; |
2067 | default: | |
5e617bc2 | 2068 | abort (); |
14473664 SM |
2069 | } |
2070 | } | |
fa9a63c5 | 2071 | |
14473664 SM |
2072 | /* Return a bit-pattern to use in the range-table bits to match multibyte |
2073 | chars of class CC. */ | |
2074 | static int | |
971de7fb | 2075 | re_wctype_to_bit (re_wctype_t cc) |
14473664 SM |
2076 | { |
2077 | switch (cc) | |
2078 | { | |
2079 | case RECC_NONASCII: case RECC_PRINT: case RECC_GRAPH: | |
0cdd06f8 SM |
2080 | case RECC_MULTIBYTE: return BIT_MULTIBYTE; |
2081 | case RECC_ALPHA: case RECC_ALNUM: case RECC_WORD: return BIT_WORD; | |
2082 | case RECC_LOWER: return BIT_LOWER; | |
2083 | case RECC_UPPER: return BIT_UPPER; | |
2084 | case RECC_PUNCT: return BIT_PUNCT; | |
2085 | case RECC_SPACE: return BIT_SPACE; | |
14473664 | 2086 | case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL: |
0cdd06f8 SM |
2087 | case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0; |
2088 | default: | |
5e617bc2 | 2089 | abort (); |
14473664 SM |
2090 | } |
2091 | } | |
2092 | #endif | |
77d11aec RS |
2093 | \f |
2094 | /* Filling in the work area of a range. */ | |
2095 | ||
2096 | /* Actually extend the space in WORK_AREA. */ | |
2097 | ||
2098 | static void | |
971de7fb | 2099 | extend_range_table_work_area (struct range_table_work_area *work_area) |
177c0ea7 | 2100 | { |
77d11aec | 2101 | work_area->allocated += 16 * sizeof (int); |
38182d90 | 2102 | work_area->table = realloc (work_area->table, work_area->allocated); |
77d11aec RS |
2103 | } |
2104 | ||
8f924df7 | 2105 | #if 0 |
77d11aec RS |
2106 | #ifdef emacs |
2107 | ||
2108 | /* Carefully find the ranges of codes that are equivalent | |
2109 | under case conversion to the range start..end when passed through | |
2110 | TRANSLATE. Handle the case where non-letters can come in between | |
2111 | two upper-case letters (which happens in Latin-1). | |
2112 | Also handle the case of groups of more than 2 case-equivalent chars. | |
2113 | ||
2114 | The basic method is to look at consecutive characters and see | |
2115 | if they can form a run that can be handled as one. | |
2116 | ||
2117 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2118 | ||
2119 | static int | |
1dae0f0a AS |
2120 | set_image_of_range_1 (struct range_table_work_area *work_area, |
2121 | re_wchar_t start, re_wchar_t end, | |
2122 | RE_TRANSLATE_TYPE translate) | |
77d11aec RS |
2123 | { |
2124 | /* `one_case' indicates a character, or a run of characters, | |
2125 | each of which is an isolate (no case-equivalents). | |
2126 | This includes all ASCII non-letters. | |
2127 | ||
2128 | `two_case' indicates a character, or a run of characters, | |
2129 | each of which has two case-equivalent forms. | |
2130 | This includes all ASCII letters. | |
2131 | ||
2132 | `strange' indicates a character that has more than one | |
2133 | case-equivalent. */ | |
177c0ea7 | 2134 | |
77d11aec RS |
2135 | enum case_type {one_case, two_case, strange}; |
2136 | ||
2137 | /* Describe the run that is in progress, | |
2138 | which the next character can try to extend. | |
2139 | If run_type is strange, that means there really is no run. | |
2140 | If run_type is one_case, then run_start...run_end is the run. | |
2141 | If run_type is two_case, then the run is run_start...run_end, | |
2142 | and the case-equivalents end at run_eqv_end. */ | |
2143 | ||
2144 | enum case_type run_type = strange; | |
2145 | int run_start, run_end, run_eqv_end; | |
2146 | ||
2147 | Lisp_Object eqv_table; | |
2148 | ||
2149 | if (!RE_TRANSLATE_P (translate)) | |
2150 | { | |
b7c12565 | 2151 | EXTEND_RANGE_TABLE (work_area, 2); |
77d11aec RS |
2152 | work_area->table[work_area->used++] = (start); |
2153 | work_area->table[work_area->used++] = (end); | |
b7c12565 | 2154 | return -1; |
77d11aec RS |
2155 | } |
2156 | ||
2157 | eqv_table = XCHAR_TABLE (translate)->extras[2]; | |
99633e97 | 2158 | |
77d11aec RS |
2159 | for (; start <= end; start++) |
2160 | { | |
2161 | enum case_type this_type; | |
2162 | int eqv = RE_TRANSLATE (eqv_table, start); | |
2163 | int minchar, maxchar; | |
2164 | ||
2165 | /* Classify this character */ | |
2166 | if (eqv == start) | |
2167 | this_type = one_case; | |
2168 | else if (RE_TRANSLATE (eqv_table, eqv) == start) | |
2169 | this_type = two_case; | |
2170 | else | |
2171 | this_type = strange; | |
2172 | ||
2173 | if (start < eqv) | |
2174 | minchar = start, maxchar = eqv; | |
2175 | else | |
2176 | minchar = eqv, maxchar = start; | |
2177 | ||
2178 | /* Can this character extend the run in progress? */ | |
2179 | if (this_type == strange || this_type != run_type | |
2180 | || !(minchar == run_end + 1 | |
2181 | && (run_type == two_case | |
2182 | ? maxchar == run_eqv_end + 1 : 1))) | |
2183 | { | |
2184 | /* No, end the run. | |
2185 | Record each of its equivalent ranges. */ | |
2186 | if (run_type == one_case) | |
2187 | { | |
2188 | EXTEND_RANGE_TABLE (work_area, 2); | |
2189 | work_area->table[work_area->used++] = run_start; | |
2190 | work_area->table[work_area->used++] = run_end; | |
2191 | } | |
2192 | else if (run_type == two_case) | |
2193 | { | |
2194 | EXTEND_RANGE_TABLE (work_area, 4); | |
2195 | work_area->table[work_area->used++] = run_start; | |
2196 | work_area->table[work_area->used++] = run_end; | |
2197 | work_area->table[work_area->used++] | |
2198 | = RE_TRANSLATE (eqv_table, run_start); | |
2199 | work_area->table[work_area->used++] | |
2200 | = RE_TRANSLATE (eqv_table, run_end); | |
2201 | } | |
2202 | run_type = strange; | |
2203 | } | |
177c0ea7 | 2204 | |
77d11aec RS |
2205 | if (this_type == strange) |
2206 | { | |
2207 | /* For a strange character, add each of its equivalents, one | |
2208 | by one. Don't start a range. */ | |
2209 | do | |
2210 | { | |
2211 | EXTEND_RANGE_TABLE (work_area, 2); | |
2212 | work_area->table[work_area->used++] = eqv; | |
2213 | work_area->table[work_area->used++] = eqv; | |
2214 | eqv = RE_TRANSLATE (eqv_table, eqv); | |
2215 | } | |
2216 | while (eqv != start); | |
2217 | } | |
2218 | ||
2219 | /* Add this char to the run, or start a new run. */ | |
2220 | else if (run_type == strange) | |
2221 | { | |
2222 | /* Initialize a new range. */ | |
2223 | run_type = this_type; | |
2224 | run_start = start; | |
2225 | run_end = start; | |
2226 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2227 | } | |
2228 | else | |
2229 | { | |
2230 | /* Extend a running range. */ | |
2231 | run_end = minchar; | |
2232 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | |
2233 | } | |
2234 | } | |
2235 | ||
2236 | /* If a run is still in progress at the end, finish it now | |
2237 | by recording its equivalent ranges. */ | |
2238 | if (run_type == one_case) | |
2239 | { | |
2240 | EXTEND_RANGE_TABLE (work_area, 2); | |
2241 | work_area->table[work_area->used++] = run_start; | |
2242 | work_area->table[work_area->used++] = run_end; | |
2243 | } | |
2244 | else if (run_type == two_case) | |
2245 | { | |
2246 | EXTEND_RANGE_TABLE (work_area, 4); | |
2247 | work_area->table[work_area->used++] = run_start; | |
2248 | work_area->table[work_area->used++] = run_end; | |
2249 | work_area->table[work_area->used++] | |
2250 | = RE_TRANSLATE (eqv_table, run_start); | |
2251 | work_area->table[work_area->used++] | |
2252 | = RE_TRANSLATE (eqv_table, run_end); | |
2253 | } | |
2254 | ||
2255 | return -1; | |
2256 | } | |
36595814 | 2257 | |
77d11aec | 2258 | #endif /* emacs */ |
36595814 | 2259 | |
2b34df4e | 2260 | /* Record the image of the range start..end when passed through |
36595814 SM |
2261 | TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end) |
2262 | and is not even necessarily contiguous. | |
b7c12565 RS |
2263 | Normally we approximate it with the smallest contiguous range that contains |
2264 | all the chars we need. However, for Latin-1 we go to extra effort | |
2265 | to do a better job. | |
2266 | ||
2267 | This function is not called for ASCII ranges. | |
77d11aec RS |
2268 | |
2269 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | |
2270 | ||
2271 | static int | |
1dae0f0a AS |
2272 | set_image_of_range (struct range_table_work_area *work_area, |
2273 | re_wchar_t start, re_wchar_t end, | |
2274 | RE_TRANSLATE_TYPE translate) | |
36595814 | 2275 | { |
77d11aec RS |
2276 | re_wchar_t cmin, cmax; |
2277 | ||
2278 | #ifdef emacs | |
2279 | /* For Latin-1 ranges, use set_image_of_range_1 | |
2280 | to get proper handling of ranges that include letters and nonletters. | |
b7c12565 | 2281 | For a range that includes the whole of Latin-1, this is not necessary. |
77d11aec | 2282 | For other character sets, we don't bother to get this right. */ |
b7c12565 RS |
2283 | if (RE_TRANSLATE_P (translate) && start < 04400 |
2284 | && !(start < 04200 && end >= 04377)) | |
77d11aec | 2285 | { |
b7c12565 | 2286 | int newend; |
77d11aec | 2287 | int tem; |
b7c12565 RS |
2288 | newend = end; |
2289 | if (newend > 04377) | |
2290 | newend = 04377; | |
2291 | tem = set_image_of_range_1 (work_area, start, newend, translate); | |
77d11aec RS |
2292 | if (tem > 0) |
2293 | return tem; | |
2294 | ||
2295 | start = 04400; | |
2296 | if (end < 04400) | |
2297 | return -1; | |
2298 | } | |
2299 | #endif | |
2300 | ||
b7c12565 RS |
2301 | EXTEND_RANGE_TABLE (work_area, 2); |
2302 | work_area->table[work_area->used++] = (start); | |
2303 | work_area->table[work_area->used++] = (end); | |
2304 | ||
2305 | cmin = -1, cmax = -1; | |
77d11aec | 2306 | |
36595814 | 2307 | if (RE_TRANSLATE_P (translate)) |
b7c12565 RS |
2308 | { |
2309 | int ch; | |
77d11aec | 2310 | |
b7c12565 RS |
2311 | for (ch = start; ch <= end; ch++) |
2312 | { | |
2313 | re_wchar_t c = TRANSLATE (ch); | |
2314 | if (! (start <= c && c <= end)) | |
2315 | { | |
2316 | if (cmin == -1) | |
2317 | cmin = c, cmax = c; | |
2318 | else | |
2319 | { | |
2320 | cmin = MIN (cmin, c); | |
2321 | cmax = MAX (cmax, c); | |
2322 | } | |
2323 | } | |
2324 | } | |
2325 | ||
2326 | if (cmin != -1) | |
2327 | { | |
2328 | EXTEND_RANGE_TABLE (work_area, 2); | |
2329 | work_area->table[work_area->used++] = (cmin); | |
2330 | work_area->table[work_area->used++] = (cmax); | |
2331 | } | |
2332 | } | |
36595814 | 2333 | |
77d11aec RS |
2334 | return -1; |
2335 | } | |
8f924df7 | 2336 | #endif /* 0 */ |
fa9a63c5 RM |
2337 | \f |
2338 | #ifndef MATCH_MAY_ALLOCATE | |
2339 | ||
2340 | /* If we cannot allocate large objects within re_match_2_internal, | |
2341 | we make the fail stack and register vectors global. | |
2342 | The fail stack, we grow to the maximum size when a regexp | |
2343 | is compiled. | |
2344 | The register vectors, we adjust in size each time we | |
2345 | compile a regexp, according to the number of registers it needs. */ | |
2346 | ||
2347 | static fail_stack_type fail_stack; | |
2348 | ||
2349 | /* Size with which the following vectors are currently allocated. | |
2350 | That is so we can make them bigger as needed, | |
4bb91c68 | 2351 | but never make them smaller. */ |
fa9a63c5 RM |
2352 | static int regs_allocated_size; |
2353 | ||
66f0296e SM |
2354 | static re_char ** regstart, ** regend; |
2355 | static re_char **best_regstart, **best_regend; | |
fa9a63c5 RM |
2356 | |
2357 | /* Make the register vectors big enough for NUM_REGS registers, | |
4bb91c68 | 2358 | but don't make them smaller. */ |
fa9a63c5 RM |
2359 | |
2360 | static | |
1dae0f0a | 2361 | regex_grow_registers (int num_regs) |
fa9a63c5 RM |
2362 | { |
2363 | if (num_regs > regs_allocated_size) | |
2364 | { | |
66f0296e SM |
2365 | RETALLOC_IF (regstart, num_regs, re_char *); |
2366 | RETALLOC_IF (regend, num_regs, re_char *); | |
2367 | RETALLOC_IF (best_regstart, num_regs, re_char *); | |
2368 | RETALLOC_IF (best_regend, num_regs, re_char *); | |
fa9a63c5 RM |
2369 | |
2370 | regs_allocated_size = num_regs; | |
2371 | } | |
2372 | } | |
2373 | ||
2374 | #endif /* not MATCH_MAY_ALLOCATE */ | |
2375 | \f | |
261cb4bb PE |
2376 | static boolean group_in_compile_stack (compile_stack_type compile_stack, |
2377 | regnum_t regnum); | |
99633e97 | 2378 | |
fa9a63c5 RM |
2379 | /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. |
2380 | Returns one of error codes defined in `regex.h', or zero for success. | |
2381 | ||
2382 | Assumes the `allocated' (and perhaps `buffer') and `translate' | |
2383 | fields are set in BUFP on entry. | |
2384 | ||
2385 | If it succeeds, results are put in BUFP (if it returns an error, the | |
2386 | contents of BUFP are undefined): | |
2387 | `buffer' is the compiled pattern; | |
2388 | `syntax' is set to SYNTAX; | |
2389 | `used' is set to the length of the compiled pattern; | |
2390 | `fastmap_accurate' is zero; | |
2391 | `re_nsub' is the number of subexpressions in PATTERN; | |
2392 | `not_bol' and `not_eol' are zero; | |
5e69f11e | 2393 | |
c0f9ea08 | 2394 | The `fastmap' field is neither examined nor set. */ |
fa9a63c5 | 2395 | |
505bde11 SM |
2396 | /* Insert the `jump' from the end of last alternative to "here". |
2397 | The space for the jump has already been allocated. */ | |
2398 | #define FIXUP_ALT_JUMP() \ | |
2399 | do { \ | |
2400 | if (fixup_alt_jump) \ | |
2401 | STORE_JUMP (jump, fixup_alt_jump, b); \ | |
2402 | } while (0) | |
2403 | ||
2404 | ||
fa9a63c5 RM |
2405 | /* Return, freeing storage we allocated. */ |
2406 | #define FREE_STACK_RETURN(value) \ | |
b18215fc RS |
2407 | do { \ |
2408 | FREE_RANGE_TABLE_WORK_AREA (range_table_work); \ | |
2409 | free (compile_stack.stack); \ | |
2410 | return value; \ | |
2411 | } while (0) | |
fa9a63c5 RM |
2412 | |
2413 | static reg_errcode_t | |
29abe551 PE |
2414 | regex_compile (const_re_char *pattern, size_t size, reg_syntax_t syntax, |
2415 | struct re_pattern_buffer *bufp) | |
fa9a63c5 | 2416 | { |
01618498 SM |
2417 | /* We fetch characters from PATTERN here. */ |
2418 | register re_wchar_t c, c1; | |
5e69f11e | 2419 | |
fa9a63c5 RM |
2420 | /* Points to the end of the buffer, where we should append. */ |
2421 | register unsigned char *b; | |
5e69f11e | 2422 | |
fa9a63c5 RM |
2423 | /* Keeps track of unclosed groups. */ |
2424 | compile_stack_type compile_stack; | |
2425 | ||
2426 | /* Points to the current (ending) position in the pattern. */ | |
22336245 RS |
2427 | #ifdef AIX |
2428 | /* `const' makes AIX compiler fail. */ | |
66f0296e | 2429 | unsigned char *p = pattern; |
22336245 | 2430 | #else |
66f0296e | 2431 | re_char *p = pattern; |
22336245 | 2432 | #endif |
66f0296e | 2433 | re_char *pend = pattern + size; |
5e69f11e | 2434 | |
fa9a63c5 | 2435 | /* How to translate the characters in the pattern. */ |
6676cb1c | 2436 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 RM |
2437 | |
2438 | /* Address of the count-byte of the most recently inserted `exactn' | |
2439 | command. This makes it possible to tell if a new exact-match | |
2440 | character can be added to that command or if the character requires | |
2441 | a new `exactn' command. */ | |
2442 | unsigned char *pending_exact = 0; | |
2443 | ||
2444 | /* Address of start of the most recently finished expression. | |
2445 | This tells, e.g., postfix * where to find the start of its | |
2446 | operand. Reset at the beginning of groups and alternatives. */ | |
2447 | unsigned char *laststart = 0; | |
2448 | ||
2449 | /* Address of beginning of regexp, or inside of last group. */ | |
2450 | unsigned char *begalt; | |
2451 | ||
2452 | /* Place in the uncompiled pattern (i.e., the {) to | |
2453 | which to go back if the interval is invalid. */ | |
66f0296e | 2454 | re_char *beg_interval; |
5e69f11e | 2455 | |
fa9a63c5 | 2456 | /* Address of the place where a forward jump should go to the end of |
7814e705 | 2457 | the containing expression. Each alternative of an `or' -- except the |
fa9a63c5 RM |
2458 | last -- ends with a forward jump of this sort. */ |
2459 | unsigned char *fixup_alt_jump = 0; | |
2460 | ||
b18215fc RS |
2461 | /* Work area for range table of charset. */ |
2462 | struct range_table_work_area range_table_work; | |
2463 | ||
2d1675e4 SM |
2464 | /* If the object matched can contain multibyte characters. */ |
2465 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | |
2466 | ||
f9b0fd99 RS |
2467 | /* Nonzero if we have pushed down into a subpattern. */ |
2468 | int in_subpattern = 0; | |
2469 | ||
2470 | /* These hold the values of p, pattern, and pend from the main | |
2471 | pattern when we have pushed into a subpattern. */ | |
da053e48 PE |
2472 | re_char *main_p IF_LINT (= NULL); |
2473 | re_char *main_pattern IF_LINT (= NULL); | |
2474 | re_char *main_pend IF_LINT (= NULL); | |
f9b0fd99 | 2475 | |
fa9a63c5 | 2476 | #ifdef DEBUG |
99633e97 | 2477 | debug++; |
dc4a2ee0 | 2478 | DEBUG_PRINT ("\nCompiling pattern: "); |
99633e97 | 2479 | if (debug > 0) |
fa9a63c5 RM |
2480 | { |
2481 | unsigned debug_count; | |
5e69f11e | 2482 | |
fa9a63c5 | 2483 | for (debug_count = 0; debug_count < size; debug_count++) |
25fe55af | 2484 | putchar (pattern[debug_count]); |
fa9a63c5 RM |
2485 | putchar ('\n'); |
2486 | } | |
2487 | #endif /* DEBUG */ | |
2488 | ||
2489 | /* Initialize the compile stack. */ | |
2490 | compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); | |
2491 | if (compile_stack.stack == NULL) | |
2492 | return REG_ESPACE; | |
2493 | ||
2494 | compile_stack.size = INIT_COMPILE_STACK_SIZE; | |
2495 | compile_stack.avail = 0; | |
2496 | ||
b18215fc RS |
2497 | range_table_work.table = 0; |
2498 | range_table_work.allocated = 0; | |
2499 | ||
fa9a63c5 RM |
2500 | /* Initialize the pattern buffer. */ |
2501 | bufp->syntax = syntax; | |
2502 | bufp->fastmap_accurate = 0; | |
2503 | bufp->not_bol = bufp->not_eol = 0; | |
6224b623 | 2504 | bufp->used_syntax = 0; |
fa9a63c5 RM |
2505 | |
2506 | /* Set `used' to zero, so that if we return an error, the pattern | |
2507 | printer (for debugging) will think there's no pattern. We reset it | |
2508 | at the end. */ | |
2509 | bufp->used = 0; | |
5e69f11e | 2510 | |
fa9a63c5 | 2511 | /* Always count groups, whether or not bufp->no_sub is set. */ |
5e69f11e | 2512 | bufp->re_nsub = 0; |
fa9a63c5 | 2513 | |
0b32bf0e | 2514 | #if !defined emacs && !defined SYNTAX_TABLE |
fa9a63c5 RM |
2515 | /* Initialize the syntax table. */ |
2516 | init_syntax_once (); | |
2517 | #endif | |
2518 | ||
2519 | if (bufp->allocated == 0) | |
2520 | { | |
2521 | if (bufp->buffer) | |
2522 | { /* If zero allocated, but buffer is non-null, try to realloc | |
25fe55af | 2523 | enough space. This loses if buffer's address is bogus, but |
7814e705 | 2524 | that is the user's responsibility. */ |
25fe55af RS |
2525 | RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); |
2526 | } | |
fa9a63c5 | 2527 | else |
7814e705 | 2528 | { /* Caller did not allocate a buffer. Do it for them. */ |
25fe55af RS |
2529 | bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); |
2530 | } | |
fa9a63c5 RM |
2531 | if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE); |
2532 | ||
2533 | bufp->allocated = INIT_BUF_SIZE; | |
2534 | } | |
2535 | ||
2536 | begalt = b = bufp->buffer; | |
2537 | ||
2538 | /* Loop through the uncompiled pattern until we're at the end. */ | |
f9b0fd99 | 2539 | while (1) |
fa9a63c5 | 2540 | { |
f9b0fd99 RS |
2541 | if (p == pend) |
2542 | { | |
2543 | /* If this is the end of an included regexp, | |
2544 | pop back to the main regexp and try again. */ | |
2545 | if (in_subpattern) | |
2546 | { | |
2547 | in_subpattern = 0; | |
2548 | pattern = main_pattern; | |
2549 | p = main_p; | |
2550 | pend = main_pend; | |
2551 | continue; | |
2552 | } | |
2553 | /* If this is the end of the main regexp, we are done. */ | |
2554 | break; | |
2555 | } | |
2556 | ||
fa9a63c5 RM |
2557 | PATFETCH (c); |
2558 | ||
2559 | switch (c) | |
25fe55af | 2560 | { |
f9b0fd99 RS |
2561 | case ' ': |
2562 | { | |
2563 | re_char *p1 = p; | |
2564 | ||
2565 | /* If there's no special whitespace regexp, treat | |
4fb680cd RS |
2566 | spaces normally. And don't try to do this recursively. */ |
2567 | if (!whitespace_regexp || in_subpattern) | |
f9b0fd99 RS |
2568 | goto normal_char; |
2569 | ||
2570 | /* Peek past following spaces. */ | |
2571 | while (p1 != pend) | |
2572 | { | |
2573 | if (*p1 != ' ') | |
2574 | break; | |
2575 | p1++; | |
2576 | } | |
2577 | /* If the spaces are followed by a repetition op, | |
2578 | treat them normally. */ | |
c721eee5 RS |
2579 | if (p1 != pend |
2580 | && (*p1 == '*' || *p1 == '+' || *p1 == '?' | |
f9b0fd99 RS |
2581 | || (*p1 == '\\' && p1 + 1 != pend && p1[1] == '{'))) |
2582 | goto normal_char; | |
2583 | ||
2584 | /* Replace the spaces with the whitespace regexp. */ | |
2585 | in_subpattern = 1; | |
2586 | main_p = p1; | |
2587 | main_pend = pend; | |
2588 | main_pattern = pattern; | |
2589 | p = pattern = whitespace_regexp; | |
5b0534c8 | 2590 | pend = p + strlen ((const char *) p); |
f9b0fd99 | 2591 | break; |
7814e705 | 2592 | } |
f9b0fd99 | 2593 | |
25fe55af RS |
2594 | case '^': |
2595 | { | |
7814e705 | 2596 | if ( /* If at start of pattern, it's an operator. */ |
25fe55af | 2597 | p == pattern + 1 |
7814e705 | 2598 | /* If context independent, it's an operator. */ |
25fe55af | 2599 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
7814e705 | 2600 | /* Otherwise, depends on what's come before. */ |
25fe55af | 2601 | || at_begline_loc_p (pattern, p, syntax)) |
c0f9ea08 | 2602 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? begbuf : begline); |
25fe55af RS |
2603 | else |
2604 | goto normal_char; | |
2605 | } | |
2606 | break; | |
2607 | ||
2608 | ||
2609 | case '$': | |
2610 | { | |
2611 | if ( /* If at end of pattern, it's an operator. */ | |
2612 | p == pend | |
7814e705 | 2613 | /* If context independent, it's an operator. */ |
25fe55af RS |
2614 | || syntax & RE_CONTEXT_INDEP_ANCHORS |
2615 | /* Otherwise, depends on what's next. */ | |
2616 | || at_endline_loc_p (p, pend, syntax)) | |
c0f9ea08 | 2617 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? endbuf : endline); |
25fe55af RS |
2618 | else |
2619 | goto normal_char; | |
2620 | } | |
2621 | break; | |
fa9a63c5 RM |
2622 | |
2623 | ||
2624 | case '+': | |
25fe55af RS |
2625 | case '?': |
2626 | if ((syntax & RE_BK_PLUS_QM) | |
2627 | || (syntax & RE_LIMITED_OPS)) | |
2628 | goto normal_char; | |
2629 | handle_plus: | |
2630 | case '*': | |
5ac2eb34 | 2631 | /* If there is no previous pattern... */ |
25fe55af RS |
2632 | if (!laststart) |
2633 | { | |
2634 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
2635 | FREE_STACK_RETURN (REG_BADRPT); | |
2636 | else if (!(syntax & RE_CONTEXT_INDEP_OPS)) | |
2637 | goto normal_char; | |
2638 | } | |
2639 | ||
2640 | { | |
7814e705 | 2641 | /* 1 means zero (many) matches is allowed. */ |
66f0296e SM |
2642 | boolean zero_times_ok = 0, many_times_ok = 0; |
2643 | boolean greedy = 1; | |
25fe55af RS |
2644 | |
2645 | /* If there is a sequence of repetition chars, collapse it | |
2646 | down to just one (the right one). We can't combine | |
2647 | interval operators with these because of, e.g., `a{2}*', | |
7814e705 | 2648 | which should only match an even number of `a's. */ |
25fe55af RS |
2649 | |
2650 | for (;;) | |
2651 | { | |
0b32bf0e | 2652 | if ((syntax & RE_FRUGAL) |
1c8c6d39 DL |
2653 | && c == '?' && (zero_times_ok || many_times_ok)) |
2654 | greedy = 0; | |
2655 | else | |
2656 | { | |
2657 | zero_times_ok |= c != '+'; | |
2658 | many_times_ok |= c != '?'; | |
2659 | } | |
25fe55af RS |
2660 | |
2661 | if (p == pend) | |
2662 | break; | |
ed0767d8 SM |
2663 | else if (*p == '*' |
2664 | || (!(syntax & RE_BK_PLUS_QM) | |
2665 | && (*p == '+' || *p == '?'))) | |
25fe55af | 2666 | ; |
ed0767d8 | 2667 | else if (syntax & RE_BK_PLUS_QM && *p == '\\') |
25fe55af | 2668 | { |
ed0767d8 SM |
2669 | if (p+1 == pend) |
2670 | FREE_STACK_RETURN (REG_EESCAPE); | |
2671 | if (p[1] == '+' || p[1] == '?') | |
2672 | PATFETCH (c); /* Gobble up the backslash. */ | |
2673 | else | |
2674 | break; | |
25fe55af RS |
2675 | } |
2676 | else | |
ed0767d8 | 2677 | break; |
25fe55af | 2678 | /* If we get here, we found another repeat character. */ |
ed0767d8 SM |
2679 | PATFETCH (c); |
2680 | } | |
25fe55af RS |
2681 | |
2682 | /* Star, etc. applied to an empty pattern is equivalent | |
2683 | to an empty pattern. */ | |
4e8a9132 | 2684 | if (!laststart || laststart == b) |
25fe55af RS |
2685 | break; |
2686 | ||
2687 | /* Now we know whether or not zero matches is allowed | |
7814e705 | 2688 | and also whether or not two or more matches is allowed. */ |
1c8c6d39 DL |
2689 | if (greedy) |
2690 | { | |
99633e97 | 2691 | if (many_times_ok) |
4e8a9132 SM |
2692 | { |
2693 | boolean simple = skip_one_char (laststart) == b; | |
d1dfb56c | 2694 | size_t startoffset = 0; |
f6a3f532 | 2695 | re_opcode_t ofj = |
01618498 | 2696 | /* Check if the loop can match the empty string. */ |
6df42991 SM |
2697 | (simple || !analyse_first (laststart, b, NULL, 0)) |
2698 | ? on_failure_jump : on_failure_jump_loop; | |
4e8a9132 | 2699 | assert (skip_one_char (laststart) <= b); |
177c0ea7 | 2700 | |
4e8a9132 SM |
2701 | if (!zero_times_ok && simple) |
2702 | { /* Since simple * loops can be made faster by using | |
2703 | on_failure_keep_string_jump, we turn simple P+ | |
2704 | into PP* if P is simple. */ | |
2705 | unsigned char *p1, *p2; | |
2706 | startoffset = b - laststart; | |
2707 | GET_BUFFER_SPACE (startoffset); | |
2708 | p1 = b; p2 = laststart; | |
2709 | while (p2 < p1) | |
2710 | *b++ = *p2++; | |
2711 | zero_times_ok = 1; | |
99633e97 | 2712 | } |
4e8a9132 SM |
2713 | |
2714 | GET_BUFFER_SPACE (6); | |
2715 | if (!zero_times_ok) | |
2716 | /* A + loop. */ | |
f6a3f532 | 2717 | STORE_JUMP (ofj, b, b + 6); |
99633e97 | 2718 | else |
4e8a9132 SM |
2719 | /* Simple * loops can use on_failure_keep_string_jump |
2720 | depending on what follows. But since we don't know | |
2721 | that yet, we leave the decision up to | |
2722 | on_failure_jump_smart. */ | |
f6a3f532 | 2723 | INSERT_JUMP (simple ? on_failure_jump_smart : ofj, |
4e8a9132 | 2724 | laststart + startoffset, b + 6); |
99633e97 | 2725 | b += 3; |
4e8a9132 | 2726 | STORE_JUMP (jump, b, laststart + startoffset); |
99633e97 SM |
2727 | b += 3; |
2728 | } | |
2729 | else | |
2730 | { | |
4e8a9132 SM |
2731 | /* A simple ? pattern. */ |
2732 | assert (zero_times_ok); | |
2733 | GET_BUFFER_SPACE (3); | |
2734 | INSERT_JUMP (on_failure_jump, laststart, b + 3); | |
99633e97 SM |
2735 | b += 3; |
2736 | } | |
1c8c6d39 DL |
2737 | } |
2738 | else /* not greedy */ | |
5ac2eb34 | 2739 | { /* I wish the greedy and non-greedy cases could be merged. */ |
1c8c6d39 | 2740 | |
0683b6fa | 2741 | GET_BUFFER_SPACE (7); /* We might use less. */ |
1c8c6d39 DL |
2742 | if (many_times_ok) |
2743 | { | |
f6a3f532 SM |
2744 | boolean emptyp = analyse_first (laststart, b, NULL, 0); |
2745 | ||
6df42991 SM |
2746 | /* The non-greedy multiple match looks like |
2747 | a repeat..until: we only need a conditional jump | |
2748 | at the end of the loop. */ | |
f6a3f532 SM |
2749 | if (emptyp) BUF_PUSH (no_op); |
2750 | STORE_JUMP (emptyp ? on_failure_jump_nastyloop | |
2751 | : on_failure_jump, b, laststart); | |
1c8c6d39 DL |
2752 | b += 3; |
2753 | if (zero_times_ok) | |
2754 | { | |
2755 | /* The repeat...until naturally matches one or more. | |
2756 | To also match zero times, we need to first jump to | |
6df42991 | 2757 | the end of the loop (its conditional jump). */ |
1c8c6d39 DL |
2758 | INSERT_JUMP (jump, laststart, b); |
2759 | b += 3; | |
2760 | } | |
2761 | } | |
2762 | else | |
2763 | { | |
2764 | /* non-greedy a?? */ | |
1c8c6d39 DL |
2765 | INSERT_JUMP (jump, laststart, b + 3); |
2766 | b += 3; | |
2767 | INSERT_JUMP (on_failure_jump, laststart, laststart + 6); | |
2768 | b += 3; | |
2769 | } | |
2770 | } | |
2771 | } | |
4e8a9132 | 2772 | pending_exact = 0; |
fa9a63c5 RM |
2773 | break; |
2774 | ||
2775 | ||
2776 | case '.': | |
25fe55af RS |
2777 | laststart = b; |
2778 | BUF_PUSH (anychar); | |
2779 | break; | |
fa9a63c5 RM |
2780 | |
2781 | ||
25fe55af RS |
2782 | case '[': |
2783 | { | |
19ed5445 PE |
2784 | re_char *p1; |
2785 | ||
b18215fc | 2786 | CLEAR_RANGE_TABLE_WORK_USED (range_table_work); |
fa9a63c5 | 2787 | |
25fe55af | 2788 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2789 | |
25fe55af RS |
2790 | /* Ensure that we have enough space to push a charset: the |
2791 | opcode, the length count, and the bitset; 34 bytes in all. */ | |
fa9a63c5 RM |
2792 | GET_BUFFER_SPACE (34); |
2793 | ||
25fe55af | 2794 | laststart = b; |
e318085a | 2795 | |
25fe55af | 2796 | /* We test `*p == '^' twice, instead of using an if |
7814e705 | 2797 | statement, so we only need one BUF_PUSH. */ |
25fe55af RS |
2798 | BUF_PUSH (*p == '^' ? charset_not : charset); |
2799 | if (*p == '^') | |
2800 | p++; | |
e318085a | 2801 | |
25fe55af RS |
2802 | /* Remember the first position in the bracket expression. */ |
2803 | p1 = p; | |
e318085a | 2804 | |
7814e705 | 2805 | /* Push the number of bytes in the bitmap. */ |
25fe55af | 2806 | BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2807 | |
25fe55af | 2808 | /* Clear the whole map. */ |
72af86bd | 2809 | memset (b, 0, (1 << BYTEWIDTH) / BYTEWIDTH); |
e318085a | 2810 | |
25fe55af RS |
2811 | /* charset_not matches newline according to a syntax bit. */ |
2812 | if ((re_opcode_t) b[-2] == charset_not | |
2813 | && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) | |
2814 | SET_LIST_BIT ('\n'); | |
fa9a63c5 | 2815 | |
7814e705 | 2816 | /* Read in characters and ranges, setting map bits. */ |
25fe55af RS |
2817 | for (;;) |
2818 | { | |
b18215fc | 2819 | boolean escaped_char = false; |
2d1675e4 | 2820 | const unsigned char *p2 = p; |
abbd1bcf | 2821 | re_wchar_t ch; |
e318085a | 2822 | |
25fe55af | 2823 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
e318085a | 2824 | |
36595814 SM |
2825 | /* Don't translate yet. The range TRANSLATE(X..Y) cannot |
2826 | always be determined from TRANSLATE(X) and TRANSLATE(Y) | |
2827 | So the translation is done later in a loop. Example: | |
2828 | (let ((case-fold-search t)) (string-match "[A-_]" "A")) */ | |
25fe55af | 2829 | PATFETCH (c); |
e318085a | 2830 | |
25fe55af RS |
2831 | /* \ might escape characters inside [...] and [^...]. */ |
2832 | if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') | |
2833 | { | |
2834 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
e318085a RS |
2835 | |
2836 | PATFETCH (c); | |
b18215fc | 2837 | escaped_char = true; |
25fe55af | 2838 | } |
b18215fc RS |
2839 | else |
2840 | { | |
7814e705 | 2841 | /* Could be the end of the bracket expression. If it's |
657fcfbd RS |
2842 | not (i.e., when the bracket expression is `[]' so |
2843 | far), the ']' character bit gets set way below. */ | |
2d1675e4 | 2844 | if (c == ']' && p2 != p1) |
657fcfbd | 2845 | break; |
25fe55af | 2846 | } |
b18215fc | 2847 | |
25fe55af RS |
2848 | /* See if we're at the beginning of a possible character |
2849 | class. */ | |
b18215fc | 2850 | |
2d1675e4 SM |
2851 | if (!escaped_char && |
2852 | syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') | |
657fcfbd | 2853 | { |
7814e705 | 2854 | /* Leave room for the null. */ |
14473664 | 2855 | unsigned char str[CHAR_CLASS_MAX_LENGTH + 1]; |
ed0767d8 | 2856 | const unsigned char *class_beg; |
b18215fc | 2857 | |
25fe55af RS |
2858 | PATFETCH (c); |
2859 | c1 = 0; | |
ed0767d8 | 2860 | class_beg = p; |
b18215fc | 2861 | |
25fe55af RS |
2862 | /* If pattern is `[[:'. */ |
2863 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | |
b18215fc | 2864 | |
25fe55af RS |
2865 | for (;;) |
2866 | { | |
14473664 SM |
2867 | PATFETCH (c); |
2868 | if ((c == ':' && *p == ']') || p == pend) | |
2869 | break; | |
2870 | if (c1 < CHAR_CLASS_MAX_LENGTH) | |
2871 | str[c1++] = c; | |
2872 | else | |
2873 | /* This is in any case an invalid class name. */ | |
2874 | str[0] = '\0'; | |
25fe55af RS |
2875 | } |
2876 | str[c1] = '\0'; | |
b18215fc RS |
2877 | |
2878 | /* If isn't a word bracketed by `[:' and `:]': | |
2879 | undo the ending character, the letters, and | |
2880 | leave the leading `:' and `[' (but set bits for | |
2881 | them). */ | |
25fe55af RS |
2882 | if (c == ':' && *p == ']') |
2883 | { | |
abbd1bcf | 2884 | re_wctype_t cc = re_wctype (str); |
14473664 SM |
2885 | |
2886 | if (cc == 0) | |
fa9a63c5 RM |
2887 | FREE_STACK_RETURN (REG_ECTYPE); |
2888 | ||
14473664 SM |
2889 | /* Throw away the ] at the end of the character |
2890 | class. */ | |
2891 | PATFETCH (c); | |
fa9a63c5 | 2892 | |
14473664 | 2893 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); |
fa9a63c5 | 2894 | |
cf9c99bc KH |
2895 | #ifndef emacs |
2896 | for (ch = 0; ch < (1 << BYTEWIDTH); ++ch) | |
8f924df7 KH |
2897 | if (re_iswctype (btowc (ch), cc)) |
2898 | { | |
2899 | c = TRANSLATE (ch); | |
ed00c2ac KH |
2900 | if (c < (1 << BYTEWIDTH)) |
2901 | SET_LIST_BIT (c); | |
8f924df7 | 2902 | } |
cf9c99bc KH |
2903 | #else /* emacs */ |
2904 | /* Most character classes in a multibyte match | |
2905 | just set a flag. Exceptions are is_blank, | |
2906 | is_digit, is_cntrl, and is_xdigit, since | |
2907 | they can only match ASCII characters. We | |
2908 | don't need to handle them for multibyte. | |
2909 | They are distinguished by a negative wctype. */ | |
96cc36cc | 2910 | |
254c06a8 SM |
2911 | /* Setup the gl_state object to its buffer-defined |
2912 | value. This hardcodes the buffer-global | |
2913 | syntax-table for ASCII chars, while the other chars | |
2914 | will obey syntax-table properties. It's not ideal, | |
2915 | but it's the way it's been done until now. */ | |
d48cd3f4 | 2916 | SETUP_BUFFER_SYNTAX_TABLE (); |
254c06a8 | 2917 | |
cf9c99bc | 2918 | for (ch = 0; ch < 256; ++ch) |
25fe55af | 2919 | { |
cf9c99bc KH |
2920 | c = RE_CHAR_TO_MULTIBYTE (ch); |
2921 | if (! CHAR_BYTE8_P (c) | |
2922 | && re_iswctype (c, cc)) | |
8f924df7 | 2923 | { |
cf9c99bc KH |
2924 | SET_LIST_BIT (ch); |
2925 | c1 = TRANSLATE (c); | |
2926 | if (c1 == c) | |
2927 | continue; | |
2928 | if (ASCII_CHAR_P (c1)) | |
2929 | SET_LIST_BIT (c1); | |
2930 | else if ((c1 = RE_CHAR_TO_UNIBYTE (c1)) >= 0) | |
2931 | SET_LIST_BIT (c1); | |
8f924df7 | 2932 | } |
25fe55af | 2933 | } |
cf9c99bc KH |
2934 | SET_RANGE_TABLE_WORK_AREA_BIT |
2935 | (range_table_work, re_wctype_to_bit (cc)); | |
2936 | #endif /* emacs */ | |
6224b623 SM |
2937 | /* In most cases the matching rule for char classes |
2938 | only uses the syntax table for multibyte chars, | |
2939 | so that the content of the syntax-table it is not | |
2940 | hardcoded in the range_table. SPACE and WORD are | |
2941 | the two exceptions. */ | |
2942 | if ((1 << cc) & ((1 << RECC_SPACE) | (1 << RECC_WORD))) | |
2943 | bufp->used_syntax = 1; | |
2944 | ||
b18215fc RS |
2945 | /* Repeat the loop. */ |
2946 | continue; | |
25fe55af RS |
2947 | } |
2948 | else | |
2949 | { | |
ed0767d8 SM |
2950 | /* Go back to right after the "[:". */ |
2951 | p = class_beg; | |
25fe55af | 2952 | SET_LIST_BIT ('['); |
b18215fc RS |
2953 | |
2954 | /* Because the `:' may starts the range, we | |
2955 | can't simply set bit and repeat the loop. | |
7814e705 | 2956 | Instead, just set it to C and handle below. */ |
b18215fc | 2957 | c = ':'; |
25fe55af RS |
2958 | } |
2959 | } | |
b18215fc RS |
2960 | |
2961 | if (p < pend && p[0] == '-' && p[1] != ']') | |
2962 | { | |
2963 | ||
2964 | /* Discard the `-'. */ | |
2965 | PATFETCH (c1); | |
2966 | ||
2967 | /* Fetch the character which ends the range. */ | |
2968 | PATFETCH (c1); | |
cf9c99bc KH |
2969 | #ifdef emacs |
2970 | if (CHAR_BYTE8_P (c1) | |
2971 | && ! ASCII_CHAR_P (c) && ! CHAR_BYTE8_P (c)) | |
2972 | /* Treat the range from a multibyte character to | |
2973 | raw-byte character as empty. */ | |
2974 | c = c1 + 1; | |
2975 | #endif /* emacs */ | |
e318085a | 2976 | } |
25fe55af | 2977 | else |
b18215fc RS |
2978 | /* Range from C to C. */ |
2979 | c1 = c; | |
2980 | ||
cf9c99bc | 2981 | if (c > c1) |
25fe55af | 2982 | { |
cf9c99bc KH |
2983 | if (syntax & RE_NO_EMPTY_RANGES) |
2984 | FREE_STACK_RETURN (REG_ERANGEX); | |
2985 | /* Else, repeat the loop. */ | |
bf216479 | 2986 | } |
6fdd04b0 | 2987 | else |
25fe55af | 2988 | { |
cf9c99bc KH |
2989 | #ifndef emacs |
2990 | /* Set the range into bitmap */ | |
8f924df7 | 2991 | for (; c <= c1; c++) |
b18215fc | 2992 | { |
cf9c99bc KH |
2993 | ch = TRANSLATE (c); |
2994 | if (ch < (1 << BYTEWIDTH)) | |
2995 | SET_LIST_BIT (ch); | |
2996 | } | |
2997 | #else /* emacs */ | |
2998 | if (c < 128) | |
2999 | { | |
3000 | ch = MIN (127, c1); | |
3001 | SETUP_ASCII_RANGE (range_table_work, c, ch); | |
3002 | c = ch + 1; | |
3003 | if (CHAR_BYTE8_P (c1)) | |
3004 | c = BYTE8_TO_CHAR (128); | |
3005 | } | |
3006 | if (c <= c1) | |
3007 | { | |
3008 | if (CHAR_BYTE8_P (c)) | |
3009 | { | |
3010 | c = CHAR_TO_BYTE8 (c); | |
3011 | c1 = CHAR_TO_BYTE8 (c1); | |
3012 | for (; c <= c1; c++) | |
3013 | SET_LIST_BIT (c); | |
3014 | } | |
3015 | else if (multibyte) | |
3016 | { | |
3017 | SETUP_MULTIBYTE_RANGE (range_table_work, c, c1); | |
3018 | } | |
3019 | else | |
3020 | { | |
3021 | SETUP_UNIBYTE_RANGE (range_table_work, c, c1); | |
3022 | } | |
e934739e | 3023 | } |
cf9c99bc | 3024 | #endif /* emacs */ |
25fe55af | 3025 | } |
e318085a RS |
3026 | } |
3027 | ||
25fe55af | 3028 | /* Discard any (non)matching list bytes that are all 0 at the |
7814e705 | 3029 | end of the map. Decrease the map-length byte too. */ |
25fe55af RS |
3030 | while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) |
3031 | b[-1]--; | |
3032 | b += b[-1]; | |
fa9a63c5 | 3033 | |
96cc36cc RS |
3034 | /* Build real range table from work area. */ |
3035 | if (RANGE_TABLE_WORK_USED (range_table_work) | |
3036 | || RANGE_TABLE_WORK_BITS (range_table_work)) | |
b18215fc RS |
3037 | { |
3038 | int i; | |
3039 | int used = RANGE_TABLE_WORK_USED (range_table_work); | |
fa9a63c5 | 3040 | |
b18215fc | 3041 | /* Allocate space for COUNT + RANGE_TABLE. Needs two |
96cc36cc | 3042 | bytes for flags, two for COUNT, and three bytes for |
5ac2eb34 | 3043 | each character. */ |
96cc36cc | 3044 | GET_BUFFER_SPACE (4 + used * 3); |
fa9a63c5 | 3045 | |
b18215fc RS |
3046 | /* Indicate the existence of range table. */ |
3047 | laststart[1] |= 0x80; | |
fa9a63c5 | 3048 | |
96cc36cc RS |
3049 | /* Store the character class flag bits into the range table. |
3050 | If not in emacs, these flag bits are always 0. */ | |
3051 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) & 0xff; | |
3052 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) >> 8; | |
3053 | ||
b18215fc RS |
3054 | STORE_NUMBER_AND_INCR (b, used / 2); |
3055 | for (i = 0; i < used; i++) | |
3056 | STORE_CHARACTER_AND_INCR | |
3057 | (b, RANGE_TABLE_WORK_ELT (range_table_work, i)); | |
3058 | } | |
25fe55af RS |
3059 | } |
3060 | break; | |
fa9a63c5 RM |
3061 | |
3062 | ||
b18215fc | 3063 | case '(': |
25fe55af RS |
3064 | if (syntax & RE_NO_BK_PARENS) |
3065 | goto handle_open; | |
3066 | else | |
3067 | goto normal_char; | |
fa9a63c5 RM |
3068 | |
3069 | ||
25fe55af RS |
3070 | case ')': |
3071 | if (syntax & RE_NO_BK_PARENS) | |
3072 | goto handle_close; | |
3073 | else | |
3074 | goto normal_char; | |
e318085a RS |
3075 | |
3076 | ||
25fe55af RS |
3077 | case '\n': |
3078 | if (syntax & RE_NEWLINE_ALT) | |
3079 | goto handle_alt; | |
3080 | else | |
3081 | goto normal_char; | |
e318085a RS |
3082 | |
3083 | ||
b18215fc | 3084 | case '|': |
25fe55af RS |
3085 | if (syntax & RE_NO_BK_VBAR) |
3086 | goto handle_alt; | |
3087 | else | |
3088 | goto normal_char; | |
3089 | ||
3090 | ||
3091 | case '{': | |
3092 | if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) | |
3093 | goto handle_interval; | |
3094 | else | |
3095 | goto normal_char; | |
3096 | ||
3097 | ||
3098 | case '\\': | |
3099 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | |
3100 | ||
3101 | /* Do not translate the character after the \, so that we can | |
3102 | distinguish, e.g., \B from \b, even if we normally would | |
3103 | translate, e.g., B to b. */ | |
36595814 | 3104 | PATFETCH (c); |
25fe55af RS |
3105 | |
3106 | switch (c) | |
3107 | { | |
3108 | case '(': | |
3109 | if (syntax & RE_NO_BK_PARENS) | |
3110 | goto normal_backslash; | |
3111 | ||
3112 | handle_open: | |
505bde11 SM |
3113 | { |
3114 | int shy = 0; | |
c69b0314 | 3115 | regnum_t regnum = 0; |
505bde11 SM |
3116 | if (p+1 < pend) |
3117 | { | |
3118 | /* Look for a special (?...) construct */ | |
ed0767d8 | 3119 | if ((syntax & RE_SHY_GROUPS) && *p == '?') |
505bde11 | 3120 | { |
ed0767d8 | 3121 | PATFETCH (c); /* Gobble up the '?'. */ |
c69b0314 | 3122 | while (!shy) |
505bde11 | 3123 | { |
c69b0314 SM |
3124 | PATFETCH (c); |
3125 | switch (c) | |
3126 | { | |
3127 | case ':': shy = 1; break; | |
3128 | case '0': | |
3129 | /* An explicitly specified regnum must start | |
3130 | with non-0. */ | |
3131 | if (regnum == 0) | |
3132 | FREE_STACK_RETURN (REG_BADPAT); | |
3133 | case '1': case '2': case '3': case '4': | |
3134 | case '5': case '6': case '7': case '8': case '9': | |
3135 | regnum = 10*regnum + (c - '0'); break; | |
3136 | default: | |
3137 | /* Only (?:...) is supported right now. */ | |
3138 | FREE_STACK_RETURN (REG_BADPAT); | |
3139 | } | |
505bde11 SM |
3140 | } |
3141 | } | |
505bde11 SM |
3142 | } |
3143 | ||
3144 | if (!shy) | |
c69b0314 SM |
3145 | regnum = ++bufp->re_nsub; |
3146 | else if (regnum) | |
3147 | { /* It's actually not shy, but explicitly numbered. */ | |
3148 | shy = 0; | |
3149 | if (regnum > bufp->re_nsub) | |
3150 | bufp->re_nsub = regnum; | |
3151 | else if (regnum > bufp->re_nsub | |
3152 | /* Ideally, we'd want to check that the specified | |
3153 | group can't have matched (i.e. all subgroups | |
3154 | using the same regnum are in other branches of | |
3155 | OR patterns), but we don't currently keep track | |
3156 | of enough info to do that easily. */ | |
3157 | || group_in_compile_stack (compile_stack, regnum)) | |
3158 | FREE_STACK_RETURN (REG_BADPAT); | |
505bde11 | 3159 | } |
c69b0314 SM |
3160 | else |
3161 | /* It's really shy. */ | |
3162 | regnum = - bufp->re_nsub; | |
25fe55af | 3163 | |
99633e97 SM |
3164 | if (COMPILE_STACK_FULL) |
3165 | { | |
3166 | RETALLOC (compile_stack.stack, compile_stack.size << 1, | |
3167 | compile_stack_elt_t); | |
3168 | if (compile_stack.stack == NULL) return REG_ESPACE; | |
25fe55af | 3169 | |
99633e97 SM |
3170 | compile_stack.size <<= 1; |
3171 | } | |
25fe55af | 3172 | |
99633e97 | 3173 | /* These are the values to restore when we hit end of this |
7814e705 | 3174 | group. They are all relative offsets, so that if the |
99633e97 SM |
3175 | whole pattern moves because of realloc, they will still |
3176 | be valid. */ | |
3177 | COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; | |
3178 | COMPILE_STACK_TOP.fixup_alt_jump | |
3179 | = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; | |
3180 | COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer; | |
c69b0314 | 3181 | COMPILE_STACK_TOP.regnum = regnum; |
99633e97 | 3182 | |
c69b0314 SM |
3183 | /* Do not push a start_memory for groups beyond the last one |
3184 | we can represent in the compiled pattern. */ | |
3185 | if (regnum <= MAX_REGNUM && regnum > 0) | |
99633e97 SM |
3186 | BUF_PUSH_2 (start_memory, regnum); |
3187 | ||
3188 | compile_stack.avail++; | |
3189 | ||
3190 | fixup_alt_jump = 0; | |
3191 | laststart = 0; | |
3192 | begalt = b; | |
3193 | /* If we've reached MAX_REGNUM groups, then this open | |
3194 | won't actually generate any code, so we'll have to | |
3195 | clear pending_exact explicitly. */ | |
3196 | pending_exact = 0; | |
3197 | break; | |
505bde11 | 3198 | } |
25fe55af RS |
3199 | |
3200 | case ')': | |
3201 | if (syntax & RE_NO_BK_PARENS) goto normal_backslash; | |
3202 | ||
3203 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3204 | { |
3205 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3206 | goto normal_backslash; | |
3207 | else | |
3208 | FREE_STACK_RETURN (REG_ERPAREN); | |
3209 | } | |
25fe55af RS |
3210 | |
3211 | handle_close: | |
505bde11 | 3212 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3213 | |
3214 | /* See similar code for backslashed left paren above. */ | |
3215 | if (COMPILE_STACK_EMPTY) | |
505bde11 SM |
3216 | { |
3217 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | |
3218 | goto normal_char; | |
3219 | else | |
3220 | FREE_STACK_RETURN (REG_ERPAREN); | |
3221 | } | |
25fe55af RS |
3222 | |
3223 | /* Since we just checked for an empty stack above, this | |
3224 | ``can't happen''. */ | |
3225 | assert (compile_stack.avail != 0); | |
3226 | { | |
3227 | /* We don't just want to restore into `regnum', because | |
3228 | later groups should continue to be numbered higher, | |
7814e705 | 3229 | as in `(ab)c(de)' -- the second group is #2. */ |
c69b0314 | 3230 | regnum_t regnum; |
25fe55af RS |
3231 | |
3232 | compile_stack.avail--; | |
3233 | begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; | |
3234 | fixup_alt_jump | |
3235 | = COMPILE_STACK_TOP.fixup_alt_jump | |
3236 | ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 | |
3237 | : 0; | |
3238 | laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; | |
c69b0314 | 3239 | regnum = COMPILE_STACK_TOP.regnum; |
b18215fc RS |
3240 | /* If we've reached MAX_REGNUM groups, then this open |
3241 | won't actually generate any code, so we'll have to | |
3242 | clear pending_exact explicitly. */ | |
3243 | pending_exact = 0; | |
e318085a | 3244 | |
25fe55af | 3245 | /* We're at the end of the group, so now we know how many |
7814e705 | 3246 | groups were inside this one. */ |
c69b0314 SM |
3247 | if (regnum <= MAX_REGNUM && regnum > 0) |
3248 | BUF_PUSH_2 (stop_memory, regnum); | |
25fe55af RS |
3249 | } |
3250 | break; | |
3251 | ||
3252 | ||
3253 | case '|': /* `\|'. */ | |
3254 | if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) | |
3255 | goto normal_backslash; | |
3256 | handle_alt: | |
3257 | if (syntax & RE_LIMITED_OPS) | |
3258 | goto normal_char; | |
3259 | ||
3260 | /* Insert before the previous alternative a jump which | |
7814e705 | 3261 | jumps to this alternative if the former fails. */ |
25fe55af RS |
3262 | GET_BUFFER_SPACE (3); |
3263 | INSERT_JUMP (on_failure_jump, begalt, b + 6); | |
3264 | pending_exact = 0; | |
3265 | b += 3; | |
3266 | ||
3267 | /* The alternative before this one has a jump after it | |
3268 | which gets executed if it gets matched. Adjust that | |
3269 | jump so it will jump to this alternative's analogous | |
3270 | jump (put in below, which in turn will jump to the next | |
3271 | (if any) alternative's such jump, etc.). The last such | |
3272 | jump jumps to the correct final destination. A picture: | |
3273 | _____ _____ | |
3274 | | | | | | |
3275 | | v | v | |
d1dfb56c | 3276 | a | b | c |
25fe55af RS |
3277 | |
3278 | If we are at `b', then fixup_alt_jump right now points to a | |
3279 | three-byte space after `a'. We'll put in the jump, set | |
3280 | fixup_alt_jump to right after `b', and leave behind three | |
3281 | bytes which we'll fill in when we get to after `c'. */ | |
3282 | ||
505bde11 | 3283 | FIXUP_ALT_JUMP (); |
25fe55af RS |
3284 | |
3285 | /* Mark and leave space for a jump after this alternative, | |
3286 | to be filled in later either by next alternative or | |
3287 | when know we're at the end of a series of alternatives. */ | |
3288 | fixup_alt_jump = b; | |
3289 | GET_BUFFER_SPACE (3); | |
3290 | b += 3; | |
3291 | ||
3292 | laststart = 0; | |
3293 | begalt = b; | |
3294 | break; | |
3295 | ||
3296 | ||
3297 | case '{': | |
3298 | /* If \{ is a literal. */ | |
3299 | if (!(syntax & RE_INTERVALS) | |
3300 | /* If we're at `\{' and it's not the open-interval | |
3301 | operator. */ | |
4bb91c68 | 3302 | || (syntax & RE_NO_BK_BRACES)) |
25fe55af RS |
3303 | goto normal_backslash; |
3304 | ||
3305 | handle_interval: | |
3306 | { | |
3307 | /* If got here, then the syntax allows intervals. */ | |
3308 | ||
3309 | /* At least (most) this many matches must be made. */ | |
99633e97 | 3310 | int lower_bound = 0, upper_bound = -1; |
25fe55af | 3311 | |
ed0767d8 | 3312 | beg_interval = p; |
25fe55af | 3313 | |
25fe55af RS |
3314 | GET_UNSIGNED_NUMBER (lower_bound); |
3315 | ||
3316 | if (c == ',') | |
ed0767d8 | 3317 | GET_UNSIGNED_NUMBER (upper_bound); |
25fe55af RS |
3318 | else |
3319 | /* Interval such as `{1}' => match exactly once. */ | |
3320 | upper_bound = lower_bound; | |
3321 | ||
3322 | if (lower_bound < 0 || upper_bound > RE_DUP_MAX | |
ed0767d8 | 3323 | || (upper_bound >= 0 && lower_bound > upper_bound)) |
4bb91c68 | 3324 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3325 | |
3326 | if (!(syntax & RE_NO_BK_BRACES)) | |
3327 | { | |
4bb91c68 SM |
3328 | if (c != '\\') |
3329 | FREE_STACK_RETURN (REG_BADBR); | |
c72b0edd SM |
3330 | if (p == pend) |
3331 | FREE_STACK_RETURN (REG_EESCAPE); | |
25fe55af RS |
3332 | PATFETCH (c); |
3333 | } | |
3334 | ||
3335 | if (c != '}') | |
4bb91c68 | 3336 | FREE_STACK_RETURN (REG_BADBR); |
25fe55af RS |
3337 | |
3338 | /* We just parsed a valid interval. */ | |
3339 | ||
3340 | /* If it's invalid to have no preceding re. */ | |
3341 | if (!laststart) | |
3342 | { | |
3343 | if (syntax & RE_CONTEXT_INVALID_OPS) | |
3344 | FREE_STACK_RETURN (REG_BADRPT); | |
3345 | else if (syntax & RE_CONTEXT_INDEP_OPS) | |
3346 | laststart = b; | |
3347 | else | |
3348 | goto unfetch_interval; | |
3349 | } | |
3350 | ||
6df42991 SM |
3351 | if (upper_bound == 0) |
3352 | /* If the upper bound is zero, just drop the sub pattern | |
3353 | altogether. */ | |
3354 | b = laststart; | |
3355 | else if (lower_bound == 1 && upper_bound == 1) | |
3356 | /* Just match it once: nothing to do here. */ | |
3357 | ; | |
3358 | ||
3359 | /* Otherwise, we have a nontrivial interval. When | |
3360 | we're all done, the pattern will look like: | |
3361 | set_number_at <jump count> <upper bound> | |
3362 | set_number_at <succeed_n count> <lower bound> | |
3363 | succeed_n <after jump addr> <succeed_n count> | |
3364 | <body of loop> | |
3365 | jump_n <succeed_n addr> <jump count> | |
3366 | (The upper bound and `jump_n' are omitted if | |
3367 | `upper_bound' is 1, though.) */ | |
3368 | else | |
3369 | { /* If the upper bound is > 1, we need to insert | |
3370 | more at the end of the loop. */ | |
3371 | unsigned int nbytes = (upper_bound < 0 ? 3 | |
3372 | : upper_bound > 1 ? 5 : 0); | |
3373 | unsigned int startoffset = 0; | |
3374 | ||
3375 | GET_BUFFER_SPACE (20); /* We might use less. */ | |
3376 | ||
3377 | if (lower_bound == 0) | |
3378 | { | |
3379 | /* A succeed_n that starts with 0 is really a | |
3380 | a simple on_failure_jump_loop. */ | |
3381 | INSERT_JUMP (on_failure_jump_loop, laststart, | |
3382 | b + 3 + nbytes); | |
3383 | b += 3; | |
3384 | } | |
3385 | else | |
3386 | { | |
3387 | /* Initialize lower bound of the `succeed_n', even | |
3388 | though it will be set during matching by its | |
3389 | attendant `set_number_at' (inserted next), | |
3390 | because `re_compile_fastmap' needs to know. | |
3391 | Jump to the `jump_n' we might insert below. */ | |
3392 | INSERT_JUMP2 (succeed_n, laststart, | |
3393 | b + 5 + nbytes, | |
3394 | lower_bound); | |
3395 | b += 5; | |
3396 | ||
3397 | /* Code to initialize the lower bound. Insert | |
7814e705 | 3398 | before the `succeed_n'. The `5' is the last two |
6df42991 SM |
3399 | bytes of this `set_number_at', plus 3 bytes of |
3400 | the following `succeed_n'. */ | |
3401 | insert_op2 (set_number_at, laststart, 5, lower_bound, b); | |
3402 | b += 5; | |
3403 | startoffset += 5; | |
3404 | } | |
3405 | ||
3406 | if (upper_bound < 0) | |
3407 | { | |
3408 | /* A negative upper bound stands for infinity, | |
3409 | in which case it degenerates to a plain jump. */ | |
3410 | STORE_JUMP (jump, b, laststart + startoffset); | |
3411 | b += 3; | |
3412 | } | |
3413 | else if (upper_bound > 1) | |
3414 | { /* More than one repetition is allowed, so | |
3415 | append a backward jump to the `succeed_n' | |
3416 | that starts this interval. | |
3417 | ||
3418 | When we've reached this during matching, | |
3419 | we'll have matched the interval once, so | |
3420 | jump back only `upper_bound - 1' times. */ | |
3421 | STORE_JUMP2 (jump_n, b, laststart + startoffset, | |
3422 | upper_bound - 1); | |
3423 | b += 5; | |
3424 | ||
3425 | /* The location we want to set is the second | |
3426 | parameter of the `jump_n'; that is `b-2' as | |
3427 | an absolute address. `laststart' will be | |
3428 | the `set_number_at' we're about to insert; | |
3429 | `laststart+3' the number to set, the source | |
3430 | for the relative address. But we are | |
3431 | inserting into the middle of the pattern -- | |
3432 | so everything is getting moved up by 5. | |
3433 | Conclusion: (b - 2) - (laststart + 3) + 5, | |
3434 | i.e., b - laststart. | |
3435 | ||
3436 | We insert this at the beginning of the loop | |
3437 | so that if we fail during matching, we'll | |
3438 | reinitialize the bounds. */ | |
3439 | insert_op2 (set_number_at, laststart, b - laststart, | |
3440 | upper_bound - 1, b); | |
3441 | b += 5; | |
3442 | } | |
3443 | } | |
25fe55af RS |
3444 | pending_exact = 0; |
3445 | beg_interval = NULL; | |
3446 | } | |
3447 | break; | |
3448 | ||
3449 | unfetch_interval: | |
3450 | /* If an invalid interval, match the characters as literals. */ | |
3451 | assert (beg_interval); | |
3452 | p = beg_interval; | |
3453 | beg_interval = NULL; | |
3454 | ||
3455 | /* normal_char and normal_backslash need `c'. */ | |
ed0767d8 | 3456 | c = '{'; |
25fe55af RS |
3457 | |
3458 | if (!(syntax & RE_NO_BK_BRACES)) | |
3459 | { | |
ed0767d8 SM |
3460 | assert (p > pattern && p[-1] == '\\'); |
3461 | goto normal_backslash; | |
25fe55af | 3462 | } |
ed0767d8 SM |
3463 | else |
3464 | goto normal_char; | |
e318085a | 3465 | |
b18215fc | 3466 | #ifdef emacs |
25fe55af | 3467 | /* There is no way to specify the before_dot and after_dot |
7814e705 | 3468 | operators. rms says this is ok. --karl */ |
25fe55af | 3469 | case '=': |
5ac2eb34 | 3470 | laststart = b; |
25fe55af RS |
3471 | BUF_PUSH (at_dot); |
3472 | break; | |
3473 | ||
3474 | case 's': | |
3475 | laststart = b; | |
3476 | PATFETCH (c); | |
3477 | BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); | |
3478 | break; | |
3479 | ||
3480 | case 'S': | |
3481 | laststart = b; | |
3482 | PATFETCH (c); | |
3483 | BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); | |
3484 | break; | |
b18215fc RS |
3485 | |
3486 | case 'c': | |
3487 | laststart = b; | |
36595814 | 3488 | PATFETCH (c); |
b18215fc RS |
3489 | BUF_PUSH_2 (categoryspec, c); |
3490 | break; | |
e318085a | 3491 | |
b18215fc RS |
3492 | case 'C': |
3493 | laststart = b; | |
36595814 | 3494 | PATFETCH (c); |
b18215fc RS |
3495 | BUF_PUSH_2 (notcategoryspec, c); |
3496 | break; | |
3497 | #endif /* emacs */ | |
e318085a | 3498 | |
e318085a | 3499 | |
25fe55af | 3500 | case 'w': |
4bb91c68 SM |
3501 | if (syntax & RE_NO_GNU_OPS) |
3502 | goto normal_char; | |
25fe55af | 3503 | laststart = b; |
1fb352e0 | 3504 | BUF_PUSH_2 (syntaxspec, Sword); |
25fe55af | 3505 | break; |
e318085a | 3506 | |
e318085a | 3507 | |
25fe55af | 3508 | case 'W': |
4bb91c68 SM |
3509 | if (syntax & RE_NO_GNU_OPS) |
3510 | goto normal_char; | |
25fe55af | 3511 | laststart = b; |
1fb352e0 | 3512 | BUF_PUSH_2 (notsyntaxspec, Sword); |
25fe55af | 3513 | break; |
e318085a RS |
3514 | |
3515 | ||
25fe55af | 3516 | case '<': |
4bb91c68 SM |
3517 | if (syntax & RE_NO_GNU_OPS) |
3518 | goto normal_char; | |
5ac2eb34 | 3519 | laststart = b; |
25fe55af RS |
3520 | BUF_PUSH (wordbeg); |
3521 | break; | |
e318085a | 3522 | |
25fe55af | 3523 | case '>': |
4bb91c68 SM |
3524 | if (syntax & RE_NO_GNU_OPS) |
3525 | goto normal_char; | |
5ac2eb34 | 3526 | laststart = b; |
25fe55af RS |
3527 | BUF_PUSH (wordend); |
3528 | break; | |
e318085a | 3529 | |
669fa600 SM |
3530 | case '_': |
3531 | if (syntax & RE_NO_GNU_OPS) | |
3532 | goto normal_char; | |
3533 | laststart = b; | |
3534 | PATFETCH (c); | |
3535 | if (c == '<') | |
3536 | BUF_PUSH (symbeg); | |
3537 | else if (c == '>') | |
3538 | BUF_PUSH (symend); | |
3539 | else | |
3540 | FREE_STACK_RETURN (REG_BADPAT); | |
3541 | break; | |
3542 | ||
25fe55af | 3543 | case 'b': |
4bb91c68 SM |
3544 | if (syntax & RE_NO_GNU_OPS) |
3545 | goto normal_char; | |
25fe55af RS |
3546 | BUF_PUSH (wordbound); |
3547 | break; | |
e318085a | 3548 | |
25fe55af | 3549 | case 'B': |
4bb91c68 SM |
3550 | if (syntax & RE_NO_GNU_OPS) |
3551 | goto normal_char; | |
25fe55af RS |
3552 | BUF_PUSH (notwordbound); |
3553 | break; | |
fa9a63c5 | 3554 | |
25fe55af | 3555 | case '`': |
4bb91c68 SM |
3556 | if (syntax & RE_NO_GNU_OPS) |
3557 | goto normal_char; | |
25fe55af RS |
3558 | BUF_PUSH (begbuf); |
3559 | break; | |
e318085a | 3560 | |
25fe55af | 3561 | case '\'': |
4bb91c68 SM |
3562 | if (syntax & RE_NO_GNU_OPS) |
3563 | goto normal_char; | |
25fe55af RS |
3564 | BUF_PUSH (endbuf); |
3565 | break; | |
e318085a | 3566 | |
25fe55af RS |
3567 | case '1': case '2': case '3': case '4': case '5': |
3568 | case '6': case '7': case '8': case '9': | |
0cdd06f8 SM |
3569 | { |
3570 | regnum_t reg; | |
e318085a | 3571 | |
0cdd06f8 SM |
3572 | if (syntax & RE_NO_BK_REFS) |
3573 | goto normal_backslash; | |
e318085a | 3574 | |
0cdd06f8 | 3575 | reg = c - '0'; |
e318085a | 3576 | |
c69b0314 SM |
3577 | if (reg > bufp->re_nsub || reg < 1 |
3578 | /* Can't back reference to a subexp before its end. */ | |
3579 | || group_in_compile_stack (compile_stack, reg)) | |
0cdd06f8 | 3580 | FREE_STACK_RETURN (REG_ESUBREG); |
e318085a | 3581 | |
0cdd06f8 SM |
3582 | laststart = b; |
3583 | BUF_PUSH_2 (duplicate, reg); | |
3584 | } | |
25fe55af | 3585 | break; |
e318085a | 3586 | |
e318085a | 3587 | |
25fe55af RS |
3588 | case '+': |
3589 | case '?': | |
3590 | if (syntax & RE_BK_PLUS_QM) | |
3591 | goto handle_plus; | |
3592 | else | |
3593 | goto normal_backslash; | |
3594 | ||
3595 | default: | |
3596 | normal_backslash: | |
3597 | /* You might think it would be useful for \ to mean | |
3598 | not to translate; but if we don't translate it | |
4bb91c68 | 3599 | it will never match anything. */ |
25fe55af RS |
3600 | goto normal_char; |
3601 | } | |
3602 | break; | |
fa9a63c5 RM |
3603 | |
3604 | ||
3605 | default: | |
25fe55af | 3606 | /* Expects the character in `c'. */ |
fa9a63c5 | 3607 | normal_char: |
36595814 | 3608 | /* If no exactn currently being built. */ |
25fe55af | 3609 | if (!pending_exact |
fa9a63c5 | 3610 | |
25fe55af RS |
3611 | /* If last exactn not at current position. */ |
3612 | || pending_exact + *pending_exact + 1 != b | |
5e69f11e | 3613 | |
25fe55af | 3614 | /* We have only one byte following the exactn for the count. */ |
2d1675e4 | 3615 | || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH |
fa9a63c5 | 3616 | |
7814e705 | 3617 | /* If followed by a repetition operator. */ |
9d99031f | 3618 | || (p != pend && (*p == '*' || *p == '^')) |
fa9a63c5 | 3619 | || ((syntax & RE_BK_PLUS_QM) |
9d99031f RS |
3620 | ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?') |
3621 | : p != pend && (*p == '+' || *p == '?')) | |
fa9a63c5 | 3622 | || ((syntax & RE_INTERVALS) |
25fe55af | 3623 | && ((syntax & RE_NO_BK_BRACES) |
9d99031f RS |
3624 | ? p != pend && *p == '{' |
3625 | : p + 1 < pend && p[0] == '\\' && p[1] == '{'))) | |
fa9a63c5 RM |
3626 | { |
3627 | /* Start building a new exactn. */ | |
5e69f11e | 3628 | |
25fe55af | 3629 | laststart = b; |
fa9a63c5 RM |
3630 | |
3631 | BUF_PUSH_2 (exactn, 0); | |
3632 | pending_exact = b - 1; | |
25fe55af | 3633 | } |
5e69f11e | 3634 | |
2d1675e4 SM |
3635 | GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH); |
3636 | { | |
e0277a47 KH |
3637 | int len; |
3638 | ||
cf9c99bc | 3639 | if (multibyte) |
6fdd04b0 | 3640 | { |
cf9c99bc | 3641 | c = TRANSLATE (c); |
6fdd04b0 KH |
3642 | len = CHAR_STRING (c, b); |
3643 | b += len; | |
3644 | } | |
e0277a47 | 3645 | else |
6fdd04b0 | 3646 | { |
cf9c99bc KH |
3647 | c1 = RE_CHAR_TO_MULTIBYTE (c); |
3648 | if (! CHAR_BYTE8_P (c1)) | |
3649 | { | |
3650 | re_wchar_t c2 = TRANSLATE (c1); | |
3651 | ||
3652 | if (c1 != c2 && (c1 = RE_CHAR_TO_UNIBYTE (c2)) >= 0) | |
3653 | c = c1; | |
409f2919 | 3654 | } |
6fdd04b0 KH |
3655 | *b++ = c; |
3656 | len = 1; | |
3657 | } | |
2d1675e4 SM |
3658 | (*pending_exact) += len; |
3659 | } | |
3660 | ||
fa9a63c5 | 3661 | break; |
25fe55af | 3662 | } /* switch (c) */ |
fa9a63c5 RM |
3663 | } /* while p != pend */ |
3664 | ||
5e69f11e | 3665 | |
fa9a63c5 | 3666 | /* Through the pattern now. */ |
5e69f11e | 3667 | |
505bde11 | 3668 | FIXUP_ALT_JUMP (); |
fa9a63c5 | 3669 | |
5e69f11e | 3670 | if (!COMPILE_STACK_EMPTY) |
fa9a63c5 RM |
3671 | FREE_STACK_RETURN (REG_EPAREN); |
3672 | ||
3673 | /* If we don't want backtracking, force success | |
3674 | the first time we reach the end of the compiled pattern. */ | |
3675 | if (syntax & RE_NO_POSIX_BACKTRACKING) | |
3676 | BUF_PUSH (succeed); | |
3677 | ||
fa9a63c5 RM |
3678 | /* We have succeeded; set the length of the buffer. */ |
3679 | bufp->used = b - bufp->buffer; | |
3680 | ||
3681 | #ifdef DEBUG | |
99633e97 | 3682 | if (debug > 0) |
fa9a63c5 | 3683 | { |
505bde11 | 3684 | re_compile_fastmap (bufp); |
dc4a2ee0 | 3685 | DEBUG_PRINT ("\nCompiled pattern: \n"); |
fa9a63c5 RM |
3686 | print_compiled_pattern (bufp); |
3687 | } | |
99633e97 | 3688 | debug--; |
fa9a63c5 RM |
3689 | #endif /* DEBUG */ |
3690 | ||
3691 | #ifndef MATCH_MAY_ALLOCATE | |
3692 | /* Initialize the failure stack to the largest possible stack. This | |
3693 | isn't necessary unless we're trying to avoid calling alloca in | |
3694 | the search and match routines. */ | |
3695 | { | |
3696 | int num_regs = bufp->re_nsub + 1; | |
3697 | ||
320a2a73 | 3698 | if (fail_stack.size < re_max_failures * TYPICAL_FAILURE_SIZE) |
fa9a63c5 | 3699 | { |
a26f4ccd | 3700 | fail_stack.size = re_max_failures * TYPICAL_FAILURE_SIZE; |
38182d90 PE |
3701 | falk_stack.stack = realloc (fail_stack.stack, |
3702 | fail_stack.size * sizeof *falk_stack.stack); | |
fa9a63c5 RM |
3703 | } |
3704 | ||
3705 | regex_grow_registers (num_regs); | |
3706 | } | |
3707 | #endif /* not MATCH_MAY_ALLOCATE */ | |
3708 | ||
839966f3 | 3709 | FREE_STACK_RETURN (REG_NOERROR); |
fa9a63c5 RM |
3710 | } /* regex_compile */ |
3711 | \f | |
3712 | /* Subroutines for `regex_compile'. */ | |
3713 | ||
7814e705 | 3714 | /* Store OP at LOC followed by two-byte integer parameter ARG. */ |
fa9a63c5 RM |
3715 | |
3716 | static void | |
971de7fb | 3717 | store_op1 (re_opcode_t op, unsigned char *loc, int arg) |
fa9a63c5 RM |
3718 | { |
3719 | *loc = (unsigned char) op; | |
3720 | STORE_NUMBER (loc + 1, arg); | |
3721 | } | |
3722 | ||
3723 | ||
3724 | /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3725 | ||
3726 | static void | |
971de7fb | 3727 | store_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2) |
fa9a63c5 RM |
3728 | { |
3729 | *loc = (unsigned char) op; | |
3730 | STORE_NUMBER (loc + 1, arg1); | |
3731 | STORE_NUMBER (loc + 3, arg2); | |
3732 | } | |
3733 | ||
3734 | ||
3735 | /* Copy the bytes from LOC to END to open up three bytes of space at LOC | |
3736 | for OP followed by two-byte integer parameter ARG. */ | |
3737 | ||
3738 | static void | |
971de7fb | 3739 | insert_op1 (re_opcode_t op, unsigned char *loc, int arg, unsigned char *end) |
fa9a63c5 RM |
3740 | { |
3741 | register unsigned char *pfrom = end; | |
3742 | register unsigned char *pto = end + 3; | |
3743 | ||
3744 | while (pfrom != loc) | |
3745 | *--pto = *--pfrom; | |
5e69f11e | 3746 | |
fa9a63c5 RM |
3747 | store_op1 (op, loc, arg); |
3748 | } | |
3749 | ||
3750 | ||
3751 | /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ | |
3752 | ||
3753 | static void | |
971de7fb | 3754 | insert_op2 (re_opcode_t op, unsigned char *loc, int arg1, int arg2, unsigned char *end) |
fa9a63c5 RM |
3755 | { |
3756 | register unsigned char *pfrom = end; | |
3757 | register unsigned char *pto = end + 5; | |
3758 | ||
3759 | while (pfrom != loc) | |
3760 | *--pto = *--pfrom; | |
5e69f11e | 3761 | |
fa9a63c5 RM |
3762 | store_op2 (op, loc, arg1, arg2); |
3763 | } | |
3764 | ||
3765 | ||
3766 | /* P points to just after a ^ in PATTERN. Return true if that ^ comes | |
3767 | after an alternative or a begin-subexpression. We assume there is at | |
3768 | least one character before the ^. */ | |
3769 | ||
3770 | static boolean | |
29abe551 | 3771 | at_begline_loc_p (const_re_char *pattern, const_re_char *p, reg_syntax_t syntax) |
fa9a63c5 | 3772 | { |
01618498 | 3773 | re_char *prev = p - 2; |
95988fcf | 3774 | boolean odd_backslashes; |
5e69f11e | 3775 | |
95988fcf AS |
3776 | /* After a subexpression? */ |
3777 | if (*prev == '(') | |
3778 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3779 | ||
3780 | /* After an alternative? */ | |
3781 | else if (*prev == '|') | |
3782 | odd_backslashes = (syntax & RE_NO_BK_VBAR) == 0; | |
3783 | ||
3784 | /* After a shy subexpression? */ | |
3785 | else if (*prev == ':' && (syntax & RE_SHY_GROUPS)) | |
3786 | { | |
3787 | /* Skip over optional regnum. */ | |
3788 | while (prev - 1 >= pattern && prev[-1] >= '0' && prev[-1] <= '9') | |
3789 | --prev; | |
3790 | ||
3791 | if (!(prev - 2 >= pattern | |
3792 | && prev[-1] == '?' && prev[-2] == '(')) | |
3793 | return false; | |
3794 | prev -= 2; | |
3795 | odd_backslashes = (syntax & RE_NO_BK_PARENS) == 0; | |
3796 | } | |
3797 | else | |
3798 | return false; | |
3799 | ||
3800 | /* Count the number of preceding backslashes. */ | |
3801 | p = prev; | |
3802 | while (prev - 1 >= pattern && prev[-1] == '\\') | |
3803 | --prev; | |
3804 | return (p - prev) & odd_backslashes; | |
fa9a63c5 RM |
3805 | } |
3806 | ||
3807 | ||
3808 | /* The dual of at_begline_loc_p. This one is for $. We assume there is | |
3809 | at least one character after the $, i.e., `P < PEND'. */ | |
3810 | ||
3811 | static boolean | |
29abe551 | 3812 | at_endline_loc_p (const_re_char *p, const_re_char *pend, reg_syntax_t syntax) |
fa9a63c5 | 3813 | { |
01618498 | 3814 | re_char *next = p; |
fa9a63c5 | 3815 | boolean next_backslash = *next == '\\'; |
01618498 | 3816 | re_char *next_next = p + 1 < pend ? p + 1 : 0; |
5e69f11e | 3817 | |
fa9a63c5 RM |
3818 | return |
3819 | /* Before a subexpression? */ | |
3820 | (syntax & RE_NO_BK_PARENS ? *next == ')' | |
25fe55af | 3821 | : next_backslash && next_next && *next_next == ')') |
fa9a63c5 RM |
3822 | /* Before an alternative? */ |
3823 | || (syntax & RE_NO_BK_VBAR ? *next == '|' | |
25fe55af | 3824 | : next_backslash && next_next && *next_next == '|'); |
fa9a63c5 RM |
3825 | } |
3826 | ||
3827 | ||
5e69f11e | 3828 | /* Returns true if REGNUM is in one of COMPILE_STACK's elements and |
fa9a63c5 RM |
3829 | false if it's not. */ |
3830 | ||
3831 | static boolean | |
971de7fb | 3832 | group_in_compile_stack (compile_stack_type compile_stack, regnum_t regnum) |
fa9a63c5 | 3833 | { |
d1dfb56c | 3834 | ssize_t this_element; |
fa9a63c5 | 3835 | |
5e69f11e RM |
3836 | for (this_element = compile_stack.avail - 1; |
3837 | this_element >= 0; | |
fa9a63c5 RM |
3838 | this_element--) |
3839 | if (compile_stack.stack[this_element].regnum == regnum) | |
3840 | return true; | |
3841 | ||
3842 | return false; | |
3843 | } | |
fa9a63c5 | 3844 | \f |
f6a3f532 SM |
3845 | /* analyse_first. |
3846 | If fastmap is non-NULL, go through the pattern and fill fastmap | |
3847 | with all the possible leading chars. If fastmap is NULL, don't | |
3848 | bother filling it up (obviously) and only return whether the | |
3849 | pattern could potentially match the empty string. | |
3850 | ||
3851 | Return 1 if p..pend might match the empty string. | |
3852 | Return 0 if p..pend matches at least one char. | |
01618498 | 3853 | Return -1 if fastmap was not updated accurately. */ |
f6a3f532 SM |
3854 | |
3855 | static int | |
29abe551 PE |
3856 | analyse_first (const_re_char *p, const_re_char *pend, char *fastmap, |
3857 | const int multibyte) | |
fa9a63c5 | 3858 | { |
505bde11 | 3859 | int j, k; |
1fb352e0 | 3860 | boolean not; |
fa9a63c5 | 3861 | |
b18215fc | 3862 | /* If all elements for base leading-codes in fastmap is set, this |
7814e705 | 3863 | flag is set true. */ |
b18215fc RS |
3864 | boolean match_any_multibyte_characters = false; |
3865 | ||
f6a3f532 | 3866 | assert (p); |
5e69f11e | 3867 | |
505bde11 SM |
3868 | /* The loop below works as follows: |
3869 | - It has a working-list kept in the PATTERN_STACK and which basically | |
3870 | starts by only containing a pointer to the first operation. | |
3871 | - If the opcode we're looking at is a match against some set of | |
3872 | chars, then we add those chars to the fastmap and go on to the | |
3873 | next work element from the worklist (done via `break'). | |
3874 | - If the opcode is a control operator on the other hand, we either | |
3875 | ignore it (if it's meaningless at this point, such as `start_memory') | |
3876 | or execute it (if it's a jump). If the jump has several destinations | |
3877 | (i.e. `on_failure_jump'), then we push the other destination onto the | |
3878 | worklist. | |
3879 | We guarantee termination by ignoring backward jumps (more or less), | |
3880 | so that `p' is monotonically increasing. More to the point, we | |
3881 | never set `p' (or push) anything `<= p1'. */ | |
3882 | ||
01618498 | 3883 | while (p < pend) |
fa9a63c5 | 3884 | { |
505bde11 SM |
3885 | /* `p1' is used as a marker of how far back a `on_failure_jump' |
3886 | can go without being ignored. It is normally equal to `p' | |
3887 | (which prevents any backward `on_failure_jump') except right | |
3888 | after a plain `jump', to allow patterns such as: | |
3889 | 0: jump 10 | |
3890 | 3..9: <body> | |
3891 | 10: on_failure_jump 3 | |
3892 | as used for the *? operator. */ | |
01618498 | 3893 | re_char *p1 = p; |
5e69f11e | 3894 | |
7393bcbb | 3895 | switch (*p++) |
fa9a63c5 | 3896 | { |
f6a3f532 | 3897 | case succeed: |
01618498 | 3898 | return 1; |
fa9a63c5 | 3899 | |
fa9a63c5 | 3900 | case duplicate: |
505bde11 SM |
3901 | /* If the first character has to match a backreference, that means |
3902 | that the group was empty (since it already matched). Since this | |
3903 | is the only case that interests us here, we can assume that the | |
3904 | backreference must match the empty string. */ | |
3905 | p++; | |
3906 | continue; | |
fa9a63c5 RM |
3907 | |
3908 | ||
3909 | /* Following are the cases which match a character. These end | |
7814e705 | 3910 | with `break'. */ |
fa9a63c5 RM |
3911 | |
3912 | case exactn: | |
e0277a47 | 3913 | if (fastmap) |
cf9c99bc KH |
3914 | { |
3915 | /* If multibyte is nonzero, the first byte of each | |
3916 | character is an ASCII or a leading code. Otherwise, | |
3917 | each byte is a character. Thus, this works in both | |
3918 | cases. */ | |
3919 | fastmap[p[1]] = 1; | |
3920 | if (! multibyte) | |
3921 | { | |
3922 | /* For the case of matching this unibyte regex | |
3923 | against multibyte, we must set a leading code of | |
3924 | the corresponding multibyte character. */ | |
3925 | int c = RE_CHAR_TO_MULTIBYTE (p[1]); | |
3926 | ||
86e893e3 | 3927 | fastmap[CHAR_LEADING_CODE (c)] = 1; |
cf9c99bc KH |
3928 | } |
3929 | } | |
fa9a63c5 RM |
3930 | break; |
3931 | ||
3932 | ||
1fb352e0 SM |
3933 | case anychar: |
3934 | /* We could put all the chars except for \n (and maybe \0) | |
3935 | but we don't bother since it is generally not worth it. */ | |
f6a3f532 | 3936 | if (!fastmap) break; |
01618498 | 3937 | return -1; |
fa9a63c5 RM |
3938 | |
3939 | ||
b18215fc | 3940 | case charset_not: |
1fb352e0 | 3941 | if (!fastmap) break; |
bf216479 KH |
3942 | { |
3943 | /* Chars beyond end of bitmap are possible matches. */ | |
bf216479 | 3944 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH; |
cf9c99bc | 3945 | j < (1 << BYTEWIDTH); j++) |
bf216479 KH |
3946 | fastmap[j] = 1; |
3947 | } | |
3948 | ||
1fb352e0 SM |
3949 | /* Fallthrough */ |
3950 | case charset: | |
3951 | if (!fastmap) break; | |
3952 | not = (re_opcode_t) *(p - 1) == charset_not; | |
b18215fc RS |
3953 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH - 1, p++; |
3954 | j >= 0; j--) | |
1fb352e0 | 3955 | if (!!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) ^ not) |
49da453b | 3956 | fastmap[j] = 1; |
b18215fc | 3957 | |
6482db2e KH |
3958 | #ifdef emacs |
3959 | if (/* Any leading code can possibly start a character | |
1fb352e0 | 3960 | which doesn't match the specified set of characters. */ |
6482db2e | 3961 | not |
409f2919 | 3962 | || |
6482db2e KH |
3963 | /* If we can match a character class, we can match any |
3964 | multibyte characters. */ | |
3965 | (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) | |
3966 | && CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0)) | |
3967 | ||
b18215fc | 3968 | { |
b18215fc RS |
3969 | if (match_any_multibyte_characters == false) |
3970 | { | |
6482db2e KH |
3971 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
3972 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
6fdd04b0 | 3973 | fastmap[j] = 1; |
b18215fc RS |
3974 | match_any_multibyte_characters = true; |
3975 | } | |
3976 | } | |
b18215fc | 3977 | |
1fb352e0 SM |
3978 | else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) |
3979 | && match_any_multibyte_characters == false) | |
3980 | { | |
bf216479 | 3981 | /* Set fastmap[I] to 1 where I is a leading code of each |
51e4f4a8 | 3982 | multibyte character in the range table. */ |
1fb352e0 | 3983 | int c, count; |
bf216479 | 3984 | unsigned char lc1, lc2; |
b18215fc | 3985 | |
1fb352e0 | 3986 | /* Make P points the range table. `+ 2' is to skip flag |
0b32bf0e | 3987 | bits for a character class. */ |
1fb352e0 | 3988 | p += CHARSET_BITMAP_SIZE (&p[-2]) + 2; |
b18215fc | 3989 | |
1fb352e0 SM |
3990 | /* Extract the number of ranges in range table into COUNT. */ |
3991 | EXTRACT_NUMBER_AND_INCR (count, p); | |
cf9c99bc | 3992 | for (; count > 0; count--, p += 3) |
1fb352e0 | 3993 | { |
9117d724 KH |
3994 | /* Extract the start and end of each range. */ |
3995 | EXTRACT_CHARACTER (c, p); | |
bf216479 | 3996 | lc1 = CHAR_LEADING_CODE (c); |
9117d724 | 3997 | p += 3; |
1fb352e0 | 3998 | EXTRACT_CHARACTER (c, p); |
bf216479 KH |
3999 | lc2 = CHAR_LEADING_CODE (c); |
4000 | for (j = lc1; j <= lc2; j++) | |
9117d724 | 4001 | fastmap[j] = 1; |
1fb352e0 SM |
4002 | } |
4003 | } | |
6482db2e | 4004 | #endif |
b18215fc RS |
4005 | break; |
4006 | ||
1fb352e0 SM |
4007 | case syntaxspec: |
4008 | case notsyntaxspec: | |
4009 | if (!fastmap) break; | |
4010 | #ifndef emacs | |
4011 | not = (re_opcode_t)p[-1] == notsyntaxspec; | |
4012 | k = *p++; | |
4013 | for (j = 0; j < (1 << BYTEWIDTH); j++) | |
990b2375 | 4014 | if ((SYNTAX (j) == (enum syntaxcode) k) ^ not) |
b18215fc | 4015 | fastmap[j] = 1; |
b18215fc | 4016 | break; |
1fb352e0 | 4017 | #else /* emacs */ |
b18215fc RS |
4018 | /* This match depends on text properties. These end with |
4019 | aborting optimizations. */ | |
01618498 | 4020 | return -1; |
b18215fc RS |
4021 | |
4022 | case categoryspec: | |
b18215fc | 4023 | case notcategoryspec: |
1fb352e0 SM |
4024 | if (!fastmap) break; |
4025 | not = (re_opcode_t)p[-1] == notcategoryspec; | |
b18215fc | 4026 | k = *p++; |
6482db2e | 4027 | for (j = (1 << BYTEWIDTH); j >= 0; j--) |
1fb352e0 | 4028 | if ((CHAR_HAS_CATEGORY (j, k)) ^ not) |
b18215fc RS |
4029 | fastmap[j] = 1; |
4030 | ||
6482db2e KH |
4031 | /* Any leading code can possibly start a character which |
4032 | has or doesn't has the specified category. */ | |
4033 | if (match_any_multibyte_characters == false) | |
6fdd04b0 | 4034 | { |
6482db2e KH |
4035 | for (j = MIN_MULTIBYTE_LEADING_CODE; |
4036 | j <= MAX_MULTIBYTE_LEADING_CODE; j++) | |
4037 | fastmap[j] = 1; | |
4038 | match_any_multibyte_characters = true; | |
6fdd04b0 | 4039 | } |
b18215fc RS |
4040 | break; |
4041 | ||
fa9a63c5 | 4042 | /* All cases after this match the empty string. These end with |
25fe55af | 4043 | `continue'. */ |
fa9a63c5 | 4044 | |
fa9a63c5 RM |
4045 | case before_dot: |
4046 | case at_dot: | |
4047 | case after_dot: | |
1fb352e0 | 4048 | #endif /* !emacs */ |
25fe55af RS |
4049 | case no_op: |
4050 | case begline: | |
4051 | case endline: | |
fa9a63c5 RM |
4052 | case begbuf: |
4053 | case endbuf: | |
4054 | case wordbound: | |
4055 | case notwordbound: | |
4056 | case wordbeg: | |
4057 | case wordend: | |
669fa600 SM |
4058 | case symbeg: |
4059 | case symend: | |
25fe55af | 4060 | continue; |
fa9a63c5 RM |
4061 | |
4062 | ||
fa9a63c5 | 4063 | case jump: |
25fe55af | 4064 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 SM |
4065 | if (j < 0) |
4066 | /* Backward jumps can only go back to code that we've already | |
4067 | visited. `re_compile' should make sure this is true. */ | |
4068 | break; | |
25fe55af | 4069 | p += j; |
7393bcbb | 4070 | switch (*p) |
505bde11 SM |
4071 | { |
4072 | case on_failure_jump: | |
4073 | case on_failure_keep_string_jump: | |
505bde11 | 4074 | case on_failure_jump_loop: |
0683b6fa | 4075 | case on_failure_jump_nastyloop: |
505bde11 SM |
4076 | case on_failure_jump_smart: |
4077 | p++; | |
4078 | break; | |
4079 | default: | |
4080 | continue; | |
4081 | }; | |
4082 | /* Keep `p1' to allow the `on_failure_jump' we are jumping to | |
4083 | to jump back to "just after here". */ | |
4084 | /* Fallthrough */ | |
fa9a63c5 | 4085 | |
25fe55af RS |
4086 | case on_failure_jump: |
4087 | case on_failure_keep_string_jump: | |
0683b6fa | 4088 | case on_failure_jump_nastyloop: |
505bde11 SM |
4089 | case on_failure_jump_loop: |
4090 | case on_failure_jump_smart: | |
25fe55af | 4091 | EXTRACT_NUMBER_AND_INCR (j, p); |
505bde11 | 4092 | if (p + j <= p1) |
ed0767d8 | 4093 | ; /* Backward jump to be ignored. */ |
01618498 SM |
4094 | else |
4095 | { /* We have to look down both arms. | |
4096 | We first go down the "straight" path so as to minimize | |
4097 | stack usage when going through alternatives. */ | |
4098 | int r = analyse_first (p, pend, fastmap, multibyte); | |
4099 | if (r) return r; | |
4100 | p += j; | |
4101 | } | |
25fe55af | 4102 | continue; |
fa9a63c5 RM |
4103 | |
4104 | ||
ed0767d8 SM |
4105 | case jump_n: |
4106 | /* This code simply does not properly handle forward jump_n. */ | |
4107 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p); assert (j < 0)); | |
4108 | p += 4; | |
4109 | /* jump_n can either jump or fall through. The (backward) jump | |
4110 | case has already been handled, so we only need to look at the | |
4111 | fallthrough case. */ | |
4112 | continue; | |
177c0ea7 | 4113 | |
fa9a63c5 | 4114 | case succeed_n: |
ed0767d8 SM |
4115 | /* If N == 0, it should be an on_failure_jump_loop instead. */ |
4116 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0)); | |
4117 | p += 4; | |
4118 | /* We only care about one iteration of the loop, so we don't | |
4119 | need to consider the case where this behaves like an | |
4120 | on_failure_jump. */ | |
25fe55af | 4121 | continue; |
fa9a63c5 RM |
4122 | |
4123 | ||
4124 | case set_number_at: | |
25fe55af RS |
4125 | p += 4; |
4126 | continue; | |
fa9a63c5 RM |
4127 | |
4128 | ||
4129 | case start_memory: | |
25fe55af | 4130 | case stop_memory: |
505bde11 | 4131 | p += 1; |
fa9a63c5 RM |
4132 | continue; |
4133 | ||
4134 | ||
4135 | default: | |
25fe55af RS |
4136 | abort (); /* We have listed all the cases. */ |
4137 | } /* switch *p++ */ | |
fa9a63c5 RM |
4138 | |
4139 | /* Getting here means we have found the possible starting | |
25fe55af | 4140 | characters for one path of the pattern -- and that the empty |
7814e705 | 4141 | string does not match. We need not follow this path further. */ |
01618498 | 4142 | return 0; |
fa9a63c5 RM |
4143 | } /* while p */ |
4144 | ||
01618498 SM |
4145 | /* We reached the end without matching anything. */ |
4146 | return 1; | |
4147 | ||
f6a3f532 SM |
4148 | } /* analyse_first */ |
4149 | \f | |
4150 | /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in | |
4151 | BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible | |
4152 | characters can start a string that matches the pattern. This fastmap | |
4153 | is used by re_search to skip quickly over impossible starting points. | |
4154 | ||
4155 | Character codes above (1 << BYTEWIDTH) are not represented in the | |
4156 | fastmap, but the leading codes are represented. Thus, the fastmap | |
4157 | indicates which character sets could start a match. | |
4158 | ||
4159 | The caller must supply the address of a (1 << BYTEWIDTH)-byte data | |
4160 | area as BUFP->fastmap. | |
4161 | ||
4162 | We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in | |
4163 | the pattern buffer. | |
4164 | ||
4165 | Returns 0 if we succeed, -2 if an internal error. */ | |
4166 | ||
4167 | int | |
971de7fb | 4168 | re_compile_fastmap (struct re_pattern_buffer *bufp) |
f6a3f532 SM |
4169 | { |
4170 | char *fastmap = bufp->fastmap; | |
4171 | int analysis; | |
4172 | ||
4173 | assert (fastmap && bufp->buffer); | |
4174 | ||
72af86bd | 4175 | memset (fastmap, 0, 1 << BYTEWIDTH); /* Assume nothing's valid. */ |
f6a3f532 SM |
4176 | bufp->fastmap_accurate = 1; /* It will be when we're done. */ |
4177 | ||
4178 | analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used, | |
2d1675e4 | 4179 | fastmap, RE_MULTIBYTE_P (bufp)); |
c0f9ea08 | 4180 | bufp->can_be_null = (analysis != 0); |
fa9a63c5 RM |
4181 | return 0; |
4182 | } /* re_compile_fastmap */ | |
4183 | \f | |
4184 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | |
4185 | ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | |
4186 | this memory for recording register information. STARTS and ENDS | |
4187 | must be allocated using the malloc library routine, and must each | |
4188 | be at least NUM_REGS * sizeof (regoff_t) bytes long. | |
4189 | ||
4190 | If NUM_REGS == 0, then subsequent matches should allocate their own | |
4191 | register data. | |
4192 | ||
4193 | Unless this function is called, the first search or match using | |
4194 | PATTERN_BUFFER will allocate its own register data, without | |
4195 | freeing the old data. */ | |
4196 | ||
4197 | void | |
971de7fb | 4198 | re_set_registers (struct re_pattern_buffer *bufp, struct re_registers *regs, unsigned int num_regs, regoff_t *starts, regoff_t *ends) |
fa9a63c5 RM |
4199 | { |
4200 | if (num_regs) | |
4201 | { | |
4202 | bufp->regs_allocated = REGS_REALLOCATE; | |
4203 | regs->num_regs = num_regs; | |
4204 | regs->start = starts; | |
4205 | regs->end = ends; | |
4206 | } | |
4207 | else | |
4208 | { | |
4209 | bufp->regs_allocated = REGS_UNALLOCATED; | |
4210 | regs->num_regs = 0; | |
7d652d97 | 4211 | regs->start = regs->end = 0; |
fa9a63c5 RM |
4212 | } |
4213 | } | |
c0f9ea08 | 4214 | WEAK_ALIAS (__re_set_registers, re_set_registers) |
fa9a63c5 | 4215 | \f |
7814e705 | 4216 | /* Searching routines. */ |
fa9a63c5 RM |
4217 | |
4218 | /* Like re_search_2, below, but only one string is specified, and | |
4219 | doesn't let you say where to stop matching. */ | |
4220 | ||
d1dfb56c EZ |
4221 | regoff_t |
4222 | re_search (struct re_pattern_buffer *bufp, const char *string, size_t size, | |
4223 | ssize_t startpos, ssize_t range, struct re_registers *regs) | |
fa9a63c5 | 4224 | { |
5e69f11e | 4225 | return re_search_2 (bufp, NULL, 0, string, size, startpos, range, |
fa9a63c5 RM |
4226 | regs, size); |
4227 | } | |
c0f9ea08 | 4228 | WEAK_ALIAS (__re_search, re_search) |
fa9a63c5 | 4229 | |
70806df6 KH |
4230 | /* Head address of virtual concatenation of string. */ |
4231 | #define HEAD_ADDR_VSTRING(P) \ | |
4232 | (((P) >= size1 ? string2 : string1)) | |
4233 | ||
b18215fc RS |
4234 | /* Address of POS in the concatenation of virtual string. */ |
4235 | #define POS_ADDR_VSTRING(POS) \ | |
4236 | (((POS) >= size1 ? string2 - size1 : string1) + (POS)) | |
fa9a63c5 RM |
4237 | |
4238 | /* Using the compiled pattern in BUFP->buffer, first tries to match the | |
4239 | virtual concatenation of STRING1 and STRING2, starting first at index | |
4240 | STARTPOS, then at STARTPOS + 1, and so on. | |
5e69f11e | 4241 | |
fa9a63c5 | 4242 | STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. |
5e69f11e | 4243 | |
fa9a63c5 RM |
4244 | RANGE is how far to scan while trying to match. RANGE = 0 means try |
4245 | only at STARTPOS; in general, the last start tried is STARTPOS + | |
4246 | RANGE. | |
5e69f11e | 4247 | |
fa9a63c5 RM |
4248 | In REGS, return the indices of the virtual concatenation of STRING1 |
4249 | and STRING2 that matched the entire BUFP->buffer and its contained | |
4250 | subexpressions. | |
5e69f11e | 4251 | |
fa9a63c5 RM |
4252 | Do not consider matching one past the index STOP in the virtual |
4253 | concatenation of STRING1 and STRING2. | |
4254 | ||
4255 | We return either the position in the strings at which the match was | |
4256 | found, -1 if no match, or -2 if error (such as failure | |
4257 | stack overflow). */ | |
4258 | ||
d1dfb56c EZ |
4259 | regoff_t |
4260 | re_search_2 (struct re_pattern_buffer *bufp, const char *str1, size_t size1, | |
4261 | const char *str2, size_t size2, ssize_t startpos, ssize_t range, | |
4262 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4263 | { |
d1dfb56c | 4264 | regoff_t val; |
66f0296e SM |
4265 | re_char *string1 = (re_char*) str1; |
4266 | re_char *string2 = (re_char*) str2; | |
fa9a63c5 | 4267 | register char *fastmap = bufp->fastmap; |
6676cb1c | 4268 | register RE_TRANSLATE_TYPE translate = bufp->translate; |
d1dfb56c EZ |
4269 | size_t total_size = size1 + size2; |
4270 | ssize_t endpos = startpos + range; | |
c0f9ea08 | 4271 | boolean anchored_start; |
cf9c99bc KH |
4272 | /* Nonzero if we are searching multibyte string. */ |
4273 | const boolean multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
b18215fc | 4274 | |
fa9a63c5 RM |
4275 | /* Check for out-of-range STARTPOS. */ |
4276 | if (startpos < 0 || startpos > total_size) | |
4277 | return -1; | |
5e69f11e | 4278 | |
fa9a63c5 | 4279 | /* Fix up RANGE if it might eventually take us outside |
34597fa9 | 4280 | the virtual concatenation of STRING1 and STRING2. |
5e69f11e | 4281 | Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ |
34597fa9 RS |
4282 | if (endpos < 0) |
4283 | range = 0 - startpos; | |
fa9a63c5 RM |
4284 | else if (endpos > total_size) |
4285 | range = total_size - startpos; | |
4286 | ||
4287 | /* If the search isn't to be a backwards one, don't waste time in a | |
7b140fd7 | 4288 | search for a pattern anchored at beginning of buffer. */ |
fa9a63c5 RM |
4289 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) |
4290 | { | |
4291 | if (startpos > 0) | |
4292 | return -1; | |
4293 | else | |
7b140fd7 | 4294 | range = 0; |
fa9a63c5 RM |
4295 | } |
4296 | ||
ae4788a8 RS |
4297 | #ifdef emacs |
4298 | /* In a forward search for something that starts with \=. | |
4299 | don't keep searching past point. */ | |
4300 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) | |
4301 | { | |
7b140fd7 RS |
4302 | range = PT_BYTE - BEGV_BYTE - startpos; |
4303 | if (range < 0) | |
ae4788a8 RS |
4304 | return -1; |
4305 | } | |
4306 | #endif /* emacs */ | |
4307 | ||
fa9a63c5 RM |
4308 | /* Update the fastmap now if not correct already. */ |
4309 | if (fastmap && !bufp->fastmap_accurate) | |
01618498 | 4310 | re_compile_fastmap (bufp); |
5e69f11e | 4311 | |
c8499ba5 | 4312 | /* See whether the pattern is anchored. */ |
c0f9ea08 | 4313 | anchored_start = (bufp->buffer[0] == begline); |
c8499ba5 | 4314 | |
b18215fc | 4315 | #ifdef emacs |
d48cd3f4 | 4316 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
cc9b4df2 | 4317 | { |
d1dfb56c | 4318 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (startpos)); |
cc9b4df2 KH |
4319 | |
4320 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); | |
4321 | } | |
b18215fc RS |
4322 | #endif |
4323 | ||
fa9a63c5 RM |
4324 | /* Loop through the string, looking for a place to start matching. */ |
4325 | for (;;) | |
5e69f11e | 4326 | { |
c8499ba5 RS |
4327 | /* If the pattern is anchored, |
4328 | skip quickly past places we cannot match. | |
4329 | We don't bother to treat startpos == 0 specially | |
4330 | because that case doesn't repeat. */ | |
4331 | if (anchored_start && startpos > 0) | |
4332 | { | |
c0f9ea08 SM |
4333 | if (! ((startpos <= size1 ? string1[startpos - 1] |
4334 | : string2[startpos - size1 - 1]) | |
4335 | == '\n')) | |
c8499ba5 RS |
4336 | goto advance; |
4337 | } | |
4338 | ||
fa9a63c5 | 4339 | /* If a fastmap is supplied, skip quickly over characters that |
25fe55af RS |
4340 | cannot be the start of a match. If the pattern can match the |
4341 | null string, however, we don't need to skip characters; we want | |
7814e705 | 4342 | the first null string. */ |
fa9a63c5 RM |
4343 | if (fastmap && startpos < total_size && !bufp->can_be_null) |
4344 | { | |
66f0296e | 4345 | register re_char *d; |
01618498 | 4346 | register re_wchar_t buf_ch; |
e934739e RS |
4347 | |
4348 | d = POS_ADDR_VSTRING (startpos); | |
4349 | ||
7814e705 | 4350 | if (range > 0) /* Searching forwards. */ |
fa9a63c5 | 4351 | { |
fa9a63c5 | 4352 | register int lim = 0; |
d1dfb56c | 4353 | ssize_t irange = range; |
fa9a63c5 | 4354 | |
25fe55af RS |
4355 | if (startpos < size1 && startpos + range >= size1) |
4356 | lim = range - (size1 - startpos); | |
fa9a63c5 | 4357 | |
25fe55af RS |
4358 | /* Written out as an if-else to avoid testing `translate' |
4359 | inside the loop. */ | |
28ae27ae AS |
4360 | if (RE_TRANSLATE_P (translate)) |
4361 | { | |
e934739e RS |
4362 | if (multibyte) |
4363 | while (range > lim) | |
4364 | { | |
4365 | int buf_charlen; | |
4366 | ||
62a6e103 | 4367 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 4368 | buf_ch = RE_TRANSLATE (translate, buf_ch); |
bf216479 | 4369 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
e934739e RS |
4370 | break; |
4371 | ||
4372 | range -= buf_charlen; | |
4373 | d += buf_charlen; | |
4374 | } | |
4375 | else | |
bf216479 | 4376 | while (range > lim) |
33c46939 | 4377 | { |
cf9c99bc KH |
4378 | register re_wchar_t ch, translated; |
4379 | ||
bf216479 | 4380 | buf_ch = *d; |
cf9c99bc KH |
4381 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); |
4382 | translated = RE_TRANSLATE (translate, ch); | |
4383 | if (translated != ch | |
4384 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4385 | buf_ch = ch; | |
6fdd04b0 | 4386 | if (fastmap[buf_ch]) |
bf216479 | 4387 | break; |
33c46939 RS |
4388 | d++; |
4389 | range--; | |
4390 | } | |
e934739e | 4391 | } |
fa9a63c5 | 4392 | else |
6fdd04b0 KH |
4393 | { |
4394 | if (multibyte) | |
4395 | while (range > lim) | |
4396 | { | |
4397 | int buf_charlen; | |
fa9a63c5 | 4398 | |
62a6e103 | 4399 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
6fdd04b0 KH |
4400 | if (fastmap[CHAR_LEADING_CODE (buf_ch)]) |
4401 | break; | |
4402 | range -= buf_charlen; | |
4403 | d += buf_charlen; | |
4404 | } | |
e934739e | 4405 | else |
6fdd04b0 | 4406 | while (range > lim && !fastmap[*d]) |
33c46939 RS |
4407 | { |
4408 | d++; | |
4409 | range--; | |
4410 | } | |
e934739e | 4411 | } |
fa9a63c5 RM |
4412 | startpos += irange - range; |
4413 | } | |
7814e705 | 4414 | else /* Searching backwards. */ |
fa9a63c5 | 4415 | { |
ba5e343c KH |
4416 | if (multibyte) |
4417 | { | |
62a6e103 | 4418 | buf_ch = STRING_CHAR (d); |
ba5e343c KH |
4419 | buf_ch = TRANSLATE (buf_ch); |
4420 | if (! fastmap[CHAR_LEADING_CODE (buf_ch)]) | |
4421 | goto advance; | |
4422 | } | |
4423 | else | |
4424 | { | |
cf9c99bc KH |
4425 | register re_wchar_t ch, translated; |
4426 | ||
4427 | buf_ch = *d; | |
4428 | ch = RE_CHAR_TO_MULTIBYTE (buf_ch); | |
4429 | translated = TRANSLATE (ch); | |
4430 | if (translated != ch | |
4431 | && (ch = RE_CHAR_TO_UNIBYTE (translated)) >= 0) | |
4432 | buf_ch = ch; | |
4433 | if (! fastmap[TRANSLATE (buf_ch)]) | |
ba5e343c KH |
4434 | goto advance; |
4435 | } | |
fa9a63c5 RM |
4436 | } |
4437 | } | |
4438 | ||
4439 | /* If can't match the null string, and that's all we have left, fail. */ | |
4440 | if (range >= 0 && startpos == total_size && fastmap | |
25fe55af | 4441 | && !bufp->can_be_null) |
fa9a63c5 RM |
4442 | return -1; |
4443 | ||
4444 | val = re_match_2_internal (bufp, string1, size1, string2, size2, | |
4445 | startpos, regs, stop); | |
fa9a63c5 RM |
4446 | |
4447 | if (val >= 0) | |
4448 | return startpos; | |
5e69f11e | 4449 | |
fa9a63c5 RM |
4450 | if (val == -2) |
4451 | return -2; | |
4452 | ||
4453 | advance: | |
5e69f11e | 4454 | if (!range) |
25fe55af | 4455 | break; |
5e69f11e | 4456 | else if (range > 0) |
25fe55af | 4457 | { |
b18215fc RS |
4458 | /* Update STARTPOS to the next character boundary. */ |
4459 | if (multibyte) | |
4460 | { | |
66f0296e | 4461 | re_char *p = POS_ADDR_VSTRING (startpos); |
aa3830c4 | 4462 | int len = BYTES_BY_CHAR_HEAD (*p); |
b18215fc RS |
4463 | |
4464 | range -= len; | |
4465 | if (range < 0) | |
4466 | break; | |
4467 | startpos += len; | |
4468 | } | |
4469 | else | |
4470 | { | |
b560c397 RS |
4471 | range--; |
4472 | startpos++; | |
4473 | } | |
e318085a | 4474 | } |
fa9a63c5 | 4475 | else |
25fe55af RS |
4476 | { |
4477 | range++; | |
4478 | startpos--; | |
b18215fc RS |
4479 | |
4480 | /* Update STARTPOS to the previous character boundary. */ | |
4481 | if (multibyte) | |
4482 | { | |
70806df6 KH |
4483 | re_char *p = POS_ADDR_VSTRING (startpos) + 1; |
4484 | re_char *p0 = p; | |
4485 | re_char *phead = HEAD_ADDR_VSTRING (startpos); | |
b18215fc RS |
4486 | |
4487 | /* Find the head of multibyte form. */ | |
70806df6 KH |
4488 | PREV_CHAR_BOUNDARY (p, phead); |
4489 | range += p0 - 1 - p; | |
4490 | if (range > 0) | |
4491 | break; | |
b18215fc | 4492 | |
70806df6 | 4493 | startpos -= p0 - 1 - p; |
b18215fc | 4494 | } |
25fe55af | 4495 | } |
fa9a63c5 RM |
4496 | } |
4497 | return -1; | |
4498 | } /* re_search_2 */ | |
c0f9ea08 | 4499 | WEAK_ALIAS (__re_search_2, re_search_2) |
fa9a63c5 RM |
4500 | \f |
4501 | /* Declarations and macros for re_match_2. */ | |
4502 | ||
261cb4bb PE |
4503 | static int bcmp_translate (re_char *s1, re_char *s2, |
4504 | register ssize_t len, | |
4505 | RE_TRANSLATE_TYPE translate, | |
4506 | const int multibyte); | |
fa9a63c5 RM |
4507 | |
4508 | /* This converts PTR, a pointer into one of the search strings `string1' | |
4509 | and `string2' into an offset from the beginning of that string. */ | |
4510 | #define POINTER_TO_OFFSET(ptr) \ | |
4511 | (FIRST_STRING_P (ptr) \ | |
dc4a2ee0 PE |
4512 | ? (ptr) - string1 \ |
4513 | : (ptr) - string2 + (ptrdiff_t) size1) | |
fa9a63c5 | 4514 | |
fa9a63c5 | 4515 | /* Call before fetching a character with *d. This switches over to |
419d1c74 SM |
4516 | string2 if necessary. |
4517 | Check re_match_2_internal for a discussion of why end_match_2 might | |
4518 | not be within string2 (but be equal to end_match_1 instead). */ | |
fa9a63c5 | 4519 | #define PREFETCH() \ |
25fe55af | 4520 | while (d == dend) \ |
fa9a63c5 RM |
4521 | { \ |
4522 | /* End of string2 => fail. */ \ | |
25fe55af RS |
4523 | if (dend == end_match_2) \ |
4524 | goto fail; \ | |
4bb91c68 | 4525 | /* End of string1 => advance to string2. */ \ |
25fe55af | 4526 | d = string2; \ |
fa9a63c5 RM |
4527 | dend = end_match_2; \ |
4528 | } | |
4529 | ||
f1ad044f SM |
4530 | /* Call before fetching a char with *d if you already checked other limits. |
4531 | This is meant for use in lookahead operations like wordend, etc.. | |
4532 | where we might need to look at parts of the string that might be | |
4533 | outside of the LIMITs (i.e past `stop'). */ | |
4534 | #define PREFETCH_NOLIMIT() \ | |
4535 | if (d == end1) \ | |
4536 | { \ | |
4537 | d = string2; \ | |
4538 | dend = end_match_2; \ | |
4539 | } \ | |
fa9a63c5 RM |
4540 | |
4541 | /* Test if at very beginning or at very end of the virtual concatenation | |
7814e705 | 4542 | of `string1' and `string2'. If only one string, it's `string2'. */ |
fa9a63c5 | 4543 | #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) |
5e69f11e | 4544 | #define AT_STRINGS_END(d) ((d) == end2) |
fa9a63c5 | 4545 | |
9121ca40 | 4546 | /* Disabled due to a compiler bug -- see comment at case wordbound */ |
b18215fc RS |
4547 | |
4548 | /* The comment at case wordbound is following one, but we don't use | |
4549 | AT_WORD_BOUNDARY anymore to support multibyte form. | |
4550 | ||
4551 | The DEC Alpha C compiler 3.x generates incorrect code for the | |
25fe55af | 4552 | test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of |
7814e705 | 4553 | AT_WORD_BOUNDARY, so this code is disabled. Expanding the |
b18215fc RS |
4554 | macro and introducing temporary variables works around the bug. */ |
4555 | ||
9121ca40 | 4556 | #if 0 |
b313f9d8 PE |
4557 | /* Test if D points to a character which is word-constituent. We have |
4558 | two special cases to check for: if past the end of string1, look at | |
4559 | the first character in string2; and if before the beginning of | |
4560 | string2, look at the last character in string1. */ | |
4561 | #define WORDCHAR_P(d) \ | |
4562 | (SYNTAX ((d) == end1 ? *string2 \ | |
4563 | : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ | |
4564 | == Sword) | |
4565 | ||
fa9a63c5 RM |
4566 | /* Test if the character before D and the one at D differ with respect |
4567 | to being word-constituent. */ | |
4568 | #define AT_WORD_BOUNDARY(d) \ | |
4569 | (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ | |
4570 | || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) | |
9121ca40 | 4571 | #endif |
fa9a63c5 RM |
4572 | |
4573 | /* Free everything we malloc. */ | |
4574 | #ifdef MATCH_MAY_ALLOCATE | |
952db0d7 PE |
4575 | # define FREE_VAR(var) \ |
4576 | do { \ | |
4577 | if (var) \ | |
4578 | { \ | |
4579 | REGEX_FREE (var); \ | |
4580 | var = NULL; \ | |
4581 | } \ | |
4582 | } while (0) | |
0b32bf0e | 4583 | # define FREE_VARIABLES() \ |
fa9a63c5 RM |
4584 | do { \ |
4585 | REGEX_FREE_STACK (fail_stack.stack); \ | |
4586 | FREE_VAR (regstart); \ | |
4587 | FREE_VAR (regend); \ | |
fa9a63c5 RM |
4588 | FREE_VAR (best_regstart); \ |
4589 | FREE_VAR (best_regend); \ | |
fa9a63c5 RM |
4590 | } while (0) |
4591 | #else | |
0b32bf0e | 4592 | # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */ |
fa9a63c5 RM |
4593 | #endif /* not MATCH_MAY_ALLOCATE */ |
4594 | ||
505bde11 SM |
4595 | \f |
4596 | /* Optimization routines. */ | |
4597 | ||
4e8a9132 SM |
4598 | /* If the operation is a match against one or more chars, |
4599 | return a pointer to the next operation, else return NULL. */ | |
01618498 | 4600 | static re_char * |
29abe551 | 4601 | skip_one_char (const_re_char *p) |
4e8a9132 | 4602 | { |
7393bcbb | 4603 | switch (*p++) |
4e8a9132 SM |
4604 | { |
4605 | case anychar: | |
4606 | break; | |
177c0ea7 | 4607 | |
4e8a9132 SM |
4608 | case exactn: |
4609 | p += *p + 1; | |
4610 | break; | |
4611 | ||
4612 | case charset_not: | |
4613 | case charset: | |
4614 | if (CHARSET_RANGE_TABLE_EXISTS_P (p - 1)) | |
4615 | { | |
4616 | int mcnt; | |
4617 | p = CHARSET_RANGE_TABLE (p - 1); | |
4618 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4619 | p = CHARSET_RANGE_TABLE_END (p, mcnt); | |
4620 | } | |
4621 | else | |
4622 | p += 1 + CHARSET_BITMAP_SIZE (p - 1); | |
4623 | break; | |
177c0ea7 | 4624 | |
4e8a9132 SM |
4625 | case syntaxspec: |
4626 | case notsyntaxspec: | |
1fb352e0 | 4627 | #ifdef emacs |
4e8a9132 SM |
4628 | case categoryspec: |
4629 | case notcategoryspec: | |
4630 | #endif /* emacs */ | |
4631 | p++; | |
4632 | break; | |
4633 | ||
4634 | default: | |
4635 | p = NULL; | |
4636 | } | |
4637 | return p; | |
4638 | } | |
4639 | ||
4640 | ||
505bde11 | 4641 | /* Jump over non-matching operations. */ |
839966f3 | 4642 | static re_char * |
29abe551 | 4643 | skip_noops (const_re_char *p, const_re_char *pend) |
505bde11 SM |
4644 | { |
4645 | int mcnt; | |
4646 | while (p < pend) | |
4647 | { | |
7393bcbb | 4648 | switch (*p) |
505bde11 SM |
4649 | { |
4650 | case start_memory: | |
505bde11 SM |
4651 | case stop_memory: |
4652 | p += 2; break; | |
4653 | case no_op: | |
4654 | p += 1; break; | |
4655 | case jump: | |
4656 | p += 1; | |
4657 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
4658 | p += mcnt; | |
4659 | break; | |
4660 | default: | |
4661 | return p; | |
4662 | } | |
4663 | } | |
4664 | assert (p == pend); | |
4665 | return p; | |
4666 | } | |
4667 | ||
4668 | /* Non-zero if "p1 matches something" implies "p2 fails". */ | |
4669 | static int | |
29abe551 PE |
4670 | mutually_exclusive_p (struct re_pattern_buffer *bufp, const_re_char *p1, |
4671 | const_re_char *p2) | |
505bde11 | 4672 | { |
4e8a9132 | 4673 | re_opcode_t op2; |
2d1675e4 | 4674 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
505bde11 SM |
4675 | unsigned char *pend = bufp->buffer + bufp->used; |
4676 | ||
4e8a9132 | 4677 | assert (p1 >= bufp->buffer && p1 < pend |
505bde11 SM |
4678 | && p2 >= bufp->buffer && p2 <= pend); |
4679 | ||
4680 | /* Skip over open/close-group commands. | |
4681 | If what follows this loop is a ...+ construct, | |
4682 | look at what begins its body, since we will have to | |
4683 | match at least one of that. */ | |
4e8a9132 SM |
4684 | p2 = skip_noops (p2, pend); |
4685 | /* The same skip can be done for p1, except that this function | |
4686 | is only used in the case where p1 is a simple match operator. */ | |
4687 | /* p1 = skip_noops (p1, pend); */ | |
4688 | ||
4689 | assert (p1 >= bufp->buffer && p1 < pend | |
4690 | && p2 >= bufp->buffer && p2 <= pend); | |
4691 | ||
4692 | op2 = p2 == pend ? succeed : *p2; | |
4693 | ||
7393bcbb | 4694 | switch (op2) |
505bde11 | 4695 | { |
4e8a9132 SM |
4696 | case succeed: |
4697 | case endbuf: | |
4698 | /* If we're at the end of the pattern, we can change. */ | |
4699 | if (skip_one_char (p1)) | |
505bde11 | 4700 | { |
dc4a2ee0 | 4701 | DEBUG_PRINT (" End of pattern: fast loop.\n"); |
505bde11 | 4702 | return 1; |
505bde11 | 4703 | } |
4e8a9132 | 4704 | break; |
177c0ea7 | 4705 | |
4e8a9132 | 4706 | case endline: |
4e8a9132 SM |
4707 | case exactn: |
4708 | { | |
01618498 | 4709 | register re_wchar_t c |
4e8a9132 | 4710 | = (re_opcode_t) *p2 == endline ? '\n' |
62a6e103 | 4711 | : RE_STRING_CHAR (p2 + 2, multibyte); |
505bde11 | 4712 | |
4e8a9132 SM |
4713 | if ((re_opcode_t) *p1 == exactn) |
4714 | { | |
62a6e103 | 4715 | if (c != RE_STRING_CHAR (p1 + 2, multibyte)) |
4e8a9132 | 4716 | { |
dc4a2ee0 | 4717 | DEBUG_PRINT (" '%c' != '%c' => fast loop.\n", c, p1[2]); |
4e8a9132 SM |
4718 | return 1; |
4719 | } | |
4720 | } | |
505bde11 | 4721 | |
4e8a9132 SM |
4722 | else if ((re_opcode_t) *p1 == charset |
4723 | || (re_opcode_t) *p1 == charset_not) | |
4724 | { | |
4725 | int not = (re_opcode_t) *p1 == charset_not; | |
505bde11 | 4726 | |
4e8a9132 SM |
4727 | /* Test if C is listed in charset (or charset_not) |
4728 | at `p1'. */ | |
6fdd04b0 | 4729 | if (! multibyte || IS_REAL_ASCII (c)) |
4e8a9132 SM |
4730 | { |
4731 | if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH | |
4732 | && p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | |
4733 | not = !not; | |
4734 | } | |
4735 | else if (CHARSET_RANGE_TABLE_EXISTS_P (p1)) | |
4736 | CHARSET_LOOKUP_RANGE_TABLE (not, c, p1); | |
505bde11 | 4737 | |
4e8a9132 SM |
4738 | /* `not' is equal to 1 if c would match, which means |
4739 | that we can't change to pop_failure_jump. */ | |
4740 | if (!not) | |
4741 | { | |
dc4a2ee0 | 4742 | DEBUG_PRINT (" No match => fast loop.\n"); |
4e8a9132 SM |
4743 | return 1; |
4744 | } | |
4745 | } | |
4746 | else if ((re_opcode_t) *p1 == anychar | |
4747 | && c == '\n') | |
4748 | { | |
dc4a2ee0 | 4749 | DEBUG_PRINT (" . != \\n => fast loop.\n"); |
4e8a9132 SM |
4750 | return 1; |
4751 | } | |
4752 | } | |
4753 | break; | |
505bde11 | 4754 | |
4e8a9132 | 4755 | case charset: |
4e8a9132 SM |
4756 | { |
4757 | if ((re_opcode_t) *p1 == exactn) | |
4758 | /* Reuse the code above. */ | |
4759 | return mutually_exclusive_p (bufp, p2, p1); | |
505bde11 | 4760 | |
505bde11 SM |
4761 | /* It is hard to list up all the character in charset |
4762 | P2 if it includes multibyte character. Give up in | |
4763 | such case. */ | |
4764 | else if (!multibyte || !CHARSET_RANGE_TABLE_EXISTS_P (p2)) | |
4765 | { | |
4766 | /* Now, we are sure that P2 has no range table. | |
4767 | So, for the size of bitmap in P2, `p2[1]' is | |
7814e705 | 4768 | enough. But P1 may have range table, so the |
505bde11 SM |
4769 | size of bitmap table of P1 is extracted by |
4770 | using macro `CHARSET_BITMAP_SIZE'. | |
4771 | ||
6fdd04b0 KH |
4772 | In a multibyte case, we know that all the character |
4773 | listed in P2 is ASCII. In a unibyte case, P1 has only a | |
4774 | bitmap table. So, in both cases, it is enough to test | |
4775 | only the bitmap table of P1. */ | |
505bde11 | 4776 | |
411e4203 | 4777 | if ((re_opcode_t) *p1 == charset) |
505bde11 SM |
4778 | { |
4779 | int idx; | |
4780 | /* We win if the charset inside the loop | |
4781 | has no overlap with the one after the loop. */ | |
4782 | for (idx = 0; | |
4783 | (idx < (int) p2[1] | |
4784 | && idx < CHARSET_BITMAP_SIZE (p1)); | |
4785 | idx++) | |
4786 | if ((p2[2 + idx] & p1[2 + idx]) != 0) | |
4787 | break; | |
4788 | ||
4789 | if (idx == p2[1] | |
4790 | || idx == CHARSET_BITMAP_SIZE (p1)) | |
4791 | { | |
dc4a2ee0 | 4792 | DEBUG_PRINT (" No match => fast loop.\n"); |
505bde11 SM |
4793 | return 1; |
4794 | } | |
4795 | } | |
411e4203 | 4796 | else if ((re_opcode_t) *p1 == charset_not) |
505bde11 SM |
4797 | { |
4798 | int idx; | |
4799 | /* We win if the charset_not inside the loop lists | |
7814e705 | 4800 | every character listed in the charset after. */ |
505bde11 SM |
4801 | for (idx = 0; idx < (int) p2[1]; idx++) |
4802 | if (! (p2[2 + idx] == 0 | |
4803 | || (idx < CHARSET_BITMAP_SIZE (p1) | |
4804 | && ((p2[2 + idx] & ~ p1[2 + idx]) == 0)))) | |
4805 | break; | |
4806 | ||
d1dfb56c EZ |
4807 | if (idx == p2[1]) |
4808 | { | |
dc4a2ee0 | 4809 | DEBUG_PRINT (" No match => fast loop.\n"); |
d1dfb56c EZ |
4810 | return 1; |
4811 | } | |
4e8a9132 SM |
4812 | } |
4813 | } | |
4814 | } | |
609b757a | 4815 | break; |
177c0ea7 | 4816 | |
411e4203 | 4817 | case charset_not: |
7393bcbb | 4818 | switch (*p1) |
411e4203 SM |
4819 | { |
4820 | case exactn: | |
4821 | case charset: | |
4822 | /* Reuse the code above. */ | |
4823 | return mutually_exclusive_p (bufp, p2, p1); | |
4824 | case charset_not: | |
4825 | /* When we have two charset_not, it's very unlikely that | |
4826 | they don't overlap. The union of the two sets of excluded | |
4827 | chars should cover all possible chars, which, as a matter of | |
4828 | fact, is virtually impossible in multibyte buffers. */ | |
36595814 | 4829 | break; |
411e4203 SM |
4830 | } |
4831 | break; | |
4832 | ||
4e8a9132 | 4833 | case wordend: |
669fa600 SM |
4834 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == Sword); |
4835 | case symend: | |
4e8a9132 | 4836 | return ((re_opcode_t) *p1 == syntaxspec |
669fa600 SM |
4837 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4838 | case notsyntaxspec: | |
4839 | return ((re_opcode_t) *p1 == syntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4840 | |
4841 | case wordbeg: | |
669fa600 SM |
4842 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == Sword); |
4843 | case symbeg: | |
4e8a9132 | 4844 | return ((re_opcode_t) *p1 == notsyntaxspec |
669fa600 SM |
4845 | && (p1[1] == Ssymbol || p1[1] == Sword)); |
4846 | case syntaxspec: | |
4847 | return ((re_opcode_t) *p1 == notsyntaxspec && p1[1] == p2[1]); | |
4e8a9132 SM |
4848 | |
4849 | case wordbound: | |
4850 | return (((re_opcode_t) *p1 == notsyntaxspec | |
4851 | || (re_opcode_t) *p1 == syntaxspec) | |
4852 | && p1[1] == Sword); | |
4853 | ||
1fb352e0 | 4854 | #ifdef emacs |
4e8a9132 SM |
4855 | case categoryspec: |
4856 | return ((re_opcode_t) *p1 == notcategoryspec && p1[1] == p2[1]); | |
4857 | case notcategoryspec: | |
4858 | return ((re_opcode_t) *p1 == categoryspec && p1[1] == p2[1]); | |
4859 | #endif /* emacs */ | |
4860 | ||
4861 | default: | |
4862 | ; | |
505bde11 SM |
4863 | } |
4864 | ||
4865 | /* Safe default. */ | |
4866 | return 0; | |
4867 | } | |
4868 | ||
fa9a63c5 RM |
4869 | \f |
4870 | /* Matching routines. */ | |
4871 | ||
25fe55af | 4872 | #ifndef emacs /* Emacs never uses this. */ |
fa9a63c5 RM |
4873 | /* re_match is like re_match_2 except it takes only a single string. */ |
4874 | ||
d1dfb56c | 4875 | regoff_t |
d2762c86 | 4876 | re_match (struct re_pattern_buffer *bufp, const char *string, |
d1dfb56c | 4877 | size_t size, ssize_t pos, struct re_registers *regs) |
fa9a63c5 | 4878 | { |
d1dfb56c EZ |
4879 | regoff_t result = re_match_2_internal (bufp, NULL, 0, (re_char*) string, |
4880 | size, pos, regs, size); | |
fa9a63c5 RM |
4881 | return result; |
4882 | } | |
c0f9ea08 | 4883 | WEAK_ALIAS (__re_match, re_match) |
fa9a63c5 RM |
4884 | #endif /* not emacs */ |
4885 | ||
b18215fc RS |
4886 | #ifdef emacs |
4887 | /* In Emacs, this is the string or buffer in which we | |
7814e705 | 4888 | are matching. It is used for looking up syntax properties. */ |
b18215fc RS |
4889 | Lisp_Object re_match_object; |
4890 | #endif | |
fa9a63c5 RM |
4891 | |
4892 | /* re_match_2 matches the compiled pattern in BUFP against the | |
4893 | the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 | |
4894 | and SIZE2, respectively). We start matching at POS, and stop | |
4895 | matching at STOP. | |
5e69f11e | 4896 | |
fa9a63c5 | 4897 | If REGS is non-null and the `no_sub' field of BUFP is nonzero, we |
7814e705 | 4898 | store offsets for the substring each group matched in REGS. See the |
fa9a63c5 RM |
4899 | documentation for exactly how many groups we fill. |
4900 | ||
4901 | We return -1 if no match, -2 if an internal error (such as the | |
7814e705 | 4902 | failure stack overflowing). Otherwise, we return the length of the |
fa9a63c5 RM |
4903 | matched substring. */ |
4904 | ||
d1dfb56c EZ |
4905 | regoff_t |
4906 | re_match_2 (struct re_pattern_buffer *bufp, const char *string1, | |
4907 | size_t size1, const char *string2, size_t size2, ssize_t pos, | |
4908 | struct re_registers *regs, ssize_t stop) | |
fa9a63c5 | 4909 | { |
d1dfb56c | 4910 | regoff_t result; |
25fe55af | 4911 | |
b18215fc | 4912 | #ifdef emacs |
d1dfb56c | 4913 | ssize_t charpos; |
d48cd3f4 | 4914 | gl_state.object = re_match_object; /* Used by SYNTAX_TABLE_BYTE_TO_CHAR. */ |
99633e97 | 4915 | charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (pos)); |
cc9b4df2 | 4916 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); |
b18215fc RS |
4917 | #endif |
4918 | ||
4bb91c68 SM |
4919 | result = re_match_2_internal (bufp, (re_char*) string1, size1, |
4920 | (re_char*) string2, size2, | |
cc9b4df2 | 4921 | pos, regs, stop); |
fa9a63c5 RM |
4922 | return result; |
4923 | } | |
c0f9ea08 | 4924 | WEAK_ALIAS (__re_match_2, re_match_2) |
fa9a63c5 | 4925 | |
bf216479 | 4926 | |
fa9a63c5 | 4927 | /* This is a separate function so that we can force an alloca cleanup |
7814e705 | 4928 | afterwards. */ |
d1dfb56c | 4929 | static regoff_t |
29abe551 PE |
4930 | re_match_2_internal (struct re_pattern_buffer *bufp, const_re_char *string1, |
4931 | size_t size1, const_re_char *string2, size_t size2, | |
d1dfb56c | 4932 | ssize_t pos, struct re_registers *regs, ssize_t stop) |
fa9a63c5 RM |
4933 | { |
4934 | /* General temporaries. */ | |
dc4a2ee0 | 4935 | int mcnt; |
01618498 | 4936 | size_t reg; |
fa9a63c5 RM |
4937 | |
4938 | /* Just past the end of the corresponding string. */ | |
66f0296e | 4939 | re_char *end1, *end2; |
fa9a63c5 RM |
4940 | |
4941 | /* Pointers into string1 and string2, just past the last characters in | |
7814e705 | 4942 | each to consider matching. */ |
66f0296e | 4943 | re_char *end_match_1, *end_match_2; |
fa9a63c5 RM |
4944 | |
4945 | /* Where we are in the data, and the end of the current string. */ | |
66f0296e | 4946 | re_char *d, *dend; |
5e69f11e | 4947 | |
99633e97 SM |
4948 | /* Used sometimes to remember where we were before starting matching |
4949 | an operator so that we can go back in case of failure. This "atomic" | |
4950 | behavior of matching opcodes is indispensable to the correctness | |
4951 | of the on_failure_keep_string_jump optimization. */ | |
4952 | re_char *dfail; | |
4953 | ||
fa9a63c5 | 4954 | /* Where we are in the pattern, and the end of the pattern. */ |
01618498 SM |
4955 | re_char *p = bufp->buffer; |
4956 | re_char *pend = p + bufp->used; | |
fa9a63c5 | 4957 | |
25fe55af | 4958 | /* We use this to map every character in the string. */ |
6676cb1c | 4959 | RE_TRANSLATE_TYPE translate = bufp->translate; |
fa9a63c5 | 4960 | |
cf9c99bc | 4961 | /* Nonzero if BUFP is setup from a multibyte regex. */ |
2d1675e4 | 4962 | const boolean multibyte = RE_MULTIBYTE_P (bufp); |
b18215fc | 4963 | |
cf9c99bc KH |
4964 | /* Nonzero if STRING1/STRING2 are multibyte. */ |
4965 | const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp); | |
4966 | ||
fa9a63c5 RM |
4967 | /* Failure point stack. Each place that can handle a failure further |
4968 | down the line pushes a failure point on this stack. It consists of | |
505bde11 | 4969 | regstart, and regend for all registers corresponding to |
fa9a63c5 RM |
4970 | the subexpressions we're currently inside, plus the number of such |
4971 | registers, and, finally, two char *'s. The first char * is where | |
4972 | to resume scanning the pattern; the second one is where to resume | |
7814e705 JB |
4973 | scanning the strings. */ |
4974 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | |
fa9a63c5 RM |
4975 | fail_stack_type fail_stack; |
4976 | #endif | |
dc4a2ee0 | 4977 | #ifdef DEBUG_COMPILES_ARGUMENTS |
fa9a63c5 RM |
4978 | unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; |
4979 | #endif | |
4980 | ||
0b32bf0e | 4981 | #if defined REL_ALLOC && defined REGEX_MALLOC |
fa9a63c5 RM |
4982 | /* This holds the pointer to the failure stack, when |
4983 | it is allocated relocatably. */ | |
4984 | fail_stack_elt_t *failure_stack_ptr; | |
99633e97 | 4985 | #endif |
fa9a63c5 RM |
4986 | |
4987 | /* We fill all the registers internally, independent of what we | |
7814e705 | 4988 | return, for use in backreferences. The number here includes |
fa9a63c5 | 4989 | an element for register zero. */ |
4bb91c68 | 4990 | size_t num_regs = bufp->re_nsub + 1; |
5e69f11e | 4991 | |
fa9a63c5 RM |
4992 | /* Information on the contents of registers. These are pointers into |
4993 | the input strings; they record just what was matched (on this | |
4994 | attempt) by a subexpression part of the pattern, that is, the | |
4995 | regnum-th regstart pointer points to where in the pattern we began | |
4996 | matching and the regnum-th regend points to right after where we | |
4997 | stopped matching the regnum-th subexpression. (The zeroth register | |
4998 | keeps track of what the whole pattern matches.) */ | |
4999 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5000 | re_char **regstart, **regend; |
fa9a63c5 RM |
5001 | #endif |
5002 | ||
fa9a63c5 | 5003 | /* The following record the register info as found in the above |
5e69f11e | 5004 | variables when we find a match better than any we've seen before. |
fa9a63c5 RM |
5005 | This happens as we backtrack through the failure points, which in |
5006 | turn happens only if we have not yet matched the entire string. */ | |
5007 | unsigned best_regs_set = false; | |
5008 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | |
66f0296e | 5009 | re_char **best_regstart, **best_regend; |
fa9a63c5 | 5010 | #endif |
5e69f11e | 5011 | |
fa9a63c5 RM |
5012 | /* Logically, this is `best_regend[0]'. But we don't want to have to |
5013 | allocate space for that if we're not allocating space for anything | |
7814e705 | 5014 | else (see below). Also, we never need info about register 0 for |
fa9a63c5 RM |
5015 | any of the other register vectors, and it seems rather a kludge to |
5016 | treat `best_regend' differently than the rest. So we keep track of | |
5017 | the end of the best match so far in a separate variable. We | |
5018 | initialize this to NULL so that when we backtrack the first time | |
5019 | and need to test it, it's not garbage. */ | |
66f0296e | 5020 | re_char *match_end = NULL; |
fa9a63c5 | 5021 | |
dc4a2ee0 | 5022 | #ifdef DEBUG_COMPILES_ARGUMENTS |
fa9a63c5 | 5023 | /* Counts the total number of registers pushed. */ |
5e69f11e | 5024 | unsigned num_regs_pushed = 0; |
fa9a63c5 RM |
5025 | #endif |
5026 | ||
dc4a2ee0 | 5027 | DEBUG_PRINT ("\n\nEntering re_match_2.\n"); |
5e69f11e | 5028 | |
fa9a63c5 | 5029 | INIT_FAIL_STACK (); |
5e69f11e | 5030 | |
fa9a63c5 RM |
5031 | #ifdef MATCH_MAY_ALLOCATE |
5032 | /* Do not bother to initialize all the register variables if there are | |
5033 | no groups in the pattern, as it takes a fair amount of time. If | |
5034 | there are groups, we include space for register 0 (the whole | |
5035 | pattern), even though we never use it, since it simplifies the | |
5036 | array indexing. We should fix this. */ | |
5037 | if (bufp->re_nsub) | |
5038 | { | |
66f0296e SM |
5039 | regstart = REGEX_TALLOC (num_regs, re_char *); |
5040 | regend = REGEX_TALLOC (num_regs, re_char *); | |
5041 | best_regstart = REGEX_TALLOC (num_regs, re_char *); | |
5042 | best_regend = REGEX_TALLOC (num_regs, re_char *); | |
fa9a63c5 | 5043 | |
505bde11 | 5044 | if (!(regstart && regend && best_regstart && best_regend)) |
25fe55af RS |
5045 | { |
5046 | FREE_VARIABLES (); | |
5047 | return -2; | |
5048 | } | |
fa9a63c5 RM |
5049 | } |
5050 | else | |
5051 | { | |
5052 | /* We must initialize all our variables to NULL, so that | |
25fe55af | 5053 | `FREE_VARIABLES' doesn't try to free them. */ |
505bde11 | 5054 | regstart = regend = best_regstart = best_regend = NULL; |
fa9a63c5 RM |
5055 | } |
5056 | #endif /* MATCH_MAY_ALLOCATE */ | |
5057 | ||
5058 | /* The starting position is bogus. */ | |
5059 | if (pos < 0 || pos > size1 + size2) | |
5060 | { | |
5061 | FREE_VARIABLES (); | |
5062 | return -1; | |
5063 | } | |
5e69f11e | 5064 | |
fa9a63c5 RM |
5065 | /* Initialize subexpression text positions to -1 to mark ones that no |
5066 | start_memory/stop_memory has been seen for. Also initialize the | |
5067 | register information struct. */ | |
01618498 SM |
5068 | for (reg = 1; reg < num_regs; reg++) |
5069 | regstart[reg] = regend[reg] = NULL; | |
99633e97 | 5070 | |
fa9a63c5 | 5071 | /* We move `string1' into `string2' if the latter's empty -- but not if |
7814e705 | 5072 | `string1' is null. */ |
fa9a63c5 RM |
5073 | if (size2 == 0 && string1 != NULL) |
5074 | { | |
5075 | string2 = string1; | |
5076 | size2 = size1; | |
5077 | string1 = 0; | |
5078 | size1 = 0; | |
5079 | } | |
5080 | end1 = string1 + size1; | |
5081 | end2 = string2 + size2; | |
5082 | ||
5e69f11e | 5083 | /* `p' scans through the pattern as `d' scans through the data. |
fa9a63c5 RM |
5084 | `dend' is the end of the input string that `d' points within. `d' |
5085 | is advanced into the following input string whenever necessary, but | |
5086 | this happens before fetching; therefore, at the beginning of the | |
5087 | loop, `d' can be pointing at the end of a string, but it cannot | |
5088 | equal `string2'. */ | |
419d1c74 | 5089 | if (pos >= size1) |
fa9a63c5 | 5090 | { |
419d1c74 SM |
5091 | /* Only match within string2. */ |
5092 | d = string2 + pos - size1; | |
5093 | dend = end_match_2 = string2 + stop - size1; | |
5094 | end_match_1 = end1; /* Just to give it a value. */ | |
fa9a63c5 RM |
5095 | } |
5096 | else | |
5097 | { | |
f1ad044f | 5098 | if (stop < size1) |
419d1c74 SM |
5099 | { |
5100 | /* Only match within string1. */ | |
5101 | end_match_1 = string1 + stop; | |
5102 | /* BEWARE! | |
5103 | When we reach end_match_1, PREFETCH normally switches to string2. | |
5104 | But in the present case, this means that just doing a PREFETCH | |
5105 | makes us jump from `stop' to `gap' within the string. | |
5106 | What we really want here is for the search to stop as | |
5107 | soon as we hit end_match_1. That's why we set end_match_2 | |
5108 | to end_match_1 (since PREFETCH fails as soon as we hit | |
5109 | end_match_2). */ | |
5110 | end_match_2 = end_match_1; | |
5111 | } | |
5112 | else | |
f1ad044f SM |
5113 | { /* It's important to use this code when stop == size so that |
5114 | moving `d' from end1 to string2 will not prevent the d == dend | |
5115 | check from catching the end of string. */ | |
419d1c74 SM |
5116 | end_match_1 = end1; |
5117 | end_match_2 = string2 + stop - size1; | |
5118 | } | |
5119 | d = string1 + pos; | |
5120 | dend = end_match_1; | |
fa9a63c5 RM |
5121 | } |
5122 | ||
dc4a2ee0 | 5123 | DEBUG_PRINT ("The compiled pattern is: "); |
fa9a63c5 | 5124 | DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); |
dc4a2ee0 | 5125 | DEBUG_PRINT ("The string to match is: `"); |
fa9a63c5 | 5126 | DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); |
dc4a2ee0 | 5127 | DEBUG_PRINT ("'\n"); |
5e69f11e | 5128 | |
7814e705 | 5129 | /* This loops over pattern commands. It exits by returning from the |
fa9a63c5 RM |
5130 | function if the match is complete, or it drops through if the match |
5131 | fails at this starting point in the input data. */ | |
5132 | for (;;) | |
5133 | { | |
dc4a2ee0 | 5134 | DEBUG_PRINT ("\n%p: ", p); |
fa9a63c5 RM |
5135 | |
5136 | if (p == pend) | |
dc4a2ee0 PE |
5137 | { |
5138 | ptrdiff_t dcnt; | |
5139 | ||
5140 | /* End of pattern means we might have succeeded. */ | |
5141 | DEBUG_PRINT ("end of pattern ... "); | |
5e69f11e | 5142 | |
fa9a63c5 | 5143 | /* If we haven't matched the entire string, and we want the |
25fe55af RS |
5144 | longest match, try backtracking. */ |
5145 | if (d != end_match_2) | |
fa9a63c5 RM |
5146 | { |
5147 | /* 1 if this match ends in the same string (string1 or string2) | |
5148 | as the best previous match. */ | |
d42f4f0f PE |
5149 | boolean same_str_p = (FIRST_STRING_P (match_end) |
5150 | == FIRST_STRING_P (d)); | |
fa9a63c5 RM |
5151 | /* 1 if this match is the best seen so far. */ |
5152 | boolean best_match_p; | |
5153 | ||
5154 | /* AIX compiler got confused when this was combined | |
7814e705 | 5155 | with the previous declaration. */ |
fa9a63c5 RM |
5156 | if (same_str_p) |
5157 | best_match_p = d > match_end; | |
5158 | else | |
99633e97 | 5159 | best_match_p = !FIRST_STRING_P (d); |
fa9a63c5 | 5160 | |
dc4a2ee0 | 5161 | DEBUG_PRINT ("backtracking.\n"); |
25fe55af RS |
5162 | |
5163 | if (!FAIL_STACK_EMPTY ()) | |
5164 | { /* More failure points to try. */ | |
5165 | ||
5166 | /* If exceeds best match so far, save it. */ | |
5167 | if (!best_regs_set || best_match_p) | |
5168 | { | |
5169 | best_regs_set = true; | |
5170 | match_end = d; | |
5171 | ||
dc4a2ee0 | 5172 | DEBUG_PRINT ("\nSAVING match as best so far.\n"); |
25fe55af | 5173 | |
01618498 | 5174 | for (reg = 1; reg < num_regs; reg++) |
25fe55af | 5175 | { |
01618498 SM |
5176 | best_regstart[reg] = regstart[reg]; |
5177 | best_regend[reg] = regend[reg]; | |
25fe55af RS |
5178 | } |
5179 | } | |
5180 | goto fail; | |
5181 | } | |
5182 | ||
5183 | /* If no failure points, don't restore garbage. And if | |
5184 | last match is real best match, don't restore second | |
5185 | best one. */ | |
5186 | else if (best_regs_set && !best_match_p) | |
5187 | { | |
5188 | restore_best_regs: | |
5189 | /* Restore best match. It may happen that `dend == | |
5190 | end_match_1' while the restored d is in string2. | |
5191 | For example, the pattern `x.*y.*z' against the | |
5192 | strings `x-' and `y-z-', if the two strings are | |
7814e705 | 5193 | not consecutive in memory. */ |
dc4a2ee0 | 5194 | DEBUG_PRINT ("Restoring best registers.\n"); |
25fe55af RS |
5195 | |
5196 | d = match_end; | |
5197 | dend = ((d >= string1 && d <= end1) | |
5198 | ? end_match_1 : end_match_2); | |
fa9a63c5 | 5199 | |
01618498 | 5200 | for (reg = 1; reg < num_regs; reg++) |
fa9a63c5 | 5201 | { |
01618498 SM |
5202 | regstart[reg] = best_regstart[reg]; |
5203 | regend[reg] = best_regend[reg]; | |
fa9a63c5 | 5204 | } |
25fe55af RS |
5205 | } |
5206 | } /* d != end_match_2 */ | |
fa9a63c5 RM |
5207 | |
5208 | succeed_label: | |
dc4a2ee0 | 5209 | DEBUG_PRINT ("Accepting match.\n"); |
fa9a63c5 | 5210 | |
25fe55af RS |
5211 | /* If caller wants register contents data back, do it. */ |
5212 | if (regs && !bufp->no_sub) | |
fa9a63c5 | 5213 | { |
25fe55af RS |
5214 | /* Have the register data arrays been allocated? */ |
5215 | if (bufp->regs_allocated == REGS_UNALLOCATED) | |
7814e705 | 5216 | { /* No. So allocate them with malloc. We need one |
25fe55af RS |
5217 | extra element beyond `num_regs' for the `-1' marker |
5218 | GNU code uses. */ | |
5219 | regs->num_regs = MAX (RE_NREGS, num_regs + 1); | |
5220 | regs->start = TALLOC (regs->num_regs, regoff_t); | |
5221 | regs->end = TALLOC (regs->num_regs, regoff_t); | |
5222 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5223 | { |
5224 | FREE_VARIABLES (); | |
5225 | return -2; | |
5226 | } | |
25fe55af RS |
5227 | bufp->regs_allocated = REGS_REALLOCATE; |
5228 | } | |
5229 | else if (bufp->regs_allocated == REGS_REALLOCATE) | |
5230 | { /* Yes. If we need more elements than were already | |
5231 | allocated, reallocate them. If we need fewer, just | |
5232 | leave it alone. */ | |
5233 | if (regs->num_regs < num_regs + 1) | |
5234 | { | |
5235 | regs->num_regs = num_regs + 1; | |
5236 | RETALLOC (regs->start, regs->num_regs, regoff_t); | |
5237 | RETALLOC (regs->end, regs->num_regs, regoff_t); | |
5238 | if (regs->start == NULL || regs->end == NULL) | |
fa9a63c5 RM |
5239 | { |
5240 | FREE_VARIABLES (); | |
5241 | return -2; | |
5242 | } | |
25fe55af RS |
5243 | } |
5244 | } | |
5245 | else | |
fa9a63c5 RM |
5246 | { |
5247 | /* These braces fend off a "empty body in an else-statement" | |
7814e705 | 5248 | warning under GCC when assert expands to nothing. */ |
fa9a63c5 RM |
5249 | assert (bufp->regs_allocated == REGS_FIXED); |
5250 | } | |
5251 | ||
25fe55af RS |
5252 | /* Convert the pointer data in `regstart' and `regend' to |
5253 | indices. Register zero has to be set differently, | |
5254 | since we haven't kept track of any info for it. */ | |
5255 | if (regs->num_regs > 0) | |
5256 | { | |
5257 | regs->start[0] = pos; | |
99633e97 | 5258 | regs->end[0] = POINTER_TO_OFFSET (d); |
25fe55af | 5259 | } |
5e69f11e | 5260 | |
25fe55af RS |
5261 | /* Go through the first `min (num_regs, regs->num_regs)' |
5262 | registers, since that is all we initialized. */ | |
01618498 | 5263 | for (reg = 1; reg < MIN (num_regs, regs->num_regs); reg++) |
fa9a63c5 | 5264 | { |
01618498 SM |
5265 | if (REG_UNSET (regstart[reg]) || REG_UNSET (regend[reg])) |
5266 | regs->start[reg] = regs->end[reg] = -1; | |
25fe55af RS |
5267 | else |
5268 | { | |
dc4a2ee0 PE |
5269 | regs->start[reg] = POINTER_TO_OFFSET (regstart[reg]); |
5270 | regs->end[reg] = POINTER_TO_OFFSET (regend[reg]); | |
25fe55af | 5271 | } |
fa9a63c5 | 5272 | } |
5e69f11e | 5273 | |
25fe55af RS |
5274 | /* If the regs structure we return has more elements than |
5275 | were in the pattern, set the extra elements to -1. If | |
5276 | we (re)allocated the registers, this is the case, | |
5277 | because we always allocate enough to have at least one | |
7814e705 | 5278 | -1 at the end. */ |
01618498 SM |
5279 | for (reg = num_regs; reg < regs->num_regs; reg++) |
5280 | regs->start[reg] = regs->end[reg] = -1; | |
fa9a63c5 RM |
5281 | } /* regs && !bufp->no_sub */ |
5282 | ||
dc4a2ee0 PE |
5283 | DEBUG_PRINT ("%u failure points pushed, %u popped (%u remain).\n", |
5284 | nfailure_points_pushed, nfailure_points_popped, | |
5285 | nfailure_points_pushed - nfailure_points_popped); | |
5286 | DEBUG_PRINT ("%u registers pushed.\n", num_regs_pushed); | |
fa9a63c5 | 5287 | |
dc4a2ee0 | 5288 | dcnt = POINTER_TO_OFFSET (d) - pos; |
fa9a63c5 | 5289 | |
dc4a2ee0 | 5290 | DEBUG_PRINT ("Returning %td from re_match_2.\n", dcnt); |
fa9a63c5 | 5291 | |
25fe55af | 5292 | FREE_VARIABLES (); |
dc4a2ee0 | 5293 | return dcnt; |
25fe55af | 5294 | } |
fa9a63c5 | 5295 | |
7814e705 | 5296 | /* Otherwise match next pattern command. */ |
7393bcbb | 5297 | switch (*p++) |
fa9a63c5 | 5298 | { |
25fe55af RS |
5299 | /* Ignore these. Used to ignore the n of succeed_n's which |
5300 | currently have n == 0. */ | |
5301 | case no_op: | |
dc4a2ee0 | 5302 | DEBUG_PRINT ("EXECUTING no_op.\n"); |
25fe55af | 5303 | break; |
fa9a63c5 RM |
5304 | |
5305 | case succeed: | |
dc4a2ee0 | 5306 | DEBUG_PRINT ("EXECUTING succeed.\n"); |
fa9a63c5 RM |
5307 | goto succeed_label; |
5308 | ||
7814e705 | 5309 | /* Match the next n pattern characters exactly. The following |
25fe55af | 5310 | byte in the pattern defines n, and the n bytes after that |
7814e705 | 5311 | are the characters to match. */ |
fa9a63c5 RM |
5312 | case exactn: |
5313 | mcnt = *p++; | |
dc4a2ee0 | 5314 | DEBUG_PRINT ("EXECUTING exactn %d.\n", mcnt); |
fa9a63c5 | 5315 | |
99633e97 SM |
5316 | /* Remember the start point to rollback upon failure. */ |
5317 | dfail = d; | |
5318 | ||
6fdd04b0 | 5319 | #ifndef emacs |
25fe55af RS |
5320 | /* This is written out as an if-else so we don't waste time |
5321 | testing `translate' inside the loop. */ | |
28703c16 | 5322 | if (RE_TRANSLATE_P (translate)) |
6fdd04b0 KH |
5323 | do |
5324 | { | |
5325 | PREFETCH (); | |
5326 | if (RE_TRANSLATE (translate, *d) != *p++) | |
e934739e | 5327 | { |
6fdd04b0 KH |
5328 | d = dfail; |
5329 | goto fail; | |
e934739e | 5330 | } |
6fdd04b0 KH |
5331 | d++; |
5332 | } | |
5333 | while (--mcnt); | |
fa9a63c5 | 5334 | else |
6fdd04b0 KH |
5335 | do |
5336 | { | |
5337 | PREFETCH (); | |
5338 | if (*d++ != *p++) | |
bf216479 | 5339 | { |
6fdd04b0 KH |
5340 | d = dfail; |
5341 | goto fail; | |
bf216479 | 5342 | } |
6fdd04b0 KH |
5343 | } |
5344 | while (--mcnt); | |
5345 | #else /* emacs */ | |
5346 | /* The cost of testing `translate' is comparatively small. */ | |
cf9c99bc | 5347 | if (target_multibyte) |
6fdd04b0 KH |
5348 | do |
5349 | { | |
5350 | int pat_charlen, buf_charlen; | |
cf9c99bc | 5351 | int pat_ch, buf_ch; |
e934739e | 5352 | |
6fdd04b0 | 5353 | PREFETCH (); |
cf9c99bc | 5354 | if (multibyte) |
62a6e103 | 5355 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
cf9c99bc KH |
5356 | else |
5357 | { | |
5358 | pat_ch = RE_CHAR_TO_MULTIBYTE (*p); | |
5359 | pat_charlen = 1; | |
5360 | } | |
62a6e103 | 5361 | buf_ch = STRING_CHAR_AND_LENGTH (d, buf_charlen); |
e934739e | 5362 | |
6fdd04b0 | 5363 | if (TRANSLATE (buf_ch) != pat_ch) |
e934739e | 5364 | { |
6fdd04b0 KH |
5365 | d = dfail; |
5366 | goto fail; | |
e934739e | 5367 | } |
bf216479 | 5368 | |
6fdd04b0 KH |
5369 | p += pat_charlen; |
5370 | d += buf_charlen; | |
5371 | mcnt -= pat_charlen; | |
5372 | } | |
5373 | while (mcnt > 0); | |
fa9a63c5 | 5374 | else |
6fdd04b0 KH |
5375 | do |
5376 | { | |
abbd1bcf | 5377 | int pat_charlen; |
cf9c99bc | 5378 | int pat_ch, buf_ch; |
bf216479 | 5379 | |
6fdd04b0 | 5380 | PREFETCH (); |
cf9c99bc KH |
5381 | if (multibyte) |
5382 | { | |
62a6e103 | 5383 | pat_ch = STRING_CHAR_AND_LENGTH (p, pat_charlen); |
2afc21f5 | 5384 | pat_ch = RE_CHAR_TO_UNIBYTE (pat_ch); |
cf9c99bc KH |
5385 | } |
5386 | else | |
5387 | { | |
5388 | pat_ch = *p; | |
5389 | pat_charlen = 1; | |
5390 | } | |
5391 | buf_ch = RE_CHAR_TO_MULTIBYTE (*d); | |
5392 | if (! CHAR_BYTE8_P (buf_ch)) | |
5393 | { | |
5394 | buf_ch = TRANSLATE (buf_ch); | |
5395 | buf_ch = RE_CHAR_TO_UNIBYTE (buf_ch); | |
5396 | if (buf_ch < 0) | |
5397 | buf_ch = *d; | |
5398 | } | |
0e2501ed AS |
5399 | else |
5400 | buf_ch = *d; | |
cf9c99bc | 5401 | if (buf_ch != pat_ch) |
6fdd04b0 KH |
5402 | { |
5403 | d = dfail; | |
5404 | goto fail; | |
bf216479 | 5405 | } |
cf9c99bc KH |
5406 | p += pat_charlen; |
5407 | d++; | |
6fdd04b0 KH |
5408 | } |
5409 | while (--mcnt); | |
5410 | #endif | |
25fe55af | 5411 | break; |
fa9a63c5 RM |
5412 | |
5413 | ||
25fe55af | 5414 | /* Match any character except possibly a newline or a null. */ |
fa9a63c5 | 5415 | case anychar: |
e934739e RS |
5416 | { |
5417 | int buf_charlen; | |
01618498 | 5418 | re_wchar_t buf_ch; |
fa9a63c5 | 5419 | |
dc4a2ee0 | 5420 | DEBUG_PRINT ("EXECUTING anychar.\n"); |
fa9a63c5 | 5421 | |
e934739e | 5422 | PREFETCH (); |
62a6e103 | 5423 | buf_ch = RE_STRING_CHAR_AND_LENGTH (d, buf_charlen, |
cf9c99bc | 5424 | target_multibyte); |
e934739e RS |
5425 | buf_ch = TRANSLATE (buf_ch); |
5426 | ||
5427 | if ((!(bufp->syntax & RE_DOT_NEWLINE) | |
5428 | && buf_ch == '\n') | |
5429 | || ((bufp->syntax & RE_DOT_NOT_NULL) | |
5430 | && buf_ch == '\000')) | |
5431 | goto fail; | |
5432 | ||
dc4a2ee0 | 5433 | DEBUG_PRINT (" Matched `%d'.\n", *d); |
e934739e RS |
5434 | d += buf_charlen; |
5435 | } | |
fa9a63c5 RM |
5436 | break; |
5437 | ||
5438 | ||
5439 | case charset: | |
5440 | case charset_not: | |
5441 | { | |
b18215fc | 5442 | register unsigned int c; |
fa9a63c5 | 5443 | boolean not = (re_opcode_t) *(p - 1) == charset_not; |
b18215fc RS |
5444 | int len; |
5445 | ||
5446 | /* Start of actual range_table, or end of bitmap if there is no | |
5447 | range table. */ | |
da053e48 | 5448 | re_char *range_table IF_LINT (= NULL); |
b18215fc | 5449 | |
96cc36cc | 5450 | /* Nonzero if there is a range table. */ |
b18215fc RS |
5451 | int range_table_exists; |
5452 | ||
96cc36cc RS |
5453 | /* Number of ranges of range table. This is not included |
5454 | in the initial byte-length of the command. */ | |
5455 | int count = 0; | |
fa9a63c5 | 5456 | |
f5020181 AS |
5457 | /* Whether matching against a unibyte character. */ |
5458 | boolean unibyte_char = false; | |
5459 | ||
dc4a2ee0 | 5460 | DEBUG_PRINT ("EXECUTING charset%s.\n", not ? "_not" : ""); |
fa9a63c5 | 5461 | |
b18215fc | 5462 | range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]); |
96cc36cc | 5463 | |
b18215fc | 5464 | if (range_table_exists) |
96cc36cc RS |
5465 | { |
5466 | range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap. */ | |
5467 | EXTRACT_NUMBER_AND_INCR (count, range_table); | |
5468 | } | |
b18215fc | 5469 | |
2d1675e4 | 5470 | PREFETCH (); |
62a6e103 | 5471 | c = RE_STRING_CHAR_AND_LENGTH (d, len, target_multibyte); |
cf9c99bc KH |
5472 | if (target_multibyte) |
5473 | { | |
5474 | int c1; | |
b18215fc | 5475 | |
cf9c99bc KH |
5476 | c = TRANSLATE (c); |
5477 | c1 = RE_CHAR_TO_UNIBYTE (c); | |
5478 | if (c1 >= 0) | |
f5020181 AS |
5479 | { |
5480 | unibyte_char = true; | |
5481 | c = c1; | |
5482 | } | |
cf9c99bc KH |
5483 | } |
5484 | else | |
5485 | { | |
5486 | int c1 = RE_CHAR_TO_MULTIBYTE (c); | |
5487 | ||
5488 | if (! CHAR_BYTE8_P (c1)) | |
5489 | { | |
5490 | c1 = TRANSLATE (c1); | |
5491 | c1 = RE_CHAR_TO_UNIBYTE (c1); | |
5492 | if (c1 >= 0) | |
f5020181 AS |
5493 | { |
5494 | unibyte_char = true; | |
5495 | c = c1; | |
5496 | } | |
cf9c99bc | 5497 | } |
0b8be006 AS |
5498 | else |
5499 | unibyte_char = true; | |
cf9c99bc KH |
5500 | } |
5501 | ||
f5020181 | 5502 | if (unibyte_char && c < (1 << BYTEWIDTH)) |
b18215fc | 5503 | { /* Lookup bitmap. */ |
b18215fc RS |
5504 | /* Cast to `unsigned' instead of `unsigned char' in |
5505 | case the bit list is a full 32 bytes long. */ | |
5506 | if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH) | |
96cc36cc RS |
5507 | && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) |
5508 | not = !not; | |
b18215fc | 5509 | } |
96cc36cc | 5510 | #ifdef emacs |
b18215fc | 5511 | else if (range_table_exists) |
96cc36cc RS |
5512 | { |
5513 | int class_bits = CHARSET_RANGE_TABLE_BITS (&p[-1]); | |
5514 | ||
14473664 SM |
5515 | if ( (class_bits & BIT_LOWER && ISLOWER (c)) |
5516 | | (class_bits & BIT_MULTIBYTE) | |
96cc36cc RS |
5517 | | (class_bits & BIT_PUNCT && ISPUNCT (c)) |
5518 | | (class_bits & BIT_SPACE && ISSPACE (c)) | |
5519 | | (class_bits & BIT_UPPER && ISUPPER (c)) | |
5520 | | (class_bits & BIT_WORD && ISWORD (c))) | |
5521 | not = !not; | |
5522 | else | |
5523 | CHARSET_LOOKUP_RANGE_TABLE_RAW (not, c, range_table, count); | |
5524 | } | |
5525 | #endif /* emacs */ | |
fa9a63c5 | 5526 | |
96cc36cc RS |
5527 | if (range_table_exists) |
5528 | p = CHARSET_RANGE_TABLE_END (range_table, count); | |
5529 | else | |
5530 | p += CHARSET_BITMAP_SIZE (&p[-1]) + 1; | |
fa9a63c5 RM |
5531 | |
5532 | if (!not) goto fail; | |
5e69f11e | 5533 | |
b18215fc | 5534 | d += len; |
fa9a63c5 | 5535 | } |
8fb31792 | 5536 | break; |
fa9a63c5 RM |
5537 | |
5538 | ||
25fe55af | 5539 | /* The beginning of a group is represented by start_memory. |
505bde11 | 5540 | The argument is the register number. The text |
25fe55af | 5541 | matched within the group is recorded (in the internal |
7814e705 | 5542 | registers data structure) under the register number. */ |
25fe55af | 5543 | case start_memory: |
dc4a2ee0 | 5544 | DEBUG_PRINT ("EXECUTING start_memory %d:\n", *p); |
505bde11 SM |
5545 | |
5546 | /* In case we need to undo this operation (via backtracking). */ | |
dc4a2ee0 | 5547 | PUSH_FAILURE_REG (*p); |
fa9a63c5 | 5548 | |
25fe55af | 5549 | regstart[*p] = d; |
4bb91c68 | 5550 | regend[*p] = NULL; /* probably unnecessary. -sm */ |
dc4a2ee0 | 5551 | DEBUG_PRINT (" regstart: %td\n", POINTER_TO_OFFSET (regstart[*p])); |
fa9a63c5 | 5552 | |
25fe55af | 5553 | /* Move past the register number and inner group count. */ |
505bde11 | 5554 | p += 1; |
25fe55af | 5555 | break; |
fa9a63c5 RM |
5556 | |
5557 | ||
25fe55af | 5558 | /* The stop_memory opcode represents the end of a group. Its |
505bde11 | 5559 | argument is the same as start_memory's: the register number. */ |
fa9a63c5 | 5560 | case stop_memory: |
dc4a2ee0 | 5561 | DEBUG_PRINT ("EXECUTING stop_memory %d:\n", *p); |
505bde11 SM |
5562 | |
5563 | assert (!REG_UNSET (regstart[*p])); | |
5564 | /* Strictly speaking, there should be code such as: | |
177c0ea7 | 5565 | |
0b32bf0e | 5566 | assert (REG_UNSET (regend[*p])); |
505bde11 SM |
5567 | PUSH_FAILURE_REGSTOP ((unsigned int)*p); |
5568 | ||
5569 | But the only info to be pushed is regend[*p] and it is known to | |
5570 | be UNSET, so there really isn't anything to push. | |
5571 | Not pushing anything, on the other hand deprives us from the | |
5572 | guarantee that regend[*p] is UNSET since undoing this operation | |
5573 | will not reset its value properly. This is not important since | |
5574 | the value will only be read on the next start_memory or at | |
5575 | the very end and both events can only happen if this stop_memory | |
5576 | is *not* undone. */ | |
fa9a63c5 | 5577 | |
25fe55af | 5578 | regend[*p] = d; |
dc4a2ee0 | 5579 | DEBUG_PRINT (" regend: %td\n", POINTER_TO_OFFSET (regend[*p])); |
fa9a63c5 | 5580 | |
25fe55af | 5581 | /* Move past the register number and the inner group count. */ |
505bde11 | 5582 | p += 1; |
25fe55af | 5583 | break; |
fa9a63c5 RM |
5584 | |
5585 | ||
5586 | /* \<digit> has been turned into a `duplicate' command which is | |
25fe55af RS |
5587 | followed by the numeric value of <digit> as the register number. */ |
5588 | case duplicate: | |
fa9a63c5 | 5589 | { |
66f0296e | 5590 | register re_char *d2, *dend2; |
7814e705 | 5591 | int regno = *p++; /* Get which register to match against. */ |
dc4a2ee0 | 5592 | DEBUG_PRINT ("EXECUTING duplicate %d.\n", regno); |
fa9a63c5 | 5593 | |
7814e705 | 5594 | /* Can't back reference a group which we've never matched. */ |
25fe55af RS |
5595 | if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) |
5596 | goto fail; | |
5e69f11e | 5597 | |
7814e705 | 5598 | /* Where in input to try to start matching. */ |
25fe55af | 5599 | d2 = regstart[regno]; |
5e69f11e | 5600 | |
99633e97 SM |
5601 | /* Remember the start point to rollback upon failure. */ |
5602 | dfail = d; | |
5603 | ||
25fe55af RS |
5604 | /* Where to stop matching; if both the place to start and |
5605 | the place to stop matching are in the same string, then | |
5606 | set to the place to stop, otherwise, for now have to use | |
5607 | the end of the first string. */ | |
fa9a63c5 | 5608 | |
25fe55af | 5609 | dend2 = ((FIRST_STRING_P (regstart[regno]) |
fa9a63c5 RM |
5610 | == FIRST_STRING_P (regend[regno])) |
5611 | ? regend[regno] : end_match_1); | |
5612 | for (;;) | |
5613 | { | |
dc4a2ee0 PE |
5614 | ptrdiff_t dcnt; |
5615 | ||
fa9a63c5 | 5616 | /* If necessary, advance to next segment in register |
25fe55af | 5617 | contents. */ |
fa9a63c5 RM |
5618 | while (d2 == dend2) |
5619 | { | |
5620 | if (dend2 == end_match_2) break; | |
5621 | if (dend2 == regend[regno]) break; | |
5622 | ||
25fe55af RS |
5623 | /* End of string1 => advance to string2. */ |
5624 | d2 = string2; | |
5625 | dend2 = regend[regno]; | |
fa9a63c5 RM |
5626 | } |
5627 | /* At end of register contents => success */ | |
5628 | if (d2 == dend2) break; | |
5629 | ||
5630 | /* If necessary, advance to next segment in data. */ | |
5631 | PREFETCH (); | |
5632 | ||
5633 | /* How many characters left in this segment to match. */ | |
dc4a2ee0 | 5634 | dcnt = dend - d; |
5e69f11e | 5635 | |
fa9a63c5 | 5636 | /* Want how many consecutive characters we can match in |
25fe55af | 5637 | one shot, so, if necessary, adjust the count. */ |
dc4a2ee0 PE |
5638 | if (dcnt > dend2 - d2) |
5639 | dcnt = dend2 - d2; | |
5e69f11e | 5640 | |
fa9a63c5 | 5641 | /* Compare that many; failure if mismatch, else move |
25fe55af | 5642 | past them. */ |
28703c16 | 5643 | if (RE_TRANSLATE_P (translate) |
dc4a2ee0 PE |
5644 | ? bcmp_translate (d, d2, dcnt, translate, target_multibyte) |
5645 | : memcmp (d, d2, dcnt)) | |
99633e97 SM |
5646 | { |
5647 | d = dfail; | |
5648 | goto fail; | |
5649 | } | |
dc4a2ee0 | 5650 | d += dcnt, d2 += dcnt; |
fa9a63c5 RM |
5651 | } |
5652 | } | |
5653 | break; | |
5654 | ||
5655 | ||
25fe55af | 5656 | /* begline matches the empty string at the beginning of the string |
c0f9ea08 | 5657 | (unless `not_bol' is set in `bufp'), and after newlines. */ |
fa9a63c5 | 5658 | case begline: |
dc4a2ee0 | 5659 | DEBUG_PRINT ("EXECUTING begline.\n"); |
5e69f11e | 5660 | |
25fe55af RS |
5661 | if (AT_STRINGS_BEG (d)) |
5662 | { | |
5663 | if (!bufp->not_bol) break; | |
5664 | } | |
419d1c74 | 5665 | else |
25fe55af | 5666 | { |
bf216479 | 5667 | unsigned c; |
419d1c74 | 5668 | GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2); |
c0f9ea08 | 5669 | if (c == '\n') |
419d1c74 | 5670 | break; |
25fe55af RS |
5671 | } |
5672 | /* In all other cases, we fail. */ | |
5673 | goto fail; | |
fa9a63c5 RM |
5674 | |
5675 | ||
25fe55af | 5676 | /* endline is the dual of begline. */ |
fa9a63c5 | 5677 | case endline: |
dc4a2ee0 | 5678 | DEBUG_PRINT ("EXECUTING endline.\n"); |
fa9a63c5 | 5679 | |
25fe55af RS |
5680 | if (AT_STRINGS_END (d)) |
5681 | { | |
5682 | if (!bufp->not_eol) break; | |
5683 | } | |
f1ad044f | 5684 | else |
25fe55af | 5685 | { |
f1ad044f | 5686 | PREFETCH_NOLIMIT (); |
c0f9ea08 | 5687 | if (*d == '\n') |
f1ad044f | 5688 | break; |
25fe55af RS |
5689 | } |
5690 | goto fail; | |
fa9a63c5 RM |
5691 | |
5692 | ||
5693 | /* Match at the very beginning of the data. */ | |
25fe55af | 5694 | case begbuf: |
dc4a2ee0 | 5695 | DEBUG_PRINT ("EXECUTING begbuf.\n"); |
25fe55af RS |
5696 | if (AT_STRINGS_BEG (d)) |
5697 | break; | |
5698 | goto fail; | |
fa9a63c5 RM |
5699 | |
5700 | ||
5701 | /* Match at the very end of the data. */ | |
25fe55af | 5702 | case endbuf: |
dc4a2ee0 | 5703 | DEBUG_PRINT ("EXECUTING endbuf.\n"); |
fa9a63c5 RM |
5704 | if (AT_STRINGS_END (d)) |
5705 | break; | |
25fe55af | 5706 | goto fail; |
5e69f11e | 5707 | |
5e69f11e | 5708 | |
25fe55af RS |
5709 | /* on_failure_keep_string_jump is used to optimize `.*\n'. It |
5710 | pushes NULL as the value for the string on the stack. Then | |
505bde11 | 5711 | `POP_FAILURE_POINT' will keep the current value for the |
25fe55af | 5712 | string, instead of restoring it. To see why, consider |
7814e705 | 5713 | matching `foo\nbar' against `.*\n'. The .* matches the foo; |
25fe55af RS |
5714 | then the . fails against the \n. But the next thing we want |
5715 | to do is match the \n against the \n; if we restored the | |
5716 | string value, we would be back at the foo. | |
5717 | ||
5718 | Because this is used only in specific cases, we don't need to | |
5719 | check all the things that `on_failure_jump' does, to make | |
5720 | sure the right things get saved on the stack. Hence we don't | |
5721 | share its code. The only reason to push anything on the | |
5722 | stack at all is that otherwise we would have to change | |
5723 | `anychar's code to do something besides goto fail in this | |
5724 | case; that seems worse than this. */ | |
5725 | case on_failure_keep_string_jump: | |
505bde11 | 5726 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5727 | DEBUG_PRINT ("EXECUTING on_failure_keep_string_jump %d (to %p):\n", |
5728 | mcnt, p + mcnt); | |
fa9a63c5 | 5729 | |
505bde11 SM |
5730 | PUSH_FAILURE_POINT (p - 3, NULL); |
5731 | break; | |
5732 | ||
0683b6fa SM |
5733 | /* A nasty loop is introduced by the non-greedy *? and +?. |
5734 | With such loops, the stack only ever contains one failure point | |
5735 | at a time, so that a plain on_failure_jump_loop kind of | |
5736 | cycle detection cannot work. Worse yet, such a detection | |
5737 | can not only fail to detect a cycle, but it can also wrongly | |
5738 | detect a cycle (between different instantiations of the same | |
6df42991 | 5739 | loop). |
0683b6fa SM |
5740 | So the method used for those nasty loops is a little different: |
5741 | We use a special cycle-detection-stack-frame which is pushed | |
5742 | when the on_failure_jump_nastyloop failure-point is *popped*. | |
5743 | This special frame thus marks the beginning of one iteration | |
5744 | through the loop and we can hence easily check right here | |
5745 | whether something matched between the beginning and the end of | |
5746 | the loop. */ | |
5747 | case on_failure_jump_nastyloop: | |
5748 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
dc4a2ee0 PE |
5749 | DEBUG_PRINT ("EXECUTING on_failure_jump_nastyloop %d (to %p):\n", |
5750 | mcnt, p + mcnt); | |
0683b6fa SM |
5751 | |
5752 | assert ((re_opcode_t)p[-4] == no_op); | |
6df42991 SM |
5753 | { |
5754 | int cycle = 0; | |
5755 | CHECK_INFINITE_LOOP (p - 4, d); | |
5756 | if (!cycle) | |
5757 | /* If there's a cycle, just continue without pushing | |
5758 | this failure point. The failure point is the "try again" | |
5759 | option, which shouldn't be tried. | |
5760 | We want (x?)*?y\1z to match both xxyz and xxyxz. */ | |
5761 | PUSH_FAILURE_POINT (p - 3, d); | |
5762 | } | |
0683b6fa SM |
5763 | break; |
5764 | ||
4e8a9132 SM |
5765 | /* Simple loop detecting on_failure_jump: just check on the |
5766 | failure stack if the same spot was already hit earlier. */ | |
505bde11 SM |
5767 | case on_failure_jump_loop: |
5768 | on_failure: | |
5769 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | |
dc4a2ee0 PE |
5770 | DEBUG_PRINT ("EXECUTING on_failure_jump_loop %d (to %p):\n", |
5771 | mcnt, p + mcnt); | |
6df42991 SM |
5772 | { |
5773 | int cycle = 0; | |
5774 | CHECK_INFINITE_LOOP (p - 3, d); | |
5775 | if (cycle) | |
5776 | /* If there's a cycle, get out of the loop, as if the matching | |
5777 | had failed. We used to just `goto fail' here, but that was | |
5778 | aborting the search a bit too early: we want to keep the | |
5779 | empty-loop-match and keep matching after the loop. | |
5780 | We want (x?)*y\1z to match both xxyz and xxyxz. */ | |
5781 | p += mcnt; | |
5782 | else | |
5783 | PUSH_FAILURE_POINT (p - 3, d); | |
5784 | } | |
25fe55af | 5785 | break; |
fa9a63c5 RM |
5786 | |
5787 | ||
5788 | /* Uses of on_failure_jump: | |
5e69f11e | 5789 | |
25fe55af RS |
5790 | Each alternative starts with an on_failure_jump that points |
5791 | to the beginning of the next alternative. Each alternative | |
5792 | except the last ends with a jump that in effect jumps past | |
5793 | the rest of the alternatives. (They really jump to the | |
5794 | ending jump of the following alternative, because tensioning | |
5795 | these jumps is a hassle.) | |
fa9a63c5 | 5796 | |
25fe55af RS |
5797 | Repeats start with an on_failure_jump that points past both |
5798 | the repetition text and either the following jump or | |
5799 | pop_failure_jump back to this on_failure_jump. */ | |
fa9a63c5 | 5800 | case on_failure_jump: |
25fe55af | 5801 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5802 | DEBUG_PRINT ("EXECUTING on_failure_jump %d (to %p):\n", |
5803 | mcnt, p + mcnt); | |
25fe55af | 5804 | |
505bde11 | 5805 | PUSH_FAILURE_POINT (p -3, d); |
25fe55af RS |
5806 | break; |
5807 | ||
4e8a9132 | 5808 | /* This operation is used for greedy *. |
505bde11 SM |
5809 | Compare the beginning of the repeat with what in the |
5810 | pattern follows its end. If we can establish that there | |
5811 | is nothing that they would both match, i.e., that we | |
5812 | would have to backtrack because of (as in, e.g., `a*a') | |
5813 | then we can use a non-backtracking loop based on | |
4e8a9132 | 5814 | on_failure_keep_string_jump instead of on_failure_jump. */ |
505bde11 | 5815 | case on_failure_jump_smart: |
25fe55af | 5816 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 PE |
5817 | DEBUG_PRINT ("EXECUTING on_failure_jump_smart %d (to %p).\n", |
5818 | mcnt, p + mcnt); | |
25fe55af | 5819 | { |
01618498 | 5820 | re_char *p1 = p; /* Next operation. */ |
6dcf2d0e SM |
5821 | /* Here, we discard `const', making re_match non-reentrant. */ |
5822 | unsigned char *p2 = (unsigned char*) p + mcnt; /* Jump dest. */ | |
5823 | unsigned char *p3 = (unsigned char*) p - 3; /* opcode location. */ | |
fa9a63c5 | 5824 | |
505bde11 SM |
5825 | p -= 3; /* Reset so that we will re-execute the |
5826 | instruction once it's been changed. */ | |
fa9a63c5 | 5827 | |
4e8a9132 SM |
5828 | EXTRACT_NUMBER (mcnt, p2 - 2); |
5829 | ||
5830 | /* Ensure this is a indeed the trivial kind of loop | |
5831 | we are expecting. */ | |
5832 | assert (skip_one_char (p1) == p2 - 3); | |
5833 | assert ((re_opcode_t) p2[-3] == jump && p2 + mcnt == p); | |
99633e97 | 5834 | DEBUG_STATEMENT (debug += 2); |
505bde11 | 5835 | if (mutually_exclusive_p (bufp, p1, p2)) |
fa9a63c5 | 5836 | { |
505bde11 | 5837 | /* Use a fast `on_failure_keep_string_jump' loop. */ |
dc4a2ee0 | 5838 | DEBUG_PRINT (" smart exclusive => fast loop.\n"); |
01618498 | 5839 | *p3 = (unsigned char) on_failure_keep_string_jump; |
4e8a9132 | 5840 | STORE_NUMBER (p2 - 2, mcnt + 3); |
25fe55af | 5841 | } |
505bde11 | 5842 | else |
fa9a63c5 | 5843 | { |
505bde11 | 5844 | /* Default to a safe `on_failure_jump' loop. */ |
dc4a2ee0 | 5845 | DEBUG_PRINT (" smart default => slow loop.\n"); |
01618498 | 5846 | *p3 = (unsigned char) on_failure_jump; |
fa9a63c5 | 5847 | } |
99633e97 | 5848 | DEBUG_STATEMENT (debug -= 2); |
25fe55af | 5849 | } |
505bde11 | 5850 | break; |
25fe55af RS |
5851 | |
5852 | /* Unconditionally jump (without popping any failure points). */ | |
5853 | case jump: | |
fa9a63c5 | 5854 | unconditional_jump: |
5b370c2b | 5855 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 5856 | EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ |
dc4a2ee0 | 5857 | DEBUG_PRINT ("EXECUTING jump %d ", mcnt); |
7814e705 | 5858 | p += mcnt; /* Do the jump. */ |
dc4a2ee0 | 5859 | DEBUG_PRINT ("(to %p).\n", p); |
25fe55af RS |
5860 | break; |
5861 | ||
5862 | ||
25fe55af RS |
5863 | /* Have to succeed matching what follows at least n times. |
5864 | After that, handle like `on_failure_jump'. */ | |
5865 | case succeed_n: | |
01618498 | 5866 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af | 5867 | EXTRACT_NUMBER (mcnt, p + 2); |
dc4a2ee0 | 5868 | DEBUG_PRINT ("EXECUTING succeed_n %d.\n", mcnt); |
5e69f11e | 5869 | |
dc1e502d SM |
5870 | /* Originally, mcnt is how many times we HAVE to succeed. */ |
5871 | if (mcnt != 0) | |
25fe55af | 5872 | { |
6dcf2d0e SM |
5873 | /* Here, we discard `const', making re_match non-reentrant. */ |
5874 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5875 | mcnt--; |
01618498 SM |
5876 | p += 4; |
5877 | PUSH_NUMBER (p2, mcnt); | |
25fe55af | 5878 | } |
dc1e502d SM |
5879 | else |
5880 | /* The two bytes encoding mcnt == 0 are two no_op opcodes. */ | |
5881 | goto on_failure; | |
25fe55af RS |
5882 | break; |
5883 | ||
5884 | case jump_n: | |
01618498 | 5885 | /* Signedness doesn't matter since we only compare MCNT to 0. */ |
25fe55af | 5886 | EXTRACT_NUMBER (mcnt, p + 2); |
dc4a2ee0 | 5887 | DEBUG_PRINT ("EXECUTING jump_n %d.\n", mcnt); |
25fe55af RS |
5888 | |
5889 | /* Originally, this is how many times we CAN jump. */ | |
dc1e502d | 5890 | if (mcnt != 0) |
25fe55af | 5891 | { |
6dcf2d0e SM |
5892 | /* Here, we discard `const', making re_match non-reentrant. */ |
5893 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | |
dc1e502d | 5894 | mcnt--; |
01618498 | 5895 | PUSH_NUMBER (p2, mcnt); |
dc1e502d | 5896 | goto unconditional_jump; |
25fe55af RS |
5897 | } |
5898 | /* If don't have to jump any more, skip over the rest of command. */ | |
5e69f11e RM |
5899 | else |
5900 | p += 4; | |
25fe55af | 5901 | break; |
5e69f11e | 5902 | |
fa9a63c5 RM |
5903 | case set_number_at: |
5904 | { | |
01618498 | 5905 | unsigned char *p2; /* Location of the counter. */ |
dc4a2ee0 | 5906 | DEBUG_PRINT ("EXECUTING set_number_at.\n"); |
fa9a63c5 | 5907 | |
25fe55af | 5908 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
6dcf2d0e SM |
5909 | /* Here, we discard `const', making re_match non-reentrant. */ |
5910 | p2 = (unsigned char*) p + mcnt; | |
01618498 | 5911 | /* Signedness doesn't matter since we only copy MCNT's bits . */ |
25fe55af | 5912 | EXTRACT_NUMBER_AND_INCR (mcnt, p); |
dc4a2ee0 | 5913 | DEBUG_PRINT (" Setting %p to %d.\n", p2, mcnt); |
01618498 | 5914 | PUSH_NUMBER (p2, mcnt); |
25fe55af RS |
5915 | break; |
5916 | } | |
9121ca40 KH |
5917 | |
5918 | case wordbound: | |
66f0296e | 5919 | case notwordbound: |
19ed5445 PE |
5920 | { |
5921 | boolean not = (re_opcode_t) *(p - 1) == notwordbound; | |
dc4a2ee0 | 5922 | DEBUG_PRINT ("EXECUTING %swordbound.\n", not ? "not" : ""); |
fa9a63c5 | 5923 | |
19ed5445 | 5924 | /* We SUCCEED (or FAIL) in one of the following cases: */ |
9121ca40 | 5925 | |
19ed5445 PE |
5926 | /* Case 1: D is at the beginning or the end of string. */ |
5927 | if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) | |
5928 | not = !not; | |
5929 | else | |
5930 | { | |
5931 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5932 | is the character at D, and S2 is the syntax of C2. */ | |
5933 | re_wchar_t c1, c2; | |
5934 | int s1, s2; | |
5935 | int dummy; | |
b18215fc | 5936 | #ifdef emacs |
d1dfb56c EZ |
5937 | ssize_t offset = PTR_TO_OFFSET (d - 1); |
5938 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 | 5939 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5940 | #endif |
19ed5445 PE |
5941 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
5942 | s1 = SYNTAX (c1); | |
b18215fc | 5943 | #ifdef emacs |
19ed5445 | 5944 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
25fe55af | 5945 | #endif |
19ed5445 PE |
5946 | PREFETCH_NOLIMIT (); |
5947 | GET_CHAR_AFTER (c2, d, dummy); | |
5948 | s2 = SYNTAX (c2); | |
5949 | ||
5950 | if (/* Case 2: Only one of S1 and S2 is Sword. */ | |
5951 | ((s1 == Sword) != (s2 == Sword)) | |
5952 | /* Case 3: Both of S1 and S2 are Sword, and macro | |
5953 | WORD_BOUNDARY_P (C1, C2) returns nonzero. */ | |
5954 | || ((s1 == Sword) && WORD_BOUNDARY_P (c1, c2))) | |
5955 | not = !not; | |
5956 | } | |
5957 | if (not) | |
5958 | break; | |
5959 | else | |
5960 | goto fail; | |
5961 | } | |
fa9a63c5 RM |
5962 | |
5963 | case wordbeg: | |
dc4a2ee0 | 5964 | DEBUG_PRINT ("EXECUTING wordbeg.\n"); |
fa9a63c5 | 5965 | |
b18215fc RS |
5966 | /* We FAIL in one of the following cases: */ |
5967 | ||
7814e705 | 5968 | /* Case 1: D is at the end of string. */ |
b18215fc | 5969 | if (AT_STRINGS_END (d)) |
99633e97 | 5970 | goto fail; |
b18215fc RS |
5971 | else |
5972 | { | |
5973 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
5974 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
5975 | re_wchar_t c1, c2; |
5976 | int s1, s2; | |
bf216479 | 5977 | int dummy; |
fa9a63c5 | 5978 | #ifdef emacs |
d1dfb56c EZ |
5979 | ssize_t offset = PTR_TO_OFFSET (d); |
5980 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 5981 | UPDATE_SYNTAX_TABLE (charpos); |
25fe55af | 5982 | #endif |
99633e97 | 5983 | PREFETCH (); |
6fdd04b0 | 5984 | GET_CHAR_AFTER (c2, d, dummy); |
b18215fc | 5985 | s2 = SYNTAX (c2); |
177c0ea7 | 5986 | |
b18215fc RS |
5987 | /* Case 2: S2 is not Sword. */ |
5988 | if (s2 != Sword) | |
5989 | goto fail; | |
5990 | ||
5991 | /* Case 3: D is not at the beginning of string ... */ | |
5992 | if (!AT_STRINGS_BEG (d)) | |
5993 | { | |
5994 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
5995 | #ifdef emacs | |
5d967c7a | 5996 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); |
25fe55af | 5997 | #endif |
b18215fc RS |
5998 | s1 = SYNTAX (c1); |
5999 | ||
6000 | /* ... and S1 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6001 | returns 0. */ |
b18215fc RS |
6002 | if ((s1 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
6003 | goto fail; | |
6004 | } | |
6005 | } | |
e318085a RS |
6006 | break; |
6007 | ||
b18215fc | 6008 | case wordend: |
dc4a2ee0 | 6009 | DEBUG_PRINT ("EXECUTING wordend.\n"); |
b18215fc RS |
6010 | |
6011 | /* We FAIL in one of the following cases: */ | |
6012 | ||
6013 | /* Case 1: D is at the beginning of string. */ | |
6014 | if (AT_STRINGS_BEG (d)) | |
e318085a | 6015 | goto fail; |
b18215fc RS |
6016 | else |
6017 | { | |
6018 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6019 | is the character at D, and S2 is the syntax of C2. */ | |
01618498 SM |
6020 | re_wchar_t c1, c2; |
6021 | int s1, s2; | |
bf216479 | 6022 | int dummy; |
5d967c7a | 6023 | #ifdef emacs |
d1dfb56c EZ |
6024 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6025 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
92432794 | 6026 | UPDATE_SYNTAX_TABLE (charpos); |
5d967c7a | 6027 | #endif |
99633e97 | 6028 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); |
b18215fc RS |
6029 | s1 = SYNTAX (c1); |
6030 | ||
6031 | /* Case 2: S1 is not Sword. */ | |
6032 | if (s1 != Sword) | |
6033 | goto fail; | |
6034 | ||
6035 | /* Case 3: D is not at the end of string ... */ | |
6036 | if (!AT_STRINGS_END (d)) | |
6037 | { | |
f1ad044f | 6038 | PREFETCH_NOLIMIT (); |
6fdd04b0 | 6039 | GET_CHAR_AFTER (c2, d, dummy); |
5d967c7a RS |
6040 | #ifdef emacs |
6041 | UPDATE_SYNTAX_TABLE_FORWARD (charpos); | |
6042 | #endif | |
b18215fc RS |
6043 | s2 = SYNTAX (c2); |
6044 | ||
6045 | /* ... and S2 is Sword, and WORD_BOUNDARY_P (C1, C2) | |
7814e705 | 6046 | returns 0. */ |
b18215fc | 6047 | if ((s2 == Sword) && !WORD_BOUNDARY_P (c1, c2)) |
25fe55af | 6048 | goto fail; |
b18215fc RS |
6049 | } |
6050 | } | |
e318085a RS |
6051 | break; |
6052 | ||
669fa600 | 6053 | case symbeg: |
dc4a2ee0 | 6054 | DEBUG_PRINT ("EXECUTING symbeg.\n"); |
669fa600 SM |
6055 | |
6056 | /* We FAIL in one of the following cases: */ | |
6057 | ||
7814e705 | 6058 | /* Case 1: D is at the end of string. */ |
669fa600 SM |
6059 | if (AT_STRINGS_END (d)) |
6060 | goto fail; | |
6061 | else | |
6062 | { | |
6063 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6064 | is the character at D, and S2 is the syntax of C2. */ | |
6065 | re_wchar_t c1, c2; | |
6066 | int s1, s2; | |
6067 | #ifdef emacs | |
d1dfb56c EZ |
6068 | ssize_t offset = PTR_TO_OFFSET (d); |
6069 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6070 | UPDATE_SYNTAX_TABLE (charpos); |
6071 | #endif | |
6072 | PREFETCH (); | |
62a6e103 | 6073 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6074 | s2 = SYNTAX (c2); |
7814e705 | 6075 | |
669fa600 SM |
6076 | /* Case 2: S2 is neither Sword nor Ssymbol. */ |
6077 | if (s2 != Sword && s2 != Ssymbol) | |
6078 | goto fail; | |
6079 | ||
6080 | /* Case 3: D is not at the beginning of string ... */ | |
6081 | if (!AT_STRINGS_BEG (d)) | |
6082 | { | |
6083 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6084 | #ifdef emacs | |
6085 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); | |
6086 | #endif | |
6087 | s1 = SYNTAX (c1); | |
6088 | ||
6089 | /* ... and S1 is Sword or Ssymbol. */ | |
6090 | if (s1 == Sword || s1 == Ssymbol) | |
6091 | goto fail; | |
6092 | } | |
6093 | } | |
6094 | break; | |
6095 | ||
6096 | case symend: | |
dc4a2ee0 | 6097 | DEBUG_PRINT ("EXECUTING symend.\n"); |
669fa600 SM |
6098 | |
6099 | /* We FAIL in one of the following cases: */ | |
6100 | ||
6101 | /* Case 1: D is at the beginning of string. */ | |
6102 | if (AT_STRINGS_BEG (d)) | |
6103 | goto fail; | |
6104 | else | |
6105 | { | |
6106 | /* C1 is the character before D, S1 is the syntax of C1, C2 | |
6107 | is the character at D, and S2 is the syntax of C2. */ | |
6108 | re_wchar_t c1, c2; | |
6109 | int s1, s2; | |
6110 | #ifdef emacs | |
d1dfb56c EZ |
6111 | ssize_t offset = PTR_TO_OFFSET (d) - 1; |
6112 | ssize_t charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
669fa600 SM |
6113 | UPDATE_SYNTAX_TABLE (charpos); |
6114 | #endif | |
6115 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | |
6116 | s1 = SYNTAX (c1); | |
6117 | ||
6118 | /* Case 2: S1 is neither Ssymbol nor Sword. */ | |
6119 | if (s1 != Sword && s1 != Ssymbol) | |
6120 | goto fail; | |
6121 | ||
6122 | /* Case 3: D is not at the end of string ... */ | |
6123 | if (!AT_STRINGS_END (d)) | |
6124 | { | |
6125 | PREFETCH_NOLIMIT (); | |
62a6e103 | 6126 | c2 = RE_STRING_CHAR (d, target_multibyte); |
669fa600 | 6127 | #ifdef emacs |
134579f2 | 6128 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); |
669fa600 SM |
6129 | #endif |
6130 | s2 = SYNTAX (c2); | |
6131 | ||
6132 | /* ... and S2 is Sword or Ssymbol. */ | |
6133 | if (s2 == Sword || s2 == Ssymbol) | |
6134 | goto fail; | |
b18215fc RS |
6135 | } |
6136 | } | |
e318085a RS |
6137 | break; |
6138 | ||
fa9a63c5 | 6139 | case syntaxspec: |
1fb352e0 | 6140 | case notsyntaxspec: |
b18215fc | 6141 | { |
19ed5445 PE |
6142 | boolean not = (re_opcode_t) *(p - 1) == notsyntaxspec; |
6143 | mcnt = *p++; | |
dc4a2ee0 PE |
6144 | DEBUG_PRINT ("EXECUTING %ssyntaxspec %d.\n", not ? "not" : "", |
6145 | mcnt); | |
19ed5445 PE |
6146 | PREFETCH (); |
6147 | #ifdef emacs | |
6148 | { | |
d1dfb56c EZ |
6149 | ssize_t offset = PTR_TO_OFFSET (d); |
6150 | ssize_t pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | |
19ed5445 PE |
6151 | UPDATE_SYNTAX_TABLE (pos1); |
6152 | } | |
25fe55af | 6153 | #endif |
19ed5445 PE |
6154 | { |
6155 | int len; | |
6156 | re_wchar_t c; | |
b18215fc | 6157 | |
19ed5445 PE |
6158 | GET_CHAR_AFTER (c, d, len); |
6159 | if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not) | |
6160 | goto fail; | |
6161 | d += len; | |
6162 | } | |
b18215fc | 6163 | } |
8fb31792 | 6164 | break; |
fa9a63c5 | 6165 | |
b18215fc | 6166 | #ifdef emacs |
1fb352e0 | 6167 | case before_dot: |
dc4a2ee0 | 6168 | DEBUG_PRINT ("EXECUTING before_dot.\n"); |
1fb352e0 | 6169 | if (PTR_BYTE_POS (d) >= PT_BYTE) |
fa9a63c5 | 6170 | goto fail; |
b18215fc RS |
6171 | break; |
6172 | ||
1fb352e0 | 6173 | case at_dot: |
dc4a2ee0 | 6174 | DEBUG_PRINT ("EXECUTING at_dot.\n"); |
1fb352e0 SM |
6175 | if (PTR_BYTE_POS (d) != PT_BYTE) |
6176 | goto fail; | |
6177 | break; | |
b18215fc | 6178 | |
1fb352e0 | 6179 | case after_dot: |
dc4a2ee0 | 6180 | DEBUG_PRINT ("EXECUTING after_dot.\n"); |
1fb352e0 SM |
6181 | if (PTR_BYTE_POS (d) <= PT_BYTE) |
6182 | goto fail; | |
e318085a | 6183 | break; |
fa9a63c5 | 6184 | |
1fb352e0 | 6185 | case categoryspec: |
b18215fc | 6186 | case notcategoryspec: |
b18215fc | 6187 | { |
8fb31792 PE |
6188 | boolean not = (re_opcode_t) *(p - 1) == notcategoryspec; |
6189 | mcnt = *p++; | |
dc4a2ee0 PE |
6190 | DEBUG_PRINT ("EXECUTING %scategoryspec %d.\n", |
6191 | not ? "not" : "", mcnt); | |
8fb31792 | 6192 | PREFETCH (); |
01618498 | 6193 | |
8fb31792 PE |
6194 | { |
6195 | int len; | |
6196 | re_wchar_t c; | |
6197 | GET_CHAR_AFTER (c, d, len); | |
6198 | if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not) | |
6199 | goto fail; | |
6200 | d += len; | |
6201 | } | |
b18215fc | 6202 | } |
fa9a63c5 | 6203 | break; |
5e69f11e | 6204 | |
1fb352e0 | 6205 | #endif /* emacs */ |
5e69f11e | 6206 | |
0b32bf0e SM |
6207 | default: |
6208 | abort (); | |
fa9a63c5 | 6209 | } |
b18215fc | 6210 | continue; /* Successfully executed one pattern command; keep going. */ |
fa9a63c5 RM |
6211 | |
6212 | ||
6213 | /* We goto here if a matching operation fails. */ | |
6214 | fail: | |
5b370c2b | 6215 | IMMEDIATE_QUIT_CHECK; |
fa9a63c5 | 6216 | if (!FAIL_STACK_EMPTY ()) |
505bde11 | 6217 | { |
01618498 | 6218 | re_char *str, *pat; |
505bde11 | 6219 | /* A restart point is known. Restore to that state. */ |
dc4a2ee0 | 6220 | DEBUG_PRINT ("\nFAIL:\n"); |
0b32bf0e | 6221 | POP_FAILURE_POINT (str, pat); |
7393bcbb | 6222 | switch (*pat++) |
505bde11 SM |
6223 | { |
6224 | case on_failure_keep_string_jump: | |
6225 | assert (str == NULL); | |
6226 | goto continue_failure_jump; | |
6227 | ||
0683b6fa SM |
6228 | case on_failure_jump_nastyloop: |
6229 | assert ((re_opcode_t)pat[-2] == no_op); | |
6230 | PUSH_FAILURE_POINT (pat - 2, str); | |
6231 | /* Fallthrough */ | |
6232 | ||
505bde11 SM |
6233 | case on_failure_jump_loop: |
6234 | case on_failure_jump: | |
6235 | case succeed_n: | |
6236 | d = str; | |
6237 | continue_failure_jump: | |
6238 | EXTRACT_NUMBER_AND_INCR (mcnt, pat); | |
6239 | p = pat + mcnt; | |
6240 | break; | |
b18215fc | 6241 | |
0683b6fa SM |
6242 | case no_op: |
6243 | /* A special frame used for nastyloops. */ | |
6244 | goto fail; | |
6245 | ||
505bde11 | 6246 | default: |
5e617bc2 | 6247 | abort (); |
505bde11 | 6248 | } |
fa9a63c5 | 6249 | |
505bde11 | 6250 | assert (p >= bufp->buffer && p <= pend); |
b18215fc | 6251 | |
0b32bf0e | 6252 | if (d >= string1 && d <= end1) |
fa9a63c5 | 6253 | dend = end_match_1; |
0b32bf0e | 6254 | } |
fa9a63c5 | 6255 | else |
0b32bf0e | 6256 | break; /* Matching at this starting point really fails. */ |
fa9a63c5 RM |
6257 | } /* for (;;) */ |
6258 | ||
6259 | if (best_regs_set) | |
6260 | goto restore_best_regs; | |
6261 | ||
6262 | FREE_VARIABLES (); | |
6263 | ||
b18215fc | 6264 | return -1; /* Failure to match. */ |
dc4a2ee0 | 6265 | } |
fa9a63c5 RM |
6266 | \f |
6267 | /* Subroutine definitions for re_match_2. */ | |
6268 | ||
fa9a63c5 RM |
6269 | /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN |
6270 | bytes; nonzero otherwise. */ | |
5e69f11e | 6271 | |
fa9a63c5 | 6272 | static int |
29abe551 | 6273 | bcmp_translate (const_re_char *s1, const_re_char *s2, register ssize_t len, |
438105ed | 6274 | RE_TRANSLATE_TYPE translate, const int target_multibyte) |
fa9a63c5 | 6275 | { |
2d1675e4 SM |
6276 | register re_char *p1 = s1, *p2 = s2; |
6277 | re_char *p1_end = s1 + len; | |
6278 | re_char *p2_end = s2 + len; | |
e934739e | 6279 | |
4bb91c68 SM |
6280 | /* FIXME: Checking both p1 and p2 presumes that the two strings might have |
6281 | different lengths, but relying on a single `len' would break this. -sm */ | |
6282 | while (p1 < p1_end && p2 < p2_end) | |
fa9a63c5 | 6283 | { |
e934739e | 6284 | int p1_charlen, p2_charlen; |
01618498 | 6285 | re_wchar_t p1_ch, p2_ch; |
e934739e | 6286 | |
6fdd04b0 KH |
6287 | GET_CHAR_AFTER (p1_ch, p1, p1_charlen); |
6288 | GET_CHAR_AFTER (p2_ch, p2, p2_charlen); | |
e934739e RS |
6289 | |
6290 | if (RE_TRANSLATE (translate, p1_ch) | |
6291 | != RE_TRANSLATE (translate, p2_ch)) | |
bc192b5b | 6292 | return 1; |
e934739e RS |
6293 | |
6294 | p1 += p1_charlen, p2 += p2_charlen; | |
fa9a63c5 | 6295 | } |
e934739e RS |
6296 | |
6297 | if (p1 != p1_end || p2 != p2_end) | |
6298 | return 1; | |
6299 | ||
fa9a63c5 RM |
6300 | return 0; |
6301 | } | |
6302 | \f | |
6303 | /* Entry points for GNU code. */ | |
6304 | ||
6305 | /* re_compile_pattern is the GNU regular expression compiler: it | |
6306 | compiles PATTERN (of length SIZE) and puts the result in BUFP. | |
6307 | Returns 0 if the pattern was valid, otherwise an error string. | |
5e69f11e | 6308 | |
fa9a63c5 RM |
6309 | Assumes the `allocated' (and perhaps `buffer') and `translate' fields |
6310 | are set in BUFP on entry. | |
5e69f11e | 6311 | |
b18215fc | 6312 | We call regex_compile to do the actual compilation. */ |
fa9a63c5 RM |
6313 | |
6314 | const char * | |
d1dfb56c EZ |
6315 | re_compile_pattern (const char *pattern, size_t length, |
6316 | struct re_pattern_buffer *bufp) | |
fa9a63c5 RM |
6317 | { |
6318 | reg_errcode_t ret; | |
5e69f11e | 6319 | |
fa9a63c5 RM |
6320 | /* GNU code is written to assume at least RE_NREGS registers will be set |
6321 | (and at least one extra will be -1). */ | |
6322 | bufp->regs_allocated = REGS_UNALLOCATED; | |
5e69f11e | 6323 | |
fa9a63c5 RM |
6324 | /* And GNU code determines whether or not to get register information |
6325 | by passing null for the REGS argument to re_match, etc., not by | |
6326 | setting no_sub. */ | |
6327 | bufp->no_sub = 0; | |
5e69f11e | 6328 | |
4bb91c68 | 6329 | ret = regex_compile ((re_char*) pattern, length, re_syntax_options, bufp); |
fa9a63c5 RM |
6330 | |
6331 | if (!ret) | |
6332 | return NULL; | |
6333 | return gettext (re_error_msgid[(int) ret]); | |
5e69f11e | 6334 | } |
c0f9ea08 | 6335 | WEAK_ALIAS (__re_compile_pattern, re_compile_pattern) |
fa9a63c5 | 6336 | \f |
b18215fc RS |
6337 | /* Entry points compatible with 4.2 BSD regex library. We don't define |
6338 | them unless specifically requested. */ | |
fa9a63c5 | 6339 | |
0b32bf0e | 6340 | #if defined _REGEX_RE_COMP || defined _LIBC |
fa9a63c5 RM |
6341 | |
6342 | /* BSD has one and only one pattern buffer. */ | |
6343 | static struct re_pattern_buffer re_comp_buf; | |
6344 | ||
6345 | char * | |
0b32bf0e | 6346 | # ifdef _LIBC |
48afdd44 RM |
6347 | /* Make these definitions weak in libc, so POSIX programs can redefine |
6348 | these names if they don't use our functions, and still use | |
6349 | regcomp/regexec below without link errors. */ | |
6350 | weak_function | |
0b32bf0e | 6351 | # endif |
31011111 | 6352 | re_comp (const char *s) |
fa9a63c5 RM |
6353 | { |
6354 | reg_errcode_t ret; | |
5e69f11e | 6355 | |
fa9a63c5 RM |
6356 | if (!s) |
6357 | { | |
6358 | if (!re_comp_buf.buffer) | |
0b32bf0e | 6359 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
a60198e5 | 6360 | return (char *) gettext ("No previous regular expression"); |
fa9a63c5 RM |
6361 | return 0; |
6362 | } | |
6363 | ||
6364 | if (!re_comp_buf.buffer) | |
6365 | { | |
38182d90 | 6366 | re_comp_buf.buffer = malloc (200); |
fa9a63c5 | 6367 | if (re_comp_buf.buffer == NULL) |
0b32bf0e SM |
6368 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6369 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6370 | re_comp_buf.allocated = 200; |
6371 | ||
38182d90 | 6372 | re_comp_buf.fastmap = malloc (1 << BYTEWIDTH); |
fa9a63c5 | 6373 | if (re_comp_buf.fastmap == NULL) |
a60198e5 SM |
6374 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ |
6375 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | |
fa9a63c5 RM |
6376 | } |
6377 | ||
6378 | /* Since `re_exec' always passes NULL for the `regs' argument, we | |
6379 | don't need to initialize the pattern buffer fields which affect it. */ | |
6380 | ||
fa9a63c5 | 6381 | ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); |
5e69f11e | 6382 | |
fa9a63c5 RM |
6383 | if (!ret) |
6384 | return NULL; | |
6385 | ||
6386 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | |
6387 | return (char *) gettext (re_error_msgid[(int) ret]); | |
6388 | } | |
6389 | ||
6390 | ||
31011111 | 6391 | int |
0b32bf0e | 6392 | # ifdef _LIBC |
48afdd44 | 6393 | weak_function |
0b32bf0e | 6394 | # endif |
d1dfb56c | 6395 | re_exec (const char *s) |
fa9a63c5 | 6396 | { |
d1dfb56c | 6397 | const size_t len = strlen (s); |
7d652d97 | 6398 | return re_search (&re_comp_buf, s, len, 0, len, 0) >= 0; |
fa9a63c5 RM |
6399 | } |
6400 | #endif /* _REGEX_RE_COMP */ | |
6401 | \f | |
6402 | /* POSIX.2 functions. Don't define these for Emacs. */ | |
6403 | ||
6404 | #ifndef emacs | |
6405 | ||
6406 | /* regcomp takes a regular expression as a string and compiles it. | |
6407 | ||
b18215fc | 6408 | PREG is a regex_t *. We do not expect any fields to be initialized, |
fa9a63c5 RM |
6409 | since POSIX says we shouldn't. Thus, we set |
6410 | ||
6411 | `buffer' to the compiled pattern; | |
6412 | `used' to the length of the compiled pattern; | |
6413 | `syntax' to RE_SYNTAX_POSIX_EXTENDED if the | |
6414 | REG_EXTENDED bit in CFLAGS is set; otherwise, to | |
6415 | RE_SYNTAX_POSIX_BASIC; | |
c0f9ea08 SM |
6416 | `fastmap' to an allocated space for the fastmap; |
6417 | `fastmap_accurate' to zero; | |
fa9a63c5 RM |
6418 | `re_nsub' to the number of subexpressions in PATTERN. |
6419 | ||
6420 | PATTERN is the address of the pattern string. | |
6421 | ||
6422 | CFLAGS is a series of bits which affect compilation. | |
6423 | ||
6424 | If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we | |
6425 | use POSIX basic syntax. | |
6426 | ||
6427 | If REG_NEWLINE is set, then . and [^...] don't match newline. | |
6428 | Also, regexec will try a match beginning after every newline. | |
6429 | ||
6430 | If REG_ICASE is set, then we considers upper- and lowercase | |
6431 | versions of letters to be equivalent when matching. | |
6432 | ||
6433 | If REG_NOSUB is set, then when PREG is passed to regexec, that | |
6434 | routine will report only success or failure, and nothing about the | |
6435 | registers. | |
6436 | ||
b18215fc | 6437 | It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for |
fa9a63c5 RM |
6438 | the return codes and their meanings.) */ |
6439 | ||
d1dfb56c | 6440 | reg_errcode_t |
29abe551 | 6441 | regcomp (regex_t *_Restrict_ preg, const char *_Restrict_ pattern, |
d2762c86 | 6442 | int cflags) |
fa9a63c5 RM |
6443 | { |
6444 | reg_errcode_t ret; | |
4bb91c68 | 6445 | reg_syntax_t syntax |
fa9a63c5 RM |
6446 | = (cflags & REG_EXTENDED) ? |
6447 | RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; | |
6448 | ||
6449 | /* regex_compile will allocate the space for the compiled pattern. */ | |
6450 | preg->buffer = 0; | |
6451 | preg->allocated = 0; | |
6452 | preg->used = 0; | |
5e69f11e | 6453 | |
c0f9ea08 | 6454 | /* Try to allocate space for the fastmap. */ |
38182d90 | 6455 | preg->fastmap = malloc (1 << BYTEWIDTH); |
5e69f11e | 6456 | |
fa9a63c5 RM |
6457 | if (cflags & REG_ICASE) |
6458 | { | |
6459 | unsigned i; | |
5e69f11e | 6460 | |
38182d90 | 6461 | preg->translate = malloc (CHAR_SET_SIZE * sizeof *preg->translate); |
fa9a63c5 | 6462 | if (preg->translate == NULL) |
0b32bf0e | 6463 | return (int) REG_ESPACE; |
fa9a63c5 RM |
6464 | |
6465 | /* Map uppercase characters to corresponding lowercase ones. */ | |
6466 | for (i = 0; i < CHAR_SET_SIZE; i++) | |
4bb91c68 | 6467 | preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i; |
fa9a63c5 RM |
6468 | } |
6469 | else | |
6470 | preg->translate = NULL; | |
6471 | ||
6472 | /* If REG_NEWLINE is set, newlines are treated differently. */ | |
6473 | if (cflags & REG_NEWLINE) | |
6474 | { /* REG_NEWLINE implies neither . nor [^...] match newline. */ | |
6475 | syntax &= ~RE_DOT_NEWLINE; | |
6476 | syntax |= RE_HAT_LISTS_NOT_NEWLINE; | |
fa9a63c5 RM |
6477 | } |
6478 | else | |
c0f9ea08 | 6479 | syntax |= RE_NO_NEWLINE_ANCHOR; |
fa9a63c5 RM |
6480 | |
6481 | preg->no_sub = !!(cflags & REG_NOSUB); | |
6482 | ||
5e69f11e | 6483 | /* POSIX says a null character in the pattern terminates it, so we |
fa9a63c5 | 6484 | can use strlen here in compiling the pattern. */ |
4bb91c68 | 6485 | ret = regex_compile ((re_char*) pattern, strlen (pattern), syntax, preg); |
5e69f11e | 6486 | |
fa9a63c5 RM |
6487 | /* POSIX doesn't distinguish between an unmatched open-group and an |
6488 | unmatched close-group: both are REG_EPAREN. */ | |
c0f9ea08 SM |
6489 | if (ret == REG_ERPAREN) |
6490 | ret = REG_EPAREN; | |
6491 | ||
6492 | if (ret == REG_NOERROR && preg->fastmap) | |
6493 | { /* Compute the fastmap now, since regexec cannot modify the pattern | |
6494 | buffer. */ | |
6495 | re_compile_fastmap (preg); | |
6496 | if (preg->can_be_null) | |
6497 | { /* The fastmap can't be used anyway. */ | |
6498 | free (preg->fastmap); | |
6499 | preg->fastmap = NULL; | |
6500 | } | |
6501 | } | |
d1dfb56c | 6502 | return ret; |
fa9a63c5 | 6503 | } |
c0f9ea08 | 6504 | WEAK_ALIAS (__regcomp, regcomp) |
fa9a63c5 RM |
6505 | |
6506 | ||
6507 | /* regexec searches for a given pattern, specified by PREG, in the | |
6508 | string STRING. | |
5e69f11e | 6509 | |
fa9a63c5 | 6510 | If NMATCH is zero or REG_NOSUB was set in the cflags argument to |
b18215fc | 6511 | `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at |
fa9a63c5 RM |
6512 | least NMATCH elements, and we set them to the offsets of the |
6513 | corresponding matched substrings. | |
5e69f11e | 6514 | |
fa9a63c5 RM |
6515 | EFLAGS specifies `execution flags' which affect matching: if |
6516 | REG_NOTBOL is set, then ^ does not match at the beginning of the | |
6517 | string; if REG_NOTEOL is set, then $ does not match at the end. | |
5e69f11e | 6518 | |
fa9a63c5 RM |
6519 | We return 0 if we find a match and REG_NOMATCH if not. */ |
6520 | ||
d1dfb56c | 6521 | reg_errcode_t |
29abe551 PE |
6522 | regexec (const regex_t *_Restrict_ preg, const char *_Restrict_ string, |
6523 | size_t nmatch, regmatch_t pmatch[_Restrict_arr_], int eflags) | |
fa9a63c5 | 6524 | { |
31011111 | 6525 | regoff_t ret; |
fa9a63c5 RM |
6526 | struct re_registers regs; |
6527 | regex_t private_preg; | |
d1dfb56c | 6528 | size_t len = strlen (string); |
c0f9ea08 | 6529 | boolean want_reg_info = !preg->no_sub && nmatch > 0 && pmatch; |
fa9a63c5 RM |
6530 | |
6531 | private_preg = *preg; | |
5e69f11e | 6532 | |
fa9a63c5 RM |
6533 | private_preg.not_bol = !!(eflags & REG_NOTBOL); |
6534 | private_preg.not_eol = !!(eflags & REG_NOTEOL); | |
5e69f11e | 6535 | |
fa9a63c5 RM |
6536 | /* The user has told us exactly how many registers to return |
6537 | information about, via `nmatch'. We have to pass that on to the | |
b18215fc | 6538 | matching routines. */ |
fa9a63c5 | 6539 | private_preg.regs_allocated = REGS_FIXED; |
5e69f11e | 6540 | |
fa9a63c5 RM |
6541 | if (want_reg_info) |
6542 | { | |
6543 | regs.num_regs = nmatch; | |
4bb91c68 SM |
6544 | regs.start = TALLOC (nmatch * 2, regoff_t); |
6545 | if (regs.start == NULL) | |
d1dfb56c | 6546 | return REG_NOMATCH; |
4bb91c68 | 6547 | regs.end = regs.start + nmatch; |
fa9a63c5 RM |
6548 | } |
6549 | ||
c0f9ea08 SM |
6550 | /* Instead of using not_eol to implement REG_NOTEOL, we could simply |
6551 | pass (&private_preg, string, len + 1, 0, len, ...) pretending the string | |
6552 | was a little bit longer but still only matching the real part. | |
6553 | This works because the `endline' will check for a '\n' and will find a | |
6554 | '\0', correctly deciding that this is not the end of a line. | |
6555 | But it doesn't work out so nicely for REG_NOTBOL, since we don't have | |
6556 | a convenient '\0' there. For all we know, the string could be preceded | |
6557 | by '\n' which would throw things off. */ | |
6558 | ||
fa9a63c5 RM |
6559 | /* Perform the searching operation. */ |
6560 | ret = re_search (&private_preg, string, len, | |
0b32bf0e | 6561 | /* start: */ 0, /* range: */ len, |
7d652d97 | 6562 | want_reg_info ? ®s : 0); |
5e69f11e | 6563 | |
fa9a63c5 RM |
6564 | /* Copy the register information to the POSIX structure. */ |
6565 | if (want_reg_info) | |
6566 | { | |
6567 | if (ret >= 0) | |
0b32bf0e SM |
6568 | { |
6569 | unsigned r; | |
fa9a63c5 | 6570 | |
0b32bf0e SM |
6571 | for (r = 0; r < nmatch; r++) |
6572 | { | |
6573 | pmatch[r].rm_so = regs.start[r]; | |
6574 | pmatch[r].rm_eo = regs.end[r]; | |
6575 | } | |
6576 | } | |
fa9a63c5 | 6577 | |
b18215fc | 6578 | /* If we needed the temporary register info, free the space now. */ |
fa9a63c5 | 6579 | free (regs.start); |
fa9a63c5 RM |
6580 | } |
6581 | ||
6582 | /* We want zero return to mean success, unlike `re_search'. */ | |
d1dfb56c | 6583 | return ret >= 0 ? REG_NOERROR : REG_NOMATCH; |
fa9a63c5 | 6584 | } |
c0f9ea08 | 6585 | WEAK_ALIAS (__regexec, regexec) |
fa9a63c5 RM |
6586 | |
6587 | ||
ec869672 JR |
6588 | /* Returns a message corresponding to an error code, ERR_CODE, returned |
6589 | from either regcomp or regexec. We don't use PREG here. | |
6590 | ||
6591 | ERR_CODE was previously called ERRCODE, but that name causes an | |
6592 | error with msvc8 compiler. */ | |
fa9a63c5 RM |
6593 | |
6594 | size_t | |
d2762c86 | 6595 | regerror (int err_code, const regex_t *preg, char *errbuf, size_t errbuf_size) |
fa9a63c5 RM |
6596 | { |
6597 | const char *msg; | |
6598 | size_t msg_size; | |
6599 | ||
ec869672 JR |
6600 | if (err_code < 0 |
6601 | || err_code >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0]))) | |
5e69f11e | 6602 | /* Only error codes returned by the rest of the code should be passed |
b18215fc | 6603 | to this routine. If we are given anything else, or if other regex |
fa9a63c5 RM |
6604 | code generates an invalid error code, then the program has a bug. |
6605 | Dump core so we can fix it. */ | |
6606 | abort (); | |
6607 | ||
ec869672 | 6608 | msg = gettext (re_error_msgid[err_code]); |
fa9a63c5 RM |
6609 | |
6610 | msg_size = strlen (msg) + 1; /* Includes the null. */ | |
5e69f11e | 6611 | |
fa9a63c5 RM |
6612 | if (errbuf_size != 0) |
6613 | { | |
6614 | if (msg_size > errbuf_size) | |
0b32bf0e | 6615 | { |
e99a530f | 6616 | memcpy (errbuf, msg, errbuf_size - 1); |
0b32bf0e SM |
6617 | errbuf[errbuf_size - 1] = 0; |
6618 | } | |
fa9a63c5 | 6619 | else |
0b32bf0e | 6620 | strcpy (errbuf, msg); |
fa9a63c5 RM |
6621 | } |
6622 | ||
6623 | return msg_size; | |
6624 | } | |
c0f9ea08 | 6625 | WEAK_ALIAS (__regerror, regerror) |
fa9a63c5 RM |
6626 | |
6627 | ||
6628 | /* Free dynamically allocated space used by PREG. */ | |
6629 | ||
6630 | void | |
d2762c86 | 6631 | regfree (regex_t *preg) |
fa9a63c5 | 6632 | { |
c2cd06e6 | 6633 | free (preg->buffer); |
fa9a63c5 | 6634 | preg->buffer = NULL; |
5e69f11e | 6635 | |
fa9a63c5 RM |
6636 | preg->allocated = 0; |
6637 | preg->used = 0; | |
6638 | ||
c2cd06e6 | 6639 | free (preg->fastmap); |
fa9a63c5 RM |
6640 | preg->fastmap = NULL; |
6641 | preg->fastmap_accurate = 0; | |
6642 | ||
c2cd06e6 | 6643 | free (preg->translate); |
fa9a63c5 RM |
6644 | preg->translate = NULL; |
6645 | } | |
c0f9ea08 | 6646 | WEAK_ALIAS (__regfree, regfree) |
fa9a63c5 RM |
6647 | |
6648 | #endif /* not emacs */ |