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

Contents of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1350 - (show annotations)
Fri Jul 26 10:03:38 2013 UTC (6 years, 2 months ago) by ph10
File MIME type: text/plain
File size: 219413 byte(s)
Error occurred while calculating annotation data.
Fix backup bugs with \X repeat matches.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2013 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40 /* This module contains pcre_exec(), the externally visible function that does
41 pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42 possible. There are also some static supporting functions. */
43
44 #ifdef HAVE_CONFIG_H
45 #include "config.h"
46 #endif
47
48 #define NLBLOCK md /* Block containing newline information */
49 #define PSSTART start_subject /* Field containing processed string start */
50 #define PSEND end_subject /* Field containing processed string end */
51
52 #include "pcre_internal.h"
53
54 /* Undefine some potentially clashing cpp symbols */
55
56 #undef min
57 #undef max
58
59 /* The md->capture_last field uses the lower 16 bits for the last captured
60 substring (which can never be greater than 65535) and a bit in the top half
61 to mean "capture vector overflowed". This odd way of doing things was
62 implemented when it was realized that preserving and restoring the overflow bit
63 whenever the last capture number was saved/restored made for a neater
64 interface, and doing it this way saved on (a) another variable, which would
65 have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66 separate set of save/restore instructions. The following defines are used in
67 implementing this. */
68
69 #define CAPLMASK 0x0000ffff /* The bits used for last_capture */
70 #define OVFLMASK 0xffff0000 /* The bits used for the overflow flag */
71 #define OVFLBIT 0x00010000 /* The bit that is set for overflow */
72
73 /* Values for setting in md->match_function_type to indicate two special types
74 of call to match(). We do it this way to save on using another stack variable,
75 as stack usage is to be discouraged. */
76
77 #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */
78 #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */
79
80 /* Non-error returns from the match() function. Error returns are externally
81 defined PCRE_ERROR_xxx codes, which are all negative. */
82
83 #define MATCH_MATCH 1
84 #define MATCH_NOMATCH 0
85
86 /* Special internal returns from the match() function. Make them sufficiently
87 negative to avoid the external error codes. */
88
89 #define MATCH_ACCEPT (-999)
90 #define MATCH_KETRPOS (-998)
91 #define MATCH_ONCE (-997)
92 /* The next 5 must be kept together and in sequence so that a test that checks
93 for any one of them can use a range. */
94 #define MATCH_COMMIT (-996)
95 #define MATCH_PRUNE (-995)
96 #define MATCH_SKIP (-994)
97 #define MATCH_SKIP_ARG (-993)
98 #define MATCH_THEN (-992)
99 #define MATCH_BACKTRACK_MAX MATCH_THEN
100 #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101
102 /* Maximum number of ints of offset to save on the stack for recursive calls.
103 If the offset vector is bigger, malloc is used. This should be a multiple of 3,
104 because the offset vector is always a multiple of 3 long. */
105
106 #define REC_STACK_SAVE_MAX 30
107
108 /* Min and max values for the common repeats; for the maxima, 0 => infinity */
109
110 static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
111 static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
112
113 #ifdef PCRE_DEBUG
114 /*************************************************
115 * Debugging function to print chars *
116 *************************************************/
117
118 /* Print a sequence of chars in printable format, stopping at the end of the
119 subject if the requested.
120
121 Arguments:
122 p points to characters
123 length number to print
124 is_subject TRUE if printing from within md->start_subject
125 md pointer to matching data block, if is_subject is TRUE
126
127 Returns: nothing
128 */
129
130 static void
131 pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132 {
133 pcre_uint32 c;
134 BOOL utf = md->utf;
135 if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136 while (length-- > 0)
137 if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138 }
139 #endif
140
141
142
143 /*************************************************
144 * Match a back-reference *
145 *************************************************/
146
147 /* Normally, if a back reference hasn't been set, the length that is passed is
148 negative, so the match always fails. However, in JavaScript compatibility mode,
149 the length passed is zero. Note that in caseless UTF-8 mode, the number of
150 subject bytes matched may be different to the number of reference bytes.
151
152 Arguments:
153 offset index into the offset vector
154 eptr pointer into the subject
155 length length of reference to be matched (number of bytes)
156 md points to match data block
157 caseless TRUE if caseless
158
159 Returns: >= 0 the number of subject bytes matched
160 -1 no match
161 -2 partial match; always given if at end subject
162 */
163
164 static int
165 match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166 BOOL caseless)
167 {
168 PCRE_PUCHAR eptr_start = eptr;
169 register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170 #ifdef SUPPORT_UTF
171 BOOL utf = md->utf;
172 #endif
173
174 #ifdef PCRE_DEBUG
175 if (eptr >= md->end_subject)
176 printf("matching subject <null>");
177 else
178 {
179 printf("matching subject ");
180 pchars(eptr, length, TRUE, md);
181 }
182 printf(" against backref ");
183 pchars(p, length, FALSE, md);
184 printf("\n");
185 #endif
186
187 /* Always fail if reference not set (and not JavaScript compatible - in that
188 case the length is passed as zero). */
189
190 if (length < 0) return -1;
191
192 /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193 properly if Unicode properties are supported. Otherwise, we can check only
194 ASCII characters. */
195
196 if (caseless)
197 {
198 #ifdef SUPPORT_UTF
199 #ifdef SUPPORT_UCP
200 if (utf)
201 {
202 /* Match characters up to the end of the reference. NOTE: the number of
203 data units matched may differ, because in UTF-8 there are some characters
204 whose upper and lower case versions code have different numbers of bytes.
205 For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206 (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207 sequence of two of the latter. It is important, therefore, to check the
208 length along the reference, not along the subject (earlier code did this
209 wrong). */
210
211 PCRE_PUCHAR endptr = p + length;
212 while (p < endptr)
213 {
214 pcre_uint32 c, d;
215 const ucd_record *ur;
216 if (eptr >= md->end_subject) return -2; /* Partial match */
217 GETCHARINC(c, eptr);
218 GETCHARINC(d, p);
219 ur = GET_UCD(d);
220 if (c != d && c != d + ur->other_case)
221 {
222 const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223 for (;;)
224 {
225 if (c < *pp) return -1;
226 if (c == *pp++) break;
227 }
228 }
229 }
230 }
231 else
232 #endif
233 #endif
234
235 /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236 is no UCP support. */
237 {
238 while (length-- > 0)
239 {
240 pcre_uint32 cc, cp;
241 if (eptr >= md->end_subject) return -2; /* Partial match */
242 cc = RAWUCHARTEST(eptr);
243 cp = RAWUCHARTEST(p);
244 if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245 p++;
246 eptr++;
247 }
248 }
249 }
250
251 /* In the caseful case, we can just compare the bytes, whether or not we
252 are in UTF-8 mode. */
253
254 else
255 {
256 while (length-- > 0)
257 {
258 if (eptr >= md->end_subject) return -2; /* Partial match */
259 if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260 }
261 }
262
263 return (int)(eptr - eptr_start);
264 }
265
266
267
268 /***************************************************************************
269 ****************************************************************************
270 RECURSION IN THE match() FUNCTION
271
272 The match() function is highly recursive, though not every recursive call
273 increases the recursive depth. Nevertheless, some regular expressions can cause
274 it to recurse to a great depth. I was writing for Unix, so I just let it call
275 itself recursively. This uses the stack for saving everything that has to be
276 saved for a recursive call. On Unix, the stack can be large, and this works
277 fine.
278
279 It turns out that on some non-Unix-like systems there are problems with
280 programs that use a lot of stack. (This despite the fact that every last chip
281 has oodles of memory these days, and techniques for extending the stack have
282 been known for decades.) So....
283
284 There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
285 calls by keeping local variables that need to be preserved in blocks of memory
286 obtained from malloc() instead instead of on the stack. Macros are used to
287 achieve this so that the actual code doesn't look very different to what it
288 always used to.
289
290 The original heap-recursive code used longjmp(). However, it seems that this
291 can be very slow on some operating systems. Following a suggestion from Stan
292 Switzer, the use of longjmp() has been abolished, at the cost of having to
293 provide a unique number for each call to RMATCH. There is no way of generating
294 a sequence of numbers at compile time in C. I have given them names, to make
295 them stand out more clearly.
296
297 Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
298 FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
299 tests. Furthermore, not using longjmp() means that local dynamic variables
300 don't have indeterminate values; this has meant that the frame size can be
301 reduced because the result can be "passed back" by straight setting of the
302 variable instead of being passed in the frame.
303 ****************************************************************************
304 ***************************************************************************/
305
306 /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307 below must be updated in sync. */
308
309 enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10,
310 RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311 RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312 RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313 RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314 RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315 RM61, RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316
317 /* These versions of the macros use the stack, as normal. There are debugging
318 versions and production versions. Note that the "rw" argument of RMATCH isn't
319 actually used in this definition. */
320
321 #ifndef NO_RECURSE
322 #define REGISTER register
323
324 #ifdef PCRE_DEBUG
325 #define RMATCH(ra,rb,rc,rd,re,rw) \
326 { \
327 printf("match() called in line %d\n", __LINE__); \
328 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329 printf("to line %d\n", __LINE__); \
330 }
331 #define RRETURN(ra) \
332 { \
333 printf("match() returned %d from line %d\n", ra, __LINE__); \
334 return ra; \
335 }
336 #else
337 #define RMATCH(ra,rb,rc,rd,re,rw) \
338 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339 #define RRETURN(ra) return ra
340 #endif
341
342 #else
343
344
345 /* These versions of the macros manage a private stack on the heap. Note that
346 the "rd" argument of RMATCH isn't actually used in this definition. It's the md
347 argument of match(), which never changes. */
348
349 #define REGISTER
350
351 #define RMATCH(ra,rb,rc,rd,re,rw)\
352 {\
353 heapframe *newframe = frame->Xnextframe;\
354 if (newframe == NULL)\
355 {\
356 newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357 if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358 newframe->Xnextframe = NULL;\
359 frame->Xnextframe = newframe;\
360 }\
361 frame->Xwhere = rw;\
362 newframe->Xeptr = ra;\
363 newframe->Xecode = rb;\
364 newframe->Xmstart = mstart;\
365 newframe->Xoffset_top = rc;\
366 newframe->Xeptrb = re;\
367 newframe->Xrdepth = frame->Xrdepth + 1;\
368 newframe->Xprevframe = frame;\
369 frame = newframe;\
370 DPRINTF(("restarting from line %d\n", __LINE__));\
371 goto HEAP_RECURSE;\
372 L_##rw:\
373 DPRINTF(("jumped back to line %d\n", __LINE__));\
374 }
375
376 #define RRETURN(ra)\
377 {\
378 heapframe *oldframe = frame;\
379 frame = oldframe->Xprevframe;\
380 if (frame != NULL)\
381 {\
382 rrc = ra;\
383 goto HEAP_RETURN;\
384 }\
385 return ra;\
386 }
387
388
389 /* Structure for remembering the local variables in a private frame */
390
391 typedef struct heapframe {
392 struct heapframe *Xprevframe;
393 struct heapframe *Xnextframe;
394
395 /* Function arguments that may change */
396
397 PCRE_PUCHAR Xeptr;
398 const pcre_uchar *Xecode;
399 PCRE_PUCHAR Xmstart;
400 int Xoffset_top;
401 eptrblock *Xeptrb;
402 unsigned int Xrdepth;
403
404 /* Function local variables */
405
406 PCRE_PUCHAR Xcallpat;
407 #ifdef SUPPORT_UTF
408 PCRE_PUCHAR Xcharptr;
409 #endif
410 PCRE_PUCHAR Xdata;
411 PCRE_PUCHAR Xnext;
412 PCRE_PUCHAR Xpp;
413 PCRE_PUCHAR Xprev;
414 PCRE_PUCHAR Xsaved_eptr;
415
416 recursion_info Xnew_recursive;
417
418 BOOL Xcur_is_word;
419 BOOL Xcondition;
420 BOOL Xprev_is_word;
421
422 #ifdef SUPPORT_UCP
423 int Xprop_type;
424 unsigned int Xprop_value;
425 int Xprop_fail_result;
426 int Xoclength;
427 pcre_uchar Xocchars[6];
428 #endif
429
430 int Xcodelink;
431 int Xctype;
432 unsigned int Xfc;
433 int Xfi;
434 int Xlength;
435 int Xmax;
436 int Xmin;
437 unsigned int Xnumber;
438 int Xoffset;
439 unsigned int Xop;
440 pcre_int32 Xsave_capture_last;
441 int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442 int Xstacksave[REC_STACK_SAVE_MAX];
443
444 eptrblock Xnewptrb;
445
446 /* Where to jump back to */
447
448 int Xwhere;
449
450 } heapframe;
451
452 #endif
453
454
455 /***************************************************************************
456 ***************************************************************************/
457
458
459
460 /*************************************************
461 * Match from current position *
462 *************************************************/
463
464 /* This function is called recursively in many circumstances. Whenever it
465 returns a negative (error) response, the outer incarnation must also return the
466 same response. */
467
468 /* These macros pack up tests that are used for partial matching, and which
469 appear several times in the code. We set the "hit end" flag if the pointer is
470 at the end of the subject and also past the start of the subject (i.e.
471 something has been matched). For hard partial matching, we then return
472 immediately. The second one is used when we already know we are past the end of
473 the subject. */
474
475 #define CHECK_PARTIAL()\
476 if (md->partial != 0 && eptr >= md->end_subject && \
477 eptr > md->start_used_ptr) \
478 { \
479 md->hitend = TRUE; \
480 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481 }
482
483 #define SCHECK_PARTIAL()\
484 if (md->partial != 0 && eptr > md->start_used_ptr) \
485 { \
486 md->hitend = TRUE; \
487 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488 }
489
490
491 /* Performance note: It might be tempting to extract commonly used fields from
492 the md structure (e.g. utf, end_subject) into individual variables to improve
493 performance. Tests using gcc on a SPARC disproved this; in the first case, it
494 made performance worse.
495
496 Arguments:
497 eptr pointer to current character in subject
498 ecode pointer to current position in compiled code
499 mstart pointer to the current match start position (can be modified
500 by encountering \K)
501 offset_top current top pointer
502 md pointer to "static" info for the match
503 eptrb pointer to chain of blocks containing eptr at start of
504 brackets - for testing for empty matches
505 rdepth the recursion depth
506
507 Returns: MATCH_MATCH if matched ) these values are >= 0
508 MATCH_NOMATCH if failed to match )
509 a negative MATCH_xxx value for PRUNE, SKIP, etc
510 a negative PCRE_ERROR_xxx value if aborted by an error condition
511 (e.g. stopped by repeated call or recursion limit)
512 */
513
514 static int
515 match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516 PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517 unsigned int rdepth)
518 {
519 /* These variables do not need to be preserved over recursion in this function,
520 so they can be ordinary variables in all cases. Mark some of them with
521 "register" because they are used a lot in loops. */
522
523 register int rrc; /* Returns from recursive calls */
524 register int i; /* Used for loops not involving calls to RMATCH() */
525 register pcre_uint32 c; /* Character values not kept over RMATCH() calls */
526 register BOOL utf; /* Local copy of UTF flag for speed */
527
528 BOOL minimize, possessive; /* Quantifier options */
529 BOOL caseless;
530 int condcode;
531
532 /* When recursion is not being used, all "local" variables that have to be
533 preserved over calls to RMATCH() are part of a "frame". We set up the top-level
534 frame on the stack here; subsequent instantiations are obtained from the heap
535 whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536 the top-level on the stack rather than malloc-ing them all gives a performance
537 boost in many cases where there is not much "recursion". */
538
539 #ifdef NO_RECURSE
540 heapframe *frame = (heapframe *)md->match_frames_base;
541
542 /* Copy in the original argument variables */
543
544 frame->Xeptr = eptr;
545 frame->Xecode = ecode;
546 frame->Xmstart = mstart;
547 frame->Xoffset_top = offset_top;
548 frame->Xeptrb = eptrb;
549 frame->Xrdepth = rdepth;
550
551 /* This is where control jumps back to to effect "recursion" */
552
553 HEAP_RECURSE:
554
555 /* Macros make the argument variables come from the current frame */
556
557 #define eptr frame->Xeptr
558 #define ecode frame->Xecode
559 #define mstart frame->Xmstart
560 #define offset_top frame->Xoffset_top
561 #define eptrb frame->Xeptrb
562 #define rdepth frame->Xrdepth
563
564 /* Ditto for the local variables */
565
566 #ifdef SUPPORT_UTF
567 #define charptr frame->Xcharptr
568 #endif
569 #define callpat frame->Xcallpat
570 #define codelink frame->Xcodelink
571 #define data frame->Xdata
572 #define next frame->Xnext
573 #define pp frame->Xpp
574 #define prev frame->Xprev
575 #define saved_eptr frame->Xsaved_eptr
576
577 #define new_recursive frame->Xnew_recursive
578
579 #define cur_is_word frame->Xcur_is_word
580 #define condition frame->Xcondition
581 #define prev_is_word frame->Xprev_is_word
582
583 #ifdef SUPPORT_UCP
584 #define prop_type frame->Xprop_type
585 #define prop_value frame->Xprop_value
586 #define prop_fail_result frame->Xprop_fail_result
587 #define oclength frame->Xoclength
588 #define occhars frame->Xocchars
589 #endif
590
591 #define ctype frame->Xctype
592 #define fc frame->Xfc
593 #define fi frame->Xfi
594 #define length frame->Xlength
595 #define max frame->Xmax
596 #define min frame->Xmin
597 #define number frame->Xnumber
598 #define offset frame->Xoffset
599 #define op frame->Xop
600 #define save_capture_last frame->Xsave_capture_last
601 #define save_offset1 frame->Xsave_offset1
602 #define save_offset2 frame->Xsave_offset2
603 #define save_offset3 frame->Xsave_offset3
604 #define stacksave frame->Xstacksave
605
606 #define newptrb frame->Xnewptrb
607
608 /* When recursion is being used, local variables are allocated on the stack and
609 get preserved during recursion in the normal way. In this environment, fi and
610 i, and fc and c, can be the same variables. */
611
612 #else /* NO_RECURSE not defined */
613 #define fi i
614 #define fc c
615
616 /* Many of the following variables are used only in small blocks of the code.
617 My normal style of coding would have declared them within each of those blocks.
618 However, in order to accommodate the version of this code that uses an external
619 "stack" implemented on the heap, it is easier to declare them all here, so the
620 declarations can be cut out in a block. The only declarations within blocks
621 below are for variables that do not have to be preserved over a recursive call
622 to RMATCH(). */
623
624 #ifdef SUPPORT_UTF
625 const pcre_uchar *charptr;
626 #endif
627 const pcre_uchar *callpat;
628 const pcre_uchar *data;
629 const pcre_uchar *next;
630 PCRE_PUCHAR pp;
631 const pcre_uchar *prev;
632 PCRE_PUCHAR saved_eptr;
633
634 recursion_info new_recursive;
635
636 BOOL cur_is_word;
637 BOOL condition;
638 BOOL prev_is_word;
639
640 #ifdef SUPPORT_UCP
641 int prop_type;
642 unsigned int prop_value;
643 int prop_fail_result;
644 int oclength;
645 pcre_uchar occhars[6];
646 #endif
647
648 int codelink;
649 int ctype;
650 int length;
651 int max;
652 int min;
653 unsigned int number;
654 int offset;
655 unsigned int op;
656 pcre_int32 save_capture_last;
657 int save_offset1, save_offset2, save_offset3;
658 int stacksave[REC_STACK_SAVE_MAX];
659
660 eptrblock newptrb;
661
662 /* There is a special fudge for calling match() in a way that causes it to
663 measure the size of its basic stack frame when the stack is being used for
664 recursion. The second argument (ecode) being NULL triggers this behaviour. It
665 cannot normally ever be NULL. The return is the negated value of the frame
666 size. */
667
668 if (ecode == NULL)
669 {
670 if (rdepth == 0)
671 return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672 else
673 {
674 int len = (char *)&rdepth - (char *)eptr;
675 return (len > 0)? -len : len;
676 }
677 }
678 #endif /* NO_RECURSE */
679
680 /* To save space on the stack and in the heap frame, I have doubled up on some
681 of the local variables that are used only in localised parts of the code, but
682 still need to be preserved over recursive calls of match(). These macros define
683 the alternative names that are used. */
684
685 #define allow_zero cur_is_word
686 #define cbegroup condition
687 #define code_offset codelink
688 #define condassert condition
689 #define matched_once prev_is_word
690 #define foc number
691 #define save_mark data
692
693 /* These statements are here to stop the compiler complaining about unitialized
694 variables. */
695
696 #ifdef SUPPORT_UCP
697 prop_value = 0;
698 prop_fail_result = 0;
699 #endif
700
701
702 /* This label is used for tail recursion, which is used in a few cases even
703 when NO_RECURSE is not defined, in order to reduce the amount of stack that is
704 used. Thanks to Ian Taylor for noticing this possibility and sending the
705 original patch. */
706
707 TAIL_RECURSE:
708
709 /* OK, now we can get on with the real code of the function. Recursive calls
710 are specified by the macro RMATCH and RRETURN is used to return. When
711 NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712 and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713 defined). However, RMATCH isn't like a function call because it's quite a
714 complicated macro. It has to be used in one particular way. This shouldn't,
715 however, impact performance when true recursion is being used. */
716
717 #ifdef SUPPORT_UTF
718 utf = md->utf; /* Local copy of the flag */
719 #else
720 utf = FALSE;
721 #endif
722
723 /* First check that we haven't called match() too many times, or that we
724 haven't exceeded the recursive call limit. */
725
726 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728
729 /* At the start of a group with an unlimited repeat that may match an empty
730 string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731 done this way to save having to use another function argument, which would take
732 up space on the stack. See also MATCH_CONDASSERT below.
733
734 When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735 such remembered pointers, to be checked when we hit the closing ket, in order
736 to break infinite loops that match no characters. When match() is called in
737 other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738 NOT be used with tail recursion, because the memory block that is used is on
739 the stack, so a new one may be required for each match(). */
740
741 if (md->match_function_type == MATCH_CBEGROUP)
742 {
743 newptrb.epb_saved_eptr = eptr;
744 newptrb.epb_prev = eptrb;
745 eptrb = &newptrb;
746 md->match_function_type = 0;
747 }
748
749 /* Now start processing the opcodes. */
750
751 for (;;)
752 {
753 minimize = possessive = FALSE;
754 op = *ecode;
755
756 switch(op)
757 {
758 case OP_MARK:
759 md->nomatch_mark = ecode + 2;
760 md->mark = NULL; /* In case previously set by assertion */
761 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762 eptrb, RM55);
763 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764 md->mark == NULL) md->mark = ecode + 2;
765
766 /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767 argument, and we must check whether that argument matches this MARK's
768 argument. It is passed back in md->start_match_ptr (an overloading of that
769 variable). If it does match, we reset that variable to the current subject
770 position and return MATCH_SKIP. Otherwise, pass back the return code
771 unaltered. */
772
773 else if (rrc == MATCH_SKIP_ARG &&
774 STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775 {
776 md->start_match_ptr = eptr;
777 RRETURN(MATCH_SKIP);
778 }
779 RRETURN(rrc);
780
781 case OP_FAIL:
782 RRETURN(MATCH_NOMATCH);
783
784 case OP_COMMIT:
785 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786 eptrb, RM52);
787 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788 RRETURN(MATCH_COMMIT);
789
790 case OP_PRUNE:
791 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792 eptrb, RM51);
793 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794 RRETURN(MATCH_PRUNE);
795
796 case OP_PRUNE_ARG:
797 md->nomatch_mark = ecode + 2;
798 md->mark = NULL; /* In case previously set by assertion */
799 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800 eptrb, RM56);
801 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
802 md->mark == NULL) md->mark = ecode + 2;
803 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
804 RRETURN(MATCH_PRUNE);
805
806 case OP_SKIP:
807 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
808 eptrb, RM53);
809 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
810 md->start_match_ptr = eptr; /* Pass back current position */
811 RRETURN(MATCH_SKIP);
812
813 /* Note that, for Perl compatibility, SKIP with an argument does NOT set
814 nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
815 not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
816 that failed and any that precede it (either they also failed, or were not
817 triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
818 SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
819 set to the count of the one that failed. */
820
821 case OP_SKIP_ARG:
822 md->skip_arg_count++;
823 if (md->skip_arg_count <= md->ignore_skip_arg)
824 {
825 ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
826 break;
827 }
828 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
829 eptrb, RM57);
830 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831
832 /* Pass back the current skip name by overloading md->start_match_ptr and
833 returning the special MATCH_SKIP_ARG return code. This will either be
834 caught by a matching MARK, or get to the top, where it causes a rematch
835 with md->ignore_skip_arg set to the value of md->skip_arg_count. */
836
837 md->start_match_ptr = ecode + 2;
838 RRETURN(MATCH_SKIP_ARG);
839
840 /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
841 the branch in which it occurs can be determined. Overload the start of
842 match pointer to do this. */
843
844 case OP_THEN:
845 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
846 eptrb, RM54);
847 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
848 md->start_match_ptr = ecode;
849 RRETURN(MATCH_THEN);
850
851 case OP_THEN_ARG:
852 md->nomatch_mark = ecode + 2;
853 md->mark = NULL; /* In case previously set by assertion */
854 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
855 md, eptrb, RM58);
856 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
857 md->mark == NULL) md->mark = ecode + 2;
858 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859 md->start_match_ptr = ecode;
860 RRETURN(MATCH_THEN);
861
862 /* Handle an atomic group that does not contain any capturing parentheses.
863 This can be handled like an assertion. Prior to 8.13, all atomic groups
864 were handled this way. In 8.13, the code was changed as below for ONCE, so
865 that backups pass through the group and thereby reset captured values.
866 However, this uses a lot more stack, so in 8.20, atomic groups that do not
867 contain any captures generate OP_ONCE_NC, which can be handled in the old,
868 less stack intensive way.
869
870 Check the alternative branches in turn - the matching won't pass the KET
871 for this kind of subpattern. If any one branch matches, we carry on as at
872 the end of a normal bracket, leaving the subject pointer, but resetting
873 the start-of-match value in case it was changed by \K. */
874
875 case OP_ONCE_NC:
876 prev = ecode;
877 saved_eptr = eptr;
878 save_mark = md->mark;
879 do
880 {
881 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
882 if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */
883 {
884 mstart = md->start_match_ptr;
885 break;
886 }
887 if (rrc == MATCH_THEN)
888 {
889 next = ecode + GET(ecode,1);
890 if (md->start_match_ptr < next &&
891 (*ecode == OP_ALT || *next == OP_ALT))
892 rrc = MATCH_NOMATCH;
893 }
894
895 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
896 ecode += GET(ecode,1);
897 md->mark = save_mark;
898 }
899 while (*ecode == OP_ALT);
900
901 /* If hit the end of the group (which could be repeated), fail */
902
903 if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
904
905 /* Continue as from after the group, updating the offsets high water
906 mark, since extracts may have been taken. */
907
908 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
909
910 offset_top = md->end_offset_top;
911 eptr = md->end_match_ptr;
912
913 /* For a non-repeating ket, just continue at this level. This also
914 happens for a repeating ket if no characters were matched in the group.
915 This is the forcible breaking of infinite loops as implemented in Perl
916 5.005. */
917
918 if (*ecode == OP_KET || eptr == saved_eptr)
919 {
920 ecode += 1+LINK_SIZE;
921 break;
922 }
923
924 /* The repeating kets try the rest of the pattern or restart from the
925 preceding bracket, in the appropriate order. The second "call" of match()
926 uses tail recursion, to avoid using another stack frame. */
927
928 if (*ecode == OP_KETRMIN)
929 {
930 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
931 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
932 ecode = prev;
933 goto TAIL_RECURSE;
934 }
935 else /* OP_KETRMAX */
936 {
937 RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
938 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
939 ecode += 1 + LINK_SIZE;
940 goto TAIL_RECURSE;
941 }
942 /* Control never gets here */
943
944 /* Handle a capturing bracket, other than those that are possessive with an
945 unlimited repeat. If there is space in the offset vector, save the current
946 subject position in the working slot at the top of the vector. We mustn't
947 change the current values of the data slot, because they may be set from a
948 previous iteration of this group, and be referred to by a reference inside
949 the group. A failure to match might occur after the group has succeeded,
950 if something later on doesn't match. For this reason, we need to restore
951 the working value and also the values of the final offsets, in case they
952 were set by a previous iteration of the same bracket.
953
954 If there isn't enough space in the offset vector, treat this as if it were
955 a non-capturing bracket. Don't worry about setting the flag for the error
956 case here; that is handled in the code for KET. */
957
958 case OP_CBRA:
959 case OP_SCBRA:
960 number = GET2(ecode, 1+LINK_SIZE);
961 offset = number << 1;
962
963 #ifdef PCRE_DEBUG
964 printf("start bracket %d\n", number);
965 printf("subject=");
966 pchars(eptr, 16, TRUE, md);
967 printf("\n");
968 #endif
969
970 if (offset < md->offset_max)
971 {
972 save_offset1 = md->offset_vector[offset];
973 save_offset2 = md->offset_vector[offset+1];
974 save_offset3 = md->offset_vector[md->offset_end - number];
975 save_capture_last = md->capture_last;
976 save_mark = md->mark;
977
978 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
979 md->offset_vector[md->offset_end - number] =
980 (int)(eptr - md->start_subject);
981
982 for (;;)
983 {
984 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
986 eptrb, RM1);
987 if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */
988
989 /* If we backed up to a THEN, check whether it is within the current
990 branch by comparing the address of the THEN that is passed back with
991 the end of the branch. If it is within the current branch, and the
992 branch is one of two or more alternatives (it either starts or ends
993 with OP_ALT), we have reached the limit of THEN's action, so convert
994 the return code to NOMATCH, which will cause normal backtracking to
995 happen from now on. Otherwise, THEN is passed back to an outer
996 alternative. This implements Perl's treatment of parenthesized groups,
997 where a group not containing | does not affect the current alternative,
998 that is, (X) is NOT the same as (X|(*F)). */
999
1000 if (rrc == MATCH_THEN)
1001 {
1002 next = ecode + GET(ecode,1);
1003 if (md->start_match_ptr < next &&
1004 (*ecode == OP_ALT || *next == OP_ALT))
1005 rrc = MATCH_NOMATCH;
1006 }
1007
1008 /* Anything other than NOMATCH is passed back. */
1009
1010 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1011 md->capture_last = save_capture_last;
1012 ecode += GET(ecode, 1);
1013 md->mark = save_mark;
1014 if (*ecode != OP_ALT) break;
1015 }
1016
1017 DPRINTF(("bracket %d failed\n", number));
1018 md->offset_vector[offset] = save_offset1;
1019 md->offset_vector[offset+1] = save_offset2;
1020 md->offset_vector[md->offset_end - number] = save_offset3;
1021
1022 /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1023
1024 RRETURN(rrc);
1025 }
1026
1027 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1028 as a non-capturing bracket. */
1029
1030 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1031 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1032
1033 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1034
1035 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037
1038 /* Non-capturing or atomic group, except for possessive with unlimited
1039 repeat and ONCE group with no captures. Loop for all the alternatives.
1040
1041 When we get to the final alternative within the brackets, we used to return
1042 the result of a recursive call to match() whatever happened so it was
1043 possible to reduce stack usage by turning this into a tail recursion,
1044 except in the case of a possibly empty group. However, now that there is
1045 the possiblity of (*THEN) occurring in the final alternative, this
1046 optimization is no longer always possible.
1047
1048 We can optimize if we know there are no (*THEN)s in the pattern; at present
1049 this is the best that can be done.
1050
1051 MATCH_ONCE is returned when the end of an atomic group is successfully
1052 reached, but subsequent matching fails. It passes back up the tree (causing
1053 captured values to be reset) until the original atomic group level is
1054 reached. This is tested by comparing md->once_target with the start of the
1055 group. At this point, the return is converted into MATCH_NOMATCH so that
1056 previous backup points can be taken. */
1057
1058 case OP_ONCE:
1059 case OP_BRA:
1060 case OP_SBRA:
1061 DPRINTF(("start non-capturing bracket\n"));
1062
1063 for (;;)
1064 {
1065 if (op >= OP_SBRA || op == OP_ONCE)
1066 md->match_function_type = MATCH_CBEGROUP;
1067
1068 /* If this is not a possibly empty group, and there are no (*THEN)s in
1069 the pattern, and this is the final alternative, optimize as described
1070 above. */
1071
1072 else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073 {
1074 ecode += PRIV(OP_lengths)[*ecode];
1075 goto TAIL_RECURSE;
1076 }
1077
1078 /* In all other cases, we have to make another call to match(). */
1079
1080 save_mark = md->mark;
1081 save_capture_last = md->capture_last;
1082 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1083 RM2);
1084
1085 /* See comment in the code for capturing groups above about handling
1086 THEN. */
1087
1088 if (rrc == MATCH_THEN)
1089 {
1090 next = ecode + GET(ecode,1);
1091 if (md->start_match_ptr < next &&
1092 (*ecode == OP_ALT || *next == OP_ALT))
1093 rrc = MATCH_NOMATCH;
1094 }
1095
1096 if (rrc != MATCH_NOMATCH)
1097 {
1098 if (rrc == MATCH_ONCE)
1099 {
1100 const pcre_uchar *scode = ecode;
1101 if (*scode != OP_ONCE) /* If not at start, find it */
1102 {
1103 while (*scode == OP_ALT) scode += GET(scode, 1);
1104 scode -= GET(scode, 1);
1105 }
1106 if (md->once_target == scode) rrc = MATCH_NOMATCH;
1107 }
1108 RRETURN(rrc);
1109 }
1110 ecode += GET(ecode, 1);
1111 md->mark = save_mark;
1112 if (*ecode != OP_ALT) break;
1113 md->capture_last = save_capture_last;
1114 }
1115
1116 RRETURN(MATCH_NOMATCH);
1117
1118 /* Handle possessive capturing brackets with an unlimited repeat. We come
1119 here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120 handled similarly to the normal case above. However, the matching is
1121 different. The end of these brackets will always be OP_KETRPOS, which
1122 returns MATCH_KETRPOS without going further in the pattern. By this means
1123 we can handle the group by iteration rather than recursion, thereby
1124 reducing the amount of stack needed. */
1125
1126 case OP_CBRAPOS:
1127 case OP_SCBRAPOS:
1128 allow_zero = FALSE;
1129
1130 POSSESSIVE_CAPTURE:
1131 number = GET2(ecode, 1+LINK_SIZE);
1132 offset = number << 1;
1133
1134 #ifdef PCRE_DEBUG
1135 printf("start possessive bracket %d\n", number);
1136 printf("subject=");
1137 pchars(eptr, 16, TRUE, md);
1138 printf("\n");
1139 #endif
1140
1141 if (offset < md->offset_max)
1142 {
1143 matched_once = FALSE;
1144 code_offset = (int)(ecode - md->start_code);
1145
1146 save_offset1 = md->offset_vector[offset];
1147 save_offset2 = md->offset_vector[offset+1];
1148 save_offset3 = md->offset_vector[md->offset_end - number];
1149 save_capture_last = md->capture_last;
1150
1151 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1152
1153 /* Each time round the loop, save the current subject position for use
1154 when the group matches. For MATCH_MATCH, the group has matched, so we
1155 restart it with a new subject starting position, remembering that we had
1156 at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157 usual. If we haven't matched any alternatives in any iteration, check to
1158 see if a previous iteration matched. If so, the group has matched;
1159 continue from afterwards. Otherwise it has failed; restore the previous
1160 capture values before returning NOMATCH. */
1161
1162 for (;;)
1163 {
1164 md->offset_vector[md->offset_end - number] =
1165 (int)(eptr - md->start_subject);
1166 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168 eptrb, RM63);
1169 if (rrc == MATCH_KETRPOS)
1170 {
1171 offset_top = md->end_offset_top;
1172 eptr = md->end_match_ptr;
1173 ecode = md->start_code + code_offset;
1174 save_capture_last = md->capture_last;
1175 matched_once = TRUE;
1176 continue;
1177 }
1178
1179 /* See comment in the code for capturing groups above about handling
1180 THEN. */
1181
1182 if (rrc == MATCH_THEN)
1183 {
1184 next = ecode + GET(ecode,1);
1185 if (md->start_match_ptr < next &&
1186 (*ecode == OP_ALT || *next == OP_ALT))
1187 rrc = MATCH_NOMATCH;
1188 }
1189
1190 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1191 md->capture_last = save_capture_last;
1192 ecode += GET(ecode, 1);
1193 if (*ecode != OP_ALT) break;
1194 }
1195
1196 if (!matched_once)
1197 {
1198 md->offset_vector[offset] = save_offset1;
1199 md->offset_vector[offset+1] = save_offset2;
1200 md->offset_vector[md->offset_end - number] = save_offset3;
1201 }
1202
1203 if (allow_zero || matched_once)
1204 {
1205 ecode += 1 + LINK_SIZE;
1206 break;
1207 }
1208
1209 RRETURN(MATCH_NOMATCH);
1210 }
1211
1212 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213 as a non-capturing bracket. */
1214
1215 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1217
1218 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1219
1220 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1222
1223 /* Non-capturing possessive bracket with unlimited repeat. We come here
1224 from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1225 without the capturing complication. It is written out separately for speed
1226 and cleanliness. */
1227
1228 case OP_BRAPOS:
1229 case OP_SBRAPOS:
1230 allow_zero = FALSE;
1231
1232 POSSESSIVE_NON_CAPTURE:
1233 matched_once = FALSE;
1234 code_offset = (int)(ecode - md->start_code);
1235 save_capture_last = md->capture_last;
1236
1237 for (;;)
1238 {
1239 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1241 eptrb, RM48);
1242 if (rrc == MATCH_KETRPOS)
1243 {
1244 offset_top = md->end_offset_top;
1245 eptr = md->end_match_ptr;
1246 ecode = md->start_code + code_offset;
1247 matched_once = TRUE;
1248 continue;
1249 }
1250
1251 /* See comment in the code for capturing groups above about handling
1252 THEN. */
1253
1254 if (rrc == MATCH_THEN)
1255 {
1256 next = ecode + GET(ecode,1);
1257 if (md->start_match_ptr < next &&
1258 (*ecode == OP_ALT || *next == OP_ALT))
1259 rrc = MATCH_NOMATCH;
1260 }
1261
1262 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1263 ecode += GET(ecode, 1);
1264 if (*ecode != OP_ALT) break;
1265 md->capture_last = save_capture_last;
1266 }
1267
1268 if (matched_once || allow_zero)
1269 {
1270 ecode += 1 + LINK_SIZE;
1271 break;
1272 }
1273 RRETURN(MATCH_NOMATCH);
1274
1275 /* Control never reaches here. */
1276
1277 /* Conditional group: compilation checked that there are no more than
1278 two branches. If the condition is false, skipping the first branch takes us
1279 past the end if there is only one branch, but that's OK because that is
1280 exactly what going to the ket would do. */
1281
1282 case OP_COND:
1283 case OP_SCOND:
1284 codelink = GET(ecode, 1);
1285
1286 /* Because of the way auto-callout works during compile, a callout item is
1287 inserted between OP_COND and an assertion condition. */
1288
1289 if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1290 {
1291 if (PUBL(callout) != NULL)
1292 {
1293 PUBL(callout_block) cb;
1294 cb.version = 2; /* Version 1 of the callout block */
1295 cb.callout_number = ecode[LINK_SIZE+2];
1296 cb.offset_vector = md->offset_vector;
1297 #if defined COMPILE_PCRE8
1298 cb.subject = (PCRE_SPTR)md->start_subject;
1299 #elif defined COMPILE_PCRE16
1300 cb.subject = (PCRE_SPTR16)md->start_subject;
1301 #elif defined COMPILE_PCRE32
1302 cb.subject = (PCRE_SPTR32)md->start_subject;
1303 #endif
1304 cb.subject_length = (int)(md->end_subject - md->start_subject);
1305 cb.start_match = (int)(mstart - md->start_subject);
1306 cb.current_position = (int)(eptr - md->start_subject);
1307 cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1308 cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1309 cb.capture_top = offset_top/2;
1310 cb.capture_last = md->capture_last & CAPLMASK;
1311 /* Internal change requires this for API compatibility. */
1312 if (cb.capture_last == 0) cb.capture_last = -1;
1313 cb.callout_data = md->callout_data;
1314 cb.mark = md->nomatch_mark;
1315 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1316 if (rrc < 0) RRETURN(rrc);
1317 }
1318 ecode += PRIV(OP_lengths)[OP_CALLOUT];
1319 codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1320 }
1321
1322 condcode = ecode[LINK_SIZE+1];
1323
1324 /* Now see what the actual condition is */
1325
1326 if (condcode == OP_RREF || condcode == OP_NRREF) /* Recursion test */
1327 {
1328 if (md->recursive == NULL) /* Not recursing => FALSE */
1329 {
1330 condition = FALSE;
1331 ecode += GET(ecode, 1);
1332 }
1333 else
1334 {
1335 unsigned int recno = GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/
1336 condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1337
1338 /* If the test is for recursion into a specific subpattern, and it is
1339 false, but the test was set up by name, scan the table to see if the
1340 name refers to any other numbers, and test them. The condition is true
1341 if any one is set. */
1342
1343 if (!condition && condcode == OP_NRREF)
1344 {
1345 pcre_uchar *slotA = md->name_table;
1346 for (i = 0; i < md->name_count; i++)
1347 {
1348 if (GET2(slotA, 0) == recno) break;
1349 slotA += md->name_entry_size;
1350 }
1351
1352 /* Found a name for the number - there can be only one; duplicate
1353 names for different numbers are allowed, but not vice versa. First
1354 scan down for duplicates. */
1355
1356 if (i < md->name_count)
1357 {
1358 pcre_uchar *slotB = slotA;
1359 while (slotB > md->name_table)
1360 {
1361 slotB -= md->name_entry_size;
1362 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1363 {
1364 condition = GET2(slotB, 0) == md->recursive->group_num;
1365 if (condition) break;
1366 }
1367 else break;
1368 }
1369
1370 /* Scan up for duplicates */
1371
1372 if (!condition)
1373 {
1374 slotB = slotA;
1375 for (i++; i < md->name_count; i++)
1376 {
1377 slotB += md->name_entry_size;
1378 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1379 {
1380 condition = GET2(slotB, 0) == md->recursive->group_num;
1381 if (condition) break;
1382 }
1383 else break;
1384 }
1385 }
1386 }
1387 }
1388
1389 /* Chose branch according to the condition */
1390
1391 ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1392 }
1393 }
1394
1395 else if (condcode == OP_CREF || condcode == OP_NCREF) /* Group used test */
1396 {
1397 offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */
1398 condition = offset < offset_top && md->offset_vector[offset] >= 0;
1399
1400 /* If the numbered capture is unset, but the reference was by name,
1401 scan the table to see if the name refers to any other numbers, and test
1402 them. The condition is true if any one is set. This is tediously similar
1403 to the code above, but not close enough to try to amalgamate. */
1404
1405 if (!condition && condcode == OP_NCREF)
1406 {
1407 unsigned int refno = offset >> 1;
1408 pcre_uchar *slotA = md->name_table;
1409
1410 for (i = 0; i < md->name_count; i++)
1411 {
1412 if (GET2(slotA, 0) == refno) break;
1413 slotA += md->name_entry_size;
1414 }
1415
1416 /* Found a name for the number - there can be only one; duplicate names
1417 for different numbers are allowed, but not vice versa. First scan down
1418 for duplicates. */
1419
1420 if (i < md->name_count)
1421 {
1422 pcre_uchar *slotB = slotA;
1423 while (slotB > md->name_table)
1424 {
1425 slotB -= md->name_entry_size;
1426 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1427 {
1428 offset = GET2(slotB, 0) << 1;
1429 condition = offset < offset_top &&
1430 md->offset_vector[offset] >= 0;
1431 if (condition) break;
1432 }
1433 else break;
1434 }
1435
1436 /* Scan up for duplicates */
1437
1438 if (!condition)
1439 {
1440 slotB = slotA;
1441 for (i++; i < md->name_count; i++)
1442 {
1443 slotB += md->name_entry_size;
1444 if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1445 {
1446 offset = GET2(slotB, 0) << 1;
1447 condition = offset < offset_top &&
1448 md->offset_vector[offset] >= 0;
1449 if (condition) break;
1450 }
1451 else break;
1452 }
1453 }
1454 }
1455 }
1456
1457 /* Chose branch according to the condition */
1458
1459 ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1460 }
1461
1462 else if (condcode == OP_DEF) /* DEFINE - always false */
1463 {
1464 condition = FALSE;
1465 ecode += GET(ecode, 1);
1466 }
1467
1468 /* The condition is an assertion. Call match() to evaluate it - setting
1469 md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1470 an assertion. */
1471
1472 else
1473 {
1474 md->match_function_type = MATCH_CONDASSERT;
1475 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1476 if (rrc == MATCH_MATCH)
1477 {
1478 if (md->end_offset_top > offset_top)
1479 offset_top = md->end_offset_top; /* Captures may have happened */
1480 condition = TRUE;
1481 ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1482 while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1483 }
1484
1485 /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1486 assertion; it is therefore treated as NOMATCH. */
1487
1488 else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1489 {
1490 RRETURN(rrc); /* Need braces because of following else */
1491 }
1492 else
1493 {
1494 condition = FALSE;
1495 ecode += codelink;
1496 }
1497 }
1498
1499 /* We are now at the branch that is to be obeyed. As there is only one, can
1500 use tail recursion to avoid using another stack frame, except when there is
1501 unlimited repeat of a possibly empty group. In the latter case, a recursive
1502 call to match() is always required, unless the second alternative doesn't
1503 exist, in which case we can just plough on. Note that, for compatibility
1504 with Perl, the | in a conditional group is NOT treated as creating two
1505 alternatives. If a THEN is encountered in the branch, it propagates out to
1506 the enclosing alternative (unless nested in a deeper set of alternatives,
1507 of course). */
1508
1509 if (condition || *ecode == OP_ALT)
1510 {
1511 if (op != OP_SCOND)
1512 {
1513 ecode += 1 + LINK_SIZE;
1514 goto TAIL_RECURSE;
1515 }
1516
1517 md->match_function_type = MATCH_CBEGROUP;
1518 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1519 RRETURN(rrc);
1520 }
1521
1522 /* Condition false & no alternative; continue after the group. */
1523
1524 else
1525 {
1526 ecode += 1 + LINK_SIZE;
1527 }
1528 break;
1529
1530
1531 /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1532 to close any currently open capturing brackets. */
1533
1534 case OP_CLOSE:
1535 number = GET2(ecode, 1); /* Must be less than 65536 */
1536 offset = number << 1;
1537
1538 #ifdef PCRE_DEBUG
1539 printf("end bracket %d at *ACCEPT", number);
1540 printf("\n");
1541 #endif
1542
1543 md->capture_last = (md->capture_last & OVFLMASK) | number;
1544 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1545 {
1546 md->offset_vector[offset] =
1547 md->offset_vector[md->offset_end - number];
1548 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1549 if (offset_top <= offset) offset_top = offset + 2;
1550 }
1551 ecode += 1 + IMM2_SIZE;
1552 break;
1553
1554
1555 /* End of the pattern, either real or forced. */
1556
1557 case OP_END:
1558 case OP_ACCEPT:
1559 case OP_ASSERT_ACCEPT:
1560
1561 /* If we have matched an empty string, fail if not in an assertion and not
1562 in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1563 is set and we have matched at the start of the subject. In both cases,
1564 backtracking will then try other alternatives, if any. */
1565
1566 if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1567 md->recursive == NULL &&
1568 (md->notempty ||
1569 (md->notempty_atstart &&
1570 mstart == md->start_subject + md->start_offset)))
1571 RRETURN(MATCH_NOMATCH);
1572
1573 /* Otherwise, we have a match. */
1574
1575 md->end_match_ptr = eptr; /* Record where we ended */
1576 md->end_offset_top = offset_top; /* and how many extracts were taken */
1577 md->start_match_ptr = mstart; /* and the start (\K can modify) */
1578
1579 /* For some reason, the macros don't work properly if an expression is
1580 given as the argument to RRETURN when the heap is in use. */
1581
1582 rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1583 RRETURN(rrc);
1584
1585 /* Assertion brackets. Check the alternative branches in turn - the
1586 matching won't pass the KET for an assertion. If any one branch matches,
1587 the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1588 start of each branch to move the current point backwards, so the code at
1589 this level is identical to the lookahead case. When the assertion is part
1590 of a condition, we want to return immediately afterwards. The caller of
1591 this incarnation of the match() function will have set MATCH_CONDASSERT in
1592 md->match_function type, and one of these opcodes will be the first opcode
1593 that is processed. We use a local variable that is preserved over calls to
1594 match() to remember this case. */
1595
1596 case OP_ASSERT:
1597 case OP_ASSERTBACK:
1598 save_mark = md->mark;
1599 if (md->match_function_type == MATCH_CONDASSERT)
1600 {
1601 condassert = TRUE;
1602 md->match_function_type = 0;
1603 }
1604 else condassert = FALSE;
1605
1606 /* Loop for each branch */
1607
1608 do
1609 {
1610 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1611
1612 /* A match means that the assertion is true; break out of the loop
1613 that matches its alternatives. */
1614
1615 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1616 {
1617 mstart = md->start_match_ptr; /* In case \K reset it */
1618 break;
1619 }
1620
1621 /* If not matched, restore the previous mark setting. */
1622
1623 md->mark = save_mark;
1624
1625 /* See comment in the code for capturing groups above about handling
1626 THEN. */
1627
1628 if (rrc == MATCH_THEN)
1629 {
1630 next = ecode + GET(ecode,1);
1631 if (md->start_match_ptr < next &&
1632 (*ecode == OP_ALT || *next == OP_ALT))
1633 rrc = MATCH_NOMATCH;
1634 }
1635
1636 /* Anything other than NOMATCH causes the entire assertion to fail,
1637 passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1638 uncaptured THEN, which means they take their normal effect. This
1639 consistent approach does not always have exactly the same effect as in
1640 Perl. */
1641
1642 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1643 ecode += GET(ecode, 1);
1644 }
1645 while (*ecode == OP_ALT); /* Continue for next alternative */
1646
1647 /* If we have tried all the alternative branches, the assertion has
1648 failed. If not, we broke out after a match. */
1649
1650 if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1651
1652 /* If checking an assertion for a condition, return MATCH_MATCH. */
1653
1654 if (condassert) RRETURN(MATCH_MATCH);
1655
1656 /* Continue from after a successful assertion, updating the offsets high
1657 water mark, since extracts may have been taken during the assertion. */
1658
1659 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1660 ecode += 1 + LINK_SIZE;
1661 offset_top = md->end_offset_top;
1662 continue;
1663
1664 /* Negative assertion: all branches must fail to match for the assertion to
1665 succeed. */
1666
1667 case OP_ASSERT_NOT:
1668 case OP_ASSERTBACK_NOT:
1669 save_mark = md->mark;
1670 if (md->match_function_type == MATCH_CONDASSERT)
1671 {
1672 condassert = TRUE;
1673 md->match_function_type = 0;
1674 }
1675 else condassert = FALSE;
1676
1677 /* Loop for each alternative branch. */
1678
1679 do
1680 {
1681 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1682 md->mark = save_mark; /* Always restore the mark setting */
1683
1684 switch(rrc)
1685 {
1686 case MATCH_MATCH: /* A successful match means */
1687 case MATCH_ACCEPT: /* the assertion has failed. */
1688 RRETURN(MATCH_NOMATCH);
1689
1690 case MATCH_NOMATCH: /* Carry on with next branch */
1691 break;
1692
1693 /* See comment in the code for capturing groups above about handling
1694 THEN. */
1695
1696 case MATCH_THEN:
1697 next = ecode + GET(ecode,1);
1698 if (md->start_match_ptr < next &&
1699 (*ecode == OP_ALT || *next == OP_ALT))
1700 {
1701 rrc = MATCH_NOMATCH;
1702 break;
1703 }
1704 /* Otherwise fall through. */
1705
1706 /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1707 assertion to fail to match, without considering any more alternatives.
1708 Failing to match means the assertion is true. This is a consistent
1709 approach, but does not always have the same effect as in Perl. */
1710
1711 case MATCH_COMMIT:
1712 case MATCH_SKIP:
1713 case MATCH_SKIP_ARG:
1714 case MATCH_PRUNE:
1715 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1716 goto NEG_ASSERT_TRUE; /* Break out of alternation loop */
1717
1718 /* Anything else is an error */
1719
1720 default:
1721 RRETURN(rrc);
1722 }
1723
1724 /* Continue with next branch */
1725
1726 ecode += GET(ecode,1);
1727 }
1728 while (*ecode == OP_ALT);
1729
1730 /* All branches in the assertion failed to match. */
1731
1732 NEG_ASSERT_TRUE:
1733 if (condassert) RRETURN(MATCH_MATCH); /* Condition assertion */
1734 ecode += 1 + LINK_SIZE; /* Continue with current branch */
1735 continue;
1736
1737 /* Move the subject pointer back. This occurs only at the start of
1738 each branch of a lookbehind assertion. If we are too close to the start to
1739 move back, this match function fails. When working with UTF-8 we move
1740 back a number of characters, not bytes. */
1741
1742 case OP_REVERSE:
1743 #ifdef SUPPORT_UTF
1744 if (utf)
1745 {
1746 i = GET(ecode, 1);
1747 while (i-- > 0)
1748 {
1749 eptr--;
1750 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1751 BACKCHAR(eptr);
1752 }
1753 }
1754 else
1755 #endif
1756
1757 /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1758
1759 {
1760 eptr -= GET(ecode, 1);
1761 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1762 }
1763
1764 /* Save the earliest consulted character, then skip to next op code */
1765
1766 if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1767 ecode += 1 + LINK_SIZE;
1768 break;
1769
1770 /* The callout item calls an external function, if one is provided, passing
1771 details of the match so far. This is mainly for debugging, though the
1772 function is able to force a failure. */
1773
1774 case OP_CALLOUT:
1775 if (PUBL(callout) != NULL)
1776 {
1777 PUBL(callout_block) cb;
1778 cb.version = 2; /* Version 1 of the callout block */
1779 cb.callout_number = ecode[1];
1780 cb.offset_vector = md->offset_vector;
1781 #if defined COMPILE_PCRE8
1782 cb.subject = (PCRE_SPTR)md->start_subject;
1783 #elif defined COMPILE_PCRE16
1784 cb.subject = (PCRE_SPTR16)md->start_subject;
1785 #elif defined COMPILE_PCRE32
1786 cb.subject = (PCRE_SPTR32)md->start_subject;
1787 #endif
1788 cb.subject_length = (int)(md->end_subject - md->start_subject);
1789 cb.start_match = (int)(mstart - md->start_subject);
1790 cb.current_position = (int)(eptr - md->start_subject);
1791 cb.pattern_position = GET(ecode, 2);
1792 cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1793 cb.capture_top = offset_top/2;
1794 cb.capture_last = md->capture_last & CAPLMASK;
1795 /* Internal change requires this for API compatibility. */
1796 if (cb.capture_last == 0) cb.capture_last = -1;
1797 cb.callout_data = md->callout_data;
1798 cb.mark = md->nomatch_mark;
1799 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1800 if (rrc < 0) RRETURN(rrc);
1801 }
1802 ecode += 2 + 2*LINK_SIZE;
1803 break;
1804
1805 /* Recursion either matches the current regex, or some subexpression. The
1806 offset data is the offset to the starting bracket from the start of the
1807 whole pattern. (This is so that it works from duplicated subpatterns.)
1808
1809 The state of the capturing groups is preserved over recursion, and
1810 re-instated afterwards. We don't know how many are started and not yet
1811 finished (offset_top records the completed total) so we just have to save
1812 all the potential data. There may be up to 65535 such values, which is too
1813 large to put on the stack, but using malloc for small numbers seems
1814 expensive. As a compromise, the stack is used when there are no more than
1815 REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
1816
1817 There are also other values that have to be saved. We use a chained
1818 sequence of blocks that actually live on the stack. Thanks to Robin Houston
1819 for the original version of this logic. It has, however, been hacked around
1820 a lot, so he is not to blame for the current way it works. */
1821
1822 case OP_RECURSE:
1823 {
1824 recursion_info *ri;
1825 unsigned int recno;
1826
1827 callpat = md->start_code + GET(ecode, 1);
1828 recno = (callpat == md->start_code)? 0 :
1829 GET2(callpat, 1 + LINK_SIZE);
1830
1831 /* Check for repeating a recursion without advancing the subject pointer.
1832 This should catch convoluted mutual recursions. (Some simple cases are
1833 caught at compile time.) */
1834
1835 for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1836 if (recno == ri->group_num && eptr == ri->subject_position)
1837 RRETURN(PCRE_ERROR_RECURSELOOP);
1838
1839 /* Add to "recursing stack" */
1840
1841 new_recursive.group_num = recno;
1842 new_recursive.saved_capture_last = md->capture_last;
1843 new_recursive.subject_position = eptr;
1844 new_recursive.prevrec = md->recursive;
1845 md->recursive = &new_recursive;
1846
1847 /* Where to continue from afterwards */
1848
1849 ecode += 1 + LINK_SIZE;
1850
1851 /* Now save the offset data */
1852
1853 new_recursive.saved_max = md->offset_end;
1854 if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
1855 new_recursive.offset_save = stacksave;
1856 else
1857 {
1858 new_recursive.offset_save =
1859 (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1860 if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1861 }
1862 memcpy(new_recursive.offset_save, md->offset_vector,
1863 new_recursive.saved_max * sizeof(int));
1864
1865 /* OK, now we can do the recursion. After processing each alternative,
1866 restore the offset data and the last captured value. If there were nested
1867 recursions, md->recursive might be changed, so reset it before looping.
1868 */
1869
1870 DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1871 cbegroup = (*callpat >= OP_SBRA);
1872 do
1873 {
1874 if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1875 RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1876 md, eptrb, RM6);
1877 memcpy(md->offset_vector, new_recursive.offset_save,
1878 new_recursive.saved_max * sizeof(int));
1879 md->capture_last = new_recursive.saved_capture_last;
1880 md->recursive = new_recursive.prevrec;
1881 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1882 {
1883 DPRINTF(("Recursion matched\n"));
1884 if (new_recursive.offset_save != stacksave)
1885 (PUBL(free))(new_recursive.offset_save);
1886
1887 /* Set where we got to in the subject, and reset the start in case
1888 it was changed by \K. This *is* propagated back out of a recursion,
1889 for Perl compatibility. */
1890
1891 eptr = md->end_match_ptr;
1892 mstart = md->start_match_ptr;
1893 goto RECURSION_MATCHED; /* Exit loop; end processing */
1894 }
1895
1896 /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1897 recursion; they cause a NOMATCH for the entire recursion. These codes
1898 are defined in a range that can be tested for. */
1899
1900 if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1901 RRETURN(MATCH_NOMATCH);
1902
1903 /* Any return code other than NOMATCH is an error. */
1904
1905 if (rrc != MATCH_NOMATCH)
1906 {
1907 DPRINTF(("Recursion gave error %d\n", rrc));
1908 if (new_recursive.offset_save != stacksave)
1909 (PUBL(free))(new_recursive.offset_save);
1910 RRETURN(rrc);
1911 }
1912
1913 md->recursive = &new_recursive;
1914 callpat += GET(callpat, 1);
1915 }
1916 while (*callpat == OP_ALT);
1917
1918 DPRINTF(("Recursion didn't match\n"));
1919 md->recursive = new_recursive.prevrec;
1920 if (new_recursive.offset_save != stacksave)
1921 (PUBL(free))(new_recursive.offset_save);
1922 RRETURN(MATCH_NOMATCH);
1923 }
1924
1925 RECURSION_MATCHED:
1926 break;
1927
1928 /* An alternation is the end of a branch; scan along to find the end of the
1929 bracketed group and go to there. */
1930
1931 case OP_ALT:
1932 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1933 break;
1934
1935 /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1936 indicating that it may occur zero times. It may repeat infinitely, or not
1937 at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1938 with fixed upper repeat limits are compiled as a number of copies, with the
1939 optional ones preceded by BRAZERO or BRAMINZERO. */
1940
1941 case OP_BRAZERO:
1942 next = ecode + 1;
1943 RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1944 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1945 do next += GET(next, 1); while (*next == OP_ALT);
1946 ecode = next + 1 + LINK_SIZE;
1947 break;
1948
1949 case OP_BRAMINZERO:
1950 next = ecode + 1;
1951 do next += GET(next, 1); while (*next == OP_ALT);
1952 RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1953 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1954 ecode++;
1955 break;
1956
1957 case OP_SKIPZERO:
1958 next = ecode+1;
1959 do next += GET(next,1); while (*next == OP_ALT);
1960 ecode = next + 1 + LINK_SIZE;
1961 break;
1962
1963 /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1964 here; just jump to the group, with allow_zero set TRUE. */
1965
1966 case OP_BRAPOSZERO:
1967 op = *(++ecode);
1968 allow_zero = TRUE;
1969 if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1970 goto POSSESSIVE_NON_CAPTURE;
1971
1972 /* End of a group, repeated or non-repeating. */
1973
1974 case OP_KET:
1975 case OP_KETRMIN:
1976 case OP_KETRMAX:
1977 case OP_KETRPOS:
1978 prev = ecode - GET(ecode, 1);
1979
1980 /* If this was a group that remembered the subject start, in order to break
1981 infinite repeats of empty string matches, retrieve the subject start from
1982 the chain. Otherwise, set it NULL. */
1983
1984 if (*prev >= OP_SBRA || *prev == OP_ONCE)
1985 {
1986 saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */
1987 eptrb = eptrb->epb_prev; /* Backup to previous group */
1988 }
1989 else saved_eptr = NULL;
1990
1991 /* If we are at the end of an assertion group or a non-capturing atomic
1992 group, stop matching and return MATCH_MATCH, but record the current high
1993 water mark for use by positive assertions. We also need to record the match
1994 start in case it was changed by \K. */
1995
1996 if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1997 *prev == OP_ONCE_NC)
1998 {
1999 md->end_match_ptr = eptr; /* For ONCE_NC */
2000 md->end_offset_top = offset_top;
2001 md->start_match_ptr = mstart;
2002 RRETURN(MATCH_MATCH); /* Sets md->mark */
2003 }
2004
2005 /* For capturing groups we have to check the group number back at the start
2006 and if necessary complete handling an extraction by setting the offsets and
2007 bumping the high water mark. Whole-pattern recursion is coded as a recurse
2008 into group 0, so it won't be picked up here. Instead, we catch it when the
2009 OP_END is reached. Other recursion is handled here. We just have to record
2010 the current subject position and start match pointer and give a MATCH
2011 return. */
2012
2013 if (*prev == OP_CBRA || *prev == OP_SCBRA ||
2014 *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
2015 {
2016 number = GET2(prev, 1+LINK_SIZE);
2017 offset = number << 1;
2018
2019 #ifdef PCRE_DEBUG
2020 printf("end bracket %d", number);
2021 printf("\n");
2022 #endif
2023
2024 /* Handle a recursively called group. */
2025
2026 if (md->recursive != NULL && md->recursive->group_num == number)
2027 {
2028 md->end_match_ptr = eptr;
2029 md->start_match_ptr = mstart;
2030 RRETURN(MATCH_MATCH);
2031 }
2032
2033 /* Deal with capturing */
2034
2035 md->capture_last = (md->capture_last & OVFLMASK) | number;
2036 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
2037 {
2038 /* If offset is greater than offset_top, it means that we are
2039 "skipping" a capturing group, and that group's offsets must be marked
2040 unset. In earlier versions of PCRE, all the offsets were unset at the
2041 start of matching, but this doesn't work because atomic groups and
2042 assertions can cause a value to be set that should later be unset.
2043 Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
2044 part of the atomic group, but this is not on the final matching path,
2045 so must be unset when 2 is set. (If there is no group 2, there is no
2046 problem, because offset_top will then be 2, indicating no capture.) */
2047
2048 if (offset > offset_top)
2049 {
2050 register int *iptr = md->offset_vector + offset_top;
2051 register int *iend = md->offset_vector + offset;
2052 while (iptr < iend) *iptr++ = -1;
2053 }
2054
2055 /* Now make the extraction */
2056
2057 md->offset_vector[offset] =
2058 md->offset_vector[md->offset_end - number];
2059 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2060 if (offset_top <= offset) offset_top = offset + 2;
2061 }
2062 }
2063
2064 /* For an ordinary non-repeating ket, just continue at this level. This
2065 also happens for a repeating ket if no characters were matched in the
2066 group. This is the forcible breaking of infinite loops as implemented in
2067 Perl 5.005. For a non-repeating atomic group that includes captures,
2068 establish a backup point by processing the rest of the pattern at a lower
2069 level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2070 original OP_ONCE level, thereby bypassing intermediate backup points, but
2071 resetting any captures that happened along the way. */
2072
2073 if (*ecode == OP_KET || eptr == saved_eptr)
2074 {
2075 if (*prev == OP_ONCE)
2076 {
2077 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2078 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2079 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
2080 RRETURN(MATCH_ONCE);
2081 }
2082 ecode += 1 + LINK_SIZE; /* Carry on at this level */
2083 break;
2084 }
2085
2086 /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2087 and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2088 at a time from the outer level, thus saving stack. */
2089
2090 if (*ecode == OP_KETRPOS)
2091 {
2092 md->end_match_ptr = eptr;
2093 md->end_offset_top = offset_top;
2094 RRETURN(MATCH_KETRPOS);
2095 }
2096
2097 /* The normal repeating kets try the rest of the pattern or restart from
2098 the preceding bracket, in the appropriate order. In the second case, we can
2099 use tail recursion to avoid using another stack frame, unless we have an
2100 an atomic group or an unlimited repeat of a group that can match an empty
2101 string. */
2102
2103 if (*ecode == OP_KETRMIN)
2104 {
2105 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2106 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2107 if (*prev == OP_ONCE)
2108 {
2109 RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2110 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2111 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
2112 RRETURN(MATCH_ONCE);
2113 }
2114 if (*prev >= OP_SBRA) /* Could match an empty string */
2115 {
2116 RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2117 RRETURN(rrc);
2118 }
2119 ecode = prev;
2120 goto TAIL_RECURSE;
2121 }
2122 else /* OP_KETRMAX */
2123 {
2124 RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2125 if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2126 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2127 if (*prev == OP_ONCE)
2128 {
2129 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2130 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2131 md->once_target = prev;
2132 RRETURN(MATCH_ONCE);
2133 }
2134 ecode += 1 + LINK_SIZE;
2135 goto TAIL_RECURSE;
2136 }
2137 /* Control never gets here */
2138
2139 /* Not multiline mode: start of subject assertion, unless notbol. */
2140
2141 case OP_CIRC:
2142 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2143
2144 /* Start of subject assertion */
2145
2146 case OP_SOD:
2147 if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
2148 ecode++;
2149 break;
2150
2151 /* Multiline mode: start of subject unless notbol, or after any newline. */
2152
2153 case OP_CIRCM:
2154 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2155 if (eptr != md->start_subject &&
2156 (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2157 RRETURN(MATCH_NOMATCH);
2158 ecode++;
2159 break;
2160
2161 /* Start of match assertion */
2162
2163 case OP_SOM:
2164 if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
2165 ecode++;
2166 break;
2167
2168 /* Reset the start of match point */
2169
2170 case OP_SET_SOM:
2171 mstart = eptr;
2172 ecode++;
2173 break;
2174
2175 /* Multiline mode: assert before any newline, or before end of subject
2176 unless noteol is set. */
2177
2178 case OP_DOLLM:
2179 if (eptr < md->end_subject)
2180 {
2181 if (!IS_NEWLINE(eptr))
2182 {
2183 if (md->partial != 0 &&
2184 eptr + 1 >= md->end_subject &&
2185 NLBLOCK->nltype == NLTYPE_FIXED &&
2186 NLBLOCK->nllen == 2 &&
2187 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2188 {
2189 md->hitend = TRUE;
2190 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2191 }
2192 RRETURN(MATCH_NOMATCH);
2193 }
2194 }
2195 else
2196 {
2197 if (md->noteol) RRETURN(MATCH_NOMATCH);
2198 SCHECK_PARTIAL();
2199 }
2200 ecode++;
2201 break;
2202
2203 /* Not multiline mode: assert before a terminating newline or before end of
2204 subject unless noteol is set. */
2205
2206 case OP_DOLL:
2207 if (md->noteol) RRETURN(MATCH_NOMATCH);
2208 if (!md->endonly) goto ASSERT_NL_OR_EOS;
2209
2210 /* ... else fall through for endonly */
2211
2212 /* End of subject assertion (\z) */
2213
2214 case OP_EOD:
2215 if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2216 SCHECK_PARTIAL();
2217 ecode++;
2218 break;
2219
2220 /* End of subject or ending \n assertion (\Z) */
2221
2222 case OP_EODN:
2223 ASSERT_NL_OR_EOS:
2224 if (eptr < md->end_subject &&
2225 (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2226 {
2227 if (md->partial != 0 &&
2228 eptr + 1 >= md->end_subject &&
2229 NLBLOCK->nltype == NLTYPE_FIXED &&
2230 NLBLOCK->nllen == 2 &&
2231 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2232 {
2233 md->hitend = TRUE;
2234 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2235 }
2236 RRETURN(MATCH_NOMATCH);
2237 }
2238
2239 /* Either at end of string or \n before end. */
2240
2241 SCHECK_PARTIAL();
2242 ecode++;
2243 break;
2244
2245 /* Word boundary assertions */
2246
2247 case OP_NOT_WORD_BOUNDARY:
2248 case OP_WORD_BOUNDARY:
2249 {
2250
2251 /* Find out if the previous and current characters are "word" characters.
2252 It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2253 be "non-word" characters. Remember the earliest consulted character for
2254 partial matching. */
2255
2256 #ifdef SUPPORT_UTF
2257 if (utf)
2258 {
2259 /* Get status of previous character */
2260
2261 if (eptr == md->start_subject) prev_is_word = FALSE; else
2262 {
2263 PCRE_PUCHAR lastptr = eptr - 1;
2264 BACKCHAR(lastptr);
2265 if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2266 GETCHAR(c, lastptr);
2267 #ifdef SUPPORT_UCP
2268 if (md->use_ucp)
2269 {
2270 if (c == '_') prev_is_word = TRUE; else
2271 {
2272 int cat = UCD_CATEGORY(c);
2273 prev_is_word = (cat == ucp_L || cat == ucp_N);
2274 }
2275 }
2276 else
2277 #endif
2278 prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2279 }
2280
2281 /* Get status of next character */
2282
2283 if (eptr >= md->end_subject)
2284 {
2285 SCHECK_PARTIAL();
2286 cur_is_word = FALSE;
2287 }
2288 else
2289 {
2290 GETCHAR(c, eptr);
2291 #ifdef SUPPORT_UCP
2292 if (md->use_ucp)
2293 {
2294 if (c == '_') cur_is_word = TRUE; else
2295 {
2296 int cat = UCD_CATEGORY(c);
2297 cur_is_word = (cat == ucp_L || cat == ucp_N);
2298 }
2299 }
2300 else
2301 #endif
2302 cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2303 }
2304 }
2305 else
2306 #endif
2307
2308 /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2309 consistency with the behaviour of \w we do use it in this case. */
2310
2311 {
2312 /* Get status of previous character */
2313
2314 if (eptr == md->start_subject) prev_is_word = FALSE; else
2315 {
2316 if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2317 #ifdef SUPPORT_UCP
2318 if (md->use_ucp)
2319 {
2320 c = eptr[-1];
2321 if (c == '_') prev_is_word = TRUE; else
2322 {
2323 int cat = UCD_CATEGORY(c);
2324 prev_is_word = (cat == ucp_L || cat == ucp_N);
2325 }
2326 }
2327 else
2328 #endif
2329 prev_is_word = MAX_255(eptr[-1])
2330 && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2331 }
2332
2333 /* Get status of next character */
2334
2335 if (eptr >= md->end_subject)
2336 {
2337 SCHECK_PARTIAL();
2338 cur_is_word = FALSE;
2339 }
2340 else
2341 #ifdef SUPPORT_UCP
2342 if (md->use_ucp)
2343 {
2344 c = *eptr;
2345 if (c == '_') cur_is_word = TRUE; else
2346 {
2347 int cat = UCD_CATEGORY(c);
2348 cur_is_word = (cat == ucp_L || cat == ucp_N);
2349 }
2350 }
2351 else
2352 #endif
2353 cur_is_word = MAX_255(*eptr)
2354 && ((md->ctypes[*eptr] & ctype_word) != 0);
2355 }
2356
2357 /* Now see if the situation is what we want */
2358
2359 if ((*ecode++ == OP_WORD_BOUNDARY)?
2360 cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2361 RRETURN(MATCH_NOMATCH);
2362 }
2363 break;
2364
2365 /* Match any single character type except newline; have to take care with
2366 CRLF newlines and partial matching. */
2367
2368 case OP_ANY:
2369 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2370 if (md->partial != 0 &&
2371 eptr + 1 >= md->end_subject &&
2372 NLBLOCK->nltype == NLTYPE_FIXED &&
2373 NLBLOCK->nllen == 2 &&
2374 RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2375 {
2376 md->hitend = TRUE;
2377 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2378 }
2379
2380 /* Fall through */
2381
2382 /* Match any single character whatsoever. */
2383
2384 case OP_ALLANY:
2385 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2386 { /* not be updated before SCHECK_PARTIAL. */
2387 SCHECK_PARTIAL();
2388 RRETURN(MATCH_NOMATCH);
2389 }
2390 eptr++;
2391 #ifdef SUPPORT_UTF
2392 if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2393 #endif
2394 ecode++;
2395 break;
2396
2397 /* Match a single byte, even in UTF-8 mode. This opcode really does match
2398 any byte, even newline, independent of the setting of PCRE_DOTALL. */
2399
2400 case OP_ANYBYTE:
2401 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2402 { /* not be updated before SCHECK_PARTIAL. */
2403 SCHECK_PARTIAL();
2404 RRETURN(MATCH_NOMATCH);
2405 }
2406 eptr++;
2407 ecode++;
2408 break;
2409
2410 case OP_NOT_DIGIT:
2411 if (eptr >= md->end_subject)
2412 {
2413 SCHECK_PARTIAL();
2414 RRETURN(MATCH_NOMATCH);
2415 }
2416 GETCHARINCTEST(c, eptr);
2417 if (
2418 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2419 c < 256 &&
2420 #endif
2421 (md->ctypes[c] & ctype_digit) != 0
2422 )
2423 RRETURN(MATCH_NOMATCH);
2424 ecode++;
2425 break;
2426
2427 case OP_DIGIT:
2428 if (eptr >= md->end_subject)
2429 {
2430 SCHECK_PARTIAL();
2431 RRETURN(MATCH_NOMATCH);
2432 }
2433 GETCHARINCTEST(c, eptr);
2434 if (
2435 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2436 c > 255 ||
2437 #endif
2438 (md->ctypes[c] & ctype_digit) == 0
2439 )
2440 RRETURN(MATCH_NOMATCH);
2441 ecode++;
2442 break;
2443
2444 case OP_NOT_WHITESPACE:
2445 if (eptr >= md->end_subject)
2446 {
2447 SCHECK_PARTIAL();
2448 RRETURN(MATCH_NOMATCH);
2449 }
2450 GETCHARINCTEST(c, eptr);
2451 if (
2452 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2453 c < 256 &&
2454 #endif
2455 (md->ctypes[c] & ctype_space) != 0
2456 )
2457 RRETURN(MATCH_NOMATCH);
2458 ecode++;
2459 break;
2460
2461 case OP_WHITESPACE:
2462 if (eptr >= md->end_subject)
2463 {
2464 SCHECK_PARTIAL();
2465 RRETURN(MATCH_NOMATCH);
2466 }
2467 GETCHARINCTEST(c, eptr);
2468 if (
2469 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2470 c > 255 ||
2471 #endif
2472 (md->ctypes[c] & ctype_space) == 0
2473 )
2474 RRETURN(MATCH_NOMATCH);
2475 ecode++;
2476 break;
2477
2478 case OP_NOT_WORDCHAR:
2479 if (eptr >= md->end_subject)
2480 {
2481 SCHECK_PARTIAL();
2482 RRETURN(MATCH_NOMATCH);
2483 }
2484 GETCHARINCTEST(c, eptr);
2485 if (
2486 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2487 c < 256 &&
2488 #endif
2489 (md->ctypes[c] & ctype_word) != 0
2490 )
2491 RRETURN(MATCH_NOMATCH);
2492 ecode++;
2493 break;
2494
2495 case OP_WORDCHAR:
2496 if (eptr >= md->end_subject)
2497 {
2498 SCHECK_PARTIAL();
2499 RRETURN(MATCH_NOMATCH);
2500 }
2501 GETCHARINCTEST(c, eptr);
2502 if (
2503 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2504 c > 255 ||
2505 #endif
2506 (md->ctypes[c] & ctype_word) == 0
2507 )
2508 RRETURN(MATCH_NOMATCH);
2509 ecode++;
2510 break;
2511
2512 case OP_ANYNL:
2513 if (eptr >= md->end_subject)
2514 {
2515 SCHECK_PARTIAL();
2516 RRETURN(MATCH_NOMATCH);
2517 }
2518 GETCHARINCTEST(c, eptr);
2519 switch(c)
2520 {
2521 default: RRETURN(MATCH_NOMATCH);
2522
2523 case CHAR_CR:
2524 if (eptr >= md->end_subject)
2525 {
2526 SCHECK_PARTIAL();
2527 }
2528 else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2529 break;
2530
2531 case CHAR_LF:
2532 break;
2533
2534 case CHAR_VT:
2535 case CHAR_FF:
2536 case CHAR_NEL:
2537 #ifndef EBCDIC
2538 case 0x2028:
2539 case 0x2029:
2540 #endif /* Not EBCDIC */
2541 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2542 break;
2543 }
2544 ecode++;
2545 break;
2546
2547 case OP_NOT_HSPACE:
2548 if (eptr >= md->end_subject)
2549 {
2550 SCHECK_PARTIAL();
2551 RRETURN(MATCH_NOMATCH);
2552 }
2553 GETCHARINCTEST(c, eptr);
2554 switch(c)
2555 {
2556 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
2557 default: break;
2558 }
2559 ecode++;
2560 break;
2561
2562 case OP_HSPACE:
2563 if (eptr >= md->end_subject)
2564 {
2565 SCHECK_PARTIAL();
2566 RRETURN(MATCH_NOMATCH);
2567 }
2568 GETCHARINCTEST(c, eptr);
2569 switch(c)
2570 {
2571 HSPACE_CASES: break; /* Byte and multibyte cases */
2572 default: RRETURN(MATCH_NOMATCH);
2573 }
2574 ecode++;
2575 break;
2576
2577 case OP_NOT_VSPACE:
2578 if (eptr >= md->end_subject)
2579 {
2580 SCHECK_PARTIAL();
2581 RRETURN(MATCH_NOMATCH);
2582 }
2583 GETCHARINCTEST(c, eptr);
2584 switch(c)
2585 {
2586 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2587 default: break;
2588 }
2589 ecode++;
2590 break;
2591
2592 case OP_VSPACE:
2593 if (eptr >= md->end_subject)
2594 {
2595 SCHECK_PARTIAL();
2596 RRETURN(MATCH_NOMATCH);
2597 }
2598 GETCHARINCTEST(c, eptr);
2599 switch(c)
2600 {
2601 VSPACE_CASES: break;
2602 default: RRETURN(MATCH_NOMATCH);
2603 }
2604 ecode++;
2605 break;
2606
2607 #ifdef SUPPORT_UCP
2608 /* Check the next character by Unicode property. We will get here only
2609 if the support is in the binary; otherwise a compile-time error occurs. */
2610
2611 case OP_PROP:
2612 case OP_NOTPROP:
2613 if (eptr >= md->end_subject)
2614 {
2615 SCHECK_PARTIAL();
2616 RRETURN(MATCH_NOMATCH);
2617 }
2618 GETCHARINCTEST(c, eptr);
2619 {
2620 const pcre_uint32 *cp;
2621 const ucd_record *prop = GET_UCD(c);
2622
2623 switch(ecode[1])
2624 {
2625 case PT_ANY:
2626 if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2627 break;
2628
2629 case PT_LAMP:
2630 if ((prop->chartype == ucp_Lu ||
2631 prop->chartype == ucp_Ll ||
2632 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2633 RRETURN(MATCH_NOMATCH);
2634 break;
2635
2636 case PT_GC:
2637 if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2638 RRETURN(MATCH_NOMATCH);
2639 break;
2640
2641 case PT_PC:
2642 if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2643 RRETURN(MATCH_NOMATCH);
2644 break;
2645
2646 case PT_SC:
2647 if ((ecode[2] != prop->script) == (op == OP_PROP))
2648 RRETURN(MATCH_NOMATCH);
2649 break;
2650
2651 /* These are specials */
2652
2653 case PT_ALNUM:
2654 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2655 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2656 RRETURN(MATCH_NOMATCH);
2657 break;
2658
2659 case PT_SPACE: /* Perl space */
2660 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2661 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2662 == (op == OP_NOTPROP))
2663 RRETURN(MATCH_NOMATCH);
2664 break;
2665
2666 case PT_PXSPACE: /* POSIX space */
2667 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2668 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2669 c == CHAR_FF || c == CHAR_CR)
2670 == (op == OP_NOTPROP))
2671 RRETURN(MATCH_NOMATCH);
2672 break;
2673
2674 case PT_WORD:
2675 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2676 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2677 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2678 RRETURN(MATCH_NOMATCH);
2679 break;
2680
2681 case PT_CLIST:
2682 cp = PRIV(ucd_caseless_sets) + ecode[2];
2683 for (;;)
2684 {
2685 if (c < *cp)
2686 { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2687 if (c == *cp++)
2688 { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2689 }
2690 break;
2691
2692 case PT_UCNC:
2693 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2694 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2695 c >= 0xe000) == (op == OP_NOTPROP))
2696 RRETURN(MATCH_NOMATCH);
2697 break;
2698
2699 /* This should never occur */
2700
2701 default:
2702 RRETURN(PCRE_ERROR_INTERNAL);
2703 }
2704
2705 ecode += 3;
2706 }
2707 break;
2708
2709 /* Match an extended Unicode sequence. We will get here only if the support
2710 is in the binary; otherwise a compile-time error occurs. */
2711
2712 case OP_EXTUNI:
2713 if (eptr >= md->end_subject)
2714 {
2715 SCHECK_PARTIAL();
2716 RRETURN(MATCH_NOMATCH);
2717 }
2718 else
2719 {
2720 int lgb, rgb;
2721 GETCHARINCTEST(c, eptr);
2722 lgb = UCD_GRAPHBREAK(c);
2723 while (eptr < md->end_subject)
2724 {
2725 int len = 1;
2726 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2727 rgb = UCD_GRAPHBREAK(c);
2728 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2729 lgb = rgb;
2730 eptr += len;
2731 }
2732 }
2733 CHECK_PARTIAL();
2734 ecode++;
2735 break;
2736 #endif /* SUPPORT_UCP */
2737
2738
2739 /* Match a back reference, possibly repeatedly. Look past the end of the
2740 item to see if there is repeat information following. The code is similar
2741 to that for character classes, but repeated for efficiency. Then obey
2742 similar code to character type repeats - written out again for speed.
2743 However, if the referenced string is the empty string, always treat
2744 it as matched, any number of times (otherwise there could be infinite
2745 loops). */
2746
2747 case OP_REF:
2748 case OP_REFI:
2749 caseless = op == OP_REFI;
2750 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
2751 ecode += 1 + IMM2_SIZE;
2752
2753 /* If the reference is unset, there are two possibilities:
2754
2755 (a) In the default, Perl-compatible state, set the length negative;
2756 this ensures that every attempt at a match fails. We can't just fail
2757 here, because of the possibility of quantifiers with zero minima.
2758
2759 (b) If the JavaScript compatibility flag is set, set the length to zero
2760 so that the back reference matches an empty string.
2761
2762 Otherwise, set the length to the length of what was matched by the
2763 referenced subpattern. */
2764
2765 if (offset >= offset_top || md->offset_vector[offset] < 0)
2766 length = (md->jscript_compat)? 0 : -1;
2767 else
2768 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2769
2770 /* Set up for repetition, or handle the non-repeated case */
2771
2772 switch (*ecode)
2773 {
2774 case OP_CRSTAR:
2775 case OP_CRMINSTAR:
2776 case OP_CRPLUS:
2777 case OP_CRMINPLUS:
2778 case OP_CRQUERY:
2779 case OP_CRMINQUERY:
2780 c = *ecode++ - OP_CRSTAR;
2781 minimize = (c & 1) != 0;
2782 min = rep_min[c]; /* Pick up values from tables; */
2783 max = rep_max[c]; /* zero for max => infinity */
2784 if (max == 0) max = INT_MAX;
2785 break;
2786
2787 case OP_CRRANGE:
2788 case OP_CRMINRANGE:
2789 minimize = (*ecode == OP_CRMINRANGE);
2790 min = GET2(ecode, 1);
2791 max = GET2(ecode, 1 + IMM2_SIZE);
2792 if (max == 0) max = INT_MAX;
2793 ecode += 1 + 2 * IMM2_SIZE;
2794 break;
2795
2796 default: /* No repeat follows */
2797 if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2798 {
2799 if (length == -2) eptr = md->end_subject; /* Partial match */
2800 CHECK_PARTIAL();
2801 RRETURN(MATCH_NOMATCH);
2802 }
2803 eptr += length;
2804 continue; /* With the main loop */
2805 }
2806
2807 /* Handle repeated back references. If the length of the reference is
2808 zero, just continue with the main loop. If the length is negative, it
2809 means the reference is unset in non-Java-compatible mode. If the minimum is
2810 zero, we can continue at the same level without recursion. For any other
2811 minimum, carrying on will result in NOMATCH. */
2812
2813 if (length == 0) continue;
2814 if (length < 0 && min == 0) continue;
2815
2816 /* First, ensure the minimum number of matches are present. We get back
2817 the length of the reference string explicitly rather than passing the
2818 address of eptr, so that eptr can be a register variable. */
2819
2820 for (i = 1; i <= min; i++)
2821 {
2822 int slength;
2823 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2824 {
2825 if (slength == -2) eptr = md->end_subject; /* Partial match */
2826 CHECK_PARTIAL();
2827 RRETURN(MATCH_NOMATCH);
2828 }
2829 eptr += slength;
2830 }
2831
2832 /* If min = max, continue at the same level without recursion.
2833 They are not both allowed to be zero. */
2834
2835 if (min == max) continue;
2836
2837 /* If minimizing, keep trying and advancing the pointer */
2838
2839 if (minimize)
2840 {
2841 for (fi = min;; fi++)
2842 {
2843 int slength;
2844 RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2845 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2846 if (fi >= max) RRETURN(MATCH_NOMATCH);
2847 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2848 {
2849 if (slength == -2) eptr = md->end_subject; /* Partial match */
2850 CHECK_PARTIAL();
2851 RRETURN(MATCH_NOMATCH);
2852 }
2853 eptr += slength;
2854 }
2855 /* Control never gets here */
2856 }
2857
2858 /* If maximizing, find the longest string and work backwards */
2859
2860 else
2861 {
2862 pp = eptr;
2863 for (i = min; i < max; i++)
2864 {
2865 int slength;
2866 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2867 {
2868 /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2869 the soft partial matching case. */
2870
2871 if (slength == -2 && md->partial != 0 &&
2872 md->end_subject > md->start_used_ptr)
2873 {
2874 md->hitend = TRUE;
2875 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2876 }
2877 break;
2878 }
2879 eptr += slength;
2880 }
2881
2882 while (eptr >= pp)
2883 {
2884 RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2885 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2886 eptr -= length;
2887 }
2888 RRETURN(MATCH_NOMATCH);
2889 }
2890 /* Control never gets here */
2891
2892 /* Match a bit-mapped character class, possibly repeatedly. This op code is
2893 used when all the characters in the class have values in the range 0-255,
2894 and either the matching is caseful, or the characters are in the range
2895 0-127 when UTF-8 processing is enabled. The only difference between
2896 OP_CLASS and OP_NCLASS occurs when a data character outside the range is
2897 encountered.
2898
2899 First, look past the end of the item to see if there is repeat information
2900 following. Then obey similar code to character type repeats - written out
2901 again for speed. */
2902
2903 case OP_NCLASS:
2904 case OP_CLASS:
2905 {
2906 /* The data variable is saved across frames, so the byte map needs to
2907 be stored there. */
2908 #define BYTE_MAP ((pcre_uint8 *)data)
2909 data = ecode + 1; /* Save for matching */
2910 ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2911
2912 switch (*ecode)
2913 {
2914 case OP_CRSTAR:
2915 case OP_CRMINSTAR:
2916 case OP_CRPLUS:
2917 case OP_CRMINPLUS:
2918 case OP_CRQUERY:
2919 case OP_CRMINQUERY:
2920 c = *ecode++ - OP_CRSTAR;
2921 minimize = (c & 1) != 0;
2922 min = rep_min[c]; /* Pick up values from tables; */
2923 max = rep_max[c]; /* zero for max => infinity */
2924 if (max == 0) max = INT_MAX;
2925 break;
2926
2927 case OP_CRRANGE:
2928 case OP_CRMINRANGE:
2929 minimize = (*ecode == OP_CRMINRANGE);
2930 min = GET2(ecode, 1);
2931 max = GET2(ecode, 1 + IMM2_SIZE);
2932 if (max == 0) max = INT_MAX;
2933 ecode += 1 + 2 * IMM2_SIZE;
2934 break;
2935
2936 default: /* No repeat follows */
2937 min = max = 1;
2938 break;
2939 }
2940
2941 /* First, ensure the minimum number of matches are present. */
2942
2943 #ifdef SUPPORT_UTF
2944 if (utf)
2945 {
2946 for (i = 1; i <= min; i++)
2947 {
2948 if (eptr >= md->end_subject)
2949 {
2950 SCHECK_PARTIAL();
2951 RRETURN(MATCH_NOMATCH);
2952 }
2953 GETCHARINC(c, eptr);
2954 if (c > 255)
2955 {
2956 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2957 }
2958 else
2959 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2960 }
2961 }
2962 else
2963 #endif
2964 /* Not UTF mode */
2965 {
2966 for (i = 1; i <= min; i++)
2967 {
2968 if (eptr >= md->end_subject)
2969 {
2970 SCHECK_PARTIAL();
2971 RRETURN(MATCH_NOMATCH);
2972 }
2973 c = *eptr++;
2974 #ifndef COMPILE_PCRE8
2975 if (c > 255)
2976 {
2977 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2978 }
2979 else
2980 #endif
2981 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2982 }
2983 }
2984
2985 /* If max == min we can continue with the main loop without the
2986 need to recurse. */
2987
2988 if (min == max) continue;
2989
2990 /* If minimizing, keep testing the rest of the expression and advancing
2991 the pointer while it matches the class. */
2992
2993 if (minimize)
2994 {
2995 #ifdef SUPPORT_UTF
2996 if (utf)
2997 {
2998 for (fi = min;; fi++)
2999 {
3000 RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
3001 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3002 if (fi >= max) RRETURN(MATCH_NOMATCH);
3003 if (eptr >= md->end_subject)
3004 {
3005 SCHECK_PARTIAL();
3006 RRETURN(MATCH_NOMATCH);
3007 }
3008 GETCHARINC(c, eptr);
3009 if (c > 255)
3010 {
3011 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3012 }
3013 else
3014 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3015 }
3016 }
3017 else
3018 #endif
3019 /* Not UTF mode */
3020 {
3021 for (fi = min;; fi++)
3022 {
3023 RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
3024 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3025 if (fi >= max) RRETURN(MATCH_NOMATCH);
3026 if (eptr >= md->end_subject)
3027 {
3028 SCHECK_PARTIAL();
3029 RRETURN(MATCH_NOMATCH);
3030 }
3031 c = *eptr++;
3032 #ifndef COMPILE_PCRE8
3033 if (c > 255)
3034 {
3035 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3036 }
3037 else
3038 #endif
3039 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3040 }
3041 }
3042 /* Control never gets here */
3043 }
3044
3045 /* If maximizing, find the longest possible run, then work backwards. */
3046
3047 else
3048 {
3049 pp = eptr;
3050
3051 #ifdef SUPPORT_UTF
3052 if (utf)
3053 {
3054 for (i = min; i < max; i++)
3055 {
3056 int len = 1;
3057 if (eptr >= md->end_subject)
3058 {
3059 SCHECK_PARTIAL();
3060 break;
3061 }
3062 GETCHARLEN(c, eptr, len);
3063 if (c > 255)
3064 {
3065 if (op == OP_CLASS) break;
3066 }
3067 else
3068 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3069 eptr += len;
3070 }
3071 for (;;)
3072 {
3073 RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3074 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3075 if (eptr-- == pp) break; /* Stop if tried at original pos */
3076 BACKCHAR(eptr);
3077 }
3078 }
3079 else
3080 #endif
3081 /* Not UTF mode */
3082 {
3083 for (i = min; i < max; i++)
3084 {
3085 if (eptr >= md->end_subject)
3086 {
3087 SCHECK_PARTIAL();
3088 break;
3089 }
3090 c = *eptr;
3091 #ifndef COMPILE_PCRE8
3092 if (c > 255)
3093 {
3094 if (op == OP_CLASS) break;
3095 }
3096 else
3097 #endif
3098 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3099 eptr++;
3100 }
3101 while (eptr >= pp)
3102 {
3103 RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3104 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3105 eptr--;
3106 }
3107 }
3108
3109 RRETURN(MATCH_NOMATCH);
3110 }
3111 #undef BYTE_MAP
3112 }
3113 /* Control never gets here */
3114
3115
3116 /* Match an extended character class. This opcode is encountered only
3117 when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
3118 mode, because Unicode properties are supported in non-UTF-8 mode. */
3119
3120 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3121 case OP_XCLASS:
3122 {
3123 data = ecode + 1 + LINK_SIZE; /* Save for matching */
3124 ecode += GET(ecode, 1); /* Advance past the item */
3125
3126 switch (*ecode)
3127 {
3128 case OP_CRSTAR:
3129 case OP_CRMINSTAR:
3130 case OP_CRPLUS:
3131 case OP_CRMINPLUS:
3132 case OP_CRQUERY:
3133 case OP_CRMINQUERY:
3134 c = *ecode++ - OP_CRSTAR;
3135 minimize = (c & 1) != 0;
3136 min = rep_min[c]; /* Pick up values from tables; */
3137 max = rep_max[c]; /* zero for max => infinity */
3138 if (max == 0) max = INT_MAX;
3139 break;
3140
3141 case OP_CRRANGE:
3142 case OP_CRMINRANGE:
3143 minimize = (*ecode == OP_CRMINRANGE);
3144 min = GET2(ecode, 1);
3145 max = GET2(ecode, 1 + IMM2_SIZE);
3146 if (max == 0) max = INT_MAX;
3147 ecode += 1 + 2 * IMM2_SIZE;
3148 break;
3149
3150 default: /* No repeat follows */
3151 min = max = 1;
3152 break;
3153 }
3154
3155 /* First, ensure the minimum number of matches are present. */
3156
3157 for (i = 1; i <= min; i++)
3158 {
3159 if (eptr >= md->end_subject)
3160 {
3161 SCHECK_PARTIAL();
3162 RRETURN(MATCH_NOMATCH);
3163 }
3164 GETCHARINCTEST(c, eptr);
3165 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3166 }
3167
3168 /* If max == min we can continue with the main loop without the
3169 need to recurse. */
3170
3171 if (min == max) continue;
3172
3173 /* If minimizing, keep testing the rest of the expression and advancing
3174 the pointer while it matches the class. */
3175
3176 if (minimize)
3177 {
3178 for (fi = min;; fi++)
3179 {
3180 RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3181 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3182 if (fi >= max) RRETURN(MATCH_NOMATCH);
3183 if (eptr >= md->end_subject)
3184 {
3185 SCHECK_PARTIAL();
3186 RRETURN(MATCH_NOMATCH);
3187 }
3188 GETCHARINCTEST(c, eptr);
3189 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3190 }
3191 /* Control never gets here */
3192 }
3193
3194 /* If maximizing, find the longest possible run, then work backwards. */
3195
3196 else
3197 {
3198 pp = eptr;
3199 for (i = min; i < max; i++)
3200 {
3201 int len = 1;
3202 if (eptr >= md->end_subject)
3203 {
3204 SCHECK_PARTIAL();
3205 break;
3206 }
3207 #ifdef SUPPORT_UTF
3208 GETCHARLENTEST(c, eptr, len);
3209 #else
3210 c = *eptr;
3211 #endif
3212 if (!PRIV(xclass)(c, data, utf)) break;
3213 eptr += len;
3214 }
3215 for(;;)
3216 {
3217 RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3218 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3219 if (eptr-- == pp) break; /* Stop if tried at original pos */
3220 #ifdef SUPPORT_UTF
3221 if (utf) BACKCHAR(eptr);
3222 #endif
3223 }
3224 RRETURN(MATCH_NOMATCH);
3225 }
3226
3227 /* Control never gets here */
3228 }
3229 #endif /* End of XCLASS */
3230
3231 /* Match a single character, casefully */
3232
3233 case OP_CHAR:
3234 #ifdef SUPPORT_UTF
3235 if (utf)
3236 {
3237 length = 1;
3238 ecode++;
3239 GETCHARLEN(fc, ecode, length);
3240 if (length > md->end_subject - eptr)
3241 {
3242 CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */
3243 RRETURN(MATCH_NOMATCH);
3244 }
3245 while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3246 }
3247 else
3248 #endif
3249 /* Not UTF mode */
3250 {
3251 if (md->end_subject - eptr < 1)
3252 {
3253 SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */
3254 RRETURN(MATCH_NOMATCH);
3255 }
3256 if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3257 ecode += 2;
3258 }
3259 break;
3260
3261 /* Match a single character, caselessly. If we are at the end of the
3262 subject, give up immediately. */
3263
3264 case OP_CHARI:
3265 if (eptr >= md->end_subject)
3266 {
3267 SCHECK_PARTIAL();
3268 RRETURN(MATCH_NOMATCH);
3269 }
3270
3271 #ifdef SUPPORT_UTF
3272 if (utf)
3273 {
3274 length = 1;
3275 ecode++;
3276 GETCHARLEN(fc, ecode, length);
3277
3278 /* If the pattern character's value is < 128, we have only one byte, and
3279 we know that its other case must also be one byte long, so we can use the
3280 fast lookup table. We know that there is at least one byte left in the
3281 subject. */
3282
3283 if (fc < 128)
3284 {
3285 pcre_uint32 cc = RAWUCHAR(eptr);
3286 if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3287 ecode++;
3288 eptr++;
3289 }
3290
3291 /* Otherwise we must pick up the subject character. Note that we cannot
3292 use the value of "length" to check for sufficient bytes left, because the
3293 other case of the character may have more or fewer bytes. */
3294
3295 else
3296 {
3297 pcre_uint32 dc;
3298 GETCHARINC(dc, eptr);
3299 ecode += length;
3300
3301 /* If we have Unicode property support, we can use it to test the other
3302 case of the character, if there is one. */
3303
3304 if (fc != dc)
3305 {
3306 #ifdef SUPPORT_UCP
3307 if (dc != UCD_OTHERCASE(fc))
3308 #endif
3309 RRETURN(MATCH_NOMATCH);
3310 }
3311 }
3312 }
3313 else
3314 #endif /* SUPPORT_UTF */
3315
3316 /* Not UTF mode */
3317 {
3318 if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3319 != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3320 eptr++;
3321 ecode += 2;
3322 }
3323 break;
3324
3325 /* Match a single character repeatedly. */
3326
3327 case OP_EXACT:
3328 case OP_EXACTI:
3329 min = max = GET2(ecode, 1);
3330 ecode += 1 + IMM2_SIZE;
3331 goto REPEATCHAR;
3332
3333 case OP_POSUPTO:
3334 case OP_POSUPTOI:
3335 possessive = TRUE;
3336 /* Fall through */
3337
3338 case OP_UPTO:
3339 case OP_UPTOI:
3340 case OP_MINUPTO:
3341 case OP_MINUPTOI:
3342 min = 0;
3343 max = GET2(ecode, 1);
3344 minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3345 ecode += 1 + IMM2_SIZE;
3346 goto REPEATCHAR;
3347
3348 case OP_POSSTAR:
3349 case OP_POSSTARI:
3350 possessive = TRUE;
3351 min = 0;
3352 max = INT_MAX;
3353 ecode++;
3354 goto REPEATCHAR;
3355
3356 case OP_POSPLUS:
3357 case OP_POSPLUSI:
3358 possessive = TRUE;
3359 min = 1;
3360 max = INT_MAX;
3361 ecode++;
3362 goto REPEATCHAR;
3363
3364 case OP_POSQUERY:
3365 case OP_POSQUERYI:
3366 possessive = TRUE;
3367 min = 0;
3368 max = 1;
3369 ecode++;
3370 goto REPEATCHAR;
3371
3372 case OP_STAR:
3373 case OP_STARI:
3374 case OP_MINSTAR:
3375 case OP_MINSTARI:
3376 case OP_PLUS:
3377 case OP_PLUSI:
3378 case OP_MINPLUS:
3379 case OP_MINPLUSI:
3380 case OP_QUERY:
3381 case OP_QUERYI:
3382 case OP_MINQUERY:
3383 case OP_MINQUERYI:
3384 c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3385 minimize = (c & 1) != 0;
3386 min = rep_min[c]; /* Pick up values from tables; */
3387 max = rep_max[c]; /* zero for max => infinity */
3388 if (max == 0) max = INT_MAX;
3389
3390 /* Common code for all repeated single-character matches. We first check
3391 for the minimum number of characters. If the minimum equals the maximum, we
3392 are done. Otherwise, if minimizing, check the rest of the pattern for a
3393 match; if there isn't one, advance up to the maximum, one character at a
3394 time.
3395
3396 If maximizing, advance up to the maximum number of matching characters,
3397 until eptr is past the end of the maximum run. If possessive, we are
3398 then done (no backing up). Otherwise, match at this position; anything
3399 other than no match is immediately returned. For nomatch, back up one
3400 character, unless we are matching \R and the last thing matched was
3401 \r\n, in which case, back up two bytes. When we reach the first optional
3402 character position, we can save stack by doing a tail recurse.
3403
3404 The various UTF/non-UTF and caseful/caseless cases are handled separately,
3405 for speed. */
3406
3407 REPEATCHAR:
3408 #ifdef SUPPORT_UTF
3409 if (utf)
3410 {
3411 length = 1;
3412 charptr = ecode;
3413 GETCHARLEN(fc, ecode, length);
3414 ecode += length;
3415
3416 /* Handle multibyte character matching specially here. There is
3417 support for caseless matching if UCP support is present. */
3418
3419 if (length > 1)
3420 {
3421 #ifdef SUPPORT_UCP
3422 pcre_uint32 othercase;
3423 if (op >= OP_STARI && /* Caseless */
3424 (othercase = UCD_OTHERCASE(fc)) != fc)
3425 oclength = PRIV(ord2utf)(othercase, occhars);
3426 else oclength = 0;
3427 #endif /* SUPPORT_UCP */
3428
3429 for (i = 1; i <= min; i++)
3430 {
3431 if (eptr <= md->end_subject - length &&
3432 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3433 #ifdef SUPPORT_UCP
3434 else if (oclength > 0 &&
3435 eptr <= md->end_subject - oclength &&
3436 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3437 #endif /* SUPPORT_UCP */
3438 else
3439 {
3440 CHECK_PARTIAL();
3441 RRETURN(MATCH_NOMATCH);
3442 }
3443 }
3444
3445 if (min == max) continue;
3446
3447 if (minimize)
3448 {
3449 for (fi = min;; fi++)
3450 {
3451 RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3452 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3453 if (fi >= max) RRETURN(MATCH_NOMATCH);
3454 if (eptr <= md->end_subject - length &&
3455 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3456 #ifdef SUPPORT_UCP
3457 else if (oclength > 0 &&
3458 eptr <= md->end_subject - oclength &&
3459 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3460 #endif /* SUPPORT_UCP */
3461 else
3462 {
3463 CHECK_PARTIAL();
3464 RRETURN(MATCH_NOMATCH);
3465 }
3466 }
3467 /* Control never gets here */
3468 }
3469
3470 else /* Maximize */
3471 {
3472 pp = eptr;
3473 for (i = min; i < max; i++)
3474 {
3475 if (eptr <= md->end_subject - length &&
3476 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3477 #ifdef SUPPORT_UCP
3478 else if (oclength > 0 &&
3479 eptr <= md->end_subject - oclength &&
3480 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3481 #endif /* SUPPORT_UCP */
3482 else
3483 {
3484 CHECK_PARTIAL();
3485 break;
3486 }
3487 }
3488
3489 if (possessive) continue; /* No backtracking */
3490 for(;;)
3491 {
3492 if (eptr == pp) goto TAIL_RECURSE;
3493 RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3494 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3495 #ifdef SUPPORT_UCP
3496 eptr--;
3497 BACKCHAR(eptr);
3498 #else /* without SUPPORT_UCP */
3499 eptr -= length;
3500 #endif /* SUPPORT_UCP */
3501 }
3502 }
3503 /* Control never gets here */
3504 }
3505
3506 /* If the length of a UTF-8 character is 1, we fall through here, and
3507 obey the code as for non-UTF-8 characters below, though in this case the
3508 value of fc will always be < 128. */
3509 }
3510 else
3511 #endif /* SUPPORT_UTF */
3512 /* When not in UTF-8 mode, load a single-byte character. */
3513 fc = *ecode++;
3514
3515 /* The value of fc at this point is always one character, though we may
3516 or may not be in UTF mode. The code is duplicated for the caseless and
3517 caseful cases, for speed, since matching characters is likely to be quite
3518 common. First, ensure the minimum number of matches are present. If min =
3519 max, continue at the same level without recursing. Otherwise, if
3520 minimizing, keep trying the rest of the expression and advancing one
3521 matching character if failing, up to the maximum. Alternatively, if
3522 maximizing, find the maximum number of characters and work backwards. */
3523
3524 DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3525 max, (char *)eptr));
3526
3527 if (op >= OP_STARI) /* Caseless */
3528 {
3529 #ifdef COMPILE_PCRE8
3530 /* fc must be < 128 if UTF is enabled. */
3531 foc = md->fcc[fc];
3532 #else
3533 #ifdef SUPPORT_UTF
3534 #ifdef SUPPORT_UCP
3535 if (utf && fc > 127)
3536 foc = UCD_OTHERCASE(fc);
3537 #else
3538 if (utf && fc > 127)
3539 foc = fc;
3540 #endif /* SUPPORT_UCP */
3541 else
3542 #endif /* SUPPORT_UTF */
3543 foc = TABLE_GET(fc, md->fcc, fc);
3544 #endif /* COMPILE_PCRE8 */
3545
3546 for (i = 1; i <= min; i++)
3547 {
3548 pcre_uint32 cc; /* Faster than pcre_uchar */
3549 if (eptr >= md->end_subject)
3550 {
3551 SCHECK_PARTIAL();
3552 RRETURN(MATCH_NOMATCH);
3553 }
3554 cc = RAWUCHARTEST(eptr);
3555 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3556 eptr++;
3557 }
3558 if (min == max) continue;
3559 if (minimize)
3560 {
3561 for (fi = min;; fi++)
3562 {
3563 pcre_uint32 cc; /* Faster than pcre_uchar */
3564 RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3565 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3566 if (fi >= max) RRETURN(MATCH_NOMATCH);
3567 if (eptr >= md->end_subject)
3568 {
3569 SCHECK_PARTIAL();
3570 RRETURN(MATCH_NOMATCH);
3571 }
3572 cc = RAWUCHARTEST(eptr);
3573 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3574 eptr++;
3575 }
3576 /* Control never gets here */
3577 }
3578 else /* Maximize */
3579 {
3580 pp = eptr;
3581 for (i = min; i < max; i++)
3582 {
3583 pcre_uint32 cc; /* Faster than pcre_uchar */
3584 if (eptr >= md->end_subject)
3585 {
3586 SCHECK_PARTIAL();
3587 break;
3588 }
3589 cc = RAWUCHARTEST(eptr);
3590 if (fc != cc && foc != cc) break;
3591 eptr++;
3592 }
3593 if (possessive) continue; /* No backtracking */
3594 for (;;)
3595 {
3596 if (eptr == pp) goto TAIL_RECURSE;
3597 RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3598 eptr--;
3599 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3600 }
3601 /* Control never gets here */
3602 }
3603 }
3604
3605 /* Caseful comparisons (includes all multi-byte characters) */
3606
3607 else
3608 {
3609 for (i = 1; i <= min; i++)
3610 {
3611 if (eptr >= md->end_subject)
3612 {
3613 SCHECK_PARTIAL();
3614 RRETURN(MATCH_NOMATCH);
3615 }
3616 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3617 }
3618
3619 if (min == max) continue;
3620
3621 if (minimize)
3622 {
3623 for (fi = min;; fi++)
3624 {
3625 RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3626 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3627 if (fi >= max) RRETURN(MATCH_NOMATCH);
3628 if (eptr >= md->end_subject)
3629 {
3630 SCHECK_PARTIAL();
3631 RRETURN(MATCH_NOMATCH);
3632 }
3633 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3634 }
3635 /* Control never gets here */
3636 }
3637 else /* Maximize */
3638 {
3639 pp = eptr;
3640 for (i = min; i < max; i++)
3641 {
3642 if (eptr >= md->end_subject)
3643 {
3644 SCHECK_PARTIAL();
3645 break;
3646 }
3647 if (fc != RAWUCHARTEST(eptr)) break;
3648 eptr++;
3649 }
3650 if (possessive) continue; /* No backtracking */
3651 for (;;)
3652 {
3653 if (eptr == pp) goto TAIL_RECURSE;
3654 RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3655 eptr--;
3656 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3657 }
3658 /* Control never gets here */
3659 }
3660 }
3661 /* Control never gets here */
3662
3663 /* Match a negated single one-byte character. The character we are
3664 checking can be multibyte. */
3665
3666 case OP_NOT:
3667 case OP_NOTI:
3668 if (eptr >= md->end_subject)
3669 {
3670 SCHECK_PARTIAL();
3671 RRETURN(MATCH_NOMATCH);
3672 }
3673 #ifdef SUPPORT_UTF
3674 if (utf)
3675 {
3676 register pcre_uint32 ch, och;
3677
3678 ecode++;
3679 GETCHARINC(ch, ecode);
3680 GETCHARINC(c, eptr);
3681
3682 if (op == OP_NOT)
3683 {
3684 if (ch == c) RRETURN(MATCH_NOMATCH);
3685 }
3686 else
3687 {
3688 #ifdef SUPPORT_UCP
3689 if (ch > 127)
3690 och = UCD_OTHERCASE(ch);
3691 #else
3692 if (ch > 127)
3693 och = ch;
3694 #endif /* SUPPORT_UCP */
3695 else
3696 och = TABLE_GET(ch, md->fcc, ch);
3697 if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3698 }
3699 }
3700 else
3701 #endif
3702 {
3703 register pcre_uint32 ch = ecode[1];
3704 c = *eptr++;
3705 if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3706 RRETURN(MATCH_NOMATCH);
3707 ecode += 2;
3708 }
3709 break;
3710
3711 /* Match a negated single one-byte character repeatedly. This is almost a
3712 repeat of the code for a repeated single character, but I haven't found a
3713 nice way of commoning these up that doesn't require a test of the
3714 positive/negative option for each character match. Maybe that wouldn't add
3715 very much to the time taken, but character matching *is* what this is all
3716 about... */
3717
3718 case OP_NOTEXACT:
3719 case OP_NOTEXACTI:
3720 min = max = GET2(ecode, 1);
3721 ecode += 1 + IMM2_SIZE;
3722 goto REPEATNOTCHAR;
3723
3724 case OP_NOTUPTO:
3725 case OP_NOTUPTOI:
3726 case OP_NOTMINUPTO:
3727 case OP_NOTMINUPTOI:
3728 min = 0;
3729 max = GET2(ecode, 1);
3730 minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3731 ecode += 1 + IMM2_SIZE;
3732 goto REPEATNOTCHAR;
3733
3734 case OP_NOTPOSSTAR:
3735 case OP_NOTPOSSTARI:
3736 possessive = TRUE;
3737 min = 0;
3738 max = INT_MAX;
3739 ecode++;
3740 goto REPEATNOTCHAR;
3741
3742 case OP_NOTPOSPLUS:
3743 case OP_NOTPOSPLUSI:
3744 possessive = TRUE;
3745 min = 1;
3746 max = INT_MAX;
3747 ecode++;
3748 goto REPEATNOTCHAR;
3749
3750 case OP_NOTPOSQUERY:
3751 case OP_NOTPOSQUERYI:
3752 possessive = TRUE;
3753 min = 0;
3754 max = 1;
3755 ecode++;
3756 goto REPEATNOTCHAR;
3757
3758 case OP_NOTPOSUPTO:
3759 case OP_NOTPOSUPTOI:
3760 possessive = TRUE;
3761 min = 0;
3762 max = GET2(ecode, 1);
3763 ecode += 1 + IMM2_SIZE;
3764 goto REPEATNOTCHAR;
3765
3766 case OP_NOTSTAR:
3767 case OP_NOTSTARI:
3768 case OP_NOTMINSTAR:
3769 case OP_NOTMINSTARI:
3770 case OP_NOTPLUS:
3771 case OP_NOTPLUSI:
3772 case OP_NOTMINPLUS:
3773 case OP_NOTMINPLUSI:
3774 case OP_NOTQUERY:
3775 case OP_NOTQUERYI:
3776 case OP_NOTMINQUERY:
3777 case OP_NOTMINQUERYI:
3778 c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3779 minimize = (c & 1) != 0;
3780 min = rep_min[c]; /* Pick up values from tables; */
3781 max = rep_max[c]; /* zero for max => infinity */
3782 if (max == 0) max = INT_MAX;
3783
3784 /* Common code for all repeated single-byte matches. */
3785
3786 REPEATNOTCHAR:
3787 GETCHARINCTEST(fc, ecode);
3788
3789 /* The code is duplicated for the caseless and caseful cases, for speed,
3790 since matching characters is likely to be quite common. First, ensure the
3791 minimum number of matches are present. If min = max, continue at the same
3792 level without recursing. Otherwise, if minimizing, keep trying the rest of
3793 the expression and advancing one matching character if failing, up to the
3794 maximum. Alternatively, if maximizing, find the maximum number of
3795 characters and work backwards. */
3796
3797 DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3798 max, (char *)eptr));
3799
3800 if (op >= OP_NOTSTARI) /* Caseless */
3801 {
3802 #ifdef SUPPORT_UTF
3803 #ifdef SUPPORT_UCP
3804 if (utf && fc > 127)
3805 foc = UCD_OTHERCASE(fc);
3806 #else
3807 if (utf && fc > 127)
3808 foc = fc;
3809 #endif /* SUPPORT_UCP */
3810 else
3811 #endif /* SUPPORT_UTF */
3812 foc = TABLE_GET(fc, md->fcc, fc);
3813
3814 #ifdef SUPPORT_UTF
3815 if (utf)
3816 {
3817 register pcre_uint32 d;
3818 for (i = 1; i <= min; i++)
3819 {
3820 if (eptr >= md->end_subject)
3821 {
3822 SCHECK_PARTIAL();
3823 RRETURN(MATCH_NOMATCH);
3824 }
3825 GETCHARINC(d, eptr);
3826 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3827 }
3828 }
3829 else
3830 #endif /* SUPPORT_UTF */
3831 /* Not UTF mode */
3832 {
3833 for (i = 1; i <= min; i++)
3834 {
3835 if (eptr >= md->end_subject)
3836 {
3837 SCHECK_PARTIAL();
3838 RRETURN(MATCH_NOMATCH);
3839 }
3840 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3841 eptr++;
3842 }
3843 }
3844
3845 if (min == max) continue;
3846
3847 if (minimize)
3848 {
3849 #ifdef SUPPORT_UTF
3850 if (utf)
3851 {
3852 register pcre_uint32 d;
3853 for (fi = min;; fi++)
3854 {
3855 RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3856 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3857 if (fi >= max) RRETURN(MATCH_NOMATCH);
3858 if (eptr >= md->end_subject)
3859 {
3860 SCHECK_PARTIAL();
3861 RRETURN(MATCH_NOMATCH);
3862 }
3863 GETCHARINC(d, eptr);
3864 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3865 }
3866 }
3867 else
3868 #endif /*SUPPORT_UTF */
3869 /* Not UTF mode */
3870 {
3871 for (fi = min;; fi++)
3872 {
3873 RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3874 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3875 if (fi >= max) RRETURN(MATCH_NOMATCH);
3876 if (eptr >= md->end_subject)
3877 {
3878 SCHECK_PARTIAL();
3879 RRETURN(MATCH_NOMATCH);
3880 }
3881 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3882 eptr++;
3883 }
3884 }
3885 /* Control never gets here */
3886 }
3887
3888 /* Maximize case */
3889
3890 else
3891 {
3892 pp = eptr;
3893
3894 #ifdef SUPPORT_UTF
3895 if (utf)
3896 {
3897 register pcre_uint32 d;
3898 for (i = min; i < max; i++)
3899 {
3900 int len = 1;
3901 if (eptr >= md->end_subject)
3902 {
3903 SCHECK_PARTIAL();
3904 break;
3905 }
3906 GETCHARLEN(d, eptr, len);
3907 if (fc == d || (unsigned int)foc == d) break;
3908 eptr += len;
3909 }
3910 if (possessive) continue; /* No backtracking */
3911 for(;;)
3912 {
3913 if (eptr == pp) goto TAIL_RECURSE;
3914 RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3915 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3916 eptr--;
3917 BACKCHAR(eptr);
3918 }
3919 }
3920 else
3921 #endif /* SUPPORT_UTF */
3922 /* Not UTF mode */
3923 {
3924 for (i = min; i < max; i++)
3925 {
3926 if (eptr >= md->end_subject)
3927 {
3928 SCHECK_PARTIAL();
3929 break;
3930 }
3931 if (fc == *eptr || foc == *eptr) break;
3932 eptr++;
3933 }
3934 if (possessive) continue; /* No backtracking */
3935 for (;;)
3936 {
3937 if (eptr == pp) goto TAIL_RECURSE;
3938 RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3939 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3940 eptr--;
3941 }
3942 }
3943 /* Control never gets here */
3944 }
3945 }
3946
3947 /* Caseful comparisons */
3948
3949 else
3950 {
3951 #ifdef SUPPORT_UTF
3952 if (utf)
3953 {
3954 register pcre_uint32 d;
3955 for (i = 1; i <= min; i++)
3956 {
3957 if (eptr >= md->end_subject)
3958 {
3959 SCHECK_PARTIAL();
3960 RRETURN(MATCH_NOMATCH);
3961 }
3962 GETCHARINC(d, eptr);
3963 if (fc == d) RRETURN(MATCH_NOMATCH);
3964 }
3965 }
3966 else
3967 #endif
3968 /* Not UTF mode */
3969 {
3970 for (i = 1; i <= min; i++)
3971 {
3972 if (eptr >= md->end_subject)
3973 {
3974 SCHECK_PARTIAL();
3975 RRETURN(MATCH_NOMATCH);
3976 }
3977 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3978 }
3979 }
3980
3981 if (min == max) continue;
3982
3983 if (minimize)
3984 {
3985 #ifdef SUPPORT_UTF
3986 if (utf)
3987 {
3988 register pcre_uint32 d;
3989 for (fi = min;; fi++)
3990 {
3991 RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3992 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3993 if (fi >= max) RRETURN(MATCH_NOMATCH);
3994 if (eptr >= md->end_subject)
3995 {
3996 SCHECK_PARTIAL();
3997 RRETURN(MATCH_NOMATCH);
3998 }
3999 GETCHARINC(d, eptr);
4000 if (fc == d) RRETURN(MATCH_NOMATCH);
4001 }
4002 }
4003 else
4004 #endif
4005 /* Not UTF mode */
4006 {
4007 for (fi = min;; fi++)
4008 {
4009 RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
4010 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4011 if (fi >= max) RRETURN(MATCH_NOMATCH);
4012 if (eptr >= md->end_subject)
4013 {
4014 SCHECK_PARTIAL();
4015 RRETURN(MATCH_NOMATCH);
4016 }
4017 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
4018 }
4019 }
4020 /* Control never gets here */
4021 }
4022
4023 /* Maximize case */
4024
4025 else
4026 {
4027 pp = eptr;
4028
4029 #ifdef SUPPORT_UTF
4030 if (utf)
4031 {
4032 register pcre_uint32 d;
4033 for (i = min; i < max; i++)
4034 {
4035 int len = 1;
4036 if (eptr >= md->end_subject)
4037 {
4038 SCHECK_PARTIAL();
4039 break;
4040 }
4041 GETCHARLEN(d, eptr, len);
4042 if (fc == d) break;
4043 eptr += len;
4044 }
4045 if (possessive) continue; /* No backtracking */
4046 for(;;)
4047 {
4048 if (eptr == pp) goto TAIL_RECURSE;
4049 RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
4050 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4051 eptr--;
4052 BACKCHAR(eptr);
4053 }
4054 }
4055 else
4056 #endif
4057 /* Not UTF mode */
4058 {
4059 for (i = min; i < max; i++)
4060 {
4061 if (eptr >= md->end_subject)
4062 {
4063 SCHECK_PARTIAL();
4064 break;
4065 }
4066 if (fc == *eptr) break;
4067 eptr++;
4068 }
4069 if (possessive) continue; /* No backtracking */
4070 for (;;)
4071 {
4072 if (eptr == pp) goto TAIL_RECURSE;
4073 RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
4074 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4075 eptr--;
4076 }
4077 }
4078 /* Control never gets here */
4079 }
4080 }
4081 /* Control never gets here */
4082
4083 /* Match a single character type repeatedly; several different opcodes
4084 share code. This is very similar to the code for single characters, but we
4085 repeat it in the interests of efficiency. */
4086
4087 case OP_TYPEEXACT:
4088 min = max = GET2(ecode, 1);
4089 minimize = TRUE;
4090 ecode += 1 + IMM2_SIZE;
4091 goto REPEATTYPE;
4092
4093 case OP_TYPEUPTO:
4094 case OP_TYPEMINUPTO:
4095 min = 0;
4096 max = GET2(ecode, 1);
4097 minimize = *ecode == OP_TYPEMINUPTO;
4098 ecode += 1 + IMM2_SIZE;
4099 goto REPEATTYPE;
4100
4101 case OP_TYPEPOSSTAR:
4102 possessive = TRUE;
4103 min = 0;
4104 max = INT_MAX;
4105 ecode++;
4106 goto REPEATTYPE;
4107
4108 case OP_TYPEPOSPLUS:
4109 possessive = TRUE;
4110 min = 1;
4111 max = INT_MAX;
4112 ecode++;
4113 goto REPEATTYPE;
4114
4115 case OP_TYPEPOSQUERY:
4116 possessive = TRUE;
4117 min = 0;
4118 max = 1;
4119 ecode++;
4120 goto REPEATTYPE;
4121
4122 case OP_TYPEPOSUPTO:
4123 possessive = TRUE;
4124 min = 0;
4125 max = GET2(ecode, 1);
4126 ecode += 1 + IMM2_SIZE;
4127 goto REPEATTYPE;
4128
4129 case OP_TYPESTAR:
4130 case OP_TYPEMINSTAR:
4131 case OP_TYPEPLUS:
4132 case OP_TYPEMINPLUS:
4133 case OP_TYPEQUERY:
4134 case OP_TYPEMINQUERY:
4135 c = *ecode++ - OP_TYPESTAR;
4136 minimize = (c & 1) != 0;
4137 min = rep_min[c]; /* Pick up values from tables; */
4138 max = rep_max[c]; /* zero for max => infinity */
4139 if (max == 0) max = INT_MAX;
4140
4141 /* Common code for all repeated single character type matches. Note that
4142 in UTF-8 mode, '.' matches a character of any length, but for the other
4143 character types, the valid characters are all one-byte long. */
4144
4145 REPEATTYPE:
4146 ctype = *ecode++; /* Code for the character type */
4147
4148 #ifdef SUPPORT_UCP
4149 if (ctype == OP_PROP || ctype == OP_NOTPROP)
4150 {
4151 prop_fail_result = ctype == OP_NOTPROP;
4152 prop_type = *ecode++;
4153 prop_value = *ecode++;
4154 }
4155 else prop_type = -1;
4156 #endif
4157
4158 /* First, ensure the minimum number of matches are present. Use inline
4159 code for maximizing the speed, and do the type test once at the start
4160 (i.e. keep it out of the loop). Separate the UTF-8 code completely as that
4161 is tidier. Also separate the UCP code, which can be the same for both UTF-8
4162 and single-bytes. */
4163
4164 if (min > 0)
4165 {
4166 #ifdef SUPPORT_UCP
4167 if (prop_type >= 0)
4168 {
4169 switch(prop_type)
4170 {
4171 case PT_ANY:
4172 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4173 for (i = 1; i <= min; i++)
4174 {
4175 if (eptr >= md->end_subject)
4176 {
4177 SCHECK_PARTIAL();
4178 RRETURN(MATCH_NOMATCH);
4179 }
4180 GETCHARINCTEST(c, eptr);
4181 }
4182 break;
4183
4184 case PT_LAMP:
4185 for (i = 1; i <= min; i++)
4186 {
4187 int chartype;
4188 if (eptr >= md->end_subject)
4189 {
4190 SCHECK_PARTIAL();
4191 RRETURN(MATCH_NOMATCH);
4192 }
4193 GETCHARINCTEST(c, eptr);
4194 chartype = UCD_CHARTYPE(c);
4195 if ((chartype == ucp_Lu ||
4196 chartype == ucp_Ll ||
4197 chartype == ucp_Lt) == prop_fail_result)
4198 RRETURN(MATCH_NOMATCH);
4199 }
4200 break;
4201
4202 case PT_GC:
4203 for (i = 1; i <= min; i++)
4204 {
4205 if (eptr >= md->end_subject)
4206 {
4207 SCHECK_PARTIAL();
4208 RRETURN(MATCH_NOMATCH);
4209 }
4210 GETCHARINCTEST(c, eptr);
4211 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4212 RRETURN(MATCH_NOMATCH);
4213 }
4214 break;
4215
4216 case PT_PC:
4217 for (i = 1; i <= min; i++)
4218 {
4219 if (eptr >= md->end_subject)
4220 {
4221 SCHECK_PARTIAL();
4222 RRETURN(MATCH_NOMATCH);
4223 }
4224 GETCHARINCTEST(c, eptr);
4225 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4226 RRETURN(MATCH_NOMATCH);
4227 }
4228 break;
4229
4230 case PT_SC:
4231 for (i = 1; i <= min; i++)
4232 {
4233 if (eptr >= md->end_subject)
4234 {
4235 SCHECK_PARTIAL();
4236 RRETURN(MATCH_NOMATCH);
4237 }
4238 GETCHARINCTEST(c, eptr);
4239 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4240 RRETURN(MATCH_NOMATCH);
4241 }
4242 break;
4243
4244 case PT_ALNUM:
4245 for (i = 1; i <= min; i++)
4246 {
4247 int category;
4248 if (eptr >= md->end_subject)
4249 {
4250 SCHECK_PARTIAL();
4251 RRETURN(MATCH_NOMATCH);
4252 }
4253 GETCHARINCTEST(c, eptr);
4254 category = UCD_CATEGORY(c);
4255 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4256 RRETURN(MATCH_NOMATCH);
4257 }
4258 break;
4259
4260 case PT_SPACE: /* Perl space */
4261 for (i = 1; i <= min; i++)
4262 {
4263 if (eptr >= md->end_subject)
4264 {
4265 SCHECK_PARTIAL();
4266 RRETURN(MATCH_NOMATCH);
4267 }
4268 GETCHARINCTEST(c, eptr);
4269 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
4270 c == CHAR_FF || c == CHAR_CR)
4271 == prop_fail_result)
4272 RRETURN(MATCH_NOMATCH);
4273 }
4274 break;
4275
4276 case PT_PXSPACE: /* POSIX space */
4277 for (i = 1; i <= min; i++)
4278 {
4279 if (eptr >= md->end_subject)
4280 {
4281 SCHECK_PARTIAL();
4282 RRETURN(MATCH_NOMATCH);
4283 }
4284 GETCHARINCTEST(c, eptr);
4285 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
4286 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4287 == prop_fail_result)
4288 RRETURN(MATCH_NOMATCH);
4289 }
4290 break;
4291
4292 case PT_WORD:
4293 for (i = 1; i <= min; i++)
4294 {
4295 int category;
4296 if (eptr >= md->end_subject)
4297 {
4298 SCHECK_PARTIAL();
4299 RRETURN(MATCH_NOMATCH);
4300 }
4301 GETCHARINCTEST(c, eptr);
4302 category = UCD_CATEGORY(c);
4303 if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
4304 == prop_fail_result)
4305 RRETURN(MATCH_NOMATCH);
4306 }
4307 break;
4308
4309 case PT_CLIST:
4310 for (i = 1; i <= min; i++)
4311 {
4312 const pcre_uint32 *cp;
4313 if (eptr >= md->end_subject)
4314 {
4315 SCHECK_PARTIAL();
4316 RRETURN(MATCH_NOMATCH);
4317 }
4318 GETCHARINCTEST(c, eptr);
4319 cp = PRIV(ucd_caseless_sets) + prop_value;
4320 for (;;)
4321 {
4322 if (c < *cp)
4323 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
4324 if (c == *cp++)
4325 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
4326 }
4327 }
4328 break;
4329
4330 case PT_UCNC:
4331 for (i = 1; i <= min; i++)
4332 {
4333 if (eptr >= md->end_subject)
4334 {
4335 SCHECK_PARTIAL();
4336 RRETURN(MATCH_NOMATCH);
4337 }
4338 GETCHARINCTEST(c, eptr);
4339 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
4340 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
4341 c >= 0xe000) == prop_fail_result)
4342 RRETURN(MATCH_NOMATCH);
4343 }
4344 break;
4345
4346 /* This should not occur */
4347
4348 default:
4349 RRETURN(PCRE_ERROR_INTERNAL);
4350 }
4351 }
4352
4353 /* Match extended Unicode sequences. We will get here only if the
4354 support is in the binary; otherwise a compile-time error occurs. */
4355
4356 else if (ctype == OP_EXTUNI)
4357 {
4358 for (i = 1; i <= min; i++)
4359 {
4360 if (eptr >= md->end_subject)
4361 {
4362 SCHECK_PARTIAL();
4363 RRETURN(MATCH_NOMATCH);
4364 }
4365 else
4366 {
4367 int lgb, rgb;
4368 GETCHARINCTEST(c, eptr);
4369 lgb = UCD_GRAPHBREAK(c);
4370 while (eptr < md->end_subject)
4371 {
4372 int len = 1;
4373 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4374 rgb = UCD_GRAPHBREAK(c);
4375 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
4376 lgb = rgb;
4377 eptr += len;
4378 }
4379 }
4380 CHECK_PARTIAL();
4381 }
4382 }
4383
4384 else
4385 #endif /* SUPPORT_UCP */
4386
4387 /* Handle all other cases when the coding is UTF-8 */
4388
4389 #ifdef SUPPORT_UTF
4390 if (utf) switch(ctype)
4391 {
4392 case OP_ANY:
4393 for (i = 1; i <= min; i++)
4394 {
4395 if (eptr >= md->end_subject)
4396 {
4397 SCHECK_PARTIAL();
4398 RRETURN(MATCH_NOMATCH);
4399 }
4400 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4401 if (md->partial != 0 &&
4402 eptr + 1 >= md->end_subject &&
4403 NLBLOCK->nltype == NLTYPE_FIXED &&
4404 NLBLOCK->nllen == 2 &&
4405 RAWUCHAR(eptr) == NLBLOCK->nl[0])
4406 {
4407 md->hitend = TRUE;
4408 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4409 }
4410 eptr++;
4411 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4412 }
4413 break;
4414
4415 case OP_ALLANY:
4416 for (i = 1; i <= min; i++)
4417 {
4418 if (eptr >= md->end_subject)
4419 {
4420 SCHECK_PARTIAL();
4421 RRETURN(MATCH_NOMATCH);
4422 }
4423 eptr++;
4424 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4425 }
4426 break;
4427
4428 case OP_ANYBYTE:
4429 if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH);
4430 eptr += min;
4431 break;
4432
4433 case OP_ANYNL:
4434 for (i = 1; i <= min; i++)
4435 {
4436 if (eptr >= md->end_subject)
4437 {
4438 SCHECK_PARTIAL();
4439 RRETURN(MATCH_NOMATCH);
4440 }
4441 GETCHARINC(c, eptr);
4442 switch(c)
4443 {
4444 default: RRETURN(MATCH_NOMATCH);
4445
4446 case CHAR_CR:
4447 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
4448 break;
4449
4450 case CHAR_LF:
4451 break;
4452
4453 case CHAR_VT:
4454 case CHAR_FF:
4455 case CHAR_NEL:
4456 #ifndef EBCDIC
4457 case 0x2028:
4458 case 0x2029:
4459 #endif /* Not EBCDIC */
4460 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4461 break;
4462 }
4463 }
4464 break;
4465
4466 case OP_NOT_HSPACE:
4467 for (i = 1; i <= min; i++)
4468 {
4469 if (eptr >= md->end_subject)
4470 {
4471 SCHECK_PARTIAL();
4472 RRETURN(MATCH_NOMATCH);
4473 }
4474 GETCHARINC(c, eptr);
4475 switch(c)
4476 {
4477 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
4478 default: break;
4479 }
4480 }
4481 break;
4482
4483 case OP_HSPACE:
4484 for (i = 1; i <= min; i++)
4485 {
4486 if (eptr >= md->end_subject)
4487 {
4488 SCHECK_PARTIAL();
4489 RRETURN(MATCH_NOMATCH);
4490 }
4491 GETCHARINC(c, eptr);
4492 switch(c)
4493 {
4494 HSPACE_CASES: break; /* Byte and multibyte cases */
4495 default: RRETURN(MATCH_NOMATCH);
4496 }
4497 }
4498 break;
4499
4500 case OP_NOT_VSPACE:
4501 for (i = 1; i <= min; i++)
4502 {
4503 if (eptr >= md->end_subject)
4504 {
4505 SCHECK_PARTIAL();
4506 RRETURN(MATCH_NOMATCH);
4507 }
4508 GETCHARINC(c, eptr);
4509 switch(c)
4510 {
4511 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
4512 default: break;
4513 }
4514 }
4515 break;
4516
4517 case OP_VSPACE:
4518 for (i = 1; i <= min; i++)
4519 {
4520 if (eptr >= md->end_subject)
4521 {
4522 SCHECK_PARTIAL();
4523 RRETURN(MATCH_NOMATCH);
4524 }
4525 GETCHARINC(c, eptr);
4526 switch(c)
4527 {
4528 VSPACE_CASES: break;
4529 default: RRETURN(MATCH_NOMATCH);
4530 }
4531 }
4532 break;
4533
4534 case OP_NOT_DIGIT:
4535 for (i = 1; i <= min; i++)
4536 {
4537 if (eptr >= md->end_subject)
4538 {
4539 SCHECK_PARTIAL();
4540 RRETURN(MATCH_NOMATCH);
4541 }
4542 GETCHARINC(c, eptr);
4543 if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
4544 RRETURN(MATCH_NOMATCH);
4545 }
4546 break;
4547
4548 case OP_DIGIT:
4549 for (i = 1; i <= min; i++)
4550 {
4551 pcre_uint32 cc;
4552 if (eptr >= md->end_subject)
4553 {
4554 SCHECK_PARTIAL();
4555 RRETURN(MATCH_NOMATCH);
4556 }
4557 cc = RAWUCHAR(eptr);
4558 if (cc >= 128 || (md->ctypes[cc] & ctype_digit) == 0)
4559 RRETURN(MATCH_NOMATCH);
4560 eptr++;
4561 /* No need to skip more bytes - we know it's a 1-byte character */
4562 }
4563 break;
4564
4565 case OP_NOT_WHITESPACE:
4566 for (i = 1; i <= min; i++)
4567 {
4568 pcre_uint32 cc;
4569 if (eptr >= md->end_subject)
4570 {
4571 SCHECK_PARTIAL();
4572 RRETURN(MATCH_NOMATCH);
4573 }
4574 cc = RAWUCHAR(eptr);
4575 if (cc < 128 && (md->ctypes[cc] & ctype_space) != 0)
4576 RRETURN(MATCH_NOMATCH);
4577 eptr++;
4578 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4579 }
4580 break;
4581
4582 case OP_WHITESPACE:
4583 for (i = 1; i <= min; i++)
4584 {
4585 pcre_uint32 cc;
4586 if (eptr >= md->end_subject)
4587 {
4588 SCHECK_PARTIAL();
4589 RRETURN(MATCH_NOMATCH);
4590 }
4591 cc = RAWUCHAR(eptr);
4592 if (cc >= 128 || (md->ctypes[cc] & ctype_space) == 0)
4593 RRETURN(MATCH_NOMATCH);
4594 eptr++;
4595 /* No need to skip more bytes - we know it's a 1-byte character */
4596 }
4597 break;
4598
4599 case OP_NOT_WORDCHAR:
4600 for (i = 1; i <= min; i++)
4601 {
4602 pcre_uint32 cc;
4603 if (eptr >= md->end_subject)
4604 {
4605 SCHECK_PARTIAL();
4606 RRETURN(MATCH_NOMATCH);
4607 }
4608 cc = RAWUCHAR(eptr);
4609 if (cc < 128 && (md->ctypes[cc] & ctype_word) != 0)
4610 RRETURN(MATCH_NOMATCH);
4611 eptr++;
4612 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4613 }
4614 break;
4615
4616 case OP_WORDCHAR:
4617 for (i = 1; i <= min; i++)
4618 {
4619 pcre_uint32 cc;
4620 if (eptr >= md->end_subject)
4621 {
4622 SCHECK_PARTIAL();
4623 RRETURN(MATCH_NOMATCH);
4624 }
4625 cc = RAWUCHAR(eptr);
4626 if (cc >= 128 || (md->ctypes[cc] & ctype_word) == 0)
4627 RRETURN(MATCH_NOMATCH);
4628 eptr++;
4629 /* No need to skip more bytes - we know it's a 1-byte character */
4630 }
4631 break;
4632
4633 default:
4634 RRETURN(PCRE_ERROR_INTERNAL);
4635 } /* End switch(ctype) */
4636
4637 else
4638 #endif /* SUPPORT_UTF */
4639
4640 /* Code for the non-UTF-8 case for minimum matching of operators other
4641 than OP_PROP and OP_NOTPROP. */
4642
4643 switch(ctype)
4644 {
4645 case OP_ANY:
4646 for (i = 1; i <= min; i++)
4647 {
4648 if (eptr >= md->end_subject)
4649 {
4650 SCHECK_PARTIAL();
4651 RRETURN(MATCH_NOMATCH);
4652 }
4653 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4654 if (md->partial != 0 &&
4655 eptr + 1 >= md->end_subject &&
4656 NLBLOCK->nltype == NLTYPE_FIXED &&
4657 NLBLOCK->nllen == 2 &&
4658 *eptr == NLBLOCK->nl[0])
4659 {
4660 md->hitend = TRUE;
4661 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4662 }
4663 eptr++;
4664 }
4665 break;
4666
4667 case OP_ALLANY:
4668 if (eptr > md->end_subject - min)
4669 {
4670 SCHECK_PARTIAL();
4671 RRETURN(MATCH_NOMATCH);
4672 }
4673 eptr += min;
4674 break;
4675
4676 case OP_ANYBYTE:
4677 if (eptr > md->end_subject - min)
4678 {
4679 SCHECK_PARTIAL();
4680 RRETURN(MATCH_NOMATCH);
4681 }
4682 eptr += min;
4683 break;
4684
4685 case OP_ANYNL:
4686 for (i = 1; i <= min; i++)
4687 {
4688 if (eptr >= md->end_subject)
4689 {
4690 SCHECK_PARTIAL();
4691 RRETURN(MATCH_NOMATCH);
4692 }
4693 switch(*eptr++)
4694 {
4695 default: RRETURN(MATCH_NOMATCH);
4696
4697 case CHAR_CR:
4698 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
4699 break;
4700
4701 case CHAR_LF:
4702 break;
4703
4704 case CHAR_VT:
4705 case CHAR_FF:
4706 case CHAR_NEL:
4707 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4708 case 0x2028:
4709 case 0x2029:
4710 #endif
4711 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4712 break;
4713 }
4714 }
4715 break;
4716
4717 case OP_NOT_HSPACE:
4718 for (i = 1; i <= min; i++)
4719 {
4720 if (eptr >= md->end_subject)
4721 {
4722 SCHECK_PARTIAL();
4723 RRETURN(MATCH_NOMATCH);
4724 }
4725 switch(*eptr++)
4726 {
4727 default: break;
4728 HSPACE_BYTE_CASES:
4729 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4730 HSPACE_MULTIBYTE_CASES:
4731 #endif
4732 RRETURN(MATCH_NOMATCH);
4733 }
4734 }
4735 break;
4736
4737 case OP_HSPACE:
4738 for (i = 1; i <= min; i++)
4739 {
4740 if (eptr >= md->end_subject)
4741 {
4742 SCHECK_PARTIAL();
4743 RRETURN(MATCH_NOMATCH);
4744 }
4745 switch(*eptr++)
4746 {
4747 default: RRETURN(MATCH_NOMATCH);
4748 HSPACE_BYTE_CASES:
4749 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4750 HSPACE_MULTIBYTE_CASES:
4751 #endif
4752 break;
4753 }
4754 }
4755 break;
4756
4757 case OP_NOT_VSPACE:
4758 for (i = 1; i <= min; i++)
4759 {
4760 if (eptr >= md->end_subject)
4761 {
4762 SCHECK_PARTIAL();
4763 RRETURN(MATCH_NOMATCH);
4764 }
4765 switch(*eptr++)
4766 {
4767 VSPACE_BYTE_CASES:
4768 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4769 VSPACE_MULTIBYTE_CASES:
4770 #endif
4771 RRETURN(MATCH_NOMATCH);
4772 default: break;
4773 }
4774 }
4775 break;
4776
4777 case OP_VSPACE:
4778 for (i = 1; i <= min; i++)
4779 {
4780 if (eptr >= md->end_subject)
4781 {
4782 SCHECK_PARTIAL();
4783 RRETURN(MATCH_NOMATCH);
4784 }
4785 switch(*eptr++)
4786 {
4787 default: RRETURN(MATCH_NOMATCH);
4788 VSPACE_BYTE_CASES:
4789 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4790 VSPACE_MULTIBYTE_CASES:
4791 #endif
4792 break;
4793 }
4794 }
4795 break;
4796
4797 case OP_NOT_DIGIT:
4798 for (i = 1; i <= min; i++)
4799 {
4800 if (eptr >= md->end_subject)
4801 {
4802 SCHECK_PARTIAL();
4803 RRETURN(MATCH_NOMATCH);
4804 }
4805 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0)
4806 RRETURN(MATCH_NOMATCH);
4807 eptr++;
4808 }
4809 break;
4810
4811 case OP_DIGIT:
4812 for (i = 1; i <= min; i++)
4813 {
4814 if (eptr >= md->end_subject)
4815 {
4816 SCHECK_PARTIAL();
4817 RRETURN(MATCH_NOMATCH);
4818 }
4819 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0)
4820 RRETURN(MATCH_NOMATCH);
4821 eptr++;
4822 }
4823 break;
4824
4825 case OP_NOT_WHITESPACE:
4826 for (i = 1; i <= min; i++)
4827 {
4828 if (eptr >= md->end_subject)
4829 {
4830 SCHECK_PARTIAL();
4831 RRETURN(MATCH_NOMATCH);
4832 }
4833 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0)
4834 RRETURN(MATCH_NOMATCH);
4835 eptr++;
4836 }
4837 break;
4838
4839 case OP_WHITESPACE:
4840 for (i = 1; i <= min; i++)
4841 {
4842 if (eptr >= md->end_subject)
4843 {
4844 SCHECK_PARTIAL();
4845 RRETURN(MATCH_NOMATCH);
4846 }
4847 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0)
4848 RRETURN(MATCH_NOMATCH);
4849 eptr++;
4850 }
4851 break;
4852
4853 case OP_NOT_WORDCHAR:
4854 for (i = 1; i <= min; i++)
4855 {
4856 if (eptr >= md->end_subject)
4857 {
4858 SCHECK_PARTIAL();
4859 RRETURN(MATCH_NOMATCH);
4860 }
4861 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0)
4862 RRETURN(MATCH_NOMATCH);
4863 eptr++;
4864 }
4865 break;
4866
4867 case OP_WORDCHAR:
4868 for (i = 1; i <= min; i++)
4869 {
4870 if (eptr >= md->end_subject)
4871 {
4872 SCHECK_PARTIAL();
4873 RRETURN(MATCH_NOMATCH);
4874 }
4875 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0)
4876 RRETURN(MATCH_NOMATCH);
4877 eptr++;
4878 }
4879 break;
4880
4881 default:
4882 RRETURN(PCRE_ERROR_INTERNAL);
4883 }
4884 }
4885
4886 /* If min = max, continue at the same level without recursing */
4887
4888 if (min == max) continue;
4889
4890 /* If minimizing, we have to test the rest of the pattern before each
4891 subsequent match. Again, separate the UTF-8 case for speed, and also
4892 separate the UCP cases. */
4893
4894 if (minimize)
4895 {
4896 #ifdef SUPPORT_UCP
4897 if (prop_type >= 0)
4898 {
4899 switch(prop_type)
4900 {
4901 case PT_ANY:
4902 for (fi = min;; fi++)
4903 {
4904 RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4905 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4906 if (fi >= max) RRETURN(MATCH_NOMATCH);
4907 if (eptr >= md->end_subject)
4908 {
4909 SCHECK_PARTIAL();
4910 RRETURN(MATCH_NOMATCH);
4911 }
4912 GETCHARINCTEST(c, eptr);
4913 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4914 }
4915 /* Control never gets here */
4916
4917 case PT_LAMP:
4918 for (fi = min;; fi++)
4919 {
4920 int chartype;
4921 RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4922 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4923 if (fi >= max) RRETURN(MATCH_NOMATCH);
4924 if (eptr >= md->end_subject)
4925 {
4926 SCHECK_PARTIAL();
4927 RRETURN(MATCH_NOMATCH);
4928 }
4929 GETCHARINCTEST(c, eptr);
4930 chartype = UCD_CHARTYPE(c);
4931 if ((chartype == ucp_Lu ||
4932 chartype == ucp_Ll ||
4933 chartype == ucp_Lt) == prop_fail_result)
4934 RRETURN(MATCH_NOMATCH);
4935 }
4936 /* Control never gets here */
4937
4938 case PT_GC:
4939 for (fi = min;; fi++)
4940 {
4941 RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4942 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4943 if (fi >= max) RRETURN(MATCH_NOMATCH);
4944 if (eptr >= md->end_subject)
4945 {
4946 SCHECK_PARTIAL();
4947 RRETURN(MATCH_NOMATCH);
4948 }
4949 GETCHARINCTEST(c, eptr);
4950 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4951 RRETURN(MATCH_NOMATCH);
4952 }
4953 /* Control never gets here */
4954
4955 case PT_PC:
4956 for (fi = min;; fi++)
4957 {
4958 RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4959 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4960 if (fi >= max) RRETURN(MATCH_NOMATCH);
4961 if (eptr >= md->end_subject)
4962 {
4963 SCHECK_PARTIAL();
4964 RRETURN(MATCH_NOMATCH);
4965 }
4966 GETCHARINCTEST(c, eptr);
4967 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4968 RRETURN(MATCH_NOMATCH);
4969 }
4970 /* Control never gets here */
4971
4972 case PT_SC:
4973 for (fi = min;; fi++)
4974 {
4975 RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
4976 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4977 if (fi >= max) RRETURN(MATCH_NOMATCH);
4978 if (eptr >= md->end_subject)
4979 {
4980 SCHECK_PARTIAL();
4981 RRETURN(MATCH_NOMATCH);
4982 }
4983 GETCHARINCTEST(c, eptr);
4984 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4985 RRETURN(MATCH_NOMATCH);
4986 }
4987 /* Control never gets here */
4988
4989 case PT_ALNUM:
4990 for (fi = min;; fi++)
4991 {
4992 int category;
4993 RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4994 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4995 if (fi >= max) RRETURN(MATCH_NOMATCH);
4996 if (eptr >= md->end_subject)
4997 {
4998 SCHECK_PARTIAL();
4999 RRETURN(MATCH_NOMATCH);
5000 }
5001 GETCHARINCTEST(c, eptr);
5002 category = UCD_CATEGORY(c);
5003 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5004 RRETURN(MATCH_NOMATCH);
5005 }
5006 /* Control never gets here */
5007
5008 case PT_SPACE: /* Perl space */
5009 for (fi = min;; fi++)
5010 {
5011 RMATCH(eptr, ecode, offset_top, md, eptrb, RM60);
5012 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5013 if (fi >= max) RRETURN(MATCH_NOMATCH);
5014 if (eptr >= md->end_subject)
5015 {
5016 SCHECK_PARTIAL();
5017 RRETURN(MATCH_NOMATCH);
5018 }
5019 GETCHARINCTEST(c, eptr);
5020 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5021 c == CHAR_FF || c == CHAR_CR)
5022 == prop_fail_result)
5023 RRETURN(MATCH_NOMATCH);
5024 }
5025 /* Control never gets here */
5026
5027 case PT_PXSPACE: /* POSIX space */
5028 for (fi = min;; fi++)
5029 {
5030 RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
5031 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5032 if (fi >= max) RRETURN(MATCH_NOMATCH);
5033 if (eptr >= md->end_subject)
5034 {
5035 SCHECK_PARTIAL();
5036 RRETURN(MATCH_NOMATCH);
5037 }
5038 GETCHARINCTEST(c, eptr);
5039 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5040 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5041 == prop_fail_result)
5042 RRETURN(MATCH_NOMATCH);
5043 }
5044 /* Control never gets here */
5045
5046 case PT_WORD:
5047 for (fi = min;; fi++)
5048 {
5049 int category;
5050 RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
5051 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5052 if (fi >= max) RRETURN(MATCH_NOMATCH);
5053 if (eptr >= md->end_subject)
5054 {
5055 SCHECK_PARTIAL();
5056 RRETURN(MATCH_NOMATCH);
5057 }
5058 GETCHARINCTEST(c, eptr);
5059 category = UCD_CATEGORY(c);
5060 if ((category == ucp_L ||
5061 category == ucp_N ||
5062 c == CHAR_UNDERSCORE)
5063 == prop_fail_result)
5064 RRETURN(MATCH_NOMATCH);
5065 }
5066 /* Control never gets here */
5067
5068 case PT_CLIST:
5069 for (fi = min;; fi++)
5070 {
5071 const pcre_uint32 *cp;
5072 RMATCH(eptr, ecode, offset_top, md, eptrb, RM67);
5073 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5074 if (fi >= max) RRETURN(MATCH_NOMATCH);
5075 if (eptr >= md->end_subject)
5076 {
5077 SCHECK_PARTIAL();
5078 RRETURN(MATCH_NOMATCH);
5079 }
5080 GETCHARINCTEST(c, eptr);
5081 cp = PRIV(ucd_caseless_sets) + prop_value;
5082 for (;;)
5083 {
5084 if (c < *cp)
5085 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
5086 if (c == *cp++)
5087 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
5088 }
5089 }
5090 /* Control never gets here */
5091
5092 case PT_UCNC:
5093 for (fi = min;; fi++)
5094 {
5095 RMATCH(eptr, ecode, offset_top, md, eptrb, RM68);
5096 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5097 if (fi >= max) RRETURN(MATCH_NOMATCH);
5098 if (eptr >= md->end_subject)
5099 {
5100 SCHECK_PARTIAL();
5101 RRETURN(MATCH_NOMATCH);
5102 }
5103 GETCHARINCTEST(c, eptr);
5104 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5105 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5106 c >= 0xe000) == prop_fail_result)
5107 RRETURN(MATCH_NOMATCH);
5108 }
5109 /* Control never gets here */
5110
5111 /* This should never occur */
5112 default:
5113 RRETURN(PCRE_ERROR_INTERNAL);
5114 }
5115 }
5116
5117 /* Match extended Unicode sequences. We will get here only if the
5118 support is in the binary; otherwise a compile-time error occurs. */
5119
5120 else if (ctype == OP_EXTUNI)
5121 {
5122 for (fi = min;; fi++)
5123 {
5124 RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
5125 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5126 if (fi >= max) RRETURN(MATCH_NOMATCH);
5127 if (eptr >= md->end_subject)
5128 {
5129 SCHECK_PARTIAL();
5130 RRETURN(MATCH_NOMATCH);
5131 }
5132 else
5133 {
5134 int lgb, rgb;
5135 GETCHARINCTEST(c, eptr);
5136 lgb = UCD_GRAPHBREAK(c);
5137 while (eptr < md->end_subject)
5138 {
5139 int len = 1;
5140 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5141 rgb = UCD_GRAPHBREAK(c);
5142 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5143 lgb = rgb;
5144 eptr += len;
5145 }
5146 }
5147 CHECK_PARTIAL();
5148 }
5149 }
5150 else
5151 #endif /* SUPPORT_UCP */
5152
5153 #ifdef SUPPORT_UTF
5154 if (utf)
5155 {
5156 for (fi = min;; fi++)
5157 {
5158 RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
5159 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5160 if (fi >= max) RRETURN(MATCH_NOMATCH);
5161 if (eptr >= md->end_subject)
5162 {
5163 SCHECK_PARTIAL();
5164 RRETURN(MATCH_NOMATCH);
5165 }
5166 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5167 RRETURN(MATCH_NOMATCH);
5168 GETCHARINC(c, eptr);
5169 switch(ctype)
5170 {
5171 case OP_ANY: /* This is the non-NL case */
5172 if (md->partial != 0 && /* Take care with CRLF partial */
5173 eptr >= md->end_subject &&
5174 NLBLOCK->nltype == NLTYPE_FIXED &&
5175 NLBLOCK->nllen == 2 &&
5176 c == NLBLOCK->nl[0])
5177 {
5178 md->hitend = TRUE;
5179 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5180 }
5181 break;
5182
5183 case OP_ALLANY:
5184 case OP_ANYBYTE:
5185 break;
5186
5187 case OP_ANYNL:
5188 switch(c)
5189 {
5190 default: RRETURN(MATCH_NOMATCH);
5191 case CHAR_CR:
5192 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
5193 break;
5194
5195 case CHAR_LF:
5196 break;
5197
5198 case CHAR_VT:
5199 case CHAR_FF:
5200 case CHAR_NEL:
5201 #ifndef EBCDIC
5202 case 0x2028:
5203 case 0x2029:
5204 #endif /* Not EBCDIC */
5205 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5206 break;
5207 }
5208 break;
5209
5210 case OP_NOT_HSPACE:
5211 switch(c)
5212 {
5213 HSPACE_CASES: RRETURN(MATCH_NOMATCH);
5214 default: break;
5215 }
5216 break;
5217
5218 case OP_HSPACE:
5219 switch(c)
5220 {
5221 HSPACE_CASES: break;
5222 default: RRETURN(MATCH_NOMATCH);
5223 }
5224 break;
5225
5226 case OP_NOT_VSPACE:
5227 switch(c)
5228 {
5229 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
5230 default: break;
5231 }
5232 break;
5233
5234 case OP_VSPACE:
5235 switch(c)
5236 {
5237 VSPACE_CASES: break;
5238 default: RRETURN(MATCH_NOMATCH);
5239 }
5240 break;
5241
5242 case OP_NOT_DIGIT:
5243 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
5244 RRETURN(MATCH_NOMATCH);
5245 break;
5246
5247 case OP_DIGIT:
5248 if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0)
5249 RRETURN(MATCH_NOMATCH);
5250 break;
5251
5252 case OP_NOT_WHITESPACE:
5253 if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
5254 RRETURN(MATCH_NOMATCH);
5255 break;
5256
5257 case OP_WHITESPACE:
5258 if (c >= 256 || (md->ctypes[c] & ctype_space) == 0)
5259 RRETURN(MATCH_NOMATCH);
5260 break;
5261
5262 case OP_NOT_WORDCHAR:
5263 if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
5264 RRETURN(MATCH_NOMATCH);
5265 break;
5266
5267 case OP_WORDCHAR:
5268 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0)
5269 RRETURN(MATCH_NOMATCH);
5270 break;
5271
5272 default:
5273 RRETURN(PCRE_ERROR_INTERNAL);
5274 }
5275 }
5276 }
5277 else
5278 #endif
5279 /* Not UTF mode */
5280 {
5281 for (fi = min;; fi++)
5282 {
5283 RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
5284 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5285 if (fi >= max) RRETURN(MATCH_NOMATCH);
5286 if (eptr >= md->end_subject)
5287 {
5288 SCHECK_PARTIAL();
5289 RRETURN(MATCH_NOMATCH);
5290 }
5291 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5292 RRETURN(MATCH_NOMATCH);
5293 c = *eptr++;
5294 switch(ctype)
5295 {
5296 case OP_ANY: /* This is the non-NL case */
5297 if (md->partial != 0 && /* Take care with CRLF partial */
5298 eptr >= md->end_subject &&
5299 NLBLOCK->nltype == NLTYPE_FIXED &&
5300 NLBLOCK->nllen == 2 &&
5301 c == NLBLOCK->nl[0])
5302 {
5303 md->hitend = TRUE;
5304 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5305 }
5306 break;
5307
5308 case OP_ALLANY:
5309 case OP_ANYBYTE:
5310 break;
5311
5312 case OP_ANYNL:
5313 switch(c)
5314 {
5315 default: RRETURN(MATCH_NOMATCH);
5316 case CHAR_CR:
5317 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
5318 break;
5319
5320 case CHAR_LF:
5321 break;
5322
5323 case CHAR_VT:
5324 case CHAR_FF:
5325 case CHAR_NEL:
5326 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5327 case 0x2028:
5328 case 0x2029:
5329 #endif
5330 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5331 break;
5332 }
5333 break;
5334
5335 case OP_NOT_HSPACE:
5336 switch(c)
5337 {
5338 default: break;
5339 HSPACE_BYTE_CASES:
5340 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5341 HSPACE_MULTIBYTE_CASES:
5342 #endif
5343 RRETURN(MATCH_NOMATCH);
5344 }
5345 break;
5346
5347 case OP_HSPACE:
5348 switch(c)
5349 {
5350 default: RRETURN(MATCH_NOMATCH);
5351 HSPACE_BYTE_CASES:
5352 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5353 HSPACE_MULTIBYTE_CASES:
5354 #endif
5355 break;
5356 }
5357 break;
5358
5359 case OP_NOT_VSPACE:
5360 switch(c)
5361 {
5362 default: break;
5363 VSPACE_BYTE_CASES:
5364 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5365 VSPACE_MULTIBYTE_CASES:
5366 #endif
5367 RRETURN(MATCH_NOMATCH);
5368 }
5369 break;
5370
5371 case OP_VSPACE:
5372 switch(c)
5373 {
5374 default: RRETURN(MATCH_NOMATCH);
5375 VSPACE_BYTE_CASES:
5376 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5377 VSPACE_MULTIBYTE_CASES:
5378 #endif
5379 break;
5380 }
5381 break;
5382
5383 case OP_NOT_DIGIT:
5384 if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
5385 break;
5386
5387 case OP_DIGIT:
5388 if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
5389 break;
5390
5391 case OP_NOT_WHITESPACE:
5392 if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
5393 break;
5394
5395 case OP_WHITESPACE:
5396 if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
5397 break;
5398
5399 case OP_NOT_WORDCHAR:
5400 if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH);
5401 break;
5402
5403 case OP_WORDCHAR:
5404 if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH);
5405 break;
5406
5407 default:
5408 RRETURN(PCRE_ERROR_INTERNAL);
5409 }
5410 }
5411 }
5412 /* Control never gets here */
5413 }
5414
5415 /* If maximizing, it is worth using inline code for speed, doing the type
5416 test once at the start (i.e. keep it out of the loop). Again, keep the
5417 UTF-8 and UCP stuff separate. */
5418
5419 else
5420 {
5421 pp = eptr; /* Remember where we started */
5422
5423 #ifdef SUPPORT_UCP
5424 if (prop_type >= 0)
5425 {
5426 switch(prop_type)
5427 {
5428 case PT_ANY:
5429 for (i = min; i < max; i++)
5430 {
5431 int len = 1;
5432 if (eptr >= md->end_subject)
5433 {
5434 SCHECK_PARTIAL();
5435 break;
5436 }
5437 GETCHARLENTEST(c, eptr, len);
5438 if (prop_fail_result) break;
5439 eptr+= len;
5440 }
5441 break;
5442
5443 case PT_LAMP:
5444 for (i = min; i < max; i++)
5445 {
5446 int chartype;
5447 int len = 1;
5448 if (eptr >= md->end_subject)
5449 {
5450 SCHECK_PARTIAL();
5451 break;
5452 }
5453 GETCHARLENTEST(c, eptr, len);
5454 chartype = UCD_CHARTYPE(c);
5455 if ((chartype == ucp_Lu ||
5456 chartype == ucp_Ll ||
5457 chartype == ucp_Lt) == prop_fail_result)
5458 break;
5459 eptr+= len;
5460 }
5461 break;
5462
5463 case PT_GC:
5464 for (i = min; i < max; i++)
5465 {
5466 int len = 1;
5467 if (eptr >= md->end_subject)
5468 {
5469 SCHECK_PARTIAL();
5470 break;
5471 }
5472 GETCHARLENTEST(c, eptr, len);
5473 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
5474 eptr+= len;
5475 }
5476 break;
5477
5478 case PT_PC:
5479 for (i = min; i < max; i++)
5480 {
5481 int len = 1;
5482 if (eptr >= md->end_subject)
5483 {
5484 SCHECK_PARTIAL();
5485 break;
5486 }
5487 GETCHARLENTEST(c, eptr, len);
5488 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
5489 eptr+= len;
5490 }
5491 break;
5492
5493 case PT_SC:
5494 for (i = min; i < max; i++)
5495 {
5496 int len = 1;
5497 if (eptr >= md->end_subject)
5498 {
5499 SCHECK_PARTIAL();
5500 break;
5501 }
5502 GETCHARLENTEST(c, eptr, len);
5503 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
5504 eptr+= len;
5505 }
5506 break;
5507
5508 case PT_ALNUM:
5509 for (i = min; i < max; i++)
5510 {
5511 int category;
5512 int len = 1;
5513 if (eptr >= md->end_subject)
5514 {
5515 SCHECK_PARTIAL();
5516 break;
5517 }
5518 GETCHARLENTEST(c, eptr, len);
5519 category = UCD_CATEGORY(c);
5520 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5521 break;
5522 eptr+= len;
5523 }
5524 break;
5525
5526 case PT_SPACE: /* Perl space */
5527 for (i = min; i < max; i++)
5528 {
5529 int len = 1;
5530 if (eptr >= md->end_subject)
5531 {
5532 SCHECK_PARTIAL();
5533 break;
5534 }
5535 GETCHARLENTEST(c, eptr, len);
5536 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5537 c == CHAR_FF || c == CHAR_CR)
5538 == prop_fail_result)
5539 break;
5540 eptr+= len;
5541 }
5542 break;
5543
5544 case PT_PXSPACE: /* POSIX space */
5545 for (i = min; i < max; i++)
5546 {
5547 int len = 1;
5548 if (eptr >= md->end_subject)
5549 {
5550 SCHECK_PARTIAL();
5551 break;
5552 }
5553 GETCHARLENTEST(c, eptr, len);
5554 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5555 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5556 == prop_fail_result)
5557 break;
5558 eptr+= len;
5559 }
5560 break;
5561
5562 case PT_WORD:
5563 for (i = min; i < max; i++)
5564 {
5565 int category;
5566 int len = 1;
5567 if (eptr >= md->end_subject)
5568 {
5569 SCHECK_PARTIAL();
5570 break;
5571 }
5572 GETCHARLENTEST(c, eptr, len);
5573 category = UCD_CATEGORY(c);
5574 if ((category == ucp_L || category == ucp_N ||
5575 c == CHAR_UNDERSCORE) == prop_fail_result)
5576 break;
5577 eptr+= len;
5578 }
5579 break;
5580
5581 case PT_CLIST:
5582 for (i = min; i < max; i++)
5583 {
5584 const pcre_uint32 *cp;
5585 int len = 1;
5586 if (eptr >= md->end_subject)
5587 {
5588 SCHECK_PARTIAL();
5589 break;
5590 }
5591 GETCHARLENTEST(c, eptr, len);
5592 cp = PRIV(ucd_caseless_sets) + prop_value;
5593 for (;;)
5594 {
5595 if (c < *cp)
5596 { if (prop_fail_result) break; else goto GOT_MAX; }
5597 if (c == *cp++)
5598 { if (prop_fail_result) goto GOT_MAX; else break; }
5599 }
5600 eptr += len;
5601 }
5602 GOT_MAX:
5603 break;
5604
5605 case PT_UCNC:
5606 for (i = min; i < max; i++)
5607 {
5608 int len = 1;
5609 if (eptr >= md->end_subject)
5610 {
5611 SCHECK_PARTIAL();
5612 break;
5613 }
5614 GETCHARLENTEST(c, eptr, len);
5615 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5616 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5617 c >= 0xe000) == prop_fail_result)
5618 break;
5619 eptr += len;
5620 }
5621 break;
5622
5623 default:
5624 RRETURN(PCRE_ERROR_INTERNAL);
5625 }
5626
5627 /* eptr is now past the end of the maximum run */
5628
5629 if (possessive) continue; /* No backtracking */
5630 for(;;)
5631 {
5632 if (eptr == pp) goto TAIL_RECURSE;
5633 RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5634 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5635 eptr--;
5636 if (utf) BACKCHAR(eptr);
5637 }
5638 }
5639
5640 /* Match extended Unicode grapheme clusters. We will get here only if the
5641 support is in the binary; otherwise a compile-time error occurs. */
5642
5643 else if (ctype == OP_EXTUNI)
5644 {
5645 for (i = min; i < max; i++)
5646 {
5647 if (eptr >= md->end_subject)
5648 {
5649 SCHECK_PARTIAL();
5650 break;
5651 }
5652 else
5653 {
5654 int lgb, rgb;
5655 GETCHARINCTEST(c, eptr);
5656 lgb = UCD_GRAPHBREAK(c);
5657 while (eptr < md->end_subject)
5658 {
5659 int len = 1;
5660 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5661 rgb = UCD_GRAPHBREAK(c);
5662 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5663 lgb = rgb;
5664 eptr += len;
5665 }
5666 }
5667 CHECK_PARTIAL();
5668 }
5669
5670 /* eptr is now past the end of the maximum run */
5671
5672 if (possessive) continue; /* No backtracking */
5673
5674 for(;;)
5675 {
5676 int lgb, rgb;
5677 PCRE_PUCHAR fptr;
5678
5679 if (eptr == pp) goto TAIL_RECURSE; /* At start of char run */
5680 RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5681 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5682
5683 /* Backtracking over an extended grapheme cluster involves inspecting
5684 the previous two characters (if present) to see if a break is
5685 permitted between them. */
5686
5687 eptr--;
5688 if (!utf) c = *eptr; else
5689 {
5690 BACKCHAR(eptr);
5691 GETCHAR(c, eptr);
5692 }
5693 rgb = UCD_GRAPHBREAK(c);
5694
5695 for (;;)
5696 {
5697 if (eptr == pp) goto TAIL_RECURSE; /* At start of char run */
5698 fptr = eptr - 1;
5699 if (!utf) c = *fptr; else
5700 {
5701 BACKCHAR(fptr);
5702 GETCHAR(c, fptr);
5703 }
5704 lgb = UCD_GRAPHBREAK(c);
5705 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5706 eptr = fptr;
5707 rgb = lgb;
5708 }
5709 }
5710 }
5711
5712 else
5713 #endif /* SUPPORT_UCP */
5714
5715 #ifdef SUPPORT_UTF
5716 if (utf)
5717 {
5718 switch(ctype)
5719 {
5720 case OP_ANY:
5721 if (max < INT_MAX)
5722 {
5723 for (i = min; i < max; i++)
5724 {
5725 if (eptr >= md->end_subject)
5726 {
5727 SCHECK_PARTIAL();
5728 break;
5729 }
5730 if (IS_NEWLINE(eptr)) break;
5731 if (md->partial != 0 && /* Take care with CRLF partial */
5732 eptr + 1 >= md->end_subject &&
5733 NLBLOCK->nltype == NLTYPE_FIXED &&
5734 NLBLOCK->nllen == 2 &&
5735 RAWUCHAR(eptr) == NLBLOCK->nl[0])
5736 {
5737 md->hitend = TRUE;
5738 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5739 }
5740 eptr++;
5741 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5742 }
5743 }
5744
5745 /* Handle unlimited UTF-8 repeat */
5746
5747 else
5748 {
5749 for (i = min; i < max; i++)
5750 {
5751 if (eptr >= md->end_subject)
5752 {
5753 SCHECK_PARTIAL();
5754 break;
5755 }
5756 if (IS_NEWLINE(eptr)) break;
5757 if (md->partial != 0 && /* Take care with CRLF partial */
5758 eptr + 1 >= md->end_subject &&
5759 NLBLOCK->nltype == NLTYPE_FIXED &&
5760 NLBLOCK->nllen == 2 &&
5761 RAWUCHAR(eptr) == NLBLOCK->nl[0])
5762 {
5763 md->hitend = TRUE;
5764 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5765 }
5766 eptr++;
5767 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5768 }
5769 }
5770 break;
5771
5772 case OP_ALLANY:
5773 if (max < INT_MAX)
5774 {
5775 for (i = min; i < max; i++)
5776 {
5777 if (eptr >= md->end_subject)
5778 {
5779 SCHECK_PARTIAL();
5780 break;
5781 }
5782 eptr++;
5783 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5784 }
5785 }
5786 else
5787 {
5788 eptr = md->end_subject; /* Unlimited UTF-8 repeat */
5789 SCHECK_PARTIAL();
5790 }
5791 break;
5792
5793 /* The byte case is the same as non-UTF8 */
5794
5795 case OP_ANYBYTE:
5796 c = max - min;
5797 if (c > (unsigned int)(md->end_subject - eptr))
5798 {
5799 eptr = md->end_subject;
5800 SCHECK_PARTIAL();
5801 }
5802 else eptr += c;
5803 break;
5804
5805 case OP_ANYNL:
5806 for (i = min; i < max; i++)
5807 {
5808 int len = 1;
5809 if (eptr >= md->end_subject)
5810 {
5811 SCHECK_PARTIAL();
5812 break;
5813 }
5814 GETCHARLEN(c, eptr, len);
5815 if (c == CHAR_CR)
5816 {
5817 if (++eptr >= md->end_subject) break;
5818 if (RAWUCHAR(eptr) == CHAR_LF) eptr++;
5819 }
5820 else
5821 {
5822 if (c != CHAR_LF &&
5823 (md->bsr_anycrlf ||
5824 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
5825 #ifndef EBCDIC
5826 && c != 0x2028 && c != 0x2029
5827 #endif /* Not EBCDIC */
5828 )))
5829 break;
5830 eptr += len;
5831 }
5832 }
5833 break;
5834
5835 case OP_NOT_HSPACE:
5836 case OP_HSPACE:
5837 for (i = min; i < max; i++)
5838 {
5839 BOOL gotspace;
5840 int len = 1;
5841 if (eptr >= md->end_subject)
5842 {
5843 SCHECK_PARTIAL();
5844 break;
5845 }
5846 GETCHARLEN(c, eptr, len);
5847 switch(c)
5848 {
5849 HSPACE_CASES: gotspace = TRUE; break;
5850 default: gotspace = FALSE; break;
5851 }
5852 if (gotspace == (ctype == OP_NOT_HSPACE)) break;
5853 eptr += len;
5854 }
5855 break;
5856
5857 case OP_NOT_VSPACE:
5858 case OP_VSPACE:
5859 for (i = min; i < max; i++)
5860 {
5861 BOOL gotspace;
5862 int len = 1;
5863 if (eptr >= md->end_subject)
5864 {
5865 SCHECK_PARTIAL();
5866 break;
5867 }
5868 GETCHARLEN(c, eptr, len);
5869 switch(c)
5870 {
5871 VSPACE_CASES: gotspace = TRUE; break;
5872 default: gotspace = FALSE; break;
5873 }
5874 if (gotspace == (ctype == OP_NOT_VSPACE)) break;
5875 eptr += len;
5876 }
5877 break;
5878
5879 case OP_NOT_DIGIT:
5880 for (i = min; i < max; i++)
5881 {
5882 int len = 1;
5883 if (eptr >= md->end_subject)
5884 {
5885 SCHECK_PARTIAL();
5886 break;
5887 }
5888 GETCHARLEN(c, eptr, len);
5889 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
5890 eptr+= len;
5891 }
5892 break;
5893
5894 case OP_DIGIT:
5895 for (i = min; i < max; i++)
5896 {
5897 int len = 1;
5898 if (eptr >= md->end_subject)
5899 {
5900 SCHECK_PARTIAL();
5901 break;
5902 }
5903 GETCHARLEN(c, eptr, len);
5904 if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break;
5905 eptr+= len;
5906 }
5907 break;
5908
5909 case OP_NOT_WHITESPACE:
5910 for (i = min; i < max; i++)
5911 {
5912 int len = 1;
5913 if (eptr >= md->end_subject)
5914 {
5915 SCHECK_PARTIAL();
5916 break;
5917 }
5918 GETCHARLEN(c, eptr, len);
5919 if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
5920 eptr+= len;
5921 }
5922 break;
5923
5924 case OP_WHITESPACE:
5925 for (i = min; i < max; i++)
5926 {
5927 int len = 1;
5928 if (eptr >= md->end_subject)
5929 {
5930 SCHECK_PARTIAL();
5931 break;
5932 }
5933 GETCHARLEN(c, eptr, len);
5934 if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break;
5935 eptr+= len;
5936 }
5937 break;
5938
5939 case OP_NOT_WORDCHAR:
5940 for (i = min; i < max; i++)
5941 {
5942 int len = 1;
5943 if (eptr >= md->end_subject)
5944 {
5945 SCHECK_PARTIAL();
5946 break;
5947 }
5948 GETCHARLEN(c, eptr, len);
5949 if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
5950 eptr+= len;
5951 }
5952 break;
5953
5954 case OP_WORDCHAR:
5955 for (i = min; i < max; i++)
5956 {
5957 int len = 1;
5958 if (eptr >= md->end_subject)
5959 {
5960 SCHECK_PARTIAL();
5961 break;
5962 }
5963 GETCHARLEN(c, eptr, len);
5964 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break;
5965 eptr+= len;
5966 }
5967 break;
5968
5969 default:
5970 RRETURN(PCRE_ERROR_INTERNAL);
5971 }
5972
5973 if (possessive) continue; /* No backtracking */
5974 for(;;)
5975 {
5976 if (eptr == pp) goto TAIL_RECURSE;
5977 RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
5978 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5979 eptr--;
5980 BACKCHAR(eptr);
5981 if (ctype == OP_ANYNL && eptr > pp && RAWUCHAR(eptr) == CHAR_NL &&
5982 RAWUCHAR(eptr - 1) == CHAR_CR) eptr--;
5983 }
5984 }
5985 else
5986 #endif /* SUPPORT_UTF */
5987 /* Not UTF mode */
5988 {
5989 switch(ctype)
5990 {
5991 case OP_ANY:
5992 for (i = min; i < max; i++)
5993 {
5994 if (eptr >= md->end_subject)
5995 {
5996 SCHECK_PARTIAL();
5997 break;
5998 }
5999 if (IS_NEWLINE(eptr)) break;
6000 if (md->partial != 0 && /* Take care with CRLF partial */
6001 eptr + 1 >= md->end_subject &&
6002 NLBLOCK->nltype == NLTYPE_FIXED &&
6003 NLBLOCK->nllen == 2 &&
6004 *eptr == NLBLOCK->nl[0])
6005 {
6006 md->hitend = TRUE;
6007 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
6008 }
6009 eptr++;
6010 }
6011 break;
6012
6013 case OP_ALLANY:
6014 case OP_ANYBYTE:
6015 c = max - min;
6016 if (c > (unsigned int)(md->end_subject - eptr))
6017 {
6018 eptr = md->end_subject;
6019 SCHECK_PARTIAL();
6020 }
6021 else eptr += c;
6022 break;
6023
6024 case OP_ANYNL:
6025 for (i = min; i < max; i++)
6026 {
6027 if (eptr >= md->end_subject)
6028 {
6029 SCHECK_PARTIAL();
6030 break;
6031 }
6032 c = *eptr;
6033 if (c == CHAR_CR)
6034 {
6035 if (++eptr >= md->end_subject) break;
6036 if (*eptr == CHAR_LF) eptr++;
6037 }
6038 else
6039 {
6040 if (c != CHAR_LF && (md->bsr_anycrlf ||
6041 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
6042 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6043 && c != 0x2028 && c != 0x2029
6044 #endif
6045 ))) break;
6046 eptr++;
6047 }
6048 }
6049 break;
6050
6051 case OP_NOT_HSPACE:
6052 for (i = min; i < max; i++)
6053 {
6054 if (eptr >= md->end_subject)
6055 {
6056 SCHECK_PARTIAL();
6057 break;
6058 }
6059 switch(*eptr)
6060 {
6061 default: eptr++; break;
6062 HSPACE_BYTE_CASES:
6063 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6064 HSPACE_MULTIBYTE_CASES:
6065 #endif
6066 goto ENDLOOP00;
6067 }
6068 }
6069 ENDLOOP00:
6070 break;
6071
6072 case OP_HSPACE:
6073 for (i = min; i < max; i++)
6074 {
6075 if (eptr >= md->end_subject)
6076 {
6077 SCHECK_PARTIAL();
6078 break;
6079 }
6080 switch(*eptr)
6081 {
6082 default: goto ENDLOOP01;
6083 HSPACE_BYTE_CASES:
6084 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6085 HSPACE_MULTIBYTE_CASES:
6086 #endif
6087 eptr++; break;
6088 }
6089 }
6090 ENDLOOP01:
6091 break;
6092
6093 case OP_NOT_VSPACE:
6094 for (i = min; i < max; i++)
6095 {
6096 if (eptr >= md->end_subject)
6097 {
6098 SCHECK_PARTIAL();
6099 break;
6100 }
6101 switch(*eptr)
6102 {
6103 default: eptr++; break;
6104 VSPACE_BYTE_CASES:
6105 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6106 VSPACE_MULTIBYTE_CASES:
6107 #endif
6108 goto ENDLOOP02;
6109 }
6110 }
6111 ENDLOOP02:
6112 break;
6113
6114 case OP_VSPACE:
6115 for (i = min; i < max; i++)
6116 {
6117 if (eptr >= md->end_subject)
6118 {
6119 SCHECK_PARTIAL();
6120 break;
6121 }
6122 switch(*eptr)
6123 {
6124 default: goto ENDLOOP03;
6125 VSPACE_BYTE_CASES:
6126 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6127 VSPACE_MULTIBYTE_CASES:
6128 #endif
6129 eptr++; break;
6130 }
6131 }
6132 ENDLOOP03:
6133 break;
6134
6135 case OP_NOT_DIGIT:
6136 for (i = min; i < max; i++)
6137 {
6138 if (eptr >= md->end_subject)
6139 {
6140 SCHECK_PARTIAL();
6141 break;
6142 }
6143 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) break;
6144 eptr++;
6145 }
6146 break;
6147
6148 case OP_DIGIT:
6149 for (i = min; i < max; i++)
6150 {
6151 if (eptr >= md->end_subject)
6152 {
6153 SCHECK_PARTIAL();
6154 break;
6155 }
6156 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) break;
6157 eptr++;
6158 }
6159 break;
6160
6161 case OP_NOT_WHITESPACE:
6162 for (i = min; i < max; i++)
6163 {
6164 if (eptr >= md->end_subject)
6165 {
6166 SCHECK_PARTIAL();
6167 break;
6168 }
6169 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break;
6170 eptr++;
6171 }
6172 break;
6173
6174 case OP_WHITESPACE:
6175 for (i = min; i < max; i++)
6176 {
6177 if (eptr >= md->end_subject)
6178 {
6179 SCHECK_PARTIAL();
6180 break;
6181 }
6182 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break;
6183 eptr++;
6184 }
6185 break;
6186
6187 case OP_NOT_WORDCHAR:
6188 for (i = min; i < max; i++)
6189 {
6190 if (eptr >= md->end_subject)
6191 {
6192 SCHECK_PARTIAL();
6193 break;
6194 }
6195 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break;
6196 eptr++;
6197 }
6198 break;
6199
6200 case OP_WORDCHAR:
6201 for (i = min; i < max; i++)
6202 {
6203 if (eptr >= md->end_subject)
6204 {
6205 SCHECK_PARTIAL();
6206 break;
6207 }
6208 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break;
6209 eptr++;
6210 }
6211 break;
6212
6213 default:
6214 RRETURN(PCRE_ERROR_INTERNAL);
6215 }
6216
6217 if (possessive) continue; /* No backtracking */
6218 for (;;)
6219 {
6220 if (eptr == pp) goto TAIL_RECURSE;
6221 RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
6222 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
6223 eptr--;
6224 if (ctype == OP_ANYNL && eptr > pp && *eptr == CHAR_LF &&
6225 eptr[-1] == CHAR_CR) eptr--;
6226 }
6227 }
6228
6229 /* Control never gets here */
6230 }
6231
6232 /* There's been some horrible disaster. Arrival here can only mean there is
6233 something seriously wrong in the code above or the OP_xxx definitions. */
6234
6235 default:
6236 DPRINTF(("Unknown opcode %d\n", *ecode));
6237 RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
6238 }
6239
6240 /* Do not stick any code in here without much thought; it is assumed
6241 that "continue" in the code above comes out to here to repeat the main
6242 loop. */
6243
6244 } /* End of main loop */
6245 /* Control never reaches here */
6246
6247
6248 /* When compiling to use the heap rather than the stack for recursive calls to
6249 match(), the RRETURN() macro jumps here. The number that is saved in
6250 frame->Xwhere indicates which label we actually want to return to. */
6251
6252 #ifdef NO_RECURSE
6253 #define LBL(val) case val: goto L_RM##val;
6254 HEAP_RETURN:
6255 switch (frame->Xwhere)
6256 {
6257 LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
6258 LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
6259 LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
6260 LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
6261 LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64)
6262 LBL(65) LBL(66)
6263 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6264 LBL(21)
6265 #endif
6266 #ifdef SUPPORT_UTF
6267 LBL(16) LBL(18) LBL(20)
6268 LBL(22) LBL(23) LBL(28) LBL(30)
6269 LBL(32) LBL(34) LBL(42) LBL(46)
6270 #ifdef SUPPORT_UCP
6271 LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
6272 LBL(59) LBL(60) LBL(61) LBL(62) LBL(67) LBL(68)
6273 #endif /* SUPPORT_UCP */
6274 #endif /* SUPPORT_UTF */
6275 default:
6276 DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
6277 return PCRE_ERROR_INTERNAL;
6278 }
6279 #undef LBL
6280 #endif /* NO_RECURSE */
6281 }
6282
6283
6284 /***************************************************************************
6285 ****************************************************************************
6286 RECURSION IN THE match() FUNCTION
6287
6288 Undefine all the macros that were defined above to handle this. */
6289
6290 #ifdef NO_RECURSE
6291 #undef eptr
6292 #undef ecode
6293 #undef mstart
6294 #undef offset_top
6295 #undef eptrb
6296 #undef flags
6297
6298 #undef callpat
6299 #undef charptr
6300 #undef data
6301 #undef next
6302 #undef pp
6303 #undef prev
6304 #undef saved_eptr
6305
6306 #undef new_recursive
6307
6308 #undef cur_is_word
6309 #undef condition
6310 #undef prev_is_word
6311
6312 #undef ctype
6313 #undef length
6314 #undef max
6315 #undef min
6316 #undef number
6317 #undef offset
6318 #undef op
6319 #undef save_capture_last
6320 #undef save_offset1
6321 #undef save_offset2
6322 #undef save_offset3
6323 #undef stacksave
6324
6325 #undef newptrb
6326
6327 #endif
6328
6329 /* These two are defined as macros in both cases */
6330
6331 #undef fc
6332 #undef fi
6333
6334 /***************************************************************************
6335 ***************************************************************************/
6336
6337
6338 #ifdef NO_RECURSE
6339 /*************************************************
6340 * Release allocated heap frames *
6341 *************************************************/
6342
6343 /* This function releases all the allocated frames. The base frame is on the
6344 machine stack, and so must not be freed.
6345
6346 Argument: the address of the base frame
6347 Returns: nothing
6348 */
6349
6350 static void
6351 release_match_heapframes (heapframe *frame_base)
6352 {
6353 heapframe *nextframe = frame_base->Xnextframe;
6354 while (nextframe != NULL)
6355 {
6356 heapframe *oldframe = nextframe;
6357 nextframe = nextframe->Xnextframe;
6358 (PUBL(stack_free))(oldframe);
6359 }
6360 }
6361 #endif
6362
6363
6364 /*************************************************
6365 * Execute a Regular Expression *
6366 *************************************************/
6367
6368 /* This function applies a compiled re to a subject string and picks out
6369 portions of the string if it matches. Two elements in the vector are set for
6370 each substring: the offsets to the start and end of the substring.
6371
6372 Arguments:
6373 argument_re points to the compiled expression
6374 extra_data points to extra data or is NULL
6375 subject points to the subject string
6376 length length of subject string (may contain binary zeros)
6377 start_offset where to start in the subject string
6378 options option bits
6379 offsets points to a vector of ints to be filled in with offsets
6380 offsetcount the number of elements in the vector
6381
6382 Returns: > 0 => success; value is the number of elements filled in
6383 = 0 => success, but offsets is not big enough
6384 -1 => failed to match
6385 < -1 => some kind of unexpected problem
6386 */
6387
6388 #if defined COMPILE_PCRE8
6389 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6390 pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
6391 PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
6392 int offsetcount)
6393 #elif defined COMPILE_PCRE16
6394 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6395 pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data,
6396 PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets,
6397 int offsetcount)
6398 #elif defined COMPILE_PCRE32
6399 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6400 pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data,
6401 PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets,
6402 int offsetcount)
6403 #endif
6404 {
6405 int rc, ocount, arg_offset_max;
6406 int newline;
6407 BOOL using_temporary_offsets = FALSE;
6408 BOOL anchored;
6409 BOOL startline;
6410 BOOL firstline;
6411 BOOL utf;
6412 BOOL has_first_char = FALSE;
6413 BOOL has_req_char = FALSE;
6414 pcre_uchar first_char = 0;
6415 pcre_uchar first_char2 = 0;
6416 pcre_uchar req_char = 0;
6417 pcre_uchar req_char2 = 0;
6418 match_data match_block;
6419 match_data *md = &match_block;
6420 const pcre_uint8 *tables;
6421 const pcre_uint8 *start_bits = NULL;
6422 PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset;
6423 PCRE_PUCHAR end_subject;
6424 PCRE_PUCHAR start_partial = NULL;
6425 PCRE_PUCHAR match_partial = NULL;
6426 PCRE_PUCHAR req_char_ptr = start_match - 1;
6427
6428 const pcre_study_data *study;
6429 const REAL_PCRE *re = (const REAL_PCRE *)argument_re;
6430
6431 #ifdef NO_RECURSE
6432 heapframe frame_zero;
6433 frame_zero.Xprevframe = NULL; /* Marks the top level */
6434 frame_zero.Xnextframe = NULL; /* None are allocated yet */
6435 md->match_frames_base = &frame_zero;
6436 #endif
6437
6438 /* Check for the special magic call that measures the size of the stack used
6439 per recursive call of match(). Without the funny casting for sizeof, a Windows
6440 compiler gave this error: "unary minus operator applied to unsigned type,
6441 result still unsigned". Hopefully the cast fixes that. */
6442
6443 if (re == NULL && extra_data == NULL && subject == NULL && length == -999 &&
6444 start_offset == -999)
6445 #ifdef NO_RECURSE
6446 return -((int)sizeof(heapframe));
6447 #else
6448 return match(NULL, NULL, NULL, 0, NULL, NULL, 0);
6449 #endif
6450
6451 /* Plausibility checks */
6452
6453 if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
6454 if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0))
6455 return PCRE_ERROR_NULL;
6456 if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
6457 if (length < 0) return PCRE_ERROR_BADLENGTH;
6458 if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
6459
6460 /* Check that the first field in the block is the magic number. If it is not,
6461 return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to
6462 REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which
6463 means that the pattern is likely compiled with different endianness. */
6464
6465 if (re->magic_number != MAGIC_NUMBER)
6466 return re->magic_number == REVERSED_MAGIC_NUMBER?
6467 PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC;
6468 if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE;
6469
6470 /* These two settings are used in the code for checking a UTF-8 string that
6471 follows immediately afterwards. Other values in the md block are used only
6472 during "normal" pcre_exec() processing, not when the JIT support is in use,
6473 so they are set up later. */
6474
6475 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
6476 utf = md->utf = (re->options & PCRE_UTF8) != 0;
6477 md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
6478 ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
6479
6480 /* Check a UTF-8 string if required. Pass back the character offset and error
6481 code for an invalid string if a results vector is available. */
6482
6483 #ifdef SUPPORT_UTF
6484 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0)
6485 {
6486 int erroroffset;
6487 int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset);
6488 if (errorcode != 0)
6489 {
6490 if (offsetcount >= 2)
6491 {
6492 offsets[0] = erroroffset;
6493 offsets[1] = errorcode;
6494 }
6495 #if defined COMPILE_PCRE8
6496 return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
6497 PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
6498 #elif defined COMPILE_PCRE16
6499 return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)?
6500 PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16;
6501 #elif defined COMPILE_PCRE32
6502 return PCRE_ERROR_BADUTF32;
6503 #endif
6504 }
6505 #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16
6506 /* Check that a start_offset points to the start of a UTF character. */
6507 if (start_offset > 0 && start_offset < length &&
6508 NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset]))
6509 return PCRE_ERROR_BADUTF8_OFFSET;
6510 #endif
6511 }
6512 #endif
6513
6514 /* If the pattern was successfully studied with JIT support, run the JIT
6515 executable instead of the rest of this function. Most options must be set at
6516 compile time for the JIT code to be usable. Fallback to the normal code path if
6517 an unsupported flag is set. */
6518
6519 #ifdef SUPPORT_JIT
6520 if (extra_data != NULL
6521 && (extra_data->flags & (PCRE_EXTRA_EXECUTABLE_JIT |
6522 PCRE_EXTRA_TABLES)) == PCRE_EXTRA_EXECUTABLE_JIT
6523 && extra_data->executable_jit != NULL
6524 && (options & ~PUBLIC_JIT_EXEC_OPTIONS) == 0)
6525 {
6526 rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length,
6527 start_offset, options, offsets, offsetcount);
6528
6529 /* PCRE_ERROR_NULL means that the selected normal or partial matching
6530 mode is not compiled. In this case we simply fallback to interpreter. */
6531
6532 if (rc != PCRE_ERROR_JIT_BADOPTION) return rc;
6533 }
6534 #endif
6535
6536 /* Carry on with non-JIT matching. This information is for finding all the
6537 numbers associated with a given name, for condition testing. */
6538
6539 md->name_table = (pcre_uchar *)re + re->name_table_offset;
6540 md->name_count = re->name_count;
6541 md->name_entry_size = re->name_entry_size;
6542
6543 /* Fish out the optional data from the extra_data structure, first setting
6544 the default values. */
6545
6546 study = NULL;
6547 md->match_limit = MATCH_LIMIT;
6548 md->match_limit_recursion = MATCH_LIMIT_RECURSION;
6549 md->callout_data = NULL;
6550
6551 /* The table pointer is always in native byte order. */
6552
6553 tables = re->tables;
6554
6555 /* The two limit values override the defaults, whatever their value. */
6556
6557 if (extra_data != NULL)
6558 {
6559 register unsigned int flags = extra_data->flags;
6560 if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
6561 study = (const pcre_study_data *)extra_data->study_data;
6562 if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
6563 md->match_limit = extra_data->match_limit;
6564 if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
6565 md->match_limit_recursion = extra_data->match_limit_recursion;
6566 if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
6567 md->callout_data = extra_data->callout_data;
6568 if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
6569 }
6570
6571 /* Limits in the regex override only if they are smaller. */
6572
6573 if ((re->flags & PCRE_MLSET) != 0 && re->limit_match < md->match_limit)
6574 md->match_limit = re->limit_match;
6575
6576 if ((re->flags & PCRE_RLSET) != 0 &&
6577 re->limit_recursion < md->match_limit_recursion)
6578 md->match_limit_recursion = re->limit_recursion;
6579
6580 /* If the exec call supplied NULL for tables, use the inbuilt ones. This
6581 is a feature that makes it possible to save compiled regex and re-use them
6582 in other programs later. */
6583
6584 if (tables == NULL) tables = PRIV(default_tables);
6585
6586 /* Set up other data */
6587
6588 anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
6589 startline = (re->flags & PCRE_STARTLINE) != 0;
6590 firstline = (re->options & PCRE_FIRSTLINE) != 0;
6591
6592 /* The code starts after the real_pcre block and the capture name table. */
6593
6594 md->start_code = (const pcre_uchar *)re + re->name_table_offset +
6595 re->name_count * re->name_entry_size;
6596
6597 md->start_subject = (PCRE_PUCHAR)subject;
6598 md->start_offset = start_offset;
6599 md->end_subject = md->start_subject + length;
6600 end_subject = md->end_subject;
6601
6602 md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
6603 md->use_ucp = (re->options & PCRE_UCP) != 0;
6604 md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6605 md->ignore_skip_arg = 0;
6606
6607 /* Some options are unpacked into BOOL variables in the hope that testing
6608 them will be faster than individual option bits. */
6609
6610 md->notbol = (options & PCRE_NOTBOL) != 0;
6611 md->noteol = (options & PCRE_NOTEOL) != 0;
6612 md->notempty = (options & PCRE_NOTEMPTY) != 0;
6613 md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6614
6615 md->hitend = FALSE;
6616 md->mark = md->nomatch_mark = NULL; /* In case never set */
6617
6618 md->recursive = NULL; /* No recursion at top level */
6619 md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6620
6621 md->lcc = tables + lcc_offset;
6622 md->fcc = tables + fcc_offset;
6623 md->ctypes = tables + ctypes_offset;
6624
6625 /* Handle different \R options. */
6626
6627 switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6628 {
6629 case 0:
6630 if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
6631 md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
6632 else
6633 #ifdef BSR_ANYCRLF
6634 md->bsr_anycrlf = TRUE;
6635 #else
6636 md->bsr_anycrlf = FALSE;
6637 #endif
6638 break;
6639
6640 case PCRE_BSR_ANYCRLF:
6641 md->bsr_anycrlf = TRUE;
6642 break;
6643
6644 case PCRE_BSR_UNICODE:
6645 md->bsr_anycrlf = FALSE;
6646 break;
6647
6648 default: return PCRE_ERROR_BADNEWLINE;
6649 }
6650
6651 /* Handle different types of newline. The three bits give eight cases. If
6652 nothing is set at run time, whatever was used at compile time applies. */
6653
6654 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
6655 (pcre_uint32)options) & PCRE_NEWLINE_BITS)
6656 {
6657 case 0: newline = NEWLINE; break; /* Compile-time default */
6658 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6659 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6660 case PCRE_NEWLINE_CR+
6661 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6662 case PCRE_NEWLINE_ANY: newline = -1; break;
6663 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6664 default: return PCRE_ERROR_BADNEWLINE;
6665 }
6666
6667 if (newline == -2)
6668 {
6669 md->nltype = NLTYPE_ANYCRLF;
6670 }
6671 else if (newline < 0)
6672 {
6673 md->nltype = NLTYPE_ANY;
6674 }
6675 else
6676 {
6677 md->nltype = NLTYPE_FIXED;
6678 if (newline > 255)
6679 {
6680 md->nllen = 2;
6681 md->nl[0] = (newline >> 8) & 255;
6682 md->nl[1] = newline & 255;
6683 }
6684 else
6685 {
6686 md->nllen = 1;
6687 md->nl[0] = newline;
6688 }
6689 }
6690
6691 /* Partial matching was originally supported only for a restricted set of
6692 regexes; from release 8.00 there are no restrictions, but the bits are still
6693 defined (though never set). So there's no harm in leaving this code. */
6694
6695 if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6696 return PCRE_ERROR_BADPARTIAL;
6697
6698 /* If the expression has got more back references than the offsets supplied can
6699 hold, we get a temporary chunk of working store to use during the matching.
6700 Otherwise, we can use the vector supplied, rounding down its size to a multiple
6701 of 3. */
6702
6703 ocount = offsetcount - (offsetcount % 3);
6704 arg_offset_max = (2*ocount)/3;
6705
6706 if (re->top_backref > 0 && re->top_backref >= ocount/3)
6707 {
6708 ocount = re->top_backref * 3 + 3;
6709 md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int));
6710 if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
6711 using_temporary_offsets = TRUE;
6712 DPRINTF(("Got memory to hold back references\n"));
6713 }
6714 else md->offset_vector = offsets;
6715 md->offset_end = ocount;
6716 md->offset_max = (2*ocount)/3;
6717 md->capture_last = 0;
6718
6719 /* Reset the working variable associated with each extraction. These should
6720 never be used unless previously set, but they get saved and restored, and so we
6721 initialize them to avoid reading uninitialized locations. Also, unset the
6722 offsets for the matched string. This is really just for tidiness with callouts,
6723 in case they inspect these fields. */
6724
6725 if (md->offset_vector != NULL)
6726 {
6727 register int *iptr = md->offset_vector + ocount;
6728 register int *iend = iptr - re->top_bracket;
6729 if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6730 while (--iptr >= iend) *iptr = -1;
6731 md->offset_vector[0] = md->offset_vector[1] = -1;
6732 }
6733
6734 /* Set up the first character to match, if available. The first_char value is
6735 never set for an anchored regular expression, but the anchoring may be forced
6736 at run time, so we have to test for anchoring. The first char may be unset for
6737 an unanchored pattern, of course. If there's no first char and the pattern was
6738 studied, there may be a bitmap of possible first characters. */
6739
6740 if (!anchored)
6741 {
6742 if ((re->flags & PCRE_FIRSTSET) != 0)
6743 {
6744 has_first_char = TRUE;
6745 first_char = first_char2 = (pcre_uchar)(re->first_char);
6746 if ((re->flags & PCRE_FCH_CASELESS) != 0)
6747 {
6748 first_char2 = TABLE_GET(first_char, md->fcc, first_char);
6749 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
6750 if (utf && first_char > 127)
6751 first_char2 = UCD_OTHERCASE(first_char);
6752 #endif
6753 }
6754 }
6755 else
6756