tgupdate: merge pcreposix-compat base into pcreposix-compat
[pcreposix-compat.git] / pcre_exec.c
blob3fd58cbe31c1e430943bd8e31f0f93e693572527
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
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.
8 Written by Philip Hazel
9 Copyright (c) 1997-2018 University of Cambridge
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
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.
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.
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 -----------------------------------------------------------------------------
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. */
44 #ifdef HAVE_CONFIG_H
45 #include "config.h"
46 #endif
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 */
52 #include "pcre_internal.h"
54 /* Undefine some potentially clashing cpp symbols */
56 #undef min
57 #undef max
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. */
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 */
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. */
77 #define MATCH_CONDASSERT 1 /* Called to check a condition assertion */
78 #define MATCH_CBEGROUP 2 /* Could-be-empty unlimited repeat group */
80 /* Non-error returns from the match() function. Error returns are externally
81 defined PCRE_ERROR_xxx codes, which are all negative. */
83 #define MATCH_MATCH 1
84 #define MATCH_NOMATCH 0
86 /* Special internal returns from the match() function. Make them sufficiently
87 negative to avoid the external error codes. */
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
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. */
106 #define REC_STACK_SAVE_MAX 30
108 /* Min and max values for the common repeats; for the maxima, 0 => infinity */
110 static const char rep_min[] = { 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, };
111 static const char rep_max[] = { 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, };
113 #ifdef PCRE_DEBUG
114 /*************************************************
115 * Debugging function to print chars *
116 *************************************************/
118 /* Print a sequence of chars in printable format, stopping at the end of the
119 subject if the requested.
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
127 Returns: nothing
130 static void
131 pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
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 = UCHAR21INCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
139 #endif
143 /*************************************************
144 * Match a back-reference *
145 *************************************************/
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.
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
159 Returns: >= 0 the number of subject bytes matched
160 -1 no match
161 -2 partial match; always given if at end subject
164 static int
165 match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166 BOOL caseless)
168 PCRE_PUCHAR eptr_start = eptr;
169 register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170 #if defined SUPPORT_UTF && defined SUPPORT_UCP
171 BOOL utf = md->utf;
172 #endif
174 #ifdef PCRE_DEBUG
175 if (eptr >= md->end_subject)
176 printf("matching subject <null>");
177 else
179 printf("matching subject ");
180 pchars(eptr, length, TRUE, md);
182 printf(" against backref ");
183 pchars(p, length, FALSE, md);
184 printf("\n");
185 #endif
187 /* Always fail if reference not set (and not JavaScript compatible - in that
188 case the length is passed as zero). */
190 if (length < 0) return -1;
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. */
196 if (caseless)
198 #if defined SUPPORT_UTF && defined SUPPORT_UCP
199 if (utf)
201 /* Match characters up to the end of the reference. NOTE: the number of
202 data units matched may differ, because in UTF-8 there are some characters
203 whose upper and lower case versions code have different numbers of bytes.
204 For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
205 (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
206 sequence of two of the latter. It is important, therefore, to check the
207 length along the reference, not along the subject (earlier code did this
208 wrong). */
210 PCRE_PUCHAR endptr = p + length;
211 while (p < endptr)
213 pcre_uint32 c, d;
214 const ucd_record *ur;
215 if (eptr >= md->end_subject) return -2; /* Partial match */
216 GETCHARINC(c, eptr);
217 GETCHARINC(d, p);
218 ur = GET_UCD(d);
219 if (c != d && c != d + ur->other_case)
221 const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
222 for (;;)
224 if (c < *pp) return -1;
225 if (c == *pp++) break;
230 else
231 #endif
233 /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
234 is no UCP support. */
236 while (length-- > 0)
238 pcre_uint32 cc, cp;
239 if (eptr >= md->end_subject) return -2; /* Partial match */
240 cc = UCHAR21TEST(eptr);
241 cp = UCHAR21TEST(p);
242 if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
243 p++;
244 eptr++;
249 /* In the caseful case, we can just compare the bytes, whether or not we
250 are in UTF-8 mode. */
252 else
254 while (length-- > 0)
256 if (eptr >= md->end_subject) return -2; /* Partial match */
257 if (UCHAR21INCTEST(p) != UCHAR21INCTEST(eptr)) return -1;
261 return (int)(eptr - eptr_start);
266 /***************************************************************************
267 ****************************************************************************
268 RECURSION IN THE match() FUNCTION
270 The match() function is highly recursive, though not every recursive call
271 increases the recursive depth. Nevertheless, some regular expressions can cause
272 it to recurse to a great depth. I was writing for Unix, so I just let it call
273 itself recursively. This uses the stack for saving everything that has to be
274 saved for a recursive call. On Unix, the stack can be large, and this works
275 fine.
277 It turns out that on some non-Unix-like systems there are problems with
278 programs that use a lot of stack. (This despite the fact that every last chip
279 has oodles of memory these days, and techniques for extending the stack have
280 been known for decades.) So....
282 There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
283 calls by keeping local variables that need to be preserved in blocks of memory
284 obtained from malloc() instead instead of on the stack. Macros are used to
285 achieve this so that the actual code doesn't look very different to what it
286 always used to.
288 The original heap-recursive code used longjmp(). However, it seems that this
289 can be very slow on some operating systems. Following a suggestion from Stan
290 Switzer, the use of longjmp() has been abolished, at the cost of having to
291 provide a unique number for each call to RMATCH. There is no way of generating
292 a sequence of numbers at compile time in C. I have given them names, to make
293 them stand out more clearly.
295 Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
296 FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
297 tests. Furthermore, not using longjmp() means that local dynamic variables
298 don't have indeterminate values; this has meant that the frame size can be
299 reduced because the result can be "passed back" by straight setting of the
300 variable instead of being passed in the frame.
301 ****************************************************************************
302 ***************************************************************************/
304 /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
305 below must be updated in sync. */
307 enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10,
308 RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
309 RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
310 RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
311 RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
312 RM51, RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
313 RM61, RM62, RM63, RM64, RM65, RM66, RM67 };
315 /* These versions of the macros use the stack, as normal. There are debugging
316 versions and production versions. Note that the "rw" argument of RMATCH isn't
317 actually used in this definition. */
319 #ifndef NO_RECURSE
320 #define REGISTER register
322 #ifdef PCRE_DEBUG
323 #define RMATCH(ra,rb,rc,rd,re,rw) \
325 printf("match() called in line %d\n", __LINE__); \
326 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
327 printf("to line %d\n", __LINE__); \
329 #define RRETURN(ra) \
331 printf("match() returned %d from line %d\n", ra, __LINE__); \
332 return ra; \
334 #else
335 #define RMATCH(ra,rb,rc,rd,re,rw) \
336 rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
337 #define RRETURN(ra) return ra
338 #endif
340 #else
343 /* These versions of the macros manage a private stack on the heap. Note that
344 the "rd" argument of RMATCH isn't actually used in this definition. It's the md
345 argument of match(), which never changes. */
347 #define REGISTER
349 #define RMATCH(ra,rb,rc,rd,re,rw)\
351 heapframe *newframe = frame->Xnextframe;\
352 if (newframe == NULL)\
354 newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
355 if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
356 newframe->Xnextframe = NULL;\
357 frame->Xnextframe = newframe;\
359 frame->Xwhere = rw;\
360 newframe->Xeptr = ra;\
361 newframe->Xecode = rb;\
362 newframe->Xmstart = mstart;\
363 newframe->Xoffset_top = rc;\
364 newframe->Xeptrb = re;\
365 newframe->Xrdepth = frame->Xrdepth + 1;\
366 newframe->Xprevframe = frame;\
367 frame = newframe;\
368 DPRINTF(("restarting from line %d\n", __LINE__));\
369 goto HEAP_RECURSE;\
370 L_##rw:\
371 DPRINTF(("jumped back to line %d\n", __LINE__));\
374 #define RRETURN(ra)\
376 heapframe *oldframe = frame;\
377 frame = oldframe->Xprevframe;\
378 if (frame != NULL)\
380 rrc = ra;\
381 goto HEAP_RETURN;\
383 return ra;\
387 /* Structure for remembering the local variables in a private frame */
389 typedef struct heapframe {
390 struct heapframe *Xprevframe;
391 struct heapframe *Xnextframe;
393 /* Function arguments that may change */
395 PCRE_PUCHAR Xeptr;
396 const pcre_uchar *Xecode;
397 PCRE_PUCHAR Xmstart;
398 int Xoffset_top;
399 eptrblock *Xeptrb;
400 unsigned int Xrdepth;
402 /* Function local variables */
404 PCRE_PUCHAR Xcallpat;
405 #ifdef SUPPORT_UTF
406 PCRE_PUCHAR Xcharptr;
407 #endif
408 PCRE_PUCHAR Xdata;
409 PCRE_PUCHAR Xnext;
410 PCRE_PUCHAR Xpp;
411 PCRE_PUCHAR Xprev;
412 PCRE_PUCHAR Xsaved_eptr;
414 recursion_info Xnew_recursive;
416 BOOL Xcur_is_word;
417 BOOL Xcondition;
418 BOOL Xprev_is_word;
420 #ifdef SUPPORT_UCP
421 int Xprop_type;
422 unsigned int Xprop_value;
423 int Xprop_fail_result;
424 int Xoclength;
425 pcre_uchar Xocchars[6];
426 #endif
428 int Xcodelink;
429 int Xctype;
430 unsigned int Xfc;
431 int Xfi;
432 int Xlength;
433 int Xmax;
434 int Xmin;
435 unsigned int Xnumber;
436 int Xoffset;
437 unsigned int Xop;
438 pcre_int32 Xsave_capture_last;
439 int Xsave_offset1, Xsave_offset2, Xsave_offset3;
440 int Xstacksave[REC_STACK_SAVE_MAX];
442 eptrblock Xnewptrb;
444 /* Where to jump back to */
446 int Xwhere;
448 } heapframe;
450 #endif
453 /***************************************************************************
454 ***************************************************************************/
458 /*************************************************
459 * Match from current position *
460 *************************************************/
462 /* This function is called recursively in many circumstances. Whenever it
463 returns a negative (error) response, the outer incarnation must also return the
464 same response. */
466 /* These macros pack up tests that are used for partial matching, and which
467 appear several times in the code. We set the "hit end" flag if the pointer is
468 at the end of the subject and also past the start of the subject (i.e.
469 something has been matched). For hard partial matching, we then return
470 immediately. The second one is used when we already know we are past the end of
471 the subject. */
473 #define CHECK_PARTIAL()\
474 if (md->partial != 0 && eptr >= md->end_subject && \
475 eptr > md->start_used_ptr) \
477 md->hitend = TRUE; \
478 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481 #define SCHECK_PARTIAL()\
482 if (md->partial != 0 && eptr > md->start_used_ptr) \
484 md->hitend = TRUE; \
485 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
489 /* Performance note: It might be tempting to extract commonly used fields from
490 the md structure (e.g. utf, end_subject) into individual variables to improve
491 performance. Tests using gcc on a SPARC disproved this; in the first case, it
492 made performance worse.
494 Arguments:
495 eptr pointer to current character in subject
496 ecode pointer to current position in compiled code
497 mstart pointer to the current match start position (can be modified
498 by encountering \K)
499 offset_top current top pointer
500 md pointer to "static" info for the match
501 eptrb pointer to chain of blocks containing eptr at start of
502 brackets - for testing for empty matches
503 rdepth the recursion depth
505 Returns: MATCH_MATCH if matched ) these values are >= 0
506 MATCH_NOMATCH if failed to match )
507 a negative MATCH_xxx value for PRUNE, SKIP, etc
508 a negative PCRE_ERROR_xxx value if aborted by an error condition
509 (e.g. stopped by repeated call or recursion limit)
512 static int
513 match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
514 PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
515 unsigned int rdepth)
517 /* These variables do not need to be preserved over recursion in this function,
518 so they can be ordinary variables in all cases. Mark some of them with
519 "register" because they are used a lot in loops. */
521 register int rrc; /* Returns from recursive calls */
522 register int i; /* Used for loops not involving calls to RMATCH() */
523 register pcre_uint32 c; /* Character values not kept over RMATCH() calls */
524 register BOOL utf; /* Local copy of UTF flag for speed */
526 BOOL minimize, possessive; /* Quantifier options */
527 BOOL caseless;
528 int condcode;
530 /* When recursion is not being used, all "local" variables that have to be
531 preserved over calls to RMATCH() are part of a "frame". We set up the top-level
532 frame on the stack here; subsequent instantiations are obtained from the heap
533 whenever RMATCH() does a "recursion". See the macro definitions above. Putting
534 the top-level on the stack rather than malloc-ing them all gives a performance
535 boost in many cases where there is not much "recursion". */
537 #ifdef NO_RECURSE
538 heapframe *frame = (heapframe *)md->match_frames_base;
540 /* Copy in the original argument variables */
542 frame->Xeptr = eptr;
543 frame->Xecode = ecode;
544 frame->Xmstart = mstart;
545 frame->Xoffset_top = offset_top;
546 frame->Xeptrb = eptrb;
547 frame->Xrdepth = rdepth;
549 /* This is where control jumps back to to effect "recursion" */
551 HEAP_RECURSE:
553 /* Macros make the argument variables come from the current frame */
555 #define eptr frame->Xeptr
556 #define ecode frame->Xecode
557 #define mstart frame->Xmstart
558 #define offset_top frame->Xoffset_top
559 #define eptrb frame->Xeptrb
560 #define rdepth frame->Xrdepth
562 /* Ditto for the local variables */
564 #ifdef SUPPORT_UTF
565 #define charptr frame->Xcharptr
566 #endif
567 #define callpat frame->Xcallpat
568 #define codelink frame->Xcodelink
569 #define data frame->Xdata
570 #define next frame->Xnext
571 #define pp frame->Xpp
572 #define prev frame->Xprev
573 #define saved_eptr frame->Xsaved_eptr
575 #define new_recursive frame->Xnew_recursive
577 #define cur_is_word frame->Xcur_is_word
578 #define condition frame->Xcondition
579 #define prev_is_word frame->Xprev_is_word
581 #ifdef SUPPORT_UCP
582 #define prop_type frame->Xprop_type
583 #define prop_value frame->Xprop_value
584 #define prop_fail_result frame->Xprop_fail_result
585 #define oclength frame->Xoclength
586 #define occhars frame->Xocchars
587 #endif
589 #define ctype frame->Xctype
590 #define fc frame->Xfc
591 #define fi frame->Xfi
592 #define length frame->Xlength
593 #define max frame->Xmax
594 #define min frame->Xmin
595 #define number frame->Xnumber
596 #define offset frame->Xoffset
597 #define op frame->Xop
598 #define save_capture_last frame->Xsave_capture_last
599 #define save_offset1 frame->Xsave_offset1
600 #define save_offset2 frame->Xsave_offset2
601 #define save_offset3 frame->Xsave_offset3
602 #define stacksave frame->Xstacksave
604 #define newptrb frame->Xnewptrb
606 /* When recursion is being used, local variables are allocated on the stack and
607 get preserved during recursion in the normal way. In this environment, fi and
608 i, and fc and c, can be the same variables. */
610 #else /* NO_RECURSE not defined */
611 #define fi i
612 #define fc c
614 /* Many of the following variables are used only in small blocks of the code.
615 My normal style of coding would have declared them within each of those blocks.
616 However, in order to accommodate the version of this code that uses an external
617 "stack" implemented on the heap, it is easier to declare them all here, so the
618 declarations can be cut out in a block. The only declarations within blocks
619 below are for variables that do not have to be preserved over a recursive call
620 to RMATCH(). */
622 #ifdef SUPPORT_UTF
623 const pcre_uchar *charptr;
624 #endif
625 const pcre_uchar *callpat;
626 const pcre_uchar *data;
627 const pcre_uchar *next;
628 PCRE_PUCHAR pp;
629 const pcre_uchar *prev;
630 PCRE_PUCHAR saved_eptr;
632 recursion_info new_recursive;
634 BOOL cur_is_word;
635 BOOL condition;
636 BOOL prev_is_word;
638 #ifdef SUPPORT_UCP
639 int prop_type;
640 unsigned int prop_value;
641 int prop_fail_result;
642 int oclength;
643 pcre_uchar occhars[6];
644 #endif
646 int codelink;
647 int ctype;
648 int length;
649 int max;
650 int min;
651 unsigned int number;
652 int offset;
653 unsigned int op;
654 pcre_int32 save_capture_last;
655 int save_offset1, save_offset2, save_offset3;
656 int stacksave[REC_STACK_SAVE_MAX];
658 eptrblock newptrb;
660 /* There is a special fudge for calling match() in a way that causes it to
661 measure the size of its basic stack frame when the stack is being used for
662 recursion. The second argument (ecode) being NULL triggers this behaviour. It
663 cannot normally ever be NULL. The return is the negated value of the frame
664 size. */
666 if (ecode == NULL)
668 if (rdepth == 0)
669 return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
670 else
672 int len = (int)((char *)&rdepth - (char *)eptr);
673 return (len > 0)? -len : len;
676 #endif /* NO_RECURSE */
678 /* To save space on the stack and in the heap frame, I have doubled up on some
679 of the local variables that are used only in localised parts of the code, but
680 still need to be preserved over recursive calls of match(). These macros define
681 the alternative names that are used. */
683 #define allow_zero cur_is_word
684 #define cbegroup condition
685 #define code_offset codelink
686 #define condassert condition
687 #define matched_once prev_is_word
688 #define foc number
689 #define save_mark data
691 /* These statements are here to stop the compiler complaining about unitialized
692 variables. */
694 #ifdef SUPPORT_UCP
695 prop_value = 0;
696 prop_fail_result = 0;
697 #endif
700 /* This label is used for tail recursion, which is used in a few cases even
701 when NO_RECURSE is not defined, in order to reduce the amount of stack that is
702 used. Thanks to Ian Taylor for noticing this possibility and sending the
703 original patch. */
705 TAIL_RECURSE:
707 /* OK, now we can get on with the real code of the function. Recursive calls
708 are specified by the macro RMATCH and RRETURN is used to return. When
709 NO_RECURSE is *not* defined, these just turn into a recursive call to match()
710 and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
711 defined). However, RMATCH isn't like a function call because it's quite a
712 complicated macro. It has to be used in one particular way. This shouldn't,
713 however, impact performance when true recursion is being used. */
715 #ifdef SUPPORT_UTF
716 utf = md->utf; /* Local copy of the flag */
717 #else
718 utf = FALSE;
719 #endif
721 /* First check that we haven't called match() too many times, or that we
722 haven't exceeded the recursive call limit. */
724 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
725 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
727 /* At the start of a group with an unlimited repeat that may match an empty
728 string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
729 done this way to save having to use another function argument, which would take
730 up space on the stack. See also MATCH_CONDASSERT below.
732 When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
733 such remembered pointers, to be checked when we hit the closing ket, in order
734 to break infinite loops that match no characters. When match() is called in
735 other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
736 NOT be used with tail recursion, because the memory block that is used is on
737 the stack, so a new one may be required for each match(). */
739 if (md->match_function_type == MATCH_CBEGROUP)
741 newptrb.epb_saved_eptr = eptr;
742 newptrb.epb_prev = eptrb;
743 eptrb = &newptrb;
744 md->match_function_type = 0;
747 /* Now start processing the opcodes. */
749 for (;;)
751 minimize = possessive = FALSE;
752 op = *ecode;
754 switch(op)
756 case OP_MARK:
757 md->nomatch_mark = ecode + 2;
758 md->mark = NULL; /* In case previously set by assertion */
759 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
760 eptrb, RM55);
761 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
762 md->mark == NULL) md->mark = ecode + 2;
764 /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
765 argument, and we must check whether that argument matches this MARK's
766 argument. It is passed back in md->start_match_ptr (an overloading of that
767 variable). If it does match, we reset that variable to the current subject
768 position and return MATCH_SKIP. Otherwise, pass back the return code
769 unaltered. */
771 else if (rrc == MATCH_SKIP_ARG &&
772 STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
774 md->start_match_ptr = eptr;
775 RRETURN(MATCH_SKIP);
777 RRETURN(rrc);
779 case OP_FAIL:
780 RRETURN(MATCH_NOMATCH);
782 case OP_COMMIT:
783 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
784 eptrb, RM52);
785 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
786 RRETURN(MATCH_COMMIT);
788 case OP_PRUNE:
789 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
790 eptrb, RM51);
791 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
792 RRETURN(MATCH_PRUNE);
794 case OP_PRUNE_ARG:
795 md->nomatch_mark = ecode + 2;
796 md->mark = NULL; /* In case previously set by assertion */
797 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
798 eptrb, RM56);
799 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
800 md->mark == NULL) md->mark = ecode + 2;
801 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
802 RRETURN(MATCH_PRUNE);
804 case OP_SKIP:
805 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
806 eptrb, RM53);
807 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
808 md->start_match_ptr = eptr; /* Pass back current position */
809 RRETURN(MATCH_SKIP);
811 /* Note that, for Perl compatibility, SKIP with an argument does NOT set
812 nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
813 not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
814 that failed and any that precede it (either they also failed, or were not
815 triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
816 SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
817 set to the count of the one that failed. */
819 case OP_SKIP_ARG:
820 md->skip_arg_count++;
821 if (md->skip_arg_count <= md->ignore_skip_arg)
823 ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
824 break;
826 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
827 eptrb, RM57);
828 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
830 /* Pass back the current skip name by overloading md->start_match_ptr and
831 returning the special MATCH_SKIP_ARG return code. This will either be
832 caught by a matching MARK, or get to the top, where it causes a rematch
833 with md->ignore_skip_arg set to the value of md->skip_arg_count. */
835 md->start_match_ptr = ecode + 2;
836 RRETURN(MATCH_SKIP_ARG);
838 /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
839 the branch in which it occurs can be determined. Overload the start of
840 match pointer to do this. */
842 case OP_THEN:
843 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
844 eptrb, RM54);
845 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
846 md->start_match_ptr = ecode;
847 RRETURN(MATCH_THEN);
849 case OP_THEN_ARG:
850 md->nomatch_mark = ecode + 2;
851 md->mark = NULL; /* In case previously set by assertion */
852 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
853 md, eptrb, RM58);
854 if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
855 md->mark == NULL) md->mark = ecode + 2;
856 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
857 md->start_match_ptr = ecode;
858 RRETURN(MATCH_THEN);
860 /* Handle an atomic group that does not contain any capturing parentheses.
861 This can be handled like an assertion. Prior to 8.13, all atomic groups
862 were handled this way. In 8.13, the code was changed as below for ONCE, so
863 that backups pass through the group and thereby reset captured values.
864 However, this uses a lot more stack, so in 8.20, atomic groups that do not
865 contain any captures generate OP_ONCE_NC, which can be handled in the old,
866 less stack intensive way.
868 Check the alternative branches in turn - the matching won't pass the KET
869 for this kind of subpattern. If any one branch matches, we carry on as at
870 the end of a normal bracket, leaving the subject pointer, but resetting
871 the start-of-match value in case it was changed by \K. */
873 case OP_ONCE_NC:
874 prev = ecode;
875 saved_eptr = eptr;
876 save_mark = md->mark;
879 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
880 if (rrc == MATCH_MATCH) /* Note: _not_ MATCH_ACCEPT */
882 mstart = md->start_match_ptr;
883 break;
885 if (rrc == MATCH_THEN)
887 next = ecode + GET(ecode,1);
888 if (md->start_match_ptr < next &&
889 (*ecode == OP_ALT || *next == OP_ALT))
890 rrc = MATCH_NOMATCH;
893 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
894 ecode += GET(ecode,1);
895 md->mark = save_mark;
897 while (*ecode == OP_ALT);
899 /* If hit the end of the group (which could be repeated), fail */
901 if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
903 /* Continue as from after the group, updating the offsets high water
904 mark, since extracts may have been taken. */
906 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
908 offset_top = md->end_offset_top;
909 eptr = md->end_match_ptr;
911 /* For a non-repeating ket, just continue at this level. This also
912 happens for a repeating ket if no characters were matched in the group.
913 This is the forcible breaking of infinite loops as implemented in Perl
914 5.005. */
916 if (*ecode == OP_KET || eptr == saved_eptr)
918 ecode += 1+LINK_SIZE;
919 break;
922 /* The repeating kets try the rest of the pattern or restart from the
923 preceding bracket, in the appropriate order. The second "call" of match()
924 uses tail recursion, to avoid using another stack frame. */
926 if (*ecode == OP_KETRMIN)
928 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
929 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
930 ecode = prev;
931 goto TAIL_RECURSE;
933 else /* OP_KETRMAX */
935 RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
936 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
937 ecode += 1 + LINK_SIZE;
938 goto TAIL_RECURSE;
940 /* Control never gets here */
942 /* Handle a capturing bracket, other than those that are possessive with an
943 unlimited repeat. If there is space in the offset vector, save the current
944 subject position in the working slot at the top of the vector. We mustn't
945 change the current values of the data slot, because they may be set from a
946 previous iteration of this group, and be referred to by a reference inside
947 the group. A failure to match might occur after the group has succeeded,
948 if something later on doesn't match. For this reason, we need to restore
949 the working value and also the values of the final offsets, in case they
950 were set by a previous iteration of the same bracket.
952 If there isn't enough space in the offset vector, treat this as if it were
953 a non-capturing bracket. Don't worry about setting the flag for the error
954 case here; that is handled in the code for KET. */
956 case OP_CBRA:
957 case OP_SCBRA:
958 number = GET2(ecode, 1+LINK_SIZE);
959 offset = number << 1;
961 #ifdef PCRE_DEBUG
962 printf("start bracket %d\n", number);
963 printf("subject=");
964 pchars(eptr, 16, TRUE, md);
965 printf("\n");
966 #endif
968 if (offset < md->offset_max)
970 save_offset1 = md->offset_vector[offset];
971 save_offset2 = md->offset_vector[offset+1];
972 save_offset3 = md->offset_vector[md->offset_end - number];
973 save_capture_last = md->capture_last;
974 save_mark = md->mark;
976 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
977 md->offset_vector[md->offset_end - number] =
978 (int)(eptr - md->start_subject);
980 for (;;)
982 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
983 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
984 eptrb, RM1);
985 if (rrc == MATCH_ONCE) break; /* Backing up through an atomic group */
987 /* If we backed up to a THEN, check whether it is within the current
988 branch by comparing the address of the THEN that is passed back with
989 the end of the branch. If it is within the current branch, and the
990 branch is one of two or more alternatives (it either starts or ends
991 with OP_ALT), we have reached the limit of THEN's action, so convert
992 the return code to NOMATCH, which will cause normal backtracking to
993 happen from now on. Otherwise, THEN is passed back to an outer
994 alternative. This implements Perl's treatment of parenthesized groups,
995 where a group not containing | does not affect the current alternative,
996 that is, (X) is NOT the same as (X|(*F)). */
998 if (rrc == MATCH_THEN)
1000 next = ecode + GET(ecode,1);
1001 if (md->start_match_ptr < next &&
1002 (*ecode == OP_ALT || *next == OP_ALT))
1003 rrc = MATCH_NOMATCH;
1006 /* Anything other than NOMATCH is passed back. */
1008 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1009 md->capture_last = save_capture_last;
1010 ecode += GET(ecode, 1);
1011 md->mark = save_mark;
1012 if (*ecode != OP_ALT) break;
1015 DPRINTF(("bracket %d failed\n", number));
1016 md->offset_vector[offset] = save_offset1;
1017 md->offset_vector[offset+1] = save_offset2;
1018 md->offset_vector[md->offset_end - number] = save_offset3;
1020 /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1022 RRETURN(rrc);
1025 /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1026 as a non-capturing bracket. */
1028 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1029 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1031 DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1033 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1034 /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036 /* Non-capturing or atomic group, except for possessive with unlimited
1037 repeat and ONCE group with no captures. Loop for all the alternatives.
1039 When we get to the final alternative within the brackets, we used to return
1040 the result of a recursive call to match() whatever happened so it was
1041 possible to reduce stack usage by turning this into a tail recursion,
1042 except in the case of a possibly empty group. However, now that there is
1043 the possiblity of (*THEN) occurring in the final alternative, this
1044 optimization is no longer always possible.
1046 We can optimize if we know there are no (*THEN)s in the pattern; at present
1047 this is the best that can be done.
1049 MATCH_ONCE is returned when the end of an atomic group is successfully
1050 reached, but subsequent matching fails. It passes back up the tree (causing
1051 captured values to be reset) until the original atomic group level is
1052 reached. This is tested by comparing md->once_target with the start of the
1053 group. At this point, the return is converted into MATCH_NOMATCH so that
1054 previous backup points can be taken. */
1056 case OP_ONCE:
1057 case OP_BRA:
1058 case OP_SBRA:
1059 DPRINTF(("start non-capturing bracket\n"));
1061 for (;;)
1063 if (op >= OP_SBRA || op == OP_ONCE)
1064 md->match_function_type = MATCH_CBEGROUP;
1066 /* If this is not a possibly empty group, and there are no (*THEN)s in
1067 the pattern, and this is the final alternative, optimize as described
1068 above. */
1070 else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1072 ecode += PRIV(OP_lengths)[*ecode];
1073 goto TAIL_RECURSE;
1076 /* In all other cases, we have to make another call to match(). */
1078 save_mark = md->mark;
1079 save_capture_last = md->capture_last;
1080 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1081 RM2);
1083 /* See comment in the code for capturing groups above about handling
1084 THEN. */
1086 if (rrc == MATCH_THEN)
1088 next = ecode + GET(ecode,1);
1089 if (md->start_match_ptr < next &&
1090 (*ecode == OP_ALT || *next == OP_ALT))
1091 rrc = MATCH_NOMATCH;
1094 if (rrc != MATCH_NOMATCH)
1096 if (rrc == MATCH_ONCE)
1098 const pcre_uchar *scode = ecode;
1099 if (*scode != OP_ONCE) /* If not at start, find it */
1101 while (*scode == OP_ALT) scode += GET(scode, 1);
1102 scode -= GET(scode, 1);
1104 if (md->once_target == scode) rrc = MATCH_NOMATCH;
1106 RRETURN(rrc);
1108 ecode += GET(ecode, 1);
1109 md->mark = save_mark;
1110 if (*ecode != OP_ALT) break;
1111 md->capture_last = save_capture_last;
1114 RRETURN(MATCH_NOMATCH);
1116 /* Handle possessive capturing brackets with an unlimited repeat. We come
1117 here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1118 handled similarly to the normal case above. However, the matching is
1119 different. The end of these brackets will always be OP_KETRPOS, which
1120 returns MATCH_KETRPOS without going further in the pattern. By this means
1121 we can handle the group by iteration rather than recursion, thereby
1122 reducing the amount of stack needed. */
1124 case OP_CBRAPOS:
1125 case OP_SCBRAPOS:
1126 allow_zero = FALSE;
1128 POSSESSIVE_CAPTURE:
1129 number = GET2(ecode, 1+LINK_SIZE);
1130 offset = number << 1;
1132 #ifdef PCRE_DEBUG
1133 printf("start possessive bracket %d\n", number);
1134 printf("subject=");
1135 pchars(eptr, 16, TRUE, md);
1136 printf("\n");
1137 #endif
1139 if (offset >= md->offset_max) goto POSSESSIVE_NON_CAPTURE;
1141 matched_once = FALSE;
1142 code_offset = (int)(ecode - md->start_code);
1144 save_offset1 = md->offset_vector[offset];
1145 save_offset2 = md->offset_vector[offset+1];
1146 save_offset3 = md->offset_vector[md->offset_end - number];
1147 save_capture_last = md->capture_last;
1149 DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1151 /* Each time round the loop, save the current subject position for use
1152 when the group matches. For MATCH_MATCH, the group has matched, so we
1153 restart it with a new subject starting position, remembering that we had
1154 at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1155 usual. If we haven't matched any alternatives in any iteration, check to
1156 see if a previous iteration matched. If so, the group has matched;
1157 continue from afterwards. Otherwise it has failed; restore the previous
1158 capture values before returning NOMATCH. */
1160 for (;;)
1162 md->offset_vector[md->offset_end - number] =
1163 (int)(eptr - md->start_subject);
1164 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1165 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1166 eptrb, RM63);
1167 if (rrc == MATCH_KETRPOS)
1169 offset_top = md->end_offset_top;
1170 ecode = md->start_code + code_offset;
1171 save_capture_last = md->capture_last;
1172 matched_once = TRUE;
1173 mstart = md->start_match_ptr; /* In case \K changed it */
1174 if (eptr == md->end_match_ptr) /* Matched an empty string */
1176 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
1177 break;
1179 eptr = md->end_match_ptr;
1180 continue;
1183 /* See comment in the code for capturing groups above about handling
1184 THEN. */
1186 if (rrc == MATCH_THEN)
1188 next = ecode + GET(ecode,1);
1189 if (md->start_match_ptr < next &&
1190 (*ecode == OP_ALT || *next == OP_ALT))
1191 rrc = MATCH_NOMATCH;
1194 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1195 md->capture_last = save_capture_last;
1196 ecode += GET(ecode, 1);
1197 if (*ecode != OP_ALT) break;
1200 if (!matched_once)
1202 md->offset_vector[offset] = save_offset1;
1203 md->offset_vector[offset+1] = save_offset2;
1204 md->offset_vector[md->offset_end - number] = save_offset3;
1207 if (allow_zero || matched_once)
1209 ecode += 1 + LINK_SIZE;
1210 break;
1213 RRETURN(MATCH_NOMATCH);
1215 /* Non-capturing possessive bracket with unlimited repeat. We come here
1216 from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1217 without the capturing complication. It is written out separately for speed
1218 and cleanliness. */
1220 case OP_BRAPOS:
1221 case OP_SBRAPOS:
1222 allow_zero = FALSE;
1224 POSSESSIVE_NON_CAPTURE:
1225 matched_once = FALSE;
1226 code_offset = (int)(ecode - md->start_code);
1227 save_capture_last = md->capture_last;
1229 for (;;)
1231 if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1232 RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1233 eptrb, RM48);
1234 if (rrc == MATCH_KETRPOS)
1236 offset_top = md->end_offset_top;
1237 ecode = md->start_code + code_offset;
1238 matched_once = TRUE;
1239 mstart = md->start_match_ptr; /* In case \K reset it */
1240 if (eptr == md->end_match_ptr) /* Matched an empty string */
1242 do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
1243 break;
1245 eptr = md->end_match_ptr;
1246 continue;
1249 /* See comment in the code for capturing groups above about handling
1250 THEN. */
1252 if (rrc == MATCH_THEN)
1254 next = ecode + GET(ecode,1);
1255 if (md->start_match_ptr < next &&
1256 (*ecode == OP_ALT || *next == OP_ALT))
1257 rrc = MATCH_NOMATCH;
1260 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1261 ecode += GET(ecode, 1);
1262 if (*ecode != OP_ALT) break;
1263 md->capture_last = save_capture_last;
1266 if (matched_once || allow_zero)
1268 ecode += 1 + LINK_SIZE;
1269 break;
1271 RRETURN(MATCH_NOMATCH);
1273 /* Control never reaches here. */
1275 /* Conditional group: compilation checked that there are no more than two
1276 branches. If the condition is false, skipping the first branch takes us
1277 past the end of the item if there is only one branch, but that's exactly
1278 what we want. */
1280 case OP_COND:
1281 case OP_SCOND:
1283 /* The variable codelink will be added to ecode when the condition is
1284 false, to get to the second branch. Setting it to the offset to the ALT
1285 or KET, then incrementing ecode achieves this effect. We now have ecode
1286 pointing to the condition or callout. */
1288 codelink = GET(ecode, 1); /* Offset to the second branch */
1289 ecode += 1 + LINK_SIZE; /* From this opcode */
1291 /* Because of the way auto-callout works during compile, a callout item is
1292 inserted between OP_COND and an assertion condition. */
1294 if (*ecode == OP_CALLOUT)
1296 if (PUBL(callout) != NULL)
1298 PUBL(callout_block) cb;
1299 cb.version = 2; /* Version 1 of the callout block */
1300 cb.callout_number = ecode[1];
1301 cb.offset_vector = md->offset_vector;
1302 #if defined COMPILE_PCRE8
1303 cb.subject = (PCRE_SPTR)md->start_subject;
1304 #elif defined COMPILE_PCRE16
1305 cb.subject = (PCRE_SPTR16)md->start_subject;
1306 #elif defined COMPILE_PCRE32
1307 cb.subject = (PCRE_SPTR32)md->start_subject;
1308 #endif
1309 cb.subject_length = (int)(md->end_subject - md->start_subject);
1310 cb.start_match = (int)(mstart - md->start_subject);
1311 cb.current_position = (int)(eptr - md->start_subject);
1312 cb.pattern_position = GET(ecode, 2);
1313 cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1314 cb.capture_top = offset_top/2;
1315 cb.capture_last = md->capture_last & CAPLMASK;
1316 /* Internal change requires this for API compatibility. */
1317 if (cb.capture_last == 0) cb.capture_last = -1;
1318 cb.callout_data = md->callout_data;
1319 cb.mark = md->nomatch_mark;
1320 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1321 if (rrc < 0) RRETURN(rrc);
1324 /* Advance ecode past the callout, so it now points to the condition. We
1325 must adjust codelink so that the value of ecode+codelink is unchanged. */
1327 ecode += PRIV(OP_lengths)[OP_CALLOUT];
1328 codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1331 /* Test the various possible conditions */
1333 condition = FALSE;
1334 switch(condcode = *ecode)
1336 case OP_RREF: /* Numbered group recursion test */
1337 if (md->recursive != NULL) /* Not recursing => FALSE */
1339 unsigned int recno = GET2(ecode, 1); /* Recursion group number*/
1340 condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1342 break;
1344 case OP_DNRREF: /* Duplicate named group recursion test */
1345 if (md->recursive != NULL)
1347 int count = GET2(ecode, 1 + IMM2_SIZE);
1348 pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1349 while (count-- > 0)
1351 unsigned int recno = GET2(slot, 0);
1352 condition = recno == md->recursive->group_num;
1353 if (condition) break;
1354 slot += md->name_entry_size;
1357 break;
1359 case OP_CREF: /* Numbered group used test */
1360 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
1361 condition = offset < offset_top && md->offset_vector[offset] >= 0;
1362 break;
1364 case OP_DNCREF: /* Duplicate named group used test */
1366 int count = GET2(ecode, 1 + IMM2_SIZE);
1367 pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1368 while (count-- > 0)
1370 offset = GET2(slot, 0) << 1;
1371 condition = offset < offset_top && md->offset_vector[offset] >= 0;
1372 if (condition) break;
1373 slot += md->name_entry_size;
1376 break;
1378 case OP_DEF: /* DEFINE - always false */
1379 case OP_FAIL: /* From optimized (?!) condition */
1380 break;
1382 /* The condition is an assertion. Call match() to evaluate it - setting
1383 md->match_function_type to MATCH_CONDASSERT causes it to stop at the end
1384 of an assertion. */
1386 default:
1387 md->match_function_type = MATCH_CONDASSERT;
1388 RMATCH(eptr, ecode, offset_top, md, NULL, RM3);
1389 if (rrc == MATCH_MATCH)
1391 if (md->end_offset_top > offset_top)
1392 offset_top = md->end_offset_top; /* Captures may have happened */
1393 condition = TRUE;
1395 /* Advance ecode past the assertion to the start of the first branch,
1396 but adjust it so that the general choosing code below works. If the
1397 assertion has a quantifier that allows zero repeats we must skip over
1398 the BRAZERO. This is a lunatic thing to do, but somebody did! */
1400 if (*ecode == OP_BRAZERO) ecode++;
1401 ecode += GET(ecode, 1);
1402 while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1403 ecode += 1 + LINK_SIZE - PRIV(OP_lengths)[condcode];
1406 /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1407 assertion; it is therefore treated as NOMATCH. Any other return is an
1408 error. */
1410 else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1412 RRETURN(rrc); /* Need braces because of following else */
1414 break;
1417 /* Choose branch according to the condition */
1419 ecode += condition? PRIV(OP_lengths)[condcode] : codelink;
1421 /* We are now at the branch that is to be obeyed. As there is only one, we
1422 can use tail recursion to avoid using another stack frame, except when
1423 there is unlimited repeat of a possibly empty group. In the latter case, a
1424 recursive call to match() is always required, unless the second alternative
1425 doesn't exist, in which case we can just plough on. Note that, for
1426 compatibility with Perl, the | in a conditional group is NOT treated as
1427 creating two alternatives. If a THEN is encountered in the branch, it
1428 propagates out to the enclosing alternative (unless nested in a deeper set
1429 of alternatives, of course). */
1431 if (condition || ecode[-(1+LINK_SIZE)] == OP_ALT)
1433 if (op != OP_SCOND)
1435 goto TAIL_RECURSE;
1438 md->match_function_type = MATCH_CBEGROUP;
1439 RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);
1440 RRETURN(rrc);
1443 /* Condition false & no alternative; continue after the group. */
1445 else
1448 break;
1451 /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1452 to close any currently open capturing brackets. */
1454 case OP_CLOSE:
1455 number = GET2(ecode, 1); /* Must be less than 65536 */
1456 offset = number << 1;
1458 #ifdef PCRE_DEBUG
1459 printf("end bracket %d at *ACCEPT", number);
1460 printf("\n");
1461 #endif
1463 md->capture_last = (md->capture_last & OVFLMASK) | number;
1464 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1466 md->offset_vector[offset] =
1467 md->offset_vector[md->offset_end - number];
1468 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1470 /* If this group is at or above the current highwater mark, ensure that
1471 any groups between the current high water mark and this group are marked
1472 unset and then update the high water mark. */
1474 if (offset >= offset_top)
1476 register int *iptr = md->offset_vector + offset_top;
1477 register int *iend = md->offset_vector + offset;
1478 while (iptr < iend) *iptr++ = -1;
1479 offset_top = offset + 2;
1482 ecode += 1 + IMM2_SIZE;
1483 break;
1486 /* End of the pattern, either real or forced. */
1488 case OP_END:
1489 case OP_ACCEPT:
1490 case OP_ASSERT_ACCEPT:
1492 /* If we have matched an empty string, fail if not in an assertion and not
1493 in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1494 is set and we have matched at the start of the subject. In both cases,
1495 backtracking will then try other alternatives, if any. */
1497 if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1498 md->recursive == NULL &&
1499 (md->notempty ||
1500 (md->notempty_atstart &&
1501 mstart == md->start_subject + md->start_offset)))
1502 RRETURN(MATCH_NOMATCH);
1504 /* Otherwise, we have a match. */
1506 md->end_match_ptr = eptr; /* Record where we ended */
1507 md->end_offset_top = offset_top; /* and how many extracts were taken */
1508 md->start_match_ptr = mstart; /* and the start (\K can modify) */
1510 /* For some reason, the macros don't work properly if an expression is
1511 given as the argument to RRETURN when the heap is in use. */
1513 rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1514 RRETURN(rrc);
1516 /* Assertion brackets. Check the alternative branches in turn - the
1517 matching won't pass the KET for an assertion. If any one branch matches,
1518 the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1519 start of each branch to move the current point backwards, so the code at
1520 this level is identical to the lookahead case. When the assertion is part
1521 of a condition, we want to return immediately afterwards. The caller of
1522 this incarnation of the match() function will have set MATCH_CONDASSERT in
1523 md->match_function type, and one of these opcodes will be the first opcode
1524 that is processed. We use a local variable that is preserved over calls to
1525 match() to remember this case. */
1527 case OP_ASSERT:
1528 case OP_ASSERTBACK:
1529 save_mark = md->mark;
1530 if (md->match_function_type == MATCH_CONDASSERT)
1532 condassert = TRUE;
1533 md->match_function_type = 0;
1535 else condassert = FALSE;
1537 /* Loop for each branch */
1541 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1543 /* A match means that the assertion is true; break out of the loop
1544 that matches its alternatives. */
1546 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1548 mstart = md->start_match_ptr; /* In case \K reset it */
1549 break;
1552 /* If not matched, restore the previous mark setting. */
1554 md->mark = save_mark;
1556 /* See comment in the code for capturing groups above about handling
1557 THEN. */
1559 if (rrc == MATCH_THEN)
1561 next = ecode + GET(ecode,1);
1562 if (md->start_match_ptr < next &&
1563 (*ecode == OP_ALT || *next == OP_ALT))
1564 rrc = MATCH_NOMATCH;
1567 /* Anything other than NOMATCH causes the entire assertion to fail,
1568 passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1569 uncaptured THEN, which means they take their normal effect. This
1570 consistent approach does not always have exactly the same effect as in
1571 Perl. */
1573 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1574 ecode += GET(ecode, 1);
1576 while (*ecode == OP_ALT); /* Continue for next alternative */
1578 /* If we have tried all the alternative branches, the assertion has
1579 failed. If not, we broke out after a match. */
1581 if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1583 /* If checking an assertion for a condition, return MATCH_MATCH. */
1585 if (condassert) RRETURN(MATCH_MATCH);
1587 /* Continue from after a successful assertion, updating the offsets high
1588 water mark, since extracts may have been taken during the assertion. */
1590 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1591 ecode += 1 + LINK_SIZE;
1592 offset_top = md->end_offset_top;
1593 continue;
1595 /* Negative assertion: all branches must fail to match for the assertion to
1596 succeed. */
1598 case OP_ASSERT_NOT:
1599 case OP_ASSERTBACK_NOT:
1600 save_mark = md->mark;
1601 if (md->match_function_type == MATCH_CONDASSERT)
1603 condassert = TRUE;
1604 md->match_function_type = 0;
1606 else condassert = FALSE;
1608 /* Loop for each alternative branch. */
1612 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1613 md->mark = save_mark; /* Always restore the mark setting */
1615 switch(rrc)
1617 case MATCH_MATCH: /* A successful match means */
1618 case MATCH_ACCEPT: /* the assertion has failed. */
1619 RRETURN(MATCH_NOMATCH);
1621 case MATCH_NOMATCH: /* Carry on with next branch */
1622 break;
1624 /* See comment in the code for capturing groups above about handling
1625 THEN. */
1627 case MATCH_THEN:
1628 next = ecode + GET(ecode,1);
1629 if (md->start_match_ptr < next &&
1630 (*ecode == OP_ALT || *next == OP_ALT))
1632 rrc = MATCH_NOMATCH;
1633 break;
1635 /* Otherwise fall through. */
1637 /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1638 assertion to fail to match, without considering any more alternatives.
1639 Failing to match means the assertion is true. This is a consistent
1640 approach, but does not always have the same effect as in Perl. */
1642 case MATCH_COMMIT:
1643 case MATCH_SKIP:
1644 case MATCH_SKIP_ARG:
1645 case MATCH_PRUNE:
1646 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1647 goto NEG_ASSERT_TRUE; /* Break out of alternation loop */
1649 /* Anything else is an error */
1651 default:
1652 RRETURN(rrc);
1655 /* Continue with next branch */
1657 ecode += GET(ecode,1);
1659 while (*ecode == OP_ALT);
1661 /* All branches in the assertion failed to match. */
1663 NEG_ASSERT_TRUE:
1664 if (condassert) RRETURN(MATCH_MATCH); /* Condition assertion */
1665 ecode += 1 + LINK_SIZE; /* Continue with current branch */
1666 continue;
1668 /* Move the subject pointer back. This occurs only at the start of
1669 each branch of a lookbehind assertion. If we are too close to the start to
1670 move back, this match function fails. When working with UTF-8 we move
1671 back a number of characters, not bytes. */
1673 case OP_REVERSE:
1674 #ifdef SUPPORT_UTF
1675 if (utf)
1677 i = GET(ecode, 1);
1678 while (i-- > 0)
1680 eptr--;
1681 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1682 BACKCHAR(eptr);
1685 else
1686 #endif
1688 /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1691 eptr -= GET(ecode, 1);
1692 if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1695 /* Save the earliest consulted character, then skip to next op code */
1697 if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1698 ecode += 1 + LINK_SIZE;
1699 break;
1701 /* The callout item calls an external function, if one is provided, passing
1702 details of the match so far. This is mainly for debugging, though the
1703 function is able to force a failure. */
1705 case OP_CALLOUT:
1706 if (PUBL(callout) != NULL)
1708 PUBL(callout_block) cb;
1709 cb.version = 2; /* Version 1 of the callout block */
1710 cb.callout_number = ecode[1];
1711 cb.offset_vector = md->offset_vector;
1712 #if defined COMPILE_PCRE8
1713 cb.subject = (PCRE_SPTR)md->start_subject;
1714 #elif defined COMPILE_PCRE16
1715 cb.subject = (PCRE_SPTR16)md->start_subject;
1716 #elif defined COMPILE_PCRE32
1717 cb.subject = (PCRE_SPTR32)md->start_subject;
1718 #endif
1719 cb.subject_length = (int)(md->end_subject - md->start_subject);
1720 cb.start_match = (int)(mstart - md->start_subject);
1721 cb.current_position = (int)(eptr - md->start_subject);
1722 cb.pattern_position = GET(ecode, 2);
1723 cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1724 cb.capture_top = offset_top/2;
1725 cb.capture_last = md->capture_last & CAPLMASK;
1726 /* Internal change requires this for API compatibility. */
1727 if (cb.capture_last == 0) cb.capture_last = -1;
1728 cb.callout_data = md->callout_data;
1729 cb.mark = md->nomatch_mark;
1730 if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1731 if (rrc < 0) RRETURN(rrc);
1733 ecode += 2 + 2*LINK_SIZE;
1734 break;
1736 /* Recursion either matches the current regex, or some subexpression. The
1737 offset data is the offset to the starting bracket from the start of the
1738 whole pattern. (This is so that it works from duplicated subpatterns.)
1740 The state of the capturing groups is preserved over recursion, and
1741 re-instated afterwards. We don't know how many are started and not yet
1742 finished (offset_top records the completed total) so we just have to save
1743 all the potential data. There may be up to 65535 such values, which is too
1744 large to put on the stack, but using malloc for small numbers seems
1745 expensive. As a compromise, the stack is used when there are no more than
1746 REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
1748 There are also other values that have to be saved. We use a chained
1749 sequence of blocks that actually live on the stack. Thanks to Robin Houston
1750 for the original version of this logic. It has, however, been hacked around
1751 a lot, so he is not to blame for the current way it works. */
1753 case OP_RECURSE:
1755 recursion_info *ri;
1756 unsigned int recno;
1758 callpat = md->start_code + GET(ecode, 1);
1759 recno = (callpat == md->start_code)? 0 :
1760 GET2(callpat, 1 + LINK_SIZE);
1762 /* Check for repeating a recursion without advancing the subject pointer.
1763 This should catch convoluted mutual recursions. (Some simple cases are
1764 caught at compile time.) */
1766 for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1767 if (recno == ri->group_num && eptr == ri->subject_position)
1768 RRETURN(PCRE_ERROR_RECURSELOOP);
1770 /* Add to "recursing stack" */
1772 new_recursive.group_num = recno;
1773 new_recursive.saved_capture_last = md->capture_last;
1774 new_recursive.subject_position = eptr;
1775 new_recursive.prevrec = md->recursive;
1776 md->recursive = &new_recursive;
1778 /* Where to continue from afterwards */
1780 ecode += 1 + LINK_SIZE;
1782 /* Now save the offset data */
1784 new_recursive.saved_max = md->offset_end;
1785 if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
1786 new_recursive.offset_save = stacksave;
1787 else
1789 new_recursive.offset_save =
1790 (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1791 if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1793 memcpy(new_recursive.offset_save, md->offset_vector,
1794 new_recursive.saved_max * sizeof(int));
1796 /* OK, now we can do the recursion. After processing each alternative,
1797 restore the offset data and the last captured value. If there were nested
1798 recursions, md->recursive might be changed, so reset it before looping.
1801 DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1802 cbegroup = (*callpat >= OP_SBRA);
1805 if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1806 RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1807 md, eptrb, RM6);
1808 memcpy(md->offset_vector, new_recursive.offset_save,
1809 new_recursive.saved_max * sizeof(int));
1810 md->capture_last = new_recursive.saved_capture_last;
1811 md->recursive = new_recursive.prevrec;
1812 if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1814 DPRINTF(("Recursion matched\n"));
1815 if (new_recursive.offset_save != stacksave)
1816 (PUBL(free))(new_recursive.offset_save);
1818 /* Set where we got to in the subject, and reset the start in case
1819 it was changed by \K. This *is* propagated back out of a recursion,
1820 for Perl compatibility. */
1822 eptr = md->end_match_ptr;
1823 mstart = md->start_match_ptr;
1824 goto RECURSION_MATCHED; /* Exit loop; end processing */
1827 /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1828 recursion; they cause a NOMATCH for the entire recursion. These codes
1829 are defined in a range that can be tested for. */
1831 if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1833 if (new_recursive.offset_save != stacksave)
1834 (PUBL(free))(new_recursive.offset_save);
1835 RRETURN(MATCH_NOMATCH);
1838 /* Any return code other than NOMATCH is an error. */
1840 if (rrc != MATCH_NOMATCH)
1842 DPRINTF(("Recursion gave error %d\n", rrc));
1843 if (new_recursive.offset_save != stacksave)
1844 (PUBL(free))(new_recursive.offset_save);
1845 RRETURN(rrc);
1848 md->recursive = &new_recursive;
1849 callpat += GET(callpat, 1);
1851 while (*callpat == OP_ALT);
1853 DPRINTF(("Recursion didn't match\n"));
1854 md->recursive = new_recursive.prevrec;
1855 if (new_recursive.offset_save != stacksave)
1856 (PUBL(free))(new_recursive.offset_save);
1857 RRETURN(MATCH_NOMATCH);
1860 RECURSION_MATCHED:
1861 break;
1863 /* An alternation is the end of a branch; scan along to find the end of the
1864 bracketed group and go to there. */
1866 case OP_ALT:
1867 do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1868 break;
1870 /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1871 indicating that it may occur zero times. It may repeat infinitely, or not
1872 at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1873 with fixed upper repeat limits are compiled as a number of copies, with the
1874 optional ones preceded by BRAZERO or BRAMINZERO. */
1876 case OP_BRAZERO:
1877 next = ecode + 1;
1878 RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1879 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1880 do next += GET(next, 1); while (*next == OP_ALT);
1881 ecode = next + 1 + LINK_SIZE;
1882 break;
1884 case OP_BRAMINZERO:
1885 next = ecode + 1;
1886 do next += GET(next, 1); while (*next == OP_ALT);
1887 RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1888 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1889 ecode++;
1890 break;
1892 case OP_SKIPZERO:
1893 next = ecode+1;
1894 do next += GET(next,1); while (*next == OP_ALT);
1895 ecode = next + 1 + LINK_SIZE;
1896 break;
1898 /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1899 here; just jump to the group, with allow_zero set TRUE. */
1901 case OP_BRAPOSZERO:
1902 op = *(++ecode);
1903 allow_zero = TRUE;
1904 if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1905 goto POSSESSIVE_NON_CAPTURE;
1907 /* End of a group, repeated or non-repeating. */
1909 case OP_KET:
1910 case OP_KETRMIN:
1911 case OP_KETRMAX:
1912 case OP_KETRPOS:
1913 prev = ecode - GET(ecode, 1);
1915 /* If this was a group that remembered the subject start, in order to break
1916 infinite repeats of empty string matches, retrieve the subject start from
1917 the chain. Otherwise, set it NULL. */
1919 if (*prev >= OP_SBRA || *prev == OP_ONCE)
1921 saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */
1922 eptrb = eptrb->epb_prev; /* Backup to previous group */
1924 else saved_eptr = NULL;
1926 /* If we are at the end of an assertion group or a non-capturing atomic
1927 group, stop matching and return MATCH_MATCH, but record the current high
1928 water mark for use by positive assertions. We also need to record the match
1929 start in case it was changed by \K. */
1931 if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1932 *prev == OP_ONCE_NC)
1934 md->end_match_ptr = eptr; /* For ONCE_NC */
1935 md->end_offset_top = offset_top;
1936 md->start_match_ptr = mstart;
1937 RRETURN(MATCH_MATCH); /* Sets md->mark */
1940 /* For capturing groups we have to check the group number back at the start
1941 and if necessary complete handling an extraction by setting the offsets and
1942 bumping the high water mark. Whole-pattern recursion is coded as a recurse
1943 into group 0, so it won't be picked up here. Instead, we catch it when the
1944 OP_END is reached. Other recursion is handled here. We just have to record
1945 the current subject position and start match pointer and give a MATCH
1946 return. */
1948 if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1949 *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1951 number = GET2(prev, 1+LINK_SIZE);
1952 offset = number << 1;
1954 #ifdef PCRE_DEBUG
1955 printf("end bracket %d", number);
1956 printf("\n");
1957 #endif
1959 /* Handle a recursively called group. */
1961 if (md->recursive != NULL && md->recursive->group_num == number)
1963 md->end_match_ptr = eptr;
1964 md->start_match_ptr = mstart;
1965 RRETURN(MATCH_MATCH);
1968 /* Deal with capturing */
1970 md->capture_last = (md->capture_last & OVFLMASK) | number;
1971 if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1973 /* If offset is greater than offset_top, it means that we are
1974 "skipping" a capturing group, and that group's offsets must be marked
1975 unset. In earlier versions of PCRE, all the offsets were unset at the
1976 start of matching, but this doesn't work because atomic groups and
1977 assertions can cause a value to be set that should later be unset.
1978 Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1979 part of the atomic group, but this is not on the final matching path,
1980 so must be unset when 2 is set. (If there is no group 2, there is no
1981 problem, because offset_top will then be 2, indicating no capture.) */
1983 if (offset > offset_top)
1985 register int *iptr = md->offset_vector + offset_top;
1986 register int *iend = md->offset_vector + offset;
1987 while (iptr < iend) *iptr++ = -1;
1990 /* Now make the extraction */
1992 md->offset_vector[offset] =
1993 md->offset_vector[md->offset_end - number];
1994 md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1995 if (offset_top <= offset) offset_top = offset + 2;
1999 /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2000 and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2001 at a time from the outer level, thus saving stack. This must precede the
2002 empty string test - in this case that test is done at the outer level. */
2004 if (*ecode == OP_KETRPOS)
2006 md->start_match_ptr = mstart; /* In case \K reset it */
2007 md->end_match_ptr = eptr;
2008 md->end_offset_top = offset_top;
2009 RRETURN(MATCH_KETRPOS);
2012 /* For an ordinary non-repeating ket, just continue at this level. This
2013 also happens for a repeating ket if no characters were matched in the
2014 group. This is the forcible breaking of infinite loops as implemented in
2015 Perl 5.005. For a non-repeating atomic group that includes captures,
2016 establish a backup point by processing the rest of the pattern at a lower
2017 level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2018 original OP_ONCE level, thereby bypassing intermediate backup points, but
2019 resetting any captures that happened along the way. */
2021 if (*ecode == OP_KET || eptr == saved_eptr)
2023 if (*prev == OP_ONCE)
2025 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2026 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2027 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
2028 RRETURN(MATCH_ONCE);
2030 ecode += 1 + LINK_SIZE; /* Carry on at this level */
2031 break;
2034 /* The normal repeating kets try the rest of the pattern or restart from
2035 the preceding bracket, in the appropriate order. In the second case, we can
2036 use tail recursion to avoid using another stack frame, unless we have an
2037 an atomic group or an unlimited repeat of a group that can match an empty
2038 string. */
2040 if (*ecode == OP_KETRMIN)
2042 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2043 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2044 if (*prev == OP_ONCE)
2046 RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2047 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2048 md->once_target = prev; /* Level at which to change to MATCH_NOMATCH */
2049 RRETURN(MATCH_ONCE);
2051 if (*prev >= OP_SBRA) /* Could match an empty string */
2053 RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2054 RRETURN(rrc);
2056 ecode = prev;
2057 goto TAIL_RECURSE;
2059 else /* OP_KETRMAX */
2061 RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2062 if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2063 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2064 if (*prev == OP_ONCE)
2066 RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2067 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2068 md->once_target = prev;
2069 RRETURN(MATCH_ONCE);
2071 ecode += 1 + LINK_SIZE;
2072 goto TAIL_RECURSE;
2074 /* Control never gets here */
2076 /* Not multiline mode: start of subject assertion, unless notbol. */
2078 case OP_CIRC:
2079 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2081 /* Start of subject assertion */
2083 case OP_SOD:
2084 if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
2085 ecode++;
2086 break;
2088 /* Multiline mode: start of subject unless notbol, or after any newline. */
2090 case OP_CIRCM:
2091 if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2092 if (eptr != md->start_subject &&
2093 (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2094 RRETURN(MATCH_NOMATCH);
2095 ecode++;
2096 break;
2098 /* Start of match assertion */
2100 case OP_SOM:
2101 if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
2102 ecode++;
2103 break;
2105 /* Reset the start of match point */
2107 case OP_SET_SOM:
2108 mstart = eptr;
2109 ecode++;
2110 break;
2112 /* Multiline mode: assert before any newline, or before end of subject
2113 unless noteol is set. */
2115 case OP_DOLLM:
2116 if (eptr < md->end_subject)
2118 if (!IS_NEWLINE(eptr))
2120 if (md->partial != 0 &&
2121 eptr + 1 >= md->end_subject &&
2122 NLBLOCK->nltype == NLTYPE_FIXED &&
2123 NLBLOCK->nllen == 2 &&
2124 UCHAR21TEST(eptr) == NLBLOCK->nl[0])
2126 md->hitend = TRUE;
2127 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2129 RRETURN(MATCH_NOMATCH);
2132 else
2134 if (md->noteol) RRETURN(MATCH_NOMATCH);
2135 SCHECK_PARTIAL();
2137 ecode++;
2138 break;
2140 /* Not multiline mode: assert before a terminating newline or before end of
2141 subject unless noteol is set. */
2143 case OP_DOLL:
2144 if (md->noteol) RRETURN(MATCH_NOMATCH);
2145 if (!md->endonly) goto ASSERT_NL_OR_EOS;
2147 /* ... else fall through for endonly */
2149 /* End of subject assertion (\z) */
2151 case OP_EOD:
2152 if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2153 SCHECK_PARTIAL();
2154 ecode++;
2155 break;
2157 /* End of subject or ending \n assertion (\Z) */
2159 case OP_EODN:
2160 ASSERT_NL_OR_EOS:
2161 if (eptr < md->end_subject &&
2162 (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2164 if (md->partial != 0 &&
2165 eptr + 1 >= md->end_subject &&
2166 NLBLOCK->nltype == NLTYPE_FIXED &&
2167 NLBLOCK->nllen == 2 &&
2168 UCHAR21TEST(eptr) == NLBLOCK->nl[0])
2170 md->hitend = TRUE;
2171 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2173 RRETURN(MATCH_NOMATCH);
2176 /* Either at end of string or \n before end. */
2178 SCHECK_PARTIAL();
2179 ecode++;
2180 break;
2182 /* Word boundary assertions */
2184 case OP_NOT_WORD_BOUNDARY:
2185 case OP_WORD_BOUNDARY:
2188 /* Find out if the previous and current characters are "word" characters.
2189 It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2190 be "non-word" characters. Remember the earliest consulted character for
2191 partial matching. */
2193 #ifdef SUPPORT_UTF
2194 if (utf)
2196 /* Get status of previous character */
2198 if (eptr == md->start_subject) prev_is_word = FALSE; else
2200 PCRE_PUCHAR lastptr = eptr - 1;
2201 BACKCHAR(lastptr);
2202 if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2203 GETCHAR(c, lastptr);
2204 #ifdef SUPPORT_UCP
2205 if (md->use_ucp)
2207 if (c == '_') prev_is_word = TRUE; else
2209 int cat = UCD_CATEGORY(c);
2210 prev_is_word = (cat == ucp_L || cat == ucp_N);
2213 else
2214 #endif
2215 prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2218 /* Get status of next character */
2220 if (eptr >= md->end_subject)
2222 SCHECK_PARTIAL();
2223 cur_is_word = FALSE;
2225 else
2227 GETCHAR(c, eptr);
2228 #ifdef SUPPORT_UCP
2229 if (md->use_ucp)
2231 if (c == '_') cur_is_word = TRUE; else
2233 int cat = UCD_CATEGORY(c);
2234 cur_is_word = (cat == ucp_L || cat == ucp_N);
2237 else
2238 #endif
2239 cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2242 else
2243 #endif
2245 /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2246 consistency with the behaviour of \w we do use it in this case. */
2249 /* Get status of previous character */
2251 if (eptr == md->start_subject) prev_is_word = FALSE; else
2253 if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2254 #ifdef SUPPORT_UCP
2255 if (md->use_ucp)
2257 c = eptr[-1];
2258 if (c == '_') prev_is_word = TRUE; else
2260 int cat = UCD_CATEGORY(c);
2261 prev_is_word = (cat == ucp_L || cat == ucp_N);
2264 else
2265 #endif
2266 prev_is_word = MAX_255(eptr[-1])
2267 && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2270 /* Get status of next character */
2272 if (eptr >= md->end_subject)
2274 SCHECK_PARTIAL();
2275 cur_is_word = FALSE;
2277 else
2278 #ifdef SUPPORT_UCP
2279 if (md->use_ucp)
2281 c = *eptr;
2282 if (c == '_') cur_is_word = TRUE; else
2284 int cat = UCD_CATEGORY(c);
2285 cur_is_word = (cat == ucp_L || cat == ucp_N);
2288 else
2289 #endif
2290 cur_is_word = MAX_255(*eptr)
2291 && ((md->ctypes[*eptr] & ctype_word) != 0);
2294 /* Now see if the situation is what we want */
2296 if ((*ecode++ == OP_WORD_BOUNDARY)?
2297 cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2298 RRETURN(MATCH_NOMATCH);
2300 break;
2302 /* Match any single character type except newline; have to take care with
2303 CRLF newlines and partial matching. */
2305 case OP_ANY:
2306 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2307 if (md->partial != 0 &&
2308 eptr == md->end_subject - 1 &&
2309 NLBLOCK->nltype == NLTYPE_FIXED &&
2310 NLBLOCK->nllen == 2 &&
2311 UCHAR21TEST(eptr) == NLBLOCK->nl[0])
2313 md->hitend = TRUE;
2314 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2317 /* Fall through */
2319 /* Match any single character whatsoever. */
2321 case OP_ALLANY:
2322 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2323 { /* not be updated before SCHECK_PARTIAL. */
2324 SCHECK_PARTIAL();
2325 RRETURN(MATCH_NOMATCH);
2327 eptr++;
2328 #ifdef SUPPORT_UTF
2329 if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2330 #endif
2331 ecode++;
2332 break;
2334 /* Match a single byte, even in UTF-8 mode. This opcode really does match
2335 any byte, even newline, independent of the setting of PCRE_DOTALL. */
2337 case OP_ANYBYTE:
2338 if (eptr >= md->end_subject) /* DO NOT merge the eptr++ here; it must */
2339 { /* not be updated before SCHECK_PARTIAL. */
2340 SCHECK_PARTIAL();
2341 RRETURN(MATCH_NOMATCH);
2343 eptr++;
2344 ecode++;
2345 break;
2347 case OP_NOT_DIGIT:
2348 if (eptr >= md->end_subject)
2350 SCHECK_PARTIAL();
2351 RRETURN(MATCH_NOMATCH);
2353 GETCHARINCTEST(c, eptr);
2354 if (
2355 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2356 c < 256 &&
2357 #endif
2358 (md->ctypes[c] & ctype_digit) != 0
2360 RRETURN(MATCH_NOMATCH);
2361 ecode++;
2362 break;
2364 case OP_DIGIT:
2365 if (eptr >= md->end_subject)
2367 SCHECK_PARTIAL();
2368 RRETURN(MATCH_NOMATCH);
2370 GETCHARINCTEST(c, eptr);
2371 if (
2372 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2373 c > 255 ||
2374 #endif
2375 (md->ctypes[c] & ctype_digit) == 0
2377 RRETURN(MATCH_NOMATCH);
2378 ecode++;
2379 break;
2381 case OP_NOT_WHITESPACE:
2382 if (eptr >= md->end_subject)
2384 SCHECK_PARTIAL();
2385 RRETURN(MATCH_NOMATCH);
2387 GETCHARINCTEST(c, eptr);
2388 if (
2389 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2390 c < 256 &&
2391 #endif
2392 (md->ctypes[c] & ctype_space) != 0
2394 RRETURN(MATCH_NOMATCH);
2395 ecode++;
2396 break;
2398 case OP_WHITESPACE:
2399 if (eptr >= md->end_subject)
2401 SCHECK_PARTIAL();
2402 RRETURN(MATCH_NOMATCH);
2404 GETCHARINCTEST(c, eptr);
2405 if (
2406 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2407 c > 255 ||
2408 #endif
2409 (md->ctypes[c] & ctype_space) == 0
2411 RRETURN(MATCH_NOMATCH);
2412 ecode++;
2413 break;
2415 case OP_NOT_WORDCHAR:
2416 if (eptr >= md->end_subject)
2418 SCHECK_PARTIAL();
2419 RRETURN(MATCH_NOMATCH);
2421 GETCHARINCTEST(c, eptr);
2422 if (
2423 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2424 c < 256 &&
2425 #endif
2426 (md->ctypes[c] & ctype_word) != 0
2428 RRETURN(MATCH_NOMATCH);
2429 ecode++;
2430 break;
2432 case OP_WORDCHAR:
2433 if (eptr >= md->end_subject)
2435 SCHECK_PARTIAL();
2436 RRETURN(MATCH_NOMATCH);
2438 GETCHARINCTEST(c, eptr);
2439 if (
2440 #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2441 c > 255 ||
2442 #endif
2443 (md->ctypes[c] & ctype_word) == 0
2445 RRETURN(MATCH_NOMATCH);
2446 ecode++;
2447 break;
2449 case OP_ANYNL:
2450 if (eptr >= md->end_subject)
2452 SCHECK_PARTIAL();
2453 RRETURN(MATCH_NOMATCH);
2455 GETCHARINCTEST(c, eptr);
2456 switch(c)
2458 default: RRETURN(MATCH_NOMATCH);
2460 case CHAR_CR:
2461 if (eptr >= md->end_subject)
2463 SCHECK_PARTIAL();
2465 else if (UCHAR21TEST(eptr) == CHAR_LF) eptr++;
2466 break;
2468 case CHAR_LF:
2469 break;
2471 case CHAR_VT:
2472 case CHAR_FF:
2473 case CHAR_NEL:
2474 #ifndef EBCDIC
2475 case 0x2028:
2476 case 0x2029:
2477 #endif /* Not EBCDIC */
2478 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2479 break;
2481 ecode++;
2482 break;
2484 case OP_NOT_HSPACE:
2485 if (eptr >= md->end_subject)
2487 SCHECK_PARTIAL();
2488 RRETURN(MATCH_NOMATCH);
2490 GETCHARINCTEST(c, eptr);
2491 switch(c)
2493 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
2494 default: break;
2496 ecode++;
2497 break;
2499 case OP_HSPACE:
2500 if (eptr >= md->end_subject)
2502 SCHECK_PARTIAL();
2503 RRETURN(MATCH_NOMATCH);
2505 GETCHARINCTEST(c, eptr);
2506 switch(c)
2508 HSPACE_CASES: break; /* Byte and multibyte cases */
2509 default: RRETURN(MATCH_NOMATCH);
2511 ecode++;
2512 break;
2514 case OP_NOT_VSPACE:
2515 if (eptr >= md->end_subject)
2517 SCHECK_PARTIAL();
2518 RRETURN(MATCH_NOMATCH);
2520 GETCHARINCTEST(c, eptr);
2521 switch(c)
2523 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2524 default: break;
2526 ecode++;
2527 break;
2529 case OP_VSPACE:
2530 if (eptr >= md->end_subject)
2532 SCHECK_PARTIAL();
2533 RRETURN(MATCH_NOMATCH);
2535 GETCHARINCTEST(c, eptr);
2536 switch(c)
2538 VSPACE_CASES: break;
2539 default: RRETURN(MATCH_NOMATCH);
2541 ecode++;
2542 break;
2544 #ifdef SUPPORT_UCP
2545 /* Check the next character by Unicode property. We will get here only
2546 if the support is in the binary; otherwise a compile-time error occurs. */
2548 case OP_PROP:
2549 case OP_NOTPROP:
2550 if (eptr >= md->end_subject)
2552 SCHECK_PARTIAL();
2553 RRETURN(MATCH_NOMATCH);
2555 GETCHARINCTEST(c, eptr);
2557 const pcre_uint32 *cp;
2558 const ucd_record *prop = GET_UCD(c);
2560 switch(ecode[1])
2562 case PT_ANY:
2563 if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2564 break;
2566 case PT_LAMP:
2567 if ((prop->chartype == ucp_Lu ||
2568 prop->chartype == ucp_Ll ||
2569 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2570 RRETURN(MATCH_NOMATCH);
2571 break;
2573 case PT_GC:
2574 if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2575 RRETURN(MATCH_NOMATCH);
2576 break;
2578 case PT_PC:
2579 if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2580 RRETURN(MATCH_NOMATCH);
2581 break;
2583 case PT_SC:
2584 if ((ecode[2] != prop->script) == (op == OP_PROP))
2585 RRETURN(MATCH_NOMATCH);
2586 break;
2588 /* These are specials */
2590 case PT_ALNUM:
2591 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2592 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2593 RRETURN(MATCH_NOMATCH);
2594 break;
2596 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
2597 which means that Perl space and POSIX space are now identical. PCRE
2598 was changed at release 8.34. */
2600 case PT_SPACE: /* Perl space */
2601 case PT_PXSPACE: /* POSIX space */
2602 switch(c)
2604 HSPACE_CASES:
2605 VSPACE_CASES:
2606 if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2607 break;
2609 default:
2610 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z) ==
2611 (op == OP_NOTPROP)) RRETURN(MATCH_NOMATCH);
2612 break;
2614 break;
2616 case PT_WORD:
2617 if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2618 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2619 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2620 RRETURN(MATCH_NOMATCH);
2621 break;
2623 case PT_CLIST:
2624 cp = PRIV(ucd_caseless_sets) + ecode[2];
2625 for (;;)
2627 if (c < *cp)
2628 { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2629 if (c == *cp++)
2630 { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2632 break;
2634 case PT_UCNC:
2635 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2636 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2637 c >= 0xe000) == (op == OP_NOTPROP))
2638 RRETURN(MATCH_NOMATCH);
2639 break;
2641 /* This should never occur */
2643 default:
2644 RRETURN(PCRE_ERROR_INTERNAL);
2647 ecode += 3;
2649 break;
2651 /* Match an extended Unicode sequence. We will get here only if the support
2652 is in the binary; otherwise a compile-time error occurs. */
2654 case OP_EXTUNI:
2655 if (eptr >= md->end_subject)
2657 SCHECK_PARTIAL();
2658 RRETURN(MATCH_NOMATCH);
2660 else
2662 int lgb, rgb;
2663 GETCHARINCTEST(c, eptr);
2664 lgb = UCD_GRAPHBREAK(c);
2665 while (eptr < md->end_subject)
2667 int len = 1;
2668 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2669 rgb = UCD_GRAPHBREAK(c);
2670 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2671 lgb = rgb;
2672 eptr += len;
2675 CHECK_PARTIAL();
2676 ecode++;
2677 break;
2678 #endif /* SUPPORT_UCP */
2681 /* Match a back reference, possibly repeatedly. Look past the end of the
2682 item to see if there is repeat information following. The code is similar
2683 to that for character classes, but repeated for efficiency. Then obey
2684 similar code to character type repeats - written out again for speed.
2685 However, if the referenced string is the empty string, always treat
2686 it as matched, any number of times (otherwise there could be infinite
2687 loops). If the reference is unset, there are two possibilities:
2689 (a) In the default, Perl-compatible state, set the length negative;
2690 this ensures that every attempt at a match fails. We can't just fail
2691 here, because of the possibility of quantifiers with zero minima.
2693 (b) If the JavaScript compatibility flag is set, set the length to zero
2694 so that the back reference matches an empty string.
2696 Otherwise, set the length to the length of what was matched by the
2697 referenced subpattern.
2699 The OP_REF and OP_REFI opcodes are used for a reference to a numbered group
2700 or to a non-duplicated named group. For a duplicated named group, OP_DNREF
2701 and OP_DNREFI are used. In this case we must scan the list of groups to
2702 which the name refers, and use the first one that is set. */
2704 case OP_DNREF:
2705 case OP_DNREFI:
2706 caseless = op == OP_DNREFI;
2708 int count = GET2(ecode, 1+IMM2_SIZE);
2709 pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
2710 ecode += 1 + 2*IMM2_SIZE;
2712 /* Setting the default length first and initializing 'offset' avoids
2713 compiler warnings in the REF_REPEAT code. */
2715 length = (md->jscript_compat)? 0 : -1;
2716 offset = 0;
2718 while (count-- > 0)
2720 offset = GET2(slot, 0) << 1;
2721 if (offset < offset_top && md->offset_vector[offset] >= 0)
2723 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2724 break;
2726 slot += md->name_entry_size;
2729 goto REF_REPEAT;
2731 case OP_REF:
2732 case OP_REFI:
2733 caseless = op == OP_REFI;
2734 offset = GET2(ecode, 1) << 1; /* Doubled ref number */
2735 ecode += 1 + IMM2_SIZE;
2736 if (offset >= offset_top || md->offset_vector[offset] < 0)
2737 length = (md->jscript_compat)? 0 : -1;
2738 else
2739 length = md->offset_vector[offset+1] - md->offset_vector[offset];
2741 /* Set up for repetition, or handle the non-repeated case */
2743 REF_REPEAT:
2744 switch (*ecode)
2746 case OP_CRSTAR:
2747 case OP_CRMINSTAR:
2748 case OP_CRPLUS:
2749 case OP_CRMINPLUS:
2750 case OP_CRQUERY:
2751 case OP_CRMINQUERY:
2752 c = *ecode++ - OP_CRSTAR;
2753 minimize = (c & 1) != 0;
2754 min = rep_min[c]; /* Pick up values from tables; */
2755 max = rep_max[c]; /* zero for max => infinity */
2756 if (max == 0) max = INT_MAX;
2757 break;
2759 case OP_CRRANGE:
2760 case OP_CRMINRANGE:
2761 minimize = (*ecode == OP_CRMINRANGE);
2762 min = GET2(ecode, 1);
2763 max = GET2(ecode, 1 + IMM2_SIZE);
2764 if (max == 0) max = INT_MAX;
2765 ecode += 1 + 2 * IMM2_SIZE;
2766 break;
2768 default: /* No repeat follows */
2769 if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2771 if (length == -2) eptr = md->end_subject; /* Partial match */
2772 CHECK_PARTIAL();
2773 RRETURN(MATCH_NOMATCH);
2775 eptr += length;
2776 continue; /* With the main loop */
2779 /* Handle repeated back references. If the length of the reference is
2780 zero, just continue with the main loop. If the length is negative, it
2781 means the reference is unset in non-Java-compatible mode. If the minimum is
2782 zero, we can continue at the same level without recursion. For any other
2783 minimum, carrying on will result in NOMATCH. */
2785 if (length == 0) continue;
2786 if (length < 0 && min == 0) continue;
2788 /* First, ensure the minimum number of matches are present. We get back
2789 the length of the reference string explicitly rather than passing the
2790 address of eptr, so that eptr can be a register variable. */
2792 for (i = 1; i <= min; i++)
2794 int slength;
2795 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2797 if (slength == -2) eptr = md->end_subject; /* Partial match */
2798 CHECK_PARTIAL();
2799 RRETURN(MATCH_NOMATCH);
2801 eptr += slength;
2804 /* If min = max, continue at the same level without recursion.
2805 They are not both allowed to be zero. */
2807 if (min == max) continue;
2809 /* If minimizing, keep trying and advancing the pointer */
2811 if (minimize)
2813 for (fi = min;; fi++)
2815 int slength;
2816 RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2817 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2818 if (fi >= max) RRETURN(MATCH_NOMATCH);
2819 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2821 if (slength == -2) eptr = md->end_subject; /* Partial match */
2822 CHECK_PARTIAL();
2823 RRETURN(MATCH_NOMATCH);
2825 eptr += slength;
2827 /* Control never gets here */
2830 /* If maximizing, find the longest string and work backwards */
2832 else
2834 pp = eptr;
2835 for (i = min; i < max; i++)
2837 int slength;
2838 if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2840 /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2841 the soft partial matching case. */
2843 if (slength == -2 && md->partial != 0 &&
2844 md->end_subject > md->start_used_ptr)
2846 md->hitend = TRUE;
2847 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2849 break;
2851 eptr += slength;
2854 while (eptr >= pp)
2856 RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2857 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2858 eptr -= length;
2860 RRETURN(MATCH_NOMATCH);
2862 /* Control never gets here */
2864 /* Match a bit-mapped character class, possibly repeatedly. This op code is
2865 used when all the characters in the class have values in the range 0-255,
2866 and either the matching is caseful, or the characters are in the range
2867 0-127 when UTF-8 processing is enabled. The only difference between
2868 OP_CLASS and OP_NCLASS occurs when a data character outside the range is
2869 encountered.
2871 First, look past the end of the item to see if there is repeat information
2872 following. Then obey similar code to character type repeats - written out
2873 again for speed. */
2875 case OP_NCLASS:
2876 case OP_CLASS:
2878 /* The data variable is saved across frames, so the byte map needs to
2879 be stored there. */
2880 #define BYTE_MAP ((pcre_uint8 *)data)
2881 data = ecode + 1; /* Save for matching */
2882 ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2884 switch (*ecode)
2886 case OP_CRSTAR:
2887 case OP_CRMINSTAR:
2888 case OP_CRPLUS:
2889 case OP_CRMINPLUS:
2890 case OP_CRQUERY:
2891 case OP_CRMINQUERY:
2892 case OP_CRPOSSTAR:
2893 case OP_CRPOSPLUS:
2894 case OP_CRPOSQUERY:
2895 c = *ecode++ - OP_CRSTAR;
2896 if (c < OP_CRPOSSTAR - OP_CRSTAR) minimize = (c & 1) != 0;
2897 else possessive = TRUE;
2898 min = rep_min[c]; /* Pick up values from tables; */
2899 max = rep_max[c]; /* zero for max => infinity */
2900 if (max == 0) max = INT_MAX;
2901 break;
2903 case OP_CRRANGE:
2904 case OP_CRMINRANGE:
2905 case OP_CRPOSRANGE:
2906 minimize = (*ecode == OP_CRMINRANGE);
2907 possessive = (*ecode == OP_CRPOSRANGE);
2908 min = GET2(ecode, 1);
2909 max = GET2(ecode, 1 + IMM2_SIZE);
2910 if (max == 0) max = INT_MAX;
2911 ecode += 1 + 2 * IMM2_SIZE;
2912 break;
2914 default: /* No repeat follows */
2915 min = max = 1;
2916 break;
2919 /* First, ensure the minimum number of matches are present. */
2921 #ifdef SUPPORT_UTF
2922 if (utf)
2924 for (i = 1; i <= min; i++)
2926 if (eptr >= md->end_subject)
2928 SCHECK_PARTIAL();
2929 RRETURN(MATCH_NOMATCH);
2931 GETCHARINC(c, eptr);
2932 if (c > 255)
2934 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2936 else
2937 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2940 else
2941 #endif
2942 /* Not UTF mode */
2944 for (i = 1; i <= min; i++)
2946 if (eptr >= md->end_subject)
2948 SCHECK_PARTIAL();
2949 RRETURN(MATCH_NOMATCH);
2951 c = *eptr++;
2952 #ifndef COMPILE_PCRE8
2953 if (c > 255)
2955 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2957 else
2958 #endif
2959 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2963 /* If max == min we can continue with the main loop without the
2964 need to recurse. */
2966 if (min == max) continue;
2968 /* If minimizing, keep testing the rest of the expression and advancing
2969 the pointer while it matches the class. */
2971 if (minimize)
2973 #ifdef SUPPORT_UTF
2974 if (utf)
2976 for (fi = min;; fi++)
2978 RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2979 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2980 if (fi >= max) RRETURN(MATCH_NOMATCH);
2981 if (eptr >= md->end_subject)
2983 SCHECK_PARTIAL();
2984 RRETURN(MATCH_NOMATCH);
2986 GETCHARINC(c, eptr);
2987 if (c > 255)
2989 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2991 else
2992 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2995 else
2996 #endif
2997 /* Not UTF mode */
2999 for (fi = min;; fi++)
3001 RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
3002 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3003 if (fi >= max) RRETURN(MATCH_NOMATCH);
3004 if (eptr >= md->end_subject)
3006 SCHECK_PARTIAL();
3007 RRETURN(MATCH_NOMATCH);
3009 c = *eptr++;
3010 #ifndef COMPILE_PCRE8
3011 if (c > 255)
3013 if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3015 else
3016 #endif
3017 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3020 /* Control never gets here */
3023 /* If maximizing, find the longest possible run, then work backwards. */
3025 else
3027 pp = eptr;
3029 #ifdef SUPPORT_UTF
3030 if (utf)
3032 for (i = min; i < max; i++)
3034 int len = 1;
3035 if (eptr >= md->end_subject)
3037 SCHECK_PARTIAL();
3038 break;
3040 GETCHARLEN(c, eptr, len);
3041 if (c > 255)
3043 if (op == OP_CLASS) break;
3045 else
3046 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3047 eptr += len;
3050 if (possessive) continue; /* No backtracking */
3052 for (;;)
3054 RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3055 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3056 if (eptr-- <= pp) break; /* Stop if tried at original pos */
3057 BACKCHAR(eptr);
3060 else
3061 #endif
3062 /* Not UTF mode */
3064 for (i = min; i < max; i++)
3066 if (eptr >= md->end_subject)
3068 SCHECK_PARTIAL();
3069 break;
3071 c = *eptr;
3072 #ifndef COMPILE_PCRE8
3073 if (c > 255)
3075 if (op == OP_CLASS) break;
3077 else
3078 #endif
3079 if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3080 eptr++;
3083 if (possessive) continue; /* No backtracking */
3085 while (eptr >= pp)
3087 RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3088 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3089 eptr--;
3093 RRETURN(MATCH_NOMATCH);
3095 #undef BYTE_MAP
3097 /* Control never gets here */
3100 /* Match an extended character class. In the 8-bit library, this opcode is
3101 encountered only when UTF-8 mode mode is supported. In the 16-bit and
3102 32-bit libraries, codepoints greater than 255 may be encountered even when
3103 UTF is not supported. */
3105 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3106 case OP_XCLASS:
3108 data = ecode + 1 + LINK_SIZE; /* Save for matching */
3109 ecode += GET(ecode, 1); /* Advance past the item */
3111 switch (*ecode)
3113 case OP_CRSTAR:
3114 case OP_CRMINSTAR:
3115 case OP_CRPLUS:
3116 case OP_CRMINPLUS:
3117 case OP_CRQUERY:
3118 case OP_CRMINQUERY:
3119 case OP_CRPOSSTAR:
3120 case OP_CRPOSPLUS:
3121 case OP_CRPOSQUERY:
3122 c = *ecode++ - OP_CRSTAR;
3123 if (c < OP_CRPOSSTAR - OP_CRSTAR) minimize = (c & 1) != 0;
3124 else possessive = TRUE;
3125 min = rep_min[c]; /* Pick up values from tables; */
3126 max = rep_max[c]; /* zero for max => infinity */
3127 if (max == 0) max = INT_MAX;
3128 break;
3130 case OP_CRRANGE:
3131 case OP_CRMINRANGE:
3132 case OP_CRPOSRANGE:
3133 minimize = (*ecode == OP_CRMINRANGE);
3134 possessive = (*ecode == OP_CRPOSRANGE);
3135 min = GET2(ecode, 1);
3136 max = GET2(ecode, 1 + IMM2_SIZE);
3137 if (max == 0) max = INT_MAX;
3138 ecode += 1 + 2 * IMM2_SIZE;
3139 break;
3141 default: /* No repeat follows */
3142 min = max = 1;
3143 break;
3146 /* First, ensure the minimum number of matches are present. */
3148 for (i = 1; i <= min; i++)
3150 if (eptr >= md->end_subject)
3152 SCHECK_PARTIAL();
3153 RRETURN(MATCH_NOMATCH);
3155 GETCHARINCTEST(c, eptr);
3156 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3159 /* If max == min we can continue with the main loop without the
3160 need to recurse. */
3162 if (min == max) continue;
3164 /* If minimizing, keep testing the rest of the expression and advancing
3165 the pointer while it matches the class. */
3167 if (minimize)
3169 for (fi = min;; fi++)
3171 RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3172 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3173 if (fi >= max) RRETURN(MATCH_NOMATCH);
3174 if (eptr >= md->end_subject)
3176 SCHECK_PARTIAL();
3177 RRETURN(MATCH_NOMATCH);
3179 GETCHARINCTEST(c, eptr);
3180 if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3182 /* Control never gets here */
3185 /* If maximizing, find the longest possible run, then work backwards. */
3187 else
3189 pp = eptr;
3190 for (i = min; i < max; i++)
3192 int len = 1;
3193 if (eptr >= md->end_subject)
3195 SCHECK_PARTIAL();
3196 break;
3198 #ifdef SUPPORT_UTF
3199 GETCHARLENTEST(c, eptr, len);
3200 #else
3201 c = *eptr;
3202 #endif
3203 if (!PRIV(xclass)(c, data, utf)) break;
3204 eptr += len;
3207 if (possessive) continue; /* No backtracking */
3209 for(;;)
3211 RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3212 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3213 if (eptr-- <= pp) break; /* Stop if tried at original pos */
3214 #ifdef SUPPORT_UTF
3215 if (utf) BACKCHAR(eptr);
3216 #endif
3218 RRETURN(MATCH_NOMATCH);
3221 /* Control never gets here */
3223 #endif /* End of XCLASS */
3225 /* Match a single character, casefully */
3227 case OP_CHAR:
3228 #ifdef SUPPORT_UTF
3229 if (utf)
3231 length = 1;
3232 ecode++;
3233 GETCHARLEN(fc, ecode, length);
3234 if (length > md->end_subject - eptr)
3236 CHECK_PARTIAL(); /* Not SCHECK_PARTIAL() */
3237 RRETURN(MATCH_NOMATCH);
3239 while (length-- > 0) if (*ecode++ != UCHAR21INC(eptr)) RRETURN(MATCH_NOMATCH);
3241 else
3242 #endif
3243 /* Not UTF mode */
3245 if (md->end_subject - eptr < 1)
3247 SCHECK_PARTIAL(); /* This one can use SCHECK_PARTIAL() */
3248 RRETURN(MATCH_NOMATCH);
3250 if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3251 ecode += 2;
3253 break;
3255 /* Match a single character, caselessly. If we are at the end of the
3256 subject, give up immediately. */
3258 case OP_CHARI:
3259 if (eptr >= md->end_subject)
3261 SCHECK_PARTIAL();
3262 RRETURN(MATCH_NOMATCH);
3265 #ifdef SUPPORT_UTF
3266 if (utf)
3268 length = 1;
3269 ecode++;
3270 GETCHARLEN(fc, ecode, length);
3272 /* If the pattern character's value is < 128, we have only one byte, and
3273 we know that its other case must also be one byte long, so we can use the
3274 fast lookup table. We know that there is at least one byte left in the
3275 subject. */
3277 if (fc < 128)
3279 pcre_uint32 cc = UCHAR21(eptr);
3280 if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3281 ecode++;
3282 eptr++;
3285 /* Otherwise we must pick up the subject character. Note that we cannot
3286 use the value of "length" to check for sufficient bytes left, because the
3287 other case of the character may have more or fewer bytes. */
3289 else
3291 pcre_uint32 dc;
3292 GETCHARINC(dc, eptr);
3293 ecode += length;
3295 /* If we have Unicode property support, we can use it to test the other
3296 case of the character, if there is one. */
3298 if (fc != dc)
3300 #ifdef SUPPORT_UCP
3301 if (dc != UCD_OTHERCASE(fc))
3302 #endif
3303 RRETURN(MATCH_NOMATCH);
3307 else
3308 #endif /* SUPPORT_UTF */
3310 /* Not UTF mode */
3312 if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3313 != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3314 eptr++;
3315 ecode += 2;
3317 break;
3319 /* Match a single character repeatedly. */
3321 case OP_EXACT:
3322 case OP_EXACTI:
3323 min = max = GET2(ecode, 1);
3324 ecode += 1 + IMM2_SIZE;
3325 goto REPEATCHAR;
3327 case OP_POSUPTO:
3328 case OP_POSUPTOI:
3329 possessive = TRUE;
3330 /* Fall through */
3332 case OP_UPTO:
3333 case OP_UPTOI:
3334 case OP_MINUPTO:
3335 case OP_MINUPTOI:
3336 min = 0;
3337 max = GET2(ecode, 1);
3338 minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3339 ecode += 1 + IMM2_SIZE;
3340 goto REPEATCHAR;
3342 case OP_POSSTAR:
3343 case OP_POSSTARI:
3344 possessive = TRUE;
3345 min = 0;
3346 max = INT_MAX;
3347 ecode++;
3348 goto REPEATCHAR;
3350 case OP_POSPLUS:
3351 case OP_POSPLUSI:
3352 possessive = TRUE;
3353 min = 1;
3354 max = INT_MAX;
3355 ecode++;
3356 goto REPEATCHAR;
3358 case OP_POSQUERY:
3359 case OP_POSQUERYI:
3360 possessive = TRUE;
3361 min = 0;
3362 max = 1;
3363 ecode++;
3364 goto REPEATCHAR;
3366 case OP_STAR:
3367 case OP_STARI:
3368 case OP_MINSTAR:
3369 case OP_MINSTARI:
3370 case OP_PLUS:
3371 case OP_PLUSI:
3372 case OP_MINPLUS:
3373 case OP_MINPLUSI:
3374 case OP_QUERY:
3375 case OP_QUERYI:
3376 case OP_MINQUERY:
3377 case OP_MINQUERYI:
3378 c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3379 minimize = (c & 1) != 0;
3380 min = rep_min[c]; /* Pick up values from tables; */
3381 max = rep_max[c]; /* zero for max => infinity */
3382 if (max == 0) max = INT_MAX;
3384 /* Common code for all repeated single-character matches. We first check
3385 for the minimum number of characters. If the minimum equals the maximum, we
3386 are done. Otherwise, if minimizing, check the rest of the pattern for a
3387 match; if there isn't one, advance up to the maximum, one character at a
3388 time.
3390 If maximizing, advance up to the maximum number of matching characters,
3391 until eptr is past the end of the maximum run. If possessive, we are
3392 then done (no backing up). Otherwise, match at this position; anything
3393 other than no match is immediately returned. For nomatch, back up one
3394 character, unless we are matching \R and the last thing matched was
3395 \r\n, in which case, back up two bytes. When we reach the first optional
3396 character position, we can save stack by doing a tail recurse.
3398 The various UTF/non-UTF and caseful/caseless cases are handled separately,
3399 for speed. */
3401 REPEATCHAR:
3402 #ifdef SUPPORT_UTF
3403 if (utf)
3405 length = 1;
3406 charptr = ecode;
3407 GETCHARLEN(fc, ecode, length);
3408 ecode += length;
3410 /* Handle multibyte character matching specially here. There is
3411 support for caseless matching if UCP support is present. */
3413 if (length > 1)
3415 #ifdef SUPPORT_UCP
3416 pcre_uint32 othercase;
3417 if (op >= OP_STARI && /* Caseless */
3418 (othercase = UCD_OTHERCASE(fc)) != fc)
3419 oclength = PRIV(ord2utf)(othercase, occhars);
3420 else oclength = 0;
3421 #endif /* SUPPORT_UCP */
3423 for (i = 1; i <= min; i++)
3425 if (eptr <= md->end_subject - length &&
3426 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3427 #ifdef SUPPORT_UCP
3428 else if (oclength > 0 &&
3429 eptr <= md->end_subject - oclength &&
3430 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3431 #endif /* SUPPORT_UCP */
3432 else
3434 CHECK_PARTIAL();
3435 RRETURN(MATCH_NOMATCH);
3439 if (min == max) continue;
3441 if (minimize)
3443 for (fi = min;; fi++)
3445 RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3446 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3447 if (fi >= max) RRETURN(MATCH_NOMATCH);
3448 if (eptr <= md->end_subject - length &&
3449 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3450 #ifdef SUPPORT_UCP
3451 else if (oclength > 0 &&
3452 eptr <= md->end_subject - oclength &&
3453 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3454 #endif /* SUPPORT_UCP */
3455 else
3457 CHECK_PARTIAL();
3458 RRETURN(MATCH_NOMATCH);
3461 /* Control never gets here */
3464 else /* Maximize */
3466 pp = eptr;
3467 for (i = min; i < max; i++)
3469 if (eptr <= md->end_subject - length &&
3470 memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3471 #ifdef SUPPORT_UCP
3472 else if (oclength > 0 &&
3473 eptr <= md->end_subject - oclength &&
3474 memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3475 #endif /* SUPPORT_UCP */
3476 else
3478 CHECK_PARTIAL();
3479 break;
3483 if (possessive) continue; /* No backtracking */
3484 for(;;)
3486 if (eptr <= pp) goto TAIL_RECURSE;
3487 RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3488 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3489 #ifdef SUPPORT_UCP
3490 eptr--;
3491 BACKCHAR(eptr);
3492 #else /* without SUPPORT_UCP */
3493 eptr -= length;
3494 #endif /* SUPPORT_UCP */
3497 /* Control never gets here */
3500 /* If the length of a UTF-8 character is 1, we fall through here, and
3501 obey the code as for non-UTF-8 characters below, though in this case the
3502 value of fc will always be < 128. */
3504 else
3505 #endif /* SUPPORT_UTF */
3506 /* When not in UTF-8 mode, load a single-byte character. */
3507 fc = *ecode++;
3509 /* The value of fc at this point is always one character, though we may
3510 or may not be in UTF mode. The code is duplicated for the caseless and
3511 caseful cases, for speed, since matching characters is likely to be quite
3512 common. First, ensure the minimum number of matches are present. If min =
3513 max, continue at the same level without recursing. Otherwise, if
3514 minimizing, keep trying the rest of the expression and advancing one
3515 matching character if failing, up to the maximum. Alternatively, if
3516 maximizing, find the maximum number of characters and work backwards. */
3518 DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3519 max, (char *)eptr));
3521 if (op >= OP_STARI) /* Caseless */
3523 #ifdef COMPILE_PCRE8
3524 /* fc must be < 128 if UTF is enabled. */
3525 foc = md->fcc[fc];
3526 #else
3527 #ifdef SUPPORT_UTF
3528 #ifdef SUPPORT_UCP
3529 if (utf && fc > 127)
3530 foc = UCD_OTHERCASE(fc);
3531 #else
3532 if (utf && fc > 127)
3533 foc = fc;
3534 #endif /* SUPPORT_UCP */
3535 else
3536 #endif /* SUPPORT_UTF */
3537 foc = TABLE_GET(fc, md->fcc, fc);
3538 #endif /* COMPILE_PCRE8 */
3540 for (i = 1; i <= min; i++)
3542 pcre_uint32 cc; /* Faster than pcre_uchar */
3543 if (eptr >= md->end_subject)
3545 SCHECK_PARTIAL();
3546 RRETURN(MATCH_NOMATCH);
3548 cc = UCHAR21TEST(eptr);
3549 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3550 eptr++;
3552 if (min == max) continue;
3553 if (minimize)
3555 for (fi = min;; fi++)
3557 pcre_uint32 cc; /* Faster than pcre_uchar */
3558 RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3559 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3560 if (fi >= max) RRETURN(MATCH_NOMATCH);
3561 if (eptr >= md->end_subject)
3563 SCHECK_PARTIAL();
3564 RRETURN(MATCH_NOMATCH);
3566 cc = UCHAR21TEST(eptr);
3567 if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3568 eptr++;
3570 /* Control never gets here */
3572 else /* Maximize */
3574 pp = eptr;
3575 for (i = min; i < max; i++)
3577 pcre_uint32 cc; /* Faster than pcre_uchar */
3578 if (eptr >= md->end_subject)
3580 SCHECK_PARTIAL();
3581 break;
3583 cc = UCHAR21TEST(eptr);
3584 if (fc != cc && foc != cc) break;
3585 eptr++;
3587 if (possessive) continue; /* No backtracking */
3588 for (;;)
3590 if (eptr == pp) goto TAIL_RECURSE;
3591 RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3592 eptr--;
3593 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3595 /* Control never gets here */
3599 /* Caseful comparisons (includes all multi-byte characters) */
3601 else
3603 for (i = 1; i <= min; i++)
3605 if (eptr >= md->end_subject)
3607 SCHECK_PARTIAL();
3608 RRETURN(MATCH_NOMATCH);
3610 if (fc != UCHAR21INCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3613 if (min == max) continue;
3615 if (minimize)
3617 for (fi = min;; fi++)
3619 RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3620 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3621 if (fi >= max) RRETURN(MATCH_NOMATCH);
3622 if (eptr >= md->end_subject)
3624 SCHECK_PARTIAL();
3625 RRETURN(MATCH_NOMATCH);
3627 if (fc != UCHAR21INCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3629 /* Control never gets here */
3631 else /* Maximize */
3633 pp = eptr;
3634 for (i = min; i < max; i++)
3636 if (eptr >= md->end_subject)
3638 SCHECK_PARTIAL();
3639 break;
3641 if (fc != UCHAR21TEST(eptr)) break;
3642 eptr++;
3644 if (possessive) continue; /* No backtracking */
3645 for (;;)
3647 if (eptr == pp) goto TAIL_RECURSE;
3648 RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3649 eptr--;
3650 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3652 /* Control never gets here */
3655 /* Control never gets here */
3657 /* Match a negated single one-byte character. The character we are
3658 checking can be multibyte. */
3660 case OP_NOT:
3661 case OP_NOTI:
3662 if (eptr >= md->end_subject)
3664 SCHECK_PARTIAL();
3665 RRETURN(MATCH_NOMATCH);
3667 #ifdef SUPPORT_UTF
3668 if (utf)
3670 register pcre_uint32 ch, och;
3672 ecode++;
3673 GETCHARINC(ch, ecode);
3674 GETCHARINC(c, eptr);
3676 if (op == OP_NOT)
3678 if (ch == c) RRETURN(MATCH_NOMATCH);
3680 else
3682 #ifdef SUPPORT_UCP
3683 if (ch > 127)
3684 och = UCD_OTHERCASE(ch);
3685 #else
3686 if (ch > 127)
3687 och = ch;
3688 #endif /* SUPPORT_UCP */
3689 else
3690 och = TABLE_GET(ch, md->fcc, ch);
3691 if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3694 else
3695 #endif
3697 register pcre_uint32 ch = ecode[1];
3698 c = *eptr++;
3699 if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3700 RRETURN(MATCH_NOMATCH);
3701 ecode += 2;
3703 break;
3705 /* Match a negated single one-byte character repeatedly. This is almost a
3706 repeat of the code for a repeated single character, but I haven't found a
3707 nice way of commoning these up that doesn't require a test of the
3708 positive/negative option for each character match. Maybe that wouldn't add
3709 very much to the time taken, but character matching *is* what this is all
3710 about... */
3712 case OP_NOTEXACT:
3713 case OP_NOTEXACTI:
3714 min = max = GET2(ecode, 1);
3715 ecode += 1 + IMM2_SIZE;
3716 goto REPEATNOTCHAR;
3718 case OP_NOTUPTO:
3719 case OP_NOTUPTOI:
3720 case OP_NOTMINUPTO:
3721 case OP_NOTMINUPTOI:
3722 min = 0;
3723 max = GET2(ecode, 1);
3724 minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3725 ecode += 1 + IMM2_SIZE;
3726 goto REPEATNOTCHAR;
3728 case OP_NOTPOSSTAR:
3729 case OP_NOTPOSSTARI:
3730 possessive = TRUE;
3731 min = 0;
3732 max = INT_MAX;
3733 ecode++;
3734 goto REPEATNOTCHAR;
3736 case OP_NOTPOSPLUS:
3737 case OP_NOTPOSPLUSI:
3738 possessive = TRUE;
3739 min = 1;
3740 max = INT_MAX;
3741 ecode++;
3742 goto REPEATNOTCHAR;
3744 case OP_NOTPOSQUERY:
3745 case OP_NOTPOSQUERYI:
3746 possessive = TRUE;
3747 min = 0;
3748 max = 1;
3749 ecode++;
3750 goto REPEATNOTCHAR;
3752 case OP_NOTPOSUPTO:
3753 case OP_NOTPOSUPTOI:
3754 possessive = TRUE;
3755 min = 0;
3756 max = GET2(ecode, 1);
3757 ecode += 1 + IMM2_SIZE;
3758 goto REPEATNOTCHAR;
3760 case OP_NOTSTAR:
3761 case OP_NOTSTARI:
3762 case OP_NOTMINSTAR:
3763 case OP_NOTMINSTARI:
3764 case OP_NOTPLUS:
3765 case OP_NOTPLUSI:
3766 case OP_NOTMINPLUS:
3767 case OP_NOTMINPLUSI:
3768 case OP_NOTQUERY:
3769 case OP_NOTQUERYI:
3770 case OP_NOTMINQUERY:
3771 case OP_NOTMINQUERYI:
3772 c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3773 minimize = (c & 1) != 0;
3774 min = rep_min[c]; /* Pick up values from tables; */
3775 max = rep_max[c]; /* zero for max => infinity */
3776 if (max == 0) max = INT_MAX;
3778 /* Common code for all repeated single-byte matches. */
3780 REPEATNOTCHAR:
3781 GETCHARINCTEST(fc, ecode);
3783 /* The code is duplicated for the caseless and caseful cases, for speed,
3784 since matching characters is likely to be quite common. First, ensure the
3785 minimum number of matches are present. If min = max, continue at the same
3786 level without recursing. Otherwise, if minimizing, keep trying the rest of
3787 the expression and advancing one matching character if failing, up to the
3788 maximum. Alternatively, if maximizing, find the maximum number of
3789 characters and work backwards. */
3791 DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3792 max, (char *)eptr));
3794 if (op >= OP_NOTSTARI) /* Caseless */
3796 #ifdef SUPPORT_UTF
3797 #ifdef SUPPORT_UCP
3798 if (utf && fc > 127)
3799 foc = UCD_OTHERCASE(fc);
3800 #else
3801 if (utf && fc > 127)
3802 foc = fc;
3803 #endif /* SUPPORT_UCP */
3804 else
3805 #endif /* SUPPORT_UTF */
3806 foc = TABLE_GET(fc, md->fcc, fc);
3808 #ifdef SUPPORT_UTF
3809 if (utf)
3811 register pcre_uint32 d;
3812 for (i = 1; i <= min; i++)
3814 if (eptr >= md->end_subject)
3816 SCHECK_PARTIAL();
3817 RRETURN(MATCH_NOMATCH);
3819 GETCHARINC(d, eptr);
3820 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3823 else
3824 #endif /* SUPPORT_UTF */
3825 /* Not UTF mode */
3827 for (i = 1; i <= min; i++)
3829 if (eptr >= md->end_subject)
3831 SCHECK_PARTIAL();
3832 RRETURN(MATCH_NOMATCH);
3834 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3835 eptr++;
3839 if (min == max) continue;
3841 if (minimize)
3843 #ifdef SUPPORT_UTF
3844 if (utf)
3846 register pcre_uint32 d;
3847 for (fi = min;; fi++)
3849 RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3850 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3851 if (fi >= max) RRETURN(MATCH_NOMATCH);
3852 if (eptr >= md->end_subject)
3854 SCHECK_PARTIAL();
3855 RRETURN(MATCH_NOMATCH);
3857 GETCHARINC(d, eptr);
3858 if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3861 else
3862 #endif /*SUPPORT_UTF */
3863 /* Not UTF mode */
3865 for (fi = min;; fi++)
3867 RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3868 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3869 if (fi >= max) RRETURN(MATCH_NOMATCH);
3870 if (eptr >= md->end_subject)
3872 SCHECK_PARTIAL();
3873 RRETURN(MATCH_NOMATCH);
3875 if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3876 eptr++;
3879 /* Control never gets here */
3882 /* Maximize case */
3884 else
3886 pp = eptr;
3888 #ifdef SUPPORT_UTF
3889 if (utf)
3891 register pcre_uint32 d;
3892 for (i = min; i < max; i++)
3894 int len = 1;
3895 if (eptr >= md->end_subject)
3897 SCHECK_PARTIAL();
3898 break;
3900 GETCHARLEN(d, eptr, len);
3901 if (fc == d || (unsigned int)foc == d) break;
3902 eptr += len;
3904 if (possessive) continue; /* No backtracking */
3905 for(;;)
3907 if (eptr <= pp) goto TAIL_RECURSE;
3908 RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3909 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3910 eptr--;
3911 BACKCHAR(eptr);
3914 else
3915 #endif /* SUPPORT_UTF */
3916 /* Not UTF mode */
3918 for (i = min; i < max; i++)
3920 if (eptr >= md->end_subject)
3922 SCHECK_PARTIAL();
3923 break;
3925 if (fc == *eptr || foc == *eptr) break;
3926 eptr++;
3928 if (possessive) continue; /* No backtracking */
3929 for (;;)
3931 if (eptr == pp) goto TAIL_RECURSE;
3932 RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3933 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3934 eptr--;
3937 /* Control never gets here */
3941 /* Caseful comparisons */
3943 else
3945 #ifdef SUPPORT_UTF
3946 if (utf)
3948 register pcre_uint32 d;
3949 for (i = 1; i <= min; i++)
3951 if (eptr >= md->end_subject)
3953 SCHECK_PARTIAL();
3954 RRETURN(MATCH_NOMATCH);
3956 GETCHARINC(d, eptr);
3957 if (fc == d) RRETURN(MATCH_NOMATCH);
3960 else
3961 #endif
3962 /* Not UTF mode */
3964 for (i = 1; i <= min; i++)
3966 if (eptr >= md->end_subject)
3968 SCHECK_PARTIAL();
3969 RRETURN(MATCH_NOMATCH);
3971 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3975 if (min == max) continue;
3977 if (minimize)
3979 #ifdef SUPPORT_UTF
3980 if (utf)
3982 register pcre_uint32 d;
3983 for (fi = min;; fi++)
3985 RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3986 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3987 if (fi >= max) RRETURN(MATCH_NOMATCH);
3988 if (eptr >= md->end_subject)
3990 SCHECK_PARTIAL();
3991 RRETURN(MATCH_NOMATCH);
3993 GETCHARINC(d, eptr);
3994 if (fc == d) RRETURN(MATCH_NOMATCH);
3997 else
3998 #endif
3999 /* Not UTF mode */
4001 for (fi = min;; fi++)
4003 RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
4004 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4005 if (fi >= max) RRETURN(MATCH_NOMATCH);
4006 if (eptr >= md->end_subject)
4008 SCHECK_PARTIAL();
4009 RRETURN(MATCH_NOMATCH);
4011 if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
4014 /* Control never gets here */
4017 /* Maximize case */
4019 else
4021 pp = eptr;
4023 #ifdef SUPPORT_UTF
4024 if (utf)
4026 register pcre_uint32 d;
4027 for (i = min; i < max; i++)
4029 int len = 1;
4030 if (eptr >= md->end_subject)
4032 SCHECK_PARTIAL();
4033 break;
4035 GETCHARLEN(d, eptr, len);
4036 if (fc == d) break;
4037 eptr += len;
4039 if (possessive) continue; /* No backtracking */
4040 for(;;)
4042 if (eptr <= pp) goto TAIL_RECURSE;
4043 RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
4044 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4045 eptr--;
4046 BACKCHAR(eptr);
4049 else
4050 #endif
4051 /* Not UTF mode */
4053 for (i = min; i < max; i++)
4055 if (eptr >= md->end_subject)
4057 SCHECK_PARTIAL();
4058 break;
4060 if (fc == *eptr) break;
4061 eptr++;
4063 if (possessive) continue; /* No backtracking */
4064 for (;;)
4066 if (eptr == pp) goto TAIL_RECURSE;
4067 RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
4068 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4069 eptr--;
4072 /* Control never gets here */
4075 /* Control never gets here */
4077 /* Match a single character type repeatedly; several different opcodes
4078 share code. This is very similar to the code for single characters, but we
4079 repeat it in the interests of efficiency. */
4081 case OP_TYPEEXACT:
4082 min = max = GET2(ecode, 1);
4083 minimize = TRUE;
4084 ecode += 1 + IMM2_SIZE;
4085 goto REPEATTYPE;
4087 case OP_TYPEUPTO:
4088 case OP_TYPEMINUPTO:
4089 min = 0;
4090 max = GET2(ecode, 1);
4091 minimize = *ecode == OP_TYPEMINUPTO;
4092 ecode += 1 + IMM2_SIZE;
4093 goto REPEATTYPE;
4095 case OP_TYPEPOSSTAR:
4096 possessive = TRUE;
4097 min = 0;
4098 max = INT_MAX;
4099 ecode++;
4100 goto REPEATTYPE;
4102 case OP_TYPEPOSPLUS:
4103 possessive = TRUE;
4104 min = 1;
4105 max = INT_MAX;
4106 ecode++;
4107 goto REPEATTYPE;
4109 case OP_TYPEPOSQUERY:
4110 possessive = TRUE;
4111 min = 0;
4112 max = 1;
4113 ecode++;
4114 goto REPEATTYPE;
4116 case OP_TYPEPOSUPTO:
4117 possessive = TRUE;
4118 min = 0;
4119 max = GET2(ecode, 1);
4120 ecode += 1 + IMM2_SIZE;
4121 goto REPEATTYPE;
4123 case OP_TYPESTAR:
4124 case OP_TYPEMINSTAR:
4125 case OP_TYPEPLUS:
4126 case OP_TYPEMINPLUS:
4127 case OP_TYPEQUERY:
4128 case OP_TYPEMINQUERY:
4129 c = *ecode++ - OP_TYPESTAR;
4130 minimize = (c & 1) != 0;
4131 min = rep_min[c]; /* Pick up values from tables; */
4132 max = rep_max[c]; /* zero for max => infinity */
4133 if (max == 0) max = INT_MAX;
4135 /* Common code for all repeated single character type matches. Note that
4136 in UTF-8 mode, '.' matches a character of any length, but for the other
4137 character types, the valid characters are all one-byte long. */
4139 REPEATTYPE:
4140 ctype = *ecode++; /* Code for the character type */
4142 #ifdef SUPPORT_UCP
4143 if (ctype == OP_PROP || ctype == OP_NOTPROP)
4145 prop_fail_result = ctype == OP_NOTPROP;
4146 prop_type = *ecode++;
4147 prop_value = *ecode++;
4149 else prop_type = -1;
4150 #endif
4152 /* First, ensure the minimum number of matches are present. Use inline
4153 code for maximizing the speed, and do the type test once at the start
4154 (i.e. keep it out of the loop). Separate the UTF-8 code completely as that
4155 is tidier. Also separate the UCP code, which can be the same for both UTF-8
4156 and single-bytes. */
4158 if (min > 0)
4160 #ifdef SUPPORT_UCP
4161 if (prop_type >= 0)
4163 switch(prop_type)
4165 case PT_ANY:
4166 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4167 for (i = 1; i <= min; i++)
4169 if (eptr >= md->end_subject)
4171 SCHECK_PARTIAL();
4172 RRETURN(MATCH_NOMATCH);
4174 GETCHARINCTEST(c, eptr);
4176 break;
4178 case PT_LAMP:
4179 for (i = 1; i <= min; i++)
4181 int chartype;
4182 if (eptr >= md->end_subject)
4184 SCHECK_PARTIAL();
4185 RRETURN(MATCH_NOMATCH);
4187 GETCHARINCTEST(c, eptr);
4188 chartype = UCD_CHARTYPE(c);
4189 if ((chartype == ucp_Lu ||
4190 chartype == ucp_Ll ||
4191 chartype == ucp_Lt) == prop_fail_result)
4192 RRETURN(MATCH_NOMATCH);
4194 break;
4196 case PT_GC:
4197 for (i = 1; i <= min; i++)
4199 if (eptr >= md->end_subject)
4201 SCHECK_PARTIAL();
4202 RRETURN(MATCH_NOMATCH);
4204 GETCHARINCTEST(c, eptr);
4205 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4206 RRETURN(MATCH_NOMATCH);
4208 break;
4210 case PT_PC:
4211 for (i = 1; i <= min; i++)
4213 if (eptr >= md->end_subject)
4215 SCHECK_PARTIAL();
4216 RRETURN(MATCH_NOMATCH);
4218 GETCHARINCTEST(c, eptr);
4219 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4220 RRETURN(MATCH_NOMATCH);
4222 break;
4224 case PT_SC:
4225 for (i = 1; i <= min; i++)
4227 if (eptr >= md->end_subject)
4229 SCHECK_PARTIAL();
4230 RRETURN(MATCH_NOMATCH);
4232 GETCHARINCTEST(c, eptr);
4233 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4234 RRETURN(MATCH_NOMATCH);
4236 break;
4238 case PT_ALNUM:
4239 for (i = 1; i <= min; i++)
4241 int category;
4242 if (eptr >= md->end_subject)
4244 SCHECK_PARTIAL();
4245 RRETURN(MATCH_NOMATCH);
4247 GETCHARINCTEST(c, eptr);
4248 category = UCD_CATEGORY(c);
4249 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4250 RRETURN(MATCH_NOMATCH);
4252 break;
4254 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
4255 which means that Perl space and POSIX space are now identical. PCRE
4256 was changed at release 8.34. */
4258 case PT_SPACE: /* Perl space */
4259 case PT_PXSPACE: /* POSIX space */
4260 for (i = 1; i <= min; i++)
4262 if (eptr >= md->end_subject)
4264 SCHECK_PARTIAL();
4265 RRETURN(MATCH_NOMATCH);
4267 GETCHARINCTEST(c, eptr);
4268 switch(c)
4270 HSPACE_CASES:
4271 VSPACE_CASES:
4272 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4273 break;
4275 default:
4276 if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
4277 RRETURN(MATCH_NOMATCH);
4278 break;
4281 break;
4283 case PT_WORD:
4284 for (i = 1; i <= min; i++)
4286 int category;
4287 if (eptr >= md->end_subject)
4289 SCHECK_PARTIAL();
4290 RRETURN(MATCH_NOMATCH);
4292 GETCHARINCTEST(c, eptr);
4293 category = UCD_CATEGORY(c);
4294 if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
4295 == prop_fail_result)
4296 RRETURN(MATCH_NOMATCH);
4298 break;
4300 case PT_CLIST:
4301 for (i = 1; i <= min; i++)
4303 const pcre_uint32 *cp;
4304 if (eptr >= md->end_subject)
4306 SCHECK_PARTIAL();
4307 RRETURN(MATCH_NOMATCH);
4309 GETCHARINCTEST(c, eptr);
4310 cp = PRIV(ucd_caseless_sets) + prop_value;
4311 for (;;)
4313 if (c < *cp)
4314 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
4315 if (c == *cp++)
4316 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
4319 break;
4321 case PT_UCNC:
4322 for (i = 1; i <= min; i++)
4324 if (eptr >= md->end_subject)
4326 SCHECK_PARTIAL();
4327 RRETURN(MATCH_NOMATCH);
4329 GETCHARINCTEST(c, eptr);
4330 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
4331 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
4332 c >= 0xe000) == prop_fail_result)
4333 RRETURN(MATCH_NOMATCH);
4335 break;
4337 /* This should not occur */
4339 default:
4340 RRETURN(PCRE_ERROR_INTERNAL);
4344 /* Match extended Unicode sequences. We will get here only if the
4345 support is in the binary; otherwise a compile-time error occurs. */
4347 else if (ctype == OP_EXTUNI)
4349 for (i = 1; i <= min; i++)
4351 if (eptr >= md->end_subject)
4353 SCHECK_PARTIAL();
4354 RRETURN(MATCH_NOMATCH);
4356 else
4358 int lgb, rgb;
4359 GETCHARINCTEST(c, eptr);
4360 lgb = UCD_GRAPHBREAK(c);
4361 while (eptr < md->end_subject)
4363 int len = 1;
4364 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4365 rgb = UCD_GRAPHBREAK(c);
4366 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
4367 lgb = rgb;
4368 eptr += len;
4371 CHECK_PARTIAL();
4375 else
4376 #endif /* SUPPORT_UCP */
4378 /* Handle all other cases when the coding is UTF-8 */
4380 #ifdef SUPPORT_UTF
4381 if (utf) switch(ctype)
4383 case OP_ANY:
4384 for (i = 1; i <= min; i++)
4386 if (eptr >= md->end_subject)
4388 SCHECK_PARTIAL();
4389 RRETURN(MATCH_NOMATCH);
4391 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4392 if (md->partial != 0 &&
4393 eptr + 1 >= md->end_subject &&
4394 NLBLOCK->nltype == NLTYPE_FIXED &&
4395 NLBLOCK->nllen == 2 &&
4396 UCHAR21(eptr) == NLBLOCK->nl[0])
4398 md->hitend = TRUE;
4399 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4401 eptr++;
4402 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4404 break;
4406 case OP_ALLANY:
4407 for (i = 1; i <= min; i++)
4409 if (eptr >= md->end_subject)
4411 SCHECK_PARTIAL();
4412 RRETURN(MATCH_NOMATCH);
4414 eptr++;
4415 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4417 break;
4419 case OP_ANYBYTE:
4420 if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH);
4421 eptr += min;
4422 break;
4424 case OP_ANYNL:
4425 for (i = 1; i <= min; i++)
4427 if (eptr >= md->end_subject)
4429 SCHECK_PARTIAL();
4430 RRETURN(MATCH_NOMATCH);
4432 GETCHARINC(c, eptr);
4433 switch(c)
4435 default: RRETURN(MATCH_NOMATCH);
4437 case CHAR_CR:
4438 if (eptr < md->end_subject && UCHAR21(eptr) == CHAR_LF) eptr++;
4439 break;
4441 case CHAR_LF:
4442 break;
4444 case CHAR_VT:
4445 case CHAR_FF:
4446 case CHAR_NEL:
4447 #ifndef EBCDIC
4448 case 0x2028:
4449 case 0x2029:
4450 #endif /* Not EBCDIC */
4451 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4452 break;
4455 break;
4457 case OP_NOT_HSPACE:
4458 for (i = 1; i <= min; i++)
4460 if (eptr >= md->end_subject)
4462 SCHECK_PARTIAL();
4463 RRETURN(MATCH_NOMATCH);
4465 GETCHARINC(c, eptr);
4466 switch(c)
4468 HSPACE_CASES: RRETURN(MATCH_NOMATCH); /* Byte and multibyte cases */
4469 default: break;
4472 break;
4474 case OP_HSPACE:
4475 for (i = 1; i <= min; i++)
4477 if (eptr >= md->end_subject)
4479 SCHECK_PARTIAL();
4480 RRETURN(MATCH_NOMATCH);
4482 GETCHARINC(c, eptr);
4483 switch(c)
4485 HSPACE_CASES: break; /* Byte and multibyte cases */
4486 default: RRETURN(MATCH_NOMATCH);
4489 break;
4491 case OP_NOT_VSPACE:
4492 for (i = 1; i <= min; i++)
4494 if (eptr >= md->end_subject)
4496 SCHECK_PARTIAL();
4497 RRETURN(MATCH_NOMATCH);
4499 GETCHARINC(c, eptr);
4500 switch(c)
4502 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
4503 default: break;
4506 break;
4508 case OP_VSPACE:
4509 for (i = 1; i <= min; i++)
4511 if (eptr >= md->end_subject)
4513 SCHECK_PARTIAL();
4514 RRETURN(MATCH_NOMATCH);
4516 GETCHARINC(c, eptr);
4517 switch(c)
4519 VSPACE_CASES: break;
4520 default: RRETURN(MATCH_NOMATCH);
4523 break;
4525 case OP_NOT_DIGIT:
4526 for (i = 1; i <= min; i++)
4528 if (eptr >= md->end_subject)
4530 SCHECK_PARTIAL();
4531 RRETURN(MATCH_NOMATCH);
4533 GETCHARINC(c, eptr);
4534 if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
4535 RRETURN(MATCH_NOMATCH);
4537 break;
4539 case OP_DIGIT:
4540 for (i = 1; i <= min; i++)
4542 pcre_uint32 cc;
4543 if (eptr >= md->end_subject)
4545 SCHECK_PARTIAL();
4546 RRETURN(MATCH_NOMATCH);
4548 cc = UCHAR21(eptr);
4549 if (cc >= 128 || (md->ctypes[cc] & ctype_digit) == 0)
4550 RRETURN(MATCH_NOMATCH);
4551 eptr++;
4552 /* No need to skip more bytes - we know it's a 1-byte character */
4554 break;
4556 case OP_NOT_WHITESPACE:
4557 for (i = 1; i <= min; i++)
4559 pcre_uint32 cc;
4560 if (eptr >= md->end_subject)
4562 SCHECK_PARTIAL();
4563 RRETURN(MATCH_NOMATCH);
4565 cc = UCHAR21(eptr);
4566 if (cc < 128 && (md->ctypes[cc] & ctype_space) != 0)
4567 RRETURN(MATCH_NOMATCH);
4568 eptr++;
4569 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4571 break;
4573 case OP_WHITESPACE:
4574 for (i = 1; i <= min; i++)
4576 pcre_uint32 cc;
4577 if (eptr >= md->end_subject)
4579 SCHECK_PARTIAL();
4580 RRETURN(MATCH_NOMATCH);
4582 cc = UCHAR21(eptr);
4583 if (cc >= 128 || (md->ctypes[cc] & ctype_space) == 0)
4584 RRETURN(MATCH_NOMATCH);
4585 eptr++;
4586 /* No need to skip more bytes - we know it's a 1-byte character */
4588 break;
4590 case OP_NOT_WORDCHAR:
4591 for (i = 1; i <= min; i++)
4593 pcre_uint32 cc;
4594 if (eptr >= md->end_subject)
4596 SCHECK_PARTIAL();
4597 RRETURN(MATCH_NOMATCH);
4599 cc = UCHAR21(eptr);
4600 if (cc < 128 && (md->ctypes[cc] & ctype_word) != 0)
4601 RRETURN(MATCH_NOMATCH);
4602 eptr++;
4603 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4605 break;
4607 case OP_WORDCHAR:
4608 for (i = 1; i <= min; i++)
4610 pcre_uint32 cc;
4611 if (eptr >= md->end_subject)
4613 SCHECK_PARTIAL();
4614 RRETURN(MATCH_NOMATCH);
4616 cc = UCHAR21(eptr);
4617 if (cc >= 128 || (md->ctypes[cc] & ctype_word) == 0)
4618 RRETURN(MATCH_NOMATCH);
4619 eptr++;
4620 /* No need to skip more bytes - we know it's a 1-byte character */
4622 break;
4624 default:
4625 RRETURN(PCRE_ERROR_INTERNAL);
4626 } /* End switch(ctype) */
4628 else
4629 #endif /* SUPPORT_UTF */
4631 /* Code for the non-UTF-8 case for minimum matching of operators other
4632 than OP_PROP and OP_NOTPROP. */
4634 switch(ctype)
4636 case OP_ANY:
4637 for (i = 1; i <= min; i++)
4639 if (eptr >= md->end_subject)
4641 SCHECK_PARTIAL();
4642 RRETURN(MATCH_NOMATCH);
4644 if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
4645 if (md->partial != 0 &&
4646 eptr + 1 >= md->end_subject &&
4647 NLBLOCK->nltype == NLTYPE_FIXED &&
4648 NLBLOCK->nllen == 2 &&
4649 *eptr == NLBLOCK->nl[0])
4651 md->hitend = TRUE;
4652 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4654 eptr++;
4656 break;
4658 case OP_ALLANY:
4659 if (eptr > md->end_subject - min)
4661 SCHECK_PARTIAL();
4662 RRETURN(MATCH_NOMATCH);
4664 eptr += min;
4665 break;
4667 case OP_ANYBYTE:
4668 if (eptr > md->end_subject - min)
4670 SCHECK_PARTIAL();
4671 RRETURN(MATCH_NOMATCH);
4673 eptr += min;
4674 break;
4676 case OP_ANYNL:
4677 for (i = 1; i <= min; i++)
4679 if (eptr >= md->end_subject)
4681 SCHECK_PARTIAL();
4682 RRETURN(MATCH_NOMATCH);
4684 switch(*eptr++)
4686 default: RRETURN(MATCH_NOMATCH);
4688 case CHAR_CR:
4689 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
4690 break;
4692 case CHAR_LF:
4693 break;
4695 case CHAR_VT:
4696 case CHAR_FF:
4697 case CHAR_NEL:
4698 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4699 case 0x2028:
4700 case 0x2029:
4701 #endif
4702 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
4703 break;
4706 break;
4708 case OP_NOT_HSPACE:
4709 for (i = 1; i <= min; i++)
4711 if (eptr >= md->end_subject)
4713 SCHECK_PARTIAL();
4714 RRETURN(MATCH_NOMATCH);
4716 switch(*eptr++)
4718 default: break;
4719 HSPACE_BYTE_CASES:
4720 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4721 HSPACE_MULTIBYTE_CASES:
4722 #endif
4723 RRETURN(MATCH_NOMATCH);
4726 break;
4728 case OP_HSPACE:
4729 for (i = 1; i <= min; i++)
4731 if (eptr >= md->end_subject)
4733 SCHECK_PARTIAL();
4734 RRETURN(MATCH_NOMATCH);
4736 switch(*eptr++)
4738 default: RRETURN(MATCH_NOMATCH);
4739 HSPACE_BYTE_CASES:
4740 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4741 HSPACE_MULTIBYTE_CASES:
4742 #endif
4743 break;
4746 break;
4748 case OP_NOT_VSPACE:
4749 for (i = 1; i <= min; i++)
4751 if (eptr >= md->end_subject)
4753 SCHECK_PARTIAL();
4754 RRETURN(MATCH_NOMATCH);
4756 switch(*eptr++)
4758 VSPACE_BYTE_CASES:
4759 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4760 VSPACE_MULTIBYTE_CASES:
4761 #endif
4762 RRETURN(MATCH_NOMATCH);
4763 default: break;
4766 break;
4768 case OP_VSPACE:
4769 for (i = 1; i <= min; i++)
4771 if (eptr >= md->end_subject)
4773 SCHECK_PARTIAL();
4774 RRETURN(MATCH_NOMATCH);
4776 switch(*eptr++)
4778 default: RRETURN(MATCH_NOMATCH);
4779 VSPACE_BYTE_CASES:
4780 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
4781 VSPACE_MULTIBYTE_CASES:
4782 #endif
4783 break;
4786 break;
4788 case OP_NOT_DIGIT:
4789 for (i = 1; i <= min; i++)
4791 if (eptr >= md->end_subject)
4793 SCHECK_PARTIAL();
4794 RRETURN(MATCH_NOMATCH);
4796 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0)
4797 RRETURN(MATCH_NOMATCH);
4798 eptr++;
4800 break;
4802 case OP_DIGIT:
4803 for (i = 1; i <= min; i++)
4805 if (eptr >= md->end_subject)
4807 SCHECK_PARTIAL();
4808 RRETURN(MATCH_NOMATCH);
4810 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0)
4811 RRETURN(MATCH_NOMATCH);
4812 eptr++;
4814 break;
4816 case OP_NOT_WHITESPACE:
4817 for (i = 1; i <= min; i++)
4819 if (eptr >= md->end_subject)
4821 SCHECK_PARTIAL();
4822 RRETURN(MATCH_NOMATCH);
4824 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0)
4825 RRETURN(MATCH_NOMATCH);
4826 eptr++;
4828 break;
4830 case OP_WHITESPACE:
4831 for (i = 1; i <= min; i++)
4833 if (eptr >= md->end_subject)
4835 SCHECK_PARTIAL();
4836 RRETURN(MATCH_NOMATCH);
4838 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0)
4839 RRETURN(MATCH_NOMATCH);
4840 eptr++;
4842 break;
4844 case OP_NOT_WORDCHAR:
4845 for (i = 1; i <= min; i++)
4847 if (eptr >= md->end_subject)
4849 SCHECK_PARTIAL();
4850 RRETURN(MATCH_NOMATCH);
4852 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0)
4853 RRETURN(MATCH_NOMATCH);
4854 eptr++;
4856 break;
4858 case OP_WORDCHAR:
4859 for (i = 1; i <= min; i++)
4861 if (eptr >= md->end_subject)
4863 SCHECK_PARTIAL();
4864 RRETURN(MATCH_NOMATCH);
4866 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0)
4867 RRETURN(MATCH_NOMATCH);
4868 eptr++;
4870 break;
4872 default:
4873 RRETURN(PCRE_ERROR_INTERNAL);
4877 /* If min = max, continue at the same level without recursing */
4879 if (min == max) continue;
4881 /* If minimizing, we have to test the rest of the pattern before each
4882 subsequent match. Again, separate the UTF-8 case for speed, and also
4883 separate the UCP cases. */
4885 if (minimize)
4887 #ifdef SUPPORT_UCP
4888 if (prop_type >= 0)
4890 switch(prop_type)
4892 case PT_ANY:
4893 for (fi = min;; fi++)
4895 RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4896 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4897 if (fi >= max) RRETURN(MATCH_NOMATCH);
4898 if (eptr >= md->end_subject)
4900 SCHECK_PARTIAL();
4901 RRETURN(MATCH_NOMATCH);
4903 GETCHARINCTEST(c, eptr);
4904 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4906 /* Control never gets here */
4908 case PT_LAMP:
4909 for (fi = min;; fi++)
4911 int chartype;
4912 RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4913 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4914 if (fi >= max) RRETURN(MATCH_NOMATCH);
4915 if (eptr >= md->end_subject)
4917 SCHECK_PARTIAL();
4918 RRETURN(MATCH_NOMATCH);
4920 GETCHARINCTEST(c, eptr);
4921 chartype = UCD_CHARTYPE(c);
4922 if ((chartype == ucp_Lu ||
4923 chartype == ucp_Ll ||
4924 chartype == ucp_Lt) == prop_fail_result)
4925 RRETURN(MATCH_NOMATCH);
4927 /* Control never gets here */
4929 case PT_GC:
4930 for (fi = min;; fi++)
4932 RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4933 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4934 if (fi >= max) RRETURN(MATCH_NOMATCH);
4935 if (eptr >= md->end_subject)
4937 SCHECK_PARTIAL();
4938 RRETURN(MATCH_NOMATCH);
4940 GETCHARINCTEST(c, eptr);
4941 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4942 RRETURN(MATCH_NOMATCH);
4944 /* Control never gets here */
4946 case PT_PC:
4947 for (fi = min;; fi++)
4949 RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4950 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4951 if (fi >= max) RRETURN(MATCH_NOMATCH);
4952 if (eptr >= md->end_subject)
4954 SCHECK_PARTIAL();
4955 RRETURN(MATCH_NOMATCH);
4957 GETCHARINCTEST(c, eptr);
4958 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4959 RRETURN(MATCH_NOMATCH);
4961 /* Control never gets here */
4963 case PT_SC:
4964 for (fi = min;; fi++)
4966 RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
4967 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4968 if (fi >= max) RRETURN(MATCH_NOMATCH);
4969 if (eptr >= md->end_subject)
4971 SCHECK_PARTIAL();
4972 RRETURN(MATCH_NOMATCH);
4974 GETCHARINCTEST(c, eptr);
4975 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4976 RRETURN(MATCH_NOMATCH);
4978 /* Control never gets here */
4980 case PT_ALNUM:
4981 for (fi = min;; fi++)
4983 int category;
4984 RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4985 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4986 if (fi >= max) RRETURN(MATCH_NOMATCH);
4987 if (eptr >= md->end_subject)
4989 SCHECK_PARTIAL();
4990 RRETURN(MATCH_NOMATCH);
4992 GETCHARINCTEST(c, eptr);
4993 category = UCD_CATEGORY(c);
4994 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4995 RRETURN(MATCH_NOMATCH);
4997 /* Control never gets here */
4999 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
5000 which means that Perl space and POSIX space are now identical. PCRE
5001 was changed at release 8.34. */
5003 case PT_SPACE: /* Perl space */
5004 case PT_PXSPACE: /* POSIX space */
5005 for (fi = min;; fi++)
5007 RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
5008 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5009 if (fi >= max) RRETURN(MATCH_NOMATCH);
5010 if (eptr >= md->end_subject)
5012 SCHECK_PARTIAL();
5013 RRETURN(MATCH_NOMATCH);
5015 GETCHARINCTEST(c, eptr);
5016 switch(c)
5018 HSPACE_CASES:
5019 VSPACE_CASES:
5020 if (prop_fail_result) RRETURN(MATCH_NOMATCH);
5021 break;
5023 default:
5024 if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
5025 RRETURN(MATCH_NOMATCH);
5026 break;
5029 /* Control never gets here */
5031 case PT_WORD:
5032 for (fi = min;; fi++)
5034 int category;
5035 RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
5036 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5037 if (fi >= max) RRETURN(MATCH_NOMATCH);
5038 if (eptr >= md->end_subject)
5040 SCHECK_PARTIAL();
5041 RRETURN(MATCH_NOMATCH);
5043 GETCHARINCTEST(c, eptr);
5044 category = UCD_CATEGORY(c);
5045 if ((category == ucp_L ||
5046 category == ucp_N ||
5047 c == CHAR_UNDERSCORE)
5048 == prop_fail_result)
5049 RRETURN(MATCH_NOMATCH);
5051 /* Control never gets here */
5053 case PT_CLIST:
5054 for (fi = min;; fi++)
5056 const pcre_uint32 *cp;
5057 RMATCH(eptr, ecode, offset_top, md, eptrb, RM67);
5058 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5059 if (fi >= max) RRETURN(MATCH_NOMATCH);
5060 if (eptr >= md->end_subject)
5062 SCHECK_PARTIAL();
5063 RRETURN(MATCH_NOMATCH);
5065 GETCHARINCTEST(c, eptr);
5066 cp = PRIV(ucd_caseless_sets) + prop_value;
5067 for (;;)
5069 if (c < *cp)
5070 { if (prop_fail_result) break; else { RRETURN(MATCH_NOMATCH); } }
5071 if (c == *cp++)
5072 { if (prop_fail_result) { RRETURN(MATCH_NOMATCH); } else break; }
5075 /* Control never gets here */
5077 case PT_UCNC:
5078 for (fi = min;; fi++)
5080 RMATCH(eptr, ecode, offset_top, md, eptrb, RM60);
5081 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5082 if (fi >= max) RRETURN(MATCH_NOMATCH);
5083 if (eptr >= md->end_subject)
5085 SCHECK_PARTIAL();
5086 RRETURN(MATCH_NOMATCH);
5088 GETCHARINCTEST(c, eptr);
5089 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5090 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5091 c >= 0xe000) == prop_fail_result)
5092 RRETURN(MATCH_NOMATCH);
5094 /* Control never gets here */
5096 /* This should never occur */
5097 default:
5098 RRETURN(PCRE_ERROR_INTERNAL);
5102 /* Match extended Unicode sequences. We will get here only if the
5103 support is in the binary; otherwise a compile-time error occurs. */
5105 else if (ctype == OP_EXTUNI)
5107 for (fi = min;; fi++)
5109 RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
5110 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5111 if (fi >= max) RRETURN(MATCH_NOMATCH);
5112 if (eptr >= md->end_subject)
5114 SCHECK_PARTIAL();
5115 RRETURN(MATCH_NOMATCH);
5117 else
5119 int lgb, rgb;
5120 GETCHARINCTEST(c, eptr);
5121 lgb = UCD_GRAPHBREAK(c);
5122 while (eptr < md->end_subject)
5124 int len = 1;
5125 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5126 rgb = UCD_GRAPHBREAK(c);
5127 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5128 lgb = rgb;
5129 eptr += len;
5132 CHECK_PARTIAL();
5135 else
5136 #endif /* SUPPORT_UCP */
5138 #ifdef SUPPORT_UTF
5139 if (utf)
5141 for (fi = min;; fi++)
5143 RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
5144 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5145 if (fi >= max) RRETURN(MATCH_NOMATCH);
5146 if (eptr >= md->end_subject)
5148 SCHECK_PARTIAL();
5149 RRETURN(MATCH_NOMATCH);
5151 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5152 RRETURN(MATCH_NOMATCH);
5153 GETCHARINC(c, eptr);
5154 switch(ctype)
5156 case OP_ANY: /* This is the non-NL case */
5157 if (md->partial != 0 && /* Take care with CRLF partial */
5158 eptr >= md->end_subject &&
5159 NLBLOCK->nltype == NLTYPE_FIXED &&
5160 NLBLOCK->nllen == 2 &&
5161 c == NLBLOCK->nl[0])
5163 md->hitend = TRUE;
5164 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5166 break;
5168 case OP_ALLANY:
5169 case OP_ANYBYTE:
5170 break;
5172 case OP_ANYNL:
5173 switch(c)
5175 default: RRETURN(MATCH_NOMATCH);
5176 case CHAR_CR:
5177 if (eptr < md->end_subject && UCHAR21(eptr) == CHAR_LF) eptr++;
5178 break;
5180 case CHAR_LF:
5181 break;
5183 case CHAR_VT:
5184 case CHAR_FF:
5185 case CHAR_NEL:
5186 #ifndef EBCDIC
5187 case 0x2028:
5188 case 0x2029:
5189 #endif /* Not EBCDIC */
5190 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5191 break;
5193 break;
5195 case OP_NOT_HSPACE:
5196 switch(c)
5198 HSPACE_CASES: RRETURN(MATCH_NOMATCH);
5199 default: break;
5201 break;
5203 case OP_HSPACE:
5204 switch(c)
5206 HSPACE_CASES: break;
5207 default: RRETURN(MATCH_NOMATCH);
5209 break;
5211 case OP_NOT_VSPACE:
5212 switch(c)
5214 VSPACE_CASES: RRETURN(MATCH_NOMATCH);
5215 default: break;
5217 break;
5219 case OP_VSPACE:
5220 switch(c)
5222 VSPACE_CASES: break;
5223 default: RRETURN(MATCH_NOMATCH);
5225 break;
5227 case OP_NOT_DIGIT:
5228 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
5229 RRETURN(MATCH_NOMATCH);
5230 break;
5232 case OP_DIGIT:
5233 if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0)
5234 RRETURN(MATCH_NOMATCH);
5235 break;
5237 case OP_NOT_WHITESPACE:
5238 if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
5239 RRETURN(MATCH_NOMATCH);
5240 break;
5242 case OP_WHITESPACE:
5243 if (c >= 256 || (md->ctypes[c] & ctype_space) == 0)
5244 RRETURN(MATCH_NOMATCH);
5245 break;
5247 case OP_NOT_WORDCHAR:
5248 if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
5249 RRETURN(MATCH_NOMATCH);
5250 break;
5252 case OP_WORDCHAR:
5253 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0)
5254 RRETURN(MATCH_NOMATCH);
5255 break;
5257 default:
5258 RRETURN(PCRE_ERROR_INTERNAL);
5262 else
5263 #endif
5264 /* Not UTF mode */
5266 for (fi = min;; fi++)
5268 RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
5269 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5270 if (fi >= max) RRETURN(MATCH_NOMATCH);
5271 if (eptr >= md->end_subject)
5273 SCHECK_PARTIAL();
5274 RRETURN(MATCH_NOMATCH);
5276 if (ctype == OP_ANY && IS_NEWLINE(eptr))
5277 RRETURN(MATCH_NOMATCH);
5278 c = *eptr++;
5279 switch(ctype)
5281 case OP_ANY: /* This is the non-NL case */
5282 if (md->partial != 0 && /* Take care with CRLF partial */
5283 eptr >= md->end_subject &&
5284 NLBLOCK->nltype == NLTYPE_FIXED &&
5285 NLBLOCK->nllen == 2 &&
5286 c == NLBLOCK->nl[0])
5288 md->hitend = TRUE;
5289 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5291 break;
5293 case OP_ALLANY:
5294 case OP_ANYBYTE:
5295 break;
5297 case OP_ANYNL:
5298 switch(c)
5300 default: RRETURN(MATCH_NOMATCH);
5301 case CHAR_CR:
5302 if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
5303 break;
5305 case CHAR_LF:
5306 break;
5308 case CHAR_VT:
5309 case CHAR_FF:
5310 case CHAR_NEL:
5311 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5312 case 0x2028:
5313 case 0x2029:
5314 #endif
5315 if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
5316 break;
5318 break;
5320 case OP_NOT_HSPACE:
5321 switch(c)
5323 default: break;
5324 HSPACE_BYTE_CASES:
5325 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5326 HSPACE_MULTIBYTE_CASES:
5327 #endif
5328 RRETURN(MATCH_NOMATCH);
5330 break;
5332 case OP_HSPACE:
5333 switch(c)
5335 default: RRETURN(MATCH_NOMATCH);
5336 HSPACE_BYTE_CASES:
5337 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5338 HSPACE_MULTIBYTE_CASES:
5339 #endif
5340 break;
5342 break;
5344 case OP_NOT_VSPACE:
5345 switch(c)
5347 default: break;
5348 VSPACE_BYTE_CASES:
5349 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5350 VSPACE_MULTIBYTE_CASES:
5351 #endif
5352 RRETURN(MATCH_NOMATCH);
5354 break;
5356 case OP_VSPACE:
5357 switch(c)
5359 default: RRETURN(MATCH_NOMATCH);
5360 VSPACE_BYTE_CASES:
5361 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
5362 VSPACE_MULTIBYTE_CASES:
5363 #endif
5364 break;
5366 break;
5368 case OP_NOT_DIGIT:
5369 if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
5370 break;
5372 case OP_DIGIT:
5373 if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
5374 break;
5376 case OP_NOT_WHITESPACE:
5377 if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
5378 break;
5380 case OP_WHITESPACE:
5381 if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
5382 break;
5384 case OP_NOT_WORDCHAR:
5385 if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH);
5386 break;
5388 case OP_WORDCHAR:
5389 if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH);
5390 break;
5392 default:
5393 RRETURN(PCRE_ERROR_INTERNAL);
5397 /* Control never gets here */
5400 /* If maximizing, it is worth using inline code for speed, doing the type
5401 test once at the start (i.e. keep it out of the loop). Again, keep the
5402 UTF-8 and UCP stuff separate. */
5404 else
5406 pp = eptr; /* Remember where we started */
5408 #ifdef SUPPORT_UCP
5409 if (prop_type >= 0)
5411 switch(prop_type)
5413 case PT_ANY:
5414 for (i = min; i < max; i++)
5416 int len = 1;
5417 if (eptr >= md->end_subject)
5419 SCHECK_PARTIAL();
5420 break;
5422 GETCHARLENTEST(c, eptr, len);
5423 if (prop_fail_result) break;
5424 eptr+= len;
5426 break;
5428 case PT_LAMP:
5429 for (i = min; i < max; i++)
5431 int chartype;
5432 int len = 1;
5433 if (eptr >= md->end_subject)
5435 SCHECK_PARTIAL();
5436 break;
5438 GETCHARLENTEST(c, eptr, len);
5439 chartype = UCD_CHARTYPE(c);
5440 if ((chartype == ucp_Lu ||
5441 chartype == ucp_Ll ||
5442 chartype == ucp_Lt) == prop_fail_result)
5443 break;
5444 eptr+= len;
5446 break;
5448 case PT_GC:
5449 for (i = min; i < max; i++)
5451 int len = 1;
5452 if (eptr >= md->end_subject)
5454 SCHECK_PARTIAL();
5455 break;
5457 GETCHARLENTEST(c, eptr, len);
5458 if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
5459 eptr+= len;
5461 break;
5463 case PT_PC:
5464 for (i = min; i < max; i++)
5466 int len = 1;
5467 if (eptr >= md->end_subject)
5469 SCHECK_PARTIAL();
5470 break;
5472 GETCHARLENTEST(c, eptr, len);
5473 if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
5474 eptr+= len;
5476 break;
5478 case PT_SC:
5479 for (i = min; i < max; i++)
5481 int len = 1;
5482 if (eptr >= md->end_subject)
5484 SCHECK_PARTIAL();
5485 break;
5487 GETCHARLENTEST(c, eptr, len);
5488 if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
5489 eptr+= len;
5491 break;
5493 case PT_ALNUM:
5494 for (i = min; i < max; i++)
5496 int category;
5497 int len = 1;
5498 if (eptr >= md->end_subject)
5500 SCHECK_PARTIAL();
5501 break;
5503 GETCHARLENTEST(c, eptr, len);
5504 category = UCD_CATEGORY(c);
5505 if ((category == ucp_L || category == ucp_N) == prop_fail_result)
5506 break;
5507 eptr+= len;
5509 break;
5511 /* Perl space used to exclude VT, but from Perl 5.18 it is included,
5512 which means that Perl space and POSIX space are now identical. PCRE
5513 was changed at release 8.34. */
5515 case PT_SPACE: /* Perl space */
5516 case PT_PXSPACE: /* POSIX space */
5517 for (i = min; i < max; i++)
5519 int len = 1;
5520 if (eptr >= md->end_subject)
5522 SCHECK_PARTIAL();
5523 break;
5525 GETCHARLENTEST(c, eptr, len);
5526 switch(c)
5528 HSPACE_CASES:
5529 VSPACE_CASES:
5530 if (prop_fail_result) goto ENDLOOP99; /* Break the loop */
5531 break;
5533 default:
5534 if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
5535 goto ENDLOOP99; /* Break the loop */
5536 break;
5538 eptr+= len;
5540 ENDLOOP99:
5541 break;
5543 case PT_WORD:
5544 for (i = min; i < max; i++)
5546 int category;
5547 int len = 1;
5548 if (eptr >= md->end_subject)
5550 SCHECK_PARTIAL();
5551 break;
5553 GETCHARLENTEST(c, eptr, len);
5554 category = UCD_CATEGORY(c);
5555 if ((category == ucp_L || category == ucp_N ||
5556 c == CHAR_UNDERSCORE) == prop_fail_result)
5557 break;
5558 eptr+= len;
5560 break;
5562 case PT_CLIST:
5563 for (i = min; i < max; i++)
5565 const pcre_uint32 *cp;
5566 int len = 1;
5567 if (eptr >= md->end_subject)
5569 SCHECK_PARTIAL();
5570 break;
5572 GETCHARLENTEST(c, eptr, len);
5573 cp = PRIV(ucd_caseless_sets) + prop_value;
5574 for (;;)
5576 if (c < *cp)
5577 { if (prop_fail_result) break; else goto GOT_MAX; }
5578 if (c == *cp++)
5579 { if (prop_fail_result) goto GOT_MAX; else break; }
5581 eptr += len;
5583 GOT_MAX:
5584 break;
5586 case PT_UCNC:
5587 for (i = min; i < max; i++)
5589 int len = 1;
5590 if (eptr >= md->end_subject)
5592 SCHECK_PARTIAL();
5593 break;
5595 GETCHARLENTEST(c, eptr, len);
5596 if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
5597 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
5598 c >= 0xe000) == prop_fail_result)
5599 break;
5600 eptr += len;
5602 break;
5604 default:
5605 RRETURN(PCRE_ERROR_INTERNAL);
5608 /* eptr is now past the end of the maximum run */
5610 if (possessive) continue; /* No backtracking */
5611 for(;;)
5613 if (eptr <= pp) goto TAIL_RECURSE;
5614 RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5615 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5616 eptr--;
5617 if (utf) BACKCHAR(eptr);
5621 /* Match extended Unicode grapheme clusters. We will get here only if the
5622 support is in the binary; otherwise a compile-time error occurs. */
5624 else if (ctype == OP_EXTUNI)
5626 for (i = min; i < max; i++)
5628 if (eptr >= md->end_subject)
5630 SCHECK_PARTIAL();
5631 break;
5633 else
5635 int lgb, rgb;
5636 GETCHARINCTEST(c, eptr);
5637 lgb = UCD_GRAPHBREAK(c);
5638 while (eptr < md->end_subject)
5640 int len = 1;
5641 if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5642 rgb = UCD_GRAPHBREAK(c);
5643 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5644 lgb = rgb;
5645 eptr += len;
5648 CHECK_PARTIAL();
5651 /* eptr is now past the end of the maximum run */
5653 if (possessive) continue; /* No backtracking */
5655 /* We use <= pp rather than == pp to detect the start of the run while
5656 backtracking because the use of \C in UTF mode can cause BACKCHAR to
5657 move back past pp. This is just palliative; the use of \C in UTF mode
5658 is fraught with danger. */
5660 for(;;)
5662 int lgb, rgb;
5663 PCRE_PUCHAR fptr;
5665 if (eptr <= pp) goto TAIL_RECURSE; /* At start of char run */
5666 RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5667 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5669 /* Backtracking over an extended grapheme cluster involves inspecting
5670 the previous two characters (if present) to see if a break is
5671 permitted between them. */
5673 eptr--;
5674 if (!utf) c = *eptr; else
5676 BACKCHAR(eptr);
5677 GETCHAR(c, eptr);
5679 rgb = UCD_GRAPHBREAK(c);
5681 for (;;)
5683 if (eptr <= pp) goto TAIL_RECURSE; /* At start of char run */
5684 fptr = eptr - 1;
5685 if (!utf) c = *fptr; else
5687 BACKCHAR(fptr);
5688 GETCHAR(c, fptr);
5690 lgb = UCD_GRAPHBREAK(c);
5691 if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5692 eptr = fptr;
5693 rgb = lgb;
5698 else
5699 #endif /* SUPPORT_UCP */
5701 #ifdef SUPPORT_UTF
5702 if (utf)
5704 switch(ctype)
5706 case OP_ANY:
5707 for (i = min; i < max; i++)
5709 if (eptr >= md->end_subject)
5711 SCHECK_PARTIAL();
5712 break;
5714 if (IS_NEWLINE(eptr)) break;
5715 if (md->partial != 0 && /* Take care with CRLF partial */
5716 eptr + 1 >= md->end_subject &&
5717 NLBLOCK->nltype == NLTYPE_FIXED &&
5718 NLBLOCK->nllen == 2 &&
5719 UCHAR21(eptr) == NLBLOCK->nl[0])
5721 md->hitend = TRUE;
5722 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5724 eptr++;
5725 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5727 break;
5729 case OP_ALLANY:
5730 if (max < INT_MAX)
5732 for (i = min; i < max; i++)
5734 if (eptr >= md->end_subject)
5736 SCHECK_PARTIAL();
5737 break;
5739 eptr++;
5740 ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5743 else
5745 eptr = md->end_subject; /* Unlimited UTF-8 repeat */
5746 SCHECK_PARTIAL();
5748 break;
5750 /* The byte case is the same as non-UTF8 */
5752 case OP_ANYBYTE:
5753 c = max - min;
5754 if (c > (unsigned int)(md->end_subject - eptr))
5756 eptr = md->end_subject;
5757 SCHECK_PARTIAL();
5759 else eptr += c;
5760 break;
5762 case OP_ANYNL:
5763 for (i = min; i < max; i++)
5765 int len = 1;
5766 if (eptr >= md->end_subject)
5768 SCHECK_PARTIAL();
5769 break;
5771 GETCHARLEN(c, eptr, len);
5772 if (c == CHAR_CR)
5774 if (++eptr >= md->end_subject) break;
5775 if (UCHAR21(eptr) == CHAR_LF) eptr++;
5777 else
5779 if (c != CHAR_LF &&
5780 (md->bsr_anycrlf ||
5781 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
5782 #ifndef EBCDIC
5783 && c != 0x2028 && c != 0x2029
5784 #endif /* Not EBCDIC */
5786 break;
5787 eptr += len;
5790 break;
5792 case OP_NOT_HSPACE:
5793 case OP_HSPACE:
5794 for (i = min; i < max; i++)
5796 BOOL gotspace;
5797 int len = 1;
5798 if (eptr >= md->end_subject)
5800 SCHECK_PARTIAL();
5801 break;
5803 GETCHARLEN(c, eptr, len);
5804 switch(c)
5806 HSPACE_CASES: gotspace = TRUE; break;
5807 default: gotspace = FALSE; break;
5809 if (gotspace == (ctype == OP_NOT_HSPACE)) break;
5810 eptr += len;
5812 break;
5814 case OP_NOT_VSPACE:
5815 case OP_VSPACE:
5816 for (i = min; i < max; i++)
5818 BOOL gotspace;
5819 int len = 1;
5820 if (eptr >= md->end_subject)
5822 SCHECK_PARTIAL();
5823 break;
5825 GETCHARLEN(c, eptr, len);
5826 switch(c)
5828 VSPACE_CASES: gotspace = TRUE; break;
5829 default: gotspace = FALSE; break;
5831 if (gotspace == (ctype == OP_NOT_VSPACE)) break;
5832 eptr += len;
5834 break;
5836 case OP_NOT_DIGIT:
5837 for (i = min; i < max; i++)
5839 int len = 1;
5840 if (eptr >= md->end_subject)
5842 SCHECK_PARTIAL();
5843 break;
5845 GETCHARLEN(c, eptr, len);
5846 if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
5847 eptr+= len;
5849 break;
5851 case OP_DIGIT:
5852 for (i = min; i < max; i++)
5854 int len = 1;
5855 if (eptr >= md->end_subject)
5857 SCHECK_PARTIAL();
5858 break;
5860 GETCHARLEN(c, eptr, len);
5861 if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break;
5862 eptr+= len;
5864 break;
5866 case OP_NOT_WHITESPACE:
5867 for (i = min; i < max; i++)
5869 int len = 1;
5870 if (eptr >= md->end_subject)
5872 SCHECK_PARTIAL();
5873 break;
5875 GETCHARLEN(c, eptr, len);
5876 if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
5877 eptr+= len;
5879 break;
5881 case OP_WHITESPACE:
5882 for (i = min; i < max; i++)
5884 int len = 1;
5885 if (eptr >= md->end_subject)
5887 SCHECK_PARTIAL();
5888 break;
5890 GETCHARLEN(c, eptr, len);
5891 if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break;
5892 eptr+= len;
5894 break;
5896 case OP_NOT_WORDCHAR:
5897 for (i = min; i < max; i++)
5899 int len = 1;
5900 if (eptr >= md->end_subject)
5902 SCHECK_PARTIAL();
5903 break;
5905 GETCHARLEN(c, eptr, len);
5906 if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
5907 eptr+= len;
5909 break;
5911 case OP_WORDCHAR:
5912 for (i = min; i < max; i++)
5914 int len = 1;
5915 if (eptr >= md->end_subject)
5917 SCHECK_PARTIAL();
5918 break;
5920 GETCHARLEN(c, eptr, len);
5921 if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break;
5922 eptr+= len;
5924 break;
5926 default:
5927 RRETURN(PCRE_ERROR_INTERNAL);
5930 if (possessive) continue; /* No backtracking */
5931 for(;;)
5933 if (eptr <= pp) goto TAIL_RECURSE;
5934 RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
5935 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5936 eptr--;
5937 BACKCHAR(eptr);
5938 if (ctype == OP_ANYNL && eptr > pp && UCHAR21(eptr) == CHAR_NL &&
5939 UCHAR21(eptr - 1) == CHAR_CR) eptr--;
5942 else
5943 #endif /* SUPPORT_UTF */
5944 /* Not UTF mode */
5946 switch(ctype)
5948 case OP_ANY:
5949 for (i = min; i < max; i++)
5951 if (eptr >= md->end_subject)
5953 SCHECK_PARTIAL();
5954 break;
5956 if (IS_NEWLINE(eptr)) break;
5957 if (md->partial != 0 && /* Take care with CRLF partial */
5958 eptr + 1 >= md->end_subject &&
5959 NLBLOCK->nltype == NLTYPE_FIXED &&
5960 NLBLOCK->nllen == 2 &&
5961 *eptr == NLBLOCK->nl[0])
5963 md->hitend = TRUE;
5964 if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5966 eptr++;
5968 break;
5970 case OP_ALLANY:
5971 case OP_ANYBYTE:
5972 c = max - min;
5973 if (c > (unsigned int)(md->end_subject - eptr))
5975 eptr = md->end_subject;
5976 SCHECK_PARTIAL();
5978 else eptr += c;
5979 break;
5981 case OP_ANYNL:
5982 for (i = min; i < max; i++)
5984 if (eptr >= md->end_subject)
5986 SCHECK_PARTIAL();
5987 break;
5989 c = *eptr;
5990 if (c == CHAR_CR)
5992 if (++eptr >= md->end_subject) break;
5993 if (*eptr == CHAR_LF) eptr++;
5995 else
5997 if (c != CHAR_LF && (md->bsr_anycrlf ||
5998 (c != CHAR_VT && c != CHAR_FF && c != CHAR_NEL
5999 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6000 && c != 0x2028 && c != 0x2029
6001 #endif
6002 ))) break;
6003 eptr++;
6006 break;
6008 case OP_NOT_HSPACE:
6009 for (i = min; i < max; i++)
6011 if (eptr >= md->end_subject)
6013 SCHECK_PARTIAL();
6014 break;
6016 switch(*eptr)
6018 default: eptr++; break;
6019 HSPACE_BYTE_CASES:
6020 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6021 HSPACE_MULTIBYTE_CASES:
6022 #endif
6023 goto ENDLOOP00;
6026 ENDLOOP00:
6027 break;
6029 case OP_HSPACE:
6030 for (i = min; i < max; i++)
6032 if (eptr >= md->end_subject)
6034 SCHECK_PARTIAL();
6035 break;
6037 switch(*eptr)
6039 default: goto ENDLOOP01;
6040 HSPACE_BYTE_CASES:
6041 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6042 HSPACE_MULTIBYTE_CASES:
6043 #endif
6044 eptr++; break;
6047 ENDLOOP01:
6048 break;
6050 case OP_NOT_VSPACE:
6051 for (i = min; i < max; i++)
6053 if (eptr >= md->end_subject)
6055 SCHECK_PARTIAL();
6056 break;
6058 switch(*eptr)
6060 default: eptr++; break;
6061 VSPACE_BYTE_CASES:
6062 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6063 VSPACE_MULTIBYTE_CASES:
6064 #endif
6065 goto ENDLOOP02;
6068 ENDLOOP02:
6069 break;
6071 case OP_VSPACE:
6072 for (i = min; i < max; i++)
6074 if (eptr >= md->end_subject)
6076 SCHECK_PARTIAL();
6077 break;
6079 switch(*eptr)
6081 default: goto ENDLOOP03;
6082 VSPACE_BYTE_CASES:
6083 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
6084 VSPACE_MULTIBYTE_CASES:
6085 #endif
6086 eptr++; break;
6089 ENDLOOP03:
6090 break;
6092 case OP_NOT_DIGIT:
6093 for (i = min; i < max; i++)
6095 if (eptr >= md->end_subject)
6097 SCHECK_PARTIAL();
6098 break;
6100 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) break;
6101 eptr++;
6103 break;
6105 case OP_DIGIT:
6106 for (i = min; i < max; i++)
6108 if (eptr >= md->end_subject)
6110 SCHECK_PARTIAL();
6111 break;
6113 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) break;
6114 eptr++;
6116 break;
6118 case OP_NOT_WHITESPACE:
6119 for (i = min; i < max; i++)
6121 if (eptr >= md->end_subject)
6123 SCHECK_PARTIAL();
6124 break;
6126 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break;
6127 eptr++;
6129 break;
6131 case OP_WHITESPACE:
6132 for (i = min; i < max; i++)
6134 if (eptr >= md->end_subject)
6136 SCHECK_PARTIAL();
6137 break;
6139 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break;
6140 eptr++;
6142 break;
6144 case OP_NOT_WORDCHAR:
6145 for (i = min; i < max; i++)
6147 if (eptr >= md->end_subject)
6149 SCHECK_PARTIAL();
6150 break;
6152 if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break;
6153 eptr++;
6155 break;
6157 case OP_WORDCHAR:
6158 for (i = min; i < max; i++)
6160 if (eptr >= md->end_subject)
6162 SCHECK_PARTIAL();
6163 break;
6165 if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break;
6166 eptr++;
6168 break;
6170 default:
6171 RRETURN(PCRE_ERROR_INTERNAL);
6174 if (possessive) continue; /* No backtracking */
6175 for (;;)
6177 if (eptr == pp) goto TAIL_RECURSE;
6178 RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
6179 if (rrc != MATCH_NOMATCH) RRETURN(rrc);
6180 eptr--;
6181 if (ctype == OP_ANYNL && eptr > pp && *eptr == CHAR_LF &&
6182 eptr[-1] == CHAR_CR) eptr--;
6186 /* Control never gets here */
6189 /* There's been some horrible disaster. Arrival here can only mean there is
6190 something seriously wrong in the code above or the OP_xxx definitions. */
6192 default:
6193 DPRINTF(("Unknown opcode %d\n", *ecode));
6194 RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
6197 /* Do not stick any code in here without much thought; it is assumed
6198 that "continue" in the code above comes out to here to repeat the main
6199 loop. */
6201 } /* End of main loop */
6202 /* Control never reaches here */
6205 /* When compiling to use the heap rather than the stack for recursive calls to
6206 match(), the RRETURN() macro jumps here. The number that is saved in
6207 frame->Xwhere indicates which label we actually want to return to. */
6209 #ifdef NO_RECURSE
6210 #define LBL(val) case val: goto L_RM##val;
6211 HEAP_RETURN:
6212 switch (frame->Xwhere)
6214 LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
6215 LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
6216 LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
6217 LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
6218 LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64)
6219 LBL(65) LBL(66)
6220 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6221 LBL(20) LBL(21)
6222 #endif
6223 #ifdef SUPPORT_UTF
6224 LBL(16) LBL(18)
6225 LBL(22) LBL(23) LBL(28) LBL(30)
6226 LBL(32) LBL(34) LBL(42) LBL(46)
6227 #ifdef SUPPORT_UCP
6228 LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
6229 LBL(59) LBL(60) LBL(61) LBL(62) LBL(67)
6230 #endif /* SUPPORT_UCP */
6231 #endif /* SUPPORT_UTF */
6232 default:
6233 DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
6234 return PCRE_ERROR_INTERNAL;
6236 #undef LBL
6237 #endif /* NO_RECURSE */
6241 /***************************************************************************
6242 ****************************************************************************
6243 RECURSION IN THE match() FUNCTION
6245 Undefine all the macros that were defined above to handle this. */
6247 #ifdef NO_RECURSE
6248 #undef eptr
6249 #undef ecode
6250 #undef mstart
6251 #undef offset_top
6252 #undef eptrb
6253 #undef flags
6255 #undef callpat
6256 #undef charptr
6257 #undef data
6258 #undef next
6259 #undef pp
6260 #undef prev
6261 #undef saved_eptr
6263 #undef new_recursive
6265 #undef cur_is_word
6266 #undef condition
6267 #undef prev_is_word
6269 #undef ctype
6270 #undef length
6271 #undef max
6272 #undef min
6273 #undef number
6274 #undef offset
6275 #undef op
6276 #undef save_capture_last
6277 #undef save_offset1
6278 #undef save_offset2
6279 #undef save_offset3
6280 #undef stacksave
6282 #undef newptrb
6284 #endif
6286 /* These two are defined as macros in both cases */
6288 #undef fc
6289 #undef fi
6291 /***************************************************************************
6292 ***************************************************************************/
6295 #ifdef NO_RECURSE
6296 /*************************************************
6297 * Release allocated heap frames *
6298 *************************************************/
6300 /* This function releases all the allocated frames. The base frame is on the
6301 machine stack, and so must not be freed.
6303 Argument: the address of the base frame
6304 Returns: nothing
6307 static void
6308 release_match_heapframes (heapframe *frame_base)
6310 heapframe *nextframe = frame_base->Xnextframe;
6311 while (nextframe != NULL)
6313 heapframe *oldframe = nextframe;
6314 nextframe = nextframe->Xnextframe;
6315 (PUBL(stack_free))(oldframe);
6318 #endif
6321 /*************************************************
6322 * Execute a Regular Expression *
6323 *************************************************/
6325 /* This function applies a compiled re to a subject string and picks out
6326 portions of the string if it matches. Two elements in the vector are set for
6327 each substring: the offsets to the start and end of the substring.
6329 Arguments:
6330 argument_re points to the compiled expression
6331 extra_data points to extra data or is NULL
6332 subject points to the subject string
6333 length length of subject string (may contain binary zeros)
6334 start_offset where to start in the subject string
6335 options option bits
6336 offsets points to a vector of ints to be filled in with offsets
6337 offsetcount the number of elements in the vector
6339 Returns: > 0 => success; value is the number of elements filled in
6340 = 0 => success, but offsets is not big enough
6341 -1 => failed to match
6342 < -1 => some kind of unexpected problem
6345 #if defined COMPILE_PCRE8
6346 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6347 pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
6348 PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
6349 int offsetcount)
6350 #elif defined COMPILE_PCRE16
6351 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6352 pcre16_exec(const pcre16 *argument_re, const pcre16_extra *extra_data,
6353 PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets,
6354 int offsetcount)
6355 #elif defined COMPILE_PCRE32
6356 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
6357 pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data,
6358 PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets,
6359 int offsetcount)
6360 #endif
6362 int rc, ocount, arg_offset_max;
6363 int newline;
6364 BOOL using_temporary_offsets = FALSE;
6365 BOOL anchored;
6366 BOOL startline;
6367 BOOL firstline;
6368 BOOL utf;
6369 BOOL has_first_char = FALSE;
6370 BOOL has_req_char = FALSE;
6371 pcre_uchar first_char = 0;
6372 pcre_uchar first_char2 = 0;
6373 pcre_uchar req_char = 0;
6374 pcre_uchar req_char2 = 0;
6375 match_data match_block;
6376 match_data *md = &match_block;
6377 const pcre_uint8 *tables;
6378 const pcre_uint8 *start_bits = NULL;
6379 PCRE_PUCHAR start_match = (PCRE_PUCHAR)subject + start_offset;
6380 PCRE_PUCHAR end_subject;
6381 PCRE_PUCHAR start_partial = NULL;
6382 PCRE_PUCHAR match_partial = NULL;
6383 PCRE_PUCHAR req_char_ptr = start_match - 1;
6385 const pcre_study_data *study;
6386 const REAL_PCRE *re = (const REAL_PCRE *)argument_re;
6388 #ifdef NO_RECURSE
6389 heapframe frame_zero;
6390 frame_zero.Xprevframe = NULL; /* Marks the top level */
6391 frame_zero.Xnextframe = NULL; /* None are allocated yet */
6392 md->match_frames_base = &frame_zero;
6393 #endif
6395 /* Check for the special magic call that measures the size of the stack used
6396 per recursive call of match(). Without the funny casting for sizeof, a Windows
6397 compiler gave this error: "unary minus operator applied to unsigned type,
6398 result still unsigned". Hopefully the cast fixes that. */
6400 if (re == NULL && extra_data == NULL && subject == NULL && length == -999 &&
6401 start_offset == -999)
6402 #ifdef NO_RECURSE
6403 return -((int)sizeof(heapframe));
6404 #else
6405 return match(NULL, NULL, NULL, 0, NULL, NULL, 0);
6406 #endif
6408 /* Plausibility checks */
6410 if ((options & ~PUBLIC_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
6411 if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0))
6412 return PCRE_ERROR_NULL;
6413 if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
6414 if (length < 0) return PCRE_ERROR_BADLENGTH;
6415 if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
6417 /* Check that the first field in the block is the magic number. If it is not,
6418 return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to
6419 REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which
6420 means that the pattern is likely compiled with different endianness. */
6422 if (re->magic_number != MAGIC_NUMBER)
6423 return re->magic_number == REVERSED_MAGIC_NUMBER?
6424 PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC;
6425 if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE;
6427 /* These two settings are used in the code for checking a UTF-8 string that
6428 follows immediately afterwards. Other values in the md block are used only
6429 during "normal" pcre_exec() processing, not when the JIT support is in use,
6430 so they are set up later. */
6432 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
6433 utf = md->utf = (re->options & PCRE_UTF8) != 0;
6434 md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
6435 ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
6437 /* Check a UTF-8 string if required. Pass back the character offset and error
6438 code for an invalid string if a results vector is available. */
6440 #ifdef SUPPORT_UTF
6441 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0)
6443 int erroroffset;
6444 int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset);
6445 if (errorcode != 0)
6447 if (offsetcount >= 2)
6449 offsets[0] = erroroffset;
6450 offsets[1] = errorcode;
6452 #if defined COMPILE_PCRE8
6453 return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
6454 PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
6455 #elif defined COMPILE_PCRE16
6456 return (errorcode <= PCRE_UTF16_ERR1 && md->partial > 1)?
6457 PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16;
6458 #elif defined COMPILE_PCRE32
6459 return PCRE_ERROR_BADUTF32;
6460 #endif
6462 #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16
6463 /* Check that a start_offset points to the start of a UTF character. */
6464 if (start_offset > 0 && start_offset < length &&
6465 NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset]))
6466 return PCRE_ERROR_BADUTF8_OFFSET;
6467 #endif
6469 #endif
6471 /* If the pattern was successfully studied with JIT support, run the JIT
6472 executable instead of the rest of this function. Most options must be set at
6473 compile time for the JIT code to be usable. Fallback to the normal code path if
6474 an unsupported flag is set. */
6476 #ifdef SUPPORT_JIT
6477 if (extra_data != NULL
6478 && (extra_data->flags & (PCRE_EXTRA_EXECUTABLE_JIT |
6479 PCRE_EXTRA_TABLES)) == PCRE_EXTRA_EXECUTABLE_JIT
6480 && extra_data->executable_jit != NULL
6481 && (options & ~PUBLIC_JIT_EXEC_OPTIONS) == 0)
6483 rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length,
6484 start_offset, options, offsets, offsetcount);
6486 /* PCRE_ERROR_NULL means that the selected normal or partial matching
6487 mode is not compiled. In this case we simply fallback to interpreter. */
6489 if (rc != PCRE_ERROR_JIT_BADOPTION) return rc;
6491 #endif
6493 /* Carry on with non-JIT matching. This information is for finding all the
6494 numbers associated with a given name, for condition testing. */
6496 md->name_table = (pcre_uchar *)re + re->name_table_offset;
6497 md->name_count = re->name_count;
6498 md->name_entry_size = re->name_entry_size;
6500 /* Fish out the optional data from the extra_data structure, first setting
6501 the default values. */
6503 study = NULL;
6504 md->match_limit = MATCH_LIMIT;
6505 md->match_limit_recursion = MATCH_LIMIT_RECURSION;
6506 md->callout_data = NULL;
6508 /* The table pointer is always in native byte order. */
6510 tables = re->tables;
6512 /* The two limit values override the defaults, whatever their value. */
6514 if (extra_data != NULL)
6516 unsigned long int flags = extra_data->flags;
6517 if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
6518 study = (const pcre_study_data *)extra_data->study_data;
6519 if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
6520 md->match_limit = extra_data->match_limit;
6521 if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
6522 md->match_limit_recursion = extra_data->match_limit_recursion;
6523 if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
6524 md->callout_data = extra_data->callout_data;
6525 if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
6528 /* Limits in the regex override only if they are smaller. */
6530 if ((re->flags & PCRE_MLSET) != 0 && re->limit_match < md->match_limit)
6531 md->match_limit = re->limit_match;
6533 if ((re->flags & PCRE_RLSET) != 0 &&
6534 re->limit_recursion < md->match_limit_recursion)
6535 md->match_limit_recursion = re->limit_recursion;
6537 /* If the exec call supplied NULL for tables, use the inbuilt ones. This
6538 is a feature that makes it possible to save compiled regex and re-use them
6539 in other programs later. */
6541 if (tables == NULL) tables = PRIV(default_tables);
6543 /* Set up other data */
6545 anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
6546 startline = (re->flags & PCRE_STARTLINE) != 0;
6547 firstline = (re->options & PCRE_FIRSTLINE) != 0;
6549 /* The code starts after the real_pcre block and the capture name table. */
6551 md->start_code = (const pcre_uchar *)re + re->name_table_offset +
6552 re->name_count * re->name_entry_size;
6554 md->start_subject = (PCRE_PUCHAR)subject;
6555 md->start_offset = start_offset;
6556 md->end_subject = md->start_subject + length;
6557 end_subject = md->end_subject;
6559 md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
6560 md->use_ucp = (re->options & PCRE_UCP) != 0;
6561 md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6562 md->ignore_skip_arg = 0;
6564 /* Some options are unpacked into BOOL variables in the hope that testing
6565 them will be faster than individual option bits. */
6567 md->notbol = (options & PCRE_NOTBOL) != 0;
6568 md->noteol = (options & PCRE_NOTEOL) != 0;
6569 md->notempty = (options & PCRE_NOTEMPTY) != 0;
6570 md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6572 md->hitend = FALSE;
6573 md->mark = md->nomatch_mark = NULL; /* In case never set */
6575 md->recursive = NULL; /* No recursion at top level */
6576 md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6578 md->lcc = tables + lcc_offset;
6579 md->fcc = tables + fcc_offset;
6580 md->ctypes = tables + ctypes_offset;
6582 /* Handle different \R options. */
6584 switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6586 case 0:
6587 if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
6588 md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
6589 else
6590 #ifdef BSR_ANYCRLF
6591 md->bsr_anycrlf = TRUE;
6592 #else
6593 md->bsr_anycrlf = FALSE;
6594 #endif
6595 break;
6597 case PCRE_BSR_ANYCRLF:
6598 md->bsr_anycrlf = TRUE;
6599 break;
6601 case PCRE_BSR_UNICODE:
6602 md->bsr_anycrlf = FALSE;
6603 break;
6605 default: return PCRE_ERROR_BADNEWLINE;
6608 /* Handle different types of newline. The three bits give eight cases. If
6609 nothing is set at run time, whatever was used at compile time applies. */
6611 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
6612 (pcre_uint32)options) & PCRE_NEWLINE_BITS)
6614 case 0: newline = NEWLINE; break; /* Compile-time default */
6615 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6616 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6617 case PCRE_NEWLINE_CR+
6618 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6619 case PCRE_NEWLINE_ANY: newline = -1; break;
6620 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6621 default: return PCRE_ERROR_BADNEWLINE;
6624 if (newline == -2)
6626 md->nltype = NLTYPE_ANYCRLF;
6628 else if (newline < 0)
6630 md->nltype = NLTYPE_ANY;
6632 else
6634 md->nltype = NLTYPE_FIXED;
6635 if (newline > 255)
6637 md->nllen = 2;
6638 md->nl[0] = (newline >> 8) & 255;
6639 md->nl[1] = newline & 255;
6641 else
6643 md->nllen = 1;
6644 md->nl[0] = newline;
6648 /* Partial matching was originally supported only for a restricted set of
6649 regexes; from release 8.00 there are no restrictions, but the bits are still
6650 defined (though never set). So there's no harm in leaving this code. */
6652 if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6653 return PCRE_ERROR_BADPARTIAL;
6655 /* If the expression has got more back references than the offsets supplied can
6656 hold, we get a temporary chunk of working store to use during the matching.
6657 Otherwise, we can use the vector supplied, rounding down its size to a multiple
6658 of 3. */
6660 ocount = offsetcount - (offsetcount % 3);
6661 arg_offset_max = (2*ocount)/3;
6663 if (re->top_backref > 0 && re->top_backref >= ocount/3)
6665 ocount = re->top_backref * 3 + 3;
6666 md->offset_vector = (int *)(PUBL(malloc))(ocount * sizeof(int));
6667 if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
6668 using_temporary_offsets = TRUE;
6669 DPRINTF(("Got memory to hold back references\n"));
6671 else md->offset_vector = offsets;
6672 md->offset_end = ocount;
6673 md->offset_max = (2*ocount)/3;
6674 md->capture_last = 0;
6676 /* Reset the working variable associated with each extraction. These should
6677 never be used unless previously set, but they get saved and restored, and so we
6678 initialize them to avoid reading uninitialized locations. Also, unset the
6679 offsets for the matched string. This is really just for tidiness with callouts,
6680 in case they inspect these fields. */
6682 if (md->offset_vector != NULL)
6684 register int *iptr = md->offset_vector + ocount;
6685 register int *iend = iptr - re->top_bracket;
6686 if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6687 while (--iptr >= iend) *iptr = -1;
6688 if (offsetcount > 0) md->offset_vector[0] = -1;
6689 if (offsetcount > 1) md->offset_vector[1] = -1;
6692 /* Set up the first character to match, if available. The first_char value is
6693 never set for an anchored regular expression, but the anchoring may be forced
6694 at run time, so we have to test for anchoring. The first char may be unset for
6695 an unanchored pattern, of course. If there's no first char and the pattern was
6696 studied, there may be a bitmap of possible first characters. */
6698 if (!anchored)
6700 if ((re->flags & PCRE_FIRSTSET) != 0)
6702 has_first_char = TRUE;
6703 first_char = first_char2 = (pcre_uchar)(re->first_char);
6704 if ((re->flags & PCRE_FCH_CASELESS) != 0)
6706 first_char2 = TABLE_GET(first_char, md->fcc, first_char);
6707 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
6708 if (utf && first_char > 127)
6709 first_char2 = UCD_OTHERCASE(first_char);
6710 #endif
6713 else
6714 if (!startline && study != NULL &&
6715 (study->flags & PCRE_STUDY_MAPPED) != 0)
6716 start_bits = study->start_bits;
6719 /* For anchored or unanchored matches, there may be a "last known required
6720 character" set. */
6722 if ((re->flags & PCRE_REQCHSET) != 0)
6724 has_req_char = TRUE;
6725 req_char = req_char2 = (pcre_uchar)(re->req_char);
6726 if ((re->flags & PCRE_RCH_CASELESS) != 0)
6728 req_char2 = TABLE_GET(req_char, md->fcc, req_char);
6729 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
6730 if (utf && req_char > 127)
6731 req_char2 = UCD_OTHERCASE(req_char);
6732 #endif
6737 /* ==========================================================================*/
6739 /* Loop for handling unanchored repeated matching attempts; for anchored regexs
6740 the loop runs just once. */
6742 for(;;)
6744 PCRE_PUCHAR save_end_subject = end_subject;
6745 PCRE_PUCHAR new_start_match;
6747 /* If firstline is TRUE, the start of the match is constrained to the first
6748 line of a multiline string. That is, the match must be before or at the first
6749 newline. Implement this by temporarily adjusting end_subject so that we stop
6750 scanning at a newline. If the match fails at the newline, later code breaks
6751 this loop. */
6753 if (firstline)
6755 PCRE_PUCHAR t = start_match;
6756 #ifdef SUPPORT_UTF
6757 if (utf)
6759 while (t < md->end_subject && !IS_NEWLINE(t))
6761 t++;
6762 ACROSSCHAR(t < end_subject, *t, t++);
6765 else
6766 #endif
6767 while (t < md->end_subject && !IS_NEWLINE(t)) t++;
6768 end_subject = t;
6771 /* There are some optimizations that avoid running the match if a known
6772 starting point is not found, or if a known later character is not present.
6773 However, there is an option that disables these, for testing and for ensuring
6774 that all callouts do actually occur. The option can be set in the regex by
6775 (*NO_START_OPT) or passed in match-time options. */
6777 if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0)
6779 /* Advance to a unique first char if there is one. */
6781 if (has_first_char)
6783 pcre_uchar smc;
6785 if (first_char != first_char2)
6786 while (start_match < end_subject &&
6787 (smc = UCHAR21TEST(start_match)) != first_char && smc != first_char2)
6788 start_match++;
6789 else
6790 while (start_match < end_subject && UCHAR21TEST(start_match) != first_char)
6791 start_match++;
6794 /* Or to just after a linebreak for a multiline match */
6796 else if (startline)
6798 if (start_match > md->start_subject + start_offset)
6800 #ifdef SUPPORT_UTF
6801 if (utf)
6803 while (start_match < end_subject && !WAS_NEWLINE(start_match))
6805 start_match++;
6806 ACROSSCHAR(start_match < end_subject, *start_match,
6807 start_match++);
6810 else
6811 #endif
6812 while (start_match < end_subject && !WAS_NEWLINE(start_match))
6813 start_match++;
6815 /* If we have just passed a CR and the newline option is ANY or ANYCRLF,
6816 and we are now at a LF, advance the match position by one more character.
6819 if (start_match[-1] == CHAR_CR &&
6820 (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&
6821 start_match < end_subject &&
6822 UCHAR21TEST(start_match) == CHAR_NL)
6823 start_match++;
6827 /* Or to a non-unique first byte after study */
6829 else if (start_bits != NULL)
6831 while (start_match < end_subject)
6833 register pcre_uint32 c = UCHAR21TEST(start_match);
6834 #ifndef COMPILE_PCRE8
6835 if (c > 255) c = 255;
6836 #endif
6837 if ((start_bits[c/8] & (1 << (c&7))) != 0) break;
6838 start_match++;
6841 } /* Starting optimizations */
6843 /* Restore fudged end_subject */
6845 end_subject = save_end_subject;
6847 /* The following two optimizations are disabled for partial matching or if
6848 disabling is explicitly requested. */
6850 if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0 && !md->partial)
6852 /* If the pattern was studied, a minimum subject length may be set. This is
6853 a lower bound; no actual string of that length may actually match the
6854 pattern. Although the value is, strictly, in characters, we treat it as
6855 bytes to avoid spending too much time in this optimization. */
6857 if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
6858 (pcre_uint32)(end_subject - start_match) < study->minlength)
6860 rc = MATCH_NOMATCH;
6861 break;
6864 /* If req_char is set, we know that that character must appear in the
6865 subject for the match to succeed. If the first character is set, req_char
6866 must be later in the subject; otherwise the test starts at the match point.
6867 This optimization can save a huge amount of backtracking in patterns with
6868 nested unlimited repeats that aren't going to match. Writing separate code
6869 for cased/caseless versions makes it go faster, as does using an
6870 autoincrement and backing off on a match.
6872 HOWEVER: when the subject string is very, very long, searching to its end
6873 can take a long time, and give bad performance on quite ordinary patterns.
6874 This showed up when somebody was matching something like /^\d+C/ on a
6875 32-megabyte string... so we don't do this when the string is sufficiently
6876 long. */
6878 if (has_req_char && end_subject - start_match < REQ_BYTE_MAX)
6880 register PCRE_PUCHAR p = start_match + (has_first_char? 1:0);
6882 /* We don't need to repeat the search if we haven't yet reached the
6883 place we found it at last time. */
6885 if (p > req_char_ptr)
6887 if (req_char != req_char2)
6889 while (p < end_subject)
6891 register pcre_uint32 pp = UCHAR21INCTEST(p);
6892 if (pp == req_char || pp == req_char2) { p--; break; }
6895 else
6897 while (p < end_subject)
6899 if (UCHAR21INCTEST(p) == req_char) { p--; break; }
6903 /* If we can't find the required character, break the matching loop,
6904 forcing a match failure. */
6906 if (p >= end_subject)
6908 rc = MATCH_NOMATCH;
6909 break;
6912 /* If we have found the required character, save the point where we
6913 found it, so that we don't search again next time round the loop if
6914 the start hasn't passed this character yet. */
6916 req_char_ptr = p;
6921 #ifdef PCRE_DEBUG /* Sigh. Some compilers never learn. */
6922 printf(">>>> Match against: ");
6923 pchars(start_match, end_subject - start_match, TRUE, md);
6924 printf("\n");
6925 #endif
6927 /* OK, we can now run the match. If "hitend" is set afterwards, remember the
6928 first starting point for which a partial match was found. */
6930 md->start_match_ptr = start_match;
6931 md->start_used_ptr = start_match;
6932 md->match_call_count = 0;
6933 md->match_function_type = 0;
6934 md->end_offset_top = 0;
6935 md->skip_arg_count = 0;
6936 rc = match(start_match, md->start_code, start_match, 2, md, NULL, 0);
6937 if (md->hitend && start_partial == NULL)
6939 start_partial = md->start_used_ptr;
6940 match_partial = start_match;
6943 switch(rc)
6945 /* If MATCH_SKIP_ARG reaches this level it means that a MARK that matched
6946 the SKIP's arg was not found. In this circumstance, Perl ignores the SKIP
6947 entirely. The only way we can do that is to re-do the match at the same
6948 point, with a flag to force SKIP with an argument to be ignored. Just
6949 treating this case as NOMATCH does not work because it does not check other
6950 alternatives in patterns such as A(*SKIP:A)B|AC when the subject is AC. */
6952 case MATCH_SKIP_ARG:
6953 new_start_match = start_match;
6954 md->ignore_skip_arg = md->skip_arg_count;
6955 break;
6957 /* SKIP passes back the next starting point explicitly, but if it is no
6958 greater than the match we have just done, treat it as NOMATCH. */
6960 case MATCH_SKIP:
6961 if (md->start_match_ptr > start_match)
6963 new_start_match = md->start_match_ptr;
6964 break;
6966 /* Fall through */
6968 /* NOMATCH and PRUNE advance by one character. THEN at this level acts
6969 exactly like PRUNE. Unset ignore SKIP-with-argument. */
6971 case MATCH_NOMATCH:
6972 case MATCH_PRUNE:
6973 case MATCH_THEN:
6974 md->ignore_skip_arg = 0;
6975 new_start_match = start_match + 1;
6976 #ifdef SUPPORT_UTF
6977 if (utf)
6978 ACROSSCHAR(new_start_match < end_subject, *new_start_match,
6979 new_start_match++);
6980 #endif
6981 break;
6983 /* COMMIT disables the bumpalong, but otherwise behaves as NOMATCH. */
6985 case MATCH_COMMIT:
6986 rc = MATCH_NOMATCH;
6987 goto ENDLOOP;
6989 /* Any other return is either a match, or some kind of error. */
6991 default:
6992 goto ENDLOOP;
6995 /* Control reaches here for the various types of "no match at this point"
6996 result. Reset the code to MATCH_NOMATCH for subsequent checking. */
6998 rc = MATCH_NOMATCH;
7000 /* If PCRE_FIRSTLINE is set, the match must happen before or at the first
7001 newline in the subject (though it may continue over the newline). Therefore,
7002 if we have just failed to match, starting at a newline, do not continue. */
7004 if (firstline && IS_NEWLINE(start_match)) break;
7006 /* Advance to new matching position */
7008 start_match = new_start_match;
7010 /* Break the loop if the pattern is anchored or if we have passed the end of
7011 the subject. */
7013 if (anchored || start_match > end_subject) break;
7015 /* If we have just passed a CR and we are now at a LF, and the pattern does
7016 not contain any explicit matches for \r or \n, and the newline option is CRLF
7017 or ANY or ANYCRLF, advance the match position by one more character. In
7018 normal matching start_match will aways be greater than the first position at
7019 this stage, but a failed *SKIP can cause a return at the same point, which is
7020 why the first test exists. */
7022 if (start_match > (PCRE_PUCHAR)subject + start_offset &&
7023 start_match[-1] == CHAR_CR &&
7024 start_match < end_subject &&
7025 *start_match == CHAR_NL &&
7026 (re->flags & PCRE_HASCRORLF) == 0 &&
7027 (md->nltype == NLTYPE_ANY ||
7028 md->nltype == NLTYPE_ANYCRLF ||
7029 md->nllen == 2))
7030 start_match++;
7032 md->mark = NULL; /* Reset for start of next match attempt */
7033 } /* End of for(;;) "bumpalong" loop */
7035 /* ==========================================================================*/
7037 /* We reach here when rc is not MATCH_NOMATCH, or if one of the stopping
7038 conditions is true:
7040 (1) The pattern is anchored or the match was failed by (*COMMIT);
7042 (2) We are past the end of the subject;
7044 (3) PCRE_FIRSTLINE is set and we have failed to match at a newline, because
7045 this option requests that a match occur at or before the first newline in
7046 the subject.
7048 When we have a match and the offset vector is big enough to deal with any
7049 backreferences, captured substring offsets will already be set up. In the case
7050 where we had to get some local store to hold offsets for backreference
7051 processing, copy those that we can. In this case there need not be overflow if
7052 certain parts of the pattern were not used, even though there are more
7053 capturing parentheses than vector slots. */
7055 ENDLOOP:
7057 if (rc == MATCH_MATCH || rc == MATCH_ACCEPT)
7059 if (using_temporary_offsets)
7061 if (arg_offset_max >= 4)
7063 memcpy(offsets + 2, md->offset_vector + 2,
7064 (arg_offset_max - 2) * sizeof(int));
7065 DPRINTF(("Copied offsets from temporary memory\n"));
7067 if (md->end_offset_top > arg_offset_max) md->capture_last |= OVFLBIT;
7068 DPRINTF(("Freeing temporary memory\n"));
7069 (PUBL(free))(md->offset_vector);
7072 /* Set the return code to the number of captured strings, or 0 if there were
7073 too many to fit into the vector. */
7075 rc = ((md->capture_last & OVFLBIT) != 0 &&
7076 md->end_offset_top >= arg_offset_max)?
7077 0 : md->end_offset_top/2;
7079 /* If there is space in the offset vector, set any unused pairs at the end of
7080 the pattern to -1 for backwards compatibility. It is documented that this
7081 happens. In earlier versions, the whole set of potential capturing offsets
7082 was set to -1 each time round the loop, but this is handled differently now.
7083 "Gaps" are set to -1 dynamically instead (this fixes a bug). Thus, it is only
7084 those at the end that need unsetting here. We can't just unset them all at
7085 the start of the whole thing because they may get set in one branch that is
7086 not the final matching branch. */
7088 if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
7090 register int *iptr, *iend;
7091 int resetcount = 2 + re->top_bracket * 2;
7092 if (resetcount > offsetcount) resetcount = offsetcount;
7093 iptr = offsets + md->end_offset_top;
7094 iend = offsets + resetcount;
7095 while (iptr < iend) *iptr++ = -1;
7098 /* If there is space, set up the whole thing as substring 0. The value of
7099 md->start_match_ptr might be modified if \K was encountered on the success
7100 matching path. */
7102 if (offsetcount < 2) rc = 0; else
7104 offsets[0] = (int)(md->start_match_ptr - md->start_subject);
7105 offsets[1] = (int)(md->end_match_ptr - md->start_subject);
7108 /* Return MARK data if requested */
7110 if (extra_data != NULL && (extra_data->flags & PCRE_EXTRA_MARK) != 0)
7111 *(extra_data->mark) = (pcre_uchar *)md->mark;
7112 DPRINTF((">>>> returning %d\n", rc));
7113 #ifdef NO_RECURSE
7114 release_match_heapframes(&frame_zero);
7115 #endif
7116 return rc;
7119 /* Control gets here if there has been an error, or if the overall match
7120 attempt has failed at all permitted starting positions. */
7122 if (using_temporary_offsets)
7124 DPRINTF(("Freeing temporary memory\n"));
7125 (PUBL(free))(md->offset_vector);
7128 /* For anything other than nomatch or partial match, just return the code. */
7130 if (rc != MATCH_NOMATCH && rc != PCRE_ERROR_PARTIAL)
7132 DPRINTF((">>>> error: returning %d\n", rc));
7133 #ifdef NO_RECURSE
7134 release_match_heapframes(&frame_zero);
7135 #endif
7136 return rc;
7139 /* Handle partial matches - disable any mark data */
7141 if (match_partial != NULL)
7143 DPRINTF((">>>> returning PCRE_ERROR_PARTIAL\n"));
7144 md->mark = NULL;
7145 if (offsetcount > 1)
7147 offsets[0] = (int)(start_partial - (PCRE_PUCHAR)subject);
7148 offsets[1] = (int)(end_subject - (PCRE_PUCHAR)subject);
7149 if (offsetcount > 2)
7150 offsets[2] = (int)(match_partial - (PCRE_PUCHAR)subject);
7152 rc = PCRE_ERROR_PARTIAL;
7155 /* This is the classic nomatch case */
7157 else
7159 DPRINTF((">>>> returning PCRE_ERROR_NOMATCH\n"));
7160 rc = PCRE_ERROR_NOMATCH;
7163 /* Return the MARK data if it has been requested. */
7165 if (extra_data != NULL && (extra_data->flags & PCRE_EXTRA_MARK) != 0)
7166 *(extra_data->mark) = (pcre_uchar *)md->nomatch_mark;
7167 #ifdef NO_RECURSE
7168 release_match_heapframes(&frame_zero);
7169 #endif
7170 return rc;
7173 /* End of pcre_exec.c */