Parent Directory
|
Revision Log
Add pcre16 prefix to 16 bit structs
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 | |
41 | /* This module contains pcre_exec(), the externally visible function that does |
42 | pattern matching using an NFA algorithm, trying to mimic Perl as closely as |
43 | possible. There are also some static supporting functions. */ |
44 | |
45 | #ifdef HAVE_CONFIG_H |
46 | #include "config.h" |
47 | #endif |
48 | |
49 | #define NLBLOCK md /* Block containing newline information */ |
50 | #define PSSTART start_subject /* Field containing processed string start */ |
51 | #define PSEND end_subject /* Field containing processed string end */ |
52 | |
53 | #include "pcre_internal.h" |
54 | |
55 | /* Undefine some potentially clashing cpp symbols */ |
56 | |
57 | #undef min |
58 | #undef max |
59 | |
60 | /* Values for setting in md->match_function_type to indicate two special types |
61 | of call to match(). We do it this way to save on using another stack variable, |
62 | as stack usage is to be discouraged. */ |
63 | |
64 | #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */ |
65 | #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */ |
66 | |
67 | /* Non-error returns from the match() function. Error returns are externally |
68 | defined PCRE_ERROR_xxx codes, which are all negative. */ |
69 | |
70 | #define MATCH_MATCH 1 |
71 | #define MATCH_NOMATCH 0 |
72 | |
73 | /* Special internal returns from the match() function. Make them sufficiently |
74 | negative to avoid the external error codes. */ |
75 | |
76 | #define MATCH_ACCEPT (-999) |
77 | #define MATCH_COMMIT (-998) |
78 | #define MATCH_KETRPOS (-997) |
79 | #define MATCH_ONCE (-996) |
80 | #define MATCH_PRUNE (-995) |
81 | #define MATCH_SKIP (-994) |
82 | #define MATCH_SKIP_ARG (-993) |
83 | #define MATCH_THEN (-992) |
84 | |
85 | /* Maximum number of ints of offset to save on the stack for recursive calls. |
86 | If the offset vector is bigger, malloc is used. This should be a multiple of 3, |
87 | because the offset vector is always a multiple of 3 long. */ |
88 | |
89 | #define REC_STACK_SAVE_MAX 30 |
90 | |
91 | /* Min and max values for the common repeats; for the maxima, 0 => infinity */ |
92 | |
93 | static const char rep_min[] = { 0, 0, 1, 1, 0, 0 }; |
94 | static const char rep_max[] = { 0, 0, 0, 0, 1, 1 }; |
95 | |
96 | |
97 | |
98 | #ifdef PCRE_DEBUG |
99 | /************************************************* |
100 | * Debugging function to print chars * |
101 | *************************************************/ |
102 | |
103 | /* Print a sequence of chars in printable format, stopping at the end of the |
104 | subject if the requested. |
105 | |
106 | Arguments: |
107 | p points to characters |
108 | length number to print |
109 | is_subject TRUE if printing from within md->start_subject |
110 | md pointer to matching data block, if is_subject is TRUE |
111 | |
112 | Returns: nothing |
113 | */ |
114 | |
115 | static void |
116 | pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md) |
117 | { |
118 | unsigned int c; |
119 | if (is_subject && length > md->end_subject - p) length = md->end_subject - p; |
120 | while (length-- > 0) |
121 | if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c); |
122 | } |
123 | #endif |
124 | |
125 | |
126 | |
127 | /************************************************* |
128 | * Match a back-reference * |
129 | *************************************************/ |
130 | |
131 | /* Normally, if a back reference hasn't been set, the length that is passed is |
132 | negative, so the match always fails. However, in JavaScript compatibility mode, |
133 | the length passed is zero. Note that in caseless UTF-8 mode, the number of |
134 | subject bytes matched may be different to the number of reference bytes. |
135 | |
136 | Arguments: |
137 | offset index into the offset vector |
138 | eptr pointer into the subject |
139 | length length of reference to be matched (number of bytes) |
140 | md points to match data block |
141 | caseless TRUE if caseless |
142 | |
143 | Returns: < 0 if not matched, otherwise the number of subject bytes matched |
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 | |
153 | #ifdef PCRE_DEBUG |
154 | if (eptr >= md->end_subject) |
155 | printf("matching subject <null>"); |
156 | else |
157 | { |
158 | printf("matching subject "); |
159 | pchars(eptr, length, TRUE, md); |
160 | } |
161 | printf(" against backref "); |
162 | pchars(p, length, FALSE, md); |
163 | printf("\n"); |
164 | #endif |
165 | |
166 | /* Always fail if reference not set (and not JavaScript compatible). */ |
167 | |
168 | if (length < 0) return -1; |
169 | |
170 | /* Separate the caseless case for speed. In UTF-8 mode we can only do this |
171 | properly if Unicode properties are supported. Otherwise, we can check only |
172 | ASCII characters. */ |
173 | |
174 | if (caseless) |
175 | { |
176 | #ifdef SUPPORT_UTF |
177 | #ifdef SUPPORT_UCP |
178 | if (md->utf) |
179 | { |
180 | /* Match characters up to the end of the reference. NOTE: the number of |
181 | bytes matched may differ, because there are some characters whose upper and |
182 | lower case versions code as different numbers of bytes. For example, U+023A |
183 | (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8); |
184 | a sequence of 3 of the former uses 6 bytes, as does a sequence of two of |
185 | the latter. It is important, therefore, to check the length along the |
186 | reference, not along the subject (earlier code did this wrong). */ |
187 | |
188 | PCRE_PUCHAR endptr = p + length; |
189 | while (p < endptr) |
190 | { |
191 | int c, d; |
192 | if (eptr >= md->end_subject) return -1; |
193 | GETCHARINC(c, eptr); |
194 | GETCHARINC(d, p); |
195 | if (c != d && c != UCD_OTHERCASE(d)) return -1; |
196 | } |
197 | } |
198 | else |
199 | #endif |
200 | #endif |
201 | |
202 | /* The same code works when not in UTF-8 mode and in UTF-8 mode when there |
203 | is no UCP support. */ |
204 | { |
205 | if (eptr + length > md->end_subject) return -1; |
206 | while (length-- > 0) |
207 | { |
208 | if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1; |
209 | p++; |
210 | eptr++; |
211 | } |
212 | } |
213 | } |
214 | |
215 | /* In the caseful case, we can just compare the bytes, whether or not we |
216 | are in UTF-8 mode. */ |
217 | |
218 | else |
219 | { |
220 | if (eptr + length > md->end_subject) return -1; |
221 | while (length-- > 0) if (*p++ != *eptr++) return -1; |
222 | } |
223 | |
224 | return (int)(eptr - eptr_start); |
225 | } |
226 | |
227 | |
228 | |
229 | /*************************************************************************** |
230 | **************************************************************************** |
231 | RECURSION IN THE match() FUNCTION |
232 | |
233 | The match() function is highly recursive, though not every recursive call |
234 | increases the recursive depth. Nevertheless, some regular expressions can cause |
235 | it to recurse to a great depth. I was writing for Unix, so I just let it call |
236 | itself recursively. This uses the stack for saving everything that has to be |
237 | saved for a recursive call. On Unix, the stack can be large, and this works |
238 | fine. |
239 | |
240 | It turns out that on some non-Unix-like systems there are problems with |
241 | programs that use a lot of stack. (This despite the fact that every last chip |
242 | has oodles of memory these days, and techniques for extending the stack have |
243 | been known for decades.) So.... |
244 | |
245 | There is a fudge, triggered by defining NO_RECURSE, which avoids recursive |
246 | calls by keeping local variables that need to be preserved in blocks of memory |
247 | obtained from malloc() instead instead of on the stack. Macros are used to |
248 | achieve this so that the actual code doesn't look very different to what it |
249 | always used to. |
250 | |
251 | The original heap-recursive code used longjmp(). However, it seems that this |
252 | can be very slow on some operating systems. Following a suggestion from Stan |
253 | Switzer, the use of longjmp() has been abolished, at the cost of having to |
254 | provide a unique number for each call to RMATCH. There is no way of generating |
255 | a sequence of numbers at compile time in C. I have given them names, to make |
256 | them stand out more clearly. |
257 | |
258 | Crude tests on x86 Linux show a small speedup of around 5-8%. However, on |
259 | FreeBSD, avoiding longjmp() more than halves the time taken to run the standard |
260 | tests. Furthermore, not using longjmp() means that local dynamic variables |
261 | don't have indeterminate values; this has meant that the frame size can be |
262 | reduced because the result can be "passed back" by straight setting of the |
263 | variable instead of being passed in the frame. |
264 | **************************************************************************** |
265 | ***************************************************************************/ |
266 | |
267 | /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN |
268 | below must be updated in sync. */ |
269 | |
270 | enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10, |
271 | RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20, |
272 | RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30, |
273 | RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40, |
274 | RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50, |
275 | RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60, |
276 | RM61, RM62, RM63, RM64, RM65, RM66 }; |
277 | |
278 | /* These versions of the macros use the stack, as normal. There are debugging |
279 | versions and production versions. Note that the "rw" argument of RMATCH isn't |
280 | actually used in this definition. */ |
281 | |
282 | #ifndef NO_RECURSE |
283 | #define REGISTER register |
284 | |
285 | #ifdef PCRE_DEBUG |
286 | #define RMATCH(ra,rb,rc,rd,re,rw) \ |
287 | { \ |
288 | printf("match() called in line %d\n", __LINE__); \ |
289 | rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \ |
290 | printf("to line %d\n", __LINE__); \ |
291 | } |
292 | #define RRETURN(ra) \ |
293 | { \ |
294 | printf("match() returned %d from line %d ", ra, __LINE__); \ |
295 | return ra; \ |
296 | } |
297 | #else |
298 | #define RMATCH(ra,rb,rc,rd,re,rw) \ |
299 | rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1) |
300 | #define RRETURN(ra) return ra |
301 | #endif |
302 | |
303 | #else |
304 | |
305 | |
306 | /* These versions of the macros manage a private stack on the heap. Note that |
307 | the "rd" argument of RMATCH isn't actually used in this definition. It's the md |
308 | argument of match(), which never changes. */ |
309 | |
310 | #define REGISTER |
311 | |
312 | #define RMATCH(ra,rb,rc,rd,re,rw)\ |
313 | {\ |
314 | heapframe *newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\ |
315 | if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\ |
316 | frame->Xwhere = rw; \ |
317 | newframe->Xeptr = ra;\ |
318 | newframe->Xecode = rb;\ |
319 | newframe->Xmstart = mstart;\ |
320 | newframe->Xoffset_top = rc;\ |
321 | newframe->Xeptrb = re;\ |
322 | newframe->Xrdepth = frame->Xrdepth + 1;\ |
323 | newframe->Xprevframe = frame;\ |
324 | frame = newframe;\ |
325 | DPRINTF(("restarting from line %d\n", __LINE__));\ |
326 | goto HEAP_RECURSE;\ |
327 | L_##rw:\ |
328 | DPRINTF(("jumped back to line %d\n", __LINE__));\ |
329 | } |
330 | |
331 | #define RRETURN(ra)\ |
332 | {\ |
333 | heapframe *oldframe = frame;\ |
334 | frame = oldframe->Xprevframe;\ |
335 | (PUBL(stack_free))(oldframe);\ |
336 | if (frame != NULL)\ |
337 | {\ |
338 | rrc = ra;\ |
339 | goto HEAP_RETURN;\ |
340 | }\ |
341 | return ra;\ |
342 | } |
343 | |
344 | |
345 | /* Structure for remembering the local variables in a private frame */ |
346 | |
347 | typedef struct heapframe { |
348 | struct heapframe *Xprevframe; |
349 | |
350 | /* Function arguments that may change */ |
351 | |
352 | PCRE_PUCHAR Xeptr; |
353 | const pcre_uchar *Xecode; |
354 | PCRE_PUCHAR Xmstart; |
355 | int Xoffset_top; |
356 | eptrblock *Xeptrb; |
357 | unsigned int Xrdepth; |
358 | |
359 | /* Function local variables */ |
360 | |
361 | PCRE_PUCHAR Xcallpat; |
362 | #ifdef SUPPORT_UTF |
363 | PCRE_PUCHAR Xcharptr; |
364 | #endif |
365 | PCRE_PUCHAR Xdata; |
366 | PCRE_PUCHAR Xnext; |
367 | PCRE_PUCHAR Xpp; |
368 | PCRE_PUCHAR Xprev; |
369 | PCRE_PUCHAR Xsaved_eptr; |
370 | |
371 | recursion_info Xnew_recursive; |
372 | |
373 | BOOL Xcur_is_word; |
374 | BOOL Xcondition; |
375 | BOOL Xprev_is_word; |
376 | |
377 | #ifdef SUPPORT_UCP |
378 | int Xprop_type; |
379 | int Xprop_value; |
380 | int Xprop_fail_result; |
381 | int Xoclength; |
382 | pcre_uchar Xocchars[6]; |
383 | #endif |
384 | |
385 | int Xcodelink; |
386 | int Xctype; |
387 | unsigned int Xfc; |
388 | int Xfi; |
389 | int Xlength; |
390 | int Xmax; |
391 | int Xmin; |
392 | int Xnumber; |
393 | int Xoffset; |
394 | int Xop; |
395 | int Xsave_capture_last; |
396 | int Xsave_offset1, Xsave_offset2, Xsave_offset3; |
397 | int Xstacksave[REC_STACK_SAVE_MAX]; |
398 | |
399 | eptrblock Xnewptrb; |
400 | |
401 | /* Where to jump back to */ |
402 | |
403 | int Xwhere; |
404 | |
405 | } heapframe; |
406 | |
407 | #endif |
408 | |
409 | |
410 | /*************************************************************************** |
411 | ***************************************************************************/ |
412 | |
413 | |
414 | |
415 | /************************************************* |
416 | * Match from current position * |
417 | *************************************************/ |
418 | |
419 | /* This function is called recursively in many circumstances. Whenever it |
420 | returns a negative (error) response, the outer incarnation must also return the |
421 | same response. */ |
422 | |
423 | /* These macros pack up tests that are used for partial matching, and which |
424 | appear several times in the code. We set the "hit end" flag if the pointer is |
425 | at the end of the subject and also past the start of the subject (i.e. |
426 | something has been matched). For hard partial matching, we then return |
427 | immediately. The second one is used when we already know we are past the end of |
428 | the subject. */ |
429 | |
430 | #define CHECK_PARTIAL()\ |
431 | if (md->partial != 0 && eptr >= md->end_subject && \ |
432 | eptr > md->start_used_ptr) \ |
433 | { \ |
434 | md->hitend = TRUE; \ |
435 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \ |
436 | } |
437 | |
438 | #define SCHECK_PARTIAL()\ |
439 | if (md->partial != 0 && eptr > md->start_used_ptr) \ |
440 | { \ |
441 | md->hitend = TRUE; \ |
442 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \ |
443 | } |
444 | |
445 | |
446 | /* Performance note: It might be tempting to extract commonly used fields from |
447 | the md structure (e.g. utf, end_subject) into individual variables to improve |
448 | performance. Tests using gcc on a SPARC disproved this; in the first case, it |
449 | made performance worse. |
450 | |
451 | Arguments: |
452 | eptr pointer to current character in subject |
453 | ecode pointer to current position in compiled code |
454 | mstart pointer to the current match start position (can be modified |
455 | by encountering \K) |
456 | offset_top current top pointer |
457 | md pointer to "static" info for the match |
458 | eptrb pointer to chain of blocks containing eptr at start of |
459 | brackets - for testing for empty matches |
460 | rdepth the recursion depth |
461 | |
462 | Returns: MATCH_MATCH if matched ) these values are >= 0 |
463 | MATCH_NOMATCH if failed to match ) |
464 | a negative MATCH_xxx value for PRUNE, SKIP, etc |
465 | a negative PCRE_ERROR_xxx value if aborted by an error condition |
466 | (e.g. stopped by repeated call or recursion limit) |
467 | */ |
468 | |
469 | static int |
470 | match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode, |
471 | PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb, |
472 | unsigned int rdepth) |
473 | { |
474 | /* These variables do not need to be preserved over recursion in this function, |
475 | so they can be ordinary variables in all cases. Mark some of them with |
476 | "register" because they are used a lot in loops. */ |
477 | |
478 | register int rrc; /* Returns from recursive calls */ |
479 | register int i; /* Used for loops not involving calls to RMATCH() */ |
480 | register unsigned int c; /* Character values not kept over RMATCH() calls */ |
481 | register BOOL utf; /* Local copy of UTF flag for speed */ |
482 | |
483 | BOOL minimize, possessive; /* Quantifier options */ |
484 | BOOL caseless; |
485 | int condcode; |
486 | |
487 | /* When recursion is not being used, all "local" variables that have to be |
488 | preserved over calls to RMATCH() are part of a "frame" which is obtained from |
489 | heap storage. Set up the top-level frame here; others are obtained from the |
490 | heap whenever RMATCH() does a "recursion". See the macro definitions above. */ |
491 | |
492 | #ifdef NO_RECURSE |
493 | heapframe *frame = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe)); |
494 | if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY); |
495 | frame->Xprevframe = NULL; /* Marks the top level */ |
496 | |
497 | /* Copy in the original argument variables */ |
498 | |
499 | frame->Xeptr = eptr; |
500 | frame->Xecode = ecode; |
501 | frame->Xmstart = mstart; |
502 | frame->Xoffset_top = offset_top; |
503 | frame->Xeptrb = eptrb; |
504 | frame->Xrdepth = rdepth; |
505 | |
506 | /* This is where control jumps back to to effect "recursion" */ |
507 | |
508 | HEAP_RECURSE: |
509 | |
510 | /* Macros make the argument variables come from the current frame */ |
511 | |
512 | #define eptr frame->Xeptr |
513 | #define ecode frame->Xecode |
514 | #define mstart frame->Xmstart |
515 | #define offset_top frame->Xoffset_top |
516 | #define eptrb frame->Xeptrb |
517 | #define rdepth frame->Xrdepth |
518 | |
519 | /* Ditto for the local variables */ |
520 | |
521 | #ifdef SUPPORT_UTF |
522 | #define charptr frame->Xcharptr |
523 | #endif |
524 | #define callpat frame->Xcallpat |
525 | #define codelink frame->Xcodelink |
526 | #define data frame->Xdata |
527 | #define next frame->Xnext |
528 | #define pp frame->Xpp |
529 | #define prev frame->Xprev |
530 | #define saved_eptr frame->Xsaved_eptr |
531 | |
532 | #define new_recursive frame->Xnew_recursive |
533 | |
534 | #define cur_is_word frame->Xcur_is_word |
535 | #define condition frame->Xcondition |
536 | #define prev_is_word frame->Xprev_is_word |
537 | |
538 | #ifdef SUPPORT_UCP |
539 | #define prop_type frame->Xprop_type |
540 | #define prop_value frame->Xprop_value |
541 | #define prop_fail_result frame->Xprop_fail_result |
542 | #define oclength frame->Xoclength |
543 | #define occhars frame->Xocchars |
544 | #endif |
545 | |
546 | #define ctype frame->Xctype |
547 | #define fc frame->Xfc |
548 | #define fi frame->Xfi |
549 | #define length frame->Xlength |
550 | #define max frame->Xmax |
551 | #define min frame->Xmin |
552 | #define number frame->Xnumber |
553 | #define offset frame->Xoffset |
554 | #define op frame->Xop |
555 | #define save_capture_last frame->Xsave_capture_last |
556 | #define save_offset1 frame->Xsave_offset1 |
557 | #define save_offset2 frame->Xsave_offset2 |
558 | #define save_offset3 frame->Xsave_offset3 |
559 | #define stacksave frame->Xstacksave |
560 | |
561 | #define newptrb frame->Xnewptrb |
562 | |
563 | /* When recursion is being used, local variables are allocated on the stack and |
564 | get preserved during recursion in the normal way. In this environment, fi and |
565 | i, and fc and c, can be the same variables. */ |
566 | |
567 | #else /* NO_RECURSE not defined */ |
568 | #define fi i |
569 | #define fc c |
570 | |
571 | /* Many of the following variables are used only in small blocks of the code. |
572 | My normal style of coding would have declared them within each of those blocks. |
573 | However, in order to accommodate the version of this code that uses an external |
574 | "stack" implemented on the heap, it is easier to declare them all here, so the |
575 | declarations can be cut out in a block. The only declarations within blocks |
576 | below are for variables that do not have to be preserved over a recursive call |
577 | to RMATCH(). */ |
578 | |
579 | #ifdef SUPPORT_UTF |
580 | const pcre_uchar *charptr; |
581 | #endif |
582 | const pcre_uchar *callpat; |
583 | const pcre_uchar *data; |
584 | const pcre_uchar *next; |
585 | PCRE_PUCHAR pp; |
586 | const pcre_uchar *prev; |
587 | PCRE_PUCHAR saved_eptr; |
588 | |
589 | recursion_info new_recursive; |
590 | |
591 | BOOL cur_is_word; |
592 | BOOL condition; |
593 | BOOL prev_is_word; |
594 | |
595 | #ifdef SUPPORT_UCP |
596 | int prop_type; |
597 | int prop_value; |
598 | int prop_fail_result; |
599 | int oclength; |
600 | pcre_uchar occhars[6]; |
601 | #endif |
602 | |
603 | int codelink; |
604 | int ctype; |
605 | int length; |
606 | int max; |
607 | int min; |
608 | int number; |
609 | int offset; |
610 | int op; |
611 | int save_capture_last; |
612 | int save_offset1, save_offset2, save_offset3; |
613 | int stacksave[REC_STACK_SAVE_MAX]; |
614 | |
615 | eptrblock newptrb; |
616 | #endif /* NO_RECURSE */ |
617 | |
618 | /* To save space on the stack and in the heap frame, I have doubled up on some |
619 | of the local variables that are used only in localised parts of the code, but |
620 | still need to be preserved over recursive calls of match(). These macros define |
621 | the alternative names that are used. */ |
622 | |
623 | #define allow_zero cur_is_word |
624 | #define cbegroup condition |
625 | #define code_offset codelink |
626 | #define condassert condition |
627 | #define matched_once prev_is_word |
628 | #define foc number |
629 | |
630 | /* These statements are here to stop the compiler complaining about unitialized |
631 | variables. */ |
632 | |
633 | #ifdef SUPPORT_UCP |
634 | prop_value = 0; |
635 | prop_fail_result = 0; |
636 | #endif |
637 | |
638 | |
639 | /* This label is used for tail recursion, which is used in a few cases even |
640 | when NO_RECURSE is not defined, in order to reduce the amount of stack that is |
641 | used. Thanks to Ian Taylor for noticing this possibility and sending the |
642 | original patch. */ |
643 | |
644 | TAIL_RECURSE: |
645 | |
646 | /* OK, now we can get on with the real code of the function. Recursive calls |
647 | are specified by the macro RMATCH and RRETURN is used to return. When |
648 | NO_RECURSE is *not* defined, these just turn into a recursive call to match() |
649 | and a "return", respectively (possibly with some debugging if PCRE_DEBUG is |
650 | defined). However, RMATCH isn't like a function call because it's quite a |
651 | complicated macro. It has to be used in one particular way. This shouldn't, |
652 | however, impact performance when true recursion is being used. */ |
653 | |
654 | #ifdef SUPPORT_UTF |
655 | utf = md->utf; /* Local copy of the flag */ |
656 | #else |
657 | utf = FALSE; |
658 | #endif |
659 | |
660 | /* First check that we haven't called match() too many times, or that we |
661 | haven't exceeded the recursive call limit. */ |
662 | |
663 | if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT); |
664 | if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT); |
665 | |
666 | /* At the start of a group with an unlimited repeat that may match an empty |
667 | string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is |
668 | done this way to save having to use another function argument, which would take |
669 | up space on the stack. See also MATCH_CONDASSERT below. |
670 | |
671 | When MATCH_CBEGROUP is set, add the current subject pointer to the chain of |
672 | such remembered pointers, to be checked when we hit the closing ket, in order |
673 | to break infinite loops that match no characters. When match() is called in |
674 | other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must |
675 | NOT be used with tail recursion, because the memory block that is used is on |
676 | the stack, so a new one may be required for each match(). */ |
677 | |
678 | if (md->match_function_type == MATCH_CBEGROUP) |
679 | { |
680 | newptrb.epb_saved_eptr = eptr; |
681 | newptrb.epb_prev = eptrb; |
682 | eptrb = &newptrb; |
683 | md->match_function_type = 0; |
684 | } |
685 | |
686 | /* Now start processing the opcodes. */ |
687 | |
688 | for (;;) |
689 | { |
690 | minimize = possessive = FALSE; |
691 | op = *ecode; |
692 | |
693 | switch(op) |
694 | { |
695 | case OP_MARK: |
696 | md->nomatch_mark = ecode + 2; |
697 | md->mark = NULL; /* In case previously set by assertion */ |
698 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
699 | eptrb, RM55); |
700 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
701 | md->mark == NULL) md->mark = ecode + 2; |
702 | |
703 | /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an |
704 | argument, and we must check whether that argument matches this MARK's |
705 | argument. It is passed back in md->start_match_ptr (an overloading of that |
706 | variable). If it does match, we reset that variable to the current subject |
707 | position and return MATCH_SKIP. Otherwise, pass back the return code |
708 | unaltered. */ |
709 | |
710 | else if (rrc == MATCH_SKIP_ARG && |
711 | STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0) |
712 | { |
713 | md->start_match_ptr = eptr; |
714 | RRETURN(MATCH_SKIP); |
715 | } |
716 | RRETURN(rrc); |
717 | |
718 | case OP_FAIL: |
719 | RRETURN(MATCH_NOMATCH); |
720 | |
721 | /* COMMIT overrides PRUNE, SKIP, and THEN */ |
722 | |
723 | case OP_COMMIT: |
724 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
725 | eptrb, RM52); |
726 | if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && |
727 | rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG && |
728 | rrc != MATCH_THEN) |
729 | RRETURN(rrc); |
730 | RRETURN(MATCH_COMMIT); |
731 | |
732 | /* PRUNE overrides THEN */ |
733 | |
734 | case OP_PRUNE: |
735 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
736 | eptrb, RM51); |
737 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
738 | RRETURN(MATCH_PRUNE); |
739 | |
740 | case OP_PRUNE_ARG: |
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, RM56); |
745 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
746 | md->mark == NULL) md->mark = ecode + 2; |
747 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
748 | RRETURN(MATCH_PRUNE); |
749 | |
750 | /* SKIP overrides PRUNE and THEN */ |
751 | |
752 | case OP_SKIP: |
753 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
754 | eptrb, RM53); |
755 | if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN) |
756 | RRETURN(rrc); |
757 | md->start_match_ptr = eptr; /* Pass back current position */ |
758 | RRETURN(MATCH_SKIP); |
759 | |
760 | /* Note that, for Perl compatibility, SKIP with an argument does NOT set |
761 | nomatch_mark. There is a flag that disables this opcode when re-matching a |
762 | pattern that ended with a SKIP for which there was not a matching MARK. */ |
763 | |
764 | case OP_SKIP_ARG: |
765 | if (md->ignore_skip_arg) |
766 | { |
767 | ecode += PRIV(OP_lengths)[*ecode] + ecode[1]; |
768 | break; |
769 | } |
770 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
771 | eptrb, RM57); |
772 | if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN) |
773 | RRETURN(rrc); |
774 | |
775 | /* Pass back the current skip name by overloading md->start_match_ptr and |
776 | returning the special MATCH_SKIP_ARG return code. This will either be |
777 | caught by a matching MARK, or get to the top, where it causes a rematch |
778 | with the md->ignore_skip_arg flag set. */ |
779 | |
780 | md->start_match_ptr = ecode + 2; |
781 | RRETURN(MATCH_SKIP_ARG); |
782 | |
783 | /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that |
784 | the branch in which it occurs can be determined. Overload the start of |
785 | match pointer to do this. */ |
786 | |
787 | case OP_THEN: |
788 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
789 | eptrb, RM54); |
790 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
791 | md->start_match_ptr = ecode; |
792 | RRETURN(MATCH_THEN); |
793 | |
794 | case OP_THEN_ARG: |
795 | md->nomatch_mark = ecode + 2; |
796 | md->mark = NULL; /* In case previously set by assertion */ |
797 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, |
798 | md, eptrb, RM58); |
799 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
800 | md->mark == NULL) md->mark = ecode + 2; |
801 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
802 | md->start_match_ptr = ecode; |
803 | RRETURN(MATCH_THEN); |
804 | |
805 | /* Handle an atomic group that does not contain any capturing parentheses. |
806 | This can be handled like an assertion. Prior to 8.13, all atomic groups |
807 | were handled this way. In 8.13, the code was changed as below for ONCE, so |
808 | that backups pass through the group and thereby reset captured values. |
809 | However, this uses a lot more stack, so in 8.20, atomic groups that do not |
810 | contain any captures generate OP_ONCE_NC, which can be handled in the old, |
811 | less stack intensive way. |
812 | |
813 | Check the alternative branches in turn - the matching won't pass the KET |
814 | for this kind of subpattern. If any one branch matches, we carry on as at |
815 | the end of a normal bracket, leaving the subject pointer, but resetting |
816 | the start-of-match value in case it was changed by \K. */ |
817 | |
818 | case OP_ONCE_NC: |
819 | prev = ecode; |
820 | saved_eptr = eptr; |
821 | do |
822 | { |
823 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64); |
824 | if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */ |
825 | { |
826 | mstart = md->start_match_ptr; |
827 | break; |
828 | } |
829 | if (rrc == MATCH_THEN) |
830 | { |
831 | next = ecode + GET(ecode,1); |
832 | if (md->start_match_ptr < next && |
833 | (*ecode == OP_ALT || *next == OP_ALT)) |
834 | rrc = MATCH_NOMATCH; |
835 | } |
836 | |
837 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
838 | ecode += GET(ecode,1); |
839 | } |
840 | while (*ecode == OP_ALT); |
841 | |
842 | /* If hit the end of the group (which could be repeated), fail */ |
843 | |
844 | if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH); |
845 | |
846 | /* Continue as from after the group, updating the offsets high water |
847 | mark, since extracts may have been taken. */ |
848 | |
849 | do ecode += GET(ecode, 1); while (*ecode == OP_ALT); |
850 | |
851 | offset_top = md->end_offset_top; |
852 | eptr = md->end_match_ptr; |
853 | |
854 | /* For a non-repeating ket, just continue at this level. This also |
855 | happens for a repeating ket if no characters were matched in the group. |
856 | This is the forcible breaking of infinite loops as implemented in Perl |
857 | 5.005. */ |
858 | |
859 | if (*ecode == OP_KET || eptr == saved_eptr) |
860 | { |
861 | ecode += 1+LINK_SIZE; |
862 | break; |
863 | } |
864 | |
865 | /* The repeating kets try the rest of the pattern or restart from the |
866 | preceding bracket, in the appropriate order. The second "call" of match() |
867 | uses tail recursion, to avoid using another stack frame. */ |
868 | |
869 | if (*ecode == OP_KETRMIN) |
870 | { |
871 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65); |
872 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
873 | ecode = prev; |
874 | goto TAIL_RECURSE; |
875 | } |
876 | else /* OP_KETRMAX */ |
877 | { |
878 | md->match_function_type = MATCH_CBEGROUP; |
879 | RMATCH(eptr, prev, offset_top, md, eptrb, RM66); |
880 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
881 | ecode += 1 + LINK_SIZE; |
882 | goto TAIL_RECURSE; |
883 | } |
884 | /* Control never gets here */ |
885 | |
886 | /* Handle a capturing bracket, other than those that are possessive with an |
887 | unlimited repeat. If there is space in the offset vector, save the current |
888 | subject position in the working slot at the top of the vector. We mustn't |
889 | change the current values of the data slot, because they may be set from a |
890 | previous iteration of this group, and be referred to by a reference inside |
891 | the group. A failure to match might occur after the group has succeeded, |
892 | if something later on doesn't match. For this reason, we need to restore |
893 | the working value and also the values of the final offsets, in case they |
894 | were set by a previous iteration of the same bracket. |
895 | |
896 | If there isn't enough space in the offset vector, treat this as if it were |
897 | a non-capturing bracket. Don't worry about setting the flag for the error |
898 | case here; that is handled in the code for KET. */ |
899 | |
900 | case OP_CBRA: |
901 | case OP_SCBRA: |
902 | number = GET2(ecode, 1+LINK_SIZE); |
903 | offset = number << 1; |
904 | |
905 | #ifdef PCRE_DEBUG |
906 | printf("start bracket %d\n", number); |
907 | printf("subject="); |
908 | pchars(eptr, 16, TRUE, md); |
909 | printf("\n"); |
910 | #endif |
911 | |
912 | if (offset < md->offset_max) |
913 | { |
914 | save_offset1 = md->offset_vector[offset]; |
915 | save_offset2 = md->offset_vector[offset+1]; |
916 | save_offset3 = md->offset_vector[md->offset_end - number]; |
917 | save_capture_last = md->capture_last; |
918 | |
919 | DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); |
920 | md->offset_vector[md->offset_end - number] = |
921 | (int)(eptr - md->start_subject); |
922 | |
923 | for (;;) |
924 | { |
925 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
926 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
927 | eptrb, RM1); |
928 | if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */ |
929 | |
930 | /* If we backed up to a THEN, check whether it is within the current |
931 | branch by comparing the address of the THEN that is passed back with |
932 | the end of the branch. If it is within the current branch, and the |
933 | branch is one of two or more alternatives (it either starts or ends |
934 | with OP_ALT), we have reached the limit of THEN's action, so convert |
935 | the return code to NOMATCH, which will cause normal backtracking to |
936 | happen from now on. Otherwise, THEN is passed back to an outer |
937 | alternative. This implements Perl's treatment of parenthesized groups, |
938 | where a group not containing | does not affect the current alternative, |
939 | that is, (X) is NOT the same as (X|(*F)). */ |
940 | |
941 | if (rrc == MATCH_THEN) |
942 | { |
943 | next = ecode + GET(ecode,1); |
944 | if (md->start_match_ptr < next && |
945 | (*ecode == OP_ALT || *next == OP_ALT)) |
946 | rrc = MATCH_NOMATCH; |
947 | } |
948 | |
949 | /* Anything other than NOMATCH is passed back. */ |
950 | |
951 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
952 | md->capture_last = save_capture_last; |
953 | ecode += GET(ecode, 1); |
954 | if (*ecode != OP_ALT) break; |
955 | } |
956 | |
957 | DPRINTF(("bracket %d failed\n", number)); |
958 | md->offset_vector[offset] = save_offset1; |
959 | md->offset_vector[offset+1] = save_offset2; |
960 | md->offset_vector[md->offset_end - number] = save_offset3; |
961 | |
962 | /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */ |
963 | |
964 | RRETURN(rrc); |
965 | } |
966 | |
967 | /* FALL THROUGH ... Insufficient room for saving captured contents. Treat |
968 | as a non-capturing bracket. */ |
969 | |
970 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
971 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
972 | |
973 | DPRINTF(("insufficient capture room: treat as non-capturing\n")); |
974 | |
975 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
976 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
977 | |
978 | /* Non-capturing or atomic group, except for possessive with unlimited |
979 | repeat and ONCE group with no captures. Loop for all the alternatives. |
980 | |
981 | When we get to the final alternative within the brackets, we used to return |
982 | the result of a recursive call to match() whatever happened so it was |
983 | possible to reduce stack usage by turning this into a tail recursion, |
984 | except in the case of a possibly empty group. However, now that there is |
985 | the possiblity of (*THEN) occurring in the final alternative, this |
986 | optimization is no longer always possible. |
987 | |
988 | We can optimize if we know there are no (*THEN)s in the pattern; at present |
989 | this is the best that can be done. |
990 | |
991 | MATCH_ONCE is returned when the end of an atomic group is successfully |
992 | reached, but subsequent matching fails. It passes back up the tree (causing |
993 | captured values to be reset) until the original atomic group level is |
994 | reached. This is tested by comparing md->once_target with the start of the |
995 | group. At this point, the return is converted into MATCH_NOMATCH so that |
996 | previous backup points can be taken. */ |
997 | |
998 | case OP_ONCE: |
999 | case OP_BRA: |
1000 | case OP_SBRA: |
1001 | DPRINTF(("start non-capturing bracket\n")); |
1002 | |
1003 | for (;;) |
1004 | { |
1005 | if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP; |
1006 | |
1007 | /* If this is not a possibly empty group, and there are no (*THEN)s in |
1008 | the pattern, and this is the final alternative, optimize as described |
1009 | above. */ |
1010 | |
1011 | else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT) |
1012 | { |
1013 | ecode += PRIV(OP_lengths)[*ecode]; |
1014 | goto TAIL_RECURSE; |
1015 | } |
1016 | |
1017 | /* In all other cases, we have to make another call to match(). */ |
1018 | |
1019 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb, |
1020 | RM2); |
1021 | |
1022 | /* See comment in the code for capturing groups above about handling |
1023 | THEN. */ |
1024 | |
1025 | if (rrc == MATCH_THEN) |
1026 | { |
1027 | next = ecode + GET(ecode,1); |
1028 | if (md->start_match_ptr < next && |
1029 | (*ecode == OP_ALT || *next == OP_ALT)) |
1030 | rrc = MATCH_NOMATCH; |
1031 | } |
1032 | |
1033 | if (rrc != MATCH_NOMATCH) |
1034 | { |
1035 | if (rrc == MATCH_ONCE) |
1036 | { |
1037 | const pcre_uchar *scode = ecode; |
1038 | if (*scode != OP_ONCE) /* If not at start, find it */ |
1039 | { |
1040 | while (*scode == OP_ALT) scode += GET(scode, 1); |
1041 | scode -= GET(scode, 1); |
1042 | } |
1043 | if (md->once_target == scode) rrc = MATCH_NOMATCH; |
1044 | } |
1045 | RRETURN(rrc); |
1046 | } |
1047 | ecode += GET(ecode, 1); |
1048 | if (*ecode != OP_ALT) break; |
1049 | } |
1050 | |
1051 | RRETURN(MATCH_NOMATCH); |
1052 | |
1053 | /* Handle possessive capturing brackets with an unlimited repeat. We come |
1054 | here from BRAZERO with allow_zero set TRUE. The offset_vector values are |
1055 | handled similarly to the normal case above. However, the matching is |
1056 | different. The end of these brackets will always be OP_KETRPOS, which |
1057 | returns MATCH_KETRPOS without going further in the pattern. By this means |
1058 | we can handle the group by iteration rather than recursion, thereby |
1059 | reducing the amount of stack needed. */ |
1060 | |
1061 | case OP_CBRAPOS: |
1062 | case OP_SCBRAPOS: |
1063 | allow_zero = FALSE; |
1064 | |
1065 | POSSESSIVE_CAPTURE: |
1066 | number = GET2(ecode, 1+LINK_SIZE); |
1067 | offset = number << 1; |
1068 | |
1069 | #ifdef PCRE_DEBUG |
1070 | printf("start possessive bracket %d\n", number); |
1071 | printf("subject="); |
1072 | pchars(eptr, 16, TRUE, md); |
1073 | printf("\n"); |
1074 | #endif |
1075 | |
1076 | if (offset < md->offset_max) |
1077 | { |
1078 | matched_once = FALSE; |
1079 | code_offset = (int)(ecode - md->start_code); |
1080 | |
1081 | save_offset1 = md->offset_vector[offset]; |
1082 | save_offset2 = md->offset_vector[offset+1]; |
1083 | save_offset3 = md->offset_vector[md->offset_end - number]; |
1084 | save_capture_last = md->capture_last; |
1085 | |
1086 | DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); |
1087 | |
1088 | /* Each time round the loop, save the current subject position for use |
1089 | when the group matches. For MATCH_MATCH, the group has matched, so we |
1090 | restart it with a new subject starting position, remembering that we had |
1091 | at least one match. For MATCH_NOMATCH, carry on with the alternatives, as |
1092 | usual. If we haven't matched any alternatives in any iteration, check to |
1093 | see if a previous iteration matched. If so, the group has matched; |
1094 | continue from afterwards. Otherwise it has failed; restore the previous |
1095 | capture values before returning NOMATCH. */ |
1096 | |
1097 | for (;;) |
1098 | { |
1099 | md->offset_vector[md->offset_end - number] = |
1100 | (int)(eptr - md->start_subject); |
1101 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
1102 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
1103 | eptrb, RM63); |
1104 | if (rrc == MATCH_KETRPOS) |
1105 | { |
1106 | offset_top = md->end_offset_top; |
1107 | eptr = md->end_match_ptr; |
1108 | ecode = md->start_code + code_offset; |
1109 | save_capture_last = md->capture_last; |
1110 | matched_once = TRUE; |
1111 | continue; |
1112 | } |
1113 | |
1114 | /* See comment in the code for capturing groups above about handling |
1115 | THEN. */ |
1116 | |
1117 | if (rrc == MATCH_THEN) |
1118 | { |
1119 | next = ecode + GET(ecode,1); |
1120 | if (md->start_match_ptr < next && |
1121 | (*ecode == OP_ALT || *next == OP_ALT)) |
1122 | rrc = MATCH_NOMATCH; |
1123 | } |
1124 | |
1125 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1126 | md->capture_last = save_capture_last; |
1127 | ecode += GET(ecode, 1); |
1128 | if (*ecode != OP_ALT) break; |
1129 | } |
1130 | |
1131 | if (!matched_once) |
1132 | { |
1133 | md->offset_vector[offset] = save_offset1; |
1134 | md->offset_vector[offset+1] = save_offset2; |
1135 | md->offset_vector[md->offset_end - number] = save_offset3; |
1136 | } |
1137 | |
1138 | if (allow_zero || matched_once) |
1139 | { |
1140 | ecode += 1 + LINK_SIZE; |
1141 | break; |
1142 | } |
1143 | |
1144 | RRETURN(MATCH_NOMATCH); |
1145 | } |
1146 | |
1147 | /* FALL THROUGH ... Insufficient room for saving captured contents. Treat |
1148 | as a non-capturing bracket. */ |
1149 | |
1150 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1151 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1152 | |
1153 | DPRINTF(("insufficient capture room: treat as non-capturing\n")); |
1154 | |
1155 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1156 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1157 | |
1158 | /* Non-capturing possessive bracket with unlimited repeat. We come here |
1159 | from BRAZERO with allow_zero = TRUE. The code is similar to the above, |
1160 | without the capturing complication. It is written out separately for speed |
1161 | and cleanliness. */ |
1162 | |
1163 | case OP_BRAPOS: |
1164 | case OP_SBRAPOS: |
1165 | allow_zero = FALSE; |
1166 | |
1167 | POSSESSIVE_NON_CAPTURE: |
1168 | matched_once = FALSE; |
1169 | code_offset = (int)(ecode - md->start_code); |
1170 | |
1171 | for (;;) |
1172 | { |
1173 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
1174 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
1175 | eptrb, RM48); |
1176 | if (rrc == MATCH_KETRPOS) |
1177 | { |
1178 | offset_top = md->end_offset_top; |
1179 | eptr = md->end_match_ptr; |
1180 | ecode = md->start_code + code_offset; |
1181 | matched_once = TRUE; |
1182 | continue; |
1183 | } |
1184 | |
1185 | /* See comment in the code for capturing groups above about handling |
1186 | THEN. */ |
1187 | |
1188 | if (rrc == MATCH_THEN) |
1189 | { |
1190 | next = ecode + GET(ecode,1); |
1191 | if (md->start_match_ptr < next && |
1192 | (*ecode == OP_ALT || *next == OP_ALT)) |
1193 | rrc = MATCH_NOMATCH; |
1194 | } |
1195 | |
1196 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1197 | ecode += GET(ecode, 1); |
1198 | if (*ecode != OP_ALT) break; |
1199 | } |
1200 | |
1201 | if (matched_once || allow_zero) |
1202 | { |
1203 | ecode += 1 + LINK_SIZE; |
1204 | break; |
1205 | } |
1206 | RRETURN(MATCH_NOMATCH); |
1207 | |
1208 | /* Control never reaches here. */ |
1209 | |
1210 | /* Conditional group: compilation checked that there are no more than |
1211 | two branches. If the condition is false, skipping the first branch takes us |
1212 | past the end if there is only one branch, but that's OK because that is |
1213 | exactly what going to the ket would do. */ |
1214 | |
1215 | case OP_COND: |
1216 | case OP_SCOND: |
1217 | codelink = GET(ecode, 1); |
1218 | |
1219 | /* Because of the way auto-callout works during compile, a callout item is |
1220 | inserted between OP_COND and an assertion condition. */ |
1221 | |
1222 | if (ecode[LINK_SIZE+1] == OP_CALLOUT) |
1223 | { |
1224 | if (PUBL(callout) != NULL) |
1225 | { |
1226 | PUBL(callout_block) cb; |
1227 | cb.version = 2; /* Version 1 of the callout block */ |
1228 | cb.callout_number = ecode[LINK_SIZE+2]; |
1229 | cb.offset_vector = md->offset_vector; |
1230 | #ifdef COMPILE_PCRE8 |
1231 | cb.subject = (PCRE_SPTR)md->start_subject; |
1232 | #else |
1233 | cb.subject = (PCRE_SPTR16)md->start_subject; |
1234 | #endif |
1235 | cb.subject_length = (int)(md->end_subject - md->start_subject); |
1236 | cb.start_match = (int)(mstart - md->start_subject); |
1237 | cb.current_position = (int)(eptr - md->start_subject); |
1238 | cb.pattern_position = GET(ecode, LINK_SIZE + 3); |
1239 | cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE); |
1240 | cb.capture_top = offset_top/2; |
1241 | cb.capture_last = md->capture_last; |
1242 | cb.callout_data = md->callout_data; |
1243 | cb.mark = md->nomatch_mark; |
1244 | if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH); |
1245 | if (rrc < 0) RRETURN(rrc); |
1246 | } |
1247 | ecode += PRIV(OP_lengths)[OP_CALLOUT]; |
1248 | } |
1249 | |
1250 | condcode = ecode[LINK_SIZE+1]; |
1251 | |
1252 | /* Now see what the actual condition is */ |
1253 | |
1254 | if (condcode == OP_RREF || condcode == OP_NRREF) /* Recursion test */ |
1255 | { |
1256 | if (md->recursive == NULL) /* Not recursing => FALSE */ |
1257 | { |
1258 | condition = FALSE; |
1259 | ecode += GET(ecode, 1); |
1260 | } |
1261 | else |
1262 | { |
1263 | int recno = GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/ |
1264 | condition = (recno == RREF_ANY || recno == md->recursive->group_num); |
1265 | |
1266 | /* If the test is for recursion into a specific subpattern, and it is |
1267 | false, but the test was set up by name, scan the table to see if the |
1268 | name refers to any other numbers, and test them. The condition is true |
1269 | if any one is set. */ |
1270 | |
1271 | if (!condition && condcode == OP_NRREF) |
1272 | { |
1273 | pcre_uchar *slotA = md->name_table; |
1274 | for (i = 0; i < md->name_count; i++) |
1275 | { |
1276 | if (GET2(slotA, 0) == recno) break; |
1277 | slotA += md->name_entry_size; |
1278 | } |
1279 | |
1280 | /* Found a name for the number - there can be only one; duplicate |
1281 | names for different numbers are allowed, but not vice versa. First |
1282 | scan down for duplicates. */ |
1283 | |
1284 | if (i < md->name_count) |
1285 | { |
1286 | pcre_uchar *slotB = slotA; |
1287 | while (slotB > md->name_table) |
1288 | { |
1289 | slotB -= md->name_entry_size; |
1290 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1291 | { |
1292 | condition = GET2(slotB, 0) == md->recursive->group_num; |
1293 | if (condition) break; |
1294 | } |
1295 | else break; |
1296 | } |
1297 | |
1298 | /* Scan up for duplicates */ |
1299 | |
1300 | if (!condition) |
1301 | { |
1302 | slotB = slotA; |
1303 | for (i++; i < md->name_count; i++) |
1304 | { |
1305 | slotB += md->name_entry_size; |
1306 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1307 | { |
1308 | condition = GET2(slotB, 0) == md->recursive->group_num; |
1309 | if (condition) break; |
1310 | } |
1311 | else break; |
1312 | } |
1313 | } |
1314 | } |
1315 | } |
1316 | |
1317 | /* Chose branch according to the condition */ |
1318 | |
1319 | ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1); |
1320 | } |
1321 | } |
1322 | |
1323 | else if (condcode == OP_CREF || condcode == OP_NCREF) /* Group used test */ |
1324 | { |
1325 | offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */ |
1326 | condition = offset < offset_top && md->offset_vector[offset] >= 0; |
1327 | |
1328 | /* If the numbered capture is unset, but the reference was by name, |
1329 | scan the table to see if the name refers to any other numbers, and test |
1330 | them. The condition is true if any one is set. This is tediously similar |
1331 | to the code above, but not close enough to try to amalgamate. */ |
1332 | |
1333 | if (!condition && condcode == OP_NCREF) |
1334 | { |
1335 | int refno = offset >> 1; |
1336 | pcre_uchar *slotA = md->name_table; |
1337 | |
1338 | for (i = 0; i < md->name_count; i++) |
1339 | { |
1340 | if (GET2(slotA, 0) == refno) break; |
1341 | slotA += md->name_entry_size; |
1342 | } |
1343 | |
1344 | /* Found a name for the number - there can be only one; duplicate names |
1345 | for different numbers are allowed, but not vice versa. First scan down |
1346 | for duplicates. */ |
1347 | |
1348 | if (i < md->name_count) |
1349 | { |
1350 | pcre_uchar *slotB = slotA; |
1351 | while (slotB > md->name_table) |
1352 | { |
1353 | slotB -= md->name_entry_size; |
1354 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1355 | { |
1356 | offset = GET2(slotB, 0) << 1; |
1357 | condition = offset < offset_top && |
1358 | md->offset_vector[offset] >= 0; |
1359 | if (condition) break; |
1360 | } |
1361 | else break; |
1362 | } |
1363 | |
1364 | /* Scan up for duplicates */ |
1365 | |
1366 | if (!condition) |
1367 | { |
1368 | slotB = slotA; |
1369 | for (i++; i < md->name_count; i++) |
1370 | { |
1371 | slotB += md->name_entry_size; |
1372 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1373 | { |
1374 | offset = GET2(slotB, 0) << 1; |
1375 | condition = offset < offset_top && |
1376 | md->offset_vector[offset] >= 0; |
1377 | if (condition) break; |
1378 | } |
1379 | else break; |
1380 | } |
1381 | } |
1382 | } |
1383 | } |
1384 | |
1385 | /* Chose branch according to the condition */ |
1386 | |
1387 | ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1); |
1388 | } |
1389 | |
1390 | else if (condcode == OP_DEF) /* DEFINE - always false */ |
1391 | { |
1392 | condition = FALSE; |
1393 | ecode += GET(ecode, 1); |
1394 | } |
1395 | |
1396 | /* The condition is an assertion. Call match() to evaluate it - setting |
1397 | md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of |
1398 | an assertion. */ |
1399 | |
1400 | else |
1401 | { |
1402 | md->match_function_type = MATCH_CONDASSERT; |
1403 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3); |
1404 | if (rrc == MATCH_MATCH) |
1405 | { |
1406 | if (md->end_offset_top > offset_top) |
1407 | offset_top = md->end_offset_top; /* Captures may have happened */ |
1408 | condition = TRUE; |
1409 | ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2); |
1410 | while (*ecode == OP_ALT) ecode += GET(ecode, 1); |
1411 | } |
1412 | |
1413 | /* PCRE doesn't allow the effect of (*THEN) to escape beyond an |
1414 | assertion; it is therefore treated as NOMATCH. */ |
1415 | |
1416 | else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) |
1417 | { |
1418 | RRETURN(rrc); /* Need braces because of following else */ |
1419 | } |
1420 | else |
1421 | { |
1422 | condition = FALSE; |
1423 | ecode += codelink; |
1424 | } |
1425 | } |
1426 | |
1427 | /* We are now at the branch that is to be obeyed. As there is only one, can |
1428 | use tail recursion to avoid using another stack frame, except when there is |
1429 | unlimited repeat of a possibly empty group. In the latter case, a recursive |
1430 | call to match() is always required, unless the second alternative doesn't |
1431 | exist, in which case we can just plough on. Note that, for compatibility |
1432 | with Perl, the | in a conditional group is NOT treated as creating two |
1433 | alternatives. If a THEN is encountered in the branch, it propagates out to |
1434 | the enclosing alternative (unless nested in a deeper set of alternatives, |
1435 | of course). */ |
1436 | |
1437 | if (condition || *ecode == OP_ALT) |
1438 | { |
1439 | if (op != OP_SCOND) |
1440 | { |
1441 | ecode += 1 + LINK_SIZE; |
1442 | goto TAIL_RECURSE; |
1443 | } |
1444 | |
1445 | md->match_function_type = MATCH_CBEGROUP; |
1446 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49); |
1447 | RRETURN(rrc); |
1448 | } |
1449 | |
1450 | /* Condition false & no alternative; continue after the group. */ |
1451 | |
1452 | else |
1453 | { |
1454 | ecode += 1 + LINK_SIZE; |
1455 | } |
1456 | break; |
1457 | |
1458 | |
1459 | /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes, |
1460 | to close any currently open capturing brackets. */ |
1461 | |
1462 | case OP_CLOSE: |
1463 | number = GET2(ecode, 1); |
1464 | offset = number << 1; |
1465 | |
1466 | #ifdef PCRE_DEBUG |
1467 | printf("end bracket %d at *ACCEPT", number); |
1468 | printf("\n"); |
1469 | #endif |
1470 | |
1471 | md->capture_last = number; |
1472 | if (offset >= md->offset_max) md->offset_overflow = TRUE; else |
1473 | { |
1474 | md->offset_vector[offset] = |
1475 | md->offset_vector[md->offset_end - number]; |
1476 | md->offset_vector[offset+1] = (int)(eptr - md->start_subject); |
1477 | if (offset_top <= offset) offset_top = offset + 2; |
1478 | } |
1479 | ecode += 1 + IMM2_SIZE; |
1480 | break; |
1481 | |
1482 | |
1483 | /* End of the pattern, either real or forced. */ |
1484 | |
1485 | case OP_END: |
1486 | case OP_ACCEPT: |
1487 | case OP_ASSERT_ACCEPT: |
1488 | |
1489 | /* If we have matched an empty string, fail if not in an assertion and not |
1490 | in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART |
1491 | is set and we have matched at the start of the subject. In both cases, |
1492 | backtracking will then try other alternatives, if any. */ |
1493 | |
1494 | if (eptr == mstart && op != OP_ASSERT_ACCEPT && |
1495 | md->recursive == NULL && |
1496 | (md->notempty || |
1497 | (md->notempty_atstart && |
1498 | mstart == md->start_subject + md->start_offset))) |
1499 | RRETURN(MATCH_NOMATCH); |
1500 | |
1501 | /* Otherwise, we have a match. */ |
1502 | |
1503 | md->end_match_ptr = eptr; /* Record where we ended */ |
1504 | md->end_offset_top = offset_top; /* and how many extracts were taken */ |
1505 | md->start_match_ptr = mstart; /* and the start (\K can modify) */ |
1506 | |
1507 | /* For some reason, the macros don't work properly if an expression is |
1508 | given as the argument to RRETURN when the heap is in use. */ |
1509 | |
1510 | rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT; |
1511 | RRETURN(rrc); |
1512 | |
1513 | /* Assertion brackets. Check the alternative branches in turn - the |
1514 | matching won't pass the KET for an assertion. If any one branch matches, |
1515 | the assertion is true. Lookbehind assertions have an OP_REVERSE item at the |
1516 | start of each branch to move the current point backwards, so the code at |
1517 | this level is identical to the lookahead case. When the assertion is part |
1518 | of a condition, we want to return immediately afterwards. The caller of |
1519 | this incarnation of the match() function will have set MATCH_CONDASSERT in |
1520 | md->match_function type, and one of these opcodes will be the first opcode |
1521 | that is processed. We use a local variable that is preserved over calls to |
1522 | match() to remember this case. */ |
1523 | |
1524 | case OP_ASSERT: |
1525 | case OP_ASSERTBACK: |
1526 | if (md->match_function_type == MATCH_CONDASSERT) |
1527 | { |
1528 | condassert = TRUE; |
1529 | md->match_function_type = 0; |
1530 | } |
1531 | else condassert = FALSE; |
1532 | |
1533 | do |
1534 | { |
1535 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4); |
1536 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) |
1537 | { |
1538 | mstart = md->start_match_ptr; /* In case \K reset it */ |
1539 | break; |
1540 | } |
1541 | |
1542 | /* PCRE does not allow THEN to escape beyond an assertion; it is treated |
1543 | as NOMATCH. */ |
1544 | |
1545 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
1546 | ecode += GET(ecode, 1); |
1547 | } |
1548 | while (*ecode == OP_ALT); |
1549 | |
1550 | if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH); |
1551 | |
1552 | /* If checking an assertion for a condition, return MATCH_MATCH. */ |
1553 | |
1554 | if (condassert) RRETURN(MATCH_MATCH); |
1555 | |
1556 | /* Continue from after the assertion, updating the offsets high water |
1557 | mark, since extracts may have been taken during the assertion. */ |
1558 | |
1559 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1560 | ecode += 1 + LINK_SIZE; |
1561 | offset_top = md->end_offset_top; |
1562 | continue; |
1563 | |
1564 | /* Negative assertion: all branches must fail to match. Encountering SKIP, |
1565 | PRUNE, or COMMIT means we must assume failure without checking subsequent |
1566 | branches. */ |
1567 | |
1568 | case OP_ASSERT_NOT: |
1569 | case OP_ASSERTBACK_NOT: |
1570 | if (md->match_function_type == MATCH_CONDASSERT) |
1571 | { |
1572 | condassert = TRUE; |
1573 | md->match_function_type = 0; |
1574 | } |
1575 | else condassert = FALSE; |
1576 | |
1577 | do |
1578 | { |
1579 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5); |
1580 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH); |
1581 | if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT) |
1582 | { |
1583 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1584 | break; |
1585 | } |
1586 | |
1587 | /* PCRE does not allow THEN to escape beyond an assertion; it is treated |
1588 | as NOMATCH. */ |
1589 | |
1590 | if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc); |
1591 | ecode += GET(ecode,1); |
1592 | } |
1593 | while (*ecode == OP_ALT); |
1594 | |
1595 | if (condassert) RRETURN(MATCH_MATCH); /* Condition assertion */ |
1596 | |
1597 | ecode += 1 + LINK_SIZE; |
1598 | continue; |
1599 | |
1600 | /* Move the subject pointer back. This occurs only at the start of |
1601 | each branch of a lookbehind assertion. If we are too close to the start to |
1602 | move back, this match function fails. When working with UTF-8 we move |
1603 | back a number of characters, not bytes. */ |
1604 | |
1605 | case OP_REVERSE: |
1606 | #ifdef SUPPORT_UTF |
1607 | if (utf) |
1608 | { |
1609 | i = GET(ecode, 1); |
1610 | while (i-- > 0) |
1611 | { |
1612 | eptr--; |
1613 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); |
1614 | BACKCHAR(eptr); |
1615 | } |
1616 | } |
1617 | else |
1618 | #endif |
1619 | |
1620 | /* No UTF-8 support, or not in UTF-8 mode: count is byte count */ |
1621 | |
1622 | { |
1623 | eptr -= GET(ecode, 1); |
1624 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); |
1625 | } |
1626 | |
1627 | /* Save the earliest consulted character, then skip to next op code */ |
1628 | |
1629 | if (eptr < md->start_used_ptr) md->start_used_ptr = eptr; |
1630 | ecode += 1 + LINK_SIZE; |
1631 | break; |
1632 | |
1633 | /* The callout item calls an external function, if one is provided, passing |
1634 | details of the match so far. This is mainly for debugging, though the |
1635 | function is able to force a failure. */ |
1636 | |
1637 | case OP_CALLOUT: |
1638 | if (PUBL(callout) != NULL) |
1639 | { |
1640 | PUBL(callout_block) cb; |
1641 | cb.version = 2; /* Version 1 of the callout block */ |
1642 | cb.callout_number = ecode[1]; |
1643 | cb.offset_vector = md->offset_vector; |
1644 | #ifdef COMPILE_PCRE8 |
1645 | cb.subject = (PCRE_SPTR)md->start_subject; |
1646 | #else |
1647 | cb.subject = (PCRE_SPTR16)md->start_subject; |
1648 | #endif |
1649 | cb.subject_length = (int)(md->end_subject - md->start_subject); |
1650 | cb.start_match = (int)(mstart - md->start_subject); |
1651 | cb.current_position = (int)(eptr - md->start_subject); |
1652 | cb.pattern_position = GET(ecode, 2); |
1653 | cb.next_item_length = GET(ecode, 2 + LINK_SIZE); |
1654 | cb.capture_top = offset_top/2; |
1655 | cb.capture_last = md->capture_last; |
1656 | cb.callout_data = md->callout_data; |
1657 | cb.mark = md->nomatch_mark; |
1658 | if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH); |
1659 | if (rrc < 0) RRETURN(rrc); |
1660 | } |
1661 | ecode += 2 + 2*LINK_SIZE; |
1662 | break; |
1663 | |
1664 | /* Recursion either matches the current regex, or some subexpression. The |
1665 | offset data is the offset to the starting bracket from the start of the |
1666 | whole pattern. (This is so that it works from duplicated subpatterns.) |
1667 | |
1668 | The state of the capturing groups is preserved over recursion, and |
1669 | re-instated afterwards. We don't know how many are started and not yet |
1670 | finished (offset_top records the completed total) so we just have to save |
1671 | all the potential data. There may be up to 65535 such values, which is too |
1672 | large to put on the stack, but using malloc for small numbers seems |
1673 | expensive. As a compromise, the stack is used when there are no more than |
1674 | REC_STACK_SAVE_MAX values to store; otherwise malloc is used. |
1675 | |
1676 | There are also other values that have to be saved. We use a chained |
1677 | sequence of blocks that actually live on the stack. Thanks to Robin Houston |
1678 | for the original version of this logic. It has, however, been hacked around |
1679 | a lot, so he is not to blame for the current way it works. */ |
1680 | |
1681 | case OP_RECURSE: |
1682 | { |
1683 | recursion_info *ri; |
1684 | int recno; |
1685 | |
1686 | callpat = md->start_code + GET(ecode, 1); |
1687 | recno = (callpat == md->start_code)? 0 : |
1688 | GET2(callpat, 1 + LINK_SIZE); |
1689 | |
1690 | /* Check for repeating a recursion without advancing the subject pointer. |
1691 | This should catch convoluted mutual recursions. (Some simple cases are |
1692 | caught at compile time.) */ |
1693 | |
1694 | for (ri = md->recursive; ri != NULL; ri = ri->prevrec) |
1695 | if (recno == ri->group_num && eptr == ri->subject_position) |
1696 | RRETURN(PCRE_ERROR_RECURSELOOP); |
1697 | |
1698 | /* Add to "recursing stack" */ |
1699 | |
1700 | new_recursive.group_num = recno; |
1701 | new_recursive.subject_position = eptr; |
1702 | new_recursive.prevrec = md->recursive; |
1703 | md->recursive = &new_recursive; |
1704 | |
1705 | /* Where to continue from afterwards */ |
1706 | |
1707 | ecode += 1 + LINK_SIZE; |
1708 | |
1709 | /* Now save the offset data */ |
1710 | |
1711 | new_recursive.saved_max = md->offset_end; |
1712 | if (new_recursive.saved_max <= REC_STACK_SAVE_MAX) |
1713 | new_recursive.offset_save = stacksave; |
1714 | else |
1715 | { |
1716 | new_recursive.offset_save = |
1717 | (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int)); |
1718 | if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY); |
1719 | } |
1720 | memcpy(new_recursive.offset_save, md->offset_vector, |
1721 | new_recursive.saved_max * sizeof(int)); |
1722 | |
1723 | /* OK, now we can do the recursion. After processing each alternative, |
1724 | restore the offset data. If there were nested recursions, md->recursive |
1725 | might be changed, so reset it before looping. */ |
1726 | |
1727 | DPRINTF(("Recursing into group %d\n", new_recursive.group_num)); |
1728 | cbegroup = (*callpat >= OP_SBRA); |
1729 | do |
1730 | { |
1731 | if (cbegroup) md->match_function_type = MATCH_CBEGROUP; |
1732 | RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top, |
1733 | md, eptrb, RM6); |
1734 | memcpy(md->offset_vector, new_recursive.offset_save, |
1735 | new_recursive.saved_max * sizeof(int)); |
1736 | md->recursive = new_recursive.prevrec; |
1737 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) |
1738 | { |
1739 | DPRINTF(("Recursion matched\n")); |
1740 | if (new_recursive.offset_save != stacksave) |
1741 | (PUBL(free))(new_recursive.offset_save); |
1742 | |
1743 | /* Set where we got to in the subject, and reset the start in case |
1744 | it was changed by \K. This *is* propagated back out of a recursion, |
1745 | for Perl compatibility. */ |
1746 | |
1747 | eptr = md->end_match_ptr; |
1748 | mstart = md->start_match_ptr; |
1749 | goto RECURSION_MATCHED; /* Exit loop; end processing */ |
1750 | } |
1751 | |
1752 | /* PCRE does not allow THEN to escape beyond a recursion; it is treated |
1753 | as NOMATCH. */ |
1754 | |
1755 | else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) |
1756 | { |
1757 | DPRINTF(("Recursion gave error %d\n", rrc)); |
1758 | if (new_recursive.offset_save != stacksave) |
1759 | (PUBL(free))(new_recursive.offset_save); |
1760 | RRETURN(rrc); |
1761 | } |
1762 | |
1763 | md->recursive = &new_recursive; |
1764 | callpat += GET(callpat, 1); |
1765 | } |
1766 | while (*callpat == OP_ALT); |
1767 | |
1768 | DPRINTF(("Recursion didn't match\n")); |
1769 | md->recursive = new_recursive.prevrec; |
1770 | if (new_recursive.offset_save != stacksave) |
1771 | (PUBL(free))(new_recursive.offset_save); |
1772 | RRETURN(MATCH_NOMATCH); |
1773 | } |
1774 | |
1775 | RECURSION_MATCHED: |
1776 | break; |
1777 | |
1778 | /* An alternation is the end of a branch; scan along to find the end of the |
1779 | bracketed group and go to there. */ |
1780 | |
1781 | case OP_ALT: |
1782 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1783 | break; |
1784 | |
1785 | /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group, |
1786 | indicating that it may occur zero times. It may repeat infinitely, or not |
1787 | at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets |
1788 | with fixed upper repeat limits are compiled as a number of copies, with the |
1789 | optional ones preceded by BRAZERO or BRAMINZERO. */ |
1790 | |
1791 | case OP_BRAZERO: |
1792 | next = ecode + 1; |
1793 | RMATCH(eptr, next, offset_top, md, eptrb, RM10); |
1794 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1795 | do next += GET(next, 1); while (*next == OP_ALT); |
1796 | ecode = next + 1 + LINK_SIZE; |
1797 | break; |
1798 | |
1799 | case OP_BRAMINZERO: |
1800 | next = ecode + 1; |
1801 | do next += GET(next, 1); while (*next == OP_ALT); |
1802 | RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11); |
1803 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1804 | ecode++; |
1805 | break; |
1806 | |
1807 | case OP_SKIPZERO: |
1808 | next = ecode+1; |
1809 | do next += GET(next,1); while (*next == OP_ALT); |
1810 | ecode = next + 1 + LINK_SIZE; |
1811 | break; |
1812 | |
1813 | /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything |
1814 | here; just jump to the group, with allow_zero set TRUE. */ |
1815 | |
1816 | case OP_BRAPOSZERO: |
1817 | op = *(++ecode); |
1818 | allow_zero = TRUE; |
1819 | if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE; |
1820 | goto POSSESSIVE_NON_CAPTURE; |
1821 | |
1822 | /* End of a group, repeated or non-repeating. */ |
1823 | |
1824 | case OP_KET: |
1825 | case OP_KETRMIN: |
1826 | case OP_KETRMAX: |
1827 | case OP_KETRPOS: |
1828 | prev = ecode - GET(ecode, 1); |
1829 | |
1830 | /* If this was a group that remembered the subject start, in order to break |
1831 | infinite repeats of empty string matches, retrieve the subject start from |
1832 | the chain. Otherwise, set it NULL. */ |
1833 | |
1834 | if (*prev >= OP_SBRA || *prev == OP_ONCE) |
1835 | { |
1836 | saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */ |
1837 | eptrb = eptrb->epb_prev; /* Backup to previous group */ |
1838 | } |
1839 | else saved_eptr = NULL; |
1840 | |
1841 | /* If we are at the end of an assertion group or a non-capturing atomic |
1842 | group, stop matching and return MATCH_MATCH, but record the current high |
1843 | water mark for use by positive assertions. We also need to record the match |
1844 | start in case it was changed by \K. */ |
1845 | |
1846 | if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) || |
1847 | *prev == OP_ONCE_NC) |
1848 | { |
1849 | md->end_match_ptr = eptr; /* For ONCE_NC */ |
1850 | md->end_offset_top = offset_top; |
1851 | md->start_match_ptr = mstart; |
1852 | RRETURN(MATCH_MATCH); /* Sets md->mark */ |
1853 | } |
1854 | |
1855 | /* For capturing groups we have to check the group number back at the start |
1856 | and if necessary complete handling an extraction by setting the offsets and |
1857 | bumping the high water mark. Whole-pattern recursion is coded as a recurse |
1858 | into group 0, so it won't be picked up here. Instead, we catch it when the |
1859 | OP_END is reached. Other recursion is handled here. We just have to record |
1860 | the current subject position and start match pointer and give a MATCH |
1861 | return. */ |
1862 | |
1863 | if (*prev == OP_CBRA || *prev == OP_SCBRA || |
1864 | *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS) |
1865 | { |
1866 | number = GET2(prev, 1+LINK_SIZE); |
1867 | offset = number << 1; |
1868 | |
1869 | #ifdef PCRE_DEBUG |
1870 | printf("end bracket %d", number); |
1871 | printf("\n"); |
1872 | #endif |
1873 | |
1874 | /* Handle a recursively called group. */ |
1875 | |
1876 | if (md->recursive != NULL && md->recursive->group_num == number) |
1877 | { |
1878 | md->end_match_ptr = eptr; |
1879 | md->start_match_ptr = mstart; |
1880 | RRETURN(MATCH_MATCH); |
1881 | } |
1882 | |
1883 | /* Deal with capturing */ |
1884 | |
1885 | md->capture_last = number; |
1886 | if (offset >= md->offset_max) md->offset_overflow = TRUE; else |
1887 | { |
1888 | /* If offset is greater than offset_top, it means that we are |
1889 | "skipping" a capturing group, and that group's offsets must be marked |
1890 | unset. In earlier versions of PCRE, all the offsets were unset at the |
1891 | start of matching, but this doesn't work because atomic groups and |
1892 | assertions can cause a value to be set that should later be unset. |
1893 | Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as |
1894 | part of the atomic group, but this is not on the final matching path, |
1895 | so must be unset when 2 is set. (If there is no group 2, there is no |
1896 | problem, because offset_top will then be 2, indicating no capture.) */ |
1897 | |
1898 | if (offset > offset_top) |
1899 | { |
1900 | register int *iptr = md->offset_vector + offset_top; |
1901 | register int *iend = md->offset_vector + offset; |
1902 | while (iptr < iend) *iptr++ = -1; |
1903 | } |
1904 | |
1905 | /* Now make the extraction */ |
1906 | |
1907 | md->offset_vector[offset] = |
1908 | md->offset_vector[md->offset_end - number]; |
1909 | md->offset_vector[offset+1] = (int)(eptr - md->start_subject); |
1910 | if (offset_top <= offset) offset_top = offset + 2; |
1911 | } |
1912 | } |
1913 | |
1914 | /* For an ordinary non-repeating ket, just continue at this level. This |
1915 | also happens for a repeating ket if no characters were matched in the |
1916 | group. This is the forcible breaking of infinite loops as implemented in |
1917 | Perl 5.005. For a non-repeating atomic group that includes captures, |
1918 | establish a backup point by processing the rest of the pattern at a lower |
1919 | level. If this results in a NOMATCH return, pass MATCH_ONCE back to the |
1920 | original OP_ONCE level, thereby bypassing intermediate backup points, but |
1921 | resetting any captures that happened along the way. */ |
1922 | |
1923 | if (*ecode == OP_KET || eptr == saved_eptr) |
1924 | { |
1925 | if (*prev == OP_ONCE) |
1926 | { |
1927 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12); |
1928 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1929 | md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */ |
1930 | RRETURN(MATCH_ONCE); |
1931 | } |
1932 | ecode += 1 + LINK_SIZE; /* Carry on at this level */ |
1933 | break; |
1934 | } |
1935 | |
1936 | /* OP_KETRPOS is a possessive repeating ket. Remember the current position, |
1937 | and return the MATCH_KETRPOS. This makes it possible to do the repeats one |
1938 | at a time from the outer level, thus saving stack. */ |
1939 | |
1940 | if (*ecode == OP_KETRPOS) |
1941 | { |
1942 | md->end_match_ptr = eptr; |
1943 | md->end_offset_top = offset_top; |
1944 | RRETURN(MATCH_KETRPOS); |
1945 | } |
1946 | |
1947 | /* The normal repeating kets try the rest of the pattern or restart from |
1948 | the preceding bracket, in the appropriate order. In the second case, we can |
1949 | use tail recursion to avoid using another stack frame, unless we have an |
1950 | an atomic group or an unlimited repeat of a group that can match an empty |
1951 | string. */ |
1952 | |
1953 | if (*ecode == OP_KETRMIN) |
1954 | { |
1955 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7); |
1956 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1957 | if (*prev == OP_ONCE) |
1958 | { |
1959 | RMATCH(eptr, prev, offset_top, md, eptrb, RM8); |
1960 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1961 | md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */ |
1962 | RRETURN(MATCH_ONCE); |
1963 | } |
1964 | if (*prev >= OP_SBRA) /* Could match an empty string */ |
1965 | { |
1966 | md->match_function_type = MATCH_CBEGROUP; |
1967 | RMATCH(eptr, prev, offset_top, md, eptrb, RM50); |
1968 | RRETURN(rrc); |
1969 | } |
1970 | ecode = prev; |
1971 | goto TAIL_RECURSE; |
1972 | } |
1973 | else /* OP_KETRMAX */ |
1974 | { |
1975 | if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
1976 | RMATCH(eptr, prev, offset_top, md, eptrb, RM13); |
1977 | if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH; |
1978 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1979 | if (*prev == OP_ONCE) |
1980 | { |
1981 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9); |
1982 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1983 | md->once_target = prev; |
1984 | RRETURN(MATCH_ONCE); |
1985 | } |
1986 | ecode += 1 + LINK_SIZE; |
1987 | goto TAIL_RECURSE; |
1988 | } |
1989 | /* Control never gets here */ |
1990 | |
1991 | /* Not multiline mode: start of subject assertion, unless notbol. */ |
1992 | |
1993 | case OP_CIRC: |
1994 | if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH); |
1995 | |
1996 | /* Start of subject assertion */ |
1997 | |
1998 | case OP_SOD: |
1999 | if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH); |
2000 | ecode++; |
2001 | break; |
2002 | |
2003 | /* Multiline mode: start of subject unless notbol, or after any newline. */ |
2004 | |
2005 | case OP_CIRCM: |
2006 | if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH); |
2007 | if (eptr != md->start_subject && |
2008 | (eptr == md->end_subject || !WAS_NEWLINE(eptr))) |
2009 | RRETURN(MATCH_NOMATCH); |
2010 | ecode++; |
2011 | break; |
2012 | |
2013 | /* Start of match assertion */ |
2014 | |
2015 | case OP_SOM: |
2016 | if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH); |
2017 | ecode++; |
2018 | break; |
2019 | |
2020 | /* Reset the start of match point */ |
2021 | |
2022 | case OP_SET_SOM: |
2023 | mstart = eptr; |
2024 | ecode++; |
2025 | break; |
2026 | |
2027 | /* Multiline mode: assert before any newline, or before end of subject |
2028 | unless noteol is set. */ |
2029 | |
2030 | case OP_DOLLM: |
2031 | if (eptr < md->end_subject) |
2032 | { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); } |
2033 | else |
2034 | { |
2035 | if (md->noteol) RRETURN(MATCH_NOMATCH); |
2036 | SCHECK_PARTIAL(); |
2037 | } |
2038 | ecode++; |
2039 | break; |
2040 | |
2041 | /* Not multiline mode: assert before a terminating newline or before end of |
2042 | subject unless noteol is set. */ |
2043 | |
2044 | case OP_DOLL: |
2045 | if (md->noteol) RRETURN(MATCH_NOMATCH); |
2046 | if (!md->endonly) goto ASSERT_NL_OR_EOS; |
2047 | |
2048 | /* ... else fall through for endonly */ |
2049 | |
2050 | /* End of subject assertion (\z) */ |
2051 | |
2052 | case OP_EOD: |
2053 | if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH); |
2054 | SCHECK_PARTIAL(); |
2055 | ecode++; |
2056 | break; |
2057 | |
2058 | /* End of subject or ending \n assertion (\Z) */ |
2059 | |
2060 | case OP_EODN: |
2061 | ASSERT_NL_OR_EOS: |
2062 | if (eptr < md->end_subject && |
2063 | (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen)) |
2064 | RRETURN(MATCH_NOMATCH); |
2065 | |
2066 | /* Either at end of string or \n before end. */ |
2067 | |
2068 | SCHECK_PARTIAL(); |
2069 | ecode++; |
2070 | break; |
2071 | |
2072 | /* Word boundary assertions */ |
2073 | |
2074 | case OP_NOT_WORD_BOUNDARY: |
2075 | case OP_WORD_BOUNDARY: |
2076 | { |
2077 | |
2078 | /* Find out if the previous and current characters are "word" characters. |
2079 | It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to |
2080 | be "non-word" characters. Remember the earliest consulted character for |
2081 | partial matching. */ |
2082 | |
2083 | #ifdef SUPPORT_UTF |
2084 | if (utf) |
2085 | { |
2086 | /* Get status of previous character */ |
2087 | |
2088 | if (eptr == md->start_subject) prev_is_word = FALSE; else |
2089 | { |
2090 | PCRE_PUCHAR lastptr = eptr - 1; |
2091 | BACKCHAR(lastptr); |
2092 | if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr; |
2093 | GETCHAR(c, lastptr); |
2094 | #ifdef SUPPORT_UCP |
2095 | if (md->use_ucp) |
2096 | { |
2097 | if (c == '_') prev_is_word = TRUE; else |
2098 | { |
2099 | int cat = UCD_CATEGORY(c); |
2100 | prev_is_word = (cat == ucp_L || cat == ucp_N); |
2101 | } |
2102 | } |
2103 | else |
2104 | #endif |
2105 | prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; |
2106 | } |
2107 | |
2108 | /* Get status of next character */ |
2109 | |
2110 | if (eptr >= md->end_subject) |
2111 | { |
2112 | SCHECK_PARTIAL(); |
2113 | cur_is_word = FALSE; |
2114 | } |
2115 | else |
2116 | { |
2117 | GETCHAR(c, eptr); |
2118 | #ifdef SUPPORT_UCP |
2119 | if (md->use_ucp) |
2120 | { |
2121 | if (c == '_') cur_is_word = TRUE; else |
2122 | { |
2123 | int cat = UCD_CATEGORY(c); |
2124 | cur_is_word = (cat == ucp_L || cat == ucp_N); |
2125 | } |
2126 | } |
2127 | else |
2128 | #endif |
2129 | cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; |
2130 | } |
2131 | } |
2132 | else |
2133 | #endif |
2134 | |
2135 | /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for |
2136 | consistency with the behaviour of \w we do use it in this case. */ |
2137 | |
2138 | { |
2139 | /* Get status of previous character */ |
2140 | |
2141 | if (eptr == md->start_subject) prev_is_word = FALSE; else |
2142 | { |
2143 | if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1; |
2144 | #ifdef SUPPORT_UCP |
2145 | if (md->use_ucp) |
2146 | { |
2147 | c = eptr[-1]; |
2148 | if (c == '_') prev_is_word = TRUE; else |
2149 | { |
2150 | int cat = UCD_CATEGORY(c); |
2151 | prev_is_word = (cat == ucp_L || cat == ucp_N); |
2152 | } |
2153 | } |
2154 | else |
2155 | #endif |
2156 | prev_is_word = MAX_255(eptr[-1]) |
2157 | && ((md->ctypes[eptr[-1]] & ctype_word) != 0); |
2158 | } |
2159 | |
2160 | /* Get status of next character */ |
2161 | |
2162 | if (eptr >= md->end_subject) |
2163 | { |
2164 | SCHECK_PARTIAL(); |
2165 | cur_is_word = FALSE; |
2166 | } |
2167 | else |
2168 | #ifdef SUPPORT_UCP |
2169 | if (md->use_ucp) |
2170 | { |
2171 | c = *eptr; |
2172 | if (c == '_') cur_is_word = TRUE; else |
2173 | { |
2174 | int cat = UCD_CATEGORY(c); |
2175 | cur_is_word = (cat == ucp_L || cat == ucp_N); |
2176 | } |
2177 | } |
2178 | else |
2179 | #endif |
2180 | cur_is_word = MAX_255(*eptr) |
2181 | && ((md->ctypes[*eptr] & ctype_word) != 0); |
2182 | } |
2183 | |
2184 | /* Now see if the situation is what we want */ |
2185 | |
2186 | if ((*ecode++ == OP_WORD_BOUNDARY)? |
2187 | cur_is_word == prev_is_word : cur_is_word != prev_is_word) |
2188 | RRETURN(MATCH_NOMATCH); |
2189 | } |
2190 | break; |
2191 | |
2192 | /* Match a single character type; inline for speed */ |
2193 | |
2194 | case OP_ANY: |
2195 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
2196 | /* Fall through */ |
2197 | |
2198 | case OP_ALLANY: |
2199 | if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */ |
2200 | { /* not be updated before SCHECK_PARTIAL. */ |
2201 | SCHECK_PARTIAL(); |
2202 | RRETURN(MATCH_NOMATCH); |
2203 | } |
2204 | eptr++; |
2205 | #ifdef SUPPORT_UTF |
2206 | if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
2207 | #endif |
2208 | ecode++; |
2209 | break; |
2210 | |
2211 | /* Match a single byte, even in UTF-8 mode. This opcode really does match |
2212 | any byte, even newline, independent of the setting of PCRE_DOTALL. */ |
2213 | |
2214 | case OP_ANYBYTE: |
2215 | if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */ |
2216 | { /* not be updated before SCHECK_PARTIAL. */ |
2217 | SCHECK_PARTIAL(); |
2218 | RRETURN(MATCH_NOMATCH); |
2219 | } |
2220 | eptr++; |
2221 | ecode++; |
2222 | break; |
2223 | |
2224 | case OP_NOT_DIGIT: |
2225 | if (eptr >= md->end_subject) |
2226 | { |
2227 | SCHECK_PARTIAL(); |
2228 | RRETURN(MATCH_NOMATCH); |
2229 | } |
2230 | GETCHARINCTEST(c, eptr); |
2231 | if ( |
2232 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2233 | c < 256 && |
2234 | #endif |
2235 | (md->ctypes[c] & ctype_digit) != 0 |
2236 | ) |
2237 | RRETURN(MATCH_NOMATCH); |
2238 | ecode++; |
2239 | break; |
2240 | |
2241 | case OP_DIGIT: |
2242 | if (eptr >= md->end_subject) |
2243 | { |
2244 | SCHECK_PARTIAL(); |
2245 | RRETURN(MATCH_NOMATCH); |
2246 | } |
2247 | GETCHARINCTEST(c, eptr); |
2248 | if ( |
2249 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2250 | c > 255 || |
2251 | #endif |
2252 | (md->ctypes[c] & ctype_digit) == 0 |
2253 | ) |
2254 | RRETURN(MATCH_NOMATCH); |
2255 | ecode++; |
2256 | break; |
2257 | |
2258 | case OP_NOT_WHITESPACE: |
2259 | if (eptr >= md->end_subject) |
2260 | { |
2261 | SCHECK_PARTIAL(); |
2262 | RRETURN(MATCH_NOMATCH); |
2263 | } |
2264 | GETCHARINCTEST(c, eptr); |
2265 | if ( |
2266 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2267 | c < 256 && |
2268 | #endif |
2269 | (md->ctypes[c] & ctype_space) != 0 |
2270 | ) |
2271 | RRETURN(MATCH_NOMATCH); |
2272 | ecode++; |
2273 | break; |
2274 | |
2275 | case OP_WHITESPACE: |
2276 | if (eptr >= md->end_subject) |
2277 | { |
2278 | SCHECK_PARTIAL(); |
2279 | RRETURN(MATCH_NOMATCH); |
2280 | } |
2281 | GETCHARINCTEST(c, eptr); |
2282 | if ( |
2283 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2284 | c > 255 || |
2285 | #endif |
2286 | (md->ctypes[c] & ctype_space) == 0 |
2287 | ) |
2288 | RRETURN(MATCH_NOMATCH); |
2289 | ecode++; |
2290 | break; |
2291 | |
2292 | case OP_NOT_WORDCHAR: |
2293 | if (eptr >= md->end_subject) |
2294 | { |
2295 | SCHECK_PARTIAL(); |
2296 | RRETURN(MATCH_NOMATCH); |
2297 | } |
2298 | GETCHARINCTEST(c, eptr); |
2299 | if ( |
2300 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2301 | c < 256 && |
2302 | #endif |
2303 | (md->ctypes[c] & ctype_word) != 0 |
2304 | ) |
2305 | RRETURN(MATCH_NOMATCH); |
2306 | ecode++; |
2307 | break; |
2308 | |
2309 | case OP_WORDCHAR: |
2310 | if (eptr >= md->end_subject) |
2311 | { |
2312 | SCHECK_PARTIAL(); |
2313 | RRETURN(MATCH_NOMATCH); |
2314 | } |
2315 | GETCHARINCTEST(c, eptr); |
2316 | if ( |
2317 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2318 | c > 255 || |
2319 | #endif |
2320 | (md->ctypes[c] & ctype_word) == 0 |
2321 | ) |
2322 | RRETURN(MATCH_NOMATCH); |
2323 | ecode++; |
2324 | break; |
2325 | |
2326 | case OP_ANYNL: |
2327 | if (eptr >= md->end_subject) |
2328 | { |
2329 | SCHECK_PARTIAL(); |
2330 | RRETURN(MATCH_NOMATCH); |
2331 | } |
2332 | GETCHARINCTEST(c, eptr); |
2333 | switch(c) |
2334 | { |
2335 | default: RRETURN(MATCH_NOMATCH); |
2336 | |
2337 | case 0x000d: |
2338 | if (eptr < md->end_subject && *eptr == 0x0a) eptr++; |
2339 | break; |
2340 | |
2341 | case 0x000a: |
2342 | break; |
2343 | |
2344 | case 0x000b: |
2345 | case 0x000c: |
2346 | case 0x0085: |
2347 | case 0x2028: |
2348 | case 0x2029: |
2349 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
2350 | break; |
2351 | } |
2352 | ecode++; |
2353 | break; |
2354 | |
2355 | case OP_NOT_HSPACE: |
2356 | if (eptr >= md->end_subject) |
2357 | { |
2358 | SCHECK_PARTIAL(); |
2359 | RRETURN(MATCH_NOMATCH); |
2360 | } |
2361 | GETCHARINCTEST(c, eptr); |
2362 | switch(c) |
2363 | { |
2364 | default: break; |
2365 | case 0x09: /* HT */ |
2366 | case 0x20: /* SPACE */ |
2367 | case 0xa0: /* NBSP */ |
2368 | case 0x1680: /* OGHAM SPACE MARK */ |
2369 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
2370 | case 0x2000: /* EN QUAD */ |
2371 | case 0x2001: /* EM QUAD */ |
2372 | case 0x2002: /* EN SPACE */ |
2373 | case 0x2003: /* EM SPACE */ |
2374 | case 0x2004: /* THREE-PER-EM SPACE */ |
2375 | case 0x2005: /* FOUR-PER-EM SPACE */ |
2376 | case 0x2006: /* SIX-PER-EM SPACE */ |
2377 | case 0x2007: /* FIGURE SPACE */ |
2378 | case 0x2008: /* PUNCTUATION SPACE */ |
2379 | case 0x2009: /* THIN SPACE */ |
2380 | case 0x200A: /* HAIR SPACE */ |
2381 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
2382 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
2383 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
2384 | RRETURN(MATCH_NOMATCH); |
2385 | } |
2386 | ecode++; |
2387 | break; |
2388 | |
2389 | case OP_HSPACE: |
2390 | if (eptr >= md->end_subject) |
2391 | { |
2392 | SCHECK_PARTIAL(); |
2393 | RRETURN(MATCH_NOMATCH); |
2394 | } |
2395 | GETCHARINCTEST(c, eptr); |
2396 | switch(c) |
2397 | { |
2398 | default: RRETURN(MATCH_NOMATCH); |
2399 | case 0x09: /* HT */ |
2400 | case 0x20: /* SPACE */ |
2401 | case 0xa0: /* NBSP */ |
2402 | case 0x1680: /* OGHAM SPACE MARK */ |
2403 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
2404 | case 0x2000: /* EN QUAD */ |
2405 | case 0x2001: /* EM QUAD */ |
2406 | case 0x2002: /* EN SPACE */ |
2407 | case 0x2003: /* EM SPACE */ |
2408 | case 0x2004: /* THREE-PER-EM SPACE */ |
2409 | case 0x2005: /* FOUR-PER-EM SPACE */ |
2410 | case 0x2006: /* SIX-PER-EM SPACE */ |
2411 | case 0x2007: /* FIGURE SPACE */ |
2412 | case 0x2008: /* PUNCTUATION SPACE */ |
2413 | case 0x2009: /* THIN SPACE */ |
2414 | case 0x200A: /* HAIR SPACE */ |
2415 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
2416 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
2417 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
2418 | break; |
2419 | } |
2420 | ecode++; |
2421 | break; |
2422 | |
2423 | case OP_NOT_VSPACE: |
2424 | if (eptr >= md->end_subject) |
2425 | { |
2426 | SCHECK_PARTIAL(); |
2427 | RRETURN(MATCH_NOMATCH); |
2428 | } |
2429 | GETCHARINCTEST(c, eptr); |
2430 | switch(c) |
2431 | { |
2432 | default: break; |
2433 | case 0x0a: /* LF */ |
2434 | case 0x0b: /* VT */ |
2435 | case 0x0c: /* FF */ |
2436 | case 0x0d: /* CR */ |
2437 | case 0x85: /* NEL */ |
2438 | case 0x2028: /* LINE SEPARATOR */ |
2439 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
2440 | RRETURN(MATCH_NOMATCH); |
2441 | } |
2442 | ecode++; |
2443 | break; |
2444 | |
2445 | case OP_VSPACE: |
2446 | if (eptr >= md->end_subject) |
2447 | { |
2448 | SCHECK_PARTIAL(); |
2449 | RRETURN(MATCH_NOMATCH); |
2450 | } |
2451 | GETCHARINCTEST(c, eptr); |
2452 | switch(c) |
2453 | { |
2454 | default: RRETURN(MATCH_NOMATCH); |
2455 | case 0x0a: /* LF */ |
2456 | case 0x0b: /* VT */ |
2457 | case 0x0c: /* FF */ |
2458 | case 0x0d: /* CR */ |
2459 | case 0x85: /* NEL */ |
2460 | case 0x2028: /* LINE SEPARATOR */ |
2461 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
2462 | break; |
2463 | } |
2464 | ecode++; |
2465 | break; |
2466 | |
2467 | #ifdef SUPPORT_UCP |
2468 | /* Check the next character by Unicode property. We will get here only |
2469 | if the support is in the binary; otherwise a compile-time error occurs. */ |
2470 | |
2471 | case OP_PROP: |
2472 | case OP_NOTPROP: |
2473 | if (eptr >= md->end_subject) |
2474 | { |
2475 | SCHECK_PARTIAL(); |
2476 | RRETURN(MATCH_NOMATCH); |
2477 | } |
2478 | GETCHARINCTEST(c, eptr); |
2479 | { |
2480 | const ucd_record *prop = GET_UCD(c); |
2481 | |
2482 | switch(ecode[1]) |
2483 | { |
2484 | case PT_ANY: |
2485 | if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH); |
2486 | break; |
2487 | |
2488 | case PT_LAMP: |
2489 | if ((prop->chartype == ucp_Lu || |
2490 | prop->chartype == ucp_Ll || |
2491 | prop->chartype == ucp_Lt) == (op == OP_NOTPROP)) |
2492 | RRETURN(MATCH_NOMATCH); |
2493 | break; |
2494 | |
2495 | case PT_GC: |
2496 | if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP)) |
2497 | RRETURN(MATCH_NOMATCH); |
2498 | break; |
2499 | |
2500 | case PT_PC: |
2501 | if ((ecode[2] != prop->chartype) == (op == OP_PROP)) |
2502 | RRETURN(MATCH_NOMATCH); |
2503 | break; |
2504 | |
2505 | case PT_SC: |
2506 | if ((ecode[2] != prop->script) == (op == OP_PROP)) |
2507 | RRETURN(MATCH_NOMATCH); |
2508 | break; |
2509 | |
2510 | /* These are specials */ |
2511 | |
2512 | case PT_ALNUM: |
2513 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2514 | PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP)) |
2515 | RRETURN(MATCH_NOMATCH); |
2516 | break; |
2517 | |
2518 | case PT_SPACE: /* Perl space */ |
2519 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z || |
2520 | c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR) |
2521 | == (op == OP_NOTPROP)) |
2522 | RRETURN(MATCH_NOMATCH); |
2523 | break; |
2524 | |
2525 | case PT_PXSPACE: /* POSIX space */ |
2526 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z || |
2527 | c == CHAR_HT || c == CHAR_NL || c == CHAR_VT || |
2528 | c == CHAR_FF || c == CHAR_CR) |
2529 | == (op == OP_NOTPROP)) |
2530 | RRETURN(MATCH_NOMATCH); |
2531 | break; |
2532 | |
2533 | case PT_WORD: |
2534 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2535 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
2536 | c == CHAR_UNDERSCORE) == (op == OP_NOTPROP)) |
2537 | RRETURN(MATCH_NOMATCH); |
2538 | break; |
2539 | |
2540 | /* This should never occur */ |
2541 | |
2542 | default: |
2543 | RRETURN(PCRE_ERROR_INTERNAL); |
2544 | } |
2545 | |
2546 | ecode += 3; |
2547 | } |
2548 | break; |
2549 | |
2550 | /* Match an extended Unicode sequence. We will get here only if the support |
2551 | is in the binary; otherwise a compile-time error occurs. */ |
2552 | |
2553 | case OP_EXTUNI: |
2554 | if (eptr >= md->end_subject) |
2555 | { |
2556 | SCHECK_PARTIAL(); |
2557 | RRETURN(MATCH_NOMATCH); |
2558 | } |
2559 | GETCHARINCTEST(c, eptr); |
2560 | if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH); |
2561 | while (eptr < md->end_subject) |
2562 | { |
2563 | int len = 1; |
2564 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
2565 | if (UCD_CATEGORY(c) != ucp_M) break; |
2566 | eptr += len; |
2567 | } |
2568 | ecode++; |
2569 | break; |
2570 | #endif |
2571 | |
2572 | |
2573 | /* Match a back reference, possibly repeatedly. Look past the end of the |
2574 | item to see if there is repeat information following. The code is similar |
2575 | to that for character classes, but repeated for efficiency. Then obey |
2576 | similar code to character type repeats - written out again for speed. |
2577 | However, if the referenced string is the empty string, always treat |
2578 | it as matched, any number of times (otherwise there could be infinite |
2579 | loops). */ |
2580 | |
2581 | case OP_REF: |
2582 | case OP_REFI: |
2583 | caseless = op == OP_REFI; |
2584 | offset = GET2(ecode, 1) << 1; /* Doubled ref number */ |
2585 | ecode += 1 + IMM2_SIZE; |
2586 | |
2587 | /* If the reference is unset, there are two possibilities: |
2588 | |
2589 | (a) In the default, Perl-compatible state, set the length negative; |
2590 | this ensures that every attempt at a match fails. We can't just fail |
2591 | here, because of the possibility of quantifiers with zero minima. |
2592 | |
2593 | (b) If the JavaScript compatibility flag is set, set the length to zero |
2594 | so that the back reference matches an empty string. |
2595 | |
2596 | Otherwise, set the length to the length of what was matched by the |
2597 | referenced subpattern. */ |
2598 | |
2599 | if (offset >= offset_top || md->offset_vector[offset] < 0) |
2600 | length = (md->jscript_compat)? 0 : -1; |
2601 | else |
2602 | length = md->offset_vector[offset+1] - md->offset_vector[offset]; |
2603 | |
2604 | /* Set up for repetition, or handle the non-repeated case */ |
2605 | |
2606 | switch (*ecode) |
2607 | { |
2608 | case OP_CRSTAR: |
2609 | case OP_CRMINSTAR: |
2610 | case OP_CRPLUS: |
2611 | case OP_CRMINPLUS: |
2612 | case OP_CRQUERY: |
2613 | case OP_CRMINQUERY: |
2614 | c = *ecode++ - OP_CRSTAR; |
2615 | minimize = (c & 1) != 0; |
2616 | min = rep_min[c]; /* Pick up values from tables; */ |
2617 | max = rep_max[c]; /* zero for max => infinity */ |
2618 | if (max == 0) max = INT_MAX; |
2619 | break; |
2620 | |
2621 | case OP_CRRANGE: |
2622 | case OP_CRMINRANGE: |
2623 | minimize = (*ecode == OP_CRMINRANGE); |
2624 | min = GET2(ecode, 1); |
2625 | max = GET2(ecode, 1 + IMM2_SIZE); |
2626 | if (max == 0) max = INT_MAX; |
2627 | ecode += 1 + 2 * IMM2_SIZE; |
2628 | break; |
2629 | |
2630 | default: /* No repeat follows */ |
2631 | if ((length = match_ref(offset, eptr, length, md, caseless)) < 0) |
2632 | { |
2633 | CHECK_PARTIAL(); |
2634 | RRETURN(MATCH_NOMATCH); |
2635 | } |
2636 | eptr += length; |
2637 | continue; /* With the main loop */ |
2638 | } |
2639 | |
2640 | /* Handle repeated back references. If the length of the reference is |
2641 | zero, just continue with the main loop. If the length is negative, it |
2642 | means the reference is unset in non-Java-compatible mode. If the minimum is |
2643 | zero, we can continue at the same level without recursion. For any other |
2644 | minimum, carrying on will result in NOMATCH. */ |
2645 | |
2646 | if (length == 0) continue; |
2647 | if (length < 0 && min == 0) continue; |
2648 | |
2649 | /* First, ensure the minimum number of matches are present. We get back |
2650 | the length of the reference string explicitly rather than passing the |
2651 | address of eptr, so that eptr can be a register variable. */ |
2652 | |
2653 | for (i = 1; i <= min; i++) |
2654 | { |
2655 | int slength; |
2656 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2657 | { |
2658 | CHECK_PARTIAL(); |
2659 | RRETURN(MATCH_NOMATCH); |
2660 | } |
2661 | eptr += slength; |
2662 | } |
2663 | |
2664 | /* If min = max, continue at the same level without recursion. |
2665 | They are not both allowed to be zero. */ |
2666 | |
2667 | if (min == max) continue; |
2668 | |
2669 | /* If minimizing, keep trying and advancing the pointer */ |
2670 | |
2671 | if (minimize) |
2672 | { |
2673 | for (fi = min;; fi++) |
2674 | { |
2675 | int slength; |
2676 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM14); |
2677 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2678 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
2679 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2680 | { |
2681 | CHECK_PARTIAL(); |
2682 | RRETURN(MATCH_NOMATCH); |
2683 | } |
2684 | eptr += slength; |
2685 | } |
2686 | /* Control never gets here */ |
2687 | } |
2688 | |
2689 | /* If maximizing, find the longest string and work backwards */ |
2690 | |
2691 | else |
2692 | { |
2693 | pp = eptr; |
2694 | for (i = min; i < max; i++) |
2695 | { |
2696 | int slength; |
2697 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2698 | { |
2699 | CHECK_PARTIAL(); |
2700 | break; |
2701 | } |
2702 | eptr += slength; |
2703 | } |
2704 | while (eptr >= pp) |
2705 | { |
2706 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM15); |
2707 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2708 | eptr -= length; |
2709 | } |
2710 | RRETURN(MATCH_NOMATCH); |
2711 | } |
2712 | /* Control never gets here */ |
2713 | |
2714 | /* Match a bit-mapped character class, possibly repeatedly. This op code is |
2715 | used when all the characters in the class have values in the range 0-255, |
2716 | and either the matching is caseful, or the characters are in the range |
2717 | 0-127 when UTF-8 processing is enabled. The only difference between |
2718 | OP_CLASS and OP_NCLASS occurs when a data character outside the range is |
2719 | encountered. |
2720 | |
2721 | First, look past the end of the item to see if there is repeat information |
2722 | following. Then obey similar code to character type repeats - written out |
2723 | again for speed. */ |
2724 | |
2725 | case OP_NCLASS: |
2726 | case OP_CLASS: |
2727 | { |
2728 | /* The data variable is saved across frames, so the byte map needs to |
2729 | be stored there. */ |
2730 | #define BYTE_MAP ((pcre_uint8 *)data) |
2731 | data = ecode + 1; /* Save for matching */ |
2732 | ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */ |
2733 | |
2734 | switch (*ecode) |
2735 | { |
2736 | case OP_CRSTAR: |
2737 | case OP_CRMINSTAR: |
2738 | case OP_CRPLUS: |
2739 | case OP_CRMINPLUS: |
2740 | case OP_CRQUERY: |
2741 | case OP_CRMINQUERY: |
2742 | c = *ecode++ - OP_CRSTAR; |
2743 | minimize = (c & 1) != 0; |
2744 | min = rep_min[c]; /* Pick up values from tables; */ |
2745 | max = rep_max[c]; /* zero for max => infinity */ |
2746 | if (max == 0) max = INT_MAX; |
2747 | break; |
2748 | |
2749 | case OP_CRRANGE: |
2750 | case OP_CRMINRANGE: |
2751 | minimize = (*ecode == OP_CRMINRANGE); |
2752 | min = GET2(ecode, 1); |
2753 | max = GET2(ecode, 1 + IMM2_SIZE); |
2754 | if (max == 0) max = INT_MAX; |
2755 | ecode += 1 + 2 * IMM2_SIZE; |
2756 | break; |
2757 | |
2758 | default: /* No repeat follows */ |
2759 | min = max = 1; |
2760 | break; |
2761 | } |
2762 | |
2763 | /* First, ensure the minimum number of matches are present. */ |
2764 | |
2765 | #ifdef SUPPORT_UTF |
2766 | if (utf) |
2767 | { |
2768 | for (i = 1; i <= min; i++) |
2769 | { |
2770 | if (eptr >= md->end_subject) |
2771 | { |
2772 | SCHECK_PARTIAL(); |
2773 | RRETURN(MATCH_NOMATCH); |
2774 | } |
2775 | GETCHARINC(c, eptr); |
2776 | if (c > 255) |
2777 | { |
2778 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2779 | } |
2780 | else |
2781 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2782 | } |
2783 | } |
2784 | else |
2785 | #endif |
2786 | /* Not UTF mode */ |
2787 | { |
2788 | for (i = 1; i <= min; i++) |
2789 | { |
2790 | if (eptr >= md->end_subject) |
2791 | { |
2792 | SCHECK_PARTIAL(); |
2793 | RRETURN(MATCH_NOMATCH); |
2794 | } |
2795 | c = *eptr++; |
2796 | #ifndef COMPILE_PCRE8 |
2797 | if (c > 255) |
2798 | { |
2799 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2800 | } |
2801 | else |
2802 | #endif |
2803 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2804 | } |
2805 | } |
2806 | |
2807 | /* If max == min we can continue with the main loop without the |
2808 | need to recurse. */ |
2809 | |
2810 | if (min == max) continue; |
2811 | |
2812 | /* If minimizing, keep testing the rest of the expression and advancing |
2813 | the pointer while it matches the class. */ |
2814 | |
2815 | if (minimize) |
2816 | { |
2817 | #ifdef SUPPORT_UTF |
2818 | if (utf) |
2819 | { |
2820 | for (fi = min;; fi++) |
2821 | { |
2822 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM16); |
2823 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2824 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
2825 | if (eptr >= md->end_subject) |
2826 | { |
2827 | SCHECK_PARTIAL(); |
2828 | RRETURN(MATCH_NOMATCH); |
2829 | } |
2830 | GETCHARINC(c, eptr); |
2831 | if (c > 255) |
2832 | { |
2833 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2834 | } |
2835 | else |
2836 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2837 | } |
2838 | } |
2839 | else |
2840 | #endif |
2841 | /* Not UTF mode */ |
2842 | { |
2843 | for (fi = min;; fi++) |
2844 | { |
2845 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM17); |
2846 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2847 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
2848 | if (eptr >= md->end_subject) |
2849 | { |
2850 | SCHECK_PARTIAL(); |
2851 | RRETURN(MATCH_NOMATCH); |
2852 | } |
2853 | c = *eptr++; |
2854 | #ifndef COMPILE_PCRE8 |
2855 | if (c > 255) |
2856 | { |
2857 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2858 | } |
2859 | else |
2860 | #endif |
2861 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2862 | } |
2863 | } |
2864 | /* Control never gets here */ |
2865 | } |
2866 | |
2867 | /* If maximizing, find the longest possible run, then work backwards. */ |
2868 | |
2869 | else |
2870 | { |
2871 | pp = eptr; |
2872 | |
2873 | #ifdef SUPPORT_UTF |
2874 | if (utf) |
2875 | { |
2876 | for (i = min; i < max; i++) |
2877 | { |
2878 | int len = 1; |
2879 | if (eptr >= md->end_subject) |
2880 | { |
2881 | SCHECK_PARTIAL(); |
2882 | break; |
2883 | } |
2884 | GETCHARLEN(c, eptr, len); |
2885 | if (c > 255) |
2886 | { |
2887 | if (op == OP_CLASS) break; |
2888 | } |
2889 | else |
2890 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break; |
2891 | eptr += len; |
2892 | } |
2893 | for (;;) |
2894 | { |
2895 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM18); |
2896 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2897 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
2898 | BACKCHAR(eptr); |
2899 | } |
2900 | } |
2901 | else |
2902 | #endif |
2903 | /* Not UTF mode */ |
2904 | { |
2905 | for (i = min; i < max; i++) |
2906 | { |
2907 | if (eptr >= md->end_subject) |
2908 | { |
2909 | SCHECK_PARTIAL(); |
2910 | break; |
2911 | } |
2912 | c = *eptr; |
2913 | #ifndef COMPILE_PCRE8 |
2914 | if (c > 255) |
2915 | { |
2916 | if (op == OP_CLASS) break; |
2917 | } |
2918 | else |
2919 | #endif |
2920 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break; |
2921 | eptr++; |
2922 | } |
2923 | while (eptr >= pp) |
2924 | { |
2925 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM19); |
2926 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2927 | eptr--; |
2928 | } |
2929 | } |
2930 | |
2931 | RRETURN(MATCH_NOMATCH); |
2932 | } |
2933 | #undef BYTE_MAP |
2934 | } |
2935 | /* Control never gets here */ |
2936 | |
2937 | |
2938 | /* Match an extended character class. This opcode is encountered only |
2939 | when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8 |
2940 | mode, because Unicode properties are supported in non-UTF-8 mode. */ |
2941 | |
2942 | #if defined SUPPORT_UTF || !defined COMPILE_PCRE8 |
2943 | case OP_XCLASS: |
2944 | { |
2945 | data = ecode + 1 + LINK_SIZE; /* Save for matching */ |
2946 | ecode += GET(ecode, 1); /* Advance past the item */ |
2947 | |
2948 | switch (*ecode) |
2949 | { |
2950 | case OP_CRSTAR: |
2951 | case OP_CRMINSTAR: |
2952 | case OP_CRPLUS: |
2953 | case OP_CRMINPLUS: |
2954 | case OP_CRQUERY: |
2955 | case OP_CRMINQUERY: |
2956 | c = *ecode++ - OP_CRSTAR; |
2957 | minimize = (c & 1) != 0; |
2958 | min = rep_min[c]; /* Pick up values from tables; */ |
2959 | max = rep_max[c]; /* zero for max => infinity */ |
2960 | if (max == 0) max = INT_MAX; |
2961 | break; |
2962 | |
2963 | case OP_CRRANGE: |
2964 | case OP_CRMINRANGE: |
2965 | minimize = (*ecode == OP_CRMINRANGE); |
2966 | min = GET2(ecode, 1); |
2967 | max = GET2(ecode, 1 + IMM2_SIZE); |
2968 | if (max == 0) max = INT_MAX; |
2969 | ecode += 1 + 2 * IMM2_SIZE; |
2970 | break; |
2971 | |
2972 | default: /* No repeat follows */ |
2973 | min = max = 1; |
2974 | break; |
2975 | } |
2976 | |
2977 | /* First, ensure the minimum number of matches are present. */ |
2978 | |
2979 | for (i = 1; i <= min; i++) |
2980 | { |
2981 | if (eptr >= md->end_subject) |
2982 | { |
2983 | SCHECK_PARTIAL(); |
2984 | RRETURN(MATCH_NOMATCH); |
2985 | } |
2986 | GETCHARINCTEST(c, eptr); |
2987 | if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH); |
2988 | } |
2989 | |
2990 | /* If max == min we can continue with the main loop without the |
2991 | need to recurse. */ |
2992 | |
2993 | if (min == max) continue; |
2994 | |
2995 | /* If minimizing, keep testing the rest of the expression and advancing |
2996 | the pointer while it matches the class. */ |
2997 | |
2998 | if (minimize) |
2999 | { |
3000 | for (fi = min;; fi++) |
3001 | { |
3002 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM20); |
3003 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3004 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3005 | if (eptr >= md->end_subject) |
3006 | { |
3007 | SCHECK_PARTIAL(); |
3008 | RRETURN(MATCH_NOMATCH); |
3009 | } |
3010 | GETCHARINCTEST(c, eptr); |
3011 | if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH); |
3012 | } |
3013 | /* Control never gets here */ |
3014 | } |
3015 | |
3016 | /* If maximizing, find the longest possible run, then work backwards. */ |
3017 | |
3018 | else |
3019 | { |
3020 | pp = eptr; |
3021 | for (i = min; i < max; i++) |
3022 | { |
3023 | int len = 1; |
3024 | if (eptr >= md->end_subject) |
3025 | { |
3026 | SCHECK_PARTIAL(); |
3027 | break; |
3028 | } |
3029 | #ifdef SUPPORT_UTF |
3030 | GETCHARLENTEST(c, eptr, len); |
3031 | #else |
3032 | c = *eptr; |
3033 | #endif |
3034 | if (!PRIV(xclass)(c, data, utf)) break; |
3035 | eptr += len; |
3036 | } |
3037 | for(;;) |
3038 | { |
3039 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM21); |
3040 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3041 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3042 | #ifdef SUPPORT_UTF |
3043 | if (utf) BACKCHAR(eptr); |
3044 | #endif |
3045 | } |
3046 | RRETURN(MATCH_NOMATCH); |
3047 | } |
3048 | |
3049 | /* Control never gets here */ |
3050 | } |
3051 | #endif /* End of XCLASS */ |
3052 | |
3053 | /* Match a single character, casefully */ |
3054 | |
3055 | case OP_CHAR: |
3056 | #ifdef SUPPORT_UTF |
3057 | if (utf) |
3058 | { |
3059 | length = 1; |
3060 | ecode++; |
3061 | GETCHARLEN(fc, ecode, length); |
3062 | if (length > md->end_subject - eptr) |
3063 | { |
3064 | CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */ |
3065 | RRETURN(MATCH_NOMATCH); |
3066 | } |
3067 | while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH); |
3068 | } |
3069 | else |
3070 | #endif |
3071 | /* Not UTF mode */ |
3072 | { |
3073 | if (md->end_subject - eptr < 1) |
3074 | { |
3075 | SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */ |
3076 | RRETURN(MATCH_NOMATCH); |
3077 | } |
3078 | if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH); |
3079 | ecode += 2; |
3080 | } |
3081 | break; |
3082 | |
3083 | /* Match a single character, caselessly. If we are at the end of the |
3084 | subject, give up immediately. */ |
3085 | |
3086 | case OP_CHARI: |
3087 | if (eptr >= md->end_subject) |
3088 | { |
3089 | SCHECK_PARTIAL(); |
3090 | RRETURN(MATCH_NOMATCH); |
3091 | } |
3092 | |
3093 | #ifdef SUPPORT_UTF |
3094 | if (utf) |
3095 | { |
3096 | length = 1; |
3097 | ecode++; |
3098 | GETCHARLEN(fc, ecode, length); |
3099 | |
3100 | /* If the pattern character's value is < 128, we have only one byte, and |
3101 | we know that its other case must also be one byte long, so we can use the |
3102 | fast lookup table. We know that there is at least one byte left in the |
3103 | subject. */ |
3104 | |
3105 | if (fc < 128) |
3106 | { |
3107 | if (md->lcc[fc] |
3108 | != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH); |
3109 | ecode++; |
3110 | eptr++; |
3111 | } |
3112 | |
3113 | /* Otherwise we must pick up the subject character. Note that we cannot |
3114 | use the value of "length" to check for sufficient bytes left, because the |
3115 | other case of the character may have more or fewer bytes. */ |
3116 | |
3117 | else |
3118 | { |
3119 | unsigned int dc; |
3120 | GETCHARINC(dc, eptr); |
3121 | ecode += length; |
3122 | |
3123 | /* If we have Unicode property support, we can use it to test the other |
3124 | case of the character, if there is one. */ |
3125 | |
3126 | if (fc != dc) |
3127 | { |
3128 | #ifdef SUPPORT_UCP |
3129 | if (dc != UCD_OTHERCASE(fc)) |
3130 | #endif |
3131 | RRETURN(MATCH_NOMATCH); |
3132 | } |
3133 | } |
3134 | } |
3135 | else |
3136 | #endif /* SUPPORT_UTF */ |
3137 | |
3138 | /* Not UTF mode */ |
3139 | { |
3140 | if (TABLE_GET(ecode[1], md->lcc, ecode[1]) |
3141 | != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH); |
3142 | eptr++; |
3143 | ecode += 2; |
3144 | } |
3145 | break; |
3146 | |
3147 | /* Match a single character repeatedly. */ |
3148 | |
3149 | case OP_EXACT: |
3150 | case OP_EXACTI: |
3151 | min = max = GET2(ecode, 1); |
3152 | ecode += 1 + IMM2_SIZE; |
3153 | goto REPEATCHAR; |
3154 | |
3155 | case OP_POSUPTO: |
3156 | case OP_POSUPTOI: |
3157 | possessive = TRUE; |
3158 | /* Fall through */ |
3159 | |
3160 | case OP_UPTO: |
3161 | case OP_UPTOI: |
3162 | case OP_MINUPTO: |
3163 | case OP_MINUPTOI: |
3164 | min = 0; |
3165 | max = GET2(ecode, 1); |
3166 | minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI; |
3167 | ecode += 1 + IMM2_SIZE; |
3168 | goto REPEATCHAR; |
3169 | |
3170 | case OP_POSSTAR: |
3171 | case OP_POSSTARI: |
3172 | possessive = TRUE; |
3173 | min = 0; |
3174 | max = INT_MAX; |
3175 | ecode++; |
3176 | goto REPEATCHAR; |
3177 | |
3178 | case OP_POSPLUS: |
3179 | case OP_POSPLUSI: |
3180 | possessive = TRUE; |
3181 | min = 1; |
3182 | max = INT_MAX; |
3183 | ecode++; |
3184 | goto REPEATCHAR; |
3185 | |
3186 | case OP_POSQUERY: |
3187 | case OP_POSQUERYI: |
3188 | possessive = TRUE; |
3189 | min = 0; |
3190 | max = 1; |
3191 | ecode++; |
3192 | goto REPEATCHAR; |
3193 | |
3194 | case OP_STAR: |
3195 | case OP_STARI: |
3196 | case OP_MINSTAR: |
3197 | case OP_MINSTARI: |
3198 | case OP_PLUS: |
3199 | case OP_PLUSI: |
3200 | case OP_MINPLUS: |
3201 | case OP_MINPLUSI: |
3202 | case OP_QUERY: |
3203 | case OP_QUERYI: |
3204 | case OP_MINQUERY: |
3205 | case OP_MINQUERYI: |
3206 | c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI); |
3207 | minimize = (c & 1) != 0; |
3208 | min = rep_min[c]; /* Pick up values from tables; */ |
3209 | max = rep_max[c]; /* zero for max => infinity */ |
3210 | if (max == 0) max = INT_MAX; |
3211 | |
3212 | /* Common code for all repeated single-character matches. */ |
3213 | |
3214 | REPEATCHAR: |
3215 | #ifdef SUPPORT_UTF |
3216 | if (utf) |
3217 | { |
3218 | length = 1; |
3219 | charptr = ecode; |
3220 | GETCHARLEN(fc, ecode, length); |
3221 | ecode += length; |
3222 | |
3223 | /* Handle multibyte character matching specially here. There is |
3224 | support for caseless matching if UCP support is present. */ |
3225 | |
3226 | if (length > 1) |
3227 | { |
3228 | #ifdef SUPPORT_UCP |
3229 | unsigned int othercase; |
3230 | if (op >= OP_STARI && /* Caseless */ |
3231 | (othercase = UCD_OTHERCASE(fc)) != fc) |
3232 | oclength = PRIV(ord2utf)(othercase, occhars); |
3233 | else oclength = 0; |
3234 | #endif /* SUPPORT_UCP */ |
3235 | |
3236 | for (i = 1; i <= min; i++) |
3237 | { |
3238 | if (eptr <= md->end_subject - length && |
3239 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3240 | #ifdef SUPPORT_UCP |
3241 | else if (oclength > 0 && |
3242 | eptr <= md->end_subject - oclength && |
3243 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3244 | #endif /* SUPPORT_UCP */ |
3245 | else |
3246 | { |
3247 | CHECK_PARTIAL(); |
3248 | RRETURN(MATCH_NOMATCH); |
3249 | } |
3250 | } |
3251 | |
3252 | if (min == max) continue; |
3253 | |
3254 | if (minimize) |
3255 | { |
3256 | for (fi = min;; fi++) |
3257 | { |
3258 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM22); |
3259 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3260 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3261 | if (eptr <= md->end_subject - length && |
3262 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3263 | #ifdef SUPPORT_UCP |
3264 | else if (oclength > 0 && |
3265 | eptr <= md->end_subject - oclength && |
3266 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3267 | #endif /* SUPPORT_UCP */ |
3268 | else |
3269 | { |
3270 | CHECK_PARTIAL(); |
3271 | RRETURN(MATCH_NOMATCH); |
3272 | } |
3273 | } |
3274 | /* Control never gets here */ |
3275 | } |
3276 | |
3277 | else /* Maximize */ |
3278 | { |
3279 | pp = eptr; |
3280 | for (i = min; i < max; i++) |
3281 | { |
3282 | if (eptr <= md->end_subject - length && |
3283 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3284 | #ifdef SUPPORT_UCP |
3285 | else if (oclength > 0 && |
3286 | eptr <= md->end_subject - oclength && |
3287 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3288 | #endif /* SUPPORT_UCP */ |
3289 | else |
3290 | { |
3291 | CHECK_PARTIAL(); |
3292 | break; |
3293 | } |
3294 | } |
3295 | |
3296 | if (possessive) continue; |
3297 | |
3298 | for(;;) |
3299 | { |
3300 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM23); |
3301 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3302 | if (eptr == pp) { RRETURN(MATCH_NOMATCH); } |
3303 | #ifdef SUPPORT_UCP |
3304 | eptr--; |
3305 | BACKCHAR(eptr); |
3306 | #else /* without SUPPORT_UCP */ |
3307 | eptr -= length; |
3308 | #endif /* SUPPORT_UCP */ |
3309 | } |
3310 | } |
3311 | /* Control never gets here */ |
3312 | } |
3313 | |
3314 | /* If the length of a UTF-8 character is 1, we fall through here, and |
3315 | obey the code as for non-UTF-8 characters below, though in this case the |
3316 | value of fc will always be < 128. */ |
3317 | } |
3318 | else |
3319 | #endif /* SUPPORT_UTF */ |
3320 | /* When not in UTF-8 mode, load a single-byte character. */ |
3321 | fc = *ecode++; |
3322 | |
3323 | /* The value of fc at this point is always one character, though we may |
3324 | or may not be in UTF mode. The code is duplicated for the caseless and |
3325 | caseful cases, for speed, since matching characters is likely to be quite |
3326 | common. First, ensure the minimum number of matches are present. If min = |
3327 | max, continue at the same level without recursing. Otherwise, if |
3328 | minimizing, keep trying the rest of the expression and advancing one |
3329 | matching character if failing, up to the maximum. Alternatively, if |
3330 | maximizing, find the maximum number of characters and work backwards. */ |
3331 | |
3332 | DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max, |
3333 | max, eptr)); |
3334 | |
3335 | if (op >= OP_STARI) /* Caseless */ |
3336 | { |
3337 | #ifdef COMPILE_PCRE8 |
3338 | /* fc must be < 128 if UTF is enabled. */ |
3339 | foc = md->fcc[fc]; |
3340 | #else |
3341 | #ifdef SUPPORT_UTF |
3342 | #ifdef SUPPORT_UCP |
3343 | if (utf && fc > 127) |
3344 | foc = UCD_OTHERCASE(fc); |
3345 | #else |
3346 | if (utf && fc > 127) |
3347 | foc = fc; |
3348 | #endif /* SUPPORT_UCP */ |
3349 | else |
3350 | #endif /* SUPPORT_UTF */ |
3351 | foc = TABLE_GET(fc, md->fcc, fc); |
3352 | #endif /* COMPILE_PCRE8 */ |
3353 | |
3354 | for (i = 1; i <= min; i++) |
3355 | { |
3356 | if (eptr >= md->end_subject) |
3357 | { |
3358 | SCHECK_PARTIAL(); |
3359 | RRETURN(MATCH_NOMATCH); |
3360 | } |
3361 | if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH); |
3362 | eptr++; |
3363 | } |
3364 | if (min == max) continue; |
3365 | if (minimize) |
3366 | { |
3367 | for (fi = min;; fi++) |
3368 | { |
3369 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM24); |
3370 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3371 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3372 | if (eptr >= md->end_subject) |
3373 | { |
3374 | SCHECK_PARTIAL(); |
3375 | RRETURN(MATCH_NOMATCH); |
3376 | } |
3377 | if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH); |
3378 | eptr++; |
3379 | } |
3380 | /* Control never gets here */ |
3381 | } |
3382 | else /* Maximize */ |
3383 | { |
3384 | pp = eptr; |
3385 | for (i = min; i < max; i++) |
3386 | { |
3387 | if (eptr >= md->end_subject) |
3388 | { |
3389 | SCHECK_PARTIAL(); |
3390 | break; |
3391 | } |
3392 | if (fc != *eptr && foc != *eptr) break; |
3393 | eptr++; |
3394 | } |
3395 | |
3396 | if (possessive) continue; |
3397 | |
3398 | while (eptr >= pp) |
3399 | { |
3400 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM25); |
3401 | eptr--; |
3402 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3403 | } |
3404 | RRETURN(MATCH_NOMATCH); |
3405 | } |
3406 | /* Control never gets here */ |
3407 | } |
3408 | |
3409 | /* Caseful comparisons (includes all multi-byte characters) */ |
3410 | |
3411 | else |
3412 | { |
3413 | for (i = 1; i <= min; i++) |
3414 | { |
3415 | if (eptr >= md->end_subject) |
3416 | { |
3417 | SCHECK_PARTIAL(); |
3418 | RRETURN(MATCH_NOMATCH); |
3419 | } |
3420 | if (fc != *eptr++) RRETURN(MATCH_NOMATCH); |
3421 | } |
3422 | |
3423 | if (min == max) continue; |
3424 | |
3425 | if (minimize) |
3426 | { |
3427 | for (fi = min;; fi++) |
3428 | { |
3429 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM26); |
3430 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3431 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3432 | if (eptr >= md->end_subject) |
3433 | { |
3434 | SCHECK_PARTIAL(); |
3435 | RRETURN(MATCH_NOMATCH); |
3436 | } |
3437 | if (fc != *eptr++) RRETURN(MATCH_NOMATCH); |
3438 | } |
3439 | /* Control never gets here */ |
3440 | } |
3441 | else /* Maximize */ |
3442 | { |
3443 | pp = eptr; |
3444 | for (i = min; i < max; i++) |
3445 | { |
3446 | if (eptr >= md->end_subject) |
3447 | { |
3448 | SCHECK_PARTIAL(); |
3449 | break; |
3450 | } |
3451 | if (fc != *eptr) break; |
3452 | eptr++; |
3453 | } |
3454 | if (possessive) continue; |
3455 | |
3456 | while (eptr >= pp) |
3457 | { |
3458 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM27); |
3459 | eptr--; |
3460 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3461 | } |
3462 | RRETURN(MATCH_NOMATCH); |
3463 | } |
3464 | } |
3465 | /* Control never gets here */ |
3466 | |
3467 | /* Match a negated single one-byte character. The character we are |
3468 | checking can be multibyte. */ |
3469 | |
3470 | case OP_NOT: |
3471 | case OP_NOTI: |
3472 | if (eptr >= md->end_subject) |
3473 | { |
3474 | SCHECK_PARTIAL(); |
3475 | RRETURN(MATCH_NOMATCH); |
3476 | } |
3477 | ecode++; |
3478 | GETCHARINCTEST(c, eptr); |
3479 | if (op == OP_NOTI) /* The caseless case */ |
3480 | { |
3481 | register int ch, och; |
3482 | ch = *ecode++; |
3483 | #ifdef COMPILE_PCRE8 |
3484 | /* ch must be < 128 if UTF is enabled. */ |
3485 | och = md->fcc[ch]; |
3486 | #else |
3487 | #ifdef SUPPORT_UTF |
3488 | #ifdef SUPPORT_UCP |
3489 | if (utf && ch > 127) |
3490 | och = UCD_OTHERCASE(ch); |
3491 | #else |
3492 | if (utf && ch > 127) |
3493 | och = ch; |
3494 | #endif /* SUPPORT_UCP */ |
3495 | else |
3496 | #endif /* SUPPORT_UTF */ |
3497 | och = TABLE_GET(ch, md->fcc, ch); |
3498 | #endif /* COMPILE_PCRE8 */ |
3499 | if (ch == c || och == c) RRETURN(MATCH_NOMATCH); |
3500 | } |
3501 | else /* Caseful */ |
3502 | { |
3503 | if (*ecode++ == c) RRETURN(MATCH_NOMATCH); |
3504 | } |
3505 | break; |
3506 | |
3507 | /* Match a negated single one-byte character repeatedly. This is almost a |
3508 | repeat of the code for a repeated single character, but I haven't found a |
3509 | nice way of commoning these up that doesn't require a test of the |
3510 | positive/negative option for each character match. Maybe that wouldn't add |
3511 | very much to the time taken, but character matching *is* what this is all |
3512 | about... */ |
3513 | |
3514 | case OP_NOTEXACT: |
3515 | case OP_NOTEXACTI: |
3516 | min = max = GET2(ecode, 1); |
3517 | ecode += 1 + IMM2_SIZE; |
3518 | goto REPEATNOTCHAR; |
3519 | |
3520 | case OP_NOTUPTO: |
3521 | case OP_NOTUPTOI: |
3522 | case OP_NOTMINUPTO: |
3523 | case OP_NOTMINUPTOI: |
3524 | min = 0; |
3525 | max = GET2(ecode, 1); |
3526 | minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI; |
3527 | ecode += 1 + IMM2_SIZE; |
3528 | goto REPEATNOTCHAR; |
3529 | |
3530 | case OP_NOTPOSSTAR: |
3531 | case OP_NOTPOSSTARI: |
3532 | possessive = TRUE; |
3533 | min = 0; |
3534 | max = INT_MAX; |
3535 | ecode++; |
3536 | goto REPEATNOTCHAR; |
3537 | |
3538 | case OP_NOTPOSPLUS: |
3539 | case OP_NOTPOSPLUSI: |
3540 | possessive = TRUE; |
3541 | min = 1; |
3542 | max = INT_MAX; |
3543 | ecode++; |
3544 | goto REPEATNOTCHAR; |
3545 | |
3546 | case OP_NOTPOSQUERY: |
3547 | case OP_NOTPOSQUERYI: |
3548 | possessive = TRUE; |
3549 | min = 0; |
3550 | max = 1; |
3551 | ecode++; |
3552 | goto REPEATNOTCHAR; |
3553 | |
3554 | case OP_NOTPOSUPTO: |
3555 | case OP_NOTPOSUPTOI: |
3556 | possessive = TRUE; |
3557 | min = 0; |
3558 | max = GET2(ecode, 1); |
3559 | ecode += 1 + IMM2_SIZE; |
3560 | goto REPEATNOTCHAR; |
3561 | |
3562 | case OP_NOTSTAR: |
3563 | case OP_NOTSTARI: |
3564 | case OP_NOTMINSTAR: |
3565 | case OP_NOTMINSTARI: |
3566 | case OP_NOTPLUS: |
3567 | case OP_NOTPLUSI: |
3568 | case OP_NOTMINPLUS: |
3569 | case OP_NOTMINPLUSI: |
3570 | case OP_NOTQUERY: |
3571 | case OP_NOTQUERYI: |
3572 | case OP_NOTMINQUERY: |
3573 | case OP_NOTMINQUERYI: |
3574 | c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR); |
3575 | minimize = (c & 1) != 0; |
3576 | min = rep_min[c]; /* Pick up values from tables; */ |
3577 | max = rep_max[c]; /* zero for max => infinity */ |
3578 | if (max == 0) max = INT_MAX; |
3579 | |
3580 | /* Common code for all repeated single-byte matches. */ |
3581 | |
3582 | REPEATNOTCHAR: |
3583 | fc = *ecode++; |
3584 | |
3585 | /* The code is duplicated for the caseless and caseful cases, for speed, |
3586 | since matching characters is likely to be quite common. First, ensure the |
3587 | minimum number of matches are present. If min = max, continue at the same |
3588 | level without recursing. Otherwise, if minimizing, keep trying the rest of |
3589 | the expression and advancing one matching character if failing, up to the |
3590 | maximum. Alternatively, if maximizing, find the maximum number of |
3591 | characters and work backwards. */ |
3592 | |
3593 | DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max, |
3594 | max, eptr)); |
3595 | |
3596 | if (op >= OP_NOTSTARI) /* Caseless */ |
3597 | { |
3598 | #ifdef COMPILE_PCRE8 |
3599 | /* fc must be < 128 if UTF is enabled. */ |
3600 | foc = md->fcc[fc]; |
3601 | #else |
3602 | #ifdef SUPPORT_UTF |
3603 | #ifdef SUPPORT_UCP |
3604 | if (utf && fc > 127) |
3605 | foc = UCD_OTHERCASE(fc); |
3606 | #else |
3607 | if (utf && fc > 127) |
3608 | foc = fc; |
3609 | #endif /* SUPPORT_UCP */ |
3610 | else |
3611 | #endif /* SUPPORT_UTF */ |
3612 | foc = TABLE_GET(fc, md->fcc, fc); |
3613 | #endif /* COMPILE_PCRE8 */ |
3614 | |
3615 | #ifdef SUPPORT_UTF |
3616 | if (utf) |
3617 | { |
3618 | register unsigned int d; |
3619 | for (i = 1; i <= min; i++) |
3620 | { |
3621 | if (eptr >= md->end_subject) |
3622 | { |
3623 | SCHECK_PARTIAL(); |
3624 | RRETURN(MATCH_NOMATCH); |
3625 | } |
3626 | GETCHARINC(d, eptr); |
3627 | if (fc == d || foc == d) RRETURN(MATCH_NOMATCH); |
3628 | } |
3629 | } |
3630 | else |
3631 | #endif |
3632 | /* Not UTF mode */ |
3633 | { |
3634 | for (i = 1; i <= min; i++) |
3635 | { |
3636 | if (eptr >= md->end_subject) |
3637 | { |
3638 | SCHECK_PARTIAL(); |
3639 | RRETURN(MATCH_NOMATCH); |
3640 | } |
3641 | if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH); |
3642 | eptr++; |
3643 | } |
3644 | } |
3645 | |
3646 | if (min == max) continue; |
3647 | |
3648 | if (minimize) |
3649 | { |
3650 | #ifdef SUPPORT_UTF |
3651 | if (utf) |
3652 | { |
3653 | register unsigned int d; |
3654 | for (fi = min;; fi++) |
3655 | { |
3656 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM28); |
3657 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3658 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3659 | if (eptr >= md->end_subject) |
3660 | { |
3661 | SCHECK_PARTIAL(); |
3662 | RRETURN(MATCH_NOMATCH); |
3663 | } |
3664 | GETCHARINC(d, eptr); |
3665 | if (fc == d || foc == d) RRETURN(MATCH_NOMATCH); |
3666 | } |
3667 | } |
3668 | else |
3669 | #endif |
3670 | /* Not UTF mode */ |
3671 | { |
3672 | for (fi = min;; fi++) |
3673 | { |
3674 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM29); |
3675 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3676 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3677 | if (eptr >= md->end_subject) |
3678 | { |
3679 | SCHECK_PARTIAL(); |
3680 | RRETURN(MATCH_NOMATCH); |
3681 | } |
3682 | if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH); |
3683 | eptr++; |
3684 | } |
3685 | } |
3686 | /* Control never gets here */ |
3687 | } |
3688 | |
3689 | /* Maximize case */ |
3690 | |
3691 | else |
3692 | { |
3693 | pp = eptr; |
3694 | |
3695 | #ifdef SUPPORT_UTF |
3696 | if (utf) |
3697 | { |
3698 | register unsigned int d; |
3699 | for (i = min; i < max; i++) |
3700 | { |
3701 | int len = 1; |
3702 | if (eptr >= md->end_subject) |
3703 | { |
3704 | SCHECK_PARTIAL(); |
3705 | break; |
3706 | } |
3707 | GETCHARLEN(d, eptr, len); |
3708 | if (fc == d || foc == d) break; |
3709 | eptr += len; |
3710 | } |
3711 | if (possessive) continue; |
3712 | for(;;) |
3713 | { |
3714 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM30); |
3715 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3716 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3717 | BACKCHAR(eptr); |
3718 | } |
3719 | } |
3720 | else |
3721 | #endif |
3722 | /* Not UTF mode */ |
3723 | { |
3724 | for (i = min; i < max; i++) |
3725 | { |
3726 | if (eptr >= md->end_subject) |
3727 | { |
3728 | SCHECK_PARTIAL(); |
3729 | break; |
3730 | } |
3731 | if (fc == *eptr || foc == *eptr) break; |
3732 | eptr++; |
3733 | } |
3734 | if (possessive) continue; |
3735 | while (eptr >= pp) |
3736 | { |
3737 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM31); |
3738 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3739 | eptr--; |
3740 | } |
3741 | } |
3742 | |
3743 | RRETURN(MATCH_NOMATCH); |
3744 | } |
3745 | /* Control never gets here */ |
3746 | } |
3747 | |
3748 | /* Caseful comparisons */ |
3749 | |
3750 | else |
3751 | { |
3752 | #ifdef SUPPORT_UTF |
3753 | if (utf) |
3754 | { |
3755 | register unsigned int d; |
3756 | for (i = 1; i <= min; i++) |
3757 | { |
3758 | if (eptr >= md->end_subject) |
3759 | { |
3760 | SCHECK_PARTIAL(); |
3761 | RRETURN(MATCH_NOMATCH); |
3762 | } |
3763 | GETCHARINC(d, eptr); |
3764 | if (fc == d) RRETURN(MATCH_NOMATCH); |
3765 | } |
3766 | } |
3767 | else |
3768 | #endif |
3769 | /* Not UTF mode */ |
3770 | { |
3771 | for (i = 1; i <= min; i++) |
3772 | { |
3773 | if (eptr >= md->end_subject) |
3774 | { |
3775 | SCHECK_PARTIAL(); |
3776 | RRETURN(MATCH_NOMATCH); |
3777 | } |
3778 | if (fc == *eptr++) RRETURN(MATCH_NOMATCH); |
3779 | } |
3780 | } |
3781 | |
3782 | if (min == max) continue; |
3783 | |
3784 | if (minimize) |
3785 | { |
3786 | #ifdef SUPPORT_UTF |
3787 | if (utf) |
3788 | { |
3789 | register unsigned int d; |
3790 | for (fi = min;; fi++) |
3791 | { |
3792 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM32); |
3793 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3794 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3795 | if (eptr >= md->end_subject) |
3796 | { |
3797 | SCHECK_PARTIAL(); |
3798 | RRETURN(MATCH_NOMATCH); |
3799 | } |
3800 | GETCHARINC(d, eptr); |
3801 | if (fc == d) RRETURN(MATCH_NOMATCH); |
3802 | } |
3803 | } |
3804 | else |
3805 | #endif |
3806 | /* Not UTF mode */ |
3807 | { |
3808 | for (fi = min;; fi++) |
3809 | { |
3810 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM33); |
3811 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3812 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3813 | if (eptr >= md->end_subject) |
3814 | { |
3815 | SCHECK_PARTIAL(); |
3816 | RRETURN(MATCH_NOMATCH); |
3817 | } |
3818 | if (fc == *eptr++) RRETURN(MATCH_NOMATCH); |
3819 | } |
3820 | } |
3821 | /* Control never gets here */ |
3822 | } |
3823 | |
3824 | /* Maximize case */ |
3825 | |
3826 | else |
3827 | { |
3828 | pp = eptr; |
3829 | |
3830 | #ifdef SUPPORT_UTF |
3831 | if (utf) |
3832 | { |
3833 | register unsigned int d; |
3834 | for (i = min; i < max; i++) |
3835 | { |
3836 | int len = 1; |
3837 | if (eptr >= md->end_subject) |
3838 | { |
3839 | SCHECK_PARTIAL(); |
3840 | break; |
3841 | } |
3842 | GETCHARLEN(d, eptr, len); |
3843 | if (fc == d) break; |
3844 | eptr += len; |
3845 | } |
3846 | if (possessive) continue; |
3847 | for(;;) |
3848 | { |
3849 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM34); |
3850 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3851 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3852 | BACKCHAR(eptr); |
3853 | } |
3854 | } |
3855 | else |
3856 | #endif |
3857 | /* Not UTF mode */ |
3858 | { |
3859 | for (i = min; i < max; i++) |
3860 | { |
3861 | if (eptr >= md->end_subject) |
3862 | { |
3863 | SCHECK_PARTIAL(); |
3864 | break; |
3865 | } |
3866 | if (fc == *eptr) break; |
3867 | eptr++; |
3868 | } |
3869 | if (possessive) continue; |
3870 | while (eptr >= pp) |
3871 | { |
3872 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM35); |
3873 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3874 | eptr--; |
3875 | } |
3876 | } |
3877 | |
3878 | RRETURN(MATCH_NOMATCH); |
3879 | } |
3880 | } |
3881 | /* Control never gets here */ |
3882 | |
3883 | /* Match a single character type repeatedly; several different opcodes |
3884 | share code. This is very similar to the code for single characters, but we |
3885 | repeat it in the interests of efficiency. */ |
3886 | |
3887 | case OP_TYPEEXACT: |
3888 | min = max = GET2(ecode, 1); |
3889 | minimize = TRUE; |
3890 | ecode += 1 + IMM2_SIZE; |
3891 | goto REPEATTYPE; |
3892 | |
3893 | case OP_TYPEUPTO: |
3894 | case OP_TYPEMINUPTO: |
3895 | min = 0; |
3896 | max = GET2(ecode, 1); |
3897 | minimize = *ecode == OP_TYPEMINUPTO; |
3898 | ecode += 1 + IMM2_SIZE; |
3899 | goto REPEATTYPE; |
3900 | |
3901 | case OP_TYPEPOSSTAR: |
3902 | possessive = TRUE; |
3903 | min = 0; |
3904 | max = INT_MAX; |
3905 | ecode++; |
3906 | goto REPEATTYPE; |
3907 | |
3908 | case OP_TYPEPOSPLUS: |
3909 | possessive = TRUE; |
3910 | min = 1; |
3911 | max = INT_MAX; |
3912 | ecode++; |
3913 | goto REPEATTYPE; |
3914 | |
3915 | case OP_TYPEPOSQUERY: |
3916 | possessive = TRUE; |
3917 | min = 0; |
3918 | max = 1; |
3919 | ecode++; |
3920 | goto REPEATTYPE; |
3921 | |
3922 | case OP_TYPEPOSUPTO: |
3923 | possessive = TRUE; |
3924 | min = 0; |
3925 | max = GET2(ecode, 1); |
3926 | ecode += 1 + IMM2_SIZE; |
3927 | goto REPEATTYPE; |
3928 | |
3929 | case OP_TYPESTAR: |
3930 | case OP_TYPEMINSTAR: |
3931 | case OP_TYPEPLUS: |
3932 | case OP_TYPEMINPLUS: |
3933 | case OP_TYPEQUERY: |
3934 | case OP_TYPEMINQUERY: |
3935 | c = *ecode++ - OP_TYPESTAR; |
3936 | minimize = (c & 1) != 0; |
3937 | min = rep_min[c]; /* Pick up values from tables; */ |
3938 | max = rep_max[c]; /* zero for max => infinity */ |
3939 | if (max == 0) max = INT_MAX; |
3940 | |
3941 | /* Common code for all repeated single character type matches. Note that |
3942 | in UTF-8 mode, '.' matches a character of any length, but for the other |
3943 | character types, the valid characters are all one-byte long. */ |
3944 | |
3945 | REPEATTYPE: |
3946 | ctype = *ecode++; /* Code for the character type */ |
3947 | |
3948 | #ifdef SUPPORT_UCP |
3949 | if (ctype == OP_PROP || ctype == OP_NOTPROP) |
3950 | { |
3951 | prop_fail_result = ctype == OP_NOTPROP; |
3952 | prop_type = *ecode++; |
3953 | prop_value = *ecode++; |
3954 | } |
3955 | else prop_type = -1; |
3956 | #endif |
3957 | |
3958 | /* First, ensure the minimum number of matches are present. Use inline |
3959 | code for maximizing the speed, and do the type test once at the start |
3960 | (i.e. keep it out of the loop). Separate the UTF-8 code completely as that |
3961 | is tidier. Also separate the UCP code, which can be the same for both UTF-8 |
3962 | and single-bytes. */ |
3963 | |
3964 | if (min > 0) |
3965 | { |
3966 | #ifdef SUPPORT_UCP |
3967 | if (prop_type >= 0) |
3968 | { |
3969 | switch(prop_type) |
3970 | { |
3971 | case PT_ANY: |
3972 | if (prop_fail_result) RRETURN(MATCH_NOMATCH); |
3973 | for (i = 1; i <= min; i++) |
3974 | { |
3975 | if (eptr >= md->end_subject) |
3976 | { |
3977 | SCHECK_PARTIAL(); |
3978 | RRETURN(MATCH_NOMATCH); |
3979 | } |
3980 | GETCHARINCTEST(c, eptr); |
3981 | } |
3982 | break; |
3983 | |
3984 | case PT_LAMP: |
3985 | for (i = 1; i <= min; i++) |
3986 | { |
3987 | int chartype; |
3988 | if (eptr >= md->end_subject) |
3989 | { |
3990 | SCHECK_PARTIAL(); |
3991 | RRETURN(MATCH_NOMATCH); |
3992 | } |
3993 | GETCHARINCTEST(c, eptr); |
3994 | chartype = UCD_CHARTYPE(c); |
3995 | if ((chartype == ucp_Lu || |
3996 | chartype == ucp_Ll || |
3997 | chartype == ucp_Lt) == prop_fail_result) |
3998 | RRETURN(MATCH_NOMATCH); |
3999 | } |
4000 | break; |
4001 | |
4002 | case PT_GC: |
4003 | for (i = 1; i <= min; i++) |
4004 | { |
4005 | if (eptr >= md->end_subject) |
4006 | { |
4007 | SCHECK_PARTIAL(); |
4008 | RRETURN(MATCH_NOMATCH); |
4009 | } |
4010 | GETCHARINCTEST(c, eptr); |
4011 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) |
4012 | RRETURN(MATCH_NOMATCH); |
4013 | } |
4014 | break; |
4015 | |
4016 | case PT_PC: |
4017 | for (i = 1; i <= min; i++) |
4018 | { |
4019 | if (eptr >= md->end_subject) |
4020 | { |
4021 | SCHECK_PARTIAL(); |
4022 | RRETURN(MATCH_NOMATCH); |
4023 | } |
4024 | GETCHARINCTEST(c, eptr); |
4025 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) |
4026 | RRETURN(MATCH_NOMATCH); |
4027 | } |
4028 | break; |
4029 | |
4030 | case PT_SC: |
4031 | for (i = 1; i <= min; i++) |
4032 | { |
4033 | if (eptr >= md->end_subject) |
4034 | { |
4035 | SCHECK_PARTIAL(); |
4036 | RRETURN(MATCH_NOMATCH); |
4037 | } |
4038 | GETCHARINCTEST(c, eptr); |
4039 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) |
4040 | RRETURN(MATCH_NOMATCH); |
4041 | } |
4042 | break; |
4043 | |
4044 | case PT_ALNUM: |
4045 | for (i = 1; i <= min; i++) |
4046 | { |
4047 | int category; |
4048 | if (eptr >= md->end_subject) |
4049 | { |
4050 | SCHECK_PARTIAL(); |
4051 | RRETURN(MATCH_NOMATCH); |
4052 | } |
4053 | GETCHARINCTEST(c, eptr); |
4054 | category = UCD_CATEGORY(c); |
4055 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
4056 | RRETURN(MATCH_NOMATCH); |
4057 | } |
4058 | break; |
4059 | |
4060 | case PT_SPACE: /* Perl space */ |
4061 | for (i = 1; i <= min; i++) |
4062 | { |
4063 | if (eptr >= md->end_subject) |
4064 | { |
4065 | SCHECK_PARTIAL(); |
4066 | RRETURN(MATCH_NOMATCH); |
4067 | } |
4068 | GETCHARINCTEST(c, eptr); |
4069 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4070 | c == CHAR_FF || c == CHAR_CR) |
4071 | == prop_fail_result) |
4072 | RRETURN(MATCH_NOMATCH); |
4073 | } |
4074 | break; |
4075 | |
4076 | case PT_PXSPACE: /* POSIX space */ |
4077 | for (i = 1; i <= min; i++) |
4078 | { |
4079 | if (eptr >= md->end_subject) |
4080 | { |
4081 | SCHECK_PARTIAL(); |
4082 | RRETURN(MATCH_NOMATCH); |
4083 | } |
4084 | GETCHARINCTEST(c, eptr); |
4085 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4086 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
4087 | == prop_fail_result) |
4088 | RRETURN(MATCH_NOMATCH); |
4089 | } |
4090 | break; |
4091 | |
4092 | case PT_WORD: |
4093 | for (i = 1; i <= min; i++) |
4094 | { |
4095 | int category; |
4096 | if (eptr >= md->end_subject) |
4097 | { |
4098 | SCHECK_PARTIAL(); |
4099 | RRETURN(MATCH_NOMATCH); |
4100 | } |
4101 | GETCHARINCTEST(c, eptr); |
4102 | category = UCD_CATEGORY(c); |
4103 | if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE) |
4104 | == prop_fail_result) |
4105 | RRETURN(MATCH_NOMATCH); |
4106 | } |
4107 | break; |
4108 | |
4109 | /* This should not occur */ |
4110 | |
4111 | default: |
4112 | RRETURN(PCRE_ERROR_INTERNAL); |
4113 | } |
4114 | } |
4115 | |
4116 | /* Match extended Unicode sequences. We will get here only if the |
4117 | support is in the binary; otherwise a compile-time error occurs. */ |
4118 | |
4119 | else if (ctype == OP_EXTUNI) |
4120 | { |
4121 | for (i = 1; i <= min; i++) |
4122 | { |
4123 | if (eptr >= md->end_subject) |
4124 | { |
4125 | SCHECK_PARTIAL(); |
4126 | RRETURN(MATCH_NOMATCH); |
4127 | } |
4128 | GETCHARINCTEST(c, eptr); |
4129 | if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH); |
4130 | while (eptr < md->end_subject) |
4131 | { |
4132 | int len = 1; |
4133 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
4134 | if (UCD_CATEGORY(c) != ucp_M) break; |
4135 | eptr += len; |
4136 | } |
4137 | } |
4138 | } |
4139 | |
4140 | else |
4141 | #endif /* SUPPORT_UCP */ |
4142 | |
4143 | /* Handle all other cases when the coding is UTF-8 */ |
4144 | |
4145 | #ifdef SUPPORT_UTF |
4146 | if (utf) switch(ctype) |
4147 | { |
4148 | case OP_ANY: |
4149 | for (i = 1; i <= min; i++) |
4150 | { |
4151 | if (eptr >= md->end_subject) |
4152 | { |
4153 | SCHECK_PARTIAL(); |
4154 | RRETURN(MATCH_NOMATCH); |
4155 | } |
4156 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
4157 | eptr++; |
4158 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4159 | } |
4160 | break; |
4161 | |
4162 | case OP_ALLANY: |
4163 | for (i = 1; i <= min; i++) |
4164 | { |
4165 | if (eptr >= md->end_subject) |
4166 | { |
4167 | SCHECK_PARTIAL(); |
4168 | RRETURN(MATCH_NOMATCH); |
4169 | } |
4170 | eptr++; |
4171 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4172 | } |
4173 | break; |
4174 | |
4175 | case OP_ANYBYTE: |
4176 | if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH); |
4177 | eptr += min; |
4178 | break; |
4179 | |
4180 | case OP_ANYNL: |
4181 | for (i = 1; i <= min; i++) |
4182 | { |
4183 | if (eptr >= md->end_subject) |
4184 | { |
4185 | SCHECK_PARTIAL(); |
4186 | RRETURN(MATCH_NOMATCH); |
4187 | } |
4188 | GETCHARINC(c, eptr); |
4189 | switch(c) |
4190 | { |
4191 | default: RRETURN(MATCH_NOMATCH); |
4192 | |
4193 | case 0x000d: |
4194 | if (eptr < md->end_subject && *eptr == 0x0a) eptr++; |
4195 | break; |
4196 | |
4197 | case 0x000a: |
4198 | break; |
4199 | |
4200 | case 0x000b: |
4201 | case 0x000c: |
4202 | case 0x0085: |
4203 | case 0x2028: |
4204 | case 0x2029: |
4205 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
4206 | break; |
4207 | } |
4208 | } |
4209 | break; |
4210 | |
4211 | case OP_NOT_HSPACE: |
4212 | for (i = 1; i <= min; i++) |
4213 | { |
4214 | if (eptr >= md->end_subject) |
4215 | { |
4216 | SCHECK_PARTIAL(); |
4217 | RRETURN(MATCH_NOMATCH); |
4218 | } |
4219 | GETCHARINC(c, eptr); |
4220 | switch(c) |
4221 | { |
4222 | default: break; |
4223 | case 0x09: /* HT */ |
4224 | case 0x20: /* SPACE */ |
4225 | case 0xa0: /* NBSP */ |
4226 | case 0x1680: /* OGHAM SPACE MARK */ |
4227 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
4228 | case 0x2000: /* EN QUAD */ |
4229 | case 0x2001: /* EM QUAD */ |
4230 | case 0x2002: /* EN SPACE */ |
4231 | case 0x2003: /* EM SPACE */ |
4232 | case 0x2004: /* THREE-PER-EM SPACE */ |
4233 | case 0x2005: /* FOUR-PER-EM SPACE */ |
4234 | case 0x2006: /* SIX-PER-EM SPACE */ |
4235 | case 0x2007: /* FIGURE SPACE */ |
4236 | case 0x2008: /* PUNCTUATION SPACE */ |
4237 | case 0x2009: /* THIN SPACE */ |
4238 | case 0x200A: /* HAIR SPACE */ |
4239 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
4240 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
4241 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
4242 | RRETURN(MATCH_NOMATCH); |
4243 | } |
4244 | } |
4245 | break; |
4246 | |
4247 | case OP_HSPACE: |
4248 | for (i = 1; i <= min; i++) |
4249 | { |
4250 | if (eptr >= md->end_subject) |
4251 | { |
4252 | SCHECK_PARTIAL(); |
4253 | RRETURN(MATCH_NOMATCH); |
4254 | } |
4255 | GETCHARINC(c, eptr); |
4256 | switch(c) |
4257 | { |
4258 | default: RRETURN(MATCH_NOMATCH); |
4259 | case 0x09: /* HT */ |
4260 | case 0x20: /* SPACE */ |
4261 | case 0xa0: /* NBSP */ |
4262 | case 0x1680: /* OGHAM SPACE MARK */ |
4263 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
4264 | case 0x2000: /* EN QUAD */ |
4265 | case 0x2001: /* EM QUAD */ |
4266 | case 0x2002: /* EN SPACE */ |
4267 | case 0x2003: /* EM SPACE */ |
4268 | case 0x2004: /* THREE-PER-EM SPACE */ |
4269 | case 0x2005: /* FOUR-PER-EM SPACE */ |
4270 | case 0x2006: /* SIX-PER-EM SPACE */ |
4271 | case 0x2007: /* FIGURE SPACE */ |
4272 | case 0x2008: /* PUNCTUATION SPACE */ |
4273 | case 0x2009: /* THIN SPACE */ |
4274 | case 0x200A: /* HAIR SPACE */ |
4275 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
4276 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
4277 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
4278 | break; |
4279 | } |
4280 | } |
4281 | break; |
4282 | |
4283 | case OP_NOT_VSPACE: |
4284 | for (i = 1; i <= min; i++) |
4285 | { |
4286 | if (eptr >= md->end_subject) |
4287 | { |
4288 | SCHECK_PARTIAL(); |
4289 | RRETURN(MATCH_NOMATCH); |
4290 | } |
4291 | GETCHARINC(c, eptr); |
4292 | switch(c) |
4293 | { |
4294 | default: break; |
4295 | case 0x0a: /* LF */ |
4296 | case 0x0b: /* VT */ |
4297 | case 0x0c: /* FF */ |
4298 | case 0x0d: /* CR */ |
4299 | case 0x85: /* NEL */ |
4300 | case 0x2028: /* LINE SEPARATOR */ |
4301 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
4302 | RRETURN(MATCH_NOMATCH); |
4303 | } |
4304 | } |
4305 | break; |
4306 | |
4307 | case OP_VSPACE: |
4308 | for (i = 1; i <= min; i++) |
4309 | { |
4310 | if (eptr >= md->end_subject) |
4311 | { |
4312 | SCHECK_PARTIAL(); |
4313 | RRETURN(MATCH_NOMATCH); |
4314 | } |
4315 | GETCHARINC(c, eptr); |
4316 | switch(c) |
4317 | { |
4318 | default: RRETURN(MATCH_NOMATCH); |
4319 | case 0x0a: /* LF */ |
4320 | case 0x0b: /* VT */ |
4321 | case 0x0c: /* FF */ |
4322 | case 0x0d: /* CR */ |
4323 | case 0x85: /* NEL */ |
4324 | case 0x2028: /* LINE SEPARATOR */ |
4325 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
4326 | break; |
4327 | } |
4328 | } |
4329 | break; |
4330 | |
4331 | case OP_NOT_DIGIT: |
4332 | for (i = 1; i <= min; i++) |
4333 | { |
4334 | if (eptr >= md->end_subject) |
4335 | { |
4336 | SCHECK_PARTIAL(); |
4337 | RRETURN(MATCH_NOMATCH); |
4338 | } |
4339 | GETCHARINC(c, eptr); |
4340 | if (c < 128 && (md->ctypes[c] & ctype_digit) != 0) |
4341 | RRETURN(MATCH_NOMATCH); |
4342 | } |
4343 | break; |
4344 | |
4345 | case OP_DIGIT: |
4346 | for (i = 1; i <= min; i++) |
4347 | { |
4348 | if (eptr >= md->end_subject) |
4349 | { |
4350 | SCHECK_PARTIAL(); |
4351 | RRETURN(MATCH_NOMATCH); |
4352 | } |
4353 | if (*eptr >= 128 || (md->ctypes[*eptr] & ctype_digit) == 0) |
4354 | RRETURN(MATCH_NOMATCH); |
4355 | eptr++; |
4356 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4357 | } |
4358 | break; |
4359 | |
4360 | case OP_NOT_WHITESPACE: |
4361 | for (i = 1; i <= min; i++) |
4362 | { |
4363 | if (eptr >= md->end_subject) |
4364 | { |
4365 | SCHECK_PARTIAL(); |
4366 | RRETURN(MATCH_NOMATCH); |
4367 | } |
4368 | if (*eptr < 128 && (md->ctypes[*eptr] & ctype_space) != 0) |
4369 | RRETURN(MATCH_NOMATCH); |
4370 | eptr++; |
4371 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4372 | } |
4373 | break; |
4374 | |
4375 | case OP_WHITESPACE: |
4376 | for (i = 1; i <= min; i++) |
4377 | { |
4378 | if (eptr >= md->end_subject) |
4379 | { |
4380 | SCHECK_PARTIAL(); |
4381 | RRETURN(MATCH_NOMATCH); |
4382 | } |
4383 | if (*eptr >= 128 || (md->ctypes[*eptr] & ctype_space) == 0) |
4384 | RRETURN(MATCH_NOMATCH); |
4385 | eptr++; |
4386 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4387 | } |
4388 | break; |
4389 | |
4390 | case OP_NOT_WORDCHAR: |
4391 | for (i = 1; i <= min; i++) |
4392 | { |
4393 | if (eptr >= md->end_subject) |
4394 | { |
4395 | SCHECK_PARTIAL(); |
4396 | RRETURN(MATCH_NOMATCH); |
4397 | } |
4398 | if (*eptr < 128 && (md->ctypes[*eptr] & ctype_word) != 0) |
4399 | RRETURN(MATCH_NOMATCH); |
4400 | eptr++; |
4401 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4402 | } |
4403 | break; |
4404 | |
4405 | case OP_WORDCHAR: |
4406 | for (i = 1; i <= min; i++) |
4407 | { |
4408 | if (eptr >= md->end_subject) |
4409 | { |
4410 | SCHECK_PARTIAL(); |
4411 | RRETURN(MATCH_NOMATCH); |
4412 | } |
4413 | if (*eptr >= 128 || (md->ctypes[*eptr] & ctype_word) == 0) |
4414 | RRETURN(MATCH_NOMATCH); |
4415 | eptr++; |
4416 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4417 | } |
4418 | break; |
4419 | |
4420 | default: |
4421 | RRETURN(PCRE_ERROR_INTERNAL); |
4422 | } /* End switch(ctype) */ |
4423 | |
4424 | else |
4425 | #endif /* SUPPORT_UTF */ |
4426 | |
4427 | /* Code for the non-UTF-8 case for minimum matching of operators other |
4428 | than OP_PROP and OP_NOTPROP. */ |
4429 | |
4430 | switch(ctype) |
4431 | { |
4432 | case OP_ANY: |
4433 | for (i = 1; i <= min; i++) |
4434 | { |
4435 | if (eptr >= md->end_subject) |
4436 | { |
4437 | SCHECK_PARTIAL(); |
4438 | RRETURN(MATCH_NOMATCH); |
4439 | } |
4440 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
4441 | eptr++; |
4442 | } |
4443 | break; |
4444 | |
4445 | case OP_ALLANY: |
4446 | if (eptr > md->end_subject - min) |
4447 | { |
4448 | SCHECK_PARTIAL(); |
4449 | RRETURN(MATCH_NOMATCH); |
4450 | } |
4451 | eptr += min; |
4452 | break; |
4453 | |
4454 | case OP_ANYBYTE: |
4455 | if (eptr > md->end_subject - min) |
4456 | { |
4457 | SCHECK_PARTIAL(); |
4458 | RRETURN(MATCH_NOMATCH); |
4459 | } |
4460 | eptr += min; |
4461 | break; |
4462 | |
4463 | case OP_ANYNL: |
4464 | for (i = 1; i <= min; i++) |
4465 | { |
4466 | if (eptr >= md->end_subject) |
4467 | { |
4468 | SCHECK_PARTIAL(); |
4469 | RRETURN(MATCH_NOMATCH); |
4470 | } |
4471 | switch(*eptr++) |
4472 | { |
4473 | default: RRETURN(MATCH_NOMATCH); |
4474 | |
4475 | case 0x000d: |
4476 | if (eptr < md->end_subject && *eptr == 0x0a) eptr++; |
4477 | break; |
4478 | |
4479 | case 0x000a: |
4480 | break; |
4481 | |
4482 | case 0x000b: |
4483 | case 0x000c: |
4484 | case 0x0085: |
4485 | #ifdef COMPILE_PCRE16 |
4486 | case 0x2028: |
4487 | case 0x2029: |
4488 | #endif |
4489 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
4490 | break; |
4491 | } |
4492 | } |
4493 | break; |
4494 | |
4495 | case OP_NOT_HSPACE: |
4496 | for (i = 1; i <= min; i++) |
4497 | { |
4498 | if (eptr >= md->end_subject) |
4499 | { |
4500 | SCHECK_PARTIAL(); |
4501 | RRETURN(MATCH_NOMATCH); |
4502 | } |
4503 | switch(*eptr++) |
4504 | { |
4505 | default: break; |
4506 | case 0x09: /* HT */ |
4507 | case 0x20: /* SPACE */ |
4508 | case 0xa0: /* NBSP */ |
4509 | #ifdef COMPILE_PCRE16 |
4510 | case 0x1680: /* OGHAM SPACE MARK */ |
4511 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
4512 | case 0x2000: /* EN QUAD */ |
4513 | case 0x2001: /* EM QUAD */ |
4514 | case 0x2002: /* EN SPACE */ |
4515 | case 0x2003: /* EM SPACE */ |
4516 | case 0x2004: /* THREE-PER-EM SPACE */ |
4517 | case 0x2005: /* FOUR-PER-EM SPACE */ |
4518 | case 0x2006: /* SIX-PER-EM SPACE */ |
4519 | case 0x2007: /* FIGURE SPACE */ |
4520 | case 0x2008: /* PUNCTUATION SPACE */ |
4521 | case 0x2009: /* THIN SPACE */ |
4522 | case 0x200A: /* HAIR SPACE */ |
4523 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
4524 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
4525 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
4526 | #endif |
4527 | RRETURN(MATCH_NOMATCH); |
4528 | } |
4529 | } |
4530 | break; |
4531 | |
4532 | case OP_HSPACE: |
4533 | for (i = 1; i <= min; i++) |
4534 | { |
4535 | if (eptr >= md->end_subject) |
4536 | { |
4537 | SCHECK_PARTIAL(); |
4538 | RRETURN(MATCH_NOMATCH); |
4539 | } |
4540 | switch(*eptr++) |
4541 | { |
4542 | default: RRETURN(MATCH_NOMATCH); |
4543 | case 0x09: /* HT */ |
4544 | case 0x20: /* SPACE */ |
4545 | case 0xa0: /* NBSP */ |
4546 | #ifdef COMPILE_PCRE16 |
4547 | case 0x1680: /* OGHAM SPACE MARK */ |
4548 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
4549 | case 0x2000: /* EN QUAD */ |
4550 | case 0x2001: /* EM QUAD */ |
4551 | case 0x2002: /* EN SPACE */ |
4552 | case 0x2003: /* EM SPACE */ |
4553 | case 0x2004: /* THREE-PER-EM SPACE */ |
4554 | case 0x2005: /* FOUR-PER-EM SPACE */ |
4555 | case 0x2006: /* SIX-PER-EM SPACE */ |
4556 | case 0x2007: /* FIGURE SPACE */ |
4557 | case 0x2008: /* PUNCTUATION SPACE */ |
4558 | case 0x2009: /* THIN SPACE */ |
4559 | case 0x200A: /* HAIR SPACE */ |
4560 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
4561 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
4562 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
4563 | #endif |
4564 | break; |
4565 | } |
4566 | } |
4567 | break; |
4568 | |
4569 | case OP_NOT_VSPACE: |
4570 | for (i = 1; i <= min; i++) |
4571 | { |
4572 | if (eptr >= md->end_subject) |
4573 | { |
4574 | SCHECK_PARTIAL(); |
4575 | RRETURN(MATCH_NOMATCH); |
4576 | } |
4577 | switch(*eptr++) |
4578 | { |
4579 | default: break; |
4580 | case 0x0a: /* LF */ |
4581 | case 0x0b: /* VT */ |
4582 | case 0x0c: /* FF */ |
4583 | case 0x0d: /* CR */ |
4584 | case 0x85: /* NEL */ |
4585 | #ifdef COMPILE_PCRE16 |
4586 | case 0x2028: /* LINE SEPARATOR */ |
4587 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
4588 | #endif |
4589 | RRETURN(MATCH_NOMATCH); |
4590 | } |
4591 | } |
4592 | break; |
4593 | |
4594 | case OP_VSPACE: |
4595 | for (i = 1; i <= min; i++) |
4596 | { |
4597 | if (eptr >= md->end_subject) |
4598 | { |
4599 | SCHECK_PARTIAL(); |
4600 | RRETURN(MATCH_NOMATCH); |
4601 | } |
4602 | switch(*eptr++) |
4603 | { |
4604 | default: RRETURN(MATCH_NOMATCH); |
4605 | case 0x0a: /* LF */ |
4606 | case 0x0b: /* VT */ |
4607 | case 0x0c: /* FF */ |
4608 | case 0x0d: /* CR */ |
4609 | case 0x85: /* NEL */ |
4610 | #ifdef COMPILE_PCRE16 |
4611 | case 0x2028: /* LINE SEPARATOR */ |
4612 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
4613 | #endif |
4614 | break; |
4615 | } |
4616 | } |
4617 | break; |
4618 | |
4619 | case OP_NOT_DIGIT: |
4620 | for (i = 1; i <= min; i++) |
4621 | { |
4622 | if (eptr >= md->end_subject) |
4623 | { |
4624 | SCHECK_PARTIAL(); |
4625 | RRETURN(MATCH_NOMATCH); |
4626 | } |
4627 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) |
4628 | RRETURN(MATCH_NOMATCH); |
4629 | eptr++; |
4630 | } |
4631 | break; |
4632 | |
4633 | case OP_DIGIT: |
4634 | for (i = 1; i <= min; i++) |
4635 | { |
4636 | if (eptr >= md->end_subject) |
4637 | { |
4638 | SCHECK_PARTIAL(); |
4639 | RRETURN(MATCH_NOMATCH); |
4640 | } |
4641 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) |
4642 | RRETURN(MATCH_NOMATCH); |
4643 | eptr++; |
4644 | } |
4645 | break; |
4646 | |
4647 | case OP_NOT_WHITESPACE: |
4648 | for (i = 1; i <= min; i++) |
4649 | { |
4650 | if (eptr >= md->end_subject) |
4651 | { |
4652 | SCHECK_PARTIAL(); |
4653 | RRETURN(MATCH_NOMATCH); |
4654 | } |
4655 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) |
4656 | RRETURN(MATCH_NOMATCH); |
4657 | eptr++; |
4658 | } |
4659 | break; |
4660 | |
4661 | case OP_WHITESPACE: |
4662 | for (i = 1; i <= min; i++) |
4663 | { |
4664 | if (eptr >= md->end_subject) |
4665 | { |
4666 | SCHECK_PARTIAL(); |
4667 | RRETURN(MATCH_NOMATCH); |
4668 | } |
4669 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) |
4670 | RRETURN(MATCH_NOMATCH); |
4671 | eptr++; |
4672 | } |
4673 | break; |
4674 | |
4675 | case OP_NOT_WORDCHAR: |
4676 | for (i = 1; i <= min; i++) |
4677 | { |
4678 | if (eptr >= md->end_subject) |
4679 | { |
4680 | SCHECK_PARTIAL(); |
4681 | RRETURN(MATCH_NOMATCH); |
4682 | } |
4683 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) |
4684 | RRETURN(MATCH_NOMATCH); |
4685 | eptr++; |
4686 | } |
4687 | break; |
4688 | |
4689 | case OP_WORDCHAR: |
4690 | for (i = 1; i <= min; i++) |
4691 | { |
4692 | if (eptr >= md->end_subject) |
4693 | { |
4694 | SCHECK_PARTIAL(); |
4695 | RRETURN(MATCH_NOMATCH); |
4696 | } |
4697 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) |
4698 | RRETURN(MATCH_NOMATCH); |
4699 | eptr++; |
4700 | } |
4701 | break; |
4702 | |
4703 | default: |
4704 | RRETURN(PCRE_ERROR_INTERNAL); |
4705 | } |
4706 | } |
4707 | |
4708 | /* If min = max, continue at the same level without recursing */ |
4709 | |
4710 | if (min == max) continue; |
4711 | |
4712 | /* If minimizing, we have to test the rest of the pattern before each |
4713 | subsequent match. Again, separate the UTF-8 case for speed, and also |
4714 | separate the UCP cases. */ |
4715 | |
4716 | if (minimize) |
4717 | { |
4718 | #ifdef SUPPORT_UCP |
4719 | if (prop_type >= 0) |
4720 | { |
4721 | switch(prop_type) |
4722 | { |
4723 | case PT_ANY: |
4724 | for (fi = min;; fi++) |
4725 | { |
4726 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM36); |
4727 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4728 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4729 | if (eptr >= md->end_subject) |
4730 | { |
4731 | SCHECK_PARTIAL(); |
4732 | RRETURN(MATCH_NOMATCH); |
4733 | } |
4734 | GETCHARINCTEST(c, eptr); |
4735 | if (prop_fail_result) RRETURN(MATCH_NOMATCH); |
4736 | } |
4737 | /* Control never gets here */ |
4738 | |
4739 | case PT_LAMP: |
4740 | for (fi = min;; fi++) |
4741 | { |
4742 | int chartype; |
4743 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM37); |
4744 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4745 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4746 | if (eptr >= md->end_subject) |
4747 | { |
4748 | SCHECK_PARTIAL(); |
4749 | RRETURN(MATCH_NOMATCH); |
4750 | } |
4751 | GETCHARINCTEST(c, eptr); |
4752 | chartype = UCD_CHARTYPE(c); |
4753 | if ((chartype == ucp_Lu || |
4754 | chartype == ucp_Ll || |
4755 | chartype == ucp_Lt) == prop_fail_result) |
4756 | RRETURN(MATCH_NOMATCH); |
4757 | } |
4758 | /* Control never gets here */ |
4759 | |
4760 | case PT_GC: |
4761 | for (fi = min;; fi++) |
4762 | { |
4763 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM38); |
4764 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4765 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4766 | if (eptr >= md->end_subject) |
4767 | { |
4768 | SCHECK_PARTIAL(); |
4769 | RRETURN(MATCH_NOMATCH); |
4770 | } |
4771 | GETCHARINCTEST(c, eptr); |
4772 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) |
4773 | RRETURN(MATCH_NOMATCH); |
4774 | } |
4775 | /* Control never gets here */ |
4776 | |
4777 | case PT_PC: |
4778 | for (fi = min;; fi++) |
4779 | { |
4780 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM39); |
4781 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4782 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4783 | if (eptr >= md->end_subject) |
4784 | { |
4785 | SCHECK_PARTIAL(); |
4786 | RRETURN(MATCH_NOMATCH); |
4787 | } |
4788 | GETCHARINCTEST(c, eptr); |
4789 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) |
4790 | RRETURN(MATCH_NOMATCH); |
4791 | } |
4792 | /* Control never gets here */ |
4793 | |
4794 | case PT_SC: |
4795 | for (fi = min;; fi++) |
4796 | { |
4797 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM40); |
4798 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4799 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4800 | if (eptr >= md->end_subject) |
4801 | { |
4802 | SCHECK_PARTIAL(); |
4803 | RRETURN(MATCH_NOMATCH); |
4804 | } |
4805 | GETCHARINCTEST(c, eptr); |
4806 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) |
4807 | RRETURN(MATCH_NOMATCH); |
4808 | } |
4809 | /* Control never gets here */ |
4810 | |
4811 | case PT_ALNUM: |
4812 | for (fi = min;; fi++) |
4813 | { |
4814 | int category; |
4815 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM59); |
4816 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4817 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4818 | if (eptr >= md->end_subject) |
4819 | { |
4820 | SCHECK_PARTIAL(); |
4821 | RRETURN(MATCH_NOMATCH); |
4822 | } |
4823 | GETCHARINCTEST(c, eptr); |
4824 | category = UCD_CATEGORY(c); |
4825 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
4826 | RRETURN(MATCH_NOMATCH); |
4827 | } |
4828 | /* Control never gets here */ |
4829 | |
4830 | case PT_SPACE: /* Perl space */ |
4831 | for (fi = min;; fi++) |
4832 | { |
4833 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM60); |
4834 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4835 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4836 | if (eptr >= md->end_subject) |
4837 | { |
4838 | SCHECK_PARTIAL(); |
4839 | RRETURN(MATCH_NOMATCH); |
4840 | } |
4841 | GETCHARINCTEST(c, eptr); |
4842 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4843 | c == CHAR_FF || c == CHAR_CR) |
4844 | == prop_fail_result) |
4845 | RRETURN(MATCH_NOMATCH); |
4846 | } |
4847 | /* Control never gets here */ |
4848 | |
4849 | case PT_PXSPACE: /* POSIX space */ |
4850 | for (fi = min;; fi++) |
4851 | { |
4852 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM61); |
4853 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4854 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4855 | if (eptr >= md->end_subject) |
4856 | { |
4857 | SCHECK_PARTIAL(); |
4858 | RRETURN(MATCH_NOMATCH); |
4859 | } |
4860 | GETCHARINCTEST(c, eptr); |
4861 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4862 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
4863 | == prop_fail_result) |
4864 | RRETURN(MATCH_NOMATCH); |
4865 | } |
4866 | /* Control never gets here */ |
4867 | |
4868 | case PT_WORD: |
4869 | for (fi = min;; fi++) |
4870 | { |
4871 | int category; |
4872 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM62); |
4873 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4874 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4875 | if (eptr >= md->end_subject) |
4876 | { |
4877 | SCHECK_PARTIAL(); |
4878 | RRETURN(MATCH_NOMATCH); |
4879 | } |
4880 | GETCHARINCTEST(c, eptr); |
4881 | category = UCD_CATEGORY(c); |
4882 | if ((category == ucp_L || |
4883 | category == ucp_N || |
4884 | c == CHAR_UNDERSCORE) |
4885 | == prop_fail_result) |
4886 | RRETURN(MATCH_NOMATCH); |
4887 | } |
4888 | /* Control never gets here */ |
4889 | |
4890 | /* This should never occur */ |
4891 | |
4892 | default: |
4893 | RRETURN(PCRE_ERROR_INTERNAL); |
4894 | } |
4895 | } |
4896 | |
4897 | /* Match extended Unicode sequences. We will get here only if the |
4898 | support is in the binary; otherwise a compile-time error occurs. */ |
4899 | |
4900 | else if (ctype == OP_EXTUNI) |
4901 | { |
4902 | for (fi = min;; fi++) |
4903 | { |
4904 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM41); |
4905 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4906 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4907 | if (eptr >= md->end_subject) |
4908 | { |
4909 | SCHECK_PARTIAL(); |
4910 | RRETURN(MATCH_NOMATCH); |
4911 | } |
4912 | GETCHARINCTEST(c, eptr); |
4913 | if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH); |
4914 | while (eptr < md->end_subject) |
4915 | { |
4916 | int len = 1; |
4917 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
4918 | if (UCD_CATEGORY(c) != ucp_M) break; |
4919 | eptr += len; |
4920 | } |
4921 | } |
4922 | } |
4923 | else |
4924 | #endif /* SUPPORT_UCP */ |
4925 | |
4926 | #ifdef SUPPORT_UTF |
4927 | if (utf) |
4928 | { |
4929 | for (fi = min;; fi++) |
4930 | { |
4931 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM42); |
4932 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4933 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4934 | if (eptr >= md->end_subject) |
4935 | { |
4936 | SCHECK_PARTIAL(); |
4937 | RRETURN(MATCH_NOMATCH); |
4938 | } |
4939 | if (ctype == OP_ANY && IS_NEWLINE(eptr)) |
4940 | RRETURN(MATCH_NOMATCH); |
4941 | GETCHARINC(c, eptr); |
4942 | switch(ctype) |
4943 | { |
4944 | case OP_ANY: /* This is the non-NL case */ |
4945 | case OP_ALLANY: |
4946 | case OP_ANYBYTE: |
4947 | break; |
4948 | |
4949 | case OP_ANYNL: |
4950 | switch(c) |
4951 | { |
4952 | default: RRETURN(MATCH_NOMATCH); |
4953 | case 0x000d: |
4954 | if (eptr < md->end_subject && *eptr == 0x0a) eptr++; |
4955 | break; |
4956 | case 0x000a: |
4957 | break; |
4958 | |
4959 | case 0x000b: |
4960 | case 0x000c: |
4961 | case 0x0085: |
4962 | case 0x2028: |
4963 | case 0x2029: |
4964 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
4965 | break; |
4966 | } |
4967 | break; |
4968 | |
4969 | case OP_NOT_HSPACE: |
4970 | switch(c) |
4971 | { |
4972 | default: break; |
4973 | case 0x09: /* HT */ |
4974 | case 0x20: /* SPACE */ |
4975 | case 0xa0: /* NBSP */ |
4976 | case 0x1680: /* OGHAM SPACE MARK */ |
4977 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
4978 | case 0x2000: /* EN QUAD */ |
4979 | case 0x2001: /* EM QUAD */ |
4980 | case 0x2002: /* EN SPACE */ |
4981 | case 0x2003: /* EM SPACE */ |
4982 | case 0x2004: /* THREE-PER-EM SPACE */ |
4983 | case 0x2005: /* FOUR-PER-EM SPACE */ |
4984 | case 0x2006: /* SIX-PER-EM SPACE */ |
4985 | case 0x2007: /* FIGURE SPACE */ |
4986 | case 0x2008: /* PUNCTUATION SPACE */ |
4987 | case 0x2009: /* THIN SPACE */ |
4988 | case 0x200A: /* HAIR SPACE */ |
4989 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
4990 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
4991 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
4992 | RRETURN(MATCH_NOMATCH); |
4993 | } |
4994 | break; |
4995 | |
4996 | case OP_HSPACE: |
4997 | switch(c) |
4998 | { |
4999 | default: RRETURN(MATCH_NOMATCH); |
5000 | case 0x09: /* HT */ |
5001 | case 0x20: /* SPACE */ |
5002 | case 0xa0: /* NBSP */ |
5003 | case 0x1680: /* OGHAM SPACE MARK */ |
5004 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
5005 | case 0x2000: /* EN QUAD */ |
5006 | case 0x2001: /* EM QUAD */ |
5007 | case 0x2002: /* EN SPACE */ |
5008 | case 0x2003: /* EM SPACE */ |
5009 | case 0x2004: /* THREE-PER-EM SPACE */ |
5010 | case 0x2005: /* FOUR-PER-EM SPACE */ |
5011 | case 0x2006: /* SIX-PER-EM SPACE */ |
5012 | case 0x2007: /* FIGURE SPACE */ |
5013 | case 0x2008: /* PUNCTUATION SPACE */ |
5014 | case 0x2009: /* THIN SPACE */ |
5015 | case 0x200A: /* HAIR SPACE */ |
5016 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
5017 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
5018 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
5019 | break; |
5020 | } |
5021 | break; |
5022 | |
5023 | case OP_NOT_VSPACE: |
5024 | switch(c) |
5025 | { |
5026 | default: break; |
5027 | case 0x0a: /* LF */ |
5028 | case 0x0b: /* VT */ |
5029 | case 0x0c: /* FF */ |
5030 | case 0x0d: /* CR */ |
5031 | case 0x85: /* NEL */ |
5032 | case 0x2028: /* LINE SEPARATOR */ |
5033 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
5034 | RRETURN(MATCH_NOMATCH); |
5035 | } |
5036 | break; |
5037 | |
5038 | case OP_VSPACE: |
5039 | switch(c) |
5040 | { |
5041 | default: RRETURN(MATCH_NOMATCH); |
5042 | case 0x0a: /* LF */ |
5043 | case 0x0b: /* VT */ |
5044 | case 0x0c: /* FF */ |
5045 | case 0x0d: /* CR */ |
5046 | case 0x85: /* NEL */ |
5047 | case 0x2028: /* LINE SEPARATOR */ |
5048 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
5049 | break; |
5050 | } |
5051 | break; |
5052 | |
5053 | case OP_NOT_DIGIT: |
5054 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) |
5055 | RRETURN(MATCH_NOMATCH); |
5056 | break; |
5057 | |
5058 | case OP_DIGIT: |
5059 | if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0) |
5060 | RRETURN(MATCH_NOMATCH); |
5061 | break; |
5062 | |
5063 | case OP_NOT_WHITESPACE: |
5064 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) |
5065 | RRETURN(MATCH_NOMATCH); |
5066 | break; |
5067 | |
5068 | case OP_WHITESPACE: |
5069 | if (c >= 256 || (md->ctypes[c] & ctype_space) == 0) |
5070 | RRETURN(MATCH_NOMATCH); |
5071 | break; |
5072 | |
5073 | case OP_NOT_WORDCHAR: |
5074 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) |
5075 | RRETURN(MATCH_NOMATCH); |
5076 | break; |
5077 | |
5078 | case OP_WORDCHAR: |
5079 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) |
5080 | RRETURN(MATCH_NOMATCH); |
5081 | break; |
5082 | |
5083 | default: |
5084 | RRETURN(PCRE_ERROR_INTERNAL); |
5085 | } |
5086 | } |
5087 | } |
5088 | else |
5089 | #endif |
5090 | /* Not UTF mode */ |
5091 | { |
5092 | for (fi = min;; fi++) |
5093 | { |
5094 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM43); |
5095 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5096 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5097 | if (eptr >= md->end_subject) |
5098 | { |
5099 | SCHECK_PARTIAL(); |
5100 | RRETURN(MATCH_NOMATCH); |
5101 | } |
5102 | if (ctype == OP_ANY && IS_NEWLINE(eptr)) |
5103 | RRETURN(MATCH_NOMATCH); |
5104 | c = *eptr++; |
5105 | switch(ctype) |
5106 | { |
5107 | case OP_ANY: /* This is the non-NL case */ |
5108 | case OP_ALLANY: |
5109 | case OP_ANYBYTE: |
5110 | break; |
5111 | |
5112 | case OP_ANYNL: |
5113 | switch(c) |
5114 | { |
5115 | default: RRETURN(MATCH_NOMATCH); |
5116 | case 0x000d: |
5117 | if (eptr < md->end_subject && *eptr == 0x0a) eptr++; |
5118 | break; |
5119 | |
5120 | case 0x000a: |
5121 | break; |
5122 | |
5123 | case 0x000b: |
5124 | case 0x000c: |
5125 | case 0x0085: |
5126 | #ifdef COMPILE_PCRE16 |
5127 | case 0x2028: |
5128 | case 0x2029: |
5129 | #endif |
5130 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
5131 | break; |
5132 | } |
5133 | break; |
5134 | |
5135 | case OP_NOT_HSPACE: |
5136 | switch(c) |
5137 | { |
5138 | default: break; |
5139 | case 0x09: /* HT */ |
5140 | case 0x20: /* SPACE */ |
5141 | case 0xa0: /* NBSP */ |
5142 | #ifdef COMPILE_PCRE16 |
5143 | case 0x1680: /* OGHAM SPACE MARK */ |
5144 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
5145 | case 0x2000: /* EN QUAD */ |
5146 | case 0x2001: /* EM QUAD */ |
5147 | case 0x2002: /* EN SPACE */ |
5148 | case 0x2003: /* EM SPACE */ |
5149 | case 0x2004: /* THREE-PER-EM SPACE */ |
5150 | case 0x2005: /* FOUR-PER-EM SPACE */ |
5151 | case 0x2006: /* SIX-PER-EM SPACE */ |
5152 | case 0x2007: /* FIGURE SPACE */ |
5153 | case 0x2008: /* PUNCTUATION SPACE */ |
5154 | case 0x2009: /* THIN SPACE */ |
5155 | case 0x200A: /* HAIR SPACE */ |
5156 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
5157 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
5158 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
5159 | #endif |
5160 | RRETURN(MATCH_NOMATCH); |
5161 | } |
5162 | break; |
5163 | |
5164 | case OP_HSPACE: |
5165 | switch(c) |
5166 | { |
5167 | default: RRETURN(MATCH_NOMATCH); |
5168 | case 0x09: /* HT */ |
5169 | case 0x20: /* SPACE */ |
5170 | case 0xa0: /* NBSP */ |
5171 | #ifdef COMPILE_PCRE16 |
5172 | case 0x1680: /* OGHAM SPACE MARK */ |
5173 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
5174 | case 0x2000: /* EN QUAD */ |
5175 | case 0x2001: /* EM QUAD */ |
5176 | case 0x2002: /* EN SPACE */ |
5177 | case 0x2003: /* EM SPACE */ |
5178 | case 0x2004: /* THREE-PER-EM SPACE */ |
5179 | case 0x2005: /* FOUR-PER-EM SPACE */ |
5180 | case 0x2006: /* SIX-PER-EM SPACE */ |
5181 | case 0x2007: /* FIGURE SPACE */ |
5182 | case 0x2008: /* PUNCTUATION SPACE */ |
5183 | case 0x2009: /* THIN SPACE */ |
5184 | case 0x200A: /* HAIR SPACE */ |
5185 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
5186 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
5187 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
5188 | #endif |
5189 | break; |
5190 | } |
5191 | break; |
5192 | |
5193 | case OP_NOT_VSPACE: |
5194 | switch(c) |
5195 | { |
5196 | default: break; |
5197 | case 0x0a: /* LF */ |
5198 | case 0x0b: /* VT */ |
5199 | case 0x0c: /* FF */ |
5200 | case 0x0d: /* CR */ |
5201 | case 0x85: /* NEL */ |
5202 | #ifdef COMPILE_PCRE16 |
5203 | case 0x2028: /* LINE SEPARATOR */ |
5204 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
5205 | #endif |
5206 | RRETURN(MATCH_NOMATCH); |
5207 | } |
5208 | break; |
5209 | |
5210 | case OP_VSPACE: |
5211 | switch(c) |
5212 | { |
5213 | default: RRETURN(MATCH_NOMATCH); |
5214 | case 0x0a: /* LF */ |
5215 | case 0x0b: /* VT */ |
5216 | case 0x0c: /* FF */ |
5217 | case 0x0d: /* CR */ |
5218 | case 0x85: /* NEL */ |
5219 | #ifdef COMPILE_PCRE16 |
5220 | case 0x2028: /* LINE SEPARATOR */ |
5221 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
5222 | #endif |
5223 | break; |
5224 | } |
5225 | break; |
5226 | |
5227 | case OP_NOT_DIGIT: |
5228 | if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH); |
5229 | break; |
5230 | |
5231 | case OP_DIGIT: |
5232 | if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH); |
5233 | break; |
5234 | |
5235 | case OP_NOT_WHITESPACE: |
5236 | if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH); |
5237 | break; |
5238 | |
5239 | case OP_WHITESPACE: |
5240 | if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH); |
5241 | break; |
5242 | |
5243 | case OP_NOT_WORDCHAR: |
5244 | if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH); |
5245 | break; |
5246 | |
5247 | case OP_WORDCHAR: |
5248 | if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH); |
5249 | break; |
5250 | |
5251 | default: |
5252 | RRETURN(PCRE_ERROR_INTERNAL); |
5253 | } |
5254 | } |
5255 | } |
5256 | /* Control never gets here */ |
5257 | } |
5258 | |
5259 | /* If maximizing, it is worth using inline code for speed, doing the type |
5260 | test once at the start (i.e. keep it out of the loop). Again, keep the |
5261 | UTF-8 and UCP stuff separate. */ |
5262 | |
5263 | else |
5264 | { |
5265 | pp = eptr; /* Remember where we started */ |
5266 | |
5267 | #ifdef SUPPORT_UCP |
5268 | if (prop_type >= 0) |
5269 | { |
5270 | switch(prop_type) |
5271 | { |
5272 | case PT_ANY: |
5273 | for (i = min; i < max; i++) |
5274 | { |
5275 | int len = 1; |
5276 | if (eptr >= md->end_subject) |
5277 | { |
5278 | SCHECK_PARTIAL(); |
5279 | break; |
5280 | } |
5281 | GETCHARLENTEST(c, eptr, len); |
5282 | if (prop_fail_result) break; |
5283 | eptr+= len; |
5284 | } |
5285 | break; |
5286 | |
5287 | case PT_LAMP: |
5288 | for (i = min; i < max; i++) |
5289 | { |
5290 | int chartype; |
5291 | int len = 1; |
5292 | if (eptr >= md->end_subject) |
5293 | { |
5294 | SCHECK_PARTIAL(); |
5295 | break; |
5296 | } |
5297 | GETCHARLENTEST(c, eptr, len); |
5298 | chartype = UCD_CHARTYPE(c); |
5299 | if ((chartype == ucp_Lu || |
5300 | chartype == ucp_Ll || |
5301 | chartype == ucp_Lt) == prop_fail_result) |
5302 | break; |
5303 | eptr+= len; |
5304 | } |
5305 | break; |
5306 | |
5307 | case PT_GC: |
5308 | for (i = min; i < max; i++) |
5309 | { |
5310 | int len = 1; |
5311 | if (eptr >= md->end_subject) |
5312 | { |
5313 | SCHECK_PARTIAL(); |
5314 | break; |
5315 | } |
5316 | GETCHARLENTEST(c, eptr, len); |
5317 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break; |
5318 | eptr+= len; |
5319 | } |
5320 | break; |
5321 | |
5322 | case PT_PC: |
5323 | for (i = min; i < max; i++) |
5324 | { |
5325 | int len = 1; |
5326 | if (eptr >= md->end_subject) |
5327 | { |
5328 | SCHECK_PARTIAL(); |
5329 | break; |
5330 | } |
5331 | GETCHARLENTEST(c, eptr, len); |
5332 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break; |
5333 | eptr+= len; |
5334 | } |
5335 | break; |
5336 | |
5337 | case PT_SC: |
5338 | for (i = min; i < max; i++) |
5339 | { |
5340 | int len = 1; |
5341 | if (eptr >= md->end_subject) |
5342 | { |
5343 | SCHECK_PARTIAL(); |
5344 | break; |
5345 | } |
5346 | GETCHARLENTEST(c, eptr, len); |
5347 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break; |
5348 | eptr+= len; |
5349 | } |
5350 | break; |
5351 | |
5352 | case PT_ALNUM: |
5353 | for (i = min; i < max; i++) |
5354 | { |
5355 | int category; |
5356 | int len = 1; |
5357 | if (eptr >= md->end_subject) |
5358 | { |
5359 | SCHECK_PARTIAL(); |
5360 | break; |
5361 | } |
5362 | GETCHARLENTEST(c, eptr, len); |
5363 | category = UCD_CATEGORY(c); |
5364 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
5365 | break; |
5366 | eptr+= len; |
5367 | } |
5368 | break; |
5369 | |
5370 | case PT_SPACE: /* Perl space */ |
5371 | for (i = min; i < max; i++) |
5372 | { |
5373 | int len = 1; |
5374 | if (eptr >= md->end_subject) |
5375 | { |
5376 | SCHECK_PARTIAL(); |
5377 | break; |
5378 | } |
5379 | GETCHARLENTEST(c, eptr, len); |
5380 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5381 | c == CHAR_FF || c == CHAR_CR) |
5382 | == prop_fail_result) |
5383 | break; |
5384 | eptr+= len; |
5385 | } |
5386 | break; |
5387 | |
5388 | case PT_PXSPACE: /* POSIX space */ |
5389 | for (i = min; i < max; i++) |
5390 | { |
5391 | int len = 1; |
5392 | if (eptr >= md->end_subject) |
5393 | { |
5394 | SCHECK_PARTIAL(); |
5395 | break; |
5396 | } |
5397 | GETCHARLENTEST(c, eptr, len); |
5398 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5399 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
5400 | == prop_fail_result) |
5401 | break; |
5402 | eptr+= len; |
5403 | } |
5404 | break; |
5405 | |
5406 | case PT_WORD: |
5407 | for (i = min; i < max; i++) |
5408 | { |
5409 | int category; |
5410 | int len = 1; |
5411 | if (eptr >= md->end_subject) |
5412 | { |
5413 | SCHECK_PARTIAL(); |
5414 | break; |
5415 | } |
5416 | GETCHARLENTEST(c, eptr, len); |
5417 | category = UCD_CATEGORY(c); |
5418 | if ((category == ucp_L || category == ucp_N || |
5419 | c == CHAR_UNDERSCORE) == prop_fail_result) |
5420 | break; |
5421 | eptr+= len; |
5422 | } |
5423 | break; |
5424 | |
5425 | default: |
5426 | RRETURN(PCRE_ERROR_INTERNAL); |
5427 | } |
5428 | |
5429 | /* eptr is now past the end of the maximum run */ |
5430 | |
5431 | if (possessive) continue; |
5432 | for(;;) |
5433 | { |
5434 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM44); |
5435 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5436 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
5437 | if (utf) BACKCHAR(eptr); |
5438 | } |
5439 | } |
5440 | |
5441 | /* Match extended Unicode sequences. We will get here only if the |
5442 | support is in the binary; otherwise a compile-time error occurs. */ |
5443 | |
5444 | else if (ctype == OP_EXTUNI) |
5445 | { |
5446 | for (i = min; i < max; i++) |
5447 | { |
5448 | int len = 1; |
5449 | if (eptr >= md->end_subject) |
5450 | { |
5451 | SCHECK_PARTIAL(); |
5452 | break; |
5453 | } |
5454 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
5455 | if (UCD_CATEGORY(c) == ucp_M) break; |
5456 | eptr += len; |
5457 | while (eptr < md->end_subject) |
5458 | { |
5459 | len = 1; |
5460 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
5461 | if (UCD_CATEGORY(c) != ucp_M) break; |
5462 | eptr += len; |
5463 | } |
5464 | } |
5465 | |
5466 | /* eptr is now past the end of the maximum run */ |
5467 | |
5468 | if (possessive) continue; |
5469 | |
5470 | for(;;) |
5471 | { |
5472 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM45); |
5473 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5474 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
5475 | for (;;) /* Move back over one extended */ |
5476 | { |
5477 | if (!utf) c = *eptr; else |
5478 | { |
5479 | BACKCHAR(eptr); |
5480 | GETCHAR(c, eptr); |
5481 | } |
5482 | if (UCD_CATEGORY(c) != ucp_M) break; |
5483 | eptr--; |
5484 | } |
5485 | } |
5486 | } |
5487 | |
5488 | else |
5489 | #endif /* SUPPORT_UCP */ |
5490 | |
5491 | #ifdef SUPPORT_UTF |
5492 | if (utf) |
5493 | { |
5494 | switch(ctype) |
5495 | { |
5496 | case OP_ANY: |
5497 | if (max < INT_MAX) |
5498 | { |
5499 | for (i = min; i < max; i++) |
5500 | { |
5501 | if (eptr >= md->end_subject) |
5502 | { |
5503 | SCHECK_PARTIAL(); |
5504 | break; |
5505 | } |
5506 | if (IS_NEWLINE(eptr)) break; |
5507 | eptr++; |
5508 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5509 | } |
5510 | } |
5511 | |
5512 | /* Handle unlimited UTF-8 repeat */ |
5513 | |
5514 | else |
5515 | { |
5516 | for (i = min; i < max; i++) |
5517 | { |
5518 | if (eptr >= md->end_subject) |
5519 | { |
5520 | SCHECK_PARTIAL(); |
5521 | break; |
5522 | } |
5523 | if (IS_NEWLINE(eptr)) break; |
5524 | eptr++; |
5525 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5526 | } |
5527 | } |
5528 | break; |
5529 | |
5530 | case OP_ALLANY: |
5531 | if (max < INT_MAX) |
5532 | { |
5533 | for (i = min; i < max; i++) |
5534 | { |
5535 | if (eptr >= md->end_subject) |
5536 | { |
5537 | SCHECK_PARTIAL(); |
5538 | break; |
5539 | } |
5540 | eptr++; |
5541 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5542 | } |
5543 | } |
5544 | else |
5545 | { |
5546 | eptr = md->end_subject; /* Unlimited UTF-8 repeat */ |
5547 | SCHECK_PARTIAL(); |
5548 | } |
5549 | break; |
5550 | |
5551 | /* The byte case is the same as non-UTF8 */ |
5552 | |
5553 | case OP_ANYBYTE: |
5554 | c = max - min; |
5555 | if (c > (unsigned int)(md->end_subject - eptr)) |
5556 | { |
5557 | eptr = md->end_subject; |
5558 | SCHECK_PARTIAL(); |
5559 | } |
5560 | else eptr += c; |
5561 | break; |
5562 | |
5563 | case OP_ANYNL: |
5564 | for (i = min; i < max; i++) |
5565 | { |
5566 | int len = 1; |
5567 | if (eptr >= md->end_subject) |
5568 | { |
5569 | SCHECK_PARTIAL(); |
5570 | break; |
5571 | } |
5572 | GETCHARLEN(c, eptr, len); |
5573 | if (c == 0x000d) |
5574 | { |
5575 | if (++eptr >= md->end_subject) break; |
5576 | if (*eptr == 0x000a) eptr++; |
5577 | } |
5578 | else |
5579 | { |
5580 | if (c != 0x000a && |
5581 | (md->bsr_anycrlf || |
5582 | (c != 0x000b && c != 0x000c && |
5583 | c != 0x0085 && c != 0x2028 && c != 0x2029))) |
5584 | break; |
5585 | eptr += len; |
5586 | } |
5587 | } |
5588 | break; |
5589 | |
5590 | case OP_NOT_HSPACE: |
5591 | case OP_HSPACE: |
5592 | for (i = min; i < max; i++) |
5593 | { |
5594 | BOOL gotspace; |
5595 | int len = 1; |
5596 | if (eptr >= md->end_subject) |
5597 | { |
5598 | SCHECK_PARTIAL(); |
5599 | break; |
5600 | } |
5601 | GETCHARLEN(c, eptr, len); |
5602 | switch(c) |
5603 | { |
5604 | default: gotspace = FALSE; break; |
5605 | case 0x09: /* HT */ |
5606 | case 0x20: /* SPACE */ |
5607 | case 0xa0: /* NBSP */ |
5608 | case 0x1680: /* OGHAM SPACE MARK */ |
5609 | case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */ |
5610 | case 0x2000: /* EN QUAD */ |
5611 | case 0x2001: /* EM QUAD */ |
5612 | case 0x2002: /* EN SPACE */ |
5613 | case 0x2003: /* EM SPACE */ |
5614 | case 0x2004: /* THREE-PER-EM SPACE */ |
5615 | case 0x2005: /* FOUR-PER-EM SPACE */ |
5616 | case 0x2006: /* SIX-PER-EM SPACE */ |
5617 | case 0x2007: /* FIGURE SPACE */ |
5618 | case 0x2008: /* PUNCTUATION SPACE */ |
5619 | case 0x2009: /* THIN SPACE */ |
5620 | case 0x200A: /* HAIR SPACE */ |
5621 | case 0x202f: /* NARROW NO-BREAK SPACE */ |
5622 | case 0x205f: /* MEDIUM MATHEMATICAL SPACE */ |
5623 | case 0x3000: /* IDEOGRAPHIC SPACE */ |
5624 | gotspace = TRUE; |
5625 | break; |
5626 | } |
5627 | if (gotspace == (ctype == OP_NOT_HSPACE)) break; |
5628 | eptr += len; |
5629 | } |
5630 | break; |
5631 | |
5632 | case OP_NOT_VSPACE: |
5633 | case OP_VSPACE: |
5634 | for (i = min; i < max; i++) |
5635 | { |
5636 | BOOL gotspace; |
5637 | int len = 1; |
5638 | if (eptr >= md->end_subject) |
5639 | { |
5640 | SCHECK_PARTIAL(); |
5641 | break; |
5642 | } |
5643 | GETCHARLEN(c, eptr, len); |
5644 | switch(c) |
5645 | { |
5646 | default: gotspace = FALSE; break; |
5647 | case 0x0a: /* LF */ |
5648 | case 0x0b: /* VT */ |
5649 | case 0x0c: /* FF */ |
5650 | case 0x0d: /* CR */ |
5651 | case 0x85: /* NEL */ |
5652 | case 0x2028: /* LINE SEPARATOR */ |
5653 | case 0x2029: /* PARAGRAPH SEPARATOR */ |
5654 | gotspace = TRUE; |
5655 | break; |
5656 | } |
5657 | if (gotspace == (ctype == OP_NOT_VSPACE)) break; |
5658 | eptr += len; |
5659 | } |
5660 | break; |
5661 | |
5662 | case OP_NOT_DIGIT: |
5663 | for (i = min; i < max; i++) |
5664 | { |
5665 | int len = 1; |
5666 | if (eptr >= md->end_subject) |
5667 | { |
5668 | SCHECK_PARTIAL(); |
5669 | break; |
5670 | } |
5671 | GETCHARLEN(c, eptr, len); |
5672 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break; |
5673 | eptr+= len; |
5674 | } |
5675 | break; |
5676 | |
5677 | case OP_DIGIT: |
5678 | for (i = min; i < max; i++) |
5679 | { |
5680 | int len = 1; |
5681 | if (eptr >= md->end_subject) |
5682 | { |
5683 | SCHECK_PARTIAL(); |
5684 | break; |
5685 | } |
5686 | GETCHARLEN(c, eptr, len); |
5687 | if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break; |
5688 | eptr+= len; |
5689 | } |
5690 | break; |
5691 | |
5692 | case OP_NOT_WHITESPACE: |
5693 | for (i = min; i < max; i++) |
5694 | { |
5695 | int len = 1; |
5696 | if (eptr >= md->end_subject) |
5697 | { |
5698 | SCHECK_PARTIAL(); |
5699 | break; |
5700 | } |
5701 | GETCHARLEN(c, eptr, len); |
5702 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break; |
5703 | eptr+= len; |
5704 | } |
5705 | break; |
5706 | |
5707 | case OP_WHITESPACE: |
5708 | for (i = min; i < max; i++) |
5709 | { |
5710 | int len = 1; |
5711 | if (eptr >= md->end_subject) |
5712 | { |
5713 | SCHECK_PARTIAL(); |
5714 | break; |
5715 | } |
5716 | GETCHARLEN(c, eptr, len); |
5717 | if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break; |
5718 | eptr+= len; |
5719 | } |
5720 | break; |
5721 | |
5722 | case OP_NOT_WORDCHAR: |
5723 | for (i = min; i < max; i++) |
5724 | { |
5725 | int len = 1; |
5726 | if (eptr >= md->end_subject) |
5727 | { |
5728 | SCHECK_PARTIAL(); |
5729 | break; |
5730 | } |
5731 | GETCHARLEN(c, eptr, len); |
5732 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break; |
5733 | eptr+= len; |
5734 | } |
5735 | break; |
5736 | |
5737 | case OP_WORDCHAR: |
5738 | for (i = min; i < max; i++) |
5739 | { |
5740 | int len = 1; |
5741 | if (eptr >= md->end_subject) |
5742 | { |
5743 | SCHECK_PARTIAL(); |
5744 | break; |
5745 | } |
5746 | GETCHARLEN(c, eptr, len); |
5747 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break; |
5748 | eptr+= len; |
5749 | } |
5750 | break; |
5751 | |
5752 | default: |
5753 | RRETURN(PCRE_ERROR_INTERNAL); |
5754 | } |
5755 | |
5756 | /* eptr is now past the end of the maximum run. If possessive, we are |
5757 | done (no backing up). Otherwise, match at this position; anything other |
5758 | than no match is immediately returned. For nomatch, back up one |
5759 | character, unless we are matching \R and the last thing matched was |
5760 | \r\n, in which case, back up two bytes. */ |
5761 | |
5762 | if (possessive) continue; |
5763 | for(;;) |
5764 | { |
5765 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM46); |
5766 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5767 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
5768 | BACKCHAR(eptr); |
5769 | if (ctype == OP_ANYNL && eptr > pp && *eptr == '\n' && |
5770 | eptr[-1] == '\r') eptr--; |
5771 | } |
5772 | } |
5773 | else |
5774 | #endif /* SUPPORT_UTF */ |
5775 | /* Not UTF mode */ |
5776 | { |
5777 | switch(ctype) |
5778 | { |
5779 | case OP_ANY: |
5780 | for (i = min; i < max; i++) |
5781 | { |
5782 | if (eptr >= md->end_subject) |
5783 | { |
5784 | SCHECK_PARTIAL(); |
5785 | break; |
5786 | } |
5787 | if (IS_NEWLINE(eptr)) break; |
5788 | eptr++; |
5789 | } |
5790 | break; |
5791 | |
5792 | case OP_ALLANY: |
5793 | case OP_ANYBYTE: |
5794 | c = max - min; |
5795 | if (c > (unsigned int)(md->end_subject - eptr)) |
5796 | { |
5797 | eptr = md->end_subject; |
5798 | SCHECK_PARTIAL(); |
5799 | } |
5800 | else eptr += c; |
5801 | break; |
5802 | |
5803 | case OP_ANYNL: |
5804 | for (i = min; i < max; i++) |
5805 | { |
5806 | if (eptr >= md->end_subject) |
5807 | { |
5808 | SCHECK_PARTIAL(); |
5809 | break; |
5810 | } |
5811 | c = *eptr; |
5812 | if (c == 0x000d) |
5813 | { |
5814 | if (++eptr >= md->end_subject) break; |
5815 | if (*eptr == 0x000a) eptr++; |
5816 | } |
5817 | else |
5818 | { |
5819 | if (c != 0x000a && (md->bsr_anycrlf || |
5820 | (c != 0x000b && c != 0x000c && c != 0x0085 |
5821 | #ifdef COMPILE_PCRE16 |
5822 | && c != 0x2028 && c != 0x2029 |
5823 | #endif |
5824 | ))) break; |
5825 | eptr++; |
5826 | } |
5827 | } |
5828 | break; |
5829 | |
5830 | case OP_NOT_HSPACE: |
5831 | for (i = min; i < max; i++) |
5832 | { |
5833 | if (eptr >= md->end_subject) |
5834 | { |
5835 | SCHECK_PARTIAL(); |
5836 | break; |
5837 | } |
5838 | c = *eptr; |
5839 | if (c == 0x09 || c == 0x20 || c == 0xa0 |
5840 | #ifdef COMPILE_PCRE16 |
5841 | || c == 0x1680 || c == 0x180e || (c >= 0x2000 && c <= 0x200A) |
5842 | || c == 0x202f || c == 0x205f || c == 0x3000 |
5843 | #endif |
5844 | ) break; |
5845 | eptr++; |
5846 | } |
5847 | break; |
5848 | |
5849 | case OP_HSPACE: |
5850 | for (i = min; i < max; i++) |
5851 | { |
5852 | if (eptr >= md->end_subject) |
5853 | { |
5854 | SCHECK_PARTIAL(); |
5855 | break; |
5856 | } |
5857 | c = *eptr; |
5858 | if (c != 0x09 && c != 0x20 && c != 0xa0 |
5859 | #ifdef COMPILE_PCRE16 |
5860 | && c != 0x1680 && c != 0x180e && (c < 0x2000 || c > 0x200A) |
5861 | && c != 0x202f && c != 0x205f && c != 0x3000 |
5862 | #endif |
5863 | ) break; |
5864 | eptr++; |
5865 | } |
5866 | break; |
5867 | |
5868 | case OP_NOT_VSPACE: |
5869 | for (i = min; i < max; i++) |
5870 | { |
5871 | if (eptr >= md->end_subject) |
5872 | { |
5873 | SCHECK_PARTIAL(); |
5874 | break; |
5875 | } |
5876 | c = *eptr; |
5877 | if (c == 0x0a || c == 0x0b || c == 0x0c || c == 0x0d || c == 0x85 |
5878 | #ifdef COMPILE_PCRE16 |
5879 | || c == 0x2028 || c == 0x2029 |
5880 | #endif |
5881 | ) break; |
5882 | eptr++; |
5883 | } |
5884 | break; |
5885 | |
5886 | case OP_VSPACE: |
5887 | for (i = min; i < max; i++) |
5888 | { |
5889 | if (eptr >= md->end_subject) |
5890 | { |
5891 | SCHECK_PARTIAL(); |
5892 | break; |
5893 | } |
5894 | c = *eptr; |
5895 | if (c != 0x0a && c != 0x0b && c != 0x0c && c != 0x0d && c != 0x85 |
5896 | #ifdef COMPILE_PCRE16 |
5897 | && c != 0x2028 && c != 0x2029 |
5898 | #endif |
5899 | ) break; |
5900 | eptr++; |
5901 | } |
5902 | break; |
5903 | |
5904 | case OP_NOT_DIGIT: |
5905 | for (i = min; i < max; i++) |
5906 | { |
5907 | if (eptr >= md->end_subject) |
5908 | { |
5909 | SCHECK_PARTIAL(); |
5910 | break; |
5911 | } |
5912 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) break; |
5913 | eptr++; |
5914 | } |
5915 | break; |
5916 | |
5917 | case OP_DIGIT: |
5918 | for (i = min; i < max; i++) |
5919 | { |
5920 | if (eptr >= md->end_subject) |
5921 | { |
5922 | SCHECK_PARTIAL(); |
5923 | break; |
5924 | } |
5925 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) break; |
5926 | eptr++; |
5927 | } |
5928 | break; |
5929 | |
5930 | case OP_NOT_WHITESPACE: |
5931 | for (i = min; i < max; i++) |
5932 | { |
5933 | if (eptr >= md->end_subject) |
5934 | { |
5935 | SCHECK_PARTIAL(); |
5936 | break; |
5937 | } |
5938 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break; |
5939 | eptr++; |
5940 | } |
5941 | break; |
5942 | |
5943 | case OP_WHITESPACE: |
5944 | for (i = min; i < max; i++) |
5945 | { |
5946 | if (eptr >= md->end_subject) |
5947 | { |
5948 | SCHECK_PARTIAL(); |
5949 | break; |
5950 | } |
5951 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break; |
5952 | eptr++; |
5953 | } |
5954 | break; |
5955 | |
5956 | case OP_NOT_WORDCHAR: |
5957 | for (i = min; i < max; i++) |
5958 | { |
5959 | if (eptr >= md->end_subject) |
5960 | { |
5961 | SCHECK_PARTIAL(); |
5962 | break; |
5963 | } |
5964 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break; |
5965 | eptr++; |
5966 | } |
5967 | break; |
5968 | |
5969 | case OP_WORDCHAR: |
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_word) == 0) break; |
5978 | eptr++; |
5979 | } |
5980 | break; |
5981 | |
5982 | default: |
5983 | RRETURN(PCRE_ERROR_INTERNAL); |
5984 | } |
5985 | |
5986 | /* eptr is now past the end of the maximum run. If possessive, we are |
5987 | done (no backing up). Otherwise, match at this position; anything other |
5988 | than no match is immediately returned. For nomatch, back up one |
5989 | character (byte), unless we are matching \R and the last thing matched |
5990 | was \r\n, in which case, back up two bytes. */ |
5991 | |
5992 | if (possessive) continue; |
5993 | while (eptr >= pp) |
5994 | { |
5995 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM47); |
5996 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5997 | eptr--; |
5998 | if (ctype == OP_ANYNL && eptr > pp && *eptr == '\n' && |
5999 | eptr[-1] == '\r') eptr--; |
6000 | } |
6001 | } |
6002 | |
6003 | /* Get here if we can't make it match with any permitted repetitions */ |
6004 | |
6005 | RRETURN(MATCH_NOMATCH); |
6006 | } |
6007 | /* Control never gets here */ |
6008 | |
6009 | /* There's been some horrible disaster. Arrival here can only mean there is |
6010 | something seriously wrong in the code above or the OP_xxx definitions. */ |
6011 | |
6012 | default: |
6013 | DPRINTF(("Unknown opcode %d\n", *ecode)); |
6014 | RRETURN(PCRE_ERROR_UNKNOWN_OPCODE); |
6015 | } |
6016 | |
6017 | /* Do not stick any code in here without much thought; it is assumed |
6018 | that "continue" in the code above comes out to here to repeat the main |
6019 | loop. */ |
6020 | |
6021 | } /* End of main loop */ |
6022 | /* Control never reaches here */ |
6023 | |
6024 | |
6025 | /* When compiling to use the heap rather than the stack for recursive calls to |
6026 | match(), the RRETURN() macro jumps here. The number that is saved in |
6027 | frame->Xwhere indicates which label we actually want to return to. */ |
6028 | |
6029 | #ifdef NO_RECURSE |
6030 | #define LBL(val) case val: goto L_RM##val; |
6031 | HEAP_RETURN: |
6032 | switch (frame->Xwhere) |
6033 | { |
6034 | LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8) |
6035 | LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17) |
6036 | LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33) |
6037 | LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52) |
6038 | LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64) |
6039 | LBL(65) LBL(66) |
6040 | #if defined SUPPORT_UTF || !defined COMPILE_PCRE8 |
6041 | LBL(21) |
6042 | #endif |
6043 | #ifdef SUPPORT_UTF |
6044 | LBL(16) LBL(18) LBL(20) |
6045 | LBL(22) LBL(23) LBL(28) LBL(30) |
6046 | LBL(32) LBL(34) LBL(42) LBL(46) |
6047 | #ifdef SUPPORT_UCP |
6048 | LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45) |
6049 | LBL(59) LBL(60) LBL(61) LBL(62) |
6050 | #endif /* SUPPORT_UCP */ |
6051 | #endif /* SUPPORT_UTF */ |
6052 | default: |
6053 | DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere)); |
6054 | |
6055 | printf("+++jump error in pcre match: label %d non-existent\n", frame->Xwhere); |
6056 | |
6057 | return PCRE_ERROR_INTERNAL; |
6058 | } |
6059 | #undef LBL |
6060 | #endif /* NO_RECURSE */ |
6061 | } |
6062 | |
6063 | |
6064 | /*************************************************************************** |
6065 | **************************************************************************** |
6066 | RECURSION IN THE match() FUNCTION |
6067 | |
6068 | Undefine all the macros that were defined above to handle this. */ |
6069 | |
6070 | #ifdef NO_RECURSE |
6071 | #undef eptr |
6072 | #undef ecode |
6073 | #undef mstart |
6074 | #undef offset_top |
6075 | #undef eptrb |
6076 | #undef flags |
6077 | |
6078 | #undef callpat |
6079 | #undef charptr |
6080 | #undef data |
6081 | #undef next |
6082 | #undef pp |
6083 | #undef prev |
6084 | #undef saved_eptr |
6085 | |
6086 | #undef new_recursive |
6087 | |
6088 | #undef cur_is_word |
6089 | #undef condition |
6090 | #undef prev_is_word |
6091 | |
6092 | #undef ctype |
6093 | #undef length |
6094 | #undef max |
6095 | #undef min |
6096 | #undef number |
6097 | #undef offset |
6098 | #undef op |
6099 | #undef save_capture_last |
6100 | #undef save_offset1 |
6101 | #undef save_offset2 |
6102 | #undef save_offset3 |
6103 | #undef stacksave |
6104 | |
6105 | #undef newptrb |
6106 | |
6107 | #endif |
6108 | |
6109 | /* These two are defined as macros in both cases */ |
6110 | |
6111 | #undef fc |
6112 | #undef fi |
6113 | |
6114 | /*************************************************************************** |
6115 | ***************************************************************************/ |
6116 | |
6117 | |
6118 | |
6119 | /************************************************* |
6120 | * Execute a Regular Expression * |
6121 | *************************************************/ |
6122 | |
6123 | /* This function applies a compiled re to a subject string and picks out |
6124 | portions of the string if it matches. Two elements in the vector are set for |
6125 | each substring: the offsets to the start and end of the substring. |
6126 | |
6127 | Arguments: |
6128 | argument_re points to the compiled expression |
6129 | extra_data points to extra data or is NULL |
6130 | subject points to the subject string |
6131 | length length of subject string (may contain binary zeros) |
6132 | start_offset where to start in the subject string |
6133 | options option bits |
6134 | offsets points to a vector of ints to be filled in with offsets |
6135 | offsetcount the number of elements in the vector |
6136 | |
6137 | Returns: > 0 => success; value is the number of elements filled in |
6138 | = 0 => success, but offsets is not big enough |
6139 | -1 => failed to match |
6140 | < -1 => some kind of unexpected problem |
6141 | */ |
6142 | |
6143 | #ifdef COMPILE_PCRE8 |
6144 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6145 | pcre_exec(const pcre *argument_re, const pcre_extra *extra_data, |
6146 | PCRE_SPTR subject, int length, int start_offset, int options, int *offsets, |
6147 | int offsetcount) |
6148 | #else |
6149 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6150 | pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data, |
6151 | PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets, |
6152 | int offsetcount) |
6153 | #endif |
6154 | { |
6155 | int rc, ocount, arg_offset_max; |
6156 | int newline; |
6157 | BOOL using_temporary_offsets = FALSE; |
6158 | BOOL anchored; |
6159 | BOOL startline; |
6160 | BOOL firstline; |
6161 | BOOL utf; |
6162 | BOOL has_first_char = FALSE; |
6163 | BOOL has_req_char = FALSE; |
6164 | pcre_uchar first_char = 0; |
6165 | pcre_uchar first_char2 = 0; |
6166 | pcre_uchar req_char = 0; |
6167 | pcre_uchar req_char2 = 0; |
6168 | match_data match_block; |
6169 | match_data *md = &match_block; |
6170 | const pcre_uint8 *tables; |
6171 | const pcre_uint8 *start_bits = NULL; |
6172 | PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset; |
6173 | PCRE_PUCHAR end_subject; |
6174 | PCRE_PUCHAR start_partial = NULL; |
6175 | PCRE_PUCHAR req_char_ptr = start_match - 1; |
6176 | |
6177 | const pcre_study_data *study; |
6178 | const REAL_PCRE *re = (const REAL_PCRE *)argument_re; |
6179 | |
6180 | /* Plausibility checks */ |
6181 | |
6182 | if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION; |
6183 | if (re == NULL || subject == NULL || |
6184 | (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL; |
6185 | if (offsetcount < 0) return PCRE_ERROR_BADCOUNT; |
6186 | if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET; |
6187 | |
6188 | /* These two settings are used in the code for checking a UTF-8 string that |
6189 | follows immediately afterwards. Other values in the md block are used only |
6190 | during "normal" pcre_exec() processing, not when the JIT support is in use, |
6191 | so they are set up later. */ |
6192 | |
6193 | /* PCRE_UTF16 has the same value as PCRE_UTF8. */ |
6194 | utf = md->utf = (re->options & PCRE_UTF8) != 0; |
6195 | md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 : |
6196 | ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0; |
6197 | |
6198 | /* Check a UTF-8 string if required. Pass back the character offset and error |
6199 | code for an invalid string if a results vector is available. */ |
6200 | |
6201 | #ifdef SUPPORT_UTF |
6202 | if (utf && (options & PCRE_NO_UTF8_CHECK) == 0) |
6203 | { |
6204 | int erroroffset; |
6205 | int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset); |
6206 | if (errorcode != 0) |
6207 | { |
6208 | if (offsetcount >= 2) |
6209 | { |
6210 | offsets[0] = erroroffset; |
6211 | offsets[1] = errorcode; |
6212 | } |
6213 | #ifdef COMPILE_PCRE16 |
6214 | return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)? |
6215 | PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16; |
6216 | #else |
6217 | return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)? |
6218 | PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8; |
6219 | #endif |
6220 | } |
6221 | |
6222 | /* Check that a start_offset points to the start of a UTF character. */ |
6223 | if (start_offset > 0 && start_offset < length && |
6224 | NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset])) |
6225 | return PCRE_ERROR_BADUTF8_OFFSET; |
6226 | } |
6227 | #endif |
6228 | |
6229 | /* If the pattern was successfully studied with JIT support, run the JIT |
6230 | executable instead of the rest of this function. Most options must be set at |
6231 | compile time for the JIT code to be usable. Fallback to the normal code path if |
6232 | an unsupported flag is set. In particular, JIT does not support partial |
6233 | matching. */ |
6234 | |
6235 | #ifdef SUPPORT_JIT |
6236 | if (extra_data != NULL |
6237 | && (extra_data->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 |
6238 | && extra_data->executable_jit != NULL |
6239 | && (extra_data->flags & PCRE_EXTRA_TABLES) == 0 |
6240 | && (options & ~(PCRE_NO_UTF8_CHECK | PCRE_NOTBOL | PCRE_NOTEOL | |
6241 | PCRE_NOTEMPTY | PCRE_NOTEMPTY_ATSTART)) == 0) |
6242 | return PRIV(jit_exec)(re, extra_data->executable_jit, |
6243 | (const pcre_uchar *)subject, length, start_offset, options, |
6244 | ((extra_data->flags & PCRE_EXTRA_MATCH_LIMIT) == 0) |
6245 | ? MATCH_LIMIT : extra_data->match_limit, offsets, offsetcount); |
6246 | #endif |
6247 | |
6248 | /* Carry on with non-JIT matching. This information is for finding all the |
6249 | numbers associated with a given name, for condition testing. */ |
6250 | |
6251 | md->name_table = (pcre_uchar *)re + re->name_table_offset; |
6252 | md->name_count = re->name_count; |
6253 | md->name_entry_size = re->name_entry_size; |
6254 | |
6255 | /* Fish out the optional data from the extra_data structure, first setting |
6256 | the default values. */ |
6257 | |
6258 | study = NULL; |
6259 | md->match_limit = MATCH_LIMIT; |
6260 | md->match_limit_recursion = MATCH_LIMIT_RECURSION; |
6261 | md->callout_data = NULL; |
6262 | |
6263 | /* The table pointer is always in native byte order. */ |
6264 | |
6265 | tables = re->tables; |
6266 | |
6267 | if (extra_data != NULL) |
6268 | { |
6269 | register unsigned int flags = extra_data->flags; |
6270 | if ((flags & PCRE_EXTRA_STUDY_DATA) != 0) |
6271 | study = (const pcre_study_data *)extra_data->study_data; |
6272 | if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0) |
6273 | md->match_limit = extra_data->match_limit; |
6274 | if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0) |
6275 | md->match_limit_recursion = extra_data->match_limit_recursion; |
6276 | if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0) |
6277 | md->callout_data = extra_data->callout_data; |
6278 | if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables; |
6279 | } |
6280 | |
6281 | /* If the exec call supplied NULL for tables, use the inbuilt ones. This |
6282 | is a feature that makes it possible to save compiled regex and re-use them |
6283 | in other programs later. */ |
6284 | |
6285 | if (tables == NULL) tables = PRIV(default_tables); |
6286 | |
6287 | /* Check that the first field in the block is the magic number. If it is not, |
6288 | return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to |
6289 | REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which |
6290 | means that the pattern is likely compiled with different endianness. */ |
6291 | |
6292 | if (re->magic_number != MAGIC_NUMBER) |
6293 | return re->magic_number == REVERSED_MAGIC_NUMBER? |
6294 | PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC; |
6295 | if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE; |
6296 | |
6297 | /* Set up other data */ |
6298 | |
6299 | anchored = ((re->options | options) & PCRE_ANCHORED) != 0; |
6300 | startline = (re->flags & PCRE_STARTLINE) != 0; |
6301 | firstline = (re->options & PCRE_FIRSTLINE) != 0; |
6302 | |
6303 | /* The code starts after the real_pcre block and the capture name table. */ |
6304 | |
6305 | md->start_code = (const pcre_uchar *)re + re->name_table_offset + |
6306 | re->name_count * re->name_entry_size; |
6307 | |
6308 | md->start_subject = (PCRE_PUCHAR)subject; |
6309 | md->start_offset = start_offset; |
6310 | md->end_subject = md->start_subject + length; |
6311 | end_subject = md->end_subject; |
6312 | |
6313 | md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0; |
6314 | md->use_ucp = (re->options & PCRE_UCP) != 0; |
6315 | md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0; |
6316 | md->ignore_skip_arg = FALSE; |
6317 | |
6318 | /* Some options are unpacked into BOOL variables in the hope that testing |
6319 | them will be faster than individual option bits. */ |
6320 | |
6321 | md->notbol = (options & PCRE_NOTBOL) != 0; |
6322 | md->noteol = (options & PCRE_NOTEOL) != 0; |
6323 | md->notempty = (options & PCRE_NOTEMPTY) != 0; |
6324 | md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0; |
6325 | |
6326 | md->hitend = FALSE; |
6327 | md->mark = md->nomatch_mark = NULL; /* In case never set */ |
6328 | |
6329 | md->recursive = NULL; /* No recursion at top level */ |
6330 | md->hasthen = (re->flags & PCRE_HASTHEN) != 0; |
6331 | |
6332 | md->lcc = tables + lcc_offset; |
6333 | md->fcc = tables + fcc_offset; |
6334 | md->ctypes = tables + ctypes_offset; |
6335 | |
6336 | /* Handle different \R options. */ |
6337 | |
6338 | switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) |
6339 | { |
6340 | case 0: |
6341 | if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0) |
6342 | md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0; |
6343 | else |
6344 | #ifdef BSR_ANYCRLF |
6345 | md->bsr_anycrlf = TRUE; |
6346 | #else |
6347 | md->bsr_anycrlf = FALSE; |
6348 | #endif |
6349 | break; |
6350 | |
6351 | case PCRE_BSR_ANYCRLF: |
6352 | md->bsr_anycrlf = TRUE; |
6353 | break; |
6354 | |
6355 | case PCRE_BSR_UNICODE: |
6356 | md->bsr_anycrlf = FALSE; |
6357 | break; |
6358 | |
6359 | default: return PCRE_ERROR_BADNEWLINE; |
6360 | } |
6361 | |
6362 | /* Handle different types of newline. The three bits give eight cases. If |
6363 | nothing is set at run time, whatever was used at compile time applies. */ |
6364 | |
6365 | switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options : |
6366 | (pcre_uint32)options) & PCRE_NEWLINE_BITS) |
6367 | { |
6368 | case 0: newline = NEWLINE; break; /* Compile-time default */ |
6369 | case PCRE_NEWLINE_CR: newline = CHAR_CR; break; |
6370 | case PCRE_NEWLINE_LF: newline = CHAR_NL; break; |
6371 | case PCRE_NEWLINE_CR+ |
6372 | PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break; |
6373 | case PCRE_NEWLINE_ANY: newline = -1; break; |
6374 | case PCRE_NEWLINE_ANYCRLF: newline = -2; break; |
6375 | default: return PCRE_ERROR_BADNEWLINE; |
6376 | } |
6377 | |
6378 | if (newline == -2) |
6379 | { |
6380 | md->nltype = NLTYPE_ANYCRLF; |
6381 | } |
6382 | else if (newline < 0) |
6383 | { |
6384 | md->nltype = NLTYPE_ANY; |
6385 | } |
6386 | else |
6387 | { |
6388 | md->nltype = NLTYPE_FIXED; |
6389 | if (newline > 255) |
6390 | { |
6391 | md->nllen = 2; |
6392 | md->nl[0] = (newline >> 8) & 255; |
6393 | md->nl[1] = newline & 255; |
6394 | } |
6395 | else |
6396 | { |
6397 | md->nllen = 1; |
6398 | md->nl[0] = newline; |
6399 | } |
6400 | } |
6401 | |
6402 | /* Partial matching was originally supported only for a restricted set of |
6403 | regexes; from release 8.00 there are no restrictions, but the bits are still |
6404 | defined (though never set). So there's no harm in leaving this code. */ |
6405 | |
6406 | if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0) |
6407 | return PCRE_ERROR_BADPARTIAL; |
6408 | |
6409 | /* If the expression has got more back references than the offsets supplied can |
6410 | hold, we get a temporary chunk of working store to use during the matching. |
6411 | Otherwise, we can use the vector supplied, rounding down its size to a multiple |
6412 | of 3. */ |
6413 | |
6414 | ocount = offsetcount - (offsetcount % 3); |
6415 | arg_offset_max = (2*ocount)/3; |
6416 | |
6417 | if (re->top_backref > 0 && re->top_backref >= ocount/3) |
6418 | { |
6419 | ocount = re->top_backref * 3 + 3; |
6420 | md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int)); |
6421 | if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY; |
6422 | using_temporary_offsets = TRUE; |
6423 | DPRINTF(("Got memory to hold back references\n")); |
6424 | } |
6425 | else md->offset_vector = offsets; |
6426 | |
6427 | md->offset_end = ocount; |
6428 | md->offset_max = (2*ocount)/3; |
6429 | md->offset_overflow = FALSE; |
6430 | md->capture_last = -1; |
6431 | |
6432 | /* Reset the working variable associated with each extraction. These should |
6433 | never be used unless previously set, but they get saved and restored, and so we |
6434 | initialize them to avoid reading uninitialized locations. Also, unset the |
6435 | offsets for the matched string. This is really just for tidiness with callouts, |
6436 | in case they inspect these fields. */ |
6437 | |
6438 | if (md->offset_vector != NULL) |
6439 | { |
6440 | register int *iptr = md->offset_vector + ocount; |
6441 | register int *iend = iptr - re->top_bracket; |
6442 | if (iend < md->offset_vector + 2) iend = md->offset_vector + 2; |
6443 | while (--iptr >= iend) *iptr = -1; |
6444 | md->offset_vector[0] = md->offset_vector[1] = -1; |
6445 | } |
6446 | |
6447 | /* Set up the first character to match, if available. The first_char value is |
6448 | never set for an anchored regular expression, but the anchoring may be forced |
6449 | at run time, so we have to test for anchoring. The first char may be unset for |
6450 | an unanchored pattern, of course. If there's no first char and the pattern was |
6451 | studied, there may be a bitmap of possible first characters. */ |
6452 | |
6453 | if (!anchored) |
6454 | { |
6455 | if ((re->flags & PCRE_FIRSTSET) != 0) |
6456 | { |
6457 | has_first_char = TRUE; |
6458 | first_char = first_char2 = re->first_char; |
6459 | if ((re->flags & PCRE_FCH_CASELESS) != 0) |
6460 | { |
6461 | first_char2 = TABLE_GET(first_char, md->fcc, first_char); |
6462 | #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8) |
6463 | if (utf && first_char > 127) |
6464 | first_char2 = UCD_OTHERCASE(first_char); |
6465 | #endif |
6466 | } |
6467 | } |
6468 | else |
6469 | if (!startline && study != NULL && |
6470 | (study->flags & PCRE_STUDY_MAPPED) != 0) |
6471 | start_bits = study->start_bits; |
6472 | } |
6473 | |
6474 | /* For anchored or unanchored matches, there may be a "last known required |
6475 | character" set. */ |
6476 | |
6477 | if ((re->flags & PCRE_REQCHSET) != 0) |
6478 | { |
6479 | has_req_char = TRUE; |
6480 | req_char = req_char2 = re->req_char; |
6481 | if ((re->flags & PCRE_RCH_CASELESS) != 0) |
6482 | { |
6483 | req_char2 = TABLE_GET(req_char, md->fcc, req_char); |
6484 | #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8) |
6485 | if (utf && req_char > 127) |
6486 | req_char2 = UCD_OTHERCASE(req_char); |
6487 | #endif |
6488 | } |
6489 | } |
6490 | |
6491 | |
6492 | /* ==========================================================================*/ |
6493 | |
6494 | /* Loop for handling unanchored repeated matching attempts; for anchored regexs |
6495 | the loop runs just once. */ |
6496 | |
6497 | for(;;) |
6498 | { |
6499 | PCRE_PUCHAR save_end_subject = end_subject; |
6500 | PCRE_PUCHAR new_start_match; |
6501 | |
6502 | /* If firstline is TRUE, the start of the match is constrained to the first |
6503 | line of a multiline string. That is, the match must be before or at the first |
6504 | newline. Implement this by temporarily adjusting end_subject so that we stop |
6505 | scanning at a newline. If the match fails at the newline, later code breaks |
6506 | this loop. */ |
6507 | |
6508 | if (firstline) |
6509 | { |
6510 | PCRE_PUCHAR t = start_match; |
6511 | #ifdef SUPPORT_UTF |
6512 | if (utf) |
6513 | { |
6514 | while (t < md->end_subject && !IS_NEWLINE(t)) |
6515 | { |
6516 | t++; |
6517 | ACROSSCHAR(t < end_subject, *t, t++); |
6518 | } |
6519 | } |
6520 | else |
6521 | #endif |
6522 | while (t < md->end_subject && !IS_NEWLINE(t)) t++; |
6523 | end_subject = t; |
6524 | } |
6525 | |
6526 | /* There are some optimizations that avoid running the match if a known |
6527 | starting point is not found, or if a known later character is not present. |
6528 | However, there is an option that disables these, for testing and for ensuring |
6529 | that all callouts do actually occur. The option can be set in the regex by |
6530 | (*NO_START_OPT) or passed in match-time options. */ |
6531 | |
6532 | if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0) |
6533 | { |
6534 | /* Advance to a unique first char if there is one. */ |
6535 | |
6536 | if (has_first_char) |
6537 | { |
6538 | if (first_char != first_char2) |
6539 | while (start_match < end_subject && |
6540 | *start_match != first_char && *start_match != first_char2) |
6541 | start_match++; |
6542 | else |
6543 | while (start_match < end_subject && *start_match != first_char) |
6544 | start_match++; |
6545 | } |
6546 | |
6547 | /* Or to just after a linebreak for a multiline match */ |
6548 | |
6549 | else if (startline) |
6550 | { |
6551 | if (start_match > md->start_subject + start_offset) |
6552 | { |
6553 | #ifdef SUPPORT_UTF |
6554 | if (utf) |
6555 | { |
6556 | while (start_match < end_subject && !WAS_NEWLINE(start_match)) |
6557 | { |
6558 | start_match++; |
6559 | ACROSSCHAR(start_match < end_subject, *start_match, |
6560 | start_match++); |
6561 | } |
6562 | } |
6563 | else |
6564 | #endif |
6565 | while (start_match < end_subject && !WAS_NEWLINE(start_match)) |
6566 | start_match++; |
6567 | |
6568 | /* If we have just passed a CR and the newline option is ANY or ANYCRLF, |
6569 | and we are now at a LF, advance the match position by one more character. |
6570 | */ |
6571 | |
6572 | if (start_match[-1] == CHAR_CR && |
6573 | (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) && |
6574 | start_match < end_subject && |
6575 | *start_match == CHAR_NL) |
6576 | start_match++; |
6577 | } |
6578 | } |
6579 | |
6580 | /* Or to a non-unique first byte after study */ |
6581 | |
6582 | else if (start_bits != NULL) |
6583 | { |
6584 | while (start_match < end_subject) |
6585 | { |
6586 | register unsigned int c = *start_match; |
6587 | #ifndef COMPILE_PCRE8 |
6588 | if (c > 255) c = 255; |
6589 | #endif |
6590 | if ((start_bits[c/8] & (1 << (c&7))) == 0) |
6591 | { |
6592 | start_match++; |
6593 | #if defined SUPPORT_UTF && defined COMPILE_PCRE8 |
6594 | /* In non 8-bit mode, the iteration will stop for |
6595 | characters > 255 at the beginning or not stop at all. */ |
6596 | if (utf) |
6597 | ACROSSCHAR(start_match < end_subject, *start_match, |
6598 | start_match++); |
6599 | #endif |
6600 | } |
6601 | else break; |
6602 | } |
6603 | } |
6604 | } /* Starting optimizations */ |
6605 | |
6606 | /* Restore fudged end_subject */ |
6607 | |
6608 | end_subject = save_end_subject; |
6609 | |
6610 | /* The following two optimizations are disabled for partial matching or if |
6611 | disabling is explicitly requested. */ |
6612 | |
6613 | if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0 && !md->partial) |
6614 | { |
6615 | /* If the pattern was studied, a minimum subject length may be set. This is |
6616 | a lower bound; no actual string of that length may actually match the |
6617 | pattern. Although the value is, strictly, in characters, we treat it as |
6618 | bytes to avoid spending too much time in this optimization. */ |
6619 | |
6620 | if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 && |
6621 | (pcre_uint32)(end_subject - start_match) < study->minlength) |
6622 | { |
6623 | rc = MATCH_NOMATCH; |
6624 | break; |
6625 | } |
6626 | |
6627 | /* If req_char is set, we know that that character must appear in the |
6628 | subject for the match to succeed. If the first character is set, req_char |
6629 | must be later in the subject; otherwise the test starts at the match point. |
6630 | This optimization can save a huge amount of backtracking in patterns with |
6631 | nested unlimited repeats that aren't going to match. Writing separate code |
6632 | for cased/caseless versions makes it go faster, as does using an |
6633 | autoincrement and backing off on a match. |
6634 | |
6635 | HOWEVER: when the subject string is very, very long, searching to its end |
6636 | can take a long time, and give bad performance on quite ordinary patterns. |
6637 | This showed up when somebody was matching something like /^\d+C/ on a |
6638 | 32-megabyte string... so we don't do this when the string is sufficiently |
6639 | long. */ |
6640 | |
6641 | if (has_req_char && end_subject - start_match < REQ_BYTE_MAX) |
6642 | { |
6643 | register PCRE_PUCHAR p = start_match + (has_first_char? 1:0); |
6644 | |
6645 | /* We don't need to repeat the search if we haven't yet reached the |
6646 | place we found it at last time. */ |
6647 | |
6648 | if (p > req_char_ptr) |
6649 | { |
6650 | if (req_char != req_char2) |
6651 | { |
6652 | while (p < end_subject) |
6653 | { |
6654 | register int pp = *p++; |
6655 | if (pp == req_char || pp == req_char2) { p--; break; } |
6656 | } |
6657 | } |
6658 | else |
6659 | { |
6660 | while (p < end_subject) |
6661 | { |
6662 | if (*p++ == req_char) { p--; break; } |
6663 | } |
6664 | } |
6665 | |
6666 | /* If we can't find the required character, break the matching loop, |
6667 | forcing a match failure. */ |
6668 | |
6669 | if (p >= end_subject) |
6670 | { |
6671 | rc = MATCH_NOMATCH; |
6672 | break; |
6673 | } |
6674 | |
6675 | /* If we have found the required character, save the point where we |
6676 | found it, so that we don't search again next time round the loop if |
6677 | the start hasn't passed this character yet. */ |
6678 | |
6679 | req_char_ptr = p; |
6680 | } |
6681 | } |
6682 | } |
6683 | |
6684 | #ifdef PCRE_DEBUG /* Sigh. Some compilers never learn. */ |
6685 | printf(">>>> Match against: "); |
6686 | pchars(start_match, end_subject - start_match, TRUE, md); |
6687 | printf("\n"); |
6688 | #endif |
6689 | |
6690 | /* OK, we can now run the match. If "hitend" is set afterwards, remember the |
6691 | first starting point for which a partial match was found. */ |
6692 | |
6693 | md->start_match_ptr = start_match; |
6694 | md->start_used_ptr = start_match; |
6695 | md->match_call_count = 0; |
6696 | md->match_function_type = 0; |
6697 | md->end_offset_top = 0; |
6698 | rc = match(start_match, md->start_code, start_match, 2, md, NULL, 0); |
6699 | if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr; |
6700 | |
6701 | switch(rc) |
6702 | { |
6703 | /* If MATCH_SKIP_ARG reaches this level it means that a MARK that matched |
6704 | the SKIP's arg was not found. In this circumstance, Perl ignores the SKIP |
6705 | entirely. The only way we can do that is to re-do the match at the same |
6706 | point, with a flag to force SKIP with an argument to be ignored. Just |
6707 | treating this case as NOMATCH does not work because it does not check other |
6708 | alternatives in patterns such as A(*SKIP:A)B|AC when the subject is AC. */ |
6709 | |
6710 | case MATCH_SKIP_ARG: |
6711 | new_start_match = start_match; |
6712 | md->ignore_skip_arg = TRUE; |
6713 | break; |
6714 | |
6715 | /* SKIP passes back the next starting point explicitly, but if it is the |
6716 | same as the match we have just done, treat it as NOMATCH. */ |
6717 | |
6718 | case MATCH_SKIP: |
6719 | if (md->start_match_ptr != start_match) |
6720 | { |
6721 | new_start_match = md->start_match_ptr; |
6722 | break; |
6723 | } |
6724 | /* Fall through */ |
6725 | |
6726 | /* NOMATCH and PRUNE advance by one character. THEN at this level acts |
6727 | exactly like PRUNE. Unset the ignore SKIP-with-argument flag. */ |
6728 | |
6729 | case MATCH_NOMATCH: |
6730 | case MATCH_PRUNE: |
6731 | case MATCH_THEN: |
6732 | md->ignore_skip_arg = FALSE; |
6733 | new_start_match = start_match + 1; |
6734 | #ifdef SUPPORT_UTF |
6735 | if (utf) |
6736 | ACROSSCHAR(new_start_match < end_subject, *new_start_match, |
6737 | new_start_match++); |
6738 | #endif |
6739 | break; |
6740 | |
6741 | /* COMMIT disables the bumpalong, but otherwise behaves as NOMATCH. */ |
6742 | |
6743 | case MATCH_COMMIT: |
6744 | rc = MATCH_NOMATCH; |
6745 | goto ENDLOOP; |