Parent Directory
|
Revision Log
Improved support for match_limit in JIT.
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) 1997-2013 University of Cambridge |
10 | |
11 | ----------------------------------------------------------------------------- |
12 | Redistribution and use in source and binary forms, with or without |
13 | modification, are permitted provided that the following conditions are met: |
14 | |
15 | * Redistributions of source code must retain the above copyright notice, |
16 | this list of conditions and the following disclaimer. |
17 | |
18 | * Redistributions in binary form must reproduce the above copyright |
19 | notice, this list of conditions and the following disclaimer in the |
20 | documentation and/or other materials provided with the distribution. |
21 | |
22 | * Neither the name of the University of Cambridge nor the names of its |
23 | contributors may be used to endorse or promote products derived from |
24 | this software without specific prior written permission. |
25 | |
26 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
27 | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
28 | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
29 | ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
30 | LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
31 | CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
32 | SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
33 | INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
34 | CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
35 | ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
36 | POSSIBILITY OF SUCH DAMAGE. |
37 | ----------------------------------------------------------------------------- |
38 | */ |
39 | |
40 | /* This module contains pcre_exec(), the externally visible function that does |
41 | pattern matching using an NFA algorithm, trying to mimic Perl as closely as |
42 | possible. There are also some static supporting functions. */ |
43 | |
44 | #ifdef HAVE_CONFIG_H |
45 | #include "config.h" |
46 | #endif |
47 | |
48 | #define NLBLOCK md /* Block containing newline information */ |
49 | #define PSSTART start_subject /* Field containing processed string start */ |
50 | #define PSEND end_subject /* Field containing processed string end */ |
51 | |
52 | #include "pcre_internal.h" |
53 | |
54 | /* Undefine some potentially clashing cpp symbols */ |
55 | |
56 | #undef min |
57 | #undef max |
58 | |
59 | /* The md->capture_last field uses the lower 16 bits for the last captured |
60 | substring (which can never be greater than 65535) and a bit in the top half |
61 | to mean "capture vector overflowed". This odd way of doing things was |
62 | implemented when it was realized that preserving and restoring the overflow bit |
63 | whenever the last capture number was saved/restored made for a neater |
64 | interface, and doing it this way saved on (a) another variable, which would |
65 | have increased the stack frame size (a big NO-NO in PCRE) and (b) another |
66 | separate set of save/restore instructions. The following defines are used in |
67 | implementing this. */ |
68 | |
69 | #define CAPLMASK 0x0000ffff /* The bits used for last_capture */ |
70 | #define OVFLMASK 0xffff0000 /* The bits used for the overflow flag */ |
71 | #define OVFLBIT 0x00010000 /* The bit that is set for overflow */ |
72 | |
73 | /* Values for setting in md->match_function_type to indicate two special types |
74 | of call to match(). We do it this way to save on using another stack variable, |
75 | as stack usage is to be discouraged. */ |
76 | |
77 | #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */ |
78 | #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */ |
79 | |
80 | /* Non-error returns from the match() function. Error returns are externally |
81 | defined PCRE_ERROR_xxx codes, which are all negative. */ |
82 | |
83 | #define MATCH_MATCH 1 |
84 | #define MATCH_NOMATCH 0 |
85 | |
86 | /* Special internal returns from the match() function. Make them sufficiently |
87 | negative to avoid the external error codes. */ |
88 | |
89 | #define MATCH_ACCEPT (-999) |
90 | #define MATCH_KETRPOS (-998) |
91 | #define MATCH_ONCE (-997) |
92 | /* The next 5 must be kept together and in sequence so that a test that checks |
93 | for any one of them can use a range. */ |
94 | #define MATCH_COMMIT (-996) |
95 | #define MATCH_PRUNE (-995) |
96 | #define MATCH_SKIP (-994) |
97 | #define MATCH_SKIP_ARG (-993) |
98 | #define MATCH_THEN (-992) |
99 | #define MATCH_BACKTRACK_MAX MATCH_THEN |
100 | #define MATCH_BACKTRACK_MIN MATCH_COMMIT |
101 | |
102 | /* Maximum number of ints of offset to save on the stack for recursive calls. |
103 | If the offset vector is bigger, malloc is used. This should be a multiple of 3, |
104 | because the offset vector is always a multiple of 3 long. */ |
105 | |
106 | #define REC_STACK_SAVE_MAX 30 |
107 | |
108 | /* Min and max values for the common repeats; for the maxima, 0 => infinity */ |
109 | |
110 | static const char rep_min[] = { 0, 0, 1, 1, 0, 0 }; |
111 | static const char rep_max[] = { 0, 0, 0, 0, 1, 1 }; |
112 | |
113 | #ifdef PCRE_DEBUG |
114 | /************************************************* |
115 | * Debugging function to print chars * |
116 | *************************************************/ |
117 | |
118 | /* Print a sequence of chars in printable format, stopping at the end of the |
119 | subject if the requested. |
120 | |
121 | Arguments: |
122 | p points to characters |
123 | length number to print |
124 | is_subject TRUE if printing from within md->start_subject |
125 | md pointer to matching data block, if is_subject is TRUE |
126 | |
127 | Returns: nothing |
128 | */ |
129 | |
130 | static void |
131 | pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md) |
132 | { |
133 | pcre_uint32 c; |
134 | BOOL utf = md->utf; |
135 | if (is_subject && length > md->end_subject - p) length = md->end_subject - p; |
136 | while (length-- > 0) |
137 | if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c); |
138 | } |
139 | #endif |
140 | |
141 | |
142 | |
143 | /************************************************* |
144 | * Match a back-reference * |
145 | *************************************************/ |
146 | |
147 | /* Normally, if a back reference hasn't been set, the length that is passed is |
148 | negative, so the match always fails. However, in JavaScript compatibility mode, |
149 | the length passed is zero. Note that in caseless UTF-8 mode, the number of |
150 | subject bytes matched may be different to the number of reference bytes. |
151 | |
152 | Arguments: |
153 | offset index into the offset vector |
154 | eptr pointer into the subject |
155 | length length of reference to be matched (number of bytes) |
156 | md points to match data block |
157 | caseless TRUE if caseless |
158 | |
159 | Returns: >= 0 the number of subject bytes matched |
160 | -1 no match |
161 | -2 partial match; always given if at end subject |
162 | */ |
163 | |
164 | static int |
165 | match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md, |
166 | BOOL caseless) |
167 | { |
168 | PCRE_PUCHAR eptr_start = eptr; |
169 | register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset]; |
170 | #ifdef SUPPORT_UTF |
171 | BOOL utf = md->utf; |
172 | #endif |
173 | |
174 | #ifdef PCRE_DEBUG |
175 | if (eptr >= md->end_subject) |
176 | printf("matching subject <null>"); |
177 | else |
178 | { |
179 | printf("matching subject "); |
180 | pchars(eptr, length, TRUE, md); |
181 | } |
182 | printf(" against backref "); |
183 | pchars(p, length, FALSE, md); |
184 | printf("\n"); |
185 | #endif |
186 | |
187 | /* Always fail if reference not set (and not JavaScript compatible - in that |
188 | case the length is passed as zero). */ |
189 | |
190 | if (length < 0) return -1; |
191 | |
192 | /* Separate the caseless case for speed. In UTF-8 mode we can only do this |
193 | properly if Unicode properties are supported. Otherwise, we can check only |
194 | ASCII characters. */ |
195 | |
196 | if (caseless) |
197 | { |
198 | #ifdef SUPPORT_UTF |
199 | #ifdef SUPPORT_UCP |
200 | if (utf) |
201 | { |
202 | /* Match characters up to the end of the reference. NOTE: the number of |
203 | data units matched may differ, because in UTF-8 there are some characters |
204 | whose upper and lower case versions code have different numbers of bytes. |
205 | For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65 |
206 | (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a |
207 | sequence of two of the latter. It is important, therefore, to check the |
208 | length along the reference, not along the subject (earlier code did this |
209 | wrong). */ |
210 | |
211 | PCRE_PUCHAR endptr = p + length; |
212 | while (p < endptr) |
213 | { |
214 | pcre_uint32 c, d; |
215 | const ucd_record *ur; |
216 | if (eptr >= md->end_subject) return -2; /* Partial match */ |
217 | GETCHARINC(c, eptr); |
218 | GETCHARINC(d, p); |
219 | ur = GET_UCD(d); |
220 | if (c != d && c != d + ur->other_case) |
221 | { |
222 | const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset; |
223 | for (;;) |
224 | { |
225 | if (c < *pp) return -1; |
226 | if (c == *pp++) break; |
227 | } |
228 | } |
229 | } |
230 | } |
231 | else |
232 | #endif |
233 | #endif |
234 | |
235 | /* The same code works when not in UTF-8 mode and in UTF-8 mode when there |
236 | is no UCP support. */ |
237 | { |
238 | while (length-- > 0) |
239 | { |
240 | pcre_uint32 cc, cp; |
241 | if (eptr >= md->end_subject) return -2; /* Partial match */ |
242 | cc = RAWUCHARTEST(eptr); |
243 | cp = RAWUCHARTEST(p); |
244 | if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1; |
245 | p++; |
246 | eptr++; |
247 | } |
248 | } |
249 | } |
250 | |
251 | /* In the caseful case, we can just compare the bytes, whether or not we |
252 | are in UTF-8 mode. */ |
253 | |
254 | else |
255 | { |
256 | while (length-- > 0) |
257 | { |
258 | if (eptr >= md->end_subject) return -2; /* Partial match */ |
259 | if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1; |
260 | } |
261 | } |
262 | |
263 | return (int)(eptr - eptr_start); |
264 | } |
265 | |
266 | |
267 | |
268 | /*************************************************************************** |
269 | **************************************************************************** |
270 | RECURSION IN THE match() FUNCTION |
271 | |
272 | The match() function is highly recursive, though not every recursive call |
273 | increases the recursive depth. Nevertheless, some regular expressions can cause |
274 | it to recurse to a great depth. I was writing for Unix, so I just let it call |
275 | itself recursively. This uses the stack for saving everything that has to be |
276 | saved for a recursive call. On Unix, the stack can be large, and this works |
277 | fine. |
278 | |
279 | It turns out that on some non-Unix-like systems there are problems with |
280 | programs that use a lot of stack. (This despite the fact that every last chip |
281 | has oodles of memory these days, and techniques for extending the stack have |
282 | been known for decades.) So.... |
283 | |
284 | There is a fudge, triggered by defining NO_RECURSE, which avoids recursive |
285 | calls by keeping local variables that need to be preserved in blocks of memory |
286 | obtained from malloc() instead instead of on the stack. Macros are used to |
287 | achieve this so that the actual code doesn't look very different to what it |
288 | always used to. |
289 | |
290 | The original heap-recursive code used longjmp(). However, it seems that this |
291 | can be very slow on some operating systems. Following a suggestion from Stan |
292 | Switzer, the use of longjmp() has been abolished, at the cost of having to |
293 | provide a unique number for each call to RMATCH. There is no way of generating |
294 | a sequence of numbers at compile time in C. I have given them names, to make |
295 | them stand out more clearly. |
296 | |
297 | Crude tests on x86 Linux show a small speedup of around 5-8%. However, on |
298 | FreeBSD, avoiding longjmp() more than halves the time taken to run the standard |
299 | tests. Furthermore, not using longjmp() means that local dynamic variables |
300 | don't have indeterminate values; this has meant that the frame size can be |
301 | reduced because the result can be "passed back" by straight setting of the |
302 | variable instead of being passed in the frame. |
303 | **************************************************************************** |
304 | ***************************************************************************/ |
305 | |
306 | /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN |
307 | below must be updated in sync. */ |
308 | |
309 | enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10, |
310 | RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20, |
311 | RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30, |
312 | RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40, |
313 | RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50, |
314 | RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60, |
315 | RM61, RM62, RM63, RM64, RM65, RM66, RM67, RM68 }; |
316 | |
317 | /* These versions of the macros use the stack, as normal. There are debugging |
318 | versions and production versions. Note that the "rw" argument of RMATCH isn't |
319 | actually used in this definition. */ |
320 | |
321 | #ifndef NO_RECURSE |
322 | #define REGISTER register |
323 | |
324 | #ifdef PCRE_DEBUG |
325 | #define RMATCH(ra,rb,rc,rd,re,rw) \ |
326 | { \ |
327 | printf("match() called in line %d\n", __LINE__); \ |
328 | rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \ |
329 | printf("to line %d\n", __LINE__); \ |
330 | } |
331 | #define RRETURN(ra) \ |
332 | { \ |
333 | printf("match() returned %d from line %d\n", ra, __LINE__); \ |
334 | return ra; \ |
335 | } |
336 | #else |
337 | #define RMATCH(ra,rb,rc,rd,re,rw) \ |
338 | rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1) |
339 | #define RRETURN(ra) return ra |
340 | #endif |
341 | |
342 | #else |
343 | |
344 | |
345 | /* These versions of the macros manage a private stack on the heap. Note that |
346 | the "rd" argument of RMATCH isn't actually used in this definition. It's the md |
347 | argument of match(), which never changes. */ |
348 | |
349 | #define REGISTER |
350 | |
351 | #define RMATCH(ra,rb,rc,rd,re,rw)\ |
352 | {\ |
353 | heapframe *newframe = frame->Xnextframe;\ |
354 | if (newframe == NULL)\ |
355 | {\ |
356 | newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\ |
357 | if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\ |
358 | newframe->Xnextframe = NULL;\ |
359 | frame->Xnextframe = newframe;\ |
360 | }\ |
361 | frame->Xwhere = rw;\ |
362 | newframe->Xeptr = ra;\ |
363 | newframe->Xecode = rb;\ |
364 | newframe->Xmstart = mstart;\ |
365 | newframe->Xoffset_top = rc;\ |
366 | newframe->Xeptrb = re;\ |
367 | newframe->Xrdepth = frame->Xrdepth + 1;\ |
368 | newframe->Xprevframe = frame;\ |
369 | frame = newframe;\ |
370 | DPRINTF(("restarting from line %d\n", __LINE__));\ |
371 | goto HEAP_RECURSE;\ |
372 | L_##rw:\ |
373 | DPRINTF(("jumped back to line %d\n", __LINE__));\ |
374 | } |
375 | |
376 | #define RRETURN(ra)\ |
377 | {\ |
378 | heapframe *oldframe = frame;\ |
379 | frame = oldframe->Xprevframe;\ |
380 | if (frame != NULL)\ |
381 | {\ |
382 | rrc = ra;\ |
383 | goto HEAP_RETURN;\ |
384 | }\ |
385 | return ra;\ |
386 | } |
387 | |
388 | |
389 | /* Structure for remembering the local variables in a private frame */ |
390 | |
391 | typedef struct heapframe { |
392 | struct heapframe *Xprevframe; |
393 | struct heapframe *Xnextframe; |
394 | |
395 | /* Function arguments that may change */ |
396 | |
397 | PCRE_PUCHAR Xeptr; |
398 | const pcre_uchar *Xecode; |
399 | PCRE_PUCHAR Xmstart; |
400 | int Xoffset_top; |
401 | eptrblock *Xeptrb; |
402 | unsigned int Xrdepth; |
403 | |
404 | /* Function local variables */ |
405 | |
406 | PCRE_PUCHAR Xcallpat; |
407 | #ifdef SUPPORT_UTF |
408 | PCRE_PUCHAR Xcharptr; |
409 | #endif |
410 | PCRE_PUCHAR Xdata; |
411 | PCRE_PUCHAR Xnext; |
412 | PCRE_PUCHAR Xpp; |
413 | PCRE_PUCHAR Xprev; |
414 | PCRE_PUCHAR Xsaved_eptr; |
415 | |
416 | recursion_info Xnew_recursive; |
417 | |
418 | BOOL Xcur_is_word; |
419 | BOOL Xcondition; |
420 | BOOL Xprev_is_word; |
421 | |
422 | #ifdef SUPPORT_UCP |
423 | int Xprop_type; |
424 | unsigned int Xprop_value; |
425 | int Xprop_fail_result; |
426 | int Xoclength; |
427 | pcre_uchar Xocchars[6]; |
428 | #endif |
429 | |
430 | int Xcodelink; |
431 | int Xctype; |
432 | unsigned int Xfc; |
433 | int Xfi; |
434 | int Xlength; |
435 | int Xmax; |
436 | int Xmin; |
437 | unsigned int Xnumber; |
438 | int Xoffset; |
439 | unsigned int Xop; |
440 | pcre_int32 Xsave_capture_last; |
441 | int Xsave_offset1, Xsave_offset2, Xsave_offset3; |
442 | int Xstacksave[REC_STACK_SAVE_MAX]; |
443 | |
444 | eptrblock Xnewptrb; |
445 | |
446 | /* Where to jump back to */ |
447 | |
448 | int Xwhere; |
449 | |
450 | } heapframe; |
451 | |
452 | #endif |
453 | |
454 | |
455 | /*************************************************************************** |
456 | ***************************************************************************/ |
457 | |
458 | |
459 | |
460 | /************************************************* |
461 | * Match from current position * |
462 | *************************************************/ |
463 | |
464 | /* This function is called recursively in many circumstances. Whenever it |
465 | returns a negative (error) response, the outer incarnation must also return the |
466 | same response. */ |
467 | |
468 | /* These macros pack up tests that are used for partial matching, and which |
469 | appear several times in the code. We set the "hit end" flag if the pointer is |
470 | at the end of the subject and also past the start of the subject (i.e. |
471 | something has been matched). For hard partial matching, we then return |
472 | immediately. The second one is used when we already know we are past the end of |
473 | the subject. */ |
474 | |
475 | #define CHECK_PARTIAL()\ |
476 | if (md->partial != 0 && eptr >= md->end_subject && \ |
477 | eptr > md->start_used_ptr) \ |
478 | { \ |
479 | md->hitend = TRUE; \ |
480 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \ |
481 | } |
482 | |
483 | #define SCHECK_PARTIAL()\ |
484 | if (md->partial != 0 && eptr > md->start_used_ptr) \ |
485 | { \ |
486 | md->hitend = TRUE; \ |
487 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \ |
488 | } |
489 | |
490 | |
491 | /* Performance note: It might be tempting to extract commonly used fields from |
492 | the md structure (e.g. utf, end_subject) into individual variables to improve |
493 | performance. Tests using gcc on a SPARC disproved this; in the first case, it |
494 | made performance worse. |
495 | |
496 | Arguments: |
497 | eptr pointer to current character in subject |
498 | ecode pointer to current position in compiled code |
499 | mstart pointer to the current match start position (can be modified |
500 | by encountering \K) |
501 | offset_top current top pointer |
502 | md pointer to "static" info for the match |
503 | eptrb pointer to chain of blocks containing eptr at start of |
504 | brackets - for testing for empty matches |
505 | rdepth the recursion depth |
506 | |
507 | Returns: MATCH_MATCH if matched ) these values are >= 0 |
508 | MATCH_NOMATCH if failed to match ) |
509 | a negative MATCH_xxx value for PRUNE, SKIP, etc |
510 | a negative PCRE_ERROR_xxx value if aborted by an error condition |
511 | (e.g. stopped by repeated call or recursion limit) |
512 | */ |
513 | |
514 | static int |
515 | match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode, |
516 | PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb, |
517 | unsigned int rdepth) |
518 | { |
519 | /* These variables do not need to be preserved over recursion in this function, |
520 | so they can be ordinary variables in all cases. Mark some of them with |
521 | "register" because they are used a lot in loops. */ |
522 | |
523 | register int rrc; /* Returns from recursive calls */ |
524 | register int i; /* Used for loops not involving calls to RMATCH() */ |
525 | register pcre_uint32 c; /* Character values not kept over RMATCH() calls */ |
526 | register BOOL utf; /* Local copy of UTF flag for speed */ |
527 | |
528 | BOOL minimize, possessive; /* Quantifier options */ |
529 | BOOL caseless; |
530 | int condcode; |
531 | |
532 | /* When recursion is not being used, all "local" variables that have to be |
533 | preserved over calls to RMATCH() are part of a "frame". We set up the top-level |
534 | frame on the stack here; subsequent instantiations are obtained from the heap |
535 | whenever RMATCH() does a "recursion". See the macro definitions above. Putting |
536 | the top-level on the stack rather than malloc-ing them all gives a performance |
537 | boost in many cases where there is not much "recursion". */ |
538 | |
539 | #ifdef NO_RECURSE |
540 | heapframe *frame = (heapframe *)md->match_frames_base; |
541 | |
542 | /* Copy in the original argument variables */ |
543 | |
544 | frame->Xeptr = eptr; |
545 | frame->Xecode = ecode; |
546 | frame->Xmstart = mstart; |
547 | frame->Xoffset_top = offset_top; |
548 | frame->Xeptrb = eptrb; |
549 | frame->Xrdepth = rdepth; |
550 | |
551 | /* This is where control jumps back to to effect "recursion" */ |
552 | |
553 | HEAP_RECURSE: |
554 | |
555 | /* Macros make the argument variables come from the current frame */ |
556 | |
557 | #define eptr frame->Xeptr |
558 | #define ecode frame->Xecode |
559 | #define mstart frame->Xmstart |
560 | #define offset_top frame->Xoffset_top |
561 | #define eptrb frame->Xeptrb |
562 | #define rdepth frame->Xrdepth |
563 | |
564 | /* Ditto for the local variables */ |
565 | |
566 | #ifdef SUPPORT_UTF |
567 | #define charptr frame->Xcharptr |
568 | #endif |
569 | #define callpat frame->Xcallpat |
570 | #define codelink frame->Xcodelink |
571 | #define data frame->Xdata |
572 | #define next frame->Xnext |
573 | #define pp frame->Xpp |
574 | #define prev frame->Xprev |
575 | #define saved_eptr frame->Xsaved_eptr |
576 | |
577 | #define new_recursive frame->Xnew_recursive |
578 | |
579 | #define cur_is_word frame->Xcur_is_word |
580 | #define condition frame->Xcondition |
581 | #define prev_is_word frame->Xprev_is_word |
582 | |
583 | #ifdef SUPPORT_UCP |
584 | #define prop_type frame->Xprop_type |
585 | #define prop_value frame->Xprop_value |
586 | #define prop_fail_result frame->Xprop_fail_result |
587 | #define oclength frame->Xoclength |
588 | #define occhars frame->Xocchars |
589 | #endif |
590 | |
591 | #define ctype frame->Xctype |
592 | #define fc frame->Xfc |
593 | #define fi frame->Xfi |
594 | #define length frame->Xlength |
595 | #define max frame->Xmax |
596 | #define min frame->Xmin |
597 | #define number frame->Xnumber |
598 | #define offset frame->Xoffset |
599 | #define op frame->Xop |
600 | #define save_capture_last frame->Xsave_capture_last |
601 | #define save_offset1 frame->Xsave_offset1 |
602 | #define save_offset2 frame->Xsave_offset2 |
603 | #define save_offset3 frame->Xsave_offset3 |
604 | #define stacksave frame->Xstacksave |
605 | |
606 | #define newptrb frame->Xnewptrb |
607 | |
608 | /* When recursion is being used, local variables are allocated on the stack and |
609 | get preserved during recursion in the normal way. In this environment, fi and |
610 | i, and fc and c, can be the same variables. */ |
611 | |
612 | #else /* NO_RECURSE not defined */ |
613 | #define fi i |
614 | #define fc c |
615 | |
616 | /* Many of the following variables are used only in small blocks of the code. |
617 | My normal style of coding would have declared them within each of those blocks. |
618 | However, in order to accommodate the version of this code that uses an external |
619 | "stack" implemented on the heap, it is easier to declare them all here, so the |
620 | declarations can be cut out in a block. The only declarations within blocks |
621 | below are for variables that do not have to be preserved over a recursive call |
622 | to RMATCH(). */ |
623 | |
624 | #ifdef SUPPORT_UTF |
625 | const pcre_uchar *charptr; |
626 | #endif |
627 | const pcre_uchar *callpat; |
628 | const pcre_uchar *data; |
629 | const pcre_uchar *next; |
630 | PCRE_PUCHAR pp; |
631 | const pcre_uchar *prev; |
632 | PCRE_PUCHAR saved_eptr; |
633 | |
634 | recursion_info new_recursive; |
635 | |
636 | BOOL cur_is_word; |
637 | BOOL condition; |
638 | BOOL prev_is_word; |
639 | |
640 | #ifdef SUPPORT_UCP |
641 | int prop_type; |
642 | unsigned int prop_value; |
643 | int prop_fail_result; |
644 | int oclength; |
645 | pcre_uchar occhars[6]; |
646 | #endif |
647 | |
648 | int codelink; |
649 | int ctype; |
650 | int length; |
651 | int max; |
652 | int min; |
653 | unsigned int number; |
654 | int offset; |
655 | unsigned int op; |
656 | pcre_int32 save_capture_last; |
657 | int save_offset1, save_offset2, save_offset3; |
658 | int stacksave[REC_STACK_SAVE_MAX]; |
659 | |
660 | eptrblock newptrb; |
661 | |
662 | /* There is a special fudge for calling match() in a way that causes it to |
663 | measure the size of its basic stack frame when the stack is being used for |
664 | recursion. The second argument (ecode) being NULL triggers this behaviour. It |
665 | cannot normally ever be NULL. The return is the negated value of the frame |
666 | size. */ |
667 | |
668 | if (ecode == NULL) |
669 | { |
670 | if (rdepth == 0) |
671 | return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1); |
672 | else |
673 | { |
674 | int len = (char *)&rdepth - (char *)eptr; |
675 | return (len > 0)? -len : len; |
676 | } |
677 | } |
678 | #endif /* NO_RECURSE */ |
679 | |
680 | /* To save space on the stack and in the heap frame, I have doubled up on some |
681 | of the local variables that are used only in localised parts of the code, but |
682 | still need to be preserved over recursive calls of match(). These macros define |
683 | the alternative names that are used. */ |
684 | |
685 | #define allow_zero cur_is_word |
686 | #define cbegroup condition |
687 | #define code_offset codelink |
688 | #define condassert condition |
689 | #define matched_once prev_is_word |
690 | #define foc number |
691 | #define save_mark data |
692 | |
693 | /* These statements are here to stop the compiler complaining about unitialized |
694 | variables. */ |
695 | |
696 | #ifdef SUPPORT_UCP |
697 | prop_value = 0; |
698 | prop_fail_result = 0; |
699 | #endif |
700 | |
701 | |
702 | /* This label is used for tail recursion, which is used in a few cases even |
703 | when NO_RECURSE is not defined, in order to reduce the amount of stack that is |
704 | used. Thanks to Ian Taylor for noticing this possibility and sending the |
705 | original patch. */ |
706 | |
707 | TAIL_RECURSE: |
708 | |
709 | /* OK, now we can get on with the real code of the function. Recursive calls |
710 | are specified by the macro RMATCH and RRETURN is used to return. When |
711 | NO_RECURSE is *not* defined, these just turn into a recursive call to match() |
712 | and a "return", respectively (possibly with some debugging if PCRE_DEBUG is |
713 | defined). However, RMATCH isn't like a function call because it's quite a |
714 | complicated macro. It has to be used in one particular way. This shouldn't, |
715 | however, impact performance when true recursion is being used. */ |
716 | |
717 | #ifdef SUPPORT_UTF |
718 | utf = md->utf; /* Local copy of the flag */ |
719 | #else |
720 | utf = FALSE; |
721 | #endif |
722 | |
723 | /* First check that we haven't called match() too many times, or that we |
724 | haven't exceeded the recursive call limit. */ |
725 | |
726 | if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT); |
727 | if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT); |
728 | |
729 | /* At the start of a group with an unlimited repeat that may match an empty |
730 | string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is |
731 | done this way to save having to use another function argument, which would take |
732 | up space on the stack. See also MATCH_CONDASSERT below. |
733 | |
734 | When MATCH_CBEGROUP is set, add the current subject pointer to the chain of |
735 | such remembered pointers, to be checked when we hit the closing ket, in order |
736 | to break infinite loops that match no characters. When match() is called in |
737 | other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must |
738 | NOT be used with tail recursion, because the memory block that is used is on |
739 | the stack, so a new one may be required for each match(). */ |
740 | |
741 | if (md->match_function_type == MATCH_CBEGROUP) |
742 | { |
743 | newptrb.epb_saved_eptr = eptr; |
744 | newptrb.epb_prev = eptrb; |
745 | eptrb = &newptrb; |
746 | md->match_function_type = 0; |
747 | } |
748 | |
749 | /* Now start processing the opcodes. */ |
750 | |
751 | for (;;) |
752 | { |
753 | minimize = possessive = FALSE; |
754 | op = *ecode; |
755 | |
756 | switch(op) |
757 | { |
758 | case OP_MARK: |
759 | md->nomatch_mark = ecode + 2; |
760 | md->mark = NULL; /* In case previously set by assertion */ |
761 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
762 | eptrb, RM55); |
763 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
764 | md->mark == NULL) md->mark = ecode + 2; |
765 | |
766 | /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an |
767 | argument, and we must check whether that argument matches this MARK's |
768 | argument. It is passed back in md->start_match_ptr (an overloading of that |
769 | variable). If it does match, we reset that variable to the current subject |
770 | position and return MATCH_SKIP. Otherwise, pass back the return code |
771 | unaltered. */ |
772 | |
773 | else if (rrc == MATCH_SKIP_ARG && |
774 | STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0) |
775 | { |
776 | md->start_match_ptr = eptr; |
777 | RRETURN(MATCH_SKIP); |
778 | } |
779 | RRETURN(rrc); |
780 | |
781 | case OP_FAIL: |
782 | RRETURN(MATCH_NOMATCH); |
783 | |
784 | case OP_COMMIT: |
785 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
786 | eptrb, RM52); |
787 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
788 | RRETURN(MATCH_COMMIT); |
789 | |
790 | case OP_PRUNE: |
791 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
792 | eptrb, RM51); |
793 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
794 | RRETURN(MATCH_PRUNE); |
795 | |
796 | case OP_PRUNE_ARG: |
797 | md->nomatch_mark = ecode + 2; |
798 | md->mark = NULL; /* In case previously set by assertion */ |
799 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
800 | eptrb, RM56); |
801 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
802 | md->mark == NULL) md->mark = ecode + 2; |
803 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
804 | RRETURN(MATCH_PRUNE); |
805 | |
806 | case OP_SKIP: |
807 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
808 | eptrb, RM53); |
809 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
810 | md->start_match_ptr = eptr; /* Pass back current position */ |
811 | RRETURN(MATCH_SKIP); |
812 | |
813 | /* Note that, for Perl compatibility, SKIP with an argument does NOT set |
814 | nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was |
815 | not a matching mark, we have to re-run the match, ignoring the SKIP_ARG |
816 | that failed and any that preceed it (either they also failed, or were not |
817 | triggered). To do this, we maintain a count of executed SKIP_ARGs. If a |
818 | SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg |
819 | set to the count of the one that failed. */ |
820 | |
821 | case OP_SKIP_ARG: |
822 | md->skip_arg_count++; |
823 | if (md->skip_arg_count <= md->ignore_skip_arg) |
824 | { |
825 | ecode += PRIV(OP_lengths)[*ecode] + ecode[1]; |
826 | break; |
827 | } |
828 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md, |
829 | eptrb, RM57); |
830 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
831 | |
832 | /* Pass back the current skip name by overloading md->start_match_ptr and |
833 | returning the special MATCH_SKIP_ARG return code. This will either be |
834 | caught by a matching MARK, or get to the top, where it causes a rematch |
835 | with md->ignore_skip_arg set to the value of md->skip_arg_count. */ |
836 | |
837 | md->start_match_ptr = ecode + 2; |
838 | RRETURN(MATCH_SKIP_ARG); |
839 | |
840 | /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that |
841 | the branch in which it occurs can be determined. Overload the start of |
842 | match pointer to do this. */ |
843 | |
844 | case OP_THEN: |
845 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
846 | eptrb, RM54); |
847 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
848 | md->start_match_ptr = ecode; |
849 | RRETURN(MATCH_THEN); |
850 | |
851 | case OP_THEN_ARG: |
852 | md->nomatch_mark = ecode + 2; |
853 | md->mark = NULL; /* In case previously set by assertion */ |
854 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, |
855 | md, eptrb, RM58); |
856 | if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) && |
857 | md->mark == NULL) md->mark = ecode + 2; |
858 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
859 | md->start_match_ptr = ecode; |
860 | RRETURN(MATCH_THEN); |
861 | |
862 | /* Handle an atomic group that does not contain any capturing parentheses. |
863 | This can be handled like an assertion. Prior to 8.13, all atomic groups |
864 | were handled this way. In 8.13, the code was changed as below for ONCE, so |
865 | that backups pass through the group and thereby reset captured values. |
866 | However, this uses a lot more stack, so in 8.20, atomic groups that do not |
867 | contain any captures generate OP_ONCE_NC, which can be handled in the old, |
868 | less stack intensive way. |
869 | |
870 | Check the alternative branches in turn - the matching won't pass the KET |
871 | for this kind of subpattern. If any one branch matches, we carry on as at |
872 | the end of a normal bracket, leaving the subject pointer, but resetting |
873 | the start-of-match value in case it was changed by \K. */ |
874 | |
875 | case OP_ONCE_NC: |
876 | prev = ecode; |
877 | saved_eptr = eptr; |
878 | save_mark = md->mark; |
879 | do |
880 | { |
881 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64); |
882 | if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */ |
883 | { |
884 | mstart = md->start_match_ptr; |
885 | break; |
886 | } |
887 | if (rrc == MATCH_THEN) |
888 | { |
889 | next = ecode + GET(ecode,1); |
890 | if (md->start_match_ptr < next && |
891 | (*ecode == OP_ALT || *next == OP_ALT)) |
892 | rrc = MATCH_NOMATCH; |
893 | } |
894 | |
895 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
896 | ecode += GET(ecode,1); |
897 | md->mark = save_mark; |
898 | } |
899 | while (*ecode == OP_ALT); |
900 | |
901 | /* If hit the end of the group (which could be repeated), fail */ |
902 | |
903 | if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH); |
904 | |
905 | /* Continue as from after the group, updating the offsets high water |
906 | mark, since extracts may have been taken. */ |
907 | |
908 | do ecode += GET(ecode, 1); while (*ecode == OP_ALT); |
909 | |
910 | offset_top = md->end_offset_top; |
911 | eptr = md->end_match_ptr; |
912 | |
913 | /* For a non-repeating ket, just continue at this level. This also |
914 | happens for a repeating ket if no characters were matched in the group. |
915 | This is the forcible breaking of infinite loops as implemented in Perl |
916 | 5.005. */ |
917 | |
918 | if (*ecode == OP_KET || eptr == saved_eptr) |
919 | { |
920 | ecode += 1+LINK_SIZE; |
921 | break; |
922 | } |
923 | |
924 | /* The repeating kets try the rest of the pattern or restart from the |
925 | preceding bracket, in the appropriate order. The second "call" of match() |
926 | uses tail recursion, to avoid using another stack frame. */ |
927 | |
928 | if (*ecode == OP_KETRMIN) |
929 | { |
930 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65); |
931 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
932 | ecode = prev; |
933 | goto TAIL_RECURSE; |
934 | } |
935 | else /* OP_KETRMAX */ |
936 | { |
937 | RMATCH(eptr, prev, offset_top, md, eptrb, RM66); |
938 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
939 | ecode += 1 + LINK_SIZE; |
940 | goto TAIL_RECURSE; |
941 | } |
942 | /* Control never gets here */ |
943 | |
944 | /* Handle a capturing bracket, other than those that are possessive with an |
945 | unlimited repeat. If there is space in the offset vector, save the current |
946 | subject position in the working slot at the top of the vector. We mustn't |
947 | change the current values of the data slot, because they may be set from a |
948 | previous iteration of this group, and be referred to by a reference inside |
949 | the group. A failure to match might occur after the group has succeeded, |
950 | if something later on doesn't match. For this reason, we need to restore |
951 | the working value and also the values of the final offsets, in case they |
952 | were set by a previous iteration of the same bracket. |
953 | |
954 | If there isn't enough space in the offset vector, treat this as if it were |
955 | a non-capturing bracket. Don't worry about setting the flag for the error |
956 | case here; that is handled in the code for KET. */ |
957 | |
958 | case OP_CBRA: |
959 | case OP_SCBRA: |
960 | number = GET2(ecode, 1+LINK_SIZE); |
961 | offset = number << 1; |
962 | |
963 | #ifdef PCRE_DEBUG |
964 | printf("start bracket %d\n", number); |
965 | printf("subject="); |
966 | pchars(eptr, 16, TRUE, md); |
967 | printf("\n"); |
968 | #endif |
969 | |
970 | if (offset < md->offset_max) |
971 | { |
972 | save_offset1 = md->offset_vector[offset]; |
973 | save_offset2 = md->offset_vector[offset+1]; |
974 | save_offset3 = md->offset_vector[md->offset_end - number]; |
975 | save_capture_last = md->capture_last; |
976 | save_mark = md->mark; |
977 | |
978 | DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); |
979 | md->offset_vector[md->offset_end - number] = |
980 | (int)(eptr - md->start_subject); |
981 | |
982 | for (;;) |
983 | { |
984 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
985 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
986 | eptrb, RM1); |
987 | if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */ |
988 | |
989 | /* If we backed up to a THEN, check whether it is within the current |
990 | branch by comparing the address of the THEN that is passed back with |
991 | the end of the branch. If it is within the current branch, and the |
992 | branch is one of two or more alternatives (it either starts or ends |
993 | with OP_ALT), we have reached the limit of THEN's action, so convert |
994 | the return code to NOMATCH, which will cause normal backtracking to |
995 | happen from now on. Otherwise, THEN is passed back to an outer |
996 | alternative. This implements Perl's treatment of parenthesized groups, |
997 | where a group not containing | does not affect the current alternative, |
998 | that is, (X) is NOT the same as (X|(*F)). */ |
999 | |
1000 | if (rrc == MATCH_THEN) |
1001 | { |
1002 | next = ecode + GET(ecode,1); |
1003 | if (md->start_match_ptr < next && |
1004 | (*ecode == OP_ALT || *next == OP_ALT)) |
1005 | rrc = MATCH_NOMATCH; |
1006 | } |
1007 | |
1008 | /* Anything other than NOMATCH is passed back. */ |
1009 | |
1010 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1011 | md->capture_last = save_capture_last; |
1012 | ecode += GET(ecode, 1); |
1013 | md->mark = save_mark; |
1014 | if (*ecode != OP_ALT) break; |
1015 | } |
1016 | |
1017 | DPRINTF(("bracket %d failed\n", number)); |
1018 | md->offset_vector[offset] = save_offset1; |
1019 | md->offset_vector[offset+1] = save_offset2; |
1020 | md->offset_vector[md->offset_end - number] = save_offset3; |
1021 | |
1022 | /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */ |
1023 | |
1024 | RRETURN(rrc); |
1025 | } |
1026 | |
1027 | /* FALL THROUGH ... Insufficient room for saving captured contents. Treat |
1028 | as a non-capturing bracket. */ |
1029 | |
1030 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1031 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1032 | |
1033 | DPRINTF(("insufficient capture room: treat as non-capturing\n")); |
1034 | |
1035 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1036 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1037 | |
1038 | /* Non-capturing or atomic group, except for possessive with unlimited |
1039 | repeat and ONCE group with no captures. Loop for all the alternatives. |
1040 | |
1041 | When we get to the final alternative within the brackets, we used to return |
1042 | the result of a recursive call to match() whatever happened so it was |
1043 | possible to reduce stack usage by turning this into a tail recursion, |
1044 | except in the case of a possibly empty group. However, now that there is |
1045 | the possiblity of (*THEN) occurring in the final alternative, this |
1046 | optimization is no longer always possible. |
1047 | |
1048 | We can optimize if we know there are no (*THEN)s in the pattern; at present |
1049 | this is the best that can be done. |
1050 | |
1051 | MATCH_ONCE is returned when the end of an atomic group is successfully |
1052 | reached, but subsequent matching fails. It passes back up the tree (causing |
1053 | captured values to be reset) until the original atomic group level is |
1054 | reached. This is tested by comparing md->once_target with the start of the |
1055 | group. At this point, the return is converted into MATCH_NOMATCH so that |
1056 | previous backup points can be taken. */ |
1057 | |
1058 | case OP_ONCE: |
1059 | case OP_BRA: |
1060 | case OP_SBRA: |
1061 | DPRINTF(("start non-capturing bracket\n")); |
1062 | |
1063 | for (;;) |
1064 | { |
1065 | if (op >= OP_SBRA || op == OP_ONCE) |
1066 | md->match_function_type = MATCH_CBEGROUP; |
1067 | |
1068 | /* If this is not a possibly empty group, and there are no (*THEN)s in |
1069 | the pattern, and this is the final alternative, optimize as described |
1070 | above. */ |
1071 | |
1072 | else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT) |
1073 | { |
1074 | ecode += PRIV(OP_lengths)[*ecode]; |
1075 | goto TAIL_RECURSE; |
1076 | } |
1077 | |
1078 | /* In all other cases, we have to make another call to match(). */ |
1079 | |
1080 | save_mark = md->mark; |
1081 | save_capture_last = md->capture_last; |
1082 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb, |
1083 | RM2); |
1084 | |
1085 | /* See comment in the code for capturing groups above about handling |
1086 | THEN. */ |
1087 | |
1088 | if (rrc == MATCH_THEN) |
1089 | { |
1090 | next = ecode + GET(ecode,1); |
1091 | if (md->start_match_ptr < next && |
1092 | (*ecode == OP_ALT || *next == OP_ALT)) |
1093 | rrc = MATCH_NOMATCH; |
1094 | } |
1095 | |
1096 | if (rrc != MATCH_NOMATCH) |
1097 | { |
1098 | if (rrc == MATCH_ONCE) |
1099 | { |
1100 | const pcre_uchar *scode = ecode; |
1101 | if (*scode != OP_ONCE) /* If not at start, find it */ |
1102 | { |
1103 | while (*scode == OP_ALT) scode += GET(scode, 1); |
1104 | scode -= GET(scode, 1); |
1105 | } |
1106 | if (md->once_target == scode) rrc = MATCH_NOMATCH; |
1107 | } |
1108 | RRETURN(rrc); |
1109 | } |
1110 | ecode += GET(ecode, 1); |
1111 | md->mark = save_mark; |
1112 | if (*ecode != OP_ALT) break; |
1113 | md->capture_last = save_capture_last; |
1114 | } |
1115 | |
1116 | RRETURN(MATCH_NOMATCH); |
1117 | |
1118 | /* Handle possessive capturing brackets with an unlimited repeat. We come |
1119 | here from BRAZERO with allow_zero set TRUE. The offset_vector values are |
1120 | handled similarly to the normal case above. However, the matching is |
1121 | different. The end of these brackets will always be OP_KETRPOS, which |
1122 | returns MATCH_KETRPOS without going further in the pattern. By this means |
1123 | we can handle the group by iteration rather than recursion, thereby |
1124 | reducing the amount of stack needed. */ |
1125 | |
1126 | case OP_CBRAPOS: |
1127 | case OP_SCBRAPOS: |
1128 | allow_zero = FALSE; |
1129 | |
1130 | POSSESSIVE_CAPTURE: |
1131 | number = GET2(ecode, 1+LINK_SIZE); |
1132 | offset = number << 1; |
1133 | |
1134 | #ifdef PCRE_DEBUG |
1135 | printf("start possessive bracket %d\n", number); |
1136 | printf("subject="); |
1137 | pchars(eptr, 16, TRUE, md); |
1138 | printf("\n"); |
1139 | #endif |
1140 | |
1141 | if (offset < md->offset_max) |
1142 | { |
1143 | matched_once = FALSE; |
1144 | code_offset = (int)(ecode - md->start_code); |
1145 | |
1146 | save_offset1 = md->offset_vector[offset]; |
1147 | save_offset2 = md->offset_vector[offset+1]; |
1148 | save_offset3 = md->offset_vector[md->offset_end - number]; |
1149 | save_capture_last = md->capture_last; |
1150 | |
1151 | DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); |
1152 | |
1153 | /* Each time round the loop, save the current subject position for use |
1154 | when the group matches. For MATCH_MATCH, the group has matched, so we |
1155 | restart it with a new subject starting position, remembering that we had |
1156 | at least one match. For MATCH_NOMATCH, carry on with the alternatives, as |
1157 | usual. If we haven't matched any alternatives in any iteration, check to |
1158 | see if a previous iteration matched. If so, the group has matched; |
1159 | continue from afterwards. Otherwise it has failed; restore the previous |
1160 | capture values before returning NOMATCH. */ |
1161 | |
1162 | for (;;) |
1163 | { |
1164 | md->offset_vector[md->offset_end - number] = |
1165 | (int)(eptr - md->start_subject); |
1166 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
1167 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
1168 | eptrb, RM63); |
1169 | if (rrc == MATCH_KETRPOS) |
1170 | { |
1171 | offset_top = md->end_offset_top; |
1172 | eptr = md->end_match_ptr; |
1173 | ecode = md->start_code + code_offset; |
1174 | save_capture_last = md->capture_last; |
1175 | matched_once = TRUE; |
1176 | continue; |
1177 | } |
1178 | |
1179 | /* See comment in the code for capturing groups above about handling |
1180 | THEN. */ |
1181 | |
1182 | if (rrc == MATCH_THEN) |
1183 | { |
1184 | next = ecode + GET(ecode,1); |
1185 | if (md->start_match_ptr < next && |
1186 | (*ecode == OP_ALT || *next == OP_ALT)) |
1187 | rrc = MATCH_NOMATCH; |
1188 | } |
1189 | |
1190 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1191 | md->capture_last = save_capture_last; |
1192 | ecode += GET(ecode, 1); |
1193 | if (*ecode != OP_ALT) break; |
1194 | } |
1195 | |
1196 | if (!matched_once) |
1197 | { |
1198 | md->offset_vector[offset] = save_offset1; |
1199 | md->offset_vector[offset+1] = save_offset2; |
1200 | md->offset_vector[md->offset_end - number] = save_offset3; |
1201 | } |
1202 | |
1203 | if (allow_zero || matched_once) |
1204 | { |
1205 | ecode += 1 + LINK_SIZE; |
1206 | break; |
1207 | } |
1208 | |
1209 | RRETURN(MATCH_NOMATCH); |
1210 | } |
1211 | |
1212 | /* FALL THROUGH ... Insufficient room for saving captured contents. Treat |
1213 | as a non-capturing bracket. */ |
1214 | |
1215 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1216 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1217 | |
1218 | DPRINTF(("insufficient capture room: treat as non-capturing\n")); |
1219 | |
1220 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1221 | /* VVVVVVVVVVVVVVVVVVVVVVVVV */ |
1222 | |
1223 | /* Non-capturing possessive bracket with unlimited repeat. We come here |
1224 | from BRAZERO with allow_zero = TRUE. The code is similar to the above, |
1225 | without the capturing complication. It is written out separately for speed |
1226 | and cleanliness. */ |
1227 | |
1228 | case OP_BRAPOS: |
1229 | case OP_SBRAPOS: |
1230 | allow_zero = FALSE; |
1231 | |
1232 | POSSESSIVE_NON_CAPTURE: |
1233 | matched_once = FALSE; |
1234 | code_offset = (int)(ecode - md->start_code); |
1235 | save_capture_last = md->capture_last; |
1236 | |
1237 | for (;;) |
1238 | { |
1239 | if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP; |
1240 | RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, |
1241 | eptrb, RM48); |
1242 | if (rrc == MATCH_KETRPOS) |
1243 | { |
1244 | offset_top = md->end_offset_top; |
1245 | eptr = md->end_match_ptr; |
1246 | ecode = md->start_code + code_offset; |
1247 | matched_once = TRUE; |
1248 | continue; |
1249 | } |
1250 | |
1251 | /* See comment in the code for capturing groups above about handling |
1252 | THEN. */ |
1253 | |
1254 | if (rrc == MATCH_THEN) |
1255 | { |
1256 | next = ecode + GET(ecode,1); |
1257 | if (md->start_match_ptr < next && |
1258 | (*ecode == OP_ALT || *next == OP_ALT)) |
1259 | rrc = MATCH_NOMATCH; |
1260 | } |
1261 | |
1262 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1263 | ecode += GET(ecode, 1); |
1264 | if (*ecode != OP_ALT) break; |
1265 | md->capture_last = save_capture_last; |
1266 | } |
1267 | |
1268 | if (matched_once || allow_zero) |
1269 | { |
1270 | ecode += 1 + LINK_SIZE; |
1271 | break; |
1272 | } |
1273 | RRETURN(MATCH_NOMATCH); |
1274 | |
1275 | /* Control never reaches here. */ |
1276 | |
1277 | /* Conditional group: compilation checked that there are no more than |
1278 | two branches. If the condition is false, skipping the first branch takes us |
1279 | past the end if there is only one branch, but that's OK because that is |
1280 | exactly what going to the ket would do. */ |
1281 | |
1282 | case OP_COND: |
1283 | case OP_SCOND: |
1284 | codelink = GET(ecode, 1); |
1285 | |
1286 | /* Because of the way auto-callout works during compile, a callout item is |
1287 | inserted between OP_COND and an assertion condition. */ |
1288 | |
1289 | if (ecode[LINK_SIZE+1] == OP_CALLOUT) |
1290 | { |
1291 | if (PUBL(callout) != NULL) |
1292 | { |
1293 | PUBL(callout_block) cb; |
1294 | cb.version = 2; /* Version 1 of the callout block */ |
1295 | cb.callout_number = ecode[LINK_SIZE+2]; |
1296 | cb.offset_vector = md->offset_vector; |
1297 | #if defined COMPILE_PCRE8 |
1298 | cb.subject = (PCRE_SPTR)md->start_subject; |
1299 | #elif defined COMPILE_PCRE16 |
1300 | cb.subject = (PCRE_SPTR16)md->start_subject; |
1301 | #elif defined COMPILE_PCRE32 |
1302 | cb.subject = (PCRE_SPTR32)md->start_subject; |
1303 | #endif |
1304 | cb.subject_length = (int)(md->end_subject - md->start_subject); |
1305 | cb.start_match = (int)(mstart - md->start_subject); |
1306 | cb.current_position = (int)(eptr - md->start_subject); |
1307 | cb.pattern_position = GET(ecode, LINK_SIZE + 3); |
1308 | cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE); |
1309 | cb.capture_top = offset_top/2; |
1310 | cb.capture_last = md->capture_last & CAPLMASK; |
1311 | /* Internal change requires this for API compatibility. */ |
1312 | if (cb.capture_last == 0) cb.capture_last = -1; |
1313 | cb.callout_data = md->callout_data; |
1314 | cb.mark = md->nomatch_mark; |
1315 | if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH); |
1316 | if (rrc < 0) RRETURN(rrc); |
1317 | } |
1318 | ecode += PRIV(OP_lengths)[OP_CALLOUT]; |
1319 | codelink -= PRIV(OP_lengths)[OP_CALLOUT]; |
1320 | } |
1321 | |
1322 | condcode = ecode[LINK_SIZE+1]; |
1323 | |
1324 | /* Now see what the actual condition is */ |
1325 | |
1326 | if (condcode == OP_RREF || condcode == OP_NRREF) /* Recursion test */ |
1327 | { |
1328 | if (md->recursive == NULL) /* Not recursing => FALSE */ |
1329 | { |
1330 | condition = FALSE; |
1331 | ecode += GET(ecode, 1); |
1332 | } |
1333 | else |
1334 | { |
1335 | unsigned int recno = GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/ |
1336 | condition = (recno == RREF_ANY || recno == md->recursive->group_num); |
1337 | |
1338 | /* If the test is for recursion into a specific subpattern, and it is |
1339 | false, but the test was set up by name, scan the table to see if the |
1340 | name refers to any other numbers, and test them. The condition is true |
1341 | if any one is set. */ |
1342 | |
1343 | if (!condition && condcode == OP_NRREF) |
1344 | { |
1345 | pcre_uchar *slotA = md->name_table; |
1346 | for (i = 0; i < md->name_count; i++) |
1347 | { |
1348 | if (GET2(slotA, 0) == recno) break; |
1349 | slotA += md->name_entry_size; |
1350 | } |
1351 | |
1352 | /* Found a name for the number - there can be only one; duplicate |
1353 | names for different numbers are allowed, but not vice versa. First |
1354 | scan down for duplicates. */ |
1355 | |
1356 | if (i < md->name_count) |
1357 | { |
1358 | pcre_uchar *slotB = slotA; |
1359 | while (slotB > md->name_table) |
1360 | { |
1361 | slotB -= md->name_entry_size; |
1362 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1363 | { |
1364 | condition = GET2(slotB, 0) == md->recursive->group_num; |
1365 | if (condition) break; |
1366 | } |
1367 | else break; |
1368 | } |
1369 | |
1370 | /* Scan up for duplicates */ |
1371 | |
1372 | if (!condition) |
1373 | { |
1374 | slotB = slotA; |
1375 | for (i++; i < md->name_count; i++) |
1376 | { |
1377 | slotB += md->name_entry_size; |
1378 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1379 | { |
1380 | condition = GET2(slotB, 0) == md->recursive->group_num; |
1381 | if (condition) break; |
1382 | } |
1383 | else break; |
1384 | } |
1385 | } |
1386 | } |
1387 | } |
1388 | |
1389 | /* Chose branch according to the condition */ |
1390 | |
1391 | ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1); |
1392 | } |
1393 | } |
1394 | |
1395 | else if (condcode == OP_CREF || condcode == OP_NCREF) /* Group used test */ |
1396 | { |
1397 | offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */ |
1398 | condition = offset < offset_top && md->offset_vector[offset] >= 0; |
1399 | |
1400 | /* If the numbered capture is unset, but the reference was by name, |
1401 | scan the table to see if the name refers to any other numbers, and test |
1402 | them. The condition is true if any one is set. This is tediously similar |
1403 | to the code above, but not close enough to try to amalgamate. */ |
1404 | |
1405 | if (!condition && condcode == OP_NCREF) |
1406 | { |
1407 | unsigned int refno = offset >> 1; |
1408 | pcre_uchar *slotA = md->name_table; |
1409 | |
1410 | for (i = 0; i < md->name_count; i++) |
1411 | { |
1412 | if (GET2(slotA, 0) == refno) break; |
1413 | slotA += md->name_entry_size; |
1414 | } |
1415 | |
1416 | /* Found a name for the number - there can be only one; duplicate names |
1417 | for different numbers are allowed, but not vice versa. First scan down |
1418 | for duplicates. */ |
1419 | |
1420 | if (i < md->name_count) |
1421 | { |
1422 | pcre_uchar *slotB = slotA; |
1423 | while (slotB > md->name_table) |
1424 | { |
1425 | slotB -= md->name_entry_size; |
1426 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1427 | { |
1428 | offset = GET2(slotB, 0) << 1; |
1429 | condition = offset < offset_top && |
1430 | md->offset_vector[offset] >= 0; |
1431 | if (condition) break; |
1432 | } |
1433 | else break; |
1434 | } |
1435 | |
1436 | /* Scan up for duplicates */ |
1437 | |
1438 | if (!condition) |
1439 | { |
1440 | slotB = slotA; |
1441 | for (i++; i < md->name_count; i++) |
1442 | { |
1443 | slotB += md->name_entry_size; |
1444 | if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0) |
1445 | { |
1446 | offset = GET2(slotB, 0) << 1; |
1447 | condition = offset < offset_top && |
1448 | md->offset_vector[offset] >= 0; |
1449 | if (condition) break; |
1450 | } |
1451 | else break; |
1452 | } |
1453 | } |
1454 | } |
1455 | } |
1456 | |
1457 | /* Chose branch according to the condition */ |
1458 | |
1459 | ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1); |
1460 | } |
1461 | |
1462 | else if (condcode == OP_DEF) /* DEFINE - always false */ |
1463 | { |
1464 | condition = FALSE; |
1465 | ecode += GET(ecode, 1); |
1466 | } |
1467 | |
1468 | /* The condition is an assertion. Call match() to evaluate it - setting |
1469 | md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of |
1470 | an assertion. */ |
1471 | |
1472 | else |
1473 | { |
1474 | md->match_function_type = MATCH_CONDASSERT; |
1475 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3); |
1476 | if (rrc == MATCH_MATCH) |
1477 | { |
1478 | if (md->end_offset_top > offset_top) |
1479 | offset_top = md->end_offset_top; /* Captures may have happened */ |
1480 | condition = TRUE; |
1481 | ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2); |
1482 | while (*ecode == OP_ALT) ecode += GET(ecode, 1); |
1483 | } |
1484 | |
1485 | /* PCRE doesn't allow the effect of (*THEN) to escape beyond an |
1486 | assertion; it is therefore treated as NOMATCH. */ |
1487 | |
1488 | else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) |
1489 | { |
1490 | RRETURN(rrc); /* Need braces because of following else */ |
1491 | } |
1492 | else |
1493 | { |
1494 | condition = FALSE; |
1495 | ecode += codelink; |
1496 | } |
1497 | } |
1498 | |
1499 | /* We are now at the branch that is to be obeyed. As there is only one, can |
1500 | use tail recursion to avoid using another stack frame, except when there is |
1501 | unlimited repeat of a possibly empty group. In the latter case, a recursive |
1502 | call to match() is always required, unless the second alternative doesn't |
1503 | exist, in which case we can just plough on. Note that, for compatibility |
1504 | with Perl, the | in a conditional group is NOT treated as creating two |
1505 | alternatives. If a THEN is encountered in the branch, it propagates out to |
1506 | the enclosing alternative (unless nested in a deeper set of alternatives, |
1507 | of course). */ |
1508 | |
1509 | if (condition || *ecode == OP_ALT) |
1510 | { |
1511 | if (op != OP_SCOND) |
1512 | { |
1513 | ecode += 1 + LINK_SIZE; |
1514 | goto TAIL_RECURSE; |
1515 | } |
1516 | |
1517 | md->match_function_type = MATCH_CBEGROUP; |
1518 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49); |
1519 | RRETURN(rrc); |
1520 | } |
1521 | |
1522 | /* Condition false & no alternative; continue after the group. */ |
1523 | |
1524 | else |
1525 | { |
1526 | ecode += 1 + LINK_SIZE; |
1527 | } |
1528 | break; |
1529 | |
1530 | |
1531 | /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes, |
1532 | to close any currently open capturing brackets. */ |
1533 | |
1534 | case OP_CLOSE: |
1535 | number = GET2(ecode, 1); /* Must be less than 65536 */ |
1536 | offset = number << 1; |
1537 | |
1538 | #ifdef PCRE_DEBUG |
1539 | printf("end bracket %d at *ACCEPT", number); |
1540 | printf("\n"); |
1541 | #endif |
1542 | |
1543 | md->capture_last = (md->capture_last & OVFLMASK) | number; |
1544 | if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else |
1545 | { |
1546 | md->offset_vector[offset] = |
1547 | md->offset_vector[md->offset_end - number]; |
1548 | md->offset_vector[offset+1] = (int)(eptr - md->start_subject); |
1549 | if (offset_top <= offset) offset_top = offset + 2; |
1550 | } |
1551 | ecode += 1 + IMM2_SIZE; |
1552 | break; |
1553 | |
1554 | |
1555 | /* End of the pattern, either real or forced. */ |
1556 | |
1557 | case OP_END: |
1558 | case OP_ACCEPT: |
1559 | case OP_ASSERT_ACCEPT: |
1560 | |
1561 | /* If we have matched an empty string, fail if not in an assertion and not |
1562 | in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART |
1563 | is set and we have matched at the start of the subject. In both cases, |
1564 | backtracking will then try other alternatives, if any. */ |
1565 | |
1566 | if (eptr == mstart && op != OP_ASSERT_ACCEPT && |
1567 | md->recursive == NULL && |
1568 | (md->notempty || |
1569 | (md->notempty_atstart && |
1570 | mstart == md->start_subject + md->start_offset))) |
1571 | RRETURN(MATCH_NOMATCH); |
1572 | |
1573 | /* Otherwise, we have a match. */ |
1574 | |
1575 | md->end_match_ptr = eptr; /* Record where we ended */ |
1576 | md->end_offset_top = offset_top; /* and how many extracts were taken */ |
1577 | md->start_match_ptr = mstart; /* and the start (\K can modify) */ |
1578 | |
1579 | /* For some reason, the macros don't work properly if an expression is |
1580 | given as the argument to RRETURN when the heap is in use. */ |
1581 | |
1582 | rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT; |
1583 | RRETURN(rrc); |
1584 | |
1585 | /* Assertion brackets. Check the alternative branches in turn - the |
1586 | matching won't pass the KET for an assertion. If any one branch matches, |
1587 | the assertion is true. Lookbehind assertions have an OP_REVERSE item at the |
1588 | start of each branch to move the current point backwards, so the code at |
1589 | this level is identical to the lookahead case. When the assertion is part |
1590 | of a condition, we want to return immediately afterwards. The caller of |
1591 | this incarnation of the match() function will have set MATCH_CONDASSERT in |
1592 | md->match_function type, and one of these opcodes will be the first opcode |
1593 | that is processed. We use a local variable that is preserved over calls to |
1594 | match() to remember this case. */ |
1595 | |
1596 | case OP_ASSERT: |
1597 | case OP_ASSERTBACK: |
1598 | save_mark = md->mark; |
1599 | if (md->match_function_type == MATCH_CONDASSERT) |
1600 | { |
1601 | condassert = TRUE; |
1602 | md->match_function_type = 0; |
1603 | } |
1604 | else condassert = FALSE; |
1605 | |
1606 | /* Loop for each branch */ |
1607 | |
1608 | do |
1609 | { |
1610 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4); |
1611 | |
1612 | /* A match means that the assertion is true; break out of the loop |
1613 | that matches its alternatives. */ |
1614 | |
1615 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) |
1616 | { |
1617 | mstart = md->start_match_ptr; /* In case \K reset it */ |
1618 | break; |
1619 | } |
1620 | |
1621 | /* If not matched, restore the previous mark setting. */ |
1622 | |
1623 | md->mark = save_mark; |
1624 | |
1625 | /* See comment in the code for capturing groups above about handling |
1626 | THEN. */ |
1627 | |
1628 | if (rrc == MATCH_THEN) |
1629 | { |
1630 | next = ecode + GET(ecode,1); |
1631 | if (md->start_match_ptr < next && |
1632 | (*ecode == OP_ALT || *next == OP_ALT)) |
1633 | rrc = MATCH_NOMATCH; |
1634 | } |
1635 | |
1636 | /* Anything other than NOMATCH causes the entire assertion to fail, |
1637 | passing back the return code. This includes COMMIT, SKIP, PRUNE and an |
1638 | uncaptured THEN, which means they take their normal effect. This |
1639 | consistent approach does not always have exactly the same effect as in |
1640 | Perl. */ |
1641 | |
1642 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1643 | ecode += GET(ecode, 1); |
1644 | } |
1645 | while (*ecode == OP_ALT); /* Continue for next alternative */ |
1646 | |
1647 | /* If we have tried all the alternative branches, the assertion has |
1648 | failed. If not, we broke out after a match. */ |
1649 | |
1650 | if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH); |
1651 | |
1652 | /* If checking an assertion for a condition, return MATCH_MATCH. */ |
1653 | |
1654 | if (condassert) RRETURN(MATCH_MATCH); |
1655 | |
1656 | /* Continue from after a successful assertion, updating the offsets high |
1657 | water mark, since extracts may have been taken during the assertion. */ |
1658 | |
1659 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1660 | ecode += 1 + LINK_SIZE; |
1661 | offset_top = md->end_offset_top; |
1662 | continue; |
1663 | |
1664 | /* Negative assertion: all branches must fail to match for the assertion to |
1665 | succeed. */ |
1666 | |
1667 | case OP_ASSERT_NOT: |
1668 | case OP_ASSERTBACK_NOT: |
1669 | save_mark = md->mark; |
1670 | if (md->match_function_type == MATCH_CONDASSERT) |
1671 | { |
1672 | condassert = TRUE; |
1673 | md->match_function_type = 0; |
1674 | } |
1675 | else condassert = FALSE; |
1676 | |
1677 | /* Loop for each alternative branch. */ |
1678 | |
1679 | do |
1680 | { |
1681 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5); |
1682 | md->mark = save_mark; /* Always restore the mark setting */ |
1683 | |
1684 | switch(rrc) |
1685 | { |
1686 | case MATCH_MATCH: /* A successful match means */ |
1687 | case MATCH_ACCEPT: /* the assertion has failed. */ |
1688 | RRETURN(MATCH_NOMATCH); |
1689 | |
1690 | case MATCH_NOMATCH: /* Carry on with next branch */ |
1691 | break; |
1692 | |
1693 | /* See comment in the code for capturing groups above about handling |
1694 | THEN. */ |
1695 | |
1696 | case MATCH_THEN: |
1697 | next = ecode + GET(ecode,1); |
1698 | if (md->start_match_ptr < next && |
1699 | (*ecode == OP_ALT || *next == OP_ALT)) |
1700 | { |
1701 | rrc = MATCH_NOMATCH; |
1702 | break; |
1703 | } |
1704 | /* Otherwise fall through. */ |
1705 | |
1706 | /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole |
1707 | assertion to fail to match, without considering any more alternatives. |
1708 | Failing to match means the assertion is true. This is a consistent |
1709 | approach, but does not always have the same effect as in Perl. */ |
1710 | |
1711 | case MATCH_COMMIT: |
1712 | case MATCH_SKIP: |
1713 | case MATCH_SKIP_ARG: |
1714 | case MATCH_PRUNE: |
1715 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1716 | goto NEG_ASSERT_TRUE; /* Break out of alternation loop */ |
1717 | |
1718 | /* Anything else is an error */ |
1719 | |
1720 | default: |
1721 | RRETURN(rrc); |
1722 | } |
1723 | |
1724 | /* Continue with next branch */ |
1725 | |
1726 | ecode += GET(ecode,1); |
1727 | } |
1728 | while (*ecode == OP_ALT); |
1729 | |
1730 | /* All branches in the assertion failed to match. */ |
1731 | |
1732 | NEG_ASSERT_TRUE: |
1733 | if (condassert) RRETURN(MATCH_MATCH); /* Condition assertion */ |
1734 | ecode += 1 + LINK_SIZE; /* Continue with current branch */ |
1735 | continue; |
1736 | |
1737 | /* Move the subject pointer back. This occurs only at the start of |
1738 | each branch of a lookbehind assertion. If we are too close to the start to |
1739 | move back, this match function fails. When working with UTF-8 we move |
1740 | back a number of characters, not bytes. */ |
1741 | |
1742 | case OP_REVERSE: |
1743 | #ifdef SUPPORT_UTF |
1744 | if (utf) |
1745 | { |
1746 | i = GET(ecode, 1); |
1747 | while (i-- > 0) |
1748 | { |
1749 | eptr--; |
1750 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); |
1751 | BACKCHAR(eptr); |
1752 | } |
1753 | } |
1754 | else |
1755 | #endif |
1756 | |
1757 | /* No UTF-8 support, or not in UTF-8 mode: count is byte count */ |
1758 | |
1759 | { |
1760 | eptr -= GET(ecode, 1); |
1761 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); |
1762 | } |
1763 | |
1764 | /* Save the earliest consulted character, then skip to next op code */ |
1765 | |
1766 | if (eptr < md->start_used_ptr) md->start_used_ptr = eptr; |
1767 | ecode += 1 + LINK_SIZE; |
1768 | break; |
1769 | |
1770 | /* The callout item calls an external function, if one is provided, passing |
1771 | details of the match so far. This is mainly for debugging, though the |
1772 | function is able to force a failure. */ |
1773 | |
1774 | case OP_CALLOUT: |
1775 | if (PUBL(callout) != NULL) |
1776 | { |
1777 | PUBL(callout_block) cb; |
1778 | cb.version = 2; /* Version 1 of the callout block */ |
1779 | cb.callout_number = ecode[1]; |
1780 | cb.offset_vector = md->offset_vector; |
1781 | #if defined COMPILE_PCRE8 |
1782 | cb.subject = (PCRE_SPTR)md->start_subject; |
1783 | #elif defined COMPILE_PCRE16 |
1784 | cb.subject = (PCRE_SPTR16)md->start_subject; |
1785 | #elif defined COMPILE_PCRE32 |
1786 | cb.subject = (PCRE_SPTR32)md->start_subject; |
1787 | #endif |
1788 | cb.subject_length = (int)(md->end_subject - md->start_subject); |
1789 | cb.start_match = (int)(mstart - md->start_subject); |
1790 | cb.current_position = (int)(eptr - md->start_subject); |
1791 | cb.pattern_position = GET(ecode, 2); |
1792 | cb.next_item_length = GET(ecode, 2 + LINK_SIZE); |
1793 | cb.capture_top = offset_top/2; |
1794 | cb.capture_last = md->capture_last & CAPLMASK; |
1795 | /* Internal change requires this for API compatibility. */ |
1796 | if (cb.capture_last == 0) cb.capture_last = -1; |
1797 | cb.callout_data = md->callout_data; |
1798 | cb.mark = md->nomatch_mark; |
1799 | if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH); |
1800 | if (rrc < 0) RRETURN(rrc); |
1801 | } |
1802 | ecode += 2 + 2*LINK_SIZE; |
1803 | break; |
1804 | |
1805 | /* Recursion either matches the current regex, or some subexpression. The |
1806 | offset data is the offset to the starting bracket from the start of the |
1807 | whole pattern. (This is so that it works from duplicated subpatterns.) |
1808 | |
1809 | The state of the capturing groups is preserved over recursion, and |
1810 | re-instated afterwards. We don't know how many are started and not yet |
1811 | finished (offset_top records the completed total) so we just have to save |
1812 | all the potential data. There may be up to 65535 such values, which is too |
1813 | large to put on the stack, but using malloc for small numbers seems |
1814 | expensive. As a compromise, the stack is used when there are no more than |
1815 | REC_STACK_SAVE_MAX values to store; otherwise malloc is used. |
1816 | |
1817 | There are also other values that have to be saved. We use a chained |
1818 | sequence of blocks that actually live on the stack. Thanks to Robin Houston |
1819 | for the original version of this logic. It has, however, been hacked around |
1820 | a lot, so he is not to blame for the current way it works. */ |
1821 | |
1822 | case OP_RECURSE: |
1823 | { |
1824 | recursion_info *ri; |
1825 | unsigned int recno; |
1826 | |
1827 | callpat = md->start_code + GET(ecode, 1); |
1828 | recno = (callpat == md->start_code)? 0 : |
1829 | GET2(callpat, 1 + LINK_SIZE); |
1830 | |
1831 | /* Check for repeating a recursion without advancing the subject pointer. |
1832 | This should catch convoluted mutual recursions. (Some simple cases are |
1833 | caught at compile time.) */ |
1834 | |
1835 | for (ri = md->recursive; ri != NULL; ri = ri->prevrec) |
1836 | if (recno == ri->group_num && eptr == ri->subject_position) |
1837 | RRETURN(PCRE_ERROR_RECURSELOOP); |
1838 | |
1839 | /* Add to "recursing stack" */ |
1840 | |
1841 | new_recursive.group_num = recno; |
1842 | new_recursive.saved_capture_last = md->capture_last; |
1843 | new_recursive.subject_position = eptr; |
1844 | new_recursive.prevrec = md->recursive; |
1845 | md->recursive = &new_recursive; |
1846 | |
1847 | /* Where to continue from afterwards */ |
1848 | |
1849 | ecode += 1 + LINK_SIZE; |
1850 | |
1851 | /* Now save the offset data */ |
1852 | |
1853 | new_recursive.saved_max = md->offset_end; |
1854 | if (new_recursive.saved_max <= REC_STACK_SAVE_MAX) |
1855 | new_recursive.offset_save = stacksave; |
1856 | else |
1857 | { |
1858 | new_recursive.offset_save = |
1859 | (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int)); |
1860 | if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY); |
1861 | } |
1862 | memcpy(new_recursive.offset_save, md->offset_vector, |
1863 | new_recursive.saved_max * sizeof(int)); |
1864 | |
1865 | /* OK, now we can do the recursion. After processing each alternative, |
1866 | restore the offset data and the last captured value. If there were nested |
1867 | recursions, md->recursive might be changed, so reset it before looping. |
1868 | */ |
1869 | |
1870 | DPRINTF(("Recursing into group %d\n", new_recursive.group_num)); |
1871 | cbegroup = (*callpat >= OP_SBRA); |
1872 | do |
1873 | { |
1874 | if (cbegroup) md->match_function_type = MATCH_CBEGROUP; |
1875 | RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top, |
1876 | md, eptrb, RM6); |
1877 | memcpy(md->offset_vector, new_recursive.offset_save, |
1878 | new_recursive.saved_max * sizeof(int)); |
1879 | md->capture_last = new_recursive.saved_capture_last; |
1880 | md->recursive = new_recursive.prevrec; |
1881 | if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) |
1882 | { |
1883 | DPRINTF(("Recursion matched\n")); |
1884 | if (new_recursive.offset_save != stacksave) |
1885 | (PUBL(free))(new_recursive.offset_save); |
1886 | |
1887 | /* Set where we got to in the subject, and reset the start in case |
1888 | it was changed by \K. This *is* propagated back out of a recursion, |
1889 | for Perl compatibility. */ |
1890 | |
1891 | eptr = md->end_match_ptr; |
1892 | mstart = md->start_match_ptr; |
1893 | goto RECURSION_MATCHED; /* Exit loop; end processing */ |
1894 | } |
1895 | |
1896 | /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a |
1897 | recursion; they cause a NOMATCH for the entire recursion. These codes |
1898 | are defined in a range that can be tested for. */ |
1899 | |
1900 | if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX) |
1901 | RRETURN(MATCH_NOMATCH); |
1902 | |
1903 | /* Any return code other than NOMATCH is an error. */ |
1904 | |
1905 | if (rrc != MATCH_NOMATCH) |
1906 | { |
1907 | DPRINTF(("Recursion gave error %d\n", rrc)); |
1908 | if (new_recursive.offset_save != stacksave) |
1909 | (PUBL(free))(new_recursive.offset_save); |
1910 | RRETURN(rrc); |
1911 | } |
1912 | |
1913 | md->recursive = &new_recursive; |
1914 | callpat += GET(callpat, 1); |
1915 | } |
1916 | while (*callpat == OP_ALT); |
1917 | |
1918 | DPRINTF(("Recursion didn't match\n")); |
1919 | md->recursive = new_recursive.prevrec; |
1920 | if (new_recursive.offset_save != stacksave) |
1921 | (PUBL(free))(new_recursive.offset_save); |
1922 | RRETURN(MATCH_NOMATCH); |
1923 | } |
1924 | |
1925 | RECURSION_MATCHED: |
1926 | break; |
1927 | |
1928 | /* An alternation is the end of a branch; scan along to find the end of the |
1929 | bracketed group and go to there. */ |
1930 | |
1931 | case OP_ALT: |
1932 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); |
1933 | break; |
1934 | |
1935 | /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group, |
1936 | indicating that it may occur zero times. It may repeat infinitely, or not |
1937 | at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets |
1938 | with fixed upper repeat limits are compiled as a number of copies, with the |
1939 | optional ones preceded by BRAZERO or BRAMINZERO. */ |
1940 | |
1941 | case OP_BRAZERO: |
1942 | next = ecode + 1; |
1943 | RMATCH(eptr, next, offset_top, md, eptrb, RM10); |
1944 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1945 | do next += GET(next, 1); while (*next == OP_ALT); |
1946 | ecode = next + 1 + LINK_SIZE; |
1947 | break; |
1948 | |
1949 | case OP_BRAMINZERO: |
1950 | next = ecode + 1; |
1951 | do next += GET(next, 1); while (*next == OP_ALT); |
1952 | RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11); |
1953 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
1954 | ecode++; |
1955 | break; |
1956 | |
1957 | case OP_SKIPZERO: |
1958 | next = ecode+1; |
1959 | do next += GET(next,1); while (*next == OP_ALT); |
1960 | ecode = next + 1 + LINK_SIZE; |
1961 | break; |
1962 | |
1963 | /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything |
1964 | here; just jump to the group, with allow_zero set TRUE. */ |
1965 | |
1966 | case OP_BRAPOSZERO: |
1967 | op = *(++ecode); |
1968 | allow_zero = TRUE; |
1969 | if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE; |
1970 | goto POSSESSIVE_NON_CAPTURE; |
1971 | |
1972 | /* End of a group, repeated or non-repeating. */ |
1973 | |
1974 | case OP_KET: |
1975 | case OP_KETRMIN: |
1976 | case OP_KETRMAX: |
1977 | case OP_KETRPOS: |
1978 | prev = ecode - GET(ecode, 1); |
1979 | |
1980 | /* If this was a group that remembered the subject start, in order to break |
1981 | infinite repeats of empty string matches, retrieve the subject start from |
1982 | the chain. Otherwise, set it NULL. */ |
1983 | |
1984 | if (*prev >= OP_SBRA || *prev == OP_ONCE) |
1985 | { |
1986 | saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */ |
1987 | eptrb = eptrb->epb_prev; /* Backup to previous group */ |
1988 | } |
1989 | else saved_eptr = NULL; |
1990 | |
1991 | /* If we are at the end of an assertion group or a non-capturing atomic |
1992 | group, stop matching and return MATCH_MATCH, but record the current high |
1993 | water mark for use by positive assertions. We also need to record the match |
1994 | start in case it was changed by \K. */ |
1995 | |
1996 | if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) || |
1997 | *prev == OP_ONCE_NC) |
1998 | { |
1999 | md->end_match_ptr = eptr; /* For ONCE_NC */ |
2000 | md->end_offset_top = offset_top; |
2001 | md->start_match_ptr = mstart; |
2002 | RRETURN(MATCH_MATCH); /* Sets md->mark */ |
2003 | } |
2004 | |
2005 | /* For capturing groups we have to check the group number back at the start |
2006 | and if necessary complete handling an extraction by setting the offsets and |
2007 | bumping the high water mark. Whole-pattern recursion is coded as a recurse |
2008 | into group 0, so it won't be picked up here. Instead, we catch it when the |
2009 | OP_END is reached. Other recursion is handled here. We just have to record |
2010 | the current subject position and start match pointer and give a MATCH |
2011 | return. */ |
2012 | |
2013 | if (*prev == OP_CBRA || *prev == OP_SCBRA || |
2014 | *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS) |
2015 | { |
2016 | number = GET2(prev, 1+LINK_SIZE); |
2017 | offset = number << 1; |
2018 | |
2019 | #ifdef PCRE_DEBUG |
2020 | printf("end bracket %d", number); |
2021 | printf("\n"); |
2022 | #endif |
2023 | |
2024 | /* Handle a recursively called group. */ |
2025 | |
2026 | if (md->recursive != NULL && md->recursive->group_num == number) |
2027 | { |
2028 | md->end_match_ptr = eptr; |
2029 | md->start_match_ptr = mstart; |
2030 | RRETURN(MATCH_MATCH); |
2031 | } |
2032 | |
2033 | /* Deal with capturing */ |
2034 | |
2035 | md->capture_last = (md->capture_last & OVFLMASK) | number; |
2036 | if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else |
2037 | { |
2038 | /* If offset is greater than offset_top, it means that we are |
2039 | "skipping" a capturing group, and that group's offsets must be marked |
2040 | unset. In earlier versions of PCRE, all the offsets were unset at the |
2041 | start of matching, but this doesn't work because atomic groups and |
2042 | assertions can cause a value to be set that should later be unset. |
2043 | Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as |
2044 | part of the atomic group, but this is not on the final matching path, |
2045 | so must be unset when 2 is set. (If there is no group 2, there is no |
2046 | problem, because offset_top will then be 2, indicating no capture.) */ |
2047 | |
2048 | if (offset > offset_top) |
2049 | { |
2050 | register int *iptr = md->offset_vector + offset_top; |
2051 | register int *iend = md->offset_vector + offset; |
2052 | while (iptr < iend) *iptr++ = -1; |
2053 | } |
2054 | |
2055 | /* Now make the extraction */ |
2056 | |
2057 | md->offset_vector[offset] = |
2058 | md->offset_vector[md->offset_end - number]; |
2059 | md->offset_vector[offset+1] = (int)(eptr - md->start_subject); |
2060 | if (offset_top <= offset) offset_top = offset + 2; |
2061 | } |
2062 | } |
2063 | |
2064 | /* For an ordinary non-repeating ket, just continue at this level. This |
2065 | also happens for a repeating ket if no characters were matched in the |
2066 | group. This is the forcible breaking of infinite loops as implemented in |
2067 | Perl 5.005. For a non-repeating atomic group that includes captures, |
2068 | establish a backup point by processing the rest of the pattern at a lower |
2069 | level. If this results in a NOMATCH return, pass MATCH_ONCE back to the |
2070 | original OP_ONCE level, thereby bypassing intermediate backup points, but |
2071 | resetting any captures that happened along the way. */ |
2072 | |
2073 | if (*ecode == OP_KET || eptr == saved_eptr) |
2074 | { |
2075 | if (*prev == OP_ONCE) |
2076 | { |
2077 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12); |
2078 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2079 | md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */ |
2080 | RRETURN(MATCH_ONCE); |
2081 | } |
2082 | ecode += 1 + LINK_SIZE; /* Carry on at this level */ |
2083 | break; |
2084 | } |
2085 | |
2086 | /* OP_KETRPOS is a possessive repeating ket. Remember the current position, |
2087 | and return the MATCH_KETRPOS. This makes it possible to do the repeats one |
2088 | at a time from the outer level, thus saving stack. */ |
2089 | |
2090 | if (*ecode == OP_KETRPOS) |
2091 | { |
2092 | md->end_match_ptr = eptr; |
2093 | md->end_offset_top = offset_top; |
2094 | RRETURN(MATCH_KETRPOS); |
2095 | } |
2096 | |
2097 | /* The normal repeating kets try the rest of the pattern or restart from |
2098 | the preceding bracket, in the appropriate order. In the second case, we can |
2099 | use tail recursion to avoid using another stack frame, unless we have an |
2100 | an atomic group or an unlimited repeat of a group that can match an empty |
2101 | string. */ |
2102 | |
2103 | if (*ecode == OP_KETRMIN) |
2104 | { |
2105 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7); |
2106 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2107 | if (*prev == OP_ONCE) |
2108 | { |
2109 | RMATCH(eptr, prev, offset_top, md, eptrb, RM8); |
2110 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2111 | md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */ |
2112 | RRETURN(MATCH_ONCE); |
2113 | } |
2114 | if (*prev >= OP_SBRA) /* Could match an empty string */ |
2115 | { |
2116 | RMATCH(eptr, prev, offset_top, md, eptrb, RM50); |
2117 | RRETURN(rrc); |
2118 | } |
2119 | ecode = prev; |
2120 | goto TAIL_RECURSE; |
2121 | } |
2122 | else /* OP_KETRMAX */ |
2123 | { |
2124 | RMATCH(eptr, prev, offset_top, md, eptrb, RM13); |
2125 | if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH; |
2126 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2127 | if (*prev == OP_ONCE) |
2128 | { |
2129 | RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9); |
2130 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2131 | md->once_target = prev; |
2132 | RRETURN(MATCH_ONCE); |
2133 | } |
2134 | ecode += 1 + LINK_SIZE; |
2135 | goto TAIL_RECURSE; |
2136 | } |
2137 | /* Control never gets here */ |
2138 | |
2139 | /* Not multiline mode: start of subject assertion, unless notbol. */ |
2140 | |
2141 | case OP_CIRC: |
2142 | if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH); |
2143 | |
2144 | /* Start of subject assertion */ |
2145 | |
2146 | case OP_SOD: |
2147 | if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH); |
2148 | ecode++; |
2149 | break; |
2150 | |
2151 | /* Multiline mode: start of subject unless notbol, or after any newline. */ |
2152 | |
2153 | case OP_CIRCM: |
2154 | if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH); |
2155 | if (eptr != md->start_subject && |
2156 | (eptr == md->end_subject || !WAS_NEWLINE(eptr))) |
2157 | RRETURN(MATCH_NOMATCH); |
2158 | ecode++; |
2159 | break; |
2160 | |
2161 | /* Start of match assertion */ |
2162 | |
2163 | case OP_SOM: |
2164 | if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH); |
2165 | ecode++; |
2166 | break; |
2167 | |
2168 | /* Reset the start of match point */ |
2169 | |
2170 | case OP_SET_SOM: |
2171 | mstart = eptr; |
2172 | ecode++; |
2173 | break; |
2174 | |
2175 | /* Multiline mode: assert before any newline, or before end of subject |
2176 | unless noteol is set. */ |
2177 | |
2178 | case OP_DOLLM: |
2179 | if (eptr < md->end_subject) |
2180 | { |
2181 | if (!IS_NEWLINE(eptr)) |
2182 | { |
2183 | if (md->partial != 0 && |
2184 | eptr + 1 >= md->end_subject && |
2185 | NLBLOCK->nltype == NLTYPE_FIXED && |
2186 | NLBLOCK->nllen == 2 && |
2187 | RAWUCHARTEST(eptr) == NLBLOCK->nl[0]) |
2188 | { |
2189 | md->hitend = TRUE; |
2190 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2191 | } |
2192 | RRETURN(MATCH_NOMATCH); |
2193 | } |
2194 | } |
2195 | else |
2196 | { |
2197 | if (md->noteol) RRETURN(MATCH_NOMATCH); |
2198 | SCHECK_PARTIAL(); |
2199 | } |
2200 | ecode++; |
2201 | break; |
2202 | |
2203 | /* Not multiline mode: assert before a terminating newline or before end of |
2204 | subject unless noteol is set. */ |
2205 | |
2206 | case OP_DOLL: |
2207 | if (md->noteol) RRETURN(MATCH_NOMATCH); |
2208 | if (!md->endonly) goto ASSERT_NL_OR_EOS; |
2209 | |
2210 | /* ... else fall through for endonly */ |
2211 | |
2212 | /* End of subject assertion (\z) */ |
2213 | |
2214 | case OP_EOD: |
2215 | if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH); |
2216 | SCHECK_PARTIAL(); |
2217 | ecode++; |
2218 | break; |
2219 | |
2220 | /* End of subject or ending \n assertion (\Z) */ |
2221 | |
2222 | case OP_EODN: |
2223 | ASSERT_NL_OR_EOS: |
2224 | if (eptr < md->end_subject && |
2225 | (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen)) |
2226 | { |
2227 | if (md->partial != 0 && |
2228 | eptr + 1 >= md->end_subject && |
2229 | NLBLOCK->nltype == NLTYPE_FIXED && |
2230 | NLBLOCK->nllen == 2 && |
2231 | RAWUCHARTEST(eptr) == NLBLOCK->nl[0]) |
2232 | { |
2233 | md->hitend = TRUE; |
2234 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2235 | } |
2236 | RRETURN(MATCH_NOMATCH); |
2237 | } |
2238 | |
2239 | /* Either at end of string or \n before end. */ |
2240 | |
2241 | SCHECK_PARTIAL(); |
2242 | ecode++; |
2243 | break; |
2244 | |
2245 | /* Word boundary assertions */ |
2246 | |
2247 | case OP_NOT_WORD_BOUNDARY: |
2248 | case OP_WORD_BOUNDARY: |
2249 | { |
2250 | |
2251 | /* Find out if the previous and current characters are "word" characters. |
2252 | It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to |
2253 | be "non-word" characters. Remember the earliest consulted character for |
2254 | partial matching. */ |
2255 | |
2256 | #ifdef SUPPORT_UTF |
2257 | if (utf) |
2258 | { |
2259 | /* Get status of previous character */ |
2260 | |
2261 | if (eptr == md->start_subject) prev_is_word = FALSE; else |
2262 | { |
2263 | PCRE_PUCHAR lastptr = eptr - 1; |
2264 | BACKCHAR(lastptr); |
2265 | if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr; |
2266 | GETCHAR(c, lastptr); |
2267 | #ifdef SUPPORT_UCP |
2268 | if (md->use_ucp) |
2269 | { |
2270 | if (c == '_') prev_is_word = TRUE; else |
2271 | { |
2272 | int cat = UCD_CATEGORY(c); |
2273 | prev_is_word = (cat == ucp_L || cat == ucp_N); |
2274 | } |
2275 | } |
2276 | else |
2277 | #endif |
2278 | prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; |
2279 | } |
2280 | |
2281 | /* Get status of next character */ |
2282 | |
2283 | if (eptr >= md->end_subject) |
2284 | { |
2285 | SCHECK_PARTIAL(); |
2286 | cur_is_word = FALSE; |
2287 | } |
2288 | else |
2289 | { |
2290 | GETCHAR(c, eptr); |
2291 | #ifdef SUPPORT_UCP |
2292 | if (md->use_ucp) |
2293 | { |
2294 | if (c == '_') cur_is_word = TRUE; else |
2295 | { |
2296 | int cat = UCD_CATEGORY(c); |
2297 | cur_is_word = (cat == ucp_L || cat == ucp_N); |
2298 | } |
2299 | } |
2300 | else |
2301 | #endif |
2302 | cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; |
2303 | } |
2304 | } |
2305 | else |
2306 | #endif |
2307 | |
2308 | /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for |
2309 | consistency with the behaviour of \w we do use it in this case. */ |
2310 | |
2311 | { |
2312 | /* Get status of previous character */ |
2313 | |
2314 | if (eptr == md->start_subject) prev_is_word = FALSE; else |
2315 | { |
2316 | if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1; |
2317 | #ifdef SUPPORT_UCP |
2318 | if (md->use_ucp) |
2319 | { |
2320 | c = eptr[-1]; |
2321 | if (c == '_') prev_is_word = TRUE; else |
2322 | { |
2323 | int cat = UCD_CATEGORY(c); |
2324 | prev_is_word = (cat == ucp_L || cat == ucp_N); |
2325 | } |
2326 | } |
2327 | else |
2328 | #endif |
2329 | prev_is_word = MAX_255(eptr[-1]) |
2330 | && ((md->ctypes[eptr[-1]] & ctype_word) != 0); |
2331 | } |
2332 | |
2333 | /* Get status of next character */ |
2334 | |
2335 | if (eptr >= md->end_subject) |
2336 | { |
2337 | SCHECK_PARTIAL(); |
2338 | cur_is_word = FALSE; |
2339 | } |
2340 | else |
2341 | #ifdef SUPPORT_UCP |
2342 | if (md->use_ucp) |
2343 | { |
2344 | c = *eptr; |
2345 | if (c == '_') cur_is_word = TRUE; else |
2346 | { |
2347 | int cat = UCD_CATEGORY(c); |
2348 | cur_is_word = (cat == ucp_L || cat == ucp_N); |
2349 | } |
2350 | } |
2351 | else |
2352 | #endif |
2353 | cur_is_word = MAX_255(*eptr) |
2354 | && ((md->ctypes[*eptr] & ctype_word) != 0); |
2355 | } |
2356 | |
2357 | /* Now see if the situation is what we want */ |
2358 | |
2359 | if ((*ecode++ == OP_WORD_BOUNDARY)? |
2360 | cur_is_word == prev_is_word : cur_is_word != prev_is_word) |
2361 | RRETURN(MATCH_NOMATCH); |
2362 | } |
2363 | break; |
2364 | |
2365 | /* Match any single character type except newline; have to take care with |
2366 | CRLF newlines and partial matching. */ |
2367 | |
2368 | case OP_ANY: |
2369 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
2370 | if (md->partial != 0 && |
2371 | eptr + 1 >= md->end_subject && |
2372 | NLBLOCK->nltype == NLTYPE_FIXED && |
2373 | NLBLOCK->nllen == 2 && |
2374 | RAWUCHARTEST(eptr) == NLBLOCK->nl[0]) |
2375 | { |
2376 | md->hitend = TRUE; |
2377 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2378 | } |
2379 | |
2380 | /* Fall through */ |
2381 | |
2382 | /* Match any single character whatsoever. */ |
2383 | |
2384 | case OP_ALLANY: |
2385 | if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */ |
2386 | { /* not be updated before SCHECK_PARTIAL. */ |
2387 | SCHECK_PARTIAL(); |
2388 | RRETURN(MATCH_NOMATCH); |
2389 | } |
2390 | eptr++; |
2391 | #ifdef SUPPORT_UTF |
2392 | if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
2393 | #endif |
2394 | ecode++; |
2395 | break; |
2396 | |
2397 | /* Match a single byte, even in UTF-8 mode. This opcode really does match |
2398 | any byte, even newline, independent of the setting of PCRE_DOTALL. */ |
2399 | |
2400 | case OP_ANYBYTE: |
2401 | if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */ |
2402 | { /* not be updated before SCHECK_PARTIAL. */ |
2403 | SCHECK_PARTIAL(); |
2404 | RRETURN(MATCH_NOMATCH); |
2405 | } |
2406 | eptr++; |
2407 | ecode++; |
2408 | break; |
2409 | |
2410 | case OP_NOT_DIGIT: |
2411 | if (eptr >= md->end_subject) |
2412 | { |
2413 | SCHECK_PARTIAL(); |
2414 | RRETURN(MATCH_NOMATCH); |
2415 | } |
2416 | GETCHARINCTEST(c, eptr); |
2417 | if ( |
2418 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2419 | c < 256 && |
2420 | #endif |
2421 | (md->ctypes[c] & ctype_digit) != 0 |
2422 | ) |
2423 | RRETURN(MATCH_NOMATCH); |
2424 | ecode++; |
2425 | break; |
2426 | |
2427 | case OP_DIGIT: |
2428 | if (eptr >= md->end_subject) |
2429 | { |
2430 | SCHECK_PARTIAL(); |
2431 | RRETURN(MATCH_NOMATCH); |
2432 | } |
2433 | GETCHARINCTEST(c, eptr); |
2434 | if ( |
2435 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2436 | c > 255 || |
2437 | #endif |
2438 | (md->ctypes[c] & ctype_digit) == 0 |
2439 | ) |
2440 | RRETURN(MATCH_NOMATCH); |
2441 | ecode++; |
2442 | break; |
2443 | |
2444 | case OP_NOT_WHITESPACE: |
2445 | if (eptr >= md->end_subject) |
2446 | { |
2447 | SCHECK_PARTIAL(); |
2448 | RRETURN(MATCH_NOMATCH); |
2449 | } |
2450 | GETCHARINCTEST(c, eptr); |
2451 | if ( |
2452 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2453 | c < 256 && |
2454 | #endif |
2455 | (md->ctypes[c] & ctype_space) != 0 |
2456 | ) |
2457 | RRETURN(MATCH_NOMATCH); |
2458 | ecode++; |
2459 | break; |
2460 | |
2461 | case OP_WHITESPACE: |
2462 | if (eptr >= md->end_subject) |
2463 | { |
2464 | SCHECK_PARTIAL(); |
2465 | RRETURN(MATCH_NOMATCH); |
2466 | } |
2467 | GETCHARINCTEST(c, eptr); |
2468 | if ( |
2469 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2470 | c > 255 || |
2471 | #endif |
2472 | (md->ctypes[c] & ctype_space) == 0 |
2473 | ) |
2474 | RRETURN(MATCH_NOMATCH); |
2475 | ecode++; |
2476 | break; |
2477 | |
2478 | case OP_NOT_WORDCHAR: |
2479 | if (eptr >= md->end_subject) |
2480 | { |
2481 | SCHECK_PARTIAL(); |
2482 | RRETURN(MATCH_NOMATCH); |
2483 | } |
2484 | GETCHARINCTEST(c, eptr); |
2485 | if ( |
2486 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2487 | c < 256 && |
2488 | #endif |
2489 | (md->ctypes[c] & ctype_word) != 0 |
2490 | ) |
2491 | RRETURN(MATCH_NOMATCH); |
2492 | ecode++; |
2493 | break; |
2494 | |
2495 | case OP_WORDCHAR: |
2496 | if (eptr >= md->end_subject) |
2497 | { |
2498 | SCHECK_PARTIAL(); |
2499 | RRETURN(MATCH_NOMATCH); |
2500 | } |
2501 | GETCHARINCTEST(c, eptr); |
2502 | if ( |
2503 | #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8) |
2504 | c > 255 || |
2505 | #endif |
2506 | (md->ctypes[c] & ctype_word) == 0 |
2507 | ) |
2508 | RRETURN(MATCH_NOMATCH); |
2509 | ecode++; |
2510 | break; |
2511 | |
2512 | case OP_ANYNL: |
2513 | if (eptr >= md->end_subject) |
2514 | { |
2515 | SCHECK_PARTIAL(); |
2516 | RRETURN(MATCH_NOMATCH); |
2517 | } |
2518 | GETCHARINCTEST(c, eptr); |
2519 | switch(c) |
2520 | { |
2521 | default: RRETURN(MATCH_NOMATCH); |
2522 | |
2523 | case CHAR_CR: |
2524 | if (eptr >= md->end_subject) |
2525 | { |
2526 | SCHECK_PARTIAL(); |
2527 | } |
2528 | else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++; |
2529 | break; |
2530 | |
2531 | case CHAR_LF: |
2532 | break; |
2533 | |
2534 | case CHAR_VT: |
2535 | case CHAR_FF: |
2536 | case CHAR_NEL: |
2537 | #ifndef EBCDIC |
2538 | case 0x2028: |
2539 | case 0x2029: |
2540 | #endif /* Not EBCDIC */ |
2541 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
2542 | break; |
2543 | } |
2544 | ecode++; |
2545 | break; |
2546 | |
2547 | case OP_NOT_HSPACE: |
2548 | if (eptr >= md->end_subject) |
2549 | { |
2550 | SCHECK_PARTIAL(); |
2551 | RRETURN(MATCH_NOMATCH); |
2552 | } |
2553 | GETCHARINCTEST(c, eptr); |
2554 | switch(c) |
2555 | { |
2556 | HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */ |
2557 | default: break; |
2558 | } |
2559 | ecode++; |
2560 | break; |
2561 | |
2562 | case OP_HSPACE: |
2563 | if (eptr >= md->end_subject) |
2564 | { |
2565 | SCHECK_PARTIAL(); |
2566 | RRETURN(MATCH_NOMATCH); |
2567 | } |
2568 | GETCHARINCTEST(c, eptr); |
2569 | switch(c) |
2570 | { |
2571 | HSPACE_CASES: break; /* Byte and multibyte cases */ |
2572 | default: RRETURN(MATCH_NOMATCH); |
2573 | } |
2574 | ecode++; |
2575 | break; |
2576 | |
2577 | case OP_NOT_VSPACE: |
2578 | if (eptr >= md->end_subject) |
2579 | { |
2580 | SCHECK_PARTIAL(); |
2581 | RRETURN(MATCH_NOMATCH); |
2582 | } |
2583 | GETCHARINCTEST(c, eptr); |
2584 | switch(c) |
2585 | { |
2586 | VSPACE_CASES: RRETURN(MATCH_NOMATCH); |
2587 | default: break; |
2588 | } |
2589 | ecode++; |
2590 | break; |
2591 | |
2592 | case OP_VSPACE: |
2593 | if (eptr >= md->end_subject) |
2594 | { |
2595 | SCHECK_PARTIAL(); |
2596 | RRETURN(MATCH_NOMATCH); |
2597 | } |
2598 | GETCHARINCTEST(c, eptr); |
2599 | switch(c) |
2600 | { |
2601 | VSPACE_CASES: break; |
2602 | default: RRETURN(MATCH_NOMATCH); |
2603 | } |
2604 | ecode++; |
2605 | break; |
2606 | |
2607 | #ifdef SUPPORT_UCP |
2608 | /* Check the next character by Unicode property. We will get here only |
2609 | if the support is in the binary; otherwise a compile-time error occurs. */ |
2610 | |
2611 | case OP_PROP: |
2612 | case OP_NOTPROP: |
2613 | if (eptr >= md->end_subject) |
2614 | { |
2615 | SCHECK_PARTIAL(); |
2616 | RRETURN(MATCH_NOMATCH); |
2617 | } |
2618 | GETCHARINCTEST(c, eptr); |
2619 | { |
2620 | const pcre_uint32 *cp; |
2621 | const ucd_record *prop = GET_UCD(c); |
2622 | |
2623 | switch(ecode[1]) |
2624 | { |
2625 | case PT_ANY: |
2626 | if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH); |
2627 | break; |
2628 | |
2629 | case PT_LAMP: |
2630 | if ((prop->chartype == ucp_Lu || |
2631 | prop->chartype == ucp_Ll || |
2632 | prop->chartype == ucp_Lt) == (op == OP_NOTPROP)) |
2633 | RRETURN(MATCH_NOMATCH); |
2634 | break; |
2635 | |
2636 | case PT_GC: |
2637 | if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP)) |
2638 | RRETURN(MATCH_NOMATCH); |
2639 | break; |
2640 | |
2641 | case PT_PC: |
2642 | if ((ecode[2] != prop->chartype) == (op == OP_PROP)) |
2643 | RRETURN(MATCH_NOMATCH); |
2644 | break; |
2645 | |
2646 | case PT_SC: |
2647 | if ((ecode[2] != prop->script) == (op == OP_PROP)) |
2648 | RRETURN(MATCH_NOMATCH); |
2649 | break; |
2650 | |
2651 | /* These are specials */ |
2652 | |
2653 | case PT_ALNUM: |
2654 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2655 | PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP)) |
2656 | RRETURN(MATCH_NOMATCH); |
2657 | break; |
2658 | |
2659 | case PT_SPACE: /* Perl space */ |
2660 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z || |
2661 | c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR) |
2662 | == (op == OP_NOTPROP)) |
2663 | RRETURN(MATCH_NOMATCH); |
2664 | break; |
2665 | |
2666 | case PT_PXSPACE: /* POSIX space */ |
2667 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z || |
2668 | c == CHAR_HT || c == CHAR_NL || c == CHAR_VT || |
2669 | c == CHAR_FF || c == CHAR_CR) |
2670 | == (op == OP_NOTPROP)) |
2671 | RRETURN(MATCH_NOMATCH); |
2672 | break; |
2673 | |
2674 | case PT_WORD: |
2675 | if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2676 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
2677 | c == CHAR_UNDERSCORE) == (op == OP_NOTPROP)) |
2678 | RRETURN(MATCH_NOMATCH); |
2679 | break; |
2680 | |
2681 | case PT_CLIST: |
2682 | cp = PRIV(ucd_caseless_sets) + ecode[2]; |
2683 | for (;;) |
2684 | { |
2685 | if (c < *cp) |
2686 | { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; } |
2687 | if (c == *cp++) |
2688 | { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } } |
2689 | } |
2690 | break; |
2691 | |
2692 | case PT_UCNC: |
2693 | if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
2694 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
2695 | c >= 0xe000) == (op == OP_NOTPROP)) |
2696 | RRETURN(MATCH_NOMATCH); |
2697 | break; |
2698 | |
2699 | /* This should never occur */ |
2700 | |
2701 | default: |
2702 | RRETURN(PCRE_ERROR_INTERNAL); |
2703 | } |
2704 | |
2705 | ecode += 3; |
2706 | } |
2707 | break; |
2708 | |
2709 | /* Match an extended Unicode sequence. We will get here only if the support |
2710 | is in the binary; otherwise a compile-time error occurs. */ |
2711 | |
2712 | case OP_EXTUNI: |
2713 | if (eptr >= md->end_subject) |
2714 | { |
2715 | SCHECK_PARTIAL(); |
2716 | RRETURN(MATCH_NOMATCH); |
2717 | } |
2718 | else |
2719 | { |
2720 | int lgb, rgb; |
2721 | GETCHARINCTEST(c, eptr); |
2722 | lgb = UCD_GRAPHBREAK(c); |
2723 | while (eptr < md->end_subject) |
2724 | { |
2725 | int len = 1; |
2726 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
2727 | rgb = UCD_GRAPHBREAK(c); |
2728 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
2729 | lgb = rgb; |
2730 | eptr += len; |
2731 | } |
2732 | } |
2733 | CHECK_PARTIAL(); |
2734 | ecode++; |
2735 | break; |
2736 | #endif /* SUPPORT_UCP */ |
2737 | |
2738 | |
2739 | /* Match a back reference, possibly repeatedly. Look past the end of the |
2740 | item to see if there is repeat information following. The code is similar |
2741 | to that for character classes, but repeated for efficiency. Then obey |
2742 | similar code to character type repeats - written out again for speed. |
2743 | However, if the referenced string is the empty string, always treat |
2744 | it as matched, any number of times (otherwise there could be infinite |
2745 | loops). */ |
2746 | |
2747 | case OP_REF: |
2748 | case OP_REFI: |
2749 | caseless = op == OP_REFI; |
2750 | offset = GET2(ecode, 1) << 1; /* Doubled ref number */ |
2751 | ecode += 1 + IMM2_SIZE; |
2752 | |
2753 | /* If the reference is unset, there are two possibilities: |
2754 | |
2755 | (a) In the default, Perl-compatible state, set the length negative; |
2756 | this ensures that every attempt at a match fails. We can't just fail |
2757 | here, because of the possibility of quantifiers with zero minima. |
2758 | |
2759 | (b) If the JavaScript compatibility flag is set, set the length to zero |
2760 | so that the back reference matches an empty string. |
2761 | |
2762 | Otherwise, set the length to the length of what was matched by the |
2763 | referenced subpattern. */ |
2764 | |
2765 | if (offset >= offset_top || md->offset_vector[offset] < 0) |
2766 | length = (md->jscript_compat)? 0 : -1; |
2767 | else |
2768 | length = md->offset_vector[offset+1] - md->offset_vector[offset]; |
2769 | |
2770 | /* Set up for repetition, or handle the non-repeated case */ |
2771 | |
2772 | switch (*ecode) |
2773 | { |
2774 | case OP_CRSTAR: |
2775 | case OP_CRMINSTAR: |
2776 | case OP_CRPLUS: |
2777 | case OP_CRMINPLUS: |
2778 | case OP_CRQUERY: |
2779 | case OP_CRMINQUERY: |
2780 | c = *ecode++ - OP_CRSTAR; |
2781 | minimize = (c & 1) != 0; |
2782 | min = rep_min[c]; /* Pick up values from tables; */ |
2783 | max = rep_max[c]; /* zero for max => infinity */ |
2784 | if (max == 0) max = INT_MAX; |
2785 | break; |
2786 | |
2787 | case OP_CRRANGE: |
2788 | case OP_CRMINRANGE: |
2789 | minimize = (*ecode == OP_CRMINRANGE); |
2790 | min = GET2(ecode, 1); |
2791 | max = GET2(ecode, 1 + IMM2_SIZE); |
2792 | if (max == 0) max = INT_MAX; |
2793 | ecode += 1 + 2 * IMM2_SIZE; |
2794 | break; |
2795 | |
2796 | default: /* No repeat follows */ |
2797 | if ((length = match_ref(offset, eptr, length, md, caseless)) < 0) |
2798 | { |
2799 | if (length == -2) eptr = md->end_subject; /* Partial match */ |
2800 | CHECK_PARTIAL(); |
2801 | RRETURN(MATCH_NOMATCH); |
2802 | } |
2803 | eptr += length; |
2804 | continue; /* With the main loop */ |
2805 | } |
2806 | |
2807 | /* Handle repeated back references. If the length of the reference is |
2808 | zero, just continue with the main loop. If the length is negative, it |
2809 | means the reference is unset in non-Java-compatible mode. If the minimum is |
2810 | zero, we can continue at the same level without recursion. For any other |
2811 | minimum, carrying on will result in NOMATCH. */ |
2812 | |
2813 | if (length == 0) continue; |
2814 | if (length < 0 && min == 0) continue; |
2815 | |
2816 | /* First, ensure the minimum number of matches are present. We get back |
2817 | the length of the reference string explicitly rather than passing the |
2818 | address of eptr, so that eptr can be a register variable. */ |
2819 | |
2820 | for (i = 1; i <= min; i++) |
2821 | { |
2822 | int slength; |
2823 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2824 | { |
2825 | if (slength == -2) eptr = md->end_subject; /* Partial match */ |
2826 | CHECK_PARTIAL(); |
2827 | RRETURN(MATCH_NOMATCH); |
2828 | } |
2829 | eptr += slength; |
2830 | } |
2831 | |
2832 | /* If min = max, continue at the same level without recursion. |
2833 | They are not both allowed to be zero. */ |
2834 | |
2835 | if (min == max) continue; |
2836 | |
2837 | /* If minimizing, keep trying and advancing the pointer */ |
2838 | |
2839 | if (minimize) |
2840 | { |
2841 | for (fi = min;; fi++) |
2842 | { |
2843 | int slength; |
2844 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM14); |
2845 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2846 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
2847 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2848 | { |
2849 | if (slength == -2) eptr = md->end_subject; /* Partial match */ |
2850 | CHECK_PARTIAL(); |
2851 | RRETURN(MATCH_NOMATCH); |
2852 | } |
2853 | eptr += slength; |
2854 | } |
2855 | /* Control never gets here */ |
2856 | } |
2857 | |
2858 | /* If maximizing, find the longest string and work backwards */ |
2859 | |
2860 | else |
2861 | { |
2862 | pp = eptr; |
2863 | for (i = min; i < max; i++) |
2864 | { |
2865 | int slength; |
2866 | if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0) |
2867 | { |
2868 | /* Can't use CHECK_PARTIAL because we don't want to update eptr in |
2869 | the soft partial matching case. */ |
2870 | |
2871 | if (slength == -2 && md->partial != 0 && |
2872 | md->end_subject > md->start_used_ptr) |
2873 | { |
2874 | md->hitend = TRUE; |
2875 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
2876 | } |
2877 | break; |
2878 | } |
2879 | eptr += slength; |
2880 | } |
2881 | |
2882 | while (eptr >= pp) |
2883 | { |
2884 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM15); |
2885 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
2886 | eptr -= length; |
2887 | } |
2888 | RRETURN(MATCH_NOMATCH); |
2889 | } |
2890 | /* Control never gets here */ |
2891 | |
2892 | /* Match a bit-mapped character class, possibly repeatedly. This op code is |
2893 | used when all the characters in the class have values in the range 0-255, |
2894 | and either the matching is caseful, or the characters are in the range |
2895 | 0-127 when UTF-8 processing is enabled. The only difference between |
2896 | OP_CLASS and OP_NCLASS occurs when a data character outside the range is |
2897 | encountered. |
2898 | |
2899 | First, look past the end of the item to see if there is repeat information |
2900 | following. Then obey similar code to character type repeats - written out |
2901 | again for speed. */ |
2902 | |
2903 | case OP_NCLASS: |
2904 | case OP_CLASS: |
2905 | { |
2906 | /* The data variable is saved across frames, so the byte map needs to |
2907 | be stored there. */ |
2908 | #define BYTE_MAP ((pcre_uint8 *)data) |
2909 | data = ecode + 1; /* Save for matching */ |
2910 | ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */ |
2911 | |
2912 | switch (*ecode) |
2913 | { |
2914 | case OP_CRSTAR: |
2915 | case OP_CRMINSTAR: |
2916 | case OP_CRPLUS: |
2917 | case OP_CRMINPLUS: |
2918 | case OP_CRQUERY: |
2919 | case OP_CRMINQUERY: |
2920 | c = *ecode++ - OP_CRSTAR; |
2921 | minimize = (c & 1) != 0; |
2922 | min = rep_min[c]; /* Pick up values from tables; */ |
2923 | max = rep_max[c]; /* zero for max => infinity */ |
2924 | if (max == 0) max = INT_MAX; |
2925 | break; |
2926 | |
2927 | case OP_CRRANGE: |
2928 | case OP_CRMINRANGE: |
2929 | minimize = (*ecode == OP_CRMINRANGE); |
2930 | min = GET2(ecode, 1); |
2931 | max = GET2(ecode, 1 + IMM2_SIZE); |
2932 | if (max == 0) max = INT_MAX; |
2933 | ecode += 1 + 2 * IMM2_SIZE; |
2934 | break; |
2935 | |
2936 | default: /* No repeat follows */ |
2937 | min = max = 1; |
2938 | break; |
2939 | } |
2940 | |
2941 | /* First, ensure the minimum number of matches are present. */ |
2942 | |
2943 | #ifdef SUPPORT_UTF |
2944 | if (utf) |
2945 | { |
2946 | for (i = 1; i <= min; i++) |
2947 | { |
2948 | if (eptr >= md->end_subject) |
2949 | { |
2950 | SCHECK_PARTIAL(); |
2951 | RRETURN(MATCH_NOMATCH); |
2952 | } |
2953 | GETCHARINC(c, eptr); |
2954 | if (c > 255) |
2955 | { |
2956 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2957 | } |
2958 | else |
2959 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2960 | } |
2961 | } |
2962 | else |
2963 | #endif |
2964 | /* Not UTF mode */ |
2965 | { |
2966 | for (i = 1; i <= min; i++) |
2967 | { |
2968 | if (eptr >= md->end_subject) |
2969 | { |
2970 | SCHECK_PARTIAL(); |
2971 | RRETURN(MATCH_NOMATCH); |
2972 | } |
2973 | c = *eptr++; |
2974 | #ifndef COMPILE_PCRE8 |
2975 | if (c > 255) |
2976 | { |
2977 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
2978 | } |
2979 | else |
2980 | #endif |
2981 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
2982 | } |
2983 | } |
2984 | |
2985 | /* If max == min we can continue with the main loop without the |
2986 | need to recurse. */ |
2987 | |
2988 | if (min == max) continue; |
2989 | |
2990 | /* If minimizing, keep testing the rest of the expression and advancing |
2991 | the pointer while it matches the class. */ |
2992 | |
2993 | if (minimize) |
2994 | { |
2995 | #ifdef SUPPORT_UTF |
2996 | if (utf) |
2997 | { |
2998 | for (fi = min;; fi++) |
2999 | { |
3000 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM16); |
3001 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3002 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3003 | if (eptr >= md->end_subject) |
3004 | { |
3005 | SCHECK_PARTIAL(); |
3006 | RRETURN(MATCH_NOMATCH); |
3007 | } |
3008 | GETCHARINC(c, eptr); |
3009 | if (c > 255) |
3010 | { |
3011 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
3012 | } |
3013 | else |
3014 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
3015 | } |
3016 | } |
3017 | else |
3018 | #endif |
3019 | /* Not UTF mode */ |
3020 | { |
3021 | for (fi = min;; fi++) |
3022 | { |
3023 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM17); |
3024 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3025 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3026 | if (eptr >= md->end_subject) |
3027 | { |
3028 | SCHECK_PARTIAL(); |
3029 | RRETURN(MATCH_NOMATCH); |
3030 | } |
3031 | c = *eptr++; |
3032 | #ifndef COMPILE_PCRE8 |
3033 | if (c > 255) |
3034 | { |
3035 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); |
3036 | } |
3037 | else |
3038 | #endif |
3039 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); |
3040 | } |
3041 | } |
3042 | /* Control never gets here */ |
3043 | } |
3044 | |
3045 | /* If maximizing, find the longest possible run, then work backwards. */ |
3046 | |
3047 | else |
3048 | { |
3049 | pp = eptr; |
3050 | |
3051 | #ifdef SUPPORT_UTF |
3052 | if (utf) |
3053 | { |
3054 | for (i = min; i < max; i++) |
3055 | { |
3056 | int len = 1; |
3057 | if (eptr >= md->end_subject) |
3058 | { |
3059 | SCHECK_PARTIAL(); |
3060 | break; |
3061 | } |
3062 | GETCHARLEN(c, eptr, len); |
3063 | if (c > 255) |
3064 | { |
3065 | if (op == OP_CLASS) break; |
3066 | } |
3067 | else |
3068 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break; |
3069 | eptr += len; |
3070 | } |
3071 | for (;;) |
3072 | { |
3073 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM18); |
3074 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3075 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3076 | BACKCHAR(eptr); |
3077 | } |
3078 | } |
3079 | else |
3080 | #endif |
3081 | /* Not UTF mode */ |
3082 | { |
3083 | for (i = min; i < max; i++) |
3084 | { |
3085 | if (eptr >= md->end_subject) |
3086 | { |
3087 | SCHECK_PARTIAL(); |
3088 | break; |
3089 | } |
3090 | c = *eptr; |
3091 | #ifndef COMPILE_PCRE8 |
3092 | if (c > 255) |
3093 | { |
3094 | if (op == OP_CLASS) break; |
3095 | } |
3096 | else |
3097 | #endif |
3098 | if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break; |
3099 | eptr++; |
3100 | } |
3101 | while (eptr >= pp) |
3102 | { |
3103 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM19); |
3104 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3105 | eptr--; |
3106 | } |
3107 | } |
3108 | |
3109 | RRETURN(MATCH_NOMATCH); |
3110 | } |
3111 | #undef BYTE_MAP |
3112 | } |
3113 | /* Control never gets here */ |
3114 | |
3115 | |
3116 | /* Match an extended character class. This opcode is encountered only |
3117 | when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8 |
3118 | mode, because Unicode properties are supported in non-UTF-8 mode. */ |
3119 | |
3120 | #if defined SUPPORT_UTF || !defined COMPILE_PCRE8 |
3121 | case OP_XCLASS: |
3122 | { |
3123 | data = ecode + 1 + LINK_SIZE; /* Save for matching */ |
3124 | ecode += GET(ecode, 1); /* Advance past the item */ |
3125 | |
3126 | switch (*ecode) |
3127 | { |
3128 | case OP_CRSTAR: |
3129 | case OP_CRMINSTAR: |
3130 | case OP_CRPLUS: |
3131 | case OP_CRMINPLUS: |
3132 | case OP_CRQUERY: |
3133 | case OP_CRMINQUERY: |
3134 | c = *ecode++ - OP_CRSTAR; |
3135 | minimize = (c & 1) != 0; |
3136 | min = rep_min[c]; /* Pick up values from tables; */ |
3137 | max = rep_max[c]; /* zero for max => infinity */ |
3138 | if (max == 0) max = INT_MAX; |
3139 | break; |
3140 | |
3141 | case OP_CRRANGE: |
3142 | case OP_CRMINRANGE: |
3143 | minimize = (*ecode == OP_CRMINRANGE); |
3144 | min = GET2(ecode, 1); |
3145 | max = GET2(ecode, 1 + IMM2_SIZE); |
3146 | if (max == 0) max = INT_MAX; |
3147 | ecode += 1 + 2 * IMM2_SIZE; |
3148 | break; |
3149 | |
3150 | default: /* No repeat follows */ |
3151 | min = max = 1; |
3152 | break; |
3153 | } |
3154 | |
3155 | /* First, ensure the minimum number of matches are present. */ |
3156 | |
3157 | for (i = 1; i <= min; i++) |
3158 | { |
3159 | if (eptr >= md->end_subject) |
3160 | { |
3161 | SCHECK_PARTIAL(); |
3162 | RRETURN(MATCH_NOMATCH); |
3163 | } |
3164 | GETCHARINCTEST(c, eptr); |
3165 | if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH); |
3166 | } |
3167 | |
3168 | /* If max == min we can continue with the main loop without the |
3169 | need to recurse. */ |
3170 | |
3171 | if (min == max) continue; |
3172 | |
3173 | /* If minimizing, keep testing the rest of the expression and advancing |
3174 | the pointer while it matches the class. */ |
3175 | |
3176 | if (minimize) |
3177 | { |
3178 | for (fi = min;; fi++) |
3179 | { |
3180 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM20); |
3181 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3182 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3183 | if (eptr >= md->end_subject) |
3184 | { |
3185 | SCHECK_PARTIAL(); |
3186 | RRETURN(MATCH_NOMATCH); |
3187 | } |
3188 | GETCHARINCTEST(c, eptr); |
3189 | if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH); |
3190 | } |
3191 | /* Control never gets here */ |
3192 | } |
3193 | |
3194 | /* If maximizing, find the longest possible run, then work backwards. */ |
3195 | |
3196 | else |
3197 | { |
3198 | pp = eptr; |
3199 | for (i = min; i < max; i++) |
3200 | { |
3201 | int len = 1; |
3202 | if (eptr >= md->end_subject) |
3203 | { |
3204 | SCHECK_PARTIAL(); |
3205 | break; |
3206 | } |
3207 | #ifdef SUPPORT_UTF |
3208 | GETCHARLENTEST(c, eptr, len); |
3209 | #else |
3210 | c = *eptr; |
3211 | #endif |
3212 | if (!PRIV(xclass)(c, data, utf)) break; |
3213 | eptr += len; |
3214 | } |
3215 | for(;;) |
3216 | { |
3217 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM21); |
3218 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3219 | if (eptr-- == pp) break; /* Stop if tried at original pos */ |
3220 | #ifdef SUPPORT_UTF |
3221 | if (utf) BACKCHAR(eptr); |
3222 | #endif |
3223 | } |
3224 | RRETURN(MATCH_NOMATCH); |
3225 | } |
3226 | |
3227 | /* Control never gets here */ |
3228 | } |
3229 | #endif /* End of XCLASS */ |
3230 | |
3231 | /* Match a single character, casefully */ |
3232 | |
3233 | case OP_CHAR: |
3234 | #ifdef SUPPORT_UTF |
3235 | if (utf) |
3236 | { |
3237 | length = 1; |
3238 | ecode++; |
3239 | GETCHARLEN(fc, ecode, length); |
3240 | if (length > md->end_subject - eptr) |
3241 | { |
3242 | CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */ |
3243 | RRETURN(MATCH_NOMATCH); |
3244 | } |
3245 | while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH); |
3246 | } |
3247 | else |
3248 | #endif |
3249 | /* Not UTF mode */ |
3250 | { |
3251 | if (md->end_subject - eptr < 1) |
3252 | { |
3253 | SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */ |
3254 | RRETURN(MATCH_NOMATCH); |
3255 | } |
3256 | if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH); |
3257 | ecode += 2; |
3258 | } |
3259 | break; |
3260 | |
3261 | /* Match a single character, caselessly. If we are at the end of the |
3262 | subject, give up immediately. */ |
3263 | |
3264 | case OP_CHARI: |
3265 | if (eptr >= md->end_subject) |
3266 | { |
3267 | SCHECK_PARTIAL(); |
3268 | RRETURN(MATCH_NOMATCH); |
3269 | } |
3270 | |
3271 | #ifdef SUPPORT_UTF |
3272 | if (utf) |
3273 | { |
3274 | length = 1; |
3275 | ecode++; |
3276 | GETCHARLEN(fc, ecode, length); |
3277 | |
3278 | /* If the pattern character's value is < 128, we have only one byte, and |
3279 | we know that its other case must also be one byte long, so we can use the |
3280 | fast lookup table. We know that there is at least one byte left in the |
3281 | subject. */ |
3282 | |
3283 | if (fc < 128) |
3284 | { |
3285 | pcre_uint32 cc = RAWUCHAR(eptr); |
3286 | if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH); |
3287 | ecode++; |
3288 | eptr++; |
3289 | } |
3290 | |
3291 | /* Otherwise we must pick up the subject character. Note that we cannot |
3292 | use the value of "length" to check for sufficient bytes left, because the |
3293 | other case of the character may have more or fewer bytes. */ |
3294 | |
3295 | else |
3296 | { |
3297 | pcre_uint32 dc; |
3298 | GETCHARINC(dc, eptr); |
3299 | ecode += length; |
3300 | |
3301 | /* If we have Unicode property support, we can use it to test the other |
3302 | case of the character, if there is one. */ |
3303 | |
3304 | if (fc != dc) |
3305 | { |
3306 | #ifdef SUPPORT_UCP |
3307 | if (dc != UCD_OTHERCASE(fc)) |
3308 | #endif |
3309 | RRETURN(MATCH_NOMATCH); |
3310 | } |
3311 | } |
3312 | } |
3313 | else |
3314 | #endif /* SUPPORT_UTF */ |
3315 | |
3316 | /* Not UTF mode */ |
3317 | { |
3318 | if (TABLE_GET(ecode[1], md->lcc, ecode[1]) |
3319 | != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH); |
3320 | eptr++; |
3321 | ecode += 2; |
3322 | } |
3323 | break; |
3324 | |
3325 | /* Match a single character repeatedly. */ |
3326 | |
3327 | case OP_EXACT: |
3328 | case OP_EXACTI: |
3329 | min = max = GET2(ecode, 1); |
3330 | ecode += 1 + IMM2_SIZE; |
3331 | goto REPEATCHAR; |
3332 | |
3333 | case OP_POSUPTO: |
3334 | case OP_POSUPTOI: |
3335 | possessive = TRUE; |
3336 | /* Fall through */ |
3337 | |
3338 | case OP_UPTO: |
3339 | case OP_UPTOI: |
3340 | case OP_MINUPTO: |
3341 | case OP_MINUPTOI: |
3342 | min = 0; |
3343 | max = GET2(ecode, 1); |
3344 | minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI; |
3345 | ecode += 1 + IMM2_SIZE; |
3346 | goto REPEATCHAR; |
3347 | |
3348 | case OP_POSSTAR: |
3349 | case OP_POSSTARI: |
3350 | possessive = TRUE; |
3351 | min = 0; |
3352 | max = INT_MAX; |
3353 | ecode++; |
3354 | goto REPEATCHAR; |
3355 | |
3356 | case OP_POSPLUS: |
3357 | case OP_POSPLUSI: |
3358 | possessive = TRUE; |
3359 | min = 1; |
3360 | max = INT_MAX; |
3361 | ecode++; |
3362 | goto REPEATCHAR; |
3363 | |
3364 | case OP_POSQUERY: |
3365 | case OP_POSQUERYI: |
3366 | possessive = TRUE; |
3367 | min = 0; |
3368 | max = 1; |
3369 | ecode++; |
3370 | goto REPEATCHAR; |
3371 | |
3372 | case OP_STAR: |
3373 | case OP_STARI: |
3374 | case OP_MINSTAR: |
3375 | case OP_MINSTARI: |
3376 | case OP_PLUS: |
3377 | case OP_PLUSI: |
3378 | case OP_MINPLUS: |
3379 | case OP_MINPLUSI: |
3380 | case OP_QUERY: |
3381 | case OP_QUERYI: |
3382 | case OP_MINQUERY: |
3383 | case OP_MINQUERYI: |
3384 | c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI); |
3385 | minimize = (c & 1) != 0; |
3386 | min = rep_min[c]; /* Pick up values from tables; */ |
3387 | max = rep_max[c]; /* zero for max => infinity */ |
3388 | if (max == 0) max = INT_MAX; |
3389 | |
3390 | /* Common code for all repeated single-character matches. We first check |
3391 | for the minimum number of characters. If the minimum equals the maximum, we |
3392 | are done. Otherwise, if minimizing, check the rest of the pattern for a |
3393 | match; if there isn't one, advance up to the maximum, one character at a |
3394 | time. |
3395 | |
3396 | If maximizing, advance up to the maximum number of matching characters, |
3397 | until eptr is past the end of the maximum run. If possessive, we are |
3398 | then done (no backing up). Otherwise, match at this position; anything |
3399 | other than no match is immediately returned. For nomatch, back up one |
3400 | character, unless we are matching \R and the last thing matched was |
3401 | \r\n, in which case, back up two bytes. When we reach the first optional |
3402 | character position, we can save stack by doing a tail recurse. |
3403 | |
3404 | The various UTF/non-UTF and caseful/caseless cases are handled separately, |
3405 | for speed. */ |
3406 | |
3407 | REPEATCHAR: |
3408 | #ifdef SUPPORT_UTF |
3409 | if (utf) |
3410 | { |
3411 | length = 1; |
3412 | charptr = ecode; |
3413 | GETCHARLEN(fc, ecode, length); |
3414 | ecode += length; |
3415 | |
3416 | /* Handle multibyte character matching specially here. There is |
3417 | support for caseless matching if UCP support is present. */ |
3418 | |
3419 | if (length > 1) |
3420 | { |
3421 | #ifdef SUPPORT_UCP |
3422 | pcre_uint32 othercase; |
3423 | if (op >= OP_STARI && /* Caseless */ |
3424 | (othercase = UCD_OTHERCASE(fc)) != fc) |
3425 | oclength = PRIV(ord2utf)(othercase, occhars); |
3426 | else oclength = 0; |
3427 | #endif /* SUPPORT_UCP */ |
3428 | |
3429 | for (i = 1; i <= min; i++) |
3430 | { |
3431 | if (eptr <= md->end_subject - length && |
3432 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3433 | #ifdef SUPPORT_UCP |
3434 | else if (oclength > 0 && |
3435 | eptr <= md->end_subject - oclength && |
3436 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3437 | #endif /* SUPPORT_UCP */ |
3438 | else |
3439 | { |
3440 | CHECK_PARTIAL(); |
3441 | RRETURN(MATCH_NOMATCH); |
3442 | } |
3443 | } |
3444 | |
3445 | if (min == max) continue; |
3446 | |
3447 | if (minimize) |
3448 | { |
3449 | for (fi = min;; fi++) |
3450 | { |
3451 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM22); |
3452 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3453 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3454 | if (eptr <= md->end_subject - length && |
3455 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3456 | #ifdef SUPPORT_UCP |
3457 | else if (oclength > 0 && |
3458 | eptr <= md->end_subject - oclength && |
3459 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3460 | #endif /* SUPPORT_UCP */ |
3461 | else |
3462 | { |
3463 | CHECK_PARTIAL(); |
3464 | RRETURN(MATCH_NOMATCH); |
3465 | } |
3466 | } |
3467 | /* Control never gets here */ |
3468 | } |
3469 | |
3470 | else /* Maximize */ |
3471 | { |
3472 | pp = eptr; |
3473 | for (i = min; i < max; i++) |
3474 | { |
3475 | if (eptr <= md->end_subject - length && |
3476 | memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length; |
3477 | #ifdef SUPPORT_UCP |
3478 | else if (oclength > 0 && |
3479 | eptr <= md->end_subject - oclength && |
3480 | memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength; |
3481 | #endif /* SUPPORT_UCP */ |
3482 | else |
3483 | { |
3484 | CHECK_PARTIAL(); |
3485 | break; |
3486 | } |
3487 | } |
3488 | |
3489 | if (possessive) continue; /* No backtracking */ |
3490 | for(;;) |
3491 | { |
3492 | if (eptr == pp) goto TAIL_RECURSE; |
3493 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM23); |
3494 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3495 | #ifdef SUPPORT_UCP |
3496 | eptr--; |
3497 | BACKCHAR(eptr); |
3498 | #else /* without SUPPORT_UCP */ |
3499 | eptr -= length; |
3500 | #endif /* SUPPORT_UCP */ |
3501 | } |
3502 | } |
3503 | /* Control never gets here */ |
3504 | } |
3505 | |
3506 | /* If the length of a UTF-8 character is 1, we fall through here, and |
3507 | obey the code as for non-UTF-8 characters below, though in this case the |
3508 | value of fc will always be < 128. */ |
3509 | } |
3510 | else |
3511 | #endif /* SUPPORT_UTF */ |
3512 | /* When not in UTF-8 mode, load a single-byte character. */ |
3513 | fc = *ecode++; |
3514 | |
3515 | /* The value of fc at this point is always one character, though we may |
3516 | or may not be in UTF mode. The code is duplicated for the caseless and |
3517 | caseful cases, for speed, since matching characters is likely to be quite |
3518 | common. First, ensure the minimum number of matches are present. If min = |
3519 | max, continue at the same level without recursing. Otherwise, if |
3520 | minimizing, keep trying the rest of the expression and advancing one |
3521 | matching character if failing, up to the maximum. Alternatively, if |
3522 | maximizing, find the maximum number of characters and work backwards. */ |
3523 | |
3524 | DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max, |
3525 | max, (char *)eptr)); |
3526 | |
3527 | if (op >= OP_STARI) /* Caseless */ |
3528 | { |
3529 | #ifdef COMPILE_PCRE8 |
3530 | /* fc must be < 128 if UTF is enabled. */ |
3531 | foc = md->fcc[fc]; |
3532 | #else |
3533 | #ifdef SUPPORT_UTF |
3534 | #ifdef SUPPORT_UCP |
3535 | if (utf && fc > 127) |
3536 | foc = UCD_OTHERCASE(fc); |
3537 | #else |
3538 | if (utf && fc > 127) |
3539 | foc = fc; |
3540 | #endif /* SUPPORT_UCP */ |
3541 | else |
3542 | #endif /* SUPPORT_UTF */ |
3543 | foc = TABLE_GET(fc, md->fcc, fc); |
3544 | #endif /* COMPILE_PCRE8 */ |
3545 | |
3546 | for (i = 1; i <= min; i++) |
3547 | { |
3548 | pcre_uint32 cc; /* Faster than pcre_uchar */ |
3549 | if (eptr >= md->end_subject) |
3550 | { |
3551 | SCHECK_PARTIAL(); |
3552 | RRETURN(MATCH_NOMATCH); |
3553 | } |
3554 | cc = RAWUCHARTEST(eptr); |
3555 | if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH); |
3556 | eptr++; |
3557 | } |
3558 | if (min == max) continue; |
3559 | if (minimize) |
3560 | { |
3561 | for (fi = min;; fi++) |
3562 | { |
3563 | pcre_uint32 cc; /* Faster than pcre_uchar */ |
3564 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM24); |
3565 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3566 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3567 | if (eptr >= md->end_subject) |
3568 | { |
3569 | SCHECK_PARTIAL(); |
3570 | RRETURN(MATCH_NOMATCH); |
3571 | } |
3572 | cc = RAWUCHARTEST(eptr); |
3573 | if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH); |
3574 | eptr++; |
3575 | } |
3576 | /* Control never gets here */ |
3577 | } |
3578 | else /* Maximize */ |
3579 | { |
3580 | pp = eptr; |
3581 | for (i = min; i < max; i++) |
3582 | { |
3583 | pcre_uint32 cc; /* Faster than pcre_uchar */ |
3584 | if (eptr >= md->end_subject) |
3585 | { |
3586 | SCHECK_PARTIAL(); |
3587 | break; |
3588 | } |
3589 | cc = RAWUCHARTEST(eptr); |
3590 | if (fc != cc && foc != cc) break; |
3591 | eptr++; |
3592 | } |
3593 | |
3594 | if (possessive) continue; /* No backtracking */ |
3595 | for (;;) |
3596 | { |
3597 | if (eptr == pp) goto TAIL_RECURSE; |
3598 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM25); |
3599 | eptr--; |
3600 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3601 | } |
3602 | RRETURN(MATCH_NOMATCH); |
3603 | } |
3604 | /* Control never gets here */ |
3605 | } |
3606 | |
3607 | /* Caseful comparisons (includes all multi-byte characters) */ |
3608 | |
3609 | else |
3610 | { |
3611 | for (i = 1; i <= min; i++) |
3612 | { |
3613 | if (eptr >= md->end_subject) |
3614 | { |
3615 | SCHECK_PARTIAL(); |
3616 | RRETURN(MATCH_NOMATCH); |
3617 | } |
3618 | if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH); |
3619 | } |
3620 | |
3621 | if (min == max) continue; |
3622 | |
3623 | if (minimize) |
3624 | { |
3625 | for (fi = min;; fi++) |
3626 | { |
3627 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM26); |
3628 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3629 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3630 | if (eptr >= md->end_subject) |
3631 | { |
3632 | SCHECK_PARTIAL(); |
3633 | RRETURN(MATCH_NOMATCH); |
3634 | } |
3635 | if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH); |
3636 | } |
3637 | /* Control never gets here */ |
3638 | } |
3639 | else /* Maximize */ |
3640 | { |
3641 | pp = eptr; |
3642 | for (i = min; i < max; i++) |
3643 | { |
3644 | if (eptr >= md->end_subject) |
3645 | { |
3646 | SCHECK_PARTIAL(); |
3647 | break; |
3648 | } |
3649 | if (fc != RAWUCHARTEST(eptr)) break; |
3650 | eptr++; |
3651 | } |
3652 | if (possessive) continue; /* No backtracking */ |
3653 | for (;;) |
3654 | { |
3655 | if (eptr == pp) goto TAIL_RECURSE; |
3656 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM27); |
3657 | eptr--; |
3658 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3659 | } |
3660 | RRETURN(MATCH_NOMATCH); |
3661 | } |
3662 | } |
3663 | /* Control never gets here */ |
3664 | |
3665 | /* Match a negated single one-byte character. The character we are |
3666 | checking can be multibyte. */ |
3667 | |
3668 | case OP_NOT: |
3669 | case OP_NOTI: |
3670 | if (eptr >= md->end_subject) |
3671 | { |
3672 | SCHECK_PARTIAL(); |
3673 | RRETURN(MATCH_NOMATCH); |
3674 | } |
3675 | #ifdef SUPPORT_UTF |
3676 | if (utf) |
3677 | { |
3678 | register pcre_uint32 ch, och; |
3679 | |
3680 | ecode++; |
3681 | GETCHARINC(ch, ecode); |
3682 | GETCHARINC(c, eptr); |
3683 | |
3684 | if (op == OP_NOT) |
3685 | { |
3686 | if (ch == c) RRETURN(MATCH_NOMATCH); |
3687 | } |
3688 | else |
3689 | { |
3690 | #ifdef SUPPORT_UCP |
3691 | if (ch > 127) |
3692 | och = UCD_OTHERCASE(ch); |
3693 | #else |
3694 | if (ch > 127) |
3695 | och = ch; |
3696 | #endif /* SUPPORT_UCP */ |
3697 | else |
3698 | och = TABLE_GET(ch, md->fcc, ch); |
3699 | if (ch == c || och == c) RRETURN(MATCH_NOMATCH); |
3700 | } |
3701 | } |
3702 | else |
3703 | #endif |
3704 | { |
3705 | register pcre_uint32 ch = ecode[1]; |
3706 | c = *eptr++; |
3707 | if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c)) |
3708 | RRETURN(MATCH_NOMATCH); |
3709 | ecode += 2; |
3710 | } |
3711 | break; |
3712 | |
3713 | /* Match a negated single one-byte character repeatedly. This is almost a |
3714 | repeat of the code for a repeated single character, but I haven't found a |
3715 | nice way of commoning these up that doesn't require a test of the |
3716 | positive/negative option for each character match. Maybe that wouldn't add |
3717 | very much to the time taken, but character matching *is* what this is all |
3718 | about... */ |
3719 | |
3720 | case OP_NOTEXACT: |
3721 | case OP_NOTEXACTI: |
3722 | min = max = GET2(ecode, 1); |
3723 | ecode += 1 + IMM2_SIZE; |
3724 | goto REPEATNOTCHAR; |
3725 | |
3726 | case OP_NOTUPTO: |
3727 | case OP_NOTUPTOI: |
3728 | case OP_NOTMINUPTO: |
3729 | case OP_NOTMINUPTOI: |
3730 | min = 0; |
3731 | max = GET2(ecode, 1); |
3732 | minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI; |
3733 | ecode += 1 + IMM2_SIZE; |
3734 | goto REPEATNOTCHAR; |
3735 | |
3736 | case OP_NOTPOSSTAR: |
3737 | case OP_NOTPOSSTARI: |
3738 | possessive = TRUE; |
3739 | min = 0; |
3740 | max = INT_MAX; |
3741 | ecode++; |
3742 | goto REPEATNOTCHAR; |
3743 | |
3744 | case OP_NOTPOSPLUS: |
3745 | case OP_NOTPOSPLUSI: |
3746 | possessive = TRUE; |
3747 | min = 1; |
3748 | max = INT_MAX; |
3749 | ecode++; |
3750 | goto REPEATNOTCHAR; |
3751 | |
3752 | case OP_NOTPOSQUERY: |
3753 | case OP_NOTPOSQUERYI: |
3754 | possessive = TRUE; |
3755 | min = 0; |
3756 | max = 1; |
3757 | ecode++; |
3758 | goto REPEATNOTCHAR; |
3759 | |
3760 | case OP_NOTPOSUPTO: |
3761 | case OP_NOTPOSUPTOI: |
3762 | possessive = TRUE; |
3763 | min = 0; |
3764 | max = GET2(ecode, 1); |
3765 | ecode += 1 + IMM2_SIZE; |
3766 | goto REPEATNOTCHAR; |
3767 | |
3768 | case OP_NOTSTAR: |
3769 | case OP_NOTSTARI: |
3770 | case OP_NOTMINSTAR: |
3771 | case OP_NOTMINSTARI: |
3772 | case OP_NOTPLUS: |
3773 | case OP_NOTPLUSI: |
3774 | case OP_NOTMINPLUS: |
3775 | case OP_NOTMINPLUSI: |
3776 | case OP_NOTQUERY: |
3777 | case OP_NOTQUERYI: |
3778 | case OP_NOTMINQUERY: |
3779 | case OP_NOTMINQUERYI: |
3780 | c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR); |
3781 | minimize = (c & 1) != 0; |
3782 | min = rep_min[c]; /* Pick up values from tables; */ |
3783 | max = rep_max[c]; /* zero for max => infinity */ |
3784 | if (max == 0) max = INT_MAX; |
3785 | |
3786 | /* Common code for all repeated single-byte matches. */ |
3787 | |
3788 | REPEATNOTCHAR: |
3789 | GETCHARINCTEST(fc, ecode); |
3790 | |
3791 | /* The code is duplicated for the caseless and caseful cases, for speed, |
3792 | since matching characters is likely to be quite common. First, ensure the |
3793 | minimum number of matches are present. If min = max, continue at the same |
3794 | level without recursing. Otherwise, if minimizing, keep trying the rest of |
3795 | the expression and advancing one matching character if failing, up to the |
3796 | maximum. Alternatively, if maximizing, find the maximum number of |
3797 | characters and work backwards. */ |
3798 | |
3799 | DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max, |
3800 | max, (char *)eptr)); |
3801 | |
3802 | if (op >= OP_NOTSTARI) /* Caseless */ |
3803 | { |
3804 | #ifdef SUPPORT_UTF |
3805 | #ifdef SUPPORT_UCP |
3806 | if (utf && fc > 127) |
3807 | foc = UCD_OTHERCASE(fc); |
3808 | #else |
3809 | if (utf && fc > 127) |
3810 | foc = fc; |
3811 | #endif /* SUPPORT_UCP */ |
3812 | else |
3813 | #endif /* SUPPORT_UTF */ |
3814 | foc = TABLE_GET(fc, md->fcc, fc); |
3815 | |
3816 | #ifdef SUPPORT_UTF |
3817 | if (utf) |
3818 | { |
3819 | register pcre_uint32 d; |
3820 | for (i = 1; i <= min; i++) |
3821 | { |
3822 | if (eptr >= md->end_subject) |
3823 | { |
3824 | SCHECK_PARTIAL(); |
3825 | RRETURN(MATCH_NOMATCH); |
3826 | } |
3827 | GETCHARINC(d, eptr); |
3828 | if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH); |
3829 | } |
3830 | } |
3831 | else |
3832 | #endif /* SUPPORT_UTF */ |
3833 | /* Not UTF mode */ |
3834 | { |
3835 | for (i = 1; i <= min; i++) |
3836 | { |
3837 | if (eptr >= md->end_subject) |
3838 | { |
3839 | SCHECK_PARTIAL(); |
3840 | RRETURN(MATCH_NOMATCH); |
3841 | } |
3842 | if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH); |
3843 | eptr++; |
3844 | } |
3845 | } |
3846 | |
3847 | if (min == max) continue; |
3848 | |
3849 | if (minimize) |
3850 | { |
3851 | #ifdef SUPPORT_UTF |
3852 | if (utf) |
3853 | { |
3854 | register pcre_uint32 d; |
3855 | for (fi = min;; fi++) |
3856 | { |
3857 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM28); |
3858 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3859 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3860 | if (eptr >= md->end_subject) |
3861 | { |
3862 | SCHECK_PARTIAL(); |
3863 | RRETURN(MATCH_NOMATCH); |
3864 | } |
3865 | GETCHARINC(d, eptr); |
3866 | if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH); |
3867 | } |
3868 | } |
3869 | else |
3870 | #endif /*SUPPORT_UTF */ |
3871 | /* Not UTF mode */ |
3872 | { |
3873 | for (fi = min;; fi++) |
3874 | { |
3875 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM29); |
3876 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3877 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3878 | if (eptr >= md->end_subject) |
3879 | { |
3880 | SCHECK_PARTIAL(); |
3881 | RRETURN(MATCH_NOMATCH); |
3882 | } |
3883 | if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH); |
3884 | eptr++; |
3885 | } |
3886 | } |
3887 | /* Control never gets here */ |
3888 | } |
3889 | |
3890 | /* Maximize case */ |
3891 | |
3892 | else |
3893 | { |
3894 | pp = eptr; |
3895 | |
3896 | #ifdef SUPPORT_UTF |
3897 | if (utf) |
3898 | { |
3899 | register pcre_uint32 d; |
3900 | for (i = min; i < max; i++) |
3901 | { |
3902 | int len = 1; |
3903 | if (eptr >= md->end_subject) |
3904 | { |
3905 | SCHECK_PARTIAL(); |
3906 | break; |
3907 | } |
3908 | GETCHARLEN(d, eptr, len); |
3909 | if (fc == d || (unsigned int)foc == d) break; |
3910 | eptr += len; |
3911 | } |
3912 | if (possessive) continue; /* No backtracking */ |
3913 | for(;;) |
3914 | { |
3915 | if (eptr == pp) goto TAIL_RECURSE; |
3916 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM30); |
3917 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3918 | eptr--; |
3919 | BACKCHAR(eptr); |
3920 | } |
3921 | } |
3922 | else |
3923 | #endif /* SUPPORT_UTF */ |
3924 | /* Not UTF mode */ |
3925 | { |
3926 | for (i = min; i < max; i++) |
3927 | { |
3928 | if (eptr >= md->end_subject) |
3929 | { |
3930 | SCHECK_PARTIAL(); |
3931 | break; |
3932 | } |
3933 | if (fc == *eptr || foc == *eptr) break; |
3934 | eptr++; |
3935 | } |
3936 | if (possessive) continue; /* No backtracking */ |
3937 | for (;;) |
3938 | { |
3939 | if (eptr == pp) goto TAIL_RECURSE; |
3940 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM31); |
3941 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3942 | eptr--; |
3943 | } |
3944 | } |
3945 | |
3946 | RRETURN(MATCH_NOMATCH); |
3947 | } |
3948 | /* Control never gets here */ |
3949 | } |
3950 | |
3951 | /* Caseful comparisons */ |
3952 | |
3953 | else |
3954 | { |
3955 | #ifdef SUPPORT_UTF |
3956 | if (utf) |
3957 | { |
3958 | register pcre_uint32 d; |
3959 | for (i = 1; i <= min; i++) |
3960 | { |
3961 | if (eptr >= md->end_subject) |
3962 | { |
3963 | SCHECK_PARTIAL(); |
3964 | RRETURN(MATCH_NOMATCH); |
3965 | } |
3966 | GETCHARINC(d, eptr); |
3967 | if (fc == d) RRETURN(MATCH_NOMATCH); |
3968 | } |
3969 | } |
3970 | else |
3971 | #endif |
3972 | /* Not UTF mode */ |
3973 | { |
3974 | for (i = 1; i <= min; i++) |
3975 | { |
3976 | if (eptr >= md->end_subject) |
3977 | { |
3978 | SCHECK_PARTIAL(); |
3979 | RRETURN(MATCH_NOMATCH); |
3980 | } |
3981 | if (fc == *eptr++) RRETURN(MATCH_NOMATCH); |
3982 | } |
3983 | } |
3984 | |
3985 | if (min == max) continue; |
3986 | |
3987 | if (minimize) |
3988 | { |
3989 | #ifdef SUPPORT_UTF |
3990 | if (utf) |
3991 | { |
3992 | register pcre_uint32 d; |
3993 | for (fi = min;; fi++) |
3994 | { |
3995 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM32); |
3996 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
3997 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
3998 | if (eptr >= md->end_subject) |
3999 | { |
4000 | SCHECK_PARTIAL(); |
4001 | RRETURN(MATCH_NOMATCH); |
4002 | } |
4003 | GETCHARINC(d, eptr); |
4004 | if (fc == d) RRETURN(MATCH_NOMATCH); |
4005 | } |
4006 | } |
4007 | else |
4008 | #endif |
4009 | /* Not UTF mode */ |
4010 | { |
4011 | for (fi = min;; fi++) |
4012 | { |
4013 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM33); |
4014 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4015 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4016 | if (eptr >= md->end_subject) |
4017 | { |
4018 | SCHECK_PARTIAL(); |
4019 | RRETURN(MATCH_NOMATCH); |
4020 | } |
4021 | if (fc == *eptr++) RRETURN(MATCH_NOMATCH); |
4022 | } |
4023 | } |
4024 | /* Control never gets here */ |
4025 | } |
4026 | |
4027 | /* Maximize case */ |
4028 | |
4029 | else |
4030 | { |
4031 | pp = eptr; |
4032 | |
4033 | #ifdef SUPPORT_UTF |
4034 | if (utf) |
4035 | { |
4036 | register pcre_uint32 d; |
4037 | for (i = min; i < max; i++) |
4038 | { |
4039 | int len = 1; |
4040 | if (eptr >= md->end_subject) |
4041 | { |
4042 | SCHECK_PARTIAL(); |
4043 | break; |
4044 | } |
4045 | GETCHARLEN(d, eptr, len); |
4046 | if (fc == d) break; |
4047 | eptr += len; |
4048 | } |
4049 | if (possessive) continue; /* No backtracking */ |
4050 | for(;;) |
4051 | { |
4052 | if (eptr == pp) goto TAIL_RECURSE; |
4053 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM34); |
4054 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4055 | eptr--; |
4056 | BACKCHAR(eptr); |
4057 | } |
4058 | } |
4059 | else |
4060 | #endif |
4061 | /* Not UTF mode */ |
4062 | { |
4063 | for (i = min; i < max; i++) |
4064 | { |
4065 | if (eptr >= md->end_subject) |
4066 | { |
4067 | SCHECK_PARTIAL(); |
4068 | break; |
4069 | } |
4070 | if (fc == *eptr) break; |
4071 | eptr++; |
4072 | } |
4073 | if (possessive) continue; /* No backtracking */ |
4074 | for (;;) |
4075 | { |
4076 | if (eptr == pp) goto TAIL_RECURSE; |
4077 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM35); |
4078 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4079 | eptr--; |
4080 | } |
4081 | } |
4082 | |
4083 | RRETURN(MATCH_NOMATCH); |
4084 | } |
4085 | } |
4086 | /* Control never gets here */ |
4087 | |
4088 | /* Match a single character type repeatedly; several different opcodes |
4089 | share code. This is very similar to the code for single characters, but we |
4090 | repeat it in the interests of efficiency. */ |
4091 | |
4092 | case OP_TYPEEXACT: |
4093 | min = max = GET2(ecode, 1); |
4094 | minimize = TRUE; |
4095 | ecode += 1 + IMM2_SIZE; |
4096 | goto REPEATTYPE; |
4097 | |
4098 | case OP_TYPEUPTO: |
4099 | case OP_TYPEMINUPTO: |
4100 | min = 0; |
4101 | max = GET2(ecode, 1); |
4102 | minimize = *ecode == OP_TYPEMINUPTO; |
4103 | ecode += 1 + IMM2_SIZE; |
4104 | goto REPEATTYPE; |
4105 | |
4106 | case OP_TYPEPOSSTAR: |
4107 | possessive = TRUE; |
4108 | min = 0; |
4109 | max = INT_MAX; |
4110 | ecode++; |
4111 | goto REPEATTYPE; |
4112 | |
4113 | case OP_TYPEPOSPLUS: |
4114 | possessive = TRUE; |
4115 | min = 1; |
4116 | max = INT_MAX; |
4117 | ecode++; |
4118 | goto REPEATTYPE; |
4119 | |
4120 | case OP_TYPEPOSQUERY: |
4121 | possessive = TRUE; |
4122 | min = 0; |
4123 | max = 1; |
4124 | ecode++; |
4125 | goto REPEATTYPE; |
4126 | |
4127 | case OP_TYPEPOSUPTO: |
4128 | possessive = TRUE; |
4129 | min = 0; |
4130 | max = GET2(ecode, 1); |
4131 | ecode += 1 + IMM2_SIZE; |
4132 | goto REPEATTYPE; |
4133 | |
4134 | case OP_TYPESTAR: |
4135 | case OP_TYPEMINSTAR: |
4136 | case OP_TYPEPLUS: |
4137 | case OP_TYPEMINPLUS: |
4138 | case OP_TYPEQUERY: |
4139 | case OP_TYPEMINQUERY: |
4140 | c = *ecode++ - OP_TYPESTAR; |
4141 | minimize = (c & 1) != 0; |
4142 | min = rep_min[c]; /* Pick up values from tables; */ |
4143 | max = rep_max[c]; /* zero for max => infinity */ |
4144 | if (max == 0) max = INT_MAX; |
4145 | |
4146 | /* Common code for all repeated single character type matches. Note that |
4147 | in UTF-8 mode, '.' matches a character of any length, but for the other |
4148 | character types, the valid characters are all one-byte long. */ |
4149 | |
4150 | REPEATTYPE: |
4151 | ctype = *ecode++; /* Code for the character type */ |
4152 | |
4153 | #ifdef SUPPORT_UCP |
4154 | if (ctype == OP_PROP || ctype == OP_NOTPROP) |
4155 | { |
4156 | prop_fail_result = ctype == OP_NOTPROP; |
4157 | prop_type = *ecode++; |
4158 | prop_value = *ecode++; |
4159 | } |
4160 | else prop_type = -1; |
4161 | #endif |
4162 | |
4163 | /* First, ensure the minimum number of matches are present. Use inline |
4164 | code for maximizing the speed, and do the type test once at the start |
4165 | (i.e. keep it out of the loop). Separate the UTF-8 code completely as that |
4166 | is tidier. Also separate the UCP code, which can be the same for both UTF-8 |
4167 | and single-bytes. */ |
4168 | |
4169 | if (min > 0) |
4170 | { |
4171 | #ifdef SUPPORT_UCP |
4172 | if (prop_type >= 0) |
4173 | { |
4174 | switch(prop_type) |
4175 | { |
4176 | case PT_ANY: |
4177 | if (prop_fail_result) RRETURN(MATCH_NOMATCH); |
4178 | for (i = 1; i <= min; i++) |
4179 | { |
4180 | if (eptr >= md->end_subject) |
4181 | { |
4182 | SCHECK_PARTIAL(); |
4183 | RRETURN(MATCH_NOMATCH); |
4184 | } |
4185 | GETCHARINCTEST(c, eptr); |
4186 | } |
4187 | break; |
4188 | |
4189 | case PT_LAMP: |
4190 | for (i = 1; i <= min; i++) |
4191 | { |
4192 | int chartype; |
4193 | if (eptr >= md->end_subject) |
4194 | { |
4195 | SCHECK_PARTIAL(); |
4196 | RRETURN(MATCH_NOMATCH); |
4197 | } |
4198 | GETCHARINCTEST(c, eptr); |
4199 | chartype = UCD_CHARTYPE(c); |
4200 | if ((chartype == ucp_Lu || |
4201 | chartype == ucp_Ll || |
4202 | chartype == ucp_Lt) == prop_fail_result) |
4203 | RRETURN(MATCH_NOMATCH); |
4204 | } |
4205 | break; |
4206 | |
4207 | case PT_GC: |
4208 | for (i = 1; i <= min; i++) |
4209 | { |
4210 | if (eptr >= md->end_subject) |
4211 | { |
4212 | SCHECK_PARTIAL(); |
4213 | RRETURN(MATCH_NOMATCH); |
4214 | } |
4215 | GETCHARINCTEST(c, eptr); |
4216 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) |
4217 | RRETURN(MATCH_NOMATCH); |
4218 | } |
4219 | break; |
4220 | |
4221 | case PT_PC: |
4222 | for (i = 1; i <= min; i++) |
4223 | { |
4224 | if (eptr >= md->end_subject) |
4225 | { |
4226 | SCHECK_PARTIAL(); |
4227 | RRETURN(MATCH_NOMATCH); |
4228 | } |
4229 | GETCHARINCTEST(c, eptr); |
4230 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) |
4231 | RRETURN(MATCH_NOMATCH); |
4232 | } |
4233 | break; |
4234 | |
4235 | case PT_SC: |
4236 | for (i = 1; i <= min; i++) |
4237 | { |
4238 | if (eptr >= md->end_subject) |
4239 | { |
4240 | SCHECK_PARTIAL(); |
4241 | RRETURN(MATCH_NOMATCH); |
4242 | } |
4243 | GETCHARINCTEST(c, eptr); |
4244 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) |
4245 | RRETURN(MATCH_NOMATCH); |
4246 | } |
4247 | break; |
4248 | |
4249 | case PT_ALNUM: |
4250 | for (i = 1; i <= min; i++) |
4251 | { |
4252 | int category; |
4253 | if (eptr >= md->end_subject) |
4254 | { |
4255 | SCHECK_PARTIAL(); |
4256 | RRETURN(MATCH_NOMATCH); |
4257 | } |
4258 | GETCHARINCTEST(c, eptr); |
4259 | category = UCD_CATEGORY(c); |
4260 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
4261 | RRETURN(MATCH_NOMATCH); |
4262 | } |
4263 | break; |
4264 | |
4265 | case PT_SPACE: /* Perl space */ |
4266 | for (i = 1; i <= min; i++) |
4267 | { |
4268 | if (eptr >= md->end_subject) |
4269 | { |
4270 | SCHECK_PARTIAL(); |
4271 | RRETURN(MATCH_NOMATCH); |
4272 | } |
4273 | GETCHARINCTEST(c, eptr); |
4274 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4275 | c == CHAR_FF || c == CHAR_CR) |
4276 | == prop_fail_result) |
4277 | RRETURN(MATCH_NOMATCH); |
4278 | } |
4279 | break; |
4280 | |
4281 | case PT_PXSPACE: /* POSIX space */ |
4282 | for (i = 1; i <= min; i++) |
4283 | { |
4284 | if (eptr >= md->end_subject) |
4285 | { |
4286 | SCHECK_PARTIAL(); |
4287 | RRETURN(MATCH_NOMATCH); |
4288 | } |
4289 | GETCHARINCTEST(c, eptr); |
4290 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
4291 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
4292 | == prop_fail_result) |
4293 | RRETURN(MATCH_NOMATCH); |
4294 | } |
4295 | break; |
4296 | |
4297 | case PT_WORD: |
4298 | for (i = 1; i <= min; i++) |
4299 | { |
4300 | int category; |
4301 | if (eptr >= md->end_subject) |
4302 | { |
4303 | SCHECK_PARTIAL(); |
4304 | RRETURN(MATCH_NOMATCH); |
4305 | } |
4306 | GETCHARINCTEST(c, eptr); |
4307 | category = UCD_CATEGORY(c); |
4308 | if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE) |
4309 | == prop_fail_result) |
4310 | RRETURN(MATCH_NOMATCH); |
4311 | } |
4312 | break; |
4313 | |
4314 | case PT_CLIST: |
4315 | for (i = 1; i <= min; i++) |
4316 | { |
4317 | const pcre_uint32 *cp; |
4318 | if (eptr >= md->end_subject) |
4319 | { |
4320 | SCHECK_PARTIAL(); |
4321 | RRETURN(MATCH_NOMATCH); |
4322 | } |
4323 | GETCHARINCTEST(c, eptr); |
4324 | cp = PRIV(ucd_caseless_sets) + prop_value; |
4325 | for (;;) |
4326 | { |
4327 | if (c < *cp) |
4328 | { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } } |
4329 | if (c == *cp++) |
4330 | { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; } |
4331 | } |
4332 | } |
4333 | break; |
4334 | |
4335 | case PT_UCNC: |
4336 | for (i = 1; i <= min; i++) |
4337 | { |
4338 | if (eptr >= md->end_subject) |
4339 | { |
4340 | SCHECK_PARTIAL(); |
4341 | RRETURN(MATCH_NOMATCH); |
4342 | } |
4343 | GETCHARINCTEST(c, eptr); |
4344 | if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
4345 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
4346 | c >= 0xe000) == prop_fail_result) |
4347 | RRETURN(MATCH_NOMATCH); |
4348 | } |
4349 | break; |
4350 | |
4351 | /* This should not occur */ |
4352 | |
4353 | default: |
4354 | RRETURN(PCRE_ERROR_INTERNAL); |
4355 | } |
4356 | } |
4357 | |
4358 | /* Match extended Unicode sequences. We will get here only if the |
4359 | support is in the binary; otherwise a compile-time error occurs. */ |
4360 | |
4361 | else if (ctype == OP_EXTUNI) |
4362 | { |
4363 | for (i = 1; i <= min; i++) |
4364 | { |
4365 | if (eptr >= md->end_subject) |
4366 | { |
4367 | SCHECK_PARTIAL(); |
4368 | RRETURN(MATCH_NOMATCH); |
4369 | } |
4370 | else |
4371 | { |
4372 | int lgb, rgb; |
4373 | GETCHARINCTEST(c, eptr); |
4374 | lgb = UCD_GRAPHBREAK(c); |
4375 | while (eptr < md->end_subject) |
4376 | { |
4377 | int len = 1; |
4378 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
4379 | rgb = UCD_GRAPHBREAK(c); |
4380 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
4381 | lgb = rgb; |
4382 | eptr += len; |
4383 | } |
4384 | } |
4385 | CHECK_PARTIAL(); |
4386 | } |
4387 | } |
4388 | |
4389 | else |
4390 | #endif /* SUPPORT_UCP */ |
4391 | |
4392 | /* Handle all other cases when the coding is UTF-8 */ |
4393 | |
4394 | #ifdef SUPPORT_UTF |
4395 | if (utf) switch(ctype) |
4396 | { |
4397 | case OP_ANY: |
4398 | for (i = 1; i <= min; i++) |
4399 | { |
4400 | if (eptr >= md->end_subject) |
4401 | { |
4402 | SCHECK_PARTIAL(); |
4403 | RRETURN(MATCH_NOMATCH); |
4404 | } |
4405 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
4406 | if (md->partial != 0 && |
4407 | eptr + 1 >= md->end_subject && |
4408 | NLBLOCK->nltype == NLTYPE_FIXED && |
4409 | NLBLOCK->nllen == 2 && |
4410 | RAWUCHAR(eptr) == NLBLOCK->nl[0]) |
4411 | { |
4412 | md->hitend = TRUE; |
4413 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
4414 | } |
4415 | eptr++; |
4416 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4417 | } |
4418 | break; |
4419 | |
4420 | case OP_ALLANY: |
4421 | for (i = 1; i <= min; i++) |
4422 | { |
4423 | if (eptr >= md->end_subject) |
4424 | { |
4425 | SCHECK_PARTIAL(); |
4426 | RRETURN(MATCH_NOMATCH); |
4427 | } |
4428 | eptr++; |
4429 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4430 | } |
4431 | break; |
4432 | |
4433 | case OP_ANYBYTE: |
4434 | if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH); |
4435 | eptr += min; |
4436 | break; |
4437 | |
4438 | case OP_ANYNL: |
4439 | for (i = 1; i <= min; i++) |
4440 | { |
4441 | if (eptr >= md->end_subject) |
4442 | { |
4443 | SCHECK_PARTIAL(); |
4444 | RRETURN(MATCH_NOMATCH); |
4445 | } |
4446 | GETCHARINC(c, eptr); |
4447 | switch(c) |
4448 | { |
4449 | default: RRETURN(MATCH_NOMATCH); |
4450 | |
4451 | case CHAR_CR: |
4452 | if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++; |
4453 | break; |
4454 | |
4455 | case CHAR_LF: |
4456 | break; |
4457 | |
4458 | case CHAR_VT: |
4459 | case CHAR_FF: |
4460 | case CHAR_NEL: |
4461 | #ifndef EBCDIC |
4462 | case 0x2028: |
4463 | case 0x2029: |
4464 | #endif /* Not EBCDIC */ |
4465 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
4466 | break; |
4467 | } |
4468 | } |
4469 | break; |
4470 | |
4471 | case OP_NOT_HSPACE: |
4472 | for (i = 1; i <= min; i++) |
4473 | { |
4474 | if (eptr >= md->end_subject) |
4475 | { |
4476 | SCHECK_PARTIAL(); |
4477 | RRETURN(MATCH_NOMATCH); |
4478 | } |
4479 | GETCHARINC(c, eptr); |
4480 | switch(c) |
4481 | { |
4482 | HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */ |
4483 | default: break; |
4484 | } |
4485 | } |
4486 | break; |
4487 | |
4488 | case OP_HSPACE: |
4489 | for (i = 1; i <= min; i++) |
4490 | { |
4491 | if (eptr >= md->end_subject) |
4492 | { |
4493 | SCHECK_PARTIAL(); |
4494 | RRETURN(MATCH_NOMATCH); |
4495 | } |
4496 | GETCHARINC(c, eptr); |
4497 | switch(c) |
4498 | { |
4499 | HSPACE_CASES: break; /* Byte and multibyte cases */ |
4500 | default: RRETURN(MATCH_NOMATCH); |
4501 | } |
4502 | } |
4503 | break; |
4504 | |
4505 | case OP_NOT_VSPACE: |
4506 | for (i = 1; i <= min; i++) |
4507 | { |
4508 | if (eptr >= md->end_subject) |
4509 | { |
4510 | SCHECK_PARTIAL(); |
4511 | RRETURN(MATCH_NOMATCH); |
4512 | } |
4513 | GETCHARINC(c, eptr); |
4514 | switch(c) |
4515 | { |
4516 | VSPACE_CASES: RRETURN(MATCH_NOMATCH); |
4517 | default: break; |
4518 | } |
4519 | } |
4520 | break; |
4521 | |
4522 | case OP_VSPACE: |
4523 | for (i = 1; i <= min; i++) |
4524 | { |
4525 | if (eptr >= md->end_subject) |
4526 | { |
4527 | SCHECK_PARTIAL(); |
4528 | RRETURN(MATCH_NOMATCH); |
4529 | } |
4530 | GETCHARINC(c, eptr); |
4531 | switch(c) |
4532 | { |
4533 | VSPACE_CASES: break; |
4534 | default: RRETURN(MATCH_NOMATCH); |
4535 | } |
4536 | } |
4537 | break; |
4538 | |
4539 | case OP_NOT_DIGIT: |
4540 | for (i = 1; i <= min; i++) |
4541 | { |
4542 | if (eptr >= md->end_subject) |
4543 | { |
4544 | SCHECK_PARTIAL(); |
4545 | RRETURN(MATCH_NOMATCH); |
4546 | } |
4547 | GETCHARINC(c, eptr); |
4548 | if (c < 128 && (md->ctypes[c] & ctype_digit) != 0) |
4549 | RRETURN(MATCH_NOMATCH); |
4550 | } |
4551 | break; |
4552 | |
4553 | case OP_DIGIT: |
4554 | for (i = 1; i <= min; i++) |
4555 | { |
4556 | pcre_uint32 cc; |
4557 | if (eptr >= md->end_subject) |
4558 | { |
4559 | SCHECK_PARTIAL(); |
4560 | RRETURN(MATCH_NOMATCH); |
4561 | } |
4562 | cc = RAWUCHAR(eptr); |
4563 | if (cc >= 128 || (md->ctypes[cc] & ctype_digit) == 0) |
4564 | RRETURN(MATCH_NOMATCH); |
4565 | eptr++; |
4566 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4567 | } |
4568 | break; |
4569 | |
4570 | case OP_NOT_WHITESPACE: |
4571 | for (i = 1; i <= min; i++) |
4572 | { |
4573 | pcre_uint32 cc; |
4574 | if (eptr >= md->end_subject) |
4575 | { |
4576 | SCHECK_PARTIAL(); |
4577 | RRETURN(MATCH_NOMATCH); |
4578 | } |
4579 | cc = RAWUCHAR(eptr); |
4580 | if (cc < 128 && (md->ctypes[cc] & ctype_space) != 0) |
4581 | RRETURN(MATCH_NOMATCH); |
4582 | eptr++; |
4583 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4584 | } |
4585 | break; |
4586 | |
4587 | case OP_WHITESPACE: |
4588 | for (i = 1; i <= min; i++) |
4589 | { |
4590 | pcre_uint32 cc; |
4591 | if (eptr >= md->end_subject) |
4592 | { |
4593 | SCHECK_PARTIAL(); |
4594 | RRETURN(MATCH_NOMATCH); |
4595 | } |
4596 | cc = RAWUCHAR(eptr); |
4597 | if (cc >= 128 || (md->ctypes[cc] & ctype_space) == 0) |
4598 | RRETURN(MATCH_NOMATCH); |
4599 | eptr++; |
4600 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4601 | } |
4602 | break; |
4603 | |
4604 | case OP_NOT_WORDCHAR: |
4605 | for (i = 1; i <= min; i++) |
4606 | { |
4607 | pcre_uint32 cc; |
4608 | if (eptr >= md->end_subject) |
4609 | { |
4610 | SCHECK_PARTIAL(); |
4611 | RRETURN(MATCH_NOMATCH); |
4612 | } |
4613 | cc = RAWUCHAR(eptr); |
4614 | if (cc < 128 && (md->ctypes[cc] & ctype_word) != 0) |
4615 | RRETURN(MATCH_NOMATCH); |
4616 | eptr++; |
4617 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
4618 | } |
4619 | break; |
4620 | |
4621 | case OP_WORDCHAR: |
4622 | for (i = 1; i <= min; i++) |
4623 | { |
4624 | pcre_uint32 cc; |
4625 | if (eptr >= md->end_subject) |
4626 | { |
4627 | SCHECK_PARTIAL(); |
4628 | RRETURN(MATCH_NOMATCH); |
4629 | } |
4630 | cc = RAWUCHAR(eptr); |
4631 | if (cc >= 128 || (md->ctypes[cc] & ctype_word) == 0) |
4632 | RRETURN(MATCH_NOMATCH); |
4633 | eptr++; |
4634 | /* No need to skip more bytes - we know it's a 1-byte character */ |
4635 | } |
4636 | break; |
4637 | |
4638 | default: |
4639 | RRETURN(PCRE_ERROR_INTERNAL); |
4640 | } /* End switch(ctype) */ |
4641 | |
4642 | else |
4643 | #endif /* SUPPORT_UTF */ |
4644 | |
4645 | /* Code for the non-UTF-8 case for minimum matching of operators other |
4646 | than OP_PROP and OP_NOTPROP. */ |
4647 | |
4648 | switch(ctype) |
4649 | { |
4650 | case OP_ANY: |
4651 | for (i = 1; i <= min; i++) |
4652 | { |
4653 | if (eptr >= md->end_subject) |
4654 | { |
4655 | SCHECK_PARTIAL(); |
4656 | RRETURN(MATCH_NOMATCH); |
4657 | } |
4658 | if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); |
4659 | if (md->partial != 0 && |
4660 | eptr + 1 >= md->end_subject && |
4661 | NLBLOCK->nltype == NLTYPE_FIXED && |
4662 | NLBLOCK->nllen == 2 && |
4663 | *eptr == NLBLOCK->nl[0]) |
4664 | { |
4665 | md->hitend = TRUE; |
4666 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
4667 | } |
4668 | eptr++; |
4669 | } |
4670 | break; |
4671 | |
4672 | case OP_ALLANY: |
4673 | if (eptr > md->end_subject - min) |
4674 | { |
4675 | SCHECK_PARTIAL(); |
4676 | RRETURN(MATCH_NOMATCH); |
4677 | } |
4678 | eptr += min; |
4679 | break; |
4680 | |
4681 | case OP_ANYBYTE: |
4682 | if (eptr > md->end_subject - min) |
4683 | { |
4684 | SCHECK_PARTIAL(); |
4685 | RRETURN(MATCH_NOMATCH); |
4686 | } |
4687 | eptr += min; |
4688 | break; |
4689 | |
4690 | case OP_ANYNL: |
4691 | for (i = 1; i <= min; i++) |
4692 | { |
4693 | if (eptr >= md->end_subject) |
4694 | { |
4695 | SCHECK_PARTIAL(); |
4696 | RRETURN(MATCH_NOMATCH); |
4697 | } |
4698 | switch(*eptr++) |
4699 | { |
4700 | default: RRETURN(MATCH_NOMATCH); |
4701 | |
4702 | case CHAR_CR: |
4703 | if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++; |
4704 | break; |
4705 | |
4706 | case CHAR_LF: |
4707 | break; |
4708 | |
4709 | case CHAR_VT: |
4710 | case CHAR_FF: |
4711 | case CHAR_NEL: |
4712 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4713 | case 0x2028: |
4714 | case 0x2029: |
4715 | #endif |
4716 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
4717 | break; |
4718 | } |
4719 | } |
4720 | break; |
4721 | |
4722 | case OP_NOT_HSPACE: |
4723 | for (i = 1; i <= min; i++) |
4724 | { |
4725 | if (eptr >= md->end_subject) |
4726 | { |
4727 | SCHECK_PARTIAL(); |
4728 | RRETURN(MATCH_NOMATCH); |
4729 | } |
4730 | switch(*eptr++) |
4731 | { |
4732 | default: break; |
4733 | HSPACE_BYTE_CASES: |
4734 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4735 | HSPACE_MULTIBYTE_CASES: |
4736 | #endif |
4737 | RRETURN(MATCH_NOMATCH); |
4738 | } |
4739 | } |
4740 | break; |
4741 | |
4742 | case OP_HSPACE: |
4743 | for (i = 1; i <= min; i++) |
4744 | { |
4745 | if (eptr >= md->end_subject) |
4746 | { |
4747 | SCHECK_PARTIAL(); |
4748 | RRETURN(MATCH_NOMATCH); |
4749 | } |
4750 | switch(*eptr++) |
4751 | { |
4752 | default: RRETURN(MATCH_NOMATCH); |
4753 | HSPACE_BYTE_CASES: |
4754 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4755 | HSPACE_MULTIBYTE_CASES: |
4756 | #endif |
4757 | break; |
4758 | } |
4759 | } |
4760 | break; |
4761 | |
4762 | case OP_NOT_VSPACE: |
4763 | for (i = 1; i <= min; i++) |
4764 | { |
4765 | if (eptr >= md->end_subject) |
4766 | { |
4767 | SCHECK_PARTIAL(); |
4768 | RRETURN(MATCH_NOMATCH); |
4769 | } |
4770 | switch(*eptr++) |
4771 | { |
4772 | VSPACE_BYTE_CASES: |
4773 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4774 | VSPACE_MULTIBYTE_CASES: |
4775 | #endif |
4776 | RRETURN(MATCH_NOMATCH); |
4777 | default: break; |
4778 | } |
4779 | } |
4780 | break; |
4781 | |
4782 | case OP_VSPACE: |
4783 | for (i = 1; i <= min; i++) |
4784 | { |
4785 | if (eptr >= md->end_subject) |
4786 | { |
4787 | SCHECK_PARTIAL(); |
4788 | RRETURN(MATCH_NOMATCH); |
4789 | } |
4790 | switch(*eptr++) |
4791 | { |
4792 | default: RRETURN(MATCH_NOMATCH); |
4793 | VSPACE_BYTE_CASES: |
4794 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
4795 | VSPACE_MULTIBYTE_CASES: |
4796 | #endif |
4797 | break; |
4798 | } |
4799 | } |
4800 | break; |
4801 | |
4802 | case OP_NOT_DIGIT: |
4803 | for (i = 1; i <= min; i++) |
4804 | { |
4805 | if (eptr >= md->end_subject) |
4806 | { |
4807 | SCHECK_PARTIAL(); |
4808 | RRETURN(MATCH_NOMATCH); |
4809 | } |
4810 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) |
4811 | RRETURN(MATCH_NOMATCH); |
4812 | eptr++; |
4813 | } |
4814 | break; |
4815 | |
4816 | case OP_DIGIT: |
4817 | for (i = 1; i <= min; i++) |
4818 | { |
4819 | if (eptr >= md->end_subject) |
4820 | { |
4821 | SCHECK_PARTIAL(); |
4822 | RRETURN(MATCH_NOMATCH); |
4823 | } |
4824 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) |
4825 | RRETURN(MATCH_NOMATCH); |
4826 | eptr++; |
4827 | } |
4828 | break; |
4829 | |
4830 | case OP_NOT_WHITESPACE: |
4831 | for (i = 1; i <= min; i++) |
4832 | { |
4833 | if (eptr >= md->end_subject) |
4834 | { |
4835 | SCHECK_PARTIAL(); |
4836 | RRETURN(MATCH_NOMATCH); |
4837 | } |
4838 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) |
4839 | RRETURN(MATCH_NOMATCH); |
4840 | eptr++; |
4841 | } |
4842 | break; |
4843 | |
4844 | case OP_WHITESPACE: |
4845 | for (i = 1; i <= min; i++) |
4846 | { |
4847 | if (eptr >= md->end_subject) |
4848 | { |
4849 | SCHECK_PARTIAL(); |
4850 | RRETURN(MATCH_NOMATCH); |
4851 | } |
4852 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) |
4853 | RRETURN(MATCH_NOMATCH); |
4854 | eptr++; |
4855 | } |
4856 | break; |
4857 | |
4858 | case OP_NOT_WORDCHAR: |
4859 | for (i = 1; i <= min; i++) |
4860 | { |
4861 | if (eptr >= md->end_subject) |
4862 | { |
4863 | SCHECK_PARTIAL(); |
4864 | RRETURN(MATCH_NOMATCH); |
4865 | } |
4866 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) |
4867 | RRETURN(MATCH_NOMATCH); |
4868 | eptr++; |
4869 | } |
4870 | break; |
4871 | |
4872 | case OP_WORDCHAR: |
4873 | for (i = 1; i <= min; i++) |
4874 | { |
4875 | if (eptr >= md->end_subject) |
4876 | { |
4877 | SCHECK_PARTIAL(); |
4878 | RRETURN(MATCH_NOMATCH); |
4879 | } |
4880 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) |
4881 | RRETURN(MATCH_NOMATCH); |
4882 | eptr++; |
4883 | } |
4884 | break; |
4885 | |
4886 | default: |
4887 | RRETURN(PCRE_ERROR_INTERNAL); |
4888 | } |
4889 | } |
4890 | |
4891 | /* If min = max, continue at the same level without recursing */ |
4892 | |
4893 | if (min == max) continue; |
4894 | |
4895 | /* If minimizing, we have to test the rest of the pattern before each |
4896 | subsequent match. Again, separate the UTF-8 case for speed, and also |
4897 | separate the UCP cases. */ |
4898 | |
4899 | if (minimize) |
4900 | { |
4901 | #ifdef SUPPORT_UCP |
4902 | if (prop_type >= 0) |
4903 | { |
4904 | switch(prop_type) |
4905 | { |
4906 | case PT_ANY: |
4907 | for (fi = min;; fi++) |
4908 | { |
4909 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM36); |
4910 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4911 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4912 | if (eptr >= md->end_subject) |
4913 | { |
4914 | SCHECK_PARTIAL(); |
4915 | RRETURN(MATCH_NOMATCH); |
4916 | } |
4917 | GETCHARINCTEST(c, eptr); |
4918 | if (prop_fail_result) RRETURN(MATCH_NOMATCH); |
4919 | } |
4920 | /* Control never gets here */ |
4921 | |
4922 | case PT_LAMP: |
4923 | for (fi = min;; fi++) |
4924 | { |
4925 | int chartype; |
4926 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM37); |
4927 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4928 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4929 | if (eptr >= md->end_subject) |
4930 | { |
4931 | SCHECK_PARTIAL(); |
4932 | RRETURN(MATCH_NOMATCH); |
4933 | } |
4934 | GETCHARINCTEST(c, eptr); |
4935 | chartype = UCD_CHARTYPE(c); |
4936 | if ((chartype == ucp_Lu || |
4937 | chartype == ucp_Ll || |
4938 | chartype == ucp_Lt) == prop_fail_result) |
4939 | RRETURN(MATCH_NOMATCH); |
4940 | } |
4941 | /* Control never gets here */ |
4942 | |
4943 | case PT_GC: |
4944 | for (fi = min;; fi++) |
4945 | { |
4946 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM38); |
4947 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4948 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4949 | if (eptr >= md->end_subject) |
4950 | { |
4951 | SCHECK_PARTIAL(); |
4952 | RRETURN(MATCH_NOMATCH); |
4953 | } |
4954 | GETCHARINCTEST(c, eptr); |
4955 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) |
4956 | RRETURN(MATCH_NOMATCH); |
4957 | } |
4958 | /* Control never gets here */ |
4959 | |
4960 | case PT_PC: |
4961 | for (fi = min;; fi++) |
4962 | { |
4963 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM39); |
4964 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4965 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4966 | if (eptr >= md->end_subject) |
4967 | { |
4968 | SCHECK_PARTIAL(); |
4969 | RRETURN(MATCH_NOMATCH); |
4970 | } |
4971 | GETCHARINCTEST(c, eptr); |
4972 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) |
4973 | RRETURN(MATCH_NOMATCH); |
4974 | } |
4975 | /* Control never gets here */ |
4976 | |
4977 | case PT_SC: |
4978 | for (fi = min;; fi++) |
4979 | { |
4980 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM40); |
4981 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
4982 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
4983 | if (eptr >= md->end_subject) |
4984 | { |
4985 | SCHECK_PARTIAL(); |
4986 | RRETURN(MATCH_NOMATCH); |
4987 | } |
4988 | GETCHARINCTEST(c, eptr); |
4989 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) |
4990 | RRETURN(MATCH_NOMATCH); |
4991 | } |
4992 | /* Control never gets here */ |
4993 | |
4994 | case PT_ALNUM: |
4995 | for (fi = min;; fi++) |
4996 | { |
4997 | int category; |
4998 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM59); |
4999 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5000 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5001 | if (eptr >= md->end_subject) |
5002 | { |
5003 | SCHECK_PARTIAL(); |
5004 | RRETURN(MATCH_NOMATCH); |
5005 | } |
5006 | GETCHARINCTEST(c, eptr); |
5007 | category = UCD_CATEGORY(c); |
5008 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
5009 | RRETURN(MATCH_NOMATCH); |
5010 | } |
5011 | /* Control never gets here */ |
5012 | |
5013 | case PT_SPACE: /* Perl space */ |
5014 | for (fi = min;; fi++) |
5015 | { |
5016 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM60); |
5017 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5018 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5019 | if (eptr >= md->end_subject) |
5020 | { |
5021 | SCHECK_PARTIAL(); |
5022 | RRETURN(MATCH_NOMATCH); |
5023 | } |
5024 | GETCHARINCTEST(c, eptr); |
5025 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5026 | c == CHAR_FF || c == CHAR_CR) |
5027 | == prop_fail_result) |
5028 | RRETURN(MATCH_NOMATCH); |
5029 | } |
5030 | /* Control never gets here */ |
5031 | |
5032 | case PT_PXSPACE: /* POSIX space */ |
5033 | for (fi = min;; fi++) |
5034 | { |
5035 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM61); |
5036 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5037 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5038 | if (eptr >= md->end_subject) |
5039 | { |
5040 | SCHECK_PARTIAL(); |
5041 | RRETURN(MATCH_NOMATCH); |
5042 | } |
5043 | GETCHARINCTEST(c, eptr); |
5044 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5045 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
5046 | == prop_fail_result) |
5047 | RRETURN(MATCH_NOMATCH); |
5048 | } |
5049 | /* Control never gets here */ |
5050 | |
5051 | case PT_WORD: |
5052 | for (fi = min;; fi++) |
5053 | { |
5054 | int category; |
5055 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM62); |
5056 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5057 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5058 | if (eptr >= md->end_subject) |
5059 | { |
5060 | SCHECK_PARTIAL(); |
5061 | RRETURN(MATCH_NOMATCH); |
5062 | } |
5063 | GETCHARINCTEST(c, eptr); |
5064 | category = UCD_CATEGORY(c); |
5065 | if ((category == ucp_L || |
5066 | category == ucp_N || |
5067 | c == CHAR_UNDERSCORE) |
5068 | == prop_fail_result) |
5069 | RRETURN(MATCH_NOMATCH); |
5070 | } |
5071 | /* Control never gets here */ |
5072 | |
5073 | case PT_CLIST: |
5074 | for (fi = min;; fi++) |
5075 | { |
5076 | const pcre_uint32 *cp; |
5077 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM67); |
5078 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5079 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5080 | if (eptr >= md->end_subject) |
5081 | { |
5082 | SCHECK_PARTIAL(); |
5083 | RRETURN(MATCH_NOMATCH); |
5084 | } |
5085 | GETCHARINCTEST(c, eptr); |
5086 | cp = PRIV(ucd_caseless_sets) + prop_value; |
5087 | for (;;) |
5088 | { |
5089 | if (c < *cp) |
5090 | { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } } |
5091 | if (c == *cp++) |
5092 | { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; } |
5093 | } |
5094 | } |
5095 | /* Control never gets here */ |
5096 | |
5097 | case PT_UCNC: |
5098 | for (fi = min;; fi++) |
5099 | { |
5100 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM68); |
5101 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5102 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5103 | if (eptr >= md->end_subject) |
5104 | { |
5105 | SCHECK_PARTIAL(); |
5106 | RRETURN(MATCH_NOMATCH); |
5107 | } |
5108 | GETCHARINCTEST(c, eptr); |
5109 | if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
5110 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
5111 | c >= 0xe000) == prop_fail_result) |
5112 | RRETURN(MATCH_NOMATCH); |
5113 | } |
5114 | /* Control never gets here */ |
5115 | |
5116 | /* This should never occur */ |
5117 | default: |
5118 | RRETURN(PCRE_ERROR_INTERNAL); |
5119 | } |
5120 | } |
5121 | |
5122 | /* Match extended Unicode sequences. We will get here only if the |
5123 | support is in the binary; otherwise a compile-time error occurs. */ |
5124 | |
5125 | else if (ctype == OP_EXTUNI) |
5126 | { |
5127 | for (fi = min;; fi++) |
5128 | { |
5129 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM41); |
5130 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5131 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5132 | if (eptr >= md->end_subject) |
5133 | { |
5134 | SCHECK_PARTIAL(); |
5135 | RRETURN(MATCH_NOMATCH); |
5136 | } |
5137 | else |
5138 | { |
5139 | int lgb, rgb; |
5140 | GETCHARINCTEST(c, eptr); |
5141 | lgb = UCD_GRAPHBREAK(c); |
5142 | while (eptr < md->end_subject) |
5143 | { |
5144 | int len = 1; |
5145 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
5146 | rgb = UCD_GRAPHBREAK(c); |
5147 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
5148 | lgb = rgb; |
5149 | eptr += len; |
5150 | } |
5151 | } |
5152 | CHECK_PARTIAL(); |
5153 | } |
5154 | } |
5155 | else |
5156 | #endif /* SUPPORT_UCP */ |
5157 | |
5158 | #ifdef SUPPORT_UTF |
5159 | if (utf) |
5160 | { |
5161 | for (fi = min;; fi++) |
5162 | { |
5163 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM42); |
5164 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5165 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5166 | if (eptr >= md->end_subject) |
5167 | { |
5168 | SCHECK_PARTIAL(); |
5169 | RRETURN(MATCH_NOMATCH); |
5170 | } |
5171 | if (ctype == OP_ANY && IS_NEWLINE(eptr)) |
5172 | RRETURN(MATCH_NOMATCH); |
5173 | GETCHARINC(c, eptr); |
5174 | switch(ctype) |
5175 | { |
5176 | case OP_ANY: /* This is the non-NL case */ |
5177 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5178 | eptr >= md->end_subject && |
5179 | NLBLOCK->nltype == NLTYPE_FIXED && |
5180 | NLBLOCK->nllen == 2 && |
5181 | c == NLBLOCK->nl[0]) |
5182 | { |
5183 | md->hitend = TRUE; |
5184 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5185 | } |
5186 | break; |
5187 | |
5188 | case OP_ALLANY: |
5189 | case OP_ANYBYTE: |
5190 | break; |
5191 | |
5192 | case OP_ANYNL: |
5193 | switch(c) |
5194 | { |
5195 | default: RRETURN(MATCH_NOMATCH); |
5196 | case CHAR_CR: |
5197 | if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++; |
5198 | break; |
5199 | |
5200 | case CHAR_LF: |
5201 | break; |
5202 | |
5203 | case CHAR_VT: |
5204 | case CHAR_FF: |
5205 | case CHAR_NEL: |
5206 | #ifndef EBCDIC |
5207 | case 0x2028: |
5208 | case 0x2029: |
5209 | #endif /* Not EBCDIC */ |
5210 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
5211 | break; |
5212 | } |
5213 | break; |
5214 | |
5215 | case OP_NOT_HSPACE: |
5216 | switch(c) |
5217 | { |
5218 | HSPACE_CASES: RRETURN(MATCH_NOMATCH); |
5219 | default: break; |
5220 | } |
5221 | break; |
5222 | |
5223 | case OP_HSPACE: |
5224 | switch(c) |
5225 | { |
5226 | HSPACE_CASES: break; |
5227 | default: RRETURN(MATCH_NOMATCH); |
5228 | } |
5229 | break; |
5230 | |
5231 | case OP_NOT_VSPACE: |
5232 | switch(c) |
5233 | { |
5234 | VSPACE_CASES: RRETURN(MATCH_NOMATCH); |
5235 | default: break; |
5236 | } |
5237 | break; |
5238 | |
5239 | case OP_VSPACE: |
5240 | switch(c) |
5241 | { |
5242 | VSPACE_CASES: break; |
5243 | default: RRETURN(MATCH_NOMATCH); |
5244 | } |
5245 | break; |
5246 | |
5247 | case OP_NOT_DIGIT: |
5248 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) |
5249 | RRETURN(MATCH_NOMATCH); |
5250 | break; |
5251 | |
5252 | case OP_DIGIT: |
5253 | if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0) |
5254 | RRETURN(MATCH_NOMATCH); |
5255 | break; |
5256 | |
5257 | case OP_NOT_WHITESPACE: |
5258 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) |
5259 | RRETURN(MATCH_NOMATCH); |
5260 | break; |
5261 | |
5262 | case OP_WHITESPACE: |
5263 | if (c >= 256 || (md->ctypes[c] & ctype_space) == 0) |
5264 | RRETURN(MATCH_NOMATCH); |
5265 | break; |
5266 | |
5267 | case OP_NOT_WORDCHAR: |
5268 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) |
5269 | RRETURN(MATCH_NOMATCH); |
5270 | break; |
5271 | |
5272 | case OP_WORDCHAR: |
5273 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) |
5274 | RRETURN(MATCH_NOMATCH); |
5275 | break; |
5276 | |
5277 | default: |
5278 | RRETURN(PCRE_ERROR_INTERNAL); |
5279 | } |
5280 | } |
5281 | } |
5282 | else |
5283 | #endif |
5284 | /* Not UTF mode */ |
5285 | { |
5286 | for (fi = min;; fi++) |
5287 | { |
5288 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM43); |
5289 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5290 | if (fi >= max) RRETURN(MATCH_NOMATCH); |
5291 | if (eptr >= md->end_subject) |
5292 | { |
5293 | SCHECK_PARTIAL(); |
5294 | RRETURN(MATCH_NOMATCH); |
5295 | } |
5296 | if (ctype == OP_ANY && IS_NEWLINE(eptr)) |
5297 | RRETURN(MATCH_NOMATCH); |
5298 | c = *eptr++; |
5299 | switch(ctype) |
5300 | { |
5301 | case OP_ANY: /* This is the non-NL case */ |
5302 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5303 | eptr >= md->end_subject && |
5304 | NLBLOCK->nltype == NLTYPE_FIXED && |
5305 | NLBLOCK->nllen == 2 && |
5306 | c == NLBLOCK->nl[0]) |
5307 | { |
5308 | md->hitend = TRUE; |
5309 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5310 | } |
5311 | break; |
5312 | |
5313 | case OP_ALLANY: |
5314 | case OP_ANYBYTE: |
5315 | break; |
5316 | |
5317 | case OP_ANYNL: |
5318 | switch(c) |
5319 | { |
5320 | default: RRETURN(MATCH_NOMATCH); |
5321 | case CHAR_CR: |
5322 | if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++; |
5323 | break; |
5324 | |
5325 | case CHAR_LF: |
5326 | break; |
5327 | |
5328 | case CHAR_VT: |
5329 | case CHAR_FF: |
5330 | case CHAR_NEL: |
5331 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5332 | case 0x2028: |
5333 | case 0x2029: |
5334 | #endif |
5335 | if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH); |
5336 | break; |
5337 | } |
5338 | break; |
5339 | |
5340 | case OP_NOT_HSPACE: |
5341 | switch(c) |
5342 | { |
5343 | default: break; |
5344 | HSPACE_BYTE_CASES: |
5345 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5346 | HSPACE_MULTIBYTE_CASES: |
5347 | #endif |
5348 | RRETURN(MATCH_NOMATCH); |
5349 | } |
5350 | break; |
5351 | |
5352 | case OP_HSPACE: |
5353 | switch(c) |
5354 | { |
5355 | default: RRETURN(MATCH_NOMATCH); |
5356 | HSPACE_BYTE_CASES: |
5357 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5358 | HSPACE_MULTIBYTE_CASES: |
5359 | #endif |
5360 | break; |
5361 | } |
5362 | break; |
5363 | |
5364 | case OP_NOT_VSPACE: |
5365 | switch(c) |
5366 | { |
5367 | default: break; |
5368 | VSPACE_BYTE_CASES: |
5369 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5370 | VSPACE_MULTIBYTE_CASES: |
5371 | #endif |
5372 | RRETURN(MATCH_NOMATCH); |
5373 | } |
5374 | break; |
5375 | |
5376 | case OP_VSPACE: |
5377 | switch(c) |
5378 | { |
5379 | default: RRETURN(MATCH_NOMATCH); |
5380 | VSPACE_BYTE_CASES: |
5381 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
5382 | VSPACE_MULTIBYTE_CASES: |
5383 | #endif |
5384 | break; |
5385 | } |
5386 | break; |
5387 | |
5388 | case OP_NOT_DIGIT: |
5389 | if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH); |
5390 | break; |
5391 | |
5392 | case OP_DIGIT: |
5393 | if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH); |
5394 | break; |
5395 | |
5396 | case OP_NOT_WHITESPACE: |
5397 | if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH); |
5398 | break; |
5399 | |
5400 | case OP_WHITESPACE: |
5401 | if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH); |
5402 | break; |
5403 | |
5404 | case OP_NOT_WORDCHAR: |
5405 | if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH); |
5406 | break; |
5407 | |
5408 | case OP_WORDCHAR: |
5409 | if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH); |
5410 | break; |
5411 | |
5412 | default: |
5413 | RRETURN(PCRE_ERROR_INTERNAL); |
5414 | } |
5415 | } |
5416 | } |
5417 | /* Control never gets here */ |
5418 | } |
5419 | |
5420 | /* If maximizing, it is worth using inline code for speed, doing the type |
5421 | test once at the start (i.e. keep it out of the loop). Again, keep the |
5422 | UTF-8 and UCP stuff separate. */ |
5423 | |
5424 | else |
5425 | { |
5426 | pp = eptr; /* Remember where we started */ |
5427 | |
5428 | #ifdef SUPPORT_UCP |
5429 | if (prop_type >= 0) |
5430 | { |
5431 | switch(prop_type) |
5432 | { |
5433 | case PT_ANY: |
5434 | for (i = min; i < max; i++) |
5435 | { |
5436 | int len = 1; |
5437 | if (eptr >= md->end_subject) |
5438 | { |
5439 | SCHECK_PARTIAL(); |
5440 | break; |
5441 | } |
5442 | GETCHARLENTEST(c, eptr, len); |
5443 | if (prop_fail_result) break; |
5444 | eptr+= len; |
5445 | } |
5446 | break; |
5447 | |
5448 | case PT_LAMP: |
5449 | for (i = min; i < max; i++) |
5450 | { |
5451 | int chartype; |
5452 | int len = 1; |
5453 | if (eptr >= md->end_subject) |
5454 | { |
5455 | SCHECK_PARTIAL(); |
5456 | break; |
5457 | } |
5458 | GETCHARLENTEST(c, eptr, len); |
5459 | chartype = UCD_CHARTYPE(c); |
5460 | if ((chartype == ucp_Lu || |
5461 | chartype == ucp_Ll || |
5462 | chartype == ucp_Lt) == prop_fail_result) |
5463 | break; |
5464 | eptr+= len; |
5465 | } |
5466 | break; |
5467 | |
5468 | case PT_GC: |
5469 | for (i = min; i < max; i++) |
5470 | { |
5471 | int len = 1; |
5472 | if (eptr >= md->end_subject) |
5473 | { |
5474 | SCHECK_PARTIAL(); |
5475 | break; |
5476 | } |
5477 | GETCHARLENTEST(c, eptr, len); |
5478 | if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break; |
5479 | eptr+= len; |
5480 | } |
5481 | break; |
5482 | |
5483 | case PT_PC: |
5484 | for (i = min; i < max; i++) |
5485 | { |
5486 | int len = 1; |
5487 | if (eptr >= md->end_subject) |
5488 | { |
5489 | SCHECK_PARTIAL(); |
5490 | break; |
5491 | } |
5492 | GETCHARLENTEST(c, eptr, len); |
5493 | if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break; |
5494 | eptr+= len; |
5495 | } |
5496 | break; |
5497 | |
5498 | case PT_SC: |
5499 | for (i = min; i < max; i++) |
5500 | { |
5501 | int len = 1; |
5502 | if (eptr >= md->end_subject) |
5503 | { |
5504 | SCHECK_PARTIAL(); |
5505 | break; |
5506 | } |
5507 | GETCHARLENTEST(c, eptr, len); |
5508 | if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break; |
5509 | eptr+= len; |
5510 | } |
5511 | break; |
5512 | |
5513 | case PT_ALNUM: |
5514 | for (i = min; i < max; i++) |
5515 | { |
5516 | int category; |
5517 | int len = 1; |
5518 | if (eptr >= md->end_subject) |
5519 | { |
5520 | SCHECK_PARTIAL(); |
5521 | break; |
5522 | } |
5523 | GETCHARLENTEST(c, eptr, len); |
5524 | category = UCD_CATEGORY(c); |
5525 | if ((category == ucp_L || category == ucp_N) == prop_fail_result) |
5526 | break; |
5527 | eptr+= len; |
5528 | } |
5529 | break; |
5530 | |
5531 | case PT_SPACE: /* Perl space */ |
5532 | for (i = min; i < max; i++) |
5533 | { |
5534 | int len = 1; |
5535 | if (eptr >= md->end_subject) |
5536 | { |
5537 | SCHECK_PARTIAL(); |
5538 | break; |
5539 | } |
5540 | GETCHARLENTEST(c, eptr, len); |
5541 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5542 | c == CHAR_FF || c == CHAR_CR) |
5543 | == prop_fail_result) |
5544 | break; |
5545 | eptr+= len; |
5546 | } |
5547 | break; |
5548 | |
5549 | case PT_PXSPACE: /* POSIX space */ |
5550 | for (i = min; i < max; i++) |
5551 | { |
5552 | int len = 1; |
5553 | if (eptr >= md->end_subject) |
5554 | { |
5555 | SCHECK_PARTIAL(); |
5556 | break; |
5557 | } |
5558 | GETCHARLENTEST(c, eptr, len); |
5559 | if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL || |
5560 | c == CHAR_VT || c == CHAR_FF || c == CHAR_CR) |
5561 | == prop_fail_result) |
5562 | break; |
5563 | eptr+= len; |
5564 | } |
5565 | break; |
5566 | |
5567 | case PT_WORD: |
5568 | for (i = min; i < max; i++) |
5569 | { |
5570 | int category; |
5571 | int len = 1; |
5572 | if (eptr >= md->end_subject) |
5573 | { |
5574 | SCHECK_PARTIAL(); |
5575 | break; |
5576 | } |
5577 | GETCHARLENTEST(c, eptr, len); |
5578 | category = UCD_CATEGORY(c); |
5579 | if ((category == ucp_L || category == ucp_N || |
5580 | c == CHAR_UNDERSCORE) == prop_fail_result) |
5581 | break; |
5582 | eptr+= len; |
5583 | } |
5584 | break; |
5585 | |
5586 | case PT_CLIST: |
5587 | for (i = min; i < max; i++) |
5588 | { |
5589 | const pcre_uint32 *cp; |
5590 | int len = 1; |
5591 | if (eptr >= md->end_subject) |
5592 | { |
5593 | SCHECK_PARTIAL(); |
5594 | break; |
5595 | } |
5596 | GETCHARLENTEST(c, eptr, len); |
5597 | cp = PRIV(ucd_caseless_sets) + prop_value; |
5598 | for (;;) |
5599 | { |
5600 | if (c < *cp) |
5601 | { if (prop_fail_result) break; else goto GOT_MAX; } |
5602 | if (c == *cp++) |
5603 | { if (prop_fail_result) goto GOT_MAX; else break; } |
5604 | } |
5605 | eptr += len; |
5606 | } |
5607 | GOT_MAX: |
5608 | break; |
5609 | |
5610 | case PT_UCNC: |
5611 | for (i = min; i < max; i++) |
5612 | { |
5613 | int len = 1; |
5614 | if (eptr >= md->end_subject) |
5615 | { |
5616 | SCHECK_PARTIAL(); |
5617 | break; |
5618 | } |
5619 | GETCHARLENTEST(c, eptr, len); |
5620 | if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
5621 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
5622 | c >= 0xe000) == prop_fail_result) |
5623 | break; |
5624 | eptr += len; |
5625 | } |
5626 | break; |
5627 | |
5628 | default: |
5629 | RRETURN(PCRE_ERROR_INTERNAL); |
5630 | } |
5631 | |
5632 | /* eptr is now past the end of the maximum run */ |
5633 | |
5634 | if (possessive) continue; /* No backtracking */ |
5635 | for(;;) |
5636 | { |
5637 | if (eptr == pp) goto TAIL_RECURSE; |
5638 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM44); |
5639 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5640 | eptr--; |
5641 | if (utf) BACKCHAR(eptr); |
5642 | } |
5643 | } |
5644 | |
5645 | /* Match extended Unicode sequences. We will get here only if the |
5646 | support is in the binary; otherwise a compile-time error occurs. */ |
5647 | |
5648 | else if (ctype == OP_EXTUNI) |
5649 | { |
5650 | for (i = min; i < max; i++) |
5651 | { |
5652 | if (eptr >= md->end_subject) |
5653 | { |
5654 | SCHECK_PARTIAL(); |
5655 | break; |
5656 | } |
5657 | else |
5658 | { |
5659 | int lgb, rgb; |
5660 | GETCHARINCTEST(c, eptr); |
5661 | lgb = UCD_GRAPHBREAK(c); |
5662 | while (eptr < md->end_subject) |
5663 | { |
5664 | int len = 1; |
5665 | if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); } |
5666 | rgb = UCD_GRAPHBREAK(c); |
5667 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
5668 | lgb = rgb; |
5669 | eptr += len; |
5670 | } |
5671 | } |
5672 | CHECK_PARTIAL(); |
5673 | } |
5674 | |
5675 | /* eptr is now past the end of the maximum run */ |
5676 | |
5677 | if (possessive) continue; /* No backtracking */ |
5678 | for(;;) |
5679 | { |
5680 | if (eptr == pp) goto TAIL_RECURSE; |
5681 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM45); |
5682 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5683 | eptr--; |
5684 | for (;;) /* Move back over one extended */ |
5685 | { |
5686 | if (!utf) c = *eptr; else |
5687 | { |
5688 | BACKCHAR(eptr); |
5689 | GETCHAR(c, eptr); |
5690 | } |
5691 | if (UCD_CATEGORY(c) != ucp_M) break; |
5692 | eptr--; |
5693 | } |
5694 | } |
5695 | } |
5696 | |
5697 | else |
5698 | #endif /* SUPPORT_UCP */ |
5699 | |
5700 | #ifdef SUPPORT_UTF |
5701 | if (utf) |
5702 | { |
5703 | switch(ctype) |
5704 | { |
5705 | case OP_ANY: |
5706 | if (max < INT_MAX) |
5707 | { |
5708 | for (i = min; i < max; i++) |
5709 | { |
5710 | if (eptr >= md->end_subject) |
5711 | { |
5712 | SCHECK_PARTIAL(); |
5713 | break; |
5714 | } |
5715 | if (IS_NEWLINE(eptr)) break; |
5716 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5717 | eptr + 1 >= md->end_subject && |
5718 | NLBLOCK->nltype == NLTYPE_FIXED && |
5719 | NLBLOCK->nllen == 2 && |
5720 | RAWUCHAR(eptr) == NLBLOCK->nl[0]) |
5721 | { |
5722 | md->hitend = TRUE; |
5723 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5724 | } |
5725 | eptr++; |
5726 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5727 | } |
5728 | } |
5729 | |
5730 | /* Handle unlimited UTF-8 repeat */ |
5731 | |
5732 | else |
5733 | { |
5734 | for (i = min; i < max; i++) |
5735 | { |
5736 | if (eptr >= md->end_subject) |
5737 | { |
5738 | SCHECK_PARTIAL(); |
5739 | break; |
5740 | } |
5741 | if (IS_NEWLINE(eptr)) break; |
5742 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5743 | eptr + 1 >= md->end_subject && |
5744 | NLBLOCK->nltype == NLTYPE_FIXED && |
5745 | NLBLOCK->nllen == 2 && |
5746 | RAWUCHAR(eptr) == NLBLOCK->nl[0]) |
5747 | { |
5748 | md->hitend = TRUE; |
5749 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5750 | } |
5751 | eptr++; |
5752 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5753 | } |
5754 | } |
5755 | break; |
5756 | |
5757 | case OP_ALLANY: |
5758 | if (max < INT_MAX) |
5759 | { |
5760 | for (i = min; i < max; i++) |
5761 | { |
5762 | if (eptr >= md->end_subject) |
5763 | { |
5764 | SCHECK_PARTIAL(); |
5765 | break; |
5766 | } |
5767 | eptr++; |
5768 | ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++); |
5769 | } |
5770 | } |
5771 | else |
5772 | { |
5773 | eptr = md->end_subject; /* Unlimited UTF-8 repeat */ |
5774 | SCHECK_PARTIAL(); |
5775 | } |
5776 | break; |
5777 | |
5778 | /* The byte case is the same as non-UTF8 */ |
5779 | |
5780 | case OP_ANYBYTE: |
5781 | c = max - min; |
5782 | if (c > (unsigned int)(md->end_subject - eptr)) |
5783 | { |
5784 | eptr = md->end_subject; |
5785 | SCHECK_PARTIAL(); |
5786 | } |
5787 | else eptr += c; |
5788 | break; |
5789 | |
5790 | case OP_ANYNL: |
5791 | for (i = min; i < max; i++) |
5792 | { |
5793 | int len = 1; |
5794 | if (eptr >= md->end_subject) |
5795 | { |
5796 | SCHECK_PARTIAL(); |
5797 | break; |
5798 | } |
5799 | GETCHARLEN(c, eptr, len); |
5800 | if (c == CHAR_CR) |
5801 | { |
5802 | if (++eptr >= md->end_subject) break; |
5803 | if (RAWUCHAR(eptr) == CHAR_LF) eptr++; |
5804 | } |
5805 | else |
5806 | { |
5807 | if (c != CHAR_LF && |
5808 | (md->bsr_anycrlf || |
5809 | (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL |
5810 | #ifndef EBCDIC |
5811 | && c != 0x2028 && c != 0x2029 |
5812 | #endif /* Not EBCDIC */ |
5813 | ))) |
5814 | break; |
5815 | eptr += len; |
5816 | } |
5817 | } |
5818 | break; |
5819 | |
5820 | case OP_NOT_HSPACE: |
5821 | case OP_HSPACE: |
5822 | for (i = min; i < max; i++) |
5823 | { |
5824 | BOOL gotspace; |
5825 | int len = 1; |
5826 | if (eptr >= md->end_subject) |
5827 | { |
5828 | SCHECK_PARTIAL(); |
5829 | break; |
5830 | } |
5831 | GETCHARLEN(c, eptr, len); |
5832 | switch(c) |
5833 | { |
5834 | HSPACE_CASES: gotspace = TRUE; break; |
5835 | default: gotspace = FALSE; break; |
5836 | } |
5837 | if (gotspace == (ctype == OP_NOT_HSPACE)) break; |
5838 | eptr += len; |
5839 | } |
5840 | break; |
5841 | |
5842 | case OP_NOT_VSPACE: |
5843 | case OP_VSPACE: |
5844 | for (i = min; i < max; i++) |
5845 | { |
5846 | BOOL gotspace; |
5847 | int len = 1; |
5848 | if (eptr >= md->end_subject) |
5849 | { |
5850 | SCHECK_PARTIAL(); |
5851 | break; |
5852 | } |
5853 | GETCHARLEN(c, eptr, len); |
5854 | switch(c) |
5855 | { |
5856 | VSPACE_CASES: gotspace = TRUE; break; |
5857 | default: gotspace = FALSE; break; |
5858 | } |
5859 | if (gotspace == (ctype == OP_NOT_VSPACE)) break; |
5860 | eptr += len; |
5861 | } |
5862 | break; |
5863 | |
5864 | case OP_NOT_DIGIT: |
5865 | for (i = min; i < max; i++) |
5866 | { |
5867 | int len = 1; |
5868 | if (eptr >= md->end_subject) |
5869 | { |
5870 | SCHECK_PARTIAL(); |
5871 | break; |
5872 | } |
5873 | GETCHARLEN(c, eptr, len); |
5874 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break; |
5875 | eptr+= len; |
5876 | } |
5877 | break; |
5878 | |
5879 | case OP_DIGIT: |
5880 | for (i = min; i < max; i++) |
5881 | { |
5882 | int len = 1; |
5883 | if (eptr >= md->end_subject) |
5884 | { |
5885 | SCHECK_PARTIAL(); |
5886 | break; |
5887 | } |
5888 | GETCHARLEN(c, eptr, len); |
5889 | if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break; |
5890 | eptr+= len; |
5891 | } |
5892 | break; |
5893 | |
5894 | case OP_NOT_WHITESPACE: |
5895 | for (i = min; i < max; i++) |
5896 | { |
5897 | int len = 1; |
5898 | if (eptr >= md->end_subject) |
5899 | { |
5900 | SCHECK_PARTIAL(); |
5901 | break; |
5902 | } |
5903 | GETCHARLEN(c, eptr, len); |
5904 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break; |
5905 | eptr+= len; |
5906 | } |
5907 | break; |
5908 | |
5909 | case OP_WHITESPACE: |
5910 | for (i = min; i < max; i++) |
5911 | { |
5912 | int len = 1; |
5913 | if (eptr >= md->end_subject) |
5914 | { |
5915 | SCHECK_PARTIAL(); |
5916 | break; |
5917 | } |
5918 | GETCHARLEN(c, eptr, len); |
5919 | if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break; |
5920 | eptr+= len; |
5921 | } |
5922 | break; |
5923 | |
5924 | case OP_NOT_WORDCHAR: |
5925 | for (i = min; i < max; i++) |
5926 | { |
5927 | int len = 1; |
5928 | if (eptr >= md->end_subject) |
5929 | { |
5930 | SCHECK_PARTIAL(); |
5931 | break; |
5932 | } |
5933 | GETCHARLEN(c, eptr, len); |
5934 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break; |
5935 | eptr+= len; |
5936 | } |
5937 | break; |
5938 | |
5939 | case OP_WORDCHAR: |
5940 | for (i = min; i < max; i++) |
5941 | { |
5942 | int len = 1; |
5943 | if (eptr >= md->end_subject) |
5944 | { |
5945 | SCHECK_PARTIAL(); |
5946 | break; |
5947 | } |
5948 | GETCHARLEN(c, eptr, len); |
5949 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break; |
5950 | eptr+= len; |
5951 | } |
5952 | break; |
5953 | |
5954 | default: |
5955 | RRETURN(PCRE_ERROR_INTERNAL); |
5956 | } |
5957 | |
5958 | if (possessive) continue; /* No backtracking */ |
5959 | for(;;) |
5960 | { |
5961 | if (eptr == pp) goto TAIL_RECURSE; |
5962 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM46); |
5963 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
5964 | eptr--; |
5965 | BACKCHAR(eptr); |
5966 | if (ctype == OP_ANYNL && eptr > pp && RAWUCHAR(eptr) == CHAR_NL && |
5967 | RAWUCHAR(eptr - 1) == CHAR_CR) eptr--; |
5968 | } |
5969 | } |
5970 | else |
5971 | #endif /* SUPPORT_UTF */ |
5972 | /* Not UTF mode */ |
5973 | { |
5974 | switch(ctype) |
5975 | { |
5976 | case OP_ANY: |
5977 | for (i = min; i < max; i++) |
5978 | { |
5979 | if (eptr >= md->end_subject) |
5980 | { |
5981 | SCHECK_PARTIAL(); |
5982 | break; |
5983 | } |
5984 | if (IS_NEWLINE(eptr)) break; |
5985 | if (md->partial != 0 && /* Take care with CRLF partial */ |
5986 | eptr + 1 >= md->end_subject && |
5987 | NLBLOCK->nltype == NLTYPE_FIXED && |
5988 | NLBLOCK->nllen == 2 && |
5989 | *eptr == NLBLOCK->nl[0]) |
5990 | { |
5991 | md->hitend = TRUE; |
5992 | if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); |
5993 | } |
5994 | eptr++; |
5995 | } |
5996 | break; |
5997 | |
5998 | case OP_ALLANY: |
5999 | case OP_ANYBYTE: |
6000 | c = max - min; |
6001 | if (c > (unsigned int)(md->end_subject - eptr)) |
6002 | { |
6003 | eptr = md->end_subject; |
6004 | SCHECK_PARTIAL(); |
6005 | } |
6006 | else eptr += c; |
6007 | break; |
6008 | |
6009 | case OP_ANYNL: |
6010 | for (i = min; i < max; i++) |
6011 | { |
6012 | if (eptr >= md->end_subject) |
6013 | { |
6014 | SCHECK_PARTIAL(); |
6015 | break; |
6016 | } |
6017 | c = *eptr; |
6018 | if (c == CHAR_CR) |
6019 | { |
6020 | if (++eptr >= md->end_subject) break; |
6021 | if (*eptr == CHAR_LF) eptr++; |
6022 | } |
6023 | else |
6024 | { |
6025 | if (c != CHAR_LF && (md->bsr_anycrlf || |
6026 | (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL |
6027 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
6028 | && c != 0x2028 && c != 0x2029 |
6029 | #endif |
6030 | ))) break; |
6031 | eptr++; |
6032 | } |
6033 | } |
6034 | break; |
6035 | |
6036 | case OP_NOT_HSPACE: |
6037 | for (i = min; i < max; i++) |
6038 | { |
6039 | if (eptr >= md->end_subject) |
6040 | { |
6041 | SCHECK_PARTIAL(); |
6042 | break; |
6043 | } |
6044 | switch(*eptr) |
6045 | { |
6046 | default: eptr++; break; |
6047 | HSPACE_BYTE_CASES: |
6048 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
6049 | HSPACE_MULTIBYTE_CASES: |
6050 | #endif |
6051 | goto ENDLOOP00; |
6052 | } |
6053 | } |
6054 | ENDLOOP00: |
6055 | break; |
6056 | |
6057 | case OP_HSPACE: |
6058 | for (i = min; i < max; i++) |
6059 | { |
6060 | if (eptr >= md->end_subject) |
6061 | { |
6062 | SCHECK_PARTIAL(); |
6063 | break; |
6064 | } |
6065 | switch(*eptr) |
6066 | { |
6067 | default: goto ENDLOOP01; |
6068 | HSPACE_BYTE_CASES: |
6069 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
6070 | HSPACE_MULTIBYTE_CASES: |
6071 | #endif |
6072 | eptr++; break; |
6073 | } |
6074 | } |
6075 | ENDLOOP01: |
6076 | break; |
6077 | |
6078 | case OP_NOT_VSPACE: |
6079 | for (i = min; i < max; i++) |
6080 | { |
6081 | if (eptr >= md->end_subject) |
6082 | { |
6083 | SCHECK_PARTIAL(); |
6084 | break; |
6085 | } |
6086 | switch(*eptr) |
6087 | { |
6088 | default: eptr++; break; |
6089 | VSPACE_BYTE_CASES: |
6090 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
6091 | VSPACE_MULTIBYTE_CASES: |
6092 | #endif |
6093 | goto ENDLOOP02; |
6094 | } |
6095 | } |
6096 | ENDLOOP02: |
6097 | break; |
6098 | |
6099 | case OP_VSPACE: |
6100 | for (i = min; i < max; i++) |
6101 | { |
6102 | if (eptr >= md->end_subject) |
6103 | { |
6104 | SCHECK_PARTIAL(); |
6105 | break; |
6106 | } |
6107 | switch(*eptr) |
6108 | { |
6109 | default: goto ENDLOOP03; |
6110 | VSPACE_BYTE_CASES: |
6111 | #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32 |
6112 | VSPACE_MULTIBYTE_CASES: |
6113 | #endif |
6114 | eptr++; break; |
6115 | } |
6116 | } |
6117 | ENDLOOP03: |
6118 | break; |
6119 | |
6120 | case OP_NOT_DIGIT: |
6121 | for (i = min; i < max; i++) |
6122 | { |
6123 | if (eptr >= md->end_subject) |
6124 | { |
6125 | SCHECK_PARTIAL(); |
6126 | break; |
6127 | } |
6128 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) break; |
6129 | eptr++; |
6130 | } |
6131 | break; |
6132 | |
6133 | case OP_DIGIT: |
6134 | for (i = min; i < max; i++) |
6135 | { |
6136 | if (eptr >= md->end_subject) |
6137 | { |
6138 | SCHECK_PARTIAL(); |
6139 | break; |
6140 | } |
6141 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) break; |
6142 | eptr++; |
6143 | } |
6144 | break; |
6145 | |
6146 | case OP_NOT_WHITESPACE: |
6147 | for (i = min; i < max; i++) |
6148 | { |
6149 | if (eptr >= md->end_subject) |
6150 | { |
6151 | SCHECK_PARTIAL(); |
6152 | break; |
6153 | } |
6154 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break; |
6155 | eptr++; |
6156 | } |
6157 | break; |
6158 | |
6159 | case OP_WHITESPACE: |
6160 | for (i = min; i < max; i++) |
6161 | { |
6162 | if (eptr >= md->end_subject) |
6163 | { |
6164 | SCHECK_PARTIAL(); |
6165 | break; |
6166 | } |
6167 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break; |
6168 | eptr++; |
6169 | } |
6170 | break; |
6171 | |
6172 | case OP_NOT_WORDCHAR: |
6173 | for (i = min; i < max; i++) |
6174 | { |
6175 | if (eptr >= md->end_subject) |
6176 | { |
6177 | SCHECK_PARTIAL(); |
6178 | break; |
6179 | } |
6180 | if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break; |
6181 | eptr++; |
6182 | } |
6183 | break; |
6184 | |
6185 | case OP_WORDCHAR: |
6186 | for (i = min; i < max; i++) |
6187 | { |
6188 | if (eptr >= md->end_subject) |
6189 | { |
6190 | SCHECK_PARTIAL(); |
6191 | break; |
6192 | } |
6193 | if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break; |
6194 | eptr++; |
6195 | } |
6196 | break; |
6197 | |
6198 | default: |
6199 | RRETURN(PCRE_ERROR_INTERNAL); |
6200 | } |
6201 | |
6202 | if (possessive) continue; /* No backtracking */ |
6203 | for (;;) |
6204 | { |
6205 | if (eptr == pp) goto TAIL_RECURSE; |
6206 | RMATCH(eptr, ecode, offset_top, md, eptrb, RM47); |
6207 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); |
6208 | eptr--; |
6209 | if (ctype == OP_ANYNL && eptr > pp && *eptr == CHAR_LF && |
6210 | eptr[-1] == CHAR_CR) eptr--; |
6211 | } |
6212 | } |
6213 | |
6214 | /* Get here if we can't make it match with any permitted repetitions */ |
6215 | |
6216 | RRETURN(MATCH_NOMATCH); |
6217 | } |
6218 | /* Control never gets here */ |
6219 | |
6220 | /* There's been some horrible disaster. Arrival here can only mean there is |
6221 | something seriously wrong in the code above or the OP_xxx definitions. */ |
6222 | |
6223 | default: |
6224 | DPRINTF(("Unknown opcode %d\n", *ecode)); |
6225 | RRETURN(PCRE_ERROR_UNKNOWN_OPCODE); |
6226 | } |
6227 | |
6228 | /* Do not stick any code in here without much thought; it is assumed |
6229 | that "continue" in the code above comes out to here to repeat the main |
6230 | loop. */ |
6231 | |
6232 | } /* End of main loop */ |
6233 | /* Control never reaches here */ |
6234 | |
6235 | |
6236 | /* When compiling to use the heap rather than the stack for recursive calls to |
6237 | match(), the RRETURN() macro jumps here. The number that is saved in |
6238 | frame->Xwhere indicates which label we actually want to return to. */ |
6239 | |
6240 | #ifdef NO_RECURSE |
6241 | #define LBL(val) case val: goto L_RM##val; |
6242 | HEAP_RETURN: |
6243 | switch (frame->Xwhere) |
6244 | { |
6245 | LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8) |
6246 | LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17) |
6247 | LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33) |
6248 | LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52) |
6249 | LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64) |
6250 | LBL(65) LBL(66) |
6251 | #if defined SUPPORT_UTF || !defined COMPILE_PCRE8 |
6252 | LBL(21) |
6253 | #endif |
6254 | #ifdef SUPPORT_UTF |
6255 | LBL(16) LBL(18) LBL(20) |
6256 | LBL(22) LBL(23) LBL(28) LBL(30) |
6257 | LBL(32) LBL(34) LBL(42) LBL(46) |
6258 | #ifdef SUPPORT_UCP |
6259 | LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45) |
6260 | LBL(59) LBL(60) LBL(61) LBL(62) LBL(67) LBL(68) |
6261 | #endif /* SUPPORT_UCP */ |
6262 | #endif /* SUPPORT_UTF */ |
6263 | default: |
6264 | DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere)); |
6265 | return PCRE_ERROR_INTERNAL; |
6266 | } |
6267 | #undef LBL |
6268 | #endif /* NO_RECURSE */ |
6269 | } |
6270 | |
6271 | |
6272 | /*************************************************************************** |
6273 | **************************************************************************** |
6274 | RECURSION IN THE match() FUNCTION |
6275 | |
6276 | Undefine all the macros that were defined above to handle this. */ |
6277 | |
6278 | #ifdef NO_RECURSE |
6279 | #undef eptr |
6280 | #undef ecode |
6281 | #undef mstart |
6282 | #undef offset_top |
6283 | #undef eptrb |
6284 | #undef flags |
6285 | |
6286 | #undef callpat |
6287 | #undef charptr |
6288 | #undef data |
6289 | #undef next |
6290 | #undef pp |
6291 | #undef prev |
6292 | #undef saved_eptr |
6293 | |
6294 | #undef new_recursive |
6295 | |
6296 | #undef cur_is_word |
6297 | #undef condition |
6298 | #undef prev_is_word |
6299 | |
6300 | #undef ctype |
6301 | #undef length |
6302 | #undef max |
6303 | #undef min |
6304 | #undef number |
6305 | #undef offset |
6306 | #undef op |
6307 | #undef save_capture_last |
6308 | #undef save_offset1 |
6309 | #undef save_offset2 |
6310 | #undef save_offset3 |
6311 | #undef stacksave |
6312 | |
6313 | #undef newptrb |
6314 | |
6315 | #endif |
6316 | |
6317 | /* These two are defined as macros in both cases */ |
6318 | |
6319 | #undef fc |
6320 | #undef fi |
6321 | |
6322 | /*************************************************************************** |
6323 | ***************************************************************************/ |
6324 | |
6325 | |
6326 | #ifdef NO_RECURSE |
6327 | /************************************************* |
6328 | * Release allocated heap frames * |
6329 | *************************************************/ |
6330 | |
6331 | /* This function releases all the allocated frames. The base frame is on the |
6332 | machine stack, and so must not be freed. |
6333 | |
6334 | Argument: the address of the base frame |
6335 | Returns: nothing |
6336 | */ |
6337 | |
6338 | static void |
6339 | release_match_heapframes (heapframe *frame_base) |
6340 | { |
6341 | heapframe *nextframe = frame_base->Xnextframe; |
6342 | while (nextframe != NULL) |
6343 | { |
6344 | heapframe *oldframe = nextframe; |
6345 | nextframe = nextframe->Xnextframe; |
6346 | (PUBL(stack_free))(oldframe); |
6347 | } |
6348 | } |
6349 | #endif |
6350 | |
6351 | |
6352 | /************************************************* |
6353 | * Execute a Regular Expression * |
6354 | *************************************************/ |
6355 | |
6356 | /* This function applies a compiled re to a subject string and picks out |
6357 | portions of the string if it matches. Two elements in the vector are set for |
6358 | each substring: the offsets to the start and end of the substring. |
6359 | |
6360 | Arguments: |
6361 | argument_re points to the compiled expression |
6362 | extra_data points to extra data or is NULL |
6363 | subject points to the subject string |
6364 | length length of subject string (may contain binary zeros) |
6365 | start_offset where to start in the subject string |
6366 | options option bits |
6367 | offsets points to a vector of ints to be filled in with offsets |
6368 | offsetcount the number of elements in the vector |
6369 | |
6370 | Returns: > 0 => success; value is the number of elements filled in |
6371 | = 0 => success, but offsets is not big enough |
6372 | -1 => failed to match |
6373 | < -1 => some kind of unexpected problem |
6374 | */ |
6375 | |
6376 | #if defined COMPILE_PCRE8 |
6377 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6378 | pcre_exec(const pcre *argument_re, const pcre_extra *extra_data, |
6379 | PCRE_SPTR subject, int length, int start_offset, int options, int *offsets, |
6380 | int offsetcount) |
6381 | #elif defined COMPILE_PCRE16 |
6382 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6383 | pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data, |
6384 | PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets, |
6385 | int offsetcount) |
6386 | #elif defined COMPILE_PCRE32 |
6387 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
6388 | pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data, |
6389 | PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets, |
6390 | int offsetcount) |
6391 | #endif |
6392 | { |
6393 | int rc, ocount, arg_offset_max; |
6394 | int newline; |
6395 | BOOL using_temporary_offsets = FALSE; |
6396 | BOOL anchored; |
6397 | BOOL startline; |
6398 | BOOL firstline; |
6399 | BOOL utf; |
6400 | BOOL has_first_char = FALSE; |
6401 | BOOL has_req_char = FALSE; |
6402 | pcre_uchar first_char = 0; |
6403 | pcre_uchar first_char2 = 0; |
6404 | pcre_uchar req_char = 0; |
6405 | pcre_uchar req_char2 = 0; |
6406 | match_data match_block; |
6407 | match_data *md = &match_block; |
6408 | const pcre_uint8 *tables; |
6409 | const pcre_uint8 *start_bits = NULL; |
6410 | PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset; |
6411 | PCRE_PUCHAR end_subject; |
6412 | PCRE_PUCHAR start_partial = NULL; |
6413 | PCRE_PUCHAR match_partial; |
6414 | PCRE_PUCHAR req_char_ptr = start_match - 1; |
6415 | |
6416 | const pcre_study_data *study; |
6417 | const REAL_PCRE *re = (const REAL_PCRE *)argument_re; |
6418 | |
6419 | #ifdef NO_RECURSE |
6420 | heapframe frame_zero; |
6421 | frame_zero.Xprevframe = NULL; /* Marks the top level */ |
6422 | frame_zero.Xnextframe = NULL; /* None are allocated yet */ |
6423 | md->match_frames_base = &frame_zero; |
6424 | #endif |
6425 | |
6426 | /* Check for the special magic call that measures the size of the stack used |
6427 | per recursive call of match(). Without the funny casting for sizeof, a Windows |
6428 | compiler gave this error: "unary minus operator applied to unsigned type, |
6429 | result still unsigned". Hopefully the cast fixes that. */ |
6430 | |
6431 | if (re == NULL && extra_data == NULL && subject == NULL && length == -999 && |
6432 | start_offset == -999) |
6433 | #ifdef NO_RECURSE |
6434 | return -((int)sizeof(heapframe)); |
6435 | #else |
6436 | return match(NULL, NULL, NULL, 0, NULL, NULL, 0); |
6437 | #endif |
6438 | |
6439 | /* Plausibility checks */ |
6440 | |
6441 | if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION; |
6442 | if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0)) |
6443 | return PCRE_ERROR_NULL; |
6444 | if (offsetcount < 0) return PCRE_ERROR_BADCOUNT; |
6445 | if (length < 0) return PCRE_ERROR_BADLENGTH; |
6446 | if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET; |
6447 | |
6448 | /* Check that the first field in the block is the magic number. If it is not, |
6449 | return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to |
6450 | REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which |
6451 | means that the pattern is likely compiled with different endianness. */ |
6452 | |
6453 | if (re->magic_number != MAGIC_NUMBER) |
6454 | return re->magic_number == REVERSED_MAGIC_NUMBER? |
6455 | PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC; |
6456 | if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE; |
6457 | |
6458 | /* These two settings are used in the code for checking a UTF-8 string that |
6459 | follows immediately afterwards. Other values in the md block are used only |
6460 | during "normal" pcre_exec() processing, not when the JIT support is in use, |
6461 | so they are set up later. */ |
6462 | |
6463 | /* PCRE_UTF16 has the same value as PCRE_UTF8. */ |
6464 | utf = md->utf = (re->options & PCRE_UTF8) != 0; |
6465 | md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 : |
6466 | ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0; |
6467 | |
6468 | /* Check a UTF-8 string if required. Pass back the character offset and error |
6469 | code for an invalid string if a results vector is available. */ |
6470 | |
6471 | #ifdef SUPPORT_UTF |
6472 | if (utf && (options & PCRE_NO_UTF8_CHECK) == 0) |
6473 | { |
6474 | int erroroffset; |
6475 | int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset); |
6476 | if (errorcode != 0) |
6477 | { |
6478 | if (offsetcount >= 2) |
6479 | { |
6480 | offsets[0] = erroroffset; |
6481 | offsets[1] = errorcode; |
6482 | } |
6483 | #if defined COMPILE_PCRE8 |
6484 | return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)? |
6485 | PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8; |
6486 | #elif defined COMPILE_PCRE16 |
6487 | return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)? |
6488 | PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16; |
6489 | #elif defined COMPILE_PCRE32 |
6490 | return PCRE_ERROR_BADUTF32; |
6491 | #endif |
6492 | } |
6493 | #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16 |
6494 | /* Check that a start_offset points to the start of a UTF character. */ |
6495 | if (start_offset > 0 && start_offset < length && |
6496 | NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset])) |
6497 | return PCRE_ERROR_BADUTF8_OFFSET; |
6498 | #endif |
6499 | } |
6500 | #endif |
6501 | |
6502 | /* If the pattern was successfully studied with JIT support, run the JIT |
6503 | executable instead of the rest of this function. Most options must be set at |
6504 | compile time for the JIT code to be usable. Fallback to the normal code path if |
6505 | an unsupported flag is set. */ |
6506 | |
6507 | #ifdef SUPPORT_JIT |
6508 | if (extra_data != NULL |
6509 | && (extra_data->flags & (PCRE_EXTRA_EXECUTABLE_JIT | |
6510 | PCRE_EXTRA_TABLES)) == PCRE_EXTRA_EXECUTABLE_JIT |
6511 | && extra_data->executable_jit != NULL |
6512 | && (options & ~PUBLIC_JIT_EXEC_OPTIONS) == 0) |
6513 | { |
6514 | rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length, |
6515 | start_offset, options, offsets, offsetcount); |
6516 | |
6517 | /* PCRE_ERROR_NULL means that the selected normal or partial matching |
6518 | mode is not compiled. In this case we simply fallback to interpreter. */ |
6519 | |
6520 | if (rc != PCRE_ERROR_JIT_BADOPTION) return rc; |
6521 | } |
6522 | #endif |
6523 | |
6524 | /* Carry on with non-JIT matching. This information is for finding all the |
6525 | numbers associated with a given name, for condition testing. */ |
6526 | |
6527 | md->name_table = (pcre_uchar *)re + re->name_table_offset; |
6528 | md->name_count = re->name_count; |
6529 | md->name_entry_size = re->name_entry_size; |
6530 | |
6531 | /* Fish out the optional data from the extra_data structure, first setting |
6532 | the default values. */ |
6533 | |
6534 | study = NULL; |
6535 | md->match_limit = MATCH_LIMIT; |
6536 | md->match_limit_recursion = MATCH_LIMIT_RECURSION; |
6537 | md->callout_data = NULL; |
6538 | |
6539 | /* The table pointer is always in native byte order. */ |
6540 | |
6541 | tables = re->tables; |
6542 | |
6543 | /* The two limit values override the defaults, whatever their value. */ |
6544 | |
6545 | if (extra_data != NULL) |
6546 | { |
6547 | register unsigned int flags = extra_data->flags; |
6548 | if ((flags & PCRE_EXTRA_STUDY_DATA) != 0) |
6549 | study = (const pcre_study_data *)extra_data->study_data; |
6550 | if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0) |
6551 | md->match_limit = extra_data->match_limit; |
6552 | if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0) |
6553 | md->match_limit_recursion = extra_data->match_limit_recursion; |
6554 | if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0) |
6555 | md->callout_data = extra_data->callout_data; |
6556 | if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables; |
6557 | } |
6558 | |
6559 | /* Limits in the regex override only if they are smaller. */ |
6560 | |
6561 | if ((re->flags & PCRE_MLSET) != 0 && re->limit_match < md->match_limit) |
6562 | md->match_limit = re->limit_match; |
6563 | |
6564 | if ((re->flags & PCRE_RLSET) != 0 && |
6565 | re->limit_recursion < md->match_limit_recursion) |
6566 | md->match_limit_recursion = re->limit_recursion; |
6567 | |
6568 | /* If the exec call supplied NULL for tables, use the inbuilt ones. This |
6569 | is a feature that makes it possible to save compiled regex and re-use them |
6570 | in other programs later. */ |
6571 | |
6572 | if (tables == NULL) tables = PRIV(default_tables); |
6573 | |
6574 | /* Set up other data */ |
6575 | |
6576 | anchored = ((re->options | options) & PCRE_ANCHORED) != 0; |
6577 | startline = (re->flags & PCRE_STARTLINE) != 0; |
6578 | firstline = (re->options & PCRE_FIRSTLINE) != 0; |
6579 | |
6580 | /* The code starts after the real_pcre block and the capture name table. */ |
6581 | |
6582 | md->start_code = (const pcre_uchar *)re + re->name_table_offset + |
6583 | re->name_count * re->name_entry_size; |
6584 | |
6585 | md->start_subject = (PCRE_PUCHAR)subject; |
6586 | md->start_offset = start_offset; |
6587 | md->end_subject = md->start_subject + length; |
6588 | end_subject = md->end_subject; |
6589 | |
6590 | md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0; |
6591 | md->use_ucp = (re->options & PCRE_UCP) != 0; |
6592 | md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0; |
6593 | md->ignore_skip_arg = 0; |
6594 | |
6595 | /* Some options are unpacked into BOOL variables in the hope that testing |
6596 | them will be faster than individual option bits. */ |
6597 | |
6598 | md->notbol = (options & PCRE_NOTBOL) != 0; |
6599 | md->noteol = (options & PCRE_NOTEOL) != 0; |
6600 | md->notempty = (options & PCRE_NOTEMPTY) != 0; |
6601 | md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0; |
6602 | |
6603 | md->hitend = FALSE; |
6604 | md->mark = md->nomatch_mark = NULL; /* In case never set */ |
6605 | |
6606 | md->recursive = NULL; /* No recursion at top level */ |
6607 | md->hasthen = (re->flags & PCRE_HASTHEN) != 0; |
6608 | |
6609 | md->lcc = tables + lcc_offset; |
6610 | md->fcc = tables + fcc_offset; |
6611 | md->ctypes = tables + ctypes_offset; |
6612 | |
6613 | /* Handle different \R options. */ |
6614 | |
6615 | switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) |
6616 | { |
6617 | case 0: |
6618 | if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0) |
6619 | md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0; |
6620 | else |
6621 | #ifdef BSR_ANYCRLF |
6622 | md->bsr_anycrlf = TRUE; |
6623 | #else |
6624 | md->bsr_anycrlf = FALSE; |
6625 | #endif |
6626 | break; |
6627 | |
6628 | case PCRE_BSR_ANYCRLF: |
6629 | md->bsr_anycrlf = TRUE; |
6630 | break; |
6631 | |
6632 | case PCRE_BSR_UNICODE: |
6633 | md->bsr_anycrlf = FALSE; |
6634 | break; |
6635 | |
6636 | default: return PCRE_ERROR_BADNEWLINE; |
6637 | } |
6638 | |
6639 | /* Handle different types of newline. The three bits give eight cases. If |
6640 | nothing is set at run time, whatever was used at compile time applies. */ |
6641 | |
6642 | switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options : |
6643 | (pcre_uint32)options) & PCRE_NEWLINE_BITS) |
6644 | { |
6645 | case 0: newline = NEWLINE; break; /* Compile-time default */ |
6646 | case PCRE_NEWLINE_CR: newline = CHAR_CR; break; |
6647 | case PCRE_NEWLINE_LF: newline = CHAR_NL; break; |
6648 | case PCRE_NEWLINE_CR+ |
6649 | PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break; |
6650 | case PCRE_NEWLINE_ANY: newline = -1; break; |
6651 | case PCRE_NEWLINE_ANYCRLF: newline = -2; break; |
6652 | default: return PCRE_ERROR_BADNEWLINE; |
6653 | } |
6654 | |
6655 | if (newline == -2) |
6656 | { |
6657 | md->nltype = NLTYPE_ANYCRLF; |
6658 | } |
6659 | else if (newline < 0) |
6660 | { |
6661 | md->nltype = NLTYPE_ANY; |
6662 | } |
6663 | else |
6664 | { |
6665 | md->nltype = NLTYPE_FIXED; |
6666 | if (newline > 255) |
6667 | { |
6668 | md->nllen = 2; |
6669 | md->nl[0] = (newline >> 8) & 255; |
6670 | md->nl[1] = newline & 255; |
6671 | } |
6672 | else |
6673 | { |
6674 | md->nllen = 1; |
6675 | md->nl[0] = newline; |
6676 | } |
6677 | } |
6678 | |
6679 | /* Partial matching was originally supported only for a restricted set of |
6680 | regexes; from release 8.00 there are no restrictions, but the bits are still |
6681 | defined (though never set). So there's no harm in leaving this code. */ |
6682 | |
6683 | if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0) |
6684 | return PCRE_ERROR_BADPARTIAL; |
6685 | |
6686 | /* If the expression has got more back references than the offsets supplied can |
6687 | hold, we get a temporary chunk of working store to use during the matching. |
6688 | Otherwise, we can use the vector supplied, rounding down its size to a multiple |
6689 | of 3. */ |
6690 | |
6691 | ocount = offsetcount - (offsetcount % 3); |
6692 | arg_offset_max = (2*ocount)/3; |
6693 | |
6694 | if (re->top_backref > 0 && re->top_backref >= ocount/3) |
6695 | { |
6696 | ocount = re->top_backref * 3 + 3; |
6697 | md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int)); |
6698 | if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY; |
6699 | using_temporary_offsets = TRUE; |
6700 | DPRINTF(("Got memory to hold back references\n")); |
6701 | } |
6702 | else md->offset_vector = offsets; |
6703 | md->offset_end = ocount; |
6704 | md->offset_max = (2*ocount)/3; |
6705 | md->capture_last = 0; |
6706 | |
6707 | /* Reset the working variable associated with each extraction. These should |
6708 | never be used unless previously set, but they get saved and restored, and so we |
6709 | initialize them to avoid reading uninitialized locations. Also, unset the |
6710 | offsets for the matched string. This is really just for tidiness with callouts, |
6711 | in case they inspect these fields. */ |
6712 | |
6713 | if (md->offset_vector != NULL) |
6714 | { |
6715 | register int *iptr = md->offset_vector + ocount; |
6716 | register int *iend = iptr - re->top_bracket; |
6717 | if (iend < md->offset_vector + 2) iend = md->offset_vector + 2; |
6718 | while (--iptr >= iend) *iptr = -1; |
6719 | md->offset_vector[0] = md->offset_vector[1] = -1; |
6720 | } |
6721 | |
6722 | /* Set up the first character to match, if available. The first_char value is |
6723 | never set for an anchored regular expression, but the anchoring may be forced |
6724 | at run time, so we have to test for anchoring. The first char may be unset for |
6725 | an unanchored pattern, of course. If there's no first char and the pattern was |
6726 | studied, there may be a bitmap of possible first characters. */ |
6727 | |
6728 | if (!anchored) |
6729 | { |
6730 | if ((re->flags & PCRE_FIRSTSET) != 0) |
6731 | { |
6732 | has_first_char = TRUE; |
6733 | first_char = first_char2 = (pcre_uchar)(re->first_char); |
6734 | if ((re->flags & PCRE_FCH_CASELESS) != 0) |
6735 | { |
6736 | first_char2 = TABLE_GET(first_char, md->fcc, first_char); |
6737 | #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8) |
6738 | if (utf && first_char > 127) |
6739 | first_char2 = UCD_OTHERCASE(first_char); |
6740 | #endif |
6741 | } |
6742 | } |
6743 | else |
6744 | if (!startline && study != NULL && |
6745 | (study->flags & PCRE_STUDY_MAPPED) != 0) |
6746 | start_bits = study->start_bits; |
6747 | } |
6748 | |
6749 | /* For anchored or unanchored matches, there may be a "last known required |
6750 | character" set. */ |
6751 | |
6752 | if ((re->flags & PCRE_REQCHSET) != 0) |
6753 | { |
6754 | has_req_char = TRUE; |