Parent Directory
|
Revision Log
File tidies, preparing for 8.32-RC1.
1 | /************************************************* |
2 | * Perl-Compatible Regular Expressions * |
3 | *************************************************/ |
4 | |
5 | /* PCRE is a library of functions to support regular expressions whose syntax |
6 | and semantics are as close as possible to those of the Perl 5 language. |
7 | |
8 | Written by Philip Hazel |
9 | Copyright (c) 1997-2012 University of Cambridge |
10 | |
11 | ----------------------------------------------------------------------------- |
12 | Redistribution and use in source and binary forms, with or without |
13 | modification, are permitted provided that the following conditions are met: |
14 | |
15 | * Redistributions of source code must retain the above copyright notice, |
16 | this list of conditions and the following disclaimer. |
17 | |
18 | * Redistributions in binary form must reproduce the above copyright |
19 | notice, this list of conditions and the following disclaimer in the |
20 | documentation and/or other materials provided with the distribution. |
21 | |
22 | * Neither the name of the University of Cambridge nor the names of its |
23 | contributors may be used to endorse or promote products derived from |
24 | this software without specific prior written permission. |
25 | |
26 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
27 | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
28 | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
29 | ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
30 | LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
31 | CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
32 | SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
33 | INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
34 | CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
35 | ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
36 | POSSIBILITY OF SUCH DAMAGE. |
37 | ----------------------------------------------------------------------------- |
38 | */ |
39 | |
40 | /* This module contains pcre_exec(), the externally visible function that does |
41 | pattern matching using an NFA algorithm, trying to mimic Perl as closely as |
42 | possible. There are also some static supporting functions. */ |
43 | |
44 | #ifdef HAVE_CONFIG_H |
45 | #include "config.h" |
46 | #endif |
47 | |
48 | #define NLBLOCK md /* Block containing newline information */ |
49 | #define PSSTART start_subject /* Field containing processed string start */ |
50 | #define PSEND end_subject /* Field containing processed string end */ |
51 | |
52 | #include "pcre_internal.h" |
53 | |
54 | /* Undefine some potentially clashing cpp symbols */ |
55 | |
56 | #undef min |
57 | #undef max |
58 | |
59 | /* Values for setting in md->match_function_type to indicate two special types |
60 | of call to match(). We do it this way to save on using another stack variable, |
61 | as stack usage is to be discouraged. */ |
62 | |
63 | #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */ |
64 | #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */ |
65 | |
66 | /* Non-error returns from the match() function. Error returns are externally |
67 | defined PCRE_ERROR_xxx codes, which are all negative. */ |
68 | |
69 | #define MATCH_MATCH 1 |
70 | #define MATCH_NOMATCH 0 |
71 | |
72 | /* Special internal returns from the match() function. Make them sufficiently |
73 | negative to avoid the external error codes. */ |
74 | |
75 | #define MATCH_ACCEPT (-999) |
76 | #define MATCH_COMMIT (-998) |
77 | #define MATCH_KETRPOS (-997) |
78 | #define MATCH_ONCE (-996) |
79 | #define MATCH_PRUNE (-995) |
80 | #define MATCH_SKIP (-994) |
81 | #define MATCH_SKIP_ARG (-993) |
82 | #define MATCH_THEN (-992) |
83 | |
84 | /* Maximum number of ints of offset to save on the stack for recursive calls. |
85 | If the offset vector is bigger, malloc is used. This should be a multiple of 3, |
86 | because the offset vector is always a multiple of 3 long. */ |
87 | |
88 | #define REC_STACK_SAVE_MAX 30 |
89 | |
90 | /* Min and max values for the common repeats; for the maxima, 0 => infinity */ |
91 | |
92 | static const char rep_min[] = { 0, 0, 1, 1, 0, 0 }; |
93 | static const char rep_max[] = { 0, 0, 0, 0, 1, 1 }; |
94 | |
95 | #ifdef PCRE_DEBUG |
96 | /************************************************* |
97 | * Debugging function to print chars * |
98 | *************************************************/ |
99 | |
100 | /* Print a sequence of chars in printable format, stopping at the end of the |
101 | subject if the requested. |
102 | |
103 | Arguments: |
104 | p points to characters |
105 | length number to print |
106 | is_subject TRUE if printing from within md->start_subject |
107 | md pointer to matching data block, if is_subject is TRUE |
108 | |
109 | Returns: nothing |
110 | */ |
111 | |
112 | static void |
113 | pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md) |
114 | { |
115 | pcre_uint32 c; |
116 | BOOL utf = md->utf; |
117 | if (is_subject && length > md->end_subject - p) length = md->end_subject - p; |
118 | while (length-- > 0) |
119 | if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c); |
120 | } |
121 | #endif |
122 | |
123 | |
124 | |
125 | /************************************************* |
126 | * Match a back-reference * |
127 | *************************************************/ |
128 | |
129 | /* Normally, if a back reference hasn't been set, the length that is passed is |
130 | negative, so the match always fails. However, in JavaScript compatibility mode, |
131 | the length passed is zero. Note that in caseless UTF-8 mode, the number of |
132 | subject bytes matched may be different to the number of reference bytes. |
133 | |
134 | Arguments: |
135 | offset index into the offset vector |
136 | eptr pointer into the subject |
137 | length length of reference to be matched (number of bytes) |
138 | md points to match data block |
139 | caseless TRUE if caseless |
140 | |
141 | Returns: >= 0 the number of subject bytes matched |
142 | -1 no match |
143 | -2 partial match; always given if at end subject |
144 | */ |
145 | |
146 | static int |
147 | match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md, |
148 | BOOL caseless) |
149 | { |
150 | PCRE_PUCHAR eptr_start = eptr; |
151 | register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset]; |
152 | #ifdef SUPPORT_UTF |
153 | BOOL utf = md->utf; |
154 | #endif |
155 | |
156 | #ifdef PCRE_DEBUG |
157 | if (eptr >= md->end_subject) |
158 | printf("matching subject <null>"); |
159 | else |
160 | { |
161 | printf("matching subject "); |
162 | pchars(eptr, length, TRUE, md); |
163 | } |
164 | printf(" against backref "); |
165 | pchars(p, length, FALSE, md); |
166 | printf("\n"); |
167 | #endif |
168 | |
169 | /* Always fail if reference not set (and not JavaScript compatible - in that |
170 | case the length is passed as zero). */ |
171 | |
172 | if (length < 0) return -1; |
173 | |
174 | /* Separate the caseless case for speed. In UTF-8 mode we can only do this |
175 | properly if Unicode properties are supported. Otherwise, we can check only |
176 | ASCII characters. */ |
177 | |
178 | if (caseless) |
179 | { |
180 | #ifdef SUPPORT_UTF |
181 | #ifdef SUPPORT_UCP |
182 | if (utf) |
183 | { |
184 | /* Match characters up to the end of the reference. NOTE: the number of |
185 | data units matched may differ, because in UTF-8 there are some characters |
186 | whose upper and lower case versions code have different numbers of bytes. |
187 | For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65 |
188 | (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a |
189 | sequence of two of the latter. It is important, therefore, to check the |
190 | length along the reference, not along the subject (earlier code did this |
191 | wrong). */ |
192 | |
193 | PCRE_PUCHAR endptr = p + length; |
194 | while (p < endptr) |
195 | { |
196 | pcre_uint32 c, d; |
197 | const ucd_record *ur; |
198 | if (eptr >= md->end_subject) return -2; /* Partial match */ |
199 | GETCHARINC(c, eptr); |
200 | GETCHARINC(d, p); |
201 | ur = GET_UCD(d); |
202 | if (c != d && c != d + ur->other_case) |
203 | { |
204 | const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset; |
205 | for (;;) |
206 | { |
207 | if (c < *pp) return -1; |
208 | if (c == *pp++) break; |
209 | } |
210 | } |
211 | } |
212 | } |
213 | else |
214 | #endif |
215 | #endif |
216 | |
217 | /* The same code works when not in UTF-8 mode and in UTF-8 mode when there |
218 | is no UCP support. */ |
219 | { |
220 | while (length-- > 0) |
221 | { |
222 | pcre_uchar cc, cp; |
223 | if (eptr >= md->end_subject) return -2; /* Partial match */ |
224 | cc = RAWUCHARTEST(eptr); |
225 | cp = RAWUCHARTEST(p); |
226 | if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1; |
227 | p++; |
228 | eptr++; |
229 | } |
230 | } |
231 | } |
232 | |
233 | /* In the caseful case, we can just compare the bytes, whether or not we |
234 | are in UTF-8 mode. */ |
235 | |
236 | else |
237 | { |
238 | while (length-- > 0) |
239 | { |
240 | if (eptr >= md->end_subject) return -2; /* Partial match */ |
241 | if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1; |
242 | } |
243 | } |
244 | |
245 | return (int)(eptr - eptr_start); |
246 | } |
247 | |
248 | |
249 | |
250 | /*************************************************************************** |
251 | **************************************************************************** |
252 | RECURSION IN THE match() FUNCTION |
253 | |
254 | The match() function is highly recursive, though not every recursive call |
255 | increases the recursive depth. Nevertheless, some regular expressions can cause |
256 | it to recurse to a great depth. I was writing for Unix, so I just let it call |
257 | itself recursively. This uses the stack for saving everything that has to be |
258 | saved for a recursive call. On Unix, the stack can be large, and this works |
259 | fine. |
260 | |
261 | It turns out that on some non-Unix-like systems there are problems with |
262 | programs that use a lot of stack. (This despite the fact that every last chip |
263 | has oodles of memory these days, and techniques for extending the stack have |
264 | been known for decades.) So.... |
265 | |
266 | There is a fudge, triggered by defining NO_RECURSE, which avoids recursive |
267 | calls by keeping local variables that need to be preserved in blocks of memory |
268 | obtained from malloc() instead instead of on the stack. Macros are used to |
269 | achieve this so that the actual code doesn't look very different to what it |
270 | always used to. |
271 | |
272 | The original heap-recursive code used longjmp(). However, it seems that this |
273 | can be very slow on some operating systems. Following a suggestion from Stan |
274 | Switzer, the use of longjmp() has been abolished, at the cost of having to |
275 | provide a unique number for each call to RMATCH. There is no way of generating |
276 | a sequence of numbers at compile time in C. I have given them names, to make |
277 | them stand out more clearly. |
278 | |
279 | Crude tests on x86 Linux show a small speedup of around 5-8%. However, on |
280 | FreeBSD, avoiding longjmp() more than halves the time taken to run the standard |
281 | tests. Furthermore, not using longjmp() means that local dynamic variables |
282 | don't have indeterminate values; this has meant that the frame size can be |
283 | reduced because the result can be "passed back" by straight setting of the |
284 | variable instead of being passed in the frame. |
285 | **************************************************************************** |
286 | ***************************************************************************/ |
287 | |
288 | /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN |
289 | below must be updated in sync. */ |
290 | |
291 | enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10, |
292 | RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20, |
293 | RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30, |
294 | RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40, |
295 | RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50, |
296 | RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60, |
297 | RM61, RM62, RM63, RM64, RM65, RM66, RM67 }; |
298 | |
299 | /* These versions of the macros use the stack, as normal. There are debugging |
300 | versions and production versions. Note that the "rw" argument of RMATCH isn't |
301 | actually used in this definition. */ |
302 | |
303 | #ifndef NO_RECURSE |
304 | #define REGISTER register |
305 | |
306 | #ifdef PCRE_DEBUG |
307 | #define RMATCH(ra,rb,rc,rd,re,rw) \ |
308 | { \ |
309 | printf("match() called in line %d\n", __LINE__); \ |
310 | rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \ |
311 | printf("to line %d\n", __LINE__); \ |
312 | } |
313 | #define RRETURN(ra) \ |
314 | { \ |
315 | printf("match() returned %d from line %d\n", ra, __LINE__); \ |
316 | return ra; \ |
317 | } |
318 | #else |
319 | #define RMATCH(ra,rb,rc,rd,re,rw) \ |
320 | rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1) |
321 | #define RRETURN(ra) return ra |
322 | #endif |
323 | |
324 | #else |
325 | |
326 | |
327 | /* These versions of the macros manage a private stack on the heap. Note that |
328 | the "rd" argument of RMATCH isn't actually used in this definition. It's the md |
329 | argument of match(), which never changes. */ |
330 | |
331 | #define REGISTER |
332 | |
333 | #define RMATCH(ra,rb,rc,rd,re,rw)\ |
334 | {\ |
335 | heapframe *newframe = frame->Xnextframe;\ |
336 | if (newframe == NULL)\ |
337 | {\ |
338 | newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\ |
339 | if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\ |
340 | newframe->Xnextframe = NULL;\ |
341 | frame->Xnextframe = newframe;\ |
342 | }\ |
343 | frame->Xwhere = rw;\ |
344 | newframe->Xeptr = ra;\ |
345 | newframe->Xecode = rb;\ |
346 | newframe->Xmstart = mstart;\ |
347 | newframe->Xoffset_top = rc;\ |
348 | newframe->Xeptrb = re;\ |
349 | newframe->Xrdepth = frame->Xrdepth + 1;\ |
350 | newframe->Xprevframe = frame;\ |
351 | frame = newframe;\ |
352 | DPRINTF(("restarting from line %d\n", __LINE__));\ |
353 | goto HEAP_RECURSE;\ |
354 | L_##rw:\ |
355 | DPRINTF(("jumped back to line %d\n", __LINE__));\ |
356 | } |
357 | |
358 | #define RRETURN(ra)\ |
359 | {\ |
360 | heapframe *oldframe = frame;\ |
361 | frame = oldframe->Xprevframe;\ |
362 | if (frame != NULL)\ |
363 | {\ |
364 | rrc = ra;\ |
365 | goto HEAP_RETURN;\ |
366 | }\ |
367 | return ra;\ |
368 | } |
369 | |
370 | |
371 | /* Structure for remembering the local variables in a private frame */ |
372 | |
373 | typedef struct heapframe { |
374 | struct heapframe *Xprevframe; |
375 | struct heapframe *Xnextframe; |
376 | |
377 | /* Function arguments that may change */ |
378 | |
379 | PCRE_PUCHAR Xeptr; |
380 | const pcre_uchar *Xecode; |
381 | PCRE_PUCHAR Xmstart; |
382 | int Xoffset_top; |
383 | eptrblock *Xeptrb; |
384 | unsigned int Xrdepth; |
385 | |
386 | /* Function local variables */ |
387 | |
388 | PCRE_PUCHAR Xcallpat; |
389 | #ifdef SUPPORT_UTF |
390 | PCRE_PUCHAR Xcharptr; |
391 | #endif |
392 | PCRE_PUCHAR Xdata; |
393 | PCRE_PUCHAR Xnext; |
394 | PCRE_PUCHAR Xpp; |
395 | PCRE_PUCHAR Xprev; |
396 | PCRE_PUCHAR Xsaved_eptr; |
397 | |
398 | recursion_info Xnew_recursive; |
399 | |
400 | BOOL Xcur_is_word; |
401 | BOOL Xcondition; |
402 | BOOL Xprev_is_word; |
403 | |
404 | #ifdef SUPPORT_UCP |
405 | int Xprop_type; |
406 | unsigned int Xprop_value; |
407 | int Xprop_fail_result; |
408 | int Xoclength; |
409 | pcre_uchar Xocchars[6]; |
410 | #endif |
411 | |
412 | int Xcodelink; |
413 | int Xctype; |
414 | unsigned int Xfc; |
415 | int Xfi; |
416 | int Xlength; |
417 | int Xmax; |
418 | int Xmin; |
419 | int Xnumber; |
420 | int Xoffset; |
421 | int Xop; |
422 | int Xsave_capture_last; |
423 | int Xsave_offset1, Xsave_offset2, Xsave_offset3; |
424 | int Xstacksave[REC_STACK_SAVE_MAX]; |
425 | |
426 | eptrblock Xnewptrb; |
427 | |
428 | /* Where to jump back to */ |
429 | |
430 | int Xwhere; |
431 | |
432 | } heapframe; |
433 | |
434 | #endif |
435 | |
436 | |
437 | /*************************************************************************** |
438 | ***************************************************************************/ |
439 | |
440 | |
441 | |
442 | /************************************************* |
443 | * Match from current position * |
444 | *************************************************/ |
445 | |
446 | /* This function is called recursively in many circumstances. Whenever it |
447 | returns a negative (error) response, the outer incarnation must also return the |
448 | same response. */ |
449 | |
450 | /* These macros pack up tests that are used for partial matching, and which |
451 | appear several times in the code. We set the "hit end" flag if the pointer is |
452 | at the end of the subject and also past the start of the subject (i.e. |
453 | something has been matched). For hard partial matching, we then return |
454 | immediately. The second one is used when we already know we are past the end of |
455 | the subject. */ |
456 | |
457 | #define CHECK_PARTIAL()\ |
458 | if (md->partial != 0 && eptr >= md->end_subject && \ |
459 | eptr > md->start_used_ptr) \ |
460 | { \ |
461 | md->hitend = TRUE; \ |
462 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \ |
463 | } |
464 | |
465 | #define SCHECK_PARTIAL()\ |
466 | if (md->partial != 0 && eptr > md->start_used_ptr) \ |
467 | { \ |
468 | md->hitend = TRUE; \ |
469 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \ |
470 | } |
471 | |
472 | |
473 | /* Performance note: It might be tempting to extract commonly used fields from |
474 | the md structure (e.g. utf, end_subject) into individual variables to improve |
475 | performance. Tests using gcc on a SPARC disproved this; in the first case, it |
476 | made performance worse. |
477 | |
478 | Arguments: |
479 | eptr pointer to current character in subject |
480 | ecode pointer to current position in compiled code |
481 | mstart pointer to the current match start position (can be modified |
482 | by encountering \K) |
483 | offset_top current top pointer |
484 | md pointer to "static" info for the match |
485 | eptrb pointer to chain of blocks containing eptr at start of |
486 | brackets - for testing for empty matches |
487 | rdepth the recursion depth |
488 | |
489 | Returns: MATCH_MATCH if matched ) these values are >= 0 |
490 | MATCH_NOMATCH if failed to match ) |
491 | a negative MATCH_xxx value for PRUNE, SKIP, etc |
492 | a negative PCRE_ERROR_xxx value if aborted by an error condition |
493 | (e.g. stopped by repeated call or recursion limit) |
494 | */ |
495 | |
496 | static int |
497 | match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode, |
498 | PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb, |
499 | unsigned int rdepth) |
500 | { |
501 | /* These variables do not need to be preserved over recursion in this function, |
502 | so they can be ordinary variables in all cases. Mark some of them with |
503 | "register" because they are used a lot in loops. */ |
504 | |
505 | register int rrc; /* Returns from recursive calls */ |
506 | register int i; /* Used for loops not involving calls to RMATCH() */ |
507 | register pcre_uint32 c; /* Character values not kept over RMATCH() calls */ |
508 | register BOOL utf; /* Local copy of UTF flag for speed */ |
509 | |
510 | BOOL minimize, possessive; /* Quantifier options */ |
511 | BOOL caseless; |
512 | int condcode; |
513 | |
514 | /* When recursion is not being used, all "local" variables that have to be |
515 | preserved over calls to RMATCH() are part of a "frame". We set up the top-level |
516 | frame on the stack here; subsequent instantiations are obtained from the heap |
517 | whenever RMATCH() does a "recursion". See the macro definitions above. Putting |
518 | the top-level on the stack rather than malloc-ing them all gives a performance |
519 | boost in many cases where there is not much "recursion". */ |
520 | |
521 | #ifdef NO_RECURSE |
522 | heapframe *frame = (heapframe *)md->match_frames_base; |
523 | |
524 | /* Copy in the original argument variables */ |
525 | |
526 | frame->Xeptr = eptr; |
527 | frame->Xecode = ecode; |
528 | frame->Xmstart = mstart; |
529 | frame->Xoffset_top = offset_top; |
530 | frame->Xeptrb = eptrb; |
531 | frame->Xrdepth = rdepth; |
532 | |
533 | /* This is where control jumps back to to effect "recursion" */ |
534 | |
535 | HEAP_RECURSE: |
536 | |
537 | /* Macros make the argument variables come from the current frame */ |
538 | |
539 | #define eptr frame->Xeptr |
540 | #define ecode frame->Xecode |
541 | #define mstart frame->Xmstart |
542 | #define offset_top frame->Xoffset_top |
543 | #define eptrb frame->Xeptrb |
544 | #define rdepth frame->Xrdepth |
545 | |
546 | /* Ditto for the local variables */ |
547 | |
548 | #ifdef SUPPORT_UTF |
549 | #define charptr frame->Xcharptr |
550 | #endif |
551 | #define callpat frame->Xcallpat |
552 | #define codelink frame->Xcodelink |
553 | #define data frame->Xdata |
554 | #define next frame->Xnext |
555 | #define pp frame->Xpp |
556 | #define prev frame->Xprev |
557 | #define saved_eptr frame->Xsaved_eptr |
558 | |
559 | #define new_recursive frame->Xnew_recursive |
560 | |
561 | #define cur_is_word frame->Xcur_is_word |
562 | #define condition frame->Xcondition |
563 | #define prev_is_word frame->Xprev_is_word |
564 | |
565 | #ifdef SUPPORT_UCP |
566 | #define prop_type frame->Xprop_type |
567 | #define prop_value frame->Xprop_value |
568 | #define prop_fail_result frame->Xprop_fail_result |
569 | #define oclength frame->Xoclength |
570 | #define occhars frame->Xocchars |
571 | #endif |
572 | |
573 | #define ctype frame->Xctype |
574 | #define fc frame->Xfc |
575 | #define fi frame->Xfi |
576 | #define length frame->Xlength |
577 | #define max frame->Xmax |
578 | #define min frame->Xmin |
579 | #define number frame->Xnumber |
580 | #define offset frame->Xoffset |
581 | #define op frame->Xop |
582 | #define save_capture_last frame->Xsave_capture_last |
583 | #define save_offset1 frame->Xsave_offset1 |
584 | #define save_offset2 frame->Xsave_offset2 |
585 | #define save_offset3 frame->Xsave_offset3 |
586 | #define stacksave frame->Xstacksave |
587 | |
588 | #define newptrb frame->Xnewptrb |
589 | |
590 | /* When recursion is being used, local variables are allocated on the stack and |
591 | get preserved during recursion in the normal way. In this environment, fi and |
592 | i, and fc and c, can be the same variables. */ |
593 | |
594 | #else /* NO_RECURSE not defined */ |
595 | #define fi i |
596 | #define fc c |
597 | |
598 | /* Many of the following variables are used only in small blocks of the code. |
599 | My normal style of coding would have declared them within each of those blocks. |
600 | However, in order to accommodate the version of this code that uses an external |
601 | "stack" implemented on the heap, it is easier to declare them all here, so the |
602 | declarations can be cut out in a block. The only declarations within blocks |
603 | below are for variables that do not have to be preserved over a recursive call |
604 | to RMATCH(). */ |
605 | |
606 | #ifdef SUPPORT_UTF |
607 | const pcre_uchar *charptr; |
608 | #endif |
609 | const pcre_uchar *callpat; |
610 | const pcre_uchar *data; |
611 | const pcre_uchar *next; |
612 | PCRE_PUCHAR pp; |
613 | const pcre_uchar *prev; |
614 | PCRE_PUCHAR saved_eptr; |
615 | |
616 | recursion_info new_recursive; |
617 | |
618 | BOOL cur_is_word; |
619 | BOOL condition; |
620 | BOOL prev_is_word; |
621 | |
622 | #ifdef SUPPORT_UCP |
623 | int prop_type; |
624 | unsigned int prop_value; |
625 | int prop_fail_result; |
626 | int oclength; |
627 | pcre_uchar occhars[6]; |
628 | #endif |
629 | |
630 | int codelink; |
631 | int ctype; |
632 | int length; |
633 | int max; |
634 | int min; |
635 | unsigned int number; |
636 | int offset; |
637 | pcre_uchar op; |
638 | int save_capture_last; |
639 | int save_offset1, save_offset2, save_offset3; |
640 | int stacksave[REC_STACK_SAVE_MAX]; |
641 | |
642 | eptrblock newptrb; |
643 | |
644 | /* There is a special fudge for calling match() in a way that causes it to |
645 | measure the size of its basic stack frame when the stack is being used for |
646 | recursion. The second argument (ecode) being NULL triggers this behaviour. It |
647 | cannot normally ever be NULL. The return is the negated value of the frame |
648 | size. */ |
649 | |
650 | if (ecode == NULL) |
651 | { |
652 | if (rdepth == 0) |
653 | return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1); |
654 | else |
655 | { |
656 | int len = (char *)&rdepth - (char *)eptr; |
657 | return (len > 0)? -len : len; |
658 | } |
659 | } |
660 | #endif /* NO_RECURSE */ |
661 | |
662 | /* To save space on the stack and in the heap frame, I have doubled up on some |
663 | of the local variables that are used only in localised parts of the code, but |
664 | still need to be preserved over recursive calls of match(). These macros define |
665 | the alternative names that are used. */ |
666 | |
667 | #define allow_zero cur_is_word |
668 | #define cbegroup condition |
669 | #define code_offset codelink |
670 | #define condassert condition |
671 | #define matched_once prev_is_word |
672 | #define foc number |
673 | #define save_mark data |
674 | |
675 | /* These statements are here to stop the compiler complaining about unitialized |
676 | variables. */ |
677 | |
678 | #ifdef SUPPORT_UCP |
679 | prop_value = 0; |
680 | prop_fail_result = 0; |
681 | #endif |
682 | |
683 | |
684 | /* This label is used for tail recursion, which is used in a few cases even |
685 | when NO_RECURSE is not defined, in order to reduce the amount of stack that is |
686 | used. Thanks to Ian Taylor for noticing this possibility and sending the |
687 | original patch. */ |
688 | |
689 | TAIL_RECURSE: |
690 | |
691 | /* OK, now we can get on with the real code of the function. Recursive calls |
692 | are specified by the macro RMATCH and RRETURN is used to return. When |
693 | NO_RECURSE is *not* defined, these just turn into a recursive call to match() |
694 | and a "return", respectively (possibly with some debugging if PCRE_DEBUG is |
695 | defined). However, RMATCH isn't like a function call because it's quite a |
696 | complicated macro. It has to be used in one particular way. This shouldn't, |
697 | however, impact performance when true recursion is being used. */ |
698 | |
699 | #ifdef SUPPORT_UTF |
700 | utf = md->utf; /* Local copy of the flag */ |
701 | #else |
702 | utf = FALSE; |
703 | #endif |
704 | |
705 | /* First check that we haven't called match() too many times, or that we |
706 | haven't exceeded the recursive call limit. */ |
707 | |
708 | if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT); |
709 | if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT); |
710 | |
711 | /* At the start of a group with an unlimited repeat that may match an empty |
712 | string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is |
713 | done this way to save having to use another function argument, which would take |
714 | up space on the stack. See also MATCH_CONDASSERT below. |
715 | |
716 | When MATCH_CBEGROUP is set, add the current subject pointer to the chain of |
717 | such remembered pointers, to be checked when we hit the closing ket, in order |
718 | to break infinite loops that match no characters. When match() is called in |
719 | other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must |
720 | NOT be used with tail recursion, because the memory block that is used is on |
721 | the stack, so a new one may be required for each match(). */ |
722 | |
723 | if (md->match_function_type == MATCH_CBEGROUP) |
724 | { |
725 | newptrb.epb_saved_eptr = eptr; |
726 | newptrb.epb_prev = eptrb; |
727 | eptrb = &newptrb; |
728 | md->match_function_type = 0; |
729 | } |
730 | |
731 | /* Now start processing the opcodes. */ |
732 | |
733 | for (;;) |
734 | { |
735 | minimize = possessive = FALSE; |
736 | op = *ecode; |
737 | |
738 | switch(op) |
739 | { |
740 | case OP_MARK: |
741 | md->nomatch_mark = ecode + 2; |
742 | md->mark = NULL; /* In case previously set by assertion */ |
743 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
744 | eptrb, RM55); |
745 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
746 | md->mark == NULL) md->mark = ecode + 2; |
747 | |
748 | /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an |
749 | argument, and we must check whether that argument matches this MARK's |
750 | argument. It is passed back in md->start_match_ptr (an overloading of that |
751 | variable). If it does match, we reset that variable to the current subject |
752 | position and return MATCH_SKIP. Otherwise, pass back the return code |
753 | unaltered. */ |
754 | |
755 | else if (rrc == MATCH_SKIP_ARG && |
756 | STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0) |
757 | { |
758 | md->start_match_ptr = eptr; |
759 | RRETURN(MATCH_SKIP); |
760 | } |
761 | RRETURN(rrc); |
762 | |
763 | case OP_FAIL: |
764 | RRETURN(MATCH_NOMATCH); |
765 | |
766 | /* COMMIT overrides PRUNE, SKIP, and THEN */ |
767 | |
768 | case OP_COMMIT: |
769 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
770 | eptrb, RM52); |
771 | if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && |
772 | rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG && |
773 | rrc != MATCH_THEN) |
774 | RRETURN(rrc); |
775 | RRETURN(MATCH_COMMIT); |
776 | |
777 | /* PRUNE overrides THEN */ |
778 | |
779 | case OP_PRUNE: |
780 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
781 | eptrb, RM51); |
782 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
783 | RRETURN(MATCH_PRUNE); |
784 | |
785 | case OP_PRUNE_ARG: |
786 | md->nomatch_mark = ecode + 2; |
787 | md->mark = NULL; /* In case previously set by assertion */ |
788 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
789 | eptrb, RM56); |
790 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
791 | md->mark == NULL) md->mark = ecode + 2; |
792 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
793 | RRETURN(MATCH_PRUNE); |
794 | |
795 | /* SKIP overrides PRUNE and THEN */ |
796 | |
797 | case OP_SKIP: |
798 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
799 | eptrb, RM53); |
800 | if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN) |
801 | RRETURN(rrc); |
802 | md->start_match_ptr = eptr; /* Pass back current position */ |
803 | RRETURN(MATCH_SKIP); |
804 | |
805 | /* Note that, for Perl compatibility, SKIP with an argument does NOT set |
806 | nomatch_mark. There is a flag that disables this opcode when re-matching a |
807 | pattern that ended with a SKIP for which there was not a matching MARK. */ |
808 | |
809 | case OP_SKIP_ARG: |
810 | if (md->ignore_skip_arg) |
811 | { |
812 | ecode += PRIV(OP_lengths)[*ecode] + ecode[1]; |
813 | break; |
814 | } |
815 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
816 | eptrb, RM57); |
817 | if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN) |
818 | RRETURN(rrc); |
819 | |
820 | /* Pass back the current skip name by overloading md->start_match_ptr and |
821 | returning the special MATCH_SKIP_ARG return code. This will either be |
822 | caught by a matching MARK, or get to the top, where it causes a rematch |
823 | with the md->ignore_skip_arg flag set. */ |
824 | |
825 | md->start_match_ptr = ecode + 2; |
826 | RRETURN(MATCH_SKIP_ARG); |
827 | |
828 | /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that |
829 | the branch in which it occurs can be determined. Overload the start of |
830 | match pointer to do this. */ |
831 | |
832 | case OP_THEN: |
833 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
834 | eptrb, RM54); |
835 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
836 | md->start_match_ptr = ecode; |
837 | RRETURN(MATCH_THEN); |
838 | |
839 | case OP_THEN_ARG: |
840 | md->nomatch_mark = ecode + 2; |
841 | md->mark = NULL; /* In case previously set by assertion */ |
842 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, |
843 | md, eptrb, RM58); |
844 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
845 | md->mark == NULL) md->mark = ecode + 2; |
846 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
847 | md->start_match_ptr = ecode; |
848 | RRETURN(MATCH_THEN); |
849 | |
850 | /* Handle an atomic group that does not contain any capturing parentheses. |
851 | This can be handled like an assertion. Prior to 8.13, all atomic groups |
852 | were handled this way. In 8.13, the code was changed as below for ONCE, so |
853 | that backups pass through the group and thereby reset captured values. |
854 | However, this uses a lot more stack, so in 8.20, atomic groups that do not |
855 | contain any captures generate OP_ONCE_NC, which can be handled in the old, |
856 | less stack intensive way. |
857 | |
858 | Check the alternative branches in turn - the matching won't pass the KET |
859 | for this kind of subpattern. If any one branch matches, we carry on as at |
860 | the end of a normal bracket, leaving the subject pointer, but resetting |
861 | the start-of-match value in case it was changed by \K. */ |
862 | |
863 | case OP_ONCE_NC: |
864 | prev = ecode; |
865 | saved_eptr = eptr; |
866 | save_mark = md->mark; |
867 | do |
868 | { |
869 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64); |
870 | if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */ |
871 | { |
872 | mstart = md->start_match_ptr; |
873 | break; |
874 | } |
875 | if (rrc == MATCH_THEN) |
876 | { |
877 | next = ecode + GET(ecode,1); |
878 | if (md->start_match_ptr < next && |
879 | (*ecode == OP_ALT || *next == OP_ALT)) |
880 | rrc = MATCH_NOMATCH; |
881 | } |
882 | |
883 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
884 | ecode += GET(ecode,1); |
885 | md->mark = save_mark; |
886 | } |
887 | while (*ecode == OP_ALT); |
888 | |
889 | /* If hit the end of the group (which could be repeated), fail */ |
890 | |
891 | if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH); |
892 | |
893 | /* Continue as from after the group, updating the offsets high water |
894 | mark, since extracts may have been taken. */ |
895 | |
896 | do ecode += GET(ecode, 1); while (*ecode == OP_ALT); |
897 | |
898 | offset_top = md->end_offset_top; |
899 | eptr = md->end_match_ptr; |
900 | |
901 | /* For a non-repeating ket, just continue at this level. This also |
902 | happens for a repeating ket if no characters were matched in the group. |
903 | This is the forcible breaking of infinite loops as implemented in Perl |
904 | 5.005. */ |
905 | |
906 | if (*ecode == OP_KET || eptr == saved_eptr) |
907 | { |
908 | ecode += 1+LINK_SIZE; |
909 | break; |
910 | } |
911 | |
912 | /* The repeating kets try the rest of the pattern or restart from the |
913 | preceding bracket, in the appropriate order. The second "call" of match() |
914 | uses tail recursion, to avoid using another stack frame. */ |
915 | |
916 | if (*ecode == OP_KETRMIN) |
917 | { |
918 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65); |
919 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
920 | ecode = prev; |
921 | goto TAIL_RECURSE; |
922 | } |
923 | else /* OP_KETRMAX */ |
924 | { |
925 | RMATCH(eptr, prev, offset_top, md, eptrb, RM66); |
926 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
927 | ecode += 1 + LINK_SIZE; |
928 | goto TAIL_RECURSE; |
929 | } |
930 | /* Control never gets here */ |
931 | |
932 | /* Handle a capturing bracket, other than those that are possessive with an |
933 | unlimited repeat. If there is space in the offset vector, save the current |
934 | subject position in the working slot at the top of the vector. We mustn't |
935 | change the current values of the data slot, because they may be set from a |
936 | previous iteration of this group, and be referred to by a reference inside |
937 | the group. A failure to match might occur after the group has succeeded, |
938 | if something later on doesn't match. For this reason, we need to restore |
939 | the working value and also the values of the final offsets, in case they |
940 | were set by a previous iteration of the same bracket. |
941 | |
942 | If there isn't enough space in the offset vector, treat this as if it were |
943 | a non-capturing bracket. Don't worry about setting the flag for the error |
944 | case here; that is handled in the code for KET. */ |
945 | |
946 | case OP_CBRA: |
947 | case OP_SCBRA: |
948 | number = GET2(ecode, 1+LINK_SIZE); |
949 | offset = number << 1; |
950 | |
951 | #ifdef PCRE_DEBUG |
952 | printf("start bracket %d\n", number); |
953 | printf("subject="); |
954 | pchars(eptr, 16, TRUE, md); |
955 | printf("\n"); |
956 | #endif |
957 | |
958 | if (offset < md->offset_max) |
959 | { |
960 | save_offset1 = md->offset_vector[offset]; |
961 | save_offset2 = md->offset_vector[offset+1]; |
962 | save_offset3 = md->offset_vector[md->offset_end - number]; |
963 | save_capture_last = md->capture_last; |
964 | save_mark = md->mark; |
965 | |
966 | DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); |
967 | md->offset_vector[md->offset_end - number] = |
968 | (int)(eptr - md->start_subject); |
969 | |
970 | for (;;) |
971 | { |
972 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
973 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
974 | eptrb, RM1); |
975 | if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */ |
976 | |
977 | /* If we backed up to a THEN, check whether it is within the current |
978 | branch by comparing the address of the THEN that is passed back with |
979 | the end of the branch. If it is within the current branch, and the |
980 | branch is one of two or more alternatives (it either starts or ends |
981 | with OP_ALT), we have reached the limit of THEN's action, so convert |
982 | the return code to NOMATCH, which will cause normal backtracking to |
983 | happen from now on. Otherwise, THEN is passed back to an outer |
984 | alternative. This implements Perl's treatment of parenthesized groups, |
985 | where a group not containing | does not affect the current alternative, |
986 | that is, (X) is NOT the same as (X|(*F)). */ |
987 | |
988 | if (rrc == MATCH_THEN) |
989 | { |
990 | next = ecode + GET(ecode,1); |
991 | if (md->start_match_ptr < next && |
992 | (*ecode == OP_ALT || *next == OP_ALT)) |
993 | rrc = MATCH_NOMATCH; |
994 | } |
995 | |
996 | /* Anything other than NOMATCH is passed back. */ |
997 | |
998 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
999 | md->capture_last = save_capture_last; |
1000 | ecode += GET(ecode, 1); |
1001 | md->mark = save_mark; |
1002 | if (*ecode != OP_ALT) break; |
1003 | } |
1004 | |
1005 | DPRINTF(("bracket %d failed\n", number)); |
1006 | md->offset_vector[offset] = save_offset1; |
1007 | md->offset_vector[offset+1] = save_offset2; |
1008 | md->offset_vector[md->offset_end - number] = save_offset3; |
1009 | |
1010 | /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */ |
1011 | |
1012 | RRETURN(rrc); |
1013 | } |
1014 | |
1015 | /* FALL THROUGH ... Insufficient room for saving captured contents. Treat |
1016 | as a non-capturing bracket. */ |
1017 | |
1018 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1019 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1020 | |
1021 | DPRINTF(("insufficient capture room: treat as non-capturing\n")); |
1022 | |
1023 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1024 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1025 | |
1026 | /* Non-capturing or atomic group, except for possessive with unlimited |
1027 | repeat and ONCE group with no captures. Loop for all the alternatives. |
1028 | |
1029 | When we get to the final alternative within the brackets, we used to return |
1030 | the result of a recursive call to match() whatever happened so it was |
1031 | possible to reduce stack usage by turning this into a tail recursion, |
1032 | except in the case of a possibly empty group. However, now that there is |
1033 | the possiblity of (*THEN) occurring in the final alternative, this |
1034 | optimization is no longer always possible. |
1035 | |
1036 | We can optimize if we know there are no (*THEN)s in the pattern; at present |
1037 | this is the best that can be done. |
1038 | |
1039 | MATCH_ONCE is returned when the end of an atomic group is successfully |
1040 | reached, but subsequent matching fails. It passes back up the tree (causing |
1041 | captured values to be reset) until the original atomic group level is |
1042 | reached. This is tested by comparing md->once_target with the start of the |
1043 | group. At this point, the return is converted into MATCH_NOMATCH so that |
1044 | previous backup points can be taken. */ |
1045 | |
1046 | case OP_ONCE: |
1047 | case OP_BRA: |
1048 | case OP_SBRA: |
1049 | DPRINTF(("start non-capturing bracket\n")); |
1050 | |
1051 | for (;;) |
1052 | { |
1053 | if (op >= OP_SBRA || op == OP_ONCE) |
1054 | md->match_function_type = MATCH_CBEGROUP; |
1055 | |
1056 | /* If this is not a possibly empty group, and there are no (*THEN)s in |
1057 | the pattern, and this is the final alternative, optimize as described |
1058 | above. */ |
1059 | |
1060 | else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT) |
1061 | { |
1062 | ecode += PRIV(OP_lengths)[*ecode]; |
1063 | goto TAIL_RECURSE; |
1064 | } |
1065 | |
1066 | /* In all other cases, we have to make another call to match(). */ |
1067 | |
1068 | save_mark = md->mark; |
1069 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb, |
1070 | RM2); |
1071 | |
1072 | /* See comment in the code for capturing groups above about handling |
1073 | THEN. */ |
1074 | |
1075 | if (rrc == MATCH_THEN) |
1076 | { |
1077 | next = ecode + GET(ecode,1); |
1078 | if (md->start_match_ptr < next && |
1079 | (*ecode == OP_ALT || *next == OP_ALT)) |
1080 | rrc = MATCH_NOMATCH; |
1081 | } |
1082 | |
1083 | if (rrc != MATCH_NOMATCH) |
1084 | { |
1085 | if (rrc == MATCH_ONCE) |
1086 | { |
1087 | const pcre_uchar *scode = ecode; |
1088 | if (*scode != OP_ONCE) /* If not at start, find it */ |
1089 | { |
1090 | while (*scode == OP_ALT) scode += GET(scode, 1); |
1091 | scode -= GET(scode, 1); |
1092 | } |
1093 | if (md->once_target == scode) rrc = MATCH_NOMATCH; |
1094 | } |
1095 | RRETURN(rrc); |
1096 | } |
1097 | ecode += GET(ecode, 1); |
1098 | md->mark = save_mark; |
1099 | if (*ecode != OP_ALT) break; |
1100 | } |
1101 | |
1102 | RRETURN(MATCH_NOMATCH); |
1103 | |
1104 | /* Handle possessive capturing brackets with an unlimited repeat. We come |
1105 | here from BRAZERO with allow_zero set TRUE. The offset_vector values are |
1106 | handled similarly to the normal case above. However, the matching is |
1107 | different. The end of these brackets will always be OP_KETRPOS, which |
1108 | returns MATCH_KETRPOS without going further in the pattern. By this means |
1109 | we can handle the group by iteration rather than recursion, thereby |
1110 | reducing the amount of stack needed. */ |
1111 | |
1112 | case OP_CBRAPOS: |
1113 | case OP_SCBRAPOS: |
1114 | allow_zero = FALSE; |
1115 | |
1116 | POSSESSIVE_CAPTURE: |
1117 | number = GET2(ecode, 1+LINK_SIZE); |
1118 | offset = number << 1; |
1119 | |
1120 | #ifdef PCRE_DEBUG |
1121 | printf("start possessive bracket %d\n", number); |
1122 | printf("subject="); |
1123 | pchars(eptr, 16, TRUE, md); |
1124 | printf("\n"); |
1125 | #endif |
1126 | |
1127 | if (offset < md->offset_max) |
1128 | { |
1129 | matched_once = FALSE; |
1130 | code_offset = (int)(ecode - md->start_code); |
1131 | |
1132 | save_offset1 = md->offset_vector[offset]; |
1133 | save_offset2 = md->offset_vector[offset+1]; |
1134 | save_offset3 = md->offset_vector[md->offset_end - number]; |
1135 | save_capture_last = md->capture_last; |
1136 | |
1137 | DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); |
1138 | |
1139 | /* Each time round the loop, save the current subject position for use |
1140 | when the group matches. For MATCH_MATCH, the group has matched, so we |
1141 | restart it with a new subject starting position, remembering that we had |
1142 | at least one match. For MATCH_NOMATCH, carry on with the alternatives, as |
1143 | usual. If we haven't matched any alternatives in any iteration, check to |
1144 | see if a previous iteration matched. If so, the group has matched; |
1145 | continue from afterwards. Otherwise it has failed; restore the previous |
1146 | capture values before returning NOMATCH. */ |
1147 | |
1148 | for (;;) |
1149 | { |
1150 | md->offset_vector[md->offset_end - number] = |
1151 | (int)(eptr - md->start_subject); |
1152 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
1153 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
1154 | eptrb, RM63); |
1155 | if (rrc == MATCH_KETRPOS) |
1156 | { |
1157 | offset_top = md->end_offset_top; |
1158 | eptr = md->end_match_ptr; |
1159 | ecode = md->start_code + code_offset; |
1160 | save_capture_last = md->capture_last; |
1161 | matched_once = TRUE; |
1162 | continue; |
1163 | } |
1164 | |
1165 | /* See comment in the code for capturing groups above about handling |
1166 | THEN. */ |
1167 | |
1168 | if (rrc == MATCH_THEN) |
1169 | { |
1170 | next = ecode + GET(ecode,1); |
1171 | if (md->start_match_ptr < next && |
1172 | (*ecode == OP_ALT || *next == OP_ALT)) |
1173 | rrc = MATCH_NOMATCH; |
1174 | } |
1175 | |
1176 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1177 | md->capture_last = save_capture_last; |
1178 | ecode += GET(ecode, 1); |
1179 | if (*ecode != OP_ALT) break; |
1180 | } |
1181 | |
1182 | if (!matched_once) |
1183 | { |
1184 | md->offset_vector[offset] = save_offset1; |
1185 | md->offset_vector[offset+1] = save_offset2; |
1186 | md->offset_vector[md->offset_end - number] = save_offset3; |
1187 | } |
1188 | |
1189 | if (allow_zero || matched_once) |
1190 | { |
1191 | ecode += 1 + LINK_SIZE; |
1192 | break; |
1193 | } |
1194 | |
1195 | RRETURN(MATCH_NOMATCH); |
1196 | } |
1197 | |
1198 | /* FALL THROUGH ... Insufficient room for saving captured contents. Treat |
1199 | as a non-capturing bracket. */ |
1200 | |
1201 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1202 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1203 | |
1204 | DPRINTF(("insufficient capture room: treat as non-capturing\n")); |
1205 | |
1206 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1207 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1208 | |
1209 | /* Non-capturing possessive bracket with unlimited repeat. We come here |
1210 | from BRAZERO with allow_zero = TRUE. The code is similar to the above, |
1211 | without the capturing complication. It is written out separately for speed |
1212 | and cleanliness. */ |
1213 | |
1214 | case OP_BRAPOS: |
1215 | case OP_SBRAPOS: |
1216 | allow_zero = FALSE; |
1217 | |
1218 | POSSESSIVE_NON_CAPTURE: |
1219 | matched_once = FALSE; |
1220 | code_offset = (int)(ecode - md->start_code); |
1221 | |
1222 | for (;;) |
1223 | { |
1224 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
1225 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
1226 | eptrb, RM48); |
1227 | if (rrc == MATCH_KETRPOS) |
1228 | { |
1229 | offset_top = md->end_offset_top; |
1230 | eptr = md->end_match_ptr; |
1231 | ecode = md->start_code + code_offset; |
1232 | matched_once = TRUE; |
1233 | continue; |
1234 | } |
1235 | |
1236 | /* See comment in the code for capturing groups above about handling |
1237 | THEN. */ |
1238 | |
1239 | if (rrc == MATCH_THEN) |
1240 | { |
1241 | next = ecode + GET(ecode,1); |
1242 | if (md->start_match_ptr < next && |
1243 | (*ecode == OP_ALT || *next == OP_ALT)) |
1244 | rrc = MATCH_NOMATCH; |
1245 | } |
1246 | |
1247 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1248 | ecode += GET(ecode, 1); |
1249 | if (*ecode != OP_ALT) break; |
1250 | } |
1251 | |
1252 | if (matched_once || allow_zero) |
1253 | { |
1254 | ecode += 1 + LINK_SIZE; |
1255 | break; |
1256 | } |
1257 | RRETURN(MATCH_NOMATCH); |
1258 | |
1259 | /* Control never reaches here. */ |
1260 | |
1261 | /* Conditional group: compilation checked that there are no more than |
1262 | two branches. If the condition is false, skipping the first branch takes us |
1263 | past the end if there is only one branch, but that's OK because that is |
1264 | exactly what going to the ket would do. */ |
1265 | |
1266 | case OP_COND: |
1267 | case OP_SCOND: |
1268 | codelink = GET(ecode, 1); |
1269 | |
1270 | /* Because of the way auto-callout works during compile, a callout item is |
1271 | inserted between OP_COND and an assertion condition. */ |
1272 | |
1273 | if (ecode[LINK_SIZE+1] == OP_CALLOUT) |
1274 | { |
1275 | if (PUBL(callout) != NULL) |
1276 | { |
1277 | PUBL(callout_block) cb; |
1278 | cb.version = 2; /* Version 1 of the callout block */ |
1279 | cb.callout_number = ecode[LINK_SIZE+2]; |
1280 | cb.offset_vector = md->offset_vector; |
1281 | #if defined COMPILE_PCRE8 |
1282 | cb.subject = (PCRE_SPTR)md->start_subject; |
1283 | #elif defined COMPILE_PCRE16 |
1284 | cb.subject = (PCRE_SPTR16)md->start_subject; |
1285 | #elif defined COMPILE_PCRE32 |
1286 | cb.subject = (PCRE_SPTR32)md->start_subject; |
1287 | #endif |
1288 | cb.subject_length = (int)(md->end_subject - md->start_subject); |
1289 | cb.start_match = (int)(mstart - md->start_subject); |
1290 | cb.current_position = (int)(eptr - md->start_subject); |
1291 | cb.pattern_position = GET(ecode, LINK_SIZE + 3); |
1292 | cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE); |
1293 | cb.capture_top = offset_top/2; |
1294 | cb.capture_last = md->capture_last; |
1295 | cb.callout_data = md->callout_data; |
1296 | cb.mark = md->nomatch_mark; |
1297 | if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH); |
1298 | if (rrc < 0) RRETURN(rrc); |
1299 | } |
1300 | ecode += PRIV(OP_lengths)[OP_CALLOUT]; |
1301 | } |
1302 | |
1303 | condcode = ecode[LINK_SIZE+1]; |
1304 | |
1305 | /* Now see what the actual condition is */ |
1306 | |
1307 | if (condcode == OP_RREF || condcode == OP_NRREF) /* Recursion test */ |
1308 | { |
1309 | if (md->recursive == NULL) /* Not recursing => FALSE */ |
1310 | { |
1311 | condition = FALSE; |
1312 | ecode += GET(ecode, 1); |
1313 | } |
1314 | else |
1315 | { |
1316 | unsigned int recno = GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/ |
1317 | condition = (recno == RREF_ANY || recno == md->recursive->group_num); |
1318 | |
1319 | /* If the test is for recursion into a specific subpattern, and it is |
1320 | false, but the test was set up by name, scan the table to see if the |
1321 | name refers to any other numbers, and test them. The condition is true |
1322 | if any one is set. */ |
1323 | |
1324 | if (!condition && condcode == OP_NRREF) |
1325 | { |
1326 | pcre_uchar *slotA = md->name_table; |
1327 | for (i = 0; i < md->name_count; i++) |
1328 | { |
1329 | if (GET2(slotA, 0) == recno) break; |
1330 | slotA += md->name_entry_size; |
1331 | } |
1332 | |
1333 | /* Found a name for the number - there can be only one; duplicate |
1334 | names for different numbers are allowed, but not vice versa. First |
1335 | scan down for duplicates. */ |
1336 | |
1337 | if (i < md->name_count) |
1338 | { |
1339 | pcre_uchar *slotB = slotA; |
1340 | while (slotB > md->name_table) |
1341 | { |
1342 | slotB -= md->name_entry_size; |
1343 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1344 | { |
1345 | condition = GET2(slotB, 0) == md->recursive->group_num; |
1346 | if (condition) break; |
1347 | } |
1348 | else break; |
1349 | } |
1350 | |
1351 | /* Scan up for duplicates */ |
1352 | |
1353 | if (!condition) |
1354 | { |
1355 | slotB = slotA; |
1356 | for (i++; i < md->name_count; i++) |
1357 | { |
1358 | slotB += md->name_entry_size; |
1359 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1360 | { |
1361 | condition = GET2(slotB, 0) == md->recursive->group_num; |
1362 | if (condition) break; |
1363 | } |
1364 | else break; |
1365 | } |
1366 | } |
1367 | } |
1368 | } |
1369 | |
1370 | /* Chose branch according to the condition */ |
1371 | |
1372 | ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1); |
1373 | } |
1374 | } |
1375 | |
1376 | else if (condcode == OP_CREF || condcode == OP_NCREF) /* Group used test */ |
1377 | { |
1378 | offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */ |
1379 | condition = offset < offset_top && md->offset_vector[offset] >= 0; |
1380 | |
1381 | /* If the numbered capture is unset, but the reference was by name, |
1382 | scan the table to see if the name refers to any other numbers, and test |
1383 | them. The condition is true if any one is set. This is tediously similar |
1384 | to the code above, but not close enough to try to amalgamate. */ |
1385 | |
1386 | if (!condition && condcode == OP_NCREF) |
1387 | { |
1388 | unsigned int refno = offset >> 1; |
1389 | pcre_uchar *slotA = md->name_table; |
1390 | |
1391 | for (i = 0; i < md->name_count; i++) |
1392 | { |
1393 | if (GET2(slotA, 0) == refno) break; |
1394 | slotA += md->name_entry_size; |
1395 | } |
1396 | |
1397 | /* Found a name for the number - there can be only one; duplicate names |
1398 | for different numbers are allowed, but not vice versa. First scan down |
1399 | for duplicates. */ |
1400 | |
1401 | if (i < md->name_count) |
1402 | { |
1403 | pcre_uchar *slotB = slotA; |
1404 | while (slotB > md->name_table) |
1405 | { |
1406 | slotB -= md->name_entry_size; |
1407 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1408 | { |
1409 | offset = GET2(slotB, 0) << 1; |
1410 | condition = offset < offset_top && |
1411 | md->offset_vector[offset] >= 0; |
1412 | if (condition) break; |
1413 | } |
1414 | else break; |
1415 | } |
1416 | |
1417 | /* Scan up for duplicates */ |
1418 | |
1419 | if (!condition) |
1420 | { |
1421 | slotB = slotA; |
1422 | for (i++; i < md->name_count; i++) |
1423 | { |
1424 | slotB += md->name_entry_size; |
1425 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1426 | { |
1427 | offset = GET2(slotB, 0) << 1; |
1428 | condition = offset < offset_top && |
1429 | md->offset_vector[offset] >= 0; |
1430 | if (condition) break; |
1431 | } |
1432 | else break; |
1433 | } |
1434 | } |
1435 | } |
1436 | } |
1437 | |
1438 | /* Chose branch according to the condition */ |
1439 | |
1440 | ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1); |
1441 | } |
1442 | |
1443 | else if (condcode == OP_DEF) /* DEFINE - always false */ |
1444 | { |
1445 | condition = FALSE; |
1446 | ecode += GET(ecode, 1); |
1447 | } |
1448 | |
1449 | /* The condition is an assertion. Call match() to evaluate it - setting |
1450 | md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of |
1451 | an assertion. */ |
1452 | |
1453 | else |
1454 | { |
1455 | md->match_function_type = MATCH_CONDASSERT; |
1456 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3); |
1457 | if (rrc == MATCH_MATCH) |
1458 | { |
1459 | if (md->end_offset_top > offset_top) |
1460 | offset_top = md->end_offset_top; /* Captures may have happened */ |
1461 | condition = TRUE; |
1462 | ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2); |
1463 | while (*ecode == OP_ALT) ecode += GET(ecode, 1); |
1464 | } |
1465 | |
1466 | /* PCRE doesn't allow the effect of (*THEN) to escape beyond an |
1467 | assertion; it is therefore treated as NOMATCH. */ |
1468 | |
1469 | else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) |
1470 | { |
1471 | RRETURN(rrc); /* Need braces because of following else */ |
1472 | } |
1473 | else |
1474 | { |
1475 | condition = FALSE; |
1476 | ecode += codelink; |
1477 | } |
1478 | } |
1479 | |
1480 | /* We are now at the branch that is to be obeyed. As there is only one, can |
1481 | use tail recursion to avoid using another stack frame, except when there is |
1482 | unlimited repeat of a possibly empty group. In the latter case, a recursive |
1483 | call to match() is always required, unless the second alternative doesn't |
1484 | exist, in which case we can just plough on. Note that, for compatibility |
1485 | with Perl, the | in a conditional group is NOT treated as creating two |
1486 | alternatives. If a THEN is encountered in the branch, it propagates out to |
1487 | the enclosing alternative (unless nested in a deeper set of alternatives, |
1488 | of course). */ |
1489 | |
1490 | if (condition || *ecode == OP_ALT) |
1491 | { |
1492 | if (op != OP_SCOND) |
1493 | { |
1494 | ecode += 1 + LINK_SIZE; |
1495 | goto TAIL_RECURSE; |
1496 | } |
1497 | |
1498 | md->match_function_type = MATCH_CBEGROUP; |
1499 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49); |
1500 | RRETURN(rrc); |
1501 | } |
1502 | |
1503 | /* Condition false & no alternative; continue after the group. */ |
1504 | |
1505 | else |
1506 | { |
1507 | ecode += 1 + LINK_SIZE; |
1508 | } |
1509 | break; |
1510 | |
1511 | |
1512 | /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes, |
1513 | to close any currently open capturing brackets. */ |
1514 | |
1515 | case OP_CLOSE: |
1516 | number = GET2(ecode, 1); |
1517 | offset = number << 1; |
1518 | |
1519 | #ifdef PCRE_DEBUG |
1520 | printf("end bracket %d at *ACCEPT", number); |
1521 | printf("\n"); |
1522 | #endif |
1523 | |
1524 | md->capture_last = number; |
1525 | if (offset >= md->offset_max) md->offset_overflow = TRUE; else |
1526 | { |
1527 | md->offset_vector[offset] = |
1528 | md->offset_vector[md->offset_end - number]; |
1529 | md->offset_vector[offset+1] = (int)(eptr - md->start_subject); |
1530 | if (offset_top <= offset) offset_top = offset + 2; |
1531 | } |
1532 | ecode += 1 + IMM2_SIZE; |
1533 | break; |
1534 | |
1535 | |
1536 | /* End of the pattern, either real or forced. */ |
1537 | |
1538 | case OP_END: |
1539 | case OP_ACCEPT: |
1540 | case OP_ASSERT_ACCEPT: |
1541 | |
1542 | /* If we have matched an empty string, fail if not in an assertion and not |
1543 | in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART |
1544 | is set and we have matched at the start of the subject. In both cases, |
1545 | backtracking will then try other alternatives, if any. */ |
1546 | |
1547 | if (eptr == mstart && op != OP_ASSERT_ACCEPT && |
1548 | md->recursive == NULL && |
1549 | (md->notempty || |
1550 | (md->notempty_atstart && |
1551 | mstart == md->start_subject + md->start_offset))) |
1552 | RRETURN(MATCH_NOMATCH); |
1553 | |
1554 | /* Otherwise, we have a match. */ |
1555 | |
1556 | md->end_match_ptr = eptr; /* Record where we ended */ |
1557 | md->end_offset_top = offset_top; /* and how many extracts were taken */ |
1558 | md->start_match_ptr = mstart; /* and the start (\K can modify) */ |
1559 | |
1560 | /* For some reason, the macros don't work properly if an expression is |
1561 | given as the argument to RRETURN when the heap is in use. */ |
1562 | |
1563 | rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT; |
1564 | RRETURN(rrc); |
1565 | |
1566 | /* Assertion brackets. Check the alternative branches in turn - the |
1567 | matching won't pass the KET for an assertion. If any one branch matches, |
1568 | the assertion is true. Lookbehind assertions have an OP_REVERSE item at the |
1569 | start of each branch to move the current point backwards, so the code at |
1570 | this level is identical to the lookahead case. When the assertion is part |
1571 | of a condition, we want to return immediately afterwards. The caller of |
1572 | this incarnation of the match() function will have set MATCH_CONDASSERT in |
1573 | md->match_function type, and one of these opcodes will be the first opcode |
1574 | that is processed. We use a local variable that is preserved over calls to |
1575 | match() to remember this case. */ |
1576 | |
1577 | case OP_ASSERT: |
1578 | case OP_ASSERTBACK: |
1579 | save_mark = md->mark; |
1580 | if (md->match_function_type == MATCH_CONDASSERT) |
1581 | { |
1582 | condassert = TRUE; |
1583 | md->match_function_type = 0; |
1584 | } |
1585 | else condassert = FALSE; |
1586 | |
1587 | do |
1588 | { |
1589 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4); |
1590 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) |
1591 | { |
1592 | mstart = md->start_match_ptr; /* In case \K reset it */ |
1593 | break; |
1594 | } |
1595 | md->mark = save_mark; |
1596 | |
1597 | /* A COMMIT failure must fail the entire assertion, without trying any |
1598 | subsequent branches. */ |
1599 | |
1600 | if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH); |
1601 | |
1602 | /* PCRE does not allow THEN to escape beyond an assertion; it |
1603 | is treated as NOMATCH. */ |
1604 | |
1605 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
1606 | ecode += GET(ecode, 1); |
1607 | } |
1608 | while (*ecode == OP_ALT); |
1609 | |
1610 | if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH); |
1611 | |
1612 | /* If checking an assertion for a condition, return MATCH_MATCH. */ |
1613 | |
1614 | if (condassert) RRETURN(MATCH_MATCH); |
1615 | |
1616 | /* Continue from after the assertion, updating the offsets high water |
1617 | mark, since extracts may have been taken during the assertion. */ |
1618 | |
1619 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1620 | ecode += 1 + LINK_SIZE; |
1621 | offset_top = md->end_offset_top; |
1622 | continue; |
1623 | |
1624 | /* Negative assertion: all branches must fail to match. Encountering SKIP, |
1625 | PRUNE, or COMMIT means we must assume failure without checking subsequent |
1626 | branches. */ |
1627 | |
1628 | case OP_ASSERT_NOT: |
1629 | case OP_ASSERTBACK_NOT: |
1630 | save_mark = md->mark; |
1631 | if (md->match_function_type == MATCH_CONDASSERT) |
1632 | { |
1633 | condassert = TRUE; |
1634 | md->match_function_type = 0; |
1635 | } |
1636 | else condassert = FALSE; |
1637 | |
1638 | do |
1639 | { |
1640 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5); |
1641 | md->mark = save_mark; |
1642 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH); |
1643 | if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT) |
1644 | { |
1645 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1646 | break; |
1647 | } |
1648 | |
1649 | /* PCRE does not allow THEN to escape beyond an assertion; it is treated |
1650 | as NOMATCH. */ |
1651 | |
1652 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
1653 | ecode += GET(ecode,1); |
1654 | } |
1655 | while (*ecode == OP_ALT); |
1656 | |
1657 | if (condassert) RRETURN(MATCH_MATCH); /* Condition assertion */ |
1658 | |
1659 | ecode += 1 + LINK_SIZE; |
1660 | continue; |
1661 | |
1662 | /* Move the subject pointer back. This occurs only at the start of |
1663 | each branch of a lookbehind assertion. If we are too close to the start to |
1664 | move back, this match function fails. When working with UTF-8 we move |
1665 | back a number of characters, not bytes. */ |
1666 | |
1667 | case OP_REVERSE: |
1668 | #ifdef SUPPORT_UTF |
1669 | if (utf) |
1670 | { |
1671 | i = GET(ecode, 1); |
1672 | while (i-- > 0) |
1673 | { |
1674 | eptr--; |
1675 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); |
1676 | BACKCHAR(eptr); |
1677 | } |
1678 | } |
1679 | else |
1680 | #endif |
1681 | |
1682 | /* No UTF-8 support, or not in UTF-8 mode: count is byte count */ |
1683 | |
1684 | { |
1685 | eptr -= GET(ecode, 1); |
1686 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); |
1687 | } |
1688 | |
1689 | /* Save the earliest consulted character, then skip to next op code */ |
1690 | |
1691 | if (eptr < md->start_used_ptr) md->start_used_ptr = eptr; |
1692 | ecode += 1 + LINK_SIZE; |
1693 | break; |
1694 | |
1695 | /* The callout item calls an external function, if one is provided, passing |
1696 | details of the match so far. This is mainly for debugging, though the |
1697 | function is able to force a failure. */ |
1698 | |
1699 | case OP_CALLOUT: |
1700 | if (PUBL(callout) != NULL) |
1701 | { |
1702 | PUBL(callout_block) cb; |
1703 | cb.version = 2; /* Version 1 of the callout block */ |
1704 | cb.callout_number = ecode[1]; |
1705 | cb.offset_vector = md->offset_vector; |
1706 | #if defined COMPILE_PCRE8 |
1707 | cb.subject = (PCRE_SPTR)md->start_subject; |
1708 | #elif defined COMPILE_PCRE16 |
1709 | cb.subject = (PCRE_SPTR16)md->start_subject; |
1710 | #elif defined COMPILE_PCRE32 |
1711 | cb.subject = (PCRE_SPTR32)md->start_subject; |
1712 | #endif |
1713 | cb.subject_length = (int)(md->end_subject - md->start_subject); |
1714 | cb.start_match = (int)(mstart - md->start_subject); |
1715 | cb.current_position = (int)(eptr - md->start_subject); |
1716 | cb.pattern_position = GET(ecode, 2); |
1717 | cb.next_item_length = GET(ecode, 2 + LINK_SIZE); |
1718 | cb.capture_top = offset_top/2; |
1719 | cb.capture_last = md->capture_last; |
1720 | cb.callout_data = md->callout_data; |
1721 | cb.mark = md->nomatch_mark; |
1722 | if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH); |
1723 | if (rrc < 0) RRETURN(rrc); |
1724 | } |
1725 | ecode += 2 + 2*LINK_SIZE; |
1726 | break; |
1727 | |
1728 | /* Recursion either matches the current regex, or some subexpression. The |
1729 | offset data is the offset to the starting bracket from the start of the |
1730 | whole pattern. (This is so that it works from duplicated subpatterns.) |
1731 | |
1732 | The state of the capturing groups is preserved over recursion, and |
1733 | re-instated afterwards. We don't know how many are started and not yet |
1734 | finished (offset_top records the completed total) so we just have to save |
1735 | all the potential data. There may be up to 65535 such values, which is too |
1736 | large to put on the stack, but using malloc for small numbers seems |
1737 | expensive. As a compromise, the stack is used when there are no more than |
1738 | REC_STACK_SAVE_MAX values to store; otherwise malloc is used. |
1739 | |
1740 | There are also other values that have to be saved. We use a chained |
1741 | sequence of blocks that actually live on the stack. Thanks to Robin Houston |
1742 | for the original version of this logic. It has, however, been hacked around |
1743 | a lot, so he is not to blame for the current way it works. */ |
1744 | |
1745 | case OP_RECURSE: |
1746 | { |
1747 | recursion_info *ri; |
1748 | unsigned int recno; |
1749 | |
1750 | callpat = md->start_code + GET(ecode, 1); |
1751 | recno = (callpat == md->start_code)? 0 : |
1752 | GET2(callpat, 1 + LINK_SIZE); |
1753 | |
1754 | /* Check for repeating a recursion without advancing the subject pointer. |
1755 | This should catch convoluted mutual recursions. (Some simple cases are |
1756 | caught at compile time.) */ |
1757 | |
1758 | for (ri = md->recursive; ri != NULL; ri = ri->prevrec) |
1759 | if (recno == ri->group_num && eptr == ri->subject_position) |
1760 | RRETURN(PCRE_ERROR_RECURSELOOP); |
1761 | |
1762 | /* Add to "recursing stack" */ |
1763 | |
1764 | new_recursive.group_num = recno; |
1765 | new_recursive.subject_position = eptr; |
1766 | new_recursive.prevrec = md->recursive; |
1767 | md->recursive = &new_recursive; |
1768 | |
1769 | /* Where to continue from afterwards */ |
1770 | |
1771 | ecode += 1 + LINK_SIZE; |
1772 | |
1773 | /* Now save the offset data */ |
1774 | |
1775 | new_recursive.saved_max = md->offset_end; |
1776 | if (new_recursive.saved_max <= REC_STACK_SAVE_MAX) |
1777 | new_recursive.offset_save = stacksave; |
1778 | else |
1779 | { |
1780 | new_recursive.offset_save = |
1781 | (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int)); |
1782 | if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY); |
1783 | } |
1784 | memcpy(new_recursive.offset_save, md->offset_vector, |
1785 | new_recursive.saved_max * sizeof(int)); |
1786 | |
1787 | /* OK, now we can do the recursion. After processing each alternative, |
1788 | restore the offset data. If there were nested recursions, md->recursive |
1789 | might be changed, so reset it before looping. */ |
1790 | |
1791 | DPRINTF(("Recursing into group %d\n", new_recursive.group_num)); |
1792 | cbegroup = (*callpat >= OP_SBRA); |
1793 | do |
1794 | { |
1795 | if (cbegroup) md->match_function_type = MATCH_CBEGROUP; |
1796 | RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top, |
1797 | md, eptrb, RM6); |
1798 | memcpy(md->offset_vector, new_recursive.offset_save, |
1799 | new_recursive.saved_max * sizeof(int)); |
1800 | md->recursive = new_recursive.prevrec; |
1801 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) |
1802 | { |
1803 | DPRINTF(("Recursion matched\n")); |
1804 | if (new_recursive.offset_save != stacksave) |
1805 | (PUBL(free))(new_recursive.offset_save); |
1806 | |
1807 | /* Set where we got to in the subject, and reset the start in case |
1808 | it was changed by \K. This *is* propagated back out of a recursion, |
1809 | for Perl compatibility. */ |
1810 | |
1811 | eptr = md->end_match_ptr; |
1812 | mstart = md->start_match_ptr; |
1813 | goto RECURSION_MATCHED; /* Exit loop; end processing */ |
1814 | } |
1815 | |
1816 | /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it |
1817 | is treated as NOMATCH. */ |
1818 | |
1819 | else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN && |
1820 | rrc != MATCH_COMMIT) |
1821 | { |
1822 | DPRINTF(("Recursion gave error %d\n", rrc)); |
1823 | if (new_recursive.offset_save != stacksave) |
1824 | (PUBL(free))(new_recursive.offset_save); |
1825 | RRETURN(rrc); |
1826 | } |
1827 | |
1828 | md->recursive = &new_recursive; |
1829 | callpat += GET(callpat, 1); |
1830 | } |
1831 | while (*callpat == OP_ALT); |
1832 | |
1833 | DPRINTF(("Recursion didn't match\n")); |
1834 | md->recursive = new_recursive.prevrec; |
1835 | if (new_recursive.offset_save != stacksave) |
1836 | (PUBL(free))(new_recursive.offset_save); |
1837 | RRETURN(MATCH_NOMATCH); |
1838 | } |
1839 | |
1840 | RECURSION_MATCHED: |
1841 | break; |
1842 | |
1843 | /* An alternation is the end of a branch; scan along to find the end of the |
1844 | bracketed group and go to there. */ |
1845 | |
1846 | case OP_ALT: |
1847 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1848 | break; |
1849 | |
1850 | /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group, |
1851 | indicating that it may occur zero times. It may repeat infinitely, or not |
1852 | at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets |
1853 | with fixed upper repeat limits are compiled as a number of copies, with the |
1854 | optional ones preceded by BRAZERO or BRAMINZERO. */ |
1855 | |
1856 | case OP_BRAZERO: |
1857 | next = ecode + 1; |
1858 | RMATCH(eptr, next, offset_top, md, eptrb, RM10); |
1859 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1860 | do next += GET(next, 1); while (*next == OP_ALT); |
1861 | ecode = next + 1 + LINK_SIZE; |
1862 | break; |
1863 | |
1864 | case OP_BRAMINZERO: |
1865 | next = ecode + 1; |
1866 | do next += GET(next, 1); while (*next == OP_ALT); |
1867 | RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11); |
1868 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1869 | ecode++; |
1870 | break; |
1871 | |
1872 | case OP_SKIPZERO: |
1873 | next = ecode+1; |
1874 | do next += GET(next,1); while (*next == OP_ALT); |
1875 | ecode = next + 1 + LINK_SIZE; |
1876 | break; |
1877 | |
1878 | /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything |
1879 | here; just jump to the group, with allow_zero set TRUE. */ |
1880 | |
1881 | case OP_BRAPOSZERO: |
1882 | op = *(++ecode); |
1883 | allow_zero = TRUE; |
1884 | if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE; |
1885 | goto POSSESSIVE_NON_CAPTURE; |
1886 | |
1887 | /* End of a group, repeated or non-repeating. */ |
1888 | |
1889 | case OP_KET: |
1890 | case OP_KETRMIN: |
1891 | case OP_KETRMAX: |
1892 | case OP_KETRPOS: |
1893 | prev = ecode - GET(ecode, 1); |
1894 | |
1895 | /* If this was a group that remembered the subject start, in order to break |
1896 | infinite repeats of empty string matches, retrieve the subject start from |
1897 | the chain. Otherwise, set it NULL. */ |
1898 | |
1899 | if (*prev >= OP_SBRA || *prev == OP_ONCE) |
1900 | { |
1901 | saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */ |
1902 | eptrb = eptrb->epb_prev; /* Backup to previous group */ |
1903 | } |
1904 | else saved_eptr = NULL; |
1905 | |
1906 | /* If we are at the end of an assertion group or a non-capturing atomic |
1907 | group, stop matching and return MATCH_MATCH, but record the current high |
1908 | water mark for use by positive assertions. We also need to record the match |
1909 | start in case it was changed by \K. */ |
1910 | |
1911 | if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) || |
1912 | *prev == OP_ONCE_NC) |
1913 | { |
1914 | md->end_match_ptr = eptr; /* For ONCE_NC */ |
1915 | md->end_offset_top = offset_top; |
1916 | md->start_match_ptr = mstart; |
1917 | RRETURN(MATCH_MATCH); /* Sets md->mark */ |
1918 | } |
1919 | |
1920 | /* For capturing groups we have to check the group number back at the start |
1921 | and if necessary complete handling an extraction by setting the offsets and |
1922 | bumping the high water mark. Whole-pattern recursion is coded as a recurse |
1923 | into group 0, so it won't be picked up here. Instead, we catch it when the |
1924 | OP_END is reached. Other recursion is handled here. We just have to record |
1925 | the current subject position and start match pointer and give a MATCH |
1926 | return. */ |
1927 | |
1928 | if (*prev == OP_CBRA || *prev == OP_SCBRA || |
1929 | *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS) |
1930 | { |
1931 | number = GET2(prev, 1+LINK_SIZE); |
1932 | offset = number << 1; |
1933 | |
1934 | #ifdef PCRE_DEBUG |
1935 | printf("end bracket %d", number); |
1936 | printf("\n"); |
1937 | #endif |
1938 | |
1939 | /* Handle a recursively called group. */ |
1940 | |
1941 | if (md->recursive != NULL && md->recursive->group_num == number) |
1942 | { |
1943 | md->end_match_ptr = eptr; |
1944 | md->start_match_ptr = mstart; |
1945 | RRETURN(MATCH_MATCH); |
1946 | } |
1947 | |
1948 | /* Deal with capturing */ |
1949 | |
1950 | md->capture_last = number; |
1951 | if (offset >= md->offset_max) md->offset_overflow = TRUE; else |
1952 | { |
1953 | /* If offset is greater than offset_top, it means that we are |
1954 | "skipping" a capturing group, and that group's offsets must be marked |
1955 | unset. In earlier versions of PCRE, all the offsets were unset at the |
1956 | start of matching, but this doesn't work because atomic groups and |
1957 | assertions can cause a value to be set that should later be unset. |
1958 | Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as |
1959 | part of the atomic group, but this is not on the final matching path, |
1960 | so must be unset when 2 is set. (If there is no group 2, there is no |
1961 | problem, because offset_top will then be 2, indicating no capture.) */ |
1962 | |
1963 | if (offset > offset_top) |
1964 | { |
1965 | register int *iptr = md->offset_vector + offset_top; |
1966 | register int *iend = md->offset_vector + offset; |
1967 | while (iptr < iend) *iptr++ = -1; |
1968 | } |
1969 | |
1970 | /* Now make the extraction */ |
1971 | |
1972 | md->offset_vector[offset] = |
1973 | md->offset_vector[md->offset_end - number]; |
1974 | md->offset_vector[offset+1] = (int)(eptr - md->start_subject); |
1975 | if (offset_top <= offset) offset_top = offset + 2; |
1976 | } |
1977 | } |
1978 | |
1979 | /* For an ordinary non-repeating ket, just continue at this level. This |
1980 | also happens for a repeating ket if no characters were matched in the |
1981 | group. This is the forcible breaking of infinite loops as implemented in |
1982 | Perl 5.005. For a non-repeating atomic group that includes captures, |
1983 | establish a backup point by processing the rest of the pattern at a lower |
1984 | level. If this results in a NOMATCH return, pass MATCH_ONCE back to the |
1985 | original OP_ONCE level, thereby bypassing intermediate backup points, but |
1986 | resetting any captures that happened along the way. */ |
1987 | |
1988 | if (*ecode == OP_KET || eptr == saved_eptr) |
1989 | { |
1990 | if (*prev == OP_ONCE) |
1991 | { |
1992 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12); |
1993 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1994 | md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */ |
1995 | RRETURN(MATCH_ONCE); |
1996 | } |
1997 | ecode += 1 + LINK_SIZE; /* Carry on at this level */ |
1998 | break; |
1999 | } |
2000 | |
2001 | /* OP_KETRPOS is a possessive repeating ket. Remember the current position, |
2002 | and return the MATCH_KETRPOS. This makes it possible to do the repeats one |
2003 | at a time from the outer level, thus saving stack. */ |
2004 | |
2005 | if (*ecode == OP_KETRPOS) |
2006 | { |
2007 | md->end_match_ptr = eptr; |
2008 | md->end_offset_top = offset_top; |
2009 | RRETURN(MATCH_KETRPOS); |
2010 | } |
2011 | |
2012 | /* The normal repeating kets try the rest of the pattern or restart from |
2013 | the preceding bracket, in the appropriate order. In the second case, we can |
2014 | use tail recursion to avoid using another stack frame, unless we have an |
2015 | an atomic group or an unlimited repeat of a group that can match an empty |
2016 | string. */ |
2017 | |
2018 | if (*ecode == OP_KETRMIN) |
2019 | { |
2020 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7); |
2021 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2022 | if (*prev == OP_ONCE) |
2023 | { |
2024 | RMATCH(eptr, prev, offset_top, md, eptrb, RM8); |
2025 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2026 | md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */ |
2027 | RRETURN(MATCH_ONCE); |
2028 | } |
2029 | if (*prev >= OP_SBRA) /* Could match an empty string */ |
2030 | { |
2031 | RMATCH(eptr, prev, offset_top, md, eptrb, RM50); |
2032 | RRETURN(rrc); |
2033 | } |
2034 | ecode = prev; |
2035 | goto TAIL_RECURSE; |
2036 | } |
2037 | else /* OP_KETRMAX */ |
2038 | { |
2039 | RMATCH(eptr, prev, offset_top, md, eptrb, RM13); |
2040 | if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH; |
2041 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2042 | if (*prev == OP_ONCE) |
2043 | { |
2044 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9); |
2045 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2046 | md->once_target = prev; |
2047 | RRETURN(MATCH_ONCE); |
2048 | } |
2049 | ecode += 1 + LINK_SIZE; |
2050 | goto TAIL_RECURSE; |
2051 | } |
2052 | /* Control never gets here */ |
2053 | |
2054 | /* Not multiline mode: start of subject assertion, unless notbol. */ |
2055 | |
2056 | case OP_CIRC: |
2057 | if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH); |
2058 | |
2059 | /* Start of subject assertion */ |
2060 | |
2061 | case OP_SOD: |
2062 | if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH); |
2063 | ecode++; |
2064 | break; |
2065 | |
2066 | /* Multiline mode: start of subject unless notbol, or after any newline. */ |
2067 | |
2068 | case OP_CIRCM: |
2069 | if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH); |
2070 | if (eptr != md->start_subject && |
2071 | (eptr == md->end_subject || !WAS_NEWLINE(eptr))) |
2072 | RRETURN(MATCH_NOMATCH); |
2073 | ecode++; |
2074 | break; |
2075 | |
2076 | /* Start of match assertion */ |
2077 | |
2078 | case OP_SOM: |
2079 | if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH); |
2080 | ecode++; |
2081 | break; |
2082 | |
2083 | /* Reset the start of match point */ |
2084 | |
2085 | case OP_SET_SOM: |
2086 | mstart = eptr; |
2087 | ecode++; |
2088 | break; |
2089 | |
2090 | /* Multiline mode: assert before any newline, or before end of subject |
2091 | unless noteol is set. */ |
2092 | |
2093 | case OP_DOLLM: |
2094 | if (eptr < md->end_subject) |
2095 | { |
2096 | if (!IS_NEWLINE(eptr)) |
2097 | { |
2098 | if (md->partial != 0 && |
2099 | eptr + 1 >= md->end_subject && |
2100 | NLBLOCK->nltype == NLTYPE_FIXED && |
2101 | NLBLOCK->nllen == 2 && |
2102 | RAWUCHARTEST(eptr) == NLBLOCK->nl[0]) |
2103 | { |
2104 | md->hitend = TRUE; |
2105 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2106 | } |
2107 | RRETURN(MATCH_NOMATCH); |
2108 | } |
2109 | } |
2110 | else |
2111 | { |
2112 | if (md->noteol) RRETURN(MATCH_NOMATCH); |
2113 | SCHECK_PARTIAL(); |
2114 | } |
2115 | ecode++; |
2116 | break; |
2117 | |
2118 | /* Not multiline mode: assert before a terminating newline or before end of |
2119 | subject unless noteol is set. */ |
2120 | |
2121 | case OP_DOLL: |
2122 | if (md->noteol) RRETURN(MATCH_NOMATCH); |
2123 | if (!md->endonly) goto ASSERT_NL_OR_EOS; |
2124 | |
2125 | /* ... else fall through for endonly */ |
2126 | |
2127 | /* End of subject assertion (\z) */ |
2128 | |
2129 | case OP_EOD: |
2130 | if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH); |
2131 | SCHECK_PARTIAL(); |
2132 | ecode++; |
2133 | break; |
2134 | |
2135 | /* End of subject or ending \n assertion (\Z) */ |
2136 | |
2137 | case OP_EODN: |
2138 | ASSERT_NL_OR_EOS: |
2139 | if (eptr < md->end_subject && |
2140 | (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen)) |
2141 | { |
2142 | if (md->partial != 0 && |
2143 | eptr + 1 >= md->end_subject && |
2144 | NLBLOCK->nltype == NLTYPE_FIXED && |
2145 | NLBLOCK->nllen == 2 && |
2146 | RAWUCHARTEST(eptr) == NLBLOCK->nl[0]) |
2147 | { |
2148 | md->hitend = TRUE; |
2149 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2150 | } |
2151 | RRETURN(MATCH_NOMATCH); |
2152 | } |
2153 | |
2154 | /* Either at end of string or \n before end. */ |
2155 | |
2156 | SCHECK_PARTIAL(); |
2157 | ecode++; |
2158 | break; |
2159 | |
2160 | /* Word boundary assertions */ |
2161 | |
2162 | case OP_NOT_WORD_BOUNDARY: |
2163 | case OP_WORD_BOUNDARY: |
2164 | { |
2165 | |
2166 | /* Find out if the previous and current characters are "word" characters. |
2167 | It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to |
2168 | be "non-word" characters. Remember the earliest consulted character for |
2169 | partial matching. */ |
2170 | |
2171 | #ifdef SUPPORT_UTF |
2172 | if (utf) |
2173 | { |
2174 | /* Get status of previous character */ |
2175 | |
2176 | if (eptr == md->start_subject) prev_is_word = FALSE; else |
2177 | { |
2178 | PCRE_PUCHAR lastptr = eptr - 1; |
2179 | BACKCHAR(lastptr); |
2180 | if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr; |
2181 | GETCHAR(c, lastptr); |
2182 | #ifdef SUPPORT_UCP |
2183 | if (md->use_ucp) |
2184 | { |
2185 | if (c == '_') prev_is_word = TRUE; else |
2186 | { |
2187 | int cat = UCD_CATEGORY(c); |
2188 | prev_is_word = (cat == ucp_L || cat == ucp_N); |
2189 | } |
2190 | } |
2191 | else |
2192 | #endif |
2193 | prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; |
2194 | } |
2195 | |
2196 | /* Get status of next character */ |
2197 | |
2198 | if (eptr >= md->end_subject) |
2199 | { |
2200 | SCHECK_PARTIAL(); |
2201 | cur_is_word = FALSE; |
2202 | } |
2203 | else |
2204 | { |
2205 | GETCHAR(c, eptr); |
2206 | #ifdef SUPPORT_UCP |
2207 | if (md->use_ucp) |
2208 | { |
2209 | if (c == '_') cur_is_word = TRUE; else |
2210 | { |
2211 | int cat = UCD_CATEGORY(c); |
2212 | cur_is_word = (cat == ucp_L || cat == ucp_N); |
2213 | } |
2214 | } |
2215 | else |
2216 | #endif |
2217 | cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; |
2218 | } |
2219 | } |
2220 | else |
2221 | #endif |
2222 | |
2223 | /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for |
2224 | consistency with the behaviour of \w we do use it in this case. */ |
2225 | |
2226 | { |
2227 | /* Get status of previous character */ |
2228 | |
2229 | if (eptr == md->start_subject) prev_is_word = FALSE; else |
2230 | { |
2231 | if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1; |
2232 | #ifdef SUPPORT_UCP |
2233 | if (md->use_ucp) |
2234 | { |
2235 | c = eptr[-1]; |
2236 | if (c == '_') prev_is_word = TRUE; else |
2237 | { |
2238 | int cat = UCD_CATEGORY(c); |
2239 | prev_is_word = (cat == ucp_L || cat == ucp_N); |
2240 | } |
2241 | } |
2242 | else |
2243 | #endif |
2244 | prev_is_word = MAX_255(eptr[-1]) |
2245 | && ((md->ctypes[eptr[-1]] & ctype_word) != 0); |
2246 | } |
2247 | |
2248 | /* Get status of next character */ |
2249 | |
2250 | if (eptr >= md->end_subject) |
2251 | { |
2252 | SCHECK_PARTIAL(); |
2253 | cur_is_word = FALSE; |
2254 | } |
2255 | else |
2256 | #ifdef SUPPORT_UCP |
2257 | if (md->use_ucp) |
2258 | { |
2259 | c = *eptr; |
2260 | if (c == '_') cur_is_word = TRUE; else |
2261 | { |
2262 | int cat = UCD_CATEGORY(c); |
2263 | cur_is_word = (cat == ucp_L || cat == ucp_N); |
2264 | } |
2265 | } |
2266 | else |
2267 | #endif |
2268 | cur_is_word = MAX_255(*eptr) |
2269 | && ((md->ctypes[*eptr] & ctype_word) != 0); |
2270 | } |
2271 | |
2272 | /* Now see if the situation is what we want */ |
2273 | |
2274 | if ((*ecode++ == OP_WORD_BOUNDARY)? |
2275 | cur_is_word == prev_is_word : cur_is_word != prev_is_word) |
2276 | RRETURN(MATCH_NOMATCH); |
2277 | } |
2278 | break; |
2279 | |
2280 | /* Match any single character type except newline; have to take care with |
2281 | CRLF newlines and partial matching. */ |
2282 | |
2283 | case OP_ANY: |
2284 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
2285 | if (md->partial != 0 && |
2286 | eptr + 1 >= md->end_subject && |
2287 | NLBLOCK->nltype == NLTYPE_FIXED && |
2288 | NLBLOCK->nllen == 2 && |
2289 | RAWUCHARTEST(eptr) == NLBLOCK->nl[0]) |
2290 | { |
2291 | md->hitend = TRUE; |
2292 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2293 | } |
2294 | |
2295 | /* Fall through */ |
2296 | |
2297 | /* Match any single character whatsoever. */ |
2298 | |
2299 | case OP_ALLANY: |
2300 | if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */ |
2301 | { /* not be updated before SCHECK_PARTIAL. */ |
2302 | SCHECK_PARTIAL(); |
2303 | RRETURN(MATCH_NOMATCH); |
2304 | } |
2305 | eptr++; |
2306 | #ifdef SUPPORT_UTF |
2307 | if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
2308 | #endif |
2309 | ecode++; |
2310 | break; |
2311 | |
2312 | /* Match a single byte, even in UTF-8 mode. This opcode really does match |
2313 | any byte, even newline, independent of the setting of PCRE_DOTALL. */ |
2314 | |
2315 | case OP_ANYBYTE: |
2316 | if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */ |
2317 | { /* not be updated before SCHECK_PARTIAL. */ |
2318 | SCHECK_PARTIAL(); |
2319 | RRETURN(MATCH_NOMATCH); |
2320 | } |
2321 | eptr++; |
2322 | ecode++; |
2323 | break; |
2324 | |
2325 | case OP_NOT_DIGIT: |
2326 | if (eptr >= md->end_subject) |
2327 | { |
2328 | SCHECK_PARTIAL(); |
2329 | RRETURN(MATCH_NOMATCH); |
2330 | } |
2331 | GETCHARINCTEST(c, eptr); |
2332 | if ( |
2333 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2334 | c < 256 && |
2335 | #endif |
2336 | (md->ctypes[c] & ctype_digit) != 0 |
2337 | ) |
2338 | RRETURN(MATCH_NOMATCH); |
2339 | ecode++; |
2340 | break; |
2341 | |
2342 | case OP_DIGIT: |
2343 | if (eptr >= md->end_subject) |
2344 | { |
2345 | SCHECK_PARTIAL(); |
2346 | RRETURN(MATCH_NOMATCH); |
2347 | } |
2348 | GETCHARINCTEST(c, eptr); |
2349 | if ( |
2350 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2351 | c > 255 || |
2352 | #endif |
2353 | (md->ctypes[c] & ctype_digit) == 0 |
2354 | ) |
2355 | RRETURN(MATCH_NOMATCH); |
2356 | ecode++; |
2357 | break; |
2358 | |
2359 | case OP_NOT_WHITESPACE: |
2360 | if (eptr >= md->end_subject) |
2361 | { |
2362 | SCHECK_PARTIAL(); |
2363 | RRETURN(MATCH_NOMATCH); |
2364 | } |
2365 | GETCHARINCTEST(c, eptr); |
2366 | if ( |
2367 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2368 | c < 256 && |
2369 | #endif |
2370 | (md->ctypes[c] & ctype_space) != 0 |
2371 | ) |
2372 | RRETURN(MATCH_NOMATCH); |
2373 | ecode++; |
2374 | break; |
2375 | |
2376 | case OP_WHITESPACE: |
2377 | if (eptr >= md->end_subject) |
2378 | { |
2379 | SCHECK_PARTIAL(); |
2380 | RRETURN(MATCH_NOMATCH); |
2381 | } |
2382 | GETCHARINCTEST(c, eptr); |
2383 | if ( |
2384 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2385 | c > 255 || |
2386 | #endif |
2387 | (md->ctypes[c] & ctype_space) == 0 |
2388 | ) |
2389 | RRETURN(MATCH_NOMATCH); |
2390 | ecode++; |
2391 | break; |
2392 | |
2393 | case OP_NOT_WORDCHAR: |
2394 | if (eptr >= md->end_subject) |
2395 | { |
2396 | SCHECK_PARTIAL(); |
2397 | RRETURN(MATCH_NOMATCH); |
2398 | } |
2399 | GETCHARINCTEST(c, eptr); |
2400 | if ( |
2401 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2402 | c < 256 && |
2403 | #endif |
2404 | (md->ctypes[c] & ctype_word) != 0 |
2405 | ) |
2406 | RRETURN(MATCH_NOMATCH); |
2407 | ecode++; |
2408 | break; |
2409 | |
2410 | case OP_WORDCHAR: |
2411 | if (eptr >= md->end_subject) |
2412 | { |
2413 | SCHECK_PARTIAL(); |
2414 | RRETURN(MATCH_NOMATCH); |
2415 | } |
2416 | GETCHARINCTEST(c, eptr); |
2417 | if ( |
2418 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2419 | c > 255 || |
2420 | #endif |
2421 | (md->ctypes[c] & ctype_word) == 0 |
2422 | ) |
2423 | RRETURN(MATCH_NOMATCH); |
2424 | ecode++; |
2425 | break; |
2426 | |
2427 | case OP_ANYNL: |
2428 | if (eptr >= md->end_subject) |
2429 | { |
2430 | SCHECK_PARTIAL(); |
2431 | RRETURN(MATCH_NOMATCH); |
2432 | } |
2433 | GETCHARINCTEST(c, eptr); |
2434 | switch(c) |
2435 | { |
2436 | default: RRETURN(MATCH_NOMATCH); |
2437 | |
2438 | case CHAR_CR: |
2439 | if (eptr >= md->end_subject) |
2440 | { |
2441 | SCHECK_PARTIAL(); |
2442 | } |
2443 | else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++; |
2444 | break; |
2445 | |
2446 | case CHAR_LF: |
2447 | break; |
2448 | |
2449 | case CHAR_VT: |
2450 | case CHAR_FF: |
2451 | case CHAR_NEL: |
2452 | #ifndef EBCDIC |
2453 | case 0x2028: |
2454 | case 0x2029: |
2455 | #endif /* Not EBCDIC */ |
2456 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
2457 | break; |
2458 | } |
2459 | ecode++; |
2460 | break; |
2461 | |
2462 | case OP_NOT_HSPACE: |
2463 | if (eptr >= md->end_subject) |
2464 | { |
2465 | SCHECK_PARTIAL(); |
2466 | RRETURN(MATCH_NOMATCH); |
2467 | } |
2468 | GETCHARINCTEST(c, eptr); |
2469 | switch(c) |
2470 | { |
2471 | HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */ |
2472 | default: break; |
2473 | } |
2474 | ecode++; |
2475 | break; |
2476 | |
2477 | case OP_HSPACE: |
2478 | if (eptr >= md->end_subject) |
2479 | { |
2480 | SCHECK_PARTIAL(); |
2481 | RRETURN(MATCH_NOMATCH); |
2482 | } |
2483 | GETCHARINCTEST(c, eptr); |
2484 | switch(c) |
2485 | { |
2486 | HSPACE_CASES: break; /* Byte and multibyte cases */ |
2487 | default: RRETURN(MATCH_NOMATCH); |
2488 | } |
2489 | ecode++; |
2490 | break; |
2491 | |
2492 | case OP_NOT_VSPACE: |
2493 | if (eptr >= md->end_subject) |
2494 | { |
2495 | SCHECK_PARTIAL(); |
2496 | RRETURN(MATCH_NOMATCH); |
2497 | } |
2498 | GETCHARINCTEST(c, eptr); |
2499 | switch(c) |
2500 | { |
2501 | VSPACE_CASES: RRETURN(MATCH_NOMATCH); |
2502 | default: break; |
2503 | } |
2504 | ecode++; |
2505 | break; |
2506 | |
2507 | case OP_VSPACE: |
2508 | if (eptr >= md->end_subject) |
2509 | { |
2510 | SCHECK_PARTIAL(); |
2511 | RRETURN(MATCH_NOMATCH); |
2512 | } |
2513 | GETCHARINCTEST(c, eptr); |
2514 | switch(c) |
2515 | { |
2516 | VSPACE_CASES: break; |
2517 | default: RRETURN(MATCH_NOMATCH); |
2518 | } |
2519 | ecode++; |
2520 | break; |
2521 | |
2522 | #ifdef SUPPORT_UCP |
2523 | /* Check the next character by Unicode property. We will get here only |
2524 | if the support is in the binary; otherwise a compile-time error occurs. */ |
2525 | |
2526 | case OP_PROP: |
2527 | case OP_NOTPROP: |
2528 | if (eptr >= md->end_subject) |
2529 | { |
2530 | SCHECK_PARTIAL(); |
2531 | RRETURN(MATCH_NOMATCH); |
2532 | } |
2533 | GETCHARINCTEST(c, eptr); |
2534 | { |
2535 | const pcre_uint32 *cp; |
2536 | const ucd_record *prop = GET_UCD(c); |
2537 | |
2538 | switch(ecode[1]) |
2539 | { |
2540 | case PT_ANY: |
2541 | if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH); |
2542 | break; |
2543 | |
2544 | case PT_LAMP: |
2545 | if ((prop->chartype == ucp_Lu || |
2546 | prop->chartype == ucp_Ll || |
2547 | prop->chartype == ucp_Lt) == (op == OP_NOTPROP)) |
2548 | RRETURN(MATCH_NOMATCH); |
2549 | break; |
2550 | |
2551 | case PT_GC: |
2552 | if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP)) |
2553 | RRETURN(MATCH_NOMATCH); |
2554 | break; |
2555 | |
2556 | case PT_PC: |
2557 | if ((ecode[2] != prop->chartype) == (op == OP_PROP)) |
2558 | RRETURN(MATCH_NOMATCH); |
2559 | break; |
2560 | |
2561 | case PT_SC: |
2562 | if ((ecode[2] != prop->script) == (op == OP_PROP)) |
2563 | RRETURN(MATCH_NOMATCH); |
2564 | break; |
2565 | |
2566 | /* These are specials */ |
2567 | |
2568 | case PT_ALNUM: |
2569 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2570 | PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP)) |
2571 | RRETURN(MATCH_NOMATCH); |
2572 | break; |
2573 | |
2574 | case PT_SPACE: /* Perl space */ |
2575 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z || |
2576 | c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR) |
2577 | == (op == OP_NOTPROP)) |
2578 | RRETURN(MATCH_NOMATCH); |
2579 | break; |
2580 | |
2581 | case PT_PXSPACE: /* POSIX space */ |
2582 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z || |
2583 | c == CHAR_HT || c == CHAR_NL || c == CHAR_VT || |
2584 | c == CHAR_FF || c == CHAR_CR) |
2585 | == (op == OP_NOTPROP)) |
2586 | RRETURN(MATCH_NOMATCH); |
2587 | break; |
2588 | |
2589 | case PT_WORD: |
2590 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2591 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
2592 | c == CHAR_UNDERSCORE) == (op == OP_NOTPROP)) |
2593 | RRETURN(MATCH_NOMATCH); |
2594 | break; |
2595 | |
2596 | case PT_CLIST: |
2597 | cp = PRIV(ucd_caseless_sets) + ecode[2]; |
2598 | for (;;) |
2599 | { |
2600 | if (c < *cp) |
2601 | { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; } |
2602 | if (c == *cp++) |
2603 | { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } } |
2604 | } |
2605 | break; |
2606 | |
2607 | /* This should never occur */ |
2608 | |
2609 | default: |
2610 | RRETURN(PCRE_ERROR_INTERNAL); |
2611 | } |
2612 | |
2613 | ecode += 3; |
2614 | } |
2615 | break; |
2616 | |
2617 | /* Match an extended Unicode sequence. We will get here only if the support |
2618 | is in the binary; otherwise a compile-time error occurs. */ |
2619 | |
2620 | case OP_EXTUNI: |
2621 | if (eptr >= md->end_subject) |
2622 | { |
2623 | SCHECK_PARTIAL(); |
2624 | RRETURN(MATCH_NOMATCH); |
2625 | } |
2626 | else |
2627 | { |
2628 | int lgb, rgb; |
2629 | GETCHARINCTEST(c, eptr); |
2630 | lgb = UCD_GRAPHBREAK(c); |
2631 | while (eptr < md->end_subject) |
2632 | { |
2633 | int len = 1; |
2634 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
2635 | rgb = UCD_GRAPHBREAK(c); |
2636 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
2637 | lgb = rgb; |
2638 | eptr += len; |
2639 | } |
2640 | } |
2641 | CHECK_PARTIAL(); |
2642 | ecode++; |
2643 | break; |
2644 | #endif /* SUPPORT_UCP */ |
2645 | |
2646 | |
2647 | /* Match a back reference, possibly repeatedly. Look past the end of the |
2648 | item to see if there is repeat information following. The code is similar |
2649 | to that for character classes, but repeated for efficiency. Then obey |
2650 | similar code to character type repeats - written out again for speed. |
2651 | However, if the referenced string is the empty string, always treat |
2652 | it as matched, any number of times (otherwise there could be infinite |
2653 | loops). */ |
2654 | |
2655 | case OP_REF: |
2656 | case OP_REFI: |
2657 | caseless = op == OP_REFI; |
2658 | offset = GET2(ecode, 1) << 1; /* Doubled ref number */ |
2659 | ecode += 1 + IMM2_SIZE; |
2660 | |
2661 | /* If the reference is unset, there are two possibilities: |
2662 | |
2663 | (a) In the default, Perl-compatible state, set the length negative; |
2664 | this ensures that every attempt at a match fails. We can't just fail |
2665 | here, because of the possibility of quantifiers with zero minima. |
2666 | |
2667 | (b) If the JavaScript compatibility flag is set, set the length to zero |
2668 | so that the back reference matches an empty string. |
2669 | |
2670 | Otherwise, set the length to the length of what was matched by the |
2671 | referenced subpattern. */ |
2672 | |
2673 | if (offset >= offset_top || md->offset_vector[offset] < 0) |
2674 | length = (md->jscript_compat)? 0 : -1; |
2675 | else |
2676 | length = md->offset_vector[offset+1] - md->offset_vector[offset]; |
2677 | |
2678 | /* Set up for repetition, or handle the non-repeated case */ |
2679 | |
2680 | switch (*ecode) |
2681 | { |
2682 | case OP_CRSTAR: |
2683 | case OP_CRMINSTAR: |
2684 | case OP_CRPLUS: |
2685 | case OP_CRMINPLUS: |
2686 | case OP_CRQUERY: |
2687 | case OP_CRMINQUERY: |
2688 | c = *ecode++ - OP_CRSTAR; |
2689 | minimize = (c & 1) != 0; |
2690 | min = rep_min[c]; /* Pick up values from tables; */ |
2691 | max = rep_max[c]; /* zero for max => infinity */ |
2692 | if (max == 0) max = INT_MAX; |
2693 | break; |
2694 | |
2695 | case OP_CRRANGE: |
2696 | case OP_CRMINRANGE: |
2697 | minimize = (*ecode == OP_CRMINRANGE); |
2698 | min = GET2(ecode, 1); |
2699 | max = GET2(ecode, 1 + IMM2_SIZE); |
2700 | if (max == 0) max = INT_MAX; |
2701 | ecode += 1 + 2 * IMM2_SIZE; |
2702 | break; |
2703 | |
2704 | default: /* No repeat follows */ |
2705 | if ((length = match_ref(offset, eptr, length, md, caseless)) < 0) |
2706 | { |
2707 | if (length == -2) eptr = md->end_subject; /* Partial match */ |
2708 | CHECK_PARTIAL(); |
2709 | RRETURN(MATCH_NOMATCH); |
2710 | } |
2711 | eptr += length; |
2712 | continue; /* With the main loop */ |
2713 | } |
2714 | |
2715 | /* Handle repeated back references. If the length of the reference is |
2716 | zero, just continue with the main loop. If the length is negative, it |
2717 | means the reference is unset in non-Java-compatible mode. If the minimum is |
2718 | zero, we can continue at the same level without recursion. For any other |
2719 | minimum, carrying on will result in NOMATCH. */ |
2720 | |
2721 | if (length == 0) continue; |
2722 | if (length < 0 && min == 0) continue; |
2723 | |
2724 | /* First, ensure the minimum number of matches are present. We get back |
2725 | the length of the reference string explicitly rather than passing the |
2726 | address of eptr, so that eptr can be a register variable. */ |
2727 | |
2728 | for (i = 1; i <= min; i++) |
2729 | { |
2730 | int slength; |
2731 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2732 | { |
2733 | if (slength == -2) eptr = md->end_subject; /* Partial match */ |
2734 | CHECK_PARTIAL(); |
2735 | RRETURN(MATCH_NOMATCH); |
2736 | } |
2737 | eptr += slength; |
2738 | } |
2739 | |
2740 | /* If min = max, continue at the same level without recursion. |
2741 | They are not both allowed to be zero. */ |
2742 | |
2743 | if (min == max) continue; |
2744 | |
2745 | /* If minimizing, keep trying and advancing the pointer */ |
2746 | |
2747 | if (minimize) |
2748 | { |
2749 | for (fi = min;; fi++) |
2750 | { |
2751 | int slength; |
2752 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM14); |
2753 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2754 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
2755 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2756 | { |
2757 | if (slength == -2) eptr = md->end_subject; /* Partial match */ |
2758 | CHECK_PARTIAL(); |
2759 | RRETURN(MATCH_NOMATCH); |
2760 | } |
2761 | eptr += slength; |
2762 | } |
2763 | /* Control never gets here */ |
2764 | } |
2765 | |
2766 | /* If maximizing, find the longest string and work backwards */ |
2767 | |
2768 | else |
2769 | { |
2770 | pp = eptr; |
2771 | for (i = min; i < max; i++) |
2772 | { |
2773 | int slength; |
2774 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2775 | { |
2776 | /* Can't use CHECK_PARTIAL because we don't want to update eptr in |
2777 | the soft partial matching case. */ |
2778 | |
2779 | if (slength == -2 && md->partial != 0 && |
2780 | md->end_subject > md->start_used_ptr) |
2781 | { |
2782 | md->hitend = TRUE; |
2783 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2784 | } |
2785 | break; |
2786 | } |
2787 | eptr += slength; |
2788 | } |
2789 | |
2790 | while (eptr >= pp) |
2791 | { |
2792 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM15); |
2793 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2794 | eptr -= length; |
2795 | } |
2796 | RRETURN(MATCH_NOMATCH); |
2797 | } |
2798 | /* Control never gets here */ |
2799 | |
2800 | /* Match a bit-mapped character class, possibly repeatedly. This op code is |
2801 | used when all the characters in the class have values in the range 0-255, |
2802 | and either the matching is caseful, or the characters are in the range |
2803 | 0-127 when UTF-8 processing is enabled. The only difference between |
2804 | OP_CLASS and OP_NCLASS occurs when a data character outside the range is |
2805 | encountered. |
2806 | |
2807 | First, look past the end of the item to see if there is repeat information |
2808 | following. Then obey similar code to character type repeats - written out |
2809 | again for speed. */ |
2810 | |
2811 | case OP_NCLASS: |
2812 | case OP_CLASS: |
2813 | { |
2814 | /* The data variable is saved across frames, so the byte map needs to |
2815 | be stored there. */ |
2816 | #define BYTE_MAP ((pcre_uint8 *)data) |
2817 | data = ecode + 1; /* Save for matching */ |
2818 | ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */ |
2819 | |
2820 | switch (*ecode) |
2821 | { |
2822 | case OP_CRSTAR: |
2823 | case OP_CRMINSTAR: |
2824 | case OP_CRPLUS: |
2825 | case OP_CRMINPLUS: |
2826 | case OP_CRQUERY: |
2827 | case OP_CRMINQUERY: |
2828 | c = *ecode++ - OP_CRSTAR; |
2829 | minimize = (c & 1) != 0; |
2830 | min = rep_min[c]; /* Pick up values from tables; */ |
2831 | max = rep_max[c]; /* zero for max => infinity */ |
2832 | if (max == 0) max = INT_MAX; |
2833 | break; |
2834 | |
2835 | case OP_CRRANGE: |
2836 | case OP_CRMINRANGE: |
2837 | minimize = (*ecode == OP_CRMINRANGE); |
2838 | min = GET2(ecode, 1); |
2839 | max = GET2(ecode, 1 + IMM2_SIZE); |
2840 | if (max == 0) max = INT_MAX; |
2841 | ecode += 1 + 2 * IMM2_SIZE; |
2842 | break; |
2843 | |
2844 | default: /* No repeat follows */ |
2845 | min = max = 1; |
2846 | break; |
2847 | } |
2848 | |
2849 | /* First, ensure the minimum number of matches are present. */ |
2850 | |
2851 | #ifdef SUPPORT_UTF |
2852 | if (utf) |
2853 | { |
2854 | for (i = 1; i <= min; i++) |
2855 | { |
2856 | if (eptr >= md->end_subject) |
2857 | { |
2858 | SCHECK_PARTIAL(); |
2859 | RRETURN(MATCH_NOMATCH); |
2860 | } |
2861 | GETCHARINC(c, eptr); |
2862 | if (c > 255) |
2863 | { |
2864 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2865 | } |
2866 | else |
2867 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2868 | } |
2869 | } |
2870 | else |
2871 | #endif |
2872 | /* Not UTF mode */ |
2873 | { |
2874 | for (i = 1; i <= min; i++) |
2875 | { |
2876 | if (eptr >= md->end_subject) |
2877 | { |
2878 | SCHECK_PARTIAL(); |
2879 | RRETURN(MATCH_NOMATCH); |
2880 | } |
2881 | c = *eptr++; |
2882 | #ifndef COMPILE_PCRE8 |
2883 | if (c > 255) |
2884 | { |
2885 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2886 | } |
2887 | else |
2888 | #endif |
2889 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2890 | } |
2891 | } |
2892 | |
2893 | /* If max == min we can continue with the main loop without the |
2894 | need to recurse. */ |
2895 | |
2896 | if (min == max) continue; |
2897 | |
2898 | /* If minimizing, keep testing the rest of the expression and advancing |
2899 | the pointer while it matches the class. */ |
2900 | |
2901 | if (minimize) |
2902 | { |
2903 | #ifdef SUPPORT_UTF |
2904 | if (utf) |
2905 | { |
2906 | for (fi = min;; fi++) |
2907 | { |
2908 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM16); |
2909 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2910 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
2911 | if (eptr >= md->end_subject) |
2912 | { |
2913 | SCHECK_PARTIAL(); |
2914 | RRETURN(MATCH_NOMATCH); |
2915 | } |
2916 | GETCHARINC(c, eptr); |
2917 | if (c > 255) |
2918 | { |
2919 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2920 | } |
2921 | else |
2922 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2923 | } |
2924 | } |
2925 | else |
2926 | #endif |
2927 | /* Not UTF mode */ |
2928 | { |
2929 | for (fi = min;; fi++) |
2930 | { |
2931 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM17); |
2932 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2933 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
2934 | if (eptr >= md->end_subject) |
2935 | { |
2936 | SCHECK_PARTIAL(); |
2937 | RRETURN(MATCH_NOMATCH); |
2938 | } |
2939 | c = *eptr++; |
2940 | #ifndef COMPILE_PCRE8 |
2941 | if (c > 255) |
2942 | { |
2943 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2944 | } |
2945 | else |
2946 | #endif |
2947 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2948 | } |
2949 | } |
2950 | /* Control never gets here */ |
2951 | } |
2952 | |
2953 | /* If maximizing, find the longest possible run, then work backwards. */ |
2954 | |
2955 | else |
2956 | { |
2957 | pp = eptr; |
2958 | |
2959 | #ifdef SUPPORT_UTF |
2960 | if (utf) |
2961 | { |
2962 | for (i = min; i < max; i++) |
2963 | { |
2964 | int len = 1; |
2965 | if (eptr >= md->end_subject) |
2966 | { |
2967 | SCHECK_PARTIAL(); |
2968 | break; |
2969 | } |
2970 | GETCHARLEN(c, eptr, len); |
2971 | if (c > 255) |
2972 | { |
2973 | if (op == OP_CLASS) break; |
2974 | } |
2975 | else |
2976 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break; |
2977 | eptr += len; |
2978 | } |
2979 | for (;;) |
2980 | { |
2981 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM18); |
2982 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2983 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
2984 | BACKCHAR(eptr); |
2985 | } |
2986 | } |
2987 | else |
2988 | #endif |
2989 | /* Not UTF mode */ |
2990 | { |
2991 | for (i = min; i < max; i++) |
2992 | { |
2993 | if (eptr >= md->end_subject) |
2994 | { |
2995 | SCHECK_PARTIAL(); |
2996 | break; |
2997 | } |
2998 | c = *eptr; |
2999 | #ifndef COMPILE_PCRE8 |
3000 | if (c > 255) |
3001 | { |
3002 | if (op == OP_CLASS) break; |
3003 | } |
3004 | else |
3005 | #endif |
3006 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break; |
3007 | eptr++; |
3008 | } |
3009 | while (eptr >= pp) |
3010 | { |
3011 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM19); |
3012 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3013 | eptr--; |
3014 | } |
3015 | } |
3016 | |
3017 | RRETURN(MATCH_NOMATCH); |
3018 | } |
3019 | #undef BYTE_MAP |
3020 | } |
3021 | /* Control never gets here */ |
3022 | |
3023 | |
3024 | /* Match an extended character class. This opcode is encountered only |
3025 | when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8 |
3026 | mode, because Unicode properties are supported in non-UTF-8 mode. */ |
3027 | |
3028 | #if defined SUPPORT_UTF || !defined COMPILE_PCRE8 |
3029 | case OP_XCLASS: |
3030 | { |
3031 | data = ecode + 1 + LINK_SIZE; /* Save for matching */ |
3032 | ecode += GET(ecode, 1); /* Advance past the item */ |
3033 | |
3034 | switch (*ecode) |
3035 | { |
3036 | case OP_CRSTAR: |
3037 | case OP_CRMINSTAR: |
3038 | case OP_CRPLUS: |
3039 | case OP_CRMINPLUS: |
3040 | case OP_CRQUERY: |
3041 | case OP_CRMINQUERY: |
3042 | c = *ecode++ - OP_CRSTAR; |
3043 | minimize = (c & 1) != 0; |
3044 | min = rep_min[c]; /* Pick up values from tables; */ |
3045 | max = rep_max[c]; /* zero for max => infinity */ |
3046 | if (max == 0) max = INT_MAX; |
3047 | break; |
3048 | |
3049 | case OP_CRRANGE: |
3050 | case OP_CRMINRANGE: |
3051 | minimize = (*ecode == OP_CRMINRANGE); |
3052 | min = GET2(ecode, 1); |
3053 | max = GET2(ecode, 1 + IMM2_SIZE); |
3054 | if (max == 0) max = INT_MAX; |
3055 | ecode += 1 + 2 * IMM2_SIZE; |
3056 | break; |
3057 | |
3058 | default: /* No repeat follows */ |
3059 | min = max = 1; |
3060 | break; |
3061 | } |
3062 | |
3063 | /* First, ensure the minimum number of matches are present. */ |
3064 | |
3065 | for (i = 1; i <= min; i++) |
3066 | { |
3067 | if (eptr >= md->end_subject) |
3068 | { |
3069 | SCHECK_PARTIAL(); |
3070 | RRETURN(MATCH_NOMATCH); |
3071 | } |
3072 | GETCHARINCTEST(c, eptr); |
3073 | if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH); |
3074 | } |
3075 | |
3076 | /* If max == min we can continue with the main loop without the |
3077 | need to recurse. */ |
3078 | |
3079 | if (min == max) continue; |
3080 | |
3081 | /* If minimizing, keep testing the rest of the expression and advancing |
3082 | the pointer while it matches the class. */ |
3083 | |
3084 | if (minimize) |
3085 | { |
3086 | for (fi = min;; fi++) |
3087 | { |
3088 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM20); |
3089 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3090 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3091 | if (eptr >= md->end_subject) |
3092 | { |
3093 | SCHECK_PARTIAL(); |
3094 | RRETURN(MATCH_NOMATCH); |
3095 | } |
3096 | GETCHARINCTEST(c, eptr); |
3097 | if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH); |
3098 | } |
3099 | /* Control never gets here */ |
3100 | } |
3101 | |
3102 | /* If maximizing, find the longest possible run, then work backwards. */ |
3103 | |
3104 | else |
3105 | { |
3106 | pp = eptr; |
3107 | for (i = min; i < max; i++) |
3108 | { |
3109 | int len = 1; |
3110 | if (eptr >= md->end_subject) |
3111 | { |
3112 | SCHECK_PARTIAL(); |
3113 | break; |
3114 | } |
3115 | #ifdef SUPPORT_UTF |
3116 | GETCHARLENTEST(c, eptr, len); |
3117 | #else |
3118 | c = *eptr; |
3119 | #endif |
3120 | if (!PRIV(xclass)(c, data, utf)) break; |
3121 | eptr += len; |
3122 | } |
3123 | for(;;) |
3124 | { |
3125 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM21); |
3126 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3127 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3128 | #ifdef SUPPORT_UTF |
3129 | if (utf) BACKCHAR(eptr); |
3130 | #endif |
3131 | } |
3132 | RRETURN(MATCH_NOMATCH); |
3133 | } |
3134 | |
3135 | /* Control never gets here */ |
3136 | } |
3137 | #endif /* End of XCLASS */ |
3138 | |
3139 | /* Match a single character, casefully */ |
3140 | |
3141 | case OP_CHAR: |
3142 | #ifdef SUPPORT_UTF |
3143 | if (utf) |
3144 | { |
3145 | length = 1; |
3146 | ecode++; |
3147 | GETCHARLEN(fc, ecode, length); |
3148 | if (length > md->end_subject - eptr) |
3149 | { |
3150 | CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */ |
3151 | RRETURN(MATCH_NOMATCH); |
3152 | } |
3153 | while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH); |
3154 | } |
3155 | else |
3156 | #endif |
3157 | /* Not UTF mode */ |
3158 | { |
3159 | if (md->end_subject - eptr < 1) |
3160 | { |
3161 | SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */ |
3162 | RRETURN(MATCH_NOMATCH); |
3163 | } |
3164 | if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH); |
3165 | ecode += 2; |
3166 | } |
3167 | break; |
3168 | |
3169 | /* Match a single character, caselessly. If we are at the end of the |
3170 | subject, give up immediately. */ |
3171 | |
3172 | case OP_CHARI: |
3173 | if (eptr >= md->end_subject) |
3174 | { |
3175 | SCHECK_PARTIAL(); |
3176 | RRETURN(MATCH_NOMATCH); |
3177 | } |
3178 | |
3179 | #ifdef SUPPORT_UTF |
3180 | if (utf) |
3181 | { |
3182 | length = 1; |
3183 | ecode++; |
3184 | GETCHARLEN(fc, ecode, length); |
3185 | |
3186 | /* If the pattern character's value is < 128, we have only one byte, and |
3187 | we know that its other case must also be one byte long, so we can use the |
3188 | fast lookup table. We know that there is at least one byte left in the |
3189 | subject. */ |
3190 | |
3191 | if (fc < 128) |
3192 | { |
3193 | pcre_uchar cc = RAWUCHAR(eptr); |
3194 | if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH); |
3195 | ecode++; |
3196 | eptr++; |
3197 | } |
3198 | |
3199 | /* Otherwise we must pick up the subject character. Note that we cannot |
3200 | use the value of "length" to check for sufficient bytes left, because the |
3201 | other case of the character may have more or fewer bytes. */ |
3202 | |
3203 | else |
3204 | { |
3205 | pcre_uint32 dc; |
3206 | GETCHARINC(dc, eptr); |
3207 | ecode += length; |
3208 | |
3209 | /* If we have Unicode property support, we can use it to test the other |
3210 | case of the character, if there is one. */ |
3211 | |
3212 | if (fc != dc) |
3213 | { |
3214 | #ifdef SUPPORT_UCP |
3215 | if (dc != UCD_OTHERCASE(fc)) |
3216 | #endif |
3217 | RRETURN(MATCH_NOMATCH); |
3218 | } |
3219 | } |
3220 | } |
3221 | else |
3222 | #endif /* SUPPORT_UTF */ |
3223 | |
3224 | /* Not UTF mode */ |
3225 | { |
3226 | if (TABLE_GET(ecode[1], md->lcc, ecode[1]) |
3227 | != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH); |
3228 | eptr++; |
3229 | ecode += 2; |
3230 | } |
3231 | break; |
3232 | |
3233 | /* Match a single character repeatedly. */ |
3234 | |
3235 | case OP_EXACT: |
3236 | case OP_EXACTI: |
3237 | min = max = GET2(ecode, 1); |
3238 | ecode += 1 + IMM2_SIZE; |
3239 | goto REPEATCHAR; |
3240 | |
3241 | case OP_POSUPTO: |
3242 | case OP_POSUPTOI: |
3243 | possessive = TRUE; |
3244 | /* Fall through */ |
3245 | |
3246 | case OP_UPTO: |
3247 | case OP_UPTOI: |
3248 | case OP_MINUPTO: |
3249 | case OP_MINUPTOI: |
3250 | min = 0; |
3251 | max = GET2(ecode, 1); |
3252 | minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI; |
3253 | ecode += 1 + IMM2_SIZE; |
3254 | goto REPEATCHAR; |
3255 | |
3256 | case OP_POSSTAR: |
3257 | case OP_POSSTARI: |
3258 | possessive = TRUE; |
3259 | min = 0; |
3260 | max = INT_MAX; |
3261 | ecode++; |
3262 | goto REPEATCHAR; |
3263 | |
3264 | case OP_POSPLUS: |
3265 | case OP_POSPLUSI: |
3266 | possessive = TRUE; |
3267 | min = 1; |
3268 | max = INT_MAX; |
3269 | ecode++; |
3270 | goto REPEATCHAR; |
3271 | |
3272 | case OP_POSQUERY: |
3273 | case OP_POSQUERYI: |
3274 | possessive = TRUE; |
3275 | min = 0; |
3276 | max = 1; |
3277 | ecode++; |
3278 | goto REPEATCHAR; |
3279 | |
3280 | case OP_STAR: |
3281 | case OP_STARI: |
3282 | case OP_MINSTAR: |
3283 | case OP_MINSTARI: |
3284 | case OP_PLUS: |
3285 | case OP_PLUSI: |
3286 | case OP_MINPLUS: |
3287 | case OP_MINPLUSI: |
3288 | case OP_QUERY: |
3289 | case OP_QUERYI: |
3290 | case OP_MINQUERY: |
3291 | case OP_MINQUERYI: |
3292 | c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI); |
3293 | minimize = (c & 1) != 0; |
3294 | min = rep_min[c]; /* Pick up values from tables; */ |
3295 | max = rep_max[c]; /* zero for max => infinity */ |
3296 | if (max == 0) max = INT_MAX; |
3297 | |
3298 | /* Common code for all repeated single-character matches. */ |
3299 | |
3300 | REPEATCHAR: |
3301 | #ifdef SUPPORT_UTF |
3302 | if (utf) |
3303 | { |
3304 | length = 1; |
3305 | charptr = ecode; |
3306 | GETCHARLEN(fc, ecode, length); |
3307 | ecode += length; |
3308 | |
3309 | /* Handle multibyte character matching specially here. There is |
3310 | support for caseless matching if UCP support is present. */ |
3311 | |
3312 | if (length > 1) |
3313 | { |
3314 | #ifdef SUPPORT_UCP |
3315 | pcre_uint32 othercase; |
3316 | if (op >= OP_STARI && /* Caseless */ |
3317 | (othercase = UCD_OTHERCASE(fc)) != fc) |
3318 | oclength = PRIV(ord2utf)(othercase, occhars); |
3319 | else oclength = 0; |
3320 | #endif /* SUPPORT_UCP */ |
3321 | |
3322 | for (i = 1; i <= min; i++) |
3323 | { |
3324 | if (eptr <= md->end_subject - length && |
3325 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3326 | #ifdef SUPPORT_UCP |
3327 | else if (oclength > 0 && |
3328 | eptr <= md->end_subject - oclength && |
3329 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3330 | #endif /* SUPPORT_UCP */ |
3331 | else |
3332 | { |
3333 | CHECK_PARTIAL(); |
3334 | RRETURN(MATCH_NOMATCH); |
3335 | } |
3336 | } |
3337 | |
3338 | if (min == max) continue; |
3339 | |
3340 | if (minimize) |
3341 | { |
3342 | for (fi = min;; fi++) |
3343 | { |
3344 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM22); |
3345 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3346 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3347 | if (eptr <= md->end_subject - length && |
3348 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3349 | #ifdef SUPPORT_UCP |
3350 | else if (oclength > 0 && |
3351 | eptr <= md->end_subject - oclength && |
3352 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3353 | #endif /* SUPPORT_UCP */ |
3354 | else |
3355 | { |
3356 | CHECK_PARTIAL(); |
3357 | RRETURN(MATCH_NOMATCH); |
3358 | } |
3359 | } |
3360 | /* Control never gets here */ |
3361 | } |
3362 | |
3363 | else /* Maximize */ |
3364 | { |
3365 | pp = eptr; |
3366 | for (i = min; i < max; i++) |
3367 | { |
3368 | if (eptr <= md->end_subject - length && |
3369 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3370 | #ifdef SUPPORT_UCP |
3371 | else if (oclength > 0 && |
3372 | eptr <= md->end_subject - oclength && |
3373 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3374 | #endif /* SUPPORT_UCP */ |
3375 | else |
3376 | { |
3377 | CHECK_PARTIAL(); |
3378 | break; |
3379 | } |
3380 | } |
3381 | |
3382 | if (possessive) continue; |
3383 | |
3384 | for(;;) |
3385 | { |
3386 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM23); |
3387 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3388 | if (eptr == pp) { RRETURN(MATCH_NOMATCH); } |
3389 | #ifdef SUPPORT_UCP |
3390 | eptr--; |
3391 | BACKCHAR(eptr); |
3392 | #else /* without SUPPORT_UCP */ |
3393 | eptr -= length; |
3394 | #endif /* SUPPORT_UCP */ |
3395 | } |
3396 | } |
3397 | /* Control never gets here */ |
3398 | } |
3399 | |
3400 | /* If the length of a UTF-8 character is 1, we fall through here, and |
3401 | obey the code as for non-UTF-8 characters below, though in this case the |
3402 | value of fc will always be < 128. */ |
3403 | } |
3404 | else |
3405 | #endif /* SUPPORT_UTF */ |
3406 | /* When not in UTF-8 mode, load a single-byte character. */ |
3407 | fc = *ecode++; |
3408 | |
3409 | /* The value of fc at this point is always one character, though we may |
3410 | or may not be in UTF mode. The code is duplicated for the caseless and |
3411 | caseful cases, for speed, since matching characters is likely to be quite |
3412 | common. First, ensure the minimum number of matches are present. If min = |
3413 | max, continue at the same level without recursing. Otherwise, if |
3414 | minimizing, keep trying the rest of the expression and advancing one |
3415 | matching character if failing, up to the maximum. Alternatively, if |
3416 | maximizing, find the maximum number of characters and work backwards. */ |
3417 | |
3418 | DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max, |
3419 | max, (char *)eptr)); |
3420 | |
3421 | if (op >= OP_STARI) /* Caseless */ |
3422 | { |
3423 | #ifdef COMPILE_PCRE8 |
3424 | /* fc must be < 128 if UTF is enabled. */ |
3425 | foc = md->fcc[fc]; |
3426 | #else |
3427 | #ifdef SUPPORT_UTF |
3428 | #ifdef SUPPORT_UCP |
3429 | if (utf && fc > 127) |
3430 | foc = UCD_OTHERCASE(fc); |
3431 | #else |
3432 | if (utf && fc > 127) |
3433 | foc = fc; |
3434 | #endif /* SUPPORT_UCP */ |
3435 | else |
3436 | #endif /* SUPPORT_UTF */ |
3437 | foc = TABLE_GET(fc, md->fcc, fc); |
3438 | #endif /* COMPILE_PCRE8 */ |
3439 | |
3440 | for (i = 1; i <= min; i++) |
3441 | { |
3442 | pcre_uchar cc; |
3443 | |
3444 | if (eptr >= md->end_subject) |
3445 | { |
3446 | SCHECK_PARTIAL(); |
3447 | RRETURN(MATCH_NOMATCH); |
3448 | } |
3449 | cc = RAWUCHARTEST(eptr); |
3450 | if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH); |
3451 | eptr++; |
3452 | } |
3453 | if (min == max) continue; |
3454 | if (minimize) |
3455 | { |
3456 | for (fi = min;; fi++) |
3457 | { |
3458 | pcre_uchar cc; |
3459 | |
3460 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM24); |
3461 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3462 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3463 | if (eptr >= md->end_subject) |
3464 | { |
3465 | SCHECK_PARTIAL(); |
3466 | RRETURN(MATCH_NOMATCH); |
3467 | } |
3468 | cc = RAWUCHARTEST(eptr); |
3469 | if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH); |
3470 | eptr++; |
3471 | } |
3472 | /* Control never gets here */ |
3473 | } |
3474 | else /* Maximize */ |
3475 | { |
3476 | pp = eptr; |
3477 | for (i = min; i < max; i++) |
3478 | { |
3479 | pcre_uchar cc; |
3480 | |
3481 | if (eptr >= md->end_subject) |
3482 | { |
3483 | SCHECK_PARTIAL(); |
3484 | break; |
3485 | } |
3486 | cc = RAWUCHARTEST(eptr); |
3487 | if (fc != cc && foc != cc) break; |
3488 | eptr++; |
3489 | } |
3490 | |
3491 | if (possessive) continue; |
3492 | |
3493 | while (eptr >= pp) |
3494 | { |
3495 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM25); |
3496 | eptr--; |
3497 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3498 | } |
3499 | RRETURN(MATCH_NOMATCH); |
3500 | } |
3501 | /* Control never gets here */ |
3502 | } |
3503 | |
3504 | /* Caseful comparisons (includes all multi-byte characters) */ |
3505 | |
3506 | else |
3507 | { |
3508 | for (i = 1; i <= min; i++) |
3509 | { |
3510 | if (eptr >= md->end_subject) |
3511 | { |
3512 | SCHECK_PARTIAL(); |
3513 | RRETURN(MATCH_NOMATCH); |
3514 | } |
3515 | if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH); |
3516 | } |
3517 | |
3518 | if (min == max) continue; |
3519 | |
3520 | if (minimize) |
3521 | { |
3522 | for (fi = min;; fi++) |
3523 | { |
3524 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM26); |
3525 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3526 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3527 | if (eptr >= md->end_subject) |
3528 | { |
3529 | SCHECK_PARTIAL(); |
3530 | RRETURN(MATCH_NOMATCH); |
3531 | } |
3532 | if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH); |
3533 | } |
3534 | /* Control never gets here */ |
3535 | } |
3536 | else /* Maximize */ |
3537 | { |
3538 | pp = eptr; |
3539 | for (i = min; i < max; i++) |
3540 | { |
3541 | if (eptr >= md->end_subject) |
3542 | { |
3543 | SCHECK_PARTIAL(); |
3544 | break; |
3545 | } |
3546 | if (fc != RAWUCHARTEST(eptr)) break; |
3547 | eptr++; |
3548 | } |
3549 | if (possessive) continue; |
3550 | |
3551 | while (eptr >= pp) |
3552 | { |
3553 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM27); |
3554 | eptr--; |
3555 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3556 | } |
3557 | RRETURN(MATCH_NOMATCH); |
3558 | } |
3559 | } |
3560 | /* Control never gets here */ |
3561 | |
3562 | /* Match a negated single one-byte character. The character we are |
3563 | checking can be multibyte. */ |
3564 | |
3565 | case OP_NOT: |
3566 | case OP_NOTI: |
3567 | if (eptr >= md->end_subject) |
3568 | { |
3569 | SCHECK_PARTIAL(); |
3570 | RRETURN(MATCH_NOMATCH); |
3571 | } |
3572 | #ifdef SUPPORT_UTF |
3573 | if (utf) |
3574 | { |
3575 | register pcre_uint32 ch, och; |
3576 | |
3577 | ecode++; |
3578 | GETCHARINC(ch, ecode); |
3579 | GETCHARINC(c, eptr); |
3580 | |
3581 | if (op == OP_NOT) |
3582 | { |
3583 | if (ch == c) RRETURN(MATCH_NOMATCH); |
3584 | } |
3585 | else |
3586 | { |
3587 | #ifdef SUPPORT_UCP |
3588 | if (ch > 127) |
3589 | och = UCD_OTHERCASE(ch); |
3590 | #else |
3591 | if (ch > 127) |
3592 | och = ch; |
3593 | #endif /* SUPPORT_UCP */ |
3594 | else |
3595 | och = TABLE_GET(ch, md->fcc, ch); |
3596 | if (ch == c || och == c) RRETURN(MATCH_NOMATCH); |
3597 | } |
3598 | } |
3599 | else |
3600 | #endif |
3601 | { |
3602 | register pcre_uint32 ch = ecode[1]; |
3603 | c = *eptr++; |
3604 | if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c)) |
3605 | RRETURN(MATCH_NOMATCH); |
3606 | ecode += 2; |
3607 | } |
3608 | break; |
3609 | |
3610 | /* Match a negated single one-byte character repeatedly. This is almost a |
3611 | repeat of the code for a repeated single character, but I haven't found a |
3612 | nice way of commoning these up that doesn't require a test of the |
3613 | positive/negative option for each character match. Maybe that wouldn't add |
3614 | very much to the time taken, but character matching *is* what this is all |
3615 | about... */ |
3616 | |
3617 | case OP_NOTEXACT: |
3618 | case OP_NOTEXACTI: |
3619 | min = max = GET2(ecode, 1); |
3620 | ecode += 1 + IMM2_SIZE; |
3621 | goto REPEATNOTCHAR; |
3622 | |
3623 | case OP_NOTUPTO: |
3624 | case OP_NOTUPTOI: |
3625 | case OP_NOTMINUPTO: |
3626 | case OP_NOTMINUPTOI: |
3627 | min = 0; |
3628 | max = GET2(ecode, 1); |
3629 | minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI; |
3630 | ecode += 1 + IMM2_SIZE; |
3631 | goto REPEATNOTCHAR; |
3632 | |
3633 | case OP_NOTPOSSTAR: |
3634 | case OP_NOTPOSSTARI: |
3635 | possessive = TRUE; |
3636 | min = 0; |
3637 | max = INT_MAX; |
3638 | ecode++; |
3639 | goto REPEATNOTCHAR; |
3640 | |
3641 | case OP_NOTPOSPLUS: |
3642 | case OP_NOTPOSPLUSI: |
3643 | possessive = TRUE; |
3644 | min = 1; |
3645 | max = INT_MAX; |
3646 | ecode++; |
3647 | goto REPEATNOTCHAR; |
3648 | |
3649 | case OP_NOTPOSQUERY: |
3650 | case OP_NOTPOSQUERYI: |
3651 | possessive = TRUE; |
3652 | min = 0; |
3653 | max = 1; |
3654 | ecode++; |
3655 | goto REPEATNOTCHAR; |
3656 | |
3657 | case OP_NOTPOSUPTO: |
3658 | case OP_NOTPOSUPTOI: |
3659 | possessive = TRUE; |
3660 | min = 0; |
3661 | max = GET2(ecode, 1); |
3662 | ecode += 1 + IMM2_SIZE; |
3663 | goto REPEATNOTCHAR; |
3664 | |
3665 | case OP_NOTSTAR: |
3666 | case OP_NOTSTARI: |
3667 | case OP_NOTMINSTAR: |
3668 | case OP_NOTMINSTARI: |
3669 | case OP_NOTPLUS: |
3670 | case OP_NOTPLUSI: |
3671 | case OP_NOTMINPLUS: |
3672 | case OP_NOTMINPLUSI: |
3673 | case OP_NOTQUERY: |
3674 | case OP_NOTQUERYI: |
3675 | case OP_NOTMINQUERY: |
3676 | case OP_NOTMINQUERYI: |
3677 | c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR); |
3678 | minimize = (c & 1) != 0; |
3679 | min = rep_min[c]; /* Pick up values from tables; */ |
3680 | max = rep_max[c]; /* zero for max => infinity */ |
3681 | if (max == 0) max = INT_MAX; |
3682 | |
3683 | /* Common code for all repeated single-byte matches. */ |
3684 | |
3685 | REPEATNOTCHAR: |
3686 | GETCHARINCTEST(fc, ecode); |
3687 | |
3688 | /* The code is duplicated for the caseless and caseful cases, for speed, |
3689 | since matching characters is likely to be quite common. First, ensure the |
3690 | minimum number of matches are present. If min = max, continue at the same |
3691 | level without recursing. Otherwise, if minimizing, keep trying the rest of |
3692 | the expression and advancing one matching character if failing, up to the |
3693 | maximum. Alternatively, if maximizing, find the maximum number of |
3694 | characters and work backwards. */ |
3695 | |
3696 | DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max, |
3697 | max, (char *)eptr)); |
3698 | |
3699 | if (op >= OP_NOTSTARI) /* Caseless */ |
3700 | { |
3701 | #ifdef SUPPORT_UTF |
3702 | #ifdef SUPPORT_UCP |
3703 | if (utf && fc > 127) |
3704 | foc = UCD_OTHERCASE(fc); |
3705 | #else |
3706 | if (utf && fc > 127) |
3707 | foc = fc; |
3708 | #endif /* SUPPORT_UCP */ |
3709 | else |
3710 | #endif /* SUPPORT_UTF */ |
3711 | foc = TABLE_GET(fc, md->fcc, fc); |
3712 | |
3713 | #ifdef SUPPORT_UTF |
3714 | if (utf) |
3715 | { |
3716 | register pcre_uint32 d; |
3717 | for (i = 1; i <= min; i++) |
3718 | { |
3719 | if (eptr >= md->end_subject) |
3720 | { |
3721 | SCHECK_PARTIAL(); |
3722 | RRETURN(MATCH_NOMATCH); |
3723 | } |
3724 | GETCHARINC(d, eptr); |
3725 | if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH); |
3726 | } |
3727 | } |
3728 | else |
3729 | #endif |
3730 | /* Not UTF mode */ |
3731 | { |
3732 | for (i = 1; i <= min; i++) |
3733 | { |
3734 | if (eptr >= md->end_subject) |
3735 | { |
3736 | SCHECK_PARTIAL(); |
3737 | RRETURN(MATCH_NOMATCH); |
3738 | } |
3739 | if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH); |
3740 | eptr++; |
3741 | } |
3742 | } |
3743 | |
3744 | if (min == max) continue; |
3745 | |
3746 | if (minimize) |
3747 | { |
3748 | #ifdef SUPPORT_UTF |
3749 | if (utf) |
3750 | { |
3751 | register pcre_uint32 d; |
3752 | for (fi = min;; fi++) |
3753 | { |
3754 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM28); |
3755 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3756 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3757 | if (eptr >= md->end_subject) |
3758 | { |
3759 | SCHECK_PARTIAL(); |
3760 | RRETURN(MATCH_NOMATCH); |
3761 | } |
3762 | GETCHARINC(d, eptr); |
3763 | if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH); |
3764 | } |
3765 | } |
3766 | else |
3767 | #endif |
3768 | /* Not UTF mode */ |
3769 | { |
3770 | for (fi = min;; fi++) |
3771 | { |
3772 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM29); |
3773 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3774 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3775 | if (eptr >= md->end_subject) |
3776 | { |
3777 | SCHECK_PARTIAL(); |
3778 | RRETURN(MATCH_NOMATCH); |
3779 | } |
3780 | if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH); |
3781 | eptr++; |
3782 | } |
3783 | } |
3784 | /* Control never gets here */ |
3785 | } |
3786 | |
3787 | /* Maximize case */ |
3788 | |
3789 | else |
3790 | { |
3791 | pp = eptr; |
3792 | |
3793 | #ifdef SUPPORT_UTF |
3794 | if (utf) |
3795 | { |
3796 | register pcre_uint32 d; |
3797 | for (i = min; i < max; i++) |
3798 | { |
3799 | int len = 1; |
3800 | if (eptr >= md->end_subject) |
3801 | { |
3802 | SCHECK_PARTIAL(); |
3803 | break; |
3804 | } |
3805 | GETCHARLEN(d, eptr, len); |
3806 | if (fc == d || (unsigned int)foc == d) break; |
3807 | eptr += len; |
3808 | } |
3809 | if (possessive) continue; |
3810 | for(;;) |
3811 | { |
3812 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM30); |
3813 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3814 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3815 | BACKCHAR(eptr); |
3816 | } |
3817 | } |
3818 | else |
3819 | #endif |
3820 | /* Not UTF mode */ |
3821 | { |
3822 | for (i = min; i < max; i++) |
3823 | { |
3824 | if (eptr >= md->end_subject) |
3825 | { |
3826 | SCHECK_PARTIAL(); |
3827 | break; |
3828 | } |
3829 | if (fc == *eptr || foc == *eptr) break; |
3830 | eptr++; |
3831 | } |
3832 | if (possessive) continue; |
3833 | while (eptr >= pp) |
3834 | { |
3835 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM31); |
3836 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3837 | eptr--; |
3838 | } |
3839 | } |
3840 | |
3841 | RRETURN(MATCH_NOMATCH); |
3842 | } |
3843 | /* Control never gets here */ |
3844 | } |
3845 | |
3846 | /* Caseful comparisons */ |
3847 | |
3848 | else |
3849 | { |
3850 | #ifdef SUPPORT_UTF |
3851 | if (utf) |
3852 | { |
3853 | register pcre_uint32 d; |
3854 | for (i = 1; i <= min; i++) |
3855 | { |
3856 | if (eptr >= md->end_subject) |
3857 | { |
3858 | SCHECK_PARTIAL(); |
3859 | RRETURN(MATCH_NOMATCH); |
3860 | } |
3861 | GETCHARINC(d, eptr); |
3862 | if (fc == d) RRETURN(MATCH_NOMATCH); |
3863 | } |
3864 | } |
3865 | else |
3866 | #endif |
3867 | /* Not UTF mode */ |
3868 | { |
3869 | for (i = 1; i <= min; i++) |
3870 | { |
3871 | if (eptr >= md->end_subject) |
3872 | { |
3873 | SCHECK_PARTIAL(); |
3874 | RRETURN(MATCH_NOMATCH); |
3875 | } |
3876 | if (fc == *eptr++) RRETURN(MATCH_NOMATCH); |
3877 | } |
3878 | } |
3879 | |
3880 | if (min == max) continue; |
3881 | |
3882 | if (minimize) |
3883 | { |
3884 | #ifdef SUPPORT_UTF |
3885 | if (utf) |
3886 | { |
3887 | register pcre_uint32 d; |
3888 | for (fi = min;; fi++) |
3889 | { |
3890 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM32); |
3891 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3892 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3893 | if (eptr >= md->end_subject) |
3894 | { |
3895 | SCHECK_PARTIAL(); |
3896 | RRETURN(MATCH_NOMATCH); |
3897 | } |
3898 | GETCHARINC(d, eptr); |
3899 | if (fc == d) RRETURN(MATCH_NOMATCH); |
3900 | } |
3901 | } |
3902 | else |
3903 | #endif |
3904 | /* Not UTF mode */ |
3905 | { |
3906 | for (fi = min;; fi++) |
3907 | { |
3908 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM33); |
3909 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3910 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3911 | if (eptr >= md->end_subject) |
3912 | { |
3913 | SCHECK_PARTIAL(); |
3914 | RRETURN(MATCH_NOMATCH); |
3915 | } |
3916 | if (fc == *eptr++) RRETURN(MATCH_NOMATCH); |
3917 | } |
3918 | } |
3919 | /* Control never gets here */ |
3920 | } |
3921 | |
3922 | /* Maximize case */ |
3923 | |
3924 | else |
3925 | { |
3926 | pp = eptr; |
3927 | |
3928 | #ifdef SUPPORT_UTF |
3929 | if (utf) |
3930 | { |
3931 | register pcre_uint32 d; |
3932 | for (i = min; i < max; i++) |
3933 | { |
3934 | int len = 1; |
3935 | if (eptr >= md->end_subject) |
3936 | { |
3937 | SCHECK_PARTIAL(); |
3938 | break; |
3939 | } |
3940 | GETCHARLEN(d, eptr, len); |
3941 | if (fc == d) break; |
3942 | eptr += len; |
3943 | } |
3944 | if (possessive) continue; |
3945 | for(;;) |
3946 | { |
3947 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM34); |
3948 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3949 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3950 | BACKCHAR(eptr); |
3951 | } |
3952 | } |
3953 | else |
3954 | #endif |
3955 | /* Not UTF mode */ |
3956 | { |
3957 | for (i = min; i < max; i++) |
3958 | { |
3959 | if (eptr >= md->end_subject) |
3960 | { |
3961 | SCHECK_PARTIAL(); |
3962 | break; |
3963 | } |
3964 | if (fc == *eptr) break; |
3965 | eptr++; |
3966 | } |
3967 | if (possessive) continue; |
3968 | while (eptr >= pp) |
3969 | { |
3970 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM35); |
3971 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3972 | eptr--; |
3973 | } |
3974 | } |
3975 | |
3976 | RRETURN(MATCH_NOMATCH); |
3977 | } |
3978 | } |
3979 | /* Control never gets here */ |
3980 | |
3981 | /* Match a single character type repeatedly; several different opcodes |
3982 | share code. This is very similar to the code for single characters, but we |
3983 | repeat it in the interests of efficiency. */ |
3984 | |
3985 | case OP_TYPEEXACT: |
3986 | min = max = GET2(ecode, 1); |
3987 | minimize = TRUE; |
3988 | ecode += 1 + IMM2_SIZE; |
3989 | goto REPEATTYPE; |
3990 | |
3991 | case OP_TYPEUPTO: |
3992 | case OP_TYPEMINUPTO: |
3993 | min = 0; |
3994 | max = GET2(ecode, 1); |
3995 | minimize = *ecode == OP_TYPEMINUPTO; |
3996 | ecode += 1 + IMM2_SIZE; |
3997 | goto REPEATTYPE; |
3998 | |
3999 | case OP_TYPEPOSSTAR: |
4000 | possessive = TRUE; |
4001 | min = 0; |
4002 | max = INT_MAX; |
4003 | ecode++; |
4004 | goto REPEATTYPE; |
4005 | |
4006 | case OP_TYPEPOSPLUS: |
4007 | possessive = TRUE; |
4008 | min = 1; |
4009 | max = INT_MAX; |
4010 | ecode++; |
4011 | goto REPEATTYPE; |
4012 | |
4013 | case OP_TYPEPOSQUERY: |
4014 | possessive = TRUE; |
4015 | min = 0; |
4016 | max = 1; |
4017 | ecode++; |
4018 | goto REPEATTYPE; |
4019 | |
4020 | case OP_TYPEPOSUPTO: |
4021 | possessive = TRUE; |
4022 | min = 0; |
4023 | max = GET2(ecode, 1); |
4024 | ecode += 1 + IMM2_SIZE; |
4025 | goto REPEATTYPE; |
4026 | |
4027 | case OP_TYPESTAR: |
4028 | case OP_TYPEMINSTAR: |
4029 | case OP_TYPEPLUS: |
4030 | case OP_TYPEMINPLUS: |
4031 | case OP_TYPEQUERY: |
4032 | case OP_TYPEMINQUERY: |
4033 | c = *ecode++ - OP_TYPESTAR; |
4034 | minimize = (c & 1) != 0; |
4035 | min = rep_min[c]; /* Pick up values from tables; */ |
4036 | max = rep_max[c]; /* zero for max => infinity */ |
4037 | if (max == 0) max = INT_MAX; |
4038 | |
4039 | /* Common code for all repeated single character type matches. Note that |
4040 | in UTF-8 mode, '.' matches a character of any length, but for the other |
4041 | character types, the valid characters are all one-byte long. */ |
4042 | |
4043 | REPEATTYPE: |
4044 | ctype = *ecode++; /* Code for the character type */ |
4045 | |
4046 | #ifdef SUPPORT_UCP |
4047 | if (ctype == OP_PROP || ctype == OP_NOTPROP) |
4048 | { |
4049 | prop_fail_result = ctype == OP_NOTPROP; |
4050 | prop_type = *ecode++; |
4051 | prop_value = *ecode++; |
4052 | } |
4053 | else prop_type = -1; |
4054 | #endif |
4055 | |
4056 | /* First, ensure the minimum number of matches are present. Use inline |
4057 | code for maximizing the speed, and do the type test once at the start |
4058 | (i.e. keep it out of the loop). Separate the UTF-8 code completely as that |
4059 | is tidier. Also separate the UCP code, which can be the same for both UTF-8 |
4060 | and single-bytes. */ |
4061 | |
4062 | if (min > 0) |
4063 | { |
4064 | #ifdef SUPPORT_UCP |
4065 | if (prop_type >= 0) |
4066 | { |
4067 | switch(prop_type) |
4068 | { |
4069 | case PT_ANY: |
4070 | if (prop_fail_result) RRETURN(MATCH_NOMATCH); |
4071 | for (i = 1; i <= min; i++) |
4072 | { |
4073 | if (eptr >= md->end_subject) |
4074 | { |
4075 | SCHECK_PARTIAL(); |
4076 | RRETURN(MATCH_NOMATCH); |
4077 | } |
4078 | GETCHARINCTEST(c, eptr); |
4079 | } |
4080 | break; |
4081 | |
4082 | case PT_LAMP: |
4083 | for (i = 1; i <= min; i++) |
4084 | { |
4085 | int chartype; |
4086 | if (eptr >= md->end_subject) |
4087 | { |
4088 | SCHECK_PARTIAL(); |
4089 | RRETURN(MATCH_NOMATCH); |
4090 | } |
4091 | GETCHARINCTEST(c, eptr); |
4092 | chartype = UCD_CHARTYPE(c); |
4093 | if ((chartype == ucp_Lu || |
4094 | chartype == ucp_Ll || |
4095 | chartype == ucp_Lt) == prop_fail_result) |
4096 | RRETURN(MATCH_NOMATCH); |
4097 | } |
4098 | break; |
4099 | |
4100 | case PT_GC: |
4101 | for (i = 1; i <= min; i++) |
4102 | { |
4103 | if (eptr >= md->end_subject) |
4104 | { |
4105 | SCHECK_PARTIAL(); |
4106 | RRETURN(MATCH_NOMATCH); |
4107 | } |
4108 | GETCHARINCTEST(c, eptr); |
4109 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) |
4110 | RRETURN(MATCH_NOMATCH); |
4111 | } |
4112 | break; |
4113 | |
4114 | case PT_PC: |
4115 | for (i = 1; i <= min; i++) |
4116 | { |
4117 | if (eptr >= md->end_subject) |
4118 | { |
4119 | SCHECK_PARTIAL(); |
4120 | RRETURN(MATCH_NOMATCH); |
4121 | } |
4122 | GETCHARINCTEST(c, eptr); |
4123 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) |
4124 | RRETURN(MATCH_NOMATCH); |
4125 | } |
4126 | break; |
4127 | |
4128 | case PT_SC: |
4129 | for (i = 1; i <= min; i++) |
4130 | { |
4131 | if (eptr >= md->end_subject) |
4132 | { |
4133 | SCHECK_PARTIAL(); |
4134 | RRETURN(MATCH_NOMATCH); |
4135 | } |
4136 | GETCHARINCTEST(c, eptr); |
4137 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) |
4138 | RRETURN(MATCH_NOMATCH); |
4139 | } |
4140 | break; |
4141 | |
4142 | case PT_ALNUM: |
4143 | for (i = 1; i <= min; i++) |
4144 | { |
4145 | int category; |
4146 | if (eptr >= md->end_subject) |
4147 | { |
4148 | SCHECK_PARTIAL(); |
4149 | RRETURN(MATCH_NOMATCH); |
4150 | } |
4151 | GETCHARINCTEST(c, eptr); |
4152 | category = UCD_CATEGORY(c); |
4153 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
4154 | RRETURN(MATCH_NOMATCH); |
4155 | } |
4156 | break; |
4157 | |
4158 | case PT_SPACE: /* Perl space */ |
4159 | for (i = 1; i <= min; i++) |
4160 | { |
4161 | if (eptr >= md->end_subject) |
4162 | { |
4163 | SCHECK_PARTIAL(); |
4164 | RRETURN(MATCH_NOMATCH); |
4165 | } |
4166 | GETCHARINCTEST(c, eptr); |
4167 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4168 | c == CHAR_FF || c == CHAR_CR) |
4169 | == prop_fail_result) |
4170 | RRETURN(MATCH_NOMATCH); |
4171 | } |
4172 | break; |
4173 | |
4174 | case PT_PXSPACE: /* POSIX space */ |
4175 | for (i = 1; i <= min; i++) |
4176 | { |
4177 | if (eptr >= md->end_subject) |
4178 | { |
4179 | SCHECK_PARTIAL(); |
4180 | RRETURN(MATCH_NOMATCH); |
4181 | } |
4182 | GETCHARINCTEST(c, eptr); |
4183 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4184 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
4185 | == prop_fail_result) |
4186 | RRETURN(MATCH_NOMATCH); |
4187 | } |
4188 | break; |
4189 | |
4190 | case PT_WORD: |
4191 | for (i = 1; i <= min; i++) |
4192 | { |
4193 | int category; |
4194 | if (eptr >= md->end_subject) |
4195 | { |
4196 | SCHECK_PARTIAL(); |
4197 | RRETURN(MATCH_NOMATCH); |
4198 | } |
4199 | GETCHARINCTEST(c, eptr); |
4200 | category = UCD_CATEGORY(c); |
4201 | if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE) |
4202 | == prop_fail_result) |
4203 | RRETURN(MATCH_NOMATCH); |
4204 | } |
4205 | break; |
4206 | |
4207 | case PT_CLIST: |
4208 | for (i = 1; i <= min; i++) |
4209 | { |
4210 | const pcre_uint32 *cp; |
4211 | if (eptr >= md->end_subject) |
4212 | { |
4213 | SCHECK_PARTIAL(); |
4214 | RRETURN(MATCH_NOMATCH); |
4215 | } |
4216 | GETCHARINCTEST(c, eptr); |
4217 | cp = PRIV(ucd_caseless_sets) + prop_value; |
4218 | for (;;) |
4219 | { |
4220 | if (c < *cp) |
4221 | { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } } |
4222 | if (c == *cp++) |
4223 | { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; } |
4224 | } |
4225 | } |
4226 | break; |
4227 | |
4228 | /* This should not occur */ |
4229 | |
4230 | default: |
4231 | RRETURN(PCRE_ERROR_INTERNAL); |
4232 | } |
4233 | } |
4234 | |
4235 | /* Match extended Unicode sequences. We will get here only if the |
4236 | support is in the binary; otherwise a compile-time error occurs. */ |
4237 | |
4238 | else if (ctype == OP_EXTUNI) |
4239 | { |
4240 | for (i = 1; i <= min; i++) |
4241 | { |
4242 | if (eptr >= md->end_subject) |
4243 | { |
4244 | SCHECK_PARTIAL(); |
4245 | RRETURN(MATCH_NOMATCH); |
4246 | } |
4247 | else |
4248 | { |
4249 | int lgb, rgb; |
4250 | GETCHARINCTEST(c, eptr); |
4251 | lgb = UCD_GRAPHBREAK(c); |
4252 | while (eptr < md->end_subject) |
4253 | { |
4254 | int len = 1; |
4255 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
4256 | rgb = UCD_GRAPHBREAK(c); |
4257 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
4258 | lgb = rgb; |
4259 | eptr += len; |
4260 | } |
4261 | } |
4262 | CHECK_PARTIAL(); |
4263 | } |
4264 | } |
4265 | |
4266 | else |
4267 | #endif /* SUPPORT_UCP */ |
4268 | |
4269 | /* Handle all other cases when the coding is UTF-8 */ |
4270 | |
4271 | #ifdef SUPPORT_UTF |
4272 | if (utf) switch(ctype) |
4273 | { |
4274 | case OP_ANY: |
4275 | for (i = 1; i <= min; i++) |
4276 | { |
4277 | if (eptr >= md->end_subject) |
4278 | { |
4279 | SCHECK_PARTIAL(); |
4280 | RRETURN(MATCH_NOMATCH); |
4281 | } |
4282 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
4283 | if (md->partial != 0 && |
4284 | eptr + 1 >= md->end_subject && |
4285 | NLBLOCK->nltype == NLTYPE_FIXED && |
4286 | NLBLOCK->nllen == 2 && |
4287 | RAWUCHAR(eptr) == NLBLOCK->nl[0]) |
4288 | { |
4289 | md->hitend = TRUE; |
4290 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
4291 | } |
4292 | eptr++; |
4293 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4294 | } |
4295 | break; |
4296 | |
4297 | case OP_ALLANY: |
4298 | for (i = 1; i <= min; i++) |
4299 | { |
4300 | if (eptr >= md->end_subject) |
4301 | { |
4302 | SCHECK_PARTIAL(); |
4303 | RRETURN(MATCH_NOMATCH); |
4304 | } |
4305 | eptr++; |
4306 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4307 | } |
4308 | break; |
4309 | |
4310 | case OP_ANYBYTE: |
4311 | if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH); |
4312 | eptr += min; |
4313 | break; |
4314 | |
4315 | case OP_ANYNL: |
4316 | for (i = 1; i <= min; i++) |
4317 | { |
4318 | if (eptr >= md->end_subject) |
4319 | { |
4320 | SCHECK_PARTIAL(); |
4321 | RRETURN(MATCH_NOMATCH); |
4322 | } |
4323 | GETCHARINC(c, eptr); |
4324 | switch(c) |
4325 | { |
4326 | default: RRETURN(MATCH_NOMATCH); |
4327 | |
4328 | case CHAR_CR: |
4329 | if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++; |
4330 | break; |
4331 | |
4332 | case CHAR_LF: |
4333 | break; |
4334 | |
4335 | case CHAR_VT: |
4336 | case CHAR_FF: |
4337 | case CHAR_NEL: |
4338 | #ifndef EBCDIC |
4339 | case 0x2028: |
4340 | case 0x2029: |
4341 | #endif /* Not EBCDIC */ |
4342 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
4343 | break; |
4344 | } |
4345 | } |
4346 | break; |
4347 | |
4348 | case OP_NOT_HSPACE: |
4349 | for (i = 1; i <= min; i++) |
4350 | { |
4351 | if (eptr >= md->end_subject) |
4352 | { |
4353 | SCHECK_PARTIAL(); |
4354 | RRETURN(MATCH_NOMATCH); |
4355 | } |
4356 | GETCHARINC(c, eptr); |
4357 | switch(c) |
4358 | { |
4359 | HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */ |
4360 | default: break; |
4361 | } |
4362 | } |
4363 | break; |
4364 | |
4365 | case OP_HSPACE: |
4366 | for (i = 1; i <= min; i++) |
4367 | { |
4368 | if (eptr >= md->end_subject) |
4369 | { |
4370 | SCHECK_PARTIAL(); |
4371 | RRETURN(MATCH_NOMATCH); |
4372 | } |
4373 | GETCHARINC(c, eptr); |
4374 | switch(c) |
4375 | { |
4376 | HSPACE_CASES: break; /* Byte and multibyte cases */ |
4377 | default: RRETURN(MATCH_NOMATCH); |
4378 | } |
4379 | } |
4380 | break; |
4381 | |
4382 | case OP_NOT_VSPACE: |
4383 | for (i = 1; i <= min; i++) |
4384 | { |
4385 | if (eptr >= md->end_subject) |
4386 | { |
4387 | SCHECK_PARTIAL(); |
4388 | RRETURN(MATCH_NOMATCH); |
4389 | } |
4390 | GETCHARINC(c, eptr); |
4391 | switch(c) |
4392 | { |
4393 | VSPACE_CASES: RRETURN(MATCH_NOMATCH); |
4394 | default: break; |
4395 | } |
4396 | } |
4397 | break; |
4398 | |
4399 | case OP_VSPACE: |
4400 | for (i = 1; i <= min; i++) |
4401 | { |
4402 | if (eptr >= md->end_subject) |
4403 | { |
4404 | SCHECK_PARTIAL(); |
4405 | RRETURN(MATCH_NOMATCH); |
4406 | } |
4407 | GETCHARINC(c, eptr); |
4408 | switch(c) |
4409 | { |
4410 | VSPACE_CASES: break; |
4411 | default: RRETURN(MATCH_NOMATCH); |
4412 | } |
4413 | } |
4414 | break; |
4415 | |
4416 | case OP_NOT_DIGIT: |
4417 | for (i = 1; i <= min; i++) |
4418 | { |
4419 | if (eptr >= md->end_subject) |
4420 | { |
4421 | SCHECK_PARTIAL(); |
4422 | RRETURN(MATCH_NOMATCH); |
4423 | } |
4424 | GETCHARINC(c, eptr); |
4425 | if (c < 128 && (md->ctypes[c] & ctype_digit) != 0) |
4426 | RRETURN(MATCH_NOMATCH); |
4427 | } |
4428 | break; |
4429 | |
4430 | case OP_DIGIT: |
4431 | for (i = 1; i <= min; i++) |
4432 | { |
4433 | pcre_uchar cc; |
4434 | |
4435 | if (eptr >= md->end_subject) |
4436 | { |
4437 | SCHECK_PARTIAL(); |
4438 | RRETURN(MATCH_NOMATCH); |
4439 | } |
4440 | cc = RAWUCHAR(eptr); |
4441 | if (cc >= 128 || (md->ctypes[cc] & ctype_digit) == 0) |
4442 | RRETURN(MATCH_NOMATCH); |
4443 | eptr++; |
4444 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4445 | } |
4446 | break; |
4447 | |
4448 | case OP_NOT_WHITESPACE: |
4449 | for (i = 1; i <= min; i++) |
4450 | { |
4451 | pcre_uchar cc; |
4452 | |
4453 | if (eptr >= md->end_subject) |
4454 | { |
4455 | SCHECK_PARTIAL(); |
4456 | RRETURN(MATCH_NOMATCH); |
4457 | } |
4458 | cc = RAWUCHAR(eptr); |
4459 | if (cc < 128 && (md->ctypes[cc] & ctype_space) != 0) |
4460 | RRETURN(MATCH_NOMATCH); |
4461 | eptr++; |
4462 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4463 | } |
4464 | break; |
4465 | |
4466 | case OP_WHITESPACE: |
4467 | for (i = 1; i <= min; i++) |
4468 | { |
4469 | pcre_uchar cc; |
4470 | |
4471 | if (eptr >= md->end_subject) |
4472 | { |
4473 | SCHECK_PARTIAL(); |
4474 | RRETURN(MATCH_NOMATCH); |
4475 | } |
4476 | cc = RAWUCHAR(eptr); |
4477 | if (cc >= 128 || (md->ctypes[cc] & ctype_space) == 0) |
4478 | RRETURN(MATCH_NOMATCH); |
4479 | eptr++; |
4480 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4481 | } |
4482 | break; |
4483 | |
4484 | case OP_NOT_WORDCHAR: |
4485 | for (i = 1; i <= min; i++) |
4486 | { |
4487 | pcre_uchar cc; |
4488 | |
4489 | if (eptr >= md->end_subject) |
4490 | { |
4491 | SCHECK_PARTIAL(); |
4492 | RRETURN(MATCH_NOMATCH); |
4493 | } |
4494 | cc = RAWUCHAR(eptr); |
4495 | if (cc < 128 && (md->ctypes[cc] & ctype_word) != 0) |
4496 | RRETURN(MATCH_NOMATCH); |
4497 | eptr++; |
4498 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4499 | } |
4500 | break; |
4501 | |
4502 | case OP_WORDCHAR: |
4503 | for (i = 1; i <= min; i++) |
4504 | { |
4505 | pcre_uchar cc; |
4506 | |
4507 | if (eptr >= md->end_subject) |
4508 | { |
4509 | SCHECK_PARTIAL(); |
4510 | RRETURN(MATCH_NOMATCH); |
4511 | } |
4512 | cc = RAWUCHAR(eptr); |
4513 | if (cc >= 128 || (md->ctypes[cc] & ctype_word) == 0) |
4514 | RRETURN(MATCH_NOMATCH); |
4515 | eptr++; |
4516 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4517 | } |
4518 | break; |
4519 | |
4520 | default: |
4521 | RRETURN(PCRE_ERROR_INTERNAL); |
4522 | } /* End switch(ctype) */ |
4523 | |
4524 | else |
4525 | #endif /* SUPPORT_UTF */ |
4526 | |
4527 | /* Code for the non-UTF-8 case for minimum matching of operators other |
4528 | than OP_PROP and OP_NOTPROP. */ |
4529 | |
4530 | switch(ctype) |
4531 | { |
4532 | case OP_ANY: |
4533 | for (i = 1; i <= min; i++) |
4534 | { |
4535 | if (eptr >= md->end_subject) |
4536 | { |
4537 | SCHECK_PARTIAL(); |
4538 | RRETURN(MATCH_NOMATCH); |
4539 | } |
4540 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
4541 | if (md->partial != 0 && |
4542 | eptr + 1 >= md->end_subject && |
4543 | NLBLOCK->nltype == NLTYPE_FIXED && |
4544 | NLBLOCK->nllen == 2 && |
4545 | *eptr == NLBLOCK->nl[0]) |
4546 | { |
4547 | md->hitend = TRUE; |
4548 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
4549 | } |
4550 | eptr++; |
4551 | } |
4552 | break; |
4553 | |
4554 | case OP_ALLANY: |
4555 | if (eptr > md->end_subject - min) |
4556 | { |
4557 | SCHECK_PARTIAL(); |
4558 | RRETURN(MATCH_NOMATCH); |
4559 | } |
4560 | eptr += min; |
4561 | break; |
4562 | |
4563 | case OP_ANYBYTE: |
4564 | if (eptr > md->end_subject - min) |
4565 | { |
4566 | SCHECK_PARTIAL(); |
4567 | RRETURN(MATCH_NOMATCH); |
4568 | } |
4569 | eptr += min; |
4570 | break; |
4571 | |
4572 | case OP_ANYNL: |
4573 | for (i = 1; i <= min; i++) |
4574 | { |
4575 | if (eptr >= md->end_subject) |
4576 | { |
4577 | SCHECK_PARTIAL(); |
4578 | RRETURN(MATCH_NOMATCH); |
4579 | } |
4580 | switch(*eptr++) |
4581 | { |
4582 | default: RRETURN(MATCH_NOMATCH); |
4583 | |
4584 | case CHAR_CR: |
4585 | if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++; |
4586 | break; |
4587 | |
4588 | case CHAR_LF: |
4589 | break; |
4590 | |
4591 | case CHAR_VT: |
4592 | case CHAR_FF: |
4593 | case CHAR_NEL: |
4594 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4595 | case 0x2028: |
4596 | case 0x2029: |
4597 | #endif |
4598 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
4599 | break; |
4600 | } |
4601 | } |
4602 | break; |
4603 | |
4604 | case OP_NOT_HSPACE: |
4605 | for (i = 1; i <= min; i++) |
4606 | { |
4607 | if (eptr >= md->end_subject) |
4608 | { |
4609 | SCHECK_PARTIAL(); |
4610 | RRETURN(MATCH_NOMATCH); |
4611 | } |
4612 | switch(*eptr++) |
4613 | { |
4614 | default: break; |
4615 | HSPACE_BYTE_CASES: |
4616 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4617 | HSPACE_MULTIBYTE_CASES: |
4618 | #endif |
4619 | RRETURN(MATCH_NOMATCH); |
4620 | } |
4621 | } |
4622 | break; |
4623 | |
4624 | case OP_HSPACE: |
4625 | for (i = 1; i <= min; i++) |
4626 | { |
4627 | if (eptr >= md->end_subject) |
4628 | { |
4629 | SCHECK_PARTIAL(); |
4630 | RRETURN(MATCH_NOMATCH); |
4631 | } |
4632 | switch(*eptr++) |
4633 | { |
4634 | default: RRETURN(MATCH_NOMATCH); |
4635 | HSPACE_BYTE_CASES: |
4636 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4637 | HSPACE_MULTIBYTE_CASES: |
4638 | #endif |
4639 | break; |
4640 | } |
4641 | } |
4642 | break; |
4643 | |
4644 | case OP_NOT_VSPACE: |
4645 | for (i = 1; i <= min; i++) |
4646 | { |
4647 | if (eptr >= md->end_subject) |
4648 | { |
4649 | SCHECK_PARTIAL(); |
4650 | RRETURN(MATCH_NOMATCH); |
4651 | } |
4652 | switch(*eptr++) |
4653 | { |
4654 | VSPACE_BYTE_CASES: |
4655 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4656 | VSPACE_MULTIBYTE_CASES: |
4657 | #endif |
4658 | RRETURN(MATCH_NOMATCH); |
4659 | default: break; |
4660 | } |
4661 | } |
4662 | break; |
4663 | |
4664 | case OP_VSPACE: |
4665 | for (i = 1; i <= min; i++) |
4666 | { |
4667 | if (eptr >= md->end_subject) |
4668 | { |
4669 | SCHECK_PARTIAL(); |
4670 | RRETURN(MATCH_NOMATCH); |
4671 | } |
4672 | switch(*eptr++) |
4673 | { |
4674 | default: RRETURN(MATCH_NOMATCH); |
4675 | VSPACE_BYTE_CASES: |
4676 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4677 | VSPACE_MULTIBYTE_CASES: |
4678 | #endif |
4679 | break; |
4680 | } |
4681 | } |
4682 | break; |
4683 | |
4684 | case OP_NOT_DIGIT: |
4685 | for (i = 1; i <= min; i++) |
4686 | { |
4687 | if (eptr >= md->end_subject) |
4688 | { |
4689 | SCHECK_PARTIAL(); |
4690 | RRETURN(MATCH_NOMATCH); |
4691 | } |
4692 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) |
4693 | RRETURN(MATCH_NOMATCH); |
4694 | eptr++; |
4695 | } |
4696 | break; |
4697 | |
4698 | case OP_DIGIT: |
4699 | for (i = 1; i <= min; i++) |
4700 | { |
4701 | if (eptr >= md->end_subject) |
4702 | { |
4703 | SCHECK_PARTIAL(); |
4704 | RRETURN(MATCH_NOMATCH); |
4705 | } |
4706 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) |
4707 | RRETURN(MATCH_NOMATCH); |
4708 | eptr++; |
4709 | } |
4710 | break; |
4711 | |
4712 | case OP_NOT_WHITESPACE: |
4713 | for (i = 1; i <= min; i++) |
4714 | { |
4715 | if (eptr >= md->end_subject) |
4716 | { |
4717 | SCHECK_PARTIAL(); |
4718 | RRETURN(MATCH_NOMATCH); |
4719 | } |
4720 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) |
4721 | RRETURN(MATCH_NOMATCH); |
4722 | eptr++; |
4723 | } |
4724 | break; |
4725 | |
4726 | case OP_WHITESPACE: |
4727 | for (i = 1; i <= min; i++) |
4728 | { |
4729 | if (eptr >= md->end_subject) |
4730 | { |
4731 | SCHECK_PARTIAL(); |
4732 | RRETURN(MATCH_NOMATCH); |
4733 | } |
4734 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) |
4735 | RRETURN(MATCH_NOMATCH); |
4736 | eptr++; |
4737 | } |
4738 | break; |
4739 | |
4740 | case OP_NOT_WORDCHAR: |
4741 | for (i = 1; i <= min; i++) |
4742 | { |
4743 | if (eptr >= md->end_subject) |
4744 | { |
4745 | SCHECK_PARTIAL(); |
4746 | RRETURN(MATCH_NOMATCH); |
4747 | } |
4748 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) |
4749 | RRETURN(MATCH_NOMATCH); |
4750 | eptr++; |
4751 | } |
4752 | break; |
4753 | |
4754 | case OP_WORDCHAR: |
4755 | for (i = 1; i <= min; i++) |
4756 | { |
4757 | if (eptr >= md->end_subject) |
4758 | { |
4759 | SCHECK_PARTIAL(); |
4760 | RRETURN(MATCH_NOMATCH); |
4761 | } |
4762 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) |
4763 | RRETURN(MATCH_NOMATCH); |
4764 | eptr++; |
4765 | } |
4766 | break; |
4767 | |
4768 | default: |
4769 | RRETURN(PCRE_ERROR_INTERNAL); |
4770 | } |
4771 | } |
4772 | |
4773 | /* If min = max, continue at the same level without recursing */ |
4774 | |
4775 | if (min == max) continue; |
4776 | |
4777 | /* If minimizing, we have to test the rest of the pattern before each |
4778 | subsequent match. Again, separate the UTF-8 case for speed, and also |
4779 | separate the UCP cases. */ |
4780 | |
4781 | if (minimize) |
4782 | { |
4783 | #ifdef SUPPORT_UCP |
4784 | if (prop_type >= 0) |
4785 | { |
4786 | switch(prop_type) |
4787 | { |
4788 | case PT_ANY: |
4789 | for (fi = min;; fi++) |
4790 | { |
4791 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM36); |
4792 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4793 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4794 | if (eptr >= md->end_subject) |
4795 | { |
4796 | SCHECK_PARTIAL(); |
4797 | RRETURN(MATCH_NOMATCH); |
4798 | } |
4799 | GETCHARINCTEST(c, eptr); |
4800 | if (prop_fail_result) RRETURN(MATCH_NOMATCH); |
4801 | } |
4802 | /* Control never gets here */ |
4803 | |
4804 | case PT_LAMP: |
4805 | for (fi = min;; fi++) |
4806 | { |
4807 | int chartype; |
4808 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM37); |
4809 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4810 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4811 | if (eptr >= md->end_subject) |
4812 | { |
4813 | SCHECK_PARTIAL(); |
4814 | RRETURN(MATCH_NOMATCH); |
4815 | } |
4816 | GETCHARINCTEST(c, eptr); |
4817 | chartype = UCD_CHARTYPE(c); |
4818 | if ((chartype == ucp_Lu || |
4819 | chartype == ucp_Ll || |
4820 | chartype == ucp_Lt) == prop_fail_result) |
4821 | RRETURN(MATCH_NOMATCH); |
4822 | } |
4823 | /* Control never gets here */ |
4824 | |
4825 | case PT_GC: |
4826 | for (fi = min;; fi++) |
4827 | { |
4828 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM38); |
4829 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4830 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4831 | if (eptr >= md->end_subject) |
4832 | { |
4833 | SCHECK_PARTIAL(); |
4834 | RRETURN(MATCH_NOMATCH); |
4835 | } |
4836 | GETCHARINCTEST(c, eptr); |
4837 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) |
4838 | RRETURN(MATCH_NOMATCH); |
4839 | } |
4840 | /* Control never gets here */ |
4841 | |
4842 | case PT_PC: |
4843 | for (fi = min;; fi++) |
4844 | { |
4845 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM39); |
4846 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4847 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4848 | if (eptr >= md->end_subject) |
4849 | { |
4850 | SCHECK_PARTIAL(); |
4851 | RRETURN(MATCH_NOMATCH); |
4852 | } |
4853 | GETCHARINCTEST(c, eptr); |
4854 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) |
4855 | RRETURN(MATCH_NOMATCH); |
4856 | } |
4857 | /* Control never gets here */ |
4858 | |
4859 | case PT_SC: |
4860 | for (fi = min;; fi++) |
4861 | { |
4862 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM40); |
4863 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4864 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4865 | if (eptr >= md->end_subject) |
4866 | { |
4867 | SCHECK_PARTIAL(); |
4868 | RRETURN(MATCH_NOMATCH); |
4869 | } |
4870 | GETCHARINCTEST(c, eptr); |
4871 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) |
4872 | RRETURN(MATCH_NOMATCH); |
4873 | } |
4874 | /* Control never gets here */ |
4875 | |
4876 | case PT_ALNUM: |
4877 | for (fi = min;; fi++) |
4878 | { |
4879 | int category; |
4880 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM59); |
4881 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4882 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4883 | if (eptr >= md->end_subject) |
4884 | { |
4885 | SCHECK_PARTIAL(); |
4886 | RRETURN(MATCH_NOMATCH); |
4887 | } |
4888 | GETCHARINCTEST(c, eptr); |
4889 | category = UCD_CATEGORY(c); |
4890 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
4891 | RRETURN(MATCH_NOMATCH); |
4892 | } |
4893 | /* Control never gets here */ |
4894 | |
4895 | case PT_SPACE: /* Perl space */ |
4896 | for (fi = min;; fi++) |
4897 | { |
4898 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM60); |
4899 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4900 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4901 | if (eptr >= md->end_subject) |
4902 | { |
4903 | SCHECK_PARTIAL(); |
4904 | RRETURN(MATCH_NOMATCH); |
4905 | } |
4906 | GETCHARINCTEST(c, eptr); |
4907 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4908 | c == CHAR_FF || c == CHAR_CR) |
4909 | == prop_fail_result) |
4910 | RRETURN(MATCH_NOMATCH); |
4911 | } |
4912 | /* Control never gets here */ |
4913 | |
4914 | case PT_PXSPACE: /* POSIX space */ |
4915 | for (fi = min;; fi++) |
4916 | { |
4917 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM61); |
4918 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4919 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4920 | if (eptr >= md->end_subject) |
4921 | { |
4922 | SCHECK_PARTIAL(); |
4923 | RRETURN(MATCH_NOMATCH); |
4924 | } |
4925 | GETCHARINCTEST(c, eptr); |
4926 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4927 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
4928 | == prop_fail_result) |
4929 | RRETURN(MATCH_NOMATCH); |
4930 | } |
4931 | /* Control never gets here */ |
4932 | |
4933 | case PT_WORD: |
4934 | for (fi = min;; fi++) |
4935 | { |
4936 | int category; |
4937 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM62); |
4938 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4939 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4940 | if (eptr >= md->end_subject) |
4941 | { |
4942 | SCHECK_PARTIAL(); |
4943 | RRETURN(MATCH_NOMATCH); |
4944 | } |
4945 | GETCHARINCTEST(c, eptr); |
4946 | category = UCD_CATEGORY(c); |
4947 | if ((category == ucp_L || |
4948 | category == ucp_N || |
4949 | c == CHAR_UNDERSCORE) |
4950 | == prop_fail_result) |
4951 | RRETURN(MATCH_NOMATCH); |
4952 | } |
4953 | /* Control never gets here */ |
4954 | |
4955 | case PT_CLIST: |
4956 | for (fi = min;; fi++) |
4957 | { |
4958 | const pcre_uint32 *cp; |
4959 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM67); |
4960 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4961 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4962 | if (eptr >= md->end_subject) |
4963 | { |
4964 | SCHECK_PARTIAL(); |
4965 | RRETURN(MATCH_NOMATCH); |
4966 | } |
4967 | GETCHARINCTEST(c, eptr); |
4968 | cp = PRIV(ucd_caseless_sets) + prop_value; |
4969 | for (;;) |
4970 | { |
4971 | if (c < *cp) |
4972 | { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } } |
4973 | if (c == *cp++) |
4974 | { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; } |
4975 | } |
4976 | } |
4977 | /* Control never gets here */ |
4978 | |
4979 | /* This should never occur */ |
4980 | default: |
4981 | RRETURN(PCRE_ERROR_INTERNAL); |
4982 | } |
4983 | } |
4984 | |
4985 | /* Match extended Unicode sequences. We will get here only if the |
4986 | support is in the binary; otherwise a compile-time error occurs. */ |
4987 | |
4988 | else if (ctype == OP_EXTUNI) |
4989 | { |
4990 | for (fi = min;; fi++) |
4991 | { |
4992 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM41); |
4993 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4994 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4995 | if (eptr >= md->end_subject) |
4996 | { |
4997 | SCHECK_PARTIAL(); |
4998 | RRETURN(MATCH_NOMATCH); |
4999 | } |
5000 | else |
5001 | { |
5002 | int lgb, rgb; |
5003 | GETCHARINCTEST(c, eptr); |
5004 | lgb = UCD_GRAPHBREAK(c); |
5005 | while (eptr < md->end_subject) |
5006 | { |
5007 | int len = 1; |
5008 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
5009 | rgb = UCD_GRAPHBREAK(c); |
5010 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
5011 | lgb = rgb; |
5012 | eptr += len; |
5013 | } |
5014 | } |
5015 | CHECK_PARTIAL(); |
5016 | } |
5017 | } |
5018 | else |
5019 | #endif /* SUPPORT_UCP */ |
5020 | |
5021 | #ifdef SUPPORT_UTF |
5022 | if (utf) |
5023 | { |
5024 | for (fi = min;; fi++) |
5025 | { |
5026 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM42); |
5027 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5028 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5029 | if (eptr >= md->end_subject) |
5030 | { |
5031 | SCHECK_PARTIAL(); |
5032 | RRETURN(MATCH_NOMATCH); |
5033 | } |
5034 | if (ctype == OP_ANY && IS_NEWLINE(eptr)) |
5035 | RRETURN(MATCH_NOMATCH); |
5036 | GETCHARINC(c, eptr); |
5037 | switch(ctype) |
5038 | { |
5039 | case OP_ANY: /* This is the non-NL case */ |
5040 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5041 | eptr >= md->end_subject && |
5042 | NLBLOCK->nltype == NLTYPE_FIXED && |
5043 | NLBLOCK->nllen == 2 && |
5044 | c == NLBLOCK->nl[0]) |
5045 | { |
5046 | md->hitend = TRUE; |
5047 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5048 | } |
5049 | break; |
5050 | |
5051 | case OP_ALLANY: |
5052 | case OP_ANYBYTE: |
5053 | break; |
5054 | |
5055 | case OP_ANYNL: |
5056 | switch(c) |
5057 | { |
5058 | default: RRETURN(MATCH_NOMATCH); |
5059 | case CHAR_CR: |
5060 | if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++; |
5061 | break; |
5062 | |
5063 | case CHAR_LF: |
5064 | break; |
5065 | |
5066 | case CHAR_VT: |
5067 | case CHAR_FF: |
5068 | case CHAR_NEL: |
5069 | #ifndef EBCDIC |
5070 | case 0x2028: |
5071 | case 0x2029: |
5072 | #endif /* Not EBCDIC */ |
5073 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
5074 | break; |
5075 | } |
5076 | break; |
5077 | |
5078 | case OP_NOT_HSPACE: |
5079 | switch(c) |
5080 | { |
5081 | HSPACE_CASES: RRETURN(MATCH_NOMATCH); |
5082 | default: break; |
5083 | } |
5084 | break; |
5085 | |
5086 | case OP_HSPACE: |
5087 | switch(c) |
5088 | { |
5089 | HSPACE_CASES: break; |
5090 | default: RRETURN(MATCH_NOMATCH); |
5091 | } |
5092 | break; |
5093 | |
5094 | case OP_NOT_VSPACE: |
5095 | switch(c) |
5096 | { |
5097 | VSPACE_CASES: RRETURN(MATCH_NOMATCH); |
5098 | default: break; |
5099 | } |
5100 | break; |
5101 | |
5102 | case OP_VSPACE: |
5103 | switch(c) |
5104 | { |
5105 | VSPACE_CASES: break; |
5106 | default: RRETURN(MATCH_NOMATCH); |
5107 | } |
5108 | break; |
5109 | |
5110 | case OP_NOT_DIGIT: |
5111 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) |
5112 | RRETURN(MATCH_NOMATCH); |
5113 | break; |
5114 | |
5115 | case OP_DIGIT: |
5116 | if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0) |
5117 | RRETURN(MATCH_NOMATCH); |
5118 | break; |
5119 | |
5120 | case OP_NOT_WHITESPACE: |
5121 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) |
5122 | RRETURN(MATCH_NOMATCH); |
5123 | break; |
5124 | |
5125 | case OP_WHITESPACE: |
5126 | if (c >= 256 || (md->ctypes[c] & ctype_space) == 0) |
5127 | RRETURN(MATCH_NOMATCH); |
5128 | break; |
5129 | |
5130 | case OP_NOT_WORDCHAR: |
5131 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) |
5132 | RRETURN(MATCH_NOMATCH); |
5133 | break; |
5134 | |
5135 | case OP_WORDCHAR: |
5136 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) |
5137 | RRETURN(MATCH_NOMATCH); |
5138 | break; |
5139 | |
5140 | default: |
5141 | RRETURN(PCRE_ERROR_INTERNAL); |
5142 | } |
5143 | } |
5144 | } |
5145 | else |
5146 | #endif |
5147 | /* Not UTF mode */ |
5148 | { |
5149 | for (fi = min;; fi++) |
5150 | { |
5151 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM43); |
5152 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5153 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5154 | if (eptr >= md->end_subject) |
5155 | { |
5156 | SCHECK_PARTIAL(); |
5157 | RRETURN(MATCH_NOMATCH); |
5158 | } |
5159 | if (ctype == OP_ANY && IS_NEWLINE(eptr)) |
5160 | RRETURN(MATCH_NOMATCH); |
5161 | c = *eptr++; |
5162 | switch(ctype) |
5163 | { |
5164 | case OP_ANY: /* This is the non-NL case */ |
5165 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5166 | eptr >= md->end_subject && |
5167 | NLBLOCK->nltype == NLTYPE_FIXED && |
5168 | NLBLOCK->nllen == 2 && |
5169 | c == NLBLOCK->nl[0]) |
5170 | { |
5171 | md->hitend = TRUE; |
5172 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5173 | } |
5174 | break; |
5175 | |
5176 | case OP_ALLANY: |
5177 | case OP_ANYBYTE: |
5178 | break; |
5179 | |
5180 | case OP_ANYNL: |
5181 | switch(c) |
5182 | { |
5183 | default: RRETURN(MATCH_NOMATCH); |
5184 | case CHAR_CR: |
5185 | if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++; |
5186 | break; |
5187 | |
5188 | case CHAR_LF: |
5189 | break; |
5190 | |
5191 | case CHAR_VT: |
5192 | case CHAR_FF: |
5193 | case CHAR_NEL: |
5194 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5195 | case 0x2028: |
5196 | case 0x2029: |
5197 | #endif |
5198 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
5199 | break; |
5200 | } |
5201 | break; |
5202 | |
5203 | case OP_NOT_HSPACE: |
5204 | switch(c) |
5205 | { |
5206 | default: break; |
5207 | HSPACE_BYTE_CASES: |
5208 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5209 | HSPACE_MULTIBYTE_CASES: |
5210 | #endif |
5211 | RRETURN(MATCH_NOMATCH); |
5212 | } |
5213 | break; |
5214 | |
5215 | case OP_HSPACE: |
5216 | switch(c) |
5217 | { |
5218 | default: RRETURN(MATCH_NOMATCH); |
5219 | HSPACE_BYTE_CASES: |
5220 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5221 | HSPACE_MULTIBYTE_CASES: |
5222 | #endif |
5223 | break; |
5224 | } |
5225 | break; |
5226 | |
5227 | case OP_NOT_VSPACE: |
5228 | switch(c) |
5229 | { |
5230 | default: break; |
5231 | VSPACE_BYTE_CASES: |
5232 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5233 | VSPACE_MULTIBYTE_CASES: |
5234 | #endif |
5235 | RRETURN(MATCH_NOMATCH); |
5236 | } |
5237 | break; |
5238 | |
5239 | case OP_VSPACE: |
5240 | switch(c) |
5241 | { |
5242 | default: RRETURN(MATCH_NOMATCH); |
5243 | VSPACE_BYTE_CASES: |
5244 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5245 | VSPACE_MULTIBYTE_CASES: |
5246 | #endif |
5247 | break; |
5248 | } |
5249 | break; |
5250 | |
5251 | case OP_NOT_DIGIT: |
5252 | if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH); |
5253 | break; |
5254 | |
5255 | case OP_DIGIT: |
5256 | if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH); |
5257 | break; |
5258 | |
5259 | case OP_NOT_WHITESPACE: |
5260 | if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH); |
5261 | break; |
5262 | |
5263 | case OP_WHITESPACE: |
5264 | if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH); |
5265 | break; |
5266 | |
5267 | case OP_NOT_WORDCHAR: |
5268 | if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH); |
5269 | break; |
5270 | |
5271 | case OP_WORDCHAR: |
5272 | if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH); |
5273 | break; |
5274 | |
5275 | default: |
5276 | RRETURN(PCRE_ERROR_INTERNAL); |
5277 | } |
5278 | } |
5279 | } |
5280 | /* Control never gets here */ |
5281 | } |
5282 | |
5283 | /* If maximizing, it is worth using inline code for speed, doing the type |
5284 | test once at the start (i.e. keep it out of the loop). Again, keep the |
5285 | UTF-8 and UCP stuff separate. */ |
5286 | |
5287 | else |
5288 | { |
5289 | pp = eptr; /* Remember where we started */ |
5290 | |
5291 | #ifdef SUPPORT_UCP |
5292 | if (prop_type >= 0) |
5293 | { |
5294 | switch(prop_type) |
5295 | { |
5296 | case PT_ANY: |
5297 | for (i = min; i < max; i++) |
5298 | { |
5299 | int len = 1; |
5300 | if (eptr >= md->end_subject) |
5301 | { |
5302 | SCHECK_PARTIAL(); |
5303 | break; |
5304 | } |
5305 | GETCHARLENTEST(c, eptr, len); |
5306 | if (prop_fail_result) break; |
5307 | eptr+= len; |
5308 | } |
5309 | break; |
5310 | |
5311 | case PT_LAMP: |
5312 | for (i = min; i < max; i++) |
5313 | { |
5314 | int chartype; |
5315 | int len = 1; |
5316 | if (eptr >= md->end_subject) |
5317 | { |
5318 | SCHECK_PARTIAL(); |
5319 | break; |
5320 | } |
5321 | GETCHARLENTEST(c, eptr, len); |
5322 | chartype = UCD_CHARTYPE(c); |
5323 | if ((chartype == ucp_Lu || |
5324 | chartype == ucp_Ll || |
5325 | chartype == ucp_Lt) == prop_fail_result) |
5326 | break; |
5327 | eptr+= len; |
5328 | } |
5329 | break; |
5330 | |
5331 | case PT_GC: |
5332 | for (i = min; i < max; i++) |
5333 | { |
5334 | int len = 1; |
5335 | if (eptr >= md->end_subject) |
5336 | { |
5337 | SCHECK_PARTIAL(); |
5338 | break; |
5339 | } |
5340 | GETCHARLENTEST(c, eptr, len); |
5341 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break; |
5342 | eptr+= len; |
5343 | } |
5344 | break; |
5345 | |
5346 | case PT_PC: |
5347 | for (i = min; i < max; i++) |
5348 | { |
5349 | int len = 1; |
5350 | if (eptr >= md->end_subject) |
5351 | { |
5352 | SCHECK_PARTIAL(); |
5353 | break; |
5354 | } |
5355 | GETCHARLENTEST(c, eptr, len); |
5356 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break; |
5357 | eptr+= len; |
5358 | } |
5359 | break; |
5360 | |
5361 | case PT_SC: |
5362 | for (i = min; i < max; i++) |
5363 | { |
5364 | int len = 1; |
5365 | if (eptr >= md->end_subject) |
5366 | { |
5367 | SCHECK_PARTIAL(); |
5368 | break; |
5369 | } |
5370 | GETCHARLENTEST(c, eptr, len); |
5371 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break; |
5372 | eptr+= len; |
5373 | } |
5374 | break; |
5375 | |
5376 | case PT_ALNUM: |
5377 | for (i = min; i < max; i++) |
5378 | { |
5379 | int category; |
5380 | int len = 1; |
5381 | if (eptr >= md->end_subject) |
5382 | { |
5383 | SCHECK_PARTIAL(); |
5384 | break; |
5385 | } |
5386 | GETCHARLENTEST(c, eptr, len); |
5387 | category = UCD_CATEGORY(c); |
5388 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
5389 | break; |
5390 | eptr+= len; |
5391 | } |
5392 | break; |
5393 | |
5394 | case PT_SPACE: /* Perl space */ |
5395 | for (i = min; i < max; i++) |
5396 | { |
5397 | int len = 1; |
5398 | if (eptr >= md->end_subject) |
5399 | { |
5400 | SCHECK_PARTIAL(); |
5401 | break; |
5402 | } |
5403 | GETCHARLENTEST(c, eptr, len); |
5404 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5405 | c == CHAR_FF || c == CHAR_CR) |
5406 | == prop_fail_result) |
5407 | break; |
5408 | eptr+= len; |
5409 | } |
5410 | break; |
5411 | |
5412 | case PT_PXSPACE: /* POSIX space */ |
5413 | for (i = min; i < max; i++) |
5414 | { |
5415 | int len = 1; |
5416 | if (eptr >= md->end_subject) |
5417 | { |
5418 | SCHECK_PARTIAL(); |
5419 | break; |
5420 | } |
5421 | GETCHARLENTEST(c, eptr, len); |
5422 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5423 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
5424 | == prop_fail_result) |
5425 | break; |
5426 | eptr+= len; |
5427 | } |
5428 | break; |
5429 | |
5430 | case PT_WORD: |
5431 | for (i = min; i < max; i++) |
5432 | { |
5433 | int category; |
5434 | int len = 1; |
5435 | if (eptr >= md->end_subject) |
5436 | { |
5437 | SCHECK_PARTIAL(); |
5438 | break; |
5439 | } |
5440 | GETCHARLENTEST(c, eptr, len); |
5441 | category = UCD_CATEGORY(c); |
5442 | if ((category == ucp_L || category == ucp_N || |
5443 | c == CHAR_UNDERSCORE) == prop_fail_result) |
5444 | break; |
5445 | eptr+= len; |
5446 | } |
5447 | break; |
5448 | |
5449 | case PT_CLIST: |
5450 | for (i = min; i < max; i++) |
5451 | { |
5452 | const pcre_uint32 *cp; |
5453 | int len = 1; |
5454 | if (eptr >= md->end_subject) |
5455 | { |
5456 | SCHECK_PARTIAL(); |
5457 | break; |
5458 | } |
5459 | GETCHARLENTEST(c, eptr, len); |
5460 | cp = PRIV(ucd_caseless_sets) + prop_value; |
5461 | for (;;) |
5462 | { |
5463 | if (c < *cp) |
5464 | { if (prop_fail_result) break; else goto GOT_MAX; } |
5465 | if (c == *cp++) |
5466 | { if (prop_fail_result) goto GOT_MAX; else break; } |
5467 | } |
5468 | eptr += len; |
5469 | } |
5470 | GOT_MAX: |
5471 | break; |
5472 | |
5473 | default: |
5474 | RRETURN(PCRE_ERROR_INTERNAL); |
5475 | } |
5476 | |
5477 | /* eptr is now past the end of the maximum run */ |
5478 | |
5479 | if (possessive) continue; |
5480 | for(;;) |
5481 | { |
5482 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM44); |
5483 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5484 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
5485 | if (utf) BACKCHAR(eptr); |
5486 | } |
5487 | } |
5488 | |
5489 | /* Match extended Unicode sequences. We will get here only if the |
5490 | support is in the binary; otherwise a compile-time error occurs. */ |
5491 | |
5492 | else if (ctype == OP_EXTUNI) |
5493 | { |
5494 | for (i = min; i < max; i++) |
5495 | { |
5496 | if (eptr >= md->end_subject) |
5497 | { |
5498 | SCHECK_PARTIAL(); |
5499 | break; |
5500 | } |
5501 | else |
5502 | { |
5503 | int lgb, rgb; |
5504 | GETCHARINCTEST(c, eptr); |
5505 | lgb = UCD_GRAPHBREAK(c); |
5506 | while (eptr < md->end_subject) |
5507 | { |
5508 | int len = 1; |
5509 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
5510 | rgb = UCD_GRAPHBREAK(c); |
5511 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
5512 | lgb = rgb; |
5513 | eptr += len; |
5514 | } |
5515 | } |
5516 | CHECK_PARTIAL(); |
5517 | } |
5518 | |
5519 | /* eptr is now past the end of the maximum run */ |
5520 | |
5521 | if (possessive) continue; |
5522 | |
5523 | for(;;) |
5524 | { |
5525 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM45); |
5526 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5527 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
5528 | for (;;) /* Move back over one extended */ |
5529 | { |
5530 | if (!utf) c = *eptr; else |
5531 | { |
5532 | BACKCHAR(eptr); |
5533 | GETCHAR(c, eptr); |
5534 | } |
5535 | if (UCD_CATEGORY(c) != ucp_M) break; |
5536 | eptr--; |
5537 | } |
5538 | } |
5539 | } |
5540 | |
5541 | else |
5542 | #endif /* SUPPORT_UCP */ |
5543 | |
5544 | #ifdef SUPPORT_UTF |
5545 | if (utf) |
5546 | { |
5547 | switch(ctype) |
5548 | { |
5549 | case OP_ANY: |
5550 | if (max < INT_MAX) |
5551 | { |
5552 | for (i = min; i < max; i++) |
5553 | { |
5554 | if (eptr >= md->end_subject) |
5555 | { |
5556 | SCHECK_PARTIAL(); |
5557 | break; |
5558 | } |
5559 | if (IS_NEWLINE(eptr)) break; |
5560 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5561 | eptr + 1 >= md->end_subject && |
5562 | NLBLOCK->nltype == NLTYPE_FIXED && |
5563 | NLBLOCK->nllen == 2 && |
5564 | RAWUCHAR(eptr) == NLBLOCK->nl[0]) |
5565 | { |
5566 | md->hitend = TRUE; |
5567 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5568 | } |
5569 | eptr++; |
5570 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5571 | } |
5572 | } |
5573 | |
5574 | /* Handle unlimited UTF-8 repeat */ |
5575 | |
5576 | else |
5577 | { |
5578 | for (i = min; i < max; i++) |
5579 | { |
5580 | if (eptr >= md->end_subject) |
5581 | { |
5582 | SCHECK_PARTIAL(); |
5583 | break; |
5584 | } |
5585 | if (IS_NEWLINE(eptr)) break; |
5586 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5587 | eptr + 1 >= md->end_subject && |
5588 | NLBLOCK->nltype == NLTYPE_FIXED && |
5589 | NLBLOCK->nllen == 2 && |
5590 | RAWUCHAR(eptr) == NLBLOCK->nl[0]) |
5591 | { |
5592 | md->hitend = TRUE; |
5593 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5594 | } |
5595 | eptr++; |
5596 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5597 | } |
5598 | } |
5599 | break; |
5600 | |
5601 | case OP_ALLANY: |
5602 | if (max < INT_MAX) |
5603 | { |
5604 | for (i = min; i < max; i++) |
5605 | { |
5606 | if (eptr >= md->end_subject) |
5607 | { |
5608 | SCHECK_PARTIAL(); |
5609 | break; |
5610 | } |
5611 | eptr++; |
5612 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5613 | } |
5614 | } |
5615 | else |
5616 | { |
5617 | eptr = md->end_subject; /* Unlimited UTF-8 repeat */ |
5618 | SCHECK_PARTIAL(); |
5619 | } |
5620 | break; |
5621 | |
5622 | /* The byte case is the same as non-UTF8 */ |
5623 | |
5624 | case OP_ANYBYTE: |
5625 | c = max - min; |
5626 | if (c > (unsigned int)(md->end_subject - eptr)) |
5627 | { |
5628 | eptr = md->end_subject; |
5629 | SCHECK_PARTIAL(); |
5630 | } |
5631 | else eptr += c; |
5632 | break; |
5633 | |
5634 | case OP_ANYNL: |
5635 | for (i = min; i < max; i++) |
5636 | { |
5637 | int len = 1; |
5638 | if (eptr >= md->end_subject) |
5639 | { |
5640 | SCHECK_PARTIAL(); |
5641 | break; |
5642 | } |
5643 | GETCHARLEN(c, eptr, len); |
5644 | if (c == CHAR_CR) |
5645 | { |
5646 | if (++eptr >= md->end_subject) break; |
5647 | if (RAWUCHAR(eptr) == CHAR_LF) eptr++; |
5648 | } |
5649 | else |
5650 | { |
5651 | if (c != CHAR_LF && |
5652 | (md->bsr_anycrlf || |
5653 | (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL |
5654 | #ifndef EBCDIC |
5655 | && c != 0x2028 && c != 0x2029 |
5656 | #endif /* Not EBCDIC */ |
5657 | ))) |
5658 | break; |
5659 | eptr += len; |
5660 | } |
5661 | } |
5662 | break; |
5663 | |
5664 | case OP_NOT_HSPACE: |
5665 | case OP_HSPACE: |
5666 | for (i = min; i < max; i++) |
5667 | { |
5668 | BOOL gotspace; |
5669 | int len = 1; |
5670 | if (eptr >= md->end_subject) |
5671 | { |
5672 | SCHECK_PARTIAL(); |
5673 | break; |
5674 | } |
5675 | GETCHARLEN(c, eptr, len); |
5676 | switch(c) |
5677 | { |
5678 | HSPACE_CASES: gotspace = TRUE; break; |
5679 | default: gotspace = FALSE; break; |
5680 | } |
5681 | if (gotspace == (ctype == OP_NOT_HSPACE)) break; |
5682 | eptr += len; |
5683 | } |
5684 | break; |
5685 | |
5686 | case OP_NOT_VSPACE: |
5687 | case OP_VSPACE: |
5688 | for (i = min; i < max; i++) |
5689 | { |
5690 | BOOL gotspace; |
5691 | int len = 1; |
5692 | if (eptr >= md->end_subject) |
5693 | { |
5694 | SCHECK_PARTIAL(); |
5695 | break; |
5696 | } |
5697 | GETCHARLEN(c, eptr, len); |
5698 | switch(c) |
5699 | { |
5700 | VSPACE_CASES: gotspace = TRUE; break; |
5701 | default: gotspace = FALSE; break; |
5702 | } |
5703 | if (gotspace == (ctype == OP_NOT_VSPACE)) break; |
5704 | eptr += len; |
5705 | } |
5706 | break; |
5707 | |
5708 | case OP_NOT_DIGIT: |
5709 | for (i = min; i < max; i++) |
5710 | { |
5711 | int len = 1; |
5712 | if (eptr >= md->end_subject) |
5713 | { |
5714 | SCHECK_PARTIAL(); |
5715 | break; |
5716 | } |
5717 | GETCHARLEN(c, eptr, len); |
5718 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break; |
5719 | eptr+= len; |
5720 | } |
5721 | break; |
5722 | |
5723 | case OP_DIGIT: |
5724 | for (i = min; i < max; i++) |
5725 | { |
5726 | int len = 1; |
5727 | if (eptr >= md->end_subject) |
5728 | { |
5729 | SCHECK_PARTIAL(); |
5730 | break; |
5731 | } |
5732 | GETCHARLEN(c, eptr, len); |
5733 | if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break; |
5734 | eptr+= len; |
5735 | } |
5736 | break; |
5737 | |
5738 | case OP_NOT_WHITESPACE: |
5739 | for (i = min; i < max; i++) |
5740 | { |
5741 | int len = 1; |
5742 | if (eptr >= md->end_subject) |
5743 | { |
5744 | SCHECK_PARTIAL(); |
5745 | break; |
5746 | } |
5747 | GETCHARLEN(c, eptr, len); |
5748 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break; |
5749 | eptr+= len; |
5750 | } |
5751 | break; |
5752 | |
5753 | case OP_WHITESPACE: |
5754 | for (i = min; i < max; i++) |
5755 | { |
5756 | int len = 1; |
5757 | if (eptr >= md->end_subject) |
5758 | { |
5759 | SCHECK_PARTIAL(); |
5760 | break; |
5761 | } |
5762 | GETCHARLEN(c, eptr, len); |
5763 | if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break; |
5764 | eptr+= len; |
5765 | } |
5766 | break; |
5767 | |
5768 | case OP_NOT_WORDCHAR: |
5769 | for (i = min; i < max; i++) |
5770 | { |
5771 | int len = 1; |
5772 | if (eptr >= md->end_subject) |
5773 | { |
5774 | SCHECK_PARTIAL(); |
5775 | break; |
5776 | } |
5777 | GETCHARLEN(c, eptr, len); |
5778 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break; |
5779 | eptr+= len; |
5780 | } |
5781 | break; |
5782 | |
5783 | case OP_WORDCHAR: |
5784 | for (i = min; i < max; i++) |
5785 | { |
5786 | int len = 1; |
5787 | if (eptr >= md->end_subject) |
5788 | { |
5789 | SCHECK_PARTIAL(); |
5790 | break; |
5791 | } |
5792 | GETCHARLEN(c, eptr, len); |
5793 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break; |
5794 | eptr+= len; |
5795 | } |
5796 | break; |
5797 | |
5798 | default: |
5799 | RRETURN(PCRE_ERROR_INTERNAL); |
5800 | } |
5801 | |
5802 | /* eptr is now past the end of the maximum run. If possessive, we are |
5803 | done (no backing up). Otherwise, match at this position; anything other |
5804 | than no match is immediately returned. For nomatch, back up one |
5805 | character, unless we are matching \R and the last thing matched was |
5806 | \r\n, in which case, back up two bytes. */ |
5807 | |
5808 | if (possessive) continue; |
5809 | for(;;) |
5810 | { |
5811 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM46); |
5812 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5813 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
5814 | BACKCHAR(eptr); |
5815 | if (ctype == OP_ANYNL && eptr > pp && RAWUCHAR(eptr) == CHAR_NL && |
5816 | RAWUCHAR(eptr - 1) == CHAR_CR) eptr--; |
5817 | } |
5818 | } |
5819 | else |
5820 | #endif /* SUPPORT_UTF */ |
5821 | /* Not UTF mode */ |
5822 | { |
5823 | switch(ctype) |
5824 | { |
5825 | case OP_ANY: |
5826 | for (i = min; i < max; i++) |
5827 | { |
5828 | if (eptr >= md->end_subject) |
5829 | { |
5830 | SCHECK_PARTIAL(); |
5831 | break; |
5832 | } |
5833 | if (IS_NEWLINE(eptr)) break; |
5834 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5835 | eptr + 1 >= md->end_subject && |
5836 | NLBLOCK->nltype == NLTYPE_FIXED && |
5837 | NLBLOCK->nllen == 2 && |
5838 | *eptr == NLBLOCK->nl[0]) |
5839 | { |
5840 | md->hitend = TRUE; |
5841 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5842 | } |
5843 | eptr++; |
5844 | } |
5845 | break; |
5846 | |
5847 | case OP_ALLANY: |
5848 | case OP_ANYBYTE: |
5849 | c = max - min; |
5850 | if (c > (unsigned int)(md->end_subject - eptr)) |
5851 | { |
5852 | eptr = md->end_subject; |
5853 | SCHECK_PARTIAL(); |
5854 | } |
5855 | else eptr += c; |
5856 | break; |
5857 | |
5858 | case OP_ANYNL: |
5859 | for (i = min; i < max; i++) |
5860 | { |
5861 | if (eptr >= md->end_subject) |
5862 | { |
5863 | SCHECK_PARTIAL(); |
5864 | break; |
5865 | } |
5866 | c = *eptr; |
5867 | if (c == CHAR_CR) |
5868 | { |
5869 | if (++eptr >= md->end_subject) break; |
5870 | if (*eptr == CHAR_LF) eptr++; |
5871 | } |
5872 | else |
5873 | { |
5874 | if (c != CHAR_LF && (md->bsr_anycrlf || |
5875 | (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL |
5876 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5877 | && c != 0x2028 && c != 0x2029 |
5878 | #endif |
5879 | ))) break; |
5880 | eptr++; |
5881 | } |
5882 | } |
5883 | break; |
5884 | |
5885 | case OP_NOT_HSPACE: |
5886 | for (i = min; i < max; i++) |
5887 | { |
5888 | if (eptr >= md->end_subject) |
5889 | { |
5890 | SCHECK_PARTIAL(); |
5891 | break; |
5892 | } |
5893 | switch(*eptr) |
5894 | { |
5895 | default: eptr++; break; |
5896 | HSPACE_BYTE_CASES: |
5897 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5898 | HSPACE_MULTIBYTE_CASES: |
5899 | #endif |
5900 | goto ENDLOOP00; |
5901 | } |
5902 | } |
5903 | ENDLOOP00: |
5904 | break; |
5905 | |
5906 | case OP_HSPACE: |
5907 | for (i = min; i < max; i++) |
5908 | { |
5909 | if (eptr >= md->end_subject) |
5910 | { |
5911 | SCHECK_PARTIAL(); |
5912 | break; |
5913 | } |
5914 | switch(*eptr) |
5915 | { |
5916 | default: goto ENDLOOP01; |
5917 | HSPACE_BYTE_CASES: |
5918 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5919 | HSPACE_MULTIBYTE_CASES: |
5920 | #endif |
5921 | eptr++; break; |
5922 | } |
5923 | } |
5924 | ENDLOOP01: |
5925 | break; |
5926 | |
5927 | case OP_NOT_VSPACE: |
5928 | for (i = min; i < max; i++) |
5929 | { |
5930 | if (eptr >= md->end_subject) |
5931 | { |
5932 | SCHECK_PARTIAL(); |
5933 | break; |
5934 | } |
5935 | switch(*eptr) |
5936 | { |
5937 | default: eptr++; break; |
5938 | VSPACE_BYTE_CASES: |
5939 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5940 | VSPACE_MULTIBYTE_CASES: |
5941 | #endif |
5942 | goto ENDLOOP02; |
5943 | } |
5944 | } |
5945 | ENDLOOP02: |
5946 | break; |
5947 | |
5948 | case OP_VSPACE: |
5949 | for (i = min; i < max; i++) |
5950 | { |
5951 | if (eptr >= md->end_subject) |
5952 | { |
5953 | SCHECK_PARTIAL(); |
5954 | break; |
5955 | } |
5956 | switch(*eptr) |
5957 | { |
5958 | default: goto ENDLOOP03; |
5959 | VSPACE_BYTE_CASES: |
5960 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5961 | VSPACE_MULTIBYTE_CASES: |
5962 | #endif |
5963 | eptr++; break; |
5964 | } |
5965 | } |
5966 | ENDLOOP03: |
5967 | break; |
5968 | |
5969 | case OP_NOT_DIGIT: |
5970 | for (i = min; i < max; i++) |
5971 | { |
5972 | if (eptr >= md->end_subject) |
5973 | { |
5974 | SCHECK_PARTIAL(); |
5975 | break; |
5976 | } |
5977 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) break; |
5978 | eptr++; |
5979 | } |
5980 | break; |
5981 | |
5982 | case OP_DIGIT: |
5983 | for (i = min; i < max; i++) |
5984 | { |
5985 | if (eptr >= md->end_subject) |
5986 | { |
5987 | SCHECK_PARTIAL(); |
5988 | break; |
5989 | } |
5990 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) break; |
5991 | eptr++; |
5992 | } |
5993 | break; |
5994 | |
5995 | case OP_NOT_WHITESPACE: |
5996 | for (i = min; i < max; i++) |
5997 | { |
5998 | if (eptr >= md->end_subject) |
5999 | { |
6000 | SCHECK_PARTIAL(); |
6001 | break; |
6002 | } |
6003 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break; |
6004 | eptr++; |
6005 | } |
6006 | break; |
6007 | |
6008 | case OP_WHITESPACE: |
6009 | for (i = min; i < max; i++) |
6010 | { |
6011 | if (eptr >= md->end_subject) |
6012 | { |
6013 | SCHECK_PARTIAL(); |
6014 | break; |
6015 | } |
6016 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break; |
6017 | eptr++; |
6018 | } |
6019 | break; |
6020 | |
6021 | case OP_NOT_WORDCHAR: |
6022 | for (i = min; i < max; i++) |
6023 | { |
6024 | if (eptr >= md->end_subject) |
6025 | { |
6026 | SCHECK_PARTIAL(); |
6027 | break; |
6028 | } |
6029 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break; |
6030 | eptr++; |
6031 | } |
6032 | break; |
6033 | |
6034 | case OP_WORDCHAR: |
6035 | for (i = min; i < max; i++) |
6036 | { |
6037 | if (eptr >= md->end_subject) |
6038 | { |
6039 | SCHECK_PARTIAL(); |
6040 | break; |
6041 | } |
6042 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break; |
6043 | eptr++; |
6044 | } |
6045 | break; |
6046 | |
6047 | default: |
6048 | RRETURN(PCRE_ERROR_INTERNAL); |
6049 | } |
6050 | |
6051 | /* eptr is now past the end of the maximum run. If possessive, we are |
6052 | done (no backing up). Otherwise, match at this position; anything other |
6053 | than no match is immediately returned. For nomatch, back up one |
6054 | character (byte), unless we are matching \R and the last thing matched |
6055 | was \r\n, in which case, back up two bytes. */ |
6056 | |
6057 | if (possessive) continue; |
6058 | while (eptr >= pp) |
6059 | { |
6060 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM47); |
6061 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
6062 | eptr--; |
6063 | if (ctype == OP_ANYNL && eptr > pp && *eptr == CHAR_LF && |
6064 | eptr[-1] == CHAR_CR) eptr--; |
6065 | } |
6066 | } |
6067 | |
6068 | /* Get here if we can't make it match with any permitted repetitions */ |
6069 | |
6070 | RRETURN(MATCH_NOMATCH); |
6071 | } |
6072 | /* Control never gets here */ |
6073 | |
6074 | /* There's been some horrible disaster. Arrival here can only mean there is |
6075 | something seriously wrong in the code above or the OP_xxx definitions. */ |
6076 | |
6077 | default: |
6078 | DPRINTF(("Unknown opcode %d\n", *ecode)); |
6079 | RRETURN(PCRE_ERROR_UNKNOWN_OPCODE); |
6080 | } |
6081 | |
6082 | /* Do not stick any code in here without much thought; it is assumed |
6083 | that "continue" in the code above comes out to here to repeat the main |
6084 | loop. */ |
6085 | |
6086 | } /* End of main loop */ |
6087 | /* Control never reaches here */ |
6088 | |
6089 | |
6090 | /* When compiling to use the heap rather than the stack for recursive calls to |
6091 | match(), the RRETURN() macro jumps here. The number that is saved in |
6092 | frame->Xwhere indicates which label we actually want to return to. */ |
6093 | |
6094 | #ifdef NO_RECURSE |
6095 | #define LBL(val) case val: goto L_RM##val; |
6096 | HEAP_RETURN: |
6097 | switch (frame->Xwhere) |
6098 | { |
6099 | LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8) |
6100 | LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17) |
6101 | LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33) |
6102 | LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52) |
6103 | LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64) |
6104 | LBL(65) LBL(66) |
6105 | #if defined SUPPORT_UTF || !defined COMPILE_PCRE8 |
6106 | LBL(21) |
6107 | #endif |
6108 | #ifdef SUPPORT_UTF |
6109 | LBL(16) LBL(18) LBL(20) |
6110 | LBL(22) LBL(23) LBL(28) LBL(30) |
6111 | LBL(32) LBL(34) LBL(42) LBL(46) |
6112 | #ifdef SUPPORT_UCP |
6113 | LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45) |
6114 | LBL(59) LBL(60) LBL(61) LBL(62) LBL(67) |
6115 | #endif /* SUPPORT_UCP */ |
6116 | #endif /* SUPPORT_UTF */ |
6117 | default: |
6118 | DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere)); |
6119 | return PCRE_ERROR_INTERNAL; |
6120 | } |
6121 | #undef LBL |
6122 | #endif /* NO_RECURSE */ |
6123 | } |
6124 | |
6125 | |
6126 | /*************************************************************************** |
6127 | **************************************************************************** |
6128 | RECURSION IN THE match() FUNCTION |
6129 | |
6130 | Undefine all the macros that were defined above to handle this. */ |
6131 | |
6132 | #ifdef NO_RECURSE |
6133 | #undef eptr |
6134 | #undef ecode |
6135 | #undef mstart |
6136 | #undef offset_top |
6137 | #undef eptrb |
6138 | #undef flags |
6139 | |
6140 | #undef callpat |
6141 | #undef charptr |
6142 | #undef data |
6143 | #undef next |
6144 | #undef pp |
6145 | #undef prev |
6146 | #undef saved_eptr |
6147 | |
6148 | #undef new_recursive |
6149 | |
6150 | #undef cur_is_word |
6151 | #undef condition |
6152 | #undef prev_is_word |
6153 | |
6154 | #undef ctype |
6155 | #undef length |
6156 | #undef max |
6157 | #undef min |
6158 | #undef number |
6159 | #undef offset |
6160 | #undef op |
6161 | #undef save_capture_last |
6162 | #undef save_offset1 |
6163 | #undef save_offset2 |
6164 | #undef save_offset3 |
6165 | #undef stacksave |
6166 | |
6167 | #undef newptrb |
6168 | |
6169 | #endif |
6170 | |
6171 | /* These two are defined as macros in both cases */ |
6172 | |
6173 | #undef fc |
6174 | #undef fi |
6175 | |
6176 | /*************************************************************************** |
6177 | ***************************************************************************/ |
6178 | |
6179 | |
6180 | #ifdef NO_RECURSE |
6181 | /************************************************* |
6182 | * Release allocated heap frames * |
6183 | *************************************************/ |
6184 | |
6185 | /* This function releases all the allocated frames. The base frame is on the |
6186 | machine stack, and so must not be freed. |
6187 | |
6188 | Argument: the address of the base frame |
6189 | Returns: nothing |
6190 | */ |
6191 | |
6192 | static void |
6193 | release_match_heapframes (heapframe *frame_base) |
6194 | { |
6195 | heapframe *nextframe = frame_base->Xnextframe; |
6196 | while (nextframe != NULL) |
6197 | { |
6198 | heapframe *oldframe = nextframe; |
6199 | nextframe = nextframe->Xnextframe; |
6200 | (PUBL(stack_free))(oldframe); |
6201 | } |
6202 | } |
6203 | #endif |
6204 | |
6205 | |
6206 | /************************************************* |
6207 | * Execute a Regular Expression * |
6208 | *************************************************/ |
6209 | |
6210 | /* This function applies a compiled re to a subject string and picks out |
6211 | portions of the string if it matches. Two elements in the vector are set for |
6212 | each substring: the offsets to the start and end of the substring. |
6213 | |
6214 | Arguments: |
6215 | argument_re points to the compiled expression |
6216 | extra_data points to extra data or is NULL |
6217 | subject points to the subject string |
6218 | length length of subject string (may contain binary zeros) |
6219 | start_offset where to start in the subject string |
6220 | options option bits |
6221 | offsets points to a vector of ints to be filled in with offsets |
6222 | offsetcount the number of elements in the vector |
6223 | |
6224 | Returns: > 0 => success; value is the number of elements filled in |
6225 | = 0 => success, but offsets is not big enough |
6226 | -1 => failed to match |
6227 | < -1 => some kind of unexpected problem |
6228 | */ |
6229 | |
6230 | #if defined COMPILE_PCRE8 |
6231 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6232 | pcre_exec(const pcre *argument_re, const pcre_extra *extra_data, |
6233 | PCRE_SPTR subject, int length, int start_offset, int options, int *offsets, |
6234 | int offsetcount) |
6235 | #elif defined COMPILE_PCRE16 |
6236 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6237 | pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data, |
6238 | PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets, |
6239 | int offsetcount) |
6240 | #elif defined COMPILE_PCRE32 |
6241 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6242 | pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data, |
6243 | PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets, |
6244 | int offsetcount) |
6245 | #endif |
6246 | { |
6247 | int rc, ocount, arg_offset_max; |
6248 | int newline; |
6249 | BOOL using_temporary_offsets = FALSE; |
6250 | BOOL anchored; |
6251 | BOOL startline; |
6252 | BOOL firstline; |
6253 | BOOL utf; |
6254 | BOOL has_first_char = FALSE; |
6255 | BOOL has_req_char = FALSE; |
6256 | pcre_uchar first_char = 0; |
6257 | pcre_uchar first_char2 = 0; |
6258 | pcre_uchar req_char = 0; |
6259 | pcre_uchar req_char2 = 0; |
6260 | match_data match_block; |
6261 | match_data *md = &match_block; |
6262 | const pcre_uint8 *tables; |
6263 | const pcre_uint8 *start_bits = NULL; |
6264 | PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset; |
6265 | PCRE_PUCHAR end_subject; |
6266 | PCRE_PUCHAR start_partial = NULL; |
6267 | PCRE_PUCHAR req_char_ptr = start_match - 1; |
6268 | |
6269 | const pcre_study_data *study; |
6270 | const REAL_PCRE *re = (const REAL_PCRE *)argument_re; |
6271 | |
6272 | #ifdef NO_RECURSE |
6273 | heapframe frame_zero; |
6274 | frame_zero.Xprevframe = NULL; /* Marks the top level */ |
6275 | frame_zero.Xnextframe = NULL; /* None are allocated yet */ |
6276 | md->match_frames_base = &frame_zero; |
6277 | #endif |
6278 | |
6279 | /* Check for the special magic call that measures the size of the stack used |
6280 | per recursive call of match(). Without the funny casting for sizeof, a Windows |
6281 | compiler gave this error: "unary minus operator applied to unsigned type, |
6282 | result still unsigned". Hopefully the cast fixes that. */ |
6283 | |
6284 | if (re == NULL && extra_data == NULL && subject == NULL && length == -999 && |
6285 | start_offset == -999) |
6286 | #ifdef NO_RECURSE |
6287 | return -((int)sizeof(heapframe)); |
6288 | #else |
6289 | return match(NULL, NULL, NULL, 0, NULL, NULL, 0); |
6290 | #endif |
6291 | |
6292 | /* Plausibility checks */ |
6293 | |
6294 | if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION; |
6295 | if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0)) |
6296 | return PCRE_ERROR_NULL; |
6297 | if (offsetcount < 0) return PCRE_ERROR_BADCOUNT; |
6298 | if (length < 0) return PCRE_ERROR_BADLENGTH; |
6299 | if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET; |
6300 | |
6301 | /* Check that the first field in the block is the magic number. If it is not, |
6302 | return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to |
6303 | REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which |
6304 | means that the pattern is likely compiled with different endianness. */ |
6305 | |
6306 | if (re->magic_number != MAGIC_NUMBER) |
6307 | return re->magic_number == REVERSED_MAGIC_NUMBER? |
6308 | PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC; |
6309 | if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE; |
6310 | |
6311 | /* These two settings are used in the code for checking a UTF-8 string that |
6312 | follows immediately afterwards. Other values in the md block are used only |
6313 | during "normal" pcre_exec() processing, not when the JIT support is in use, |
6314 | so they are set up later. */ |
6315 | |
6316 | /* PCRE_UTF16 has the same value as PCRE_UTF8. */ |
6317 | utf = md->utf = (re->options & PCRE_UTF8) != 0; |
6318 | md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 : |
6319 | ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0; |
6320 | |
6321 | /* Check a UTF-8 string if required. Pass back the character offset and error |
6322 | code for an invalid string if a results vector is available. */ |
6323 | |
6324 | #ifdef SUPPORT_UTF |
6325 | if (utf && (options & PCRE_NO_UTF8_CHECK) == 0) |
6326 | { |
6327 | int erroroffset; |
6328 | int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset); |
6329 | if (errorcode != 0) |
6330 | { |
6331 | if (offsetcount >= 2) |
6332 | { |
6333 | offsets[0] = erroroffset; |
6334 | offsets[1] = errorcode; |
6335 | } |
6336 | #if defined COMPILE_PCRE8 |
6337 | return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)? |
6338 | PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8; |
6339 | #elif defined COMPILE_PCRE16 |
6340 | return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)? |
6341 | PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16; |
6342 | #elif defined COMPILE_PCRE32 |
6343 | return PCRE_ERROR_BADUTF32; |
6344 | #endif |
6345 | } |
6346 | #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16 |
6347 | /* Check that a start_offset points to the start of a UTF character. */ |
6348 | if (start_offset > 0 && start_offset < length && |
6349 | NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset])) |
6350 | return PCRE_ERROR_BADUTF8_OFFSET; |
6351 | #endif |
6352 | } |
6353 | #endif |
6354 | |
6355 | /* If the pattern was successfully studied with JIT support, run the JIT |
6356 | executable instead of the rest of this function. Most options must be set at |
6357 | compile time for the JIT code to be usable. Fallback to the normal code path if |
6358 | an unsupported flag is set. */ |
6359 | |
6360 | #ifdef SUPPORT_JIT |
6361 | if (extra_data != NULL |
6362 | && (extra_data->flags & (PCRE_EXTRA_EXECUTABLE_JIT | |
6363 | PCRE_EXTRA_TABLES)) == PCRE_EXTRA_EXECUTABLE_JIT |
6364 | && extra_data->executable_jit != NULL |
6365 | && (options & ~PUBLIC_JIT_EXEC_OPTIONS) == 0) |
6366 | { |
6367 | rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length, |
6368 | start_offset, options, offsets, offsetcount); |
6369 | |
6370 | /* PCRE_ERROR_NULL means that the selected normal or partial matching |
6371 | mode is not compiled. In this case we simply fallback to interpreter. */ |
6372 | |
6373 | if (rc != PCRE_ERROR_JIT_BADOPTION) return rc; |
6374 | } |
6375 | #endif |
6376 | |
6377 | /* Carry on with non-JIT matching. This information is for finding all the |
6378 | numbers associated with a given name, for condition testing. */ |
6379 | |
6380 | md->name_table = (pcre_uchar *)re + re->name_table_offset; |
6381 | md->name_count = re->name_count; |
6382 | md->name_entry_size = re->name_entry_size; |
6383 | |
6384 | /* Fish out the optional data from the extra_data structure, first setting |
6385 | the default values. */ |
6386 | |
6387 | study = NULL; |
6388 | md->match_limit = MATCH_LIMIT; |
6389 | md->match_limit_recursion = MATCH_LIMIT_RECURSION; |
6390 | md->callout_data = NULL; |
6391 | |
6392 | /* The table pointer is always in native byte order. */ |
6393 | |
6394 | tables = re->tables; |
6395 | |
6396 | if (extra_data != NULL) |
6397 | { |
6398 | register unsigned int flags = extra_data->flags; |
6399 | if ((flags & PCRE_EXTRA_STUDY_DATA) != 0) |
6400 | study = (const pcre_study_data *)extra_data->study_data; |
6401 | if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0) |
6402 | md->match_limit = extra_data->match_limit; |
6403 | if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0) |
6404 | md->match_limit_recursion = extra_data->match_limit_recursion; |
6405 | if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0) |
6406 | md->callout_data = extra_data->callout_data; |
6407 | if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables; |
6408 | } |
6409 | |
6410 | /* If the exec call supplied NULL for tables, use the inbuilt ones. This |
6411 | is a feature that makes it possible to save compiled regex and re-use them |
6412 | in other programs later. */ |
6413 | |
6414 | if (tables == NULL) tables = PRIV(default_tables); |
6415 | |
6416 | /* Set up other data */ |
6417 | |
6418 | anchored = ((re->options | options) & PCRE_ANCHORED) != 0; |
6419 | startline = (re->flags & PCRE_STARTLINE) != 0; |
6420 | firstline = (re->options & PCRE_FIRSTLINE) != 0; |
6421 | |
6422 | /* The code starts after the real_pcre block and the capture name table. */ |
6423 | |
6424 | md->start_code = (const pcre_uchar *)re + re->name_table_offset + |
6425 | re->name_count * re->name_entry_size; |
6426 | |
6427 | md->start_subject = (PCRE_PUCHAR)subject; |
6428 | md->start_offset = start_offset; |
6429 | md->end_subject = md->start_subject + length; |
6430 | end_subject = md->end_subject; |
6431 | |
6432 | md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0; |
6433 | md->use_ucp = (re->options & PCRE_UCP) != 0; |
6434 | md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0; |
6435 | md->ignore_skip_arg = FALSE; |
6436 | |
6437 | /* Some options are unpacked into BOOL variables in the hope that testing |
6438 | them will be faster than individual option bits. */ |
6439 | |
6440 | md->notbol = (options & PCRE_NOTBOL) != 0; |
6441 | md->noteol = (options & PCRE_NOTEOL) != 0; |
6442 | md->notempty = (options & PCRE_NOTEMPTY) != 0; |
6443 | md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0; |
6444 | |
6445 | md->hitend = FALSE; |
6446 | md->mark = md->nomatch_mark = NULL; /* In case never set */ |
6447 | |
6448 | md->recursive = NULL; /* No recursion at top level */ |
6449 | md->hasthen = (re->flags & PCRE_HASTHEN) != 0; |
6450 | |
6451 | md->lcc = tables + lcc_offset; |
6452 | md->fcc = tables + fcc_offset; |
6453 | md->ctypes = tables + ctypes_offset; |
6454 | |
6455 | /* Handle different \R options. */ |
6456 | |
6457 | switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) |
6458 | { |
6459 | case 0: |
6460 | if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0) |
6461 | md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0; |
6462 | else |
6463 | #ifdef BSR_ANYCRLF |
6464 | md->bsr_anycrlf = TRUE; |
6465 | #else |
6466 | md->bsr_anycrlf = FALSE; |
6467 | #endif |
6468 | break; |
6469 | |
6470 | case PCRE_BSR_ANYCRLF: |
6471 | md->bsr_anycrlf = TRUE; |
6472 | break; |
6473 | |
6474 | case PCRE_BSR_UNICODE: |
6475 | md->bsr_anycrlf = FALSE; |
6476 | break; |
6477 | |
6478 | default: return PCRE_ERROR_BADNEWLINE; |
6479 | } |
6480 | |
6481 | /* Handle different types of newline. The three bits give eight cases. If |
6482 | nothing is set at run time, whatever was used at compile time applies. */ |
6483 | |
6484 | switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options : |
6485 | (pcre_uint32)options) & PCRE_NEWLINE_BITS) |
6486 | { |
6487 | case 0: newline = NEWLINE; break; /* Compile-time default */ |
6488 | case PCRE_NEWLINE_CR: newline = CHAR_CR; break; |
6489 | case PCRE_NEWLINE_LF: newline = CHAR_NL; break; |
6490 | case PCRE_NEWLINE_CR+ |
6491 | PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break; |
6492 | case PCRE_NEWLINE_ANY: newline = -1; break; |
6493 | case PCRE_NEWLINE_ANYCRLF: newline = -2; break; |
6494 | default: return PCRE_ERROR_BADNEWLINE; |
6495 | } |
6496 | |
6497 | if (newline == -2) |
6498 | { |
6499 | md->nltype = NLTYPE_ANYCRLF; |
6500 | } |
6501 | else if (newline < 0) |
6502 | { |
6503 | md->nltype = NLTYPE_ANY; |
6504 | } |
6505 | else |
6506 | { |
6507 | md->nltype = NLTYPE_FIXED; |
6508 | if (newline > 255) |
6509 | { |
6510 | md->nllen = 2; |
6511 | md->nl[0] = (newline >> 8) & 255; |
6512 | md->nl[1] = newline & 255; |
6513 | } |
6514 | else |
6515 | { |
6516 | md->nllen = 1; |
6517 | md->nl[0] = newline; |
6518 | } |
6519 | } |
6520 | |
6521 | /* Partial matching was originally supported only for a restricted set of |
6522 | regexes; from release 8.00 there are no restrictions, but the bits are still |
6523 | defined (though never set). So there's no harm in leaving this code. */ |
6524 | |
6525 | if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0) |
6526 | return PCRE_ERROR_BADPARTIAL; |
6527 | |
6528 | /* If the expression has got more back references than the offsets supplied can |
6529 | hold, we get a temporary chunk of working store to use during the matching. |
6530 | Otherwise, we can use the vector supplied, rounding down its size to a multiple |
6531 | of 3. */ |
6532 | |
6533 | ocount = offsetcount - (offsetcount % 3); |
6534 | arg_offset_max = (2*ocount)/3; |
6535 | |
6536 | if (re->top_backref > 0 && re->top_backref >= ocount/3) |
6537 | { |
6538 | ocount = re->top_backref * 3 + 3; |
6539 | md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int)); |
6540 | if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY; |
6541 | using_temporary_offsets = TRUE; |
6542 | DPRINTF(("Got memory to hold back references\n")); |
6543 | } |
6544 | else md->offset_vector = offsets; |
6545 | |
6546 | md->offset_end = ocount; |
6547 | md->offset_max = (2*ocount)/3; |
6548 | md->offset_overflow = FALSE; |
6549 | md->capture_last = -1; |
6550 | |
6551 | /* Reset the working variable associated with each extraction. These should |
6552 | never be used unless previously set, but they get saved and restored, and so we |
6553 | initialize them to avoid reading uninitialized locations. Also, unset the |
6554 | offsets for the matched string. This is really just for tidiness with callouts, |
6555 | in case they inspect these fields. */ |
6556 | |
6557 | if (md->offset_vector != NULL) |
6558 | { |
6559 | register int *iptr = md->offset_vector + ocount; |
6560 | register int *iend = iptr - re->top_bracket; |
6561 | if (iend < md->offset_vector + 2) iend = md->offset_vector + 2; |
6562 | while (--iptr >= iend) *iptr = -1; |
6563 | md->offset_vector[0] = md->offset_vector[1] = -1; |
6564 | } |
6565 | |
6566 | /* Set up the first character to match, if available. The first_char value is |
6567 | never set for an anchored regular expression, but the anchoring may be forced |
6568 | at run time, so we have to test for anchoring. The first char may be unset for |
6569 | an unanchored pattern, of course. If there's no first char and the pattern was |
6570 | studied, there may be a bitmap of possible first characters. */ |
6571 | |
6572 | if (!anchored) |
6573 | { |
6574 | if ((re->flags & PCRE_FIRSTSET) != 0) |
6575 | { |
6576 | has_first_char = TRUE; |
6577 | first_char = first_char2 = (pcre_uchar)(re->first_char); |
6578 | if ((re->flags & PCRE_FCH_CASELESS) != 0) |
6579 | { |
6580 | first_char2 = TABLE_GET(first_char, md->fcc, first_char); |
6581 | #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8) |
6582 | if (utf && first_char > 127) |
6583 | first_char2 = UCD_OTHERCASE(first_char); |
6584 | #endif |
6585 | } |
6586 | } |
6587 | else |
6588 | if (!startline && study != NULL && |
6589 | (study->flags & PCRE_STUDY_MAPPED) != 0) |
6590 | start_bits = study->start_bits; |
6591 | } |
6592 | |
6593 | /* For anchored or unanchored matches, there may be a "last known required |
6594 | character" set. */ |
6595 | |
6596 | if ((re->flags & PCRE_REQCHSET) != 0) |
6597 | { |
6598 | has_req_char = TRUE; |
6599 | req_char = req_char2 = (pcre_uchar)(re->req_char); |
6600 | if ((re->flags & PCRE_RCH_CASELESS) != 0) |
6601 | { |
6602 | req_char2 = TABLE_GET(req_char, md->fcc, req_char); |
6603 | #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8) |
6604 | if (utf && req_char > 127) |
6605 | req_char2 = UCD_OTHERCASE(req_char); |
6606 | #endif |
6607 | } |
6608 | } |
6609 | |
6610 | |
6611 | /* ==========================================================================*/ |
6612 | |
6613 | /* Loop for handling unanchored repeated matching attempts; for anchored regexs |
6614 | the loop runs just once. */ |
6615 | |
6616 | for(;;) |
6617 | { |
6618 | PCRE_PUCHAR save_end_subject = end_subject; |
6619 | PCRE_PUCHAR new_start_match; |
6620 | |
6621 | /* If firstline is TRUE, the start of the match is constrained to the first |
6622 | line of a multiline string. That is, the match must be before or at the first |
6623 | newline. Implement this by temporarily adjusting end_subject so that we stop |
6624 | scanning at a newline. If the match fails at the newline, later code breaks |
6625 | this loop. */ |
6626 | |
6627 | if (firstline) |
6628 | { |
6629 | PCRE_PUCHAR t = start_match; |
6630 | #ifdef SUPPORT_UTF |
6631 | if (utf) |
6632 | { |
6633 | while (t < md->end_subject && !IS_NEWLINE(t)) |
6634 | { |
6635 | t++; |
6636 | ACROSSCHAR(t < end_subject, *t, t++); |
6637 | } |
6638 | } |
6639 | else |
6640 | #endif |
6641 | while (t < md->end_subject && !IS_NEWLINE(t)) t++; |
6642 | end_subject = t; |
6643 | } |
6644 | |
6645 | /* There are some optimizations that avoid running the match if a known |
6646 | starting point is not found, or if a known later character is not present. |
6647 | However, there is an option that disables these, for testing and for ensuring |
6648 | that all callouts do actually occur. The option can be set in the regex by |
6649 | (*NO_START_OPT) or passed in match-time options. */ |
6650 | |
6651 | if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0) |
6652 | { |
6653 | /* Advance to a unique first char if there is one. */ |
6654 | |
6655 | if (has_first_char) |
6656 | { |
6657 | pcre_uchar smc; |
6658 | |
6659 | if (first_char != first_char2) |
6660 | while (start_match < end_subject && |
6661 | (smc = RAWUCHARTEST(start_match)) != first_char && smc != first_char2) |
6662 | start_match++; |
6663 | else |
6664 | while (start_match < end_subject && RAWUCHARTEST(start_match) != first_char) |
6665 | start_match++; |
6666 | } |
6667 | |
6668 | /* Or to just after a linebreak for a multiline match */ |
6669 | |
6670 | else if (startline) |
6671 | { |
6672 | if (start_match > md->start_subject + start_offset) |
6673 | { |
6674 | #ifdef SUPPORT_UTF |
6675 | if (utf) |
6676 | { |
6677 | while (start_match < end_subject && !WAS_NEWLINE(start_match)) |
6678 | { |
6679 | start_match++; |
6680 | ACROSSCHAR(start_match < end_subject, *start_match, |
6681 | start_match++); |
6682 | } |
6683 | } |
6684 | else |
6685 | #endif |
6686 | while (start_match < end_subject && !WAS_NEWLINE(start_match)) |
6687 | start_match++; |
6688 | |
6689 | /* If we have just passed a CR and the newline option is ANY or ANYCRLF, |
6690 | and we are now at a LF, advance the match position by one more character. |
6691 | */ |
6692 | |
6693 | if (start_match[-1] == CHAR_CR && |
6694 | (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) && |
6695 | start_match < end_subject && |
6696 | RAWUCHARTEST(start_match) == CHAR_NL) |
6697 | start_match++; |
6698 | } |
6699 | } |
6700 | |
6701 | /* Or to a non-unique first byte after study */ |
6702 | |
6703 | else if (start_bits != NULL) |
6704 | { |
6705 | while (start_match < end_subject) |
6706 | { |
6707 | register pcre_uint32 c = RAWUCHARTEST(start_match); |
6708 | #ifndef COMPILE_PCRE8 |
6709 | if (c > 255) c = 255; |
6710 | #endif |
6711 | if ((start_bits[c/8] & (1 << (c&7))) == 0) |
6712 | { |
6713 | start_match++; |
6714 | #if defined SUPPORT_UTF && defined COMPILE_PCRE8 |
6715 | /* In non 8-bit mode, the iteration will stop for |
6716 | characters > 255 at the beginning or not stop at all. */ |
6717 | if (utf) |
6718 | ACROSSCHAR(start_match < end_subject, *start_match, |
6719 | start_match++); |
6720 | #endif |
6721 | } |
6722 | else break; |
6723 | } |
6724 | } |
6725 | } /* Starting optimizations */ |
6726 | |
6727 | /* Restore fudged end_subject */ |
6728 | |
6729 | end_subject = save_end_subject; |
6730 | |
6731 | /* The following two optimizations are disabled for partial matching or if |
6732 | disabling is explicitly requested. */ |
6733 | |
6734 | if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0 && !md->partial) |
6735 | { |
6736 | /* If the pattern was studied, a minimum subject length may be set. This is |
6737 | a lower bound; no actual string of that length may actually match the |
6738 | pattern. Although the value is, strictly, in characters, we treat it as |
6739 | bytes to avoid spending too much time in this optimization. */ |
6740 | |
6741 | if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 && |
6742 | (pcre_uint32)(end_subject - start_match) < study->minlength) |
6743 | { |
6744 | rc = MATCH_NOMATCH; |
6745 | break; |
6746 | } |
6747 | |
6748 | /* If req_char is set, we know that that character must appear in the |
6749 | subject for the match to succeed. If the first character is set, req_char |
6750 | must be later in the subject; otherwise the test starts at the match point. |
6751 | This optimization can save a huge amount of backtracking in patterns with |
6752 | nested unlimited repeats that aren't going to match. Writing separate code |
6753 | for cased/caseless versions makes it go faster, as does using an |
6754 | autoincrement and backing off on a match. |
6755 | |
6756 | HOWEVER: when the subject string is very, very long, searching to its end |
6757 | can take a long time, and give bad performance on quite ordinary patterns. |
6758 | This showed up when somebody was matching something like /^\d+C/ on a |
6759 | 32-megabyte string... so we don't do this when the string is sufficiently |
6760 | long. */ |
6761 | |
6762 | if (has_req_char && end_subject - start_match < REQ_BYTE_MAX) |
6763 | { |
6764 | register PCRE_PUCHAR p = start_match + (has_first_char? 1:0); |
6765 | |
6766 | /* We don't need to repeat the search if we haven't yet reached the |