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

Contents of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1311 - (show annotations)
Mon Apr 22 17:35:23 2013 UTC (6 years, 5 months ago) by ph10
File MIME type: text/plain
File size: 218777 byte(s)
Use tail recursion in maximizing character and character type repetitions, to 
reduce stack usage.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2013 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40 /* This module contains pcre_exec(), the externally visible function that does
41 pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42 possible. There are also some static supporting functions. */
43
44 #ifdef HAVE_CONFIG_H
45 #include "config.h"
46 #endif
47
48 #define NLBLOCK md /* Block containing newline information */
49 #define PSSTART start_subject /* Field containing processed string start */
50 #define PSEND end_subject /* Field containing processed string end */
51
52 #include "pcre_internal.h"
53
54 /* Undefine some potentially clashing cpp symbols */
55
56 #undef min
57 #undef max
58
59 /* The md->capture_last field uses the lower 16 bits for the last captured
60 substring (which can never be greater than 65535) and a bit in the top half
61 to mean "capture vector overflowed". This odd way of doing things was
62 implemented when it was realized that preserving and restoring the overflow bit
63 whenever the last capture number was saved/restored made for a neater
64 interface, and doing it this way saved on (a) another variable, which would
65 have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66 separate set of save/restore instructions. The following defines are used in
67 implementing this. */
68
69 #define CAPLMASK 0x0000ffff /* The bits used for last_capture */
70 #define OVFLMASK 0xffff0000 /* The bits used for the overflow flag */
71 #define OVFLBIT 0x00010000 /* The bit that is set for overflow */
72
73 /* Values for setting in md->match_function_type to indicate two special types
74 of call to match(). We do it this way to save on using another stack variable,
75 as stack usage is to be discouraged. */
76
77 #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */
78 #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */
79
80 /* Non-error returns from the match() function. Error returns are externally
81 defined PCRE_ERROR_xxx codes, which are all negative. */
82
83 #define MATCH_MATCH 1
84 #define MATCH_NOMATCH 0
85
86 /* Special internal returns from the match() function. Make them sufficiently
87 negative to avoid the external error codes. */
88
89 #define MATCH_ACCEPT (-999)
90 #define MATCH_KETRPOS (-998)
91 #define MATCH_ONCE (-997)
92 /* The next 5 must be kept together and in sequence so that a test that checks
93 for any one of them can use a range. */
94 #define MATCH_COMMIT (-996)
95 #define MATCH_PRUNE (-995)
96 #define MATCH_SKIP (-994)
97 #define MATCH_SKIP_ARG (-993)
98 #define MATCH_THEN (-992)
99 #define MATCH_BACKTRACK_MAX MATCH_THEN
100 #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101
102 /* Maximum number of ints of offset to save on the stack for recursive calls.
103 If the offset vector is bigger, malloc is used. This should be a multiple of 3,
104 because the offset vector is always a multiple of 3 long. */
105
106 #define REC_STACK_SAVE_MAX 30
107
108 /* Min and max values for the common repeats; for the maxima, 0 => infinity */
109
110 static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
111 static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
112
113 #ifdef PCRE_DEBUG
114 /*************************************************
115 * Debugging function to print chars *
116 *************************************************/
117
118 /* Print a sequence of chars in printable format, stopping at the end of the
119 subject if the requested.
120
121 Arguments:
122 p points to characters
123 length number to print
124 is_subject TRUE if printing from within md->start_subject
125 md pointer to matching data block, if is_subject is TRUE
126
127 Returns: nothing
128 */
129
130 static void
131 pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132 {
133 pcre_uint32 c;
134 BOOL utf = md->utf;
135 if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136 while (length-- > 0)
137 if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138 }
139 #endif
140
141
142
143 /*************************************************
144 * Match a back-reference *
145 *************************************************/
146
147 /* Normally, if a back reference hasn't been set, the length that is passed is
148 negative, so the match always fails. However, in JavaScript compatibility mode,
149 the length passed is zero. Note that in caseless UTF-8 mode, the number of
150 subject bytes matched may be different to the number of reference bytes.
151
152 Arguments:
153 offset index into the offset vector
154 eptr pointer into the subject
155 length length of reference to be matched (number of bytes)
156 md points to match data block
157 caseless TRUE if caseless
158
159 Returns: >= 0 the number of subject bytes matched
160 -1 no match
161 -2 partial match; always given if at end subject
162 */
163
164 static int
165 match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166 BOOL caseless)
167 {
168 PCRE_PUCHAR eptr_start = eptr;
169 register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170 #ifdef SUPPORT_UTF
171 BOOL utf = md->utf;
172 #endif
173
174 #ifdef PCRE_DEBUG
175 if (eptr >= md->end_subject)
176 printf("matching subject <null>");
177 else
178 {
179 printf("matching subject ");
180 pchars(eptr, length, TRUE, md);
181 }
182 printf(" against backref ");
183 pchars(p, length, FALSE, md);
184 printf("\n");
185 #endif
186
187 /* Always fail if reference not set (and not JavaScript compatible - in that
188 case the length is passed as zero). */
189
190 if (length < 0) return -1;
191
192 /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193 properly if Unicode properties are supported. Otherwise, we can check only
194 ASCII characters. */
195
196 if (caseless)
197 {
198 #ifdef SUPPORT_UTF
199 #ifdef SUPPORT_UCP
200 if (utf)
201 {
202 /* Match characters up to the end of the reference. NOTE: the number of
203 data units matched may differ, because in UTF-8 there are some characters
204 whose upper and lower case versions code have different numbers of bytes.
205 For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206 (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207 sequence of two of the latter. It is important, therefore, to check the
208 length along the reference, not along the subject (earlier code did this
209 wrong). */
210
211 PCRE_PUCHAR endptr = p + length;
212 while (p < endptr)
213 {
214 pcre_uint32 c, d;
215 const ucd_record *ur;
216 if (eptr >= md->end_subject) return -2; /* Partial match */
217 GETCHARINC(c, eptr);
218 GETCHARINC(d, p);
219 ur = GET_UCD(d);
220 if (c != d && c != d + ur->other_case)
221 {
222 const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223 for (;;)
224 {
225 if (c < *pp) return -1;
226 if (c == *pp++) break;
227 }
228 }
229 }
230 }
231 else
232 #endif
233 #endif
234
235 /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236 is no UCP support. */
237 {
238 while (length-- > 0)
239 {
240 pcre_uint32 cc, cp;
241 if (eptr >= md->end_subject) return -2; /* Partial match */
242 cc = RAWUCHARTEST(eptr);
243 cp = RAWUCHARTEST(p);
244 if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245 p++;
246 eptr++;
247 }
248 }
249 }
250
251 /* In the caseful case, we can just compare the bytes, whether or not we
252 are in UTF-8 mode. */
253
254 else
255 {
256 while (length-- > 0)
257 {
258 if (eptr >= md->end_subject) return -2; /* Partial match */
259 if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260 }
261 }
262
263 return (int)(eptr - eptr_start);
264 }
265
266
267
268 /***************************************************************************
269 ****************************************************************************
270 RECURSION IN THE match() FUNCTION
271
272 The match() function is highly recursive, though not every recursive call
273 increases the recursive depth. Nevertheless, some regular expressions can cause
274 it to recurse to a great depth. I was writing for Unix, so I just let it call
275 itself recursively. This uses the stack for saving everything that has to be
276 saved for a recursive call. On Unix, the stack can be large, and this works
277 fine.
278
279 It turns out that on some non-Unix-like systems there are problems with
280 programs that use a lot of stack. (This despite the fact that every last chip
281 has oodles of memory these days, and techniques for extending the stack have
282 been known for decades.) So....
283
284 There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
285 calls by keeping local variables that need to be preserved in blocks of memory
286 obtained from malloc() instead instead of on the stack. Macros are used to
287 achieve this so that the actual code doesn't look very different to what it
288 always used to.
289
290 The original heap-recursive code used longjmp(). However, it seems that this
291 can be very slow on some operating systems. Following a suggestion from Stan
292 Switzer, the use of longjmp() has been abolished, at the cost of having to
293 provide a unique number for each call to RMATCH. There is no way of generating
294 a sequence of numbers at compile time in C. I have given them names, to make
295 them stand out more clearly.
296
297 Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
298 FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
299 tests. Furthermore, not using longjmp() means that local dynamic variables
300 don't have indeterminate values; this has meant that the frame size can be
301 reduced because the result can be "passed back" by straight setting of the
302 variable instead of being passed in the frame.
303 ****************************************************************************
304 ***************************************************************************/
305
306 /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307 below must be updated in sync. */
308
309 enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10,
310 RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311 RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312 RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313 RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314 RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315 RM61, RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316
317 /* These versions of the macros use the stack, as normal. There are debugging
318 versions and production versions. Note that the "rw" argument of RMATCH isn't
319 actually used in this definition. */
320
321 #ifndef NO_RECURSE
322 #define REGISTER register
323
324 #ifdef PCRE_DEBUG
325 #define RMATCH(ra,rb,rc,rd,re,rw) \
326 { \
327 printf("match() called in line %d\n", __LINE__); \
328 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329 printf("to line %d\n", __LINE__); \
330 }
331 #define RRETURN(ra) \
332 { \
333 printf("match() returned %d from line %d\n", ra, __LINE__); \
334 return ra; \
335 }
336 #else
337 #define RMATCH(ra,rb,rc,rd,re,rw) \
338 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339 #define RRETURN(ra) return ra
340 #endif
341
342 #else
343
344
345 /* These versions of the macros manage a private stack on the heap. Note that
346 the "rd" argument of RMATCH isn't actually used in this definition. It's the md
347 argument of match(), which never changes. */
348
349 #define REGISTER
350
351 #define RMATCH(ra,rb,rc,rd,re,rw)\
352 {\
353 heapframe *newframe = frame->Xnextframe;\
354 if (newframe == NULL)\
355 {\
356 newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357 if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358 newframe->Xnextframe = NULL;\
359 frame->Xnextframe = newframe;\
360 }\
361 frame->Xwhere = rw;\
362 newframe->Xeptr = ra;\
363 newframe->Xecode = rb;\
364 newframe->Xmstart = mstart;\
365 newframe->Xoffset_top = rc;\
366 newframe->Xeptrb = re;\
367 newframe->Xrdepth = frame->Xrdepth + 1;\
368 newframe->Xprevframe = frame;\
369 frame = newframe;\
370 DPRINTF(("restarting from line %d\n", __LINE__));\
371 goto HEAP_RECURSE;\
372 L_##rw:\
373 DPRINTF(("jumped back to line %d\n", __LINE__));\
374 }
375
376 #define RRETURN(ra)\
377 {\
378 heapframe *oldframe = frame;\
379 frame = oldframe->Xprevframe;\
380 if (frame != NULL)\
381 {\
382 rrc = ra;\
383 goto HEAP_RETURN;\
384 }\
385 return ra;\
386 }
387
388
389 /* Structure for remembering the local variables in a private frame */
390
391 typedef struct heapframe {
392 struct heapframe *Xprevframe;
393 struct heapframe *Xnextframe;
394
395 /* Function arguments that may change */
396
397 PCRE_PUCHAR Xeptr;
398 const pcre_uchar *Xecode;
399 PCRE_PUCHAR Xmstart;
400 int Xoffset_top;
401 eptrblock *Xeptrb;
402 unsigned int Xrdepth;
403
404 /* Function local variables */
405
406 PCRE_PUCHAR Xcallpat;
407 #ifdef SUPPORT_UTF
408 PCRE_PUCHAR Xcharptr;
409 #endif
410 PCRE_PUCHAR Xdata;
411 PCRE_PUCHAR Xnext;
412 PCRE_PUCHAR Xpp;
413 PCRE_PUCHAR Xprev;
414 PCRE_PUCHAR Xsaved_eptr;
415
416 recursion_info Xnew_recursive;
417
418 BOOL Xcur_is_word;
419 BOOL Xcondition;
420 BOOL Xprev_is_word;
421
422 #ifdef SUPPORT_UCP
423 int Xprop_type;
424 unsigned int Xprop_value;
425 int Xprop_fail_result;
426 int Xoclength;
427 pcre_uchar Xocchars[6];
428 #endif
429
430 int Xcodelink;
431 int Xctype;
432 unsigned int Xfc;
433 int Xfi;
434 int Xlength;
435 int Xmax;
436 int Xmin;
437 unsigned int Xnumber;
438 int Xoffset;
439 unsigned int Xop;
440 pcre_int32 Xsave_capture_last;
441 int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442 int Xstacksave[REC_STACK_SAVE_MAX];
443
444 eptrblock Xnewptrb;
445
446 /* Where to jump back to */
447
448 int Xwhere;
449
450 } heapframe;
451
452 #endif
453
454
455 /***************************************************************************
456 ***************************************************************************/
457
458
459
460 /*************************************************
461 * Match from current position *
462 *************************************************/
463
464 /* This function is called recursively in many circumstances. Whenever it
465 returns a negative (error) response, the outer incarnation must also return the
466 same response. */
467
468 /* These macros pack up tests that are used for partial matching, and which
469 appear several times in the code. We set the "hit end" flag if the pointer is
470 at the end of the subject and also past the start of the subject (i.e.
471 something has been matched). For hard partial matching, we then return
472 immediately. The second one is used when we already know we are past the end of
473 the subject. */
474
475 #define CHECK_PARTIAL()\
476 if (md->partial != 0 && eptr >= md->end_subject && \
477 eptr > md->start_used_ptr) \
478 { \
479 md->hitend = TRUE; \
480 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481 }
482
483 #define SCHECK_PARTIAL()\
484 if (md->partial != 0 && eptr > md->start_used_ptr) \
485 { \
486 md->hitend = TRUE; \
487 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488 }
489
490
491 /* Performance note: It might be tempting to extract commonly used fields from
492 the md structure (e.g. utf, end_subject) into individual variables to improve
493 performance. Tests using gcc on a SPARC disproved this; in the first case, it
494 made performance worse.
495
496 Arguments:
497 eptr pointer to current character in subject
498 ecode pointer to current position in compiled code
499 mstart pointer to the current match start position (can be modified
500 by encountering \K)
501 offset_top current top pointer
502 md pointer to "static" info for the match
503 eptrb pointer to chain of blocks containing eptr at start of
504 brackets - for testing for empty matches
505 rdepth the recursion depth
506
507 Returns: MATCH_MATCH if matched ) these values are >= 0
508 MATCH_NOMATCH if failed to match )
509 a negative MATCH_xxx value for PRUNE, SKIP, etc
510 a negative PCRE_ERROR_xxx value if aborted by an error condition
511 (e.g. stopped by repeated call or recursion limit)
512 */
513
514 static int
515 match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516 PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517 unsigned int rdepth)
518 {
519 /* These variables do not need to be preserved over recursion in this function,
520 so they can be ordinary variables in all cases. Mark some of them with
521 "register" because they are used a lot in loops. */
522
523 register int rrc; /* Returns from recursive calls */
524 register int i; /* Used for loops not involving calls to RMATCH() */
525 register pcre_uint32 c; /* Character values not kept over RMATCH() calls */
526 register BOOL utf; /* Local copy of UTF flag for speed */
527
528 BOOL minimize, possessive; /* Quantifier options */
529 BOOL caseless;
530 int condcode;
531
532 /* When recursion is not being used, all "local" variables that have to be
533 preserved over calls to RMATCH() are part of a "frame". We set up the top-level
534 frame on the stack here; subsequent instantiations are obtained from the heap
535 whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536 the top-level on the stack rather than malloc-ing them all gives a performance
537 boost in many cases where there is not much "recursion". */
538
539 #ifdef NO_RECURSE
540 heapframe *frame = (heapframe *)md->match_frames_base;
541
542 /* Copy in the original argument variables */
543
544 frame->Xeptr = eptr;
545 frame->Xecode = ecode;
546 frame->Xmstart = mstart;
547 frame->Xoffset_top = offset_top;
548 frame->Xeptrb = eptrb;
549 frame->Xrdepth = rdepth;
550
551 /* This is where control jumps back to to effect "recursion" */
552
553 HEAP_RECURSE:
554
555 /* Macros make the argument variables come from the current frame */
556
557 #define eptr frame->Xeptr
558 #define ecode frame->Xecode
559 #define mstart frame->Xmstart
560 #define offset_top frame->Xoffset_top
561 #define eptrb frame->Xeptrb
562 #define rdepth frame->Xrdepth
563
564 /* Ditto for the local variables */
565
566 #ifdef SUPPORT_UTF
567 #define charptr frame->Xcharptr
568 #endif
569 #define callpat frame->Xcallpat
570 #define codelink frame->Xcodelink
571 #define data frame->Xdata
572 #define next frame->Xnext
573 #define pp frame->Xpp
574 #define prev frame->Xprev
575 #define saved_eptr frame->Xsaved_eptr
576
577 #define new_recursive frame->Xnew_recursive
578
579 #define cur_is_word frame->Xcur_is_word
580 #define condition frame->Xcondition
581 #define prev_is_word frame->Xprev_is_word
582
583 #ifdef SUPPORT_UCP
584 #define prop_type frame->Xprop_type
585 #define prop_value frame->Xprop_value
586 #define prop_fail_result frame->Xprop_fail_result
587 #define oclength frame->Xoclength
588 #define occhars frame->Xocchars
589 #endif
590
591 #define ctype frame->Xctype
592 #define fc frame->Xfc
593 #define fi frame->Xfi
594 #define length frame->Xlength
595 #define max frame->Xmax
596 #define min frame->Xmin
597 #define number frame->Xnumber
598 #define offset frame->Xoffset
599 #define op frame->Xop
600 #define save_capture_last frame->Xsave_capture_last
601 #define save_offset1 frame->Xsave_offset1
602 #define save_offset2 frame->Xsave_offset2
603 #define save_offset3 frame->Xsave_offset3
604 #define stacksave frame->Xstacksave
605
606 #define newptrb frame->Xnewptrb
607
608 /* When recursion is being used, local variables are allocated on the stack and
609 get preserved during recursion in the normal way. In this environment, fi and
610 i, and fc and c, can be the same variables. */
611
612 #else /* NO_RECURSE not defined */
613 #define fi i
614 #define fc c
615
616 /* Many of the following variables are used only in small blocks of the code.
617 My normal style of coding would have declared them within each of those blocks.
618 However, in order to accommodate the version of this code that uses an external
619 "stack" implemented on the heap, it is easier to declare them all here, so the
620 declarations can be cut out in a block. The only declarations within blocks
621 below are for variables that do not have to be preserved over a recursive call
622 to RMATCH(). */
623
624 #ifdef SUPPORT_UTF
625 const pcre_uchar *charptr;
626 #endif
627 const pcre_uchar *callpat;
628 const pcre_uchar *data;
629 const pcre_uchar *next;
630 PCRE_PUCHAR pp;
631 const pcre_uchar *prev;
632 PCRE_PUCHAR saved_eptr;
633
634 recursion_info new_recursive;
635
636 BOOL cur_is_word;
637 BOOL condition;
638 BOOL prev_is_word;
639
640 #ifdef SUPPORT_UCP
641 int prop_type;
642 unsigned int prop_value;
643 int prop_fail_result;
644 int oclength;
645 pcre_uchar occhars[6];
646 #endif
647
648 int codelink;
649 int ctype;
650 int length;
651 int max;
652 int min;
653 unsigned int number;
654 int offset;
655 unsigned int op;
656 pcre_int32 save_capture_last;
657 int save_offset1, save_offset2, save_offset3;
658 int stacksave[REC_STACK_SAVE_MAX];
659
660 eptrblock newptrb;
661
662 /* There is a special fudge for calling match() in a way that causes it to
663 measure the size of its basic stack frame when the stack is being used for
664 recursion. The second argument (ecode) being NULL triggers this behaviour. It
665 cannot normally ever be NULL. The return is the negated value of the frame
666 size. */
667
668 if (ecode == NULL)
669 {
670 if (rdepth == 0)
671 return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672 else
673 {
674 int len = (char *)&rdepth - (char *)eptr;
675 return (len > 0)? -len : len;
676 }
677 }
678 #endif /* NO_RECURSE */
679
680 /* To save space on the stack and in the heap frame, I have doubled up on some
681 of the local variables that are used only in localised parts of the code, but
682 still need to be preserved over recursive calls of match(). These macros define
683 the alternative names that are used. */
684
685 #define allow_zero cur_is_word
686 #define cbegroup condition
687 #define code_offset codelink
688 #define condassert condition
689 #define matched_once prev_is_word
690 #define foc number
691 #define save_mark data
692
693 /* These statements are here to stop the compiler complaining about unitialized
694 variables. */
695
696 #ifdef SUPPORT_UCP
697 prop_value = 0;
698 prop_fail_result = 0;
699 #endif
700
701
702 /* This label is used for tail recursion, which is used in a few cases even
703 when NO_RECURSE is not defined, in order to reduce the amount of stack that is
704 used. Thanks to Ian Taylor for noticing this possibility and sending the
705 original patch. */
706
707 TAIL_RECURSE:
708
709 /* OK, now we can get on with the real code of the function. Recursive calls
710 are specified by the macro RMATCH and RRETURN is used to return. When
711 NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712 and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713 defined). However, RMATCH isn't like a function call because it's quite a
714 complicated macro. It has to be used in one particular way. This shouldn't,
715 however, impact performance when true recursion is being used. */
716
717 #ifdef SUPPORT_UTF
718 utf = md->utf; /* Local copy of the flag */
719 #else
720 utf = FALSE;
721 #endif
722
723 /* First check that we haven't called match() too many times, or that we
724 haven't exceeded the recursive call limit. */
725
726 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728
729 /* At the start of a group with an unlimited repeat that may match an empty
730 string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731 done this way to save having to use another function argument, which would take
732 up space on the stack. See also MATCH_CONDASSERT below.
733
734 When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735 such remembered pointers, to be checked when we hit the closing ket, in order
736 to break infinite loops that match no characters. When match() is called in
737 other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738 NOT be used with tail recursion, because the memory block that is used is on
739 the stack, so a new one may be required for each match(). */
740
741 if (md->match_function_type == MATCH_CBEGROUP)
742 {
743 newptrb.epb_saved_eptr = eptr;
744 newptrb.epb_prev = eptrb;
745 eptrb = &newptrb;
746 md->match_function_type = 0;
747 }
748
749 /* Now start processing the opcodes. */
750
751 for (;;)
752 {
753 minimize = possessive = FALSE;
754 op = *ecode;
755
756 switch(op)
757 {
758 case OP_MARK:
759 md->nomatch_mark = ecode + 2;
760 md->mark = NULL; /* In case previously set by assertion */
761 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762 eptrb, RM55);
763 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764 md->mark == NULL) md->mark = ecode + 2;
765
766 /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767 argument, and we must check whether that argument matches this MARK's
768 argument. It is passed back in md->start_match_ptr (an overloading of that
769 variable). If it does match, we reset that variable to the current subject
770 position and return MATCH_SKIP. Otherwise, pass back the return code
771 unaltered. */
772
773 else if (rrc == MATCH_SKIP_ARG &&
774 STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775 {
776 md->start_match_ptr = eptr;
777 RRETURN(MATCH_SKIP);
778 }
779 RRETURN(rrc);
780
781 case OP_FAIL:
782 RRETURN(MATCH_NOMATCH);
783
784 case OP_COMMIT:
785 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786 eptrb, RM52);
787 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788 RRETURN(MATCH_COMMIT);
789
790 case OP_PRUNE:
791 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792 eptrb, RM51);
793 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794 RRETURN(MATCH_PRUNE);
795
796 case OP_PRUNE_ARG:
797 md->nomatch_mark = ecode + 2;
798 md->mark = NULL; /* In case previously set by assertion */
799 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800 eptrb, RM56);
801 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
802 md->mark == NULL) md->mark = ecode + 2;
803 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
804 RRETURN(MATCH_PRUNE);
805
806 case OP_SKIP:
807 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
808 eptrb, RM53);
809 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
810 md->start_match_ptr = eptr; /* Pass back current position */
811 RRETURN(MATCH_SKIP);
812
813 /* Note that, for Perl compatibility, SKIP with an argument does NOT set
814 nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
815 not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
816 that failed and any that preceed it (either they also failed, or were not
817 triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
818 SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
819 set to the count of the one that failed. */
820
821 case OP_SKIP_ARG:
822 md->skip_arg_count++;
823 if (md->skip_arg_count <= md->ignore_skip_arg)
824 {
825 ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
826 break;
827 }
828 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
829 eptrb, RM57);
830 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831
832 /* Pass back the current skip name by overloading md->start_match_ptr and
833 returning the special MATCH_SKIP_ARG return code. This will either be
834 caught by a matching MARK, or get to the top, where it causes a rematch
835 with md->ignore_skip_arg set to the value of md->skip_arg_count. */
836
837 md->start_match_ptr = ecode + 2;
838 RRETURN(MATCH_SKIP_ARG);
839
840 /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
841 the branch in which it occurs can be determined. Overload the start of
842 match pointer to do this. */
843
844 case OP_THEN:
845 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
846 eptrb, RM54);
847 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
848 md->start_match_ptr = ecode;
849 RRETURN(MATCH_THEN);
850
851 case OP_THEN_ARG:
852 md->nomatch_mark = ecode + 2;
853 md->mark = NULL; /* In case previously set by assertion */
854 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
855 md, eptrb, RM58);
856 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
857 md->mark == NULL) md->mark = ecode + 2;
858 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859 md->start_match_ptr = ecode;
860 RRETURN(MATCH_THEN);
861
862 /* Handle an atomic group that does not contain any capturing parentheses.
863 This can be handled like an assertion. Prior to 8.13, all atomic groups
864 were handled this way. In 8.13, the code was changed as below for ONCE, so
865 that backups pass through the group and thereby reset captured values.
866 However, this uses a lot more stack, so in 8.20, atomic groups that do not
867 contain any captures generate OP_ONCE_NC, which can be handled in the old,
868 less stack intensive way.
869
870 Check the alternative branches in turn - the matching won't pass the KET
871 for this kind of subpattern. If any one branch matches, we carry on as at
872 the end of a normal bracket, leaving the subject pointer, but resetting
873 the start-of-match value in case it was changed by \K. */
874
875 case OP_ONCE_NC:
876 prev = ecode;
877 saved_eptr = eptr;
878 save_mark = md->mark;
879 do
880 {
881 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
882 if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */
883 {
884 mstart = md->start_match_ptr;
885 break;
886 }
887 if (rrc == MATCH_THEN)
888 {
889 next = ecode + GET(ecode,1);
890 if (md->start_match_ptr < next &&
891 (*ecode == OP_ALT || *next == OP_ALT))
892 rrc = MATCH_NOMATCH;
893 }
894
895 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
896 ecode += GET(ecode,1);
897 md->mark = save_mark;
898 }
899 while (*ecode == OP_ALT);
900
901 /* If hit the end of the group (which could be repeated), fail */
902
903 if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
904
905 /* Continue as from after the group, updating the offsets high water
906 mark, since extracts may have been taken. */
907
908 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
909
910 offset_top = md->end_offset_top;
911 eptr = md->end_match_ptr;
912
913 /* For a non-repeating ket, just continue at this level. This also
914 happens for a repeating ket if no characters were matched in the group.
915 This is the forcible breaking of infinite loops as implemented in Perl
916 5.005. */
917
918 if (*ecode == OP_KET || eptr == saved_eptr)
919 {
920 ecode += 1+LINK_SIZE;
921 break;
922 }
923
924 /* The repeating kets try the rest of the pattern or restart from the
925 preceding bracket, in the appropriate order. The second "call" of match()
926 uses tail recursion, to avoid using another stack frame. */
927
928 if (*ecode == OP_KETRMIN)
929 {
930 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
931 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
932 ecode = prev;
933 goto TAIL_RECURSE;
934 }
935 else /* OP_KETRMAX */
936 {
937 RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
938 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
939 ecode += 1 + LINK_SIZE;
940 goto TAIL_RECURSE;
941 }
942 /* Control never gets here */
943
944 /* Handle a capturing bracket, other than those that are possessive with an
945 unlimited repeat. If there is space in the offset vector, save the current
946 subject position in the working slot at the top of the vector. We mustn't
947 change the current values of the data slot, because they may be set from a
948 previous iteration of this group, and be referred to by a reference inside
949 the group. A failure to match might occur after the group has succeeded,
950 if something later on doesn't match. For this reason, we need to restore
951 the working value and also the values of the final offsets, in case they
952 were set by a previous iteration of the same bracket.
953
954 If there isn't enough space in the offset vector, treat this as if it were
955 a non-capturing bracket. Don't worry about setting the flag for the error
956 case here; that is handled in the code for KET. */
957
958 case OP_CBRA:
959 case OP_SCBRA:
960 number = GET2(ecode, 1+LINK_SIZE);
961 offset = number << 1;
962
963 #ifdef PCRE_DEBUG
964 printf("start bracket %d\n", number);
965 printf("subject=");
966 pchars(eptr, 16, TRUE, md);
967 printf("\n");
968 #endif
969
970 if (offset < md->offset_max)
971 {
972 save_offset1 = md->offset_vector[offset];
973 save_offset2 = md->offset_vector[offset+1];
974 save_offset3 = md->offset_vector[md->offset_end - number];
975 save_capture_last = md->capture_last;
976 save_mark = md->mark;
977
978 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
979 md->offset_vector[md->offset_end - number] =
980 (int)(eptr - md->start_subject);
981
982 for (;;)
983 {
984 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
986 eptrb, RM1);
987 if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */
988
989 /* If we backed up to a THEN, check whether it is within the current
990 branch by comparing the address of the THEN that is passed back with
991 the end of the branch. If it is within the current branch, and the
992 branch is one of two or more alternatives (it either starts or ends
993 with OP_ALT), we have reached the limit of THEN's action, so convert
994 the return code to NOMATCH, which will cause normal backtracking to
995 happen from now on. Otherwise, THEN is passed back to an outer
996 alternative. This implements Perl's treatment of parenthesized groups,
997 where a group not containing | does not affect the current alternative,
998 that is, (X) is NOT the same as (X|(*F)). */
999
1000 if (rrc == MATCH_THEN)
1001 {
1002 next = ecode + GET(ecode,1);
1003 if (md->start_match_ptr < next &&
1004 (*ecode == OP_ALT || *next == OP_ALT))
1005 rrc = MATCH_NOMATCH;
1006 }
1007
1008 /* Anything other than NOMATCH is passed back. */
1009
1010 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1011 md->capture_last = save_capture_last;
1012 ecode += GET(ecode, 1);
1013 md->mark = save_mark;
1014 if (*ecode != OP_ALT) break;
1015 }
1016
1017 DPRINTF(("bracket %d failed\n", number));
1018 md->offset_vector[offset] = save_offset1;
1019 md->offset_vector[offset+1] = save_offset2;
1020 md->offset_vector[md->offset_end - number] = save_offset3;
1021
1022 /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1023
1024 RRETURN(rrc);
1025 }
1026
1027 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1028 as a non-capturing bracket. */
1029
1030 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1031 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1032
1033 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1034
1035 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037
1038 /* Non-capturing or atomic group, except for possessive with unlimited
1039 repeat and ONCE group with no captures. Loop for all the alternatives.
1040
1041 When we get to the final alternative within the brackets, we used to return
1042 the result of a recursive call to match() whatever happened so it was
1043 possible to reduce stack usage by turning this into a tail recursion,
1044 except in the case of a possibly empty group. However, now that there is
1045 the possiblity of (*THEN) occurring in the final alternative, this
1046 optimization is no longer always possible.
1047
1048 We can optimize if we know there are no (*THEN)s in the pattern; at present
1049 this is the best that can be done.
1050
1051 MATCH_ONCE is returned when the end of an atomic group is successfully
1052 reached, but subsequent matching fails. It passes back up the tree (causing
1053 captured values to be reset) until the original atomic group level is
1054 reached. This is tested by comparing md->once_target with the start of the
1055 group. At this point, the return is converted into MATCH_NOMATCH so that
1056 previous backup points can be taken. */
1057
1058 case OP_ONCE:
1059 case OP_BRA:
1060 case OP_SBRA:
1061 DPRINTF(("start non-capturing bracket\n"));
1062
1063 for (;;)
1064 {
1065 if (op >= OP_SBRA || op == OP_ONCE)
1066 md->match_function_type = MATCH_CBEGROUP;
1067
1068 /* If this is not a possibly empty group, and there are no (*THEN)s in
1069 the pattern, and this is the final alternative, optimize as described
1070 above. */
1071
1072 else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073 {
1074 ecode += PRIV(OP_lengths)[*ecode];
1075 goto TAIL_RECURSE;
1076 }
1077
1078 /* In all other cases, we have to make another call to match(). */
1079
1080 save_mark = md->mark;
1081 save_capture_last = md->capture_last;
1082 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1083 RM2);
1084
1085 /* See comment in the code for capturing groups above about handling
1086 THEN. */
1087
1088 if (rrc == MATCH_THEN)
1089 {
1090 next = ecode + GET(ecode,1);
1091 if (md->start_match_ptr < next &&
1092 (*ecode == OP_ALT || *next == OP_ALT))
1093 rrc = MATCH_NOMATCH;
1094 }
1095
1096 if (rrc != MATCH_NOMATCH)
1097 {
1098 if (rrc == MATCH_ONCE)
1099 {
1100 const pcre_uchar *scode = ecode;
1101 if (*scode != OP_ONCE) /* If not at start, find it */
1102 {
1103 while (*scode == OP_ALT) scode += GET(scode, 1);
1104 scode -= GET(scode, 1);
1105 }
1106 if (md->once_target == scode) rrc = MATCH_NOMATCH;
1107 }
1108 RRETURN(rrc);
1109 }
1110 ecode += GET(ecode, 1);
1111 md->mark = save_mark;
1112 if (*ecode != OP_ALT) break;
1113 md->capture_last = save_capture_last;
1114 }
1115
1116 RRETURN(MATCH_NOMATCH);
1117
1118 /* Handle possessive capturing brackets with an unlimited repeat. We come
1119 here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120 handled similarly to the normal case above. However, the matching is
1121 different. The end of these brackets will always be OP_KETRPOS, which
1122 returns MATCH_KETRPOS without going further in the pattern. By this means
1123 we can handle the group by iteration rather than recursion, thereby
1124 reducing the amount of stack needed. */
1125
1126 case OP_CBRAPOS:
1127 case OP_SCBRAPOS:
1128 allow_zero = FALSE;
1129
1130 POSSESSIVE_CAPTURE:
1131 number = GET2(ecode, 1+LINK_SIZE);
1132 offset = number << 1;
1133
1134 #ifdef PCRE_DEBUG
1135 printf("start possessive bracket %d\n", number);
1136 printf("subject=");
1137 pchars(eptr, 16, TRUE, md);
1138 printf("\n");
1139 #endif
1140
1141 if (offset < md->offset_max)
1142 {
1143 matched_once = FALSE;
1144 code_offset = (int)(ecode - md->start_code);
1145
1146 save_offset1 = md->offset_vector[offset];
1147 save_offset2 = md->offset_vector[offset+1];
1148 save_offset3 = md->offset_vector[md->offset_end - number];
1149 save_capture_last = md->capture_last;
1150
1151 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1152
1153 /* Each time round the loop, save the current subject position for use
1154 when the group matches. For MATCH_MATCH, the group has matched, so we
1155 restart it with a new subject starting position, remembering that we had
1156 at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157 usual. If we haven't matched any alternatives in any iteration, check to
1158 see if a previous iteration matched. If so, the group has matched;
1159 continue from afterwards. Otherwise it has failed; restore the previous
1160 capture values before returning NOMATCH. */
1161
1162 for (;;)
1163 {
1164 md->offset_vector[md->offset_end - number] =
1165 (int)(eptr - md->start_subject);
1166 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168 eptrb, RM63);
1169 if (rrc == MATCH_KETRPOS)
1170 {
1171 offset_top = md->end_offset_top;
1172 eptr = md->end_match_ptr;
1173 ecode = md->start_code + code_offset;
1174 save_capture_last = md->capture_last;
1175 matched_once = TRUE;
1176 continue;
1177 }
1178
1179 /* See comment in the code for capturing groups above about handling
1180 THEN. */
1181
1182 if (rrc == MATCH_THEN)
1183 {
1184 next = ecode + GET(ecode,1);
1185 if (md->start_match_ptr < next &&
1186 (*ecode == OP_ALT || *next == OP_ALT))
1187 rrc = MATCH_NOMATCH;
1188 }
1189
1190 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1191 md->capture_last = save_capture_last;
1192 ecode += GET(ecode, 1);
1193 if (*ecode != OP_ALT) break;
1194 }
1195
1196 if (!matched_once)
1197 {
1198 md->offset_vector[offset] = save_offset1;
1199 md->offset_vector[offset+1] = save_offset2;
1200 md->offset_vector[md->offset_end - number] = save_offset3;
1201 }
1202
1203 if (allow_zero || matched_once)
1204 {
1205 ecode += 1 + LINK_SIZE;
1206 break;
1207 }
1208
1209 RRETURN(MATCH_NOMATCH);
1210 }
1211
1212 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213 as a non-capturing bracket. */
1214
1215 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1217
1218 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1219
1220 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1222
1223 /* Non-capturing possessive bracket with unlimited repeat. We come here
1224 from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1225 without the capturing complication. It is written out separately for speed
1226 and cleanliness. */
1227
1228 case OP_BRAPOS:
1229 case OP_SBRAPOS:
1230 allow_zero = FALSE;
1231
1232 POSSESSIVE_NON_CAPTURE:
1233 matched_once = FALSE;
1234 code_offset = (int)(ecode - md->start_code);
1235 save_capture_last = md->capture_last;
1236
1237 for (;;)
1238 {
1239 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1241 eptrb, RM48);
1242 if (rrc == MATCH_KETRPOS)
1243 {
1244 offset_top = md->end_offset_top;
1245 eptr = md->end_match_ptr;
1246 ecode = md->start_code + code_offset;
1247 matched_once = TRUE;
1248 continue;
1249 }
1250
1251 /* See comment in the code for capturing groups above about handling
1252 THEN. */
1253
1254 if (rrc == MATCH_THEN)
1255 {
1256 next = ecode + GET(ecode,1);
1257 if (md->start_match_ptr < next &&
1258 (*ecode == OP_ALT || *next == OP_ALT))
1259 rrc = MATCH_NOMATCH;
1260 }
1261
1262 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1263 ecode += GET(ecode, 1);
1264 if (*ecode != OP_ALT) break;
1265 md->capture_last = save_capture_last;
1266 }
1267
1268 if (matched_once || allow_zero)
1269 {
1270 ecode += 1 + LINK_SIZE;
1271 break;
1272 }
1273 RRETURN(MATCH_NOMATCH);
1274
1275 /* Control never reaches here. */
1276
1277 /* Conditional group: compilation checked that there are no more than
1278 two branches. If the condition is false, skipping the first branch takes us
1279 past the end if there is only one branch, but that's OK because that is
1280 exactly what going to the ket would do. */
1281
1282 case OP_COND:
1283 case OP_SCOND:
1284 codelink = GET(ecode, 1);
1285
1286 /* Because of the way auto-callout works during compile, a callout item is
1287 inserted between OP_COND and an assertion condition. */
1288
1289 if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1290 {
1291 if (PUBL(callout) != NULL)
1292 {
1293 PUBL(callout_block) cb;
1294 cb.version = 2; /* Version 1 of the callout block */
1295 cb.callout_number = ecode[LINK_SIZE+2];
1296 cb.offset_vector = md->offset_vector;
1297 #if defined COMPILE_PCRE8
1298 cb.subject = (PCRE_SPTR)md->start_subject;
1299 #elif defined COMPILE_PCRE16
1300 cb.subject = (PCRE_SPTR16)md->start_subject;
1301 #elif defined COMPILE_PCRE32
1302 cb.subject = (PCRE_SPTR32)md->start_subject;
1303 #endif
1304 cb.subject_length = (int)(md->end_subject - md->start_subject);
1305 cb.start_match = (int)(mstart - md->start_subject);
1306 cb.current_position = (int)(eptr - md->start_subject);
1307 cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1308 cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1309 cb.capture_top = offset_top/2;
1310 cb.capture_last = md->capture_last & CAPLMASK;
1311 /* Internal change requires this for API compatibility. */
1312 if (cb.capture_last == 0) cb.capture_last = -1;
1313 cb.callout_data = md->callout_data;
1314 cb.mark = md->nomatch_mark;
1315 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1316 if (rrc < 0) RRETURN(rrc);
1317 }
1318 ecode += PRIV(OP_lengths)[OP_CALLOUT];
1319 codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1320 }
1321
1322 condcode = ecode[LINK_SIZE+1];
1323
1324 /* Now see what the actual condition is */
1325
1326 if (condcode == OP_RREF || condcode == OP_NRREF) /* Recursion test */
1327 {
1328 if (md->recursive == NULL) /* Not recursing => FALSE */
1329 {
1330 condition = FALSE;
1331 ecode += GET(ecode, 1);
1332 }
1333 else
1334 {
1335 unsigned int recno = GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/
1336 condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1337
1338 /* If the test is for recursion into a specific subpattern, and it is
1339 false, but the test was set up by name, scan the table to see if the
1340 name refers to any other numbers, and test them. The condition is true
1341 if any one is set. */
1342
1343 if (!condition && condcode == OP_NRREF)
1344 {
1345 pcre_uchar *slotA = md->name_table;
1346 for (i = 0; i < md->name_count; i++)
1347 {
1348 if (GET2(slotA, 0) == recno) break;
1349 slotA += md->name_entry_size;
1350 }
1351
1352 /* Found a name for the number - there can be only one; duplicate
1353 names for different numbers are allowed, but not vice versa. First
1354 scan down for duplicates. */
1355
1356 if (i < md->name_count)
1357 {
1358 pcre_uchar *slotB = slotA;
1359 while (slotB > md->name_table)
1360 {
1361 slotB -= md->name_entry_size;
1362 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1363 {
1364 condition = GET2(slotB, 0) == md->recursive->group_num;
1365 if (condition) break;
1366 }
1367 else break;
1368 }
1369
1370 /* Scan up for duplicates */
1371
1372 if (!condition)
1373 {
1374 slotB = slotA;
1375 for (i++; i < md->name_count; i++)
1376 {
1377 slotB += md->name_entry_size;
1378 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1379 {
1380 condition = GET2(slotB, 0) == md->recursive->group_num;
1381 if (condition) break;
1382 }
1383 else break;
1384 }
1385 }
1386 }
1387 }
1388
1389 /* Chose branch according to the condition */
1390
1391 ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1392 }
1393 }
1394
1395 else if (condcode == OP_CREF || condcode == OP_NCREF) /* Group used test */
1396 {
1397 offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */
1398 condition = offset < offset_top && md->offset_vector[offset] >= 0;
1399
1400 /* If the numbered capture is unset, but the reference was by name,
1401 scan the table to see if the name refers to any other numbers, and test
1402 them. The condition is true if any one is set. This is tediously similar
1403 to the code above, but not close enough to try to amalgamate. */
1404
1405 if (!condition && condcode == OP_NCREF)
1406 {
1407 unsigned int refno = offset >> 1;
1408 pcre_uchar *slotA = md->name_table;
1409
1410 for (i = 0; i < md->name_count; i++)
1411 {
1412 if (GET2(slotA, 0) == refno) break;
1413 slotA += md->name_entry_size;
1414 }
1415
1416 /* Found a name for the number - there can be only one; duplicate names
1417 for different numbers are allowed, but not vice versa. First scan down
1418 for duplicates. */
1419
1420 if (i < md->name_count)
1421 {
1422 pcre_uchar *slotB = slotA;
1423 while (slotB > md->name_table)
1424 {
1425 slotB -= md->name_entry_size;
1426 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1427 {
1428 offset = GET2(slotB, 0) << 1;
1429 condition = offset < offset_top &&
1430 md->offset_vector[offset] >= 0;
1431 if (condition) break;
1432 }
1433 else break;
1434 }
1435
1436 /* Scan up for duplicates */
1437
1438 if (!condition)
1439 {
1440 slotB = slotA;
1441 for (i++; i < md->name_count; i++)
1442 {
1443 slotB += md->name_entry_size;
1444 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1445 {
1446 offset = GET2(slotB, 0) << 1;
1447 condition = offset < offset_top &&
1448 md->offset_vector[offset] >= 0;
1449 if (condition) break;
1450 }
1451 else break;
1452 }
1453 }
1454 }
1455 }
1456
1457 /* Chose branch according to the condition */
1458
1459 ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1460 }
1461
1462 else if (condcode == OP_DEF) /* DEFINE - always false */
1463 {
1464 condition = FALSE;
1465 ecode += GET(ecode, 1);
1466 }
1467
1468 /* The condition is an assertion. Call match() to evaluate it - setting
1469 md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1470 an assertion. */
1471
1472 else
1473 {
1474 md->match_function_type = MATCH_CONDASSERT;
1475 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1476 if (rrc == MATCH_MATCH)
1477 {
1478 if (md->end_offset_top > offset_top)
1479 offset_top = md->end_offset_top; /* Captures may have happened */
1480 condition = TRUE;
1481 ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1482 while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1483 }
1484
1485 /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1486 assertion; it is therefore treated as NOMATCH. */
1487
1488 else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1489 {
1490 RRETURN(rrc); /* Need braces because of following else */
1491 }
1492 else
1493 {
1494 condition = FALSE;
1495 ecode += codelink;
1496 }
1497 }
1498
1499 /* We are now at the branch that is to be obeyed. As there is only one, can
1500 use tail recursion to avoid using another stack frame, except when there is
1501 unlimited repeat of a possibly empty group. In the latter case, a recursive
1502 call to match() is always required, unless the second alternative doesn't
1503 exist, in which case we can just plough on. Note that, for compatibility
1504 with Perl, the | in a conditional group is NOT treated as creating two
1505 alternatives. If a THEN is encountered in the branch, it propagates out to
1506 the enclosing alternative (unless nested in a deeper set of alternatives,
1507 of course). */
1508
1509 if (condition || *ecode == OP_ALT)
1510 {
1511 if (op != OP_SCOND)
1512 {
1513 ecode += 1 + LINK_SIZE;
1514 goto TAIL_RECURSE;
1515 }
1516
1517 md->match_function_type = MATCH_CBEGROUP;
1518 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1519 RRETURN(rrc);
1520 }
1521
1522 /* Condition false & no alternative; continue after the group. */
1523
1524 else
1525 {
1526 ecode += 1 + LINK_SIZE;
1527 }
1528 break;
1529
1530
1531 /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1532 to close any currently open capturing brackets. */
1533
1534 case OP_CLOSE:
1535 number = GET2(ecode, 1); /* Must be less than 65536 */
1536 offset = number << 1;
1537
1538 #ifdef PCRE_DEBUG
1539 printf("end bracket %d at *ACCEPT", number);
1540 printf("\n");
1541 #endif
1542
1543 md->capture_last = (md->capture_last & OVFLMASK) | number;
1544 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1545 {
1546 md->offset_vector[offset] =
1547 md->offset_vector[md->offset_end - number];
1548 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1549 if (offset_top <= offset) offset_top = offset + 2;
1550 }
1551 ecode += 1 + IMM2_SIZE;
1552 break;
1553
1554
1555 /* End of the pattern, either real or forced. */
1556
1557 case OP_END:
1558 case OP_ACCEPT:
1559 case OP_ASSERT_ACCEPT:
1560
1561 /* If we have matched an empty string, fail if not in an assertion and not
1562 in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1563 is set and we have matched at the start of the subject. In both cases,
1564 backtracking will then try other alternatives, if any. */
1565
1566 if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1567 md->recursive == NULL &&
1568 (md->notempty ||
1569 (md->notempty_atstart &&
1570 mstart == md->start_subject + md->start_offset)))
1571 RRETURN(MATCH_NOMATCH);
1572
1573 /* Otherwise, we have a match. */
1574
1575 md->end_match_ptr = eptr; /* Record where we ended */
1576 md->end_offset_top = offset_top; /* and how many extracts were taken */
1577 md->start_match_ptr = mstart; /* and the start (\K can modify) */
1578
1579 /* For some reason, the macros don't work properly if an expression is
1580 given as the argument to RRETURN when the heap is in use. */
1581
1582 rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1583 RRETURN(rrc);
1584
1585 /* Assertion brackets. Check the alternative branches in turn - the
1586 matching won't pass the KET for an assertion. If any one branch matches,
1587 the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1588 start of each branch to move the current point backwards, so the code at
1589 this level is identical to the lookahead case. When the assertion is part
1590 of a condition, we want to return immediately afterwards. The caller of
1591 this incarnation of the match() function will have set MATCH_CONDASSERT in
1592 md->match_function type, and one of these opcodes will be the first opcode
1593 that is processed. We use a local variable that is preserved over calls to
1594 match() to remember this case. */
1595
1596 case OP_ASSERT:
1597 case OP_ASSERTBACK:
1598 save_mark = md->mark;
1599 if (md->match_function_type == MATCH_CONDASSERT)
1600 {
1601 condassert = TRUE;
1602 md->match_function_type = 0;
1603 }
1604 else condassert = FALSE;
1605
1606 /* Loop for each branch */
1607
1608 do
1609 {
1610 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1611
1612 /* A match means that the assertion is true; break out of the loop
1613 that matches its alternatives. */
1614
1615 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1616 {
1617 mstart = md->start_match_ptr; /* In case \K reset it */
1618 break;
1619 }
1620
1621 /* If not matched, restore the previous mark setting. */
1622
1623 md->mark = save_mark;
1624
1625 /* See comment in the code for capturing groups above about handling
1626 THEN. */
1627
1628 if (rrc == MATCH_THEN)
1629 {
1630 next = ecode + GET(ecode,1);
1631 if (md->start_match_ptr < next &&
1632 (*ecode == OP_ALT || *next == OP_ALT))
1633 rrc = MATCH_NOMATCH;
1634 }
1635
1636 /* Anything other than NOMATCH causes the entire assertion to fail,
1637 passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1638 uncaptured THEN, which means they take their normal effect. This
1639 consistent approach does not always have exactly the same effect as in
1640 Perl. */
1641
1642 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1643 ecode += GET(ecode, 1);
1644 }
1645 while (*ecode == OP_ALT); /* Continue for next alternative */
1646
1647 /* If we have tried all the alternative branches, the assertion has
1648 failed. If not, we broke out after a match. */
1649
1650 if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1651
1652 /* If checking an assertion for a condition, return MATCH_MATCH. */
1653
1654 if (condassert) RRETURN(MATCH_MATCH);
1655
1656 /* Continue from after a successful assertion, updating the offsets high
1657 water mark, since extracts may have been taken during the assertion. */
1658
1659 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1660 ecode += 1 + LINK_SIZE;
1661 offset_top = md->end_offset_top;
1662 continue;
1663
1664 /* Negative assertion: all branches must fail to match for the assertion to
1665 succeed. */
1666
1667 case OP_ASSERT_NOT:
1668 case OP_ASSERTBACK_NOT:
1669 save_mark = md->mark;
1670 if (md->match_function_type == MATCH_CONDASSERT)
1671 {
1672 condassert = TRUE;
1673 md->match_function_type = 0;
1674 }
1675 else condassert = FALSE;
1676
1677 /* Loop for each alternative branch. */
1678
1679 do
1680 {
1681 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1682 md->mark = save_mark; /* Always restore the mark setting */
1683
1684 switch(rrc)
1685 {
1686 case MATCH_MATCH: /* A successful match means */
1687 case MATCH_ACCEPT: /* the assertion has failed. */
1688 RRETURN(MATCH_NOMATCH);
1689
1690 case MATCH_NOMATCH: /* Carry on with next branch */
1691 break;
1692
1693 /* See comment in the code for capturing groups above about handling
1694 THEN. */
1695
1696 case MATCH_THEN:
1697 next = ecode + GET(ecode,1);
1698 if (md->start_match_ptr < next &&
1699 (*ecode == OP_ALT || *next == OP_ALT))
1700 {
1701 rrc = MATCH_NOMATCH;
1702 break;
1703 }
1704 /* Otherwise fall through. */
1705
1706 /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1707 assertion to fail to match, without considering any more alternatives.
1708 Failing to match means the assertion is true. This is a consistent
1709 approach, but does not always have the same effect as in Perl. */
1710
1711 case MATCH_COMMIT:
1712 case MATCH_SKIP:
1713 case MATCH_SKIP_ARG:
1714 case MATCH_PRUNE:
1715 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1716 goto NEG_ASSERT_TRUE; /* Break out of alternation loop */
1717
1718 /* Anything else is an error */
1719
1720 default:
1721 RRETURN(rrc);
1722 }
1723
1724 /* Continue with next branch */
1725
1726 ecode += GET(ecode,1);
1727 }
1728 while (*ecode == OP_ALT);
1729
1730 /* All branches in the assertion failed to match. */
1731
1732 NEG_ASSERT_TRUE:
1733 if (condassert) RRETURN(MATCH_MATCH); /* Condition assertion */
1734 ecode += 1 + LINK_SIZE; /* Continue with current branch */
1735 continue;
1736
1737 /* Move the subject pointer back. This occurs only at the start of
1738 each branch of a lookbehind assertion. If we are too close to the start to
1739 move back, this match function fails. When working with UTF-8 we move
1740 back a number of characters, not bytes. */
1741
1742 case OP_REVERSE:
1743 #ifdef SUPPORT_UTF
1744 if (utf)
1745 {
1746 i = GET(ecode, 1);
1747 while (i-- > 0)
1748 {
1749 eptr--;
1750 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1751 BACKCHAR(eptr);
1752 }
1753 }
1754 else
1755 #endif
1756
1757 /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1758
1759 {
1760 eptr -= GET(ecode, 1);
1761 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1762 }
1763
1764 /* Save the earliest consulted character, then skip to next op code */
1765
1766 if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1767 ecode += 1 + LINK_SIZE;
1768 break;
1769
1770 /* The callout item calls an external function, if one is provided, passing
1771 details of the match so far. This is mainly for debugging, though the
1772 function is able to force a failure. */
1773
1774 case OP_CALLOUT:
1775 if (PUBL(callout) != NULL)
1776 {
1777 PUBL(callout_block) cb;
1778 cb.version = 2; /* Version 1 of the callout block */
1779 cb.callout_number = ecode[1];
1780 cb.offset_vector = md->offset_vector;
1781 #if defined COMPILE_PCRE8
1782 cb.subject = (PCRE_SPTR)md->start_subject;
1783 #elif defined COMPILE_PCRE16
1784 cb.subject = (PCRE_SPTR16)md->start_subject;
1785 #elif defined COMPILE_PCRE32
1786 cb.subject = (PCRE_SPTR32)md->start_subject;
1787 #endif
1788 cb.subject_length = (int)(md->end_subject - md->start_subject);
1789 cb.start_match = (int)(mstart - md->start_subject);
1790 cb.current_position = (int)(eptr - md->start_subject);
1791 cb.pattern_position = GET(ecode, 2);
1792 cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1793 cb.capture_top = offset_top/2;
1794 cb.capture_last = md->capture_last & CAPLMASK;
1795 /* Internal change requires this for API compatibility. */
1796 if (cb.capture_last == 0) cb.capture_last = -1;
1797 cb.callout_data = md->callout_data;
1798 cb.mark = md->nomatch_mark;
1799 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1800 if (rrc < 0) RRETURN(rrc);
1801 }
1802 ecode += 2 + 2*LINK_SIZE;
1803 break;
1804
1805 /* Recursion either matches the current regex, or some subexpression. The
1806 offset data is the offset to the starting bracket from the start of the
1807 whole pattern. (This is so that it works from duplicated subpatterns.)
1808
1809 The state of the capturing groups is preserved over recursion, and
1810 re-instated afterwards. We don't know how many are started and not yet
1811 finished (offset_top records the completed total) so we just have to save
1812 all the potential data. There may be up to 65535 such values, which is too
1813 large to put on the stack, but using malloc for small numbers seems
1814 expensive. As a compromise, the stack is used when there are no more than
1815 REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
1816
1817 There are also other values that have to be saved. We use a chained
1818 sequence of blocks that actually live on the stack. Thanks to Robin Houston
1819 for the original version of this logic. It has, however, been hacked around
1820 a lot, so he is not to blame for the current way it works. */
1821
1822 case OP_RECURSE:
1823 {
1824 recursion_info *ri;
1825 unsigned int recno;
1826
1827 callpat = md->start_code + GET(ecode, 1);
1828 recno = (callpat == md->start_code)? 0 :
1829 GET2(callpat, 1 + LINK_SIZE);
1830
1831 /* Check for repeating a recursion without advancing the subject pointer.
1832 This should catch convoluted mutual recursions. (Some simple cases are
1833 caught at compile time.) */
1834
1835 for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1836 if (recno == ri->group_num && eptr == ri->subject_position)
1837 RRETURN(PCRE_ERROR_RECURSELOOP);
1838
1839 /* Add to "recursing stack" */
1840
1841 new_recursive.group_num = recno;
1842 new_recursive.saved_capture_last = md->capture_last;
1843 new_recursive.subject_position = eptr;
1844 new_recursive.prevrec = md->recursive;
1845 md->recursive = &new_recursive;
1846
1847 /* Where to continue from afterwards */
1848
1849 ecode += 1 + LINK_SIZE;
1850
1851 /* Now save the offset data */
1852
1853 new_recursive.saved_max = md->offset_end;
1854 if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
1855 new_recursive.offset_save = stacksave;
1856 else
1857 {
1858 new_recursive.offset_save =
1859 (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1860 if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1861 }
1862 memcpy(new_recursive.offset_save, md->offset_vector,
1863 new_recursive.saved_max * sizeof(int));
1864
1865 /* OK, now we can do the recursion. After processing each alternative,
1866 restore the offset data and the last captured value. If there were nested
1867 recursions, md->recursive might be changed, so reset it before looping.
1868 */
1869
1870 DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1871 cbegroup = (*callpat >= OP_SBRA);
1872 do
1873 {
1874 if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1875 RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1876 md, eptrb, RM6);
1877 memcpy(md->offset_vector, new_recursive.offset_save,
1878 new_recursive.saved_max * sizeof(int));
1879 md->capture_last = new_recursive.saved_capture_last;
1880 md->recursive = new_recursive.prevrec;
1881 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1882 {
1883 DPRINTF(("Recursion matched\n"));
1884 if (new_recursive.offset_save != stacksave)
1885 (PUBL(free))(new_recursive.offset_save);
1886
1887 /* Set where we got to in the subject, and reset the start in case
1888 it was changed by \K. This *is* propagated back out of a recursion,
1889 for Perl compatibility. */
1890
1891 eptr = md->end_match_ptr;
1892 mstart = md->start_match_ptr;
1893 goto RECURSION_MATCHED; /* Exit loop; end processing */
1894 }
1895
1896 /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1897 recursion; they cause a NOMATCH for the entire recursion. These codes
1898 are defined in a range that can be tested for. */
1899
1900 if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1901 RRETURN(MATCH_NOMATCH);
1902
1903 /* Any return code other than NOMATCH is an error. */
1904
1905 if (rrc != MATCH_NOMATCH)
1906 {
1907 DPRINTF(("Recursion gave error %d\n", rrc));
1908 if (new_recursive.offset_save != stacksave)
1909 (PUBL(free))(new_recursive.offset_save);
1910 RRETURN(rrc);
1911 }
1912
1913 md->recursive = &new_recursive;
1914 callpat += GET(callpat, 1);
1915 }
1916 while (*callpat == OP_ALT);
1917
1918 DPRINTF(("Recursion didn't match\n"));
1919 md->recursive = new_recursive.prevrec;
1920 if (new_recursive.offset_save != stacksave)
1921 (PUBL(free))(new_recursive.offset_save);
1922 RRETURN(MATCH_NOMATCH);
1923 }
1924
1925 RECURSION_MATCHED:
1926 break;
1927
1928 /* An alternation is the end of a branch; scan along to find the end of the
1929 bracketed group and go to there. */
1930
1931 case OP_ALT:
1932 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1933 break;
1934
1935 /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1936 indicating that it may occur zero times. It may repeat infinitely, or not
1937 at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1938 with fixed upper repeat limits are compiled as a number of copies, with the
1939 optional ones preceded by BRAZERO or BRAMINZERO. */
1940
1941 case OP_BRAZERO:
1942 next = ecode + 1;
1943 RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1944 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1945 do next += GET(next, 1); while (*next == OP_ALT);
1946 ecode = next + 1 + LINK_SIZE;
1947 break;
1948
1949 case OP_BRAMINZERO:
1950 next = ecode + 1;
1951 do next += GET(next, 1); while (*next == OP_ALT);
1952 RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1953 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1954 ecode++;
1955 break;
1956
1957 case OP_SKIPZERO:
1958 next = ecode+1;
1959 do next += GET(next,1); while (*next == OP_ALT);
1960 ecode = next + 1 + LINK_SIZE;
1961 break;
1962
1963 /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1964 here; just jump to the group, with allow_zero set TRUE. */
1965
1966 case OP_BRAPOSZERO:
1967 op = *(++ecode);
1968 allow_zero = TRUE;
1969 if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1970 goto POSSESSIVE_NON_CAPTURE;
1971
1972 /* End of a group, repeated or non-repeating. */
1973
1974 case OP_KET:
1975 case OP_KETRMIN:
1976 case OP_KETRMAX:
1977 case OP_KETRPOS:
1978 prev = ecode - GET(ecode, 1);
1979
1980 /* If this was a group that remembered the subject start, in order to break
1981 infinite repeats of empty string matches, retrieve the subject start from
1982 the chain. Otherwise, set it NULL. */
1983
1984 if (*prev >= OP_SBRA || *prev == OP_ONCE)
1985 {
1986 saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */
1987 eptrb = eptrb->epb_prev; /* Backup to previous group */
1988 }
1989 else saved_eptr = NULL;
1990
1991 /* If we are at the end of an assertion group or a non-capturing atomic
1992 group, stop matching and return MATCH_MATCH, but record the current high
1993 water mark for use by positive assertions. We also need to record the match
1994 start in case it was changed by \K. */
1995
1996 if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1997 *prev == OP_ONCE_NC)
1998 {
1999 md->end_match_ptr = eptr; /* For ONCE_NC */
2000 md->end_offset_top = offset_top;
2001 md->start_match_ptr = mstart;
2002 RRETURN(MATCH_MATCH); /* Sets md->mark */
2003 }
2004
2005 /* For capturing groups we have to check the group number back at the start
2006 and if necessary complete handling an extraction by setting the offsets and
2007 bumping the high water mark. Whole-pattern recursion is coded as a recurse
2008 into group 0, so it won't be picked up here. Instead, we catch it when the
2009 OP_END is reached. Other recursion is handled here. We just have to record
2010 the current subject position and start match pointer and give a MATCH
2011 return. */
2012
2013 if (*prev == OP_CBRA || *prev == OP_SCBRA ||
2014 *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
2015 {
2016 number = GET2(prev, 1+LINK_SIZE);
2017 offset = number << 1;
2018
2019 #ifdef PCRE_DEBUG
2020 printf("end bracket %d", number);
2021 printf("\n");
2022 #endif
2023
2024 /* Handle a recursively called group. */
2025
2026 if (md->recursive != NULL && md->recursive->group_num == number)
2027 {
2028 md->end_match_ptr = eptr;
2029 md->start_match_ptr = mstart;
2030 RRETURN(MATCH_MATCH);
2031 }
2032
2033 /* Deal with capturing */
2034
2035 md->capture_last = (md->capture_last & OVFLMASK) | number;
2036 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
2037 {
2038 /* If offset is greater than offset_top, it means that we are
2039 "skipping" a capturing group, and that group's offsets must be marked
2040 unset. In earlier versions of PCRE, all the offsets were unset at the
2041 start of matching, but this doesn't work because atomic groups and
2042 assertions can cause a value to be set that should later be unset.
2043 Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
2044 part of the atomic group, but this is not on the final matching path,
2045 so must be unset when 2 is set. (If there is no group 2, there is no
2046 problem, because offset_top will then be 2, indicating no capture.) */
2047
2048 if (offset > offset_top)
2049 {
2050 register int *iptr = md->offset_vector + offset_top;
2051 register int *iend = md->offset_vector + offset;
2052 while (iptr < iend) *iptr++ = -1;
2053 }
2054
2055 /* Now make the extraction */
2056
2057 md->offset_vector[offset] =
2058 md->offset_vector[md->offset_end - number];
2059 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2060 if (offset_top <= offset) offset_top = offset + 2;
2061 }
2062 }
2063
2064 /* For an ordinary non-repeating ket, just continue at this level. This
2065 also happens for a repeating ket if no characters were matched in the
2066 group. This is the forcible breaking of infinite loops as implemented in
2067 Perl 5.005. For a non-repeating atomic group that includes captures,
2068 establish a backup point by processing the rest of the pattern at a lower
2069 level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2070 original OP_ONCE level, thereby bypassing intermediate backup points, but
2071 resetting any captures that happened along the way. */
2072
2073 if (*ecode == OP_KET || eptr == saved_eptr)
2074 {
2075 if (*prev == OP_ONCE)
2076 {
2077 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2078 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2079 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
2080 RRETURN(MATCH_ONCE);
2081 }
2082 ecode += 1 + LINK_SIZE; /* Carry on at this level */
2083 break;
2084 }
2085
2086 /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2087 and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2088 at a time from the outer level, thus saving stack. */
2089
2090 if (*ecode == OP_KETRPOS)
2091 {
2092 md->end_match_ptr = eptr;
2093 md->end_offset_top = offset_top;
2094 RRETURN(MATCH_KETRPOS);
2095 }
2096
2097 /* The normal repeating kets try the rest of the pattern or restart from
2098 the preceding bracket, in the appropriate order. In the second case, we can
2099 use tail recursion to avoid using another stack frame, unless we have an
2100 an atomic group or an unlimited repeat of a group that can match an empty
2101 string. */
2102
2103 if (*ecode == OP_KETRMIN)
2104 {
2105 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2106 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2107 if (*prev == OP_ONCE)
2108 {
2109 RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2110 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2111 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
2112 RRETURN(MATCH_ONCE);
2113 }
2114 if (*prev >= OP_SBRA) /* Could match an empty string */
2115 {
2116 RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2117 RRETURN(rrc);
2118 }
2119 ecode = prev;
2120 goto TAIL_RECURSE;
2121 }
2122 else /* OP_KETRMAX */
2123 {
2124 RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2125 if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2126 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2127 if (*prev == OP_ONCE)
2128 {
2129 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2130 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2131 md->once_target = prev;
2132 RRETURN(MATCH_ONCE);
2133 }
2134 ecode += 1 + LINK_SIZE;
2135 goto TAIL_RECURSE;
2136 }
2137 /* Control never gets here */
2138
2139 /* Not multiline mode: start of subject assertion, unless notbol. */
2140
2141 case OP_CIRC:
2142 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2143
2144 /* Start of subject assertion */
2145
2146 case OP_SOD:
2147 if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
2148 ecode++;
2149 break;
2150
2151 /* Multiline mode: start of subject unless notbol, or after any newline. */
2152
2153 case OP_CIRCM:
2154 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2155 if (eptr != md->start_subject &&
2156 (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2157 RRETURN(MATCH_NOMATCH);
2158 ecode++;
2159 break;
2160
2161 /* Start of match assertion */
2162
2163 case OP_SOM:
2164 if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
2165 ecode++;
2166 break;
2167
2168 /* Reset the start of match point */
2169
2170 case OP_SET_SOM:
2171 mstart = eptr;
2172 ecode++;
2173 break;
2174
2175 /* Multiline mode: assert before any newline, or before end of subject
2176 unless noteol is set. */
2177
2178 case OP_DOLLM:
2179 if (eptr < md->end_subject)
2180 {
2181 if (!IS_NEWLINE(eptr))
2182 {
2183 if (md->partial != 0 &&
2184 eptr + 1 >= md->end_subject &&
2185 NLBLOCK->nltype == NLTYPE_FIXED &&
2186 NLBLOCK->nllen == 2 &&
2187 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2188 {
2189 md->hitend = TRUE;
2190 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2191 }
2192 RRETURN(MATCH_NOMATCH);
2193 }
2194 }
2195 else
2196 {
2197 if (md->noteol) RRETURN(MATCH_NOMATCH);
2198 SCHECK_PARTIAL();
2199 }
2200 ecode++;
2201 break;
2202
2203 /* Not multiline mode: assert before a terminating newline or before end of
2204 subject unless noteol is set. */
2205
2206 case OP_DOLL:
2207 if (md->noteol) RRETURN(MATCH_NOMATCH);
2208 if (!md->endonly) goto ASSERT_NL_OR_EOS;
2209
2210 /* ... else fall through for endonly */
2211
2212 /* End of subject assertion (\z) */
2213
2214 case OP_EOD:
2215 if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2216 SCHECK_PARTIAL();
2217 ecode++;
2218 break;
2219
2220 /* End of subject or ending \n assertion (\Z) */
2221
2222 case OP_EODN:
2223 ASSERT_NL_OR_EOS:
2224 if (eptr < md->end_subject &&
2225 (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2226 {
2227 if (md->partial != 0 &&
2228 eptr + 1 >= md->end_subject &&
2229 NLBLOCK->nltype == NLTYPE_FIXED &&
2230 NLBLOCK->nllen == 2 &&
2231 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2232 {
2233 md->hitend = TRUE;
2234 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2235 }
2236 RRETURN(MATCH_NOMATCH);
2237 }
2238
2239 /* Either at end of string or \n before end. */
2240
2241 SCHECK_PARTIAL();
2242 ecode++;
2243 break;
2244
2245 /* Word boundary assertions */
2246
2247 case OP_NOT_WORD_BOUNDARY:
2248 case OP_WORD_BOUNDARY:
2249 {
2250
2251 /* Find out if the previous and current characters are "word" characters.
2252 It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2253 be "non-word" characters. Remember the earliest consulted character for
2254 partial matching. */
2255
2256 #ifdef SUPPORT_UTF
2257 if (utf)
2258 {
2259 /* Get status of previous character */
2260
2261 if (eptr == md->start_subject) prev_is_word = FALSE; else
2262 {
2263 PCRE_PUCHAR lastptr = eptr - 1;
2264 BACKCHAR(lastptr);
2265 if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2266 GETCHAR(c, lastptr);
2267 #ifdef SUPPORT_UCP
2268 if (md->use_ucp)
2269 {
2270 if (c == '_') prev_is_word = TRUE; else
2271 {
2272 int cat = UCD_CATEGORY(c);
2273 prev_is_word = (cat == ucp_L || cat == ucp_N);
2274 }
2275 }
2276 else
2277 #endif
2278 prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2279 }
2280
2281 /* Get status of next character */
2282
2283 if (eptr >= md->end_subject)
2284 {
2285 SCHECK_PARTIAL();
2286 cur_is_word = FALSE;
2287 }
2288 else
2289 {
2290 GETCHAR(c, eptr);
2291 #ifdef SUPPORT_UCP
2292 if (md->use_ucp)
2293 {
2294 if (c == '_') cur_is_word = TRUE; else
2295 {
2296 int cat = UCD_CATEGORY(c);
2297 cur_is_word = (cat == ucp_L || cat == ucp_N);
2298 }
2299 }
2300 else
2301 #endif
2302 cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2303 }
2304 }
2305 else
2306 #endif
2307
2308 /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2309 consistency with the behaviour of \w we do use it in this case. */
2310
2311 {
2312 /* Get status of previous character */
2313
2314 if (eptr == md->start_subject) prev_is_word = FALSE; else
2315 {
2316 if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2317 #ifdef SUPPORT_UCP
2318 if (md->use_ucp)
2319 {
2320 c = eptr[-1];
2321 if (c == '_') prev_is_word = TRUE; else
2322 {
2323 int cat = UCD_CATEGORY(c);
2324 prev_is_word = (cat == ucp_L || cat == ucp_N);
2325 }
2326 }
2327 else
2328 #endif
2329 prev_is_word = MAX_255(eptr[-1])
2330 && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2331 }
2332
2333 /* Get status of next character */
2334
2335 if (eptr >= md->end_subject)
2336 {
2337 SCHECK_PARTIAL();
2338 cur_is_word = FALSE;
2339 }
2340 else
2341 #ifdef SUPPORT_UCP
2342 if (md->use_ucp)
2343 {
2344 c = *eptr;
2345 if (c == '_') cur_is_word = TRUE; else
2346 {
2347 int cat = UCD_CATEGORY(c);
2348 cur_is_word = (cat == ucp_L || cat == ucp_N);
2349 }
2350 }
2351 else
2352 #endif
2353 cur_is_word = MAX_255(*eptr)
2354 && ((md->ctypes[*eptr] & ctype_word) != 0);
2355 }
2356
2357 /* Now see if the situation is what we want */
2358
2359 if ((*ecode++ == OP_WORD_BOUNDARY)?
2360 cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2361 RRETURN(MATCH_NOMATCH);
2362 }
2363 break;
2364
2365 /* Match any single character type except newline; have to take care with
2366 CRLF newlines and partial matching. */
2367
2368 case OP_ANY:
2369 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2370 if (md->partial != 0 &&
2371 eptr + 1 >= md->end_subject &&
2372 NLBLOCK->nltype == NLTYPE_FIXED &&
2373 NLBLOCK->nllen == 2 &&
2374 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2375 {
2376 md->hitend = TRUE;
2377 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2378 }
2379
2380 /* Fall through */
2381
2382 /* Match any single character whatsoever. */
2383
2384 case OP_ALLANY:
2385 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2386 { /* not be updated before SCHECK_PARTIAL. */
2387 SCHECK_PARTIAL();
2388 RRETURN(MATCH_NOMATCH);
2389 }
2390 eptr++;
2391 #ifdef SUPPORT_UTF
2392 if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2393 #endif
2394 ecode++;
2395 break;
2396
2397 /* Match a single byte, even in UTF-8 mode. This opcode really does match
2398 any byte, even newline, independent of the setting of PCRE_DOTALL. */
2399
2400 case OP_ANYBYTE:
2401 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2402 { /* not be updated before SCHECK_PARTIAL. */
2403 SCHECK_PARTIAL();
2404 RRETURN(MATCH_NOMATCH);
2405 }
2406 eptr++;
2407 ecode++;
2408 break;
2409
2410 case OP_NOT_DIGIT:
2411 if (eptr >= md->end_subject)
2412 {
2413 SCHECK_PARTIAL();
2414 RRETURN(MATCH_NOMATCH);
2415 }
2416 GETCHARINCTEST(c, eptr);
2417 if (
2418 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2419 c < 256 &&
2420 #endif
2421 (md->ctypes[c] & ctype_digit) != 0
2422 )
2423 RRETURN(MATCH_NOMATCH);
2424 ecode++;
2425 break;
2426
2427 case OP_DIGIT:
2428 if (eptr >= md->end_subject)
2429 {
2430 SCHECK_PARTIAL();
2431 RRETURN(MATCH_NOMATCH);
2432 }
2433 GETCHARINCTEST(c, eptr);
2434 if (
2435 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2436 c > 255 ||
2437 #endif
2438 (md->ctypes[c] & ctype_digit) == 0
2439 )
2440 RRETURN(MATCH_NOMATCH);
2441 ecode++;
2442 break;
2443
2444 case OP_NOT_WHITESPACE:
2445 if (eptr >= md->end_subject)
2446 {
2447 SCHECK_PARTIAL();
2448 RRETURN(MATCH_NOMATCH);
2449 }
2450 GETCHARINCTEST(c, eptr);
2451 if (
2452 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2453 c < 256 &&
2454 #endif
2455 (md->ctypes[c] & ctype_space) != 0
2456 )
2457 RRETURN(MATCH_NOMATCH);
2458 ecode++;
2459 break;
2460
2461 case OP_WHITESPACE:
2462 if (eptr >= md->end_subject)
2463 {
2464 SCHECK_PARTIAL();
2465 RRETURN(MATCH_NOMATCH);
2466 }
2467 GETCHARINCTEST(c, eptr);
2468 if (
2469 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2470 c > 255 ||
2471 #endif
2472 (md->ctypes[c] & ctype_space) == 0
2473 )
2474 RRETURN(MATCH_NOMATCH);
2475 ecode++;
2476 break;
2477
2478 case OP_NOT_WORDCHAR:
2479 if (eptr >= md->end_subject)
2480 {
2481 SCHECK_PARTIAL();
2482 RRETURN(MATCH_NOMATCH);
2483 }
2484 GETCHARINCTEST(c, eptr);
2485 if (
2486 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2487 c < 256 &&
2488 #endif
2489 (md->ctypes[c] & ctype_word) != 0
2490 )
2491 RRETURN(MATCH_NOMATCH);
2492 ecode++;
2493 break;
2494
2495 case OP_WORDCHAR:
2496 if (eptr >= md->end_subject)
2497 {
2498 SCHECK_PARTIAL();
2499 RRETURN(MATCH_NOMATCH);
2500 }
2501 GETCHARINCTEST(c, eptr);
2502 if (
2503 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2504 c > 255 ||
2505 #endif
2506 (md->ctypes[c] & ctype_word) == 0
2507 )
2508 RRETURN(MATCH_NOMATCH);
2509 ecode++;
2510 break;
2511
2512 case OP_ANYNL:
2513 if (eptr >= md->end_subject)
2514 {
2515 SCHECK_PARTIAL();
2516 RRETURN(MATCH_NOMATCH);
2517 }
2518 GETCHARINCTEST(c, eptr);
2519 switch(c)
2520 {
2521 default: RRETURN(MATCH_NOMATCH);
2522
2523 case CHAR_CR:
2524 if (eptr >= md->end_subject)
2525 {
2526 SCHECK_PARTIAL();
2527 }
2528 else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2529 break;
2530
2531 case CHAR_LF:
2532 break;
2533
2534 case CHAR_VT:
2535 case CHAR_FF:
2536 case CHAR_NEL:
2537 #ifndef EBCDIC
2538 case 0x2028:
2539 case 0x2029:
2540 #endif /* Not EBCDIC */
2541 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2542 break;
2543 }
2544 ecode++;
2545 break;
2546
2547 case OP_NOT_HSPACE:
2548 if (eptr >= md->end_subject)
2549 {
2550 SCHECK_PARTIAL();
2551 RRETURN(MATCH_NOMATCH);
2552 }
2553 GETCHARINCTEST(c, eptr);
2554 switch(c)
2555 {
2556 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
2557 default: break;
2558 }
2559 ecode++;
2560 break;
2561
2562 case OP_HSPACE:
2563 if (eptr >= md->end_subject)
2564 {
2565 SCHECK_PARTIAL();
2566 RRETURN(MATCH_NOMATCH);
2567 }
2568 GETCHARINCTEST(c, eptr);
2569 switch(c)
2570 {
2571 HSPACE_CASES: break; /* Byte and multibyte cases */
2572 default: RRETURN(MATCH_NOMATCH);
2573 }
2574 ecode++;
2575 break;
2576
2577 case OP_NOT_VSPACE:
2578 if (eptr >= md->end_subject)
2579 {
2580 SCHECK_PARTIAL();
2581 RRETURN(MATCH_NOMATCH);
2582 }
2583 GETCHARINCTEST(c, eptr);
2584 switch(c)
2585 {
2586 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2587 default: break;
2588 }
2589 ecode++;
2590 break;
2591
2592 case OP_VSPACE:
2593 if (eptr >= md->end_subject)
2594 {
2595 SCHECK_PARTIAL();
2596 RRETURN(MATCH_NOMATCH);
2597 }
2598 GETCHARINCTEST(c, eptr);
2599 switch(c)
2600 {
2601 VSPACE_CASES: break;
2602 default: RRETURN(MATCH_NOMATCH);
2603 }
2604 ecode++;
2605 break;
2606
2607 #ifdef SUPPORT_UCP
2608 /* Check the next character by Unicode property. We will get here only
2609 if the support is in the binary; otherwise a compile-time error occurs. */
2610
2611 case OP_PROP:
2612 case OP_NOTPROP:
2613 if (eptr >= md->end_subject)
2614 {
2615 SCHECK_PARTIAL();
2616 RRETURN(MATCH_NOMATCH);
2617 }
2618 GETCHARINCTEST(c, eptr);
2619 {
2620 const pcre_uint32 *cp;
2621 const ucd_record *prop = GET_UCD(c);
2622
2623 switch(ecode[1])
2624 {
2625 case PT_ANY:
2626 if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2627 break;
2628
2629 case PT_LAMP:
2630 if ((prop->chartype == ucp_Lu ||
2631 prop->chartype == ucp_Ll ||
2632 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2633 RRETURN(MATCH_NOMATCH);
2634 break;
2635
2636 case PT_GC:
2637 if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2638 RRETURN(MATCH_NOMATCH);
2639 break;
2640
2641 case PT_PC:
2642 if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2643 RRETURN(MATCH_NOMATCH);
2644 break;
2645
2646 case PT_SC:
2647 if ((ecode[2] != prop->script) == (op == OP_PROP))
2648 RRETURN(MATCH_NOMATCH);
2649 break;
2650
2651 /* These are specials */
2652
2653 case PT_ALNUM:
2654 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2655 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2656 RRETURN(MATCH_NOMATCH);
2657 break;
2658
2659 case PT_SPACE: /* Perl space */
2660 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2661 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2662 == (op == OP_NOTPROP))
2663 RRETURN(MATCH_NOMATCH);
2664 break;
2665
2666 case PT_PXSPACE: /* POSIX space */
2667 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2668 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2669 c == CHAR_FF || c == CHAR_CR)
2670 == (op == OP_NOTPROP))
2671 RRETURN(MATCH_NOMATCH);
2672 break;
2673
2674 case PT_WORD:
2675 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2676 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2677 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2678 RRETURN(MATCH_NOMATCH);
2679 break;
2680
2681 case PT_CLIST:
2682 cp = PRIV(ucd_caseless_sets) + ecode[2];
2683 for (;;)
2684 {
2685 if (c < *cp)
2686 { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2687 if (c == *cp++)
2688 { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2689 }
2690 break;
2691
2692 case PT_UCNC:
2693 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2694 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2695 c >= 0xe000) == (op == OP_NOTPROP))
2696 RRETURN(MATCH_NOMATCH);
2697 break;
2698
2699 /* This should never occur */
2700
2701 default:
2702 RRETURN(PCRE_ERROR_INTERNAL);
2703 }
2704
2705 ecode += 3;
2706 }
2707 break;
2708
2709 /* Match an extended Unicode sequence. We will get here only if the support
2710 is in the binary; otherwise a compile-time error occurs. */
2711
2712 case OP_EXTUNI:
2713 if (eptr >= md->end_subject)
2714 {
2715 SCHECK_PARTIAL();
2716 RRETURN(MATCH_NOMATCH);
2717 }
2718 else
2719 {
2720 int lgb, rgb;
2721 GETCHARINCTEST(c, eptr);
2722 lgb = UCD_GRAPHBREAK(c);
2723 while (eptr < md->end_subject)
2724 {
2725 int len = 1;
2726 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2727 rgb = UCD_GRAPHBREAK(c);
2728 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2729 lgb = rgb;
2730 eptr += len;
2731 }
2732 }
2733 CHECK_PARTIAL();
2734 ecode++;
2735 break;
2736 #endif /* SUPPORT_UCP */
2737
2738
2739 /* Match a back reference, possibly repeatedly. Look past the end of the
2740 item to see if there is repeat information following. The code is similar
2741 to that for character classes, but repeated for efficiency. Then obey
2742 similar code to character type repeats - written out again for speed.
2743 However, if the referenced string is the empty string, always treat
2744 it as matched, any number of times (otherwise there could be infinite
2745 loops). */
2746
2747 case OP_REF:
2748 case OP_REFI:
2749 caseless = op == OP_REFI;
2750 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
2751 ecode += 1 + IMM2_SIZE;
2752
2753 /* If the reference is unset, there are two possibilities:
2754
2755 (a) In the default, Perl-compatible state, set the length negative;
2756 this ensures that every attempt at a match fails. We can't just fail
2757 here, because of the possibility of quantifiers with zero minima.
2758
2759 (b) If the JavaScript compatibility flag is set, set the length to zero
2760 so that the back reference matches an empty string.
2761
2762 Otherwise, set the length to the length of what was matched by the
2763 referenced subpattern. */
2764
2765 if (offset >= offset_top || md->offset_vector[offset] < 0)
2766 length = (md->jscript_compat)? 0 : -1;
2767 else
2768 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2769
2770 /* Set up for repetition, or handle the non-repeated case */
2771
2772 switch (*ecode)
2773 {
2774 case OP_CRSTAR:
2775 case OP_CRMINSTAR:
2776 case OP_CRPLUS:
2777 case OP_CRMINPLUS:
2778 case OP_CRQUERY:
2779 case OP_CRMINQUERY:
2780 c = *ecode++ - OP_CRSTAR;
2781 minimize = (c & 1) != 0;
2782 min = rep_min[c]; /* Pick up values from tables; */
2783 max = rep_max[c]; /* zero for max => infinity */
2784 if (max == 0) max = INT_MAX;
2785 break;
2786
2787 case OP_CRRANGE:
2788 case OP_CRMINRANGE:
2789 minimize = (*ecode == OP_CRMINRANGE);
2790 min = GET2(ecode, 1);
2791 max = GET2(ecode, 1 + IMM2_SIZE);
2792 if (max == 0) max = INT_MAX;
2793 ecode += 1 + 2 * IMM2_SIZE;
2794 break;
2795
2796 default: /* No repeat follows */
2797 if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2798 {
2799 if (length == -2) eptr = md->end_subject; /* Partial match */
2800 CHECK_PARTIAL();
2801 RRETURN(MATCH_NOMATCH);
2802 }
2803 eptr += length;
2804 continue; /* With the main loop */
2805 }
2806
2807 /* Handle repeated back references. If the length of the reference is
2808 zero, just continue with the main loop. If the length is negative, it
2809 means the reference is unset in non-Java-compatible mode. If the minimum is
2810 zero, we can continue at the same level without recursion. For any other
2811 minimum, carrying on will result in NOMATCH. */
2812
2813 if (length == 0) continue;
2814 if (length < 0 && min == 0) continue;
2815
2816 /* First, ensure the minimum number of matches are present. We get back
2817 the length of the reference string explicitly rather than passing the
2818 address of eptr, so that eptr can be a register variable. */
2819
2820 for (i = 1; i <= min; i++)
2821 {
2822 int slength;
2823 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2824 {
2825 if (slength == -2) eptr = md->end_subject; /* Partial match */
2826 CHECK_PARTIAL();
2827 RRETURN(MATCH_NOMATCH);
2828 }
2829 eptr += slength;
2830 }
2831
2832 /* If min = max, continue at the same level without recursion.
2833 They are not both allowed to be zero. */
2834
2835 if (min == max) continue;
2836
2837 /* If minimizing, keep trying and advancing the pointer */
2838
2839 if (minimize)
2840 {
2841 for (fi = min;; fi++)
2842 {
2843 int slength;
2844 RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2845 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2846 if (fi >= max) RRETURN(MATCH_NOMATCH);
2847 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2848 {
2849 if (slength == -2) eptr = md->end_subject; /* Partial match */
2850 CHECK_PARTIAL();
2851 RRETURN(MATCH_NOMATCH);
2852 }
2853 eptr += slength;
2854 }
2855 /* Control never gets here */
2856 }
2857
2858 /* If maximizing, find the longest string and work backwards */
2859
2860 else
2861 {
2862 pp = eptr;
2863 for (i = min; i < max; i++)
2864 {
2865 int slength;
2866 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2867 {
2868 /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2869 the soft partial matching case. */
2870
2871 if (slength == -2 && md->partial != 0 &&
2872 md->end_subject > md->start_used_ptr)
2873 {
2874 md->hitend = TRUE;
2875 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2876 }
2877 break;
2878 }
2879 eptr += slength;
2880 }
2881
2882 while (eptr >= pp)
2883 {
2884 RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2885 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2886 eptr -= length;
2887 }
2888 RRETURN(MATCH_NOMATCH);
2889 }
2890 /* Control never gets here */
2891
2892 /* Match a bit-mapped character class, possibly repeatedly. This op code is
2893 used when all the characters in the class have values in the range 0-255,
2894 and either the matching is caseful, or the characters are in the range
2895 0-127 when UTF-8 processing is enabled. The only difference between
2896 OP_CLASS and OP_NCLASS occurs when a data character outside the range is
2897 encountered.
2898
2899 First, look past the end of the item to see if there is repeat information
2900 following. Then obey similar code to character type repeats - written out
2901 again for speed. */
2902
2903 case OP_NCLASS:
2904 case OP_CLASS:
2905 {
2906 /* The data variable is saved across frames, so the byte map needs to
2907 be stored there. */
2908 #define BYTE_MAP ((pcre_uint8 *)data)
2909 data = ecode + 1; /* Save for matching */
2910 ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2911
2912 switch (*ecode)
2913 {
2914 case OP_CRSTAR:
2915 case OP_CRMINSTAR:
2916 case OP_CRPLUS:
2917 case OP_CRMINPLUS:
2918 case OP_CRQUERY:
2919 case OP_CRMINQUERY:
2920 c = *ecode++ - OP_CRSTAR;
2921 minimize = (c & 1) != 0;
2922 min = rep_min[c]; /* Pick up values from tables; */
2923 max = rep_max[c]; /* zero for max => infinity */
2924 if (max == 0) max = INT_MAX;
2925 break;
2926
2927 case OP_CRRANGE:
2928 case OP_CRMINRANGE:
2929 minimize = (*ecode == OP_CRMINRANGE);
2930 min = GET2(ecode, 1);
2931 max = GET2(ecode, 1 + IMM2_SIZE);
2932 if (max == 0) max = INT_MAX;
2933 ecode += 1 + 2 * IMM2_SIZE;
2934 break;
2935
2936 default: /* No repeat follows */
2937 min = max = 1;
2938 break;
2939 }
2940
2941 /* First, ensure the minimum number of matches are present. */
2942
2943 #ifdef SUPPORT_UTF
2944 if (utf)
2945 {
2946 for (i = 1; i <= min; i++)
2947 {
2948 if (eptr >= md->end_subject)
2949 {
2950 SCHECK_PARTIAL();
2951 RRETURN(MATCH_NOMATCH);
2952 }
2953 GETCHARINC(c, eptr);
2954 if (c > 255)
2955 {
2956 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2957 }
2958 else
2959 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2960 }
2961 }
2962 else
2963 #endif
2964 /* Not UTF mode */
2965 {
2966 for (i = 1; i <= min; i++)
2967 {
2968 if (eptr >= md->end_subject)
2969 {
2970 SCHECK_PARTIAL();
2971 RRETURN(MATCH_NOMATCH);
2972 }
2973 c = *eptr++;
2974 #ifndef COMPILE_PCRE8
2975 if (c > 255)
2976 {
2977 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2978 }
2979 else
2980 #endif
2981 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2982 }
2983 }
2984
2985 /* If max == min we can continue with the main loop without the
2986 need to recurse. */
2987
2988 if (min == max) continue;
2989
2990 /* If minimizing, keep testing the rest of the expression and advancing
2991 the pointer while it matches the class. */
2992
2993 if (minimize)
2994 {
2995 #ifdef SUPPORT_UTF
2996 if (utf)
2997 {
2998 for (fi = min;; fi++)
2999 {
3000 RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
3001 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3002 if (fi >= max) RRETURN(MATCH_NOMATCH);
3003 if (eptr >= md->end_subject)
3004 {
3005 SCHECK_PARTIAL();
3006 RRETURN(MATCH_NOMATCH);
3007 }
3008 GETCHARINC(c, eptr);
3009 if (c > 255)
3010 {
3011 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3012 }
3013 else
3014 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3015 }
3016 }
3017 else
3018 #endif
3019 /* Not UTF mode */
3020 {
3021 for (fi = min;; fi++)
3022 {
3023 RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
3024 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3025 if (fi >= max) RRETURN(MATCH_NOMATCH);
3026 if (eptr >= md->end_subject)
3027 {
3028 SCHECK_PARTIAL();
3029 RRETURN(MATCH_NOMATCH);
3030 }
3031 c = *eptr++;
3032 #ifndef COMPILE_PCRE8
3033 if (c > 255)
3034 {
3035 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3036 }
3037 else
3038 #endif
3039 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3040 }
3041 }
3042 /* Control never gets here */
3043 }
3044
3045 /* If maximizing, find the longest possible run, then work backwards. */
3046
3047 else
3048 {
3049 pp = eptr;
3050
3051 #ifdef SUPPORT_UTF
3052 if (utf)
3053 {
3054 for (i = min; i < max; i++)
3055 {
3056 int len = 1;
3057 if (eptr >= md->end_subject)
3058 {
3059 SCHECK_PARTIAL();
3060 break;
3061 }
3062 GETCHARLEN(c, eptr, len);
3063 if (c > 255)
3064 {
3065 if (op == OP_CLASS) break;
3066 }
3067 else
3068 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3069 eptr += len;
3070 }
3071 for (;;)
3072 {
3073 RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3074 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3075 if (eptr-- == pp) break; /* Stop if tried at original pos */
3076 BACKCHAR(eptr);
3077 }
3078 }
3079 else
3080 #endif
3081 /* Not UTF mode */
3082 {
3083 for (i = min; i < max; i++)
3084 {
3085 if (eptr >= md->end_subject)
3086 {
3087 SCHECK_PARTIAL();
3088 break;
3089 }
3090 c = *eptr;
3091 #ifndef COMPILE_PCRE8
3092 if (c > 255)
3093 {
3094 if (op == OP_CLASS) break;
3095 }
3096 else
3097 #endif
3098 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3099 eptr++;
3100 }
3101 while (eptr >= pp)
3102 {
3103 RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3104 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3105 eptr--;
3106 }
3107 }
3108
3109 RRETURN(MATCH_NOMATCH);
3110 }
3111 #undef BYTE_MAP
3112 }
3113 /* Control never gets here */
3114
3115
3116 /* Match an extended character class. This opcode is encountered only
3117 when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
3118 mode, because Unicode properties are supported in non-UTF-8 mode. */
3119
3120 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3121 case OP_XCLASS:
3122 {
3123 data = ecode + 1 + LINK_SIZE; /* Save for matching */
3124 ecode += GET(ecode, 1); /* Advance past the item */
3125
3126 switch (*ecode)
3127 {
3128 case OP_CRSTAR:
3129 case OP_CRMINSTAR:
3130 case OP_CRPLUS:
3131 case OP_CRMINPLUS:
3132 case OP_CRQUERY:
3133 case OP_CRMINQUERY:
3134 c = *ecode++ - OP_CRSTAR;
3135 minimize = (c & 1) != 0;
3136 min = rep_min[c]; /* Pick up values from tables; */
3137 max = rep_max[c]; /* zero for max => infinity */
3138 if (max == 0) max = INT_MAX;
3139 break;
3140
3141 case OP_CRRANGE:
3142 case OP_CRMINRANGE:
3143 minimize = (*ecode == OP_CRMINRANGE);
3144 min = GET2(ecode, 1);
3145 max = GET2(ecode, 1 + IMM2_SIZE);
3146 if (max == 0) max = INT_MAX;
3147 ecode += 1 + 2 * IMM2_SIZE;
3148 break;
3149
3150 default: /* No repeat follows */
3151 min = max = 1;
3152 break;
3153 }
3154
3155 /* First, ensure the minimum number of matches are present. */
3156
3157 for (i = 1; i <= min; i++)
3158 {
3159 if (eptr >= md->end_subject)
3160 {
3161 SCHECK_PARTIAL();
3162 RRETURN(MATCH_NOMATCH);
3163 }
3164 GETCHARINCTEST(c, eptr);
3165 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3166 }
3167
3168 /* If max == min we can continue with the main loop without the
3169 need to recurse. */
3170
3171 if (min == max) continue;
3172
3173 /* If minimizing, keep testing the rest of the expression and advancing
3174 the pointer while it matches the class. */
3175
3176 if (minimize)
3177 {
3178 for (fi = min;; fi++)
3179 {
3180 RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3181 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3182 if (fi >= max) RRETURN(MATCH_NOMATCH);
3183 if (eptr >= md->end_subject)
3184 {
3185 SCHECK_PARTIAL();
3186 RRETURN(MATCH_NOMATCH);
3187 }
3188 GETCHARINCTEST(c, eptr);
3189 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3190 }
3191 /* Control never gets here */
3192 }
3193
3194 /* If maximizing, find the longest possible run, then work backwards. */
3195
3196 else
3197 {
3198 pp = eptr;
3199 for (i = min; i < max; i++)
3200 {
3201 int len = 1;
3202 if (eptr >= md->end_subject)
3203 {
3204 SCHECK_PARTIAL();
3205 break;
3206 }
3207 #ifdef SUPPORT_UTF
3208 GETCHARLENTEST(c, eptr, len);
3209 #else
3210 c = *eptr;
3211 #endif
3212 if (!PRIV(xclass)(c, data, utf)) break;
3213 eptr += len;
3214 }
3215 for(;;)
3216 {
3217 RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3218 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3219 if (eptr-- == pp) break; /* Stop if tried at original pos */
3220 #ifdef SUPPORT_UTF
3221 if (utf) BACKCHAR(eptr);
3222 #endif
3223 }
3224 RRETURN(MATCH_NOMATCH);
3225 }
3226
3227 /* Control never gets here */
3228 }
3229 #endif /* End of XCLASS */
3230
3231 /* Match a single character, casefully */
3232
3233 case OP_CHAR:
3234 #ifdef SUPPORT_UTF
3235 if (utf)
3236 {
3237 length = 1;
3238 ecode++;
3239 GETCHARLEN(fc, ecode, length);
3240 if (length > md->end_subject - eptr)
3241 {
3242 CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */
3243 RRETURN(MATCH_NOMATCH);
3244 }
3245 while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3246 }
3247 else
3248 #endif
3249 /* Not UTF mode */
3250 {
3251 if (md->end_subject - eptr < 1)
3252 {
3253 SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */
3254 RRETURN(MATCH_NOMATCH);
3255 }
3256 if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3257 ecode += 2;
3258 }
3259 break;
3260
3261 /* Match a single character, caselessly. If we are at the end of the
3262 subject, give up immediately. */
3263
3264 case OP_CHARI:
3265 if (eptr >= md->end_subject)
3266 {
3267 SCHECK_PARTIAL();
3268 RRETURN(MATCH_NOMATCH);
3269 }
3270
3271 #ifdef SUPPORT_UTF
3272 if (utf)
3273 {
3274 length = 1;
3275 ecode++;
3276 GETCHARLEN(fc, ecode, length);
3277
3278 /* If the pattern character's value is < 128, we have only one byte, and
3279 we know that its other case must also be one byte long, so we can use the
3280 fast lookup table. We know that there is at least one byte left in the
3281 subject. */
3282
3283 if (fc < 128)
3284 {
3285 pcre_uint32 cc = RAWUCHAR(eptr);
3286 if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3287 ecode++;
3288 eptr++;
3289 }
3290
3291 /* Otherwise we must pick up the subject character. Note that we cannot
3292 use the value of "length" to check for sufficient bytes left, because the
3293 other case of the character may have more or fewer bytes. */
3294
3295 else
3296 {
3297 pcre_uint32 dc;
3298 GETCHARINC(dc, eptr);
3299 ecode += length;
3300
3301 /* If we have Unicode property support, we can use it to test the other
3302 case of the character, if there is one. */
3303
3304 if (fc != dc)
3305 {
3306 #ifdef SUPPORT_UCP
3307 if (dc != UCD_OTHERCASE(fc))
3308 #endif
3309 RRETURN(MATCH_NOMATCH);
3310 }
3311 }
3312 }
3313 else
3314 #endif /* SUPPORT_UTF */
3315
3316 /* Not UTF mode */
3317 {
3318 if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3319 != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3320 eptr++;
3321 ecode += 2;
3322 }
3323 break;
3324
3325 /* Match a single character repeatedly. */
3326
3327 case OP_EXACT:
3328 case OP_EXACTI:
3329 min = max = GET2(ecode, 1);
3330 ecode += 1 + IMM2_SIZE;
3331 goto REPEATCHAR;
3332
3333 case OP_POSUPTO:
3334 case OP_POSUPTOI:
3335 possessive = TRUE;
3336 /* Fall through */
3337
3338 case OP_UPTO:
3339 case OP_UPTOI:
3340 case OP_MINUPTO:
3341 case OP_MINUPTOI:
3342 min = 0;
3343 max = GET2(ecode, 1);
3344 minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3345 ecode += 1 + IMM2_SIZE;
3346 goto REPEATCHAR;
3347
3348 case OP_POSSTAR:
3349 case OP_POSSTARI:
3350 possessive = TRUE;
3351 min = 0;
3352 max = INT_MAX;
3353 ecode++;
3354 goto REPEATCHAR;
3355
3356 case OP_POSPLUS:
3357 case OP_POSPLUSI:
3358 possessive = TRUE;
3359 min = 1;
3360 max = INT_MAX;
3361 ecode++;
3362 goto REPEATCHAR;
3363
3364 case OP_POSQUERY:
3365 case OP_POSQUERYI:
3366 possessive = TRUE;
3367 min = 0;
3368 max = 1;
3369 ecode++;
3370 goto REPEATCHAR;
3371
3372 case OP_STAR:
3373 case OP_STARI:
3374 case OP_MINSTAR:
3375 case OP_MINSTARI:
3376 case OP_PLUS:
3377 case OP_PLUSI:
3378 case OP_MINPLUS:
3379 case OP_MINPLUSI:
3380 case OP_QUERY:
3381 case OP_QUERYI:
3382 case OP_MINQUERY:
3383 case OP_MINQUERYI:
3384 c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3385 minimize = (c & 1) != 0;
3386 min = rep_min[c]; /* Pick up values from tables; */
3387 max = rep_max[c]; /* zero for max => infinity */
3388 if (max == 0) max = INT_MAX;
3389
3390 /* Common code for all repeated single-character matches. We first check
3391 for the minimum number of characters. If the minimum equals the maximum, we
3392 are done. Otherwise, if minimizing, check the rest of the pattern for a
3393 match; if there isn't one, advance up to the maximum, one character at a
3394 time.
3395
3396 If maximizing, advance up to the maximum number of matching characters,
3397 until eptr is past the end of the maximum run. If possessive, we are
3398 then done (no backing up). Otherwise, match at this position; anything
3399 other than no match is immediately returned. For nomatch, back up one
3400 character, unless we are matching \R and the last thing matched was
3401 \r\n, in which case, back up two bytes. When we reach the first optional
3402 character position, we can save stack by doing a tail recurse.
3403
3404 The various UTF/non-UTF and caseful/caseless cases are handled separately,
3405 for speed. */
3406
3407 REPEATCHAR:
3408 #ifdef SUPPORT_UTF
3409 if (utf)
3410 {
3411 length = 1;
3412 charptr = ecode;
3413 GETCHARLEN(fc, ecode, length);
3414 ecode += length;
3415
3416 /* Handle multibyte character matching specially here. There is
3417 support for caseless matching if UCP support is present. */
3418
3419 if (length > 1)
3420 {
3421 #ifdef SUPPORT_UCP
3422 pcre_uint32 othercase;
3423 if (op >= OP_STARI && /* Caseless */
3424 (othercase = UCD_OTHERCASE(fc)) != fc)
3425 oclength = PRIV(ord2utf)(othercase, occhars);
3426 else oclength = 0;
3427 #endif /* SUPPORT_UCP */
3428
3429 for (i = 1; i <= min; i++)
3430 {
3431 if (eptr <= md->end_subject - length &&
3432 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3433 #ifdef SUPPORT_UCP
3434 else if (oclength > 0 &&
3435 eptr <= md->end_subject - oclength &&
3436 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3437 #endif /* SUPPORT_UCP */
3438 else
3439 {
3440 CHECK_PARTIAL();
3441 RRETURN(MATCH_NOMATCH);
3442 }
3443 }
3444
3445 if (min == max) continue;
3446
3447 if (minimize)
3448 {
3449 for (fi = min;; fi++)
3450 {
3451 RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3452 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3453 if (fi >= max) RRETURN(MATCH_NOMATCH);
3454 if (eptr <= md->end_subject - length &&
3455 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3456 #ifdef SUPPORT_UCP
3457 else if (oclength > 0 &&
3458 eptr <= md->end_subject - oclength &&
3459 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3460 #endif /* SUPPORT_UCP */
3461 else
3462 {
3463 CHECK_PARTIAL();
3464 RRETURN(MATCH_NOMATCH);
3465 }
3466 }
3467 /* Control never gets here */
3468 }
3469
3470 else /* Maximize */
3471 {
3472 pp = eptr;
3473 for (i = min; i < max; i++)
3474 {
3475 if (eptr <= md->end_subject - length &&
3476 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3477 #ifdef SUPPORT_UCP
3478 else if (oclength > 0 &&
3479 eptr <= md->end_subject - oclength &&
3480 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3481 #endif /* SUPPORT_UCP */
3482 else
3483 {
3484 CHECK_PARTIAL();
3485 break;
3486 }
3487 }
3488
3489 if (possessive) continue; /* No backtracking */
3490 for(;;)
3491 {
3492 if (eptr == pp) goto TAIL_RECURSE;
3493 RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3494 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3495 /* if (eptr == pp) { RRETURN(MATCH_NOMATCH); } */
3496 #ifdef SUPPORT_UCP
3497 eptr--;
3498 BACKCHAR(eptr);
3499 #else /* without SUPPORT_UCP */
3500 eptr -= length;
3501 #endif /* SUPPORT_UCP */
3502 }
3503 }
3504 /* Control never gets here */
3505 }
3506
3507 /* If the length of a UTF-8 character is 1, we fall through here, and
3508 obey the code as for non-UTF-8 characters below, though in this case the
3509 value of fc will always be < 128. */
3510 }
3511 else
3512 #endif /* SUPPORT_UTF */
3513 /* When not in UTF-8 mode, load a single-byte character. */
3514 fc = *ecode++;
3515
3516 /* The value of fc at this point is always one character, though we may
3517 or may not be in UTF mode. The code is duplicated for the caseless and
3518 caseful cases, for speed, since matching characters is likely to be quite
3519 common. First, ensure the minimum number of matches are present. If min =
3520 max, continue at the same level without recursing. Otherwise, if
3521 minimizing, keep trying the rest of the expression and advancing one
3522 matching character if failing, up to the maximum. Alternatively, if
3523 maximizing, find the maximum number of characters and work backwards. */
3524
3525 DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3526 max, (char *)eptr));
3527
3528 if (op >= OP_STARI) /* Caseless */
3529 {
3530 #ifdef COMPILE_PCRE8
3531 /* fc must be < 128 if UTF is enabled. */
3532 foc = md->fcc[fc];
3533 #else
3534 #ifdef SUPPORT_UTF
3535 #ifdef SUPPORT_UCP
3536 if (utf && fc > 127)
3537 foc = UCD_OTHERCASE(fc);
3538 #else
3539 if (utf && fc > 127)
3540 foc = fc;
3541 #endif /* SUPPORT_UCP */
3542 else
3543 #endif /* SUPPORT_UTF */
3544 foc = TABLE_GET(fc, md->fcc, fc);
3545 #endif /* COMPILE_PCRE8 */
3546
3547 for (i = 1; i <= min; i++)
3548 {
3549 pcre_uint32 cc; /* Faster than pcre_uchar */
3550 if (eptr >= md->end_subject)
3551 {
3552 SCHECK_PARTIAL();
3553 RRETURN(MATCH_NOMATCH);
3554 }
3555 cc = RAWUCHARTEST(eptr);
3556 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3557 eptr++;
3558 }
3559 if (min == max) continue;
3560 if (minimize)
3561 {
3562 for (fi = min;; fi++)
3563 {
3564 pcre_uint32 cc; /* Faster than pcre_uchar */
3565 RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3566 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3567 if (fi >= max) RRETURN(MATCH_NOMATCH);
3568 if (eptr >= md->end_subject)
3569 {
3570 SCHECK_PARTIAL();
3571 RRETURN(MATCH_NOMATCH);
3572 }
3573 cc = RAWUCHARTEST(eptr);
3574 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3575 eptr++;
3576 }
3577 /* Control never gets here */
3578 }
3579 else /* Maximize */
3580 {
3581 pp = eptr;
3582 for (i = min; i < max; i++)
3583 {
3584 pcre_uint32 cc; /* Faster than pcre_uchar */
3585 if (eptr >= md->end_subject)
3586 {
3587 SCHECK_PARTIAL();
3588 break;
3589 }
3590 cc = RAWUCHARTEST(eptr);
3591 if (fc != cc && foc != cc) break;
3592 eptr++;
3593 }
3594
3595 if (possessive) continue; /* No backtracking */
3596 for (;;)
3597 {
3598 if (eptr == pp) goto TAIL_RECURSE;
3599 RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3600 eptr--;
3601 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3602 }
3603 RRETURN(MATCH_NOMATCH);
3604 }
3605 /* Control never gets here */
3606 }
3607
3608 /* Caseful comparisons (includes all multi-byte characters) */
3609
3610 else
3611 {
3612 for (i = 1; i <= min; i++)
3613 {
3614 if (eptr >= md->end_subject)
3615 {
3616 SCHECK_PARTIAL();
3617 RRETURN(MATCH_NOMATCH);
3618 }
3619 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3620 }
3621
3622 if (min == max) continue;
3623
3624 if (minimize)
3625 {
3626 for (fi = min;; fi++)
3627 {
3628 RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3629 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3630 if (fi >= max) RRETURN(MATCH_NOMATCH);
3631 if (eptr >= md->end_subject)
3632 {
3633 SCHECK_PARTIAL();
3634 RRETURN(MATCH_NOMATCH);
3635 }
3636 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3637 }
3638 /* Control never gets here */
3639 }
3640 else /* Maximize */
3641 {
3642 pp = eptr;
3643 for (i = min; i < max; i++)
3644 {
3645 if (eptr >= md->end_subject)
3646 {
3647 SCHECK_PARTIAL();
3648 break;
3649 }
3650 if (fc != RAWUCHARTEST(eptr)) break;
3651 eptr++;
3652 }
3653 if (possessive) continue; /* No backtracking */
3654 for (;;)
3655 {
3656 if (eptr == pp) goto TAIL_RECURSE;
3657 RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3658 eptr--;
3659 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3660 }
3661 RRETURN(MATCH_NOMATCH);
3662 }
3663 }
3664 /* Control never gets here */
3665
3666 /* Match a negated single one-byte character. The character we are
3667 checking can be multibyte. */
3668
3669 case OP_NOT:
3670 case OP_NOTI:
3671 if (eptr >= md->end_subject)
3672 {
3673 SCHECK_PARTIAL();
3674 RRETURN(MATCH_NOMATCH);
3675 }
3676 #ifdef SUPPORT_UTF
3677 if (utf)
3678 {
3679 register pcre_uint32 ch, och;
3680
3681 ecode++;
3682 GETCHARINC(ch, ecode);
3683 GETCHARINC(c, eptr);
3684
3685 if (op == OP_NOT)
3686 {
3687 if (ch == c) RRETURN(MATCH_NOMATCH);
3688 }
3689 else
3690 {
3691 #ifdef SUPPORT_UCP
3692 if (ch > 127)
3693 och = UCD_OTHERCASE(ch);
3694 #else
3695 if (ch > 127)
3696 och = ch;
3697 #endif /* SUPPORT_UCP */
3698 else
3699 och = TABLE_GET(ch, md->fcc, ch);
3700 if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3701 }
3702 }
3703 else
3704 #endif
3705 {
3706 register pcre_uint32 ch = ecode[1];
3707 c = *eptr++;
3708 if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3709 RRETURN(MATCH_NOMATCH);
3710 ecode += 2;
3711 }
3712 break;
3713
3714 /* Match a negated single one-byte character repeatedly. This is almost a
3715 repeat of the code for a repeated single character, but I haven't found a
3716 nice way of commoning these up that doesn't require a test of the
3717 positive/negative option for each character match. Maybe that wouldn't add
3718 very much to the time taken, but character matching *is* what this is all
3719 about... */
3720
3721 case OP_NOTEXACT:
3722 case OP_NOTEXACTI:
3723 min = max = GET2(ecode, 1);
3724 ecode += 1 + IMM2_SIZE;
3725 goto REPEATNOTCHAR;
3726
3727 case OP_NOTUPTO:
3728 case OP_NOTUPTOI:
3729 case OP_NOTMINUPTO:
3730 case OP_NOTMINUPTOI:
3731 min = 0;
3732 max = GET2(ecode, 1);
3733 minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3734 ecode += 1 + IMM2_SIZE;
3735 goto REPEATNOTCHAR;
3736
3737 case OP_NOTPOSSTAR:
3738 case OP_NOTPOSSTARI:
3739 possessive = TRUE;
3740 min = 0;
3741 max = INT_MAX;
3742 ecode++;
3743 goto REPEATNOTCHAR;
3744
3745 case OP_NOTPOSPLUS:
3746 case OP_NOTPOSPLUSI:
3747 possessive = TRUE;
3748 min = 1;
3749 max = INT_MAX;
3750 ecode++;
3751 goto REPEATNOTCHAR;
3752
3753 case OP_NOTPOSQUERY:
3754 case OP_NOTPOSQUERYI:
3755 possessive = TRUE;
3756 min = 0;
3757 max = 1;
3758 ecode++;
3759 goto REPEATNOTCHAR;
3760
3761 case OP_NOTPOSUPTO:
3762 case OP_NOTPOSUPTOI:
3763 possessive = TRUE;
3764 min = 0;
3765 max = GET2(ecode, 1);
3766 ecode += 1 + IMM2_SIZE;
3767 goto REPEATNOTCHAR;
3768
3769 case OP_NOTSTAR:
3770 case OP_NOTSTARI:
3771 case OP_NOTMINSTAR:
3772 case OP_NOTMINSTARI:
3773 case OP_NOTPLUS:
3774 case OP_NOTPLUSI:
3775 case OP_NOTMINPLUS:
3776 case OP_NOTMINPLUSI:
3777 case OP_NOTQUERY:
3778 case OP_NOTQUERYI:
3779 case OP_NOTMINQUERY:
3780 case OP_NOTMINQUERYI:
3781 c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3782 minimize = (c & 1) != 0;
3783 min = rep_min[c]; /* Pick up values from tables; */
3784 max = rep_max[c]; /* zero for max => infinity */
3785 if (max == 0) max = INT_MAX;
3786
3787 /* Common code for all repeated single-byte matches. */
3788
3789 REPEATNOTCHAR:
3790 GETCHARINCTEST(fc, ecode);
3791
3792 /* The code is duplicated for the caseless and caseful cases, for speed,
3793 since matching characters is likely to be quite common. First, ensure the
3794 minimum number of matches are present. If min = max, continue at the same
3795 level without recursing. Otherwise, if minimizing, keep trying the rest of
3796 the expression and advancing one matching character if failing, up to the
3797 maximum. Alternatively, if maximizing, find the maximum number of
3798 characters and work backwards. */
3799
3800 DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3801 max, (char *)eptr));
3802
3803 if (op >= OP_NOTSTARI) /* Caseless */
3804 {
3805 #ifdef SUPPORT_UTF
3806 #ifdef SUPPORT_UCP
3807 if (utf && fc > 127)
3808 foc = UCD_OTHERCASE(fc);
3809 #else
3810 if (utf && fc > 127)
3811 foc = fc;
3812 #endif /* SUPPORT_UCP */
3813 else
3814 #endif /* SUPPORT_UTF */
3815 foc = TABLE_GET(fc, md->fcc, fc);
3816
3817 #ifdef SUPPORT_UTF
3818 if (utf)
3819 {
3820 register pcre_uint32 d;
3821 for (i = 1; i <= min; i++)
3822 {
3823 if (eptr >= md->end_subject)
3824 {
3825 SCHECK_PARTIAL();
3826 RRETURN(MATCH_NOMATCH);
3827 }
3828 GETCHARINC(d, eptr);
3829 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3830 }
3831 }
3832 else
3833 #endif /* SUPPORT_UTF */
3834 /* Not UTF mode */
3835 {
3836 for (i = 1; i <= min; i++)
3837 {
3838 if (eptr >= md->end_subject)
3839 {
3840 SCHECK_PARTIAL();
3841 RRETURN(MATCH_NOMATCH);
3842 }
3843 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3844 eptr++;
3845 }
3846 }
3847
3848 if (min == max) continue;
3849
3850 if (minimize)
3851 {
3852 #ifdef SUPPORT_UTF
3853 if (utf)
3854 {
3855 register pcre_uint32 d;
3856 for (fi = min;; fi++)
3857 {
3858 RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3859 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3860 if (fi >= max) RRETURN(MATCH_NOMATCH);
3861 if (eptr >= md->end_subject)
3862 {
3863 SCHECK_PARTIAL();
3864 RRETURN(MATCH_NOMATCH);
3865 }
3866 GETCHARINC(d, eptr);
3867 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3868 }
3869 }
3870 else
3871 #endif /*SUPPORT_UTF */
3872 /* Not UTF mode */
3873 {
3874 for (fi = min;; fi++)
3875 {
3876 RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3877 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3878 if (fi >= max) RRETURN(MATCH_NOMATCH);
3879 if (eptr >= md->end_subject)
3880 {
3881 SCHECK_PARTIAL();
3882 RRETURN(MATCH_NOMATCH);
3883 }
3884 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3885 eptr++;
3886 }
3887 }
3888 /* Control never gets here */
3889 }
3890
3891 /* Maximize case */
3892
3893 else
3894 {
3895 pp = eptr;
3896
3897 #ifdef SUPPORT_UTF
3898 if (utf)
3899 {
3900 register pcre_uint32 d;
3901 for (i = min; i < max; i++)
3902 {
3903 int len = 1;
3904 if (eptr >= md->end_subject)
3905 {
3906 SCHECK_PARTIAL();
3907 break;
3908 }
3909 GETCHARLEN(d, eptr, len);
3910 if (fc == d || (unsigned int)foc == d) break;
3911 eptr += len;
3912 }
3913 if (possessive) continue; /* No backtracking */
3914 for(;;)
3915 {
3916 if (eptr == pp) goto TAIL_RECURSE;
3917 RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3918 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3919 eptr--;
3920 BACKCHAR(eptr);
3921 }
3922 }
3923 else
3924 #endif /* SUPPORT_UTF */
3925 /* Not UTF mode */
3926 {
3927 for (i = min; i < max; i++)
3928 {
3929 if (eptr >= md->end_subject)
3930 {
3931 SCHECK_PARTIAL();
3932 break;
3933 }
3934 if (fc == *eptr || foc == *eptr) break;
3935 eptr++;
3936 }
3937 if (possessive) continue; /* No backtracking */
3938 for (;;)
3939 {
3940 if (eptr == pp) goto TAIL_RECURSE;
3941 RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3942 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3943 eptr--;
3944 }
3945 }
3946
3947 RRETURN(MATCH_NOMATCH);
3948 }
3949 /* Control never gets here */
3950 }
3951
3952 /* Caseful comparisons */
3953
3954 else
3955 {
3956 #ifdef SUPPORT_UTF
3957 if (utf)
3958 {
3959 register pcre_uint32 d;
3960 for (i = 1; i <= min; i++)
3961 {
3962 if (eptr >= md->end_subject)
3963 {
3964 SCHECK_PARTIAL();
3965 RRETURN(MATCH_NOMATCH);
3966 }
3967 GETCHARINC(d, eptr);
3968 if (fc == d) RRETURN(MATCH_NOMATCH);
3969 }
3970 }
3971 else
3972 #endif
3973 /* Not UTF mode */
3974 {
3975 for (i = 1; i <= min; i++)
3976 {
3977 if (eptr >= md->end_subject)
3978 {
3979 SCHECK_PARTIAL();
3980 RRETURN(MATCH_NOMATCH);
3981 }
3982 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3983 }
3984 }
3985
3986 if (min == max) continue;
3987
3988 if (minimize)
3989 {
3990 #ifdef SUPPORT_UTF
3991 if (utf)
3992 {
3993 register pcre_uint32 d;
3994 for (fi = min;; fi++)
3995 {
3996 RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3997 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3998 if (fi >= max) RRETURN(MATCH_NOMATCH);
3999 if (eptr >= md->end_subject)
4000 {
4001 SCHECK_PARTIAL();
4002 RRETURN(MATCH_NOMATCH);
4003 }
4004 GETCHARINC(d, eptr);
4005 if (fc == d) RRETURN(MATCH_NOMATCH);
4006 }
4007 }
4008 else
4009 #endif
4010 /* Not UTF mode */
4011 {
4012 for (fi = min;; fi++)
4013 {
4014 RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
4015 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4016 if (fi >= max) RRETURN(MATCH_NOMATCH);
4017 if (eptr >= md->end_subject)
4018 {
4019 SCHECK_PARTIAL();
4020 RRETURN(MATCH_NOMATCH);
4021 }
4022 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
4023 }
4024 }
4025 /* Control never gets here */
4026 }
4027
4028 /* Maximize case */
4029
4030 else
4031 {
4032 pp = eptr;
4033
4034 #ifdef SUPPORT_UTF
4035 if (utf)
4036 {
4037 register pcre_uint32 d;
4038 for (i = min; i < max; i++)
4039 {
4040 int len = 1;
4041 if (eptr >= md->end_subject)
4042 {
4043 SCHECK_PARTIAL();
4044 break;
4045 }
4046 GETCHARLEN(d, eptr, len);
4047 if (fc == d) break;
4048 eptr += len;
4049 }
4050 if (possessive) continue; /* No backtracking */
4051 for(;;)
4052 {
4053 if (eptr == pp) goto TAIL_RECURSE;
4054 RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
4055 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4056 eptr--;
4057 BACKCHAR(eptr);
4058 }
4059 }
4060 else
4061 #endif
4062 /* Not UTF mode */
4063 {
4064 for (i = min; i < max; i++)
4065 {
4066 if (eptr >= md->end_subject)
4067 {
4068 SCHECK_PARTIAL();
4069 break;
4070 }
4071 if (fc == *eptr) break;
4072 eptr++;
4073 }
4074 if (possessive) continue; /* No backtracking */
4075 for (;;)
4076 {
4077 if (eptr == pp) goto TAIL_RECURSE;
4078 RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
4079 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4080 eptr--;
4081 }
4082 }
4083
4084 RRETURN(MATCH_NOMATCH);
4085 }
4086 }
4087 /* Control never gets here */
4088
4089 /* Match a single character type repeatedly; several different opcodes
4090 share code. This is very similar to the code for single characters, but we
4091 repeat it in the interests of efficiency. */
4092
4093 case OP_TYPEEXACT:
4094 min = max = GET2(ecode, 1);
4095 minimize = TRUE;
4096 ecode += 1 + IMM2_SIZE;
4097 goto REPEATTYPE;
4098
4099 case OP_TYPEUPTO:
4100 case OP_TYPEMINUPTO:
4101 min = 0;
4102 max = GET2(ecode, 1);
4103 minimize = *ecode == OP_TYPEMINUPTO;
4104 ecode += 1 + IMM2_SIZE;
4105 goto REPEATTYPE;
4106
4107 case OP_TYPEPOSSTAR:
4108 possessive = TRUE;
4109 min = 0;
4110 max = INT_MAX;
4111 ecode++;
4112 goto REPEATTYPE;
4113
4114 case OP_TYPEPOSPLUS:
4115 possessive = TRUE;
4116 min = 1;
4117 max = INT_MAX;
4118 ecode++;
4119 goto REPEATTYPE;
4120
4121 case OP_TYPEPOSQUERY:
4122 possessive = TRUE;
4123 min = 0;
4124 max = 1;
4125 ecode++;
4126 goto REPEATTYPE;
4127
4128 case OP_TYPEPOSUPTO:
4129 possessive = TRUE;
4130 min = 0;
4131 max = GET2(ecode, 1);
4132 ecode += 1 + IMM2_SIZE;
4133 goto REPEATTYPE;
4134
4135 case OP_TYPESTAR:
4136 case OP_TYPEMINSTAR:
4137 case OP_TYPEPLUS:
4138 case OP_TYPEMINPLUS:
4139 case OP_TYPEQUERY:
4140 case OP_TYPEMINQUERY:
4141 c = *ecode++ - OP_TYPESTAR;
4142 minimize = (c & 1) != 0;
4143 min = rep_min[c]; /* Pick up values from tables; */
4144 max = rep_max[c]; /* zero for max => infinity */
4145 if (max == 0) max = INT_MAX;
4146
4147 /* Common code for all repeated single character type matches. Note that
4148 in UTF-8 mode, '.' matches a character of any length, but for the other
4149 character types, the valid characters are all one-byte long. */
4150
4151 REPEATTYPE:
4152 ctype = *ecode++; /* Code for the character type */
4153
4154 #ifdef SUPPORT_UCP
4155 if (ctype == OP_PROP || ctype == OP_NOTPROP)
4156 {
4157 prop_fail_result = ctype == OP_NOTPROP;
4158 prop_type = *ecode++;
4159 prop_value = *ecode++;
4160 }
4161 else prop_type = -1;
4162 #endif
4163
4164 /* First, ensure the minimum number of matches are present. Use inline
4165 code for maximizing the speed, and do the type test once at the start
4166 (i.e. keep it out of the loop). Separate the UTF-8 code completely as that
4167 is tidier. Also separate the UCP code, which can be the same for both UTF-8
4168 and single-bytes. */
4169
4170 if (min > 0)
4171 {
4172 #ifdef SUPPORT_UCP
4173 if (prop_type >= 0)
4174 {
4175 switch(prop_type)
4176 {
4177 case PT_ANY:
4178 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4179 for (i = 1; i <= min; i++)
4180 {
4181 if (eptr >= md->end_subject)
4182 {
4183 SCHECK_PARTIAL();
4184 RRETURN(MATCH_NOMATCH);
4185 }
4186 GETCHARINCTEST(c, eptr);
4187 }
4188 break;
4189
4190 case PT_LAMP:
4191 for (i = 1; i <= min; i++)
4192 {
4193 int chartype;
4194 if (eptr >= md->end_subject)
4195 {
4196 SCHECK_PARTIAL();
4197 RRETURN(MATCH_NOMATCH);
4198 }
4199 GETCHARINCTEST(c, eptr);
4200 chartype = UCD_CHARTYPE(c);
4201 if ((chartype == ucp_Lu ||
4202 chartype == ucp_Ll ||
4203 chartype == ucp_Lt) == prop_fail_result)
4204 RRETURN(MATCH_NOMATCH);
4205 }
4206 break;
4207
4208 case PT_GC:
4209 for (i = 1; i <= min; i++)
4210 {
4211 if (eptr >= md->end_subject)
4212 {
4213 SCHECK_PARTIAL();
4214 RRETURN(MATCH_NOMATCH);
4215 }
4216 GETCHARINCTEST(c, eptr);
4217 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4218 RRETURN(MATCH_NOMATCH);
4219 }
4220 break;
4221
4222 case PT_PC:
4223 for (i = 1; i <= min; i++)
4224 {
4225 if (eptr >= md->end_subject)
4226 {
4227 SCHECK_PARTIAL();
4228 RRETURN(MATCH_NOMATCH);
4229 }
4230 GETCHARINCTEST(c, eptr);
4231 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4232 RRETURN(MATCH_NOMATCH);
4233 }
4234 break;
4235
4236 case PT_SC:
4237 for (i = 1; i <= min; i++)
4238 {
4239 if (eptr >= md->end_subject)
4240 {
4241 SCHECK_PARTIAL();
4242 RRETURN(MATCH_NOMATCH);
4243 }
4244 GETCHARINCTEST(c, eptr);
4245 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4246 RRETURN(MATCH_NOMATCH);
4247 }
4248 break;
4249
4250 case PT_ALNUM:
4251 for (i = 1; i <= min; i++)
4252 {
4253 int category;
4254 if (eptr >= md->end_subject)
4255 {
4256 SCHECK_PARTIAL();
4257 RRETURN(MATCH_NOMATCH);
4258 }
4259 GETCHARINCTEST(c, eptr);
4260 category = UCD_CATEGORY(c);
4261 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4262 RRETURN(MATCH_NOMATCH);
4263 }
4264 break;
4265
4266 case PT_SPACE: /* Perl space */
4267 for (i = 1; i <= min; i++)
4268 {
4269 if (eptr >= md->end_subject)
4270 {
4271 SCHECK_PARTIAL();
4272 RRETURN(MATCH_NOMATCH);
4273 }
4274 GETCHARINCTEST(c, eptr);
4275 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
4276 c == CHAR_FF || c == CHAR_CR)
4277 == prop_fail_result)
4278 RRETURN(MATCH_NOMATCH);
4279 }
4280 break;
4281
4282 case PT_PXSPACE: /* POSIX space */
4283 for (i = 1; i <= min; i++)
4284 {
4285 if (eptr >= md->end_subject)
4286 {
4287 SCHECK_PARTIAL();
4288 RRETURN(MATCH_NOMATCH);
4289 }
4290 GETCHARINCTEST(c, eptr);
4291 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
4292 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4293 == prop_fail_result)
4294 RRETURN(MATCH_NOMATCH);
4295 }
4296 break;
4297
4298 case PT_WORD:
4299 for (i = 1; i <= min; i++)
4300 {
4301 int category;
4302 if (eptr >= md->end_subject)
4303 {
4304 SCHECK_PARTIAL();
4305 RRETURN(MATCH_NOMATCH);
4306 }
4307 GETCHARINCTEST(c, eptr);
4308 category = UCD_CATEGORY(c);
4309 if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
4310 == prop_fail_result)
4311 RRETURN(MATCH_NOMATCH);
4312 }
4313 break;
4314
4315 case PT_CLIST:
4316 for (i = 1; i <= min; i++)
4317 {
4318 const pcre_uint32 *cp;
4319 if (eptr >= md->end_subject)
4320 {
4321 SCHECK_PARTIAL();
4322 RRETURN(MATCH_NOMATCH);
4323 }
4324 GETCHARINCTEST(c, eptr);
4325 cp = PRIV(ucd_caseless_sets) + prop_value;
4326 for (;;)
4327 {
4328 if (c < *cp)
4329 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
4330 if (c == *cp++)
4331 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
4332 }
4333 }
4334 break;
4335
4336 case PT_UCNC:
4337 for (i = 1; i <= min; i++)
4338 {
4339 if (eptr >= md->end_subject)
4340 {
4341 SCHECK_PARTIAL();
4342 RRETURN(MATCH_NOMATCH);
4343 }
4344 GETCHARINCTEST(c, eptr);
4345 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
4346 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
4347 c >= 0xe000) == prop_fail_result)
4348 RRETURN(MATCH_NOMATCH);
4349 }
4350 break;
4351
4352 /* This should not occur */
4353
4354 default:
4355 RRETURN(PCRE_ERROR_INTERNAL);
4356 }
4357 }
4358
4359 /* Match extended Unicode sequences. We will get here only if the
4360 support is in the binary; otherwise a compile-time error occurs. */
4361
4362 else if (ctype == OP_EXTUNI)
4363 {
4364 for (i = 1; i <= min; i++)
4365 {
4366 if (eptr >= md->end_subject)
4367 {
4368 SCHECK_PARTIAL();
4369 RRETURN(MATCH_NOMATCH);
4370 }
4371 else
4372 {
4373 int lgb, rgb;
4374 GETCHARINCTEST(c, eptr);
4375 lgb = UCD_GRAPHBREAK(c);
4376 while (eptr < md->end_subject)
4377 {
4378 int len = 1;
4379 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4380 rgb = UCD_GRAPHBREAK(c);
4381 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
4382 lgb = rgb;
4383 eptr += len;
4384 }
4385 }
4386 CHECK_PARTIAL();
4387 }
4388 }
4389
4390 else
4391 #endif /* SUPPORT_UCP */
4392
4393 /* Handle all other cases when the coding is UTF-8 */
4394
4395 #ifdef SUPPORT_UTF
4396 if (utf) switch(ctype)
4397 {
4398 case OP_ANY:
4399 for (i = 1; i <= min; i++)
4400 {
4401 if (eptr >= md->end_subject)
4402 {
4403 SCHECK_PARTIAL();
4404 RRETURN(MATCH_NOMATCH);
4405 }
4406 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4407 if (md->partial != 0 &&
4408 eptr + 1 >= md->end_subject &&
4409 NLBLOCK->nltype == NLTYPE_FIXED &&
4410 NLBLOCK->nllen == 2 &&
4411 RAWUCHAR(eptr) == NLBLOCK->nl[0])
4412 {
4413 md->hitend = TRUE;
4414 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4415 }
4416 eptr++;
4417 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4418 }
4419 break;
4420
4421 case OP_ALLANY:
4422 for (i = 1; i <= min; i++)
4423 {
4424 if (eptr >= md->end_subject)
4425 {
4426 SCHECK_PARTIAL();
4427 RRETURN(MATCH_NOMATCH);
4428 }
4429 eptr++;
4430 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4431 }
4432 break;
4433
4434 case OP_ANYBYTE:
4435 if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH);
4436 eptr += min;
4437 break;
4438
4439 case OP_ANYNL:
4440 for (i = 1; i <= min; i++)
4441 {
4442 if (eptr >= md->end_subject)
4443 {
4444 SCHECK_PARTIAL();
4445 RRETURN(MATCH_NOMATCH);
4446 }
4447 GETCHARINC(c, eptr);
4448 switch(c)
4449 {
4450 default: RRETURN(MATCH_NOMATCH);
4451
4452 case CHAR_CR:
4453 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
4454 break;
4455
4456 case CHAR_LF:
4457 break;
4458
4459 case CHAR_VT:
4460 case CHAR_FF:
4461 case CHAR_NEL:
4462 #ifndef EBCDIC
4463 case 0x2028:
4464 case 0x2029:
4465 #endif /* Not EBCDIC */
4466 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4467 break;
4468 }
4469 }
4470 break;
4471
4472 case OP_NOT_HSPACE:
4473 for (i = 1; i <= min; i++)
4474 {
4475 if (eptr >= md->end_subject)
4476 {
4477 SCHECK_PARTIAL();
4478 RRETURN(MATCH_NOMATCH);
4479 }
4480 GETCHARINC(c, eptr);
4481 switch(c)
4482 {
4483 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
4484 default: break;
4485 }
4486 }
4487 break;
4488
4489 case OP_HSPACE:
4490 for (i = 1; i <= min; i++)
4491 {
4492 if (eptr >= md->end_subject)
4493 {
4494 SCHECK_PARTIAL();
4495 RRETURN(MATCH_NOMATCH);
4496 }
4497 GETCHARINC(c, eptr);
4498 switch(c)
4499 {
4500 HSPACE_CASES: break; /* Byte and multibyte cases */
4501 default: RRETURN(MATCH_NOMATCH);
4502 }
4503 }
4504 break;
4505
4506 case OP_NOT_VSPACE:
4507 for (i = 1; i <= min; i++)
4508 {
4509 if (eptr >= md->end_subject)
4510 {
4511 SCHECK_PARTIAL();
4512 RRETURN(MATCH_NOMATCH);
4513 }
4514 GETCHARINC(c, eptr);
4515 switch(c)
4516 {
4517 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
4518 default: break;
4519 }
4520 }
4521 break;
4522
4523 case OP_VSPACE:
4524 for (i = 1; i <= min; i++)
4525 {
4526 if (eptr >= md->end_subject)
4527 {
4528 SCHECK_PARTIAL();
4529 RRETURN(MATCH_NOMATCH);
4530 }
4531 GETCHARINC(c, eptr);
4532 switch(c)
4533 {
4534 VSPACE_CASES: break;
4535 default: RRETURN(MATCH_NOMATCH);
4536 }
4537 }
4538 break;
4539
4540 case OP_NOT_DIGIT:
4541 for (i = 1; i <= min; i++)
4542 {
4543 if (eptr >= md->end_subject)
4544 {
4545 SCHECK_PARTIAL();
4546 RRETURN(MATCH_NOMATCH);
4547 }
4548 GETCHARINC(c, eptr);
4549 if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
4550 RRETURN(MATCH_NOMATCH);
4551 }
4552 break;
4553
4554 case OP_DIGIT:
4555 for (i = 1; i <= min; i++)
4556 {
4557 pcre_uint32 cc;
4558 if (eptr >= md->end_subject)
4559 {
4560 SCHECK_PARTIAL();
4561 RRETURN(MATCH_NOMATCH);
4562 }
4563 cc = RAWUCHAR(eptr);
4564 if (cc >= 128 || (md->ctypes[cc] & ctype_digit) == 0)
4565 RRETURN(MATCH_NOMATCH);
4566 eptr++;
4567 /* No need to skip more bytes - we know it's a 1-byte character */
4568 }
4569 break;
4570
4571 case OP_NOT_WHITESPACE:
4572 for (i = 1; i <= min; i++)
4573 {
4574 pcre_uint32 cc;
4575 if (eptr >= md->end_subject)
4576 {
4577 SCHECK_PARTIAL();
4578 RRETURN(MATCH_NOMATCH);
4579 }
4580 cc = RAWUCHAR(eptr);
4581 if (cc < 128 && (md->ctypes[cc] & ctype_space) != 0)
4582 RRETURN(MATCH_NOMATCH);
4583 eptr++;
4584 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4585 }
4586 break;
4587
4588 case OP_WHITESPACE:
4589 for (i = 1; i <= min; i++)
4590 {
4591 pcre_uint32 cc;
4592 if (eptr >= md->end_subject)
4593 {
4594 SCHECK_PARTIAL();
4595 RRETURN(MATCH_NOMATCH);
4596 }
4597 cc = RAWUCHAR(eptr);
4598 if (cc >= 128 || (md->ctypes[cc] & ctype_space) == 0)
4599 RRETURN(MATCH_NOMATCH);
4600 eptr++;
4601 /* No need to skip more bytes - we know it's a 1-byte character */
4602 }
4603 break;
4604
4605 case OP_NOT_WORDCHAR:
4606 for (i = 1; i <= min; i++)
4607 {
4608 pcre_uint32 cc;
4609 if (eptr >= md->end_subject)
4610 {
4611 SCHECK_PARTIAL();
4612 RRETURN(MATCH_NOMATCH);
4613 }
4614 cc = RAWUCHAR(eptr);
4615 if (cc < 128 && (md->ctypes[cc] & ctype_word) != 0)
4616 RRETURN(MATCH_NOMATCH);
4617 eptr++;
4618 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4619 }
4620 break;
4621
4622 case OP_WORDCHAR:
4623 for (i = 1; i <= min; i++)
4624 {
4625 pcre_uint32 cc;
4626 if (eptr >= md->end_subject)
4627 {
4628 SCHECK_PARTIAL();
4629 RRETURN(MATCH_NOMATCH);
4630 }
4631 cc = RAWUCHAR(eptr);
4632 if (cc >= 128 || (md->ctypes[cc] & ctype_word) == 0)
4633 RRETURN(MATCH_NOMATCH);
4634 eptr++;
4635 /* No need to skip more bytes - we know it's a 1-byte character */
4636 }
4637 break;
4638
4639 default:
4640 RRETURN(PCRE_ERROR_INTERNAL);
4641 } /* End switch(ctype) */
4642
4643 else
4644 #endif /* SUPPORT_UTF */
4645
4646 /* Code for the non-UTF-8 case for minimum matching of operators other
4647 than OP_PROP and OP_NOTPROP. */
4648
4649 switch(ctype)
4650 {
4651 case OP_ANY:
4652 for (i = 1; i <= min; i++)
4653 {
4654 if (eptr >= md->end_subject)
4655 {
4656 SCHECK_PARTIAL();
4657 RRETURN(MATCH_NOMATCH);
4658 }
4659 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4660 if (md->partial != 0 &&
4661 eptr + 1 >= md->end_subject &&
4662 NLBLOCK->nltype == NLTYPE_FIXED &&
4663 NLBLOCK->nllen == 2 &&
4664 *eptr == NLBLOCK->nl[0])
4665 {
4666 md->hitend = TRUE;
4667 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4668 }
4669 eptr++;
4670 }
4671 break;
4672
4673 case OP_ALLANY:
4674 if (eptr > md->end_subject - min)
4675 {
4676 SCHECK_PARTIAL();
4677 RRETURN(MATCH_NOMATCH);
4678 }
4679 eptr += min;
4680 break;
4681
4682 case OP_ANYBYTE:
4683 if (eptr > md->end_subject - min)
4684 {
4685 SCHECK_PARTIAL();
4686 RRETURN(MATCH_NOMATCH);
4687 }
4688 eptr += min;
4689 break;
4690
4691 case OP_ANYNL:
4692 for (i = 1; i <= min; i++)
4693 {
4694 if (eptr >= md->end_subject)
4695 {
4696 SCHECK_PARTIAL();
4697 RRETURN(MATCH_NOMATCH);
4698 }
4699 switch(*eptr++)
4700 {
4701 default: RRETURN(MATCH_NOMATCH);
4702
4703 case CHAR_CR:
4704 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
4705 break;
4706
4707 case CHAR_LF:
4708 break;
4709
4710 case CHAR_VT:
4711 case CHAR_FF:
4712 case CHAR_NEL:
4713 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4714 case 0x2028:
4715 case 0x2029:
4716 #endif
4717 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4718 break;
4719 }
4720 }
4721 break;
4722
4723 case OP_NOT_HSPACE:
4724 for (i = 1; i <= min; i++)
4725 {
4726 if (eptr >= md->end_subject)
4727 {
4728 SCHECK_PARTIAL();
4729 RRETURN(MATCH_NOMATCH);
4730 }
4731 switch(*eptr++)
4732 {
4733 default: break;
4734 HSPACE_BYTE_CASES:
4735 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4736 HSPACE_MULTIBYTE_CASES:
4737 #endif
4738 RRETURN(MATCH_NOMATCH);
4739 }
4740 }
4741 break;
4742
4743 case OP_HSPACE:
4744 for (i = 1; i <= min; i++)
4745 {
4746 if (eptr >= md->end_subject)
4747 {
4748 SCHECK_PARTIAL();
4749 RRETURN(MATCH_NOMATCH);
4750 }
4751 switch(*eptr++)
4752 {
4753 default: RRETURN(MATCH_NOMATCH);
4754 HSPACE_BYTE_CASES:
4755 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4756 HSPACE_MULTIBYTE_CASES:
4757 #endif
4758 break;
4759 }
4760 }
4761 break;
4762
4763 case OP_NOT_VSPACE:
4764 for (i = 1; i <= min; i++)
4765 {
4766 if (eptr >= md->end_subject)
4767 {
4768 SCHECK_PARTIAL();
4769 RRETURN(MATCH_NOMATCH);
4770 }
4771 switch(*eptr++)
4772 {
4773 VSPACE_BYTE_CASES:
4774 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4775 VSPACE_MULTIBYTE_CASES:
4776 #endif
4777 RRETURN(MATCH_NOMATCH);
4778 default: break;
4779 }
4780 }
4781 break;
4782
4783 case OP_VSPACE:
4784 for (i = 1; i <= min; i++)
4785 {
4786 if (eptr >= md->end_subject)
4787 {
4788 SCHECK_PARTIAL();
4789 RRETURN(MATCH_NOMATCH);
4790 }
4791 switch(*eptr++)
4792 {
4793 default: RRETURN(MATCH_NOMATCH);
4794 VSPACE_BYTE_CASES:
4795 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4796 VSPACE_MULTIBYTE_CASES:
4797 #endif
4798 break;
4799 }
4800 }
4801 break;
4802
4803 case OP_NOT_DIGIT:
4804 for (i = 1; i <= min; i++)
4805 {
4806 if (eptr >= md->end_subject)
4807 {
4808 SCHECK_PARTIAL();
4809 RRETURN(MATCH_NOMATCH);
4810 }
4811 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0)
4812 RRETURN(MATCH_NOMATCH);
4813 eptr++;
4814 }
4815 break;
4816
4817 case OP_DIGIT:
4818 for (i = 1; i <= min; i++)
4819 {
4820 if (eptr >= md->end_subject)
4821 {
4822 SCHECK_PARTIAL();
4823 RRETURN(MATCH_NOMATCH);
4824 }
4825 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0)
4826 RRETURN(MATCH_NOMATCH);
4827 eptr++;
4828 }
4829 break;
4830
4831 case OP_NOT_WHITESPACE:
4832 for (i = 1; i <= min; i++)
4833 {
4834 if (eptr >= md->end_subject)
4835 {
4836 SCHECK_PARTIAL();
4837 RRETURN(MATCH_NOMATCH);
4838 }
4839 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0)
4840 RRETURN(MATCH_NOMATCH);
4841 eptr++;
4842 }
4843 break;
4844
4845 case OP_WHITESPACE:
4846 for (i = 1; i <= min; i++)
4847 {
4848 if (eptr >= md->end_subject)
4849 {
4850 SCHECK_PARTIAL();
4851 RRETURN(MATCH_NOMATCH);
4852 }
4853 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0)
4854 RRETURN(MATCH_NOMATCH);
4855 eptr++;
4856 }
4857 break;
4858
4859 case OP_NOT_WORDCHAR:
4860 for (i = 1; i <= min; i++)
4861 {
4862 if (eptr >= md->end_subject)
4863 {
4864 SCHECK_PARTIAL();
4865 RRETURN(MATCH_NOMATCH);
4866 }
4867 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0)
4868 RRETURN(MATCH_NOMATCH);
4869 eptr++;
4870 }
4871 break;
4872
4873 case OP_WORDCHAR:
4874 for (i = 1; i <= min; i++)
4875 {
4876 if (eptr >= md->end_subject)
4877 {
4878 SCHECK_PARTIAL();
4879 RRETURN(MATCH_NOMATCH);
4880 }
4881 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0)
4882 RRETURN(MATCH_NOMATCH);
4883 eptr++;
4884 }
4885 break;
4886
4887 default:
4888 RRETURN(PCRE_ERROR_INTERNAL);
4889 }
4890 }
4891
4892 /* If min = max, continue at the same level without recursing */
4893
4894 if (min == max) continue;
4895
4896 /* If minimizing, we have to test the rest of the pattern before each
4897 subsequent match. Again, separate the UTF-8 case for speed, and also
4898 separate the UCP cases. */
4899
4900 if (minimize)
4901 {
4902 #ifdef SUPPORT_UCP
4903 if (prop_type >= 0)
4904 {
4905 switch(prop_type)
4906 {
4907 case PT_ANY:
4908 for (fi = min;; fi++)
4909 {
4910 RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4911 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4912 if (fi >= max) RRETURN(MATCH_NOMATCH);
4913 if (eptr >= md->end_subject)
4914 {
4915 SCHECK_PARTIAL();
4916 RRETURN(MATCH_NOMATCH);
4917 }
4918 GETCHARINCTEST(c, eptr);
4919 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4920 }
4921 /* Control never gets here */
4922
4923 case PT_LAMP:
4924 for (fi = min;; fi++)
4925 {
4926 int chartype;
4927 RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4928 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4929 if (fi >= max) RRETURN(MATCH_NOMATCH);
4930 if (eptr >= md->end_subject)
4931 {
4932 SCHECK_PARTIAL();
4933 RRETURN(MATCH_NOMATCH);
4934 }
4935 GETCHARINCTEST(c, eptr);
4936 chartype = UCD_CHARTYPE(c);
4937 if ((chartype == ucp_Lu ||
4938 chartype == ucp_Ll ||
4939 chartype == ucp_Lt) == prop_fail_result)
4940 RRETURN(MATCH_NOMATCH);
4941 }
4942 /* Control never gets here */
4943
4944 case PT_GC:
4945 for (fi = min;; fi++)
4946 {
4947 RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4948 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4949 if (fi >= max) RRETURN(MATCH_NOMATCH);
4950 if (eptr >= md->end_subject)
4951 {
4952 SCHECK_PARTIAL();
4953 RRETURN(MATCH_NOMATCH);
4954 }
4955 GETCHARINCTEST(c, eptr);
4956 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4957 RRETURN(MATCH_NOMATCH);
4958 }
4959 /* Control never gets here */
4960
4961 case PT_PC:
4962 for (fi = min;; fi++)
4963 {
4964 RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4965 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4966 if (fi >= max) RRETURN(MATCH_NOMATCH);
4967 if (eptr >= md->end_subject)
4968 {
4969 SCHECK_PARTIAL();
4970 RRETURN(MATCH_NOMATCH);
4971 }
4972 GETCHARINCTEST(c, eptr);
4973 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4974 RRETURN(MATCH_NOMATCH);
4975 }
4976 /* Control never gets here */
4977
4978 case PT_SC:
4979 for (fi = min;; fi++)
4980 {
4981 RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
4982 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4983 if (fi >= max) RRETURN(MATCH_NOMATCH);
4984 if (eptr >= md->end_subject)
4985 {
4986 SCHECK_PARTIAL();
4987 RRETURN(MATCH_NOMATCH);
4988 }
4989 GETCHARINCTEST(c, eptr);
4990 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4991 RRETURN(MATCH_NOMATCH);
4992 }
4993 /* Control never gets here */
4994
4995 case PT_ALNUM:
4996 for (fi = min;; fi++)
4997 {
4998 int category;
4999 RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
5000 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5001 if (fi >= max) RRETURN(MATCH_NOMATCH);
5002 if (eptr >= md->end_subject)
5003 {
5004 SCHECK_PARTIAL();
5005 RRETURN(MATCH_NOMATCH);
5006 }
5007 GETCHARINCTEST(c, eptr);
5008 category = UCD_CATEGORY(c);
5009 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5010 RRETURN(MATCH_NOMATCH);
5011 }
5012 /* Control never gets here */
5013
5014 case PT_SPACE: /* Perl space */
5015 for (fi = min;; fi++)
5016 {
5017 RMATCH(eptr, ecode, offset_top, md, eptrb, RM60);
5018 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5019 if (fi >= max) RRETURN(MATCH_NOMATCH);
5020 if (eptr >= md->end_subject)
5021 {
5022 SCHECK_PARTIAL();
5023 RRETURN(MATCH_NOMATCH);
5024 }
5025 GETCHARINCTEST(c, eptr);
5026 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5027 c == CHAR_FF || c == CHAR_CR)
5028 == prop_fail_result)
5029 RRETURN(MATCH_NOMATCH);
5030 }
5031 /* Control never gets here */
5032
5033 case PT_PXSPACE: /* POSIX space */
5034 for (fi = min;; fi++)
5035 {
5036 RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
5037 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5038 if (fi >= max) RRETURN(MATCH_NOMATCH);
5039 if (eptr >= md->end_subject)
5040 {
5041 SCHECK_PARTIAL();
5042 RRETURN(MATCH_NOMATCH);
5043 }
5044 GETCHARINCTEST(c, eptr);
5045 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5046 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5047 == prop_fail_result)
5048 RRETURN(MATCH_NOMATCH);
5049 }
5050 /* Control never gets here */
5051
5052 case PT_WORD:
5053 for (fi = min;; fi++)
5054 {
5055 int category;
5056 RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
5057 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5058 if (fi >= max) RRETURN(MATCH_NOMATCH);
5059 if (eptr >= md->end_subject)
5060 {
5061 SCHECK_PARTIAL();
5062 RRETURN(MATCH_NOMATCH);
5063 }
5064 GETCHARINCTEST(c, eptr);
5065 category = UCD_CATEGORY(c);
5066 if ((category == ucp_L ||
5067 category == ucp_N ||
5068 c == CHAR_UNDERSCORE)
5069 == prop_fail_result)
5070 RRETURN(MATCH_NOMATCH);
5071 }
5072 /* Control never gets here */
5073
5074 case PT_CLIST:
5075 for (fi = min;; fi++)
5076 {
5077 const pcre_uint32 *cp;
5078 RMATCH(eptr, ecode, offset_top, md, eptrb, RM67);
5079 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5080 if (fi >= max) RRETURN(MATCH_NOMATCH);
5081 if (eptr >= md->end_subject)
5082 {
5083 SCHECK_PARTIAL();
5084 RRETURN(MATCH_NOMATCH);
5085 }
5086 GETCHARINCTEST(c, eptr);
5087 cp = PRIV(ucd_caseless_sets) + prop_value;
5088 for (;;)
5089 {
5090 if (c < *cp)
5091 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
5092 if (c == *cp++)
5093 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
5094 }
5095 }
5096 /* Control never gets here */
5097
5098 case PT_UCNC:
5099 for (fi = min;; fi++)
5100 {
5101 RMATCH(eptr, ecode, offset_top, md, eptrb, RM68);
5102 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5103 if (fi >= max) RRETURN(MATCH_NOMATCH);
5104 if (eptr >= md->end_subject)
5105 {
5106 SCHECK_PARTIAL();
5107 RRETURN(MATCH_NOMATCH);
5108 }
5109 GETCHARINCTEST(c, eptr);
5110 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5111 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5112 c >= 0xe000) == prop_fail_result)
5113 RRETURN(MATCH_NOMATCH);
5114 }
5115 /* Control never gets here */
5116
5117 /* This should never occur */
5118 default:
5119 RRETURN(PCRE_ERROR_INTERNAL);
5120 }
5121 }
5122
5123 /* Match extended Unicode sequences. We will get here only if the
5124 support is in the binary; otherwise a compile-time error occurs. */
5125
5126 else if (ctype == OP_EXTUNI)
5127 {
5128 for (fi = min;; fi++)
5129 {
5130 RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
5131 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5132 if (fi >= max) RRETURN(MATCH_NOMATCH);
5133 if (eptr >= md->end_subject)
5134 {
5135 SCHECK_PARTIAL();
5136 RRETURN(MATCH_NOMATCH);
5137 }
5138 else
5139 {
5140 int lgb, rgb;
5141 GETCHARINCTEST(c, eptr);
5142 lgb = UCD_GRAPHBREAK(c);
5143 while (eptr < md->end_subject)
5144 {
5145 int len = 1;
5146 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5147 rgb = UCD_GRAPHBREAK(c);
5148 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5149 lgb = rgb;
5150 eptr += len;
5151 }
5152 }
5153 CHECK_PARTIAL();
5154 }
5155 }
5156 else
5157 #endif /* SUPPORT_UCP */
5158
5159 #ifdef SUPPORT_UTF
5160 if (utf)
5161 {
5162 for (fi = min;; fi++)
5163 {
5164 RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
5165 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5166 if (fi >= max) RRETURN(MATCH_NOMATCH);
5167 if (eptr >= md->end_subject)
5168 {
5169 SCHECK_PARTIAL();
5170 RRETURN(MATCH_NOMATCH);
5171 }
5172 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5173 RRETURN(MATCH_NOMATCH);
5174 GETCHARINC(c, eptr);
5175 switch(ctype)
5176 {
5177 case OP_ANY: /* This is the non-NL case */
5178 if (md->partial != 0 && /* Take care with CRLF partial */
5179 eptr >= md->end_subject &&
5180 NLBLOCK->nltype == NLTYPE_FIXED &&
5181 NLBLOCK->nllen == 2 &&
5182 c == NLBLOCK->nl[0])
5183 {
5184 md->hitend = TRUE;
5185 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5186 }
5187 break;
5188
5189 case OP_ALLANY:
5190 case OP_ANYBYTE:
5191 break;
5192
5193 case OP_ANYNL:
5194 switch(c)
5195 {
5196 default: RRETURN(MATCH_NOMATCH);
5197 case CHAR_CR:
5198 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
5199 break;
5200
5201 case CHAR_LF:
5202 break;
5203
5204 case CHAR_VT:
5205 case CHAR_FF:
5206 case CHAR_NEL:
5207 #ifndef EBCDIC
5208 case 0x2028:
5209 case 0x2029:
5210 #endif /* Not EBCDIC */
5211 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5212 break;
5213 }
5214 break;
5215
5216 case OP_NOT_HSPACE:
5217 switch(c)
5218 {
5219 HSPACE_CASES: RRETURN(MATCH_NOMATCH);
5220 default: break;
5221 }
5222 break;
5223
5224 case OP_HSPACE:
5225 switch(c)
5226 {
5227 HSPACE_CASES: break;
5228 default: RRETURN(MATCH_NOMATCH);
5229 }
5230 break;
5231
5232 case OP_NOT_VSPACE:
5233 switch(c)
5234 {
5235 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
5236 default: break;
5237 }
5238 break;
5239
5240 case OP_VSPACE:
5241 switch(c)
5242 {
5243 VSPACE_CASES: break;
5244 default: RRETURN(MATCH_NOMATCH);
5245 }
5246 break;
5247
5248 case OP_NOT_DIGIT:
5249 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
5250 RRETURN(MATCH_NOMATCH);
5251 break;
5252
5253 case OP_DIGIT:
5254 if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0)
5255 RRETURN(MATCH_NOMATCH);
5256 break;
5257
5258 case OP_NOT_WHITESPACE:
5259 if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
5260 RRETURN(MATCH_NOMATCH);
5261 break;
5262
5263 case OP_WHITESPACE:
5264 if (c >= 256 || (md->ctypes[c] & ctype_space) == 0)
5265 RRETURN(MATCH_NOMATCH);
5266 break;
5267
5268 case OP_NOT_WORDCHAR:
5269 if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
5270 RRETURN(MATCH_NOMATCH);
5271 break;
5272
5273 case OP_WORDCHAR:
5274 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0)
5275 RRETURN(MATCH_NOMATCH);
5276 break;
5277
5278 default:
5279 RRETURN(PCRE_ERROR_INTERNAL);
5280 }
5281 }
5282 }
5283 else
5284 #endif
5285 /* Not UTF mode */
5286 {
5287 for (fi = min;; fi++)
5288 {
5289 RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
5290 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5291 if (fi >= max) RRETURN(MATCH_NOMATCH);
5292 if (eptr >= md->end_subject)
5293 {
5294 SCHECK_PARTIAL();
5295 RRETURN(MATCH_NOMATCH);
5296 }
5297 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5298 RRETURN(MATCH_NOMATCH);
5299 c = *eptr++;
5300 switch(ctype)
5301 {
5302 case OP_ANY: /* This is the non-NL case */
5303 if (md->partial != 0 && /* Take care with CRLF partial */
5304 eptr >= md->end_subject &&
5305 NLBLOCK->nltype == NLTYPE_FIXED &&
5306 NLBLOCK->nllen == 2 &&
5307 c == NLBLOCK->nl[0])
5308 {
5309 md->hitend = TRUE;
5310 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5311 }
5312 break;
5313
5314 case OP_ALLANY:
5315 case OP_ANYBYTE:
5316 break;
5317
5318 case OP_ANYNL:
5319 switch(c)
5320 {
5321 default: RRETURN(MATCH_NOMATCH);
5322 case CHAR_CR:
5323 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
5324 break;
5325
5326 case CHAR_LF:
5327 break;
5328
5329 case CHAR_VT:
5330 case CHAR_FF:
5331 case CHAR_NEL:
5332 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5333 case 0x2028:
5334 case 0x2029:
5335 #endif
5336 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5337 break;
5338 }
5339 break;
5340
5341 case OP_NOT_HSPACE:
5342 switch(c)
5343 {
5344 default: break;
5345 HSPACE_BYTE_CASES:
5346 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5347 HSPACE_MULTIBYTE_CASES:
5348 #endif
5349 RRETURN(MATCH_NOMATCH);
5350 }
5351 break;
5352
5353 case OP_HSPACE:
5354 switch(c)
5355 {
5356 default: RRETURN(MATCH_NOMATCH);
5357 HSPACE_BYTE_CASES:
5358 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5359 HSPACE_MULTIBYTE_CASES:
5360 #endif
5361 break;
5362 }
5363 break;
5364
5365 case OP_NOT_VSPACE:
5366 switch(c)
5367 {
5368 default: break;
5369 VSPACE_BYTE_CASES:
5370 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5371 VSPACE_MULTIBYTE_CASES:
5372 #endif
5373 RRETURN(MATCH_NOMATCH);
5374 }
5375 break;
5376
5377 case OP_VSPACE:
5378 switch(c)
5379 {
5380 default: RRETURN(MATCH_NOMATCH);
5381 VSPACE_BYTE_CASES:
5382 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5383 VSPACE_MULTIBYTE_CASES:
5384 #endif
5385 break;
5386 }
5387 break;
5388
5389 case OP_NOT_DIGIT:
5390 if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
5391 break;
5392
5393 case OP_DIGIT:
5394 if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
5395 break;
5396
5397 case OP_NOT_WHITESPACE:
5398 if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
5399 break;
5400
5401 case OP_WHITESPACE:
5402 if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
5403 break;
5404
5405 case OP_NOT_WORDCHAR:
5406 if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH);
5407 break;
5408
5409 case OP_WORDCHAR:
5410 if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH);
5411 break;
5412
5413 default:
5414 RRETURN(PCRE_ERROR_INTERNAL);
5415 }
5416 }
5417 }
5418 /* Control never gets here */
5419 }
5420
5421 /* If maximizing, it is worth using inline code for speed, doing the type
5422 test once at the start (i.e. keep it out of the loop). Again, keep the
5423 UTF-8 and UCP stuff separate. */
5424
5425 else
5426 {
5427 pp = eptr; /* Remember where we started */
5428
5429 #ifdef SUPPORT_UCP
5430 if (prop_type >= 0)
5431 {
5432 switch(prop_type)
5433 {
5434 case PT_ANY:
5435 for (i = min; i < max; i++)
5436 {
5437 int len = 1;
5438 if (eptr >= md->end_subject)
5439 {
5440 SCHECK_PARTIAL();
5441 break;
5442 }
5443 GETCHARLENTEST(c, eptr, len);
5444 if (prop_fail_result) break;
5445 eptr+= len;
5446 }
5447 break;
5448
5449 case PT_LAMP:
5450 for (i = min; i < max; i++)
5451 {
5452 int chartype;
5453 int len = 1;
5454 if (eptr >= md->end_subject)
5455 {
5456 SCHECK_PARTIAL();
5457 break;
5458 }
5459 GETCHARLENTEST(c, eptr, len);
5460 chartype = UCD_CHARTYPE(c);
5461 if ((chartype == ucp_Lu ||
5462 chartype == ucp_Ll ||
5463 chartype == ucp_Lt) == prop_fail_result)
5464 break;
5465 eptr+= len;
5466 }
5467 break;
5468
5469 case PT_GC:
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 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
5480 eptr+= len;
5481 }
5482 break;
5483
5484 case PT_PC:
5485 for (i = min; i < max; i++)
5486 {
5487 int len = 1;
5488 if (eptr >= md->end_subject)
5489 {
5490 SCHECK_PARTIAL();
5491 break;
5492 }
5493 GETCHARLENTEST(c, eptr, len);
5494 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
5495 eptr+= len;
5496 }
5497 break;
5498
5499 case PT_SC:
5500 for (i = min; i < max; i++)
5501 {
5502 int len = 1;
5503 if (eptr >= md->end_subject)
5504 {
5505 SCHECK_PARTIAL();
5506 break;
5507 }
5508 GETCHARLENTEST(c, eptr, len);
5509 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
5510 eptr+= len;
5511 }
5512 break;
5513
5514 case PT_ALNUM:
5515 for (i = min; i < max; i++)
5516 {
5517 int category;
5518 int len = 1;
5519 if (eptr >= md->end_subject)
5520 {
5521 SCHECK_PARTIAL();
5522 break;
5523 }
5524 GETCHARLENTEST(c, eptr, len);
5525 category = UCD_CATEGORY(c);
5526 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5527 break;
5528 eptr+= len;
5529 }
5530 break;
5531
5532 case PT_SPACE: /* Perl space */
5533 for (i = min; i < max; i++)
5534 {
5535 int len = 1;
5536 if (eptr >= md->end_subject)
5537 {
5538 SCHECK_PARTIAL();
5539 break;
5540 }
5541 GETCHARLENTEST(c, eptr, len);
5542 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5543 c == CHAR_FF || c == CHAR_CR)
5544 == prop_fail_result)
5545 break;
5546 eptr+= len;
5547 }
5548 break;
5549
5550 case PT_PXSPACE: /* POSIX space */
5551 for (i = min; i < max; i++)
5552 {
5553 int len = 1;
5554 if (eptr >= md->end_subject)
5555 {
5556 SCHECK_PARTIAL();
5557 break;
5558 }
5559 GETCHARLENTEST(c, eptr, len);
5560 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5561 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5562 == prop_fail_result)
5563 break;
5564 eptr+= len;
5565 }
5566 break;
5567
5568 case PT_WORD:
5569 for (i = min; i < max; i++)
5570 {
5571 int category;
5572 int len = 1;
5573 if (eptr >= md->end_subject)
5574 {
5575 SCHECK_PARTIAL();
5576 break;
5577 }
5578 GETCHARLENTEST(c, eptr, len);
5579 category = UCD_CATEGORY(c);
5580 if ((category == ucp_L || category == ucp_N ||
5581 c == CHAR_UNDERSCORE) == prop_fail_result)
5582 break;
5583 eptr+= len;
5584 }
5585 break;
5586
5587 case PT_CLIST:
5588 for (i = min; i < max; i++)
5589 {
5590 const pcre_uint32 *cp;
5591 int len = 1;
5592 if (eptr >= md->end_subject)
5593 {
5594 SCHECK_PARTIAL();
5595 break;
5596 }
5597 GETCHARLENTEST(c, eptr, len);
5598 cp = PRIV(ucd_caseless_sets) + prop_value;
5599 for (;;)
5600 {
5601 if (c < *cp)
5602 { if (prop_fail_result) break; else goto GOT_MAX; }
5603 if (c == *cp++)
5604 { if (prop_fail_result) goto GOT_MAX; else break; }
5605 }
5606 eptr += len;
5607 }
5608 GOT_MAX:
5609 break;
5610
5611 case PT_UCNC:
5612 for (i = min; i < max; i++)
5613 {
5614 int len = 1;
5615 if (eptr >= md->end_subject)
5616 {
5617 SCHECK_PARTIAL();
5618 break;
5619 }
5620 GETCHARLENTEST(c, eptr, len);
5621 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5622 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5623 c >= 0xe000) == prop_fail_result)
5624 break;
5625 eptr += len;
5626 }
5627 break;
5628
5629 default:
5630 RRETURN(PCRE_ERROR_INTERNAL);
5631 }
5632
5633 /* eptr is now past the end of the maximum run */
5634
5635 if (possessive) continue; /* No backtracking */
5636 for(;;)
5637 {
5638 if (eptr == pp) goto TAIL_RECURSE;
5639 RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5640 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5641 eptr--;
5642 if (utf) BACKCHAR(eptr);
5643 }
5644 }
5645
5646 /* Match extended Unicode sequences. We will get here only if the
5647 support is in the binary; otherwise a compile-time error occurs. */
5648
5649 else if (ctype == OP_EXTUNI)
5650 {
5651 for (i = min; i < max; i++)
5652 {
5653 if (eptr >= md->end_subject)
5654 {
5655 SCHECK_PARTIAL();
5656 break;
5657 }
5658 else
5659 {
5660 int lgb, rgb;
5661 GETCHARINCTEST(c, eptr);
5662 lgb = UCD_GRAPHBREAK(c);
5663 while (eptr < md->end_subject)
5664 {
5665 int len = 1;
5666 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5667 rgb = UCD_GRAPHBREAK(c);
5668 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5669 lgb = rgb;
5670 eptr += len;
5671 }
5672 }
5673 CHECK_PARTIAL();
5674 }
5675
5676 /* eptr is now past the end of the maximum run */
5677
5678 if (possessive) continue; /* No backtracking */
5679 for(;;)
5680 {
5681 if (eptr == pp) goto TAIL_RECURSE;
5682 RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5683 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5684 eptr--;
5685 for (;;) /* Move back over one extended */
5686 {
5687 if (!utf) c = *eptr; else
5688 {
5689 BACKCHAR(eptr);
5690 GETCHAR(c, eptr);
5691 }
5692 if (UCD_CATEGORY(c) != ucp_M) break;
5693 eptr--;
5694 }
5695 }
5696 }
5697
5698 else
5699 #endif /* SUPPORT_UCP */
5700
5701 #ifdef SUPPORT_UTF
5702 if (utf)
5703 {
5704 switch(ctype)
5705 {
5706 case OP_ANY:
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 if (IS_NEWLINE(eptr)) break;
5717 if (md->partial != 0 && /* Take care with CRLF partial */
5718 eptr + 1 >= md->end_subject &&
5719 NLBLOCK->nltype == NLTYPE_FIXED &&
5720 NLBLOCK->nllen == 2 &&
5721 RAWUCHAR(eptr) == NLBLOCK->nl[0])
5722 {
5723 md->hitend = TRUE;
5724 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5725 }
5726 eptr++;
5727 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5728 }
5729 }
5730
5731 /* Handle unlimited UTF-8 repeat */
5732
5733 else
5734 {
5735 for (i = min; i < max; i++)
5736 {
5737 if (eptr >= md->end_subject)
5738 {
5739 SCHECK_PARTIAL();
5740 break;
5741 }
5742 if (IS_NEWLINE(eptr)) break;
5743 if (md->partial != 0 && /* Take care with CRLF partial */
5744 eptr + 1 >= md->end_subject &&
5745 NLBLOCK->nltype == NLTYPE_FIXED &&
5746 NLBLOCK->nllen == 2 &&
5747 RAWUCHAR(eptr) == NLBLOCK->nl[0])
5748 {
5749 md->hitend = TRUE;
5750 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5751 }
5752 eptr++;
5753 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5754 }
5755 }
5756 break;
5757
5758 case OP_ALLANY:
5759 if (max < INT_MAX)
5760 {
5761 for (i = min; i < max; i++)
5762 {
5763 if (eptr >= md->end_subject)
5764 {
5765 SCHECK_PARTIAL();
5766 break;
5767 }
5768 eptr++;
5769 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5770 }
5771 }
5772 else
5773 {
5774 eptr = md->end_subject; /* Unlimited UTF-8 repeat */
5775 SCHECK_PARTIAL();
5776 }
5777 break;
5778
5779 /* The byte case is the same as non-UTF8 */
5780
5781 case OP_ANYBYTE:
5782 c = max - min;
5783 if (c > (unsigned int)(md->end_subject - eptr))
5784 {
5785 eptr = md->end_subject;
5786 SCHECK_PARTIAL();
5787 }
5788 else eptr += c;
5789 break;
5790
5791 case OP_ANYNL:
5792 for (i = min; i < max; i++)
5793 {
5794 int len = 1;
5795 if (eptr >= md->end_subject)
5796 {
5797 SCHECK_PARTIAL();
5798 break;
5799 }
5800 GETCHARLEN(c, eptr, len);
5801 if (c == CHAR_CR)
5802 {
5803 if (++eptr >= md->end_subject) break;
5804 if (RAWUCHAR(eptr) == CHAR_LF) eptr++;
5805 }
5806 else
5807 {
5808 if (c != CHAR_LF &&
5809 (md->bsr_anycrlf ||
5810 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
5811 #ifndef EBCDIC
5812 && c != 0x2028 && c != 0x2029
5813 #endif /* Not EBCDIC */
5814 )))
5815 break;
5816 eptr += len;
5817 }
5818 }
5819 break;
5820
5821 case OP_NOT_HSPACE:
5822 case OP_HSPACE:
5823 for (i = min; i < max; i++)
5824 {
5825 BOOL gotspace;
5826 int len = 1;
5827 if (eptr >= md->end_subject)
5828 {
5829 SCHECK_PARTIAL();
5830 break;
5831 }
5832 GETCHARLEN(c, eptr, len);
5833 switch(c)
5834 {
5835 HSPACE_CASES: gotspace = TRUE; break;
5836 default: gotspace = FALSE; break;
5837 }
5838 if (gotspace == (ctype == OP_NOT_HSPACE)) break;
5839 eptr += len;
5840 }
5841 break;
5842
5843 case OP_NOT_VSPACE:
5844 case OP_VSPACE:
5845 for (i = min; i < max; i++)
5846 {
5847 BOOL gotspace;
5848 int len = 1;
5849 if (eptr >= md->end_subject)
5850 {
5851 SCHECK_PARTIAL();
5852 break;
5853 }
5854 GETCHARLEN(c, eptr, len);
5855 switch(c)
5856 {
5857 VSPACE_CASES: gotspace = TRUE; break;
5858 default: gotspace = FALSE; break;
5859 }
5860 if (gotspace == (ctype == OP_NOT_VSPACE)) break;
5861 eptr += len;
5862 }
5863 break;
5864
5865 case OP_NOT_DIGIT:
5866 for (i = min; i < max; i++)
5867 {
5868 int len = 1;
5869 if (eptr >= md->end_subject)
5870 {
5871 SCHECK_PARTIAL();
5872 break;
5873 }
5874 GETCHARLEN(c, eptr, len);
5875 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
5876 eptr+= len;
5877 }
5878 break;
5879
5880 case OP_DIGIT:
5881 for (i = min; i < max; i++)
5882 {
5883 int len = 1;
5884 if (eptr >= md->end_subject)
5885 {
5886 SCHECK_PARTIAL();
5887 break;
5888 }
5889 GETCHARLEN(c, eptr, len);
5890 if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break;
5891 eptr+= len;
5892 }
5893 break;
5894
5895 case OP_NOT_WHITESPACE:
5896 for (i = min; i < max; i++)
5897 {
5898 int len = 1;
5899 if (eptr >= md->end_subject)
5900 {
5901 SCHECK_PARTIAL();
5902 break;
5903 }
5904 GETCHARLEN(c, eptr, len);
5905 if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
5906 eptr+= len;
5907 }
5908 break;
5909
5910 case OP_WHITESPACE:
5911 for (i = min; i < max; i++)
5912 {
5913 int len = 1;
5914 if (eptr >= md->end_subject)
5915 {
5916 SCHECK_PARTIAL();
5917 break;
5918 }
5919 GETCHARLEN(c, eptr, len);
5920 if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break;
5921 eptr+= len;
5922 }
5923 break;
5924
5925 case OP_NOT_WORDCHAR:
5926 for (i = min; i < max; i++)
5927 {
5928 int len = 1;
5929 if (eptr >= md->end_subject)
5930 {
5931 SCHECK_PARTIAL();
5932 break;
5933 }
5934 GETCHARLEN(c, eptr, len);
5935 if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
5936 eptr+= len;
5937 }
5938 break;
5939
5940 case OP_WORDCHAR:
5941 for (i = min; i < max; i++)
5942 {
5943 int len = 1;
5944 if (eptr >= md->end_subject)
5945 {
5946 SCHECK_PARTIAL();
5947 break;
5948 }
5949 GETCHARLEN(c, eptr, len);
5950 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break;
5951 eptr+= len;
5952 }
5953 break;
5954
5955 default:
5956 RRETURN(PCRE_ERROR_INTERNAL);
5957 }
5958
5959 if (possessive) continue; /* No backtracking */
5960 for(;;)
5961 {
5962 if (eptr == pp) goto TAIL_RECURSE;
5963 RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
5964 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5965 eptr--;
5966 BACKCHAR(eptr);
5967 if (ctype == OP_ANYNL && eptr > pp && RAWUCHAR(eptr) == CHAR_NL &&
5968 RAWUCHAR(eptr - 1) == CHAR_CR) eptr--;
5969 }
5970 }
5971 else
5972 #endif /* SUPPORT_UTF */
5973 /* Not UTF mode */
5974 {
5975 switch(ctype)
5976 {
5977 case OP_ANY:
5978 for (i = min; i < max; i++)
5979 {
5980 if (eptr >= md->end_subject)
5981 {
5982 SCHECK_PARTIAL();
5983 break;
5984 }
5985 if (IS_NEWLINE(eptr)) break;
5986 if (md->partial != 0 && /* Take care with CRLF partial */
5987 eptr + 1 >= md->end_subject &&
5988 NLBLOCK->nltype == NLTYPE_FIXED &&
5989 NLBLOCK->nllen == 2 &&
5990 *eptr == NLBLOCK->nl[0])
5991 {
5992 md->hitend = TRUE;
5993 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5994 }
5995 eptr++;
5996 }
5997 break;
5998
5999 case OP_ALLANY:
6000 case OP_ANYBYTE:
6001 c = max - min;
6002 if (c > (unsigned int)(md->end_subject - eptr))
6003 {
6004 eptr = md->end_subject;
6005 SCHECK_PARTIAL();
6006 }
6007 else eptr += c;
6008 break;
6009
6010 case OP_ANYNL:
6011 for (i = min; i < max; i++)
6012 {
6013 if (eptr >= md->end_subject)
6014 {
6015 SCHECK_PARTIAL();
6016 break;
6017 }
6018 c = *eptr;
6019 if (c == CHAR_CR)
6020 {
6021 if (++eptr >= md->end_subject) break;
6022 if (*eptr == CHAR_LF) eptr++;
6023 }
6024 else
6025 {
6026 if (c != CHAR_LF && (md->bsr_anycrlf ||
6027 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
6028 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6029 && c != 0x2028 && c != 0x2029
6030 #endif
6031 ))) break;
6032 eptr++;
6033 }
6034 }
6035 break;
6036
6037 case OP_NOT_HSPACE:
6038 for (i = min; i < max; i++)
6039 {
6040 if (eptr >= md->end_subject)
6041 {
6042 SCHECK_PARTIAL();
6043 break;
6044 }
6045 switch(*eptr)
6046 {
6047 default: eptr++; break;
6048 HSPACE_BYTE_CASES:
6049 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6050 HSPACE_MULTIBYTE_CASES:
6051 #endif
6052 goto ENDLOOP00;
6053 }
6054 }
6055 ENDLOOP00:
6056 break;
6057
6058 case OP_HSPACE:
6059 for (i = min; i < max; i++)
6060 {
6061 if (eptr >= md->end_subject)
6062 {
6063 SCHECK_PARTIAL();
6064 break;
6065 }
6066 switch(*eptr)
6067 {
6068 default: goto ENDLOOP01;
6069 HSPACE_BYTE_CASES:
6070 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6071 HSPACE_MULTIBYTE_CASES:
6072 #endif
6073 eptr++; break;
6074 }
6075 }
6076 ENDLOOP01:
6077 break;
6078
6079 case OP_NOT_VSPACE:
6080 for (i = min; i < max; i++)
6081 {
6082 if (eptr >= md->end_subject)
6083 {
6084 SCHECK_PARTIAL();
6085 break;
6086 }
6087 switch(*eptr)
6088 {
6089 default: eptr++; break;
6090 VSPACE_BYTE_CASES:
6091 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6092 VSPACE_MULTIBYTE_CASES:
6093 #endif
6094 goto ENDLOOP02;
6095 }
6096 }
6097 ENDLOOP02:
6098 break;
6099
6100 case OP_VSPACE:
6101 for (i = min; i < max; i++)
6102 {
6103 if (eptr >= md->end_subject)
6104 {
6105 SCHECK_PARTIAL();
6106 break;
6107 }
6108 switch(*eptr)
6109 {
6110 default: goto ENDLOOP03;
6111 VSPACE_BYTE_CASES:
6112 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6113 VSPACE_MULTIBYTE_CASES:
6114 #endif
6115 eptr++; break;
6116 }
6117 }
6118 ENDLOOP03:
6119 break;
6120
6121 case OP_NOT_DIGIT:
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_digit) != 0) break;
6130 eptr++;
6131 }
6132 break;
6133
6134 case OP_DIGIT:
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_digit) == 0) break;
6143 eptr++;
6144 }
6145 break;
6146
6147 case OP_NOT_WHITESPACE:
6148 for (i = min; i < max; i++)
6149 {
6150 if (eptr >= md->end_subject)
6151 {
6152 SCHECK_PARTIAL();
6153 break;
6154 }
6155 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break;
6156 eptr++;
6157 }
6158 break;
6159
6160 case OP_WHITESPACE:
6161 for (i = min; i < max; i++)
6162 {
6163 if (eptr >= md->end_subject)
6164 {
6165 SCHECK_PARTIAL();
6166 break;
6167 }
6168 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break;
6169 eptr++;
6170 }
6171 break;
6172
6173 case OP_NOT_WORDCHAR:
6174 for (i = min; i < max; i++)
6175 {
6176 if (eptr >= md->end_subject)
6177 {
6178 SCHECK_PARTIAL();
6179 break;
6180 }
6181 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break;
6182 eptr++;
6183 }
6184 break;
6185
6186 case OP_WORDCHAR:
6187 for (i = min; i < max; i++)
6188 {
6189 if (eptr >= md->end_subject)
6190 {
6191 SCHECK_PARTIAL();
6192 break;
6193 }
6194 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break;
6195 eptr++;
6196 }
6197 break;
6198
6199 default:
6200 RRETURN(PCRE_ERROR_INTERNAL);
6201 }
6202
6203 if (possessive) continue; /* No backtracking */
6204 for (;;)
6205 {
6206 if (eptr == pp) goto TAIL_RECURSE;
6207 RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
6208 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
6209 eptr--;
6210 if (ctype == OP_ANYNL && eptr > pp && *eptr == CHAR_LF &&
6211 eptr[-1] == CHAR_CR) eptr--;
6212 }
6213 }
6214
6215 /* Get here if we can't make it match with any permitted repetitions */
6216
6217 RRETURN(MATCH_NOMATCH);
6218 }
6219 /* Control never gets here */
6220
6221 /* There's been some horrible disaster. Arrival here can only mean there is
6222 something seriously wrong in the code above or the OP_xxx definitions. */
6223
6224 default:
6225 DPRINTF(("Unknown opcode %d\n", *ecode));
6226 RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
6227 }
6228
6229 /* Do not stick any code in here without much thought; it is assumed
6230 that "continue" in the code above comes out to here to repeat the main
6231 loop. */
6232
6233 } /* End of main loop */
6234 /* Control never reaches here */
6235
6236
6237 /* When compiling to use the heap rather than the stack for recursive calls to
6238 match(), the RRETURN() macro jumps here. The number that is saved in
6239 frame->Xwhere indicates which label we actually want to return to. */
6240
6241 #ifdef NO_RECURSE
6242 #define LBL(val) case val: goto L_RM##val;
6243 HEAP_RETURN:
6244 switch (frame->Xwhere)
6245 {
6246 LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
6247 LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
6248 LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
6249 LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
6250 LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64)
6251 LBL(65) LBL(66)
6252 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6253 LBL(21)
6254 #endif
6255 #ifdef SUPPORT_UTF
6256 LBL(16) LBL(18) LBL(20)
6257 LBL(22) LBL(23) LBL(28) LBL(30)
6258 LBL(32) LBL(34) LBL(42) LBL(46)
6259 #ifdef SUPPORT_UCP
6260 LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
6261 LBL(59) LBL(60) LBL(61) LBL(62) LBL(67) LBL(68)
6262 #endif /* SUPPORT_UCP */
6263 #endif /* SUPPORT_UTF */
6264 default:
6265 DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
6266 return PCRE_ERROR_INTERNAL;
6267 }
6268 #undef LBL
6269 #endif /* NO_RECURSE */
6270 }
6271
6272
6273 /***************************************************************************
6274 ****************************************************************************
6275 RECURSION IN THE match() FUNCTION
6276
6277 Undefine all the macros that were defined above to handle this. */
6278
6279 #ifdef NO_RECURSE
6280 #undef eptr
6281 #undef ecode
6282 #undef mstart
6283 #undef offset_top
6284 #undef eptrb
6285 #undef flags
6286
6287 #undef callpat
6288 #undef charptr
6289 #undef data
6290 #undef next
6291 #undef pp
6292 #undef prev
6293 #undef saved_eptr
6294
6295 #undef new_recursive
6296
6297 #undef cur_is_word
6298 #undef condition
6299 #undef prev_is_word
6300
6301 #undef ctype
6302 #undef length
6303 #undef max
6304 #undef min
6305 #undef number
6306 #undef offset
6307 #undef op
6308 #undef save_capture_last
6309 #undef save_offset1
6310 #undef save_offset2
6311 #undef save_offset3
6312 #undef stacksave
6313
6314 #undef newptrb
6315
6316 #endif
6317
6318 /* These two are defined as macros in both cases */
6319
6320 #undef fc
6321 #undef fi
6322
6323 /***************************************************************************
6324 ***************************************************************************/
6325
6326
6327 #ifdef NO_RECURSE
6328 /*************************************************
6329 * Release allocated heap frames *
6330 *************************************************/
6331
6332 /* This function releases all the allocated frames. The base frame is on the
6333 machine stack, and so must not be freed.
6334
6335 Argument: the address of the base frame
6336 Returns: nothing
6337 */
6338
6339 static void
6340 release_match_heapframes (heapframe *frame_base)
6341 {
6342 heapframe *nextframe = frame_base->Xnextframe;
6343 while (nextframe != NULL)
6344 {
6345 heapframe *oldframe = nextframe;
6346 nextframe = nextframe->Xnextframe;
6347 (PUBL(stack_free))(oldframe);
6348 }
6349 }
6350 #endif
6351
6352
6353 /*************************************************
6354 * Execute a Regular Expression *
6355 *************************************************/
6356
6357 /* This function applies a compiled re to a subject string and picks out
6358 portions of the string if it matches. Two elements in the vector are set for
6359 each substring: the offsets to the start and end of the substring.
6360
6361 Arguments:
6362 argument_re points to the compiled expression
6363 extra_data points to extra data or is NULL
6364 subject points to the subject string
6365 length length of subject string (may contain binary zeros)
6366 start_offset where to start in the subject string
6367 options option bits
6368 offsets points to a vector of ints to be filled in with offsets
6369 offsetcount the number of elements in the vector
6370
6371 Returns: > 0 => success; value is the number of elements filled in
6372 = 0 => success, but offsets is not big enough
6373 -1 => failed to match
6374 < -1 => some kind of unexpected problem
6375 */
6376
6377 #if defined COMPILE_PCRE8
6378 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6379 pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
6380 PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
6381 int offsetcount)
6382 #elif defined COMPILE_PCRE16
6383 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6384 pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data,
6385 PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets,
6386 int offsetcount)
6387 #elif defined COMPILE_PCRE32
6388 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6389 pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data,
6390 PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets,
6391 int offsetcount)
6392 #endif
6393 {
6394 int rc, ocount, arg_offset_max;
6395 int newline;
6396 BOOL using_temporary_offsets = FALSE;
6397 BOOL anchored;
6398 BOOL startline;
6399 BOOL firstline;
6400 BOOL utf;
6401 BOOL has_first_char = FALSE;
6402 BOOL has_req_char = FALSE;
6403 pcre_uchar first_char = 0;
6404 pcre_uchar first_char2 = 0;
6405 pcre_uchar req_char = 0;
6406 pcre_uchar req_char2 = 0;
6407 match_data match_block;
6408 match_data *md = &match_block;
6409 const pcre_uint8 *tables;
6410 const pcre_uint8 *start_bits = NULL;
6411 PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset;
6412 PCRE_PUCHAR end_subject;
6413 PCRE_PUCHAR start_partial = NULL;
6414 PCRE_PUCHAR match_partial;
6415 PCRE_PUCHAR req_char_ptr = start_match - 1;
6416
6417 const pcre_study_data *study;
6418 const REAL_PCRE *re = (const REAL_PCRE *)argument_re;
6419
6420 #ifdef NO_RECURSE
6421 heapframe frame_zero;
6422 frame_zero.Xprevframe = NULL; /* Marks the top level */
6423 frame_zero.Xnextframe = NULL; /* None are allocated yet */
6424 md->match_frames_base = &frame_zero;
6425 #endif
6426
6427 /* Check for the special magic call that measures the size of the stack used
6428 per recursive call of match(). Without the funny casting for sizeof, a Windows
6429 compiler gave this error: "unary minus operator applied to unsigned type,
6430 result still unsigned". Hopefully the cast fixes that. */
6431
6432 if (re == NULL && extra_data == NULL && subject == NULL && length == -999 &&
6433 start_offset == -999)
6434 #ifdef NO_RECURSE
6435 return -((int)sizeof(heapframe));
6436 #else
6437 return match(NULL, NULL, NULL, 0, NULL, NULL, 0);
6438 #endif
6439
6440 /* Plausibility checks */
6441
6442 if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
6443 if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0))
6444 return PCRE_ERROR_NULL;
6445 if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
6446 if (length < 0) return PCRE_ERROR_BADLENGTH;
6447 if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
6448
6449 /* Check that the first field in the block is the magic number. If it is not,
6450 return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to
6451 REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which
6452 means that the pattern is likely compiled with different endianness. */
6453
6454 if (re->magic_number != MAGIC_NUMBER)
6455 return re->magic_number == REVERSED_MAGIC_NUMBER?
6456 PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC;
6457 if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE;
6458
6459 /* These two settings are used in the code for checking a UTF-8 string that
6460 follows immediately afterwards. Other values in the md block are used only
6461 during "normal" pcre_exec() processing, not when the JIT support is in use,
6462 so they are set up later. */
6463
6464 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
6465 utf = md->utf = (re->options & PCRE_UTF8) != 0;
6466 md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
6467 ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
6468
6469 /* Check a UTF-8 string if required. Pass back the character offset and error
6470 code for an invalid string if a results vector is available. */
6471
6472 #ifdef SUPPORT_UTF
6473 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0)
6474 {
6475 int erroroffset;
6476 int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset);
6477 if (errorcode != 0)
6478 {
6479 if (offsetcount >= 2)
6480 {
6481 offsets[0] = erroroffset;
6482 offsets[1] = errorcode;
6483 }
6484 #if defined COMPILE_PCRE8
6485 return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
6486 PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
6487 #elif defined COMPILE_PCRE16
6488 return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)?
6489 PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16;
6490 #elif defined COMPILE_PCRE32
6491 return PCRE_ERROR_BADUTF32;
6492 #endif
6493 }
6494 #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16
6495 /* Check that a start_offset points to the start of a UTF character. */
6496 if (start_offset > 0 && start_offset < length &&
6497 NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset]))
6498 return PCRE_ERROR_BADUTF8_OFFSET;
6499 #endif
6500 }
6501 #endif
6502
6503 /* If the pattern was successfully studied with JIT support, run the JIT
6504 executable instead of the rest of this function. Most options must be set at
6505 compile time for the JIT code to be usable. Fallback to the normal code path if
6506 an unsupported flag is set. */
6507
6508 #ifdef SUPPORT_JIT
6509 if (extra_data != NULL
6510 && (extra_data->flags & (PCRE_EXTRA_EXECUTABLE_JIT |
6511 PCRE_EXTRA_TABLES)) == PCRE_EXTRA_EXECUTABLE_JIT
6512 && extra_data->executable_jit != NULL
6513 && (options & ~PUBLIC_JIT_EXEC_OPTIONS) == 0)
6514 {
6515 rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length,
6516 start_offset, options, offsets, offsetcount);
6517
6518 /* PCRE_ERROR_NULL means that the selected normal or partial matching
6519 mode is not compiled. In this case we simply fallback to interpreter. */
6520
6521 if (rc != PCRE_ERROR_JIT_BADOPTION) return rc;
6522 }
6523 #endif
6524
6525 /* Carry on with non-JIT matching. This information is for finding all the
6526 numbers associated with a given name, for condition testing. */
6527
6528 md->name_table = (pcre_uchar *)re + re->name_table_offset;
6529 md->name_count = re->name_count;
6530 md->name_entry_size = re->name_entry_size;
6531
6532 /* Fish out the optional data from the extra_data structure, first setting
6533 the default values. */
6534
6535 study = NULL;
6536 md->match_limit = MATCH_LIMIT;
6537 md->match_limit_recursion = MATCH_LIMIT_RECURSION;
6538 md->callout_data = NULL;
6539
6540 /* The table pointer is always in native byte order. */
6541
6542 tables = re->tables;
6543
6544 if (extra_data != NULL)
6545 {
6546 register unsigned int flags = extra_data->flags;
6547 if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
6548 study = (const pcre_study_data *)extra_data->study_data;
6549 if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
6550 md->match_limit = extra_data->match_limit;
6551 if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
6552 md->match_limit_recursion = extra_data->match_limit_recursion;
6553 if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
6554 md->callout_data = extra_data->callout_data;
6555 if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
6556 }
6557
6558 /* If the exec call supplied NULL for tables, use the inbuilt ones. This
6559 is a feature that makes it possible to save compiled regex and re-use them
6560 in other programs later. */
6561
6562 if (tables == NULL) tables = PRIV(default_tables);
6563
6564 /* Set up other data */
6565
6566 anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
6567 startline = (re->flags & PCRE_STARTLINE) != 0;
6568 firstline = (re->options & PCRE_FIRSTLINE) != 0;
6569
6570 /* The code starts after the real_pcre block and the capture name table. */
6571
6572 md->start_code = (const pcre_uchar *)re + re->name_table_offset +
6573 re->name_count * re->name_entry_size;
6574
6575 md->start_subject = (PCRE_PUCHAR)subject;
6576 md->start_offset = start_offset;
6577 md->end_subject = md->start_subject + length;
6578 end_subject = md->end_subject;
6579
6580 md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
6581 md->use_ucp = (re->options & PCRE_UCP) != 0;
6582 md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6583 md->ignore_skip_arg = 0;
6584
6585 /* Some options are unpacked into BOOL variables in the hope that testing
6586 them will be faster than individual option bits. */
6587
6588 md->notbol = (options & PCRE_NOTBOL) != 0;
6589 md->noteol = (options & PCRE_NOTEOL) != 0;
6590 md->notempty = (options & PCRE_NOTEMPTY) != 0;
6591 md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6592
6593 md->hitend = FALSE;
6594 md->mark = md->nomatch_mark = NULL; /* In case never set */
6595
6596 md->recursive = NULL; /* No recursion at top level */
6597 md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6598
6599 md->lcc = tables + lcc_offset;
6600 md->fcc = tables + fcc_offset;
6601 md->ctypes = tables + ctypes_offset;
6602
6603 /* Handle different \R options. */
6604
6605 switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6606 {
6607 case 0:
6608 if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
6609 md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
6610 else
6611 #ifdef BSR_ANYCRLF
6612 md->bsr_anycrlf = TRUE;
6613 #else
6614 md->bsr_anycrlf = FALSE;
6615 #endif
6616 break;
6617
6618 case PCRE_BSR_ANYCRLF:
6619 md->bsr_anycrlf = TRUE;
6620 break;
6621
6622 case PCRE_BSR_UNICODE:
6623 md->bsr_anycrlf = FALSE;
6624 break;
6625
6626 default: return PCRE_ERROR_BADNEWLINE;
6627 }
6628
6629 /* Handle different types of newline. The three bits give eight cases. If
6630 nothing is set at run time, whatever was used at compile time applies. */
6631
6632 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
6633 (pcre_uint32)options) & PCRE_NEWLINE_BITS)
6634 {
6635 case 0: newline = NEWLINE; break; /* Compile-time default */
6636 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6637 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6638 case PCRE_NEWLINE_CR+
6639 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6640 case PCRE_NEWLINE_ANY: newline = -1; break;
6641 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6642 default: return PCRE_ERROR_BADNEWLINE;
6643 }
6644
6645 if (newline == -2)
6646 {
6647 md->nltype = NLTYPE_ANYCRLF;
6648 }
6649 else if (newline < 0)
6650 {
6651 md->nltype = NLTYPE_ANY;
6652 }
6653 else
6654 {
6655 md->nltype = NLTYPE_FIXED;
6656 if (newline > 255)
6657 {
6658 md->nllen = 2;
6659 md->nl[0] = (newline >> 8) & 255;
6660 md->nl[1] = newline & 255;
6661 }
6662 else
6663 {
6664 md->nllen = 1;
6665 md->nl[0] = newline;
6666 }
6667 }
6668
6669 /* Partial matching was originally supported only for a restricted set of
6670 regexes; from release 8.00 there are no restrictions, but the bits are still
6671 defined (though never set). So there's no harm in leaving this code. */
6672
6673 if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6674 return PCRE_ERROR_BADPARTIAL;
6675
6676 /* If the expression has got more back references than the offsets supplied can
6677 hold, we get a temporary chunk of working store to use during the matching.
6678 Otherwise, we can use the vector supplied, rounding down its size to a multiple
6679 of 3. */
6680
6681 ocount = offsetcount - (offsetcount % 3);
6682 arg_offset_max = (2*ocount)/3;
6683
6684 if (re->top_backref > 0 && re->top_backref >= ocount/3)
6685 {
6686 ocount = re->top_backref * 3 + 3;
6687 md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int));
6688 if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
6689 using_temporary_offsets = TRUE;
6690 DPRINTF(("Got memory to hold back references\n"));
6691 }
6692 else md->offset_vector = offsets;
6693 md->offset_end = ocount;
6694 md->offset_max = (2*ocount)/3;
6695 md->capture_last = 0;
6696
6697 /* Reset the working variable associated with each extraction. These should
6698 never be used unless previously set, but they get saved and restored, and so we
6699 initialize them to avoid reading uninitialized locations. Also, unset the
6700 offsets for the matched string. This is really just for tidiness with callouts,
6701 in case they inspect these fields. */
6702
6703 if (md->offset_vector != NULL)
6704 {
6705 register int *iptr = md->offset_vector + ocount;
6706 register int *iend = iptr - re->top_bracket;
6707 if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6708 while (--iptr >= iend) *iptr = -1;
6709 md->offset_vector[0] = md->offset_vector[1] = -1;
6710 }
6711
6712 /* Set up the first character to match, if available. The first_char value is
6713 never set for an anchored regular expression, but the anchoring may be forced
6714 at run time, so we have to test for anchoring. The first char may be unset for
6715 an unanchored pattern, of course. If there's no first char and the pattern was
6716 studied, there may be a bitmap of possible first characters. */
6717
6718 if (!anchored)
6719 {
6720 if ((re->flags & PCRE_FIRSTSET) != 0)
6721 {
6722 has_first_char = TRUE;
6723 first_char = first_char2 = (pcre_uchar)(re->first_char);
6724 if ((re->flags & PCRE_FCH_CASELESS) != 0)
6725 {
6726 first_char2 = TABLE_GET(first_char, md->fcc, first_char);
6727 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
6728 if (utf && first_char > 127)
6729 first_char2 = UCD_OTHERCASE(first_char);
6730 #endif
6731 }
6732 }
6733 else
6734 if (!startline && study != NULL &&
6735 (study->flags & PCRE_STUDY_MAPPED) != 0)
6736 start_bits = study->start_bits;
6737 }
6738
6739 /* For anchored or unanchored matches, there may be a "last known required
6740 character" set. */
6741
6742 if ((re->flags & PCRE_REQCHSET) != 0)
6743 {
6744 has_req_char = TRUE;
6745 req_char = req_char2 = (pcre_uchar)(re->req_char);
6746 if ((re->flags & PCRE_RCH_CASELESS) != 0)
6747 {
6748 req_char2 = TABLE_GET(req_char, md->fcc, req_char);
6749 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
6750 if (utf && req_char > 127)
6751 req_char2 = UCD_OTHERCASE(req_char);
6752 #endif
6753 }
6754 }
6755
6756