Commit | Line | Data |
---|---|---|
92e772ff | 1 | /* $Cambridge: exim/src/src/pcre/pcre_exec.c,v 1.2 2005/08/08 10:22:14 ph10 Exp $ */ |
8ac170f3 PH |
2 | |
3 | /************************************************* | |
4 | * Perl-Compatible Regular Expressions * | |
5 | *************************************************/ | |
6 | ||
7 | /* PCRE is a library of functions to support regular expressions whose syntax | |
8 | and semantics are as close as possible to those of the Perl 5 language. | |
9 | ||
10 | Written by Philip Hazel | |
11 | Copyright (c) 1997-2005 University of Cambridge | |
12 | ||
13 | ----------------------------------------------------------------------------- | |
14 | Redistribution and use in source and binary forms, with or without | |
15 | modification, are permitted provided that the following conditions are met: | |
16 | ||
17 | * Redistributions of source code must retain the above copyright notice, | |
18 | this list of conditions and the following disclaimer. | |
19 | ||
20 | * Redistributions in binary form must reproduce the above copyright | |
21 | notice, this list of conditions and the following disclaimer in the | |
22 | documentation and/or other materials provided with the distribution. | |
23 | ||
24 | * Neither the name of the University of Cambridge nor the names of its | |
25 | contributors may be used to endorse or promote products derived from | |
26 | this software without specific prior written permission. | |
27 | ||
28 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" | |
29 | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
30 | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
31 | ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE | |
32 | LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | |
33 | CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | |
34 | SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS | |
35 | INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN | |
36 | CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) | |
37 | ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | |
38 | POSSIBILITY OF SUCH DAMAGE. | |
39 | ----------------------------------------------------------------------------- | |
40 | */ | |
41 | ||
42 | ||
43 | /* This module contains pcre_exec(), the externally visible function that does | |
44 | pattern matching using an NFA algorithm, trying to mimic Perl as closely as | |
45 | possible. There are also some static supporting functions. */ | |
46 | ||
47 | ||
48 | #include "pcre_internal.h" | |
49 | ||
50 | ||
51 | /* Structure for building a chain of data that actually lives on the | |
52 | stack, for holding the values of the subject pointer at the start of each | |
53 | subpattern, so as to detect when an empty string has been matched by a | |
54 | subpattern - to break infinite loops. When NO_RECURSE is set, these blocks | |
55 | are on the heap, not on the stack. */ | |
56 | ||
57 | typedef struct eptrblock { | |
58 | struct eptrblock *epb_prev; | |
59 | const uschar *epb_saved_eptr; | |
60 | } eptrblock; | |
61 | ||
62 | /* Flag bits for the match() function */ | |
63 | ||
64 | #define match_condassert 0x01 /* Called to check a condition assertion */ | |
65 | #define match_isgroup 0x02 /* Set if start of bracketed group */ | |
66 | ||
67 | /* Non-error returns from the match() function. Error returns are externally | |
68 | defined PCRE_ERROR_xxx codes, which are all negative. */ | |
69 | ||
70 | #define MATCH_MATCH 1 | |
71 | #define MATCH_NOMATCH 0 | |
72 | ||
73 | /* Maximum number of ints of offset to save on the stack for recursive calls. | |
74 | If the offset vector is bigger, malloc is used. This should be a multiple of 3, | |
75 | because the offset vector is always a multiple of 3 long. */ | |
76 | ||
77 | #define REC_STACK_SAVE_MAX 30 | |
78 | ||
79 | /* Min and max values for the common repeats; for the maxima, 0 => infinity */ | |
80 | ||
81 | static const char rep_min[] = { 0, 0, 1, 1, 0, 0 }; | |
82 | static const char rep_max[] = { 0, 0, 0, 0, 1, 1 }; | |
83 | ||
84 | ||
85 | ||
86 | #ifdef DEBUG | |
87 | /************************************************* | |
88 | * Debugging function to print chars * | |
89 | *************************************************/ | |
90 | ||
91 | /* Print a sequence of chars in printable format, stopping at the end of the | |
92 | subject if the requested. | |
93 | ||
94 | Arguments: | |
95 | p points to characters | |
96 | length number to print | |
97 | is_subject TRUE if printing from within md->start_subject | |
98 | md pointer to matching data block, if is_subject is TRUE | |
99 | ||
100 | Returns: nothing | |
101 | */ | |
102 | ||
103 | static void | |
104 | pchars(const uschar *p, int length, BOOL is_subject, match_data *md) | |
105 | { | |
106 | int c; | |
107 | if (is_subject && length > md->end_subject - p) length = md->end_subject - p; | |
108 | while (length-- > 0) | |
109 | if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c); | |
110 | } | |
111 | #endif | |
112 | ||
113 | ||
114 | ||
115 | /************************************************* | |
116 | * Match a back-reference * | |
117 | *************************************************/ | |
118 | ||
119 | /* If a back reference hasn't been set, the length that is passed is greater | |
120 | than the number of characters left in the string, so the match fails. | |
121 | ||
122 | Arguments: | |
123 | offset index into the offset vector | |
124 | eptr points into the subject | |
125 | length length to be matched | |
126 | md points to match data block | |
127 | ims the ims flags | |
128 | ||
129 | Returns: TRUE if matched | |
130 | */ | |
131 | ||
132 | static BOOL | |
133 | match_ref(int offset, register const uschar *eptr, int length, match_data *md, | |
134 | unsigned long int ims) | |
135 | { | |
136 | const uschar *p = md->start_subject + md->offset_vector[offset]; | |
137 | ||
138 | #ifdef DEBUG | |
139 | if (eptr >= md->end_subject) | |
140 | printf("matching subject <null>"); | |
141 | else | |
142 | { | |
143 | printf("matching subject "); | |
144 | pchars(eptr, length, TRUE, md); | |
145 | } | |
146 | printf(" against backref "); | |
147 | pchars(p, length, FALSE, md); | |
148 | printf("\n"); | |
149 | #endif | |
150 | ||
151 | /* Always fail if not enough characters left */ | |
152 | ||
153 | if (length > md->end_subject - eptr) return FALSE; | |
154 | ||
155 | /* Separate the caselesss case for speed */ | |
156 | ||
157 | if ((ims & PCRE_CASELESS) != 0) | |
158 | { | |
159 | while (length-- > 0) | |
160 | if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; | |
161 | } | |
162 | else | |
163 | { while (length-- > 0) if (*p++ != *eptr++) return FALSE; } | |
164 | ||
165 | return TRUE; | |
166 | } | |
167 | ||
168 | ||
169 | ||
170 | /*************************************************************************** | |
171 | **************************************************************************** | |
172 | RECURSION IN THE match() FUNCTION | |
173 | ||
174 | The match() function is highly recursive. Some regular expressions can cause | |
175 | it to recurse thousands of times. I was writing for Unix, so I just let it | |
176 | call itself recursively. This uses the stack for saving everything that has | |
177 | to be saved for a recursive call. On Unix, the stack can be large, and this | |
178 | works fine. | |
179 | ||
180 | It turns out that on non-Unix systems there are problems with programs that | |
181 | use a lot of stack. (This despite the fact that every last chip has oodles | |
182 | of memory these days, and techniques for extending the stack have been known | |
183 | for decades.) So.... | |
184 | ||
185 | There is a fudge, triggered by defining NO_RECURSE, which avoids recursive | |
186 | calls by keeping local variables that need to be preserved in blocks of memory | |
187 | obtained from malloc instead instead of on the stack. Macros are used to | |
188 | achieve this so that the actual code doesn't look very different to what it | |
189 | always used to. | |
190 | **************************************************************************** | |
191 | ***************************************************************************/ | |
192 | ||
193 | ||
194 | /* These versions of the macros use the stack, as normal */ | |
195 | ||
196 | #ifndef NO_RECURSE | |
197 | #define REGISTER register | |
198 | #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg) | |
199 | #define RRETURN(ra) return ra | |
200 | #else | |
201 | ||
202 | ||
203 | /* These versions of the macros manage a private stack on the heap. Note | |
204 | that the rd argument of RMATCH isn't actually used. It's the md argument of | |
205 | match(), which never changes. */ | |
206 | ||
207 | #define REGISTER | |
208 | ||
209 | #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\ | |
210 | {\ | |
211 | heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\ | |
212 | if (setjmp(frame->Xwhere) == 0)\ | |
213 | {\ | |
214 | newframe->Xeptr = ra;\ | |
215 | newframe->Xecode = rb;\ | |
216 | newframe->Xoffset_top = rc;\ | |
217 | newframe->Xims = re;\ | |
218 | newframe->Xeptrb = rf;\ | |
219 | newframe->Xflags = rg;\ | |
220 | newframe->Xprevframe = frame;\ | |
221 | frame = newframe;\ | |
222 | DPRINTF(("restarting from line %d\n", __LINE__));\ | |
223 | goto HEAP_RECURSE;\ | |
224 | }\ | |
225 | else\ | |
226 | {\ | |
227 | DPRINTF(("longjumped back to line %d\n", __LINE__));\ | |
228 | frame = md->thisframe;\ | |
229 | rx = frame->Xresult;\ | |
230 | }\ | |
231 | } | |
232 | ||
233 | #define RRETURN(ra)\ | |
234 | {\ | |
235 | heapframe *newframe = frame;\ | |
236 | frame = newframe->Xprevframe;\ | |
237 | (pcre_stack_free)(newframe);\ | |
238 | if (frame != NULL)\ | |
239 | {\ | |
240 | frame->Xresult = ra;\ | |
241 | md->thisframe = frame;\ | |
242 | longjmp(frame->Xwhere, 1);\ | |
243 | }\ | |
244 | return ra;\ | |
245 | } | |
246 | ||
247 | ||
248 | /* Structure for remembering the local variables in a private frame */ | |
249 | ||
250 | typedef struct heapframe { | |
251 | struct heapframe *Xprevframe; | |
252 | ||
253 | /* Function arguments that may change */ | |
254 | ||
255 | const uschar *Xeptr; | |
256 | const uschar *Xecode; | |
257 | int Xoffset_top; | |
258 | long int Xims; | |
259 | eptrblock *Xeptrb; | |
260 | int Xflags; | |
261 | ||
262 | /* Function local variables */ | |
263 | ||
264 | const uschar *Xcallpat; | |
265 | const uschar *Xcharptr; | |
266 | const uschar *Xdata; | |
267 | const uschar *Xnext; | |
268 | const uschar *Xpp; | |
269 | const uschar *Xprev; | |
270 | const uschar *Xsaved_eptr; | |
271 | ||
272 | recursion_info Xnew_recursive; | |
273 | ||
274 | BOOL Xcur_is_word; | |
275 | BOOL Xcondition; | |
276 | BOOL Xminimize; | |
277 | BOOL Xprev_is_word; | |
278 | ||
279 | unsigned long int Xoriginal_ims; | |
280 | ||
281 | #ifdef SUPPORT_UCP | |
282 | int Xprop_type; | |
283 | int Xprop_fail_result; | |
284 | int Xprop_category; | |
285 | int Xprop_chartype; | |
286 | int Xprop_othercase; | |
287 | int Xprop_test_against; | |
288 | int *Xprop_test_variable; | |
289 | #endif | |
290 | ||
291 | int Xctype; | |
292 | int Xfc; | |
293 | int Xfi; | |
294 | int Xlength; | |
295 | int Xmax; | |
296 | int Xmin; | |
297 | int Xnumber; | |
298 | int Xoffset; | |
299 | int Xop; | |
300 | int Xsave_capture_last; | |
301 | int Xsave_offset1, Xsave_offset2, Xsave_offset3; | |
302 | int Xstacksave[REC_STACK_SAVE_MAX]; | |
303 | ||
304 | eptrblock Xnewptrb; | |
305 | ||
306 | /* Place to pass back result, and where to jump back to */ | |
307 | ||
308 | int Xresult; | |
309 | jmp_buf Xwhere; | |
310 | ||
311 | } heapframe; | |
312 | ||
313 | #endif | |
314 | ||
315 | ||
316 | /*************************************************************************** | |
317 | ***************************************************************************/ | |
318 | ||
319 | ||
320 | ||
321 | /************************************************* | |
322 | * Match from current position * | |
323 | *************************************************/ | |
324 | ||
325 | /* On entry ecode points to the first opcode, and eptr to the first character | |
326 | in the subject string, while eptrb holds the value of eptr at the start of the | |
327 | last bracketed group - used for breaking infinite loops matching zero-length | |
328 | strings. This function is called recursively in many circumstances. Whenever it | |
329 | returns a negative (error) response, the outer incarnation must also return the | |
330 | same response. | |
331 | ||
332 | Performance note: It might be tempting to extract commonly used fields from the | |
333 | md structure (e.g. utf8, end_subject) into individual variables to improve | |
334 | performance. Tests using gcc on a SPARC disproved this; in the first case, it | |
335 | made performance worse. | |
336 | ||
337 | Arguments: | |
338 | eptr pointer in subject | |
339 | ecode position in code | |
340 | offset_top current top pointer | |
341 | md pointer to "static" info for the match | |
342 | ims current /i, /m, and /s options | |
343 | eptrb pointer to chain of blocks containing eptr at start of | |
344 | brackets - for testing for empty matches | |
345 | flags can contain | |
346 | match_condassert - this is an assertion condition | |
347 | match_isgroup - this is the start of a bracketed group | |
348 | ||
349 | Returns: MATCH_MATCH if matched ) these values are >= 0 | |
350 | MATCH_NOMATCH if failed to match ) | |
351 | a negative PCRE_ERROR_xxx value if aborted by an error condition | |
352 | (e.g. stopped by recursion limit) | |
353 | */ | |
354 | ||
355 | static int | |
356 | match(REGISTER const uschar *eptr, REGISTER const uschar *ecode, | |
357 | int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb, | |
358 | int flags) | |
359 | { | |
360 | /* These variables do not need to be preserved over recursion in this function, | |
361 | so they can be ordinary variables in all cases. Mark them with "register" | |
362 | because they are used a lot in loops. */ | |
363 | ||
364 | register int rrc; /* Returns from recursive calls */ | |
365 | register int i; /* Used for loops not involving calls to RMATCH() */ | |
366 | register int c; /* Character values not kept over RMATCH() calls */ | |
367 | register BOOL utf8; /* Local copy of UTF-8 flag for speed */ | |
368 | ||
369 | /* When recursion is not being used, all "local" variables that have to be | |
370 | preserved over calls to RMATCH() are part of a "frame" which is obtained from | |
371 | heap storage. Set up the top-level frame here; others are obtained from the | |
372 | heap whenever RMATCH() does a "recursion". See the macro definitions above. */ | |
373 | ||
374 | #ifdef NO_RECURSE | |
375 | heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe)); | |
376 | frame->Xprevframe = NULL; /* Marks the top level */ | |
377 | ||
378 | /* Copy in the original argument variables */ | |
379 | ||
380 | frame->Xeptr = eptr; | |
381 | frame->Xecode = ecode; | |
382 | frame->Xoffset_top = offset_top; | |
383 | frame->Xims = ims; | |
384 | frame->Xeptrb = eptrb; | |
385 | frame->Xflags = flags; | |
386 | ||
387 | /* This is where control jumps back to to effect "recursion" */ | |
388 | ||
389 | HEAP_RECURSE: | |
390 | ||
391 | /* Macros make the argument variables come from the current frame */ | |
392 | ||
393 | #define eptr frame->Xeptr | |
394 | #define ecode frame->Xecode | |
395 | #define offset_top frame->Xoffset_top | |
396 | #define ims frame->Xims | |
397 | #define eptrb frame->Xeptrb | |
398 | #define flags frame->Xflags | |
399 | ||
400 | /* Ditto for the local variables */ | |
401 | ||
402 | #ifdef SUPPORT_UTF8 | |
403 | #define charptr frame->Xcharptr | |
404 | #endif | |
405 | #define callpat frame->Xcallpat | |
406 | #define data frame->Xdata | |
407 | #define next frame->Xnext | |
408 | #define pp frame->Xpp | |
409 | #define prev frame->Xprev | |
410 | #define saved_eptr frame->Xsaved_eptr | |
411 | ||
412 | #define new_recursive frame->Xnew_recursive | |
413 | ||
414 | #define cur_is_word frame->Xcur_is_word | |
415 | #define condition frame->Xcondition | |
416 | #define minimize frame->Xminimize | |
417 | #define prev_is_word frame->Xprev_is_word | |
418 | ||
419 | #define original_ims frame->Xoriginal_ims | |
420 | ||
421 | #ifdef SUPPORT_UCP | |
422 | #define prop_type frame->Xprop_type | |
423 | #define prop_fail_result frame->Xprop_fail_result | |
424 | #define prop_category frame->Xprop_category | |
425 | #define prop_chartype frame->Xprop_chartype | |
426 | #define prop_othercase frame->Xprop_othercase | |
427 | #define prop_test_against frame->Xprop_test_against | |
428 | #define prop_test_variable frame->Xprop_test_variable | |
429 | #endif | |
430 | ||
431 | #define ctype frame->Xctype | |
432 | #define fc frame->Xfc | |
433 | #define fi frame->Xfi | |
434 | #define length frame->Xlength | |
435 | #define max frame->Xmax | |
436 | #define min frame->Xmin | |
437 | #define number frame->Xnumber | |
438 | #define offset frame->Xoffset | |
439 | #define op frame->Xop | |
440 | #define save_capture_last frame->Xsave_capture_last | |
441 | #define save_offset1 frame->Xsave_offset1 | |
442 | #define save_offset2 frame->Xsave_offset2 | |
443 | #define save_offset3 frame->Xsave_offset3 | |
444 | #define stacksave frame->Xstacksave | |
445 | ||
446 | #define newptrb frame->Xnewptrb | |
447 | ||
448 | /* When recursion is being used, local variables are allocated on the stack and | |
449 | get preserved during recursion in the normal way. In this environment, fi and | |
450 | i, and fc and c, can be the same variables. */ | |
451 | ||
452 | #else | |
453 | #define fi i | |
454 | #define fc c | |
455 | ||
456 | ||
457 | #ifdef SUPPORT_UTF8 /* Many of these variables are used ony */ | |
458 | const uschar *charptr; /* small blocks of the code. My normal */ | |
459 | #endif /* style of coding would have declared */ | |
460 | const uschar *callpat; /* them within each of those blocks. */ | |
461 | const uschar *data; /* However, in order to accommodate the */ | |
462 | const uschar *next; /* version of this code that uses an */ | |
463 | const uschar *pp; /* external "stack" implemented on the */ | |
464 | const uschar *prev; /* heap, it is easier to declare them */ | |
465 | const uschar *saved_eptr; /* all here, so the declarations can */ | |
466 | /* be cut out in a block. The only */ | |
467 | recursion_info new_recursive; /* declarations within blocks below are */ | |
468 | /* for variables that do not have to */ | |
469 | BOOL cur_is_word; /* be preserved over a recursive call */ | |
470 | BOOL condition; /* to RMATCH(). */ | |
471 | BOOL minimize; | |
472 | BOOL prev_is_word; | |
473 | ||
474 | unsigned long int original_ims; | |
475 | ||
476 | #ifdef SUPPORT_UCP | |
477 | int prop_type; | |
478 | int prop_fail_result; | |
479 | int prop_category; | |
480 | int prop_chartype; | |
481 | int prop_othercase; | |
482 | int prop_test_against; | |
483 | int *prop_test_variable; | |
484 | #endif | |
485 | ||
486 | int ctype; | |
487 | int length; | |
488 | int max; | |
489 | int min; | |
490 | int number; | |
491 | int offset; | |
492 | int op; | |
493 | int save_capture_last; | |
494 | int save_offset1, save_offset2, save_offset3; | |
495 | int stacksave[REC_STACK_SAVE_MAX]; | |
496 | ||
497 | eptrblock newptrb; | |
498 | #endif | |
499 | ||
500 | /* These statements are here to stop the compiler complaining about unitialized | |
501 | variables. */ | |
502 | ||
503 | #ifdef SUPPORT_UCP | |
504 | prop_fail_result = 0; | |
505 | prop_test_against = 0; | |
506 | prop_test_variable = NULL; | |
507 | #endif | |
508 | ||
509 | /* OK, now we can get on with the real code of the function. Recursion is | |
510 | specified by the macros RMATCH and RRETURN. When NO_RECURSE is *not* defined, | |
511 | these just turn into a recursive call to match() and a "return", respectively. | |
512 | However, RMATCH isn't like a function call because it's quite a complicated | |
513 | macro. It has to be used in one particular way. This shouldn't, however, impact | |
514 | performance when true recursion is being used. */ | |
515 | ||
516 | if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT); | |
517 | ||
518 | original_ims = ims; /* Save for resetting on ')' */ | |
519 | utf8 = md->utf8; /* Local copy of the flag */ | |
520 | ||
521 | /* At the start of a bracketed group, add the current subject pointer to the | |
522 | stack of such pointers, to be re-instated at the end of the group when we hit | |
523 | the closing ket. When match() is called in other circumstances, we don't add to | |
524 | this stack. */ | |
525 | ||
526 | if ((flags & match_isgroup) != 0) | |
527 | { | |
528 | newptrb.epb_prev = eptrb; | |
529 | newptrb.epb_saved_eptr = eptr; | |
530 | eptrb = &newptrb; | |
531 | } | |
532 | ||
533 | /* Now start processing the operations. */ | |
534 | ||
535 | for (;;) | |
536 | { | |
537 | op = *ecode; | |
538 | minimize = FALSE; | |
539 | ||
540 | /* For partial matching, remember if we ever hit the end of the subject after | |
541 | matching at least one subject character. */ | |
542 | ||
543 | if (md->partial && | |
544 | eptr >= md->end_subject && | |
545 | eptr > md->start_match) | |
546 | md->hitend = TRUE; | |
547 | ||
548 | /* Opening capturing bracket. If there is space in the offset vector, save | |
549 | the current subject position in the working slot at the top of the vector. We | |
550 | mustn't change the current values of the data slot, because they may be set | |
551 | from a previous iteration of this group, and be referred to by a reference | |
552 | inside the group. | |
553 | ||
554 | If the bracket fails to match, we need to restore this value and also the | |
555 | values of the final offsets, in case they were set by a previous iteration of | |
556 | the same bracket. | |
557 | ||
558 | If there isn't enough space in the offset vector, treat this as if it were a | |
559 | non-capturing bracket. Don't worry about setting the flag for the error case | |
560 | here; that is handled in the code for KET. */ | |
561 | ||
562 | if (op > OP_BRA) | |
563 | { | |
564 | number = op - OP_BRA; | |
565 | ||
566 | /* For extended extraction brackets (large number), we have to fish out the | |
567 | number from a dummy opcode at the start. */ | |
568 | ||
569 | if (number > EXTRACT_BASIC_MAX) | |
570 | number = GET2(ecode, 2+LINK_SIZE); | |
571 | offset = number << 1; | |
572 | ||
573 | #ifdef DEBUG | |
574 | printf("start bracket %d subject=", number); | |
575 | pchars(eptr, 16, TRUE, md); | |
576 | printf("\n"); | |
577 | #endif | |
578 | ||
579 | if (offset < md->offset_max) | |
580 | { | |
581 | save_offset1 = md->offset_vector[offset]; | |
582 | save_offset2 = md->offset_vector[offset+1]; | |
583 | save_offset3 = md->offset_vector[md->offset_end - number]; | |
584 | save_capture_last = md->capture_last; | |
585 | ||
586 | DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3)); | |
587 | md->offset_vector[md->offset_end - number] = eptr - md->start_subject; | |
588 | ||
589 | do | |
590 | { | |
591 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, | |
592 | match_isgroup); | |
593 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
594 | md->capture_last = save_capture_last; | |
595 | ecode += GET(ecode, 1); | |
596 | } | |
597 | while (*ecode == OP_ALT); | |
598 | ||
599 | DPRINTF(("bracket %d failed\n", number)); | |
600 | ||
601 | md->offset_vector[offset] = save_offset1; | |
602 | md->offset_vector[offset+1] = save_offset2; | |
603 | md->offset_vector[md->offset_end - number] = save_offset3; | |
604 | ||
605 | RRETURN(MATCH_NOMATCH); | |
606 | } | |
607 | ||
608 | /* Insufficient room for saving captured contents */ | |
609 | ||
610 | else op = OP_BRA; | |
611 | } | |
612 | ||
613 | /* Other types of node can be handled by a switch */ | |
614 | ||
615 | switch(op) | |
616 | { | |
617 | case OP_BRA: /* Non-capturing bracket: optimized */ | |
618 | DPRINTF(("start bracket 0\n")); | |
619 | do | |
620 | { | |
621 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, | |
622 | match_isgroup); | |
623 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
624 | ecode += GET(ecode, 1); | |
625 | } | |
626 | while (*ecode == OP_ALT); | |
627 | DPRINTF(("bracket 0 failed\n")); | |
628 | RRETURN(MATCH_NOMATCH); | |
629 | ||
630 | /* Conditional group: compilation checked that there are no more than | |
631 | two branches. If the condition is false, skipping the first branch takes us | |
632 | past the end if there is only one branch, but that's OK because that is | |
633 | exactly what going to the ket would do. */ | |
634 | ||
635 | case OP_COND: | |
636 | if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */ | |
637 | { | |
638 | offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */ | |
639 | condition = (offset == CREF_RECURSE * 2)? | |
640 | (md->recursive != NULL) : | |
641 | (offset < offset_top && md->offset_vector[offset] >= 0); | |
642 | RMATCH(rrc, eptr, ecode + (condition? | |
643 | (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1))), | |
644 | offset_top, md, ims, eptrb, match_isgroup); | |
645 | RRETURN(rrc); | |
646 | } | |
647 | ||
648 | /* The condition is an assertion. Call match() to evaluate it - setting | |
649 | the final argument TRUE causes it to stop at the end of an assertion. */ | |
650 | ||
651 | else | |
652 | { | |
653 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, | |
654 | match_condassert | match_isgroup); | |
655 | if (rrc == MATCH_MATCH) | |
656 | { | |
657 | ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2); | |
658 | while (*ecode == OP_ALT) ecode += GET(ecode, 1); | |
659 | } | |
660 | else if (rrc != MATCH_NOMATCH) | |
661 | { | |
662 | RRETURN(rrc); /* Need braces because of following else */ | |
663 | } | |
664 | else ecode += GET(ecode, 1); | |
665 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, | |
666 | match_isgroup); | |
667 | RRETURN(rrc); | |
668 | } | |
669 | /* Control never reaches here */ | |
670 | ||
671 | /* Skip over conditional reference or large extraction number data if | |
672 | encountered. */ | |
673 | ||
674 | case OP_CREF: | |
675 | case OP_BRANUMBER: | |
676 | ecode += 3; | |
677 | break; | |
678 | ||
679 | /* End of the pattern. If we are in a recursion, we should restore the | |
680 | offsets appropriately and continue from after the call. */ | |
681 | ||
682 | case OP_END: | |
683 | if (md->recursive != NULL && md->recursive->group_num == 0) | |
684 | { | |
685 | recursion_info *rec = md->recursive; | |
686 | DPRINTF(("Hit the end in a (?0) recursion\n")); | |
687 | md->recursive = rec->prevrec; | |
688 | memmove(md->offset_vector, rec->offset_save, | |
689 | rec->saved_max * sizeof(int)); | |
690 | md->start_match = rec->save_start; | |
691 | ims = original_ims; | |
692 | ecode = rec->after_call; | |
693 | break; | |
694 | } | |
695 | ||
696 | /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty | |
697 | string - backtracking will then try other alternatives, if any. */ | |
698 | ||
699 | if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH); | |
700 | md->end_match_ptr = eptr; /* Record where we ended */ | |
701 | md->end_offset_top = offset_top; /* and how many extracts were taken */ | |
702 | RRETURN(MATCH_MATCH); | |
703 | ||
704 | /* Change option settings */ | |
705 | ||
706 | case OP_OPT: | |
707 | ims = ecode[1]; | |
708 | ecode += 2; | |
709 | DPRINTF(("ims set to %02lx\n", ims)); | |
710 | break; | |
711 | ||
712 | /* Assertion brackets. Check the alternative branches in turn - the | |
713 | matching won't pass the KET for an assertion. If any one branch matches, | |
714 | the assertion is true. Lookbehind assertions have an OP_REVERSE item at the | |
715 | start of each branch to move the current point backwards, so the code at | |
716 | this level is identical to the lookahead case. */ | |
717 | ||
718 | case OP_ASSERT: | |
719 | case OP_ASSERTBACK: | |
720 | do | |
721 | { | |
722 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, | |
723 | match_isgroup); | |
724 | if (rrc == MATCH_MATCH) break; | |
725 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
726 | ecode += GET(ecode, 1); | |
727 | } | |
728 | while (*ecode == OP_ALT); | |
729 | if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH); | |
730 | ||
731 | /* If checking an assertion for a condition, return MATCH_MATCH. */ | |
732 | ||
733 | if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH); | |
734 | ||
735 | /* Continue from after the assertion, updating the offsets high water | |
736 | mark, since extracts may have been taken during the assertion. */ | |
737 | ||
738 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); | |
739 | ecode += 1 + LINK_SIZE; | |
740 | offset_top = md->end_offset_top; | |
741 | continue; | |
742 | ||
743 | /* Negative assertion: all branches must fail to match */ | |
744 | ||
745 | case OP_ASSERT_NOT: | |
746 | case OP_ASSERTBACK_NOT: | |
747 | do | |
748 | { | |
749 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, | |
750 | match_isgroup); | |
751 | if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH); | |
752 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
753 | ecode += GET(ecode,1); | |
754 | } | |
755 | while (*ecode == OP_ALT); | |
756 | ||
757 | if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH); | |
758 | ||
759 | ecode += 1 + LINK_SIZE; | |
760 | continue; | |
761 | ||
762 | /* Move the subject pointer back. This occurs only at the start of | |
763 | each branch of a lookbehind assertion. If we are too close to the start to | |
764 | move back, this match function fails. When working with UTF-8 we move | |
765 | back a number of characters, not bytes. */ | |
766 | ||
767 | case OP_REVERSE: | |
768 | #ifdef SUPPORT_UTF8 | |
769 | if (utf8) | |
770 | { | |
771 | c = GET(ecode,1); | |
772 | for (i = 0; i < c; i++) | |
773 | { | |
774 | eptr--; | |
775 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); | |
776 | BACKCHAR(eptr) | |
777 | } | |
778 | } | |
779 | else | |
780 | #endif | |
781 | ||
782 | /* No UTF-8 support, or not in UTF-8 mode: count is byte count */ | |
783 | ||
784 | { | |
785 | eptr -= GET(ecode,1); | |
786 | if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH); | |
787 | } | |
788 | ||
789 | /* Skip to next op code */ | |
790 | ||
791 | ecode += 1 + LINK_SIZE; | |
792 | break; | |
793 | ||
794 | /* The callout item calls an external function, if one is provided, passing | |
795 | details of the match so far. This is mainly for debugging, though the | |
796 | function is able to force a failure. */ | |
797 | ||
798 | case OP_CALLOUT: | |
799 | if (pcre_callout != NULL) | |
800 | { | |
801 | pcre_callout_block cb; | |
802 | cb.version = 1; /* Version 1 of the callout block */ | |
803 | cb.callout_number = ecode[1]; | |
804 | cb.offset_vector = md->offset_vector; | |
805 | cb.subject = (const char *)md->start_subject; | |
806 | cb.subject_length = md->end_subject - md->start_subject; | |
807 | cb.start_match = md->start_match - md->start_subject; | |
808 | cb.current_position = eptr - md->start_subject; | |
809 | cb.pattern_position = GET(ecode, 2); | |
810 | cb.next_item_length = GET(ecode, 2 + LINK_SIZE); | |
811 | cb.capture_top = offset_top/2; | |
812 | cb.capture_last = md->capture_last; | |
813 | cb.callout_data = md->callout_data; | |
814 | if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH); | |
815 | if (rrc < 0) RRETURN(rrc); | |
816 | } | |
817 | ecode += 2 + 2*LINK_SIZE; | |
818 | break; | |
819 | ||
820 | /* Recursion either matches the current regex, or some subexpression. The | |
821 | offset data is the offset to the starting bracket from the start of the | |
822 | whole pattern. (This is so that it works from duplicated subpatterns.) | |
823 | ||
824 | If there are any capturing brackets started but not finished, we have to | |
825 | save their starting points and reinstate them after the recursion. However, | |
826 | we don't know how many such there are (offset_top records the completed | |
827 | total) so we just have to save all the potential data. There may be up to | |
828 | 65535 such values, which is too large to put on the stack, but using malloc | |
829 | for small numbers seems expensive. As a compromise, the stack is used when | |
830 | there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc | |
831 | is used. A problem is what to do if the malloc fails ... there is no way of | |
832 | returning to the top level with an error. Save the top REC_STACK_SAVE_MAX | |
833 | values on the stack, and accept that the rest may be wrong. | |
834 | ||
835 | There are also other values that have to be saved. We use a chained | |
836 | sequence of blocks that actually live on the stack. Thanks to Robin Houston | |
837 | for the original version of this logic. */ | |
838 | ||
839 | case OP_RECURSE: | |
840 | { | |
841 | callpat = md->start_code + GET(ecode, 1); | |
842 | new_recursive.group_num = *callpat - OP_BRA; | |
843 | ||
844 | /* For extended extraction brackets (large number), we have to fish out | |
845 | the number from a dummy opcode at the start. */ | |
846 | ||
847 | if (new_recursive.group_num > EXTRACT_BASIC_MAX) | |
848 | new_recursive.group_num = GET2(callpat, 2+LINK_SIZE); | |
849 | ||
850 | /* Add to "recursing stack" */ | |
851 | ||
852 | new_recursive.prevrec = md->recursive; | |
853 | md->recursive = &new_recursive; | |
854 | ||
855 | /* Find where to continue from afterwards */ | |
856 | ||
857 | ecode += 1 + LINK_SIZE; | |
858 | new_recursive.after_call = ecode; | |
859 | ||
860 | /* Now save the offset data. */ | |
861 | ||
862 | new_recursive.saved_max = md->offset_end; | |
863 | if (new_recursive.saved_max <= REC_STACK_SAVE_MAX) | |
864 | new_recursive.offset_save = stacksave; | |
865 | else | |
866 | { | |
867 | new_recursive.offset_save = | |
868 | (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int)); | |
869 | if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY); | |
870 | } | |
871 | ||
872 | memcpy(new_recursive.offset_save, md->offset_vector, | |
873 | new_recursive.saved_max * sizeof(int)); | |
874 | new_recursive.save_start = md->start_match; | |
875 | md->start_match = eptr; | |
876 | ||
877 | /* OK, now we can do the recursion. For each top-level alternative we | |
878 | restore the offset and recursion data. */ | |
879 | ||
880 | DPRINTF(("Recursing into group %d\n", new_recursive.group_num)); | |
881 | do | |
882 | { | |
883 | RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims, | |
884 | eptrb, match_isgroup); | |
885 | if (rrc == MATCH_MATCH) | |
886 | { | |
887 | md->recursive = new_recursive.prevrec; | |
888 | if (new_recursive.offset_save != stacksave) | |
889 | (pcre_free)(new_recursive.offset_save); | |
890 | RRETURN(MATCH_MATCH); | |
891 | } | |
892 | else if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
893 | ||
894 | md->recursive = &new_recursive; | |
895 | memcpy(md->offset_vector, new_recursive.offset_save, | |
896 | new_recursive.saved_max * sizeof(int)); | |
897 | callpat += GET(callpat, 1); | |
898 | } | |
899 | while (*callpat == OP_ALT); | |
900 | ||
901 | DPRINTF(("Recursion didn't match\n")); | |
902 | md->recursive = new_recursive.prevrec; | |
903 | if (new_recursive.offset_save != stacksave) | |
904 | (pcre_free)(new_recursive.offset_save); | |
905 | RRETURN(MATCH_NOMATCH); | |
906 | } | |
907 | /* Control never reaches here */ | |
908 | ||
909 | /* "Once" brackets are like assertion brackets except that after a match, | |
910 | the point in the subject string is not moved back. Thus there can never be | |
911 | a move back into the brackets. Friedl calls these "atomic" subpatterns. | |
912 | Check the alternative branches in turn - the matching won't pass the KET | |
913 | for this kind of subpattern. If any one branch matches, we carry on as at | |
914 | the end of a normal bracket, leaving the subject pointer. */ | |
915 | ||
916 | case OP_ONCE: | |
917 | { | |
918 | prev = ecode; | |
919 | saved_eptr = eptr; | |
920 | ||
921 | do | |
922 | { | |
923 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, | |
924 | eptrb, match_isgroup); | |
925 | if (rrc == MATCH_MATCH) break; | |
926 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
927 | ecode += GET(ecode,1); | |
928 | } | |
929 | while (*ecode == OP_ALT); | |
930 | ||
931 | /* If hit the end of the group (which could be repeated), fail */ | |
932 | ||
933 | if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH); | |
934 | ||
935 | /* Continue as from after the assertion, updating the offsets high water | |
936 | mark, since extracts may have been taken. */ | |
937 | ||
938 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); | |
939 | ||
940 | offset_top = md->end_offset_top; | |
941 | eptr = md->end_match_ptr; | |
942 | ||
943 | /* For a non-repeating ket, just continue at this level. This also | |
944 | happens for a repeating ket if no characters were matched in the group. | |
945 | This is the forcible breaking of infinite loops as implemented in Perl | |
946 | 5.005. If there is an options reset, it will get obeyed in the normal | |
947 | course of events. */ | |
948 | ||
949 | if (*ecode == OP_KET || eptr == saved_eptr) | |
950 | { | |
951 | ecode += 1+LINK_SIZE; | |
952 | break; | |
953 | } | |
954 | ||
955 | /* The repeating kets try the rest of the pattern or restart from the | |
956 | preceding bracket, in the appropriate order. We need to reset any options | |
957 | that changed within the bracket before re-running it, so check the next | |
958 | opcode. */ | |
959 | ||
960 | if (ecode[1+LINK_SIZE] == OP_OPT) | |
961 | { | |
962 | ims = (ims & ~PCRE_IMS) | ecode[4]; | |
963 | DPRINTF(("ims set to %02lx at group repeat\n", ims)); | |
964 | } | |
965 | ||
966 | if (*ecode == OP_KETRMIN) | |
967 | { | |
968 | RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0); | |
969 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
970 | RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup); | |
971 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
972 | } | |
973 | else /* OP_KETRMAX */ | |
974 | { | |
975 | RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup); | |
976 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
977 | RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0); | |
978 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
979 | } | |
980 | } | |
981 | RRETURN(MATCH_NOMATCH); | |
982 | ||
983 | /* An alternation is the end of a branch; scan along to find the end of the | |
984 | bracketed group and go to there. */ | |
985 | ||
986 | case OP_ALT: | |
987 | do ecode += GET(ecode,1); while (*ecode == OP_ALT); | |
988 | break; | |
989 | ||
990 | /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating | |
991 | that it may occur zero times. It may repeat infinitely, or not at all - | |
992 | i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper | |
993 | repeat limits are compiled as a number of copies, with the optional ones | |
994 | preceded by BRAZERO or BRAMINZERO. */ | |
995 | ||
996 | case OP_BRAZERO: | |
997 | { | |
998 | next = ecode+1; | |
999 | RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup); | |
1000 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1001 | do next += GET(next,1); while (*next == OP_ALT); | |
1002 | ecode = next + 1+LINK_SIZE; | |
1003 | } | |
1004 | break; | |
1005 | ||
1006 | case OP_BRAMINZERO: | |
1007 | { | |
1008 | next = ecode+1; | |
1009 | do next += GET(next,1); while (*next == OP_ALT); | |
1010 | RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, | |
1011 | match_isgroup); | |
1012 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1013 | ecode++; | |
1014 | } | |
1015 | break; | |
1016 | ||
1017 | /* End of a group, repeated or non-repeating. If we are at the end of | |
1018 | an assertion "group", stop matching and return MATCH_MATCH, but record the | |
1019 | current high water mark for use by positive assertions. Do this also | |
1020 | for the "once" (not-backup up) groups. */ | |
1021 | ||
1022 | case OP_KET: | |
1023 | case OP_KETRMIN: | |
1024 | case OP_KETRMAX: | |
1025 | { | |
1026 | prev = ecode - GET(ecode, 1); | |
1027 | saved_eptr = eptrb->epb_saved_eptr; | |
1028 | ||
1029 | /* Back up the stack of bracket start pointers. */ | |
1030 | ||
1031 | eptrb = eptrb->epb_prev; | |
1032 | ||
1033 | if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT || | |
1034 | *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT || | |
1035 | *prev == OP_ONCE) | |
1036 | { | |
1037 | md->end_match_ptr = eptr; /* For ONCE */ | |
1038 | md->end_offset_top = offset_top; | |
1039 | RRETURN(MATCH_MATCH); | |
1040 | } | |
1041 | ||
1042 | /* In all other cases except a conditional group we have to check the | |
1043 | group number back at the start and if necessary complete handling an | |
1044 | extraction by setting the offsets and bumping the high water mark. */ | |
1045 | ||
1046 | if (*prev != OP_COND) | |
1047 | { | |
1048 | number = *prev - OP_BRA; | |
1049 | ||
1050 | /* For extended extraction brackets (large number), we have to fish out | |
1051 | the number from a dummy opcode at the start. */ | |
1052 | ||
1053 | if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE); | |
1054 | offset = number << 1; | |
1055 | ||
1056 | #ifdef DEBUG | |
1057 | printf("end bracket %d", number); | |
1058 | printf("\n"); | |
1059 | #endif | |
1060 | ||
1061 | /* Test for a numbered group. This includes groups called as a result | |
1062 | of recursion. Note that whole-pattern recursion is coded as a recurse | |
1063 | into group 0, so it won't be picked up here. Instead, we catch it when | |
1064 | the OP_END is reached. */ | |
1065 | ||
1066 | if (number > 0) | |
1067 | { | |
1068 | md->capture_last = number; | |
1069 | if (offset >= md->offset_max) md->offset_overflow = TRUE; else | |
1070 | { | |
1071 | md->offset_vector[offset] = | |
1072 | md->offset_vector[md->offset_end - number]; | |
1073 | md->offset_vector[offset+1] = eptr - md->start_subject; | |
1074 | if (offset_top <= offset) offset_top = offset + 2; | |
1075 | } | |
1076 | ||
1077 | /* Handle a recursively called group. Restore the offsets | |
1078 | appropriately and continue from after the call. */ | |
1079 | ||
1080 | if (md->recursive != NULL && md->recursive->group_num == number) | |
1081 | { | |
1082 | recursion_info *rec = md->recursive; | |
1083 | DPRINTF(("Recursion (%d) succeeded - continuing\n", number)); | |
1084 | md->recursive = rec->prevrec; | |
1085 | md->start_match = rec->save_start; | |
1086 | memcpy(md->offset_vector, rec->offset_save, | |
1087 | rec->saved_max * sizeof(int)); | |
1088 | ecode = rec->after_call; | |
1089 | ims = original_ims; | |
1090 | break; | |
1091 | } | |
1092 | } | |
1093 | } | |
1094 | ||
1095 | /* Reset the value of the ims flags, in case they got changed during | |
1096 | the group. */ | |
1097 | ||
1098 | ims = original_ims; | |
1099 | DPRINTF(("ims reset to %02lx\n", ims)); | |
1100 | ||
1101 | /* For a non-repeating ket, just continue at this level. This also | |
1102 | happens for a repeating ket if no characters were matched in the group. | |
1103 | This is the forcible breaking of infinite loops as implemented in Perl | |
1104 | 5.005. If there is an options reset, it will get obeyed in the normal | |
1105 | course of events. */ | |
1106 | ||
1107 | if (*ecode == OP_KET || eptr == saved_eptr) | |
1108 | { | |
1109 | ecode += 1 + LINK_SIZE; | |
1110 | break; | |
1111 | } | |
1112 | ||
1113 | /* The repeating kets try the rest of the pattern or restart from the | |
1114 | preceding bracket, in the appropriate order. */ | |
1115 | ||
1116 | if (*ecode == OP_KETRMIN) | |
1117 | { | |
1118 | RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0); | |
1119 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1120 | RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup); | |
1121 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1122 | } | |
1123 | else /* OP_KETRMAX */ | |
1124 | { | |
1125 | RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup); | |
1126 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1127 | RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0); | |
1128 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1129 | } | |
1130 | } | |
1131 | ||
1132 | RRETURN(MATCH_NOMATCH); | |
1133 | ||
1134 | /* Start of subject unless notbol, or after internal newline if multiline */ | |
1135 | ||
1136 | case OP_CIRC: | |
1137 | if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH); | |
1138 | if ((ims & PCRE_MULTILINE) != 0) | |
1139 | { | |
1140 | if (eptr != md->start_subject && eptr[-1] != NEWLINE) | |
1141 | RRETURN(MATCH_NOMATCH); | |
1142 | ecode++; | |
1143 | break; | |
1144 | } | |
1145 | /* ... else fall through */ | |
1146 | ||
1147 | /* Start of subject assertion */ | |
1148 | ||
1149 | case OP_SOD: | |
1150 | if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH); | |
1151 | ecode++; | |
1152 | break; | |
1153 | ||
1154 | /* Start of match assertion */ | |
1155 | ||
1156 | case OP_SOM: | |
1157 | if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH); | |
1158 | ecode++; | |
1159 | break; | |
1160 | ||
1161 | /* Assert before internal newline if multiline, or before a terminating | |
1162 | newline unless endonly is set, else end of subject unless noteol is set. */ | |
1163 | ||
1164 | case OP_DOLL: | |
1165 | if ((ims & PCRE_MULTILINE) != 0) | |
1166 | { | |
1167 | if (eptr < md->end_subject) | |
1168 | { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); } | |
1169 | else | |
1170 | { if (md->noteol) RRETURN(MATCH_NOMATCH); } | |
1171 | ecode++; | |
1172 | break; | |
1173 | } | |
1174 | else | |
1175 | { | |
1176 | if (md->noteol) RRETURN(MATCH_NOMATCH); | |
1177 | if (!md->endonly) | |
1178 | { | |
1179 | if (eptr < md->end_subject - 1 || | |
1180 | (eptr == md->end_subject - 1 && *eptr != NEWLINE)) | |
1181 | RRETURN(MATCH_NOMATCH); | |
1182 | ecode++; | |
1183 | break; | |
1184 | } | |
1185 | } | |
1186 | /* ... else fall through */ | |
1187 | ||
1188 | /* End of subject assertion (\z) */ | |
1189 | ||
1190 | case OP_EOD: | |
1191 | if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH); | |
1192 | ecode++; | |
1193 | break; | |
1194 | ||
1195 | /* End of subject or ending \n assertion (\Z) */ | |
1196 | ||
1197 | case OP_EODN: | |
1198 | if (eptr < md->end_subject - 1 || | |
1199 | (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH); | |
1200 | ecode++; | |
1201 | break; | |
1202 | ||
1203 | /* Word boundary assertions */ | |
1204 | ||
1205 | case OP_NOT_WORD_BOUNDARY: | |
1206 | case OP_WORD_BOUNDARY: | |
1207 | { | |
1208 | ||
1209 | /* Find out if the previous and current characters are "word" characters. | |
1210 | It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to | |
1211 | be "non-word" characters. */ | |
1212 | ||
1213 | #ifdef SUPPORT_UTF8 | |
1214 | if (utf8) | |
1215 | { | |
1216 | if (eptr == md->start_subject) prev_is_word = FALSE; else | |
1217 | { | |
1218 | const uschar *lastptr = eptr - 1; | |
1219 | while((*lastptr & 0xc0) == 0x80) lastptr--; | |
1220 | GETCHAR(c, lastptr); | |
1221 | prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; | |
1222 | } | |
1223 | if (eptr >= md->end_subject) cur_is_word = FALSE; else | |
1224 | { | |
1225 | GETCHAR(c, eptr); | |
1226 | cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0; | |
1227 | } | |
1228 | } | |
1229 | else | |
1230 | #endif | |
1231 | ||
1232 | /* More streamlined when not in UTF-8 mode */ | |
1233 | ||
1234 | { | |
1235 | prev_is_word = (eptr != md->start_subject) && | |
1236 | ((md->ctypes[eptr[-1]] & ctype_word) != 0); | |
1237 | cur_is_word = (eptr < md->end_subject) && | |
1238 | ((md->ctypes[*eptr] & ctype_word) != 0); | |
1239 | } | |
1240 | ||
1241 | /* Now see if the situation is what we want */ | |
1242 | ||
1243 | if ((*ecode++ == OP_WORD_BOUNDARY)? | |
1244 | cur_is_word == prev_is_word : cur_is_word != prev_is_word) | |
1245 | RRETURN(MATCH_NOMATCH); | |
1246 | } | |
1247 | break; | |
1248 | ||
1249 | /* Match a single character type; inline for speed */ | |
1250 | ||
1251 | case OP_ANY: | |
1252 | if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE) | |
1253 | RRETURN(MATCH_NOMATCH); | |
1254 | if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1255 | #ifdef SUPPORT_UTF8 | |
1256 | if (utf8) | |
1257 | while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++; | |
1258 | #endif | |
1259 | ecode++; | |
1260 | break; | |
1261 | ||
1262 | /* Match a single byte, even in UTF-8 mode. This opcode really does match | |
1263 | any byte, even newline, independent of the setting of PCRE_DOTALL. */ | |
1264 | ||
1265 | case OP_ANYBYTE: | |
1266 | if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1267 | ecode++; | |
1268 | break; | |
1269 | ||
1270 | case OP_NOT_DIGIT: | |
1271 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1272 | GETCHARINCTEST(c, eptr); | |
1273 | if ( | |
1274 | #ifdef SUPPORT_UTF8 | |
1275 | c < 256 && | |
1276 | #endif | |
1277 | (md->ctypes[c] & ctype_digit) != 0 | |
1278 | ) | |
1279 | RRETURN(MATCH_NOMATCH); | |
1280 | ecode++; | |
1281 | break; | |
1282 | ||
1283 | case OP_DIGIT: | |
1284 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1285 | GETCHARINCTEST(c, eptr); | |
1286 | if ( | |
1287 | #ifdef SUPPORT_UTF8 | |
1288 | c >= 256 || | |
1289 | #endif | |
1290 | (md->ctypes[c] & ctype_digit) == 0 | |
1291 | ) | |
1292 | RRETURN(MATCH_NOMATCH); | |
1293 | ecode++; | |
1294 | break; | |
1295 | ||
1296 | case OP_NOT_WHITESPACE: | |
1297 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1298 | GETCHARINCTEST(c, eptr); | |
1299 | if ( | |
1300 | #ifdef SUPPORT_UTF8 | |
1301 | c < 256 && | |
1302 | #endif | |
1303 | (md->ctypes[c] & ctype_space) != 0 | |
1304 | ) | |
1305 | RRETURN(MATCH_NOMATCH); | |
1306 | ecode++; | |
1307 | break; | |
1308 | ||
1309 | case OP_WHITESPACE: | |
1310 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1311 | GETCHARINCTEST(c, eptr); | |
1312 | if ( | |
1313 | #ifdef SUPPORT_UTF8 | |
1314 | c >= 256 || | |
1315 | #endif | |
1316 | (md->ctypes[c] & ctype_space) == 0 | |
1317 | ) | |
1318 | RRETURN(MATCH_NOMATCH); | |
1319 | ecode++; | |
1320 | break; | |
1321 | ||
1322 | case OP_NOT_WORDCHAR: | |
1323 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1324 | GETCHARINCTEST(c, eptr); | |
1325 | if ( | |
1326 | #ifdef SUPPORT_UTF8 | |
1327 | c < 256 && | |
1328 | #endif | |
1329 | (md->ctypes[c] & ctype_word) != 0 | |
1330 | ) | |
1331 | RRETURN(MATCH_NOMATCH); | |
1332 | ecode++; | |
1333 | break; | |
1334 | ||
1335 | case OP_WORDCHAR: | |
1336 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1337 | GETCHARINCTEST(c, eptr); | |
1338 | if ( | |
1339 | #ifdef SUPPORT_UTF8 | |
1340 | c >= 256 || | |
1341 | #endif | |
1342 | (md->ctypes[c] & ctype_word) == 0 | |
1343 | ) | |
1344 | RRETURN(MATCH_NOMATCH); | |
1345 | ecode++; | |
1346 | break; | |
1347 | ||
1348 | #ifdef SUPPORT_UCP | |
1349 | /* Check the next character by Unicode property. We will get here only | |
1350 | if the support is in the binary; otherwise a compile-time error occurs. */ | |
1351 | ||
1352 | case OP_PROP: | |
1353 | case OP_NOTPROP: | |
1354 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1355 | GETCHARINCTEST(c, eptr); | |
1356 | { | |
1357 | int chartype, rqdtype; | |
1358 | int othercase; | |
1359 | int category = ucp_findchar(c, &chartype, &othercase); | |
1360 | ||
1361 | rqdtype = *(++ecode); | |
1362 | ecode++; | |
1363 | ||
1364 | if (rqdtype >= 128) | |
1365 | { | |
1366 | if ((rqdtype - 128 != category) == (op == OP_PROP)) | |
1367 | RRETURN(MATCH_NOMATCH); | |
1368 | } | |
1369 | else | |
1370 | { | |
1371 | if ((rqdtype != chartype) == (op == OP_PROP)) | |
1372 | RRETURN(MATCH_NOMATCH); | |
1373 | } | |
1374 | } | |
1375 | break; | |
1376 | ||
1377 | /* Match an extended Unicode sequence. We will get here only if the support | |
1378 | is in the binary; otherwise a compile-time error occurs. */ | |
1379 | ||
1380 | case OP_EXTUNI: | |
1381 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1382 | GETCHARINCTEST(c, eptr); | |
1383 | { | |
1384 | int chartype; | |
1385 | int othercase; | |
1386 | int category = ucp_findchar(c, &chartype, &othercase); | |
1387 | if (category == ucp_M) RRETURN(MATCH_NOMATCH); | |
1388 | while (eptr < md->end_subject) | |
1389 | { | |
1390 | int len = 1; | |
1391 | if (!utf8) c = *eptr; else | |
1392 | { | |
1393 | GETCHARLEN(c, eptr, len); | |
1394 | } | |
1395 | category = ucp_findchar(c, &chartype, &othercase); | |
1396 | if (category != ucp_M) break; | |
1397 | eptr += len; | |
1398 | } | |
1399 | } | |
1400 | ecode++; | |
1401 | break; | |
1402 | #endif | |
1403 | ||
1404 | ||
1405 | /* Match a back reference, possibly repeatedly. Look past the end of the | |
1406 | item to see if there is repeat information following. The code is similar | |
1407 | to that for character classes, but repeated for efficiency. Then obey | |
1408 | similar code to character type repeats - written out again for speed. | |
1409 | However, if the referenced string is the empty string, always treat | |
1410 | it as matched, any number of times (otherwise there could be infinite | |
1411 | loops). */ | |
1412 | ||
1413 | case OP_REF: | |
1414 | { | |
1415 | offset = GET2(ecode, 1) << 1; /* Doubled ref number */ | |
1416 | ecode += 3; /* Advance past item */ | |
1417 | ||
1418 | /* If the reference is unset, set the length to be longer than the amount | |
1419 | of subject left; this ensures that every attempt at a match fails. We | |
1420 | can't just fail here, because of the possibility of quantifiers with zero | |
1421 | minima. */ | |
1422 | ||
1423 | length = (offset >= offset_top || md->offset_vector[offset] < 0)? | |
1424 | md->end_subject - eptr + 1 : | |
1425 | md->offset_vector[offset+1] - md->offset_vector[offset]; | |
1426 | ||
1427 | /* Set up for repetition, or handle the non-repeated case */ | |
1428 | ||
1429 | switch (*ecode) | |
1430 | { | |
1431 | case OP_CRSTAR: | |
1432 | case OP_CRMINSTAR: | |
1433 | case OP_CRPLUS: | |
1434 | case OP_CRMINPLUS: | |
1435 | case OP_CRQUERY: | |
1436 | case OP_CRMINQUERY: | |
1437 | c = *ecode++ - OP_CRSTAR; | |
1438 | minimize = (c & 1) != 0; | |
1439 | min = rep_min[c]; /* Pick up values from tables; */ | |
1440 | max = rep_max[c]; /* zero for max => infinity */ | |
1441 | if (max == 0) max = INT_MAX; | |
1442 | break; | |
1443 | ||
1444 | case OP_CRRANGE: | |
1445 | case OP_CRMINRANGE: | |
1446 | minimize = (*ecode == OP_CRMINRANGE); | |
1447 | min = GET2(ecode, 1); | |
1448 | max = GET2(ecode, 3); | |
1449 | if (max == 0) max = INT_MAX; | |
1450 | ecode += 5; | |
1451 | break; | |
1452 | ||
1453 | default: /* No repeat follows */ | |
1454 | if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH); | |
1455 | eptr += length; | |
1456 | continue; /* With the main loop */ | |
1457 | } | |
1458 | ||
1459 | /* If the length of the reference is zero, just continue with the | |
1460 | main loop. */ | |
1461 | ||
1462 | if (length == 0) continue; | |
1463 | ||
1464 | /* First, ensure the minimum number of matches are present. We get back | |
1465 | the length of the reference string explicitly rather than passing the | |
1466 | address of eptr, so that eptr can be a register variable. */ | |
1467 | ||
1468 | for (i = 1; i <= min; i++) | |
1469 | { | |
1470 | if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH); | |
1471 | eptr += length; | |
1472 | } | |
1473 | ||
1474 | /* If min = max, continue at the same level without recursion. | |
1475 | They are not both allowed to be zero. */ | |
1476 | ||
1477 | if (min == max) continue; | |
1478 | ||
1479 | /* If minimizing, keep trying and advancing the pointer */ | |
1480 | ||
1481 | if (minimize) | |
1482 | { | |
1483 | for (fi = min;; fi++) | |
1484 | { | |
1485 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1486 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1487 | if (fi >= max || !match_ref(offset, eptr, length, md, ims)) | |
1488 | RRETURN(MATCH_NOMATCH); | |
1489 | eptr += length; | |
1490 | } | |
1491 | /* Control never gets here */ | |
1492 | } | |
1493 | ||
1494 | /* If maximizing, find the longest string and work backwards */ | |
1495 | ||
1496 | else | |
1497 | { | |
1498 | pp = eptr; | |
1499 | for (i = min; i < max; i++) | |
1500 | { | |
1501 | if (!match_ref(offset, eptr, length, md, ims)) break; | |
1502 | eptr += length; | |
1503 | } | |
1504 | while (eptr >= pp) | |
1505 | { | |
1506 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1507 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1508 | eptr -= length; | |
1509 | } | |
1510 | RRETURN(MATCH_NOMATCH); | |
1511 | } | |
1512 | } | |
1513 | /* Control never gets here */ | |
1514 | ||
1515 | ||
1516 | ||
1517 | /* Match a bit-mapped character class, possibly repeatedly. This op code is | |
1518 | used when all the characters in the class have values in the range 0-255, | |
1519 | and either the matching is caseful, or the characters are in the range | |
1520 | 0-127 when UTF-8 processing is enabled. The only difference between | |
1521 | OP_CLASS and OP_NCLASS occurs when a data character outside the range is | |
1522 | encountered. | |
1523 | ||
1524 | First, look past the end of the item to see if there is repeat information | |
1525 | following. Then obey similar code to character type repeats - written out | |
1526 | again for speed. */ | |
1527 | ||
1528 | case OP_NCLASS: | |
1529 | case OP_CLASS: | |
1530 | { | |
1531 | data = ecode + 1; /* Save for matching */ | |
1532 | ecode += 33; /* Advance past the item */ | |
1533 | ||
1534 | switch (*ecode) | |
1535 | { | |
1536 | case OP_CRSTAR: | |
1537 | case OP_CRMINSTAR: | |
1538 | case OP_CRPLUS: | |
1539 | case OP_CRMINPLUS: | |
1540 | case OP_CRQUERY: | |
1541 | case OP_CRMINQUERY: | |
1542 | c = *ecode++ - OP_CRSTAR; | |
1543 | minimize = (c & 1) != 0; | |
1544 | min = rep_min[c]; /* Pick up values from tables; */ | |
1545 | max = rep_max[c]; /* zero for max => infinity */ | |
1546 | if (max == 0) max = INT_MAX; | |
1547 | break; | |
1548 | ||
1549 | case OP_CRRANGE: | |
1550 | case OP_CRMINRANGE: | |
1551 | minimize = (*ecode == OP_CRMINRANGE); | |
1552 | min = GET2(ecode, 1); | |
1553 | max = GET2(ecode, 3); | |
1554 | if (max == 0) max = INT_MAX; | |
1555 | ecode += 5; | |
1556 | break; | |
1557 | ||
1558 | default: /* No repeat follows */ | |
1559 | min = max = 1; | |
1560 | break; | |
1561 | } | |
1562 | ||
1563 | /* First, ensure the minimum number of matches are present. */ | |
1564 | ||
1565 | #ifdef SUPPORT_UTF8 | |
1566 | /* UTF-8 mode */ | |
1567 | if (utf8) | |
1568 | { | |
1569 | for (i = 1; i <= min; i++) | |
1570 | { | |
1571 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1572 | GETCHARINC(c, eptr); | |
1573 | if (c > 255) | |
1574 | { | |
1575 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); | |
1576 | } | |
1577 | else | |
1578 | { | |
1579 | if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); | |
1580 | } | |
1581 | } | |
1582 | } | |
1583 | else | |
1584 | #endif | |
1585 | /* Not UTF-8 mode */ | |
1586 | { | |
1587 | for (i = 1; i <= min; i++) | |
1588 | { | |
1589 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1590 | c = *eptr++; | |
1591 | if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); | |
1592 | } | |
1593 | } | |
1594 | ||
1595 | /* If max == min we can continue with the main loop without the | |
1596 | need to recurse. */ | |
1597 | ||
1598 | if (min == max) continue; | |
1599 | ||
1600 | /* If minimizing, keep testing the rest of the expression and advancing | |
1601 | the pointer while it matches the class. */ | |
1602 | ||
1603 | if (minimize) | |
1604 | { | |
1605 | #ifdef SUPPORT_UTF8 | |
1606 | /* UTF-8 mode */ | |
1607 | if (utf8) | |
1608 | { | |
1609 | for (fi = min;; fi++) | |
1610 | { | |
1611 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1612 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1613 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1614 | GETCHARINC(c, eptr); | |
1615 | if (c > 255) | |
1616 | { | |
1617 | if (op == OP_CLASS) RRETURN(MATCH_NOMATCH); | |
1618 | } | |
1619 | else | |
1620 | { | |
1621 | if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); | |
1622 | } | |
1623 | } | |
1624 | } | |
1625 | else | |
1626 | #endif | |
1627 | /* Not UTF-8 mode */ | |
1628 | { | |
1629 | for (fi = min;; fi++) | |
1630 | { | |
1631 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1632 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1633 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1634 | c = *eptr++; | |
1635 | if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH); | |
1636 | } | |
1637 | } | |
1638 | /* Control never gets here */ | |
1639 | } | |
1640 | ||
1641 | /* If maximizing, find the longest possible run, then work backwards. */ | |
1642 | ||
1643 | else | |
1644 | { | |
1645 | pp = eptr; | |
1646 | ||
1647 | #ifdef SUPPORT_UTF8 | |
1648 | /* UTF-8 mode */ | |
1649 | if (utf8) | |
1650 | { | |
1651 | for (i = min; i < max; i++) | |
1652 | { | |
1653 | int len = 1; | |
1654 | if (eptr >= md->end_subject) break; | |
1655 | GETCHARLEN(c, eptr, len); | |
1656 | if (c > 255) | |
1657 | { | |
1658 | if (op == OP_CLASS) break; | |
1659 | } | |
1660 | else | |
1661 | { | |
1662 | if ((data[c/8] & (1 << (c&7))) == 0) break; | |
1663 | } | |
1664 | eptr += len; | |
1665 | } | |
1666 | for (;;) | |
1667 | { | |
1668 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1669 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1670 | if (eptr-- == pp) break; /* Stop if tried at original pos */ | |
1671 | BACKCHAR(eptr); | |
1672 | } | |
1673 | } | |
1674 | else | |
1675 | #endif | |
1676 | /* Not UTF-8 mode */ | |
1677 | { | |
1678 | for (i = min; i < max; i++) | |
1679 | { | |
1680 | if (eptr >= md->end_subject) break; | |
1681 | c = *eptr; | |
1682 | if ((data[c/8] & (1 << (c&7))) == 0) break; | |
1683 | eptr++; | |
1684 | } | |
1685 | while (eptr >= pp) | |
1686 | { | |
1687 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1688 | eptr--; | |
1689 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1690 | } | |
1691 | } | |
1692 | ||
1693 | RRETURN(MATCH_NOMATCH); | |
1694 | } | |
1695 | } | |
1696 | /* Control never gets here */ | |
1697 | ||
1698 | ||
1699 | /* Match an extended character class. This opcode is encountered only | |
1700 | in UTF-8 mode, because that's the only time it is compiled. */ | |
1701 | ||
1702 | #ifdef SUPPORT_UTF8 | |
1703 | case OP_XCLASS: | |
1704 | { | |
1705 | data = ecode + 1 + LINK_SIZE; /* Save for matching */ | |
1706 | ecode += GET(ecode, 1); /* Advance past the item */ | |
1707 | ||
1708 | switch (*ecode) | |
1709 | { | |
1710 | case OP_CRSTAR: | |
1711 | case OP_CRMINSTAR: | |
1712 | case OP_CRPLUS: | |
1713 | case OP_CRMINPLUS: | |
1714 | case OP_CRQUERY: | |
1715 | case OP_CRMINQUERY: | |
1716 | c = *ecode++ - OP_CRSTAR; | |
1717 | minimize = (c & 1) != 0; | |
1718 | min = rep_min[c]; /* Pick up values from tables; */ | |
1719 | max = rep_max[c]; /* zero for max => infinity */ | |
1720 | if (max == 0) max = INT_MAX; | |
1721 | break; | |
1722 | ||
1723 | case OP_CRRANGE: | |
1724 | case OP_CRMINRANGE: | |
1725 | minimize = (*ecode == OP_CRMINRANGE); | |
1726 | min = GET2(ecode, 1); | |
1727 | max = GET2(ecode, 3); | |
1728 | if (max == 0) max = INT_MAX; | |
1729 | ecode += 5; | |
1730 | break; | |
1731 | ||
1732 | default: /* No repeat follows */ | |
1733 | min = max = 1; | |
1734 | break; | |
1735 | } | |
1736 | ||
1737 | /* First, ensure the minimum number of matches are present. */ | |
1738 | ||
1739 | for (i = 1; i <= min; i++) | |
1740 | { | |
1741 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1742 | GETCHARINC(c, eptr); | |
1743 | if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH); | |
1744 | } | |
1745 | ||
1746 | /* If max == min we can continue with the main loop without the | |
1747 | need to recurse. */ | |
1748 | ||
1749 | if (min == max) continue; | |
1750 | ||
1751 | /* If minimizing, keep testing the rest of the expression and advancing | |
1752 | the pointer while it matches the class. */ | |
1753 | ||
1754 | if (minimize) | |
1755 | { | |
1756 | for (fi = min;; fi++) | |
1757 | { | |
1758 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1759 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1760 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1761 | GETCHARINC(c, eptr); | |
1762 | if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH); | |
1763 | } | |
1764 | /* Control never gets here */ | |
1765 | } | |
1766 | ||
1767 | /* If maximizing, find the longest possible run, then work backwards. */ | |
1768 | ||
1769 | else | |
1770 | { | |
1771 | pp = eptr; | |
1772 | for (i = min; i < max; i++) | |
1773 | { | |
1774 | int len = 1; | |
1775 | if (eptr >= md->end_subject) break; | |
1776 | GETCHARLEN(c, eptr, len); | |
1777 | if (!_pcre_xclass(c, data)) break; | |
1778 | eptr += len; | |
1779 | } | |
1780 | for(;;) | |
1781 | { | |
1782 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1783 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1784 | if (eptr-- == pp) break; /* Stop if tried at original pos */ | |
1785 | BACKCHAR(eptr) | |
1786 | } | |
1787 | RRETURN(MATCH_NOMATCH); | |
1788 | } | |
1789 | ||
1790 | /* Control never gets here */ | |
1791 | } | |
1792 | #endif /* End of XCLASS */ | |
1793 | ||
1794 | /* Match a single character, casefully */ | |
1795 | ||
1796 | case OP_CHAR: | |
1797 | #ifdef SUPPORT_UTF8 | |
1798 | if (utf8) | |
1799 | { | |
1800 | length = 1; | |
1801 | ecode++; | |
1802 | GETCHARLEN(fc, ecode, length); | |
1803 | if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH); | |
1804 | while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH); | |
1805 | } | |
1806 | else | |
1807 | #endif | |
1808 | ||
1809 | /* Non-UTF-8 mode */ | |
1810 | { | |
1811 | if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH); | |
1812 | if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH); | |
1813 | ecode += 2; | |
1814 | } | |
1815 | break; | |
1816 | ||
1817 | /* Match a single character, caselessly */ | |
1818 | ||
1819 | case OP_CHARNC: | |
1820 | #ifdef SUPPORT_UTF8 | |
1821 | if (utf8) | |
1822 | { | |
1823 | length = 1; | |
1824 | ecode++; | |
1825 | GETCHARLEN(fc, ecode, length); | |
1826 | ||
1827 | if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH); | |
1828 | ||
1829 | /* If the pattern character's value is < 128, we have only one byte, and | |
1830 | can use the fast lookup table. */ | |
1831 | ||
1832 | if (fc < 128) | |
1833 | { | |
1834 | if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH); | |
1835 | } | |
1836 | ||
1837 | /* Otherwise we must pick up the subject character */ | |
1838 | ||
1839 | else | |
1840 | { | |
1841 | int dc; | |
1842 | GETCHARINC(dc, eptr); | |
1843 | ecode += length; | |
1844 | ||
1845 | /* If we have Unicode property support, we can use it to test the other | |
1846 | case of the character, if there is one. The result of ucp_findchar() is | |
1847 | < 0 if the char isn't found, and othercase is returned as zero if there | |
1848 | isn't one. */ | |
1849 | ||
1850 | if (fc != dc) | |
1851 | { | |
1852 | #ifdef SUPPORT_UCP | |
1853 | int chartype; | |
1854 | int othercase; | |
1855 | if (ucp_findchar(fc, &chartype, &othercase) < 0 || dc != othercase) | |
1856 | #endif | |
1857 | RRETURN(MATCH_NOMATCH); | |
1858 | } | |
1859 | } | |
1860 | } | |
1861 | else | |
1862 | #endif /* SUPPORT_UTF8 */ | |
1863 | ||
1864 | /* Non-UTF-8 mode */ | |
1865 | { | |
1866 | if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH); | |
1867 | if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH); | |
1868 | ecode += 2; | |
1869 | } | |
1870 | break; | |
1871 | ||
1872 | /* Match a single character repeatedly; different opcodes share code. */ | |
1873 | ||
1874 | case OP_EXACT: | |
1875 | min = max = GET2(ecode, 1); | |
1876 | ecode += 3; | |
1877 | goto REPEATCHAR; | |
1878 | ||
1879 | case OP_UPTO: | |
1880 | case OP_MINUPTO: | |
1881 | min = 0; | |
1882 | max = GET2(ecode, 1); | |
1883 | minimize = *ecode == OP_MINUPTO; | |
1884 | ecode += 3; | |
1885 | goto REPEATCHAR; | |
1886 | ||
1887 | case OP_STAR: | |
1888 | case OP_MINSTAR: | |
1889 | case OP_PLUS: | |
1890 | case OP_MINPLUS: | |
1891 | case OP_QUERY: | |
1892 | case OP_MINQUERY: | |
1893 | c = *ecode++ - OP_STAR; | |
1894 | minimize = (c & 1) != 0; | |
1895 | min = rep_min[c]; /* Pick up values from tables; */ | |
1896 | max = rep_max[c]; /* zero for max => infinity */ | |
1897 | if (max == 0) max = INT_MAX; | |
1898 | ||
1899 | /* Common code for all repeated single-character matches. We can give | |
1900 | up quickly if there are fewer than the minimum number of characters left in | |
1901 | the subject. */ | |
1902 | ||
1903 | REPEATCHAR: | |
1904 | #ifdef SUPPORT_UTF8 | |
1905 | if (utf8) | |
1906 | { | |
1907 | length = 1; | |
1908 | charptr = ecode; | |
1909 | GETCHARLEN(fc, ecode, length); | |
1910 | if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH); | |
1911 | ecode += length; | |
1912 | ||
1913 | /* Handle multibyte character matching specially here. There is | |
1914 | support for caseless matching if UCP support is present. */ | |
1915 | ||
1916 | if (length > 1) | |
1917 | { | |
1918 | int oclength = 0; | |
1919 | uschar occhars[8]; | |
1920 | ||
1921 | #ifdef SUPPORT_UCP | |
1922 | int othercase; | |
1923 | int chartype; | |
1924 | if ((ims & PCRE_CASELESS) != 0 && | |
1925 | ucp_findchar(fc, &chartype, &othercase) >= 0 && | |
1926 | othercase > 0) | |
1927 | oclength = _pcre_ord2utf8(othercase, occhars); | |
1928 | #endif /* SUPPORT_UCP */ | |
1929 | ||
1930 | for (i = 1; i <= min; i++) | |
1931 | { | |
1932 | if (memcmp(eptr, charptr, length) == 0) eptr += length; | |
1933 | /* Need braces because of following else */ | |
1934 | else if (oclength == 0) { RRETURN(MATCH_NOMATCH); } | |
1935 | else | |
1936 | { | |
1937 | if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH); | |
1938 | eptr += oclength; | |
1939 | } | |
1940 | } | |
1941 | ||
1942 | if (min == max) continue; | |
1943 | ||
1944 | if (minimize) | |
1945 | { | |
1946 | for (fi = min;; fi++) | |
1947 | { | |
1948 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1949 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1950 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
1951 | if (memcmp(eptr, charptr, length) == 0) eptr += length; | |
1952 | /* Need braces because of following else */ | |
1953 | else if (oclength == 0) { RRETURN(MATCH_NOMATCH); } | |
1954 | else | |
1955 | { | |
1956 | if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH); | |
1957 | eptr += oclength; | |
1958 | } | |
1959 | } | |
1960 | /* Control never gets here */ | |
1961 | } | |
1962 | else | |
1963 | { | |
1964 | pp = eptr; | |
1965 | for (i = min; i < max; i++) | |
1966 | { | |
1967 | if (eptr > md->end_subject - length) break; | |
1968 | if (memcmp(eptr, charptr, length) == 0) eptr += length; | |
1969 | else if (oclength == 0) break; | |
1970 | else | |
1971 | { | |
1972 | if (memcmp(eptr, occhars, oclength) != 0) break; | |
1973 | eptr += oclength; | |
1974 | } | |
1975 | } | |
1976 | while (eptr >= pp) | |
1977 | { | |
1978 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
1979 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
1980 | eptr -= length; | |
1981 | } | |
1982 | RRETURN(MATCH_NOMATCH); | |
1983 | } | |
1984 | /* Control never gets here */ | |
1985 | } | |
1986 | ||
1987 | /* If the length of a UTF-8 character is 1, we fall through here, and | |
1988 | obey the code as for non-UTF-8 characters below, though in this case the | |
1989 | value of fc will always be < 128. */ | |
1990 | } | |
1991 | else | |
1992 | #endif /* SUPPORT_UTF8 */ | |
1993 | ||
1994 | /* When not in UTF-8 mode, load a single-byte character. */ | |
1995 | { | |
1996 | if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH); | |
1997 | fc = *ecode++; | |
1998 | } | |
1999 | ||
2000 | /* The value of fc at this point is always less than 256, though we may or | |
2001 | may not be in UTF-8 mode. The code is duplicated for the caseless and | |
2002 | caseful cases, for speed, since matching characters is likely to be quite | |
2003 | common. First, ensure the minimum number of matches are present. If min = | |
2004 | max, continue at the same level without recursing. Otherwise, if | |
2005 | minimizing, keep trying the rest of the expression and advancing one | |
2006 | matching character if failing, up to the maximum. Alternatively, if | |
2007 | maximizing, find the maximum number of characters and work backwards. */ | |
2008 | ||
2009 | DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max, | |
2010 | max, eptr)); | |
2011 | ||
2012 | if ((ims & PCRE_CASELESS) != 0) | |
2013 | { | |
2014 | fc = md->lcc[fc]; | |
2015 | for (i = 1; i <= min; i++) | |
2016 | if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH); | |
2017 | if (min == max) continue; | |
2018 | if (minimize) | |
2019 | { | |
2020 | for (fi = min;; fi++) | |
2021 | { | |
2022 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2023 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2024 | if (fi >= max || eptr >= md->end_subject || | |
2025 | fc != md->lcc[*eptr++]) | |
2026 | RRETURN(MATCH_NOMATCH); | |
2027 | } | |
2028 | /* Control never gets here */ | |
2029 | } | |
2030 | else | |
2031 | { | |
2032 | pp = eptr; | |
2033 | for (i = min; i < max; i++) | |
2034 | { | |
2035 | if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break; | |
2036 | eptr++; | |
2037 | } | |
2038 | while (eptr >= pp) | |
2039 | { | |
2040 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2041 | eptr--; | |
2042 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2043 | } | |
2044 | RRETURN(MATCH_NOMATCH); | |
2045 | } | |
2046 | /* Control never gets here */ | |
2047 | } | |
2048 | ||
2049 | /* Caseful comparisons (includes all multi-byte characters) */ | |
2050 | ||
2051 | else | |
2052 | { | |
2053 | for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH); | |
2054 | if (min == max) continue; | |
2055 | if (minimize) | |
2056 | { | |
2057 | for (fi = min;; fi++) | |
2058 | { | |
2059 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2060 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2061 | if (fi >= max || eptr >= md->end_subject || fc != *eptr++) | |
2062 | RRETURN(MATCH_NOMATCH); | |
2063 | } | |
2064 | /* Control never gets here */ | |
2065 | } | |
2066 | else | |
2067 | { | |
2068 | pp = eptr; | |
2069 | for (i = min; i < max; i++) | |
2070 | { | |
2071 | if (eptr >= md->end_subject || fc != *eptr) break; | |
2072 | eptr++; | |
2073 | } | |
2074 | while (eptr >= pp) | |
2075 | { | |
2076 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2077 | eptr--; | |
2078 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2079 | } | |
2080 | RRETURN(MATCH_NOMATCH); | |
2081 | } | |
2082 | } | |
2083 | /* Control never gets here */ | |
2084 | ||
2085 | /* Match a negated single one-byte character. The character we are | |
2086 | checking can be multibyte. */ | |
2087 | ||
2088 | case OP_NOT: | |
2089 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
2090 | ecode++; | |
2091 | GETCHARINCTEST(c, eptr); | |
2092 | if ((ims & PCRE_CASELESS) != 0) | |
2093 | { | |
2094 | #ifdef SUPPORT_UTF8 | |
2095 | if (c < 256) | |
2096 | #endif | |
2097 | c = md->lcc[c]; | |
2098 | if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH); | |
2099 | } | |
2100 | else | |
2101 | { | |
2102 | if (*ecode++ == c) RRETURN(MATCH_NOMATCH); | |
2103 | } | |
2104 | break; | |
2105 | ||
2106 | /* Match a negated single one-byte character repeatedly. This is almost a | |
2107 | repeat of the code for a repeated single character, but I haven't found a | |
2108 | nice way of commoning these up that doesn't require a test of the | |
2109 | positive/negative option for each character match. Maybe that wouldn't add | |
2110 | very much to the time taken, but character matching *is* what this is all | |
2111 | about... */ | |
2112 | ||
2113 | case OP_NOTEXACT: | |
2114 | min = max = GET2(ecode, 1); | |
2115 | ecode += 3; | |
2116 | goto REPEATNOTCHAR; | |
2117 | ||
2118 | case OP_NOTUPTO: | |
2119 | case OP_NOTMINUPTO: | |
2120 | min = 0; | |
2121 | max = GET2(ecode, 1); | |
2122 | minimize = *ecode == OP_NOTMINUPTO; | |
2123 | ecode += 3; | |
2124 | goto REPEATNOTCHAR; | |
2125 | ||
2126 | case OP_NOTSTAR: | |
2127 | case OP_NOTMINSTAR: | |
2128 | case OP_NOTPLUS: | |
2129 | case OP_NOTMINPLUS: | |
2130 | case OP_NOTQUERY: | |
2131 | case OP_NOTMINQUERY: | |
2132 | c = *ecode++ - OP_NOTSTAR; | |
2133 | minimize = (c & 1) != 0; | |
2134 | min = rep_min[c]; /* Pick up values from tables; */ | |
2135 | max = rep_max[c]; /* zero for max => infinity */ | |
2136 | if (max == 0) max = INT_MAX; | |
2137 | ||
2138 | /* Common code for all repeated single-byte matches. We can give up quickly | |
2139 | if there are fewer than the minimum number of bytes left in the | |
2140 | subject. */ | |
2141 | ||
2142 | REPEATNOTCHAR: | |
2143 | if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH); | |
2144 | fc = *ecode++; | |
2145 | ||
2146 | /* The code is duplicated for the caseless and caseful cases, for speed, | |
2147 | since matching characters is likely to be quite common. First, ensure the | |
2148 | minimum number of matches are present. If min = max, continue at the same | |
2149 | level without recursing. Otherwise, if minimizing, keep trying the rest of | |
2150 | the expression and advancing one matching character if failing, up to the | |
2151 | maximum. Alternatively, if maximizing, find the maximum number of | |
2152 | characters and work backwards. */ | |
2153 | ||
2154 | DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max, | |
2155 | max, eptr)); | |
2156 | ||
2157 | if ((ims & PCRE_CASELESS) != 0) | |
2158 | { | |
2159 | fc = md->lcc[fc]; | |
2160 | ||
2161 | #ifdef SUPPORT_UTF8 | |
2162 | /* UTF-8 mode */ | |
2163 | if (utf8) | |
2164 | { | |
2165 | register int d; | |
2166 | for (i = 1; i <= min; i++) | |
2167 | { | |
2168 | GETCHARINC(d, eptr); | |
2169 | if (d < 256) d = md->lcc[d]; | |
2170 | if (fc == d) RRETURN(MATCH_NOMATCH); | |
2171 | } | |
2172 | } | |
2173 | else | |
2174 | #endif | |
2175 | ||
2176 | /* Not UTF-8 mode */ | |
2177 | { | |
2178 | for (i = 1; i <= min; i++) | |
2179 | if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH); | |
2180 | } | |
2181 | ||
2182 | if (min == max) continue; | |
2183 | ||
2184 | if (minimize) | |
2185 | { | |
2186 | #ifdef SUPPORT_UTF8 | |
2187 | /* UTF-8 mode */ | |
2188 | if (utf8) | |
2189 | { | |
2190 | register int d; | |
2191 | for (fi = min;; fi++) | |
2192 | { | |
2193 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2194 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2195 | GETCHARINC(d, eptr); | |
2196 | if (d < 256) d = md->lcc[d]; | |
2197 | if (fi >= max || eptr >= md->end_subject || fc == d) | |
2198 | RRETURN(MATCH_NOMATCH); | |
2199 | } | |
2200 | } | |
2201 | else | |
2202 | #endif | |
2203 | /* Not UTF-8 mode */ | |
2204 | { | |
2205 | for (fi = min;; fi++) | |
2206 | { | |
2207 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2208 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2209 | if (fi >= max || eptr >= md->end_subject || fc == md->lcc[*eptr++]) | |
2210 | RRETURN(MATCH_NOMATCH); | |
2211 | } | |
2212 | } | |
2213 | /* Control never gets here */ | |
2214 | } | |
2215 | ||
2216 | /* Maximize case */ | |
2217 | ||
2218 | else | |
2219 | { | |
2220 | pp = eptr; | |
2221 | ||
2222 | #ifdef SUPPORT_UTF8 | |
2223 | /* UTF-8 mode */ | |
2224 | if (utf8) | |
2225 | { | |
2226 | register int d; | |
2227 | for (i = min; i < max; i++) | |
2228 | { | |
2229 | int len = 1; | |
2230 | if (eptr >= md->end_subject) break; | |
2231 | GETCHARLEN(d, eptr, len); | |
2232 | if (d < 256) d = md->lcc[d]; | |
2233 | if (fc == d) break; | |
2234 | eptr += len; | |
2235 | } | |
2236 | for(;;) | |
2237 | { | |
2238 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2239 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2240 | if (eptr-- == pp) break; /* Stop if tried at original pos */ | |
2241 | BACKCHAR(eptr); | |
2242 | } | |
2243 | } | |
2244 | else | |
2245 | #endif | |
2246 | /* Not UTF-8 mode */ | |
2247 | { | |
2248 | for (i = min; i < max; i++) | |
2249 | { | |
2250 | if (eptr >= md->end_subject || fc == md->lcc[*eptr]) break; | |
2251 | eptr++; | |
2252 | } | |
2253 | while (eptr >= pp) | |
2254 | { | |
2255 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2256 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2257 | eptr--; | |
2258 | } | |
2259 | } | |
2260 | ||
2261 | RRETURN(MATCH_NOMATCH); | |
2262 | } | |
2263 | /* Control never gets here */ | |
2264 | } | |
2265 | ||
2266 | /* Caseful comparisons */ | |
2267 | ||
2268 | else | |
2269 | { | |
2270 | #ifdef SUPPORT_UTF8 | |
2271 | /* UTF-8 mode */ | |
2272 | if (utf8) | |
2273 | { | |
2274 | register int d; | |
2275 | for (i = 1; i <= min; i++) | |
2276 | { | |
2277 | GETCHARINC(d, eptr); | |
2278 | if (fc == d) RRETURN(MATCH_NOMATCH); | |
2279 | } | |
2280 | } | |
2281 | else | |
2282 | #endif | |
2283 | /* Not UTF-8 mode */ | |
2284 | { | |
2285 | for (i = 1; i <= min; i++) | |
2286 | if (fc == *eptr++) RRETURN(MATCH_NOMATCH); | |
2287 | } | |
2288 | ||
2289 | if (min == max) continue; | |
2290 | ||
2291 | if (minimize) | |
2292 | { | |
2293 | #ifdef SUPPORT_UTF8 | |
2294 | /* UTF-8 mode */ | |
2295 | if (utf8) | |
2296 | { | |
2297 | register int d; | |
2298 | for (fi = min;; fi++) | |
2299 | { | |
2300 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2301 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2302 | GETCHARINC(d, eptr); | |
2303 | if (fi >= max || eptr >= md->end_subject || fc == d) | |
2304 | RRETURN(MATCH_NOMATCH); | |
2305 | } | |
2306 | } | |
2307 | else | |
2308 | #endif | |
2309 | /* Not UTF-8 mode */ | |
2310 | { | |
2311 | for (fi = min;; fi++) | |
2312 | { | |
2313 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2314 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2315 | if (fi >= max || eptr >= md->end_subject || fc == *eptr++) | |
2316 | RRETURN(MATCH_NOMATCH); | |
2317 | } | |
2318 | } | |
2319 | /* Control never gets here */ | |
2320 | } | |
2321 | ||
2322 | /* Maximize case */ | |
2323 | ||
2324 | else | |
2325 | { | |
2326 | pp = eptr; | |
2327 | ||
2328 | #ifdef SUPPORT_UTF8 | |
2329 | /* UTF-8 mode */ | |
2330 | if (utf8) | |
2331 | { | |
2332 | register int d; | |
2333 | for (i = min; i < max; i++) | |
2334 | { | |
2335 | int len = 1; | |
2336 | if (eptr >= md->end_subject) break; | |
2337 | GETCHARLEN(d, eptr, len); | |
2338 | if (fc == d) break; | |
2339 | eptr += len; | |
2340 | } | |
2341 | for(;;) | |
2342 | { | |
2343 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2344 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2345 | if (eptr-- == pp) break; /* Stop if tried at original pos */ | |
2346 | BACKCHAR(eptr); | |
2347 | } | |
2348 | } | |
2349 | else | |
2350 | #endif | |
2351 | /* Not UTF-8 mode */ | |
2352 | { | |
2353 | for (i = min; i < max; i++) | |
2354 | { | |
2355 | if (eptr >= md->end_subject || fc == *eptr) break; | |
2356 | eptr++; | |
2357 | } | |
2358 | while (eptr >= pp) | |
2359 | { | |
2360 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2361 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2362 | eptr--; | |
2363 | } | |
2364 | } | |
2365 | ||
2366 | RRETURN(MATCH_NOMATCH); | |
2367 | } | |
2368 | } | |
2369 | /* Control never gets here */ | |
2370 | ||
2371 | /* Match a single character type repeatedly; several different opcodes | |
2372 | share code. This is very similar to the code for single characters, but we | |
2373 | repeat it in the interests of efficiency. */ | |
2374 | ||
2375 | case OP_TYPEEXACT: | |
2376 | min = max = GET2(ecode, 1); | |
2377 | minimize = TRUE; | |
2378 | ecode += 3; | |
2379 | goto REPEATTYPE; | |
2380 | ||
2381 | case OP_TYPEUPTO: | |
2382 | case OP_TYPEMINUPTO: | |
2383 | min = 0; | |
2384 | max = GET2(ecode, 1); | |
2385 | minimize = *ecode == OP_TYPEMINUPTO; | |
2386 | ecode += 3; | |
2387 | goto REPEATTYPE; | |
2388 | ||
2389 | case OP_TYPESTAR: | |
2390 | case OP_TYPEMINSTAR: | |
2391 | case OP_TYPEPLUS: | |
2392 | case OP_TYPEMINPLUS: | |
2393 | case OP_TYPEQUERY: | |
2394 | case OP_TYPEMINQUERY: | |
2395 | c = *ecode++ - OP_TYPESTAR; | |
2396 | minimize = (c & 1) != 0; | |
2397 | min = rep_min[c]; /* Pick up values from tables; */ | |
2398 | max = rep_max[c]; /* zero for max => infinity */ | |
2399 | if (max == 0) max = INT_MAX; | |
2400 | ||
2401 | /* Common code for all repeated single character type matches. Note that | |
2402 | in UTF-8 mode, '.' matches a character of any length, but for the other | |
2403 | character types, the valid characters are all one-byte long. */ | |
2404 | ||
2405 | REPEATTYPE: | |
2406 | ctype = *ecode++; /* Code for the character type */ | |
2407 | ||
2408 | #ifdef SUPPORT_UCP | |
2409 | if (ctype == OP_PROP || ctype == OP_NOTPROP) | |
2410 | { | |
2411 | prop_fail_result = ctype == OP_NOTPROP; | |
2412 | prop_type = *ecode++; | |
2413 | if (prop_type >= 128) | |
2414 | { | |
2415 | prop_test_against = prop_type - 128; | |
2416 | prop_test_variable = &prop_category; | |
2417 | } | |
2418 | else | |
2419 | { | |
2420 | prop_test_against = prop_type; | |
2421 | prop_test_variable = &prop_chartype; | |
2422 | } | |
2423 | } | |
2424 | else prop_type = -1; | |
2425 | #endif | |
2426 | ||
2427 | /* First, ensure the minimum number of matches are present. Use inline | |
2428 | code for maximizing the speed, and do the type test once at the start | |
2429 | (i.e. keep it out of the loop). Also we can test that there are at least | |
2430 | the minimum number of bytes before we start. This isn't as effective in | |
2431 | UTF-8 mode, but it does no harm. Separate the UTF-8 code completely as that | |
2432 | is tidier. Also separate the UCP code, which can be the same for both UTF-8 | |
2433 | and single-bytes. */ | |
2434 | ||
2435 | if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH); | |
2436 | if (min > 0) | |
2437 | { | |
2438 | #ifdef SUPPORT_UCP | |
2439 | if (prop_type > 0) | |
2440 | { | |
2441 | for (i = 1; i <= min; i++) | |
2442 | { | |
2443 | GETCHARINC(c, eptr); | |
2444 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2445 | if ((*prop_test_variable == prop_test_against) == prop_fail_result) | |
2446 | RRETURN(MATCH_NOMATCH); | |
2447 | } | |
2448 | } | |
2449 | ||
2450 | /* Match extended Unicode sequences. We will get here only if the | |
2451 | support is in the binary; otherwise a compile-time error occurs. */ | |
2452 | ||
2453 | else if (ctype == OP_EXTUNI) | |
2454 | { | |
2455 | for (i = 1; i <= min; i++) | |
2456 | { | |
2457 | GETCHARINCTEST(c, eptr); | |
2458 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2459 | if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH); | |
2460 | while (eptr < md->end_subject) | |
2461 | { | |
2462 | int len = 1; | |
2463 | if (!utf8) c = *eptr; else | |
2464 | { | |
2465 | GETCHARLEN(c, eptr, len); | |
2466 | } | |
2467 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2468 | if (prop_category != ucp_M) break; | |
2469 | eptr += len; | |
2470 | } | |
2471 | } | |
2472 | } | |
2473 | ||
2474 | else | |
2475 | #endif /* SUPPORT_UCP */ | |
2476 | ||
2477 | /* Handle all other cases when the coding is UTF-8 */ | |
2478 | ||
2479 | #ifdef SUPPORT_UTF8 | |
2480 | if (utf8) switch(ctype) | |
2481 | { | |
2482 | case OP_ANY: | |
2483 | for (i = 1; i <= min; i++) | |
2484 | { | |
2485 | if (eptr >= md->end_subject || | |
2486 | (*eptr++ == NEWLINE && (ims & PCRE_DOTALL) == 0)) | |
2487 | RRETURN(MATCH_NOMATCH); | |
2488 | while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++; | |
2489 | } | |
2490 | break; | |
2491 | ||
2492 | case OP_ANYBYTE: | |
2493 | eptr += min; | |
2494 | break; | |
2495 | ||
2496 | case OP_NOT_DIGIT: | |
2497 | for (i = 1; i <= min; i++) | |
2498 | { | |
2499 | if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
2500 | GETCHARINC(c, eptr); | |
2501 | if (c < 128 && (md->ctypes[c] & ctype_digit) != 0) | |
2502 | RRETURN(MATCH_NOMATCH); | |
2503 | } | |
2504 | break; | |
2505 | ||
2506 | case OP_DIGIT: | |
2507 | for (i = 1; i <= min; i++) | |
2508 | { | |
2509 | if (eptr >= md->end_subject || | |
2510 | *eptr >= 128 || (md->ctypes[*eptr++] & ctype_digit) == 0) | |
2511 | RRETURN(MATCH_NOMATCH); | |
2512 | /* No need to skip more bytes - we know it's a 1-byte character */ | |
2513 | } | |
2514 | break; | |
2515 | ||
2516 | case OP_NOT_WHITESPACE: | |
2517 | for (i = 1; i <= min; i++) | |
2518 | { | |
2519 | if (eptr >= md->end_subject || | |
2520 | (*eptr < 128 && (md->ctypes[*eptr++] & ctype_space) != 0)) | |
2521 | RRETURN(MATCH_NOMATCH); | |
2522 | while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++; | |
2523 | } | |
2524 | break; | |
2525 | ||
2526 | case OP_WHITESPACE: | |
2527 | for (i = 1; i <= min; i++) | |
2528 | { | |
2529 | if (eptr >= md->end_subject || | |
2530 | *eptr >= 128 || (md->ctypes[*eptr++] & ctype_space) == 0) | |
2531 | RRETURN(MATCH_NOMATCH); | |
2532 | /* No need to skip more bytes - we know it's a 1-byte character */ | |
2533 | } | |
2534 | break; | |
2535 | ||
2536 | case OP_NOT_WORDCHAR: | |
2537 | for (i = 1; i <= min; i++) | |
2538 | { | |
2539 | if (eptr >= md->end_subject || | |
2540 | (*eptr < 128 && (md->ctypes[*eptr++] & ctype_word) != 0)) | |
2541 | RRETURN(MATCH_NOMATCH); | |
2542 | while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++; | |
2543 | } | |
2544 | break; | |
2545 | ||
2546 | case OP_WORDCHAR: | |
2547 | for (i = 1; i <= min; i++) | |
2548 | { | |
2549 | if (eptr >= md->end_subject || | |
2550 | *eptr >= 128 || (md->ctypes[*eptr++] & ctype_word) == 0) | |
2551 | RRETURN(MATCH_NOMATCH); | |
2552 | /* No need to skip more bytes - we know it's a 1-byte character */ | |
2553 | } | |
2554 | break; | |
2555 | ||
2556 | default: | |
2557 | RRETURN(PCRE_ERROR_INTERNAL); | |
2558 | } /* End switch(ctype) */ | |
2559 | ||
2560 | else | |
2561 | #endif /* SUPPORT_UTF8 */ | |
2562 | ||
2563 | /* Code for the non-UTF-8 case for minimum matching of operators other | |
2564 | than OP_PROP and OP_NOTPROP. */ | |
2565 | ||
2566 | switch(ctype) | |
2567 | { | |
2568 | case OP_ANY: | |
2569 | if ((ims & PCRE_DOTALL) == 0) | |
2570 | { | |
2571 | for (i = 1; i <= min; i++) | |
2572 | if (*eptr++ == NEWLINE) RRETURN(MATCH_NOMATCH); | |
2573 | } | |
2574 | else eptr += min; | |
2575 | break; | |
2576 | ||
2577 | case OP_ANYBYTE: | |
2578 | eptr += min; | |
2579 | break; | |
2580 | ||
2581 | case OP_NOT_DIGIT: | |
2582 | for (i = 1; i <= min; i++) | |
2583 | if ((md->ctypes[*eptr++] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH); | |
2584 | break; | |
2585 | ||
2586 | case OP_DIGIT: | |
2587 | for (i = 1; i <= min; i++) | |
2588 | if ((md->ctypes[*eptr++] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH); | |
2589 | break; | |
2590 | ||
2591 | case OP_NOT_WHITESPACE: | |
2592 | for (i = 1; i <= min; i++) | |
2593 | if ((md->ctypes[*eptr++] & ctype_space) != 0) RRETURN(MATCH_NOMATCH); | |
2594 | break; | |
2595 | ||
2596 | case OP_WHITESPACE: | |
2597 | for (i = 1; i <= min; i++) | |
2598 | if ((md->ctypes[*eptr++] & ctype_space) == 0) RRETURN(MATCH_NOMATCH); | |
2599 | break; | |
2600 | ||
2601 | case OP_NOT_WORDCHAR: | |
2602 | for (i = 1; i <= min; i++) | |
2603 | if ((md->ctypes[*eptr++] & ctype_word) != 0) | |
2604 | RRETURN(MATCH_NOMATCH); | |
2605 | break; | |
2606 | ||
2607 | case OP_WORDCHAR: | |
2608 | for (i = 1; i <= min; i++) | |
2609 | if ((md->ctypes[*eptr++] & ctype_word) == 0) | |
2610 | RRETURN(MATCH_NOMATCH); | |
2611 | break; | |
2612 | ||
2613 | default: | |
2614 | RRETURN(PCRE_ERROR_INTERNAL); | |
2615 | } | |
2616 | } | |
2617 | ||
2618 | /* If min = max, continue at the same level without recursing */ | |
2619 | ||
2620 | if (min == max) continue; | |
2621 | ||
2622 | /* If minimizing, we have to test the rest of the pattern before each | |
2623 | subsequent match. Again, separate the UTF-8 case for speed, and also | |
2624 | separate the UCP cases. */ | |
2625 | ||
2626 | if (minimize) | |
2627 | { | |
2628 | #ifdef SUPPORT_UCP | |
2629 | if (prop_type > 0) | |
2630 | { | |
2631 | for (fi = min;; fi++) | |
2632 | { | |
2633 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2634 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2635 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
2636 | GETCHARINC(c, eptr); | |
2637 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2638 | if ((*prop_test_variable == prop_test_against) == prop_fail_result) | |
2639 | RRETURN(MATCH_NOMATCH); | |
2640 | } | |
2641 | } | |
2642 | ||
2643 | /* Match extended Unicode sequences. We will get here only if the | |
2644 | support is in the binary; otherwise a compile-time error occurs. */ | |
2645 | ||
2646 | else if (ctype == OP_EXTUNI) | |
2647 | { | |
2648 | for (fi = min;; fi++) | |
2649 | { | |
2650 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2651 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2652 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
2653 | GETCHARINCTEST(c, eptr); | |
2654 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2655 | if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH); | |
2656 | while (eptr < md->end_subject) | |
2657 | { | |
2658 | int len = 1; | |
2659 | if (!utf8) c = *eptr; else | |
2660 | { | |
2661 | GETCHARLEN(c, eptr, len); | |
2662 | } | |
2663 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2664 | if (prop_category != ucp_M) break; | |
2665 | eptr += len; | |
2666 | } | |
2667 | } | |
2668 | } | |
2669 | ||
2670 | else | |
2671 | #endif /* SUPPORT_UCP */ | |
2672 | ||
2673 | #ifdef SUPPORT_UTF8 | |
2674 | /* UTF-8 mode */ | |
2675 | if (utf8) | |
2676 | { | |
2677 | for (fi = min;; fi++) | |
2678 | { | |
2679 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2680 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2681 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
2682 | ||
2683 | GETCHARINC(c, eptr); | |
2684 | switch(ctype) | |
2685 | { | |
2686 | case OP_ANY: | |
2687 | if ((ims & PCRE_DOTALL) == 0 && c == NEWLINE) RRETURN(MATCH_NOMATCH); | |
2688 | break; | |
2689 | ||
2690 | case OP_ANYBYTE: | |
2691 | break; | |
2692 | ||
2693 | case OP_NOT_DIGIT: | |
2694 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) | |
2695 | RRETURN(MATCH_NOMATCH); | |
2696 | break; | |
2697 | ||
2698 | case OP_DIGIT: | |
2699 | if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0) | |
2700 | RRETURN(MATCH_NOMATCH); | |
2701 | break; | |
2702 | ||
2703 | case OP_NOT_WHITESPACE: | |
2704 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) | |
2705 | RRETURN(MATCH_NOMATCH); | |
2706 | break; | |
2707 | ||
2708 | case OP_WHITESPACE: | |
2709 | if (c >= 256 || (md->ctypes[c] & ctype_space) == 0) | |
2710 | RRETURN(MATCH_NOMATCH); | |
2711 | break; | |
2712 | ||
2713 | case OP_NOT_WORDCHAR: | |
2714 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) | |
2715 | RRETURN(MATCH_NOMATCH); | |
2716 | break; | |
2717 | ||
2718 | case OP_WORDCHAR: | |
2719 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) | |
2720 | RRETURN(MATCH_NOMATCH); | |
2721 | break; | |
2722 | ||
2723 | default: | |
2724 | RRETURN(PCRE_ERROR_INTERNAL); | |
2725 | } | |
2726 | } | |
2727 | } | |
2728 | else | |
2729 | #endif | |
2730 | /* Not UTF-8 mode */ | |
2731 | { | |
2732 | for (fi = min;; fi++) | |
2733 | { | |
2734 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2735 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2736 | if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH); | |
2737 | c = *eptr++; | |
2738 | switch(ctype) | |
2739 | { | |
2740 | case OP_ANY: | |
2741 | if ((ims & PCRE_DOTALL) == 0 && c == NEWLINE) RRETURN(MATCH_NOMATCH); | |
2742 | break; | |
2743 | ||
2744 | case OP_ANYBYTE: | |
2745 | break; | |
2746 | ||
2747 | case OP_NOT_DIGIT: | |
2748 | if ((md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH); | |
2749 | break; | |
2750 | ||
2751 | case OP_DIGIT: | |
2752 | if ((md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH); | |
2753 | break; | |
2754 | ||
2755 | case OP_NOT_WHITESPACE: | |
2756 | if ((md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH); | |
2757 | break; | |
2758 | ||
2759 | case OP_WHITESPACE: | |
2760 | if ((md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH); | |
2761 | break; | |
2762 | ||
2763 | case OP_NOT_WORDCHAR: | |
2764 | if ((md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH); | |
2765 | break; | |
2766 | ||
2767 | case OP_WORDCHAR: | |
2768 | if ((md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH); | |
2769 | break; | |
2770 | ||
2771 | default: | |
2772 | RRETURN(PCRE_ERROR_INTERNAL); | |
2773 | } | |
2774 | } | |
2775 | } | |
2776 | /* Control never gets here */ | |
2777 | } | |
2778 | ||
2779 | /* If maximizing it is worth using inline code for speed, doing the type | |
2780 | test once at the start (i.e. keep it out of the loop). Again, keep the | |
2781 | UTF-8 and UCP stuff separate. */ | |
2782 | ||
2783 | else | |
2784 | { | |
2785 | pp = eptr; /* Remember where we started */ | |
2786 | ||
2787 | #ifdef SUPPORT_UCP | |
2788 | if (prop_type > 0) | |
2789 | { | |
2790 | for (i = min; i < max; i++) | |
2791 | { | |
2792 | int len = 1; | |
2793 | if (eptr >= md->end_subject) break; | |
2794 | GETCHARLEN(c, eptr, len); | |
2795 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2796 | if ((*prop_test_variable == prop_test_against) == prop_fail_result) | |
2797 | break; | |
2798 | eptr+= len; | |
2799 | } | |
2800 | ||
2801 | /* eptr is now past the end of the maximum run */ | |
2802 | ||
2803 | for(;;) | |
2804 | { | |
2805 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2806 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2807 | if (eptr-- == pp) break; /* Stop if tried at original pos */ | |
2808 | BACKCHAR(eptr); | |
2809 | } | |
2810 | } | |
2811 | ||
2812 | /* Match extended Unicode sequences. We will get here only if the | |
2813 | support is in the binary; otherwise a compile-time error occurs. */ | |
2814 | ||
2815 | else if (ctype == OP_EXTUNI) | |
2816 | { | |
2817 | for (i = min; i < max; i++) | |
2818 | { | |
2819 | if (eptr >= md->end_subject) break; | |
2820 | GETCHARINCTEST(c, eptr); | |
2821 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2822 | if (prop_category == ucp_M) break; | |
2823 | while (eptr < md->end_subject) | |
2824 | { | |
2825 | int len = 1; | |
2826 | if (!utf8) c = *eptr; else | |
2827 | { | |
2828 | GETCHARLEN(c, eptr, len); | |
2829 | } | |
2830 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2831 | if (prop_category != ucp_M) break; | |
2832 | eptr += len; | |
2833 | } | |
2834 | } | |
2835 | ||
2836 | /* eptr is now past the end of the maximum run */ | |
2837 | ||
2838 | for(;;) | |
2839 | { | |
2840 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
2841 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
2842 | if (eptr-- == pp) break; /* Stop if tried at original pos */ | |
2843 | for (;;) /* Move back over one extended */ | |
2844 | { | |
2845 | int len = 1; | |
2846 | BACKCHAR(eptr); | |
2847 | if (!utf8) c = *eptr; else | |
2848 | { | |
2849 | GETCHARLEN(c, eptr, len); | |
2850 | } | |
2851 | prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase); | |
2852 | if (prop_category != ucp_M) break; | |
2853 | eptr--; | |
2854 | } | |
2855 | } | |
2856 | } | |
2857 | ||
2858 | else | |
2859 | #endif /* SUPPORT_UCP */ | |
2860 | ||
2861 | #ifdef SUPPORT_UTF8 | |
2862 | /* UTF-8 mode */ | |
2863 | ||
2864 | if (utf8) | |
2865 | { | |
2866 | switch(ctype) | |
2867 | { | |
2868 | case OP_ANY: | |
2869 | ||
2870 | /* Special code is required for UTF8, but when the maximum is unlimited | |
2871 | we don't need it, so we repeat the non-UTF8 code. This is probably | |
2872 | worth it, because .* is quite a common idiom. */ | |
2873 | ||
2874 | if (max < INT_MAX) | |
2875 | { | |
2876 | if ((ims & PCRE_DOTALL) == 0) | |
2877 | { | |
2878 | for (i = min; i < max; i++) | |
2879 | { | |
2880 | if (eptr >= md->end_subject || *eptr == NEWLINE) break; | |
2881 | eptr++; | |
2882 | while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++; | |
2883 | } | |
2884 | } | |
2885 | else | |
2886 | { | |
2887 | for (i = min; i < max; i++) | |
2888 | { | |
2889 | eptr++; | |
2890 | while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++; | |
2891 | } | |
2892 | } | |
2893 | } | |
2894 | ||
2895 | /* Handle unlimited UTF-8 repeat */ | |
2896 | ||
2897 | else | |
2898 | { | |
2899 | if ((ims & PCRE_DOTALL) == 0) | |
2900 | { | |
2901 | for (i = min; i < max; i++) | |
2902 | { | |
2903 | if (eptr >= md->end_subject || *eptr == NEWLINE) break; | |
2904 | eptr++; | |
2905 | } | |
2906 | break; | |
2907 | } | |
2908 | else | |
2909 | { | |
2910 | c = max - min; | |
2911 | if (c > md->end_subject - eptr) c = md->end_subject - eptr; | |
2912 | eptr += c; | |
2913 | } | |
2914 | } | |
2915 | break; | |
2916 | ||
2917 | /* The byte case is the same as non-UTF8 */ | |
2918 | ||
2919 | case OP_ANYBYTE: | |
2920 | c = max - min; | |
2921 | if (c > md->end_subject - eptr) c = md->end_subject - eptr; | |
2922 | eptr += c; | |
2923 | break; | |
2924 | ||
2925 | case OP_NOT_DIGIT: | |
2926 | for (i = min; i < max; i++) | |
2927 | { | |
2928 | int len = 1; | |
2929 | if (eptr >= md->end_subject) break; | |
2930 | GETCHARLEN(c, eptr, len); | |
2931 | if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break; | |
2932 | eptr+= len; | |
2933 | } | |
2934 | break; | |
2935 | ||
2936 | case OP_DIGIT: | |
2937 | for (i = min; i < max; i++) | |
2938 | { | |
2939 | int len = 1; | |
2940 | if (eptr >= md->end_subject) break; | |
2941 | GETCHARLEN(c, eptr, len); | |
2942 | if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break; | |
2943 | eptr+= len; | |
2944 | } | |
2945 | break; | |
2946 | ||
2947 | case OP_NOT_WHITESPACE: | |
2948 | for (i = min; i < max; i++) | |
2949 | { | |
2950 | int len = 1; | |
2951 | if (eptr >= md->end_subject) break; | |
2952 | GETCHARLEN(c, eptr, len); | |
2953 | if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break; | |
2954 | eptr+= len; | |
2955 | } | |
2956 | break; | |
2957 | ||
2958 | case OP_WHITESPACE: | |
2959 | for (i = min; i < max; i++) | |
2960 | { | |
2961 | int len = 1; | |
2962 | if (eptr >= md->end_subject) break; | |
2963 | GETCHARLEN(c, eptr, len); | |
2964 | if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break; | |
2965 | eptr+= len; | |
2966 | } | |
2967 | break; | |
2968 | ||
2969 | case OP_NOT_WORDCHAR: | |
2970 | for (i = min; i < max; i++) | |
2971 | { | |
2972 | int len = 1; | |
2973 | if (eptr >= md->end_subject) break; | |
2974 | GETCHARLEN(c, eptr, len); | |
2975 | if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break; | |
2976 | eptr+= len; | |
2977 | } | |
2978 | break; | |
2979 | ||
2980 | case OP_WORDCHAR: | |
2981 | for (i = min; i < max; i++) | |
2982 | { | |
2983 | int len = 1; | |
2984 | if (eptr >= md->end_subject) break; | |
2985 | GETCHARLEN(c, eptr, len); | |
2986 | if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break; | |
2987 | eptr+= len; | |
2988 | } | |
2989 | break; | |
2990 | ||
2991 | default: | |
2992 | RRETURN(PCRE_ERROR_INTERNAL); | |
2993 | } | |
2994 | ||
2995 | /* eptr is now past the end of the maximum run */ | |
2996 | ||
2997 | for(;;) | |
2998 | { | |
2999 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
3000 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
3001 | if (eptr-- == pp) break; /* Stop if tried at original pos */ | |
3002 | BACKCHAR(eptr); | |
3003 | } | |
3004 | } | |
3005 | else | |
3006 | #endif | |
3007 | ||
3008 | /* Not UTF-8 mode */ | |
3009 | { | |
3010 | switch(ctype) | |
3011 | { | |
3012 | case OP_ANY: | |
3013 | if ((ims & PCRE_DOTALL) == 0) | |
3014 | { | |
3015 | for (i = min; i < max; i++) | |
3016 | { | |
3017 | if (eptr >= md->end_subject || *eptr == NEWLINE) break; | |
3018 | eptr++; | |
3019 | } | |
3020 | break; | |
3021 | } | |
3022 | /* For DOTALL case, fall through and treat as \C */ | |
3023 | ||
3024 | case OP_ANYBYTE: | |
3025 | c = max - min; | |
3026 | if (c > md->end_subject - eptr) c = md->end_subject - eptr; | |
3027 | eptr += c; | |
3028 | break; | |
3029 | ||
3030 | case OP_NOT_DIGIT: | |
3031 | for (i = min; i < max; i++) | |
3032 | { | |
3033 | if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_digit) != 0) | |
3034 | break; | |
3035 | eptr++; | |
3036 | } | |
3037 | break; | |
3038 | ||
3039 | case OP_DIGIT: | |
3040 | for (i = min; i < max; i++) | |
3041 | { | |
3042 | if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_digit) == 0) | |
3043 | break; | |
3044 | eptr++; | |
3045 | } | |
3046 | break; | |
3047 | ||
3048 | case OP_NOT_WHITESPACE: | |
3049 | for (i = min; i < max; i++) | |
3050 | { | |
3051 | if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_space) != 0) | |
3052 | break; | |
3053 | eptr++; | |
3054 | } | |
3055 | break; | |
3056 | ||
3057 | case OP_WHITESPACE: | |
3058 | for (i = min; i < max; i++) | |
3059 | { | |
3060 | if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_space) == 0) | |
3061 | break; | |
3062 | eptr++; | |
3063 | } | |
3064 | break; | |
3065 | ||
3066 | case OP_NOT_WORDCHAR: | |
3067 | for (i = min; i < max; i++) | |
3068 | { | |
3069 | if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_word) != 0) | |
3070 | break; | |
3071 | eptr++; | |
3072 | } | |
3073 | break; | |
3074 | ||
3075 | case OP_WORDCHAR: | |
3076 | for (i = min; i < max; i++) | |
3077 | { | |
3078 | if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_word) == 0) | |
3079 | break; | |
3080 | eptr++; | |
3081 | } | |
3082 | break; | |
3083 | ||
3084 | default: | |
3085 | RRETURN(PCRE_ERROR_INTERNAL); | |
3086 | } | |
3087 | ||
3088 | /* eptr is now past the end of the maximum run */ | |
3089 | ||
3090 | while (eptr >= pp) | |
3091 | { | |
3092 | RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0); | |
3093 | eptr--; | |
3094 | if (rrc != MATCH_NOMATCH) RRETURN(rrc); | |
3095 | } | |
3096 | } | |
3097 | ||
3098 | /* Get here if we can't make it match with any permitted repetitions */ | |
3099 | ||
3100 | RRETURN(MATCH_NOMATCH); | |
3101 | } | |
3102 | /* Control never gets here */ | |
3103 | ||
3104 | /* There's been some horrible disaster. Since all codes > OP_BRA are | |
3105 | for capturing brackets, and there shouldn't be any gaps between 0 and | |
3106 | OP_BRA, arrival here can only mean there is something seriously wrong | |
3107 | in the code above or the OP_xxx definitions. */ | |
3108 | ||
3109 | default: | |
3110 | DPRINTF(("Unknown opcode %d\n", *ecode)); | |
3111 | RRETURN(PCRE_ERROR_UNKNOWN_NODE); | |
3112 | } | |
3113 | ||
3114 | /* Do not stick any code in here without much thought; it is assumed | |
3115 | that "continue" in the code above comes out to here to repeat the main | |
3116 | loop. */ | |
3117 | ||
3118 | } /* End of main loop */ | |
3119 | /* Control never reaches here */ | |
3120 | } | |
3121 | ||
3122 | ||
3123 | /*************************************************************************** | |
3124 | **************************************************************************** | |
3125 | RECURSION IN THE match() FUNCTION | |
3126 | ||
3127 | Undefine all the macros that were defined above to handle this. */ | |
3128 | ||
3129 | #ifdef NO_RECURSE | |
3130 | #undef eptr | |
3131 | #undef ecode | |
3132 | #undef offset_top | |
3133 | #undef ims | |
3134 | #undef eptrb | |
3135 | #undef flags | |
3136 | ||
3137 | #undef callpat | |
3138 | #undef charptr | |
3139 | #undef data | |
3140 | #undef next | |
3141 | #undef pp | |
3142 | #undef prev | |
3143 | #undef saved_eptr | |
3144 | ||
3145 | #undef new_recursive | |
3146 | ||
3147 | #undef cur_is_word | |
3148 | #undef condition | |
3149 | #undef minimize | |
3150 | #undef prev_is_word | |
3151 | ||
3152 | #undef original_ims | |
3153 | ||
3154 | #undef ctype | |
3155 | #undef length | |
3156 | #undef max | |
3157 | #undef min | |
3158 | #undef number | |
3159 | #undef offset | |
3160 | #undef op | |
3161 | #undef save_capture_last | |
3162 | #undef save_offset1 | |
3163 | #undef save_offset2 | |
3164 | #undef save_offset3 | |
3165 | #undef stacksave | |
3166 | ||
3167 | #undef newptrb | |
3168 | ||
3169 | #endif | |
3170 | ||
3171 | /* These two are defined as macros in both cases */ | |
3172 | ||
3173 | #undef fc | |
3174 | #undef fi | |
3175 | ||
3176 | /*************************************************************************** | |
3177 | ***************************************************************************/ | |
3178 | ||
3179 | ||
3180 | ||
3181 | /************************************************* | |
3182 | * Execute a Regular Expression * | |
3183 | *************************************************/ | |
3184 | ||
3185 | /* This function applies a compiled re to a subject string and picks out | |
3186 | portions of the string if it matches. Two elements in the vector are set for | |
3187 | each substring: the offsets to the start and end of the substring. | |
3188 | ||
3189 | Arguments: | |
3190 | argument_re points to the compiled expression | |
3191 | extra_data points to extra data or is NULL | |
3192 | subject points to the subject string | |
3193 | length length of subject string (may contain binary zeros) | |
3194 | start_offset where to start in the subject string | |
3195 | options option bits | |
3196 | offsets points to a vector of ints to be filled in with offsets | |
3197 | offsetcount the number of elements in the vector | |
3198 | ||
3199 | Returns: > 0 => success; value is the number of elements filled in | |
3200 | = 0 => success, but offsets is not big enough | |
3201 | -1 => failed to match | |
3202 | < -1 => some kind of unexpected problem | |
3203 | */ | |
3204 | ||
3205 | EXPORT int | |
3206 | pcre_exec(const pcre *argument_re, const pcre_extra *extra_data, | |
3207 | const char *subject, int length, int start_offset, int options, int *offsets, | |
3208 | int offsetcount) | |
3209 | { | |
3210 | int rc, resetcount, ocount; | |
3211 | int first_byte = -1; | |
3212 | int req_byte = -1; | |
3213 | int req_byte2 = -1; | |
3214 | unsigned long int ims = 0; | |
3215 | BOOL using_temporary_offsets = FALSE; | |
3216 | BOOL anchored; | |
3217 | BOOL startline; | |
3218 | BOOL firstline; | |
3219 | BOOL first_byte_caseless = FALSE; | |
3220 | BOOL req_byte_caseless = FALSE; | |
3221 | match_data match_block; | |
3222 | const uschar *tables; | |
3223 | const uschar *start_bits = NULL; | |
3224 | const uschar *start_match = (const uschar *)subject + start_offset; | |
3225 | const uschar *end_subject; | |
3226 | const uschar *req_byte_ptr = start_match - 1; | |
3227 | ||
3228 | pcre_study_data internal_study; | |
3229 | const pcre_study_data *study; | |
3230 | ||
3231 | real_pcre internal_re; | |
3232 | const real_pcre *external_re = (const real_pcre *)argument_re; | |
3233 | const real_pcre *re = external_re; | |
3234 | ||
3235 | /* Plausibility checks */ | |
3236 | ||
3237 | if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION; | |
3238 | if (re == NULL || subject == NULL || | |
3239 | (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL; | |
3240 | if (offsetcount < 0) return PCRE_ERROR_BADCOUNT; | |
3241 | ||
3242 | /* Fish out the optional data from the extra_data structure, first setting | |
3243 | the default values. */ | |
3244 | ||
3245 | study = NULL; | |
3246 | match_block.match_limit = MATCH_LIMIT; | |
3247 | match_block.callout_data = NULL; | |
3248 | ||
3249 | /* The table pointer is always in native byte order. */ | |
3250 | ||
3251 | tables = external_re->tables; | |
3252 | ||
3253 | if (extra_data != NULL) | |
3254 | { | |
3255 | register unsigned int flags = extra_data->flags; | |
3256 | if ((flags & PCRE_EXTRA_STUDY_DATA) != 0) | |
3257 | study = (const pcre_study_data *)extra_data->study_data; | |
3258 | if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0) | |
3259 | match_block.match_limit = extra_data->match_limit; | |
3260 | if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0) | |
3261 | match_block.callout_data = extra_data->callout_data; | |
3262 | if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables; | |
3263 | } | |
3264 | ||
3265 | /* If the exec call supplied NULL for tables, use the inbuilt ones. This | |
3266 | is a feature that makes it possible to save compiled regex and re-use them | |
3267 | in other programs later. */ | |
3268 | ||
3269 | if (tables == NULL) tables = _pcre_default_tables; | |
3270 | ||
3271 | /* Check that the first field in the block is the magic number. If it is not, | |
3272 | test for a regex that was compiled on a host of opposite endianness. If this is | |
3273 | the case, flipped values are put in internal_re and internal_study if there was | |
3274 | study data too. */ | |
3275 | ||
3276 | if (re->magic_number != MAGIC_NUMBER) | |
3277 | { | |
3278 | re = _pcre_try_flipped(re, &internal_re, study, &internal_study); | |
3279 | if (re == NULL) return PCRE_ERROR_BADMAGIC; | |
3280 | if (study != NULL) study = &internal_study; | |
3281 | } | |
3282 | ||
3283 | /* Set up other data */ | |
3284 | ||
3285 | anchored = ((re->options | options) & PCRE_ANCHORED) != 0; | |
3286 | startline = (re->options & PCRE_STARTLINE) != 0; | |
3287 | firstline = (re->options & PCRE_FIRSTLINE) != 0; | |
3288 | ||
3289 | /* The code starts after the real_pcre block and the capture name table. */ | |
3290 | ||
3291 | match_block.start_code = (const uschar *)external_re + re->name_table_offset + | |
3292 | re->name_count * re->name_entry_size; | |
3293 | ||
3294 | match_block.start_subject = (const uschar *)subject; | |
3295 | match_block.start_offset = start_offset; | |
3296 | match_block.end_subject = match_block.start_subject + length; | |
3297 | end_subject = match_block.end_subject; | |
3298 | ||
3299 | match_block.endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0; | |
3300 | match_block.utf8 = (re->options & PCRE_UTF8) != 0; | |
3301 | ||
3302 | match_block.notbol = (options & PCRE_NOTBOL) != 0; | |
3303 | match_block.noteol = (options & PCRE_NOTEOL) != 0; | |
3304 | match_block.notempty = (options & PCRE_NOTEMPTY) != 0; | |
3305 | match_block.partial = (options & PCRE_PARTIAL) != 0; | |
3306 | match_block.hitend = FALSE; | |
3307 | ||
3308 | match_block.recursive = NULL; /* No recursion at top level */ | |
3309 | ||
3310 | match_block.lcc = tables + lcc_offset; | |
3311 | match_block.ctypes = tables + ctypes_offset; | |
3312 | ||
3313 | /* Partial matching is supported only for a restricted set of regexes at the | |
3314 | moment. */ | |
3315 | ||
3316 | if (match_block.partial && (re->options & PCRE_NOPARTIAL) != 0) | |
3317 | return PCRE_ERROR_BADPARTIAL; | |
3318 | ||
3319 | /* Check a UTF-8 string if required. Unfortunately there's no way of passing | |
3320 | back the character offset. */ | |
3321 | ||
3322 | #ifdef SUPPORT_UTF8 | |
3323 | if (match_block.utf8 && (options & PCRE_NO_UTF8_CHECK) == 0) | |
3324 | { | |
3325 | if (_pcre_valid_utf8((uschar *)subject, length) >= 0) | |
3326 | return PCRE_ERROR_BADUTF8; | |
3327 | if (start_offset > 0 && start_offset < length) | |
3328 | { | |
3329 | int tb = ((uschar *)subject)[start_offset]; | |
3330 | if (tb > 127) | |
3331 | { | |
3332 | tb &= 0xc0; | |
3333 | if (tb != 0 && tb != 0xc0) return PCRE_ERROR_BADUTF8_OFFSET; | |
3334 | } | |
3335 | } | |
3336 | } | |
3337 | #endif | |
3338 | ||
3339 | /* The ims options can vary during the matching as a result of the presence | |
3340 | of (?ims) items in the pattern. They are kept in a local variable so that | |
3341 | restoring at the exit of a group is easy. */ | |
3342 | ||
3343 | ims = re->options & (PCRE_CASELESS|PCRE_MULTILINE|PCRE_DOTALL); | |
3344 | ||
3345 | /* If the expression has got more back references than the offsets supplied can | |
3346 | hold, we get a temporary chunk of working store to use during the matching. | |
3347 | Otherwise, we can use the vector supplied, rounding down its size to a multiple | |
3348 | of 3. */ | |
3349 | ||
3350 | ocount = offsetcount - (offsetcount % 3); | |
3351 | ||
3352 | if (re->top_backref > 0 && re->top_backref >= ocount/3) | |
3353 | { | |
3354 | ocount = re->top_backref * 3 + 3; | |
3355 | match_block.offset_vector = (int *)(pcre_malloc)(ocount * sizeof(int)); | |
3356 | if (match_block.offset_vector == NULL) return PCRE_ERROR_NOMEMORY; | |
3357 | using_temporary_offsets = TRUE; | |
3358 | DPRINTF(("Got memory to hold back references\n")); | |
3359 | } | |
3360 | else match_block.offset_vector = offsets; | |
3361 | ||
3362 | match_block.offset_end = ocount; | |
3363 | match_block.offset_max = (2*ocount)/3; | |
3364 | match_block.offset_overflow = FALSE; | |
3365 | match_block.capture_last = -1; | |
3366 | ||
3367 | /* Compute the minimum number of offsets that we need to reset each time. Doing | |
3368 | this makes a huge difference to execution time when there aren't many brackets | |
3369 | in the pattern. */ | |
3370 | ||
3371 | resetcount = 2 + re->top_bracket * 2; | |
3372 | if (resetcount > offsetcount) resetcount = ocount; | |
3373 | ||
3374 | /* Reset the working variable associated with each extraction. These should | |
3375 | never be used unless previously set, but they get saved and restored, and so we | |
3376 | initialize them to avoid reading uninitialized locations. */ | |
3377 | ||
3378 | if (match_block.offset_vector != NULL) | |
3379 | { | |
3380 | register int *iptr = match_block.offset_vector + ocount; | |
3381 | register int *iend = iptr - resetcount/2 + 1; | |
3382 | while (--iptr >= iend) *iptr = -1; | |
3383 | } | |
3384 | ||
3385 | /* Set up the first character to match, if available. The first_byte value is | |
3386 | never set for an anchored regular expression, but the anchoring may be forced | |
3387 | at run time, so we have to test for anchoring. The first char may be unset for | |
3388 | an unanchored pattern, of course. If there's no first char and the pattern was | |
3389 | studied, there may be a bitmap of possible first characters. */ | |
3390 | ||
3391 | if (!anchored) | |
3392 | { | |
3393 | if ((re->options & PCRE_FIRSTSET) != 0) | |
3394 | { | |
3395 | first_byte = re->first_byte & 255; | |
3396 | if ((first_byte_caseless = ((re->first_byte & REQ_CASELESS) != 0)) == TRUE) | |
3397 | first_byte = match_block.lcc[first_byte]; | |
3398 | } | |
3399 | else | |
3400 | if (!startline && study != NULL && | |
3401 | (study->options & PCRE_STUDY_MAPPED) != 0) | |
3402 | start_bits = study->start_bits; | |
3403 | } | |
3404 | ||
3405 | /* For anchored or unanchored matches, there may be a "last known required | |
3406 | character" set. */ | |
3407 | ||
3408 | if ((re->options & PCRE_REQCHSET) != 0) | |
3409 | { | |
3410 | req_byte = re->req_byte & 255; | |
3411 | req_byte_caseless = (re->req_byte & REQ_CASELESS) != 0; | |
3412 | req_byte2 = (tables + fcc_offset)[req_byte]; /* case flipped */ | |
3413 | } | |
3414 | ||
3415 | /* Loop for handling unanchored repeated matching attempts; for anchored regexs | |
3416 | the loop runs just once. */ | |
3417 | ||
3418 | do | |
3419 | { | |
3420 | const uschar *save_end_subject = end_subject; | |
3421 | ||
3422 | /* Reset the maximum number of extractions we might see. */ | |
3423 | ||
3424 | if (match_block.offset_vector != NULL) | |
3425 | { | |
3426 | register int *iptr = match_block.offset_vector; | |
3427 | register int *iend = iptr + resetcount; | |
3428 | while (iptr < iend) *iptr++ = -1; | |
3429 | } | |
3430 | ||
3431 | /* Advance to a unique first char if possible. If firstline is TRUE, the | |
3432 | start of the match is constrained to the first line of a multiline string. | |
3433 | Implement this by temporarily adjusting end_subject so that we stop scanning | |
3434 | at a newline. If the match fails at the newline, later code breaks this loop. | |
3435 | */ | |
3436 | ||
3437 | if (firstline) | |
3438 | { | |
3439 | const uschar *t = start_match; | |
3440 | while (t < save_end_subject && *t != '\n') t++; | |
3441 | end_subject = t; | |
3442 | } | |
3443 | ||
3444 | /* Now test for a unique first byte */ | |
3445 | ||
3446 | if (first_byte >= 0) | |
3447 | { | |
3448 | if (first_byte_caseless) | |
3449 | while (start_match < end_subject && | |
3450 | match_block.lcc[*start_match] != first_byte) | |
3451 | start_match++; | |
3452 | else | |
3453 | while (start_match < end_subject && *start_match != first_byte) | |
3454 | start_match++; | |
3455 | } | |
3456 | ||
3457 | /* Or to just after \n for a multiline match if possible */ | |
3458 | ||
3459 | else if (startline) | |
3460 | { | |
3461 | if (start_match > match_block.start_subject + start_offset) | |
3462 | { | |
3463 | while (start_match < end_subject && start_match[-1] != NEWLINE) | |
3464 | start_match++; | |
3465 | } | |
3466 | } | |
3467 | ||
3468 | /* Or to a non-unique first char after study */ | |
3469 | ||
3470 | else if (start_bits != NULL) | |
3471 | { | |
3472 | while (start_match < end_subject) | |
3473 | { | |
3474 | register unsigned int c = *start_match; | |
3475 | if ((start_bits[c/8] & (1 << (c&7))) == 0) start_match++; else break; | |
3476 | } | |
3477 | } | |
3478 | ||
3479 | /* Restore fudged end_subject */ | |
3480 | ||
3481 | end_subject = save_end_subject; | |
3482 | ||
3483 | #ifdef DEBUG /* Sigh. Some compilers never learn. */ | |
3484 | printf(">>>> Match against: "); | |
3485 | pchars(start_match, end_subject - start_match, TRUE, &match_block); | |
3486 | printf("\n"); | |
3487 | #endif | |
3488 | ||
3489 | /* If req_byte is set, we know that that character must appear in the subject | |
3490 | for the match to succeed. If the first character is set, req_byte must be | |
3491 | later in the subject; otherwise the test starts at the match point. This | |
3492 | optimization can save a huge amount of backtracking in patterns with nested | |
3493 | unlimited repeats that aren't going to match. Writing separate code for | |
3494 | cased/caseless versions makes it go faster, as does using an autoincrement | |
3495 | and backing off on a match. | |
3496 | ||
3497 | HOWEVER: when the subject string is very, very long, searching to its end can | |
3498 | take a long time, and give bad performance on quite ordinary patterns. This | |
3499 | showed up when somebody was matching /^C/ on a 32-megabyte string... so we | |
3500 | don't do this when the string is sufficiently long. | |
3501 | ||
3502 | ALSO: this processing is disabled when partial matching is requested. | |
3503 | */ | |
3504 | ||
3505 | if (req_byte >= 0 && | |
3506 | end_subject - start_match < REQ_BYTE_MAX && | |
3507 | !match_block.partial) | |
3508 | { | |
3509 | register const uschar *p = start_match + ((first_byte >= 0)? 1 : 0); | |
3510 | ||
3511 | /* We don't need to repeat the search if we haven't yet reached the | |
3512 | place we found it at last time. */ | |
3513 | ||
3514 | if (p > req_byte_ptr) | |
3515 | { | |
3516 | if (req_byte_caseless) | |
3517 | { | |
3518 | while (p < end_subject) | |
3519 | { | |
3520 | register int pp = *p++; | |
3521 | if (pp == req_byte || pp == req_byte2) { p--; break; } | |
3522 | } | |
3523 | } | |
3524 | else | |
3525 | { | |
3526 | while (p < end_subject) | |
3527 | { | |
3528 | if (*p++ == req_byte) { p--; break; } | |
3529 | } | |
3530 | } | |
3531 | ||
3532 | /* If we can't find the required character, break the matching loop */ | |
3533 | ||
3534 | if (p >= end_subject) break; | |
3535 | ||
3536 | /* If we have found the required character, save the point where we | |
3537 | found it, so that we don't search again next time round the loop if | |
3538 | the start hasn't passed this character yet. */ | |
3539 | ||
3540 | req_byte_ptr = p; | |
3541 | } | |
3542 | } | |
3543 | ||
3544 | /* When a match occurs, substrings will be set for all internal extractions; | |
3545 | we just need to set up the whole thing as substring 0 before returning. If | |
3546 | there were too many extractions, set the return code to zero. In the case | |
3547 | where we had to get some local store to hold offsets for backreferences, copy | |
3548 | those back references that we can. In this case there need not be overflow | |
3549 | if certain parts of the pattern were not used. */ | |
3550 | ||
3551 | match_block.start_match = start_match; | |
3552 | match_block.match_call_count = 0; | |
3553 | ||
3554 | rc = match(start_match, match_block.start_code, 2, &match_block, ims, NULL, | |
3555 | match_isgroup); | |
3556 | ||
3557 | /* When the result is no match, if the subject's first character was a | |
3558 | newline and the PCRE_FIRSTLINE option is set, break (which will return | |
3559 | PCRE_ERROR_NOMATCH). The option requests that a match occur before the first | |
3560 | newline in the subject. Otherwise, advance the pointer to the next character | |
3561 | and continue - but the continuation will actually happen only when the | |
3562 | pattern is not anchored. */ | |
3563 | ||
3564 | if (rc == MATCH_NOMATCH) | |
3565 | { | |
3566 | if (firstline && *start_match == NEWLINE) break; | |
3567 | start_match++; | |
3568 | #ifdef SUPPORT_UTF8 | |
3569 | if (match_block.utf8) | |
3570 | while(start_match < end_subject && (*start_match & 0xc0) == 0x80) | |
3571 | start_match++; | |
3572 | #endif | |
3573 | continue; | |
3574 | } | |
3575 | ||
3576 | if (rc != MATCH_MATCH) | |
3577 | { | |
3578 | DPRINTF((">>>> error: returning %d\n", rc)); | |
3579 | return rc; | |
3580 | } | |
3581 | ||
3582 | /* We have a match! Copy the offset information from temporary store if | |
3583 | necessary */ | |
3584 | ||
3585 | if (using_temporary_offsets) | |
3586 | { | |
3587 | if (offsetcount >= 4) | |
3588 | { | |
3589 | memcpy(offsets + 2, match_block.offset_vector + 2, | |
3590 | (offsetcount - 2) * sizeof(int)); | |
3591 | DPRINTF(("Copied offsets from temporary memory\n")); | |
3592 | } | |
3593 | if (match_block.end_offset_top > offsetcount) | |
3594 | match_block.offset_overflow = TRUE; | |
3595 | ||
3596 | DPRINTF(("Freeing temporary memory\n")); | |
3597 | (pcre_free)(match_block.offset_vector); | |
3598 | } | |
3599 | ||
3600 | rc = match_block.offset_overflow? 0 : match_block.end_offset_top/2; | |
3601 | ||
3602 | if (offsetcount < 2) rc = 0; else | |
3603 | { | |
3604 | offsets[0] = start_match - match_block.start_subject; | |
3605 | offsets[1] = match_block.end_match_ptr - match_block.start_subject; | |
3606 | } | |
3607 | ||
3608 | DPRINTF((">>>> returning %d\n", rc)); | |
3609 | return rc; | |
3610 | } | |
3611 | ||
3612 | /* This "while" is the end of the "do" above */ | |
3613 | ||
3614 | while (!anchored && start_match <= end_subject); | |
3615 | ||
3616 | if (using_temporary_offsets) | |
3617 | { | |
3618 | DPRINTF(("Freeing temporary memory\n")); | |
3619 | (pcre_free)(match_block.offset_vector); | |
3620 | } | |
3621 | ||
3622 | if (match_block.partial && match_block.hitend) | |
3623 | { | |
3624 | DPRINTF((">>>> returning PCRE_ERROR_PARTIAL\n")); | |
3625 | return PCRE_ERROR_PARTIAL; | |
3626 | } | |
3627 | else | |
3628 | { | |
3629 | DPRINTF((">>>> returning PCRE_ERROR_NOMATCH\n")); | |
3630 | return PCRE_ERROR_NOMATCH; | |
3631 | } | |
3632 | } | |
3633 | ||
3634 | /* End of pcre_exec.c */ |