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

Contents of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1376 - (show annotations)
Sat Oct 12 18:02:11 2013 UTC (6 years ago) by ph10
File MIME type: text/plain
File size: 217361 byte(s)
Error occurred while calculating annotation data.
Add U+0085 and U+180E to what \s matches in UCP mode, to match Perl.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2013 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40 /* This module contains pcre_exec(), the externally visible function that does
41 pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42 possible. There are also some static supporting functions. */
43
44 #ifdef HAVE_CONFIG_H
45 #include "config.h"
46 #endif
47
48 #define NLBLOCK md /* Block containing newline information */
49 #define PSSTART start_subject /* Field containing processed string start */
50 #define PSEND end_subject /* Field containing processed string end */
51
52 #include "pcre_internal.h"
53
54 /* Undefine some potentially clashing cpp symbols */
55
56 #undef min
57 #undef max
58
59 /* The md->capture_last field uses the lower 16 bits for the last captured
60 substring (which can never be greater than 65535) and a bit in the top half
61 to mean "capture vector overflowed". This odd way of doing things was
62 implemented when it was realized that preserving and restoring the overflow bit
63 whenever the last capture number was saved/restored made for a neater
64 interface, and doing it this way saved on (a) another variable, which would
65 have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66 separate set of save/restore instructions. The following defines are used in
67 implementing this. */
68
69 #define CAPLMASK 0x0000ffff /* The bits used for last_capture */
70 #define OVFLMASK 0xffff0000 /* The bits used for the overflow flag */
71 #define OVFLBIT 0x00010000 /* The bit that is set for overflow */
72
73 /* Values for setting in md->match_function_type to indicate two special types
74 of call to match(). We do it this way to save on using another stack variable,
75 as stack usage is to be discouraged. */
76
77 #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */
78 #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */
79
80 /* Non-error returns from the match() function. Error returns are externally
81 defined PCRE_ERROR_xxx codes, which are all negative. */
82
83 #define MATCH_MATCH 1
84 #define MATCH_NOMATCH 0
85
86 /* Special internal returns from the match() function. Make them sufficiently
87 negative to avoid the external error codes. */
88
89 #define MATCH_ACCEPT (-999)
90 #define MATCH_KETRPOS (-998)
91 #define MATCH_ONCE (-997)
92 /* The next 5 must be kept together and in sequence so that a test that checks
93 for any one of them can use a range. */
94 #define MATCH_COMMIT (-996)
95 #define MATCH_PRUNE (-995)
96 #define MATCH_SKIP (-994)
97 #define MATCH_SKIP_ARG (-993)
98 #define MATCH_THEN (-992)
99 #define MATCH_BACKTRACK_MAX MATCH_THEN
100 #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101
102 /* Maximum number of ints of offset to save on the stack for recursive calls.
103 If the offset vector is bigger, malloc is used. This should be a multiple of 3,
104 because the offset vector is always a multiple of 3 long. */
105
106 #define REC_STACK_SAVE_MAX 30
107
108 /* Min and max values for the common repeats; for the maxima, 0 => infinity */
109
110 static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
111 static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
112
113 #ifdef PCRE_DEBUG
114 /*************************************************
115 * Debugging function to print chars *
116 *************************************************/
117
118 /* Print a sequence of chars in printable format, stopping at the end of the
119 subject if the requested.
120
121 Arguments:
122 p points to characters
123 length number to print
124 is_subject TRUE if printing from within md->start_subject
125 md pointer to matching data block, if is_subject is TRUE
126
127 Returns: nothing
128 */
129
130 static void
131 pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132 {
133 pcre_uint32 c;
134 BOOL utf = md->utf;
135 if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136 while (length-- > 0)
137 if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138 }
139 #endif
140
141
142
143 /*************************************************
144 * Match a back-reference *
145 *************************************************/
146
147 /* Normally, if a back reference hasn't been set, the length that is passed is
148 negative, so the match always fails. However, in JavaScript compatibility mode,
149 the length passed is zero. Note that in caseless UTF-8 mode, the number of
150 subject bytes matched may be different to the number of reference bytes.
151
152 Arguments:
153 offset index into the offset vector
154 eptr pointer into the subject
155 length length of reference to be matched (number of bytes)
156 md points to match data block
157 caseless TRUE if caseless
158
159 Returns: >= 0 the number of subject bytes matched
160 -1 no match
161 -2 partial match; always given if at end subject
162 */
163
164 static int
165 match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166 BOOL caseless)
167 {
168 PCRE_PUCHAR eptr_start = eptr;
169 register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170 #ifdef SUPPORT_UTF
171 BOOL utf = md->utf;
172 #endif
173
174 #ifdef PCRE_DEBUG
175 if (eptr >= md->end_subject)
176 printf("matching subject <null>");
177 else
178 {
179 printf("matching subject ");
180 pchars(eptr, length, TRUE, md);
181 }
182 printf(" against backref ");
183 pchars(p, length, FALSE, md);
184 printf("\n");
185 #endif
186
187 /* Always fail if reference not set (and not JavaScript compatible - in that
188 case the length is passed as zero). */
189
190 if (length < 0) return -1;
191
192 /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193 properly if Unicode properties are supported. Otherwise, we can check only
194 ASCII characters. */
195
196 if (caseless)
197 {
198 #ifdef SUPPORT_UTF
199 #ifdef SUPPORT_UCP
200 if (utf)
201 {
202 /* Match characters up to the end of the reference. NOTE: the number of
203 data units matched may differ, because in UTF-8 there are some characters
204 whose upper and lower case versions code have different numbers of bytes.
205 For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206 (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207 sequence of two of the latter. It is important, therefore, to check the
208 length along the reference, not along the subject (earlier code did this
209 wrong). */
210
211 PCRE_PUCHAR endptr = p + length;
212 while (p < endptr)
213 {
214 pcre_uint32 c, d;
215 const ucd_record *ur;
216 if (eptr >= md->end_subject) return -2; /* Partial match */
217 GETCHARINC(c, eptr);
218 GETCHARINC(d, p);
219 ur = GET_UCD(d);
220 if (c != d && c != d + ur->other_case)
221 {
222 const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223 for (;;)
224 {
225 if (c < *pp) return -1;
226 if (c == *pp++) break;
227 }
228 }
229 }
230 }
231 else
232 #endif
233 #endif
234
235 /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236 is no UCP support. */
237 {
238 while (length-- > 0)
239 {
240 pcre_uint32 cc, cp;
241 if (eptr >= md->end_subject) return -2; /* Partial match */
242 cc = RAWUCHARTEST(eptr);
243 cp = RAWUCHARTEST(p);
244 if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245 p++;
246 eptr++;
247 }
248 }
249 }
250
251 /* In the caseful case, we can just compare the bytes, whether or not we
252 are in UTF-8 mode. */
253
254 else
255 {
256 while (length-- > 0)
257 {
258 if (eptr >= md->end_subject) return -2; /* Partial match */
259 if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260 }
261 }
262
263 return (int)(eptr - eptr_start);
264 }
265
266
267
268 /***************************************************************************
269 ****************************************************************************
270 RECURSION IN THE match() FUNCTION
271
272 The match() function is highly recursive, though not every recursive call
273 increases the recursive depth. Nevertheless, some regular expressions can cause
274 it to recurse to a great depth. I was writing for Unix, so I just let it call
275 itself recursively. This uses the stack for saving everything that has to be
276 saved for a recursive call. On Unix, the stack can be large, and this works
277 fine.
278
279 It turns out that on some non-Unix-like systems there are problems with
280 programs that use a lot of stack. (This despite the fact that every last chip
281 has oodles of memory these days, and techniques for extending the stack have
282 been known for decades.) So....
283
284 There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
285 calls by keeping local variables that need to be preserved in blocks of memory
286 obtained from malloc() instead instead of on the stack. Macros are used to
287 achieve this so that the actual code doesn't look very different to what it
288 always used to.
289
290 The original heap-recursive code used longjmp(). However, it seems that this
291 can be very slow on some operating systems. Following a suggestion from Stan
292 Switzer, the use of longjmp() has been abolished, at the cost of having to
293 provide a unique number for each call to RMATCH. There is no way of generating
294 a sequence of numbers at compile time in C. I have given them names, to make
295 them stand out more clearly.
296
297 Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
298 FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
299 tests. Furthermore, not using longjmp() means that local dynamic variables
300 don't have indeterminate values; this has meant that the frame size can be
301 reduced because the result can be "passed back" by straight setting of the
302 variable instead of being passed in the frame.
303 ****************************************************************************
304 ***************************************************************************/
305
306 /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307 below must be updated in sync. */
308
309 enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10,
310 RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311 RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312 RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313 RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314 RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315 RM61, RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316
317 /* These versions of the macros use the stack, as normal. There are debugging
318 versions and production versions. Note that the "rw" argument of RMATCH isn't
319 actually used in this definition. */
320
321 #ifndef NO_RECURSE
322 #define REGISTER register
323
324 #ifdef PCRE_DEBUG
325 #define RMATCH(ra,rb,rc,rd,re,rw) \
326 { \
327 printf("match() called in line %d\n", __LINE__); \
328 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329 printf("to line %d\n", __LINE__); \
330 }
331 #define RRETURN(ra) \
332 { \
333 printf("match() returned %d from line %d\n", ra, __LINE__); \
334 return ra; \
335 }
336 #else
337 #define RMATCH(ra,rb,rc,rd,re,rw) \
338 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339 #define RRETURN(ra) return ra
340 #endif
341
342 #else
343
344
345 /* These versions of the macros manage a private stack on the heap. Note that
346 the "rd" argument of RMATCH isn't actually used in this definition. It's the md
347 argument of match(), which never changes. */
348
349 #define REGISTER
350
351 #define RMATCH(ra,rb,rc,rd,re,rw)\
352 {\
353 heapframe *newframe = frame->Xnextframe;\
354 if (newframe == NULL)\
355 {\
356 newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357 if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358 newframe->Xnextframe = NULL;\
359 frame->Xnextframe = newframe;\
360 }\
361 frame->Xwhere = rw;\
362 newframe->Xeptr = ra;\
363 newframe->Xecode = rb;\
364 newframe->Xmstart = mstart;\
365 newframe->Xoffset_top = rc;\
366 newframe->Xeptrb = re;\
367 newframe->Xrdepth = frame->Xrdepth + 1;\
368 newframe->Xprevframe = frame;\
369 frame = newframe;\
370 DPRINTF(("restarting from line %d\n", __LINE__));\
371 goto HEAP_RECURSE;\
372 L_##rw:\
373 DPRINTF(("jumped back to line %d\n", __LINE__));\
374 }
375
376 #define RRETURN(ra)\
377 {\
378 heapframe *oldframe = frame;\
379 frame = oldframe->Xprevframe;\
380 if (frame != NULL)\
381 {\
382 rrc = ra;\
383 goto HEAP_RETURN;\
384 }\
385 return ra;\
386 }
387
388
389 /* Structure for remembering the local variables in a private frame */
390
391 typedef struct heapframe {
392 struct heapframe *Xprevframe;
393 struct heapframe *Xnextframe;
394
395 /* Function arguments that may change */
396
397 PCRE_PUCHAR Xeptr;
398 const pcre_uchar *Xecode;
399 PCRE_PUCHAR Xmstart;
400 int Xoffset_top;
401 eptrblock *Xeptrb;
402 unsigned int Xrdepth;
403
404 /* Function local variables */
405
406 PCRE_PUCHAR Xcallpat;
407 #ifdef SUPPORT_UTF
408 PCRE_PUCHAR Xcharptr;
409 #endif
410 PCRE_PUCHAR Xdata;
411 PCRE_PUCHAR Xnext;
412 PCRE_PUCHAR Xpp;
413 PCRE_PUCHAR Xprev;
414 PCRE_PUCHAR Xsaved_eptr;
415
416 recursion_info Xnew_recursive;
417
418 BOOL Xcur_is_word;
419 BOOL Xcondition;
420 BOOL Xprev_is_word;
421
422 #ifdef SUPPORT_UCP
423 int Xprop_type;
424 unsigned int Xprop_value;
425 int Xprop_fail_result;
426 int Xoclength;
427 pcre_uchar Xocchars[6];
428 #endif
429
430 int Xcodelink;
431 int Xctype;
432 unsigned int Xfc;
433 int Xfi;
434 int Xlength;
435 int Xmax;
436 int Xmin;
437 unsigned int Xnumber;
438 int Xoffset;
439 unsigned int Xop;
440 pcre_int32 Xsave_capture_last;
441 int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442 int Xstacksave[REC_STACK_SAVE_MAX];
443
444 eptrblock Xnewptrb;
445
446 /* Where to jump back to */
447
448 int Xwhere;
449
450 } heapframe;
451
452 #endif
453
454
455 /***************************************************************************
456 ***************************************************************************/
457
458
459
460 /*************************************************
461 * Match from current position *
462 *************************************************/
463
464 /* This function is called recursively in many circumstances. Whenever it
465 returns a negative (error) response, the outer incarnation must also return the
466 same response. */
467
468 /* These macros pack up tests that are used for partial matching, and which
469 appear several times in the code. We set the "hit end" flag if the pointer is
470 at the end of the subject and also past the start of the subject (i.e.
471 something has been matched). For hard partial matching, we then return
472 immediately. The second one is used when we already know we are past the end of
473 the subject. */
474
475 #define CHECK_PARTIAL()\
476 if (md->partial != 0 && eptr >= md->end_subject && \
477 eptr > md->start_used_ptr) \
478 { \
479 md->hitend = TRUE; \
480 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481 }
482
483 #define SCHECK_PARTIAL()\
484 if (md->partial != 0 && eptr > md->start_used_ptr) \
485 { \
486 md->hitend = TRUE; \
487 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488 }
489
490
491 /* Performance note: It might be tempting to extract commonly used fields from
492 the md structure (e.g. utf, end_subject) into individual variables to improve
493 performance. Tests using gcc on a SPARC disproved this; in the first case, it
494 made performance worse.
495
496 Arguments:
497 eptr pointer to current character in subject
498 ecode pointer to current position in compiled code
499 mstart pointer to the current match start position (can be modified
500 by encountering \K)
501 offset_top current top pointer
502 md pointer to "static" info for the match
503 eptrb pointer to chain of blocks containing eptr at start of
504 brackets - for testing for empty matches
505 rdepth the recursion depth
506
507 Returns: MATCH_MATCH if matched ) these values are >= 0
508 MATCH_NOMATCH if failed to match )
509 a negative MATCH_xxx value for PRUNE, SKIP, etc
510 a negative PCRE_ERROR_xxx value if aborted by an error condition
511 (e.g. stopped by repeated call or recursion limit)
512 */
513
514 static int
515 match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516 PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517 unsigned int rdepth)
518 {
519 /* These variables do not need to be preserved over recursion in this function,
520 so they can be ordinary variables in all cases. Mark some of them with
521 "register" because they are used a lot in loops. */
522
523 register int rrc; /* Returns from recursive calls */
524 register int i; /* Used for loops not involving calls to RMATCH() */
525 register pcre_uint32 c; /* Character values not kept over RMATCH() calls */
526 register BOOL utf; /* Local copy of UTF flag for speed */
527
528 BOOL minimize, possessive; /* Quantifier options */
529 BOOL caseless;
530 int condcode;
531
532 /* When recursion is not being used, all "local" variables that have to be
533 preserved over calls to RMATCH() are part of a "frame". We set up the top-level
534 frame on the stack here; subsequent instantiations are obtained from the heap
535 whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536 the top-level on the stack rather than malloc-ing them all gives a performance
537 boost in many cases where there is not much "recursion". */
538
539 #ifdef NO_RECURSE
540 heapframe *frame = (heapframe *)md->match_frames_base;
541
542 /* Copy in the original argument variables */
543
544 frame->Xeptr = eptr;
545 frame->Xecode = ecode;
546 frame->Xmstart = mstart;
547 frame->Xoffset_top = offset_top;
548 frame->Xeptrb = eptrb;
549 frame->Xrdepth = rdepth;
550
551 /* This is where control jumps back to to effect "recursion" */
552
553 HEAP_RECURSE:
554
555 /* Macros make the argument variables come from the current frame */
556
557 #define eptr frame->Xeptr
558 #define ecode frame->Xecode
559 #define mstart frame->Xmstart
560 #define offset_top frame->Xoffset_top
561 #define eptrb frame->Xeptrb
562 #define rdepth frame->Xrdepth
563
564 /* Ditto for the local variables */
565
566 #ifdef SUPPORT_UTF
567 #define charptr frame->Xcharptr
568 #endif
569 #define callpat frame->Xcallpat
570 #define codelink frame->Xcodelink
571 #define data frame->Xdata
572 #define next frame->Xnext
573 #define pp frame->Xpp
574 #define prev frame->Xprev
575 #define saved_eptr frame->Xsaved_eptr
576
577 #define new_recursive frame->Xnew_recursive
578
579 #define cur_is_word frame->Xcur_is_word
580 #define condition frame->Xcondition
581 #define prev_is_word frame->Xprev_is_word
582
583 #ifdef SUPPORT_UCP
584 #define prop_type frame->Xprop_type
585 #define prop_value frame->Xprop_value
586 #define prop_fail_result frame->Xprop_fail_result
587 #define oclength frame->Xoclength
588 #define occhars frame->Xocchars
589 #endif
590
591 #define ctype frame->Xctype
592 #define fc frame->Xfc
593 #define fi frame->Xfi
594 #define length frame->Xlength
595 #define max frame->Xmax
596 #define min frame->Xmin
597 #define number frame->Xnumber
598 #define offset frame->Xoffset
599 #define op frame->Xop
600 #define save_capture_last frame->Xsave_capture_last
601 #define save_offset1 frame->Xsave_offset1
602 #define save_offset2 frame->Xsave_offset2
603 #define save_offset3 frame->Xsave_offset3
604 #define stacksave frame->Xstacksave
605
606 #define newptrb frame->Xnewptrb
607
608 /* When recursion is being used, local variables are allocated on the stack and
609 get preserved during recursion in the normal way. In this environment, fi and
610 i, and fc and c, can be the same variables. */
611
612 #else /* NO_RECURSE not defined */
613 #define fi i
614 #define fc c
615
616 /* Many of the following variables are used only in small blocks of the code.
617 My normal style of coding would have declared them within each of those blocks.
618 However, in order to accommodate the version of this code that uses an external
619 "stack" implemented on the heap, it is easier to declare them all here, so the
620 declarations can be cut out in a block. The only declarations within blocks
621 below are for variables that do not have to be preserved over a recursive call
622 to RMATCH(). */
623
624 #ifdef SUPPORT_UTF
625 const pcre_uchar *charptr;
626 #endif
627 const pcre_uchar *callpat;
628 const pcre_uchar *data;
629 const pcre_uchar *next;
630 PCRE_PUCHAR pp;
631 const pcre_uchar *prev;
632 PCRE_PUCHAR saved_eptr;
633
634 recursion_info new_recursive;
635
636 BOOL cur_is_word;
637 BOOL condition;
638 BOOL prev_is_word;
639
640 #ifdef SUPPORT_UCP
641 int prop_type;
642 unsigned int prop_value;
643 int prop_fail_result;
644 int oclength;
645 pcre_uchar occhars[6];
646 #endif
647
648 int codelink;
649 int ctype;
650 int length;
651 int max;
652 int min;
653 unsigned int number;
654 int offset;
655 unsigned int op;
656 pcre_int32 save_capture_last;
657 int save_offset1, save_offset2, save_offset3;
658 int stacksave[REC_STACK_SAVE_MAX];
659
660 eptrblock newptrb;
661
662 /* There is a special fudge for calling match() in a way that causes it to
663 measure the size of its basic stack frame when the stack is being used for
664 recursion. The second argument (ecode) being NULL triggers this behaviour. It
665 cannot normally ever be NULL. The return is the negated value of the frame
666 size. */
667
668 if (ecode == NULL)
669 {
670 if (rdepth == 0)
671 return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672 else
673 {
674 int len = (char *)&rdepth - (char *)eptr;
675 return (len > 0)? -len : len;
676 }
677 }
678 #endif /* NO_RECURSE */
679
680 /* To save space on the stack and in the heap frame, I have doubled up on some
681 of the local variables that are used only in localised parts of the code, but
682 still need to be preserved over recursive calls of match(). These macros define
683 the alternative names that are used. */
684
685 #define allow_zero cur_is_word
686 #define cbegroup condition
687 #define code_offset codelink
688 #define condassert condition
689 #define matched_once prev_is_word
690 #define foc number
691 #define save_mark data
692
693 /* These statements are here to stop the compiler complaining about unitialized
694 variables. */
695
696 #ifdef SUPPORT_UCP
697 prop_value = 0;
698 prop_fail_result = 0;
699 #endif
700
701
702 /* This label is used for tail recursion, which is used in a few cases even
703 when NO_RECURSE is not defined, in order to reduce the amount of stack that is
704 used. Thanks to Ian Taylor for noticing this possibility and sending the
705 original patch. */
706
707 TAIL_RECURSE:
708
709 /* OK, now we can get on with the real code of the function. Recursive calls
710 are specified by the macro RMATCH and RRETURN is used to return. When
711 NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712 and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713 defined). However, RMATCH isn't like a function call because it's quite a
714 complicated macro. It has to be used in one particular way. This shouldn't,
715 however, impact performance when true recursion is being used. */
716
717 #ifdef SUPPORT_UTF
718 utf = md->utf; /* Local copy of the flag */
719 #else
720 utf = FALSE;
721 #endif
722
723 /* First check that we haven't called match() too many times, or that we
724 haven't exceeded the recursive call limit. */
725
726 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728
729 /* At the start of a group with an unlimited repeat that may match an empty
730 string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731 done this way to save having to use another function argument, which would take
732 up space on the stack. See also MATCH_CONDASSERT below.
733
734 When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735 such remembered pointers, to be checked when we hit the closing ket, in order
736 to break infinite loops that match no characters. When match() is called in
737 other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738 NOT be used with tail recursion, because the memory block that is used is on
739 the stack, so a new one may be required for each match(). */
740
741 if (md->match_function_type == MATCH_CBEGROUP)
742 {
743 newptrb.epb_saved_eptr = eptr;
744 newptrb.epb_prev = eptrb;
745 eptrb = &newptrb;
746 md->match_function_type = 0;
747 }
748
749 /* Now start processing the opcodes. */
750
751 for (;;)
752 {
753 minimize = possessive = FALSE;
754 op = *ecode;
755
756 switch(op)
757 {
758 case OP_MARK:
759 md->nomatch_mark = ecode + 2;
760 md->mark = NULL; /* In case previously set by assertion */
761 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762 eptrb, RM55);
763 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764 md->mark == NULL) md->mark = ecode + 2;
765
766 /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767 argument, and we must check whether that argument matches this MARK's
768 argument. It is passed back in md->start_match_ptr (an overloading of that
769 variable). If it does match, we reset that variable to the current subject
770 position and return MATCH_SKIP. Otherwise, pass back the return code
771 unaltered. */
772
773 else if (rrc == MATCH_SKIP_ARG &&
774 STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775 {
776 md->start_match_ptr = eptr;
777 RRETURN(MATCH_SKIP);
778 }
779 RRETURN(rrc);
780
781 case OP_FAIL:
782 RRETURN(MATCH_NOMATCH);
783
784 case OP_COMMIT:
785 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786 eptrb, RM52);
787 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788 RRETURN(MATCH_COMMIT);
789
790 case OP_PRUNE:
791 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792 eptrb, RM51);
793 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794 RRETURN(MATCH_PRUNE);
795
796 case OP_PRUNE_ARG:
797 md->nomatch_mark = ecode + 2;
798 md->mark = NULL; /* In case previously set by assertion */
799 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800 eptrb, RM56);
801 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
802 md->mark == NULL) md->mark = ecode + 2;
803 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
804 RRETURN(MATCH_PRUNE);
805
806 case OP_SKIP:
807 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
808 eptrb, RM53);
809 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
810 md->start_match_ptr = eptr; /* Pass back current position */
811 RRETURN(MATCH_SKIP);
812
813 /* Note that, for Perl compatibility, SKIP with an argument does NOT set
814 nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
815 not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
816 that failed and any that precede it (either they also failed, or were not
817 triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
818 SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
819 set to the count of the one that failed. */
820
821 case OP_SKIP_ARG:
822 md->skip_arg_count++;
823 if (md->skip_arg_count <= md->ignore_skip_arg)
824 {
825 ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
826 break;
827 }
828 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
829 eptrb, RM57);
830 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831
832 /* Pass back the current skip name by overloading md->start_match_ptr and
833 returning the special MATCH_SKIP_ARG return code. This will either be
834 caught by a matching MARK, or get to the top, where it causes a rematch
835 with md->ignore_skip_arg set to the value of md->skip_arg_count. */
836
837 md->start_match_ptr = ecode + 2;
838 RRETURN(MATCH_SKIP_ARG);
839
840 /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
841 the branch in which it occurs can be determined. Overload the start of
842 match pointer to do this. */
843
844 case OP_THEN:
845 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
846 eptrb, RM54);
847 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
848 md->start_match_ptr = ecode;
849 RRETURN(MATCH_THEN);
850
851 case OP_THEN_ARG:
852 md->nomatch_mark = ecode + 2;
853 md->mark = NULL; /* In case previously set by assertion */
854 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
855 md, eptrb, RM58);
856 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
857 md->mark == NULL) md->mark = ecode + 2;
858 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859 md->start_match_ptr = ecode;
860 RRETURN(MATCH_THEN);
861
862 /* Handle an atomic group that does not contain any capturing parentheses.
863 This can be handled like an assertion. Prior to 8.13, all atomic groups
864 were handled this way. In 8.13, the code was changed as below for ONCE, so
865 that backups pass through the group and thereby reset captured values.
866 However, this uses a lot more stack, so in 8.20, atomic groups that do not
867 contain any captures generate OP_ONCE_NC, which can be handled in the old,
868 less stack intensive way.
869
870 Check the alternative branches in turn - the matching won't pass the KET
871 for this kind of subpattern. If any one branch matches, we carry on as at
872 the end of a normal bracket, leaving the subject pointer, but resetting
873 the start-of-match value in case it was changed by \K. */
874
875 case OP_ONCE_NC:
876 prev = ecode;
877 saved_eptr = eptr;
878 save_mark = md->mark;
879 do
880 {
881 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
882 if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */
883 {
884 mstart = md->start_match_ptr;
885 break;
886 }
887 if (rrc == MATCH_THEN)
888 {
889 next = ecode + GET(ecode,1);
890 if (md->start_match_ptr < next &&
891 (*ecode == OP_ALT || *next == OP_ALT))
892 rrc = MATCH_NOMATCH;
893 }
894
895 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
896 ecode += GET(ecode,1);
897 md->mark = save_mark;
898 }
899 while (*ecode == OP_ALT);
900
901 /* If hit the end of the group (which could be repeated), fail */
902
903 if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
904
905 /* Continue as from after the group, updating the offsets high water
906 mark, since extracts may have been taken. */
907
908 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
909
910 offset_top = md->end_offset_top;
911 eptr = md->end_match_ptr;
912
913 /* For a non-repeating ket, just continue at this level. This also
914 happens for a repeating ket if no characters were matched in the group.
915 This is the forcible breaking of infinite loops as implemented in Perl
916 5.005. */
917
918 if (*ecode == OP_KET || eptr == saved_eptr)
919 {
920 ecode += 1+LINK_SIZE;
921 break;
922 }
923
924 /* The repeating kets try the rest of the pattern or restart from the
925 preceding bracket, in the appropriate order. The second "call" of match()
926 uses tail recursion, to avoid using another stack frame. */
927
928 if (*ecode == OP_KETRMIN)
929 {
930 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
931 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
932 ecode = prev;
933 goto TAIL_RECURSE;
934 }
935 else /* OP_KETRMAX */
936 {
937 RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
938 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
939 ecode += 1 + LINK_SIZE;
940 goto TAIL_RECURSE;
941 }
942 /* Control never gets here */
943
944 /* Handle a capturing bracket, other than those that are possessive with an
945 unlimited repeat. If there is space in the offset vector, save the current
946 subject position in the working slot at the top of the vector. We mustn't
947 change the current values of the data slot, because they may be set from a
948 previous iteration of this group, and be referred to by a reference inside
949 the group. A failure to match might occur after the group has succeeded,
950 if something later on doesn't match. For this reason, we need to restore
951 the working value and also the values of the final offsets, in case they
952 were set by a previous iteration of the same bracket.
953
954 If there isn't enough space in the offset vector, treat this as if it were
955 a non-capturing bracket. Don't worry about setting the flag for the error
956 case here; that is handled in the code for KET. */
957
958 case OP_CBRA:
959 case OP_SCBRA:
960 number = GET2(ecode, 1+LINK_SIZE);
961 offset = number << 1;
962
963 #ifdef PCRE_DEBUG
964 printf("start bracket %d\n", number);
965 printf("subject=");
966 pchars(eptr, 16, TRUE, md);
967 printf("\n");
968 #endif
969
970 if (offset < md->offset_max)
971 {
972 save_offset1 = md->offset_vector[offset];
973 save_offset2 = md->offset_vector[offset+1];
974 save_offset3 = md->offset_vector[md->offset_end - number];
975 save_capture_last = md->capture_last;
976 save_mark = md->mark;
977
978 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
979 md->offset_vector[md->offset_end - number] =
980 (int)(eptr - md->start_subject);
981
982 for (;;)
983 {
984 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
986 eptrb, RM1);
987 if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */
988
989 /* If we backed up to a THEN, check whether it is within the current
990 branch by comparing the address of the THEN that is passed back with
991 the end of the branch. If it is within the current branch, and the
992 branch is one of two or more alternatives (it either starts or ends
993 with OP_ALT), we have reached the limit of THEN's action, so convert
994 the return code to NOMATCH, which will cause normal backtracking to
995 happen from now on. Otherwise, THEN is passed back to an outer
996 alternative. This implements Perl's treatment of parenthesized groups,
997 where a group not containing | does not affect the current alternative,
998 that is, (X) is NOT the same as (X|(*F)). */
999
1000 if (rrc == MATCH_THEN)
1001 {
1002 next = ecode + GET(ecode,1);
1003 if (md->start_match_ptr < next &&
1004 (*ecode == OP_ALT || *next == OP_ALT))
1005 rrc = MATCH_NOMATCH;
1006 }
1007
1008 /* Anything other than NOMATCH is passed back. */
1009
1010 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1011 md->capture_last = save_capture_last;
1012 ecode += GET(ecode, 1);
1013 md->mark = save_mark;
1014 if (*ecode != OP_ALT) break;
1015 }
1016
1017 DPRINTF(("bracket %d failed\n", number));
1018 md->offset_vector[offset] = save_offset1;
1019 md->offset_vector[offset+1] = save_offset2;
1020 md->offset_vector[md->offset_end - number] = save_offset3;
1021
1022 /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1023
1024 RRETURN(rrc);
1025 }
1026
1027 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1028 as a non-capturing bracket. */
1029
1030 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1031 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1032
1033 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1034
1035 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037
1038 /* Non-capturing or atomic group, except for possessive with unlimited
1039 repeat and ONCE group with no captures. Loop for all the alternatives.
1040
1041 When we get to the final alternative within the brackets, we used to return
1042 the result of a recursive call to match() whatever happened so it was
1043 possible to reduce stack usage by turning this into a tail recursion,
1044 except in the case of a possibly empty group. However, now that there is
1045 the possiblity of (*THEN) occurring in the final alternative, this
1046 optimization is no longer always possible.
1047
1048 We can optimize if we know there are no (*THEN)s in the pattern; at present
1049 this is the best that can be done.
1050
1051 MATCH_ONCE is returned when the end of an atomic group is successfully
1052 reached, but subsequent matching fails. It passes back up the tree (causing
1053 captured values to be reset) until the original atomic group level is
1054 reached. This is tested by comparing md->once_target with the start of the
1055 group. At this point, the return is converted into MATCH_NOMATCH so that
1056 previous backup points can be taken. */
1057
1058 case OP_ONCE:
1059 case OP_BRA:
1060 case OP_SBRA:
1061 DPRINTF(("start non-capturing bracket\n"));
1062
1063 for (;;)
1064 {
1065 if (op >= OP_SBRA || op == OP_ONCE)
1066 md->match_function_type = MATCH_CBEGROUP;
1067
1068 /* If this is not a possibly empty group, and there are no (*THEN)s in
1069 the pattern, and this is the final alternative, optimize as described
1070 above. */
1071
1072 else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073 {
1074 ecode += PRIV(OP_lengths)[*ecode];
1075 goto TAIL_RECURSE;
1076 }
1077
1078 /* In all other cases, we have to make another call to match(). */
1079
1080 save_mark = md->mark;
1081 save_capture_last = md->capture_last;
1082 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1083 RM2);
1084
1085 /* See comment in the code for capturing groups above about handling
1086 THEN. */
1087
1088 if (rrc == MATCH_THEN)
1089 {
1090 next = ecode + GET(ecode,1);
1091 if (md->start_match_ptr < next &&
1092 (*ecode == OP_ALT || *next == OP_ALT))
1093 rrc = MATCH_NOMATCH;
1094 }
1095
1096 if (rrc != MATCH_NOMATCH)
1097 {
1098 if (rrc == MATCH_ONCE)
1099 {
1100 const pcre_uchar *scode = ecode;
1101 if (*scode != OP_ONCE) /* If not at start, find it */
1102 {
1103 while (*scode == OP_ALT) scode += GET(scode, 1);
1104 scode -= GET(scode, 1);
1105 }
1106 if (md->once_target == scode) rrc = MATCH_NOMATCH;
1107 }
1108 RRETURN(rrc);
1109 }
1110 ecode += GET(ecode, 1);
1111 md->mark = save_mark;
1112 if (*ecode != OP_ALT) break;
1113 md->capture_last = save_capture_last;
1114 }
1115
1116 RRETURN(MATCH_NOMATCH);
1117
1118 /* Handle possessive capturing brackets with an unlimited repeat. We come
1119 here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120 handled similarly to the normal case above. However, the matching is
1121 different. The end of these brackets will always be OP_KETRPOS, which
1122 returns MATCH_KETRPOS without going further in the pattern. By this means
1123 we can handle the group by iteration rather than recursion, thereby
1124 reducing the amount of stack needed. */
1125
1126 case OP_CBRAPOS:
1127 case OP_SCBRAPOS:
1128 allow_zero = FALSE;
1129
1130 POSSESSIVE_CAPTURE:
1131 number = GET2(ecode, 1+LINK_SIZE);
1132 offset = number << 1;
1133
1134 #ifdef PCRE_DEBUG
1135 printf("start possessive bracket %d\n", number);
1136 printf("subject=");
1137 pchars(eptr, 16, TRUE, md);
1138 printf("\n");
1139 #endif
1140
1141 if (offset < md->offset_max)
1142 {
1143 matched_once = FALSE;
1144 code_offset = (int)(ecode - md->start_code);
1145
1146 save_offset1 = md->offset_vector[offset];
1147 save_offset2 = md->offset_vector[offset+1];
1148 save_offset3 = md->offset_vector[md->offset_end - number];
1149 save_capture_last = md->capture_last;
1150
1151 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1152
1153 /* Each time round the loop, save the current subject position for use
1154 when the group matches. For MATCH_MATCH, the group has matched, so we
1155 restart it with a new subject starting position, remembering that we had
1156 at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157 usual. If we haven't matched any alternatives in any iteration, check to
1158 see if a previous iteration matched. If so, the group has matched;
1159 continue from afterwards. Otherwise it has failed; restore the previous
1160 capture values before returning NOMATCH. */
1161
1162 for (;;)
1163 {
1164 md->offset_vector[md->offset_end - number] =
1165 (int)(eptr - md->start_subject);
1166 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168 eptrb, RM63);
1169 if (rrc == MATCH_KETRPOS)
1170 {
1171 offset_top = md->end_offset_top;
1172 eptr = md->end_match_ptr;
1173 ecode = md->start_code + code_offset;
1174 save_capture_last = md->capture_last;
1175 matched_once = TRUE;
1176 continue;
1177 }
1178
1179 /* See comment in the code for capturing groups above about handling
1180 THEN. */
1181
1182 if (rrc == MATCH_THEN)
1183 {
1184 next = ecode + GET(ecode,1);
1185 if (md->start_match_ptr < next &&
1186 (*ecode == OP_ALT || *next == OP_ALT))
1187 rrc = MATCH_NOMATCH;
1188 }
1189
1190 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1191 md->capture_last = save_capture_last;
1192 ecode += GET(ecode, 1);
1193 if (*ecode != OP_ALT) break;
1194 }
1195
1196 if (!matched_once)
1197 {
1198 md->offset_vector[offset] = save_offset1;
1199 md->offset_vector[offset+1] = save_offset2;
1200 md->offset_vector[md->offset_end - number] = save_offset3;
1201 }
1202
1203 if (allow_zero || matched_once)
1204 {
1205 ecode += 1 + LINK_SIZE;
1206 break;
1207 }
1208
1209 RRETURN(MATCH_NOMATCH);
1210 }
1211
1212 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213 as a non-capturing bracket. */
1214
1215 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1217
1218 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1219
1220 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1222
1223 /* Non-capturing possessive bracket with unlimited repeat. We come here
1224 from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1225 without the capturing complication. It is written out separately for speed
1226 and cleanliness. */
1227
1228 case OP_BRAPOS:
1229 case OP_SBRAPOS:
1230 allow_zero = FALSE;
1231
1232 POSSESSIVE_NON_CAPTURE:
1233 matched_once = FALSE;
1234 code_offset = (int)(ecode - md->start_code);
1235 save_capture_last = md->capture_last;
1236
1237 for (;;)
1238 {
1239 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1241 eptrb, RM48);
1242 if (rrc == MATCH_KETRPOS)
1243 {
1244 offset_top = md->end_offset_top;
1245 eptr = md->end_match_ptr;
1246 ecode = md->start_code + code_offset;
1247 matched_once = TRUE;
1248 continue;
1249 }
1250
1251 /* See comment in the code for capturing groups above about handling
1252 THEN. */
1253
1254 if (rrc == MATCH_THEN)
1255 {
1256 next = ecode + GET(ecode,1);
1257 if (md->start_match_ptr < next &&
1258 (*ecode == OP_ALT || *next == OP_ALT))
1259 rrc = MATCH_NOMATCH;
1260 }
1261
1262 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1263 ecode += GET(ecode, 1);
1264 if (*ecode != OP_ALT) break;
1265 md->capture_last = save_capture_last;
1266 }
1267
1268 if (matched_once || allow_zero)
1269 {
1270 ecode += 1 + LINK_SIZE;
1271 break;
1272 }
1273 RRETURN(MATCH_NOMATCH);
1274
1275 /* Control never reaches here. */
1276
1277 /* Conditional group: compilation checked that there are no more than two
1278 branches. If the condition is false, skipping the first branch takes us
1279 past the end of the item if there is only one branch, but that's exactly
1280 what we want. */
1281
1282 case OP_COND:
1283 case OP_SCOND:
1284
1285 /* The variable codelink will be added to ecode when the condition is
1286 false, to get to the second branch. Setting it to the offset to the ALT
1287 or KET, then incrementing ecode achieves this effect. We now have ecode
1288 pointing to the condition or callout. */
1289
1290 codelink = GET(ecode, 1); /* Offset to the second branch */
1291 ecode += 1 + LINK_SIZE; /* From this opcode */
1292
1293 /* Because of the way auto-callout works during compile, a callout item is
1294 inserted between OP_COND and an assertion condition. */
1295
1296 if (*ecode == OP_CALLOUT)
1297 {
1298 if (PUBL(callout) != NULL)
1299 {
1300 PUBL(callout_block) cb;
1301 cb.version = 2; /* Version 1 of the callout block */
1302 cb.callout_number = ecode[1];
1303 cb.offset_vector = md->offset_vector;
1304 #if defined COMPILE_PCRE8
1305 cb.subject = (PCRE_SPTR)md->start_subject;
1306 #elif defined COMPILE_PCRE16
1307 cb.subject = (PCRE_SPTR16)md->start_subject;
1308 #elif defined COMPILE_PCRE32
1309 cb.subject = (PCRE_SPTR32)md->start_subject;
1310 #endif
1311 cb.subject_length = (int)(md->end_subject - md->start_subject);
1312 cb.start_match = (int)(mstart - md->start_subject);
1313 cb.current_position = (int)(eptr - md->start_subject);
1314 cb.pattern_position = GET(ecode, 2);
1315 cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1316 cb.capture_top = offset_top/2;
1317 cb.capture_last = md->capture_last & CAPLMASK;
1318 /* Internal change requires this for API compatibility. */
1319 if (cb.capture_last == 0) cb.capture_last = -1;
1320 cb.callout_data = md->callout_data;
1321 cb.mark = md->nomatch_mark;
1322 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1323 if (rrc < 0) RRETURN(rrc);
1324 }
1325
1326 /* Advance ecode past the callout, so it now points to the condition. We
1327 must adjust codelink so that the value of ecode+codelink is unchanged. */
1328
1329 ecode += PRIV(OP_lengths)[OP_CALLOUT];
1330 codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1331 }
1332
1333 /* Test the various possible conditions */
1334
1335 condition = FALSE;
1336 switch(condcode = *ecode)
1337 {
1338 case OP_RREF: /* Numbered group recursion test */
1339 if (md->recursive != NULL) /* Not recursing => FALSE */
1340 {
1341 unsigned int recno = GET2(ecode, 1); /* Recursion group number*/
1342 condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1343 }
1344 break;
1345
1346 case OP_DNRREF: /* Duplicate named group recursion test */
1347 if (md->recursive != NULL)
1348 {
1349 int count = GET2(ecode, 1 + IMM2_SIZE);
1350 pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1351 while (count-- > 0)
1352 {
1353 unsigned int recno = GET2(slot, 0);
1354 condition = recno == md->recursive->group_num;
1355 if (condition) break;
1356 slot += md->name_entry_size;
1357 }
1358 }
1359 break;
1360
1361 case OP_CREF: /* Numbered group used test */
1362 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
1363 condition = offset < offset_top && md->offset_vector[offset] >= 0;
1364 break;
1365
1366 case OP_DNCREF: /* Duplicate named group used test */
1367 {
1368 int count = GET2(ecode, 1 + IMM2_SIZE);
1369 pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1370 while (count-- > 0)
1371 {
1372 offset = GET2(slot, 0) << 1;
1373 condition = offset < offset_top && md->offset_vector[offset] >= 0;
1374 if (condition) break;
1375 slot += md->name_entry_size;
1376 }
1377 }
1378 break;
1379
1380 case OP_DEF: /* DEFINE - always false */
1381 break;
1382
1383 /* The condition is an assertion. Call match() to evaluate it - setting
1384 md->match_function_type to MATCH_CONDASSERT causes it to stop at the end
1385 of an assertion. */
1386
1387 default:
1388 md->match_function_type = MATCH_CONDASSERT;
1389 RMATCH(eptr, ecode, offset_top, md, NULL, RM3);
1390 if (rrc == MATCH_MATCH)
1391 {
1392 if (md->end_offset_top > offset_top)
1393 offset_top = md->end_offset_top; /* Captures may have happened */
1394 condition = TRUE;
1395
1396 /* Advance ecode past the assertion to the start of the first branch,
1397 but adjust it so that the general choosing code below works. */
1398
1399 ecode += GET(ecode, 1);
1400 while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1401 ecode += 1 + LINK_SIZE - PRIV(OP_lengths)[condcode];
1402 }
1403
1404 /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1405 assertion; it is therefore treated as NOMATCH. Any other return is an
1406 error. */
1407
1408 else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1409 {
1410 RRETURN(rrc); /* Need braces because of following else */
1411 }
1412 break;
1413 }
1414
1415 /* Choose branch according to the condition */
1416
1417 ecode += condition? PRIV(OP_lengths)[condcode] : codelink;
1418
1419 /* We are now at the branch that is to be obeyed. As there is only one, we
1420 can use tail recursion to avoid using another stack frame, except when
1421 there is unlimited repeat of a possibly empty group. In the latter case, a
1422 recursive call to match() is always required, unless the second alternative
1423 doesn't exist, in which case we can just plough on. Note that, for
1424 compatibility with Perl, the | in a conditional group is NOT treated as
1425 creating two alternatives. If a THEN is encountered in the branch, it
1426 propagates out to the enclosing alternative (unless nested in a deeper set
1427 of alternatives, of course). */
1428
1429 if (condition || ecode[-(1+LINK_SIZE)] == OP_ALT)
1430 {
1431 if (op != OP_SCOND)
1432 {
1433 goto TAIL_RECURSE;
1434 }
1435
1436 md->match_function_type = MATCH_CBEGROUP;
1437 RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);
1438 RRETURN(rrc);
1439 }
1440
1441 /* Condition false & no alternative; continue after the group. */
1442
1443 else
1444 {
1445 }
1446 break;
1447
1448
1449 /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1450 to close any currently open capturing brackets. */
1451
1452 case OP_CLOSE:
1453 number = GET2(ecode, 1); /* Must be less than 65536 */
1454 offset = number << 1;
1455
1456 #ifdef PCRE_DEBUG
1457 printf("end bracket %d at *ACCEPT", number);
1458 printf("\n");
1459 #endif
1460
1461 md->capture_last = (md->capture_last & OVFLMASK) | number;
1462 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1463 {
1464 md->offset_vector[offset] =
1465 md->offset_vector[md->offset_end - number];
1466 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1467 if (offset_top <= offset) offset_top = offset + 2;
1468 }
1469 ecode += 1 + IMM2_SIZE;
1470 break;
1471
1472
1473 /* End of the pattern, either real or forced. */
1474
1475 case OP_END:
1476 case OP_ACCEPT:
1477 case OP_ASSERT_ACCEPT:
1478
1479 /* If we have matched an empty string, fail if not in an assertion and not
1480 in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1481 is set and we have matched at the start of the subject. In both cases,
1482 backtracking will then try other alternatives, if any. */
1483
1484 if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1485 md->recursive == NULL &&
1486 (md->notempty ||
1487 (md->notempty_atstart &&
1488 mstart == md->start_subject + md->start_offset)))
1489 RRETURN(MATCH_NOMATCH);
1490
1491 /* Otherwise, we have a match. */
1492
1493 md->end_match_ptr = eptr; /* Record where we ended */
1494 md->end_offset_top = offset_top; /* and how many extracts were taken */
1495 md->start_match_ptr = mstart; /* and the start (\K can modify) */
1496
1497 /* For some reason, the macros don't work properly if an expression is
1498 given as the argument to RRETURN when the heap is in use. */
1499
1500 rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1501 RRETURN(rrc);
1502
1503 /* Assertion brackets. Check the alternative branches in turn - the
1504 matching won't pass the KET for an assertion. If any one branch matches,
1505 the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1506 start of each branch to move the current point backwards, so the code at
1507 this level is identical to the lookahead case. When the assertion is part
1508 of a condition, we want to return immediately afterwards. The caller of
1509 this incarnation of the match() function will have set MATCH_CONDASSERT in
1510 md->match_function type, and one of these opcodes will be the first opcode
1511 that is processed. We use a local variable that is preserved over calls to
1512 match() to remember this case. */
1513
1514 case OP_ASSERT:
1515 case OP_ASSERTBACK:
1516 save_mark = md->mark;
1517 if (md->match_function_type == MATCH_CONDASSERT)
1518 {
1519 condassert = TRUE;
1520 md->match_function_type = 0;
1521 }
1522 else condassert = FALSE;
1523
1524 /* Loop for each branch */
1525
1526 do
1527 {
1528 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1529
1530 /* A match means that the assertion is true; break out of the loop
1531 that matches its alternatives. */
1532
1533 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1534 {
1535 mstart = md->start_match_ptr; /* In case \K reset it */
1536 break;
1537 }
1538
1539 /* If not matched, restore the previous mark setting. */
1540
1541 md->mark = save_mark;
1542
1543 /* See comment in the code for capturing groups above about handling
1544 THEN. */
1545
1546 if (rrc == MATCH_THEN)
1547 {
1548 next = ecode + GET(ecode,1);
1549 if (md->start_match_ptr < next &&
1550 (*ecode == OP_ALT || *next == OP_ALT))
1551 rrc = MATCH_NOMATCH;
1552 }
1553
1554 /* Anything other than NOMATCH causes the entire assertion to fail,
1555 passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1556 uncaptured THEN, which means they take their normal effect. This
1557 consistent approach does not always have exactly the same effect as in
1558 Perl. */
1559
1560 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1561 ecode += GET(ecode, 1);
1562 }
1563 while (*ecode == OP_ALT); /* Continue for next alternative */
1564
1565 /* If we have tried all the alternative branches, the assertion has
1566 failed. If not, we broke out after a match. */
1567
1568 if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1569
1570 /* If checking an assertion for a condition, return MATCH_MATCH. */
1571
1572 if (condassert) RRETURN(MATCH_MATCH);
1573
1574 /* Continue from after a successful assertion, updating the offsets high
1575 water mark, since extracts may have been taken during the assertion. */
1576
1577 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1578 ecode += 1 + LINK_SIZE;
1579 offset_top = md->end_offset_top;
1580 continue;
1581
1582 /* Negative assertion: all branches must fail to match for the assertion to
1583 succeed. */
1584
1585 case OP_ASSERT_NOT:
1586 case OP_ASSERTBACK_NOT:
1587 save_mark = md->mark;
1588 if (md->match_function_type == MATCH_CONDASSERT)
1589 {
1590 condassert = TRUE;
1591 md->match_function_type = 0;
1592 }
1593 else condassert = FALSE;
1594
1595 /* Loop for each alternative branch. */
1596
1597 do
1598 {
1599 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1600 md->mark = save_mark; /* Always restore the mark setting */
1601
1602 switch(rrc)
1603 {
1604 case MATCH_MATCH: /* A successful match means */
1605 case MATCH_ACCEPT: /* the assertion has failed. */
1606 RRETURN(MATCH_NOMATCH);
1607
1608 case MATCH_NOMATCH: /* Carry on with next branch */
1609 break;
1610
1611 /* See comment in the code for capturing groups above about handling
1612 THEN. */
1613
1614 case MATCH_THEN:
1615 next = ecode + GET(ecode,1);
1616 if (md->start_match_ptr < next &&
1617 (*ecode == OP_ALT || *next == OP_ALT))
1618 {
1619 rrc = MATCH_NOMATCH;
1620 break;
1621 }
1622 /* Otherwise fall through. */
1623
1624 /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1625 assertion to fail to match, without considering any more alternatives.
1626 Failing to match means the assertion is true. This is a consistent
1627 approach, but does not always have the same effect as in Perl. */
1628
1629 case MATCH_COMMIT:
1630 case MATCH_SKIP:
1631 case MATCH_SKIP_ARG:
1632 case MATCH_PRUNE:
1633 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1634 goto NEG_ASSERT_TRUE; /* Break out of alternation loop */
1635
1636 /* Anything else is an error */
1637
1638 default:
1639 RRETURN(rrc);
1640 }
1641
1642 /* Continue with next branch */
1643
1644 ecode += GET(ecode,1);
1645 }
1646 while (*ecode == OP_ALT);
1647
1648 /* All branches in the assertion failed to match. */
1649
1650 NEG_ASSERT_TRUE:
1651 if (condassert) RRETURN(MATCH_MATCH); /* Condition assertion */
1652 ecode += 1 + LINK_SIZE; /* Continue with current branch */
1653 continue;
1654
1655 /* Move the subject pointer back. This occurs only at the start of
1656 each branch of a lookbehind assertion. If we are too close to the start to
1657 move back, this match function fails. When working with UTF-8 we move
1658 back a number of characters, not bytes. */
1659
1660 case OP_REVERSE:
1661 #ifdef SUPPORT_UTF
1662 if (utf)
1663 {
1664 i = GET(ecode, 1);
1665 while (i-- > 0)
1666 {
1667 eptr--;
1668 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1669 BACKCHAR(eptr);
1670 }
1671 }
1672 else
1673 #endif
1674
1675 /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1676
1677 {
1678 eptr -= GET(ecode, 1);
1679 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1680 }
1681
1682 /* Save the earliest consulted character, then skip to next op code */
1683
1684 if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1685 ecode += 1 + LINK_SIZE;
1686 break;
1687
1688 /* The callout item calls an external function, if one is provided, passing
1689 details of the match so far. This is mainly for debugging, though the
1690 function is able to force a failure. */
1691
1692 case OP_CALLOUT:
1693 if (PUBL(callout) != NULL)
1694 {
1695 PUBL(callout_block) cb;
1696 cb.version = 2; /* Version 1 of the callout block */
1697 cb.callout_number = ecode[1];
1698 cb.offset_vector = md->offset_vector;
1699 #if defined COMPILE_PCRE8
1700 cb.subject = (PCRE_SPTR)md->start_subject;
1701 #elif defined COMPILE_PCRE16
1702 cb.subject = (PCRE_SPTR16)md->start_subject;
1703 #elif defined COMPILE_PCRE32
1704 cb.subject = (PCRE_SPTR32)md->start_subject;
1705 #endif
1706 cb.subject_length = (int)(md->end_subject - md->start_subject);
1707 cb.start_match = (int)(mstart - md->start_subject);
1708 cb.current_position = (int)(eptr - md->start_subject);
1709 cb.pattern_position = GET(ecode, 2);
1710 cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1711 cb.capture_top = offset_top/2;
1712 cb.capture_last = md->capture_last & CAPLMASK;
1713 /* Internal change requires this for API compatibility. */
1714 if (cb.capture_last == 0) cb.capture_last = -1;
1715 cb.callout_data = md->callout_data;
1716 cb.mark = md->nomatch_mark;
1717 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1718 if (rrc < 0) RRETURN(rrc);
1719 }
1720 ecode += 2 + 2*LINK_SIZE;
1721 break;
1722
1723 /* Recursion either matches the current regex, or some subexpression. The
1724 offset data is the offset to the starting bracket from the start of the
1725 whole pattern. (This is so that it works from duplicated subpatterns.)
1726
1727 The state of the capturing groups is preserved over recursion, and
1728 re-instated afterwards. We don't know how many are started and not yet
1729 finished (offset_top records the completed total) so we just have to save
1730 all the potential data. There may be up to 65535 such values, which is too
1731 large to put on the stack, but using malloc for small numbers seems
1732 expensive. As a compromise, the stack is used when there are no more than
1733 REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
1734
1735 There are also other values that have to be saved. We use a chained
1736 sequence of blocks that actually live on the stack. Thanks to Robin Houston
1737 for the original version of this logic. It has, however, been hacked around
1738 a lot, so he is not to blame for the current way it works. */
1739
1740 case OP_RECURSE:
1741 {
1742 recursion_info *ri;
1743 unsigned int recno;
1744
1745 callpat = md->start_code + GET(ecode, 1);
1746 recno = (callpat == md->start_code)? 0 :
1747 GET2(callpat, 1 + LINK_SIZE);
1748
1749 /* Check for repeating a recursion without advancing the subject pointer.
1750 This should catch convoluted mutual recursions. (Some simple cases are
1751 caught at compile time.) */
1752
1753 for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1754 if (recno == ri->group_num && eptr == ri->subject_position)
1755 RRETURN(PCRE_ERROR_RECURSELOOP);
1756
1757 /* Add to "recursing stack" */
1758
1759 new_recursive.group_num = recno;
1760 new_recursive.saved_capture_last = md->capture_last;
1761 new_recursive.subject_position = eptr;
1762 new_recursive.prevrec = md->recursive;
1763 md->recursive = &new_recursive;
1764
1765 /* Where to continue from afterwards */
1766
1767 ecode += 1 + LINK_SIZE;
1768
1769 /* Now save the offset data */
1770
1771 new_recursive.saved_max = md->offset_end;
1772 if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
1773 new_recursive.offset_save = stacksave;
1774 else
1775 {
1776 new_recursive.offset_save =
1777 (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1778 if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1779 }
1780 memcpy(new_recursive.offset_save, md->offset_vector,
1781 new_recursive.saved_max * sizeof(int));
1782
1783 /* OK, now we can do the recursion. After processing each alternative,
1784 restore the offset data and the last captured value. If there were nested
1785 recursions, md->recursive might be changed, so reset it before looping.
1786 */
1787
1788 DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1789 cbegroup = (*callpat >= OP_SBRA);
1790 do
1791 {
1792 if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1793 RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1794 md, eptrb, RM6);
1795 memcpy(md->offset_vector, new_recursive.offset_save,
1796 new_recursive.saved_max * sizeof(int));
1797 md->capture_last = new_recursive.saved_capture_last;
1798 md->recursive = new_recursive.prevrec;
1799 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1800 {
1801 DPRINTF(("Recursion matched\n"));
1802 if (new_recursive.offset_save != stacksave)
1803 (PUBL(free))(new_recursive.offset_save);
1804
1805 /* Set where we got to in the subject, and reset the start in case
1806 it was changed by \K. This *is* propagated back out of a recursion,
1807 for Perl compatibility. */
1808
1809 eptr = md->end_match_ptr;
1810 mstart = md->start_match_ptr;
1811 goto RECURSION_MATCHED; /* Exit loop; end processing */
1812 }
1813
1814 /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1815 recursion; they cause a NOMATCH for the entire recursion. These codes
1816 are defined in a range that can be tested for. */
1817
1818 if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1819 RRETURN(MATCH_NOMATCH);
1820
1821 /* Any return code other than NOMATCH is an error. */
1822
1823 if (rrc != MATCH_NOMATCH)
1824 {
1825 DPRINTF(("Recursion gave error %d\n", rrc));
1826 if (new_recursive.offset_save != stacksave)
1827 (PUBL(free))(new_recursive.offset_save);
1828 RRETURN(rrc);
1829 }
1830
1831 md->recursive = &new_recursive;
1832 callpat += GET(callpat, 1);
1833 }
1834 while (*callpat == OP_ALT);
1835
1836 DPRINTF(("Recursion didn't match\n"));
1837 md->recursive = new_recursive.prevrec;
1838 if (new_recursive.offset_save != stacksave)
1839 (PUBL(free))(new_recursive.offset_save);
1840 RRETURN(MATCH_NOMATCH);
1841 }
1842
1843 RECURSION_MATCHED:
1844 break;
1845
1846 /* An alternation is the end of a branch; scan along to find the end of the
1847 bracketed group and go to there. */
1848
1849 case OP_ALT:
1850 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1851 break;
1852
1853 /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1854 indicating that it may occur zero times. It may repeat infinitely, or not
1855 at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1856 with fixed upper repeat limits are compiled as a number of copies, with the
1857 optional ones preceded by BRAZERO or BRAMINZERO. */
1858
1859 case OP_BRAZERO:
1860 next = ecode + 1;
1861 RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1862 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1863 do next += GET(next, 1); while (*next == OP_ALT);
1864 ecode = next + 1 + LINK_SIZE;
1865 break;
1866
1867 case OP_BRAMINZERO:
1868 next = ecode + 1;
1869 do next += GET(next, 1); while (*next == OP_ALT);
1870 RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1871 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1872 ecode++;
1873 break;
1874
1875 case OP_SKIPZERO:
1876 next = ecode+1;
1877 do next += GET(next,1); while (*next == OP_ALT);
1878 ecode = next + 1 + LINK_SIZE;
1879 break;
1880
1881 /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1882 here; just jump to the group, with allow_zero set TRUE. */
1883
1884 case OP_BRAPOSZERO:
1885 op = *(++ecode);
1886 allow_zero = TRUE;
1887 if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1888 goto POSSESSIVE_NON_CAPTURE;
1889
1890 /* End of a group, repeated or non-repeating. */
1891
1892 case OP_KET:
1893 case OP_KETRMIN:
1894 case OP_KETRMAX:
1895 case OP_KETRPOS:
1896 prev = ecode - GET(ecode, 1);
1897
1898 /* If this was a group that remembered the subject start, in order to break
1899 infinite repeats of empty string matches, retrieve the subject start from
1900 the chain. Otherwise, set it NULL. */
1901
1902 if (*prev >= OP_SBRA || *prev == OP_ONCE)
1903 {
1904 saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */
1905 eptrb = eptrb->epb_prev; /* Backup to previous group */
1906 }
1907 else saved_eptr = NULL;
1908
1909 /* If we are at the end of an assertion group or a non-capturing atomic
1910 group, stop matching and return MATCH_MATCH, but record the current high
1911 water mark for use by positive assertions. We also need to record the match
1912 start in case it was changed by \K. */
1913
1914 if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1915 *prev == OP_ONCE_NC)
1916 {
1917 md->end_match_ptr = eptr; /* For ONCE_NC */
1918 md->end_offset_top = offset_top;
1919 md->start_match_ptr = mstart;
1920 RRETURN(MATCH_MATCH); /* Sets md->mark */
1921 }
1922
1923 /* For capturing groups we have to check the group number back at the start
1924 and if necessary complete handling an extraction by setting the offsets and
1925 bumping the high water mark. Whole-pattern recursion is coded as a recurse
1926 into group 0, so it won't be picked up here. Instead, we catch it when the
1927 OP_END is reached. Other recursion is handled here. We just have to record
1928 the current subject position and start match pointer and give a MATCH
1929 return. */
1930
1931 if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1932 *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1933 {
1934 number = GET2(prev, 1+LINK_SIZE);
1935 offset = number << 1;
1936
1937 #ifdef PCRE_DEBUG
1938 printf("end bracket %d", number);
1939 printf("\n");
1940 #endif
1941
1942 /* Handle a recursively called group. */
1943
1944 if (md->recursive != NULL && md->recursive->group_num == number)
1945 {
1946 md->end_match_ptr = eptr;
1947 md->start_match_ptr = mstart;
1948 RRETURN(MATCH_MATCH);
1949 }
1950
1951 /* Deal with capturing */
1952
1953 md->capture_last = (md->capture_last & OVFLMASK) | number;
1954 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1955 {
1956 /* If offset is greater than offset_top, it means that we are
1957 "skipping" a capturing group, and that group's offsets must be marked
1958 unset. In earlier versions of PCRE, all the offsets were unset at the
1959 start of matching, but this doesn't work because atomic groups and
1960 assertions can cause a value to be set that should later be unset.
1961 Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1962 part of the atomic group, but this is not on the final matching path,
1963 so must be unset when 2 is set. (If there is no group 2, there is no
1964 problem, because offset_top will then be 2, indicating no capture.) */
1965
1966 if (offset > offset_top)
1967 {
1968 register int *iptr = md->offset_vector + offset_top;
1969 register int *iend = md->offset_vector + offset;
1970 while (iptr < iend) *iptr++ = -1;
1971 }
1972
1973 /* Now make the extraction */
1974
1975 md->offset_vector[offset] =
1976 md->offset_vector[md->offset_end - number];
1977 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1978 if (offset_top <= offset) offset_top = offset + 2;
1979 }
1980 }
1981
1982 /* For an ordinary non-repeating ket, just continue at this level. This
1983 also happens for a repeating ket if no characters were matched in the
1984 group. This is the forcible breaking of infinite loops as implemented in
1985 Perl 5.005. For a non-repeating atomic group that includes captures,
1986 establish a backup point by processing the rest of the pattern at a lower
1987 level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1988 original OP_ONCE level, thereby bypassing intermediate backup points, but
1989 resetting any captures that happened along the way. */
1990
1991 if (*ecode == OP_KET || eptr == saved_eptr)
1992 {
1993 if (*prev == OP_ONCE)
1994 {
1995 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1996 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1997 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
1998 RRETURN(MATCH_ONCE);
1999 }
2000 ecode += 1 + LINK_SIZE; /* Carry on at this level */
2001 break;
2002 }
2003
2004 /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2005 and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2006 at a time from the outer level, thus saving stack. */
2007
2008 if (*ecode == OP_KETRPOS)
2009 {
2010 md->end_match_ptr = eptr;
2011 md->end_offset_top = offset_top;
2012 RRETURN(MATCH_KETRPOS);
2013 }
2014
2015 /* The normal repeating kets try the rest of the pattern or restart from
2016 the preceding bracket, in the appropriate order. In the second case, we can
2017 use tail recursion to avoid using another stack frame, unless we have an
2018 an atomic group or an unlimited repeat of a group that can match an empty
2019 string. */
2020
2021 if (*ecode == OP_KETRMIN)
2022 {
2023 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2024 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2025 if (*prev == OP_ONCE)
2026 {
2027 RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2028 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2029 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
2030 RRETURN(MATCH_ONCE);
2031 }
2032 if (*prev >= OP_SBRA) /* Could match an empty string */
2033 {
2034 RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2035 RRETURN(rrc);
2036 }
2037 ecode = prev;
2038 goto TAIL_RECURSE;
2039 }
2040 else /* OP_KETRMAX */
2041 {
2042 RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2043 if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2044 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2045 if (*prev == OP_ONCE)
2046 {
2047 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2048 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2049 md->once_target = prev;
2050 RRETURN(MATCH_ONCE);
2051 }
2052 ecode += 1 + LINK_SIZE;
2053 goto TAIL_RECURSE;
2054 }
2055 /* Control never gets here */
2056
2057 /* Not multiline mode: start of subject assertion, unless notbol. */
2058
2059 case OP_CIRC:
2060 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2061
2062 /* Start of subject assertion */
2063
2064 case OP_SOD:
2065 if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
2066 ecode++;
2067 break;
2068
2069 /* Multiline mode: start of subject unless notbol, or after any newline. */
2070
2071 case OP_CIRCM:
2072 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2073 if (eptr != md->start_subject &&
2074 (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2075 RRETURN(MATCH_NOMATCH);
2076 ecode++;
2077 break;
2078
2079 /* Start of match assertion */
2080
2081 case OP_SOM:
2082 if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
2083 ecode++;
2084 break;
2085
2086 /* Reset the start of match point */
2087
2088 case OP_SET_SOM:
2089 mstart = eptr;
2090 ecode++;
2091 break;
2092
2093 /* Multiline mode: assert before any newline, or before end of subject
2094 unless noteol is set. */
2095
2096 case OP_DOLLM:
2097 if (eptr < md->end_subject)
2098 {
2099 if (!IS_NEWLINE(eptr))
2100 {
2101 if (md->partial != 0 &&
2102 eptr + 1 >= md->end_subject &&
2103 NLBLOCK->nltype == NLTYPE_FIXED &&
2104 NLBLOCK->nllen == 2 &&
2105 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2106 {
2107 md->hitend = TRUE;
2108 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2109 }
2110 RRETURN(MATCH_NOMATCH);
2111 }
2112 }
2113 else
2114 {
2115 if (md->noteol) RRETURN(MATCH_NOMATCH);
2116 SCHECK_PARTIAL();
2117 }
2118 ecode++;
2119 break;
2120
2121 /* Not multiline mode: assert before a terminating newline or before end of
2122 subject unless noteol is set. */
2123
2124 case OP_DOLL:
2125 if (md->noteol) RRETURN(MATCH_NOMATCH);
2126 if (!md->endonly) goto ASSERT_NL_OR_EOS;
2127
2128 /* ... else fall through for endonly */
2129
2130 /* End of subject assertion (\z) */
2131
2132 case OP_EOD:
2133 if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2134 SCHECK_PARTIAL();
2135 ecode++;
2136 break;
2137
2138 /* End of subject or ending \n assertion (\Z) */
2139
2140 case OP_EODN:
2141 ASSERT_NL_OR_EOS:
2142 if (eptr < md->end_subject &&
2143 (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2144 {
2145 if (md->partial != 0 &&
2146 eptr + 1 >= md->end_subject &&
2147 NLBLOCK->nltype == NLTYPE_FIXED &&
2148 NLBLOCK->nllen == 2 &&
2149 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2150 {
2151 md->hitend = TRUE;
2152 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2153 }
2154 RRETURN(MATCH_NOMATCH);
2155 }
2156
2157 /* Either at end of string or \n before end. */
2158
2159 SCHECK_PARTIAL();
2160 ecode++;
2161 break;
2162
2163 /* Word boundary assertions */
2164
2165 case OP_NOT_WORD_BOUNDARY:
2166 case OP_WORD_BOUNDARY:
2167 {
2168
2169 /* Find out if the previous and current characters are "word" characters.
2170 It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2171 be "non-word" characters. Remember the earliest consulted character for
2172 partial matching. */
2173
2174 #ifdef SUPPORT_UTF
2175 if (utf)
2176 {
2177 /* Get status of previous character */
2178
2179 if (eptr == md->start_subject) prev_is_word = FALSE; else
2180 {
2181 PCRE_PUCHAR lastptr = eptr - 1;
2182 BACKCHAR(lastptr);
2183 if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2184 GETCHAR(c, lastptr);
2185 #ifdef SUPPORT_UCP
2186 if (md->use_ucp)
2187 {
2188 if (c == '_') prev_is_word = TRUE; else
2189 {
2190 int cat = UCD_CATEGORY(c);
2191 prev_is_word = (cat == ucp_L || cat == ucp_N);
2192 }
2193 }
2194 else
2195 #endif
2196 prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2197 }
2198
2199 /* Get status of next character */
2200
2201 if (eptr >= md->end_subject)
2202 {
2203 SCHECK_PARTIAL();
2204 cur_is_word = FALSE;
2205 }
2206 else
2207 {
2208 GETCHAR(c, eptr);
2209 #ifdef SUPPORT_UCP
2210 if (md->use_ucp)
2211 {
2212 if (c == '_') cur_is_word = TRUE; else
2213 {
2214 int cat = UCD_CATEGORY(c);
2215 cur_is_word = (cat == ucp_L || cat == ucp_N);
2216 }
2217 }
2218 else
2219 #endif
2220 cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2221 }
2222 }
2223 else
2224 #endif
2225
2226 /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2227 consistency with the behaviour of \w we do use it in this case. */
2228
2229 {
2230 /* Get status of previous character */
2231
2232 if (eptr == md->start_subject) prev_is_word = FALSE; else
2233 {
2234 if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2235 #ifdef SUPPORT_UCP
2236 if (md->use_ucp)
2237 {
2238 c = eptr[-1];
2239 if (c == '_') prev_is_word = TRUE; else
2240 {
2241 int cat = UCD_CATEGORY(c);
2242 prev_is_word = (cat == ucp_L || cat == ucp_N);
2243 }
2244 }
2245 else
2246 #endif
2247 prev_is_word = MAX_255(eptr[-1])
2248 && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2249 }
2250
2251 /* Get status of next character */
2252
2253 if (eptr >= md->end_subject)
2254 {
2255 SCHECK_PARTIAL();
2256 cur_is_word = FALSE;
2257 }
2258 else
2259 #ifdef SUPPORT_UCP
2260 if (md->use_ucp)
2261 {
2262 c = *eptr;
2263 if (c == '_') cur_is_word = TRUE; else
2264 {
2265 int cat = UCD_CATEGORY(c);
2266 cur_is_word = (cat == ucp_L || cat == ucp_N);
2267 }
2268 }
2269 else
2270 #endif
2271 cur_is_word = MAX_255(*eptr)
2272 && ((md->ctypes[*eptr] & ctype_word) != 0);
2273 }
2274
2275 /* Now see if the situation is what we want */
2276
2277 if ((*ecode++ == OP_WORD_BOUNDARY)?
2278 cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2279 RRETURN(MATCH_NOMATCH);
2280 }
2281 break;
2282
2283 /* Match any single character type except newline; have to take care with
2284 CRLF newlines and partial matching. */
2285
2286 case OP_ANY:
2287 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2288 if (md->partial != 0 &&
2289 eptr + 1 >= md->end_subject &&
2290 NLBLOCK->nltype == NLTYPE_FIXED &&
2291 NLBLOCK->nllen == 2 &&
2292 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2293 {
2294 md->hitend = TRUE;
2295 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2296 }
2297
2298 /* Fall through */
2299
2300 /* Match any single character whatsoever. */
2301
2302 case OP_ALLANY:
2303 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2304 { /* not be updated before SCHECK_PARTIAL. */
2305 SCHECK_PARTIAL();
2306 RRETURN(MATCH_NOMATCH);
2307 }
2308 eptr++;
2309 #ifdef SUPPORT_UTF
2310 if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2311 #endif
2312 ecode++;
2313 break;
2314
2315 /* Match a single byte, even in UTF-8 mode. This opcode really does match
2316 any byte, even newline, independent of the setting of PCRE_DOTALL. */
2317
2318 case OP_ANYBYTE:
2319 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2320 { /* not be updated before SCHECK_PARTIAL. */
2321 SCHECK_PARTIAL();
2322 RRETURN(MATCH_NOMATCH);
2323 }
2324 eptr++;
2325 ecode++;
2326 break;
2327
2328 case OP_NOT_DIGIT:
2329 if (eptr >= md->end_subject)
2330 {
2331 SCHECK_PARTIAL();
2332 RRETURN(MATCH_NOMATCH);
2333 }
2334 GETCHARINCTEST(c, eptr);
2335 if (
2336 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2337 c < 256 &&
2338 #endif
2339 (md->ctypes[c] & ctype_digit) != 0
2340 )
2341 RRETURN(MATCH_NOMATCH);
2342 ecode++;
2343 break;
2344
2345 case OP_DIGIT:
2346 if (eptr >= md->end_subject)
2347 {
2348 SCHECK_PARTIAL();
2349 RRETURN(MATCH_NOMATCH);
2350 }
2351 GETCHARINCTEST(c, eptr);
2352 if (
2353 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2354 c > 255 ||
2355 #endif
2356 (md->ctypes[c] & ctype_digit) == 0
2357 )
2358 RRETURN(MATCH_NOMATCH);
2359 ecode++;
2360 break;
2361
2362 case OP_NOT_WHITESPACE:
2363 if (eptr >= md->end_subject)
2364 {
2365 SCHECK_PARTIAL();
2366 RRETURN(MATCH_NOMATCH);
2367 }
2368 GETCHARINCTEST(c, eptr);
2369 if (
2370 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2371 c < 256 &&
2372 #endif
2373 (md->ctypes[c] & ctype_space) != 0
2374 )
2375 RRETURN(MATCH_NOMATCH);
2376 ecode++;
2377 break;
2378
2379 case OP_WHITESPACE:
2380 if (eptr >= md->end_subject)
2381 {
2382 SCHECK_PARTIAL();
2383 RRETURN(MATCH_NOMATCH);
2384 }
2385 GETCHARINCTEST(c, eptr);
2386 if (
2387 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2388 c > 255 ||
2389 #endif
2390 (md->ctypes[c] & ctype_space) == 0
2391 )
2392 RRETURN(MATCH_NOMATCH);
2393 ecode++;
2394 break;
2395
2396 case OP_NOT_WORDCHAR:
2397 if (eptr >= md->end_subject)
2398 {
2399 SCHECK_PARTIAL();
2400 RRETURN(MATCH_NOMATCH);
2401 }
2402 GETCHARINCTEST(c, eptr);
2403 if (
2404 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2405 c < 256 &&
2406 #endif
2407 (md->ctypes[c] & ctype_word) != 0
2408 )
2409 RRETURN(MATCH_NOMATCH);
2410 ecode++;
2411 break;
2412
2413 case OP_WORDCHAR:
2414 if (eptr >= md->end_subject)
2415 {
2416 SCHECK_PARTIAL();
2417 RRETURN(MATCH_NOMATCH);
2418 }
2419 GETCHARINCTEST(c, eptr);
2420 if (
2421 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2422 c > 255 ||
2423 #endif
2424 (md->ctypes[c] & ctype_word) == 0
2425 )
2426 RRETURN(MATCH_NOMATCH);
2427 ecode++;
2428 break;
2429
2430 case OP_ANYNL:
2431 if (eptr >= md->end_subject)
2432 {
2433 SCHECK_PARTIAL();
2434 RRETURN(MATCH_NOMATCH);
2435 }
2436 GETCHARINCTEST(c, eptr);
2437 switch(c)
2438 {
2439 default: RRETURN(MATCH_NOMATCH);
2440
2441 case CHAR_CR:
2442 if (eptr >= md->end_subject)
2443 {
2444 SCHECK_PARTIAL();
2445 }
2446 else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2447 break;
2448
2449 case CHAR_LF:
2450 break;
2451
2452 case CHAR_VT:
2453 case CHAR_FF:
2454 case CHAR_NEL:
2455 #ifndef EBCDIC
2456 case 0x2028:
2457 case 0x2029:
2458 #endif /* Not EBCDIC */
2459 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2460 break;
2461 }
2462 ecode++;
2463 break;
2464
2465 case OP_NOT_HSPACE:
2466 if (eptr >= md->end_subject)
2467 {
2468 SCHECK_PARTIAL();
2469 RRETURN(MATCH_NOMATCH);
2470 }
2471 GETCHARINCTEST(c, eptr);
2472 switch(c)
2473 {
2474 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
2475 default: break;
2476 }
2477 ecode++;
2478 break;
2479
2480 case OP_HSPACE:
2481 if (eptr >= md->end_subject)
2482 {
2483 SCHECK_PARTIAL();
2484 RRETURN(MATCH_NOMATCH);
2485 }
2486 GETCHARINCTEST(c, eptr);
2487 switch(c)
2488 {
2489 HSPACE_CASES: break; /* Byte and multibyte cases */
2490 default: RRETURN(MATCH_NOMATCH);
2491 }
2492 ecode++;
2493 break;
2494
2495 case OP_NOT_VSPACE:
2496 if (eptr >= md->end_subject)
2497 {
2498 SCHECK_PARTIAL();
2499 RRETURN(MATCH_NOMATCH);
2500 }
2501 GETCHARINCTEST(c, eptr);
2502 switch(c)
2503 {
2504 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2505 default: break;
2506 }
2507 ecode++;
2508 break;
2509
2510 case OP_VSPACE:
2511 if (eptr >= md->end_subject)
2512 {
2513 SCHECK_PARTIAL();
2514 RRETURN(MATCH_NOMATCH);
2515 }
2516 GETCHARINCTEST(c, eptr);
2517 switch(c)
2518 {
2519 VSPACE_CASES: break;
2520 default: RRETURN(MATCH_NOMATCH);
2521 }
2522 ecode++;
2523 break;
2524
2525 #ifdef SUPPORT_UCP
2526 /* Check the next character by Unicode property. We will get here only
2527 if the support is in the binary; otherwise a compile-time error occurs. */
2528
2529 case OP_PROP:
2530 case OP_NOTPROP:
2531 if (eptr >= md->end_subject)
2532 {
2533 SCHECK_PARTIAL();
2534 RRETURN(MATCH_NOMATCH);
2535 }
2536 GETCHARINCTEST(c, eptr);
2537 {
2538 const pcre_uint32 *cp;
2539 const ucd_record *prop = GET_UCD(c);
2540
2541 switch(ecode[1])
2542 {
2543 case PT_ANY:
2544 if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2545 break;
2546
2547 case PT_LAMP:
2548 if ((prop->chartype == ucp_Lu ||
2549 prop->chartype == ucp_Ll ||
2550 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2551 RRETURN(MATCH_NOMATCH);
2552 break;
2553
2554 case PT_GC:
2555 if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2556 RRETURN(MATCH_NOMATCH);
2557 break;
2558
2559 case PT_PC:
2560 if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2561 RRETURN(MATCH_NOMATCH);
2562 break;
2563
2564 case PT_SC:
2565 if ((ecode[2] != prop->script) == (op == OP_PROP))
2566 RRETURN(MATCH_NOMATCH);
2567 break;
2568
2569 /* These are specials */
2570
2571 case PT_ALNUM:
2572 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2573 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2574 RRETURN(MATCH_NOMATCH);
2575 break;
2576
2577 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
2578 which means that Perl space and POSIX space are now identical. PCRE
2579 was changed at release 8.34. */
2580
2581 case PT_SPACE: /* Perl space */
2582 case PT_PXSPACE: /* POSIX space */
2583 switch(c)
2584 {
2585 HSPACE_CASES:
2586 VSPACE_CASES:
2587 if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2588 break;
2589
2590 default:
2591 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z) ==
2592 (op == OP_NOTPROP)) RRETURN(MATCH_NOMATCH);
2593 break;
2594 }
2595 break;
2596
2597 case PT_WORD:
2598 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2599 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2600 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2601 RRETURN(MATCH_NOMATCH);
2602 break;
2603
2604 case PT_CLIST:
2605 cp = PRIV(ucd_caseless_sets) + ecode[2];
2606 for (;;)
2607 {
2608 if (c < *cp)
2609 { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2610 if (c == *cp++)
2611 { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2612 }
2613 break;
2614
2615 case PT_UCNC:
2616 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2617 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2618 c >= 0xe000) == (op == OP_NOTPROP))
2619 RRETURN(MATCH_NOMATCH);
2620 break;
2621
2622 /* This should never occur */
2623
2624 default:
2625 RRETURN(PCRE_ERROR_INTERNAL);
2626 }
2627
2628 ecode += 3;
2629 }
2630 break;
2631
2632 /* Match an extended Unicode sequence. We will get here only if the support
2633 is in the binary; otherwise a compile-time error occurs. */
2634
2635 case OP_EXTUNI:
2636 if (eptr >= md->end_subject)
2637 {
2638 SCHECK_PARTIAL();
2639 RRETURN(MATCH_NOMATCH);
2640 }
2641 else
2642 {
2643 int lgb, rgb;
2644 GETCHARINCTEST(c, eptr);
2645 lgb = UCD_GRAPHBREAK(c);
2646 while (eptr < md->end_subject)
2647 {
2648 int len = 1;
2649 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2650 rgb = UCD_GRAPHBREAK(c);
2651 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2652 lgb = rgb;
2653 eptr += len;
2654 }
2655 }
2656 CHECK_PARTIAL();
2657 ecode++;
2658 break;
2659 #endif /* SUPPORT_UCP */
2660
2661
2662 /* Match a back reference, possibly repeatedly. Look past the end of the
2663 item to see if there is repeat information following. The code is similar
2664 to that for character classes, but repeated for efficiency. Then obey
2665 similar code to character type repeats - written out again for speed.
2666 However, if the referenced string is the empty string, always treat
2667 it as matched, any number of times (otherwise there could be infinite
2668 loops). If the reference is unset, there are two possibilities:
2669
2670 (a) In the default, Perl-compatible state, set the length negative;
2671 this ensures that every attempt at a match fails. We can't just fail
2672 here, because of the possibility of quantifiers with zero minima.
2673
2674 (b) If the JavaScript compatibility flag is set, set the length to zero
2675 so that the back reference matches an empty string.
2676
2677 Otherwise, set the length to the length of what was matched by the
2678 referenced subpattern.
2679
2680 The OP_REF and OP_REFI opcodes are used for a reference to a numbered group
2681 or to a non-duplicated named group. For a duplicated named group, OP_DNREF
2682 and OP_DNREFI are used. In this case we must scan the list of groups to
2683 which the name refers, and use the first one that is set. */
2684
2685 case OP_DNREF:
2686 case OP_DNREFI:
2687 caseless = op == OP_DNREFI;
2688 {
2689 int count = GET2(ecode, 1+IMM2_SIZE);
2690 pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
2691 ecode += 1 + 2*IMM2_SIZE;
2692
2693 while (count-- > 0)
2694 {
2695 offset = GET2(slot, 0) << 1;
2696 if (offset < offset_top && md->offset_vector[offset] >= 0) break;
2697 slot += md->name_entry_size;
2698 }
2699 if (count < 0)
2700 length = (md->jscript_compat)? 0 : -1;
2701 else
2702 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2703 }
2704 goto REF_REPEAT;
2705
2706 case OP_REF:
2707 case OP_REFI:
2708 caseless = op == OP_REFI;
2709 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
2710 ecode += 1 + IMM2_SIZE;
2711 if (offset >= offset_top || md->offset_vector[offset] < 0)
2712 length = (md->jscript_compat)? 0 : -1;
2713 else
2714 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2715
2716 /* Set up for repetition, or handle the non-repeated case */
2717
2718 REF_REPEAT:
2719 switch (*ecode)
2720 {
2721 case OP_CRSTAR:
2722 case OP_CRMINSTAR:
2723 case OP_CRPLUS:
2724 case OP_CRMINPLUS:
2725 case OP_CRQUERY:
2726 case OP_CRMINQUERY:
2727 c = *ecode++ - OP_CRSTAR;
2728 minimize = (c & 1) != 0;
2729 min = rep_min[c]; /* Pick up values from tables; */
2730 max = rep_max[c]; /* zero for max => infinity */
2731 if (max == 0) max = INT_MAX;
2732 break;
2733
2734 case OP_CRRANGE:
2735 case OP_CRMINRANGE:
2736 minimize = (*ecode == OP_CRMINRANGE);
2737 min = GET2(ecode, 1);
2738 max = GET2(ecode, 1 + IMM2_SIZE);
2739 if (max == 0) max = INT_MAX;
2740 ecode += 1 + 2 * IMM2_SIZE;
2741 break;
2742
2743 default: /* No repeat follows */
2744 if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2745 {
2746 if (length == -2) eptr = md->end_subject; /* Partial match */
2747 CHECK_PARTIAL();
2748 RRETURN(MATCH_NOMATCH);
2749 }
2750 eptr += length;
2751 continue; /* With the main loop */
2752 }
2753
2754 /* Handle repeated back references. If the length of the reference is
2755 zero, just continue with the main loop. If the length is negative, it
2756 means the reference is unset in non-Java-compatible mode. If the minimum is
2757 zero, we can continue at the same level without recursion. For any other
2758 minimum, carrying on will result in NOMATCH. */
2759
2760 if (length == 0) continue;
2761 if (length < 0 && min == 0) continue;
2762
2763 /* First, ensure the minimum number of matches are present. We get back
2764 the length of the reference string explicitly rather than passing the
2765 address of eptr, so that eptr can be a register variable. */
2766
2767 for (i = 1; i <= min; i++)
2768 {
2769 int slength;
2770 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2771 {
2772 if (slength == -2) eptr = md->end_subject; /* Partial match */
2773 CHECK_PARTIAL();
2774 RRETURN(MATCH_NOMATCH);
2775 }
2776 eptr += slength;
2777 }
2778
2779 /* If min = max, continue at the same level without recursion.
2780 They are not both allowed to be zero. */
2781
2782 if (min == max) continue;
2783
2784 /* If minimizing, keep trying and advancing the pointer */
2785
2786 if (minimize)
2787 {
2788 for (fi = min;; fi++)
2789 {
2790 int slength;
2791 RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2792 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2793 if (fi >= max) RRETURN(MATCH_NOMATCH);
2794 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2795 {
2796 if (slength == -2) eptr = md->end_subject; /* Partial match */
2797 CHECK_PARTIAL();
2798 RRETURN(MATCH_NOMATCH);
2799 }
2800 eptr += slength;
2801 }
2802 /* Control never gets here */
2803 }
2804
2805 /* If maximizing, find the longest string and work backwards */
2806
2807 else
2808 {
2809 pp = eptr;
2810 for (i = min; i < max; i++)
2811 {
2812 int slength;
2813 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2814 {
2815 /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2816 the soft partial matching case. */
2817
2818 if (slength == -2 && md->partial != 0 &&
2819 md->end_subject > md->start_used_ptr)
2820 {
2821 md->hitend = TRUE;
2822 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2823 }
2824 break;
2825 }
2826 eptr += slength;
2827 }
2828
2829 while (eptr >= pp)
2830 {
2831 RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2832 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2833 eptr -= length;
2834 }
2835 RRETURN(MATCH_NOMATCH);
2836 }
2837 /* Control never gets here */
2838
2839 /* Match a bit-mapped character class, possibly repeatedly. This op code is
2840 used when all the characters in the class have values in the range 0-255,
2841 and either the matching is caseful, or the characters are in the range
2842 0-127 when UTF-8 processing is enabled. The only difference between
2843 OP_CLASS and OP_NCLASS occurs when a data character outside the range is
2844 encountered.
2845
2846 First, look past the end of the item to see if there is repeat information
2847 following. Then obey similar code to character type repeats - written out
2848 again for speed. */
2849
2850 case OP_NCLASS:
2851 case OP_CLASS:
2852 {
2853 /* The data variable is saved across frames, so the byte map needs to
2854 be stored there. */
2855 #define BYTE_MAP ((pcre_uint8 *)data)
2856 data = ecode + 1; /* Save for matching */
2857 ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2858
2859 switch (*ecode)
2860 {
2861 case OP_CRSTAR:
2862 case OP_CRMINSTAR:
2863 case OP_CRPLUS:
2864 case OP_CRMINPLUS:
2865 case OP_CRQUERY:
2866 case OP_CRMINQUERY:
2867 c = *ecode++ - OP_CRSTAR;
2868 minimize = (c & 1) != 0;
2869 min = rep_min[c]; /* Pick up values from tables; */
2870 max = rep_max[c]; /* zero for max => infinity */
2871 if (max == 0) max = INT_MAX;
2872 break;
2873
2874 case OP_CRRANGE:
2875 case OP_CRMINRANGE:
2876 minimize = (*ecode == OP_CRMINRANGE);
2877 min = GET2(ecode, 1);
2878 max = GET2(ecode, 1 + IMM2_SIZE);
2879 if (max == 0) max = INT_MAX;
2880 ecode += 1 + 2 * IMM2_SIZE;
2881 break;
2882
2883 default: /* No repeat follows */
2884 min = max = 1;
2885 break;
2886 }
2887
2888 /* First, ensure the minimum number of matches are present. */
2889
2890 #ifdef SUPPORT_UTF
2891 if (utf)
2892 {
2893 for (i = 1; i <= min; i++)
2894 {
2895 if (eptr >= md->end_subject)
2896 {
2897 SCHECK_PARTIAL();
2898 RRETURN(MATCH_NOMATCH);
2899 }
2900 GETCHARINC(c, eptr);
2901 if (c > 255)
2902 {
2903 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2904 }
2905 else
2906 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2907 }
2908 }
2909 else
2910 #endif
2911 /* Not UTF mode */
2912 {
2913 for (i = 1; i <= min; i++)
2914 {
2915 if (eptr >= md->end_subject)
2916 {
2917 SCHECK_PARTIAL();
2918 RRETURN(MATCH_NOMATCH);
2919 }
2920 c = *eptr++;
2921 #ifndef COMPILE_PCRE8
2922 if (c > 255)
2923 {
2924 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2925 }
2926 else
2927 #endif
2928 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2929 }
2930 }
2931
2932 /* If max == min we can continue with the main loop without the
2933 need to recurse. */
2934
2935 if (min == max) continue;
2936
2937 /* If minimizing, keep testing the rest of the expression and advancing
2938 the pointer while it matches the class. */
2939
2940 if (minimize)
2941 {
2942 #ifdef SUPPORT_UTF
2943 if (utf)
2944 {
2945 for (fi = min;; fi++)
2946 {
2947 RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2948 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2949 if (fi >= max) RRETURN(MATCH_NOMATCH);
2950 if (eptr >= md->end_subject)
2951 {
2952 SCHECK_PARTIAL();
2953 RRETURN(MATCH_NOMATCH);
2954 }
2955 GETCHARINC(c, eptr);
2956 if (c > 255)
2957 {
2958 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2959 }
2960 else
2961 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2962 }
2963 }
2964 else
2965 #endif
2966 /* Not UTF mode */
2967 {
2968 for (fi = min;; fi++)
2969 {
2970 RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2971 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2972 if (fi >= max) RRETURN(MATCH_NOMATCH);
2973 if (eptr >= md->end_subject)
2974 {
2975 SCHECK_PARTIAL();
2976 RRETURN(MATCH_NOMATCH);
2977 }
2978 c = *eptr++;
2979 #ifndef COMPILE_PCRE8
2980 if (c > 255)
2981 {
2982 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2983 }
2984 else
2985 #endif
2986 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2987 }
2988 }
2989 /* Control never gets here */
2990 }
2991
2992 /* If maximizing, find the longest possible run, then work backwards. */
2993
2994 else
2995 {
2996 pp = eptr;
2997
2998 #ifdef SUPPORT_UTF
2999 if (utf)
3000 {
3001 for (i = min; i < max; i++)
3002 {
3003 int len = 1;
3004 if (eptr >= md->end_subject)
3005 {
3006 SCHECK_PARTIAL();
3007 break;
3008 }
3009 GETCHARLEN(c, eptr, len);
3010 if (c > 255)
3011 {
3012 if (op == OP_CLASS) break;
3013 }
3014 else
3015 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3016 eptr += len;
3017 }
3018 for (;;)
3019 {
3020 RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3021 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3022 if (eptr-- == pp) break; /* Stop if tried at original pos */
3023 BACKCHAR(eptr);
3024 }
3025 }
3026 else
3027 #endif
3028 /* Not UTF mode */
3029 {
3030 for (i = min; i < max; i++)
3031 {
3032 if (eptr >= md->end_subject)
3033 {
3034 SCHECK_PARTIAL();
3035 break;
3036 }
3037 c = *eptr;
3038 #ifndef COMPILE_PCRE8
3039 if (c > 255)
3040 {
3041 if (op == OP_CLASS) break;
3042 }
3043 else
3044 #endif
3045 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3046 eptr++;
3047 }
3048 while (eptr >= pp)
3049 {
3050 RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3051 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3052 eptr--;
3053 }
3054 }
3055
3056 RRETURN(MATCH_NOMATCH);
3057 }
3058 #undef BYTE_MAP
3059 }
3060 /* Control never gets here */
3061
3062
3063 /* Match an extended character class. This opcode is encountered only
3064 when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
3065 mode, because Unicode properties are supported in non-UTF-8 mode. */
3066
3067 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3068 case OP_XCLASS:
3069 {
3070 data = ecode + 1 + LINK_SIZE; /* Save for matching */
3071 ecode += GET(ecode, 1); /* Advance past the item */
3072
3073 switch (*ecode)
3074 {
3075 case OP_CRSTAR:
3076 case OP_CRMINSTAR:
3077 case OP_CRPLUS:
3078 case OP_CRMINPLUS:
3079 case OP_CRQUERY:
3080 case OP_CRMINQUERY:
3081 c = *ecode++ - OP_CRSTAR;
3082 minimize = (c & 1) != 0;
3083 min = rep_min[c]; /* Pick up values from tables; */
3084 max = rep_max[c]; /* zero for max => infinity */
3085 if (max == 0) max = INT_MAX;
3086 break;
3087
3088 case OP_CRRANGE:
3089 case OP_CRMINRANGE:
3090 minimize = (*ecode == OP_CRMINRANGE);
3091 min = GET2(ecode, 1);
3092 max = GET2(ecode, 1 + IMM2_SIZE);
3093 if (max == 0) max = INT_MAX;
3094 ecode += 1 + 2 * IMM2_SIZE;
3095 break;
3096
3097 default: /* No repeat follows */
3098 min = max = 1;
3099 break;
3100 }
3101
3102 /* First, ensure the minimum number of matches are present. */
3103
3104 for (i = 1; i <= min; i++)
3105 {
3106 if (eptr >= md->end_subject)
3107 {
3108 SCHECK_PARTIAL();
3109 RRETURN(MATCH_NOMATCH);
3110 }
3111 GETCHARINCTEST(c, eptr);
3112 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3113 }
3114
3115 /* If max == min we can continue with the main loop without the
3116 need to recurse. */
3117
3118 if (min == max) continue;
3119
3120 /* If minimizing, keep testing the rest of the expression and advancing
3121 the pointer while it matches the class. */
3122
3123 if (minimize)
3124 {
3125 for (fi = min;; fi++)
3126 {
3127 RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3128 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3129 if (fi >= max) RRETURN(MATCH_NOMATCH);
3130 if (eptr >= md->end_subject)
3131 {
3132 SCHECK_PARTIAL();
3133 RRETURN(MATCH_NOMATCH);
3134 }
3135 GETCHARINCTEST(c, eptr);
3136 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3137 }
3138 /* Control never gets here */
3139 }
3140
3141 /* If maximizing, find the longest possible run, then work backwards. */
3142
3143 else
3144 {
3145 pp = eptr;
3146 for (i = min; i < max; i++)
3147 {
3148 int len = 1;
3149 if (eptr >= md->end_subject)
3150 {
3151 SCHECK_PARTIAL();
3152 break;
3153 }
3154 #ifdef SUPPORT_UTF
3155 GETCHARLENTEST(c, eptr, len);
3156 #else
3157 c = *eptr;
3158 #endif
3159 if (!PRIV(xclass)(c, data, utf)) break;
3160 eptr += len;
3161 }
3162 for(;;)
3163 {
3164 RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3165 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3166 if (eptr-- == pp) break; /* Stop if tried at original pos */
3167 #ifdef SUPPORT_UTF
3168 if (utf) BACKCHAR(eptr);
3169 #endif
3170 }
3171 RRETURN(MATCH_NOMATCH);
3172 }
3173
3174 /* Control never gets here */
3175 }
3176 #endif /* End of XCLASS */
3177
3178 /* Match a single character, casefully */
3179
3180 case OP_CHAR:
3181 #ifdef SUPPORT_UTF
3182 if (utf)
3183 {
3184 length = 1;
3185 ecode++;
3186 GETCHARLEN(fc, ecode, length);
3187 if (length > md->end_subject - eptr)
3188 {
3189 CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */
3190 RRETURN(MATCH_NOMATCH);
3191 }
3192 while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3193 }
3194 else
3195 #endif
3196 /* Not UTF mode */
3197 {
3198 if (md->end_subject - eptr < 1)
3199 {
3200 SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */
3201 RRETURN(MATCH_NOMATCH);
3202 }
3203 if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3204 ecode += 2;
3205 }
3206 break;
3207
3208 /* Match a single character, caselessly. If we are at the end of the
3209 subject, give up immediately. */
3210
3211 case OP_CHARI:
3212 if (eptr >= md->end_subject)
3213 {
3214 SCHECK_PARTIAL();
3215 RRETURN(MATCH_NOMATCH);
3216 }
3217
3218 #ifdef SUPPORT_UTF
3219 if (utf)
3220 {
3221 length = 1;
3222 ecode++;
3223 GETCHARLEN(fc, ecode, length);
3224
3225 /* If the pattern character's value is < 128, we have only one byte, and
3226 we know that its other case must also be one byte long, so we can use the
3227 fast lookup table. We know that there is at least one byte left in the
3228 subject. */
3229
3230 if (fc < 128)
3231 {
3232 pcre_uint32 cc = RAWUCHAR(eptr);
3233 if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3234 ecode++;
3235 eptr++;
3236 }
3237
3238 /* Otherwise we must pick up the subject character. Note that we cannot
3239 use the value of "length" to check for sufficient bytes left, because the
3240 other case of the character may have more or fewer bytes. */
3241
3242 else
3243 {
3244 pcre_uint32 dc;
3245 GETCHARINC(dc, eptr);
3246 ecode += length;
3247
3248 /* If we have Unicode property support, we can use it to test the other
3249 case of the character, if there is one. */
3250
3251 if (fc != dc)
3252 {
3253 #ifdef SUPPORT_UCP
3254 if (dc != UCD_OTHERCASE(fc))
3255 #endif
3256 RRETURN(MATCH_NOMATCH);
3257 }
3258 }
3259 }
3260 else
3261 #endif /* SUPPORT_UTF */
3262
3263 /* Not UTF mode */
3264 {
3265 if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3266 != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3267 eptr++;
3268 ecode += 2;
3269 }
3270 break;
3271
3272 /* Match a single character repeatedly. */
3273
3274 case OP_EXACT:
3275 case OP_EXACTI:
3276 min = max = GET2(ecode, 1);
3277 ecode += 1 + IMM2_SIZE;
3278 goto REPEATCHAR;
3279
3280 case OP_POSUPTO:
3281 case OP_POSUPTOI:
3282 possessive = TRUE;
3283 /* Fall through */
3284
3285 case OP_UPTO:
3286 case OP_UPTOI:
3287 case OP_MINUPTO:
3288 case OP_MINUPTOI:
3289 min = 0;
3290 max = GET2(ecode, 1);
3291 minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3292 ecode += 1 + IMM2_SIZE;
3293 goto REPEATCHAR;
3294
3295 case OP_POSSTAR:
3296 case OP_POSSTARI:
3297 possessive = TRUE;
3298 min = 0;
3299 max = INT_MAX;
3300 ecode++;
3301 goto REPEATCHAR;
3302
3303 case OP_POSPLUS:
3304 case OP_POSPLUSI:
3305 possessive = TRUE;
3306 min = 1;
3307 max = INT_MAX;
3308 ecode++;
3309 goto REPEATCHAR;
3310
3311 case OP_POSQUERY:
3312 case OP_POSQUERYI:
3313 possessive = TRUE;
3314 min = 0;
3315 max = 1;
3316 ecode++;
3317 goto REPEATCHAR;
3318
3319 case OP_STAR:
3320 case OP_STARI:
3321 case OP_MINSTAR:
3322 case OP_MINSTARI:
3323 case OP_PLUS:
3324 case OP_PLUSI:
3325 case OP_MINPLUS:
3326 case OP_MINPLUSI:
3327 case OP_QUERY:
3328 case OP_QUERYI:
3329 case OP_MINQUERY:
3330 case OP_MINQUERYI:
3331 c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3332 minimize = (c & 1) != 0;
3333 min = rep_min[c]; /* Pick up values from tables; */
3334 max = rep_max[c]; /* zero for max => infinity */
3335 if (max == 0) max = INT_MAX;
3336
3337 /* Common code for all repeated single-character matches. We first check
3338 for the minimum number of characters. If the minimum equals the maximum, we
3339 are done. Otherwise, if minimizing, check the rest of the pattern for a
3340 match; if there isn't one, advance up to the maximum, one character at a
3341 time.
3342
3343 If maximizing, advance up to the maximum number of matching characters,
3344 until eptr is past the end of the maximum run. If possessive, we are
3345 then done (no backing up). Otherwise, match at this position; anything
3346 other than no match is immediately returned. For nomatch, back up one
3347 character, unless we are matching \R and the last thing matched was
3348 \r\n, in which case, back up two bytes. When we reach the first optional
3349 character position, we can save stack by doing a tail recurse.
3350
3351 The various UTF/non-UTF and caseful/caseless cases are handled separately,
3352 for speed. */
3353
3354 REPEATCHAR:
3355 #ifdef SUPPORT_UTF
3356 if (utf)
3357 {
3358 length = 1;
3359 charptr = ecode;
3360 GETCHARLEN(fc, ecode, length);
3361 ecode += length;
3362
3363 /* Handle multibyte character matching specially here. There is
3364 support for caseless matching if UCP support is present. */
3365
3366 if (length > 1)
3367 {
3368 #ifdef SUPPORT_UCP
3369 pcre_uint32 othercase;
3370 if (op >= OP_STARI && /* Caseless */
3371 (othercase = UCD_OTHERCASE(fc)) != fc)
3372 oclength = PRIV(ord2utf)(othercase, occhars);
3373 else oclength = 0;
3374 #endif /* SUPPORT_UCP */
3375
3376 for (i = 1; i <= min; i++)
3377 {
3378 if (eptr <= md->end_subject - length &&
3379 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3380 #ifdef SUPPORT_UCP
3381 else if (oclength > 0 &&
3382 eptr <= md->end_subject - oclength &&
3383 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3384 #endif /* SUPPORT_UCP */
3385 else
3386 {
3387 CHECK_PARTIAL();
3388 RRETURN(MATCH_NOMATCH);
3389 }
3390 }
3391
3392 if (min == max) continue;
3393
3394 if (minimize)
3395 {
3396 for (fi = min;; fi++)
3397 {
3398 RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3399 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3400 if (fi >= max) RRETURN(MATCH_NOMATCH);
3401 if (eptr <= md->end_subject - length &&
3402 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3403 #ifdef SUPPORT_UCP
3404 else if (oclength > 0 &&
3405 eptr <= md->end_subject - oclength &&
3406 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3407 #endif /* SUPPORT_UCP */
3408 else
3409 {
3410 CHECK_PARTIAL();
3411 RRETURN(MATCH_NOMATCH);
3412 }
3413 }
3414 /* Control never gets here */
3415 }
3416
3417 else /* Maximize */
3418 {
3419 pp = eptr;
3420 for (i = min; i < max; i++)
3421 {
3422 if (eptr <= md->end_subject - length &&
3423 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3424 #ifdef SUPPORT_UCP
3425 else if (oclength > 0 &&
3426 eptr <= md->end_subject - oclength &&
3427 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3428 #endif /* SUPPORT_UCP */
3429 else
3430 {
3431 CHECK_PARTIAL();
3432 break;
3433 }
3434 }
3435
3436 if (possessive) continue; /* No backtracking */
3437 for(;;)
3438 {
3439 if (eptr == pp) goto TAIL_RECURSE;
3440 RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3441 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3442 #ifdef SUPPORT_UCP
3443 eptr--;
3444 BACKCHAR(eptr);
3445 #else /* without SUPPORT_UCP */
3446 eptr -= length;
3447 #endif /* SUPPORT_UCP */
3448 }
3449 }
3450 /* Control never gets here */
3451 }
3452
3453 /* If the length of a UTF-8 character is 1, we fall through here, and
3454 obey the code as for non-UTF-8 characters below, though in this case the
3455 value of fc will always be < 128. */
3456 }
3457 else
3458 #endif /* SUPPORT_UTF */
3459 /* When not in UTF-8 mode, load a single-byte character. */
3460 fc = *ecode++;
3461
3462 /* The value of fc at this point is always one character, though we may
3463 or may not be in UTF mode. The code is duplicated for the caseless and
3464 caseful cases, for speed, since matching characters is likely to be quite
3465 common. First, ensure the minimum number of matches are present. If min =
3466 max, continue at the same level without recursing. Otherwise, if
3467 minimizing, keep trying the rest of the expression and advancing one
3468 matching character if failing, up to the maximum. Alternatively, if
3469 maximizing, find the maximum number of characters and work backwards. */
3470
3471 DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3472 max, (char *)eptr));
3473
3474 if (op >= OP_STARI) /* Caseless */
3475 {
3476 #ifdef COMPILE_PCRE8
3477 /* fc must be < 128 if UTF is enabled. */
3478 foc = md->fcc[fc];
3479 #else
3480 #ifdef SUPPORT_UTF
3481 #ifdef SUPPORT_UCP
3482 if (utf && fc > 127)
3483 foc = UCD_OTHERCASE(fc);
3484 #else
3485 if (utf && fc > 127)
3486 foc = fc;
3487 #endif /* SUPPORT_UCP */
3488 else
3489 #endif /* SUPPORT_UTF */
3490 foc = TABLE_GET(fc, md->fcc, fc);
3491 #endif /* COMPILE_PCRE8 */
3492
3493 for (i = 1; i <= min; i++)
3494 {
3495 pcre_uint32 cc; /* Faster than pcre_uchar */
3496 if (eptr >= md->end_subject)
3497 {
3498 SCHECK_PARTIAL();
3499 RRETURN(MATCH_NOMATCH);
3500 }
3501 cc = RAWUCHARTEST(eptr);
3502 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3503 eptr++;
3504 }
3505 if (min == max) continue;
3506 if (minimize)
3507 {
3508 for (fi = min;; fi++)
3509 {
3510 pcre_uint32 cc; /* Faster than pcre_uchar */
3511 RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3512 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3513 if (fi >= max) RRETURN(MATCH_NOMATCH);
3514 if (eptr >= md->end_subject)
3515 {
3516 SCHECK_PARTIAL();
3517 RRETURN(MATCH_NOMATCH);
3518 }
3519 cc = RAWUCHARTEST(eptr);
3520 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3521 eptr++;
3522 }
3523 /* Control never gets here */
3524 }
3525 else /* Maximize */
3526 {
3527 pp = eptr;
3528 for (i = min; i < max; i++)
3529 {
3530 pcre_uint32 cc; /* Faster than pcre_uchar */
3531 if (eptr >= md->end_subject)
3532 {
3533 SCHECK_PARTIAL();
3534 break;
3535 }
3536 cc = RAWUCHARTEST(eptr);
3537 if (fc != cc && foc != cc) break;
3538 eptr++;
3539 }
3540 if (possessive) continue; /* No backtracking */
3541 for (;;)
3542 {
3543 if (eptr == pp) goto TAIL_RECURSE;
3544 RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3545 eptr--;
3546 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3547 }
3548 /* Control never gets here */
3549 }
3550 }
3551
3552 /* Caseful comparisons (includes all multi-byte characters) */
3553
3554 else
3555 {
3556 for (i = 1; i <= min; i++)
3557 {
3558 if (eptr >= md->end_subject)
3559 {
3560 SCHECK_PARTIAL();
3561 RRETURN(MATCH_NOMATCH);
3562 }
3563 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3564 }
3565
3566 if (min == max) continue;
3567
3568 if (minimize)
3569 {
3570 for (fi = min;; fi++)
3571 {
3572 RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3573 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3574 if (fi >= max) RRETURN(MATCH_NOMATCH);
3575 if (eptr >= md->end_subject)
3576 {
3577 SCHECK_PARTIAL();
3578 RRETURN(MATCH_NOMATCH);
3579 }
3580 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3581 }
3582 /* Control never gets here */
3583 }
3584 else /* Maximize */
3585 {
3586 pp = eptr;
3587 for (i = min; i < max; i++)
3588 {
3589 if (eptr >= md->end_subject)
3590 {
3591 SCHECK_PARTIAL();
3592 break;
3593 }
3594 if (fc != RAWUCHARTEST(eptr)) break;
3595 eptr++;
3596 }
3597 if (possessive) continue; /* No backtracking */
3598 for (;;)
3599 {
3600 if (eptr == pp) goto TAIL_RECURSE;
3601 RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3602 eptr--;
3603 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3604 }
3605 /* Control never gets here */
3606 }
3607 }
3608 /* Control never gets here */
3609
3610 /* Match a negated single one-byte character. The character we are
3611 checking can be multibyte. */
3612
3613 case OP_NOT:
3614 case OP_NOTI:
3615 if (eptr >= md->end_subject)
3616 {
3617 SCHECK_PARTIAL();
3618 RRETURN(MATCH_NOMATCH);
3619 }
3620 #ifdef SUPPORT_UTF
3621 if (utf)
3622 {
3623 register pcre_uint32 ch, och;
3624
3625 ecode++;
3626 GETCHARINC(ch, ecode);
3627 GETCHARINC(c, eptr);
3628
3629 if (op == OP_NOT)
3630 {
3631 if (ch == c) RRETURN(MATCH_NOMATCH);
3632 }
3633 else
3634 {
3635 #ifdef SUPPORT_UCP
3636 if (ch > 127)
3637 och = UCD_OTHERCASE(ch);
3638 #else
3639 if (ch > 127)
3640 och = ch;
3641 #endif /* SUPPORT_UCP */
3642 else
3643 och = TABLE_GET(ch, md->fcc, ch);
3644 if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3645 }
3646 }
3647 else
3648 #endif
3649 {
3650 register pcre_uint32 ch = ecode[1];
3651 c = *eptr++;
3652 if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3653 RRETURN(MATCH_NOMATCH);
3654 ecode += 2;
3655 }
3656 break;
3657
3658 /* Match a negated single one-byte character repeatedly. This is almost a
3659 repeat of the code for a repeated single character, but I haven't found a
3660 nice way of commoning these up that doesn't require a test of the
3661 positive/negative option for each character match. Maybe that wouldn't add
3662 very much to the time taken, but character matching *is* what this is all
3663 about... */
3664
3665 case OP_NOTEXACT:
3666 case OP_NOTEXACTI:
3667 min = max = GET2(ecode, 1);
3668 ecode += 1 + IMM2_SIZE;
3669 goto REPEATNOTCHAR;
3670
3671 case OP_NOTUPTO:
3672 case OP_NOTUPTOI:
3673 case OP_NOTMINUPTO:
3674 case OP_NOTMINUPTOI:
3675 min = 0;
3676 max = GET2(ecode, 1);
3677 minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3678 ecode += 1 + IMM2_SIZE;
3679 goto REPEATNOTCHAR;
3680
3681 case OP_NOTPOSSTAR:
3682 case OP_NOTPOSSTARI:
3683 possessive = TRUE;
3684 min = 0;
3685 max = INT_MAX;
3686 ecode++;
3687 goto REPEATNOTCHAR;
3688
3689 case OP_NOTPOSPLUS:
3690 case OP_NOTPOSPLUSI:
3691 possessive = TRUE;
3692 min = 1;
3693 max = INT_MAX;
3694 ecode++;
3695 goto REPEATNOTCHAR;
3696
3697 case OP_NOTPOSQUERY:
3698 case OP_NOTPOSQUERYI:
3699 possessive = TRUE;
3700 min = 0;
3701 max = 1;
3702 ecode++;
3703 goto REPEATNOTCHAR;
3704
3705 case OP_NOTPOSUPTO:
3706 case OP_NOTPOSUPTOI:
3707 possessive = TRUE;
3708 min = 0;
3709 max = GET2(ecode, 1);
3710 ecode += 1 + IMM2_SIZE;
3711 goto REPEATNOTCHAR;
3712
3713 case OP_NOTSTAR:
3714 case OP_NOTSTARI:
3715 case OP_NOTMINSTAR:
3716 case OP_NOTMINSTARI:
3717 case OP_NOTPLUS:
3718 case OP_NOTPLUSI:
3719 case OP_NOTMINPLUS:
3720 case OP_NOTMINPLUSI:
3721 case OP_NOTQUERY:
3722 case OP_NOTQUERYI:
3723 case OP_NOTMINQUERY:
3724 case OP_NOTMINQUERYI:
3725 c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3726 minimize = (c & 1) != 0;
3727 min = rep_min[c]; /* Pick up values from tables; */
3728 max = rep_max[c]; /* zero for max => infinity */
3729 if (max == 0) max = INT_MAX;
3730
3731 /* Common code for all repeated single-byte matches. */
3732
3733 REPEATNOTCHAR:
3734 GETCHARINCTEST(fc, ecode);
3735
3736 /* The code is duplicated for the caseless and caseful cases, for speed,
3737 since matching characters is likely to be quite common. First, ensure the
3738 minimum number of matches are present. If min = max, continue at the same
3739 level without recursing. Otherwise, if minimizing, keep trying the rest of
3740 the expression and advancing one matching character if failing, up to the
3741 maximum. Alternatively, if maximizing, find the maximum number of
3742 characters and work backwards. */
3743
3744 DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3745 max, (char *)eptr));
3746
3747 if (op >= OP_NOTSTARI) /* Caseless */
3748 {
3749 #ifdef SUPPORT_UTF
3750 #ifdef SUPPORT_UCP
3751 if (utf && fc > 127)
3752 foc = UCD_OTHERCASE(fc);
3753 #else
3754 if (utf && fc > 127)
3755 foc = fc;
3756 #endif /* SUPPORT_UCP */
3757 else
3758 #endif /* SUPPORT_UTF */
3759 foc = TABLE_GET(fc, md->fcc, fc);
3760
3761 #ifdef SUPPORT_UTF
3762 if (utf)
3763 {
3764 register pcre_uint32 d;
3765 for (i = 1; i <= min; i++)
3766 {
3767 if (eptr >= md->end_subject)
3768 {
3769 SCHECK_PARTIAL();
3770 RRETURN(MATCH_NOMATCH);
3771 }
3772 GETCHARINC(d, eptr);
3773 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3774 }
3775 }
3776 else
3777 #endif /* SUPPORT_UTF */
3778 /* Not UTF mode */
3779 {
3780 for (i = 1; i <= min; i++)
3781 {
3782 if (eptr >= md->end_subject)
3783 {
3784 SCHECK_PARTIAL();
3785 RRETURN(MATCH_NOMATCH);
3786 }
3787 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3788 eptr++;
3789 }
3790 }
3791
3792 if (min == max) continue;
3793
3794 if (minimize)
3795 {
3796 #ifdef SUPPORT_UTF
3797 if (utf)
3798 {
3799 register pcre_uint32 d;
3800 for (fi = min;; fi++)
3801 {
3802 RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3803 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3804 if (fi >= max) RRETURN(MATCH_NOMATCH);
3805 if (eptr >= md->end_subject)
3806 {
3807 SCHECK_PARTIAL();
3808 RRETURN(MATCH_NOMATCH);
3809 }
3810 GETCHARINC(d, eptr);
3811 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3812 }
3813 }
3814 else
3815 #endif /*SUPPORT_UTF */
3816 /* Not UTF mode */
3817 {
3818 for (fi = min;; fi++)
3819 {
3820 RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3821 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3822 if (fi >= max) RRETURN(MATCH_NOMATCH);
3823 if (eptr >= md->end_subject)
3824 {
3825 SCHECK_PARTIAL();
3826 RRETURN(MATCH_NOMATCH);
3827 }
3828 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3829 eptr++;
3830 }
3831 }
3832 /* Control never gets here */
3833 }
3834
3835 /* Maximize case */
3836
3837 else
3838 {
3839 pp = eptr;
3840
3841 #ifdef SUPPORT_UTF
3842 if (utf)
3843 {
3844 register pcre_uint32 d;
3845 for (i = min; i < max; i++)
3846 {
3847 int len = 1;
3848 if (eptr >= md->end_subject)
3849 {
3850 SCHECK_PARTIAL();
3851 break;
3852 }
3853 GETCHARLEN(d, eptr, len);
3854 if (fc == d || (unsigned int)foc == d) break;
3855 eptr += len;
3856 }
3857 if (possessive) continue; /* No backtracking */
3858 for(;;)
3859 {
3860 if (eptr == pp) goto TAIL_RECURSE;
3861 RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3862 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3863 eptr--;
3864 BACKCHAR(eptr);
3865 }
3866 }
3867 else
3868 #endif /* SUPPORT_UTF */
3869 /* Not UTF mode */
3870 {
3871 for (i = min; i < max; i++)
3872 {
3873 if (eptr >= md->end_subject)
3874 {
3875 SCHECK_PARTIAL();
3876 break;
3877 }
3878 if (fc == *eptr || foc == *eptr) break;
3879 eptr++;
3880 }
3881 if (possessive) continue; /* No backtracking */
3882 for (;;)
3883 {
3884 if (eptr == pp) goto TAIL_RECURSE;
3885 RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3886 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3887 eptr--;
3888 }
3889 }
3890 /* Control never gets here */
3891 }
3892 }
3893
3894 /* Caseful comparisons */
3895
3896 else
3897 {
3898 #ifdef SUPPORT_UTF
3899 if (utf)
3900 {
3901 register pcre_uint32 d;
3902 for (i = 1; i <= min; i++)
3903 {
3904 if (eptr >= md->end_subject)
3905 {
3906 SCHECK_PARTIAL();
3907 RRETURN(MATCH_NOMATCH);
3908 }
3909 GETCHARINC(d, eptr);
3910 if (fc == d) RRETURN(MATCH_NOMATCH);
3911 }
3912 }
3913 else
3914 #endif
3915 /* Not UTF mode */
3916 {
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 (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3925 }
3926 }
3927
3928 if (min == max) continue;
3929
3930 if (minimize)
3931 {
3932 #ifdef SUPPORT_UTF
3933 if (utf)
3934 {
3935 register pcre_uint32 d;
3936 for (fi = min;; fi++)
3937 {
3938 RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3939 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3940 if (fi >= max) RRETURN(MATCH_NOMATCH);
3941 if (eptr >= md->end_subject)
3942 {
3943 SCHECK_PARTIAL();
3944 RRETURN(MATCH_NOMATCH);
3945 }
3946 GETCHARINC(d, eptr);
3947 if (fc == d) RRETURN(MATCH_NOMATCH);
3948 }
3949 }
3950 else
3951 #endif
3952 /* Not UTF mode */
3953 {
3954 for (fi = min;; fi++)
3955 {
3956 RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
3957 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3958 if (fi >= max) RRETURN(MATCH_NOMATCH);
3959 if (eptr >= md->end_subject)
3960 {
3961 SCHECK_PARTIAL();
3962 RRETURN(MATCH_NOMATCH);
3963 }
3964 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3965 }
3966 }
3967 /* Control never gets here */
3968 }
3969
3970 /* Maximize case */
3971
3972 else
3973 {
3974 pp = eptr;
3975
3976 #ifdef SUPPORT_UTF
3977 if (utf)
3978 {
3979 register pcre_uint32 d;
3980 for (i = min; i < max; i++)
3981 {
3982 int len = 1;
3983 if (eptr >= md->end_subject)
3984 {
3985 SCHECK_PARTIAL();
3986 break;
3987 }
3988 GETCHARLEN(d, eptr, len);
3989 if (fc == d) break;
3990 eptr += len;
3991 }
3992 if (possessive) continue; /* No backtracking */
3993 for(;;)
3994 {
3995 if (eptr == pp) goto TAIL_RECURSE;
3996 RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
3997 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3998 eptr--;
3999 BACKCHAR(eptr);
4000 }
4001 }
4002 else
4003 #endif
4004 /* Not UTF mode */
4005 {
4006 for (i = min; i < max; i++)
4007 {
4008 if (eptr >= md->end_subject)
4009 {
4010 SCHECK_PARTIAL();
4011 break;
4012 }
4013 if (fc == *eptr) break;
4014 eptr++;
4015 }
4016 if (possessive) continue; /* No backtracking */
4017 for (;;)
4018 {
4019 if (eptr == pp) goto TAIL_RECURSE;
4020 RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
4021 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4022 eptr--;
4023 }
4024 }
4025 /* Control never gets here */
4026 }
4027 }
4028 /* Control never gets here */
4029
4030 /* Match a single character type repeatedly; several different opcodes
4031 share code. This is very similar to the code for single characters, but we
4032 repeat it in the interests of efficiency. */
4033
4034 case OP_TYPEEXACT:
4035 min = max = GET2(ecode, 1);
4036 minimize = TRUE;
4037 ecode += 1 + IMM2_SIZE;
4038 goto REPEATTYPE;
4039
4040 case OP_TYPEUPTO:
4041 case OP_TYPEMINUPTO:
4042 min = 0;
4043 max = GET2(ecode, 1);
4044 minimize = *ecode == OP_TYPEMINUPTO;
4045 ecode += 1 + IMM2_SIZE;
4046 goto REPEATTYPE;
4047
4048 case OP_TYPEPOSSTAR:
4049 possessive = TRUE;
4050 min = 0;
4051 max = INT_MAX;
4052 ecode++;
4053 goto REPEATTYPE;
4054
4055 case OP_TYPEPOSPLUS:
4056 possessive = TRUE;
4057 min = 1;
4058 max = INT_MAX;
4059 ecode++;
4060 goto REPEATTYPE;
4061
4062 case OP_TYPEPOSQUERY:
4063 possessive = TRUE;
4064 min = 0;
4065 max = 1;
4066 ecode++;
4067 goto REPEATTYPE;
4068
4069 case OP_TYPEPOSUPTO:
4070 possessive = TRUE;
4071 min = 0;
4072 max = GET2(ecode, 1);
4073 ecode += 1 + IMM2_SIZE;
4074 goto REPEATTYPE;
4075
4076 case OP_TYPESTAR:
4077 case OP_TYPEMINSTAR:
4078 case OP_TYPEPLUS:
4079 case OP_TYPEMINPLUS:
4080 case OP_TYPEQUERY:
4081 case OP_TYPEMINQUERY:
4082 c = *ecode++ - OP_TYPESTAR;
4083 minimize = (c & 1) != 0;
4084 min = rep_min[c]; /* Pick up values from tables; */
4085 max = rep_max[c]; /* zero for max => infinity */
4086 if (max == 0) max = INT_MAX;
4087
4088 /* Common code for all repeated single character type matches. Note that
4089 in UTF-8 mode, '.' matches a character of any length, but for the other
4090 character types, the valid characters are all one-byte long. */
4091
4092 REPEATTYPE:
4093 ctype = *ecode++; /* Code for the character type */
4094
4095 #ifdef SUPPORT_UCP
4096 if (ctype == OP_PROP || ctype == OP_NOTPROP)
4097 {
4098 prop_fail_result = ctype == OP_NOTPROP;
4099 prop_type = *ecode++;
4100 prop_value = *ecode++;
4101 }
4102 else prop_type = -1;
4103 #endif
4104
4105 /* First, ensure the minimum number of matches are present. Use inline
4106 code for maximizing the speed, and do the type test once at the start
4107 (i.e. keep it out of the loop). Separate the UTF-8 code completely as that
4108 is tidier. Also separate the UCP code, which can be the same for both UTF-8
4109 and single-bytes. */
4110
4111 if (min > 0)
4112 {
4113 #ifdef SUPPORT_UCP
4114 if (prop_type >= 0)
4115 {
4116 switch(prop_type)
4117 {
4118 case PT_ANY:
4119 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4120 for (i = 1; i <= min; i++)
4121 {
4122 if (eptr >= md->end_subject)
4123 {
4124 SCHECK_PARTIAL();
4125 RRETURN(MATCH_NOMATCH);
4126 }
4127 GETCHARINCTEST(c, eptr);
4128 }
4129 break;
4130
4131 case PT_LAMP:
4132 for (i = 1; i <= min; i++)
4133 {
4134 int chartype;
4135 if (eptr >= md->end_subject)
4136 {
4137 SCHECK_PARTIAL();
4138 RRETURN(MATCH_NOMATCH);
4139 }
4140 GETCHARINCTEST(c, eptr);
4141 chartype = UCD_CHARTYPE(c);
4142 if ((chartype == ucp_Lu ||
4143 chartype == ucp_Ll ||
4144 chartype == ucp_Lt) == prop_fail_result)
4145 RRETURN(MATCH_NOMATCH);
4146 }
4147 break;
4148
4149 case PT_GC:
4150 for (i = 1; i <= min; i++)
4151 {
4152 if (eptr >= md->end_subject)
4153 {
4154 SCHECK_PARTIAL();
4155 RRETURN(MATCH_NOMATCH);
4156 }
4157 GETCHARINCTEST(c, eptr);
4158 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4159 RRETURN(MATCH_NOMATCH);
4160 }
4161 break;
4162
4163 case PT_PC:
4164 for (i = 1; i <= min; i++)
4165 {
4166 if (eptr >= md->end_subject)
4167 {
4168 SCHECK_PARTIAL();
4169 RRETURN(MATCH_NOMATCH);
4170 }
4171 GETCHARINCTEST(c, eptr);
4172 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4173 RRETURN(MATCH_NOMATCH);
4174 }
4175 break;
4176
4177 case PT_SC:
4178 for (i = 1; i <= min; i++)
4179 {
4180 if (eptr >= md->end_subject)
4181 {
4182 SCHECK_PARTIAL();
4183 RRETURN(MATCH_NOMATCH);
4184 }
4185 GETCHARINCTEST(c, eptr);
4186 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4187 RRETURN(MATCH_NOMATCH);
4188 }
4189 break;
4190
4191 case PT_ALNUM:
4192 for (i = 1; i <= min; i++)
4193 {
4194 int category;
4195 if (eptr >= md->end_subject)
4196 {
4197 SCHECK_PARTIAL();
4198 RRETURN(MATCH_NOMATCH);
4199 }
4200 GETCHARINCTEST(c, eptr);
4201 category = UCD_CATEGORY(c);
4202 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4203 RRETURN(MATCH_NOMATCH);
4204 }
4205 break;
4206
4207 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
4208 which means that Perl space and POSIX space are now identical. PCRE
4209 was changed at release 8.34. */
4210
4211 case PT_SPACE: /* Perl space */
4212 case PT_PXSPACE: /* POSIX space */
4213 for (i = 1; i <= min; i++)
4214 {
4215 if (eptr >= md->end_subject)
4216 {
4217 SCHECK_PARTIAL();
4218 RRETURN(MATCH_NOMATCH);
4219 }
4220 GETCHARINCTEST(c, eptr);
4221 switch(c)
4222 {
4223 HSPACE_CASES:
4224 VSPACE_CASES:
4225 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4226 break;
4227
4228 default:
4229 if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
4230 RRETURN(MATCH_NOMATCH);
4231 break;
4232 }
4233 }
4234 break;
4235
4236 case PT_WORD:
4237 for (i = 1; i <= min; i++)
4238 {
4239 int category;
4240 if (eptr >= md->end_subject)
4241 {
4242 SCHECK_PARTIAL();
4243 RRETURN(MATCH_NOMATCH);
4244 }
4245 GETCHARINCTEST(c, eptr);
4246 category = UCD_CATEGORY(c);
4247 if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
4248 == prop_fail_result)
4249 RRETURN(MATCH_NOMATCH);
4250 }
4251 break;
4252
4253 case PT_CLIST:
4254 for (i = 1; i <= min; i++)
4255 {
4256 const pcre_uint32 *cp;
4257 if (eptr >= md->end_subject)
4258 {
4259 SCHECK_PARTIAL();
4260 RRETURN(MATCH_NOMATCH);
4261 }
4262 GETCHARINCTEST(c, eptr);
4263 cp = PRIV(ucd_caseless_sets) + prop_value;
4264 for (;;)
4265 {
4266 if (c < *cp)
4267 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
4268 if (c == *cp++)
4269 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
4270 }
4271 }
4272 break;
4273
4274 case PT_UCNC:
4275 for (i = 1; i <= min; i++)
4276 {
4277 if (eptr >= md->end_subject)
4278 {
4279 SCHECK_PARTIAL();
4280 RRETURN(MATCH_NOMATCH);
4281 }
4282 GETCHARINCTEST(c, eptr);
4283 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
4284 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
4285 c >= 0xe000) == prop_fail_result)
4286 RRETURN(MATCH_NOMATCH);
4287 }
4288 break;
4289
4290 /* This should not occur */
4291
4292 default:
4293 RRETURN(PCRE_ERROR_INTERNAL);
4294 }
4295 }
4296
4297 /* Match extended Unicode sequences. We will get here only if the
4298 support is in the binary; otherwise a compile-time error occurs. */
4299
4300 else if (ctype == OP_EXTUNI)
4301 {
4302 for (i = 1; i <= min; i++)
4303 {
4304 if (eptr >= md->end_subject)
4305 {
4306 SCHECK_PARTIAL();
4307 RRETURN(MATCH_NOMATCH);
4308 }
4309 else
4310 {
4311 int lgb, rgb;
4312 GETCHARINCTEST(c, eptr);
4313 lgb = UCD_GRAPHBREAK(c);
4314 while (eptr < md->end_subject)
4315 {
4316 int len = 1;
4317 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4318 rgb = UCD_GRAPHBREAK(c);
4319 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
4320 lgb = rgb;
4321 eptr += len;
4322 }
4323 }
4324 CHECK_PARTIAL();
4325 }
4326 }
4327
4328 else
4329 #endif /* SUPPORT_UCP */
4330
4331 /* Handle all other cases when the coding is UTF-8 */
4332
4333 #ifdef SUPPORT_UTF
4334 if (utf) switch(ctype)
4335 {
4336 case OP_ANY:
4337 for (i = 1; i <= min; i++)
4338 {
4339 if (eptr >= md->end_subject)
4340 {
4341 SCHECK_PARTIAL();
4342 RRETURN(MATCH_NOMATCH);
4343 }
4344 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4345 if (md->partial != 0 &&
4346 eptr + 1 >= md->end_subject &&
4347 NLBLOCK->nltype == NLTYPE_FIXED &&
4348 NLBLOCK->nllen == 2 &&
4349 RAWUCHAR(eptr) == NLBLOCK->nl[0])
4350 {
4351 md->hitend = TRUE;
4352 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4353 }
4354 eptr++;
4355 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4356 }
4357 break;
4358
4359 case OP_ALLANY:
4360 for (i = 1; i <= min; i++)
4361 {
4362 if (eptr >= md->end_subject)
4363 {
4364 SCHECK_PARTIAL();
4365 RRETURN(MATCH_NOMATCH);
4366 }
4367 eptr++;
4368 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4369 }
4370 break;
4371
4372 case OP_ANYBYTE:
4373 if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH);
4374 eptr += min;
4375 break;
4376
4377 case OP_ANYNL:
4378 for (i = 1; i <= min; i++)
4379 {
4380 if (eptr >= md->end_subject)
4381 {
4382 SCHECK_PARTIAL();
4383 RRETURN(MATCH_NOMATCH);
4384 }
4385 GETCHARINC(c, eptr);
4386 switch(c)
4387 {
4388 default: RRETURN(MATCH_NOMATCH);
4389
4390 case CHAR_CR:
4391 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
4392 break;
4393
4394 case CHAR_LF:
4395 break;
4396
4397 case CHAR_VT:
4398 case CHAR_FF:
4399 case CHAR_NEL:
4400 #ifndef EBCDIC
4401 case 0x2028:
4402 case 0x2029:
4403 #endif /* Not EBCDIC */
4404 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4405 break;
4406 }
4407 }
4408 break;
4409
4410 case OP_NOT_HSPACE:
4411 for (i = 1; i <= min; i++)
4412 {
4413 if (eptr >= md->end_subject)
4414 {
4415 SCHECK_PARTIAL();
4416 RRETURN(MATCH_NOMATCH);
4417 }
4418 GETCHARINC(c, eptr);
4419 switch(c)
4420 {
4421 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
4422 default: break;
4423 }
4424 }
4425 break;
4426
4427 case OP_HSPACE:
4428 for (i = 1; i <= min; i++)
4429 {
4430 if (eptr >= md->end_subject)
4431 {
4432 SCHECK_PARTIAL();
4433 RRETURN(MATCH_NOMATCH);
4434 }
4435 GETCHARINC(c, eptr);
4436 switch(c)
4437 {
4438 HSPACE_CASES: break; /* Byte and multibyte cases */
4439 default: RRETURN(MATCH_NOMATCH);
4440 }
4441 }
4442 break;
4443
4444 case OP_NOT_VSPACE:
4445 for (i = 1; i <= min; i++)
4446 {
4447 if (eptr >= md->end_subject)
4448 {
4449 SCHECK_PARTIAL();
4450 RRETURN(MATCH_NOMATCH);
4451 }
4452 GETCHARINC(c, eptr);
4453 switch(c)
4454 {
4455 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
4456 default: break;
4457 }
4458 }
4459 break;
4460
4461 case OP_VSPACE:
4462 for (i = 1; i <= min; i++)
4463 {
4464 if (eptr >= md->end_subject)
4465 {
4466 SCHECK_PARTIAL();
4467 RRETURN(MATCH_NOMATCH);
4468 }
4469 GETCHARINC(c, eptr);
4470 switch(c)
4471 {
4472 VSPACE_CASES: break;
4473 default: RRETURN(MATCH_NOMATCH);
4474 }
4475 }
4476 break;
4477
4478 case OP_NOT_DIGIT:
4479 for (i = 1; i <= min; i++)
4480 {
4481 if (eptr >= md->end_subject)
4482 {
4483 SCHECK_PARTIAL();
4484 RRETURN(MATCH_NOMATCH);
4485 }
4486 GETCHARINC(c, eptr);
4487 if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
4488 RRETURN(MATCH_NOMATCH);
4489 }
4490 break;
4491
4492 case OP_DIGIT:
4493 for (i = 1; i <= min; i++)
4494 {
4495 pcre_uint32 cc;
4496 if (eptr >= md->end_subject)
4497 {
4498 SCHECK_PARTIAL();
4499 RRETURN(MATCH_NOMATCH);
4500 }
4501 cc = RAWUCHAR(eptr);
4502 if (cc >= 128 || (md->ctypes[cc] & ctype_digit) == 0)
4503 RRETURN(MATCH_NOMATCH);
4504 eptr++;
4505 /* No need to skip more bytes - we know it's a 1-byte character */
4506 }
4507 break;
4508
4509 case OP_NOT_WHITESPACE:
4510 for (i = 1; i <= min; i++)
4511 {
4512 pcre_uint32 cc;
4513 if (eptr >= md->end_subject)
4514 {
4515 SCHECK_PARTIAL();
4516 RRETURN(MATCH_NOMATCH);
4517 }
4518 cc = RAWUCHAR(eptr);
4519 if (cc < 128 && (md->ctypes[cc] & ctype_space) != 0)
4520 RRETURN(MATCH_NOMATCH);
4521 eptr++;
4522 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4523 }
4524 break;
4525
4526 case OP_WHITESPACE:
4527 for (i = 1; i <= min; i++)
4528 {
4529 pcre_uint32 cc;
4530 if (eptr >= md->end_subject)
4531 {
4532 SCHECK_PARTIAL();
4533 RRETURN(MATCH_NOMATCH);
4534 }
4535 cc = RAWUCHAR(eptr);
4536 if (cc >= 128 || (md->ctypes[cc] & ctype_space) == 0)
4537 RRETURN(MATCH_NOMATCH);
4538 eptr++;
4539 /* No need to skip more bytes - we know it's a 1-byte character */
4540 }
4541 break;
4542
4543 case OP_NOT_WORDCHAR:
4544 for (i = 1; i <= min; i++)
4545 {
4546 pcre_uint32 cc;
4547 if (eptr >= md->end_subject)
4548 {
4549 SCHECK_PARTIAL();
4550 RRETURN(MATCH_NOMATCH);
4551 }
4552 cc = RAWUCHAR(eptr);
4553 if (cc < 128 && (md->ctypes[cc] & ctype_word) != 0)
4554 RRETURN(MATCH_NOMATCH);
4555 eptr++;
4556 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4557 }
4558 break;
4559
4560 case OP_WORDCHAR:
4561 for (i = 1; i <= min; i++)
4562 {
4563 pcre_uint32 cc;
4564 if (eptr >= md->end_subject)
4565 {
4566 SCHECK_PARTIAL();
4567 RRETURN(MATCH_NOMATCH);
4568 }
4569 cc = RAWUCHAR(eptr);
4570 if (cc >= 128 || (md->ctypes[cc] & ctype_word) == 0)
4571 RRETURN(MATCH_NOMATCH);
4572 eptr++;
4573 /* No need to skip more bytes - we know it's a 1-byte character */
4574 }
4575 break;
4576
4577 default:
4578 RRETURN(PCRE_ERROR_INTERNAL);
4579 } /* End switch(ctype) */
4580
4581 else
4582 #endif /* SUPPORT_UTF */
4583
4584 /* Code for the non-UTF-8 case for minimum matching of operators other
4585 than OP_PROP and OP_NOTPROP. */
4586
4587 switch(ctype)
4588 {
4589 case OP_ANY:
4590 for (i = 1; i <= min; i++)
4591 {
4592 if (eptr >= md->end_subject)
4593 {
4594 SCHECK_PARTIAL();
4595 RRETURN(MATCH_NOMATCH);
4596 }
4597 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4598 if (md->partial != 0 &&
4599 eptr + 1 >= md->end_subject &&
4600 NLBLOCK->nltype == NLTYPE_FIXED &&
4601 NLBLOCK->nllen == 2 &&
4602 *eptr == NLBLOCK->nl[0])
4603 {
4604 md->hitend = TRUE;
4605 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4606 }
4607 eptr++;
4608 }
4609 break;
4610
4611 case OP_ALLANY:
4612 if (eptr > md->end_subject - min)
4613 {
4614 SCHECK_PARTIAL();
4615 RRETURN(MATCH_NOMATCH);
4616 }
4617 eptr += min;
4618 break;
4619
4620 case OP_ANYBYTE:
4621 if (eptr > md->end_subject - min)
4622 {
4623 SCHECK_PARTIAL();
4624 RRETURN(MATCH_NOMATCH);
4625 }
4626 eptr += min;
4627 break;
4628
4629 case OP_ANYNL:
4630 for (i = 1; i <= min; i++)
4631 {
4632 if (eptr >= md->end_subject)
4633 {
4634 SCHECK_PARTIAL();
4635 RRETURN(MATCH_NOMATCH);
4636 }
4637 switch(*eptr++)
4638 {
4639 default: RRETURN(MATCH_NOMATCH);
4640
4641 case CHAR_CR:
4642 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
4643 break;
4644
4645 case CHAR_LF:
4646 break;
4647
4648 case CHAR_VT:
4649 case CHAR_FF:
4650 case CHAR_NEL:
4651 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4652 case 0x2028:
4653 case 0x2029:
4654 #endif
4655 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4656 break;
4657 }
4658 }
4659 break;
4660
4661 case OP_NOT_HSPACE:
4662 for (i = 1; i <= min; i++)
4663 {
4664 if (eptr >= md->end_subject)
4665 {
4666 SCHECK_PARTIAL();
4667 RRETURN(MATCH_NOMATCH);
4668 }
4669 switch(*eptr++)
4670 {
4671 default: break;
4672 HSPACE_BYTE_CASES:
4673 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4674 HSPACE_MULTIBYTE_CASES:
4675 #endif
4676 RRETURN(MATCH_NOMATCH);
4677 }
4678 }
4679 break;
4680
4681 case OP_HSPACE:
4682 for (i = 1; i <= min; i++)
4683 {
4684 if (eptr >= md->end_subject)
4685 {
4686 SCHECK_PARTIAL();
4687 RRETURN(MATCH_NOMATCH);
4688 }
4689 switch(*eptr++)
4690 {
4691 default: RRETURN(MATCH_NOMATCH);
4692 HSPACE_BYTE_CASES:
4693 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4694 HSPACE_MULTIBYTE_CASES:
4695 #endif
4696 break;
4697 }
4698 }
4699 break;
4700
4701 case OP_NOT_VSPACE:
4702 for (i = 1; i <= min; i++)
4703 {
4704 if (eptr >= md->end_subject)
4705 {
4706 SCHECK_PARTIAL();
4707 RRETURN(MATCH_NOMATCH);
4708 }
4709 switch(*eptr++)
4710 {
4711 VSPACE_BYTE_CASES:
4712 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4713 VSPACE_MULTIBYTE_CASES:
4714 #endif
4715 RRETURN(MATCH_NOMATCH);
4716 default: break;
4717 }
4718 }
4719 break;
4720
4721 case OP_VSPACE:
4722 for (i = 1; i <= min; i++)
4723 {
4724 if (eptr >= md->end_subject)
4725 {
4726 SCHECK_PARTIAL();
4727 RRETURN(MATCH_NOMATCH);
4728 }
4729 switch(*eptr++)
4730 {
4731 default: RRETURN(MATCH_NOMATCH);
4732 VSPACE_BYTE_CASES:
4733 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4734 VSPACE_MULTIBYTE_CASES:
4735 #endif
4736 break;
4737 }
4738 }
4739 break;
4740
4741 case OP_NOT_DIGIT:
4742 for (i = 1; i <= min; i++)
4743 {
4744 if (eptr >= md->end_subject)
4745 {
4746 SCHECK_PARTIAL();
4747 RRETURN(MATCH_NOMATCH);
4748 }
4749 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0)
4750 RRETURN(MATCH_NOMATCH);
4751 eptr++;
4752 }
4753 break;
4754
4755 case OP_DIGIT:
4756 for (i = 1; i <= min; i++)
4757 {
4758 if (eptr >= md->end_subject)
4759 {
4760 SCHECK_PARTIAL();
4761 RRETURN(MATCH_NOMATCH);
4762 }
4763 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0)
4764 RRETURN(MATCH_NOMATCH);
4765 eptr++;
4766 }
4767 break;
4768
4769 case OP_NOT_WHITESPACE:
4770 for (i = 1; i <= min; i++)
4771 {
4772 if (eptr >= md->end_subject)
4773 {
4774 SCHECK_PARTIAL();
4775 RRETURN(MATCH_NOMATCH);
4776 }
4777 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0)
4778 RRETURN(MATCH_NOMATCH);
4779 eptr++;
4780 }
4781 break;
4782
4783 case OP_WHITESPACE:
4784 for (i = 1; i <= min; i++)
4785 {
4786 if (eptr >= md->end_subject)
4787 {
4788 SCHECK_PARTIAL();
4789 RRETURN(MATCH_NOMATCH);
4790 }
4791 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0)
4792 RRETURN(MATCH_NOMATCH);
4793 eptr++;
4794 }
4795 break;
4796
4797 case OP_NOT_WORDCHAR:
4798 for (i = 1; i <= min; i++)
4799 {
4800 if (eptr >= md->end_subject)
4801 {
4802 SCHECK_PARTIAL();
4803 RRETURN(MATCH_NOMATCH);
4804 }
4805 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0)
4806 RRETURN(MATCH_NOMATCH);
4807 eptr++;
4808 }
4809 break;
4810
4811 case OP_WORDCHAR:
4812 for (i = 1; i <= min; i++)
4813 {
4814 if (eptr >= md->end_subject)
4815 {
4816 SCHECK_PARTIAL();
4817 RRETURN(MATCH_NOMATCH);
4818 }
4819 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0)
4820 RRETURN(MATCH_NOMATCH);
4821 eptr++;
4822 }
4823 break;
4824
4825 default:
4826 RRETURN(PCRE_ERROR_INTERNAL);
4827 }
4828 }
4829
4830 /* If min = max, continue at the same level without recursing */
4831
4832 if (min == max) continue;
4833
4834 /* If minimizing, we have to test the rest of the pattern before each
4835 subsequent match. Again, separate the UTF-8 case for speed, and also
4836 separate the UCP cases. */
4837
4838 if (minimize)
4839 {
4840 #ifdef SUPPORT_UCP
4841 if (prop_type >= 0)
4842 {
4843 switch(prop_type)
4844 {
4845 case PT_ANY:
4846 for (fi = min;; fi++)
4847 {
4848 RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4849 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4850 if (fi >= max) RRETURN(MATCH_NOMATCH);
4851 if (eptr >= md->end_subject)
4852 {
4853 SCHECK_PARTIAL();
4854 RRETURN(MATCH_NOMATCH);
4855 }
4856 GETCHARINCTEST(c, eptr);
4857 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4858 }
4859 /* Control never gets here */
4860
4861 case PT_LAMP:
4862 for (fi = min;; fi++)
4863 {
4864 int chartype;
4865 RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4866 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4867 if (fi >= max) RRETURN(MATCH_NOMATCH);
4868 if (eptr >= md->end_subject)
4869 {
4870 SCHECK_PARTIAL();
4871 RRETURN(MATCH_NOMATCH);
4872 }
4873 GETCHARINCTEST(c, eptr);
4874 chartype = UCD_CHARTYPE(c);
4875 if ((chartype == ucp_Lu ||
4876 chartype == ucp_Ll ||
4877 chartype == ucp_Lt) == prop_fail_result)
4878 RRETURN(MATCH_NOMATCH);
4879 }
4880 /* Control never gets here */
4881
4882 case PT_GC:
4883 for (fi = min;; fi++)
4884 {
4885 RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4886 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4887 if (fi >= max) RRETURN(MATCH_NOMATCH);
4888 if (eptr >= md->end_subject)
4889 {
4890 SCHECK_PARTIAL();
4891 RRETURN(MATCH_NOMATCH);
4892 }
4893 GETCHARINCTEST(c, eptr);
4894 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4895 RRETURN(MATCH_NOMATCH);
4896 }
4897 /* Control never gets here */
4898
4899 case PT_PC:
4900 for (fi = min;; fi++)
4901 {
4902 RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4903 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4904 if (fi >= max) RRETURN(MATCH_NOMATCH);
4905 if (eptr >= md->end_subject)
4906 {
4907 SCHECK_PARTIAL();
4908 RRETURN(MATCH_NOMATCH);
4909 }
4910 GETCHARINCTEST(c, eptr);
4911 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4912 RRETURN(MATCH_NOMATCH);
4913 }
4914 /* Control never gets here */
4915
4916 case PT_SC:
4917 for (fi = min;; fi++)
4918 {
4919 RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
4920 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4921 if (fi >= max) RRETURN(MATCH_NOMATCH);
4922 if (eptr >= md->end_subject)
4923 {
4924 SCHECK_PARTIAL();
4925 RRETURN(MATCH_NOMATCH);
4926 }
4927 GETCHARINCTEST(c, eptr);
4928 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4929 RRETURN(MATCH_NOMATCH);
4930 }
4931 /* Control never gets here */
4932
4933 case PT_ALNUM:
4934 for (fi = min;; fi++)
4935 {
4936 int category;
4937 RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4938 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4939 if (fi >= max) RRETURN(MATCH_NOMATCH);
4940 if (eptr >= md->end_subject)
4941 {
4942 SCHECK_PARTIAL();
4943 RRETURN(MATCH_NOMATCH);
4944 }
4945 GETCHARINCTEST(c, eptr);
4946 category = UCD_CATEGORY(c);
4947 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4948 RRETURN(MATCH_NOMATCH);
4949 }
4950 /* Control never gets here */
4951
4952 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
4953 which means that Perl space and POSIX space are now identical. PCRE
4954 was changed at release 8.34. */
4955
4956 case PT_SPACE: /* Perl space */
4957 case PT_PXSPACE: /* POSIX space */
4958 for (fi = min;; fi++)
4959 {
4960 RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
4961 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4962 if (fi >= max) RRETURN(MATCH_NOMATCH);
4963 if (eptr >= md->end_subject)
4964 {
4965 SCHECK_PARTIAL();
4966 RRETURN(MATCH_NOMATCH);
4967 }
4968 GETCHARINCTEST(c, eptr);
4969 switch(c)
4970 {
4971 HSPACE_CASES:
4972 VSPACE_CASES:
4973 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4974 break;
4975
4976 default:
4977 if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
4978 RRETURN(MATCH_NOMATCH);
4979 break;
4980 }
4981 }
4982 /* Control never gets here */
4983
4984 case PT_WORD:
4985 for (fi = min;; fi++)
4986 {
4987 int category;
4988 RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4989 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4990 if (fi >= max) RRETURN(MATCH_NOMATCH);
4991 if (eptr >= md->end_subject)
4992 {
4993 SCHECK_PARTIAL();
4994 RRETURN(MATCH_NOMATCH);
4995 }
4996 GETCHARINCTEST(c, eptr);
4997 category = UCD_CATEGORY(c);
4998 if ((category == ucp_L ||
4999 category == ucp_N ||
5000 c == CHAR_UNDERSCORE)
5001 == prop_fail_result)
5002 RRETURN(MATCH_NOMATCH);
5003 }
5004 /* Control never gets here */
5005
5006 case PT_CLIST:
5007 for (fi = min;; fi++)
5008 {
5009 const pcre_uint32 *cp;
5010 RMATCH(eptr, ecode, offset_top, md, eptrb, RM67);
5011 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5012 if (fi >= max) RRETURN(MATCH_NOMATCH);
5013 if (eptr >= md->end_subject)
5014 {
5015 SCHECK_PARTIAL();
5016 RRETURN(MATCH_NOMATCH);
5017 }
5018 GETCHARINCTEST(c, eptr);
5019 cp = PRIV(ucd_caseless_sets) + prop_value;
5020 for (;;)
5021 {
5022 if (c < *cp)
5023 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
5024 if (c == *cp++)
5025 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
5026 }
5027 }
5028 /* Control never gets here */
5029
5030 case PT_UCNC:
5031 for (fi = min;; fi++)
5032 {
5033 RMATCH(eptr, ecode, offset_top, md, eptrb, RM68);
5034 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5035 if (fi >= max) RRETURN(MATCH_NOMATCH);
5036 if (eptr >= md->end_subject)
5037 {
5038 SCHECK_PARTIAL();
5039 RRETURN(MATCH_NOMATCH);
5040 }
5041 GETCHARINCTEST(c, eptr);
5042 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5043 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5044 c >= 0xe000) == prop_fail_result)
5045 RRETURN(MATCH_NOMATCH);
5046 }
5047 /* Control never gets here */
5048
5049 /* This should never occur */
5050 default:
5051 RRETURN(PCRE_ERROR_INTERNAL);
5052 }
5053 }
5054
5055 /* Match extended Unicode sequences. We will get here only if the
5056 support is in the binary; otherwise a compile-time error occurs. */
5057
5058 else if (ctype == OP_EXTUNI)
5059 {
5060 for (fi = min;; fi++)
5061 {
5062 RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
5063 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5064 if (fi >= max) RRETURN(MATCH_NOMATCH);
5065 if (eptr >= md->end_subject)
5066 {
5067 SCHECK_PARTIAL();
5068 RRETURN(MATCH_NOMATCH);
5069 }
5070 else
5071 {
5072 int lgb, rgb;
5073 GETCHARINCTEST(c, eptr);
5074 lgb = UCD_GRAPHBREAK(c);
5075 while (eptr < md->end_subject)
5076 {
5077 int len = 1;
5078 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5079 rgb = UCD_GRAPHBREAK(c);
5080 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5081 lgb = rgb;
5082 eptr += len;
5083 }
5084 }
5085 CHECK_PARTIAL();
5086 }
5087 }
5088 else
5089 #endif /* SUPPORT_UCP */
5090
5091 #ifdef SUPPORT_UTF
5092 if (utf)
5093 {
5094 for (fi = min;; fi++)
5095 {
5096 RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
5097 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5098 if (fi >= max) RRETURN(MATCH_NOMATCH);
5099 if (eptr >= md->end_subject)
5100 {
5101 SCHECK_PARTIAL();
5102 RRETURN(MATCH_NOMATCH);
5103 }
5104 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5105 RRETURN(MATCH_NOMATCH);
5106 GETCHARINC(c, eptr);
5107 switch(ctype)
5108 {
5109 case OP_ANY: /* This is the non-NL case */
5110 if (md->partial != 0 && /* Take care with CRLF partial */
5111 eptr >= md->end_subject &&
5112 NLBLOCK->nltype == NLTYPE_FIXED &&
5113 NLBLOCK->nllen == 2 &&
5114 c == NLBLOCK->nl[0])
5115 {
5116 md->hitend = TRUE;
5117 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5118 }
5119 break;
5120
5121 case OP_ALLANY:
5122 case OP_ANYBYTE:
5123 break;
5124
5125 case OP_ANYNL:
5126 switch(c)
5127 {
5128 default: RRETURN(MATCH_NOMATCH);
5129 case CHAR_CR:
5130 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
5131 break;
5132
5133 case CHAR_LF:
5134 break;
5135
5136 case CHAR_VT:
5137 case CHAR_FF:
5138 case CHAR_NEL:
5139 #ifndef EBCDIC
5140 case 0x2028:
5141 case 0x2029:
5142 #endif /* Not EBCDIC */
5143 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5144 break;
5145 }
5146 break;
5147
5148 case OP_NOT_HSPACE:
5149 switch(c)
5150 {
5151 HSPACE_CASES: RRETURN(MATCH_NOMATCH);
5152 default: break;
5153 }
5154 break;
5155
5156 case OP_HSPACE:
5157 switch(c)
5158 {
5159 HSPACE_CASES: break;
5160 default: RRETURN(MATCH_NOMATCH);
5161 }
5162 break;
5163
5164 case OP_NOT_VSPACE:
5165 switch(c)
5166 {
5167 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
5168 default: break;
5169 }
5170 break;
5171
5172 case OP_VSPACE:
5173 switch(c)
5174 {
5175 VSPACE_CASES: break;
5176 default: RRETURN(MATCH_NOMATCH);
5177 }
5178 break;
5179
5180 case OP_NOT_DIGIT:
5181 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
5182 RRETURN(MATCH_NOMATCH);
5183 break;
5184
5185 case OP_DIGIT:
5186 if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0)
5187 RRETURN(MATCH_NOMATCH);
5188 break;
5189
5190 case OP_NOT_WHITESPACE:
5191 if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
5192 RRETURN(MATCH_NOMATCH);
5193 break;
5194
5195 case OP_WHITESPACE:
5196 if (c >= 256 || (md->ctypes[c] & ctype_space) == 0)
5197 RRETURN(MATCH_NOMATCH);
5198 break;
5199
5200 case OP_NOT_WORDCHAR:
5201 if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
5202 RRETURN(MATCH_NOMATCH);
5203 break;
5204
5205 case OP_WORDCHAR:
5206 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0)
5207 RRETURN(MATCH_NOMATCH);
5208 break;
5209
5210 default:
5211 RRETURN(PCRE_ERROR_INTERNAL);
5212 }
5213 }
5214 }
5215 else
5216 #endif
5217 /* Not UTF mode */
5218 {
5219 for (fi = min;; fi++)
5220 {
5221 RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
5222 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5223 if (fi >= max) RRETURN(MATCH_NOMATCH);
5224 if (eptr >= md->end_subject)
5225 {
5226 SCHECK_PARTIAL();
5227 RRETURN(MATCH_NOMATCH);
5228 }
5229 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5230 RRETURN(MATCH_NOMATCH);
5231 c = *eptr++;
5232 switch(ctype)
5233 {
5234 case OP_ANY: /* This is the non-NL case */
5235 if (md->partial != 0 && /* Take care with CRLF partial */
5236 eptr >= md->end_subject &&
5237 NLBLOCK->nltype == NLTYPE_FIXED &&
5238 NLBLOCK->nllen == 2 &&
5239 c == NLBLOCK->nl[0])
5240 {
5241 md->hitend = TRUE;
5242 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5243 }
5244 break;
5245
5246 case OP_ALLANY:
5247 case OP_ANYBYTE:
5248 break;
5249
5250 case OP_ANYNL:
5251 switch(c)
5252 {
5253 default: RRETURN(MATCH_NOMATCH);
5254 case CHAR_CR:
5255 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
5256 break;
5257
5258 case CHAR_LF:
5259 break;
5260
5261 case CHAR_VT:
5262 case CHAR_FF:
5263 case CHAR_NEL:
5264 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5265 case 0x2028:
5266 case 0x2029:
5267 #endif
5268 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5269 break;
5270 }
5271 break;
5272
5273 case OP_NOT_HSPACE:
5274 switch(c)
5275 {
5276 default: break;
5277 HSPACE_BYTE_CASES:
5278 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5279 HSPACE_MULTIBYTE_CASES:
5280 #endif
5281 RRETURN(MATCH_NOMATCH);
5282 }
5283 break;
5284
5285 case OP_HSPACE:
5286 switch(c)
5287 {
5288 default: RRETURN(MATCH_NOMATCH);
5289 HSPACE_BYTE_CASES:
5290 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5291 HSPACE_MULTIBYTE_CASES:
5292 #endif
5293 break;
5294 }
5295 break;
5296
5297 case OP_NOT_VSPACE:
5298 switch(c)
5299 {
5300 default: break;
5301 VSPACE_BYTE_CASES:
5302 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5303 VSPACE_MULTIBYTE_CASES:
5304 #endif
5305 RRETURN(MATCH_NOMATCH);
5306 }
5307 break;
5308
5309 case OP_VSPACE:
5310 switch(c)
5311 {
5312 default: RRETURN(MATCH_NOMATCH);
5313 VSPACE_BYTE_CASES:
5314 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5315 VSPACE_MULTIBYTE_CASES:
5316 #endif
5317 break;
5318 }
5319 break;
5320
5321 case OP_NOT_DIGIT:
5322 if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
5323 break;
5324
5325 case OP_DIGIT:
5326 if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
5327 break;
5328
5329 case OP_NOT_WHITESPACE:
5330 if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
5331 break;
5332
5333 case OP_WHITESPACE:
5334 if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
5335 break;
5336
5337 case OP_NOT_WORDCHAR:
5338 if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH);
5339 break;
5340
5341 case OP_WORDCHAR:
5342 if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH);
5343 break;
5344
5345 default:
5346 RRETURN(PCRE_ERROR_INTERNAL);
5347 }
5348 }
5349 }
5350 /* Control never gets here */
5351 }
5352
5353 /* If maximizing, it is worth using inline code for speed, doing the type
5354 test once at the start (i.e. keep it out of the loop). Again, keep the
5355 UTF-8 and UCP stuff separate. */
5356
5357 else
5358 {
5359 pp = eptr; /* Remember where we started */
5360
5361 #ifdef SUPPORT_UCP
5362 if (prop_type >= 0)
5363 {
5364 switch(prop_type)
5365 {
5366 case PT_ANY:
5367 for (i = min; i < max; i++)
5368 {
5369 int len = 1;
5370 if (eptr >= md->end_subject)
5371 {
5372 SCHECK_PARTIAL();
5373 break;
5374 }
5375 GETCHARLENTEST(c, eptr, len);
5376 if (prop_fail_result) break;
5377 eptr+= len;
5378 }
5379 break;
5380
5381 case PT_LAMP:
5382 for (i = min; i < max; i++)
5383 {
5384 int chartype;
5385 int len = 1;
5386 if (eptr >= md->end_subject)
5387 {
5388 SCHECK_PARTIAL();
5389 break;
5390 }
5391 GETCHARLENTEST(c, eptr, len);
5392 chartype = UCD_CHARTYPE(c);
5393 if ((chartype == ucp_Lu ||
5394 chartype == ucp_Ll ||
5395 chartype == ucp_Lt) == prop_fail_result)
5396 break;
5397 eptr+= len;
5398 }
5399 break;
5400
5401 case PT_GC:
5402 for (i = min; i < max; i++)
5403 {
5404 int len = 1;
5405 if (eptr >= md->end_subject)
5406 {
5407 SCHECK_PARTIAL();
5408 break;
5409 }
5410 GETCHARLENTEST(c, eptr, len);
5411 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
5412 eptr+= len;
5413 }
5414 break;
5415
5416 case PT_PC:
5417 for (i = min; i < max; i++)
5418 {
5419 int len = 1;
5420 if (eptr >= md->end_subject)
5421 {
5422 SCHECK_PARTIAL();
5423 break;
5424 }
5425 GETCHARLENTEST(c, eptr, len);
5426 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
5427 eptr+= len;
5428 }
5429 break;
5430
5431 case PT_SC:
5432 for (i = min; i < max; i++)
5433 {
5434 int len = 1;
5435 if (eptr >= md->end_subject)
5436 {
5437 SCHECK_PARTIAL();
5438 break;
5439 }
5440 GETCHARLENTEST(c, eptr, len);
5441 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
5442 eptr+= len;
5443 }
5444 break;
5445
5446 case PT_ALNUM:
5447 for (i = min; i < max; i++)
5448 {
5449 int category;
5450 int len = 1;
5451 if (eptr >= md->end_subject)
5452 {
5453 SCHECK_PARTIAL();
5454 break;
5455 }
5456 GETCHARLENTEST(c, eptr, len);
5457 category = UCD_CATEGORY(c);
5458 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5459 break;
5460 eptr+= len;
5461 }
5462 break;
5463
5464 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
5465 which means that Perl space and POSIX space are now identical. PCRE
5466 was changed at release 8.34. */
5467
5468 case PT_SPACE: /* Perl space */
5469 case PT_PXSPACE: /* POSIX space */
5470 for (i = min; i < max; i++)
5471 {
5472 int len = 1;
5473 if (eptr >= md->end_subject)
5474 {
5475 SCHECK_PARTIAL();
5476 break;
5477 }
5478 GETCHARLENTEST(c, eptr, len);
5479 switch(c)
5480 {
5481 HSPACE_CASES:
5482 VSPACE_CASES:
5483 if (prop_fail_result) goto ENDLOOP99; /* Break the loop */
5484 break;
5485
5486 default:
5487 if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
5488 goto ENDLOOP99; /* Break the loop */
5489 break;
5490 }
5491 eptr+= len;
5492 }
5493 ENDLOOP99:
5494 break;
5495
5496 case PT_WORD:
5497 for (i = min; i < max; i++)
5498 {
5499 int category;
5500 int len = 1;
5501 if (eptr >= md->end_subject)
5502 {
5503 SCHECK_PARTIAL();
5504 break;
5505 }
5506 GETCHARLENTEST(c, eptr, len);
5507 category = UCD_CATEGORY(c);
5508 if ((category == ucp_L || category == ucp_N ||
5509 c == CHAR_UNDERSCORE) == prop_fail_result)
5510 break;
5511 eptr+= len;
5512 }
5513 break;
5514
5515 case PT_CLIST:
5516 for (i = min; i < max; i++)
5517 {
5518 const pcre_uint32 *cp;
5519 int len = 1;
5520 if (eptr >= md->end_subject)
5521 {
5522 SCHECK_PARTIAL();
5523 break;
5524 }
5525 GETCHARLENTEST(c, eptr, len);
5526 cp = PRIV(ucd_caseless_sets) + prop_value;
5527 for (;;)
5528 {
5529 if (c < *cp)
5530 { if (prop_fail_result) break; else goto GOT_MAX; }
5531 if (c == *cp++)
5532 { if (prop_fail_result) goto GOT_MAX; else break; }
5533 }
5534 eptr += len;
5535 }
5536 GOT_MAX:
5537 break;
5538
5539 case PT_UCNC:
5540 for (i = min; i < max; i++)
5541 {
5542 int len = 1;
5543 if (eptr >= md->end_subject)
5544 {
5545 SCHECK_PARTIAL();
5546 break;
5547 }
5548 GETCHARLENTEST(c, eptr, len);
5549 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5550 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5551 c >= 0xe000) == prop_fail_result)
5552 break;
5553 eptr += len;
5554 }
5555 break;
5556
5557 default:
5558 RRETURN(PCRE_ERROR_INTERNAL);
5559 }
5560
5561 /* eptr is now past the end of the maximum run */
5562
5563 if (possessive) continue; /* No backtracking */
5564 for(;;)
5565 {
5566 if (eptr == pp) goto TAIL_RECURSE;
5567 RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5568 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5569 eptr--;
5570 if (utf) BACKCHAR(eptr);
5571 }
5572 }
5573
5574 /* Match extended Unicode grapheme clusters. We will get here only if the
5575 support is in the binary; otherwise a compile-time error occurs. */
5576
5577 else if (ctype == OP_EXTUNI)
5578 {
5579 for (i = min; i < max; i++)
5580 {
5581 if (eptr >= md->end_subject)
5582 {
5583 SCHECK_PARTIAL();
5584 break;
5585 }
5586 else
5587 {
5588 int lgb, rgb;
5589 GETCHARINCTEST(c, eptr);
5590 lgb = UCD_GRAPHBREAK(c);
5591 while (eptr < md->end_subject)
5592 {
5593 int len = 1;
5594 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5595 rgb = UCD_GRAPHBREAK(c);
5596 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5597 lgb = rgb;
5598 eptr += len;
5599 }
5600 }
5601 CHECK_PARTIAL();
5602 }
5603
5604 /* eptr is now past the end of the maximum run */
5605
5606 if (possessive) continue; /* No backtracking */
5607
5608 for(;;)
5609 {
5610 int lgb, rgb;
5611 PCRE_PUCHAR fptr;
5612
5613 if (eptr == pp) goto TAIL_RECURSE; /* At start of char run */
5614 RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5615 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5616
5617 /* Backtracking over an extended grapheme cluster involves inspecting
5618 the previous two characters (if present) to see if a break is
5619 permitted between them. */
5620
5621 eptr--;
5622 if (!utf) c = *eptr; else
5623 {
5624 BACKCHAR(eptr);
5625 GETCHAR(c, eptr);
5626 }
5627 rgb = UCD_GRAPHBREAK(c);
5628
5629 for (;;)
5630 {
5631 if (eptr == pp) goto TAIL_RECURSE; /* At start of char run */
5632 fptr = eptr - 1;
5633 if (!utf) c = *fptr; else
5634 {
5635 BACKCHAR(fptr);
5636 GETCHAR(c, fptr);
5637 }
5638 lgb = UCD_GRAPHBREAK(c);
5639 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5640 eptr = fptr;
5641 rgb = lgb;
5642 }
5643 }
5644 }
5645
5646 else
5647 #endif /* SUPPORT_UCP */
5648
5649 #ifdef SUPPORT_UTF
5650 if (utf)
5651 {
5652 switch(ctype)
5653 {
5654 case OP_ANY:
5655 if (max < INT_MAX)
5656 {
5657 for (i = min; i < max; i++)
5658 {
5659 if (eptr >= md->end_subject)
5660 {
5661 SCHECK_PARTIAL();
5662 break;
5663 }
5664 if (IS_NEWLINE(eptr)) break;
5665 if (md->partial != 0 && /* Take care with CRLF partial */
5666 eptr + 1 >= md->end_subject &&
5667 NLBLOCK->nltype == NLTYPE_FIXED &&
5668 NLBLOCK->nllen == 2 &&
5669 RAWUCHAR(eptr) == NLBLOCK->nl[0])
5670 {
5671 md->hitend = TRUE;
5672 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5673 }
5674 eptr++;
5675 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5676 }
5677 }
5678
5679 /* Handle unlimited UTF-8 repeat */
5680
5681 else
5682 {
5683 for (i = min; i < max; i++)
5684 {
5685 if (eptr >= md->end_subject)
5686 {
5687 SCHECK_PARTIAL();
5688 break;
5689 }
5690 if (IS_NEWLINE(eptr)) break;
5691 if (md->partial != 0 && /* Take care with CRLF partial */
5692 eptr + 1 >= md->end_subject &&
5693 NLBLOCK->nltype == NLTYPE_FIXED &&
5694 NLBLOCK->nllen == 2 &&
5695 RAWUCHAR(eptr) == NLBLOCK->nl[0])
5696 {
5697 md->hitend = TRUE;
5698 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5699 }
5700 eptr++;
5701 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5702 }
5703 }
5704 break;
5705
5706 case OP_ALLANY:
5707 if (max < INT_MAX)
5708 {
5709 for (i = min; i < max; i++)
5710 {
5711 if (eptr >= md->end_subject)
5712 {
5713 SCHECK_PARTIAL();
5714 break;
5715 }
5716 eptr++;
5717 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5718 }
5719 }
5720 else
5721 {
5722 eptr = md->end_subject; /* Unlimited UTF-8 repeat */
5723 SCHECK_PARTIAL();
5724 }
5725 break;
5726
5727 /* The byte case is the same as non-UTF8 */
5728
5729 case OP_ANYBYTE:
5730 c = max - min;
5731 if (c > (unsigned int)(md->end_subject - eptr))
5732 {
5733 eptr = md->end_subject;
5734 SCHECK_PARTIAL();
5735 }
5736 else eptr += c;
5737 break;
5738
5739 case OP_ANYNL:
5740 for (i = min; i < max; i++)
5741 {
5742 int len = 1;
5743 if (eptr >= md->end_subject)
5744 {
5745 SCHECK_PARTIAL();
5746 break;
5747 }
5748 GETCHARLEN(c, eptr, len);
5749 if (c == CHAR_CR)
5750 {
5751 if (++eptr >= md->end_subject) break;
5752 if (RAWUCHAR(eptr) == CHAR_LF) eptr++;
5753 }
5754 else
5755 {
5756 if (c != CHAR_LF &&
5757 (md->bsr_anycrlf ||
5758 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
5759 #ifndef EBCDIC
5760 && c != 0x2028 && c != 0x2029
5761 #endif /* Not EBCDIC */
5762 )))
5763 break;
5764 eptr += len;
5765 }
5766 }
5767 break;
5768
5769 case OP_NOT_HSPACE:
5770 case OP_HSPACE:
5771 for (i = min; i < max; i++)
5772 {
5773 BOOL gotspace;
5774 int len = 1;
5775 if (eptr >= md->end_subject)
5776 {
5777 SCHECK_PARTIAL();
5778 break;
5779 }
5780 GETCHARLEN(c, eptr, len);
5781 switch(c)
5782 {
5783 HSPACE_CASES: gotspace = TRUE; break;
5784 default: gotspace = FALSE; break;
5785 }
5786 if (gotspace == (ctype == OP_NOT_HSPACE)) break;
5787 eptr += len;
5788 }
5789 break;
5790
5791 case OP_NOT_VSPACE:
5792 case OP_VSPACE:
5793 for (i = min; i < max; i++)
5794 {
5795 BOOL gotspace;
5796 int len = 1;
5797 if (eptr >= md->end_subject)
5798 {
5799 SCHECK_PARTIAL();
5800 break;
5801 }
5802 GETCHARLEN(c, eptr, len);
5803 switch(c)
5804 {
5805 VSPACE_CASES: gotspace = TRUE; break;
5806 default: gotspace = FALSE; break;
5807 }
5808 if (gotspace == (ctype == OP_NOT_VSPACE)) break;
5809 eptr += len;
5810 }
5811 break;
5812
5813 case OP_NOT_DIGIT:
5814 for (i = min; i < max; i++)
5815 {
5816 int len = 1;
5817 if (eptr >= md->end_subject)
5818 {
5819 SCHECK_PARTIAL();
5820 break;
5821 }
5822 GETCHARLEN(c, eptr, len);
5823 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
5824 eptr+= len;
5825 }
5826 break;
5827
5828 case OP_DIGIT:
5829 for (i = min; i < max; i++)
5830 {
5831 int len = 1;
5832 if (eptr >= md->end_subject)
5833 {
5834 SCHECK_PARTIAL();
5835 break;
5836 }
5837 GETCHARLEN(c, eptr, len);
5838 if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break;
5839 eptr+= len;
5840 }
5841 break;
5842
5843 case OP_NOT_WHITESPACE:
5844 for (i = min; i < max; i++)
5845 {
5846 int len = 1;
5847 if (eptr >= md->end_subject)
5848 {
5849 SCHECK_PARTIAL();
5850 break;
5851 }
5852 GETCHARLEN(c, eptr, len);
5853 if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
5854 eptr+= len;
5855 }
5856 break;
5857
5858 case OP_WHITESPACE:
5859 for (i = min; i < max; i++)
5860 {
5861 int len = 1;
5862 if (eptr >= md->end_subject)
5863 {
5864 SCHECK_PARTIAL();
5865 break;
5866 }
5867 GETCHARLEN(c, eptr, len);
5868 if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break;
5869 eptr+= len;
5870 }
5871 break;
5872
5873 case OP_NOT_WORDCHAR:
5874 for (i = min; i < max; i++)
5875 {
5876 int len = 1;
5877 if (eptr >= md->end_subject)
5878 {
5879 SCHECK_PARTIAL();
5880 break;
5881 }
5882 GETCHARLEN(c, eptr, len);
5883 if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
5884 eptr+= len;
5885 }
5886 break;
5887
5888 case OP_WORDCHAR:
5889 for (i = min; i < max; i++)
5890 {
5891 int len = 1;
5892 if (eptr >= md->end_subject)
5893 {
5894 SCHECK_PARTIAL();
5895 break;
5896 }
5897 GETCHARLEN(c, eptr, len);
5898 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break;
5899 eptr+= len;
5900 }
5901 break;
5902
5903 default:
5904 RRETURN(PCRE_ERROR_INTERNAL);
5905 }
5906
5907 if (possessive) continue; /* No backtracking */
5908 for(;;)
5909 {
5910 if (eptr == pp) goto TAIL_RECURSE;
5911 RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
5912 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5913 eptr--;
5914 BACKCHAR(eptr);
5915 if (ctype == OP_ANYNL && eptr > pp && RAWUCHAR(eptr) == CHAR_NL &&
5916 RAWUCHAR(eptr - 1) == CHAR_CR) eptr--;
5917 }
5918 }
5919 else
5920 #endif /* SUPPORT_UTF */
5921 /* Not UTF mode */
5922 {
5923 switch(ctype)
5924 {
5925 case OP_ANY:
5926 for (i = min; i < max; i++)
5927 {
5928 if (eptr >= md->end_subject)
5929 {
5930 SCHECK_PARTIAL();
5931 break;
5932 }
5933 if (IS_NEWLINE(eptr)) break;
5934 if (md->partial != 0 && /* Take care with CRLF partial */
5935 eptr + 1 >= md->end_subject &&
5936 NLBLOCK->nltype == NLTYPE_FIXED &&
5937 NLBLOCK->nllen == 2 &&
5938 *eptr == NLBLOCK->nl[0])
5939 {
5940 md->hitend = TRUE;
5941 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5942 }
5943 eptr++;
5944 }
5945 break;
5946
5947 case OP_ALLANY:
5948 case OP_ANYBYTE:
5949 c = max - min;
5950 if (c > (unsigned int)(md->end_subject - eptr))
5951 {
5952 eptr = md->end_subject;
5953 SCHECK_PARTIAL();
5954 }
5955 else eptr += c;
5956 break;
5957
5958 case OP_ANYNL:
5959 for (i = min; i < max; i++)
5960 {
5961 if (eptr >= md->end_subject)
5962 {
5963 SCHECK_PARTIAL();
5964 break;
5965 }
5966 c = *eptr;
5967 if (c == CHAR_CR)
5968 {
5969 if (++eptr >= md->end_subject) break;
5970 if (*eptr == CHAR_LF) eptr++;
5971 }
5972 else
5973 {
5974 if (c != CHAR_LF && (md->bsr_anycrlf ||
5975 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
5976 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5977 && c != 0x2028 && c != 0x2029
5978 #endif
5979 ))) break;
5980 eptr++;
5981 }
5982 }
5983 break;
5984
5985 case OP_NOT_HSPACE:
5986 for (i = min; i < max; i++)
5987 {
5988 if (eptr >= md->end_subject)
5989 {
5990 SCHECK_PARTIAL();
5991 break;
5992 }
5993 switch(*eptr)
5994 {
5995 default: eptr++; break;
5996 HSPACE_BYTE_CASES:
5997 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5998 HSPACE_MULTIBYTE_CASES:
5999 #endif
6000 goto ENDLOOP00;
6001 }
6002 }
6003 ENDLOOP00:
6004 break;
6005
6006 case OP_HSPACE:
6007 for (i = min; i < max; i++)
6008 {
6009 if (eptr >= md->end_subject)
6010 {
6011 SCHECK_PARTIAL();
6012 break;
6013 }
6014 switch(*eptr)
6015 {
6016 default: goto ENDLOOP01;
6017 HSPACE_BYTE_CASES:
6018 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6019 HSPACE_MULTIBYTE_CASES:
6020 #endif
6021 eptr++; break;
6022 }
6023 }
6024 ENDLOOP01:
6025 break;
6026
6027 case OP_NOT_VSPACE:
6028 for (i = min; i < max; i++)
6029 {
6030 if (eptr >= md->end_subject)
6031 {
6032 SCHECK_PARTIAL();
6033 break;
6034 }
6035 switch(*eptr)
6036 {
6037 default: eptr++; break;
6038 VSPACE_BYTE_CASES:
6039 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6040 VSPACE_MULTIBYTE_CASES:
6041 #endif
6042 goto ENDLOOP02;
6043 }
6044 }
6045 ENDLOOP02:
6046 break;
6047
6048 case OP_VSPACE:
6049 for (i = min; i < max; i++)
6050 {
6051 if (eptr >= md->end_subject)
6052 {
6053 SCHECK_PARTIAL();
6054 break;
6055 }
6056 switch(*eptr)
6057 {
6058 default: goto ENDLOOP03;
6059 VSPACE_BYTE_CASES:
6060 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6061 VSPACE_MULTIBYTE_CASES:
6062 #endif
6063 eptr++; break;
6064 }
6065 }
6066 ENDLOOP03:
6067 break;
6068
6069 case OP_NOT_DIGIT:
6070 for (i = min; i < max; i++)
6071 {
6072 if (eptr >= md->end_subject)
6073 {
6074 SCHECK_PARTIAL();
6075 break;
6076 }
6077 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) break;
6078 eptr++;
6079 }
6080 break;
6081
6082 case OP_DIGIT:
6083 for (i = min; i < max; i++)
6084 {
6085 if (eptr >= md->end_subject)
6086 {
6087 SCHECK_PARTIAL();
6088 break;
6089 }
6090 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) break;
6091 eptr++;
6092 }
6093 break;
6094
6095 case OP_NOT_WHITESPACE:
6096 for (i = min; i < max; i++)
6097 {
6098 if (eptr >= md->end_subject)
6099 {
6100 SCHECK_PARTIAL();
6101 break;
6102 }
6103 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break;
6104 eptr++;
6105 }
6106 break;
6107
6108 case OP_WHITESPACE:
6109 for (i = min; i < max; i++)
6110 {
6111 if (eptr >= md->end_subject)
6112 {
6113 SCHECK_PARTIAL();
6114 break;
6115 }
6116 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break;
6117 eptr++;
6118 }
6119 break;
6120
6121 case OP_NOT_WORDCHAR:
6122 for (i = min; i < max; i++)
6123 {
6124 if (eptr >= md->end_subject)
6125 {
6126 SCHECK_PARTIAL();
6127 break;
6128 }
6129 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break;
6130 eptr++;
6131 }
6132 break;
6133
6134 case OP_WORDCHAR:
6135 for (i = min; i < max; i++)
6136 {
6137 if (eptr >= md->end_subject)
6138 {
6139 SCHECK_PARTIAL();
6140 break;
6141 }
6142 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break;
6143 eptr++;
6144 }
6145 break;
6146
6147 default:
6148 RRETURN(PCRE_ERROR_INTERNAL);
6149 }
6150
6151 if (possessive) continue; /* No backtracking */
6152 for (;;)
6153 {
6154 if (eptr == pp) goto TAIL_RECURSE;
6155 RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
6156 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
6157 eptr--;
6158 if (ctype == OP_ANYNL && eptr > pp && *eptr == CHAR_LF &&
6159 eptr[-1] == CHAR_CR) eptr--;
6160 }
6161 }
6162
6163 /* Control never gets here */
6164 }
6165
6166 /* There's been some horrible disaster. Arrival here can only mean there is
6167 something seriously wrong in the code above or the OP_xxx definitions. */
6168
6169 default:
6170 DPRINTF(("Unknown opcode %d\n", *ecode));
6171 RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
6172 }
6173
6174 /* Do not stick any code in here without much thought; it is assumed
6175 that "continue" in the code above comes out to here to repeat the main
6176 loop. */
6177
6178 } /* End of main loop */
6179 /* Control never reaches here */
6180
6181
6182 /* When compiling to use the heap rather than the stack for recursive calls to
6183 match(), the RRETURN() macro jumps here. The number that is saved in
6184 frame->Xwhere indicates which label we actually want to return to. */
6185
6186 #ifdef NO_RECURSE
6187 #define LBL(val) case val: goto L_RM##val;
6188 HEAP_RETURN:
6189 switch (frame->Xwhere)
6190 {
6191 LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
6192 LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
6193 LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
6194 LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
6195 LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64)
6196 LBL(65) LBL(66)
6197 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6198 LBL(21)
6199 #endif
6200 #ifdef SUPPORT_UTF
6201 LBL(16) LBL(18) LBL(20)
6202 LBL(22) LBL(23) LBL(28) LBL(30)
6203 LBL(32) LBL(34) LBL(42) LBL(46)
6204 #ifdef SUPPORT_UCP
6205 LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
6206 LBL(59) LBL(60) LBL(61) LBL(62) LBL(67) LBL(68)
6207 #endif /* SUPPORT_UCP */
6208 #endif /* SUPPORT_UTF */
6209 default:
6210 DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
6211 return PCRE_ERROR_INTERNAL;
6212 }
6213 #undef LBL
6214 #endif /* NO_RECURSE */
6215 }
6216
6217
6218 /***************************************************************************
6219 ****************************************************************************
6220 RECURSION IN THE match() FUNCTION
6221
6222 Undefine all the macros that were defined above to handle this. */
6223
6224 #ifdef NO_RECURSE
6225 #undef eptr
6226 #undef ecode
6227 #undef mstart
6228 #undef offset_top
6229 #undef eptrb
6230 #undef flags
6231
6232 #undef callpat
6233 #undef charptr
6234 #undef data
6235 #undef next
6236 #undef pp
6237 #undef prev
6238 #undef saved_eptr
6239
6240 #undef new_recursive
6241
6242 #undef cur_is_word
6243 #undef condition
6244 #undef prev_is_word
6245
6246 #undef ctype
6247 #undef length
6248 #undef max
6249 #undef min
6250 #undef number
6251 #undef offset
6252 #undef op
6253 #undef save_capture_last
6254 #undef save_offset1
6255 #undef save_offset2
6256 #undef save_offset3
6257 #undef stacksave
6258
6259 #undef newptrb
6260
6261 #endif
6262
6263 /* These two are defined as macros in both cases */
6264
6265 #undef fc
6266 #undef fi
6267
6268 /***************************************************************************
6269 ***************************************************************************/
6270
6271
6272 #ifdef NO_RECURSE
6273 /*************************************************
6274 * Release allocated heap frames *
6275 *************************************************/
6276
6277 /* This function releases all the allocated frames. The base frame is on the
6278 machine stack, and so must not be freed.
6279
6280 Argument: the address of the base frame
6281 Returns: nothing
6282 */
6283
6284 static void
6285 release_match_heapframes (heapframe *frame_base)
6286 {
6287 heapframe *nextframe = frame_base->Xnextframe;
6288 while (nextframe != NULL)
6289 {
6290 heapframe *oldframe = nextframe;
6291 nextframe = nextframe->Xnextframe;
6292 (PUBL(stack_free))(oldframe);
6293 }
6294 }
6295 #endif
6296
6297
6298 /*************************************************
6299 * Execute a Regular Expression *
6300 *************************************************/
6301
6302 /* This function applies a compiled re to a subject string and picks out
6303 portions of the string if it matches. Two elements in the vector are set for
6304 each substring: the offsets to the start and end of the substring.
6305
6306 Arguments:
6307 argument_re points to the compiled expression
6308 extra_data points to extra data or is NULL
6309 subject points to the subject string
6310 length length of subject string (may contain binary zeros)
6311 start_offset where to start in the subject string
6312 options option bits
6313 offsets points to a vector of ints to be filled in with offsets
6314 offsetcount the number of elements in the vector
6315
6316 Returns: > 0 => success; value is the number of elements filled in
6317 = 0 => success, but offsets is not big enough
6318 -1 => failed to match
6319 < -1 => some kind of unexpected problem
6320 */
6321
6322 #if defined COMPILE_PCRE8
6323 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6324 pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
6325 PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
6326 int offsetcount)
6327 #elif defined COMPILE_PCRE16
6328 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6329 pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data,
6330 PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets,
6331 int offsetcount)
6332 #elif defined COMPILE_PCRE32
6333 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6334 pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data,
6335 PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets,
6336 int offsetcount)
6337 #endif
6338 {
6339 int rc, ocount, arg_offset_max;
6340 int newline;
6341 BOOL using_temporary_offsets = FALSE;
6342 BOOL anchored;
6343 BOOL startline;
6344 BOOL firstline;
6345 BOOL utf;
6346 BOOL has_first_char = FALSE;
6347 BOOL has_req_char = FALSE;
6348 pcre_uchar first_char = 0;
6349 pcre_uchar first_char2 = 0;
6350 pcre_uchar req_char = 0;
6351 pcre_uchar req_char2 = 0;
6352 match_data match_block;
6353 match_data *md = &match_block;
6354 const pcre_uint8 *tables;
6355 const pcre_uint8 *start_bits = NULL;
6356 PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset;
6357 PCRE_PUCHAR end_subject;
6358 PCRE_PUCHAR start_partial = NULL;
6359 PCRE_PUCHAR match_partial = NULL;
6360 PCRE_PUCHAR req_char_ptr = start_match - 1;
6361
6362 const pcre_study_data *study;
6363 const REAL_PCRE *re = (const REAL_PCRE *)argument_re;
6364
6365 #ifdef NO_RECURSE
6366 heapframe frame_zero;
6367 frame_zero.Xprevframe = NULL; /* Marks the top level */
6368 frame_zero.Xnextframe = NULL; /* None are allocated yet */
6369 md->match_frames_base = &frame_zero;
6370 #endif
6371
6372 /* Check for the special magic call that measures the size of the stack used
6373 per recursive call of match(). Without the funny casting for sizeof, a Windows
6374 compiler gave this error: "unary minus operator applied to unsigned type,
6375 result still unsigned". Hopefully the cast fixes that. */
6376
6377 if (re == NULL && extra_data == NULL && subject == NULL && length == -999 &&
6378 start_offset == -999)
6379 #ifdef NO_RECURSE
6380 return -((int)sizeof(heapframe));
6381 #else
6382 return match(NULL, NULL, NULL, 0, NULL, NULL, 0);
6383 #endif
6384
6385 /* Plausibility checks */
6386
6387 if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
6388 if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0))
6389 return PCRE_ERROR_NULL;
6390 if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
6391 if (length < 0) return PCRE_ERROR_BADLENGTH;
6392 if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
6393
6394 /* Check that the first field in the block is the magic number. If it is not,
6395 return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to
6396 REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which
6397 means that the pattern is likely compiled with different endianness. */
6398
6399 if (re->magic_number != MAGIC_NUMBER)
6400 return re->magic_number == REVERSED_MAGIC_NUMBER?
6401 PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC;
6402 if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE;
6403
6404 /* These two settings are used in the code for checking a UTF-8 string that
6405 follows immediately afterwards. Other values in the md block are used only
6406 during "normal" pcre_exec() processing, not when the JIT support is in use,
6407 so they are set up later. */
6408
6409 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
6410 utf = md->utf = (re->options & PCRE_UTF8) != 0;
6411 md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
6412 ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
6413
6414 /* Check a UTF-8 string if required. Pass back the character offset and error
6415 code for an invalid string if a results vector is available. */
6416
6417 #ifdef SUPPORT_UTF
6418 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0)
6419 {
6420 int erroroffset;
6421 int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset);
6422 if (errorcode != 0)
6423 {
6424 if (offsetcount >= 2)
6425 {
6426 offsets[0] = erroroffset;
6427 offsets[1] = errorcode;
6428 }
6429 #if defined COMPILE_PCRE8
6430 return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
6431 PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
6432 #elif defined COMPILE_PCRE16
6433 return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)?
6434 PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16;
6435 #elif defined COMPILE_PCRE32
6436 return PCRE_ERROR_BADUTF32;
6437 #endif
6438 }
6439 #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16
6440 /* Check that a start_offset points to the start of a UTF character. */
6441 if (start_offset > 0 && start_offset < length &&
6442 NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset]))
6443 return PCRE_ERROR_BADUTF8_OFFSET;
6444 #endif
6445 }
6446 #endif
6447
6448 /* If the pattern was successfully studied with JIT support, run the JIT
6449 executable instead of the rest of this function. Most options must be set at
6450 compile time for the JIT code to be usable. Fallback to the normal code path if
6451 an unsupported flag is set. */
6452
6453 #ifdef SUPPORT_JIT
6454 if (extra_data != NULL
6455 && (extra_data->flags & (PCRE_EXTRA_EXECUTABLE_JIT |
6456 PCRE_EXTRA_TABLES)) == PCRE_EXTRA_EXECUTABLE_JIT
6457 && extra_data->executable_jit != NULL
6458 && (options & ~PUBLIC_JIT_EXEC_OPTIONS) == 0)
6459 {
6460 rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length,
6461 start_offset, options, offsets, offsetcount);
6462
6463 /* PCRE_ERROR_NULL means that the selected normal or partial matching
6464 mode is not compiled. In this case we simply fallback to interpreter. */
6465
6466 if (rc != PCRE_ERROR_JIT_BADOPTION) return rc;
6467 }
6468 #endif
6469
6470 /* Carry on with non-JIT matching. This information is for finding all the
6471 numbers associated with a given name, for condition testing. */
6472
6473 md->name_table = (pcre_uchar *)re + re->name_table_offset;
6474 md->name_count = re->name_count;
6475 md->name_entry_size = re->name_entry_size;
6476
6477 /* Fish out the optional data from the extra_data structure, first setting
6478 the default values. */
6479
6480 study = NULL;
6481 md->match_limit = MATCH_LIMIT;
6482 md->match_limit_recursion = MATCH_LIMIT_RECURSION;
6483 md->callout_data = NULL;
6484
6485 /* The table pointer is always in native byte order. */
6486
6487 tables = re->tables;
6488
6489 /* The two limit values override the defaults, whatever their value. */
6490
6491 if (extra_data != NULL)
6492 {
6493 register unsigned int flags = extra_data->flags;
6494 if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
6495 study = (const pcre_study_data *)extra_data->study_data;
6496 if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
6497 md->match_limit = extra_data->match_limit;
6498 if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
6499 md->match_limit_recursion = extra_data->match_limit_recursion;
6500 if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
6501 md->callout_data = extra_data->callout_data;
6502 if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
6503 }
6504
6505 /* Limits in the regex override only if they are smaller. */
6506
6507 if ((re->flags & PCRE_MLSET) != 0 && re->limit_match < md->match_limit)
6508 md->match_limit = re->limit_match;
6509
6510 if ((re->flags & PCRE_RLSET) != 0 &&
6511 re->limit_recursion < md->match_limit_recursion)
6512 md->match_limit_recursion = re->limit_recursion;
6513
6514 /* If the exec call supplied NULL for tables, use the inbuilt ones. This
6515 is a feature that makes it possible to save compiled regex and re-use them
6516 in other programs later. */
6517
6518 if (tables == NULL) tables = PRIV(default_tables);
6519
6520 /* Set up other data */
6521
6522 anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
6523 startline = (re->flags & PCRE_STARTLINE) != 0;
6524 firstline = (re->options & PCRE_FIRSTLINE) != 0;
6525
6526 /* The code starts after the real_pcre block and the capture name table. */
6527
6528 md->start_code = (const pcre_uchar *)re + re->name_table_offset +
6529 re->name_count * re->name_entry_size;
6530
6531 md->start_subject = (PCRE_PUCHAR)subject;
6532 md->start_offset = start_offset;
6533 md->end_subject = md->start_subject + length;
6534 end_subject = md->end_subject;
6535
6536 md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
6537 md->use_ucp = (re->options & PCRE_UCP) != 0;
6538 md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6539 md->ignore_skip_arg = 0;
6540
6541 /* Some options are unpacked into BOOL variables in the hope that testing
6542 them will be faster than individual option bits. */
6543
6544 md->notbol = (options & PCRE_NOTBOL) != 0;
6545 md->noteol = (options & PCRE_NOTEOL) != 0;
6546 md->notempty = (options & PCRE_NOTEMPTY) != 0;
6547 md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6548
6549 md->hitend = FALSE;
6550 md->mark = md->nomatch_mark = NULL; /* In case never set */
6551
6552 md->recursive = NULL; /* No recursion at top level */
6553 md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6554
6555 md->lcc = tables + lcc_offset;
6556 md->fcc = tables + fcc_offset;
6557 md->ctypes = tables + ctypes_offset;
6558
6559 /* Handle different \R options. */
6560
6561 switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6562 {
6563 case 0:
6564 if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
6565 md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
6566 else
6567 #ifdef BSR_ANYCRLF
6568 md->bsr_anycrlf = TRUE;
6569 #else
6570 md->bsr_anycrlf = FALSE;
6571 #endif
6572 break;
6573
6574 case PCRE_BSR_ANYCRLF:
6575 md->bsr_anycrlf = TRUE;
6576 break;
6577
6578 case PCRE_BSR_UNICODE:
6579 md->bsr_anycrlf = FALSE;
6580 break;
6581
6582 default: return PCRE_ERROR_BADNEWLINE;
6583 }
6584
6585 /* Handle different types of newline. The three bits give eight cases. If
6586 nothing is set at run time, whatever was used at compile time applies. */
6587
6588 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
6589 (pcre_uint32)options) & PCRE_NEWLINE_BITS)
6590 {
6591 case 0: newline = NEWLINE; break; /* Compile-time default */
6592 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6593 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6594 case PCRE_NEWLINE_CR+
6595 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6596 case PCRE_NEWLINE_ANY: newline = -1; break;
6597 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6598 default: return PCRE_ERROR_BADNEWLINE;
6599 }
6600
6601 if (newline == -2)
6602 {
6603 md->nltype = NLTYPE_ANYCRLF;
6604 }
6605 else if (newline < 0)
6606 {
6607 md->nltype = NLTYPE_ANY;
6608 }
6609 else
6610 {
6611 md->nltype = NLTYPE_FIXED;
6612 if (newline > 255)
6613 {
6614 md->nllen = 2;
6615 md->nl[0] = (newline >> 8) & 255;
6616 md->nl[1] = newline & 255;
6617 }
6618 else
6619 {
6620 md->nllen = 1;
6621 md->nl[0] = newline;
6622 }
6623 }
6624
6625 /* Partial matching was originally supported only for a restricted set of
6626 regexes; from release 8.00 there are no restrictions, but the bits are still
6627 defined (though never set). So there's no harm in leaving this code. */
6628
6629 if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6630 return PCRE_ERROR_BADPARTIAL;
6631
6632 /* If the expression has got more back references than the offsets supplied can
6633 hold, we get a temporary chunk of working store to use during the matching.
6634 Otherwise, we can use the vector supplied, rounding down its size to a multiple
6635 of 3. */
6636
6637 ocount = offsetcount - (offsetcount % 3);
6638 arg_offset_max = (2*ocount)/3;
6639
6640 if (re->top_backref > 0 && re->top_backref >= ocount/3)
6641 {
6642 ocount = re->top_backref * 3 + 3;
6643 md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int));
6644 if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
6645 using_temporary_offsets = TRUE;
6646 DPRINTF(("Got memory to hold back references\n"));
6647 }
6648 else md->offset_vector = offsets;
6649 md->offset_end = ocount;
6650 md->offset_max = (2*ocount)/3;
6651 md->capture_last = 0;
6652
6653 /* Reset the working variable associated with each extraction. These should
6654 never be used unless previously set, but they get saved and restored, and so we
6655 initialize them to avoid reading uninitialized locations. Also, unset the
6656 offsets for the matched string. This is really just for tidiness with callouts,
6657 in case they inspect these fields. */
6658
6659 if (md->offset_vector != NULL)
6660 {
6661 register int *iptr = md->offset_vector + ocount;
6662 register int *iend = iptr - re->top_bracket;
6663 if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6664 while (--iptr >= iend) *iptr = -1;
6665 md->offset_vector[0] = md->offset_vector[1] = -1;
6666 }
6667
6668 /* Set up the first character to match, if available. The first_char value is
6669 never set for an anchored regular expression, but the anchoring may be forced
6670 at run time, so we have to test for anchoring. The first char may be unset for
6671 an unanchored pattern, of course. If there's no first char and the pattern was
6672 studied, there may be a bitmap of possible first characters. */
6673
6674 if (!anchored)
6675 {
6676 if ((re->flags & PCRE_FIRSTSET) != 0)
6677 {
6678 has_first_char = TRUE;
6679 first_char = first_char2 = (pcre_uchar)(re->first_char);
6680 if ((re->flags & PCRE_FCH_CASELESS) != 0)
6681 {
6682 first_char2 = TABLE_GET(first_char, md->fcc, first_char);
6683 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
6684 if (utf && first_char > 127)
6685 first_char2 = UCD_OTHERCASE(first_char);
6686 #endif
6687 }
6688 }
6689 else
6690 if (!startline && study != NULL &&
6691 (study->flags & PCRE_STUDY_MAPPED) != 0)
6692 start_bits = study->start_bits;
6693 }
6694
6695 /* For anchored or unanchored matches, there may be a "last known required
6696 character" set. */
6697
6698 if ((re->flags & PCRE_REQCHSET) != 0)
6699 {
6700 has_req_char = TRUE;
6701 req_char = req_char2 = (pcre_uchar)(re->req_char);
6702 if ((re->flags & PCRE_RCH_CASELESS) != 0)
6703 {
6704 req_char2 = TABLE_GET(req_char, md->fcc, req_char);
6705 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
6706 if (utf && req_char > 127)
6707 req_char2 = UCD_OTHERCASE(req_char);
6708 #endif
6709 }
6710 }
6711
6712
6713 /* ==========================================================================*/
6714
6715 /* Loop for handling unanchored repeated matching attempts; for anchored regexs
6716 the loop runs just once. */
6717
6718 for(;;)
6719 {
6720 PCRE_PUCHAR save_end_subject = end_subject;
6721 PCRE_PUCHAR new_start_match;
6722
6723 /* If firstline is TRUE, the start of the match is constrained to the first
6724 line of a multiline string. That is, the match must be before or at the first
6725 newline. Implement this by temporarily adjusting end_subject so that we stop
6726 scanning at a newline. If the match fails at the newline, later code breaks
6727 this loop. */
6728
6729 if (firstline)
6730 {
6731 PCRE_PUCHAR t = start_match;
6732 #ifdef SUPPORT_UTF
6733 if (utf)
6734 {
6735 while (t < md->end_subject && !IS_NEWLINE(t))
6736 {
6737 t++;
6738 ACROSSCHAR(t < end_subject, *t, t++);
6739 }
6740 }
6741 else
6742 #endif
6743 while (t < md->end_subject && !IS_NEWLINE(t)) t++;
6744 end_subject = t;
6745 }
6746
6747 /* There are some optimizations that avoid running the match if a known
6748 starting point is not found, or if a known later character is not present.
6749 However, there is an option that disables these, for testing and for ensuring
6750 that all callouts do actually occur. The option can be set in the regex by
6751 (*NO_START_OPT) or passed in match-time options. */
6752
6753 if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0)
6754 {
6755 /* Advance to a unique first char if there is one. */