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

Contents of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1361 - (show annotations)
Fri Sep 6 17:47:32 2013 UTC (6 years, 1 month ago) by ph10
File MIME type: text/plain
File size: 220353 byte(s)
Error occurred while calculating annotation data.
Make back references to duplicated named subpatterns more like Perl.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2013 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40 /* This module contains pcre_exec(), the externally visible function that does
41 pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42 possible. There are also some static supporting functions. */
43
44 #ifdef HAVE_CONFIG_H
45 #include "config.h"
46 #endif
47
48 #define NLBLOCK md /* Block containing newline information */
49 #define PSSTART start_subject /* Field containing processed string start */
50 #define PSEND end_subject /* Field containing processed string end */
51
52 #include "pcre_internal.h"
53
54 /* Undefine some potentially clashing cpp symbols */
55
56 #undef min
57 #undef max
58
59 /* The md->capture_last field uses the lower 16 bits for the last captured
60 substring (which can never be greater than 65535) and a bit in the top half
61 to mean "capture vector overflowed". This odd way of doing things was
62 implemented when it was realized that preserving and restoring the overflow bit
63 whenever the last capture number was saved/restored made for a neater
64 interface, and doing it this way saved on (a) another variable, which would
65 have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66 separate set of save/restore instructions. The following defines are used in
67 implementing this. */
68
69 #define CAPLMASK 0x0000ffff /* The bits used for last_capture */
70 #define OVFLMASK 0xffff0000 /* The bits used for the overflow flag */
71 #define OVFLBIT 0x00010000 /* The bit that is set for overflow */
72
73 /* Values for setting in md->match_function_type to indicate two special types
74 of call to match(). We do it this way to save on using another stack variable,
75 as stack usage is to be discouraged. */
76
77 #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */
78 #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */
79
80 /* Non-error returns from the match() function. Error returns are externally
81 defined PCRE_ERROR_xxx codes, which are all negative. */
82
83 #define MATCH_MATCH 1
84 #define MATCH_NOMATCH 0
85
86 /* Special internal returns from the match() function. Make them sufficiently
87 negative to avoid the external error codes. */
88
89 #define MATCH_ACCEPT (-999)
90 #define MATCH_KETRPOS (-998)
91 #define MATCH_ONCE (-997)
92 /* The next 5 must be kept together and in sequence so that a test that checks
93 for any one of them can use a range. */
94 #define MATCH_COMMIT (-996)
95 #define MATCH_PRUNE (-995)
96 #define MATCH_SKIP (-994)
97 #define MATCH_SKIP_ARG (-993)
98 #define MATCH_THEN (-992)
99 #define MATCH_BACKTRACK_MAX MATCH_THEN
100 #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101
102 /* Maximum number of ints of offset to save on the stack for recursive calls.
103 If the offset vector is bigger, malloc is used. This should be a multiple of 3,
104 because the offset vector is always a multiple of 3 long. */
105
106 #define REC_STACK_SAVE_MAX 30
107
108 /* Min and max values for the common repeats; for the maxima, 0 => infinity */
109
110 static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
111 static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
112
113 #ifdef PCRE_DEBUG
114 /*************************************************
115 * Debugging function to print chars *
116 *************************************************/
117
118 /* Print a sequence of chars in printable format, stopping at the end of the
119 subject if the requested.
120
121 Arguments:
122 p points to characters
123 length number to print
124 is_subject TRUE if printing from within md->start_subject
125 md pointer to matching data block, if is_subject is TRUE
126
127 Returns: nothing
128 */
129
130 static void
131 pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132 {
133 pcre_uint32 c;
134 BOOL utf = md->utf;
135 if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136 while (length-- > 0)
137 if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138 }
139 #endif
140
141
142
143 /*************************************************
144 * Match a back-reference *
145 *************************************************/
146
147 /* Normally, if a back reference hasn't been set, the length that is passed is
148 negative, so the match always fails. However, in JavaScript compatibility mode,
149 the length passed is zero. Note that in caseless UTF-8 mode, the number of
150 subject bytes matched may be different to the number of reference bytes.
151
152 Arguments:
153 offset index into the offset vector
154 eptr pointer into the subject
155 length length of reference to be matched (number of bytes)
156 md points to match data block
157 caseless TRUE if caseless
158
159 Returns: >= 0 the number of subject bytes matched
160 -1 no match
161 -2 partial match; always given if at end subject
162 */
163
164 static int
165 match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166 BOOL caseless)
167 {
168 PCRE_PUCHAR eptr_start = eptr;
169 register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170 #ifdef SUPPORT_UTF
171 BOOL utf = md->utf;
172 #endif
173
174 #ifdef PCRE_DEBUG
175 if (eptr >= md->end_subject)
176 printf("matching subject <null>");
177 else
178 {
179 printf("matching subject ");
180 pchars(eptr, length, TRUE, md);
181 }
182 printf(" against backref ");
183 pchars(p, length, FALSE, md);
184 printf("\n");
185 #endif
186
187 /* Always fail if reference not set (and not JavaScript compatible - in that
188 case the length is passed as zero). */
189
190 if (length < 0) return -1;
191
192 /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193 properly if Unicode properties are supported. Otherwise, we can check only
194 ASCII characters. */
195
196 if (caseless)
197 {
198 #ifdef SUPPORT_UTF
199 #ifdef SUPPORT_UCP
200 if (utf)
201 {
202 /* Match characters up to the end of the reference. NOTE: the number of
203 data units matched may differ, because in UTF-8 there are some characters
204 whose upper and lower case versions code have different numbers of bytes.
205 For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206 (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207 sequence of two of the latter. It is important, therefore, to check the
208 length along the reference, not along the subject (earlier code did this
209 wrong). */
210
211 PCRE_PUCHAR endptr = p + length;
212 while (p < endptr)
213 {
214 pcre_uint32 c, d;
215 const ucd_record *ur;
216 if (eptr >= md->end_subject) return -2; /* Partial match */
217 GETCHARINC(c, eptr);
218 GETCHARINC(d, p);
219 ur = GET_UCD(d);
220 if (c != d && c != d + ur->other_case)
221 {
222 const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223 for (;;)
224 {
225 if (c < *pp) return -1;
226 if (c == *pp++) break;
227 }
228 }
229 }
230 }
231 else
232 #endif
233 #endif
234
235 /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236 is no UCP support. */
237 {
238 while (length-- > 0)
239 {
240 pcre_uint32 cc, cp;
241 if (eptr >= md->end_subject) return -2; /* Partial match */
242 cc = RAWUCHARTEST(eptr);
243 cp = RAWUCHARTEST(p);
244 if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245 p++;
246 eptr++;
247 }
248 }
249 }
250
251 /* In the caseful case, we can just compare the bytes, whether or not we
252 are in UTF-8 mode. */
253
254 else
255 {
256 while (length-- > 0)
257 {
258 if (eptr >= md->end_subject) return -2; /* Partial match */
259 if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260 }
261 }
262
263 return (int)(eptr - eptr_start);
264 }
265
266
267
268 /***************************************************************************
269 ****************************************************************************
270 RECURSION IN THE match() FUNCTION
271
272 The match() function is highly recursive, though not every recursive call
273 increases the recursive depth. Nevertheless, some regular expressions can cause
274 it to recurse to a great depth. I was writing for Unix, so I just let it call
275 itself recursively. This uses the stack for saving everything that has to be
276 saved for a recursive call. On Unix, the stack can be large, and this works
277 fine.
278
279 It turns out that on some non-Unix-like systems there are problems with
280 programs that use a lot of stack. (This despite the fact that every last chip
281 has oodles of memory these days, and techniques for extending the stack have
282 been known for decades.) So....
283
284 There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
285 calls by keeping local variables that need to be preserved in blocks of memory
286 obtained from malloc() instead instead of on the stack. Macros are used to
287 achieve this so that the actual code doesn't look very different to what it
288 always used to.
289
290 The original heap-recursive code used longjmp(). However, it seems that this
291 can be very slow on some operating systems. Following a suggestion from Stan
292 Switzer, the use of longjmp() has been abolished, at the cost of having to
293 provide a unique number for each call to RMATCH. There is no way of generating
294 a sequence of numbers at compile time in C. I have given them names, to make
295 them stand out more clearly.
296
297 Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
298 FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
299 tests. Furthermore, not using longjmp() means that local dynamic variables
300 don't have indeterminate values; this has meant that the frame size can be
301 reduced because the result can be "passed back" by straight setting of the
302 variable instead of being passed in the frame.
303 ****************************************************************************
304 ***************************************************************************/
305
306 /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307 below must be updated in sync. */
308
309 enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10,
310 RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311 RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312 RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313 RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314 RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315 RM61, RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316
317 /* These versions of the macros use the stack, as normal. There are debugging
318 versions and production versions. Note that the "rw" argument of RMATCH isn't
319 actually used in this definition. */
320
321 #ifndef NO_RECURSE
322 #define REGISTER register
323
324 #ifdef PCRE_DEBUG
325 #define RMATCH(ra,rb,rc,rd,re,rw) \
326 { \
327 printf("match() called in line %d\n", __LINE__); \
328 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329 printf("to line %d\n", __LINE__); \
330 }
331 #define RRETURN(ra) \
332 { \
333 printf("match() returned %d from line %d\n", ra, __LINE__); \
334 return ra; \
335 }
336 #else
337 #define RMATCH(ra,rb,rc,rd,re,rw) \
338 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339 #define RRETURN(ra) return ra
340 #endif
341
342 #else
343
344
345 /* These versions of the macros manage a private stack on the heap. Note that
346 the "rd" argument of RMATCH isn't actually used in this definition. It's the md
347 argument of match(), which never changes. */
348
349 #define REGISTER
350
351 #define RMATCH(ra,rb,rc,rd,re,rw)\
352 {\
353 heapframe *newframe = frame->Xnextframe;\
354 if (newframe == NULL)\
355 {\
356 newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357 if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358 newframe->Xnextframe = NULL;\
359 frame->Xnextframe = newframe;\
360 }\
361 frame->Xwhere = rw;\
362 newframe->Xeptr = ra;\
363 newframe->Xecode = rb;\
364 newframe->Xmstart = mstart;\
365 newframe->Xoffset_top = rc;\
366 newframe->Xeptrb = re;\
367 newframe->Xrdepth = frame->Xrdepth + 1;\
368 newframe->Xprevframe = frame;\
369 frame = newframe;\
370 DPRINTF(("restarting from line %d\n", __LINE__));\
371 goto HEAP_RECURSE;\
372 L_##rw:\
373 DPRINTF(("jumped back to line %d\n", __LINE__));\
374 }
375
376 #define RRETURN(ra)\
377 {\
378 heapframe *oldframe = frame;\
379 frame = oldframe->Xprevframe;\
380 if (frame != NULL)\
381 {\
382 rrc = ra;\
383 goto HEAP_RETURN;\
384 }\
385 return ra;\
386 }
387
388
389 /* Structure for remembering the local variables in a private frame */
390
391 typedef struct heapframe {
392 struct heapframe *Xprevframe;
393 struct heapframe *Xnextframe;
394
395 /* Function arguments that may change */
396
397 PCRE_PUCHAR Xeptr;
398 const pcre_uchar *Xecode;
399 PCRE_PUCHAR Xmstart;
400 int Xoffset_top;
401 eptrblock *Xeptrb;
402 unsigned int Xrdepth;
403
404 /* Function local variables */
405
406 PCRE_PUCHAR Xcallpat;
407 #ifdef SUPPORT_UTF
408 PCRE_PUCHAR Xcharptr;
409 #endif
410 PCRE_PUCHAR Xdata;
411 PCRE_PUCHAR Xnext;
412 PCRE_PUCHAR Xpp;
413 PCRE_PUCHAR Xprev;
414 PCRE_PUCHAR Xsaved_eptr;
415
416 recursion_info Xnew_recursive;
417
418 BOOL Xcur_is_word;
419 BOOL Xcondition;
420 BOOL Xprev_is_word;
421
422 #ifdef SUPPORT_UCP
423 int Xprop_type;
424 unsigned int Xprop_value;
425 int Xprop_fail_result;
426 int Xoclength;
427 pcre_uchar Xocchars[6];
428 #endif
429
430 int Xcodelink;
431 int Xctype;
432 unsigned int Xfc;
433 int Xfi;
434 int Xlength;
435 int Xmax;
436 int Xmin;
437 unsigned int Xnumber;
438 int Xoffset;
439 unsigned int Xop;
440 pcre_int32 Xsave_capture_last;
441 int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442 int Xstacksave[REC_STACK_SAVE_MAX];
443
444 eptrblock Xnewptrb;
445
446 /* Where to jump back to */
447
448 int Xwhere;
449
450 } heapframe;
451
452 #endif
453
454
455 /***************************************************************************
456 ***************************************************************************/
457
458
459
460 /*************************************************
461 * Match from current position *
462 *************************************************/
463
464 /* This function is called recursively in many circumstances. Whenever it
465 returns a negative (error) response, the outer incarnation must also return the
466 same response. */
467
468 /* These macros pack up tests that are used for partial matching, and which
469 appear several times in the code. We set the "hit end" flag if the pointer is
470 at the end of the subject and also past the start of the subject (i.e.
471 something has been matched). For hard partial matching, we then return
472 immediately. The second one is used when we already know we are past the end of
473 the subject. */
474
475 #define CHECK_PARTIAL()\
476 if (md->partial != 0 && eptr >= md->end_subject && \
477 eptr > md->start_used_ptr) \
478 { \
479 md->hitend = TRUE; \
480 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481 }
482
483 #define SCHECK_PARTIAL()\
484 if (md->partial != 0 && eptr > md->start_used_ptr) \
485 { \
486 md->hitend = TRUE; \
487 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488 }
489
490
491 /* Performance note: It might be tempting to extract commonly used fields from
492 the md structure (e.g. utf, end_subject) into individual variables to improve
493 performance. Tests using gcc on a SPARC disproved this; in the first case, it
494 made performance worse.
495
496 Arguments:
497 eptr pointer to current character in subject
498 ecode pointer to current position in compiled code
499 mstart pointer to the current match start position (can be modified
500 by encountering \K)
501 offset_top current top pointer
502 md pointer to "static" info for the match
503 eptrb pointer to chain of blocks containing eptr at start of
504 brackets - for testing for empty matches
505 rdepth the recursion depth
506
507 Returns: MATCH_MATCH if matched ) these values are >= 0
508 MATCH_NOMATCH if failed to match )
509 a negative MATCH_xxx value for PRUNE, SKIP, etc
510 a negative PCRE_ERROR_xxx value if aborted by an error condition
511 (e.g. stopped by repeated call or recursion limit)
512 */
513
514 static int
515 match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516 PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517 unsigned int rdepth)
518 {
519 /* These variables do not need to be preserved over recursion in this function,
520 so they can be ordinary variables in all cases. Mark some of them with
521 "register" because they are used a lot in loops. */
522
523 register int rrc; /* Returns from recursive calls */
524 register int i; /* Used for loops not involving calls to RMATCH() */
525 register pcre_uint32 c; /* Character values not kept over RMATCH() calls */
526 register BOOL utf; /* Local copy of UTF flag for speed */
527
528 BOOL minimize, possessive; /* Quantifier options */
529 BOOL caseless;
530 int condcode;
531
532 /* When recursion is not being used, all "local" variables that have to be
533 preserved over calls to RMATCH() are part of a "frame". We set up the top-level
534 frame on the stack here; subsequent instantiations are obtained from the heap
535 whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536 the top-level on the stack rather than malloc-ing them all gives a performance
537 boost in many cases where there is not much "recursion". */
538
539 #ifdef NO_RECURSE
540 heapframe *frame = (heapframe *)md->match_frames_base;
541
542 /* Copy in the original argument variables */
543
544 frame->Xeptr = eptr;
545 frame->Xecode = ecode;
546 frame->Xmstart = mstart;
547 frame->Xoffset_top = offset_top;
548 frame->Xeptrb = eptrb;
549 frame->Xrdepth = rdepth;
550
551 /* This is where control jumps back to to effect "recursion" */
552
553 HEAP_RECURSE:
554
555 /* Macros make the argument variables come from the current frame */
556
557 #define eptr frame->Xeptr
558 #define ecode frame->Xecode
559 #define mstart frame->Xmstart
560 #define offset_top frame->Xoffset_top
561 #define eptrb frame->Xeptrb
562 #define rdepth frame->Xrdepth
563
564 /* Ditto for the local variables */
565
566 #ifdef SUPPORT_UTF
567 #define charptr frame->Xcharptr
568 #endif
569 #define callpat frame->Xcallpat
570 #define codelink frame->Xcodelink
571 #define data frame->Xdata
572 #define next frame->Xnext
573 #define pp frame->Xpp
574 #define prev frame->Xprev
575 #define saved_eptr frame->Xsaved_eptr
576
577 #define new_recursive frame->Xnew_recursive
578
579 #define cur_is_word frame->Xcur_is_word
580 #define condition frame->Xcondition
581 #define prev_is_word frame->Xprev_is_word
582
583 #ifdef SUPPORT_UCP
584 #define prop_type frame->Xprop_type
585 #define prop_value frame->Xprop_value
586 #define prop_fail_result frame->Xprop_fail_result
587 #define oclength frame->Xoclength
588 #define occhars frame->Xocchars
589 #endif
590
591 #define ctype frame->Xctype
592 #define fc frame->Xfc
593 #define fi frame->Xfi
594 #define length frame->Xlength
595 #define max frame->Xmax
596 #define min frame->Xmin
597 #define number frame->Xnumber
598 #define offset frame->Xoffset
599 #define op frame->Xop
600 #define save_capture_last frame->Xsave_capture_last
601 #define save_offset1 frame->Xsave_offset1
602 #define save_offset2 frame->Xsave_offset2
603 #define save_offset3 frame->Xsave_offset3
604 #define stacksave frame->Xstacksave
605
606 #define newptrb frame->Xnewptrb
607
608 /* When recursion is being used, local variables are allocated on the stack and
609 get preserved during recursion in the normal way. In this environment, fi and
610 i, and fc and c, can be the same variables. */
611
612 #else /* NO_RECURSE not defined */
613 #define fi i
614 #define fc c
615
616 /* Many of the following variables are used only in small blocks of the code.
617 My normal style of coding would have declared them within each of those blocks.
618 However, in order to accommodate the version of this code that uses an external
619 "stack" implemented on the heap, it is easier to declare them all here, so the
620 declarations can be cut out in a block. The only declarations within blocks
621 below are for variables that do not have to be preserved over a recursive call
622 to RMATCH(). */
623
624 #ifdef SUPPORT_UTF
625 const pcre_uchar *charptr;
626 #endif
627 const pcre_uchar *callpat;
628 const pcre_uchar *data;
629 const pcre_uchar *next;
630 PCRE_PUCHAR pp;
631 const pcre_uchar *prev;
632 PCRE_PUCHAR saved_eptr;
633
634 recursion_info new_recursive;
635
636 BOOL cur_is_word;
637 BOOL condition;
638 BOOL prev_is_word;
639
640 #ifdef SUPPORT_UCP
641 int prop_type;
642 unsigned int prop_value;
643 int prop_fail_result;
644 int oclength;
645 pcre_uchar occhars[6];
646 #endif
647
648 int codelink;
649 int ctype;
650 int length;
651 int max;
652 int min;
653 unsigned int number;
654 int offset;
655 unsigned int op;
656 pcre_int32 save_capture_last;
657 int save_offset1, save_offset2, save_offset3;
658 int stacksave[REC_STACK_SAVE_MAX];
659
660 eptrblock newptrb;
661
662 /* There is a special fudge for calling match() in a way that causes it to
663 measure the size of its basic stack frame when the stack is being used for
664 recursion. The second argument (ecode) being NULL triggers this behaviour. It
665 cannot normally ever be NULL. The return is the negated value of the frame
666 size. */
667
668 if (ecode == NULL)
669 {
670 if (rdepth == 0)
671 return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672 else
673 {
674 int len = (char *)&rdepth - (char *)eptr;
675 return (len > 0)? -len : len;
676 }
677 }
678 #endif /* NO_RECURSE */
679
680 /* To save space on the stack and in the heap frame, I have doubled up on some
681 of the local variables that are used only in localised parts of the code, but
682 still need to be preserved over recursive calls of match(). These macros define
683 the alternative names that are used. */
684
685 #define allow_zero cur_is_word
686 #define cbegroup condition
687 #define code_offset codelink
688 #define condassert condition
689 #define matched_once prev_is_word
690 #define foc number
691 #define save_mark data
692
693 /* These statements are here to stop the compiler complaining about unitialized
694 variables. */
695
696 #ifdef SUPPORT_UCP
697 prop_value = 0;
698 prop_fail_result = 0;
699 #endif
700
701
702 /* This label is used for tail recursion, which is used in a few cases even
703 when NO_RECURSE is not defined, in order to reduce the amount of stack that is
704 used. Thanks to Ian Taylor for noticing this possibility and sending the
705 original patch. */
706
707 TAIL_RECURSE:
708
709 /* OK, now we can get on with the real code of the function. Recursive calls
710 are specified by the macro RMATCH and RRETURN is used to return. When
711 NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712 and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713 defined). However, RMATCH isn't like a function call because it's quite a
714 complicated macro. It has to be used in one particular way. This shouldn't,
715 however, impact performance when true recursion is being used. */
716
717 #ifdef SUPPORT_UTF
718 utf = md->utf; /* Local copy of the flag */
719 #else
720 utf = FALSE;
721 #endif
722
723 /* First check that we haven't called match() too many times, or that we
724 haven't exceeded the recursive call limit. */
725
726 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728
729 /* At the start of a group with an unlimited repeat that may match an empty
730 string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731 done this way to save having to use another function argument, which would take
732 up space on the stack. See also MATCH_CONDASSERT below.
733
734 When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735 such remembered pointers, to be checked when we hit the closing ket, in order
736 to break infinite loops that match no characters. When match() is called in
737 other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738 NOT be used with tail recursion, because the memory block that is used is on
739 the stack, so a new one may be required for each match(). */
740
741 if (md->match_function_type == MATCH_CBEGROUP)
742 {
743 newptrb.epb_saved_eptr = eptr;
744 newptrb.epb_prev = eptrb;
745 eptrb = &newptrb;
746 md->match_function_type = 0;
747 }
748
749 /* Now start processing the opcodes. */
750
751 for (;;)
752 {
753 minimize = possessive = FALSE;
754 op = *ecode;
755
756 switch(op)
757 {
758 case OP_MARK:
759 md->nomatch_mark = ecode + 2;
760 md->mark = NULL; /* In case previously set by assertion */
761 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762 eptrb, RM55);
763 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764 md->mark == NULL) md->mark = ecode + 2;
765
766 /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767 argument, and we must check whether that argument matches this MARK's
768 argument. It is passed back in md->start_match_ptr (an overloading of that
769 variable). If it does match, we reset that variable to the current subject
770 position and return MATCH_SKIP. Otherwise, pass back the return code
771 unaltered. */
772
773 else if (rrc == MATCH_SKIP_ARG &&
774 STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775 {
776 md->start_match_ptr = eptr;
777 RRETURN(MATCH_SKIP);
778 }
779 RRETURN(rrc);
780
781 case OP_FAIL:
782 RRETURN(MATCH_NOMATCH);
783
784 case OP_COMMIT:
785 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786 eptrb, RM52);
787 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788 RRETURN(MATCH_COMMIT);
789
790 case OP_PRUNE:
791 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792 eptrb, RM51);
793 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794 RRETURN(MATCH_PRUNE);
795
796 case OP_PRUNE_ARG:
797 md->nomatch_mark = ecode + 2;
798 md->mark = NULL; /* In case previously set by assertion */
799 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800 eptrb, RM56);
801 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
802 md->mark == NULL) md->mark = ecode + 2;
803 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
804 RRETURN(MATCH_PRUNE);
805
806 case OP_SKIP:
807 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
808 eptrb, RM53);
809 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
810 md->start_match_ptr = eptr; /* Pass back current position */
811 RRETURN(MATCH_SKIP);
812
813 /* Note that, for Perl compatibility, SKIP with an argument does NOT set
814 nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
815 not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
816 that failed and any that precede it (either they also failed, or were not
817 triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
818 SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
819 set to the count of the one that failed. */
820
821 case OP_SKIP_ARG:
822 md->skip_arg_count++;
823 if (md->skip_arg_count <= md->ignore_skip_arg)
824 {
825 ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
826 break;
827 }
828 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
829 eptrb, RM57);
830 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831
832 /* Pass back the current skip name by overloading md->start_match_ptr and
833 returning the special MATCH_SKIP_ARG return code. This will either be
834 caught by a matching MARK, or get to the top, where it causes a rematch
835 with md->ignore_skip_arg set to the value of md->skip_arg_count. */
836
837 md->start_match_ptr = ecode + 2;
838 RRETURN(MATCH_SKIP_ARG);
839
840 /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
841 the branch in which it occurs can be determined. Overload the start of
842 match pointer to do this. */
843
844 case OP_THEN:
845 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
846 eptrb, RM54);
847 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
848 md->start_match_ptr = ecode;
849 RRETURN(MATCH_THEN);
850
851 case OP_THEN_ARG:
852 md->nomatch_mark = ecode + 2;
853 md->mark = NULL; /* In case previously set by assertion */
854 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
855 md, eptrb, RM58);
856 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
857 md->mark == NULL) md->mark = ecode + 2;
858 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859 md->start_match_ptr = ecode;
860 RRETURN(MATCH_THEN);
861
862 /* Handle an atomic group that does not contain any capturing parentheses.
863 This can be handled like an assertion. Prior to 8.13, all atomic groups
864 were handled this way. In 8.13, the code was changed as below for ONCE, so
865 that backups pass through the group and thereby reset captured values.
866 However, this uses a lot more stack, so in 8.20, atomic groups that do not
867 contain any captures generate OP_ONCE_NC, which can be handled in the old,
868 less stack intensive way.
869
870 Check the alternative branches in turn - the matching won't pass the KET
871 for this kind of subpattern. If any one branch matches, we carry on as at
872 the end of a normal bracket, leaving the subject pointer, but resetting
873 the start-of-match value in case it was changed by \K. */
874
875 case OP_ONCE_NC:
876 prev = ecode;
877 saved_eptr = eptr;
878 save_mark = md->mark;
879 do
880 {
881 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
882 if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */
883 {
884 mstart = md->start_match_ptr;
885 break;
886 }
887 if (rrc == MATCH_THEN)
888 {
889 next = ecode + GET(ecode,1);
890 if (md->start_match_ptr < next &&
891 (*ecode == OP_ALT || *next == OP_ALT))
892 rrc = MATCH_NOMATCH;
893 }
894
895 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
896 ecode += GET(ecode,1);
897 md->mark = save_mark;
898 }
899 while (*ecode == OP_ALT);
900
901 /* If hit the end of the group (which could be repeated), fail */
902
903 if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
904
905 /* Continue as from after the group, updating the offsets high water
906 mark, since extracts may have been taken. */
907
908 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
909
910 offset_top = md->end_offset_top;
911 eptr = md->end_match_ptr;
912
913 /* For a non-repeating ket, just continue at this level. This also
914 happens for a repeating ket if no characters were matched in the group.
915 This is the forcible breaking of infinite loops as implemented in Perl
916 5.005. */
917
918 if (*ecode == OP_KET || eptr == saved_eptr)
919 {
920 ecode += 1+LINK_SIZE;
921 break;
922 }
923
924 /* The repeating kets try the rest of the pattern or restart from the
925 preceding bracket, in the appropriate order. The second "call" of match()
926 uses tail recursion, to avoid using another stack frame. */
927
928 if (*ecode == OP_KETRMIN)
929 {
930 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
931 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
932 ecode = prev;
933 goto TAIL_RECURSE;
934 }
935 else /* OP_KETRMAX */
936 {
937 RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
938 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
939 ecode += 1 + LINK_SIZE;
940 goto TAIL_RECURSE;
941 }
942 /* Control never gets here */
943
944 /* Handle a capturing bracket, other than those that are possessive with an
945 unlimited repeat. If there is space in the offset vector, save the current
946 subject position in the working slot at the top of the vector. We mustn't
947 change the current values of the data slot, because they may be set from a
948 previous iteration of this group, and be referred to by a reference inside
949 the group. A failure to match might occur after the group has succeeded,
950 if something later on doesn't match. For this reason, we need to restore
951 the working value and also the values of the final offsets, in case they
952 were set by a previous iteration of the same bracket.
953
954 If there isn't enough space in the offset vector, treat this as if it were
955 a non-capturing bracket. Don't worry about setting the flag for the error
956 case here; that is handled in the code for KET. */
957
958 case OP_CBRA:
959 case OP_SCBRA:
960 number = GET2(ecode, 1+LINK_SIZE);
961 offset = number << 1;
962
963 #ifdef PCRE_DEBUG
964 printf("start bracket %d\n", number);
965 printf("subject=");
966 pchars(eptr, 16, TRUE, md);
967 printf("\n");
968 #endif
969
970 if (offset < md->offset_max)
971 {
972 save_offset1 = md->offset_vector[offset];
973 save_offset2 = md->offset_vector[offset+1];
974 save_offset3 = md->offset_vector[md->offset_end - number];
975 save_capture_last = md->capture_last;
976 save_mark = md->mark;
977
978 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
979 md->offset_vector[md->offset_end - number] =
980 (int)(eptr - md->start_subject);
981
982 for (;;)
983 {
984 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
986 eptrb, RM1);
987 if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */
988
989 /* If we backed up to a THEN, check whether it is within the current
990 branch by comparing the address of the THEN that is passed back with
991 the end of the branch. If it is within the current branch, and the
992 branch is one of two or more alternatives (it either starts or ends
993 with OP_ALT), we have reached the limit of THEN's action, so convert
994 the return code to NOMATCH, which will cause normal backtracking to
995 happen from now on. Otherwise, THEN is passed back to an outer
996 alternative. This implements Perl's treatment of parenthesized groups,
997 where a group not containing | does not affect the current alternative,
998 that is, (X) is NOT the same as (X|(*F)). */
999
1000 if (rrc == MATCH_THEN)
1001 {
1002 next = ecode + GET(ecode,1);
1003 if (md->start_match_ptr < next &&
1004 (*ecode == OP_ALT || *next == OP_ALT))
1005 rrc = MATCH_NOMATCH;
1006 }
1007
1008 /* Anything other than NOMATCH is passed back. */
1009
1010 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1011 md->capture_last = save_capture_last;
1012 ecode += GET(ecode, 1);
1013 md->mark = save_mark;
1014 if (*ecode != OP_ALT) break;
1015 }
1016
1017 DPRINTF(("bracket %d failed\n", number));
1018 md->offset_vector[offset] = save_offset1;
1019 md->offset_vector[offset+1] = save_offset2;
1020 md->offset_vector[md->offset_end - number] = save_offset3;
1021
1022 /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1023
1024 RRETURN(rrc);
1025 }
1026
1027 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1028 as a non-capturing bracket. */
1029
1030 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1031 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1032
1033 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1034
1035 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037
1038 /* Non-capturing or atomic group, except for possessive with unlimited
1039 repeat and ONCE group with no captures. Loop for all the alternatives.
1040
1041 When we get to the final alternative within the brackets, we used to return
1042 the result of a recursive call to match() whatever happened so it was
1043 possible to reduce stack usage by turning this into a tail recursion,
1044 except in the case of a possibly empty group. However, now that there is
1045 the possiblity of (*THEN) occurring in the final alternative, this
1046 optimization is no longer always possible.
1047
1048 We can optimize if we know there are no (*THEN)s in the pattern; at present
1049 this is the best that can be done.
1050
1051 MATCH_ONCE is returned when the end of an atomic group is successfully
1052 reached, but subsequent matching fails. It passes back up the tree (causing
1053 captured values to be reset) until the original atomic group level is
1054 reached. This is tested by comparing md->once_target with the start of the
1055 group. At this point, the return is converted into MATCH_NOMATCH so that
1056 previous backup points can be taken. */
1057
1058 case OP_ONCE:
1059 case OP_BRA:
1060 case OP_SBRA:
1061 DPRINTF(("start non-capturing bracket\n"));
1062
1063 for (;;)
1064 {
1065 if (op >= OP_SBRA || op == OP_ONCE)
1066 md->match_function_type = MATCH_CBEGROUP;
1067
1068 /* If this is not a possibly empty group, and there are no (*THEN)s in
1069 the pattern, and this is the final alternative, optimize as described
1070 above. */
1071
1072 else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073 {
1074 ecode += PRIV(OP_lengths)[*ecode];
1075 goto TAIL_RECURSE;
1076 }
1077
1078 /* In all other cases, we have to make another call to match(). */
1079
1080 save_mark = md->mark;
1081 save_capture_last = md->capture_last;
1082 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1083 RM2);
1084
1085 /* See comment in the code for capturing groups above about handling
1086 THEN. */
1087
1088 if (rrc == MATCH_THEN)
1089 {
1090 next = ecode + GET(ecode,1);
1091 if (md->start_match_ptr < next &&
1092 (*ecode == OP_ALT || *next == OP_ALT))
1093 rrc = MATCH_NOMATCH;
1094 }
1095
1096 if (rrc != MATCH_NOMATCH)
1097 {
1098 if (rrc == MATCH_ONCE)
1099 {
1100 const pcre_uchar *scode = ecode;
1101 if (*scode != OP_ONCE) /* If not at start, find it */
1102 {
1103 while (*scode == OP_ALT) scode += GET(scode, 1);
1104 scode -= GET(scode, 1);
1105 }
1106 if (md->once_target == scode) rrc = MATCH_NOMATCH;
1107 }
1108 RRETURN(rrc);
1109 }
1110 ecode += GET(ecode, 1);
1111 md->mark = save_mark;
1112 if (*ecode != OP_ALT) break;
1113 md->capture_last = save_capture_last;
1114 }
1115
1116 RRETURN(MATCH_NOMATCH);
1117
1118 /* Handle possessive capturing brackets with an unlimited repeat. We come
1119 here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120 handled similarly to the normal case above. However, the matching is
1121 different. The end of these brackets will always be OP_KETRPOS, which
1122 returns MATCH_KETRPOS without going further in the pattern. By this means
1123 we can handle the group by iteration rather than recursion, thereby
1124 reducing the amount of stack needed. */
1125
1126 case OP_CBRAPOS:
1127 case OP_SCBRAPOS:
1128 allow_zero = FALSE;
1129
1130 POSSESSIVE_CAPTURE:
1131 number = GET2(ecode, 1+LINK_SIZE);
1132 offset = number << 1;
1133
1134 #ifdef PCRE_DEBUG
1135 printf("start possessive bracket %d\n", number);
1136 printf("subject=");
1137 pchars(eptr, 16, TRUE, md);
1138 printf("\n");
1139 #endif
1140
1141 if (offset < md->offset_max)
1142 {
1143 matched_once = FALSE;
1144 code_offset = (int)(ecode - md->start_code);
1145
1146 save_offset1 = md->offset_vector[offset];
1147 save_offset2 = md->offset_vector[offset+1];
1148 save_offset3 = md->offset_vector[md->offset_end - number];
1149 save_capture_last = md->capture_last;
1150
1151 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1152
1153 /* Each time round the loop, save the current subject position for use
1154 when the group matches. For MATCH_MATCH, the group has matched, so we
1155 restart it with a new subject starting position, remembering that we had
1156 at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157 usual. If we haven't matched any alternatives in any iteration, check to
1158 see if a previous iteration matched. If so, the group has matched;
1159 continue from afterwards. Otherwise it has failed; restore the previous
1160 capture values before returning NOMATCH. */
1161
1162 for (;;)
1163 {
1164 md->offset_vector[md->offset_end - number] =
1165 (int)(eptr - md->start_subject);
1166 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168 eptrb, RM63);
1169 if (rrc == MATCH_KETRPOS)
1170 {
1171 offset_top = md->end_offset_top;
1172 eptr = md->end_match_ptr;
1173 ecode = md->start_code + code_offset;
1174 save_capture_last = md->capture_last;
1175 matched_once = TRUE;
1176 continue;
1177 }
1178
1179 /* See comment in the code for capturing groups above about handling
1180 THEN. */
1181
1182 if (rrc == MATCH_THEN)
1183 {
1184 next = ecode + GET(ecode,1);
1185 if (md->start_match_ptr < next &&
1186 (*ecode == OP_ALT || *next == OP_ALT))
1187 rrc = MATCH_NOMATCH;
1188 }
1189
1190 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1191 md->capture_last = save_capture_last;
1192 ecode += GET(ecode, 1);
1193 if (*ecode != OP_ALT) break;
1194 }
1195
1196 if (!matched_once)
1197 {
1198 md->offset_vector[offset] = save_offset1;
1199 md->offset_vector[offset+1] = save_offset2;
1200 md->offset_vector[md->offset_end - number] = save_offset3;
1201 }
1202
1203 if (allow_zero || matched_once)
1204 {
1205 ecode += 1 + LINK_SIZE;
1206 break;
1207 }
1208
1209 RRETURN(MATCH_NOMATCH);
1210 }
1211
1212 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213 as a non-capturing bracket. */
1214
1215 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1217
1218 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1219
1220 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1222
1223 /* Non-capturing possessive bracket with unlimited repeat. We come here
1224 from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1225 without the capturing complication. It is written out separately for speed
1226 and cleanliness. */
1227
1228 case OP_BRAPOS:
1229 case OP_SBRAPOS:
1230 allow_zero = FALSE;
1231
1232 POSSESSIVE_NON_CAPTURE:
1233 matched_once = FALSE;
1234 code_offset = (int)(ecode - md->start_code);
1235 save_capture_last = md->capture_last;
1236
1237 for (;;)
1238 {
1239 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1241 eptrb, RM48);
1242 if (rrc == MATCH_KETRPOS)
1243 {
1244 offset_top = md->end_offset_top;
1245 eptr = md->end_match_ptr;
1246 ecode = md->start_code + code_offset;
1247 matched_once = TRUE;
1248 continue;
1249 }
1250
1251 /* See comment in the code for capturing groups above about handling
1252 THEN. */
1253
1254 if (rrc == MATCH_THEN)
1255 {
1256 next = ecode + GET(ecode,1);
1257 if (md->start_match_ptr < next &&
1258 (*ecode == OP_ALT || *next == OP_ALT))
1259 rrc = MATCH_NOMATCH;
1260 }
1261
1262 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1263 ecode += GET(ecode, 1);
1264 if (*ecode != OP_ALT) break;
1265 md->capture_last = save_capture_last;
1266 }
1267
1268 if (matched_once || allow_zero)
1269 {
1270 ecode += 1 + LINK_SIZE;
1271 break;
1272 }
1273 RRETURN(MATCH_NOMATCH);
1274
1275 /* Control never reaches here. */
1276
1277 /* Conditional group: compilation checked that there are no more than
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). If the reference is unset, there are two possibilities:
2746
2747 (a) In the default, Perl-compatible state, set the length negative;
2748 this ensures that every attempt at a match fails. We can't just fail
2749 here, because of the possibility of quantifiers with zero minima.
2750
2751 (b) If the JavaScript compatibility flag is set, set the length to zero
2752 so that the back reference matches an empty string.
2753
2754 Otherwise, set the length to the length of what was matched by the
2755 referenced subpattern.
2756
2757 The OP_REF and OP_REFI opcodes are used for a reference to a numbered group
2758 or to a non-duplicated named group. For a duplicated named group, OP_DNREF
2759 and OP_DNREFI are used. In this case we must scan the list of groups to
2760 which the name refers, and use the first one that is set. */
2761
2762 case OP_DNREF:
2763 case OP_DNREFI:
2764 caseless = op == OP_DNREFI;
2765 {
2766 int count = GET2(ecode, 1+IMM2_SIZE);
2767 pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
2768 ecode += 1 + 2*IMM2_SIZE;
2769
2770 while (count-- > 0)
2771 {
2772 offset = GET2(slot, 0) << 1;
2773 if (offset < offset_top && md->offset_vector[offset] >= 0) break;
2774 slot += md->name_entry_size;
2775 }
2776 if (count < 0)
2777 length = (md->jscript_compat)? 0 : -1;
2778 else
2779 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2780 }
2781 goto REF_REPEAT;
2782
2783 case OP_REF:
2784 case OP_REFI:
2785 caseless = op == OP_REFI;
2786 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
2787 ecode += 1 + IMM2_SIZE;
2788
2789
2790 if (offset >= offset_top || md->offset_vector[offset] < 0)
2791 length = (md->jscript_compat)? 0 : -1;
2792 else
2793 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2794
2795 /* Set up for repetition, or handle the non-repeated case */
2796
2797 REF_REPEAT:
2798 switch (*ecode)
2799 {
2800 case OP_CRSTAR:
2801 case OP_CRMINSTAR:
2802 case OP_CRPLUS:
2803 case OP_CRMINPLUS:
2804 case OP_CRQUERY:
2805 case OP_CRMINQUERY:
2806 c = *ecode++ - OP_CRSTAR;
2807 minimize = (c & 1) != 0;
2808 min = rep_min[c]; /* Pick up values from tables; */
2809 max = rep_max[c]; /* zero for max => infinity */
2810 if (max == 0) max = INT_MAX;
2811 break;
2812
2813 case OP_CRRANGE:
2814 case OP_CRMINRANGE:
2815 minimize = (*ecode == OP_CRMINRANGE);
2816 min = GET2(ecode, 1);
2817 max = GET2(ecode, 1 + IMM2_SIZE);
2818 if (max == 0) max = INT_MAX;
2819 ecode += 1 + 2 * IMM2_SIZE;
2820 break;
2821
2822 default: /* No repeat follows */
2823 if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2824 {
2825 if (length == -2) eptr = md->end_subject; /* Partial match */
2826 CHECK_PARTIAL();
2827 RRETURN(MATCH_NOMATCH);
2828 }
2829 eptr += length;
2830 continue; /* With the main loop */
2831 }
2832
2833 /* Handle repeated back references. If the length of the reference is
2834 zero, just continue with the main loop. If the length is negative, it
2835 means the reference is unset in non-Java-compatible mode. If the minimum is
2836 zero, we can continue at the same level without recursion. For any other
2837 minimum, carrying on will result in NOMATCH. */
2838
2839 if (length == 0) continue;
2840 if (length < 0 && min == 0) continue;
2841
2842 /* First, ensure the minimum number of matches are present. We get back
2843 the length of the reference string explicitly rather than passing the
2844 address of eptr, so that eptr can be a register variable. */
2845
2846 for (i = 1; i <= min; i++)
2847 {
2848 int slength;
2849 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2850 {
2851 if (slength == -2) eptr = md->end_subject; /* Partial match */
2852 CHECK_PARTIAL();
2853 RRETURN(MATCH_NOMATCH);
2854 }
2855 eptr += slength;
2856 }
2857
2858 /* If min = max, continue at the same level without recursion.
2859 They are not both allowed to be zero. */
2860
2861 if (min == max) continue;
2862
2863 /* If minimizing, keep trying and advancing the pointer */
2864
2865 if (minimize)
2866 {
2867 for (fi = min;; fi++)
2868 {
2869 int slength;
2870 RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2871 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2872 if (fi >= max) RRETURN(MATCH_NOMATCH);
2873 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2874 {
2875 if (slength == -2) eptr = md->end_subject; /* Partial match */
2876 CHECK_PARTIAL();
2877 RRETURN(MATCH_NOMATCH);
2878 }
2879 eptr += slength;
2880 }
2881 /* Control never gets here */
2882 }
2883
2884 /* If maximizing, find the longest string and work backwards */
2885
2886 else
2887 {
2888 pp = eptr;
2889 for (i = min; i < max; i++)
2890 {
2891 int slength;
2892 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2893 {
2894 /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2895 the soft partial matching case. */
2896
2897 if (slength == -2 && md->partial != 0 &&
2898 md->end_subject > md->start_used_ptr)
2899 {
2900 md->hitend = TRUE;
2901 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2902 }
2903 break;
2904 }
2905 eptr += slength;
2906 }
2907
2908 while (eptr >= pp)
2909 {
2910 RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2911 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2912 eptr -= length;
2913 }
2914 RRETURN(MATCH_NOMATCH);
2915 }
2916 /* Control never gets here */
2917
2918 /* Match a bit-mapped character class, possibly repeatedly. This op code is
2919 used when all the characters in the class have values in the range 0-255,
2920 and either the matching is caseful, or the characters are in the range
2921 0-127 when UTF-8 processing is enabled. The only difference between
2922 OP_CLASS and OP_NCLASS occurs when a data character outside the range is
2923 encountered.
2924
2925 First, look past the end of the item to see if there is repeat information
2926 following. Then obey similar code to character type repeats - written out
2927 again for speed. */
2928
2929 case OP_NCLASS:
2930 case OP_CLASS:
2931 {
2932 /* The data variable is saved across frames, so the byte map needs to
2933 be stored there. */
2934 #define BYTE_MAP ((pcre_uint8 *)data)
2935 data = ecode + 1; /* Save for matching */
2936 ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2937
2938 switch (*ecode)
2939 {
2940 case OP_CRSTAR:
2941 case OP_CRMINSTAR:
2942 case OP_CRPLUS:
2943 case OP_CRMINPLUS:
2944 case OP_CRQUERY:
2945 case OP_CRMINQUERY:
2946 c = *ecode++ - OP_CRSTAR;
2947 minimize = (c & 1) != 0;
2948 min = rep_min[c]; /* Pick up values from tables; */
2949 max = rep_max[c]; /* zero for max => infinity */
2950 if (max == 0) max = INT_MAX;
2951 break;
2952
2953 case OP_CRRANGE:
2954 case OP_CRMINRANGE:
2955 minimize = (*ecode == OP_CRMINRANGE);
2956 min = GET2(ecode, 1);
2957 max = GET2(ecode, 1 + IMM2_SIZE);
2958 if (max == 0) max = INT_MAX;
2959 ecode += 1 + 2 * IMM2_SIZE;
2960 break;
2961
2962 default: /* No repeat follows */
2963 min = max = 1;
2964 break;
2965 }
2966
2967 /* First, ensure the minimum number of matches are present. */
2968
2969 #ifdef SUPPORT_UTF
2970 if (utf)
2971 {
2972 for (i = 1; i <= min; i++)
2973 {
2974 if (eptr >= md->end_subject)
2975 {
2976 SCHECK_PARTIAL();
2977 RRETURN(MATCH_NOMATCH);
2978 }
2979 GETCHARINC(c, eptr);
2980 if (c > 255)
2981 {
2982 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2983 }
2984 else
2985 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2986 }
2987 }
2988 else
2989 #endif
2990 /* Not UTF mode */
2991 {
2992 for (i = 1; i <= min; i++)
2993 {
2994 if (eptr >= md->end_subject)
2995 {
2996 SCHECK_PARTIAL();
2997 RRETURN(MATCH_NOMATCH);
2998 }
2999 c = *eptr++;
3000 #ifndef COMPILE_PCRE8
3001 if (c > 255)
3002 {
3003 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3004 }
3005 else
3006 #endif
3007 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3008 }
3009 }
3010
3011 /* If max == min we can continue with the main loop without the
3012 need to recurse. */
3013
3014 if (min == max) continue;
3015
3016 /* If minimizing, keep testing the rest of the expression and advancing
3017 the pointer while it matches the class. */
3018
3019 if (minimize)
3020 {
3021 #ifdef SUPPORT_UTF
3022 if (utf)
3023 {
3024 for (fi = min;; fi++)
3025 {
3026 RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
3027 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3028 if (fi >= max) RRETURN(MATCH_NOMATCH);
3029 if (eptr >= md->end_subject)
3030 {
3031 SCHECK_PARTIAL();
3032 RRETURN(MATCH_NOMATCH);
3033 }
3034 GETCHARINC(c, eptr);
3035 if (c > 255)
3036 {
3037 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3038 }
3039 else
3040 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3041 }
3042 }
3043 else
3044 #endif
3045 /* Not UTF mode */
3046 {
3047 for (fi = min;; fi++)
3048 {
3049 RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
3050 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3051 if (fi >= max) RRETURN(MATCH_NOMATCH);
3052 if (eptr >= md->end_subject)
3053 {
3054 SCHECK_PARTIAL();
3055 RRETURN(MATCH_NOMATCH);
3056 }
3057 c = *eptr++;
3058 #ifndef COMPILE_PCRE8
3059 if (c > 255)
3060 {
3061 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3062 }
3063 else
3064 #endif
3065 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3066 }
3067 }
3068 /* Control never gets here */
3069 }
3070
3071 /* If maximizing, find the longest possible run, then work backwards. */
3072
3073 else
3074 {
3075 pp = eptr;
3076
3077 #ifdef SUPPORT_UTF
3078 if (utf)
3079 {
3080 for (i = min; i < max; i++)
3081 {
3082 int len = 1;
3083 if (eptr >= md->end_subject)
3084 {
3085 SCHECK_PARTIAL();
3086 break;
3087 }
3088 GETCHARLEN(c, eptr, len);
3089 if (c > 255)
3090 {
3091 if (op == OP_CLASS) break;
3092 }
3093 else
3094 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3095 eptr += len;
3096 }
3097 for (;;)
3098 {
3099 RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3100 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3101 if (eptr-- == pp) break; /* Stop if tried at original pos */
3102 BACKCHAR(eptr);
3103 }
3104 }
3105 else
3106 #endif
3107 /* Not UTF mode */
3108 {
3109 for (i = min; i < max; i++)
3110 {
3111 if (eptr >= md->end_subject)
3112 {
3113 SCHECK_PARTIAL();
3114 break;
3115 }
3116 c = *eptr;
3117 #ifndef COMPILE_PCRE8
3118 if (c > 255)
3119 {
3120 if (op == OP_CLASS) break;
3121 }
3122 else
3123 #endif
3124 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3125 eptr++;
3126 }
3127 while (eptr >= pp)
3128 {
3129 RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3130 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3131 eptr--;
3132 }
3133 }
3134
3135 RRETURN(MATCH_NOMATCH);
3136 }
3137 #undef BYTE_MAP
3138 }
3139 /* Control never gets here */
3140
3141
3142 /* Match an extended character class. This opcode is encountered only
3143 when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
3144 mode, because Unicode properties are supported in non-UTF-8 mode. */
3145
3146 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3147 case OP_XCLASS:
3148 {
3149 data = ecode + 1 + LINK_SIZE; /* Save for matching */
3150 ecode += GET(ecode, 1); /* Advance past the item */
3151
3152 switch (*ecode)
3153 {
3154 case OP_CRSTAR:
3155 case OP_CRMINSTAR:
3156 case OP_CRPLUS:
3157 case OP_CRMINPLUS:
3158 case OP_CRQUERY:
3159 case OP_CRMINQUERY:
3160 c = *ecode++ - OP_CRSTAR;
3161 minimize = (c & 1) != 0;
3162 min = rep_min[c]; /* Pick up values from tables; */
3163 max = rep_max[c]; /* zero for max => infinity */
3164 if (max == 0) max = INT_MAX;
3165 break;
3166
3167 case OP_CRRANGE:
3168 case OP_CRMINRANGE:
3169 minimize = (*ecode == OP_CRMINRANGE);
3170 min = GET2(ecode, 1);
3171 max = GET2(ecode, 1 + IMM2_SIZE);
3172 if (max == 0) max = INT_MAX;
3173 ecode += 1 + 2 * IMM2_SIZE;
3174 break;
3175
3176 default: /* No repeat follows */
3177 min = max = 1;
3178 break;
3179 }
3180
3181 /* First, ensure the minimum number of matches are present. */
3182
3183 for (i = 1; i <= min; i++)
3184 {
3185 if (eptr >= md->end_subject)
3186 {
3187 SCHECK_PARTIAL();
3188 RRETURN(MATCH_NOMATCH);
3189 }
3190 GETCHARINCTEST(c, eptr);
3191 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3192 }
3193
3194 /* If max == min we can continue with the main loop without the
3195 need to recurse. */
3196
3197 if (min == max) continue;
3198
3199 /* If minimizing, keep testing the rest of the expression and advancing
3200 the pointer while it matches the class. */
3201
3202 if (minimize)
3203 {
3204 for (fi = min;; fi++)
3205 {
3206 RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3207 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3208 if (fi >= max) RRETURN(MATCH_NOMATCH);
3209 if (eptr >= md->end_subject)
3210 {
3211 SCHECK_PARTIAL();
3212 RRETURN(MATCH_NOMATCH);
3213 }
3214 GETCHARINCTEST(c, eptr);
3215 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3216 }
3217 /* Control never gets here */
3218 }
3219
3220 /* If maximizing, find the longest possible run, then work backwards. */
3221
3222 else
3223 {
3224 pp = eptr;
3225 for (i = min; i < max; i++)
3226 {
3227 int len = 1;
3228 if (eptr >= md->end_subject)
3229 {
3230 SCHECK_PARTIAL();
3231 break;
3232 }
3233 #ifdef SUPPORT_UTF
3234 GETCHARLENTEST(c, eptr, len);
3235 #else
3236 c = *eptr;
3237 #endif
3238 if (!PRIV(xclass)(c, data, utf)) break;
3239 eptr += len;
3240 }
3241 for(;;)
3242 {
3243 RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3244 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3245 if (eptr-- == pp) break; /* Stop if tried at original pos */
3246 #ifdef SUPPORT_UTF
3247 if (utf) BACKCHAR(eptr);
3248 #endif
3249 }
3250 RRETURN(MATCH_NOMATCH);
3251 }
3252
3253 /* Control never gets here */
3254 }
3255 #endif /* End of XCLASS */
3256
3257 /* Match a single character, casefully */
3258
3259 case OP_CHAR:
3260 #ifdef SUPPORT_UTF
3261 if (utf)
3262 {
3263 length = 1;
3264 ecode++;
3265 GETCHARLEN(fc, ecode, length);
3266 if (length > md->end_subject - eptr)
3267 {
3268 CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */
3269 RRETURN(MATCH_NOMATCH);
3270 }
3271 while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3272 }
3273 else
3274 #endif
3275 /* Not UTF mode */
3276 {
3277 if (md->end_subject - eptr < 1)
3278 {
3279 SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */
3280 RRETURN(MATCH_NOMATCH);
3281 }
3282 if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3283 ecode += 2;
3284 }
3285 break;
3286
3287 /* Match a single character, caselessly. If we are at the end of the
3288 subject, give up immediately. */
3289
3290 case OP_CHARI:
3291 if (eptr >= md->end_subject)
3292 {
3293 SCHECK_PARTIAL();
3294 RRETURN(MATCH_NOMATCH);
3295 }
3296
3297 #ifdef SUPPORT_UTF
3298 if (utf)
3299 {
3300 length = 1;
3301 ecode++;
3302 GETCHARLEN(fc, ecode, length);
3303
3304 /* If the pattern character's value is < 128, we have only one byte, and
3305 we know that its other case must also be one byte long, so we can use the
3306 fast lookup table. We know that there is at least one byte left in the
3307 subject. */
3308
3309 if (fc < 128)
3310 {
3311 pcre_uint32 cc = RAWUCHAR(eptr);
3312 if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3313 ecode++;
3314 eptr++;
3315 }
3316
3317 /* Otherwise we must pick up the subject character. Note that we cannot
3318 use the value of "length" to check for sufficient bytes left, because the
3319 other case of the character may have more or fewer bytes. */
3320
3321 else
3322 {
3323 pcre_uint32 dc;
3324 GETCHARINC(dc, eptr);
3325 ecode += length;
3326
3327 /* If we have Unicode property support, we can use it to test the other
3328 case of the character, if there is one. */
3329
3330 if (fc != dc)
3331 {
3332 #ifdef SUPPORT_UCP
3333 if (dc != UCD_OTHERCASE(fc))
3334 #endif
3335 RRETURN(MATCH_NOMATCH);
3336 }
3337 }
3338 }
3339 else
3340 #endif /* SUPPORT_UTF */
3341
3342 /* Not UTF mode */
3343 {
3344 if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3345 != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3346 eptr++;
3347 ecode += 2;
3348 }
3349 break;
3350
3351 /* Match a single character repeatedly. */
3352
3353 case OP_EXACT:
3354 case OP_EXACTI:
3355 min = max = GET2(ecode, 1);
3356 ecode += 1 + IMM2_SIZE;
3357 goto REPEATCHAR;
3358
3359 case OP_POSUPTO:
3360 case OP_POSUPTOI:
3361 possessive = TRUE;
3362 /* Fall through */
3363
3364 case OP_UPTO:
3365 case OP_UPTOI:
3366 case OP_MINUPTO:
3367 case OP_MINUPTOI:
3368 min = 0;
3369 max = GET2(ecode, 1);
3370 minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3371 ecode += 1 + IMM2_SIZE;
3372 goto REPEATCHAR;
3373
3374 case OP_POSSTAR:
3375 case OP_POSSTARI:
3376 possessive = TRUE;
3377 min = 0;
3378 max = INT_MAX;
3379 ecode++;
3380 goto REPEATCHAR;
3381
3382 case OP_POSPLUS:
3383 case OP_POSPLUSI:
3384 possessive = TRUE;
3385 min = 1;
3386 max = INT_MAX;
3387 ecode++;
3388 goto REPEATCHAR;
3389
3390 case OP_POSQUERY:
3391 case OP_POSQUERYI:
3392 possessive = TRUE;
3393 min = 0;
3394 max = 1;
3395 ecode++;
3396 goto REPEATCHAR;
3397
3398 case OP_STAR:
3399 case OP_STARI:
3400 case OP_MINSTAR:
3401 case OP_MINSTARI:
3402 case OP_PLUS:
3403 case OP_PLUSI:
3404 case OP_MINPLUS:
3405 case OP_MINPLUSI:
3406 case OP_QUERY:
3407 case OP_QUERYI:
3408 case OP_MINQUERY:
3409 case OP_MINQUERYI:
3410 c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3411 minimize = (c & 1) != 0;
3412 min = rep_min[c]; /* Pick up values from tables; */
3413 max = rep_max[c]; /* zero for max => infinity */
3414 if (max == 0) max = INT_MAX;
3415
3416 /* Common code for all repeated single-character matches. We first check
3417 for the minimum number of characters. If the minimum equals the maximum, we
3418 are done. Otherwise, if minimizing, check the rest of the pattern for a
3419 match; if there isn't one, advance up to the maximum, one character at a
3420 time.
3421
3422 If maximizing, advance up to the maximum number of matching characters,
3423 until eptr is past the end of the maximum run. If possessive, we are
3424 then done (no backing up). Otherwise, match at this position; anything
3425 other than no match is immediately returned. For nomatch, back up one
3426 character, unless we are matching \R and the last thing matched was
3427 \r\n, in which case, back up two bytes. When we reach the first optional
3428 character position, we can save stack by doing a tail recurse.
3429
3430 The various UTF/non-UTF and caseful/caseless cases are handled separately,
3431 for speed. */
3432
3433 REPEATCHAR:
3434 #ifdef SUPPORT_UTF
3435 if (utf)
3436 {
3437 length = 1;
3438 charptr = ecode;
3439 GETCHARLEN(fc, ecode, length);
3440 ecode += length;
3441
3442 /* Handle multibyte character matching specially here. There is
3443 support for caseless matching if UCP support is present. */
3444
3445 if (length > 1)
3446 {
3447 #ifdef SUPPORT_UCP
3448 pcre_uint32 othercase;
3449 if (op >= OP_STARI && /* Caseless */
3450 (othercase = UCD_OTHERCASE(fc)) != fc)
3451 oclength = PRIV(ord2utf)(othercase, occhars);
3452 else oclength = 0;
3453 #endif /* SUPPORT_UCP */
3454
3455 for (i = 1; i <= min; i++)
3456 {
3457 if (eptr <= md->end_subject - length &&
3458 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3459 #ifdef SUPPORT_UCP
3460 else if (oclength > 0 &&
3461 eptr <= md->end_subject - oclength &&
3462 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3463 #endif /* SUPPORT_UCP */
3464 else
3465 {
3466 CHECK_PARTIAL();
3467 RRETURN(MATCH_NOMATCH);
3468 }
3469 }
3470
3471 if (min == max) continue;
3472
3473 if (minimize)
3474 {
3475 for (fi = min;; fi++)
3476 {
3477 RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3478 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3479 if (fi >= max) RRETURN(MATCH_NOMATCH);
3480 if (eptr <= md->end_subject - length &&
3481 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3482 #ifdef SUPPORT_UCP
3483 else if (oclength > 0 &&
3484 eptr <= md->end_subject - oclength &&
3485 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3486 #endif /* SUPPORT_UCP */
3487 else
3488 {
3489 CHECK_PARTIAL();
3490 RRETURN(MATCH_NOMATCH);
3491 }
3492 }
3493 /* Control never gets here */
3494 }
3495
3496 else /* Maximize */
3497 {
3498 pp = eptr;
3499 for (i = min; i < max; i++)
3500 {
3501 if (eptr <= md->end_subject - length &&
3502 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3503 #ifdef SUPPORT_UCP
3504 else if (oclength > 0 &&
3505 eptr <= md->end_subject - oclength &&
3506 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3507 #endif /* SUPPORT_UCP */
3508 else
3509 {
3510 CHECK_PARTIAL();
3511 break;
3512 }
3513 }
3514
3515 if (possessive) continue; /* No backtracking */
3516 for(;;)
3517 {
3518 if (eptr == pp) goto TAIL_RECURSE;
3519 RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3520 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3521 #ifdef SUPPORT_UCP
3522 eptr--;
3523 BACKCHAR(eptr);
3524 #else /* without SUPPORT_UCP */
3525 eptr -= length;
3526 #endif /* SUPPORT_UCP */
3527 }
3528 }
3529 /* Control never gets here */
3530 }
3531
3532 /* If the length of a UTF-8 character is 1, we fall through here, and
3533 obey the code as for non-UTF-8 characters below, though in this case the
3534 value of fc will always be < 128. */
3535 }
3536 else
3537 #endif /* SUPPORT_UTF */
3538 /* When not in UTF-8 mode, load a single-byte character. */
3539 fc = *ecode++;
3540
3541 /* The value of fc at this point is always one character, though we may
3542 or may not be in UTF mode. The code is duplicated for the caseless and
3543 caseful cases, for speed, since matching characters is likely to be quite
3544 common. First, ensure the minimum number of matches are present. If min =
3545 max, continue at the same level without recursing. Otherwise, if
3546 minimizing, keep trying the rest of the expression and advancing one
3547 matching character if failing, up to the maximum. Alternatively, if
3548 maximizing, find the maximum number of characters and work backwards. */
3549
3550 DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3551 max, (char *)eptr));
3552
3553 if (op >= OP_STARI) /* Caseless */
3554 {
3555 #ifdef COMPILE_PCRE8
3556 /* fc must be < 128 if UTF is enabled. */
3557 foc = md->fcc[fc];
3558 #else
3559 #ifdef SUPPORT_UTF
3560 #ifdef SUPPORT_UCP
3561 if (utf && fc > 127)
3562 foc = UCD_OTHERCASE(fc);
3563 #else
3564 if (utf && fc > 127)
3565 foc = fc;
3566 #endif /* SUPPORT_UCP */
3567 else
3568 #endif /* SUPPORT_UTF */
3569 foc = TABLE_GET(fc, md->fcc, fc);
3570 #endif /* COMPILE_PCRE8 */
3571
3572 for (i = 1; i <= min; i++)
3573 {
3574 pcre_uint32 cc; /* Faster than pcre_uchar */
3575 if (eptr >= md->end_subject)
3576 {
3577 SCHECK_PARTIAL();
3578 RRETURN(MATCH_NOMATCH);
3579 }
3580 cc = RAWUCHARTEST(eptr);
3581 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3582 eptr++;
3583 }
3584 if (min == max) continue;
3585 if (minimize)
3586 {
3587 for (fi = min;; fi++)
3588 {
3589 pcre_uint32 cc; /* Faster than pcre_uchar */
3590 RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3591 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3592 if (fi >= max) RRETURN(MATCH_NOMATCH);
3593 if (eptr >= md->end_subject)
3594 {
3595 SCHECK_PARTIAL();
3596 RRETURN(MATCH_NOMATCH);
3597 }
3598 cc = RAWUCHARTEST(eptr);
3599 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3600 eptr++;
3601 }
3602 /* Control never gets here */
3603 }
3604 else /* Maximize */
3605 {
3606 pp = eptr;
3607 for (i = min; i < max; i++)
3608 {
3609 pcre_uint32 cc; /* Faster than pcre_uchar */
3610 if (eptr >= md->end_subject)
3611 {
3612 SCHECK_PARTIAL();
3613 break;
3614 }
3615 cc = RAWUCHARTEST(eptr);
3616 if (fc != cc && foc != cc) break;
3617 eptr++;
3618 }
3619 if (possessive) continue; /* No backtracking */
3620 for (;;)
3621 {
3622 if (eptr == pp) goto TAIL_RECURSE;
3623 RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3624 eptr--;
3625 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3626 }
3627 /* Control never gets here */
3628 }
3629 }
3630
3631 /* Caseful comparisons (includes all multi-byte characters) */
3632
3633 else
3634 {
3635 for (i = 1; i <= min; i++)
3636 {
3637 if (eptr >= md->end_subject)
3638 {
3639 SCHECK_PARTIAL();
3640 RRETURN(MATCH_NOMATCH);
3641 }
3642 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3643 }
3644
3645 if (min == max) continue;
3646
3647 if (minimize)
3648 {
3649 for (fi = min;; fi++)
3650 {
3651 RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3652 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3653 if (fi >= max) RRETURN(MATCH_NOMATCH);
3654 if (eptr >= md->end_subject)
3655 {
3656 SCHECK_PARTIAL();
3657 RRETURN(MATCH_NOMATCH);
3658 }
3659 if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3660 }
3661 /* Control never gets here */
3662 }
3663 else /* Maximize */
3664 {
3665 pp = eptr;
3666 for (i = min; i < max; i++)
3667 {
3668 if (eptr >= md->end_subject)
3669 {
3670 SCHECK_PARTIAL();
3671 break;
3672 }
3673 if (fc != RAWUCHARTEST(eptr)) break;
3674 eptr++;
3675 }
3676 if (possessive) continue; /* No backtracking */
3677 for (;;)
3678 {
3679 if (eptr == pp) goto TAIL_RECURSE;
3680 RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3681 eptr--;
3682 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3683 }
3684 /* Control never gets here */
3685 }
3686 }
3687 /* Control never gets here */
3688
3689 /* Match a negated single one-byte character. The character we are
3690 checking can be multibyte. */
3691
3692 case OP_NOT:
3693 case OP_NOTI:
3694 if (eptr >= md->end_subject)
3695 {
3696 SCHECK_PARTIAL();
3697 RRETURN(MATCH_NOMATCH);
3698 }
3699 #ifdef SUPPORT_UTF
3700 if (utf)
3701 {
3702 register pcre_uint32 ch, och;
3703
3704 ecode++;
3705 GETCHARINC(ch, ecode);
3706 GETCHARINC(c, eptr);
3707
3708 if (op == OP_NOT)
3709 {
3710 if (ch == c) RRETURN(MATCH_NOMATCH);
3711 }
3712 else
3713 {
3714 #ifdef SUPPORT_UCP
3715 if (ch > 127)
3716 och = UCD_OTHERCASE(ch);
3717 #else
3718 if (ch > 127)
3719 och = ch;
3720 #endif /* SUPPORT_UCP */
3721 else
3722 och = TABLE_GET(ch, md->fcc, ch);
3723 if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3724 }
3725 }
3726 else
3727 #endif
3728 {
3729 register pcre_uint32 ch = ecode[1];
3730 c = *eptr++;
3731 if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3732 RRETURN(MATCH_NOMATCH);
3733 ecode += 2;
3734 }
3735 break;
3736
3737 /* Match a negated single one-byte character repeatedly. This is almost a
3738 repeat of the code for a repeated single character, but I haven't found a
3739 nice way of commoning these up that doesn't require a test of the
3740 positive/negative option for each character match. Maybe that wouldn't add
3741 very much to the time taken, but character matching *is* what this is all
3742 about... */
3743
3744 case OP_NOTEXACT:
3745 case OP_NOTEXACTI:
3746 min = max = GET2(ecode, 1);
3747 ecode += 1 + IMM2_SIZE;
3748 goto REPEATNOTCHAR;
3749
3750 case OP_NOTUPTO:
3751 case OP_NOTUPTOI:
3752 case OP_NOTMINUPTO:
3753 case OP_NOTMINUPTOI:
3754 min = 0;
3755 max = GET2(ecode, 1);
3756 minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3757 ecode += 1 + IMM2_SIZE;
3758 goto REPEATNOTCHAR;
3759
3760 case OP_NOTPOSSTAR:
3761 case OP_NOTPOSSTARI:
3762 possessive = TRUE;
3763 min = 0;
3764 max = INT_MAX;
3765 ecode++;
3766 goto REPEATNOTCHAR;
3767
3768 case OP_NOTPOSPLUS:
3769 case OP_NOTPOSPLUSI:
3770 possessive = TRUE;
3771 min = 1;
3772 max = INT_MAX;
3773 ecode++;
3774 goto REPEATNOTCHAR;
3775
3776 case OP_NOTPOSQUERY:
3777 case OP_NOTPOSQUERYI:
3778 possessive = TRUE;
3779 min = 0;
3780 max = 1;
3781 ecode++;
3782 goto REPEATNOTCHAR;
3783
3784 case OP_NOTPOSUPTO:
3785 case OP_NOTPOSUPTOI:
3786 possessive = TRUE;
3787 min = 0;
3788 max = GET2(ecode, 1);
3789 ecode += 1 + IMM2_SIZE;
3790 goto REPEATNOTCHAR;
3791
3792 case OP_NOTSTAR:
3793 case OP_NOTSTARI:
3794 case OP_NOTMINSTAR:
3795 case OP_NOTMINSTARI:
3796 case OP_NOTPLUS:
3797 case OP_NOTPLUSI:
3798 case OP_NOTMINPLUS:
3799 case OP_NOTMINPLUSI:
3800 case OP_NOTQUERY:
3801 case OP_NOTQUERYI:
3802 case OP_NOTMINQUERY:
3803 case OP_NOTMINQUERYI:
3804 c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3805 minimize = (c & 1) != 0;
3806 min = rep_min[c]; /* Pick up values from tables; */
3807 max = rep_max[c]; /* zero for max => infinity */
3808 if (max == 0) max = INT_MAX;
3809
3810 /* Common code for all repeated single-byte matches. */
3811
3812 REPEATNOTCHAR:
3813 GETCHARINCTEST(fc, ecode);
3814
3815 /* The code is duplicated for the caseless and caseful cases, for speed,
3816 since matching characters is likely to be quite common. First, ensure the
3817 minimum number of matches are present. If min = max, continue at the same
3818 level without recursing. Otherwise, if minimizing, keep trying the rest of
3819 the expression and advancing one matching character if failing, up to the
3820 maximum. Alternatively, if maximizing, find the maximum number of
3821 characters and work backwards. */
3822
3823 DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3824 max, (char *)eptr));
3825
3826 if (op >= OP_NOTSTARI) /* Caseless */
3827 {
3828 #ifdef SUPPORT_UTF
3829 #ifdef SUPPORT_UCP
3830 if (utf && fc > 127)
3831 foc = UCD_OTHERCASE(fc);
3832 #else
3833 if (utf && fc > 127)
3834 foc = fc;
3835 #endif /* SUPPORT_UCP */
3836 else
3837 #endif /* SUPPORT_UTF */
3838 foc = TABLE_GET(fc, md->fcc, fc);
3839
3840 #ifdef SUPPORT_UTF
3841 if (utf)
3842 {
3843 register pcre_uint32 d;
3844 for (i = 1; i <= min; i++)
3845 {
3846 if (eptr >= md->end_subject)
3847 {
3848 SCHECK_PARTIAL();
3849 RRETURN(MATCH_NOMATCH);
3850 }
3851 GETCHARINC(d, eptr);
3852 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3853 }
3854 }
3855 else
3856 #endif /* SUPPORT_UTF */
3857 /* Not UTF mode */
3858 {
3859 for (i = 1; i <= min; i++)
3860 {
3861 if (eptr >= md->end_subject)
3862 {
3863 SCHECK_PARTIAL();
3864 RRETURN(MATCH_NOMATCH);
3865 }
3866 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3867 eptr++;
3868 }
3869 }
3870
3871 if (min == max) continue;
3872
3873 if (minimize)
3874 {
3875 #ifdef SUPPORT_UTF
3876 if (utf)
3877 {
3878 register pcre_uint32 d;
3879 for (fi = min;; fi++)
3880 {
3881 RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3882 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3883 if (fi >= max) RRETURN(MATCH_NOMATCH);
3884 if (eptr >= md->end_subject)
3885 {
3886 SCHECK_PARTIAL();
3887 RRETURN(MATCH_NOMATCH);
3888 }
3889 GETCHARINC(d, eptr);
3890 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3891 }
3892 }
3893 else
3894 #endif /*SUPPORT_UTF */
3895 /* Not UTF mode */
3896 {
3897 for (fi = min;; fi++)
3898 {
3899 RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3900 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3901 if (fi >= max) RRETURN(MATCH_NOMATCH);
3902 if (eptr >= md->end_subject)
3903 {
3904 SCHECK_PARTIAL();
3905 RRETURN(MATCH_NOMATCH);
3906 }
3907 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3908 eptr++;
3909 }
3910 }
3911 /* Control never gets here */
3912 }
3913
3914 /* Maximize case */
3915
3916 else
3917 {
3918 pp = eptr;
3919
3920 #ifdef SUPPORT_UTF
3921 if (utf)
3922 {
3923 register pcre_uint32 d;
3924 for (i = min; i < max; i++)
3925 {
3926 int len = 1;
3927 if (eptr >= md->end_subject)
3928 {
3929 SCHECK_PARTIAL();
3930 break;
3931 }
3932 GETCHARLEN(d, eptr, len);
3933 if (fc == d || (unsigned int)foc == d) break;
3934 eptr += len;
3935 }
3936 if (possessive) continue; /* No backtracking */
3937 for(;;)
3938 {
3939 if (eptr == pp) goto TAIL_RECURSE;
3940 RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3941 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3942 eptr--;
3943 BACKCHAR(eptr);
3944 }
3945 }
3946 else
3947 #endif /* SUPPORT_UTF */
3948 /* Not UTF mode */
3949 {
3950 for (i = min; i < max; i++)
3951 {
3952 if (eptr >= md->end_subject)
3953 {
3954 SCHECK_PARTIAL();
3955 break;
3956 }
3957 if (fc == *eptr || foc == *eptr) break;
3958 eptr++;
3959 }
3960 if (possessive) continue; /* No backtracking */
3961 for (;;)
3962 {
3963 if (eptr == pp) goto TAIL_RECURSE;
3964 RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3965 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3966 eptr--;
3967 }
3968 }
3969 /* Control never gets here */
3970 }
3971 }
3972
3973 /* Caseful comparisons */
3974
3975 else
3976 {
3977 #ifdef SUPPORT_UTF
3978 if (utf)
3979 {
3980 register pcre_uint32 d;
3981 for (i = 1; i <= min; i++)
3982 {
3983 if (eptr >= md->end_subject)
3984 {
3985 SCHECK_PARTIAL();
3986 RRETURN(MATCH_NOMATCH);
3987 }
3988 GETCHARINC(d, eptr);
3989 if (fc == d) RRETURN(MATCH_NOMATCH);
3990 }
3991 }
3992 else
3993 #endif
3994 /* Not UTF mode */
3995 {
3996 for (i = 1; i <= min; i++)
3997 {
3998 if (eptr >= md->end_subject)
3999 {
4000 SCHECK_PARTIAL();
4001 RRETURN(MATCH_NOMATCH);
4002 }
4003 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
4004 }
4005 }
4006
4007 if (min == max) continue;
4008
4009 if (minimize)
4010 {
4011 #ifdef SUPPORT_UTF
4012 if (utf)
4013 {
4014 register pcre_uint32 d;
4015 for (fi = min;; fi++)
4016 {
4017 RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
4018 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4019 if (fi >= max) RRETURN(MATCH_NOMATCH);
4020 if (eptr >= md->end_subject)
4021 {
4022 SCHECK_PARTIAL();
4023 RRETURN(MATCH_NOMATCH);
4024 }
4025 GETCHARINC(d, eptr);
4026 if (fc == d) RRETURN(MATCH_NOMATCH);
4027 }
4028 }
4029 else
4030 #endif
4031 /* Not UTF mode */
4032 {
4033 for (fi = min;; fi++)
4034 {
4035 RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
4036 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4037 if (fi >= max) RRETURN(MATCH_NOMATCH);
4038 if (eptr >= md->end_subject)
4039 {
4040 SCHECK_PARTIAL();
4041 RRETURN(MATCH_NOMATCH);
4042 }
4043 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
4044 }
4045 }
4046 /* Control never gets here */
4047 }
4048
4049 /* Maximize case */
4050
4051 else
4052 {
4053 pp = eptr;
4054
4055 #ifdef SUPPORT_UTF
4056 if (utf)
4057 {
4058 register pcre_uint32 d;
4059 for (i = min; i < max; i++)
4060 {
4061 int len = 1;
4062 if (eptr >= md->end_subject)
4063 {
4064 SCHECK_PARTIAL();
4065 break;
4066 }
4067 GETCHARLEN(d, eptr, len);
4068 if (fc == d) break;
4069 eptr += len;
4070 }
4071 if (possessive) continue; /* No backtracking */
4072 for(;;)
4073 {
4074 if (eptr == pp) goto TAIL_RECURSE;
4075 RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
4076 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4077 eptr--;
4078 BACKCHAR(eptr);
4079 }
4080 }
4081 else
4082 #endif
4083 /* Not UTF mode */
4084 {
4085 for (i = min; i < max; i++)
4086 {
4087 if (eptr >= md->end_subject)
4088 {
4089 SCHECK_PARTIAL();
4090 break;
4091 }
4092 if (fc == *eptr) break;
4093 eptr++;
4094 }
4095 if (possessive) continue; /* No backtracking */
4096 for (;;)
4097 {
4098 if (eptr == pp) goto TAIL_RECURSE;
4099 RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
4100 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4101 eptr--;
4102 }
4103 }
4104 /* Control never gets here */
4105 }
4106 }
4107 /* Control never gets here */
4108
4109 /* Match a single character type repeatedly; several different opcodes
4110 share code. This is very similar to the code for single characters, but we
4111 repeat it in the interests of efficiency. */
4112
4113 case OP_TYPEEXACT:
4114 min = max = GET2(ecode, 1);
4115 minimize = TRUE;
4116 ecode += 1 + IMM2_SIZE;
4117 goto REPEATTYPE;
4118
4119 case OP_TYPEUPTO:
4120 case OP_TYPEMINUPTO:
4121 min = 0;
4122 max = GET2(ecode, 1);
4123 minimize = *ecode == OP_TYPEMINUPTO;
4124 ecode += 1 + IMM2_SIZE;
4125 goto REPEATTYPE;
4126
4127 case OP_TYPEPOSSTAR:
4128 possessive = TRUE;
4129 min = 0;
4130 max = INT_MAX;
4131 ecode++;
4132 goto REPEATTYPE;
4133
4134 case OP_TYPEPOSPLUS:
4135 possessive = TRUE;
4136 min = 1;
4137 max = INT_MAX;
4138 ecode++;
4139 goto REPEATTYPE;
4140
4141 case OP_TYPEPOSQUERY:
4142 possessive = TRUE;
4143 min = 0;
4144 max = 1;
4145 ecode++;
4146 goto REPEATTYPE;
4147
4148 case OP_TYPEPOSUPTO:
4149 possessive = TRUE;
4150 min = 0;
4151 max = GET2(ecode, 1);
4152 ecode += 1 + IMM2_SIZE;
4153 goto REPEATTYPE;
4154
4155 case OP_TYPESTAR:
4156 case OP_TYPEMINSTAR:
4157 case OP_TYPEPLUS:
4158 case OP_TYPEMINPLUS:
4159 case OP_TYPEQUERY:
4160 case OP_TYPEMINQUERY:
4161 c = *ecode++ - OP_TYPESTAR;
4162 minimize = (c & 1) != 0;
4163 min = rep_min[c]; /* Pick up values from tables; */
4164 max = rep_max[c]; /* zero for max => infinity */
4165 if (max == 0) max = INT_MAX;
4166
4167 /* Common code for all repeated single character type matches. Note that
4168 in UTF-8 mode, '.' matches a character of any length, but for the other
4169 character types, the valid characters are all one-byte long. */
4170
4171 REPEATTYPE:
4172 ctype = *ecode++; /* Code for the character type */
4173
4174 #ifdef SUPPORT_UCP
4175 if (ctype == OP_PROP || ctype == OP_NOTPROP)
4176 {
4177 prop_fail_result = ctype == OP_NOTPROP;
4178 prop_type = *ecode++;
4179 prop_value = *ecode++;
4180 }
4181 else prop_type = -1;
4182 #endif
4183
4184 /* First, ensure the minimum number of matches are present. Use inline
4185 code for maximizing the speed, and do the type test once at the start
4186 (i.e. keep it out of the loop). Separate the UTF-8 code completely as that
4187 is tidier. Also separate the UCP code, which can be the same for both UTF-8
4188 and single-bytes. */
4189
4190 if (min > 0)
4191 {
4192 #ifdef SUPPORT_UCP
4193 if (prop_type >= 0)
4194 {
4195 switch(prop_type)
4196 {
4197 case PT_ANY:
4198 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4199 for (i = 1; i <= min; i++)
4200 {
4201 if (eptr >= md->end_subject)
4202 {
4203 SCHECK_PARTIAL();
4204 RRETURN(MATCH_NOMATCH);
4205 }
4206 GETCHARINCTEST(c, eptr);
4207 }
4208 break;
4209
4210 case PT_LAMP:
4211 for (i = 1; i <= min; i++)
4212 {
4213 int chartype;
4214 if (eptr >= md->end_subject)
4215 {
4216 SCHECK_PARTIAL();
4217 RRETURN(MATCH_NOMATCH);
4218 }
4219 GETCHARINCTEST(c, eptr);
4220 chartype = UCD_CHARTYPE(c);
4221 if ((chartype == ucp_Lu ||
4222 chartype == ucp_Ll ||
4223 chartype == ucp_Lt) == prop_fail_result)
4224 RRETURN(MATCH_NOMATCH);
4225 }
4226 break;
4227
4228 case PT_GC:
4229 for (i = 1; i <= min; i++)
4230 {
4231 if (eptr >= md->end_subject)
4232 {
4233 SCHECK_PARTIAL();
4234 RRETURN(MATCH_NOMATCH);
4235 }
4236 GETCHARINCTEST(c, eptr);
4237 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4238 RRETURN(MATCH_NOMATCH);
4239 }
4240 break;
4241
4242 case PT_PC:
4243 for (i = 1; i <= min; i++)
4244 {
4245 if (eptr >= md->end_subject)
4246 {
4247 SCHECK_PARTIAL();
4248 RRETURN(MATCH_NOMATCH);
4249 }
4250 GETCHARINCTEST(c, eptr);
4251 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4252 RRETURN(MATCH_NOMATCH);
4253 }
4254 break;
4255
4256 case PT_SC:
4257 for (i = 1; i <= min; i++)
4258 {
4259 if (eptr >= md->end_subject)
4260 {
4261 SCHECK_PARTIAL();
4262 RRETURN(MATCH_NOMATCH);
4263 }
4264 GETCHARINCTEST(c, eptr);
4265 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4266 RRETURN(MATCH_NOMATCH);
4267 }
4268 break;
4269
4270 case PT_ALNUM:
4271 for (i = 1; i <= min; i++)
4272 {
4273 int category;
4274 if (eptr >= md->end_subject)
4275 {
4276 SCHECK_PARTIAL();
4277 RRETURN(MATCH_NOMATCH);
4278 }
4279 GETCHARINCTEST(c, eptr);
4280 category = UCD_CATEGORY(c);
4281 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4282 RRETURN(MATCH_NOMATCH);
4283 }
4284 break;
4285
4286 case PT_SPACE: /* Perl space */
4287 for (i = 1; i <= min; i++)
4288 {
4289 if (eptr >= md->end_subject)
4290 {
4291 SCHECK_PARTIAL();
4292 RRETURN(MATCH_NOMATCH);
4293 }
4294 GETCHARINCTEST(c, eptr);
4295 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
4296 c == CHAR_FF || c == CHAR_CR)
4297 == prop_fail_result)
4298 RRETURN(MATCH_NOMATCH);
4299 }
4300 break;
4301
4302 case PT_PXSPACE: /* POSIX space */
4303 for (i = 1; i <= min; i++)
4304 {
4305 if (eptr >= md->end_subject)
4306 {
4307 SCHECK_PARTIAL();
4308 RRETURN(MATCH_NOMATCH);
4309 }
4310 GETCHARINCTEST(c, eptr);
4311 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
4312 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4313 == prop_fail_result)
4314 RRETURN(MATCH_NOMATCH);
4315 }
4316 break;
4317
4318 case PT_WORD:
4319 for (i = 1; i <= min; i++)
4320 {
4321 int category;
4322 if (eptr >= md->end_subject)
4323 {
4324 SCHECK_PARTIAL();
4325 RRETURN(MATCH_NOMATCH);
4326 }
4327 GETCHARINCTEST(c, eptr);
4328 category = UCD_CATEGORY(c);
4329 if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
4330 == prop_fail_result)
4331 RRETURN(MATCH_NOMATCH);
4332 }
4333 break;
4334
4335 case PT_CLIST:
4336 for (i = 1; i <= min; i++)
4337 {
4338 const pcre_uint32 *cp;
4339 if (eptr >= md->end_subject)
4340 {
4341 SCHECK_PARTIAL();
4342 RRETURN(MATCH_NOMATCH);
4343 }
4344 GETCHARINCTEST(c, eptr);
4345 cp = PRIV(ucd_caseless_sets) + prop_value;
4346 for (;;)
4347 {
4348 if (c < *cp)
4349 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
4350 if (c == *cp++)
4351 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
4352 }
4353 }
4354 break;
4355
4356 case PT_UCNC:
4357 for (i = 1; i <= min; i++)
4358 {
4359 if (eptr >= md->end_subject)
4360 {
4361 SCHECK_PARTIAL();
4362 RRETURN(MATCH_NOMATCH);
4363 }
4364 GETCHARINCTEST(c, eptr);
4365 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
4366 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
4367 c >= 0xe000) == prop_fail_result)
4368 RRETURN(MATCH_NOMATCH);
4369 }
4370 break;
4371
4372 /* This should not occur */
4373
4374 default:
4375 RRETURN(PCRE_ERROR_INTERNAL);
4376 }
4377 }
4378
4379 /* Match extended Unicode sequences. We will get here only if the
4380 support is in the binary; otherwise a compile-time error occurs. */
4381
4382 else if (ctype == OP_EXTUNI)
4383 {
4384 for (i = 1; i <= min; i++)
4385 {
4386 if (eptr >= md->end_subject)
4387 {
4388 SCHECK_PARTIAL();
4389 RRETURN(MATCH_NOMATCH);
4390 }
4391 else
4392 {
4393 int lgb, rgb;
4394 GETCHARINCTEST(c, eptr);
4395 lgb = UCD_GRAPHBREAK(c);
4396 while (eptr < md->end_subject)
4397 {
4398 int len = 1;
4399 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4400 rgb = UCD_GRAPHBREAK(c);
4401 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
4402 lgb = rgb;
4403 eptr += len;
4404 }
4405 }
4406 CHECK_PARTIAL();
4407 }
4408 }
4409
4410 else
4411 #endif /* SUPPORT_UCP */
4412
4413 /* Handle all other cases when the coding is UTF-8 */
4414
4415 #ifdef SUPPORT_UTF
4416 if (utf) switch(ctype)
4417 {
4418 case OP_ANY:
4419 for (i = 1; i <= min; i++)
4420 {
4421 if (eptr >= md->end_subject)
4422 {
4423 SCHECK_PARTIAL();
4424 RRETURN(MATCH_NOMATCH);
4425 }
4426 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4427 if (md->partial != 0 &&
4428 eptr + 1 >= md->end_subject &&
4429 NLBLOCK->nltype == NLTYPE_FIXED &&
4430 NLBLOCK->nllen == 2 &&
4431 RAWUCHAR(eptr) == NLBLOCK->nl[0])
4432 {
4433 md->hitend = TRUE;
4434 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4435 }
4436 eptr++;
4437 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4438 }
4439 break;
4440
4441 case OP_ALLANY:
4442 for (i = 1; i <= min; i++)
4443 {
4444 if (eptr >= md->end_subject)
4445 {
4446 SCHECK_PARTIAL();
4447 RRETURN(MATCH_NOMATCH);
4448 }
4449 eptr++;
4450 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4451 }
4452 break;
4453
4454 case OP_ANYBYTE:
4455 if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH);
4456 eptr += min;
4457 break;
4458
4459 case OP_ANYNL:
4460 for (i = 1; i <= min; i++)
4461 {
4462 if (eptr >= md->end_subject)
4463 {
4464 SCHECK_PARTIAL();
4465 RRETURN(MATCH_NOMATCH);
4466 }
4467 GETCHARINC(c, eptr);
4468 switch(c)
4469 {
4470 default: RRETURN(MATCH_NOMATCH);
4471
4472 case CHAR_CR:
4473 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
4474 break;
4475
4476 case CHAR_LF:
4477 break;
4478
4479 case CHAR_VT:
4480 case CHAR_FF:
4481 case CHAR_NEL:
4482 #ifndef EBCDIC
4483 case 0x2028:
4484 case 0x2029:
4485 #endif /* Not EBCDIC */
4486 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4487 break;
4488 }
4489 }
4490 break;
4491
4492 case OP_NOT_HSPACE:
4493 for (i = 1; i <= min; i++)
4494 {
4495 if (eptr >= md->end_subject)
4496 {
4497 SCHECK_PARTIAL();
4498 RRETURN(MATCH_NOMATCH);
4499 }
4500 GETCHARINC(c, eptr);
4501 switch(c)
4502 {
4503 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
4504 default: break;
4505 }
4506 }
4507 break;
4508
4509 case OP_HSPACE:
4510 for (i = 1; i <= min; i++)
4511 {
4512 if (eptr >= md->end_subject)
4513 {
4514 SCHECK_PARTIAL();
4515 RRETURN(MATCH_NOMATCH);
4516 }
4517 GETCHARINC(c, eptr);
4518 switch(c)
4519 {
4520 HSPACE_CASES: break; /* Byte and multibyte cases */
4521 default: RRETURN(MATCH_NOMATCH);
4522 }
4523 }
4524 break;
4525
4526 case OP_NOT_VSPACE:
4527 for (i = 1; i <= min; i++)
4528 {
4529 if (eptr >= md->end_subject)
4530 {
4531 SCHECK_PARTIAL();
4532 RRETURN(MATCH_NOMATCH);
4533 }
4534 GETCHARINC(c, eptr);
4535 switch(c)
4536 {
4537 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
4538 default: break;
4539 }
4540 }
4541 break;
4542
4543 case OP_VSPACE:
4544 for (i = 1; i <= min; i++)
4545 {
4546 if (eptr >= md->end_subject)
4547 {
4548 SCHECK_PARTIAL();
4549 RRETURN(MATCH_NOMATCH);
4550 }
4551 GETCHARINC(c, eptr);
4552 switch(c)
4553 {
4554 VSPACE_CASES: break;
4555 default: RRETURN(MATCH_NOMATCH);
4556 }
4557 }
4558 break;
4559
4560 case OP_NOT_DIGIT:
4561 for (i = 1; i <= min; i++)
4562 {
4563 if (eptr >= md->end_subject)
4564 {
4565 SCHECK_PARTIAL();
4566 RRETURN(MATCH_NOMATCH);
4567 }
4568 GETCHARINC(c, eptr);
4569 if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
4570 RRETURN(MATCH_NOMATCH);
4571 }
4572 break;
4573
4574 case OP_DIGIT:
4575 for (i = 1; i <= min; i++)
4576 {
4577 pcre_uint32 cc;
4578 if (eptr >= md->end_subject)
4579 {
4580 SCHECK_PARTIAL();
4581 RRETURN(MATCH_NOMATCH);
4582 }
4583 cc = RAWUCHAR(eptr);
4584 if (cc >= 128 || (md->ctypes[cc] & ctype_digit) == 0)
4585 RRETURN(MATCH_NOMATCH);
4586 eptr++;
4587 /* No need to skip more bytes - we know it's a 1-byte character */
4588 }
4589 break;
4590
4591 case OP_NOT_WHITESPACE:
4592 for (i = 1; i <= min; i++)
4593 {
4594 pcre_uint32 cc;
4595 if (eptr >= md->end_subject)
4596 {
4597 SCHECK_PARTIAL();
4598 RRETURN(MATCH_NOMATCH);
4599 }
4600 cc = RAWUCHAR(eptr);
4601 if (cc < 128 && (md->ctypes[cc] & ctype_space) != 0)
4602 RRETURN(MATCH_NOMATCH);
4603 eptr++;
4604 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4605 }
4606 break;
4607
4608 case OP_WHITESPACE:
4609 for (i = 1; i <= min; i++)
4610 {
4611 pcre_uint32 cc;
4612 if (eptr >= md->end_subject)
4613 {
4614 SCHECK_PARTIAL();
4615 RRETURN(MATCH_NOMATCH);
4616 }
4617 cc = RAWUCHAR(eptr);
4618 if (cc >= 128 || (md->ctypes[cc] & ctype_space) == 0)
4619 RRETURN(MATCH_NOMATCH);
4620 eptr++;
4621 /* No need to skip more bytes - we know it's a 1-byte character */
4622 }
4623 break;
4624
4625 case OP_NOT_WORDCHAR:
4626 for (i = 1; i <= min; i++)
4627 {
4628 pcre_uint32 cc;
4629 if (eptr >= md->end_subject)
4630 {
4631 SCHECK_PARTIAL();
4632 RRETURN(MATCH_NOMATCH);
4633 }
4634 cc = RAWUCHAR(eptr);
4635 if (cc < 128 && (md->ctypes[cc] & ctype_word) != 0)
4636 RRETURN(MATCH_NOMATCH);
4637 eptr++;
4638 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4639 }
4640 break;
4641
4642 case OP_WORDCHAR:
4643 for (i = 1; i <= min; i++)
4644 {
4645 pcre_uint32 cc;
4646 if (eptr >= md->end_subject)
4647 {
4648 SCHECK_PARTIAL();
4649 RRETURN(MATCH_NOMATCH);
4650 }
4651 cc = RAWUCHAR(eptr);
4652 if (cc >= 128 || (md->ctypes[cc] & ctype_word) == 0)
4653 RRETURN(MATCH_NOMATCH);
4654 eptr++;
4655 /* No need to skip more bytes - we know it's a 1-byte character */
4656 }
4657 break;
4658
4659 default:
4660 RRETURN(PCRE_ERROR_INTERNAL);
4661 } /* End switch(ctype) */
4662
4663 else
4664 #endif /* SUPPORT_UTF */
4665
4666 /* Code for the non-UTF-8 case for minimum matching of operators other
4667 than OP_PROP and OP_NOTPROP. */
4668
4669 switch(ctype)
4670 {
4671 case OP_ANY:
4672 for (i = 1; i <= min; i++)
4673 {
4674 if (eptr >= md->end_subject)
4675 {
4676 SCHECK_PARTIAL();
4677 RRETURN(MATCH_NOMATCH);
4678 }
4679 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4680 if (md->partial != 0 &&
4681 eptr + 1 >= md->end_subject &&
4682 NLBLOCK->nltype == NLTYPE_FIXED &&
4683 NLBLOCK->nllen == 2 &&
4684 *eptr == NLBLOCK->nl[0])
4685 {
4686 md->hitend = TRUE;
4687 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4688 }
4689 eptr++;
4690 }
4691 break;
4692
4693 case OP_ALLANY:
4694 if (eptr > md->end_subject - min)
4695 {
4696 SCHECK_PARTIAL();
4697 RRETURN(MATCH_NOMATCH);
4698 }
4699 eptr += min;
4700 break;
4701
4702 case OP_ANYBYTE:
4703 if (eptr > md->end_subject - min)
4704 {
4705 SCHECK_PARTIAL();
4706 RRETURN(MATCH_NOMATCH);
4707 }
4708 eptr += min;
4709 break;
4710
4711 case OP_ANYNL:
4712 for (i = 1; i <= min; i++)
4713 {
4714 if (eptr >= md->end_subject)
4715 {
4716 SCHECK_PARTIAL();
4717 RRETURN(MATCH_NOMATCH);
4718 }
4719 switch(*eptr++)
4720 {
4721 default: RRETURN(MATCH_NOMATCH);
4722
4723 case CHAR_CR:
4724 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
4725 break;
4726
4727 case CHAR_LF:
4728 break;
4729
4730 case CHAR_VT:
4731 case CHAR_FF:
4732 case CHAR_NEL:
4733 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4734 case 0x2028:
4735 case 0x2029:
4736 #endif
4737 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4738 break;
4739 }
4740 }
4741 break;
4742
4743 case OP_NOT_HSPACE:
4744 for (i = 1; i <= min; i++)
4745 {
4746 if (eptr >= md->end_subject)
4747 {
4748 SCHECK_PARTIAL();
4749 RRETURN(MATCH_NOMATCH);
4750 }
4751 switch(*eptr++)
4752 {
4753 default: break;
4754 HSPACE_BYTE_CASES:
4755 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4756 HSPACE_MULTIBYTE_CASES:
4757 #endif
4758 RRETURN(MATCH_NOMATCH);
4759 }
4760 }
4761 break;
4762
4763 case OP_HSPACE:
4764 for (i = 1; i <= min; i++)
4765 {
4766 if (eptr >= md->end_subject)
4767 {
4768 SCHECK_PARTIAL();
4769 RRETURN(MATCH_NOMATCH);
4770 }
4771 switch(*eptr++)
4772 {
4773 default: RRETURN(MATCH_NOMATCH);
4774 HSPACE_BYTE_CASES:
4775 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4776 HSPACE_MULTIBYTE_CASES:
4777 #endif
4778 break;
4779 }
4780 }
4781 break;
4782
4783 case OP_NOT_VSPACE:
4784 for (i = 1; i <= min; i++)
4785 {
4786 if (eptr >= md->end_subject)
4787 {
4788 SCHECK_PARTIAL();
4789 RRETURN(MATCH_NOMATCH);
4790 }
4791 switch(*eptr++)
4792 {
4793 VSPACE_BYTE_CASES:
4794 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4795 VSPACE_MULTIBYTE_CASES:
4796 #endif
4797 RRETURN(MATCH_NOMATCH);
4798 default: break;
4799 }
4800 }
4801 break;
4802
4803 case OP_VSPACE:
4804 for (i = 1; i <= min; i++)
4805 {
4806 if (eptr >= md->end_subject)
4807 {
4808 SCHECK_PARTIAL();
4809 RRETURN(MATCH_NOMATCH);
4810 }
4811 switch(*eptr++)
4812 {
4813 default: RRETURN(MATCH_NOMATCH);
4814 VSPACE_BYTE_CASES:
4815 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4816 VSPACE_MULTIBYTE_CASES:
4817 #endif
4818 break;
4819 }
4820 }
4821 break;
4822
4823 case OP_NOT_DIGIT:
4824 for (i = 1; i <= min; i++)
4825 {
4826 if (eptr >= md->end_subject)
4827 {
4828 SCHECK_PARTIAL();
4829 RRETURN(MATCH_NOMATCH);
4830 }
4831 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0)
4832 RRETURN(MATCH_NOMATCH);
4833 eptr++;
4834 }
4835 break;
4836
4837 case OP_DIGIT:
4838 for (i = 1; i <= min; i++)
4839 {
4840 if (eptr >= md->end_subject)
4841 {
4842 SCHECK_PARTIAL();
4843 RRETURN(MATCH_NOMATCH);
4844 }
4845 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0)
4846 RRETURN(MATCH_NOMATCH);
4847 eptr++;
4848 }
4849 break;
4850
4851 case OP_NOT_WHITESPACE:
4852 for (i = 1; i <= min; i++)
4853 {
4854 if (eptr >= md->end_subject)
4855 {
4856 SCHECK_PARTIAL();
4857 RRETURN(MATCH_NOMATCH);
4858 }
4859 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0)
4860 RRETURN(MATCH_NOMATCH);
4861 eptr++;
4862 }
4863 break;
4864
4865 case OP_WHITESPACE:
4866 for (i = 1; i <= min; i++)
4867 {
4868 if (eptr >= md->end_subject)
4869 {
4870 SCHECK_PARTIAL();
4871 RRETURN(MATCH_NOMATCH);
4872 }
4873 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0)
4874 RRETURN(MATCH_NOMATCH);
4875 eptr++;
4876 }
4877 break;
4878
4879 case OP_NOT_WORDCHAR:
4880 for (i = 1; i <= min; i++)
4881 {
4882 if (eptr >= md->end_subject)
4883 {
4884 SCHECK_PARTIAL();
4885 RRETURN(MATCH_NOMATCH);
4886 }
4887 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0)
4888 RRETURN(MATCH_NOMATCH);
4889 eptr++;
4890 }
4891 break;
4892
4893 case OP_WORDCHAR:
4894 for (i = 1; i <= min; i++)
4895 {
4896 if (eptr >= md->end_subject)
4897 {
4898 SCHECK_PARTIAL();
4899 RRETURN(MATCH_NOMATCH);
4900 }
4901 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0)
4902 RRETURN(MATCH_NOMATCH);
4903 eptr++;
4904 }
4905 break;
4906
4907 default:
4908 RRETURN(PCRE_ERROR_INTERNAL);
4909 }
4910 }
4911
4912 /* If min = max, continue at the same level without recursing */
4913
4914 if (min == max) continue;
4915
4916 /* If minimizing, we have to test the rest of the pattern before each
4917 subsequent match. Again, separate the UTF-8 case for speed, and also
4918 separate the UCP cases. */
4919
4920 if (minimize)
4921 {
4922 #ifdef SUPPORT_UCP
4923 if (prop_type >= 0)
4924 {
4925 switch(prop_type)
4926 {
4927 case PT_ANY:
4928 for (fi = min;; fi++)
4929 {
4930 RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4931 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4932 if (fi >= max) RRETURN(MATCH_NOMATCH);
4933 if (eptr >= md->end_subject)
4934 {
4935 SCHECK_PARTIAL();
4936 RRETURN(MATCH_NOMATCH);
4937 }
4938 GETCHARINCTEST(c, eptr);
4939 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4940 }
4941 /* Control never gets here */
4942
4943 case PT_LAMP:
4944 for (fi = min;; fi++)
4945 {
4946 int chartype;
4947 RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4948 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4949 if (fi >= max) RRETURN(MATCH_NOMATCH);
4950 if (eptr >= md->end_subject)
4951 {
4952 SCHECK_PARTIAL();
4953 RRETURN(MATCH_NOMATCH);
4954 }
4955 GETCHARINCTEST(c, eptr);
4956 chartype = UCD_CHARTYPE(c);
4957 if ((chartype == ucp_Lu ||
4958 chartype == ucp_Ll ||
4959 chartype == ucp_Lt) == prop_fail_result)
4960 RRETURN(MATCH_NOMATCH);
4961 }
4962 /* Control never gets here */
4963
4964 case PT_GC:
4965 for (fi = min;; fi++)
4966 {
4967 RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4968 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4969 if (fi >= max) RRETURN(MATCH_NOMATCH);
4970 if (eptr >= md->end_subject)
4971 {
4972 SCHECK_PARTIAL();
4973 RRETURN(MATCH_NOMATCH);
4974 }
4975 GETCHARINCTEST(c, eptr);
4976 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4977 RRETURN(MATCH_NOMATCH);
4978 }
4979 /* Control never gets here */
4980
4981 case PT_PC:
4982 for (fi = min;; fi++)
4983 {
4984 RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4985 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4986 if (fi >= max) RRETURN(MATCH_NOMATCH);
4987 if (eptr >= md->end_subject)
4988 {
4989 SCHECK_PARTIAL();
4990 RRETURN(MATCH_NOMATCH);
4991 }
4992 GETCHARINCTEST(c, eptr);
4993 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4994 RRETURN(MATCH_NOMATCH);
4995 }
4996 /* Control never gets here */
4997
4998 case PT_SC:
4999 for (fi = min;; fi++)
5000 {
5001 RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
5002 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5003 if (fi >= max) RRETURN(MATCH_NOMATCH);
5004 if (eptr >= md->end_subject)
5005 {
5006 SCHECK_PARTIAL();
5007 RRETURN(MATCH_NOMATCH);
5008 }
5009 GETCHARINCTEST(c, eptr);
5010 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
5011 RRETURN(MATCH_NOMATCH);
5012 }
5013 /* Control never gets here */
5014
5015 case PT_ALNUM:
5016 for (fi = min;; fi++)
5017 {
5018 int category;
5019 RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
5020 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5021 if (fi >= max) RRETURN(MATCH_NOMATCH);
5022 if (eptr >= md->end_subject)
5023 {
5024 SCHECK_PARTIAL();
5025 RRETURN(MATCH_NOMATCH);
5026 }
5027 GETCHARINCTEST(c, eptr);
5028 category = UCD_CATEGORY(c);
5029 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5030 RRETURN(MATCH_NOMATCH);
5031 }
5032 /* Control never gets here */
5033
5034 case PT_SPACE: /* Perl space */
5035 for (fi = min;; fi++)
5036 {
5037 RMATCH(eptr, ecode, offset_top, md, eptrb, RM60);
5038 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5039 if (fi >= max) RRETURN(MATCH_NOMATCH);
5040 if (eptr >= md->end_subject)
5041 {
5042 SCHECK_PARTIAL();
5043 RRETURN(MATCH_NOMATCH);
5044 }
5045 GETCHARINCTEST(c, eptr);
5046 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5047 c == CHAR_FF || c == CHAR_CR)
5048 == prop_fail_result)
5049 RRETURN(MATCH_NOMATCH);
5050 }
5051 /* Control never gets here */
5052
5053 case PT_PXSPACE: /* POSIX space */
5054 for (fi = min;; fi++)
5055 {
5056 RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
5057 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5058 if (fi >= max) RRETURN(MATCH_NOMATCH);
5059 if (eptr >= md->end_subject)
5060 {
5061 SCHECK_PARTIAL();
5062 RRETURN(MATCH_NOMATCH);
5063 }
5064 GETCHARINCTEST(c, eptr);
5065 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5066 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5067 == prop_fail_result)
5068 RRETURN(MATCH_NOMATCH);
5069 }
5070 /* Control never gets here */
5071
5072 case PT_WORD:
5073 for (fi = min;; fi++)
5074 {
5075 int category;
5076 RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
5077 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5078 if (fi >= max) RRETURN(MATCH_NOMATCH);
5079 if (eptr >= md->end_subject)
5080 {
5081 SCHECK_PARTIAL();
5082 RRETURN(MATCH_NOMATCH);
5083 }
5084 GETCHARINCTEST(c, eptr);
5085 category = UCD_CATEGORY(c);
5086 if ((category == ucp_L ||
5087 category == ucp_N ||
5088 c == CHAR_UNDERSCORE)
5089 == prop_fail_result)
5090 RRETURN(MATCH_NOMATCH);
5091 }
5092 /* Control never gets here */
5093
5094 case PT_CLIST:
5095 for (fi = min;; fi++)
5096 {
5097 const pcre_uint32 *cp;
5098 RMATCH(eptr, ecode, offset_top, md, eptrb, RM67);
5099 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5100 if (fi >= max) RRETURN(MATCH_NOMATCH);
5101 if (eptr >= md->end_subject)
5102 {
5103 SCHECK_PARTIAL();
5104 RRETURN(MATCH_NOMATCH);
5105 }
5106 GETCHARINCTEST(c, eptr);
5107 cp = PRIV(ucd_caseless_sets) + prop_value;
5108 for (;;)
5109 {
5110 if (c < *cp)
5111 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
5112 if (c == *cp++)
5113 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
5114 }
5115 }
5116 /* Control never gets here */
5117
5118 case PT_UCNC:
5119 for (fi = min;; fi++)
5120 {
5121 RMATCH(eptr, ecode, offset_top, md, eptrb, RM68);
5122 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5123 if (fi >= max) RRETURN(MATCH_NOMATCH);
5124 if (eptr >= md->end_subject)
5125 {
5126 SCHECK_PARTIAL();
5127 RRETURN(MATCH_NOMATCH);
5128 }
5129 GETCHARINCTEST(c, eptr);
5130 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5131 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5132 c >= 0xe000) == prop_fail_result)
5133 RRETURN(MATCH_NOMATCH);
5134 }
5135 /* Control never gets here */
5136
5137 /* This should never occur */
5138 default:
5139 RRETURN(PCRE_ERROR_INTERNAL);
5140 }
5141 }
5142
5143 /* Match extended Unicode sequences. We will get here only if the
5144 support is in the binary; otherwise a compile-time error occurs. */
5145
5146 else if (ctype == OP_EXTUNI)
5147 {
5148 for (fi = min;; fi++)
5149 {
5150 RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
5151 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5152 if (fi >= max) RRETURN(MATCH_NOMATCH);
5153 if (eptr >= md->end_subject)
5154 {
5155 SCHECK_PARTIAL();
5156 RRETURN(MATCH_NOMATCH);
5157 }
5158 else
5159 {
5160 int lgb, rgb;
5161 GETCHARINCTEST(c, eptr);
5162 lgb = UCD_GRAPHBREAK(c);
5163 while (eptr < md->end_subject)
5164 {
5165 int len = 1;
5166 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5167 rgb = UCD_GRAPHBREAK(c);
5168 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5169 lgb = rgb;
5170 eptr += len;
5171 }
5172 }
5173 CHECK_PARTIAL();
5174 }
5175 }
5176 else
5177 #endif /* SUPPORT_UCP */
5178
5179 #ifdef SUPPORT_UTF
5180 if (utf)
5181 {
5182 for (fi = min;; fi++)
5183 {
5184 RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
5185 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5186 if (fi >= max) RRETURN(MATCH_NOMATCH);
5187 if (eptr >= md->end_subject)
5188 {
5189 SCHECK_PARTIAL();
5190 RRETURN(MATCH_NOMATCH);
5191 }
5192 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5193 RRETURN(MATCH_NOMATCH);
5194 GETCHARINC(c, eptr);
5195 switch(ctype)
5196 {
5197 case OP_ANY: /* This is the non-NL case */
5198 if (md->partial != 0 && /* Take care with CRLF partial */
5199 eptr >= md->end_subject &&
5200 NLBLOCK->nltype == NLTYPE_FIXED &&
5201 NLBLOCK->nllen == 2 &&
5202 c == NLBLOCK->nl[0])
5203 {
5204 md->hitend = TRUE;
5205 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5206 }
5207 break;
5208
5209 case OP_ALLANY:
5210 case OP_ANYBYTE:
5211 break;
5212
5213 case OP_ANYNL:
5214 switch(c)
5215 {
5216 default: RRETURN(MATCH_NOMATCH);
5217 case CHAR_CR:
5218 if (eptr < md->end_subject && RAWUCHAR(eptr) == CHAR_LF) eptr++;
5219 break;
5220
5221 case CHAR_LF:
5222 break;
5223
5224 case CHAR_VT:
5225 case CHAR_FF:
5226 case CHAR_NEL:
5227 #ifndef EBCDIC
5228 case 0x2028:
5229 case 0x2029:
5230 #endif /* Not EBCDIC */
5231 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5232 break;
5233 }
5234 break;
5235
5236 case OP_NOT_HSPACE:
5237 switch(c)
5238 {
5239 HSPACE_CASES: RRETURN(MATCH_NOMATCH);
5240 default: break;
5241 }
5242 break;
5243
5244 case OP_HSPACE:
5245 switch(c)
5246 {
5247 HSPACE_CASES: break;
5248 default: RRETURN(MATCH_NOMATCH);
5249 }
5250 break;
5251
5252 case OP_NOT_VSPACE:
5253 switch(c)
5254 {
5255 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
5256 default: break;
5257 }
5258 break;
5259
5260 case OP_VSPACE:
5261 switch(c)
5262 {
5263 VSPACE_CASES: break;
5264 default: RRETURN(MATCH_NOMATCH);
5265 }
5266 break;
5267
5268 case OP_NOT_DIGIT:
5269 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
5270 RRETURN(MATCH_NOMATCH);
5271 break;
5272
5273 case OP_DIGIT:
5274 if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0)
5275 RRETURN(MATCH_NOMATCH);
5276 break;
5277
5278 case OP_NOT_WHITESPACE:
5279 if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
5280 RRETURN(MATCH_NOMATCH);
5281 break;
5282
5283 case OP_WHITESPACE:
5284 if (c >= 256 || (md->ctypes[c] & ctype_space) == 0)
5285 RRETURN(MATCH_NOMATCH);
5286 break;
5287
5288 case OP_NOT_WORDCHAR:
5289 if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
5290 RRETURN(MATCH_NOMATCH);
5291 break;
5292
5293 case OP_WORDCHAR:
5294 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0)
5295 RRETURN(MATCH_NOMATCH);
5296 break;
5297
5298 default:
5299 RRETURN(PCRE_ERROR_INTERNAL);
5300 }
5301 }
5302 }
5303 else
5304 #endif
5305 /* Not UTF mode */
5306 {
5307 for (fi = min;; fi++)
5308 {
5309 RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
5310 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5311 if (fi >= max) RRETURN(MATCH_NOMATCH);
5312 if (eptr >= md->end_subject)
5313 {
5314 SCHECK_PARTIAL();
5315 RRETURN(MATCH_NOMATCH);
5316 }
5317 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5318 RRETURN(MATCH_NOMATCH);
5319 c = *eptr++;
5320 switch(ctype)
5321 {
5322 case OP_ANY: /* This is the non-NL case */
5323 if (md->partial != 0 && /* Take care with CRLF partial */
5324 eptr >= md->end_subject &&
5325 NLBLOCK->nltype == NLTYPE_FIXED &&
5326 NLBLOCK->nllen == 2 &&
5327 c == NLBLOCK->nl[0])
5328 {
5329 md->hitend = TRUE;
5330 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5331 }
5332 break;
5333
5334 case OP_ALLANY:
5335 case OP_ANYBYTE:
5336 break;
5337
5338 case OP_ANYNL:
5339 switch(c)
5340 {
5341 default: RRETURN(MATCH_NOMATCH);
5342 case CHAR_CR:
5343 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
5344 break;
5345
5346 case CHAR_LF:
5347 break;
5348
5349 case CHAR_VT:
5350 case CHAR_FF:
5351 case CHAR_NEL:
5352 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5353 case 0x2028:
5354 case 0x2029:
5355 #endif
5356 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5357 break;
5358 }
5359 break;
5360
5361 case OP_NOT_HSPACE:
5362 switch(c)
5363 {
5364 default: break;
5365 HSPACE_BYTE_CASES:
5366 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5367 HSPACE_MULTIBYTE_CASES:
5368 #endif
5369 RRETURN(MATCH_NOMATCH);
5370 }
5371 break;
5372
5373 case OP_HSPACE:
5374 switch(c)
5375 {
5376 default: RRETURN(MATCH_NOMATCH);
5377 HSPACE_BYTE_CASES:
5378 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5379 HSPACE_MULTIBYTE_CASES:
5380 #endif
5381 break;
5382 }
5383 break;
5384
5385 case OP_NOT_VSPACE:
5386 switch(c)
5387 {
5388 default: break;
5389 VSPACE_BYTE_CASES:
5390 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5391 VSPACE_MULTIBYTE_CASES:
5392 #endif
5393 RRETURN(MATCH_NOMATCH);
5394 }
5395 break;
5396
5397 case OP_VSPACE:
5398 switch(c)
5399 {
5400 default: RRETURN(MATCH_NOMATCH);
5401 VSPACE_BYTE_CASES:
5402 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5403 VSPACE_MULTIBYTE_CASES:
5404 #endif
5405 break;
5406 }
5407 break;
5408
5409 case OP_NOT_DIGIT:
5410 if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
5411 break;
5412
5413 case OP_DIGIT:
5414 if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
5415 break;
5416
5417 case OP_NOT_WHITESPACE:
5418 if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
5419 break;
5420
5421 case OP_WHITESPACE:
5422 if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
5423 break;
5424
5425 case OP_NOT_WORDCHAR:
5426 if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH);
5427 break;
5428
5429 case OP_WORDCHAR:
5430 if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH);
5431 break;
5432
5433 default:
5434 RRETURN(PCRE_ERROR_INTERNAL);
5435 }
5436 }
5437 }
5438 /* Control never gets here */
5439 }
5440
5441 /* If maximizing, it is worth using inline code for speed, doing the type
5442 test once at the start (i.e. keep it out of the loop). Again, keep the
5443 UTF-8 and UCP stuff separate. */
5444
5445 else
5446 {
5447 pp = eptr; /* Remember where we started */
5448
5449 #ifdef SUPPORT_UCP
5450 if (prop_type >= 0)
5451 {
5452 switch(prop_type)
5453 {
5454 case PT_ANY:
5455 for (i = min; i < max; i++)
5456 {
5457 int len = 1;
5458 if (eptr >= md->end_subject)
5459 {
5460 SCHECK_PARTIAL();
5461 break;
5462 }
5463 GETCHARLENTEST(c, eptr, len);
5464 if (prop_fail_result) break;
5465 eptr+= len;
5466 }
5467 break;
5468
5469 case PT_LAMP:
5470 for (i = min; i < max; i++)
5471 {
5472 int chartype;
5473 int len = 1;
5474 if (eptr >= md->end_subject)
5475 {
5476 SCHECK_PARTIAL();
5477 break;
5478 }
5479 GETCHARLENTEST(c, eptr, len);
5480 chartype = UCD_CHARTYPE(c);
5481 if ((chartype == ucp_Lu ||
5482 chartype == ucp_Ll ||
5483 chartype == ucp_Lt) == prop_fail_result)
5484 break;
5485 eptr+= len;
5486 }
5487 break;
5488
5489 case PT_GC:
5490 for (i = min; i < max; i++)
5491 {
5492 int len = 1;
5493 if (eptr >= md->end_subject)
5494 {
5495 SCHECK_PARTIAL();
5496 break;
5497 }
5498 GETCHARLENTEST(c, eptr, len);
5499 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
5500 eptr+= len;
5501 }
5502 break;
5503
5504 case PT_PC:
5505 for (i = min; i < max; i++)
5506 {
5507 int len = 1;
5508 if (eptr >= md->end_subject)
5509 {
5510 SCHECK_PARTIAL();
5511 break;
5512 }
5513 GETCHARLENTEST(c, eptr, len);
5514 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
5515 eptr+= len;
5516 }
5517 break;
5518
5519 case PT_SC:
5520 for (i = min; i < max; i++)
5521 {
5522 int len = 1;
5523 if (eptr >= md->end_subject)
5524 {
5525 SCHECK_PARTIAL();
5526 break;
5527 }
5528 GETCHARLENTEST(c, eptr, len);
5529 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
5530 eptr+= len;
5531 }
5532 break;
5533
5534 case PT_ALNUM:
5535 for (i = min; i < max; i++)
5536 {
5537 int category;
5538 int len = 1;
5539 if (eptr >= md->end_subject)
5540 {
5541 SCHECK_PARTIAL();
5542 break;
5543 }
5544 GETCHARLENTEST(c, eptr, len);
5545 category = UCD_CATEGORY(c);
5546 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5547 break;
5548 eptr+= len;
5549 }
5550 break;
5551
5552 case PT_SPACE: /* Perl space */
5553 for (i = min; i < max; i++)
5554 {
5555 int len = 1;
5556 if (eptr >= md->end_subject)
5557 {
5558 SCHECK_PARTIAL();
5559 break;
5560 }
5561 GETCHARLENTEST(c, eptr, len);
5562 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5563 c == CHAR_FF || c == CHAR_CR)
5564 == prop_fail_result)
5565 break;
5566 eptr+= len;
5567 }
5568 break;
5569
5570 case PT_PXSPACE: /* POSIX space */
5571 for (i = min; i < max; i++)
5572 {
5573 int len = 1;
5574 if (eptr >= md->end_subject)
5575 {
5576 SCHECK_PARTIAL();
5577 break;
5578 }
5579 GETCHARLENTEST(c, eptr, len);
5580 if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
5581 c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5582 == prop_fail_result)
5583 break;
5584 eptr+= len;
5585 }
5586 break;
5587
5588 case PT_WORD:
5589 for (i = min; i < max; i++)
5590 {
5591 int category;
5592 int len = 1;
5593 if (eptr >= md->end_subject)
5594 {
5595 SCHECK_PARTIAL();
5596 break;
5597 }
5598 GETCHARLENTEST(c, eptr, len);
5599 category = UCD_CATEGORY(c);
5600 if ((category == ucp_L || category == ucp_N ||
5601 c == CHAR_UNDERSCORE) == prop_fail_result)
5602 break;
5603 eptr+= len;
5604 }
5605 break;
5606
5607 case PT_CLIST:
5608 for (i = min; i < max; i++)
5609 {
5610 const pcre_uint32 *cp;
5611 int len = 1;
5612 if (eptr >= md->end_subject)
5613 {
5614 SCHECK_PARTIAL();
5615 break;
5616 }
5617 GETCHARLENTEST(c, eptr, len);
5618 cp = PRIV(ucd_caseless_sets) + prop_value;
5619 for (;;)
5620 {
5621 if (c < *cp)
5622 { if (prop_fail_result) break; else goto GOT_MAX; }
5623 if (c == *cp++)
5624 { if (prop_fail_result) goto GOT_MAX; else break; }
5625 }
5626 eptr += len;
5627 }
5628 GOT_MAX:
5629 break;
5630
5631 case PT_UCNC:
5632 for (i = min; i < max; i++)
5633 {
5634 int len = 1;
5635 if (eptr >= md->end_subject)
5636 {
5637 SCHECK_PARTIAL();
5638 break;
5639 }
5640 GETCHARLENTEST(c, eptr, len);
5641 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5642 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5643 c >= 0xe000) == prop_fail_result)
5644 break;
5645 eptr += len;
5646 }
5647 break;
5648
5649 default:
5650 RRETURN(PCRE_ERROR_INTERNAL);
5651 }
5652
5653 /* eptr is now past the end of the maximum run */
5654
5655 if (possessive) continue; /* No backtracking */
5656 for(;;)
5657 {
5658 if (eptr == pp) goto TAIL_RECURSE;
5659 RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5660 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5661 eptr--;
5662 if (utf) BACKCHAR(eptr);
5663 }
5664 }
5665
5666 /* Match extended Unicode grapheme clusters. We will get here only if the
5667 support is in the binary; otherwise a compile-time error occurs. */
5668
5669 else if (ctype == OP_EXTUNI)
5670 {
5671 for (i = min; i < max; i++)
5672 {
5673 if (eptr >= md->end_subject)
5674 {
5675 SCHECK_PARTIAL();
5676 break;
5677 }
5678 else
5679 {
5680 int lgb, rgb;
5681 GETCHARINCTEST(c, eptr);
5682 lgb = UCD_GRAPHBREAK(c);
5683 while (eptr < md->end_subject)
5684 {
5685 int len = 1;
5686 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5687 rgb = UCD_GRAPHBREAK(c);
5688 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5689 lgb = rgb;
5690 eptr += len;
5691 }
5692 }
5693 CHECK_PARTIAL();
5694 }
5695
5696 /* eptr is now past the end of the maximum run */
5697
5698 if (possessive) continue; /* No backtracking */
5699
5700 for(;;)
5701 {
5702 int lgb, rgb;
5703 PCRE_PUCHAR fptr;
5704
5705 if (eptr == pp) goto TAIL_RECURSE; /* At start of char run */
5706 RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5707 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5708
5709 /* Backtracking over an extended grapheme cluster involves inspecting
5710 the previous two characters (if present) to see if a break is
5711 permitted between them. */
5712
5713 eptr--;
5714 if (!utf) c = *eptr; else
5715 {
5716 BACKCHAR(eptr);
5717 GETCHAR(c, eptr);
5718 }
5719 rgb = UCD_GRAPHBREAK(c);
5720
5721 for (;;)
5722 {
5723 if (eptr == pp) goto TAIL_RECURSE; /* At start of char run */
5724 fptr = eptr - 1;
5725 if (!utf) c = *fptr; else
5726 {
5727 BACKCHAR(fptr);
5728 GETCHAR(c, fptr);
5729 }
5730 lgb = UCD_GRAPHBREAK(c);
5731 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5732 eptr = fptr;
5733 rgb = lgb;
5734 }
5735 }
5736 }
5737
5738 else
5739 #endif /* SUPPORT_UCP */
5740
5741 #ifdef SUPPORT_UTF
5742 if (utf)
5743 {
5744 switch(ctype)
5745 {
5746 case OP_ANY:
5747 if (max < INT_MAX)
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
5771 /* Handle unlimited UTF-8 repeat */
5772
5773 else
5774 {
5775 for (i = min; i < max; i++)
5776 {
5777 if (eptr >= md->end_subject)
5778 {
5779 SCHECK_PARTIAL();
5780 break;
5781 }
5782 if (IS_NEWLINE(eptr)) break;
5783 if (md->partial != 0 && /* Take care with CRLF partial */
5784 eptr + 1 >= md->end_subject &&
5785 NLBLOCK->nltype == NLTYPE_FIXED &&
5786 NLBLOCK->nllen == 2 &&
5787 RAWUCHAR(eptr) == NLBLOCK->nl[0])
5788 {
5789 md->hitend = TRUE;
5790 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5791 }
5792 eptr++;
5793 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5794 }
5795 }
5796 break;
5797
5798 case OP_ALLANY:
5799 if (max < INT_MAX)
5800 {
5801 for (i = min; i < max; i++)
5802 {
5803 if (eptr >= md->end_subject)
5804 {
5805 SCHECK_PARTIAL();
5806 break;
5807 }
5808 eptr++;
5809 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5810 }
5811 }
5812 else
5813 {
5814 eptr = md->end_subject; /* Unlimited UTF-8 repeat */
5815 SCHECK_PARTIAL();
5816 }
5817 break;
5818
5819 /* The byte case is the same as non-UTF8 */
5820
5821 case OP_ANYBYTE:
5822 c = max - min;
5823 if (c > (unsigned int)(md->end_subject - eptr))
5824 {
5825 eptr = md->end_subject;
5826 SCHECK_PARTIAL();
5827 }
5828 else eptr += c;
5829 break;
5830
5831 case OP_ANYNL:
5832 for (i = min; i < max; i++)
5833 {
5834 int len = 1;
5835 if (eptr >= md->end_subject)
5836 {
5837 SCHECK_PARTIAL();
5838 break;
5839 }
5840 GETCHARLEN(c, eptr, len);
5841 if (c == CHAR_CR)
5842 {
5843 if (++eptr >= md->end_subject) break;
5844 if (RAWUCHAR(eptr) == CHAR_LF) eptr++;
5845 }
5846 else
5847 {
5848 if (c != CHAR_LF &&
5849 (md->bsr_anycrlf ||
5850 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
5851 #ifndef EBCDIC
5852 && c != 0x2028 && c != 0x2029
5853 #endif /* Not EBCDIC */
5854 )))
5855 break;
5856 eptr += len;
5857 }
5858 }
5859 break;
5860
5861 case OP_NOT_HSPACE:
5862 case OP_HSPACE:
5863 for (i = min; i < max; i++)
5864 {
5865 BOOL gotspace;
5866 int len = 1;
5867 if (eptr >= md->end_subject)
5868 {
5869 SCHECK_PARTIAL();
5870 break;
5871 }
5872 GETCHARLEN(c, eptr, len);
5873 switch(c)
5874 {
5875 HSPACE_CASES: gotspace = TRUE; break;
5876 default: gotspace = FALSE; break;
5877 }
5878 if (gotspace == (ctype == OP_NOT_HSPACE)) break;
5879 eptr += len;
5880 }
5881 break;
5882
5883 case OP_NOT_VSPACE:
5884 case OP_VSPACE:
5885 for (i = min; i < max; i++)
5886 {
5887 BOOL gotspace;
5888 int len = 1;
5889 if (eptr >= md->end_subject)
5890 {
5891 SCHECK_PARTIAL();
5892 break;
5893 }
5894 GETCHARLEN(c, eptr, len);
5895 switch(c)
5896 {
5897 VSPACE_CASES: gotspace = TRUE; break;
5898 default: gotspace = FALSE; break;
5899 }
5900 if (gotspace == (ctype == OP_NOT_VSPACE)) break;
5901 eptr += len;
5902 }
5903 break;
5904
5905 case OP_NOT_DIGIT:
5906 for (i = min; i < max; i++)
5907 {
5908 int len = 1;
5909 if (eptr >= md->end_subject)
5910 {
5911 SCHECK_PARTIAL();
5912 break;
5913 }
5914 GETCHARLEN(c, eptr, len);
5915 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
5916 eptr+= len;
5917 }
5918 break;
5919
5920 case OP_DIGIT:
5921 for (i = min; i < max; i++)
5922 {
5923 int len = 1;
5924 if (eptr >= md->end_subject)
5925 {
5926 SCHECK_PARTIAL();
5927 break;
5928 }
5929 GETCHARLEN(c, eptr, len);
5930 if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break;
5931 eptr+= len;
5932 }
5933 break;
5934
5935 case OP_NOT_WHITESPACE:
5936 for (i = min; i < max; i++)
5937 {
5938 int len = 1;
5939 if (eptr >= md->end_subject)
5940 {
5941 SCHECK_PARTIAL();
5942 break;
5943 }
5944 GETCHARLEN(c, eptr, len);
5945 if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
5946 eptr+= len;
5947 }
5948 break;
5949
5950 case OP_WHITESPACE:
5951 for (i = min; i < max; i++)
5952 {
5953 int len = 1;
5954 if (eptr >= md->end_subject)
5955 {
5956 SCHECK_PARTIAL();
5957 break;
5958 }
5959 GETCHARLEN(c, eptr, len);
5960 if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break;
5961 eptr+= len;
5962 }
5963 break;
5964
5965 case OP_NOT_WORDCHAR:
5966 for (i = min; i < max; i++)
5967 {
5968 int len = 1;
5969 if (eptr >= md->end_subject)
5970 {
5971 SCHECK_PARTIAL();
5972 break;
5973 }
5974 GETCHARLEN(c, eptr, len);
5975 if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
5976 eptr+= len;
5977 }
5978 break;
5979
5980 case OP_WORDCHAR:
5981 for (i = min; i < max; i++)
5982 {
5983 int len = 1;
5984 if (eptr >= md->end_subject)
5985 {
5986 SCHECK_PARTIAL();
5987 break;
5988 }
5989 GETCHARLEN(c, eptr, len);
5990 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break;
5991 eptr+= len;
5992 }
5993 break;
5994
5995 default:
5996 RRETURN(PCRE_ERROR_INTERNAL);
5997 }
5998
5999 if (possessive) continue; /* No backtracking */
6000 for(;;)
6001 {
6002 if (eptr == pp) goto TAIL_RECURSE;
6003 RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
6004 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
6005 eptr--;
6006 BACKCHAR(eptr);
6007 if (ctype == OP_ANYNL && eptr > pp && RAWUCHAR(eptr) == CHAR_NL &&
6008 RAWUCHAR(eptr - 1) == CHAR_CR) eptr--;
6009 }
6010 }
6011 else
6012 #endif /* SUPPORT_UTF */
6013 /* Not UTF mode */
6014 {
6015 switch(ctype)
6016 {
6017 case OP_ANY:
6018 for (i = min; i < max; i++)
6019 {
6020 if (eptr >= md->end_subject)
6021 {
6022 SCHECK_PARTIAL();
6023 break;
6024 }
6025 if (IS_NEWLINE(eptr)) break;
6026 if (md->partial != 0 && /* Take care with CRLF partial */
6027 eptr + 1 >= md->end_subject &&
6028 NLBLOCK->nltype == NLTYPE_FIXED &&
6029 NLBLOCK->nllen == 2 &&
6030 *eptr == NLBLOCK->nl[0])
6031 {
6032 md->hitend = TRUE;
6033 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
6034 }
6035 eptr++;
6036 }
6037 break;
6038
6039 case OP_ALLANY:
6040 case OP_ANYBYTE:
6041 c = max - min;
6042 if (c > (unsigned int)(md->end_subject - eptr))
6043 {
6044 eptr = md->end_subject;
6045 SCHECK_PARTIAL();
6046 }
6047 else eptr += c;
6048 break;
6049
6050 case OP_ANYNL:
6051 for (i = min; i < max; i++)
6052 {
6053 if (eptr >= md->end_subject)
6054 {
6055 SCHECK_PARTIAL();
6056 break;
6057 }
6058 c = *eptr;
6059 if (c == CHAR_CR)
6060 {
6061 if (++eptr >= md->end_subject) break;
6062 if (*eptr == CHAR_LF) eptr++;
6063 }
6064 else
6065 {
6066 if (c != CHAR_LF && (md->bsr_anycrlf ||
6067 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
6068 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6069 && c != 0x2028 && c != 0x2029
6070 #endif
6071 ))) break;
6072 eptr++;
6073 }
6074 }
6075 break;
6076
6077 case OP_NOT_HSPACE:
6078 for (i = min; i < max; i++)
6079 {
6080 if (eptr >= md->end_subject)
6081 {
6082 SCHECK_PARTIAL();
6083 break;
6084 }
6085 switch(*eptr)
6086 {
6087 default: eptr++; break;
6088 HSPACE_BYTE_CASES:
6089 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6090 HSPACE_MULTIBYTE_CASES:
6091 #endif
6092 goto ENDLOOP00;
6093 }
6094 }
6095 ENDLOOP00:
6096 break;
6097
6098 case OP_HSPACE:
6099 for (i = min; i < max; i++)
6100 {
6101 if (eptr >= md->end_subject)
6102 {
6103 SCHECK_PARTIAL();
6104 break;
6105 }
6106 switch(*eptr)
6107 {
6108 default: goto ENDLOOP01;
6109 HSPACE_BYTE_CASES:
6110 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6111 HSPACE_MULTIBYTE_CASES:
6112 #endif
6113 eptr++; break;
6114 }
6115 }
6116 ENDLOOP01:
6117 break;
6118
6119 case OP_NOT_VSPACE:
6120 for (i = min; i < max; i++)
6121 {
6122 if (eptr >= md->end_subject)
6123 {
6124 SCHECK_PARTIAL();
6125 break;
6126 }
6127 switch(*eptr)
6128 {
6129 default: eptr++; break;
6130 VSPACE_BYTE_CASES:
6131 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6132 VSPACE_MULTIBYTE_CASES:
6133 #endif
6134 goto ENDLOOP02;
6135 }
6136 }
6137 ENDLOOP02:
6138 break;
6139
6140 case OP_VSPACE:
6141 for (i = min; i < max; i++)
6142 {
6143 if (eptr >= md->end_subject)
6144 {
6145 SCHECK_PARTIAL();
6146 break;
6147 }
6148 switch(*eptr)
6149 {
6150 default: goto ENDLOOP03;
6151 VSPACE_BYTE_CASES:
6152 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6153 VSPACE_MULTIBYTE_CASES:
6154 #endif
6155 eptr++; break;
6156 }
6157 }
6158 ENDLOOP03:
6159 break;
6160
6161 case OP_NOT_DIGIT:
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_digit) != 0) break;
6170 eptr++;
6171 }
6172 break;
6173
6174 case OP_DIGIT:
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_digit) == 0) break;
6183 eptr++;
6184 }
6185 break;
6186
6187 case OP_NOT_WHITESPACE:
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_space) != 0) break;
6196 eptr++;
6197 }
6198 break;
6199
6200 case OP_WHITESPACE:
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_space) == 0) break;
6209 eptr++;
6210 }
6211 break;
6212
6213 case OP_NOT_WORDCHAR:
6214 for (i = min; i < max; i++)
6215 {
6216 if (eptr >= md->end_subject)
6217 {
6218 SCHECK_PARTIAL();
6219 break;
6220 }
6221 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break;
6222 eptr++;
6223 }
6224 break;
6225
6226 case OP_WORDCHAR:
6227 for (i = min; i < max; i++)
6228 {
6229 if (eptr >= md->end_subject)
6230 {
6231 SCHECK_PARTIAL();
6232 break;
6233 }
6234 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break;
6235 eptr++;
6236 }
6237 break;
6238
6239 default:
6240 RRETURN(PCRE_ERROR_INTERNAL);
6241 }
6242
6243 if (possessive) continue; /* No backtracking */
6244 for (;;)
6245 {
6246 if (eptr == pp) goto TAIL_RECURSE;
6247 RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
6248 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
6249 eptr--;
6250 if (ctype == OP_ANYNL && eptr > pp && *eptr == CHAR_LF &&
6251 eptr[-1] == CHAR_CR) eptr--;
6252 }
6253 }
6254
6255 /* Control never gets here */
6256 }
6257
6258 /* There's been some horrible disaster. Arrival here can only mean there is
6259 something seriously wrong in the code above or the OP_xxx definitions. */
6260
6261 default:
6262 DPRINTF(("Unknown opcode %d\n", *ecode));
6263 RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
6264 }
6265
6266 /* Do not stick any code in here without much thought; it is assumed
6267 that "continue" in the code above comes out to here to repeat the main
6268 loop. */
6269
6270 } /* End of main loop */
6271 /* Control never reaches here */
6272
6273
6274 /* When compiling to use the heap rather than the stack for recursive calls to
6275 match(), the RRETURN() macro jumps here. The number that is saved in
6276 frame->Xwhere indicates which label we actually want to return to. */
6277
6278 #ifdef NO_RECURSE
6279 #define LBL(val) case val: goto L_RM##val;
6280 HEAP_RETURN:
6281 switch (frame->Xwhere)
6282 {
6283 LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
6284 LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
6285 LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
6286 LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
6287 LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64)
6288 LBL(65) LBL(66)
6289 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6290 LBL(21)
6291 #endif
6292 #ifdef SUPPORT_UTF
6293 LBL(16) LBL(18) LBL(20)
6294 LBL(22) LBL(23) LBL(28) LBL(30)
6295 LBL(32) LBL(34) LBL(42) LBL(46)
6296 #ifdef SUPPORT_UCP
6297 LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
6298 LBL(59) LBL(60) LBL(61) LBL(62) LBL(67) LBL(68)
6299 #endif /* SUPPORT_UCP */
6300 #endif /* SUPPORT_UTF */
6301 default:
6302 DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
6303 return PCRE_ERROR_INTERNAL;
6304 }
6305 #undef LBL
6306 #endif /* NO_RECURSE */
6307 }
6308
6309
6310 /***************************************************************************
6311 ****************************************************************************
6312 RECURSION IN THE match() FUNCTION
6313
6314 Undefine all the macros that were defined above to handle this. */
6315
6316 #ifdef NO_RECURSE
6317 #undef eptr
6318 #undef ecode
6319 #undef mstart
6320 #undef offset_top
6321 #undef eptrb
6322 #undef flags
6323
6324 #undef callpat
6325 #undef charptr
6326 #undef data
6327 #undef next
6328 #undef pp
6329 #undef prev
6330 #undef saved_eptr
6331
6332 #undef new_recursive
6333
6334 #undef cur_is_word
6335 #undef condition
6336 #undef prev_is_word
6337
6338 #undef ctype
6339 #undef length
6340 #undef max
6341 #undef min
6342 #undef number
6343 #undef offset
6344 #undef op
6345 #undef save_capture_last
6346 #undef save_offset1
6347 #undef save_offset2
6348 #undef save_offset3
6349 #undef stacksave
6350
6351 #undef newptrb
6352
6353 #endif
6354
6355 /* These two are defined as macros in both cases */
6356
6357 #undef fc
6358 #undef fi
6359
6360 /***************************************************************************
6361 ***************************************************************************/
6362
6363
6364 #ifdef NO_RECURSE
6365 /*************************************************
6366 * Release allocated heap frames *
6367 *************************************************/
6368
6369 /* This function releases all the allocated frames. The base frame is on the
6370 machine stack, and so must not be freed.
6371
6372 Argument: the address of the base frame
6373 Returns: nothing
6374 */
6375
6376 static void
6377 release_match_heapframes (heapframe *frame_base)
6378 {
6379 heapframe *nextframe = frame_base->Xnextframe;
6380 while (nextframe != NULL)
6381 {
6382 heapframe *oldframe = nextframe;
6383 nextframe = nextframe->Xnextframe;
6384 (PUBL(stack_free))(oldframe);
6385 }
6386 }
6387 #endif
6388
6389
6390 /*************************************************
6391 * Execute a Regular Expression *
6392 *************************************************/
6393
6394 /* This function applies a compiled re to a subject string and picks out
6395 portions of the string if it matches. Two elements in the vector are set for
6396 each substring: the offsets to the start and end of the substring.
6397
6398 Arguments:
6399 argument_re points to the compiled expression
6400 extra_data points to extra data or is NULL
6401 subject points to the subject string
6402 length length of subject string (may contain binary zeros)
6403 start_offset where to start in the subject string
6404 options option bits
6405 offsets points to a vector of ints to be filled in with offsets
6406 offsetcount the number of elements in the vector
6407
6408 Returns: > 0 => success; value is the number of elements filled in
6409 = 0 => success, but offsets is not big enough
6410 -1 => failed to match
6411 < -1 => some kind of unexpected problem
6412 */
6413
6414 #if defined COMPILE_PCRE8
6415 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6416 pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
6417 PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
6418 int offsetcount)
6419 #elif defined COMPILE_PCRE16
6420 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6421 pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data,
6422 PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets,
6423 int offsetcount)
6424 #elif defined COMPILE_PCRE32
6425 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6426 pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data,
6427 PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets,
6428 int offsetcount)
6429 #endif
6430 {
6431 int rc, ocount, arg_offset_max;
6432 int newline;
6433 BOOL using_temporary_offsets = FALSE;
6434 BOOL anchored;
6435 BOOL startline;
6436 BOOL firstline;
6437 BOOL utf;
6438 BOOL has_first_char = FALSE;
6439 BOOL has_req_char = FALSE;
6440 pcre_uchar first_char = 0;
6441 pcre_uchar first_char2 = 0;
6442 pcre_uchar req_char = 0;
6443 pcre_uchar req_char2 = 0;
6444 match_data match_block;
6445 match_data *md = &match_block;
6446 const pcre_uint8 *tables;
6447 const pcre_uint8 *start_bits = NULL;
6448 PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset;
6449 PCRE_PUCHAR end_subject;
6450 PCRE_PUCHAR start_partial = NULL;
6451 PCRE_PUCHAR match_partial = NULL;
6452 PCRE_PUCHAR req_char_ptr = start_match - 1;
6453
6454 const pcre_study_data *study;
6455 const REAL_PCRE *re = (const REAL_PCRE *)argument_re;
6456
6457 #ifdef NO_RECURSE
6458 heapframe frame_zero;
6459 frame_zero.Xprevframe = NULL; /* Marks the top level */
6460 frame_zero.Xnextframe = NULL; /* None are allocated yet */
6461 md->match_frames_base = &frame_zero;
6462 #endif
6463
6464 /* Check for the special magic call that measures the size of the stack used
6465 per recursive call of match(). Without the funny casting for sizeof, a Windows
6466 compiler gave this error: "unary minus operator applied to unsigned type,
6467 result still unsigned". Hopefully the cast fixes that. */
6468
6469 if (re == NULL && extra_data == NULL && subject == NULL && length == -999 &&
6470 start_offset == -999)
6471 #ifdef NO_RECURSE
6472 return -((int)sizeof(heapframe));
6473 #else
6474 return match(NULL, NULL, NULL, 0, NULL, NULL, 0);
6475 #endif
6476
6477 /* Plausibility checks */
6478
6479 if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
6480 if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0))
6481 return PCRE_ERROR_NULL;
6482 if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
6483 if (length < 0) return PCRE_ERROR_BADLENGTH;
6484 if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
6485
6486 /* Check that the first field in the block is the magic number. If it is not,
6487 return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to
6488 REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which
6489 means that the pattern is likely compiled with different endianness. */
6490
6491 if (re->magic_number != MAGIC_NUMBER)
6492 return re->magic_number == REVERSED_MAGIC_NUMBER?
6493 PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC;
6494 if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE;
6495
6496 /* These two settings are used in the code for checking a UTF-8 string that
6497 follows immediately afterwards. Other values in the md block are used only
6498 during "normal" pcre_exec() processing, not when the JIT support is in use,
6499 so they are set up later. */
6500
6501 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
6502 utf = md->utf = (re->options & PCRE_UTF8) != 0;
6503 md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
6504 ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
6505
6506 /* Check a UTF-8 string if required. Pass back the character offset and error
6507 code for an invalid string if a results vector is available. */
6508
6509 #ifdef SUPPORT_UTF
6510 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0)
6511 {
6512 int erroroffset;
6513 int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset);
6514 if (errorcode != 0)
6515 {
6516 if (offsetcount >= 2)
6517 {
6518 offsets[0] = erroroffset;
6519 offsets[1] = errorcode;
6520 }
6521 #if defined COMPILE_PCRE8
6522 return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
6523 PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
6524 #elif defined COMPILE_PCRE16
6525 return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)?
6526 PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16;
6527 #elif defined COMPILE_PCRE32
6528 return PCRE_ERROR_BADUTF32;
6529 #endif
6530 }
6531 #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16
6532 /* Check that a start_offset points to the start of a UTF character. */
6533 if (start_offset > 0 && start_offset < length &&
6534 NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset]))
6535 return PCRE_ERROR_BADUTF8_OFFSET;
6536 #endif
6537 }
6538 #endif
6539
6540 /* If the pattern was successfully studied with JIT support, run the JIT
6541 executable instead of the rest of this function. Most options must be set at
6542 compile time for the JIT code to be usable. Fallback to the normal code path if
6543 an unsupported flag is set. */
6544
6545 #ifdef SUPPORT_JIT
6546 if (extra_data != NULL
6547 && (extra_data->flags & (PCRE_EXTRA_EXECUTABLE_JIT |
6548 PCRE_EXTRA_TABLES)) == PCRE_EXTRA_EXECUTABLE_JIT
6549 && extra_data->executable_jit != NULL
6550 && (options & ~PUBLIC_JIT_EXEC_OPTIONS) == 0)
6551 {
6552 rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length,
6553 start_offset, options, offsets, offsetcount);
6554
6555 /* PCRE_ERROR_NULL means that the selected normal or partial matching
6556 mode is not compiled. In this case we simply fallback to interpreter. */
6557
6558 if (rc != PCRE_ERROR_JIT_BADOPTION) return rc;
6559 }
6560 #endif
6561
6562 /* Carry on with non-JIT matching. This information is for finding all the
6563 numbers associated with a given name, for condition testing. */
6564
6565 md->name_table = (pcre_uchar *)re + re->name_table_offset;
6566 md->name_count = re->name_count;
6567 md->name_entry_size = re->name_entry_size;
6568
6569 /* Fish out the optional data from the extra_data structure, first setting
6570 the default values. */
6571
6572 study = NULL;
6573 md->match_limit = MATCH_LIMIT;
6574 md->match_limit_recursion = MATCH_LIMIT_RECURSION;
6575 md->callout_data = NULL;
6576
6577 /* The table pointer is always in native byte order. */
6578
6579 tables = re->tables;
6580
6581 /* The two limit values override the defaults, whatever their value. */
6582
6583 if (extra_data != NULL)
6584 {
6585 register unsigned int flags = extra_data->flags;
6586 if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
6587 study = (const pcre_study_data *)extra_data->study_data;
6588 if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
6589 md->match_limit = extra_data->match_limit;
6590 if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
6591 md->match_limit_recursion = extra_data->match_limit_recursion;
6592 if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
6593 md->callout_data = extra_data->callout_data;
6594 if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
6595 }
6596
6597 /* Limits in the regex override only if they are smaller. */
6598
6599 if ((re->flags & PCRE_MLSET) != 0 && re->limit_match < md->match_limit)
6600 md->match_limit = re->limit_match;
6601
6602 if ((re->flags & PCRE_RLSET) != 0 &&
6603 re->limit_recursion < md->match_limit_recursion)
6604 md->match_limit_recursion = re->limit_recursion;
6605
6606 /* If the exec call supplied NULL for tables, use the inbuilt ones. This
6607 is a feature that makes it possible to save compiled regex and re-use them
6608 in other programs later. */
6609
6610 if (tables == NULL) tables = PRIV(default_tables);
6611
6612 /* Set up other data */
6613
6614 anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
6615 startline = (re->flags & PCRE_STARTLINE) != 0;
6616 firstline = (re->options & PCRE_FIRSTLINE) != 0;
6617
6618 /* The code starts after the real_pcre block and the capture name table. */
6619
6620 md->start_code = (const pcre_uchar *)re + re->name_table_offset +
6621 re->name_count * re->name_entry_size;
6622
6623 md->start_subject = (PCRE_PUCHAR)subject;
6624 md->start_offset = start_offset;
6625 md->end_subject = md->start_subject + length;
6626 end_subject = md->end_subject;
6627
6628 md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
6629 md->use_ucp = (re->options & PCRE_UCP) != 0;
6630 md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6631 md->ignore_skip_arg = 0;
6632
6633 /* Some options are unpacked into BOOL variables in the hope that testing
6634 them will be faster than individual option bits. */
6635
6636 md->notbol = (options & PCRE_NOTBOL) != 0;
6637 md->noteol = (options & PCRE_NOTEOL) != 0;
6638 md->notempty = (options & PCRE_NOTEMPTY) != 0;
6639 md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6640
6641 md->hitend = FALSE;
6642 md->mark = md->nomatch_mark = NULL; /* In case never set */
6643
6644 md->recursive = NULL; /* No recursion at top level */
6645 md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6646
6647 md->lcc = tables + lcc_offset;
6648 md->fcc = tables + fcc_offset;
6649 md->ctypes = tables + ctypes_offset;
6650
6651 /* Handle different \R options. */
6652
6653 switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6654 {
6655 case 0:
6656 if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
6657 md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
6658 else
6659 #ifdef BSR_ANYCRLF
6660 md->bsr_anycrlf = TRUE;
6661 #else
6662 md->bsr_anycrlf = FALSE;
6663 #endif
6664 break;
6665
6666 case PCRE_BSR_ANYCRLF:
6667 md->bsr_anycrlf = TRUE;
6668 break;
6669
6670 case PCRE_BSR_UNICODE:
6671 md->bsr_anycrlf = FALSE;
6672 break;
6673
6674 default: return PCRE_ERROR_BADNEWLINE;
6675 }
6676
6677 /* Handle different types of newline. The three bits give eight cases. If
6678 nothing is set at run time, whatever was used at compile time applies. */
6679
6680 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
6681 (pcre_uint32)options) & PCRE_NEWLINE_BITS)
6682 {
6683 case 0: newline = NEWLINE; break; /* Compile-time default */
6684 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6685 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6686 case PCRE_NEWLINE_CR+
6687 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6688 case PCRE_NEWLINE_ANY: newline = -1; break;
6689 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6690 default: return PCRE_ERROR_BADNEWLINE;
6691 }
6692
6693 if (newline == -2)
6694 {
6695 md->nltype = NLTYPE_ANYCRLF;
6696 }
6697 else if (newline < 0)
6698 {
6699 md->nltype = NLTYPE_ANY;
6700 }
6701 else
6702 {
6703 md->nltype = NLTYPE_FIXED;
6704 if (newline > 255)
6705 {
6706 md->nllen = 2;
6707 md->nl[0] = (newline >> 8) & 255;
6708 md->nl[1] = newline & 255;
6709 }
6710 else
6711 {
6712 md->nllen = 1;
6713 md->nl[0] = newline;
6714 }
6715 }
6716
6717 /* Partial matching was originally supported only for a restricted set of
6718 regexes; from release 8.00 there are no restrictions, but the bits are still
6719 defined (though never set). So there's no harm in leaving this code. */
6720
6721 if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6722 return PCRE_ERROR_BADPARTIAL;
6723
6724 /* If the expression has got more back references than the offsets supplied can
6725 hold, we get a temporary chunk of working store to use during the matching.
6726 Otherwise, we can use the vector supplied, rounding down its size to a multiple
6727 of 3. */
6728
6729 ocount = offsetcount - (offsetcount % 3);
6730 arg_offset_max = (2*ocount)/3;
6731
6732 if (re->top_backref > 0 && re->top_backref >= ocount/3)
6733 {
6734 ocount = re->top_backref * 3 + 3;
6735 md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int));
6736 if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
6737 using_temporary_offsets = TRUE;
6738 DPRINTF(("Got memory to hold back references\n"));
6739 }
6740 else md->offset_vector = offsets;
6741 md->offset_end = ocount;
6742 md->offset_max = (2*ocount)/3;
6743 md->capture_last = 0;
6744
6745 /* Reset the working variable associated with each extraction. These should
6746 never be used unless previously set, but they get saved and restored, and so we
6747 initialize them to avoid reading uninitialized locations. Also, unset the
6748 offsets for the matched string. This is really just for tidiness with callouts,
6749 in case they inspect these fields. */
6750
6751 if (md->offset_vector != NULL)
6752 {
6753 register int *iptr = md->offset_vector + ocount;
6754 register int *iend = iptr - re->top_bracket;
6755 if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;