/[pcre]/code/trunk/pcre_exec.c
ViewVC logotype

Contents of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 475 - (show annotations) (download)
Sat Jan 2 18:21:30 2010 UTC (4 years, 8 months ago) by ph10
File MIME type: text/plain
File size: 170933 byte(s)
Tidies to allow easier embedded compilation; avoid (double) where possible.

1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2010 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40
41 /* This module contains pcre_exec(), the externally visible function that does
42 pattern matching using an NFA algorithm, trying to mimic Perl as closely as
43 possible. There are also some static supporting functions. */
44
45 #ifdef HAVE_CONFIG_H
46 #include "config.h"
47 #endif
48
49 #define NLBLOCK md /* Block containing newline information */
50 #define PSSTART start_subject /* Field containing processed string start */
51 #define PSEND end_subject /* Field containing processed string end */
52
53 #include "pcre_internal.h"
54
55 /* Undefine some potentially clashing cpp symbols */
56
57 #undef min
58 #undef max
59
60 /* Flag bits for the match() function */
61
62 #define match_condassert 0x01 /* Called to check a condition assertion */
63 #define match_cbegroup 0x02 /* Could-be-empty unlimited repeat group */
64
65 /* Non-error returns from the match() function. Error returns are externally
66 defined PCRE_ERROR_xxx codes, which are all negative. */
67
68 #define MATCH_MATCH 1
69 #define MATCH_NOMATCH 0
70
71 /* Special internal returns from the match() function. Make them sufficiently
72 negative to avoid the external error codes. */
73
74 #define MATCH_COMMIT (-999)
75 #define MATCH_PRUNE (-998)
76 #define MATCH_SKIP (-997)
77 #define MATCH_THEN (-996)
78
79 /* Maximum number of ints of offset to save on the stack for recursive calls.
80 If the offset vector is bigger, malloc is used. This should be a multiple of 3,
81 because the offset vector is always a multiple of 3 long. */
82
83 #define REC_STACK_SAVE_MAX 30
84
85 /* Min and max values for the common repeats; for the maxima, 0 => infinity */
86
87 static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
88 static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
89
90
91
92 #ifdef PCRE_DEBUG
93 /*************************************************
94 * Debugging function to print chars *
95 *************************************************/
96
97 /* Print a sequence of chars in printable format, stopping at the end of the
98 subject if the requested.
99
100 Arguments:
101 p points to characters
102 length number to print
103 is_subject TRUE if printing from within md->start_subject
104 md pointer to matching data block, if is_subject is TRUE
105
106 Returns: nothing
107 */
108
109 static void
110 pchars(const uschar *p, int length, BOOL is_subject, match_data *md)
111 {
112 unsigned int c;
113 if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
114 while (length-- > 0)
115 if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);
116 }
117 #endif
118
119
120
121 /*************************************************
122 * Match a back-reference *
123 *************************************************/
124
125 /* If a back reference hasn't been set, the length that is passed is greater
126 than the number of characters left in the string, so the match fails.
127
128 Arguments:
129 offset index into the offset vector
130 eptr points into the subject
131 length length to be matched
132 md points to match data block
133 ims the ims flags
134
135 Returns: TRUE if matched
136 */
137
138 static BOOL
139 match_ref(int offset, register USPTR eptr, int length, match_data *md,
140 unsigned long int ims)
141 {
142 USPTR p = md->start_subject + md->offset_vector[offset];
143
144 #ifdef PCRE_DEBUG
145 if (eptr >= md->end_subject)
146 printf("matching subject <null>");
147 else
148 {
149 printf("matching subject ");
150 pchars(eptr, length, TRUE, md);
151 }
152 printf(" against backref ");
153 pchars(p, length, FALSE, md);
154 printf("\n");
155 #endif
156
157 /* Always fail if not enough characters left */
158
159 if (length > md->end_subject - eptr) return FALSE;
160
161 /* Separate the caseless case for speed. In UTF-8 mode we can only do this
162 properly if Unicode properties are supported. Otherwise, we can check only
163 ASCII characters. */
164
165 if ((ims & PCRE_CASELESS) != 0)
166 {
167 #ifdef SUPPORT_UTF8
168 #ifdef SUPPORT_UCP
169 if (md->utf8)
170 {
171 USPTR endptr = eptr + length;
172 while (eptr < endptr)
173 {
174 int c, d;
175 GETCHARINC(c, eptr);
176 GETCHARINC(d, p);
177 if (c != d && c != UCD_OTHERCASE(d)) return FALSE;
178 }
179 }
180 else
181 #endif
182 #endif
183
184 /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
185 is no UCP support. */
186
187 while (length-- > 0)
188 { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }
189 }
190
191 /* In the caseful case, we can just compare the bytes, whether or not we
192 are in UTF-8 mode. */
193
194 else
195 { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }
196
197 return TRUE;
198 }
199
200
201
202 /***************************************************************************
203 ****************************************************************************
204 RECURSION IN THE match() FUNCTION
205
206 The match() function is highly recursive, though not every recursive call
207 increases the recursive depth. Nevertheless, some regular expressions can cause
208 it to recurse to a great depth. I was writing for Unix, so I just let it call
209 itself recursively. This uses the stack for saving everything that has to be
210 saved for a recursive call. On Unix, the stack can be large, and this works
211 fine.
212
213 It turns out that on some non-Unix-like systems there are problems with
214 programs that use a lot of stack. (This despite the fact that every last chip
215 has oodles of memory these days, and techniques for extending the stack have
216 been known for decades.) So....
217
218 There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
219 calls by keeping local variables that need to be preserved in blocks of memory
220 obtained from malloc() instead instead of on the stack. Macros are used to
221 achieve this so that the actual code doesn't look very different to what it
222 always used to.
223
224 The original heap-recursive code used longjmp(). However, it seems that this
225 can be very slow on some operating systems. Following a suggestion from Stan
226 Switzer, the use of longjmp() has been abolished, at the cost of having to
227 provide a unique number for each call to RMATCH. There is no way of generating
228 a sequence of numbers at compile time in C. I have given them names, to make
229 them stand out more clearly.
230
231 Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
232 FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
233 tests. Furthermore, not using longjmp() means that local dynamic variables
234 don't have indeterminate values; this has meant that the frame size can be
235 reduced because the result can be "passed back" by straight setting of the
236 variable instead of being passed in the frame.
237 ****************************************************************************
238 ***************************************************************************/
239
240 /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
241 below must be updated in sync. */
242
243 enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10,
244 RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
245 RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
246 RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
247 RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
248 RM51, RM52, RM53, RM54 };
249
250 /* These versions of the macros use the stack, as normal. There are debugging
251 versions and production versions. Note that the "rw" argument of RMATCH isn't
252 actuall used in this definition. */
253
254 #ifndef NO_RECURSE
255 #define REGISTER register
256
257 #ifdef PCRE_DEBUG
258 #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \
259 { \
260 printf("match() called in line %d\n", __LINE__); \
261 rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \
262 printf("to line %d\n", __LINE__); \
263 }
264 #define RRETURN(ra) \
265 { \
266 printf("match() returned %d from line %d ", ra, __LINE__); \
267 return ra; \
268 }
269 #else
270 #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \
271 rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)
272 #define RRETURN(ra) return ra
273 #endif
274
275 #else
276
277
278 /* These versions of the macros manage a private stack on the heap. Note that
279 the "rd" argument of RMATCH isn't actually used in this definition. It's the md
280 argument of match(), which never changes. */
281
282 #define REGISTER
283
284 #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\
285 {\
286 heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\
287 frame->Xwhere = rw; \
288 newframe->Xeptr = ra;\
289 newframe->Xecode = rb;\
290 newframe->Xmstart = mstart;\
291 newframe->Xoffset_top = rc;\
292 newframe->Xims = re;\
293 newframe->Xeptrb = rf;\
294 newframe->Xflags = rg;\
295 newframe->Xrdepth = frame->Xrdepth + 1;\
296 newframe->Xprevframe = frame;\
297 frame = newframe;\
298 DPRINTF(("restarting from line %d\n", __LINE__));\
299 goto HEAP_RECURSE;\
300 L_##rw:\
301 DPRINTF(("jumped back to line %d\n", __LINE__));\
302 }
303
304 #define RRETURN(ra)\
305 {\
306 heapframe *newframe = frame;\
307 frame = newframe->Xprevframe;\
308 (pcre_stack_free)(newframe);\
309 if (frame != NULL)\
310 {\
311 rrc = ra;\
312 goto HEAP_RETURN;\
313 }\
314 return ra;\
315 }
316
317
318 /* Structure for remembering the local variables in a private frame */
319
320 typedef struct heapframe {
321 struct heapframe *Xprevframe;
322
323 /* Function arguments that may change */
324
325 USPTR Xeptr;
326 const uschar *Xecode;
327 USPTR Xmstart;
328 int Xoffset_top;
329 long int Xims;
330 eptrblock *Xeptrb;
331 int Xflags;
332 unsigned int Xrdepth;
333
334 /* Function local variables */
335
336 USPTR Xcallpat;
337 #ifdef SUPPORT_UTF8
338 USPTR Xcharptr;
339 #endif
340 USPTR Xdata;
341 USPTR Xnext;
342 USPTR Xpp;
343 USPTR Xprev;
344 USPTR Xsaved_eptr;
345
346 recursion_info Xnew_recursive;
347
348 BOOL Xcur_is_word;
349 BOOL Xcondition;
350 BOOL Xprev_is_word;
351
352 unsigned long int Xoriginal_ims;
353
354 #ifdef SUPPORT_UCP
355 int Xprop_type;
356 int Xprop_value;
357 int Xprop_fail_result;
358 int Xprop_category;
359 int Xprop_chartype;
360 int Xprop_script;
361 int Xoclength;
362 uschar Xocchars[8];
363 #endif
364
365 int Xcodelink;
366 int Xctype;
367 unsigned int Xfc;
368 int Xfi;
369 int Xlength;
370 int Xmax;
371 int Xmin;
372 int Xnumber;
373 int Xoffset;
374 int Xop;
375 int Xsave_capture_last;
376 int Xsave_offset1, Xsave_offset2, Xsave_offset3;
377 int Xstacksave[REC_STACK_SAVE_MAX];
378
379 eptrblock Xnewptrb;
380
381 /* Where to jump back to */
382
383 int Xwhere;
384
385 } heapframe;
386
387 #endif
388
389
390 /***************************************************************************
391 ***************************************************************************/
392
393
394
395 /*************************************************
396 * Match from current position *
397 *************************************************/
398
399 /* This function is called recursively in many circumstances. Whenever it
400 returns a negative (error) response, the outer incarnation must also return the
401 same response. */
402
403 /* These macros pack up tests that are used for partial matching, and which
404 appears several times in the code. We set the "hit end" flag if the pointer is
405 at the end of the subject and also past the start of the subject (i.e.
406 something has been matched). For hard partial matching, we then return
407 immediately. The second one is used when we already know we are past the end of
408 the subject. */
409
410 #define CHECK_PARTIAL()\
411 if (md->partial != 0 && eptr >= md->end_subject && eptr > mstart)\
412 {\
413 md->hitend = TRUE;\
414 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);\
415 }
416
417 #define SCHECK_PARTIAL()\
418 if (md->partial != 0 && eptr > mstart)\
419 {\
420 md->hitend = TRUE;\
421 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);\
422 }
423
424
425 /* Performance note: It might be tempting to extract commonly used fields from
426 the md structure (e.g. utf8, end_subject) into individual variables to improve
427 performance. Tests using gcc on a SPARC disproved this; in the first case, it
428 made performance worse.
429
430 Arguments:
431 eptr pointer to current character in subject
432 ecode pointer to current position in compiled code
433 mstart pointer to the current match start position (can be modified
434 by encountering \K)
435 offset_top current top pointer
436 md pointer to "static" info for the match
437 ims current /i, /m, and /s options
438 eptrb pointer to chain of blocks containing eptr at start of
439 brackets - for testing for empty matches
440 flags can contain
441 match_condassert - this is an assertion condition
442 match_cbegroup - this is the start of an unlimited repeat
443 group that can match an empty string
444 rdepth the recursion depth
445
446 Returns: MATCH_MATCH if matched ) these values are >= 0
447 MATCH_NOMATCH if failed to match )
448 a negative PCRE_ERROR_xxx value if aborted by an error condition
449 (e.g. stopped by repeated call or recursion limit)
450 */
451
452 static int
453 match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
454 int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,
455 int flags, unsigned int rdepth)
456 {
457 /* These variables do not need to be preserved over recursion in this function,
458 so they can be ordinary variables in all cases. Mark some of them with
459 "register" because they are used a lot in loops. */
460
461 register int rrc; /* Returns from recursive calls */
462 register int i; /* Used for loops not involving calls to RMATCH() */
463 register unsigned int c; /* Character values not kept over RMATCH() calls */
464 register BOOL utf8; /* Local copy of UTF-8 flag for speed */
465
466 BOOL minimize, possessive; /* Quantifier options */
467 int condcode;
468
469 /* When recursion is not being used, all "local" variables that have to be
470 preserved over calls to RMATCH() are part of a "frame" which is obtained from
471 heap storage. Set up the top-level frame here; others are obtained from the
472 heap whenever RMATCH() does a "recursion". See the macro definitions above. */
473
474 #ifdef NO_RECURSE
475 heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));
476 frame->Xprevframe = NULL; /* Marks the top level */
477
478 /* Copy in the original argument variables */
479
480 frame->Xeptr = eptr;
481 frame->Xecode = ecode;
482 frame->Xmstart = mstart;
483 frame->Xoffset_top = offset_top;
484 frame->Xims = ims;
485 frame->Xeptrb = eptrb;
486 frame->Xflags = flags;
487 frame->Xrdepth = rdepth;
488
489 /* This is where control jumps back to to effect "recursion" */
490
491 HEAP_RECURSE:
492
493 /* Macros make the argument variables come from the current frame */
494
495 #define eptr frame->Xeptr
496 #define ecode frame->Xecode
497 #define mstart frame->Xmstart
498 #define offset_top frame->Xoffset_top
499 #define ims frame->Xims
500 #define eptrb frame->Xeptrb
501 #define flags frame->Xflags
502 #define rdepth frame->Xrdepth
503
504 /* Ditto for the local variables */
505
506 #ifdef SUPPORT_UTF8
507 #define charptr frame->Xcharptr
508 #endif
509 #define callpat frame->Xcallpat
510 #define codelink frame->Xcodelink
511 #define data frame->Xdata
512 #define next frame->Xnext
513 #define pp frame->Xpp
514 #define prev frame->Xprev
515 #define saved_eptr frame->Xsaved_eptr
516
517 #define new_recursive frame->Xnew_recursive
518
519 #define cur_is_word frame->Xcur_is_word
520 #define condition frame->Xcondition
521 #define prev_is_word frame->Xprev_is_word
522
523 #define original_ims frame->Xoriginal_ims
524
525 #ifdef SUPPORT_UCP
526 #define prop_type frame->Xprop_type
527 #define prop_value frame->Xprop_value
528 #define prop_fail_result frame->Xprop_fail_result
529 #define prop_category frame->Xprop_category
530 #define prop_chartype frame->Xprop_chartype
531 #define prop_script frame->Xprop_script
532 #define oclength frame->Xoclength
533 #define occhars frame->Xocchars
534 #endif
535
536 #define ctype frame->Xctype
537 #define fc frame->Xfc
538 #define fi frame->Xfi
539 #define length frame->Xlength
540 #define max frame->Xmax
541 #define min frame->Xmin
542 #define number frame->Xnumber
543 #define offset frame->Xoffset
544 #define op frame->Xop
545 #define save_capture_last frame->Xsave_capture_last
546 #define save_offset1 frame->Xsave_offset1
547 #define save_offset2 frame->Xsave_offset2
548 #define save_offset3 frame->Xsave_offset3
549 #define stacksave frame->Xstacksave
550
551 #define newptrb frame->Xnewptrb
552
553 /* When recursion is being used, local variables are allocated on the stack and
554 get preserved during recursion in the normal way. In this environment, fi and
555 i, and fc and c, can be the same variables. */
556
557 #else /* NO_RECURSE not defined */
558 #define fi i
559 #define fc c
560
561
562 #ifdef SUPPORT_UTF8 /* Many of these variables are used only */
563 const uschar *charptr; /* in small blocks of the code. My normal */
564 #endif /* style of coding would have declared */
565 const uschar *callpat; /* them within each of those blocks. */
566 const uschar *data; /* However, in order to accommodate the */
567 const uschar *next; /* version of this code that uses an */
568 USPTR pp; /* external "stack" implemented on the */
569 const uschar *prev; /* heap, it is easier to declare them all */
570 USPTR saved_eptr; /* here, so the declarations can be cut */
571 /* out in a block. The only declarations */
572 recursion_info new_recursive; /* within blocks below are for variables */
573 /* that do not have to be preserved over */
574 BOOL cur_is_word; /* a recursive call to RMATCH(). */
575 BOOL condition;
576 BOOL prev_is_word;
577
578 unsigned long int original_ims;
579
580 #ifdef SUPPORT_UCP
581 int prop_type;
582 int prop_value;
583 int prop_fail_result;
584 int prop_category;
585 int prop_chartype;
586 int prop_script;
587 int oclength;
588 uschar occhars[8];
589 #endif
590
591 int codelink;
592 int ctype;
593 int length;
594 int max;
595 int min;
596 int number;
597 int offset;
598 int op;
599 int save_capture_last;
600 int save_offset1, save_offset2, save_offset3;
601 int stacksave[REC_STACK_SAVE_MAX];
602
603 eptrblock newptrb;
604 #endif /* NO_RECURSE */
605
606 /* These statements are here to stop the compiler complaining about unitialized
607 variables. */
608
609 #ifdef SUPPORT_UCP
610 prop_value = 0;
611 prop_fail_result = 0;
612 #endif
613
614
615 /* This label is used for tail recursion, which is used in a few cases even
616 when NO_RECURSE is not defined, in order to reduce the amount of stack that is
617 used. Thanks to Ian Taylor for noticing this possibility and sending the
618 original patch. */
619
620 TAIL_RECURSE:
621
622 /* OK, now we can get on with the real code of the function. Recursive calls
623 are specified by the macro RMATCH and RRETURN is used to return. When
624 NO_RECURSE is *not* defined, these just turn into a recursive call to match()
625 and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
626 defined). However, RMATCH isn't like a function call because it's quite a
627 complicated macro. It has to be used in one particular way. This shouldn't,
628 however, impact performance when true recursion is being used. */
629
630 #ifdef SUPPORT_UTF8
631 utf8 = md->utf8; /* Local copy of the flag */
632 #else
633 utf8 = FALSE;
634 #endif
635
636 /* First check that we haven't called match() too many times, or that we
637 haven't exceeded the recursive call limit. */
638
639 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
640 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
641
642 original_ims = ims; /* Save for resetting on ')' */
643
644 /* At the start of a group with an unlimited repeat that may match an empty
645 string, the match_cbegroup flag is set. When this is the case, add the current
646 subject pointer to the chain of such remembered pointers, to be checked when we
647 hit the closing ket, in order to break infinite loops that match no characters.
648 When match() is called in other circumstances, don't add to the chain. The
649 match_cbegroup flag must NOT be used with tail recursion, because the memory
650 block that is used is on the stack, so a new one may be required for each
651 match(). */
652
653 if ((flags & match_cbegroup) != 0)
654 {
655 newptrb.epb_saved_eptr = eptr;
656 newptrb.epb_prev = eptrb;
657 eptrb = &newptrb;
658 }
659
660 /* Now start processing the opcodes. */
661
662 for (;;)
663 {
664 minimize = possessive = FALSE;
665 op = *ecode;
666
667 switch(op)
668 {
669 case OP_FAIL:
670 RRETURN(MATCH_NOMATCH);
671
672 case OP_PRUNE:
673 RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
674 ims, eptrb, flags, RM51);
675 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
676 RRETURN(MATCH_PRUNE);
677
678 case OP_COMMIT:
679 RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
680 ims, eptrb, flags, RM52);
681 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
682 RRETURN(MATCH_COMMIT);
683
684 case OP_SKIP:
685 RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
686 ims, eptrb, flags, RM53);
687 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
688 md->start_match_ptr = eptr; /* Pass back current position */
689 RRETURN(MATCH_SKIP);
690
691 case OP_THEN:
692 RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
693 ims, eptrb, flags, RM54);
694 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
695 RRETURN(MATCH_THEN);
696
697 /* Handle a capturing bracket. If there is space in the offset vector, save
698 the current subject position in the working slot at the top of the vector.
699 We mustn't change the current values of the data slot, because they may be
700 set from a previous iteration of this group, and be referred to by a
701 reference inside the group.
702
703 If the bracket fails to match, we need to restore this value and also the
704 values of the final offsets, in case they were set by a previous iteration
705 of the same bracket.
706
707 If there isn't enough space in the offset vector, treat this as if it were
708 a non-capturing bracket. Don't worry about setting the flag for the error
709 case here; that is handled in the code for KET. */
710
711 case OP_CBRA:
712 case OP_SCBRA:
713 number = GET2(ecode, 1+LINK_SIZE);
714 offset = number << 1;
715
716 #ifdef PCRE_DEBUG
717 printf("start bracket %d\n", number);
718 printf("subject=");
719 pchars(eptr, 16, TRUE, md);
720 printf("\n");
721 #endif
722
723 if (offset < md->offset_max)
724 {
725 save_offset1 = md->offset_vector[offset];
726 save_offset2 = md->offset_vector[offset+1];
727 save_offset3 = md->offset_vector[md->offset_end - number];
728 save_capture_last = md->capture_last;
729
730 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
731 md->offset_vector[md->offset_end - number] = eptr - md->start_subject;
732
733 flags = (op == OP_SCBRA)? match_cbegroup : 0;
734 do
735 {
736 RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
737 ims, eptrb, flags, RM1);
738 if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
739 md->capture_last = save_capture_last;
740 ecode += GET(ecode, 1);
741 }
742 while (*ecode == OP_ALT);
743
744 DPRINTF(("bracket %d failed\n", number));
745
746 md->offset_vector[offset] = save_offset1;
747 md->offset_vector[offset+1] = save_offset2;
748 md->offset_vector[md->offset_end - number] = save_offset3;
749
750 RRETURN(MATCH_NOMATCH);
751 }
752
753 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
754 as a non-capturing bracket. */
755
756 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
757 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
758
759 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
760
761 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
762 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
763
764 /* Non-capturing bracket. Loop for all the alternatives. When we get to the
765 final alternative within the brackets, we would return the result of a
766 recursive call to match() whatever happened. We can reduce stack usage by
767 turning this into a tail recursion, except in the case when match_cbegroup
768 is set.*/
769
770 case OP_BRA:
771 case OP_SBRA:
772 DPRINTF(("start non-capturing bracket\n"));
773 flags = (op >= OP_SBRA)? match_cbegroup : 0;
774 for (;;)
775 {
776 if (ecode[GET(ecode, 1)] != OP_ALT) /* Final alternative */
777 {
778 if (flags == 0) /* Not a possibly empty group */
779 {
780 ecode += _pcre_OP_lengths[*ecode];
781 DPRINTF(("bracket 0 tail recursion\n"));
782 goto TAIL_RECURSE;
783 }
784
785 /* Possibly empty group; can't use tail recursion. */
786
787 RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
788 eptrb, flags, RM48);
789 RRETURN(rrc);
790 }
791
792 /* For non-final alternatives, continue the loop for a NOMATCH result;
793 otherwise return. */
794
795 RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
796 eptrb, flags, RM2);
797 if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
798 ecode += GET(ecode, 1);
799 }
800 /* Control never reaches here. */
801
802 /* Conditional group: compilation checked that there are no more than
803 two branches. If the condition is false, skipping the first branch takes us
804 past the end if there is only one branch, but that's OK because that is
805 exactly what going to the ket would do. As there is only one branch to be
806 obeyed, we can use tail recursion to avoid using another stack frame. */
807
808 case OP_COND:
809 case OP_SCOND:
810 codelink= GET(ecode, 1);
811
812 /* Because of the way auto-callout works during compile, a callout item is
813 inserted between OP_COND and an assertion condition. */
814
815 if (ecode[LINK_SIZE+1] == OP_CALLOUT)
816 {
817 if (pcre_callout != NULL)
818 {
819 pcre_callout_block cb;
820 cb.version = 1; /* Version 1 of the callout block */
821 cb.callout_number = ecode[LINK_SIZE+2];
822 cb.offset_vector = md->offset_vector;
823 cb.subject = (PCRE_SPTR)md->start_subject;
824 cb.subject_length = md->end_subject - md->start_subject;
825 cb.start_match = mstart - md->start_subject;
826 cb.current_position = eptr - md->start_subject;
827 cb.pattern_position = GET(ecode, LINK_SIZE + 3);
828 cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
829 cb.capture_top = offset_top/2;
830 cb.capture_last = md->capture_last;
831 cb.callout_data = md->callout_data;
832 if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);
833 if (rrc < 0) RRETURN(rrc);
834 }
835 ecode += _pcre_OP_lengths[OP_CALLOUT];
836 }
837
838 condcode = ecode[LINK_SIZE+1];
839
840 /* Now see what the actual condition is */
841
842 if (condcode == OP_RREF || condcode == OP_NRREF) /* Recursion test */
843 {
844 if (md->recursive == NULL) /* Not recursing => FALSE */
845 {
846 condition = FALSE;
847 ecode += GET(ecode, 1);
848 }
849 else
850 {
851 int recno = GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/
852 condition = (recno == RREF_ANY || recno == md->recursive->group_num);
853
854 /* If the test is for recursion into a specific subpattern, and it is
855 false, but the test was set up by name, scan the table to see if the
856 name refers to any other numbers, and test them. The condition is true
857 if any one is set. */
858
859 if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
860 {
861 uschar *slotA = md->name_table;
862 for (i = 0; i < md->name_count; i++)
863 {
864 if (GET2(slotA, 0) == recno) break;
865 slotA += md->name_entry_size;
866 }
867
868 /* Found a name for the number - there can be only one; duplicate
869 names for different numbers are allowed, but not vice versa. First
870 scan down for duplicates. */
871
872 if (i < md->name_count)
873 {
874 uschar *slotB = slotA;
875 while (slotB > md->name_table)
876 {
877 slotB -= md->name_entry_size;
878 if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
879 {
880 condition = GET2(slotB, 0) == md->recursive->group_num;
881 if (condition) break;
882 }
883 else break;
884 }
885
886 /* Scan up for duplicates */
887
888 if (!condition)
889 {
890 slotB = slotA;
891 for (i++; i < md->name_count; i++)
892 {
893 slotB += md->name_entry_size;
894 if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
895 {
896 condition = GET2(slotB, 0) == md->recursive->group_num;
897 if (condition) break;
898 }
899 else break;
900 }
901 }
902 }
903 }
904
905 /* Chose branch according to the condition */
906
907 ecode += condition? 3 : GET(ecode, 1);
908 }
909 }
910
911 else if (condcode == OP_CREF || condcode == OP_NCREF) /* Group used test */
912 {
913 offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */
914 condition = offset < offset_top && md->offset_vector[offset] >= 0;
915
916 /* If the numbered capture is unset, but the reference was by name,
917 scan the table to see if the name refers to any other numbers, and test
918 them. The condition is true if any one is set. This is tediously similar
919 to the code above, but not close enough to try to amalgamate. */
920
921 if (!condition && condcode == OP_NCREF)
922 {
923 int refno = offset >> 1;
924 uschar *slotA = md->name_table;
925
926 for (i = 0; i < md->name_count; i++)
927 {
928 if (GET2(slotA, 0) == refno) break;
929 slotA += md->name_entry_size;
930 }
931
932 /* Found a name for the number - there can be only one; duplicate names
933 for different numbers are allowed, but not vice versa. First scan down
934 for duplicates. */
935
936 if (i < md->name_count)
937 {
938 uschar *slotB = slotA;
939 while (slotB > md->name_table)
940 {
941 slotB -= md->name_entry_size;
942 if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
943 {
944 offset = GET2(slotB, 0) << 1;
945 condition = offset < offset_top &&
946 md->offset_vector[offset] >= 0;
947 if (condition) break;
948 }
949 else break;
950 }
951
952 /* Scan up for duplicates */
953
954 if (!condition)
955 {
956 slotB = slotA;
957 for (i++; i < md->name_count; i++)
958 {
959 slotB += md->name_entry_size;
960 if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
961 {
962 offset = GET2(slotB, 0) << 1;
963 condition = offset < offset_top &&
964 md->offset_vector[offset] >= 0;
965 if (condition) break;
966 }
967 else break;
968 }
969 }
970 }
971 }
972
973 /* Chose branch according to the condition */
974
975 ecode += condition? 3 : GET(ecode, 1);
976 }
977
978 else if (condcode == OP_DEF) /* DEFINE - always false */
979 {
980 condition = FALSE;
981 ecode += GET(ecode, 1);
982 }
983
984 /* The condition is an assertion. Call match() to evaluate it - setting
985 the final argument match_condassert causes it to stop at the end of an
986 assertion. */
987
988 else
989 {
990 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
991 match_condassert, RM3);
992 if (rrc == MATCH_MATCH)
993 {
994 condition = TRUE;
995 ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
996 while (*ecode == OP_ALT) ecode += GET(ecode, 1);
997 }
998 else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
999 {
1000 RRETURN(rrc); /* Need braces because of following else */
1001 }
1002 else
1003 {
1004 condition = FALSE;
1005 ecode += codelink;
1006 }
1007 }
1008
1009 /* We are now at the branch that is to be obeyed. As there is only one,
1010 we can use tail recursion to avoid using another stack frame, except when
1011 match_cbegroup is required for an unlimited repeat of a possibly empty
1012 group. If the second alternative doesn't exist, we can just plough on. */
1013
1014 if (condition || *ecode == OP_ALT)
1015 {
1016 ecode += 1 + LINK_SIZE;
1017 if (op == OP_SCOND) /* Possibly empty group */
1018 {
1019 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);
1020 RRETURN(rrc);
1021 }
1022 else /* Group must match something */
1023 {
1024 flags = 0;
1025 goto TAIL_RECURSE;
1026 }
1027 }
1028 else /* Condition false & no alternative */
1029 {
1030 ecode += 1 + LINK_SIZE;
1031 }
1032 break;
1033
1034
1035 /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1036 to close any currently open capturing brackets. */
1037
1038 case OP_CLOSE:
1039 number = GET2(ecode, 1);
1040 offset = number << 1;
1041
1042 #ifdef PCRE_DEBUG
1043 printf("end bracket %d at *ACCEPT", number);
1044 printf("\n");
1045 #endif
1046
1047 md->capture_last = number;
1048 if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1049 {
1050 md->offset_vector[offset] =
1051 md->offset_vector[md->offset_end - number];
1052 md->offset_vector[offset+1] = eptr - md->start_subject;
1053 if (offset_top <= offset) offset_top = offset + 2;
1054 }
1055 ecode += 3;
1056 break;
1057
1058
1059 /* End of the pattern, either real or forced. If we are in a top-level
1060 recursion, we should restore the offsets appropriately and continue from
1061 after the call. */
1062
1063 case OP_ACCEPT:
1064 case OP_END:
1065 if (md->recursive != NULL && md->recursive->group_num == 0)
1066 {
1067 recursion_info *rec = md->recursive;
1068 DPRINTF(("End of pattern in a (?0) recursion\n"));
1069 md->recursive = rec->prevrec;
1070 memmove(md->offset_vector, rec->offset_save,
1071 rec->saved_max * sizeof(int));
1072 offset_top = rec->save_offset_top;
1073 mstart = rec->save_start;
1074 ims = original_ims;
1075 ecode = rec->after_call;
1076 break;
1077 }
1078
1079 /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is
1080 set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of
1081 the subject. In both cases, backtracking will then try other alternatives,
1082 if any. */
1083
1084 if (eptr == mstart &&
1085 (md->notempty ||
1086 (md->notempty_atstart &&
1087 mstart == md->start_subject + md->start_offset)))
1088 RRETURN(MATCH_NOMATCH);
1089
1090 /* Otherwise, we have a match. */
1091
1092 md->end_match_ptr = eptr; /* Record where we ended */
1093 md->end_offset_top = offset_top; /* and how many extracts were taken */
1094 md->start_match_ptr = mstart; /* and the start (\K can modify) */
1095 RRETURN(MATCH_MATCH);
1096
1097 /* Change option settings */
1098
1099 case OP_OPT:
1100 ims = ecode[1];
1101 ecode += 2;
1102 DPRINTF(("ims set to %02lx\n", ims));
1103 break;
1104
1105 /* Assertion brackets. Check the alternative branches in turn - the
1106 matching won't pass the KET for an assertion. If any one branch matches,
1107 the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1108 start of each branch to move the current point backwards, so the code at
1109 this level is identical to the lookahead case. */
1110
1111 case OP_ASSERT:
1112 case OP_ASSERTBACK:
1113 do
1114 {
1115 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
1116 RM4);
1117 if (rrc == MATCH_MATCH) break;
1118 if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1119 ecode += GET(ecode, 1);
1120 }
1121 while (*ecode == OP_ALT);
1122 if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1123
1124 /* If checking an assertion for a condition, return MATCH_MATCH. */
1125
1126 if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);
1127
1128 /* Continue from after the assertion, updating the offsets high water
1129 mark, since extracts may have been taken during the assertion. */
1130
1131 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1132 ecode += 1 + LINK_SIZE;
1133 offset_top = md->end_offset_top;
1134 continue;
1135
1136 /* Negative assertion: all branches must fail to match. Encountering SKIP,
1137 PRUNE, or COMMIT means we must assume failure without checking subsequent
1138 branches. */
1139
1140 case OP_ASSERT_NOT:
1141 case OP_ASSERTBACK_NOT:
1142 do
1143 {
1144 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
1145 RM5);
1146 if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);
1147 if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1148 {
1149 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1150 break;
1151 }
1152 if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1153 ecode += GET(ecode,1);
1154 }
1155 while (*ecode == OP_ALT);
1156
1157 if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);
1158
1159 ecode += 1 + LINK_SIZE;
1160 continue;
1161
1162 /* Move the subject pointer back. This occurs only at the start of
1163 each branch of a lookbehind assertion. If we are too close to the start to
1164 move back, this match function fails. When working with UTF-8 we move
1165 back a number of characters, not bytes. */
1166
1167 case OP_REVERSE:
1168 #ifdef SUPPORT_UTF8
1169 if (utf8)
1170 {
1171 i = GET(ecode, 1);
1172 while (i-- > 0)
1173 {
1174 eptr--;
1175 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1176 BACKCHAR(eptr);
1177 }
1178 }
1179 else
1180 #endif
1181
1182 /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1183
1184 {
1185 eptr -= GET(ecode, 1);
1186 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1187 }
1188
1189 /* Save the earliest consulted character, then skip to next op code */
1190
1191 if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1192 ecode += 1 + LINK_SIZE;
1193 break;
1194
1195 /* The callout item calls an external function, if one is provided, passing
1196 details of the match so far. This is mainly for debugging, though the
1197 function is able to force a failure. */
1198
1199 case OP_CALLOUT:
1200 if (pcre_callout != NULL)
1201 {
1202 pcre_callout_block cb;
1203 cb.version = 1; /* Version 1 of the callout block */
1204 cb.callout_number = ecode[1];
1205 cb.offset_vector = md->offset_vector;
1206 cb.subject = (PCRE_SPTR)md->start_subject;
1207 cb.subject_length = md->end_subject - md->start_subject;
1208 cb.start_match = mstart - md->start_subject;
1209 cb.current_position = eptr - md->start_subject;
1210 cb.pattern_position = GET(ecode, 2);
1211 cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1212 cb.capture_top = offset_top/2;
1213 cb.capture_last = md->capture_last;
1214 cb.callout_data = md->callout_data;
1215 if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1216 if (rrc < 0) RRETURN(rrc);
1217 }
1218 ecode += 2 + 2*LINK_SIZE;
1219 break;
1220
1221 /* Recursion either matches the current regex, or some subexpression. The
1222 offset data is the offset to the starting bracket from the start of the
1223 whole pattern. (This is so that it works from duplicated subpatterns.)
1224
1225 If there are any capturing brackets started but not finished, we have to
1226 save their starting points and reinstate them after the recursion. However,
1227 we don't know how many such there are (offset_top records the completed
1228 total) so we just have to save all the potential data. There may be up to
1229 65535 such values, which is too large to put on the stack, but using malloc
1230 for small numbers seems expensive. As a compromise, the stack is used when
1231 there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc
1232 is used. A problem is what to do if the malloc fails ... there is no way of
1233 returning to the top level with an error. Save the top REC_STACK_SAVE_MAX
1234 values on the stack, and accept that the rest may be wrong.
1235
1236 There are also other values that have to be saved. We use a chained
1237 sequence of blocks that actually live on the stack. Thanks to Robin Houston
1238 for the original version of this logic. */
1239
1240 case OP_RECURSE:
1241 {
1242 callpat = md->start_code + GET(ecode, 1);
1243 new_recursive.group_num = (callpat == md->start_code)? 0 :
1244 GET2(callpat, 1 + LINK_SIZE);
1245
1246 /* Add to "recursing stack" */
1247
1248 new_recursive.prevrec = md->recursive;
1249 md->recursive = &new_recursive;
1250
1251 /* Find where to continue from afterwards */
1252
1253 ecode += 1 + LINK_SIZE;
1254 new_recursive.after_call = ecode;
1255
1256 /* Now save the offset data. */
1257
1258 new_recursive.saved_max = md->offset_end;
1259 if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
1260 new_recursive.offset_save = stacksave;
1261 else
1262 {
1263 new_recursive.offset_save =
1264 (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1265 if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1266 }
1267
1268 memcpy(new_recursive.offset_save, md->offset_vector,
1269 new_recursive.saved_max * sizeof(int));
1270 new_recursive.save_start = mstart;
1271 new_recursive.save_offset_top = offset_top;
1272 mstart = eptr;
1273
1274 /* OK, now we can do the recursion. For each top-level alternative we
1275 restore the offset and recursion data. */
1276
1277 DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1278 flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;
1279 do
1280 {
1281 RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1282 md, ims, eptrb, flags, RM6);
1283 if (rrc == MATCH_MATCH)
1284 {
1285 DPRINTF(("Recursion matched\n"));
1286 md->recursive = new_recursive.prevrec;
1287 if (new_recursive.offset_save != stacksave)
1288 (pcre_free)(new_recursive.offset_save);
1289 RRETURN(MATCH_MATCH);
1290 }
1291 else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1292 {
1293 DPRINTF(("Recursion gave error %d\n", rrc));
1294 if (new_recursive.offset_save != stacksave)
1295 (pcre_free)(new_recursive.offset_save);
1296 RRETURN(rrc);
1297 }
1298
1299 md->recursive = &new_recursive;
1300 memcpy(md->offset_vector, new_recursive.offset_save,
1301 new_recursive.saved_max * sizeof(int));
1302 callpat += GET(callpat, 1);
1303 }
1304 while (*callpat == OP_ALT);
1305
1306 DPRINTF(("Recursion didn't match\n"));
1307 md->recursive = new_recursive.prevrec;
1308 if (new_recursive.offset_save != stacksave)
1309 (pcre_free)(new_recursive.offset_save);
1310 RRETURN(MATCH_NOMATCH);
1311 }
1312 /* Control never reaches here */
1313
1314 /* "Once" brackets are like assertion brackets except that after a match,
1315 the point in the subject string is not moved back. Thus there can never be
1316 a move back into the brackets. Friedl calls these "atomic" subpatterns.
1317 Check the alternative branches in turn - the matching won't pass the KET
1318 for this kind of subpattern. If any one branch matches, we carry on as at
1319 the end of a normal bracket, leaving the subject pointer. */
1320
1321 case OP_ONCE:
1322 prev = ecode;
1323 saved_eptr = eptr;
1324
1325 do
1326 {
1327 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);
1328 if (rrc == MATCH_MATCH) break;
1329 if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1330 ecode += GET(ecode,1);
1331 }
1332 while (*ecode == OP_ALT);
1333
1334 /* If hit the end of the group (which could be repeated), fail */
1335
1336 if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
1337
1338 /* Continue as from after the assertion, updating the offsets high water
1339 mark, since extracts may have been taken. */
1340
1341 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
1342
1343 offset_top = md->end_offset_top;
1344 eptr = md->end_match_ptr;
1345
1346 /* For a non-repeating ket, just continue at this level. This also
1347 happens for a repeating ket if no characters were matched in the group.
1348 This is the forcible breaking of infinite loops as implemented in Perl
1349 5.005. If there is an options reset, it will get obeyed in the normal
1350 course of events. */
1351
1352 if (*ecode == OP_KET || eptr == saved_eptr)
1353 {
1354 ecode += 1+LINK_SIZE;
1355 break;
1356 }
1357
1358 /* The repeating kets try the rest of the pattern or restart from the
1359 preceding bracket, in the appropriate order. The second "call" of match()
1360 uses tail recursion, to avoid using another stack frame. We need to reset
1361 any options that changed within the bracket before re-running it, so
1362 check the next opcode. */
1363
1364 if (ecode[1+LINK_SIZE] == OP_OPT)
1365 {
1366 ims = (ims & ~PCRE_IMS) | ecode[4];
1367 DPRINTF(("ims set to %02lx at group repeat\n", ims));
1368 }
1369
1370 if (*ecode == OP_KETRMIN)
1371 {
1372 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);
1373 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1374 ecode = prev;
1375 flags = 0;
1376 goto TAIL_RECURSE;
1377 }
1378 else /* OP_KETRMAX */
1379 {
1380 RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);
1381 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1382 ecode += 1 + LINK_SIZE;
1383 flags = 0;
1384 goto TAIL_RECURSE;
1385 }
1386 /* Control never gets here */
1387
1388 /* An alternation is the end of a branch; scan along to find the end of the
1389 bracketed group and go to there. */
1390
1391 case OP_ALT:
1392 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1393 break;
1394
1395 /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1396 indicating that it may occur zero times. It may repeat infinitely, or not
1397 at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1398 with fixed upper repeat limits are compiled as a number of copies, with the
1399 optional ones preceded by BRAZERO or BRAMINZERO. */
1400
1401 case OP_BRAZERO:
1402 {
1403 next = ecode+1;
1404 RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);
1405 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1406 do next += GET(next,1); while (*next == OP_ALT);
1407 ecode = next + 1 + LINK_SIZE;
1408 }
1409 break;
1410
1411 case OP_BRAMINZERO:
1412 {
1413 next = ecode+1;
1414 do next += GET(next, 1); while (*next == OP_ALT);
1415 RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);
1416 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1417 ecode++;
1418 }
1419 break;
1420
1421 case OP_SKIPZERO:
1422 {
1423 next = ecode+1;
1424 do next += GET(next,1); while (*next == OP_ALT);
1425 ecode = next + 1 + LINK_SIZE;
1426 }
1427 break;
1428
1429 /* End of a group, repeated or non-repeating. */
1430
1431 case OP_KET:
1432 case OP_KETRMIN:
1433 case OP_KETRMAX:
1434 prev = ecode - GET(ecode, 1);
1435
1436 /* If this was a group that remembered the subject start, in order to break
1437 infinite repeats of empty string matches, retrieve the subject start from
1438 the chain. Otherwise, set it NULL. */
1439
1440 if (*prev >= OP_SBRA)
1441 {
1442 saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */
1443 eptrb = eptrb->epb_prev; /* Backup to previous group */
1444 }
1445 else saved_eptr = NULL;
1446
1447 /* If we are at the end of an assertion group, stop matching and return
1448 MATCH_MATCH, but record the current high water mark for use by positive
1449 assertions. Do this also for the "once" (atomic) groups. */
1450
1451 if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1452 *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
1453 *prev == OP_ONCE)
1454 {
1455 md->end_match_ptr = eptr; /* For ONCE */
1456 md->end_offset_top = offset_top;
1457 RRETURN(MATCH_MATCH);
1458 }
1459
1460 /* For capturing groups we have to check the group number back at the start
1461 and if necessary complete handling an extraction by setting the offsets and
1462 bumping the high water mark. Note that whole-pattern recursion is coded as
1463 a recurse into group 0, so it won't be picked up here. Instead, we catch it
1464 when the OP_END is reached. Other recursion is handled here. */
1465
1466 if (*prev == OP_CBRA || *prev == OP_SCBRA)
1467 {
1468 number = GET2(prev, 1+LINK_SIZE);
1469 offset = number << 1;
1470
1471 #ifdef PCRE_DEBUG
1472 printf("end bracket %d", number);
1473 printf("\n");
1474 #endif
1475
1476 md->capture_last = number;
1477 if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1478 {
1479 md->offset_vector[offset] =
1480 md->offset_vector[md->offset_end - number];
1481 md->offset_vector[offset+1] = eptr - md->start_subject;
1482 if (offset_top <= offset) offset_top = offset + 2;
1483 }
1484
1485 /* Handle a recursively called group. Restore the offsets
1486 appropriately and continue from after the call. */
1487
1488 if (md->recursive != NULL && md->recursive->group_num == number)
1489 {
1490 recursion_info *rec = md->recursive;
1491 DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
1492 md->recursive = rec->prevrec;
1493 mstart = rec->save_start;
1494 memcpy(md->offset_vector, rec->offset_save,
1495 rec->saved_max * sizeof(int));
1496 offset_top = rec->save_offset_top;
1497 ecode = rec->after_call;
1498 ims = original_ims;
1499 break;
1500 }
1501 }
1502
1503 /* For both capturing and non-capturing groups, reset the value of the ims
1504 flags, in case they got changed during the group. */
1505
1506 ims = original_ims;
1507 DPRINTF(("ims reset to %02lx\n", ims));
1508
1509 /* For a non-repeating ket, just continue at this level. This also
1510 happens for a repeating ket if no characters were matched in the group.
1511 This is the forcible breaking of infinite loops as implemented in Perl
1512 5.005. If there is an options reset, it will get obeyed in the normal
1513 course of events. */
1514
1515 if (*ecode == OP_KET || eptr == saved_eptr)
1516 {
1517 ecode += 1 + LINK_SIZE;
1518 break;
1519 }
1520
1521 /* The repeating kets try the rest of the pattern or restart from the
1522 preceding bracket, in the appropriate order. In the second case, we can use
1523 tail recursion to avoid using another stack frame, unless we have an
1524 unlimited repeat of a group that can match an empty string. */
1525
1526 flags = (*prev >= OP_SBRA)? match_cbegroup : 0;
1527
1528 if (*ecode == OP_KETRMIN)
1529 {
1530 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);
1531 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1532 if (flags != 0) /* Could match an empty string */
1533 {
1534 RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);
1535 RRETURN(rrc);
1536 }
1537 ecode = prev;
1538 goto TAIL_RECURSE;
1539 }
1540 else /* OP_KETRMAX */
1541 {
1542 RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);
1543 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1544 ecode += 1 + LINK_SIZE;
1545 flags = 0;
1546 goto TAIL_RECURSE;
1547 }
1548 /* Control never gets here */
1549
1550 /* Start of subject unless notbol, or after internal newline if multiline */
1551
1552 case OP_CIRC:
1553 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
1554 if ((ims & PCRE_MULTILINE) != 0)
1555 {
1556 if (eptr != md->start_subject &&
1557 (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1558 RRETURN(MATCH_NOMATCH);
1559 ecode++;
1560 break;
1561 }
1562 /* ... else fall through */
1563
1564 /* Start of subject assertion */
1565
1566 case OP_SOD:
1567 if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
1568 ecode++;
1569 break;
1570
1571 /* Start of match assertion */
1572
1573 case OP_SOM:
1574 if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
1575 ecode++;
1576 break;
1577
1578 /* Reset the start of match point */
1579
1580 case OP_SET_SOM:
1581 mstart = eptr;
1582 ecode++;
1583 break;
1584
1585 /* Assert before internal newline if multiline, or before a terminating
1586 newline unless endonly is set, else end of subject unless noteol is set. */
1587
1588 case OP_DOLL:
1589 if ((ims & PCRE_MULTILINE) != 0)
1590 {
1591 if (eptr < md->end_subject)
1592 { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
1593 else
1594 { if (md->noteol) RRETURN(MATCH_NOMATCH); }
1595 ecode++;
1596 break;
1597 }
1598 else
1599 {
1600 if (md->noteol) RRETURN(MATCH_NOMATCH);
1601 if (!md->endonly)
1602 {
1603 if (eptr != md->end_subject &&
1604 (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1605 RRETURN(MATCH_NOMATCH);
1606 ecode++;
1607 break;
1608 }
1609 }
1610 /* ... else fall through for endonly */
1611
1612 /* End of subject assertion (\z) */
1613
1614 case OP_EOD:
1615 if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
1616 ecode++;
1617 break;
1618
1619 /* End of subject or ending \n assertion (\Z) */
1620
1621 case OP_EODN:
1622 if (eptr != md->end_subject &&
1623 (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1624 RRETURN(MATCH_NOMATCH);
1625 ecode++;
1626 break;
1627
1628 /* Word boundary assertions */
1629
1630 case OP_NOT_WORD_BOUNDARY:
1631 case OP_WORD_BOUNDARY:
1632 {
1633
1634 /* Find out if the previous and current characters are "word" characters.
1635 It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
1636 be "non-word" characters. Remember the earliest consulted character for
1637 partial matching. */
1638
1639 #ifdef SUPPORT_UTF8
1640 if (utf8)
1641 {
1642 if (eptr == md->start_subject) prev_is_word = FALSE; else
1643 {
1644 USPTR lastptr = eptr - 1;
1645 while((*lastptr & 0xc0) == 0x80) lastptr--;
1646 if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
1647 GETCHAR(c, lastptr);
1648 prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1649 }
1650 if (eptr >= md->end_subject)
1651 {
1652 SCHECK_PARTIAL();
1653 cur_is_word = FALSE;
1654 }
1655 else
1656 {
1657 GETCHAR(c, eptr);
1658 cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1659 }
1660 }
1661 else
1662 #endif
1663
1664 /* Not in UTF-8 mode */
1665
1666 {
1667 if (eptr == md->start_subject) prev_is_word = FALSE; else
1668 {
1669 if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
1670 prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
1671 }
1672 if (eptr >= md->end_subject)
1673 {
1674 SCHECK_PARTIAL();
1675 cur_is_word = FALSE;
1676 }
1677 else cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
1678 }
1679
1680 /* Now see if the situation is what we want */
1681
1682 if ((*ecode++ == OP_WORD_BOUNDARY)?
1683 cur_is_word == prev_is_word : cur_is_word != prev_is_word)
1684 RRETURN(MATCH_NOMATCH);
1685 }
1686 break;
1687
1688 /* Match a single character type; inline for speed */
1689
1690 case OP_ANY:
1691 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
1692 /* Fall through */
1693
1694 case OP_ALLANY:
1695 if (eptr++ >= md->end_subject)
1696 {
1697 SCHECK_PARTIAL();
1698 RRETURN(MATCH_NOMATCH);
1699 }
1700 if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
1701 ecode++;
1702 break;
1703
1704 /* Match a single byte, even in UTF-8 mode. This opcode really does match
1705 any byte, even newline, independent of the setting of PCRE_DOTALL. */
1706
1707 case OP_ANYBYTE:
1708 if (eptr++ >= md->end_subject)
1709 {
1710 SCHECK_PARTIAL();
1711 RRETURN(MATCH_NOMATCH);
1712 }
1713 ecode++;
1714 break;
1715
1716 case OP_NOT_DIGIT:
1717 if (eptr >= md->end_subject)
1718 {
1719 SCHECK_PARTIAL();
1720 RRETURN(MATCH_NOMATCH);
1721 }
1722 GETCHARINCTEST(c, eptr);
1723 if (
1724 #ifdef SUPPORT_UTF8
1725 c < 256 &&
1726 #endif
1727 (md->ctypes[c] & ctype_digit) != 0
1728 )
1729 RRETURN(MATCH_NOMATCH);
1730 ecode++;
1731 break;
1732
1733 case OP_DIGIT:
1734 if (eptr >= md->end_subject)
1735 {
1736 SCHECK_PARTIAL();
1737 RRETURN(MATCH_NOMATCH);
1738 }
1739 GETCHARINCTEST(c, eptr);
1740 if (
1741 #ifdef SUPPORT_UTF8
1742 c >= 256 ||
1743 #endif
1744 (md->ctypes[c] & ctype_digit) == 0
1745 )
1746 RRETURN(MATCH_NOMATCH);
1747 ecode++;
1748 break;
1749
1750 case OP_NOT_WHITESPACE:
1751 if (eptr >= md->end_subject)
1752 {
1753 SCHECK_PARTIAL();
1754 RRETURN(MATCH_NOMATCH);
1755 }
1756 GETCHARINCTEST(c, eptr);
1757 if (
1758 #ifdef SUPPORT_UTF8
1759 c < 256 &&
1760 #endif
1761 (md->ctypes[c] & ctype_space) != 0
1762 )
1763 RRETURN(MATCH_NOMATCH);
1764 ecode++;
1765 break;
1766
1767 case OP_WHITESPACE:
1768 if (eptr >= md->end_subject)
1769 {
1770 SCHECK_PARTIAL();
1771 RRETURN(MATCH_NOMATCH);
1772 }
1773 GETCHARINCTEST(c, eptr);
1774 if (
1775 #ifdef SUPPORT_UTF8
1776 c >= 256 ||
1777 #endif
1778 (md->ctypes[c] & ctype_space) == 0
1779 )
1780 RRETURN(MATCH_NOMATCH);
1781 ecode++;
1782 break;
1783
1784 case OP_NOT_WORDCHAR:
1785 if (eptr >= md->end_subject)
1786 {
1787 SCHECK_PARTIAL();
1788 RRETURN(MATCH_NOMATCH);
1789 }
1790 GETCHARINCTEST(c, eptr);
1791 if (
1792 #ifdef SUPPORT_UTF8
1793 c < 256 &&
1794 #endif
1795 (md->ctypes[c] & ctype_word) != 0
1796 )
1797 RRETURN(MATCH_NOMATCH);
1798 ecode++;
1799 break;
1800
1801 case OP_WORDCHAR:
1802 if (eptr >= md->end_subject)
1803 {
1804 SCHECK_PARTIAL();
1805 RRETURN(MATCH_NOMATCH);
1806 }
1807 GETCHARINCTEST(c, eptr);
1808 if (
1809 #ifdef SUPPORT_UTF8
1810 c >= 256 ||
1811 #endif
1812 (md->ctypes[c] & ctype_word) == 0
1813 )
1814 RRETURN(MATCH_NOMATCH);
1815 ecode++;
1816 break;
1817
1818 case OP_ANYNL:
1819 if (eptr >= md->end_subject)
1820 {
1821 SCHECK_PARTIAL();
1822 RRETURN(MATCH_NOMATCH);
1823 }
1824 GETCHARINCTEST(c, eptr);
1825 switch(c)
1826 {
1827 default: RRETURN(MATCH_NOMATCH);
1828 case 0x000d:
1829 if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
1830 break;
1831
1832 case 0x000a:
1833 break;
1834
1835 case 0x000b:
1836 case 0x000c:
1837 case 0x0085:
1838 case 0x2028:
1839 case 0x2029:
1840 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
1841 break;
1842 }
1843 ecode++;
1844 break;
1845
1846 case OP_NOT_HSPACE:
1847 if (eptr >= md->end_subject)
1848 {
1849 SCHECK_PARTIAL();
1850 RRETURN(MATCH_NOMATCH);
1851 }
1852 GETCHARINCTEST(c, eptr);
1853 switch(c)
1854 {
1855 default: break;
1856 case 0x09: /* HT */
1857 case 0x20: /* SPACE */
1858 case 0xa0: /* NBSP */
1859 case 0x1680: /* OGHAM SPACE MARK */
1860 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
1861 case 0x2000: /* EN QUAD */
1862 case 0x2001: /* EM QUAD */
1863 case 0x2002: /* EN SPACE */
1864 case 0x2003: /* EM SPACE */
1865 case 0x2004: /* THREE-PER-EM SPACE */
1866 case 0x2005: /* FOUR-PER-EM SPACE */
1867 case 0x2006: /* SIX-PER-EM SPACE */
1868 case 0x2007: /* FIGURE SPACE */
1869 case 0x2008: /* PUNCTUATION SPACE */
1870 case 0x2009: /* THIN SPACE */
1871 case 0x200A: /* HAIR SPACE */
1872 case 0x202f: /* NARROW NO-BREAK SPACE */
1873 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
1874 case 0x3000: /* IDEOGRAPHIC SPACE */
1875 RRETURN(MATCH_NOMATCH);
1876 }
1877 ecode++;
1878 break;
1879
1880 case OP_HSPACE:
1881 if (eptr >= md->end_subject)
1882 {
1883 SCHECK_PARTIAL();
1884 RRETURN(MATCH_NOMATCH);
1885 }
1886 GETCHARINCTEST(c, eptr);
1887 switch(c)
1888 {
1889 default: RRETURN(MATCH_NOMATCH);
1890 case 0x09: /* HT */
1891 case 0x20: /* SPACE */
1892 case 0xa0: /* NBSP */
1893 case 0x1680: /* OGHAM SPACE MARK */
1894 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
1895 case 0x2000: /* EN QUAD */
1896 case 0x2001: /* EM QUAD */
1897 case 0x2002: /* EN SPACE */
1898 case 0x2003: /* EM SPACE */
1899 case 0x2004: /* THREE-PER-EM SPACE */
1900 case 0x2005: /* FOUR-PER-EM SPACE */
1901 case 0x2006: /* SIX-PER-EM SPACE */
1902 case 0x2007: /* FIGURE SPACE */
1903 case 0x2008: /* PUNCTUATION SPACE */
1904 case 0x2009: /* THIN SPACE */
1905 case 0x200A: /* HAIR SPACE */
1906 case 0x202f: /* NARROW NO-BREAK SPACE */
1907 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
1908 case 0x3000: /* IDEOGRAPHIC SPACE */
1909 break;
1910 }
1911 ecode++;
1912 break;
1913
1914 case OP_NOT_VSPACE:
1915 if (eptr >= md->end_subject)
1916 {
1917 SCHECK_PARTIAL();
1918 RRETURN(MATCH_NOMATCH);
1919 }
1920 GETCHARINCTEST(c, eptr);
1921 switch(c)
1922 {
1923 default: break;
1924 case 0x0a: /* LF */
1925 case 0x0b: /* VT */
1926 case 0x0c: /* FF */
1927 case 0x0d: /* CR */
1928 case 0x85: /* NEL */
1929 case 0x2028: /* LINE SEPARATOR */
1930 case 0x2029: /* PARAGRAPH SEPARATOR */
1931 RRETURN(MATCH_NOMATCH);
1932 }
1933 ecode++;
1934 break;
1935
1936 case OP_VSPACE:
1937 if (eptr >= md->end_subject)
1938 {
1939 SCHECK_PARTIAL();
1940 RRETURN(MATCH_NOMATCH);
1941 }
1942 GETCHARINCTEST(c, eptr);
1943 switch(c)
1944 {
1945 default: RRETURN(MATCH_NOMATCH);
1946 case 0x0a: /* LF */
1947 case 0x0b: /* VT */
1948 case 0x0c: /* FF */
1949 case 0x0d: /* CR */
1950 case 0x85: /* NEL */
1951 case 0x2028: /* LINE SEPARATOR */
1952 case 0x2029: /* PARAGRAPH SEPARATOR */
1953 break;
1954 }
1955 ecode++;
1956 break;
1957
1958 #ifdef SUPPORT_UCP
1959 /* Check the next character by Unicode property. We will get here only
1960 if the support is in the binary; otherwise a compile-time error occurs. */
1961
1962 case OP_PROP:
1963 case OP_NOTPROP:
1964 if (eptr >= md->end_subject)
1965 {
1966 SCHECK_PARTIAL();
1967 RRETURN(MATCH_NOMATCH);
1968 }
1969 GETCHARINCTEST(c, eptr);
1970 {
1971 const ucd_record *prop = GET_UCD(c);
1972
1973 switch(ecode[1])
1974 {
1975 case PT_ANY:
1976 if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
1977 break;
1978
1979 case PT_LAMP:
1980 if ((prop->chartype == ucp_Lu ||
1981 prop->chartype == ucp_Ll ||
1982 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
1983 RRETURN(MATCH_NOMATCH);
1984 break;
1985
1986 case PT_GC:
1987 if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
1988 RRETURN(MATCH_NOMATCH);
1989 break;
1990
1991 case PT_PC:
1992 if ((ecode[2] != prop->chartype) == (op == OP_PROP))
1993 RRETURN(MATCH_NOMATCH);
1994 break;
1995
1996 case PT_SC:
1997 if ((ecode[2] != prop->script) == (op == OP_PROP))
1998 RRETURN(MATCH_NOMATCH);
1999 break;
2000
2001 default:
2002 RRETURN(PCRE_ERROR_INTERNAL);
2003 }
2004
2005 ecode += 3;
2006 }
2007 break;
2008
2009 /* Match an extended Unicode sequence. We will get here only if the support
2010 is in the binary; otherwise a compile-time error occurs. */
2011
2012 case OP_EXTUNI:
2013 if (eptr >= md->end_subject)
2014 {
2015 SCHECK_PARTIAL();
2016 RRETURN(MATCH_NOMATCH);
2017 }
2018 GETCHARINCTEST(c, eptr);
2019 {
2020 int category = UCD_CATEGORY(c);
2021 if (category == ucp_M) RRETURN(MATCH_NOMATCH);
2022 while (eptr < md->end_subject)
2023 {
2024 int len = 1;
2025 if (!utf8) c = *eptr; else
2026 {
2027 GETCHARLEN(c, eptr, len);
2028 }
2029 category = UCD_CATEGORY(c);
2030 if (category != ucp_M) break;
2031 eptr += len;
2032 }
2033 }
2034 ecode++;
2035 break;
2036 #endif
2037
2038
2039 /* Match a back reference, possibly repeatedly. Look past the end of the
2040 item to see if there is repeat information following. The code is similar
2041 to that for character classes, but repeated for efficiency. Then obey
2042 similar code to character type repeats - written out again for speed.
2043 However, if the referenced string is the empty string, always treat
2044 it as matched, any number of times (otherwise there could be infinite
2045 loops). */
2046
2047 case OP_REF:
2048 {
2049 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
2050 ecode += 3;
2051
2052 /* If the reference is unset, there are two possibilities:
2053
2054 (a) In the default, Perl-compatible state, set the length to be longer
2055 than the amount of subject left; this ensures that every attempt at a
2056 match fails. We can't just fail here, because of the possibility of
2057 quantifiers with zero minima.
2058
2059 (b) If the JavaScript compatibility flag is set, set the length to zero
2060 so that the back reference matches an empty string.
2061
2062 Otherwise, set the length to the length of what was matched by the
2063 referenced subpattern. */
2064
2065 if (offset >= offset_top || md->offset_vector[offset] < 0)
2066 length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;
2067 else
2068 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2069
2070 /* Set up for repetition, or handle the non-repeated case */
2071
2072 switch (*ecode)
2073 {
2074 case OP_CRSTAR:
2075 case OP_CRMINSTAR:
2076 case OP_CRPLUS:
2077 case OP_CRMINPLUS:
2078 case OP_CRQUERY:
2079 case OP_CRMINQUERY:
2080 c = *ecode++ - OP_CRSTAR;
2081 minimize = (c & 1) != 0;
2082 min = rep_min[c]; /* Pick up values from tables; */
2083 max = rep_max[c]; /* zero for max => infinity */
2084 if (max == 0) max = INT_MAX;
2085 break;
2086
2087 case OP_CRRANGE:
2088 case OP_CRMINRANGE:
2089 minimize = (*ecode == OP_CRMINRANGE);
2090 min = GET2(ecode, 1);
2091 max = GET2(ecode, 3);
2092 if (max == 0) max = INT_MAX;
2093 ecode += 5;
2094 break;
2095
2096 default: /* No repeat follows */
2097 if (!match_ref(offset, eptr, length, md, ims))
2098 {
2099 CHECK_PARTIAL();
2100 RRETURN(MATCH_NOMATCH);
2101 }
2102 eptr += length;
2103 continue; /* With the main loop */
2104 }
2105
2106 /* If the length of the reference is zero, just continue with the
2107 main loop. */
2108
2109 if (length == 0) continue;
2110
2111 /* First, ensure the minimum number of matches are present. We get back
2112 the length of the reference string explicitly rather than passing the
2113 address of eptr, so that eptr can be a register variable. */
2114
2115 for (i = 1; i <= min; i++)
2116 {
2117 if (!match_ref(offset, eptr, length, md, ims))
2118 {
2119 CHECK_PARTIAL();
2120 RRETURN(MATCH_NOMATCH);
2121 }
2122 eptr += length;
2123 }
2124
2125 /* If min = max, continue at the same level without recursion.
2126 They are not both allowed to be zero. */
2127
2128 if (min == max) continue;
2129
2130 /* If minimizing, keep trying and advancing the pointer */
2131
2132 if (minimize)
2133 {
2134 for (fi = min;; fi++)
2135 {
2136 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);
2137 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2138 if (fi >= max) RRETURN(MATCH_NOMATCH);
2139 if (!match_ref(offset, eptr, length, md, ims))
2140 {
2141 CHECK_PARTIAL();
2142 RRETURN(MATCH_NOMATCH);
2143 }
2144 eptr += length;
2145 }
2146 /* Control never gets here */
2147 }
2148
2149 /* If maximizing, find the longest string and work backwards */
2150
2151 else
2152 {
2153 pp = eptr;
2154 for (i = min; i < max; i++)
2155 {
2156 if (!match_ref(offset, eptr, length, md, ims))
2157 {
2158 CHECK_PARTIAL();
2159 break;
2160 }
2161 eptr += length;
2162 }
2163 while (eptr >= pp)
2164 {
2165 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);
2166 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2167 eptr -= length;
2168 }
2169 RRETURN(MATCH_NOMATCH);
2170 }
2171 }
2172 /* Control never gets here */
2173
2174 /* Match a bit-mapped character class, possibly repeatedly. This op code is
2175 used when all the characters in the class have values in the range 0-255,
2176 and either the matching is caseful, or the characters are in the range
2177 0-127 when UTF-8 processing is enabled. The only difference between
2178 OP_CLASS and OP_NCLASS occurs when a data character outside the range is
2179 encountered.
2180
2181 First, look past the end of the item to see if there is repeat information
2182 following. Then obey similar code to character type repeats - written out
2183 again for speed. */
2184
2185 case OP_NCLASS:
2186 case OP_CLASS:
2187 {
2188 data = ecode + 1; /* Save for matching */
2189 ecode += 33; /* Advance past the item */
2190
2191 switch (*ecode)
2192 {
2193 case OP_CRSTAR:
2194 case OP_CRMINSTAR:
2195 case OP_CRPLUS:
2196 case OP_CRMINPLUS:
2197 case OP_CRQUERY:
2198 case OP_CRMINQUERY:
2199 c = *ecode++ - OP_CRSTAR;
2200 minimize = (c & 1) != 0;
2201 min = rep_min[c]; /* Pick up values from tables; */
2202 max = rep_max[c]; /* zero for max => infinity */
2203 if (max == 0) max = INT_MAX;
2204 break;
2205
2206 case OP_CRRANGE:
2207 case OP_CRMINRANGE:
2208 minimize = (*ecode == OP_CRMINRANGE);
2209 min = GET2(ecode, 1);
2210 max = GET2(ecode, 3);
2211 if (max == 0) max = INT_MAX;
2212 ecode += 5;
2213 break;
2214
2215 default: /* No repeat follows */
2216 min = max = 1;
2217 break;
2218 }
2219
2220 /* First, ensure the minimum number of matches are present. */
2221
2222 #ifdef SUPPORT_UTF8
2223 /* UTF-8 mode */
2224 if (utf8)
2225 {
2226 for (i = 1; i <= min; i++)
2227 {
2228 if (eptr >= md->end_subject)
2229 {
2230 SCHECK_PARTIAL();
2231 RRETURN(MATCH_NOMATCH);
2232 }
2233 GETCHARINC(c, eptr);
2234 if (c > 255)
2235 {
2236 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2237 }
2238 else
2239 {
2240 if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2241 }
2242 }
2243 }
2244 else
2245 #endif
2246 /* Not UTF-8 mode */
2247 {
2248 for (i = 1; i <= min; i++)
2249 {
2250 if (eptr >= md->end_subject)
2251 {
2252 SCHECK_PARTIAL();
2253 RRETURN(MATCH_NOMATCH);
2254 }
2255 c = *eptr++;
2256 if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2257 }
2258 }
2259
2260 /* If max == min we can continue with the main loop without the
2261 need to recurse. */
2262
2263 if (min == max) continue;
2264
2265 /* If minimizing, keep testing the rest of the expression and advancing
2266 the pointer while it matches the class. */
2267
2268 if (minimize)
2269 {
2270 #ifdef SUPPORT_UTF8
2271 /* UTF-8 mode */
2272 if (utf8)
2273 {
2274 for (fi = min;; fi++)
2275 {
2276 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);
2277 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2278 if (fi >= max) RRETURN(MATCH_NOMATCH);
2279 if (eptr >= md->end_subject)
2280 {
2281 SCHECK_PARTIAL();
2282 RRETURN(MATCH_NOMATCH);
2283 }
2284 GETCHARINC(c, eptr);
2285 if (c > 255)
2286 {
2287 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2288 }
2289 else
2290 {
2291 if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2292 }
2293 }
2294 }
2295 else
2296 #endif
2297 /* Not UTF-8 mode */
2298 {
2299 for (fi = min;; fi++)
2300 {
2301 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);
2302 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2303 if (fi >= max) RRETURN(MATCH_NOMATCH);
2304 if (eptr >= md->end_subject)
2305 {
2306 SCHECK_PARTIAL();
2307 RRETURN(MATCH_NOMATCH);
2308 }
2309 c = *eptr++;
2310 if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2311 }
2312 }
2313 /* Control never gets here */
2314 }
2315
2316 /* If maximizing, find the longest possible run, then work backwards. */
2317
2318 else
2319 {
2320 pp = eptr;
2321
2322 #ifdef SUPPORT_UTF8
2323 /* UTF-8 mode */
2324 if (utf8)
2325 {
2326 for (i = min; i < max; i++)
2327 {
2328 int len = 1;
2329 if (eptr >= md->end_subject)
2330 {
2331 SCHECK_PARTIAL();
2332 break;
2333 }
2334 GETCHARLEN(c, eptr, len);
2335 if (c > 255)
2336 {
2337 if (op == OP_CLASS) break;
2338 }
2339 else
2340 {
2341 if ((data[c/8] & (1 << (c&7))) == 0) break;
2342 }
2343 eptr += len;
2344 }
2345 for (;;)
2346 {
2347 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);
2348 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2349 if (eptr-- == pp) break; /* Stop if tried at original pos */
2350 BACKCHAR(eptr);
2351 }
2352 }
2353 else
2354 #endif
2355 /* Not UTF-8 mode */
2356 {
2357 for (i = min; i < max; i++)
2358 {
2359 if (eptr >= md->end_subject)
2360 {
2361 SCHECK_PARTIAL();
2362 break;
2363 }
2364 c = *eptr;
2365 if ((data[c/8] & (1 << (c&7))) == 0) break;
2366 eptr++;
2367 }
2368 while (eptr >= pp)
2369 {
2370 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);
2371 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2372 eptr--;
2373 }
2374 }
2375
2376 RRETURN(MATCH_NOMATCH);
2377 }
2378 }
2379 /* Control never gets here */
2380
2381
2382 /* Match an extended character class. This opcode is encountered only
2383 when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
2384 mode, because Unicode properties are supported in non-UTF-8 mode. */
2385
2386 #ifdef SUPPORT_UTF8
2387 case OP_XCLASS:
2388 {
2389 data = ecode + 1 + LINK_SIZE; /* Save for matching */
2390 ecode += GET(ecode, 1); /* Advance past the item */
2391
2392 switch (*ecode)
2393 {
2394 case OP_CRSTAR:
2395 case OP_CRMINSTAR:
2396 case OP_CRPLUS:
2397 case OP_CRMINPLUS:
2398 case OP_CRQUERY:
2399 case OP_CRMINQUERY:
2400 c = *ecode++ - OP_CRSTAR;
2401 minimize = (c & 1) != 0;
2402 min = rep_min[c]; /* Pick up values from tables; */
2403 max = rep_max[c]; /* zero for max => infinity */
2404 if (max == 0) max = INT_MAX;
2405 break;
2406
2407 case OP_CRRANGE:
2408 case OP_CRMINRANGE:
2409 minimize = (*ecode == OP_CRMINRANGE);
2410 min = GET2(ecode, 1);
2411 max = GET2(ecode, 3);
2412 if (max == 0) max = INT_MAX;
2413 ecode += 5;
2414 break;
2415
2416 default: /* No repeat follows */
2417 min = max = 1;
2418 break;
2419 }
2420
2421 /* First, ensure the minimum number of matches are present. */
2422
2423 for (i = 1; i <= min; i++)
2424 {
2425 if (eptr >= md->end_subject)
2426 {
2427 SCHECK_PARTIAL();
2428 RRETURN(MATCH_NOMATCH);
2429 }
2430 GETCHARINCTEST(c, eptr);
2431 if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
2432 }
2433
2434 /* If max == min we can continue with the main loop without the
2435 need to recurse. */
2436
2437 if (min == max) continue;
2438
2439 /* If minimizing, keep testing the rest of the expression and advancing
2440 the pointer while it matches the class. */
2441
2442 if (minimize)
2443 {
2444 for (fi = min;; fi++)
2445 {
2446 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);
2447 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2448 if (fi >= max) RRETURN(MATCH_NOMATCH);
2449 if (eptr >= md->end_subject)
2450 {
2451 SCHECK_PARTIAL();
2452 RRETURN(MATCH_NOMATCH);
2453 }
2454 GETCHARINCTEST(c, eptr);
2455 if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
2456 }
2457 /* Control never gets here */
2458 }
2459
2460 /* If maximizing, find the longest possible run, then work backwards. */
2461
2462 else
2463 {
2464 pp = eptr;
2465 for (i = min; i < max; i++)
2466 {
2467 int len = 1;
2468 if (eptr >= md->end_subject)
2469 {
2470 SCHECK_PARTIAL();
2471 break;
2472 }
2473 GETCHARLENTEST(c, eptr, len);
2474 if (!_pcre_xclass(c, data)) break;
2475 eptr += len;
2476 }
2477 for(;;)
2478 {
2479 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);
2480 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2481 if (eptr-- == pp) break; /* Stop if tried at original pos */
2482 if (utf8) BACKCHAR(eptr);
2483 }
2484 RRETURN(MATCH_NOMATCH);
2485 }
2486
2487 /* Control never gets here */
2488 }
2489 #endif /* End of XCLASS */
2490
2491 /* Match a single character, casefully */
2492
2493 case OP_CHAR:
2494 #ifdef SUPPORT_UTF8
2495 if (utf8)
2496 {
2497 length = 1;
2498 ecode++;
2499 GETCHARLEN(fc, ecode, length);
2500 if (length > md->end_subject - eptr)
2501 {
2502 CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */
2503 RRETURN(MATCH_NOMATCH);
2504 }
2505 while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
2506 }
2507 else
2508 #endif
2509
2510 /* Non-UTF-8 mode */
2511 {
2512 if (md->end_subject - eptr < 1)
2513 {
2514 SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */
2515 RRETURN(MATCH_NOMATCH);
2516 }
2517 if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
2518 ecode += 2;
2519 }
2520 break;
2521
2522 /* Match a single character, caselessly */
2523
2524 case OP_CHARNC:
2525 #ifdef SUPPORT_UTF8
2526 if (utf8)
2527 {
2528 length = 1;
2529 ecode++;
2530 GETCHARLEN(fc, ecode, length);
2531
2532 if (length > md->end_subject - eptr)
2533 {
2534 CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */
2535 RRETURN(MATCH_NOMATCH);
2536 }
2537
2538 /* If the pattern character's value is < 128, we have only one byte, and
2539 can use the fast lookup table. */
2540
2541 if (fc < 128)
2542 {
2543 if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
2544 }
2545
2546 /* Otherwise we must pick up the subject character */
2547
2548 else
2549 {
2550 unsigned int dc;
2551 GETCHARINC(dc, eptr);
2552 ecode += length;
2553
2554 /* If we have Unicode property support, we can use it to test the other
2555 case of the character, if there is one. */
2556
2557 if (fc != dc)
2558 {
2559 #ifdef SUPPORT_UCP
2560 if (dc != UCD_OTHERCASE(fc))
2561 #endif
2562 RRETURN(MATCH_NOMATCH);
2563 }
2564 }
2565 }
2566 else
2567 #endif /* SUPPORT_UTF8 */
2568
2569 /* Non-UTF-8 mode */
2570 {
2571 if (md->end_subject - eptr < 1)
2572 {
2573 SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */
2574 RRETURN(MATCH_NOMATCH);
2575 }
2576 if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
2577 ecode += 2;
2578 }
2579 break;
2580
2581 /* Match a single character repeatedly. */
2582
2583 case OP_EXACT:
2584 min = max = GET2(ecode, 1);
2585 ecode += 3;
2586 goto REPEATCHAR;
2587
2588 case OP_POSUPTO:
2589 possessive = TRUE;
2590 /* Fall through */
2591
2592 case OP_UPTO:
2593 case OP_MINUPTO:
2594 min = 0;
2595 max = GET2(ecode, 1);
2596 minimize = *ecode == OP_MINUPTO;
2597 ecode += 3;
2598 goto REPEATCHAR;
2599
2600 case OP_POSSTAR:
2601 possessive = TRUE;
2602 min = 0;
2603 max = INT_MAX;
2604 ecode++;
2605 goto REPEATCHAR;
2606
2607 case OP_POSPLUS:
2608 possessive = TRUE;
2609 min = 1;
2610 max = INT_MAX;
2611 ecode++;
2612 goto REPEATCHAR;
2613
2614 case OP_POSQUERY:
2615 possessive = TRUE;
2616 min = 0;
2617 max = 1;
2618 ecode++;
2619 goto REPEATCHAR;
2620
2621 case OP_STAR:
2622 case OP_MINSTAR:
2623 case OP_PLUS:
2624 case OP_MINPLUS:
2625 case OP_QUERY:
2626 case OP_MINQUERY:
2627 c = *ecode++ - OP_STAR;
2628 minimize = (c & 1) != 0;
2629
2630 min = rep_min[c]; /* Pick up values from tables; */
2631 max = rep_max[c]; /* zero for max => infinity */
2632 if (max == 0) max = INT_MAX;
2633
2634 /* Common code for all repeated single-character matches. */
2635
2636 REPEATCHAR:
2637 #ifdef SUPPORT_UTF8
2638 if (utf8)
2639 {
2640 length = 1;
2641 charptr = ecode;
2642 GETCHARLEN(fc, ecode, length);
2643 ecode += length;
2644
2645 /* Handle multibyte character matching specially here. There is
2646 support for caseless matching if UCP support is present. */
2647
2648 if (length > 1)
2649 {
2650 #ifdef SUPPORT_UCP
2651 unsigned int othercase;
2652 if ((ims & PCRE_CASELESS) != 0 &&
2653 (othercase = UCD_OTHERCASE(fc)) != fc)
2654 oclength = _pcre_ord2utf8(othercase, occhars);
2655 else oclength = 0;
2656 #endif /* SUPPORT_UCP */
2657
2658 for (i = 1; i <= min; i++)
2659 {
2660 if (eptr <= md->end_subject - length &&
2661 memcmp(eptr, charptr, length) == 0) eptr += length;
2662 #ifdef SUPPORT_UCP
2663 else if (oclength > 0 &&
2664 eptr <= md->end_subject - oclength &&
2665 memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
2666 #endif /* SUPPORT_UCP */
2667 else
2668 {
2669 CHECK_PARTIAL();
2670 RRETURN(MATCH_NOMATCH);
2671 }
2672 }
2673
2674 if (min == max) continue;
2675
2676 if (minimize)
2677 {
2678 for (fi = min;; fi++)
2679 {
2680 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);
2681 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2682 if (fi >= max) RRETURN(MATCH_NOMATCH);
2683 if (eptr <= md->end_subject - length &&
2684 memcmp(eptr, charptr, length) == 0) eptr += length;
2685 #ifdef SUPPORT_UCP
2686 else if (oclength > 0 &&
2687 eptr <= md->end_subject - oclength &&
2688 memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
2689 #endif /* SUPPORT_UCP */
2690 else
2691 {
2692 CHECK_PARTIAL();
2693 RRETURN(MATCH_NOMATCH);
2694 }
2695 }
2696 /* Control never gets here */
2697 }
2698
2699 else /* Maximize */
2700 {
2701 pp = eptr;
2702 for (i = min; i < max; i++)
2703 {
2704 if (eptr <= md->end_subject - length &&
2705 memcmp(eptr, charptr, length) == 0) eptr += length;
2706 #ifdef SUPPORT_UCP
2707 else if (oclength > 0 &&
2708 eptr <= md->end_subject - oclength &&
2709 memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
2710 #endif /* SUPPORT_UCP */
2711 else
2712 {
2713 CHECK_PARTIAL();
2714 break;
2715 }
2716 }
2717
2718 if (possessive) continue;
2719
2720 for(;;)
2721 {
2722 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);
2723 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2724 if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
2725 #ifdef SUPPORT_UCP
2726 eptr--;
2727 BACKCHAR(eptr);
2728 #else /* without SUPPORT_UCP */
2729 eptr -= length;
2730 #endif /* SUPPORT_UCP */
2731 }
2732 }
2733 /* Control never gets here */
2734 }
2735
2736 /* If the length of a UTF-8 character is 1, we fall through here, and
2737 obey the code as for non-UTF-8 characters below, though in this case the
2738 value of fc will always be < 128. */
2739 }
2740 else
2741 #endif /* SUPPORT_UTF8 */
2742
2743 /* When not in UTF-8 mode, load a single-byte character. */
2744
2745 fc = *ecode++;
2746
2747 /* The value of fc at this point is always less than 256, though we may or
2748 may not be in UTF-8 mode. The code is duplicated for the caseless and
2749 caseful cases, for speed, since matching characters is likely to be quite
2750 common. First, ensure the minimum number of matches are present. If min =
2751 max, continue at the same level without recursing. Otherwise, if
2752 minimizing, keep trying the rest of the expression and advancing one
2753 matching character if failing, up to the maximum. Alternatively, if
2754 maximizing, find the maximum number of characters and work backwards. */
2755
2756 DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
2757 max, eptr));
2758
2759 if ((ims & PCRE_CASELESS) != 0)
2760 {
2761 fc = md->lcc[fc];
2762 for (i = 1; i <= min; i++)
2763 {
2764 if (eptr >= md->end_subject)
2765 {
2766 SCHECK_PARTIAL();
2767 RRETURN(MATCH_NOMATCH);
2768 }
2769 if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
2770 }
2771 if (min == max) continue;
2772 if (minimize)
2773 {
2774 for (fi = min;; fi++)
2775 {
2776 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);
2777 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2778 if (fi >= max) RRETURN(MATCH_NOMATCH);
2779 if (eptr >= md->end_subject)
2780 {
2781 SCHECK_PARTIAL();
2782 RRETURN(MATCH_NOMATCH);
2783 }
2784 if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
2785 }
2786 /* Control never gets here */
2787 }
2788 else /* Maximize */
2789 {
2790 pp = eptr;
2791 for (i = min; i < max; i++)
2792 {
2793 if (eptr >= md->end_subject)
2794 {
2795 SCHECK_PARTIAL();
2796 break;
2797 }
2798 if (fc != md->lcc[*eptr]) break;
2799 eptr++;
2800 }
2801
2802 if (possessive) continue;
2803
2804 while (eptr >= pp)
2805 {
2806 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);
2807 eptr--;
2808 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2809 }
2810 RRETURN(MATCH_NOMATCH);
2811 }
2812 /* Control never gets here */
2813 }
2814
2815 /* Caseful comparisons (includes all multi-byte characters) */
2816
2817 else
2818 {
2819 for (i = 1; i <= min; i++)
2820 {
2821 if (eptr >= md->end_subject)
2822 {
2823 SCHECK_PARTIAL();
2824 RRETURN(MATCH_NOMATCH);
2825 }
2826 if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
2827 }
2828
2829 if (min == max) continue;
2830
2831 if (minimize)
2832 {
2833 for (fi = min;; fi++)
2834 {
2835 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);
2836 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2837 if (fi >= max) RRETURN(MATCH_NOMATCH);
2838 if (eptr >= md->end_subject)
2839 {
2840 SCHECK_PARTIAL();
2841 RRETURN(MATCH_NOMATCH);
2842 }
2843 if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
2844 }
2845 /* Control never gets here */
2846 }
2847 else /* Maximize */
2848 {
2849 pp = eptr;
2850 for (i = min; i < max; i++)
2851 {
2852 if (eptr >= md->end_subject)
2853 {
2854 SCHECK_PARTIAL();
2855 break;
2856 }
2857 if (fc != *eptr) break;
2858 eptr++;
2859 }
2860 if (possessive) continue;
2861
2862 while (eptr >= pp)
2863 {
2864 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);
2865 eptr--;
2866 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2867 }
2868 RRETURN(MATCH_NOMATCH);
2869 }
2870 }
2871 /* Control never gets here */
2872
2873 /* Match a negated single one-byte character. The character we are
2874 checking can be multibyte. */
2875
2876 case OP_NOT:
2877 if (eptr >= md->end_subject)
2878 {
2879 SCHECK_PARTIAL();
2880 RRETURN(MATCH_NOMATCH);
2881 }
2882 ecode++;
2883 GETCHARINCTEST(c, eptr);
2884 if ((ims & PCRE_CASELESS) != 0)
2885 {
2886 #ifdef SUPPORT_UTF8
2887 if (c < 256)
2888 #endif
2889 c = md->lcc[c];
2890 if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);
2891 }
2892 else
2893 {
2894 if (*ecode++ == c) RRETURN(MATCH_NOMATCH);
2895 }
2896 break;
2897
2898 /* Match a negated single one-byte character repeatedly. This is almost a
2899 repeat of the code for a repeated single character, but I haven't found a
2900 nice way of commoning these up that doesn't require a test of the
2901 positive/negative option for each character match. Maybe that wouldn't add
2902 very much to the time taken, but character matching *is* what this is all
2903 about... */
2904
2905 case OP_NOTEXACT:
2906 min = max = GET2(ecode, 1);
2907 ecode += 3;
2908 goto REPEATNOTCHAR;
2909
2910 case OP_NOTUPTO:
2911 case OP_NOTMINUPTO:
2912 min = 0;
2913 max = GET2(ecode, 1);
2914 minimize = *ecode == OP_NOTMINUPTO;
2915 ecode += 3;
2916 goto REPEATNOTCHAR;
2917
2918 case OP_NOTPOSSTAR:
2919 possessive = TRUE;
2920 min = 0;
2921 max = INT_MAX;
2922 ecode++;
2923 goto REPEATNOTCHAR;
2924
2925 case OP_NOTPOSPLUS:
2926 possessive = TRUE;
2927 min = 1;
2928 max = INT_MAX;
2929 ecode++;
2930 goto REPEATNOTCHAR;
2931
2932 case OP_NOTPOSQUERY:
2933 possessive = TRUE;
2934 min = 0;
2935 max = 1;
2936 ecode++;
2937 goto REPEATNOTCHAR;
2938
2939 case OP_NOTPOSUPTO:
2940 possessive = TRUE;
2941 min = 0;
2942 max = GET2(ecode, 1);
2943 ecode += 3;
2944 goto REPEATNOTCHAR;
2945
2946 case OP_NOTSTAR:
2947 case OP_NOTMINSTAR:
2948 case OP_NOTPLUS:
2949 case OP_NOTMINPLUS:
2950 case OP_NOTQUERY:
2951 case OP_NOTMINQUERY:
2952 c = *ecode++ - OP_NOTSTAR;
2953 minimize = (c & 1) != 0;
2954 min = rep_min[c]; /* Pick up values from tables; */
2955 max = rep_max[c]; /* zero for max => infinity */
2956 if (max == 0) max = INT_MAX;
2957
2958 /* Common code for all repeated single-byte matches. */
2959
2960 REPEATNOTCHAR:
2961 fc = *ecode++;
2962
2963 /* The code is duplicated for the caseless and caseful cases, for speed,
2964 since matching characters is likely to be quite common. First, ensure the
2965 minimum number of matches are present. If min = max, continue at the same
2966 level without recursing. Otherwise, if minimizing, keep trying the rest of
2967 the expression and advancing one matching character if failing, up to the
2968 maximum. Alternatively, if maximizing, find the maximum number of
2969 characters and work backwards. */
2970
2971 DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
2972 max, eptr));
2973
2974 if ((ims & PCRE_CASELESS) != 0)
2975 {
2976 fc = md->lcc[fc];
2977
2978 #ifdef SUPPORT_UTF8
2979 /* UTF-8 mode */
2980 if (utf8)
2981 {
2982 register unsigned int d;
2983 for (i = 1; i <= min; i++)
2984 {
2985 if (eptr >= md->end_subject)
2986 {
2987 SCHECK_PARTIAL();
2988 RRETURN(MATCH_NOMATCH);
2989 }
2990 GETCHARINC(d, eptr);
2991 if (d < 256) d = md->lcc[d];
2992 if (fc == d) RRETURN(MATCH_NOMATCH);
2993 }
2994 }
2995 else
2996 #endif
2997
2998 /* Not UTF-8 mode */
2999 {
3000 for (i = 1; i <= min; i++)
3001 {
3002 if (eptr >= md->end_subject)
3003 {
3004 SCHECK_PARTIAL();
3005 RRETURN(MATCH_NOMATCH);
3006 }
3007 if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
3008 }
3009 }
3010
3011 if (min == max) continue;
3012
3013 if (minimize)
3014 {
3015 #ifdef SUPPORT_UTF8
3016 /* UTF-8 mode */
3017 if (utf8)
3018 {
3019 register unsigned int d;
3020 for (fi = min;; fi++)
3021 {
3022 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM28);
3023 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3024 if (fi >= max) RRETURN(MATCH_NOMATCH);
3025 if (eptr >= md->end_subject)
3026 {
3027 SCHECK_PARTIAL();
3028 RRETURN(MATCH_NOMATCH);
3029 }
3030 GETCHARINC(d, eptr);
3031 if (d < 256) d = md->lcc[d];
3032 if (fc == d) RRETURN(MATCH_NOMATCH);
3033 }
3034 }
3035 else
3036 #endif
3037 /* Not UTF-8 mode */
3038 {
3039 for (fi = min;; fi++)
3040 {
3041 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM29);
3042 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3043 if (fi >= max) RRETURN(MATCH_NOMATCH);
3044 if (eptr >= md->end_subject)
3045 {
3046 SCHECK_PARTIAL();
3047 RRETURN(MATCH_NOMATCH);
3048 }
3049 if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
3050 }
3051 }
3052 /* Control never gets here */
3053 }
3054
3055 /* Maximize case */
3056
3057 else
3058 {
3059 pp = eptr;
3060
3061 #ifdef SUPPORT_UTF8
3062 /* UTF-8 mode */
3063 if (utf8)
3064 {
3065 register unsigned int d;
3066 for (i = min; i < max; i++)
3067 {
3068 int len = 1;
3069 if (eptr >= md->end_subject)
3070 {
3071 SCHECK_PARTIAL();
3072 break;
3073 }
3074 GETCHARLEN(d, eptr, len);
3075 if (d < 256) d = md->lcc[d];
3076 if (fc == d) break;
3077 eptr += len;
3078 }
3079 if (possessive) continue;
3080 for(;;)
3081 {
3082 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM30);
3083 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3084 if (eptr-- == pp) break; /* Stop if tried at original pos */
3085 BACKCHAR(eptr);
3086 }
3087 }
3088 else
3089 #endif
3090 /* Not UTF-8 mode */
3091 {
3092 for (i = min; i < max; i++)
3093 {
3094 if (eptr >= md->end_subject)
3095 {
3096 SCHECK_PARTIAL();
3097 break;
3098 }
3099 if (fc == md->lcc[*eptr]) break;
3100 eptr++;
3101 }
3102 if (possessive) continue;
3103 while (eptr >= pp)
3104 {
3105 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM31);
3106 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3107 eptr--;
3108 }
3109 }
3110
3111 RRETURN(MATCH_NOMATCH);
3112 }
3113 /* Control never gets here */
3114 }
3115
3116 /* Caseful comparisons */
3117
3118 else
3119 {
3120 #ifdef SUPPORT_UTF8
3121 /* UTF-8 mode */
3122 if (utf8)
3123 {
3124 register unsigned int d;
3125 for (i = 1; i <= min; i++)
3126 {
3127 if (eptr >= md->end_subject)
3128 {
3129 SCHECK_PARTIAL();
3130 RRETURN(MATCH_NOMATCH);
3131 }
3132 GETCHARINC(d, eptr);
3133 if (fc == d) RRETURN(MATCH_NOMATCH);
3134 }
3135 }
3136 else
3137 #endif
3138 /* Not UTF-8 mode */
3139 {
3140 for (i = 1; i <= min; i++)
3141 {
3142 if (eptr >= md->end_subject)
3143 {
3144 SCHECK_PARTIAL();
3145 RRETURN(MATCH_NOMATCH);
3146 }
3147 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3148 }
3149 }
3150
3151 if (min == max) continue;
3152
3153 if (minimize)
3154 {
3155 #ifdef SUPPORT_UTF8
3156 /* UTF-8 mode */
3157 if (utf8)
3158 {
3159 register unsigned int d;
3160 for (fi = min;; fi++)
3161 {
3162 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM32);
3163 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3164 if (fi >= max) RRETURN(MATCH_NOMATCH);
3165 if (eptr >= md->end_subject)
3166 {
3167 SCHECK_PARTIAL();
3168 RRETURN(MATCH_NOMATCH);
3169 }
3170 GETCHARINC(d, eptr);
3171 if (fc == d) RRETURN(MATCH_NOMATCH);
3172 }
3173 }
3174 else
3175 #endif
3176 /* Not UTF-8 mode */
3177 {
3178 for (fi = min;; fi++)
3179 {
3180 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM33);
3181 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3182 if (fi >= max) RRETURN(MATCH_NOMATCH);
3183 if (eptr >= md->end_subject)
3184 {
3185 SCHECK_PARTIAL();
3186 RRETURN(MATCH_NOMATCH);
3187 }
3188 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3189 }
3190 }
3191 /* Control never gets here */
3192 }
3193
3194 /* Maximize case */
3195
3196 else
3197 {
3198 pp = eptr;
3199
3200 #ifdef SUPPORT_UTF8
3201 /* UTF-8 mode */
3202 if (utf8)
3203 {
3204 register unsigned int d;
3205 for (i = min; i < max; i++)
3206 {
3207 int len = 1;
3208 if (eptr >= md->end_subject)
3209 {
3210 SCHECK_PARTIAL();
3211 break;
3212 }
3213 GETCHARLEN(d, eptr, len);
3214 if (fc == d) break;
3215 eptr += len;
3216 }
3217 if (possessive) continue;
3218 for(;;)
3219 {
3220 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM34);
3221 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3222 if (eptr-- == pp) break; /* Stop if tried at original pos */
3223 BACKCHAR(eptr);
3224 }
3225 }
3226 else
3227 #endif
3228 /* Not UTF-8 mode */
3229 {
3230 for (i = min; i < max; i++)
3231 {
3232 if (eptr >= md->end_subject)
3233 {
3234 SCHECK_PARTIAL();
3235 break;
3236 }
3237 if (fc == *eptr) break;
3238 eptr++;
3239 }
3240 if (possessive) continue;
3241 while (eptr >= pp)
3242 {
3243 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM35);
3244 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3245 eptr--;
3246 }
3247 }
3248
3249 RRETURN(MATCH_NOMATCH);
3250 }
3251 }
3252 /* Control never gets here */
3253
3254 /* Match a single character type repeatedly; several different opcodes
3255 share code. This is very similar to the code for single characters, but we
3256 repeat it in the interests of efficiency. */
3257
3258 case OP_TYPEEXACT:
3259 min = max = GET2(ecode, 1);
3260 minimize = TRUE;
3261 ecode += 3;
3262 goto REPEATTYPE;
3263
3264 case OP_TYPEUPTO:
3265 case OP_TYPEMINUPTO:
3266 min = 0;
3267 max = GET2(ecode, 1);
3268 minimize = *ecode == OP_TYPEMINUPTO;
3269 ecode += 3;
3270 goto REPEATTYPE;
3271
3272 case OP_TYPEPOSSTAR:
3273 possessive = TRUE;
3274 min = 0;
3275 max = INT_MAX;
3276 ecode++;
3277 goto REPEATTYPE;
3278
3279 case OP_TYPEPOSPLUS:
3280 possessive = TRUE;
3281 min = 1;
3282 max = INT_MAX;
3283 ecode++;
3284 goto REPEATTYPE;
3285
3286 case OP_TYPEPOSQUERY:
3287 possessive = TRUE;
3288 min = 0;
3289 max = 1;
3290 ecode++;
3291 goto REPEATTYPE;
3292
3293 case OP_TYPEPOSUPTO:
3294 possessive = TRUE;
3295 min = 0;
3296 max = GET2(ecode, 1);
3297 ecode += 3;
3298 goto REPEATTYPE;
3299
3300 case OP_TYPESTAR:
3301 case OP_TYPEMINSTAR:
3302 case OP_TYPEPLUS:
3303 case OP_TYPEMINPLUS:
3304 case OP_TYPEQUERY:
3305 case OP_TYPEMINQUERY:
3306 c = *ecode++ - OP_TYPESTAR;
3307 minimize = (c & 1) != 0;
3308 min = rep_min[c]; /* Pick up values from tables; */
3309 max = rep_max[c]; /* zero for max => infinity */
3310 if (max == 0) max = INT_MAX;
3311
3312 /* Common code for all repeated single character type matches. Note that
3313 in UTF-8 mode, '.' matches a character of any length, but for the other
3314 character types, the valid characters are all one-byte long. */
3315
3316 REPEATTYPE:
3317 ctype = *ecode++; /* Code for the character type */
3318
3319 #ifdef SUPPORT_UCP
3320 if (ctype == OP_PROP || ctype == OP_NOTPROP)
3321 {
3322 prop_fail_result = ctype == OP_NOTPROP;
3323 prop_type = *ecode++;
3324 prop_value = *ecode++;
3325 }
3326 else prop_type = -1;
3327 #endif
3328
3329 /* First, ensure the minimum number of matches are present. Use inline
3330 code for maximizing the speed, and do the type test once at the start
3331 (i.e. keep it out of the loop). Separate the UTF-8 code completely as that
3332 is tidier. Also separate the UCP code, which can be the same for both UTF-8
3333 and single-bytes. */
3334
3335 if (min > 0)
3336 {
3337 #ifdef SUPPORT_UCP
3338 if (prop_type >= 0)
3339 {
3340 switch(prop_type)
3341 {
3342 case PT_ANY:
3343 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
3344 for (i = 1; i <= min; i++)
3345 {
3346 if (eptr >= md->end_subject)
3347 {
3348 SCHECK_PARTIAL();
3349 RRETURN(MATCH_NOMATCH);
3350 }
3351 GETCHARINCTEST(c, eptr);
3352 }
3353 break;
3354
3355 case PT_LAMP:
3356 for (i = 1; i <= min; i++)
3357 {
3358 if (eptr >= md->end_subject)
3359 {
3360 SCHECK_PARTIAL();
3361 RRETURN(MATCH_NOMATCH);
3362 }
3363 GETCHARINCTEST(c, eptr);
3364 prop_chartype = UCD_CHARTYPE(c);
3365 if ((prop_chartype == ucp_Lu ||
3366 prop_chartype == ucp_Ll ||
3367 prop_chartype == ucp_Lt) == prop_fail_result)
3368 RRETURN(MATCH_NOMATCH);
3369 }
3370 break;
3371
3372 case PT_GC:
3373 for (i = 1; i <= min; i++)
3374 {
3375 if (eptr >= md->end_subject)
3376 {
3377 SCHECK_PARTIAL();
3378 RRETURN(MATCH_NOMATCH);
3379 }
3380 GETCHARINCTEST(c, eptr);
3381 prop_category = UCD_CATEGORY(c);
3382 if ((prop_category == prop_value) == prop_fail_result)
3383 RRETURN(MATCH_NOMATCH);
3384 }
3385 break;
3386
3387 case PT_PC:
3388 for (i = 1; i <= min; i++)
3389 {
3390 if (eptr >= md->end_subject)
3391 {
3392 SCHECK_PARTIAL();
3393 RRETURN(MATCH_NOMATCH);
3394 }
3395 GETCHARINCTEST(c, eptr);
3396 prop_chartype = UCD_CHARTYPE(c);
3397 if ((prop_chartype == prop_value) == prop_fail_result)
3398 RRETURN(MATCH_NOMATCH);
3399 }
3400 break;
3401
3402 case PT_SC:
3403 for (i = 1; i <= min; i++)
3404 {
3405 if (eptr >= md->end_subject)
3406 {
3407 SCHECK_PARTIAL();
3408 RRETURN(MATCH_NOMATCH);
3409 }
3410 GETCHARINCTEST(c, eptr);
3411 prop_script = UCD_SCRIPT(c);
3412 if ((prop_script == prop_value) == prop_fail_result)
3413 RRETURN(MATCH_NOMATCH);
3414 }
3415 break;
3416
3417 default:
3418 RRETURN(PCRE_ERROR_INTERNAL);
3419 }
3420 }
3421
3422 /* Match extended Unicode sequences. We will get here only if the
3423 support is in the binary; otherwise a compile-time error occurs. */
3424
3425 else if (ctype == OP_EXTUNI)
3426 {
3427 for (i = 1; i <= min; i++)
3428 {
3429 if (eptr >= md->end_subject)
3430 {
3431 SCHECK_PARTIAL();
3432 RRETURN(MATCH_NOMATCH);
3433 }
3434 GETCHARINCTEST(c, eptr);
3435 prop_category = UCD_CATEGORY(c);
3436 if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH);
3437 while (eptr < md->end_subject)
3438 {
3439 int len = 1;
3440 if (!utf8) c = *eptr;
3441 else { GETCHARLEN(c, eptr, len); }
3442 prop_category = UCD_CATEGORY(c);
3443 if (prop_category != ucp_M) break;
3444 eptr += len;
3445 }
3446 }
3447 }
3448
3449 else
3450 #endif /* SUPPORT_UCP */
3451
3452 /* Handle all other cases when the coding is UTF-8 */
3453
3454 #ifdef SUPPORT_UTF8
3455 if (utf8) switch(ctype)
3456 {
3457 case OP_ANY:
3458 for (i = 1; i <= min; i++)
3459 {
3460 if (eptr >= md->end_subject)
3461 {
3462 SCHECK_PARTIAL();
3463 RRETURN(MATCH_NOMATCH);
3464 }
3465 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
3466 eptr++;
3467 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
3468 }
3469 break;
3470
3471 case OP_ALLANY:
3472 for (i = 1; i <= min; i++)
3473 {
3474 if (eptr >= md->end_subject)
3475 {
3476 SCHECK_PARTIAL();
3477 RRETURN(MATCH_NOMATCH);
3478 }
3479 eptr++;
3480 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
3481 }
3482 break;
3483
3484 case OP_ANYBYTE:
3485 if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH);
3486 eptr += min;
3487 break;
3488
3489 case OP_ANYNL:
3490 for (i = 1; i <= min; i++)
3491 {
3492 if (eptr >= md->end_subject)
3493 {
3494 SCHECK_PARTIAL();
3495 RRETURN(MATCH_NOMATCH);
3496 }
3497 GETCHARINC(c, eptr);
3498 switch(c)
3499 {
3500 default: RRETURN(MATCH_NOMATCH);
3501 case 0x000d:
3502 if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3503 break;
3504
3505 case 0x000a:
3506 break;
3507
3508 case 0x000b:
3509 case 0x000c:
3510 case 0x0085:
3511 case 0x2028:
3512 case 0x2029:
3513 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
3514 break;
3515 }
3516 }
3517 break;
3518
3519 case OP_NOT_HSPACE:
3520 for (i = 1; i <= min; i++)
3521 {
3522 if (eptr >= md->end_subject)
3523 {
3524 SCHECK_PARTIAL();
3525 RRETURN(MATCH_NOMATCH);
3526 }
3527 GETCHARINC(c, eptr);
3528 switch(c)
3529 {
3530 default: break;
3531 case 0x09: /* HT */
3532 case 0x20: /* SPACE */
3533 case 0xa0: /* NBSP */
3534 case 0x1680: /* OGHAM SPACE MARK */
3535 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
3536 case 0x2000: /* EN QUAD */
3537 case 0x2001: /* EM QUAD */
3538 case 0x2002: /* EN SPACE */
3539 case 0x2003: /* EM SPACE */
3540 case 0x2004: /* THREE-PER-EM SPACE */
3541 case 0x2005: /* FOUR-PER-EM SPACE */
3542 case 0x2006: /* SIX-PER-EM SPACE */
3543 case 0x2007: /* FIGURE SPACE */
3544 case 0x2008: /* PUNCTUATION SPACE */
3545 case 0x2009: /* THIN SPACE */
3546 case 0x200A: /* HAIR SPACE */
3547 case 0x202f: /* NARROW NO-BREAK SPACE */
3548 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
3549 case 0x3000: /* IDEOGRAPHIC SPACE */
3550 RRETURN(MATCH_NOMATCH);
3551 }
3552 }
3553 break;
3554
3555 case OP_HSPACE:
3556 for (i = 1; i <= min; i++)
3557 {
3558 if (eptr >= md->end_subject)
3559 {
3560 SCHECK_PARTIAL();
3561 RRETURN(MATCH_NOMATCH);
3562 }
3563 GETCHARINC(c, eptr);
3564 switch(c)
3565 {
3566 default: RRETURN(MATCH_NOMATCH);
3567 case 0x09: /* HT */
3568 case 0x20: /* SPACE */
3569 case 0xa0: /* NBSP */
3570 case 0x1680: /* OGHAM SPACE MARK */
3571 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
3572 case 0x2000: /* EN QUAD */
3573 case 0x2001: /* EM QUAD */
3574 case 0x2002: /* EN SPACE */
3575 case 0x2003: /* EM SPACE */
3576 case 0x2004: /* THREE-PER-EM SPACE */
3577 case 0x2005: /* FOUR-PER-EM SPACE */
3578 case 0x2006: /* SIX-PER-EM SPACE */
3579 case 0x2007: /* FIGURE SPACE */
3580 case 0x2008: /* PUNCTUATION SPACE */
3581 case 0x2009: /* THIN SPACE */
3582 case 0x200A: /* HAIR SPACE */
3583 case 0x202f: /* NARROW NO-BREAK SPACE */
3584 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
3585 case 0x3000: /* IDEOGRAPHIC SPACE */
3586 break;
3587 }
3588 }
3589 break;
3590
3591 case OP_NOT_VSPACE:
3592 for (i = 1; i <= min; i++)
3593 {
3594 if (eptr >= md->end_subject)
3595 {
3596 SCHECK_PARTIAL();
3597 RRETURN(MATCH_NOMATCH);
3598 }
3599 GETCHARINC(c, eptr);
3600 switch(c)
3601 {
3602 default: break;
3603 case 0x0a: /* LF */
3604 case 0x0b: /* VT */
3605 case 0x0c: /* FF */
3606 case 0x0d: /* CR */
3607 case 0x85: /* NEL */
3608 case 0x2028: /* LINE SEPARATOR */
3609 case 0x2029: /* PARAGRAPH SEPARATOR */
3610 RRETURN(MATCH_NOMATCH);
3611 }
3612 }
3613 break;
3614
3615 case OP_VSPACE:
3616 for (i = 1; i <= min; i++)
3617 {
3618 if (eptr >= md->end_subject)
3619 {
3620 SCHECK_PARTIAL();
3621 RRETURN(MATCH_NOMATCH);
3622 }
3623 GETCHARINC(c, eptr);
3624 switch(c)
3625 {
3626 default: RRETURN(MATCH_NOMATCH);
3627 case 0x0a: /* LF */
3628 case 0x0b: /* VT */
3629 case 0x0c: /* FF */
3630 case 0x0d: /* CR */
3631 case 0x85: /* NEL */
3632 case 0x2028: /* LINE SEPARATOR */
3633 case 0x2029: /* PARAGRAPH SEPARATOR */
3634 break;
3635 }
3636 }
3637 break;
3638
3639 case OP_NOT_DIGIT:
3640 for (i = 1; i <= min; i++)
3641 {
3642 if (eptr >= md->end_subject)
3643 {
3644 SCHECK_PARTIAL();
3645 RRETURN(MATCH_NOMATCH);
3646 }
3647 GETCHARINC(c, eptr);
3648 if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
3649 RRETURN(MATCH_NOMATCH);
3650 }
3651 break;
3652
3653 case OP_DIGIT:
3654 for (i = 1; i <= min; i++)
3655 {
3656 if (eptr >= md->end_subject)
3657 {
3658 SCHECK_PARTIAL();
3659 RRETURN(MATCH_NOMATCH);
3660 }
3661 if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_digit) == 0)
3662 RRETURN(MATCH_NOMATCH);
3663 /* No need to skip more bytes - we know it's a 1-byte character */
3664 }
3665 break;
3666
3667 case OP_NOT_WHITESPACE:
3668 for (i = 1; i <= min; i++)
3669 {
3670 if (eptr >= md->end_subject)
3671 {
3672 SCHECK_PARTIAL();
3673 RRETURN(MATCH_NOMATCH);
3674 }
3675 if (*eptr < 128 && (md->ctypes[*eptr] & ctype_space) != 0)
3676 RRETURN(MATCH_NOMATCH);
3677 while (++eptr < md->end_subject && (*eptr & 0xc0) == 0x80);
3678 }
3679 break;
3680
3681 case OP_WHITESPACE:
3682 for (i = 1; i <= min; i++)
3683 {
3684 if (eptr >= md->end_subject)
3685 {
3686 SCHECK_PARTIAL();
3687 RRETURN(MATCH_NOMATCH);
3688 }
3689 if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_space) == 0)
3690 RRETURN(MATCH_NOMATCH);
3691 /* No need to skip more bytes - we know it's a 1-byte character */
3692 }
3693 break;
3694
3695 case OP_NOT_WORDCHAR:
3696 for (i = 1; i <= min; i++)
3697 {
3698 if (eptr >= md->end_subject ||
3699 (*eptr < 128 && (md->ctypes[*eptr] & ctype_word) != 0))
3700 RRETURN(MATCH_NOMATCH);
3701 while (++eptr < md->end_subject && (*eptr & 0xc0) == 0x80);
3702 }
3703 break;
3704
3705 case OP_WORDCHAR:
3706 for (i = 1; i <= min; i++)
3707 {
3708 if (eptr >= md->end_subject)
3709 {
3710 SCHECK_PARTIAL();
3711 RRETURN(MATCH_NOMATCH);
3712 }
3713 if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_word) == 0)
3714 RRETURN(MATCH_NOMATCH);
3715 /* No need to skip more bytes - we know it's a 1-byte character */
3716 }
3717 break;
3718
3719 default:
3720 RRETURN(PCRE_ERROR_INTERNAL);
3721 } /* End switch(ctype) */
3722
3723 else
3724 #endif /* SUPPORT_UTF8 */
3725
3726 /* Code for the non-UTF-8 case for minimum matching of operators other
3727 than OP_PROP and OP_NOTPROP. */
3728
3729 switch(ctype)
3730 {
3731 case OP_ANY:
3732 for (i = 1; i <= min; i++)
3733 {
3734 if (eptr >= md->end_subject)
3735 {
3736 SCHECK_PARTIAL();
3737 RRETURN(MATCH_NOMATCH);
3738 }
3739 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
3740 eptr++;
3741 }
3742 break;
3743
3744 case OP_ALLANY:
3745 if (eptr > md->end_subject - min)
3746 {
3747 SCHECK_PARTIAL();
3748 RRETURN(MATCH_NOMATCH);
3749 }
3750 eptr += min;
3751 break;
3752
3753 case OP_ANYBYTE:
3754 if (eptr > md->end_subject - min)
3755 {
3756 SCHECK_PARTIAL();
3757 RRETURN(MATCH_NOMATCH);
3758 }
3759 eptr += min;
3760 break;
3761
3762 case OP_ANYNL:
3763 for (i = 1; i <= min; i++)
3764 {
3765 if (eptr >= md->end_subject)
3766 {
3767 SCHECK_PARTIAL();
3768 RRETURN(MATCH_NOMATCH);
3769 }
3770 switch(*eptr++)
3771 {
3772 default: RRETURN(MATCH_NOMATCH);
3773 case 0x000d:
3774 if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3775 break;
3776 case 0x000a:
3777 break;
3778
3779 case 0x000b:
3780 case 0x000c:
3781 case 0x0085:
3782 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
3783 break;
3784 }
3785 }
3786 break;
3787
3788 case OP_NOT_HSPACE:
3789 for (i = 1; i <= min; i++)
3790 {
3791 if (eptr >= md->end_subject)
3792 {
3793 SCHECK_PARTIAL();
3794 RRETURN(MATCH_NOMATCH);
3795 }
3796 switch(*eptr++)
3797 {
3798 default: break;
3799 case 0x09: /* HT */
3800 case 0x20: /* SPACE */
3801 case 0xa0: /* NBSP */
3802 RRETURN(MATCH_NOMATCH);
3803 }
3804 }
3805 break;
3806
3807 case OP_HSPACE:
3808 for (i = 1; i <= min; i++)
3809 {
3810 if (eptr >= md->end_subject)
3811 {
3812 SCHECK_PARTIAL();
3813 RRETURN(MATCH_NOMATCH);
3814 }
3815 switch(*eptr++)
3816 {
3817 default: RRETURN(MATCH_NOMATCH);
3818 case 0x09: /* HT */
3819 case 0x20: /* SPACE */
3820 case 0xa0: /* NBSP */
3821 break;
3822 }
3823 }
3824 break;
3825
3826 case OP_NOT_VSPACE:
3827 for (i = 1; i <= min; i++)
3828 {
3829 if (eptr >= md->end_subject)
3830 {
3831 SCHECK_PARTIAL();
3832 RRETURN(MATCH_NOMATCH);
3833 }
3834 switch(*eptr++)
3835 {
3836 default: break;
3837 case 0x0a: /* LF */
3838 case 0x0b: /* VT */
3839 case 0x0c: /* FF */
3840 case 0x0d: /* CR */
3841 case 0x85: /* NEL */
3842 RRETURN(MATCH_NOMATCH);
3843 }
3844 }
3845 break;
3846
3847 case OP_VSPACE:
3848 for (i = 1; i <= min; i++)
3849 {
3850 if (eptr >= md->end_subject)
3851 {
3852 SCHECK_PARTIAL();
3853 RRETURN(MATCH_NOMATCH);
3854 }
3855 switch(*eptr++)
3856 {
3857 default: RRETURN(MATCH_NOMATCH);
3858 case 0x0a: /* LF */
3859 case 0x0b: /* VT */
3860 case 0x0c: /* FF */
3861 case 0x0d: /* CR */
3862 case 0x85: /* NEL */
3863 break;
3864 }
3865 }
3866 break;
3867
3868 case OP_NOT_DIGIT:
3869 for (i = 1; i <= min; i++)
3870 {
3871 if (eptr >= md->end_subject)
3872 {
3873 SCHECK_PARTIAL();
3874 RRETURN(MATCH_NOMATCH);
3875 }
3876 if ((md->ctypes[*eptr++] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
3877 }
3878 break;
3879
3880 case OP_DIGIT:
3881 for (i = 1; i <= min; i++)
3882 {
3883 if (eptr >= md->end_subject)
3884 {
3885 SCHECK_PARTIAL();
3886 RRETURN(MATCH_NOMATCH);
3887 }
3888 if ((md->ctypes[*eptr++] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
3889 }
3890 break;
3891
3892 case OP_NOT_WHITESPACE:
3893 for (i = 1; i <= min; i++)
3894 {
3895 if (eptr >= md->end_subject)
3896 {
3897 SCHECK_PARTIAL();
3898 RRETURN(MATCH_NOMATCH);
3899 }
3900 if ((md->ctypes[*eptr++] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
3901 }
3902 break;
3903
3904 case OP_WHITESPACE:
3905 for (i = 1; i <= min; i++)
3906 {
3907 if (eptr >= md->end_subject)
3908 {
3909 SCHECK_PARTIAL();
3910 RRETURN(MATCH_NOMATCH);
3911 }
3912 if ((md->ctypes[*eptr++] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
3913 }
3914 break;
3915
3916 case OP_NOT_WORDCHAR:
3917 for (i = 1; i <= min; i++)
3918 {
3919 if (eptr >= md->end_subject)
3920 {
3921 SCHECK_PARTIAL();
3922 RRETURN(MATCH_NOMATCH);
3923 }
3924 if ((md->ctypes[*eptr++] & ctype_word) != 0)
3925 RRETURN(MATCH_NOMATCH);
3926 }
3927 break;
3928
3929 case OP_WORDCHAR:
3930 for (i = 1; i <= min; i++)
3931 {
3932 if (eptr >= md->end_subject)
3933 {
3934 SCHECK_PARTIAL();
3935 RRETURN(MATCH_NOMATCH);
3936 }
3937 if ((md->ctypes[*eptr++] & ctype_word) == 0)
3938 RRETURN(MATCH_NOMATCH);
3939 }
3940 break;
3941
3942 default:
3943 RRETURN(PCRE_ERROR_INTERNAL);
3944 }
3945 }
3946
3947 /* If min = max, continue at the same level without recursing */
3948
3949 if (min == max) continue;
3950
3951 /* If minimizing, we have to test the rest of the pattern before each
3952 subsequent match. Again, separate the UTF-8 case for speed, and also
3953 separate the UCP cases. */
3954
3955 if (minimize)
3956 {
3957 #ifdef SUPPORT_UCP
3958 if (prop_type >= 0)
3959 {
3960 switch(prop_type)
3961 {
3962 case PT_ANY:
3963 for (fi = min;; fi++)
3964 {
3965 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM36);
3966 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3967 if (fi >= max) RRETURN(MATCH_NOMATCH);
3968 if (eptr >= md->end_subject)
3969 {
3970 SCHECK_PARTIAL();
3971 RRETURN(MATCH_NOMATCH);
3972 }
3973 GETCHARINC(c, eptr);
3974 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
3975 }
3976 /* Control never gets here */
3977
3978 case PT_LAMP:
3979 for (fi = min;; fi++)
3980 {
3981 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM37);
3982 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3983 if (fi >= max) RRETURN(MATCH_NOMATCH);
3984 if (eptr >= md->end_subject)
3985 {
3986 SCHECK_PARTIAL();
3987 RRETURN(MATCH_NOMATCH);
3988 }
3989 GETCHARINC(c, eptr);
3990 prop_chartype = UCD_CHARTYPE(c);
3991 if ((prop_chartype == ucp_Lu ||
3992 prop_chartype == ucp_Ll ||
3993 prop_chartype == ucp_Lt) == prop_fail_result)
3994 RRETURN(MATCH_NOMATCH);
3995 }
3996 /* Control never gets here */
3997
3998 case PT_GC:
3999 for (fi = min;; fi++)
4000 {
4001 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM38);
4002 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4003 if (fi >= max) RRETURN(MATCH_NOMATCH);
4004 if (eptr >= md->end_subject)
4005 {
4006 SCHECK_PARTIAL();
4007 RRETURN(MATCH_NOMATCH);
4008 }
4009 GETCHARINC(c, eptr);
4010 prop_category = UCD_CATEGORY(c);
4011 if ((prop_category == prop_value) == prop_fail_result)
4012 RRETURN(MATCH_NOMATCH);
4013 }
4014 /* Control never gets here */
4015
4016 case PT_PC:
4017 for (fi = min;; fi++)
4018 {
4019 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM39);
4020 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4021 if (fi >= max) RRETURN(MATCH_NOMATCH);
4022 if (eptr >= md->end_subject)
4023 {
4024 SCHECK_PARTIAL();
4025 RRETURN(MATCH_NOMATCH);
4026 }
4027 GETCHARINC(c, eptr);
4028 prop_chartype = UCD_CHARTYPE(c);
4029 if ((prop_chartype == prop_value) == prop_fail_result)
4030 RRETURN(MATCH_NOMATCH);
4031 }
4032 /* Control never gets here */
4033
4034 case PT_SC:
4035 for (fi = min;; fi++)
4036 {
4037 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM40);
4038 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4039 if (fi >= max) RRETURN(MATCH_NOMATCH);
4040 if (eptr >= md->end_subject)
4041 {
4042 SCHECK_PARTIAL();
4043 RRETURN(MATCH_NOMATCH);
4044 }
4045 GETCHARINC(c, eptr);
4046 prop_script = UCD_SCRIPT(c);
4047 if ((prop_script == prop_value) == prop_fail_result)
4048 RRETURN(MATCH_NOMATCH);
4049 }
4050 /* Control never gets here */
4051
4052 default:
4053 RRETURN(PCRE_ERROR_INTERNAL);
4054 }
4055 }
4056
4057 /* Match extended Unicode sequences. We will get here only if the
4058 support is in the binary; otherwise a compile-time error occurs. */
4059
4060 else if (ctype == OP_EXTUNI)
4061 {
4062 for (fi = min;; fi++)
4063 {
4064 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM41);
4065 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4066 if (fi >= max) RRETURN(MATCH_NOMATCH);
4067 if (eptr >= md->end_subject)
4068 {
4069 SCHECK_PARTIAL();
4070 RRETURN(MATCH_NOMATCH);
4071 }
4072 GETCHARINCTEST(c, eptr);
4073 prop_category = UCD_CATEGORY(c);
4074 if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH);
4075 while (eptr < md->end_subject)
4076 {
4077 int len = 1;
4078 if (!utf8) c = *eptr;
4079 else { GETCHARLEN(c, eptr, len); }
4080 prop_category = UCD_CATEGORY(c);
4081 if (prop_category != ucp_M) break;
4082 eptr += len;
4083 }
4084 }
4085 }
4086
4087 else
4088 #endif /* SUPPORT_UCP */
4089
4090 #ifdef SUPPORT_UTF8
4091 /* UTF-8 mode */
4092 if (utf8)
4093 {
4094 for (fi = min;; fi++)
4095 {
4096 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM42);
4097 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4098 if (fi >= max) RRETURN(MATCH_NOMATCH);
4099 if (eptr >= md->end_subject)
4100 {
4101 SCHECK_PARTIAL();
4102 RRETURN(MATCH_NOMATCH);
4103 }
4104 if (ctype == OP_ANY && IS_NEWLINE(eptr))
4105 RRETURN(MATCH_NOMATCH);
4106 GETCHARINC(c, eptr);
4107 switch(ctype)
4108 {
4109 case OP_ANY: /* This is the non-NL case */
4110 case OP_ALLANY:
4111 case OP_ANYBYTE:
4112 break;
4113
4114 case OP_ANYNL:
4115 switch(c)
4116 {
4117 default: RRETURN(MATCH_NOMATCH);
4118 case 0x000d:
4119 if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4120 break;
4121 case 0x000a:
4122 break;
4123
4124 case 0x000b:
4125 case 0x000c:
4126 case 0x0085:
4127 case 0x2028:
4128 case 0x2029:
4129 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4130 break;
4131 }
4132 break;
4133
4134 case OP_NOT_HSPACE:
4135 switch(c)
4136 {
4137 default: break;
4138 case 0x09: /* HT */
4139 case 0x20: /* SPACE */
4140 case 0xa0: /* NBSP */
4141 case 0x1680: /* OGHAM SPACE MARK */
4142 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
4143 case 0x2000: /* EN QUAD */
4144 case 0x2001: /* EM QUAD */
4145 case 0x2002: /* EN SPACE */
4146 case 0x2003: /* EM SPACE */
4147 case 0x2004: /* THREE-PER-EM SPACE */
4148 case 0x2005: /* FOUR-PER-EM SPACE */
4149 case 0x2006: /* SIX-PER-EM SPACE */
4150 case 0x2007: /* FIGURE SPACE */
4151 case 0x2008: /* PUNCTUATION SPACE */
4152 case 0x2009: /* THIN SPACE */
4153 case 0x200A: /* HAIR SPACE */
4154 case 0x202f: /* NARROW NO-BREAK SPACE */
4155 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
4156 case 0x3000: /* IDEOGRAPHIC SPACE */
4157 RRETURN(MATCH_NOMATCH);
4158 }
4159 break;
4160
4161 case OP_HSPACE:
4162 switch(c)
4163 {
4164 default: RRETURN(MATCH_NOMATCH);
4165 case 0x09: /* HT */
4166 case 0x20: /* SPACE */
4167 case 0xa0: /* NBSP */
4168 case 0x1680: /* OGHAM SPACE MARK */
4169 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
4170 case 0x2000: /* EN QUAD */
4171 case 0x2001: /* EM QUAD */
4172 case 0x2002: /* EN SPACE */
4173 case 0x2003: /* EM SPACE */
4174 case 0x2004: /* THREE-PER-EM SPACE */
4175 case 0x2005: /* FOUR-PER-EM SPACE */
4176 case 0x2006: /* SIX-PER-EM SPACE */
4177 case 0x2007: /* FIGURE SPACE */
4178 case 0x2008: /* PUNCTUATION SPACE */
4179 case 0x2009: /* THIN SPACE */
4180 case 0x200A: /* HAIR SPACE */
4181 case 0x202f: /* NARROW NO-BREAK SPACE */
4182 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
4183 case 0x3000: /* IDEOGRAPHIC SPACE */
4184 break;
4185 }
4186 break;
4187
4188 case OP_NOT_VSPACE:
4189 switch(c)
4190 {
4191 default: break;
4192 case 0x0a: /* LF */
4193 case 0x0b: /* VT */
4194 case 0x0c: /* FF */
4195 case 0x0d: /* CR */
4196 case 0x85: /* NEL */
4197 case 0x2028: /* LINE SEPARATOR */
4198 case 0x2029: /* PARAGRAPH SEPARATOR */
4199 RRETURN(MATCH_NOMATCH);
4200 }
4201 break;
4202
4203 case OP_VSPACE:
4204 switch(c)
4205 {
4206 default: RRETURN(MATCH_NOMATCH);
4207 case 0x0a: /* LF */
4208 case 0x0b: /* VT */
4209 case 0x0c: /* FF */
4210 case 0x0d: /* CR */
4211 case 0x85: /* NEL */
4212 case 0x2028: /* LINE SEPARATOR */
4213 case 0x2029: /* PARAGRAPH SEPARATOR */
4214 break;
4215 }
4216 break;
4217
4218 case OP_NOT_DIGIT:
4219 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
4220 RRETURN(MATCH_NOMATCH);
4221 break;
4222
4223 case OP_DIGIT:
4224 if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0)
4225 RRETURN(MATCH_NOMATCH);
4226 break;
4227
4228 case OP_NOT_WHITESPACE:
4229 if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
4230 RRETURN(MATCH_NOMATCH);
4231 break;
4232
4233 case OP_WHITESPACE:
4234 if (c >= 256 || (md->ctypes[c] & ctype_space) == 0)
4235 RRETURN(MATCH_NOMATCH);
4236 break;
4237
4238 case OP_NOT_WORDCHAR:
4239 if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
4240 RRETURN(MATCH_NOMATCH);
4241 break;
4242
4243 case OP_WORDCHAR:
4244 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0)
4245 RRETURN(MATCH_NOMATCH);
4246 break;
4247
4248 default:
4249 RRETURN(PCRE_ERROR_INTERNAL);
4250 }
4251 }
4252 }
4253 else
4254 #endif
4255 /* Not UTF-8 mode */
4256 {
4257 for (fi = min;; fi++)
4258 {
4259 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM43);
4260 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4261 if (fi >= max) RRETURN(MATCH_NOMATCH);
4262 if (eptr >= md->end_subject)
4263 {
4264 SCHECK_PARTIAL();
4265 RRETURN(MATCH_NOMATCH);
4266 }
4267 if (ctype == OP_ANY && IS_NEWLINE(eptr))
4268 RRETURN(MATCH_NOMATCH);
4269 c = *eptr++;
4270 switch(ctype)
4271 {
4272 case OP_ANY: /* This is the non-NL case */
4273 case OP_ALLANY:
4274 case OP_ANYBYTE:
4275 break;
4276
4277 case OP_ANYNL:
4278 switch(c)
4279 {
4280 default: RRETURN(MATCH_NOMATCH);
4281 case 0x000d:
4282 if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4283 break;
4284
4285 case 0x000a:
4286 break;
4287
4288 case 0x000b:
4289 case 0x000c:
4290 case 0x0085:
4291 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4292 break;
4293 }
4294 break;
4295
4296 case OP_NOT_HSPACE:
4297 switch(c)
4298 {
4299 default: break;
4300 case 0x09: /* HT */
4301 case 0x20: /* SPACE */
4302 case 0xa0: /* NBSP */
4303 RRETURN(MATCH_NOMATCH);
4304 }
4305 break;
4306
4307 case OP_HSPACE:
4308 switch(c)
4309 {
4310 default: RRETURN(MATCH_NOMATCH);
4311 case 0x09: /* HT */
4312 case 0x20: /* SPACE */
4313 case 0xa0: /* NBSP */
4314 break;
4315 }
4316 break;
4317
4318 case OP_NOT_VSPACE:
4319 switch(c)
4320 {
4321 default: break;
4322 case 0x0a: /* LF */
4323 case 0x0b: /* VT */
4324 case 0x0c: /* FF */
4325 case 0x0d: /* CR */
4326 case 0x85: /* NEL */
4327 RRETURN(MATCH_NOMATCH);
4328 }
4329 break;
4330
4331 case OP_VSPACE:
4332 switch(c)
4333 {
4334 default: RRETURN(MATCH_NOMATCH);
4335 case 0x0a: /* LF */
4336 case 0x0b: /* VT */
4337 case 0x0c: /* FF */
4338 case 0x0d: /* CR */
4339 case 0x85: /* NEL */
4340 break;
4341 }
4342 break;
4343
4344 case OP_NOT_DIGIT:
4345 if ((md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
4346 break;
4347
4348 case OP_DIGIT:
4349 if ((md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
4350 break;
4351
4352 case OP_NOT_WHITESPACE:
4353 if ((md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
4354 break;
4355
4356 case OP_WHITESPACE:
4357 if ((md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
4358 break;
4359
4360 case OP_NOT_WORDCHAR:
4361 if ((md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH);
4362 break;
4363
4364 case OP_WORDCHAR:
4365 if ((md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH);
4366 break;
4367
4368 default:
4369 RRETURN(PCRE_ERROR_INTERNAL);
4370 }
4371 }
4372 }
4373 /* Control never gets here */
4374 }
4375
4376 /* If maximizing, it is worth using inline code for speed, doing the type
4377 test once at the start (i.e. keep it out of the loop). Again, keep the
4378 UTF-8 and UCP stuff separate. */
4379
4380 else
4381 {
4382 pp = eptr; /* Remember where we started */
4383
4384 #ifdef SUPPORT_UCP
4385 if (prop_type >= 0)
4386 {
4387 switch(prop_type)
4388 {
4389 case PT_ANY:
4390 for (i = min; i < max; i++)
4391 {
4392 int len = 1;
4393 if (eptr >= md->end_subject)
4394 {
4395 SCHECK_PARTIAL();
4396 break;
4397 }
4398 GETCHARLEN(c, eptr, len);
4399 if (prop_fail_result) break;
4400 eptr+= len;
4401 }
4402 break;
4403
4404 case PT_LAMP:
4405 for (i = min; i < max; i++)
4406 {
4407 int len = 1;
4408 if (eptr >= md->end_subject)
4409 {
4410 SCHECK_PARTIAL();
4411 break;
4412 }
4413 GETCHARLEN(c, eptr, len);
4414 prop_chartype = UCD_CHARTYPE(c);
4415 if ((prop_chartype == ucp_Lu ||
4416 prop_chartype == ucp_Ll ||
4417 prop_chartype == ucp_Lt) == prop_fail_result)
4418 break;
4419 eptr+= len;
4420 }
4421 break;
4422
4423 case PT_GC:
4424 for (i = min; i < max; i++)
4425 {
4426 int len = 1;
4427 if (eptr >= md->end_subject)
4428 {
4429 SCHECK_PARTIAL();
4430 break;
4431 }
4432 GETCHARLEN(c, eptr, len);
4433 prop_category = UCD_CATEGORY(c);
4434 if ((prop_category == prop_value) == prop_fail_result)
4435 break;
4436 eptr+= len;
4437 }
4438 break;
4439
4440 case PT_PC:
4441 for (i = min; i < max; i++)
4442 {
4443 int len = 1;
4444 if (eptr >= md->end_subject)
4445 {
4446 SCHECK_PARTIAL();
4447 break;
4448 }
4449 GETCHARLEN(c, eptr, len);
4450 prop_chartype = UCD_CHARTYPE(c);
4451 if ((prop_chartype == prop_value) == prop_fail_result)
4452 break;
4453 eptr+= len;
4454 }
4455 break;
4456
4457 case PT_SC:
4458 for (i = min; i < max; i++)
4459 {
4460 int len = 1;
4461 if (eptr >= md->end_subject)
4462 {
4463 SCHECK_PARTIAL();
4464 break;
4465 }
4466 GETCHARLEN(c, eptr, len);
4467 prop_script = UCD_SCRIPT(c);
4468 if ((prop_script == prop_value) == prop_fail_result)
4469 break;
4470 eptr+= len;
4471 }
4472 break;
4473 }
4474
4475 /* eptr is now past the end of the maximum run */
4476
4477 if (possessive) continue;
4478 for(;;)
4479 {
4480 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM44);
4481 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4482 if (eptr-- == pp) break; /* Stop if tried at original pos */
4483 if (utf8) BACKCHAR(eptr);
4484 }
4485 }
4486
4487 /* Match extended Unicode sequences. We will get here only if the
4488 support is in the binary; otherwise a compile-time error occurs. */
4489
4490 else if (ctype == OP_EXTUNI)
4491 {
4492 for (i = min; i < max; i++)
4493 {
4494 if (eptr >= md->end_subject)
4495 {
4496 SCHECK_PARTIAL();
4497 break;
4498 }
4499 GETCHARINCTEST(c, eptr);
4500 prop_category = UCD_CATEGORY(c);
4501 if (prop_category == ucp_M) break;
4502 while (eptr < md->end_subject)
4503 {
4504 int len = 1;
4505 if (!utf8) c = *eptr; else
4506 {
4507 GETCHARLEN(c, eptr, len);
4508 }
4509 prop_category = UCD_CATEGORY(c);
4510 if (prop_category != ucp_M) break;
4511 eptr += len;
4512 }
4513 }
4514
4515 /* eptr is now past the end of the maximum run */
4516
4517 if (possessive) continue;
4518
4519 for(;;)
4520 {
4521 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM45);
4522 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4523 if (eptr-- == pp) break; /* Stop if tried at original pos */
4524 for (;;) /* Move back over one extended */
4525 {
4526 int len = 1;
4527 if (!utf8) c = *eptr; else
4528 {
4529 BACKCHAR(eptr);
4530 GETCHARLEN(c, eptr, len);
4531 }
4532 prop_category = UCD_CATEGORY(c);
4533 if (prop_category != ucp_M) break;
4534 eptr--;
4535 }
4536 }
4537 }
4538
4539 else
4540 #endif /* SUPPORT_UCP */
4541
4542 #ifdef SUPPORT_UTF8
4543 /* UTF-8 mode */
4544
4545 if (utf8)
4546 {
4547 switch(ctype)
4548 {
4549 case OP_ANY:
4550 if (max < INT_MAX)
4551 {
4552 for (i = min; i < max; i++)
4553 {
4554 if (eptr >= md->end_subject)
4555 {
4556 SCHECK_PARTIAL();
4557 break;
4558 }
4559 if (IS_NEWLINE(eptr)) break;
4560 eptr++;
4561 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
4562 }
4563 }
4564
4565 /* Handle unlimited UTF-8 repeat */
4566
4567 else
4568 {
4569 for (i = min; i < max; i++)
4570 {
4571 if (eptr >= md->end_subject)
4572 {
4573 SCHECK_PARTIAL();
4574 break;
4575 }
4576 if (IS_NEWLINE(eptr)) break;
4577 eptr++;
4578 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
4579 }
4580 }
4581 break;
4582
4583 case OP_ALLANY:
4584 if (max < INT_MAX)
4585 {
4586 for (i = min; i < max; i++)
4587 {
4588 if (eptr >= md->end_subject)
4589 {
4590 SCHECK_PARTIAL();
4591 break;
4592 }
4593 eptr++;
4594 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
4595 }
4596 }
4597 else eptr = md->end_subject; /* Unlimited UTF-8 repeat */
4598 break;
4599
4600 /* The byte case is the same as non-UTF8 */
4601
4602 case OP_ANYBYTE:
4603 c = max - min;
4604 if (c > (unsigned int)(md->end_subject - eptr))
4605 {
4606 eptr = md->end_subject;
4607 SCHECK_PARTIAL();
4608 }
4609 else eptr += c;
4610 break;
4611
4612 case OP_ANYNL:
4613 for (i = min; i < max; i++)
4614 {
4615 int len = 1;
4616 if (eptr >= md->end_subject)
4617 {
4618 SCHECK_PARTIAL();
4619 break;
4620 }
4621 GETCHARLEN(c, eptr, len);
4622 if (c == 0x000d)
4623 {
4624 if (++eptr >= md->end_subject) break;
4625 if (*eptr == 0x000a) eptr++;
4626 }
4627 else
4628 {
4629 if (c != 0x000a &&
4630 (md->bsr_anycrlf ||
4631 (c != 0x000b && c != 0x000c &&
4632 c != 0x0085 && c != 0x2028 && c != 0x2029)))
4633 break;
4634 eptr += len;
4635 }
4636 }
4637 break;
4638
4639 case OP_NOT_HSPACE:
4640 case OP_HSPACE:
4641 for (i = min; i < max; i++)
4642 {
4643 BOOL gotspace;
4644 int len = 1;
4645 if (eptr >= md->end_subject)
4646 {
4647 SCHECK_PARTIAL();
4648 break;
4649 }
4650 GETCHARLEN(c, eptr, len);
4651 switch(c)
4652 {
4653 default: gotspace = FALSE; break;
4654 case 0x09: /* HT */
4655 case 0x20: /* SPACE */
4656 case 0xa0: /* NBSP */
4657 case 0x1680: /* OGHAM SPACE MARK */
4658 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
4659 case 0x2000: /* EN QUAD */
4660 case 0x2001: /* EM QUAD */
4661 case 0x2002: /* EN SPACE */
4662 case 0x2003: /* EM SPACE */
4663 case 0x2004: /* THREE-PER-EM SPACE */
4664 case 0x2005: /* FOUR-PER-EM SPACE */
4665 case 0x2006: /* SIX-PER-EM SPACE */
4666 case 0x2007: /* FIGURE SPACE */
4667 case 0x2008: /* PUNCTUATION SPACE */
4668 case 0x2009: /* THIN SPACE */
4669 case 0x200A: /* HAIR SPACE */
4670 case 0x202f: /* NARROW NO-BREAK SPACE */
4671 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
4672 case 0x3000: /* IDEOGRAPHIC SPACE */
4673 gotspace = TRUE;
4674 break;
4675 }
4676 if (gotspace == (ctype == OP_NOT_HSPACE)) break;
4677 eptr += len;
4678 }
4679 break;
4680
4681 case OP_NOT_VSPACE:
4682 case OP_VSPACE:
4683 for (i = min; i < max; i++)
4684 {
4685 BOOL gotspace;
4686 int len = 1;
4687 if (eptr >= md->end_subject)
4688 {
4689 SCHECK_PARTIAL();
4690 break;
4691 }
4692 GETCHARLEN(c, eptr, len);
4693 switch(c)
4694 {
4695 default: gotspace = FALSE; break;
4696 case 0x0a: /* LF */
4697 case 0x0b: /* VT */
4698 case 0x0c: /* FF */
4699 case 0x0d: /* CR */
4700 case 0x85: /* NEL */
4701 case 0x2028: /* LINE SEPARATOR */
4702 case 0x2029: /* PARAGRAPH SEPARATOR */
4703 gotspace = TRUE;
4704 break;
4705 }
4706 if (gotspace == (ctype == OP_NOT_VSPACE)) break;
4707 eptr += len;
4708 }
4709 break;
4710
4711 case OP_NOT_DIGIT:
4712 for (i = min; i < max; i++)
4713 {
4714 int len = 1;
4715 if (eptr >= md->end_subject)
4716 {
4717 SCHECK_PARTIAL();
4718 break;
4719 }
4720 GETCHARLEN(c, eptr, len);
4721 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
4722 eptr+= len;
4723 }
4724 break;
4725
4726 case OP_DIGIT:
4727 for (i = min; i < max; i++)
4728 {
4729 int len = 1;
4730 if (eptr >= md->end_subject)
4731 {
4732 SCHECK_PARTIAL();
4733 break;
4734 }
4735 GETCHARLEN(c, eptr, len);
4736 if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break;
4737 eptr+= len;
4738 }
4739 break;
4740
4741 case OP_NOT_WHITESPACE:
4742 for (i = min; i < max; i++)
4743 {
4744 int len = 1;
4745 if (eptr >= md->end_subject)
4746 {
4747 SCHECK_PARTIAL();
4748 break;
4749 }
4750 GETCHARLEN(c, eptr, len);
4751 if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
4752 eptr+= len;
4753 }
4754 break;
4755
4756 case OP_WHITESPACE:
4757 for (i = min; i < max; i++)
4758 {
4759 int len = 1;
4760 if (eptr >= md->end_subject)
4761 {
4762 SCHECK_PARTIAL();
4763 break;
4764 }
4765 GETCHARLEN(c, eptr, len);
4766 if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break;
4767 eptr+= len;
4768 }
4769 break;
4770
4771 case OP_NOT_WORDCHAR:
4772 for (i = min; i < max; i++)
4773 {
4774 int len = 1;
4775 if (eptr >= md->end_subject)
4776 {
4777 SCHECK_PARTIAL();
4778 break;
4779 }
4780 GETCHARLEN(c, eptr, len);
4781 if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
4782 eptr+= len;
4783 }
4784 break;
4785
4786 case OP_WORDCHAR:
4787 for (i = min; i < max; i++)
4788 {
4789 int len = 1;
4790 if (eptr >= md->end_subject)
4791 {
4792 SCHECK_PARTIAL();
4793 break;
4794 }
4795 GETCHARLEN(c, eptr, len);
4796 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break;
4797 eptr+= len;
4798 }
4799 break;
4800
4801 default:
4802 RRETURN(PCRE_ERROR_INTERNAL);
4803 }
4804
4805 /* eptr is now past the end of the maximum run */
4806
4807 if (possessive) continue;
4808 for(;;)
4809 {
4810 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM46);
4811 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4812 if (eptr-- == pp) break; /* Stop if tried at original pos */
4813 BACKCHAR(eptr);
4814 }
4815 }
4816 else
4817 #endif /* SUPPORT_UTF8 */
4818
4819 /* Not UTF-8 mode */
4820 {
4821 switch(ctype)
4822 {
4823 case OP_ANY:
4824 for (i = min; i < max; i++)
4825 {
4826 if (eptr >= md->end_subject)
4827 {
4828 SCHECK_PARTIAL();
4829 break;
4830 }
4831 if (IS_NEWLINE(eptr)) break;
4832 eptr++;
4833 }
4834 break;
4835
4836 case OP_ALLANY:
4837 case OP_ANYBYTE:
4838 c = max - min;
4839 if (c > (unsigned int)(md->end_subject - eptr))
4840 {
4841 eptr = md->end_subject;
4842 SCHECK_PARTIAL();
4843 }
4844 else eptr += c;
4845 break;
4846
4847 case OP_ANYNL:
4848 for (i = min; i < max; i++)
4849 {
4850 if (eptr >= md->end_subject)
4851 {
4852 SCHECK_PARTIAL();
4853 break;
4854 }
4855 c = *eptr;
4856 if (c == 0x000d)
4857 {
4858 if (++eptr >= md->end_subject) break;
4859 if (*eptr == 0x000a) eptr++;
4860 }
4861 else
4862 {
4863 if (c != 0x000a &&
4864 (md->bsr_anycrlf ||
4865 (c != 0x000b && c != 0x000c && c != 0x0085)))
4866 break;
4867 eptr++;
4868 }
4869 }
4870 break;
4871
4872 case OP_NOT_HSPACE:
4873 for (i = min; i < max; i++)
4874 {
4875 if (eptr >= md->end_subject)
4876 {
4877 SCHECK_PARTIAL();
4878 break;
4879 }
4880 c = *eptr;
4881 if (c == 0x09 || c == 0x20 || c == 0xa0) break;
4882 eptr++;
4883 }
4884 break;
4885
4886 case OP_HSPACE:
4887 for (i = min; i < max; i++)
4888 {
4889 if (eptr >= md->end_subject)
4890 {
4891 SCHECK_PARTIAL();
4892 break;
4893 }
4894 c = *eptr;
4895 if (c != 0x09 && c != 0x20 && c != 0xa0) break;
4896 eptr++;
4897 }
4898 break;
4899
4900 case OP_NOT_VSPACE:
4901 for (i = min; i < max; i++)
4902 {
4903 if (eptr >= md->end_subject)
4904 {
4905 SCHECK_PARTIAL();
4906 break;
4907 }
4908 c = *eptr;
4909 if (c == 0x0a || c == 0x0b || c == 0x0c || c == 0x0d || c == 0x85)
4910 break;
4911 eptr++;
4912 }
4913 break;
4914
4915 case OP_VSPACE:
4916 for (i = min; i < max; i++)
4917 {
4918 if (eptr >= md->end_subject)
4919 {
4920 SCHECK_PARTIAL();
4921 break;
4922 }
4923 c = *eptr;
4924 if (c != 0x0a && c != 0x0b && c != 0x0c && c != 0x0d && c != 0x85)
4925 break;
4926 eptr++;
4927 }
4928 break;
4929
4930 case OP_NOT_DIGIT:
4931 for (i = min; i < max; i++)
4932 {
4933 if (eptr >= md->end_subject)
4934 {
4935 SCHECK_PARTIAL();
4936 break;
4937 }
4938 if ((md->ctypes[*eptr] & ctype_digit) != 0) break;
4939 eptr++;
4940 }
4941 break;
4942
4943 case OP_DIGIT:
4944 for (i = min; i < max; i++)
4945 {
4946 if (eptr >= md->end_subject)
4947 {
4948 SCHECK_PARTIAL();
4949 break;
4950 }
4951 if ((md->ctypes[*eptr] & ctype_digit) == 0) break;
4952 eptr++;
4953 }
4954 break;
4955
4956 case OP_NOT_WHITESPACE:
4957 for (i = min; i < max; i++)
4958 {
4959 if (eptr >= md->end_subject)
4960 {
4961 SCHECK_PARTIAL();
4962 break;
4963 }
4964 if ((md->ctypes[*eptr] & ctype_space) != 0) break;
4965 eptr++;
4966 }
4967 break;
4968
4969 case OP_WHITESPACE:
4970 for (i = min; i < max; i++)
4971 {
4972 if (eptr >= md->end_subject)
4973 {
4974 SCHECK_PARTIAL();
4975 break;
4976 }
4977 if ((md->ctypes[*eptr] & ctype_space) == 0) break;
4978 eptr++;
4979 }
4980 break;
4981
4982 case OP_NOT_WORDCHAR:
4983 for (i = min; i < max; i++)
4984 {
4985 if (eptr >= md->end_subject)
4986 {
4987 SCHECK_PARTIAL();
4988 break;
4989 }
4990 if ((md->ctypes[*eptr] & ctype_word) != 0) break;
4991 eptr++;
4992 }
4993 break;
4994
4995 case OP_WORDCHAR:
4996 for (i = min; i < max; i++)
4997 {
4998 if (eptr >= md->end_subject)
4999 {
5000 SCHECK_PARTIAL();
5001 break;
5002 }
5003 if ((md->ctypes[*eptr] & ctype_word) == 0) break;
5004 eptr++;
5005 }
5006 break;
5007
5008 default:
5009 RRETURN(PCRE_ERROR_INTERNAL);
5010 }
5011
5012 /* eptr is now past the end of the maximum run */
5013
5014 if (possessive) continue;
5015 while (eptr >= pp)
5016 {
5017 RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM47);
5018 eptr--;
5019 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5020 }
5021 }
5022
5023 /* Get here if we can't make it match with any permitted repetitions */
5024
5025 RRETURN(MATCH_NOMATCH);
5026 }
5027 /* Control never gets here */
5028
5029 /* There's been some horrible disaster. Arrival here can only mean there is
5030 something seriously wrong in the code above or the OP_xxx definitions. */
5031
5032 default:
5033 DPRINTF(("Unknown opcode %d\n", *ecode));
5034 RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
5035 }
5036
5037 /* Do not stick any code in here without much thought; it is assumed
5038 that "continue" in the code above comes out to here to repeat the main
5039 loop. */
5040
5041 } /* End of main loop */
5042 /* Control never reaches here */
5043
5044
5045 /* When compiling to use the heap rather than the stack for recursive calls to
5046 match(), the RRETURN() macro jumps here. The number that is saved in
5047 frame->Xwhere indicates which label we actually want to return to. */
5048
5049 #ifdef NO_RECURSE
5050 #define LBL(val) case val: goto L_RM##val;
5051 HEAP_RETURN:
5052 switch (frame->Xwhere)
5053 {
5054 LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
5055 LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
5056 LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
5057 LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
5058 LBL(53) LBL(54)
5059 #ifdef SUPPORT_UTF8
5060 LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)
5061 LBL(32) LBL(34) LBL(42) LBL(46)
5062 #ifdef SUPPORT_UCP
5063 LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
5064 #endif /* SUPPORT_UCP */
5065 #endif /* SUPPORT_UTF8 */
5066 default:
5067 DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
5068 return PCRE_ERROR_INTERNAL;
5069 }
5070 #undef LBL
5071 #endif /* NO_RECURSE */
5072 }
5073
5074
5075 /***************************************************************************
5076 ****************************************************************************
5077 RECURSION IN THE match() FUNCTION
5078
5079 Undefine all the macros that were defined above to handle this. */
5080
5081 #ifdef NO_RECURSE
5082 #undef eptr
5083 #undef ecode
5084 #undef mstart
5085 #undef offset_top
5086 #undef ims
5087 #undef eptrb
5088 #undef flags
5089
5090 #undef callpat
5091 #undef charptr
5092 #undef data
5093 #undef next
5094 #undef pp
5095 #undef prev
5096 #undef saved_eptr
5097
5098 #undef new_recursive
5099
5100 #undef cur_is_word
5101 #undef condition
5102 #undef prev_is_word
5103
5104 #undef original_ims
5105
5106 #undef ctype
5107 #undef length
5108 #undef max
5109 #undef min
5110 #undef number
5111 #undef offset
5112 #undef op
5113 #undef save_capture_last
5114 #undef save_offset1
5115 #undef save_offset2
5116 #undef save_offset3
5117 #undef stacksave
5118
5119 #undef newptrb
5120
5121 #endif
5122
5123 /* These two are defined as macros in both cases */
5124
5125 #undef fc
5126 #undef fi
5127
5128 /***************************************************************************
5129 ***************************************************************************/
5130
5131
5132
5133 /*************************************************
5134 * Execute a Regular Expression *
5135 *************************************************/
5136
5137 /* This function applies a compiled re to a subject string and picks out
5138 portions of the string if it matches. Two elements in the vector are set for
5139 each substring: the offsets to the start and end of the substring.
5140
5141 Arguments:
5142 argument_re points to the compiled expression
5143 extra_data points to extra data or is NULL
5144 subject points to the subject string
5145 length length of subject string (may contain binary zeros)
5146 start_offset where to start in the subject string
5147 options option bits
5148 offsets points to a vector of ints to be filled in with offsets
5149 offsetcount the number of elements in the vector
5150
5151 Returns: > 0 => success; value is the number of elements filled in
5152 = 0 => success, but offsets is not big enough
5153 -1 => failed to match
5154 < -1 => some kind of unexpected problem
5155 */
5156
5157 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
5158 pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
5159 PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5160 int offsetcount)
5161 {
5162 int rc, resetcount, ocount;
5163 int first_byte = -1;
5164 int req_byte = -1;
5165 int req_byte2 = -1;
5166 int newline;
5167 unsigned long int ims;
5168 BOOL using_temporary_offsets = FALSE;
5169 BOOL anchored;
5170 BOOL startline;
5171 BOOL firstline;
5172 BOOL first_byte_caseless = FALSE;
5173 BOOL req_byte_caseless = FALSE;
5174 BOOL utf8;
5175 match_data match_block;
5176 match_data *md = &match_block;
5177 const uschar *tables;
5178 const uschar *start_bits = NULL;
5179 USPTR start_match = (USPTR)subject + start_offset;
5180 USPTR end_subject;
5181 USPTR start_partial = NULL;
5182 USPTR req_byte_ptr = start_match - 1;
5183
5184 pcre_study_data internal_study;
5185 const pcre_study_data *study;
5186
5187 real_pcre internal_re;
5188 const real_pcre *external_re = (const real_pcre *)argument_re;
5189 const real_pcre *re = external_re;
5190
5191 /* Plausibility checks */
5192
5193 if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
5194 if (re == NULL || subject == NULL ||
5195 (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL;
5196 if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
5197
5198 /* This information is for finding all the numbers associated with a given
5199 name, for condition testing. */
5200
5201 md->name_table = (uschar *)re + re->name_table_offset;
5202 md->name_count = re->name_count;
5203 md->name_entry_size = re->name_entry_size;
5204
5205 /* Fish out the optional data from the extra_data structure, first setting
5206 the default values. */
5207
5208 study = NULL;
5209 md->match_limit = MATCH_LIMIT;
5210 md->match_limit_recursion = MATCH_LIMIT_RECURSION;
5211 md->callout_data = NULL;
5212
5213 /* The table pointer is always in native byte order. */
5214
5215 tables = external_re->tables;
5216
5217 if (extra_data != NULL)
5218 {
5219 register unsigned int flags = extra_data->flags;
5220 if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
5221 study = (const pcre_study_data *)extra_data->study_data;
5222 if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
5223 md->match_limit = extra_data->match_limit;
5224 if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
5225 md->match_limit_recursion = extra_data->match_limit_recursion;
5226 if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
5227 md->callout_data = extra_data->callout_data;
5228 if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
5229 }
5230
5231 /* If the exec call supplied NULL for tables, use the inbuilt ones. This
5232 is a feature that makes it possible to save compiled regex and re-use them
5233 in other programs later. */
5234
5235 if (tables == NULL) tables = _pcre_default_tables;
5236
5237 /* Check that the first field in the block is the magic number. If it is not,
5238 test for a regex that was compiled on a host of opposite endianness. If this is
5239 the case, flipped values are put in internal_re and internal_study if there was
5240 study data too. */
5241
5242 if (re->magic_number != MAGIC_NUMBER)
5243 {
5244 re = _pcre_try_flipped(re, &internal_re, study, &internal_study);
5245 if (re == NULL) return PCRE_ERROR_BADMAGIC;
5246 if (study != NULL) study = &internal_study;
5247 }
5248
5249 /* Set up other data */
5250
5251 anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
5252 startline = (re->flags & PCRE_STARTLINE) != 0;
5253 firstline = (re->options & PCRE_FIRSTLINE) != 0;
5254
5255 /* The code starts after the real_pcre block and the capture name table. */
5256
5257 md->start_code = (const uschar *)external_re + re->name_table_offset +
5258 re->name_count * re->name_entry_size;
5259
5260 md->start_subject = (USPTR)subject;
5261 md->start_offset = start_offset;
5262 md->end_subject = md->start_subject + length;
5263 end_subject = md->end_subject;
5264
5265 md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
5266 utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
5267 md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
5268
5269 md->notbol = (options & PCRE_NOTBOL) != 0;
5270 md->noteol = (options & PCRE_NOTEOL) != 0;
5271 md->notempty = (options & PCRE_NOTEMPTY) != 0;
5272 md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
5273 md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5274 ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5275 md->hitend = FALSE;
5276
5277 md->recursive = NULL; /* No recursion at top level */
5278
5279 md->lcc = tables + lcc_offset;
5280 md->ctypes = tables + ctypes_offset;
5281
5282 /* Handle different \R options. */
5283
5284 switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
5285 {
5286 case 0:
5287 if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
5288 md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
5289 else
5290 #ifdef BSR_ANYCRLF
5291 md->bsr_anycrlf = TRUE;
5292 #else
5293 md->bsr_anycrlf = FALSE;
5294 #endif
5295 break;
5296
5297 case PCRE_BSR_ANYCRLF:
5298 md->bsr_anycrlf = TRUE;
5299 break;
5300
5301 case PCRE_BSR_UNICODE:
5302 md->bsr_anycrlf = FALSE;
5303 break;
5304
5305 default: return PCRE_ERROR_BADNEWLINE;
5306 }
5307
5308 /* Handle different types of newline. The three bits give eight cases. If
5309 nothing is set at run time, whatever was used at compile time applies. */
5310
5311 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
5312 (pcre_uint32)options) & PCRE_NEWLINE_BITS)
5313 {
5314 case 0: newline = NEWLINE; break; /* Compile-time default */
5315 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
5316 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
5317 case PCRE_NEWLINE_CR+
5318 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
5319 case PCRE_NEWLINE_ANY: newline = -1; break;
5320 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
5321 default: return PCRE_ERROR_BADNEWLINE;
5322 }
5323
5324 if (newline == -2)
5325 {
5326 md->nltype = NLTYPE_ANYCRLF;
5327 }
5328 else if (newline < 0)
5329 {
5330 md->nltype = NLTYPE_ANY;
5331 }
5332 else
5333 {
5334 md->nltype = NLTYPE_FIXED;
5335 if (newline > 255)
5336 {
5337 md->nllen = 2;
5338 md->nl[0] = (newline >> 8) & 255;
5339 md->nl[1] = newline & 255;
5340 }
5341 else
5342 {
5343 md->nllen = 1;
5344 md->nl[0] = newline;
5345 }
5346 }
5347
5348 /* Partial matching was originally supported only for a restricted set of
5349 regexes; from release 8.00 there are no restrictions, but the bits are still
5350 defined (though never set). So there's no harm in leaving this code. */
5351
5352 if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
5353 return PCRE_ERROR_BADPARTIAL;
5354
5355 /* Check a UTF-8 string if required. Unfortunately there's no way of passing
5356 back the character offset. */
5357
5358 #ifdef SUPPORT_UTF8
5359 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5360 {
5361 if (_pcre_valid_utf8((USPTR)subject, length) >= 0)
5362 return PCRE_ERROR_BADUTF8;
5363 if (start_offset > 0 && start_offset < length)
5364 {
5365 int tb = ((USPTR)subject)[start_offset];
5366 if (tb > 127)
5367 {
5368 tb &= 0xc0;
5369 if (tb != 0 && tb != 0xc0) return PCRE_ERROR_BADUTF8_OFFSET;
5370 }
5371 }
5372 }
5373 #endif
5374
5375 /* The ims options can vary during the matching as a result of the presence
5376 of (?ims) items in the pattern. They are kept in a local variable so that
5377 restoring at the exit of a group is easy. */
5378
5379 ims = re->options & (PCRE_CASELESS|PCRE_MULTILINE|PCRE_DOTALL);
5380
5381 /* If the expression has got more back references than the offsets supplied can
5382 hold, we get a temporary chunk of working store to use during the matching.
5383 Otherwise, we can use the vector supplied, rounding down its size to a multiple
5384 of 3. */
5385
5386 ocount = offsetcount - (offsetcount % 3);
5387
5388 if (re->top_backref > 0 && re->top_backref >= ocount/3)
5389 {
5390 ocount = re->top_backref * 3 + 3;
5391 md->offset_vector = (int *)(pcre_malloc)(ocount * sizeof(int));
5392 if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
5393 using_temporary_offsets = TRUE;
5394 DPRINTF(("Got memory to hold back references\n"));
5395 }
5396 else md->offset_vector = offsets;
5397
5398 md->offset_end = ocount;
5399 md->offset_max = (2*ocount)/3;
5400 md->offset_overflow = FALSE;
5401 md->capture_last = -1;
5402
5403 /* Compute the minimum number of offsets that we need to reset each time. Doing
5404 this makes a huge difference to execution time when there aren't many brackets
5405 in the pattern. */
5406
5407 resetcount = 2 + re->top_bracket * 2;
5408 if (resetcount > offsetcount) resetcount = ocount;
5409
5410 /* Reset the working variable associated with each extraction. These should
5411 never be used unless previously set, but they get saved and restored, and so we
5412 initialize them to avoid reading uninitialized locations. */
5413
5414 if (md->offset_vector != NULL)
5415 {
5416 register int *iptr = md->offset_vector + ocount;
5417 register int *iend = iptr - resetcount/2 + 1;
5418 while (--iptr >= iend) *iptr = -1;
5419 }
5420
5421 /* Set up the first character to match, if available. The first_byte value is
5422 never set for an anchored regular expression, but the anchoring may be forced
5423 at run time, so we have to test for anchoring. The first char may be unset for
5424 an unanchored pattern, of course. If there's no first char and the pattern was
5425 studied, there may be a bitmap of possible first characters. */
5426
5427 if (!anchored)
5428 {
5429 if ((re->flags & PCRE_FIRSTSET) != 0)
5430 {
5431 first_byte = re->first_byte & 255;
5432 if ((first_byte_caseless = ((re->first_byte & REQ_CASELESS) != 0)) == TRUE)
5433 first_byte = md->lcc[first_byte];
5434 }
5435 else
5436 if (!startline && study != NULL &&
5437 (study->flags & PCRE_STUDY_MAPPED) != 0)
5438 start_bits = study->start_bits;
5439 }
5440
5441 /* For anchored or unanchored matches, there may be a "last known required
5442 character" set. */
5443
5444 if ((re->flags & PCRE_REQCHSET) != 0)
5445 {
5446 req_byte = re->req_byte & 255;
5447 req_byte_caseless = (re->req_byte & REQ_CASELESS) != 0;
5448 req_byte2 = (tables + fcc_offset)[req_byte]; /* case flipped */
5449 }
5450
5451
5452 /* ==========================================================================*/
5453
5454 /* Loop for handling unanchored repeated matching attempts; for anchored regexs
5455 the loop runs just once. */
5456
5457 for(;;)
5458 {
5459 USPTR save_end_subject = end_subject;
5460 USPTR new_start_match;
5461
5462 /* Reset the maximum number of extractions we might see. */
5463
5464 if (md->offset_vector != NULL)
5465 {
5466 register int *iptr = md->offset_vector;
5467 register int *iend = iptr + resetcount;
5468 while (iptr < iend) *iptr++ = -1;
5469 }
5470
5471 /* If firstline is TRUE, the start of the match is constrained to the first
5472 line of a multiline string. That is, the match must be before or at the first
5473 newline. Implement this by temporarily adjusting end_subject so that we stop
5474 scanning at a newline. If the match fails at the newline, later code breaks
5475 this loop. */
5476
5477 if (firstline)
5478 {
5479 USPTR t = start_match;
5480 #ifdef SUPPORT_UTF8
5481 if (utf8)
5482 {
5483 while (t < md->end_subject && !IS_NEWLINE(t))
5484 {
5485 t++;
5486 while (t < end_subject && (*t & 0xc0) == 0x80) t++;
5487 }
5488 }
5489 else
5490 #endif
5491 while (t < md->end_subject && !IS_NEWLINE(t)) t++;
5492 end_subject = t;
5493 }
5494
5495 /* There are some optimizations that avoid running the match if a known
5496 starting point is not found, or if a known later character is not present.
5497 However, there is an option that disables these, for testing and for ensuring
5498 that all callouts do actually occur. */
5499
5500 if ((options & PCRE_NO_START_OPTIMIZE) == 0)
5501 {
5502 /* Advance to a unique first byte if there is one. */
5503
5504 if (first_byte >= 0)
5505 {
5506 if (first_byte_caseless)
5507 while (start_match < end_subject && md->lcc[*start_match] != first_byte)
5508 start_match++;
5509 else
5510 while (start_match < end_subject && *start_match != first_byte)
5511 start_match++;
5512 }
5513
5514 /* Or to just after a linebreak for a multiline match */
5515
5516 else if (startline)
5517 {
5518 if (start_match > md->start_subject + start_offset)
5519 {
5520 #ifdef SUPPORT_UTF8
5521 if (utf8)
5522 {
5523 while (start_match < end_subject && !WAS_NEWLINE(start_match))
5524 {
5525 start_match++;
5526 while(start_match < end_subject && (*start_match & 0xc0) == 0x80)
5527 start_match++;
5528 }
5529 }
5530 else
5531 #endif
5532 while (start_match < end_subject && !WAS_NEWLINE(start_match))
5533 start_match++;
5534
5535 /* If we have just passed a CR and the newline option is ANY or ANYCRLF,
5536 and we are now at a LF, advance the match position by one more character.
5537 */
5538
5539 if (start_match[-1] == CHAR_CR &&
5540 (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&
5541 start_match < end_subject &&
5542 *start_match == CHAR_NL)
5543 start_match++;
5544 }
5545 }
5546
5547 /* Or to a non-unique first byte after study */
5548
5549 else if (start_bits != NULL)
5550 {
5551 while (start_match < end_subject)
5552 {
5553 register unsigned int c = *start_match;
5554 if ((start_bits[c/8] & (1 << (c&7))) == 0) start_match++;
5555 else break;
5556 }
5557 }
5558 } /* Starting optimizations */
5559
5560 /* Restore fudged end_subject */
5561
5562 end_subject = save_end_subject;
5563
5564 /* The following two optimizations are disabled for partial matching or if
5565 disabling is explicitly requested. */
5566
5567 if ((options & PCRE_NO_START_OPTIMIZE) == 0 && !md->partial)
5568 {
5569 /* If the pattern was studied, a minimum subject length may be set. This is
5570 a lower bound; no actual string of that length may actually match the
5571 pattern. Although the value is, strictly, in characters, we treat it as
5572 bytes to avoid spending too much time in this optimization. */
5573
5574 if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
5575 end_subject - start_match < study->minlength)
5576 {
5577 rc = MATCH_NOMATCH;
5578 break;
5579 }
5580
5581 /* If req_byte is set, we know that that character must appear in the
5582 subject for the match to succeed. If the first character is set, req_byte
5583 must be later in the subject; otherwise the test starts at the match point.
5584 This optimization can save a huge amount of backtracking in patterns with
5585 nested unlimited repeats that aren't going to match. Writing separate code
5586 for cased/caseless versions makes it go faster, as does using an
5587 autoincrement and backing off on a match.
5588
5589 HOWEVER: when the subject string is very, very long, searching to its end
5590 can take a long time, and give bad performance on quite ordinary patterns.
5591 This showed up when somebody was matching something like /^\d+C/ on a
5592 32-megabyte string... so we don't do this when the string is sufficiently
5593 long. */
5594
5595 if (req_byte >= 0 && end_subject - start_match < REQ_BYTE_MAX)
5596 {
5597 register USPTR p = start_match + ((first_byte >= 0)? 1 : 0);
5598
5599 /* We don't need to repeat the search if we haven't yet reached the
5600 place we found it at last time. */
5601
5602 if (p > req_byte_ptr)
5603 {
5604 if (req_byte_caseless)
5605 {
5606 while (p < end_subject)
5607 {
5608 register int pp = *p++;
5609 if (pp == req_byte || pp == req_byte2) { p--; break; }
5610 }
5611 }
5612 else
5613 {
5614 while (p < end_subject)
5615 {
5616 if (*p++ == req_byte) { p--; break; }
5617 }
5618 }
5619
5620 /* If we can't find the required character, break the matching loop,
5621 forcing a match failure. */
5622
5623 if (p >= end_subject)
5624 {
5625 rc = MATCH_NOMATCH;
5626 break;
5627 }
5628
5629 /* If we have found the required character, save the point where we
5630 found it, so that we don't search again next time round the loop if
5631 the start hasn't passed this character yet. */
5632
5633 req_byte_ptr = p;
5634 }
5635 }
5636 }
5637
5638 #ifdef PCRE_DEBUG /* Sigh. Some compilers never learn. */
5639 printf(">>>> Match against: ");
5640 pchars(start_match, end_subject - start_match, TRUE, md);
5641 printf("\n");
5642 #endif
5643
5644 /* OK, we can now run the match. If "hitend" is set afterwards, remember the
5645 first starting point for which a partial match was found. */
5646
5647 md->start_match_ptr = start_match;
5648 md->start_used_ptr = start_match;
5649 md->match_call_count = 0;
5650 rc = match(start_match, md->start_code, start_match, 2, md, ims, NULL, 0, 0);
5651 if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
5652
5653 switch(rc)
5654 {
5655 /* NOMATCH and PRUNE advance by one character. THEN at this level acts
5656 exactly like PRUNE. */
5657
5658 case MATCH_NOMATCH:
5659 case MATCH_PRUNE:
5660 case MATCH_THEN:
5661 new_start_match = start_match + 1;
5662 #ifdef SUPPORT_UTF8
5663 if (utf8)
5664 while(new_start_match < end_subject && (*new_start_match & 0xc0) == 0x80)
5665 new_start_match++;
5666 #endif
5667 break;
5668
5669 /* SKIP passes back the next starting point explicitly. */
5670
5671 case MATCH_SKIP:
5672 new_start_match = md->start_match_ptr;
5673 break;
5674
5675 /* COMMIT disables the bumpalong, but otherwise behaves as NOMATCH. */
5676
5677 case MATCH_COMMIT:
5678 rc = MATCH_NOMATCH;
5679 goto ENDLOOP;
5680
5681 /* Any other return is either a match, or some kind of error. */
5682
5683 default:
5684 goto ENDLOOP;
5685 }
5686
5687 /* Control reaches here for the various types of "no match at this point"
5688 result. Reset the code to MATCH_NOMATCH for subsequent checking. */
5689
5690 rc = MATCH_NOMATCH;
5691
5692 /* If PCRE_FIRSTLINE is set, the match must happen before or at the first
5693 newline in the subject (though it may continue over the newline). Therefore,
5694 if we have just failed to match, starting at a newline, do not continue. */
5695
5696 if (firstline && IS_NEWLINE(start_match)) break;
5697
5698 /* Advance to new matching position */
5699
5700 start_match = new_start_match;
5701
5702 /* Break the loop if the pattern is anchored or if we have passed the end of
5703 the subject. */
5704
5705 if (anchored || start_match > end_subject) break;
5706
5707 /* If we have just passed a CR and we are now at a LF, and the pattern does
5708 not contain any explicit matches for \r or \n, and the newline option is CRLF
5709 or ANY or ANYCRLF, advance the match position by one more character. */
5710
5711 if (start_match[-1] == CHAR_CR &&
5712 start_match < end_subject &&
5713 *start_match == CHAR_NL &&
5714 (re->flags & PCRE_HASCRORLF) == 0 &&
5715 (md->nltype == NLTYPE_ANY ||
5716 md->nltype == NLTYPE_ANYCRLF ||
5717 md->nllen == 2))
5718 start_match++;
5719
5720 } /* End of for(;;) "bumpalong" loop */
5721
5722 /* ==========================================================================*/
5723
5724 /* We reach here when rc is not MATCH_NOMATCH, or if one of the stopping
5725 conditions is true:
5726
5727 (1) The pattern is anchored or the match was failed by (*COMMIT);
5728
5729 (2) We are past the end of the subject;
5730
5731 (3) PCRE_FIRSTLINE is set and we have failed to match at a newline, because
5732 this option requests that a match occur at or before the first newline in
5733 the subject.
5734
5735 When we have a match and the offset vector is big enough to deal with any
5736 backreferences, captured substring offsets will already be set up. In the case
5737 where we had to get some local store to hold offsets for backreference
5738 processing, copy those that we can. In this case there need not be overflow if
5739 certain parts of the pattern were not used, even though there are more
5740 capturing parentheses than vector slots. */
5741
5742 ENDLOOP:
5743
5744 if (rc == MATCH_MATCH)
5745 {
5746 if (using_temporary_offsets)
5747 {
5748 if (offsetcount >= 4)
5749 {
5750 memcpy(offsets + 2, md->offset_vector + 2,
5751 (offsetcount - 2) * sizeof(int));
5752 DPRINTF(("Copied offsets from temporary memory\n"));
5753 }
5754 if (md->end_offset_top > offsetcount) md->offset_overflow = TRUE;
5755 DPRINTF(("Freeing temporary memory\n"));
5756 (pcre_free)(md->offset_vector);
5757 }
5758
5759 /* Set the return code to the number of captured strings, or 0 if there are
5760 too many to fit into the vector. */
5761
5762 rc = md->offset_overflow? 0 : md->end_offset_top/2;
5763
5764 /* If there is space, set up the whole thing as substring 0. The value of
5765 md->start_match_ptr might be modified if \K was encountered on the success
5766 matching path. */
5767
5768 if (offsetcount < 2) rc = 0; else
5769 {
5770 offsets[0] = md->start_match_ptr - md->start_subject;
5771 offsets[1] = md->end_match_ptr - md->start_subject;
5772 }
5773
5774 DPRINTF((">>>> returning %d\n", rc));
5775 return rc;
5776 }
5777
5778 /* Control gets here if there has been an error, or if the overall match
5779 attempt has failed at all permitted starting positions. */
5780
5781 if (using_temporary_offsets)
5782 {
5783 DPRINTF(("Freeing temporary memory\n"));
5784 (pcre_free)(md->offset_vector);
5785 }
5786
5787 if (rc != MATCH_NOMATCH && rc != PCRE_ERROR_PARTIAL)
5788 {
5789 DPRINTF((">>>> error: returning %d\n", rc));
5790 return rc;
5791 }
5792 else if (start_partial != NULL)
5793 {
5794 DPRINTF((">>>> returning PCRE_ERROR_PARTIAL\n"));
5795 if (offsetcount > 1)
5796 {
5797 offsets[0] = start_partial - (USPTR)subject;
5798 offsets[1] = end_subject - (USPTR)subject;
5799 }
5800 return PCRE_ERROR_PARTIAL;
5801 }
5802 else
5803 {
5804 DPRINTF((">>>> returning PCRE_ERROR_NOMATCH\n"));
5805 return PCRE_ERROR_NOMATCH;
5806 }
5807 }
5808
5809 /* End of pcre_exec.c */

Properties

Name Value
svn:eol-style native
svn:keywords "Author Date Id Revision Url"

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12