xref: /PHP-5.6/ext/pcre/pcrelib/pcre_exec.c (revision ca02d9c2)
1 /*************************************************
2 *      Perl-Compatible Regular Expressions       *
3 *************************************************/
4 
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7 
8                        Written by Philip Hazel
9            Copyright (c) 1997-2014 University of Cambridge
10 
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14 
15     * Redistributions of source code must retain the above copyright notice,
16       this list of conditions and the following disclaimer.
17 
18     * Redistributions in binary form must reproduce the above copyright
19       notice, this list of conditions and the following disclaimer in the
20       documentation and/or other materials provided with the distribution.
21 
22     * Neither the name of the University of Cambridge nor the names of its
23       contributors may be used to endorse or promote products derived from
24       this software without specific prior written permission.
25 
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39 
40 /* This module contains pcre_exec(), the externally visible function that does
41 pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42 possible. There are also some static supporting functions. */
43 
44 #ifdef HAVE_CONFIG_H
45 #include "config.h"
46 #endif
47 
48 #define NLBLOCK md             /* Block containing newline information */
49 #define PSSTART start_subject  /* Field containing processed string start */
50 #define PSEND   end_subject    /* Field containing processed string end */
51 
52 #include "pcre_internal.h"
53 
54 /* Undefine some potentially clashing cpp symbols */
55 
56 #undef min
57 #undef max
58 
59 /* The md->capture_last field uses the lower 16 bits for the last captured
60 substring (which can never be greater than 65535) and a bit in the top half
61 to mean "capture vector overflowed". This odd way of doing things was
62 implemented when it was realized that preserving and restoring the overflow bit
63 whenever the last capture number was saved/restored made for a neater
64 interface, and doing it this way saved on (a) another variable, which would
65 have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66 separate set of save/restore instructions. The following defines are used in
67 implementing this. */
68 
69 #define CAPLMASK    0x0000ffff    /* The bits used for last_capture */
70 #define OVFLMASK    0xffff0000    /* The bits used for the overflow flag */
71 #define OVFLBIT     0x00010000    /* The bit that is set for overflow */
72 
73 /* Values for setting in md->match_function_type to indicate two special types
74 of call to match(). We do it this way to save on using another stack variable,
75 as stack usage is to be discouraged. */
76 
77 #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
78 #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
79 
80 /* Non-error returns from the match() function. Error returns are externally
81 defined PCRE_ERROR_xxx codes, which are all negative. */
82 
83 #define MATCH_MATCH        1
84 #define MATCH_NOMATCH      0
85 
86 /* Special internal returns from the match() function. Make them sufficiently
87 negative to avoid the external error codes. */
88 
89 #define MATCH_ACCEPT       (-999)
90 #define MATCH_KETRPOS      (-998)
91 #define MATCH_ONCE         (-997)
92 /* The next 5 must be kept together and in sequence so that a test that checks
93 for any one of them can use a range. */
94 #define MATCH_COMMIT       (-996)
95 #define MATCH_PRUNE        (-995)
96 #define MATCH_SKIP         (-994)
97 #define MATCH_SKIP_ARG     (-993)
98 #define MATCH_THEN         (-992)
99 #define MATCH_BACKTRACK_MAX MATCH_THEN
100 #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101 
102 /* Maximum number of ints of offset to save on the stack for recursive calls.
103 If the offset vector is bigger, malloc is used. This should be a multiple of 3,
104 because the offset vector is always a multiple of 3 long. */
105 
106 #define REC_STACK_SAVE_MAX 30
107 
108 /* Min and max values for the common repeats; for the maxima, 0 => infinity */
109 
110 static const char rep_min[] = { 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, };
111 static const char rep_max[] = { 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, };
112 
113 #ifdef PCRE_DEBUG
114 /*************************************************
115 *        Debugging function to print chars       *
116 *************************************************/
117 
118 /* Print a sequence of chars in printable format, stopping at the end of the
119 subject if the requested.
120 
121 Arguments:
122   p           points to characters
123   length      number to print
124   is_subject  TRUE if printing from within md->start_subject
125   md          pointer to matching data block, if is_subject is TRUE
126 
127 Returns:     nothing
128 */
129 
130 static void
pchars(const pcre_uchar * p,int length,BOOL is_subject,match_data * md)131 pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132 {
133 pcre_uint32 c;
134 BOOL utf = md->utf;
135 if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136 while (length-- > 0)
137   if (isprint(c = UCHAR21INCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138 }
139 #endif
140 
141 
142 
143 /*************************************************
144 *          Match a back-reference                *
145 *************************************************/
146 
147 /* Normally, if a back reference hasn't been set, the length that is passed is
148 negative, so the match always fails. However, in JavaScript compatibility mode,
149 the length passed is zero. Note that in caseless UTF-8 mode, the number of
150 subject bytes matched may be different to the number of reference bytes.
151 
152 Arguments:
153   offset      index into the offset vector
154   eptr        pointer into the subject
155   length      length of reference to be matched (number of bytes)
156   md          points to match data block
157   caseless    TRUE if caseless
158 
159 Returns:      >= 0 the number of subject bytes matched
160               -1 no match
161               -2 partial match; always given if at end subject
162 */
163 
164 static int
match_ref(int offset,register PCRE_PUCHAR eptr,int length,match_data * md,BOOL caseless)165 match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166   BOOL caseless)
167 {
168 PCRE_PUCHAR eptr_start = eptr;
169 register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170 #if defined SUPPORT_UTF && defined SUPPORT_UCP
171 BOOL utf = md->utf;
172 #endif
173 
174 #ifdef PCRE_DEBUG
175 if (eptr >= md->end_subject)
176   printf("matching subject <null>");
177 else
178   {
179   printf("matching subject ");
180   pchars(eptr, length, TRUE, md);
181   }
182 printf(" against backref ");
183 pchars(p, length, FALSE, md);
184 printf("\n");
185 #endif
186 
187 /* Always fail if reference not set (and not JavaScript compatible - in that
188 case the length is passed as zero). */
189 
190 if (length < 0) return -1;
191 
192 /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193 properly if Unicode properties are supported. Otherwise, we can check only
194 ASCII characters. */
195 
196 if (caseless)
197   {
198 #if defined SUPPORT_UTF && defined SUPPORT_UCP
199   if (utf)
200     {
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). */
209 
210     PCRE_PUCHAR endptr = p + length;
211     while (p < endptr)
212       {
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)
220         {
221         const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
222         for (;;)
223           {
224           if (c < *pp) return -1;
225           if (c == *pp++) break;
226           }
227         }
228       }
229     }
230   else
231 #endif
232 
233   /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
234   is no UCP support. */
235     {
236     while (length-- > 0)
237       {
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++;
245       }
246     }
247   }
248 
249 /* In the caseful case, we can just compare the bytes, whether or not we
250 are in UTF-8 mode. */
251 
252 else
253   {
254   while (length-- > 0)
255     {
256     if (eptr >= md->end_subject) return -2;   /* Partial match */
257     if (UCHAR21INCTEST(p) != UCHAR21INCTEST(eptr)) return -1;
258     }
259   }
260 
261 return (int)(eptr - eptr_start);
262 }
263 
264 
265 
266 /***************************************************************************
267 ****************************************************************************
268                    RECURSION IN THE match() FUNCTION
269 
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.
276 
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....
281 
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.
287 
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.
294 
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 ***************************************************************************/
303 
304 /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
305 below must be updated in sync.  */
306 
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 };
314 
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. */
318 
319 #ifndef NO_RECURSE
320 #define REGISTER register
321 
322 #ifdef PCRE_DEBUG
323 #define RMATCH(ra,rb,rc,rd,re,rw) \
324   { \
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__); \
328   }
329 #define RRETURN(ra) \
330   { \
331   printf("match() returned %d from line %d\n", ra, __LINE__); \
332   return ra; \
333   }
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
339 
340 #else
341 
342 
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. */
346 
347 #define REGISTER
348 
349 #define RMATCH(ra,rb,rc,rd,re,rw)\
350   {\
351   heapframe *newframe = frame->Xnextframe;\
352   if (newframe == NULL)\
353     {\
354     newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
355     if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
356     newframe->Xnextframe = NULL;\
357     frame->Xnextframe = newframe;\
358     }\
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__));\
372   }
373 
374 #define RRETURN(ra)\
375   {\
376   heapframe *oldframe = frame;\
377   frame = oldframe->Xprevframe;\
378   if (frame != NULL)\
379     {\
380     rrc = ra;\
381     goto HEAP_RETURN;\
382     }\
383   return ra;\
384   }
385 
386 
387 /* Structure for remembering the local variables in a private frame */
388 
389 typedef struct heapframe {
390   struct heapframe *Xprevframe;
391   struct heapframe *Xnextframe;
392 
393   /* Function arguments that may change */
394 
395   PCRE_PUCHAR Xeptr;
396   const pcre_uchar *Xecode;
397   PCRE_PUCHAR Xmstart;
398   int Xoffset_top;
399   eptrblock *Xeptrb;
400   unsigned int Xrdepth;
401 
402   /* Function local variables */
403 
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;
413 
414   recursion_info Xnew_recursive;
415 
416   BOOL Xcur_is_word;
417   BOOL Xcondition;
418   BOOL Xprev_is_word;
419 
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
427 
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];
441 
442   eptrblock Xnewptrb;
443 
444   /* Where to jump back to */
445 
446   int Xwhere;
447 
448 } heapframe;
449 
450 #endif
451 
452 
453 /***************************************************************************
454 ***************************************************************************/
455 
456 
457 
458 /*************************************************
459 *         Match from current position            *
460 *************************************************/
461 
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. */
465 
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. */
472 
473 #define CHECK_PARTIAL()\
474   if (md->partial != 0 && eptr >= md->end_subject && \
475       eptr > md->start_used_ptr) \
476     { \
477     md->hitend = TRUE; \
478     if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
479     }
480 
481 #define SCHECK_PARTIAL()\
482   if (md->partial != 0 && eptr > md->start_used_ptr) \
483     { \
484     md->hitend = TRUE; \
485     if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
486     }
487 
488 
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.
493 
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
504 
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)
510 */
511 
512 static int
match(REGISTER PCRE_PUCHAR eptr,REGISTER const pcre_uchar * ecode,PCRE_PUCHAR mstart,int offset_top,match_data * md,eptrblock * eptrb,unsigned int rdepth)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)
516 {
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. */
520 
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 */
525 
526 BOOL minimize, possessive; /* Quantifier options */
527 BOOL caseless;
528 int condcode;
529 
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". */
536 
537 #ifdef NO_RECURSE
538 heapframe *frame = (heapframe *)md->match_frames_base;
539 
540 /* Copy in the original argument variables */
541 
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;
548 
549 /* This is where control jumps back to to effect "recursion" */
550 
551 HEAP_RECURSE:
552 
553 /* Macros make the argument variables come from the current frame */
554 
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
561 
562 /* Ditto for the local variables */
563 
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
574 
575 #define new_recursive      frame->Xnew_recursive
576 
577 #define cur_is_word        frame->Xcur_is_word
578 #define condition          frame->Xcondition
579 #define prev_is_word       frame->Xprev_is_word
580 
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
588 
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
603 
604 #define newptrb            frame->Xnewptrb
605 
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. */
609 
610 #else         /* NO_RECURSE not defined */
611 #define fi i
612 #define fc c
613 
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(). */
621 
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;
631 
632 recursion_info new_recursive;
633 
634 BOOL cur_is_word;
635 BOOL condition;
636 BOOL prev_is_word;
637 
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
645 
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];
657 
658 eptrblock newptrb;
659 
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. */
665 
666 if (ecode == NULL)
667   {
668   if (rdepth == 0)
669     return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
670   else
671     {
672     int len = (char *)&rdepth - (char *)eptr;
673     return (len > 0)? -len : len;
674     }
675   }
676 #endif     /* NO_RECURSE */
677 
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. */
682 
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
690 
691 /* These statements are here to stop the compiler complaining about unitialized
692 variables. */
693 
694 #ifdef SUPPORT_UCP
695 prop_value = 0;
696 prop_fail_result = 0;
697 #endif
698 
699 
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. */
704 
705 TAIL_RECURSE:
706 
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. */
714 
715 #ifdef SUPPORT_UTF
716 utf = md->utf;       /* Local copy of the flag */
717 #else
718 utf = FALSE;
719 #endif
720 
721 /* First check that we haven't called match() too many times, or that we
722 haven't exceeded the recursive call limit. */
723 
724 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
725 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
726 
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.
731 
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(). */
738 
739 if (md->match_function_type == MATCH_CBEGROUP)
740   {
741   newptrb.epb_saved_eptr = eptr;
742   newptrb.epb_prev = eptrb;
743   eptrb = &newptrb;
744   md->match_function_type = 0;
745   }
746 
747 /* Now start processing the opcodes. */
748 
749 for (;;)
750   {
751   minimize = possessive = FALSE;
752   op = *ecode;
753 
754   switch(op)
755     {
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;
763 
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. */
770 
771     else if (rrc == MATCH_SKIP_ARG &&
772         STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
773       {
774       md->start_match_ptr = eptr;
775       RRETURN(MATCH_SKIP);
776       }
777     RRETURN(rrc);
778 
779     case OP_FAIL:
780     RRETURN(MATCH_NOMATCH);
781 
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);
787 
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);
793 
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);
803 
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);
810 
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. */
818 
819     case OP_SKIP_ARG:
820     md->skip_arg_count++;
821     if (md->skip_arg_count <= md->ignore_skip_arg)
822       {
823       ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
824       break;
825       }
826     RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
827       eptrb, RM57);
828     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
829 
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. */
834 
835     md->start_match_ptr = ecode + 2;
836     RRETURN(MATCH_SKIP_ARG);
837 
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. */
841 
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);
848 
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);
859 
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.
867 
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. */
872 
873     case OP_ONCE_NC:
874     prev = ecode;
875     saved_eptr = eptr;
876     save_mark = md->mark;
877     do
878       {
879       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
880       if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
881         {
882         mstart = md->start_match_ptr;
883         break;
884         }
885       if (rrc == MATCH_THEN)
886         {
887         next = ecode + GET(ecode,1);
888         if (md->start_match_ptr < next &&
889             (*ecode == OP_ALT || *next == OP_ALT))
890           rrc = MATCH_NOMATCH;
891         }
892 
893       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
894       ecode += GET(ecode,1);
895       md->mark = save_mark;
896       }
897     while (*ecode == OP_ALT);
898 
899     /* If hit the end of the group (which could be repeated), fail */
900 
901     if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
902 
903     /* Continue as from after the group, updating the offsets high water
904     mark, since extracts may have been taken. */
905 
906     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
907 
908     offset_top = md->end_offset_top;
909     eptr = md->end_match_ptr;
910 
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. */
915 
916     if (*ecode == OP_KET || eptr == saved_eptr)
917       {
918       ecode += 1+LINK_SIZE;
919       break;
920       }
921 
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. */
925 
926     if (*ecode == OP_KETRMIN)
927       {
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;
932       }
933     else  /* OP_KETRMAX */
934       {
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;
939       }
940     /* Control never gets here */
941 
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.
951 
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. */
955 
956     case OP_CBRA:
957     case OP_SCBRA:
958     number = GET2(ecode, 1+LINK_SIZE);
959     offset = number << 1;
960 
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
967 
968     if (offset < md->offset_max)
969       {
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;
975 
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);
979 
980       for (;;)
981         {
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 */
986 
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)). */
997 
998         if (rrc == MATCH_THEN)
999           {
1000           next = ecode + GET(ecode,1);
1001           if (md->start_match_ptr < next &&
1002               (*ecode == OP_ALT || *next == OP_ALT))
1003             rrc = MATCH_NOMATCH;
1004           }
1005 
1006         /* Anything other than NOMATCH is passed back. */
1007 
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;
1013         }
1014 
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;
1019 
1020       /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1021 
1022       RRETURN(rrc);
1023       }
1024 
1025     /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1026     as a non-capturing bracket. */
1027 
1028     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1029     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1030 
1031     DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1032 
1033     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1034     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1035 
1036     /* Non-capturing or atomic group, except for possessive with unlimited
1037     repeat and ONCE group with no captures. Loop for all the alternatives.
1038 
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 possibility of (*THEN) occurring in the final alternative, this
1044     optimization is no longer always possible.
1045 
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.
1048 
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. */
1055 
1056     case OP_ONCE:
1057     case OP_BRA:
1058     case OP_SBRA:
1059     DPRINTF(("start non-capturing bracket\n"));
1060 
1061     for (;;)
1062       {
1063       if (op >= OP_SBRA || op == OP_ONCE)
1064         md->match_function_type = MATCH_CBEGROUP;
1065 
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. */
1069 
1070       else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1071         {
1072         ecode += PRIV(OP_lengths)[*ecode];
1073         goto TAIL_RECURSE;
1074         }
1075 
1076       /* In all other cases, we have to make another call to match(). */
1077 
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);
1082 
1083       /* See comment in the code for capturing groups above about handling
1084       THEN. */
1085 
1086       if (rrc == MATCH_THEN)
1087         {
1088         next = ecode + GET(ecode,1);
1089         if (md->start_match_ptr < next &&
1090             (*ecode == OP_ALT || *next == OP_ALT))
1091           rrc = MATCH_NOMATCH;
1092         }
1093 
1094       if (rrc != MATCH_NOMATCH)
1095         {
1096         if (rrc == MATCH_ONCE)
1097           {
1098           const pcre_uchar *scode = ecode;
1099           if (*scode != OP_ONCE)           /* If not at start, find it */
1100             {
1101             while (*scode == OP_ALT) scode += GET(scode, 1);
1102             scode -= GET(scode, 1);
1103             }
1104           if (md->once_target == scode) rrc = MATCH_NOMATCH;
1105           }
1106         RRETURN(rrc);
1107         }
1108       ecode += GET(ecode, 1);
1109       md->mark = save_mark;
1110       if (*ecode != OP_ALT) break;
1111       md->capture_last = save_capture_last;
1112       }
1113 
1114     RRETURN(MATCH_NOMATCH);
1115 
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. */
1123 
1124     case OP_CBRAPOS:
1125     case OP_SCBRAPOS:
1126     allow_zero = FALSE;
1127 
1128     POSSESSIVE_CAPTURE:
1129     number = GET2(ecode, 1+LINK_SIZE);
1130     offset = number << 1;
1131 
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
1138 
1139     if (offset >= md->offset_max) goto POSSESSIVE_NON_CAPTURE;
1140 
1141     matched_once = FALSE;
1142     code_offset = (int)(ecode - md->start_code);
1143 
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;
1148 
1149     DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1150 
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. */
1159 
1160     for (;;)
1161       {
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)
1168         {
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 */
1175           {
1176           do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
1177           break;
1178           }
1179         eptr = md->end_match_ptr;
1180         continue;
1181         }
1182 
1183       /* See comment in the code for capturing groups above about handling
1184       THEN. */
1185 
1186       if (rrc == MATCH_THEN)
1187         {
1188         next = ecode + GET(ecode,1);
1189         if (md->start_match_ptr < next &&
1190             (*ecode == OP_ALT || *next == OP_ALT))
1191           rrc = MATCH_NOMATCH;
1192         }
1193 
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;
1198       }
1199 
1200     if (!matched_once)
1201       {
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;
1205       }
1206 
1207     if (allow_zero || matched_once)
1208       {
1209       ecode += 1 + LINK_SIZE;
1210       break;
1211       }
1212 
1213     RRETURN(MATCH_NOMATCH);
1214 
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. */
1219 
1220     case OP_BRAPOS:
1221     case OP_SBRAPOS:
1222     allow_zero = FALSE;
1223 
1224     POSSESSIVE_NON_CAPTURE:
1225     matched_once = FALSE;
1226     code_offset = (int)(ecode - md->start_code);
1227     save_capture_last = md->capture_last;
1228 
1229     for (;;)
1230       {
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)
1235         {
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 */
1241           {
1242           do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
1243           break;
1244           }
1245         eptr = md->end_match_ptr;
1246         continue;
1247         }
1248 
1249       /* See comment in the code for capturing groups above about handling
1250       THEN. */
1251 
1252       if (rrc == MATCH_THEN)
1253         {
1254         next = ecode + GET(ecode,1);
1255         if (md->start_match_ptr < next &&
1256             (*ecode == OP_ALT || *next == OP_ALT))
1257           rrc = MATCH_NOMATCH;
1258         }
1259 
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;
1264       }
1265 
1266     if (matched_once || allow_zero)
1267       {
1268       ecode += 1 + LINK_SIZE;
1269       break;
1270       }
1271     RRETURN(MATCH_NOMATCH);
1272 
1273     /* Control never reaches here. */
1274 
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. */
1279 
1280     case OP_COND:
1281     case OP_SCOND:
1282 
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. */
1287 
1288     codelink = GET(ecode, 1);   /* Offset to the second branch */
1289     ecode += 1 + LINK_SIZE;     /* From this opcode */
1290 
1291     /* Because of the way auto-callout works during compile, a callout item is
1292     inserted between OP_COND and an assertion condition. */
1293 
1294     if (*ecode == OP_CALLOUT)
1295       {
1296       if (PUBL(callout) != NULL)
1297         {
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);
1322         }
1323 
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. */
1326 
1327       ecode += PRIV(OP_lengths)[OP_CALLOUT];
1328       codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1329       }
1330 
1331     /* Test the various possible conditions */
1332 
1333     condition = FALSE;
1334     switch(condcode = *ecode)
1335       {
1336       case OP_RREF:         /* Numbered group recursion test */
1337       if (md->recursive != NULL)     /* Not recursing => FALSE */
1338         {
1339         unsigned int recno = GET2(ecode, 1);   /* Recursion group number*/
1340         condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1341         }
1342       break;
1343 
1344       case OP_DNRREF:       /* Duplicate named group recursion test */
1345       if (md->recursive != NULL)
1346         {
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)
1350           {
1351           unsigned int recno = GET2(slot, 0);
1352           condition = recno == md->recursive->group_num;
1353           if (condition) break;
1354           slot += md->name_entry_size;
1355           }
1356         }
1357       break;
1358 
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;
1363 
1364       case OP_DNCREF:      /* Duplicate named group used test */
1365         {
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)
1369           {
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;
1374           }
1375         }
1376       break;
1377 
1378       case OP_DEF:     /* DEFINE - always false */
1379       case OP_FAIL:    /* From optimized (?!) condition */
1380       break;
1381 
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. */
1385 
1386       default:
1387       md->match_function_type = MATCH_CONDASSERT;
1388       RMATCH(eptr, ecode, offset_top, md, NULL, RM3);
1389       if (rrc == MATCH_MATCH)
1390         {
1391         if (md->end_offset_top > offset_top)
1392           offset_top = md->end_offset_top;  /* Captures may have happened */
1393         condition = TRUE;
1394 
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! */
1399 
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];
1404         }
1405 
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. */
1409 
1410       else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1411         {
1412         RRETURN(rrc);         /* Need braces because of following else */
1413         }
1414       break;
1415       }
1416 
1417     /* Choose branch according to the condition */
1418 
1419     ecode += condition? PRIV(OP_lengths)[condcode] : codelink;
1420 
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). */
1430 
1431     if (condition || ecode[-(1+LINK_SIZE)] == OP_ALT)
1432       {
1433       if (op != OP_SCOND)
1434         {
1435         goto TAIL_RECURSE;
1436         }
1437 
1438       md->match_function_type = MATCH_CBEGROUP;
1439       RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);
1440       RRETURN(rrc);
1441       }
1442 
1443      /* Condition false & no alternative; continue after the group. */
1444 
1445     else
1446       {
1447       }
1448     break;
1449 
1450 
1451     /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1452     to close any currently open capturing brackets. */
1453 
1454     case OP_CLOSE:
1455     number = GET2(ecode, 1);   /* Must be less than 65536 */
1456     offset = number << 1;
1457 
1458 #ifdef PCRE_DEBUG
1459       printf("end bracket %d at *ACCEPT", number);
1460       printf("\n");
1461 #endif
1462 
1463     md->capture_last = (md->capture_last & OVFLMASK) | number;
1464     if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1465       {
1466       md->offset_vector[offset] =
1467         md->offset_vector[md->offset_end - number];
1468       md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1469 
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. */
1473 
1474       if (offset >= offset_top)
1475         {
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;
1480         }
1481       }
1482     ecode += 1 + IMM2_SIZE;
1483     break;
1484 
1485 
1486     /* End of the pattern, either real or forced. */
1487 
1488     case OP_END:
1489     case OP_ACCEPT:
1490     case OP_ASSERT_ACCEPT:
1491 
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. */
1496 
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);
1503 
1504     /* Otherwise, we have a match. */
1505 
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) */
1509 
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. */
1512 
1513     rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1514     RRETURN(rrc);
1515 
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. */
1526 
1527     case OP_ASSERT:
1528     case OP_ASSERTBACK:
1529     save_mark = md->mark;
1530     if (md->match_function_type == MATCH_CONDASSERT)
1531       {
1532       condassert = TRUE;
1533       md->match_function_type = 0;
1534       }
1535     else condassert = FALSE;
1536 
1537     /* Loop for each branch */
1538 
1539     do
1540       {
1541       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1542 
1543       /* A match means that the assertion is true; break out of the loop
1544       that matches its alternatives. */
1545 
1546       if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1547         {
1548         mstart = md->start_match_ptr;   /* In case \K reset it */
1549         break;
1550         }
1551 
1552       /* If not matched, restore the previous mark setting. */
1553 
1554       md->mark = save_mark;
1555 
1556       /* See comment in the code for capturing groups above about handling
1557       THEN. */
1558 
1559       if (rrc == MATCH_THEN)
1560         {
1561         next = ecode + GET(ecode,1);
1562         if (md->start_match_ptr < next &&
1563             (*ecode == OP_ALT || *next == OP_ALT))
1564           rrc = MATCH_NOMATCH;
1565         }
1566 
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. */
1572 
1573       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1574       ecode += GET(ecode, 1);
1575       }
1576     while (*ecode == OP_ALT);   /* Continue for next alternative */
1577 
1578     /* If we have tried all the alternative branches, the assertion has
1579     failed. If not, we broke out after a match. */
1580 
1581     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1582 
1583     /* If checking an assertion for a condition, return MATCH_MATCH. */
1584 
1585     if (condassert) RRETURN(MATCH_MATCH);
1586 
1587     /* Continue from after a successful assertion, updating the offsets high
1588     water mark, since extracts may have been taken during the assertion. */
1589 
1590     do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1591     ecode += 1 + LINK_SIZE;
1592     offset_top = md->end_offset_top;
1593     continue;
1594 
1595     /* Negative assertion: all branches must fail to match for the assertion to
1596     succeed. */
1597 
1598     case OP_ASSERT_NOT:
1599     case OP_ASSERTBACK_NOT:
1600     save_mark = md->mark;
1601     if (md->match_function_type == MATCH_CONDASSERT)
1602       {
1603       condassert = TRUE;
1604       md->match_function_type = 0;
1605       }
1606     else condassert = FALSE;
1607 
1608     /* Loop for each alternative branch. */
1609 
1610     do
1611       {
1612       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1613       md->mark = save_mark;   /* Always restore the mark setting */
1614 
1615       switch(rrc)
1616         {
1617         case MATCH_MATCH:            /* A successful match means */
1618         case MATCH_ACCEPT:           /* the assertion has failed. */
1619         RRETURN(MATCH_NOMATCH);
1620 
1621         case MATCH_NOMATCH:          /* Carry on with next branch */
1622         break;
1623 
1624         /* See comment in the code for capturing groups above about handling
1625         THEN. */
1626 
1627         case MATCH_THEN:
1628         next = ecode + GET(ecode,1);
1629         if (md->start_match_ptr < next &&
1630             (*ecode == OP_ALT || *next == OP_ALT))
1631           {
1632           rrc = MATCH_NOMATCH;
1633           break;
1634           }
1635         /* Otherwise fall through. */
1636 
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. */
1641 
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 */
1648 
1649         /* Anything else is an error */
1650 
1651         default:
1652         RRETURN(rrc);
1653         }
1654 
1655       /* Continue with next branch */
1656 
1657       ecode += GET(ecode,1);
1658       }
1659     while (*ecode == OP_ALT);
1660 
1661     /* All branches in the assertion failed to match. */
1662 
1663     NEG_ASSERT_TRUE:
1664     if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1665     ecode += 1 + LINK_SIZE;                /* Continue with current branch */
1666     continue;
1667 
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. */
1672 
1673     case OP_REVERSE:
1674 #ifdef SUPPORT_UTF
1675     if (utf)
1676       {
1677       i = GET(ecode, 1);
1678       while (i-- > 0)
1679         {
1680         eptr--;
1681         if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1682         BACKCHAR(eptr);
1683         }
1684       }
1685     else
1686 #endif
1687 
1688     /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1689 
1690       {
1691       eptr -= GET(ecode, 1);
1692       if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1693       }
1694 
1695     /* Save the earliest consulted character, then skip to next op code */
1696 
1697     if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1698     ecode += 1 + LINK_SIZE;
1699     break;
1700 
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. */
1704 
1705     case OP_CALLOUT:
1706     if (PUBL(callout) != NULL)
1707       {
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);
1732       }
1733     ecode += 2 + 2*LINK_SIZE;
1734     break;
1735 
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.)
1739 
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.
1747 
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. */
1752 
1753     case OP_RECURSE:
1754       {
1755       recursion_info *ri;
1756       unsigned int recno;
1757 
1758       callpat = md->start_code + GET(ecode, 1);
1759       recno = (callpat == md->start_code)? 0 :
1760         GET2(callpat, 1 + LINK_SIZE);
1761 
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.) */
1765 
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);
1769 
1770       /* Add to "recursing stack" */
1771 
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;
1777 
1778       /* Where to continue from afterwards */
1779 
1780       ecode += 1 + LINK_SIZE;
1781 
1782       /* Now save the offset data */
1783 
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
1788         {
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);
1792         }
1793       memcpy(new_recursive.offset_save, md->offset_vector,
1794             new_recursive.saved_max * sizeof(int));
1795 
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.
1799       */
1800 
1801       DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1802       cbegroup = (*callpat >= OP_SBRA);
1803       do
1804         {
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)
1813           {
1814           DPRINTF(("Recursion matched\n"));
1815           if (new_recursive.offset_save != stacksave)
1816             (PUBL(free))(new_recursive.offset_save);
1817 
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. */
1821 
1822           eptr = md->end_match_ptr;
1823           mstart = md->start_match_ptr;
1824           goto RECURSION_MATCHED;        /* Exit loop; end processing */
1825           }
1826 
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. */
1830 
1831         if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1832           {
1833           if (new_recursive.offset_save != stacksave)
1834             (PUBL(free))(new_recursive.offset_save);
1835           RRETURN(MATCH_NOMATCH);
1836           }
1837 
1838         /* Any return code other than NOMATCH is an error. */
1839 
1840         if (rrc != MATCH_NOMATCH)
1841           {
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);
1846           }
1847 
1848         md->recursive = &new_recursive;
1849         callpat += GET(callpat, 1);
1850         }
1851       while (*callpat == OP_ALT);
1852 
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);
1858       }
1859 
1860     RECURSION_MATCHED:
1861     break;
1862 
1863     /* An alternation is the end of a branch; scan along to find the end of the
1864     bracketed group and go to there. */
1865 
1866     case OP_ALT:
1867     do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1868     break;
1869 
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. */
1875 
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;
1883 
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;
1891 
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;
1897 
1898     /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1899     here; just jump to the group, with allow_zero set TRUE. */
1900 
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;
1906 
1907     /* End of a group, repeated or non-repeating. */
1908 
1909     case OP_KET:
1910     case OP_KETRMIN:
1911     case OP_KETRMAX:
1912     case OP_KETRPOS:
1913     prev = ecode - GET(ecode, 1);
1914 
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. */
1918 
1919     if (*prev >= OP_SBRA || *prev == OP_ONCE)
1920       {
1921       saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1922       eptrb = eptrb->epb_prev;              /* Backup to previous group */
1923       }
1924     else saved_eptr = NULL;
1925 
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. */
1930 
1931     if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1932          *prev == OP_ONCE_NC)
1933       {
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 */
1938       }
1939 
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. */
1947 
1948     if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1949         *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1950       {
1951       number = GET2(prev, 1+LINK_SIZE);
1952       offset = number << 1;
1953 
1954 #ifdef PCRE_DEBUG
1955       printf("end bracket %d", number);
1956       printf("\n");
1957 #endif
1958 
1959       /* Handle a recursively called group. */
1960 
1961       if (md->recursive != NULL && md->recursive->group_num == number)
1962         {
1963         md->end_match_ptr = eptr;
1964         md->start_match_ptr = mstart;
1965         RRETURN(MATCH_MATCH);
1966         }
1967 
1968       /* Deal with capturing */
1969 
1970       md->capture_last = (md->capture_last & OVFLMASK) | number;
1971       if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1972         {
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.) */
1982 
1983         if (offset > offset_top)
1984           {
1985           register int *iptr = md->offset_vector + offset_top;
1986           register int *iend = md->offset_vector + offset;
1987           while (iptr < iend) *iptr++ = -1;
1988           }
1989 
1990         /* Now make the extraction */
1991 
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;
1996         }
1997       }
1998 
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. */
2003 
2004     if (*ecode == OP_KETRPOS)
2005       {
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);
2010       }
2011 
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. */
2020 
2021     if (*ecode == OP_KET || eptr == saved_eptr)
2022       {
2023       if (*prev == OP_ONCE)
2024         {
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);
2029         }
2030       ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2031       break;
2032       }
2033 
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. */
2039 
2040     if (*ecode == OP_KETRMIN)
2041       {
2042       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2043       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2044       if (*prev == OP_ONCE)
2045         {
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);
2050         }
2051       if (*prev >= OP_SBRA)    /* Could match an empty string */
2052         {
2053         RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2054         RRETURN(rrc);
2055         }
2056       ecode = prev;
2057       goto TAIL_RECURSE;
2058       }
2059     else  /* OP_KETRMAX */
2060       {
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)
2065         {
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);
2070         }
2071       ecode += 1 + LINK_SIZE;
2072       goto TAIL_RECURSE;
2073       }
2074     /* Control never gets here */
2075 
2076     /* Not multiline mode: start of subject assertion, unless notbol. */
2077 
2078     case OP_CIRC:
2079     if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2080 
2081     /* Start of subject assertion */
2082 
2083     case OP_SOD:
2084     if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
2085     ecode++;
2086     break;
2087 
2088     /* Multiline mode: start of subject unless notbol, or after any newline. */
2089 
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;
2097 
2098     /* Start of match assertion */
2099 
2100     case OP_SOM:
2101     if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
2102     ecode++;
2103     break;
2104 
2105     /* Reset the start of match point */
2106 
2107     case OP_SET_SOM:
2108     mstart = eptr;
2109     ecode++;
2110     break;
2111 
2112     /* Multiline mode: assert before any newline, or before end of subject
2113     unless noteol is set. */
2114 
2115     case OP_DOLLM:
2116     if (eptr < md->end_subject)
2117       {
2118       if (!IS_NEWLINE(eptr))
2119         {
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])
2125           {
2126           md->hitend = TRUE;
2127           if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2128           }
2129         RRETURN(MATCH_NOMATCH);
2130         }
2131       }
2132     else
2133       {
2134       if (md->noteol) RRETURN(MATCH_NOMATCH);
2135       SCHECK_PARTIAL();
2136       }
2137     ecode++;
2138     break;
2139 
2140     /* Not multiline mode: assert before a terminating newline or before end of
2141     subject unless noteol is set. */
2142 
2143     case OP_DOLL:
2144     if (md->noteol) RRETURN(MATCH_NOMATCH);
2145     if (!md->endonly) goto ASSERT_NL_OR_EOS;
2146 
2147     /* ... else fall through for endonly */
2148 
2149     /* End of subject assertion (\z) */
2150 
2151     case OP_EOD:
2152     if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2153     SCHECK_PARTIAL();
2154     ecode++;
2155     break;
2156 
2157     /* End of subject or ending \n assertion (\Z) */
2158 
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))
2163       {
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])
2169         {
2170         md->hitend = TRUE;
2171         if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2172         }
2173       RRETURN(MATCH_NOMATCH);
2174       }
2175 
2176     /* Either at end of string or \n before end. */
2177 
2178     SCHECK_PARTIAL();
2179     ecode++;
2180     break;
2181 
2182     /* Word boundary assertions */
2183 
2184     case OP_NOT_WORD_BOUNDARY:
2185     case OP_WORD_BOUNDARY:
2186       {
2187 
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. */
2192 
2193 #ifdef SUPPORT_UTF
2194       if (utf)
2195         {
2196         /* Get status of previous character */
2197 
2198         if (eptr == md->start_subject) prev_is_word = FALSE; else
2199           {
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)
2206             {
2207             if (c == '_') prev_is_word = TRUE; else
2208               {
2209               int cat = UCD_CATEGORY(c);
2210               prev_is_word = (cat == ucp_L || cat == ucp_N);
2211               }
2212             }
2213           else
2214 #endif
2215           prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2216           }
2217 
2218         /* Get status of next character */
2219 
2220         if (eptr >= md->end_subject)
2221           {
2222           SCHECK_PARTIAL();
2223           cur_is_word = FALSE;
2224           }
2225         else
2226           {
2227           GETCHAR(c, eptr);
2228 #ifdef SUPPORT_UCP
2229           if (md->use_ucp)
2230             {
2231             if (c == '_') cur_is_word = TRUE; else
2232               {
2233               int cat = UCD_CATEGORY(c);
2234               cur_is_word = (cat == ucp_L || cat == ucp_N);
2235               }
2236             }
2237           else
2238 #endif
2239           cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2240           }
2241         }
2242       else
2243 #endif
2244 
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. */
2247 
2248         {
2249         /* Get status of previous character */
2250 
2251         if (eptr == md->start_subject) prev_is_word = FALSE; else
2252           {
2253           if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2254 #ifdef SUPPORT_UCP
2255           if (md->use_ucp)
2256             {
2257             c = eptr[-1];
2258             if (c == '_') prev_is_word = TRUE; else
2259               {
2260               int cat = UCD_CATEGORY(c);
2261               prev_is_word = (cat == ucp_L || cat == ucp_N);
2262               }
2263             }
2264           else
2265 #endif
2266           prev_is_word = MAX_255(eptr[-1])
2267             && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2268           }
2269 
2270         /* Get status of next character */
2271 
2272         if (eptr >= md->end_subject)
2273           {
2274           SCHECK_PARTIAL();
2275           cur_is_word = FALSE;
2276           }
2277         else
2278 #ifdef SUPPORT_UCP
2279         if (md->use_ucp)
2280           {
2281           c = *eptr;
2282           if (c == '_') cur_is_word = TRUE; else
2283             {
2284             int cat = UCD_CATEGORY(c);
2285             cur_is_word = (cat == ucp_L || cat == ucp_N);
2286             }
2287           }
2288         else
2289 #endif
2290         cur_is_word = MAX_255(*eptr)
2291           && ((md->ctypes[*eptr] & ctype_word) != 0);
2292         }
2293 
2294       /* Now see if the situation is what we want */
2295 
2296       if ((*ecode++ == OP_WORD_BOUNDARY)?
2297            cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2298         RRETURN(MATCH_NOMATCH);
2299       }
2300     break;
2301 
2302     /* Match any single character type except newline; have to take care with
2303     CRLF newlines and partial matching. */
2304 
2305     case OP_ANY:
2306     if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2307     if (md->partial != 0 &&
2308         eptr + 1 >= md->end_subject &&
2309         NLBLOCK->nltype == NLTYPE_FIXED &&
2310         NLBLOCK->nllen == 2 &&
2311         UCHAR21TEST(eptr) == NLBLOCK->nl[0])
2312       {
2313       md->hitend = TRUE;
2314       if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2315       }
2316 
2317     /* Fall through */
2318 
2319     /* Match any single character whatsoever. */
2320 
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);
2326       }
2327     eptr++;
2328 #ifdef SUPPORT_UTF
2329     if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2330 #endif
2331     ecode++;
2332     break;
2333 
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. */
2336 
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);
2342       }
2343     eptr++;
2344     ecode++;
2345     break;
2346 
2347     case OP_NOT_DIGIT:
2348     if (eptr >= md->end_subject)
2349       {
2350       SCHECK_PARTIAL();
2351       RRETURN(MATCH_NOMATCH);
2352       }
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
2359        )
2360       RRETURN(MATCH_NOMATCH);
2361     ecode++;
2362     break;
2363 
2364     case OP_DIGIT:
2365     if (eptr >= md->end_subject)
2366       {
2367       SCHECK_PARTIAL();
2368       RRETURN(MATCH_NOMATCH);
2369       }
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
2376        )
2377       RRETURN(MATCH_NOMATCH);
2378     ecode++;
2379     break;
2380 
2381     case OP_NOT_WHITESPACE:
2382     if (eptr >= md->end_subject)
2383       {
2384       SCHECK_PARTIAL();
2385       RRETURN(MATCH_NOMATCH);
2386       }
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
2393        )
2394       RRETURN(MATCH_NOMATCH);
2395     ecode++;
2396     break;
2397 
2398     case OP_WHITESPACE:
2399     if (eptr >= md->end_subject)
2400       {
2401       SCHECK_PARTIAL();
2402       RRETURN(MATCH_NOMATCH);
2403       }
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
2410        )
2411       RRETURN(MATCH_NOMATCH);
2412     ecode++;
2413     break;
2414 
2415     case OP_NOT_WORDCHAR:
2416     if (eptr >= md->end_subject)
2417       {
2418       SCHECK_PARTIAL();
2419       RRETURN(MATCH_NOMATCH);
2420       }
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
2427        )
2428       RRETURN(MATCH_NOMATCH);
2429     ecode++;
2430     break;
2431 
2432     case OP_WORDCHAR:
2433     if (eptr >= md->end_subject)
2434       {
2435       SCHECK_PARTIAL();
2436       RRETURN(MATCH_NOMATCH);
2437       }
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
2444        )
2445       RRETURN(MATCH_NOMATCH);
2446     ecode++;
2447     break;
2448 
2449     case OP_ANYNL:
2450     if (eptr >= md->end_subject)
2451       {
2452       SCHECK_PARTIAL();
2453       RRETURN(MATCH_NOMATCH);
2454       }
2455     GETCHARINCTEST(c, eptr);
2456     switch(c)
2457       {
2458       default: RRETURN(MATCH_NOMATCH);
2459 
2460       case CHAR_CR:
2461       if (eptr >= md->end_subject)
2462         {
2463         SCHECK_PARTIAL();
2464         }
2465       else if (UCHAR21TEST(eptr) == CHAR_LF) eptr++;
2466       break;
2467 
2468       case CHAR_LF:
2469       break;
2470 
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;
2480       }
2481     ecode++;
2482     break;
2483 
2484     case OP_NOT_HSPACE:
2485     if (eptr >= md->end_subject)
2486       {
2487       SCHECK_PARTIAL();
2488       RRETURN(MATCH_NOMATCH);
2489       }
2490     GETCHARINCTEST(c, eptr);
2491     switch(c)
2492       {
2493       HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2494       default: break;
2495       }
2496     ecode++;
2497     break;
2498 
2499     case OP_HSPACE:
2500     if (eptr >= md->end_subject)
2501       {
2502       SCHECK_PARTIAL();
2503       RRETURN(MATCH_NOMATCH);
2504       }
2505     GETCHARINCTEST(c, eptr);
2506     switch(c)
2507       {
2508       HSPACE_CASES: break;  /* Byte and multibyte cases */
2509       default: RRETURN(MATCH_NOMATCH);
2510       }
2511     ecode++;
2512     break;
2513 
2514     case OP_NOT_VSPACE:
2515     if (eptr >= md->end_subject)
2516       {
2517       SCHECK_PARTIAL();
2518       RRETURN(MATCH_NOMATCH);
2519       }
2520     GETCHARINCTEST(c, eptr);
2521     switch(c)
2522       {
2523       VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2524       default: break;
2525       }
2526     ecode++;
2527     break;
2528 
2529     case OP_VSPACE:
2530     if (eptr >= md->end_subject)
2531       {
2532       SCHECK_PARTIAL();
2533       RRETURN(MATCH_NOMATCH);
2534       }
2535     GETCHARINCTEST(c, eptr);
2536     switch(c)
2537       {
2538       VSPACE_CASES: break;
2539       default: RRETURN(MATCH_NOMATCH);
2540       }
2541     ecode++;
2542     break;
2543 
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. */
2547 
2548     case OP_PROP:
2549     case OP_NOTPROP:
2550     if (eptr >= md->end_subject)
2551       {
2552       SCHECK_PARTIAL();
2553       RRETURN(MATCH_NOMATCH);
2554       }
2555     GETCHARINCTEST(c, eptr);
2556       {
2557       const pcre_uint32 *cp;
2558       const ucd_record *prop = GET_UCD(c);
2559 
2560       switch(ecode[1])
2561         {
2562         case PT_ANY:
2563         if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2564         break;
2565 
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;
2572 
2573         case PT_GC:
2574         if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2575           RRETURN(MATCH_NOMATCH);
2576         break;
2577 
2578         case PT_PC:
2579         if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2580           RRETURN(MATCH_NOMATCH);
2581         break;
2582 
2583         case PT_SC:
2584         if ((ecode[2] != prop->script) == (op == OP_PROP))
2585           RRETURN(MATCH_NOMATCH);
2586         break;
2587 
2588         /* These are specials */
2589 
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;
2595 
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. */
2599 
2600         case PT_SPACE:    /* Perl space */
2601         case PT_PXSPACE:  /* POSIX space */
2602         switch(c)
2603           {
2604           HSPACE_CASES:
2605           VSPACE_CASES:
2606           if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2607           break;
2608 
2609           default:
2610           if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z) ==
2611             (op == OP_NOTPROP)) RRETURN(MATCH_NOMATCH);
2612           break;
2613           }
2614         break;
2615 
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;
2622 
2623         case PT_CLIST:
2624         cp = PRIV(ucd_caseless_sets) + ecode[2];
2625         for (;;)
2626           {
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); } }
2631           }
2632         break;
2633 
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;
2640 
2641         /* This should never occur */
2642 
2643         default:
2644         RRETURN(PCRE_ERROR_INTERNAL);
2645         }
2646 
2647       ecode += 3;
2648       }
2649     break;
2650 
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. */
2653 
2654     case OP_EXTUNI:
2655     if (eptr >= md->end_subject)
2656       {
2657       SCHECK_PARTIAL();
2658       RRETURN(MATCH_NOMATCH);
2659       }
2660     else
2661       {
2662       int lgb, rgb;
2663       GETCHARINCTEST(c, eptr);
2664       lgb = UCD_GRAPHBREAK(c);
2665       while (eptr < md->end_subject)
2666         {
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;
2673         }
2674       }
2675     CHECK_PARTIAL();
2676     ecode++;
2677     break;
2678 #endif  /* SUPPORT_UCP */
2679 
2680 
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:
2688 
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.
2692 
2693     (b) If the JavaScript compatibility flag is set, set the length to zero
2694     so that the back reference matches an empty string.
2695 
2696     Otherwise, set the length to the length of what was matched by the
2697     referenced subpattern.
2698 
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. */
2703 
2704     case OP_DNREF:
2705     case OP_DNREFI:
2706     caseless = op == OP_DNREFI;
2707       {
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;
2711 
2712       /* Setting the default length first and initializing 'offset' avoids
2713       compiler warnings in the REF_REPEAT code. */
2714 
2715       length = (md->jscript_compat)? 0 : -1;
2716       offset = 0;
2717 
2718       while (count-- > 0)
2719         {
2720         offset = GET2(slot, 0) << 1;
2721         if (offset < offset_top && md->offset_vector[offset] >= 0)
2722           {
2723           length = md->offset_vector[offset+1] - md->offset_vector[offset];
2724           break;
2725           }
2726         slot += md->name_entry_size;
2727         }
2728       }
2729     goto REF_REPEAT;
2730 
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];
2740 
2741     /* Set up for repetition, or handle the non-repeated case */
2742 
2743     REF_REPEAT:
2744     switch (*ecode)
2745       {
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;
2758 
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;
2767 
2768       default:               /* No repeat follows */
2769       if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2770         {
2771         if (length == -2) eptr = md->end_subject;   /* Partial match */
2772         CHECK_PARTIAL();
2773         RRETURN(MATCH_NOMATCH);
2774         }
2775       eptr += length;
2776       continue;              /* With the main loop */
2777       }
2778 
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. */
2784 
2785     if (length == 0) continue;
2786     if (length < 0 && min == 0) continue;
2787 
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. */
2791 
2792     for (i = 1; i <= min; i++)
2793       {
2794       int slength;
2795       if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2796         {
2797         if (slength == -2) eptr = md->end_subject;   /* Partial match */
2798         CHECK_PARTIAL();
2799         RRETURN(MATCH_NOMATCH);
2800         }
2801       eptr += slength;
2802       }
2803 
2804     /* If min = max, continue at the same level without recursion.
2805     They are not both allowed to be zero. */
2806 
2807     if (min == max) continue;
2808 
2809     /* If minimizing, keep trying and advancing the pointer */
2810 
2811     if (minimize)
2812       {
2813       for (fi = min;; fi++)
2814         {
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)
2820           {
2821           if (slength == -2) eptr = md->end_subject;   /* Partial match */
2822           CHECK_PARTIAL();
2823           RRETURN(MATCH_NOMATCH);
2824           }
2825         eptr += slength;
2826         }
2827       /* Control never gets here */
2828       }
2829 
2830     /* If maximizing, find the longest string and work backwards */
2831 
2832     else
2833       {
2834       pp = eptr;
2835       for (i = min; i < max; i++)
2836         {
2837         int slength;
2838         if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2839           {
2840           /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2841           the soft partial matching case. */
2842 
2843           if (slength == -2 && md->partial != 0 &&
2844               md->end_subject > md->start_used_ptr)
2845             {
2846             md->hitend = TRUE;
2847             if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2848             }
2849           break;
2850           }
2851         eptr += slength;
2852         }
2853 
2854       while (eptr >= pp)
2855         {
2856         RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2857         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2858         eptr -= length;
2859         }
2860       RRETURN(MATCH_NOMATCH);
2861       }
2862     /* Control never gets here */
2863 
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.
2870 
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. */
2874 
2875     case OP_NCLASS:
2876     case OP_CLASS:
2877       {
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 */
2883 
2884       switch (*ecode)
2885         {
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;
2902 
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;
2913 
2914         default:               /* No repeat follows */
2915         min = max = 1;
2916         break;
2917         }
2918 
2919       /* First, ensure the minimum number of matches are present. */
2920 
2921 #ifdef SUPPORT_UTF
2922       if (utf)
2923         {
2924         for (i = 1; i <= min; i++)
2925           {
2926           if (eptr >= md->end_subject)
2927             {
2928             SCHECK_PARTIAL();
2929             RRETURN(MATCH_NOMATCH);
2930             }
2931           GETCHARINC(c, eptr);
2932           if (c > 255)
2933             {
2934             if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2935             }
2936           else
2937             if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2938           }
2939         }
2940       else
2941 #endif
2942       /* Not UTF mode */
2943         {
2944         for (i = 1; i <= min; i++)
2945           {
2946           if (eptr >= md->end_subject)
2947             {
2948             SCHECK_PARTIAL();
2949             RRETURN(MATCH_NOMATCH);
2950             }
2951           c = *eptr++;
2952 #ifndef COMPILE_PCRE8
2953           if (c > 255)
2954             {
2955             if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2956             }
2957           else
2958 #endif
2959             if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2960           }
2961         }
2962 
2963       /* If max == min we can continue with the main loop without the
2964       need to recurse. */
2965 
2966       if (min == max) continue;
2967 
2968       /* If minimizing, keep testing the rest of the expression and advancing
2969       the pointer while it matches the class. */
2970 
2971       if (minimize)
2972         {
2973 #ifdef SUPPORT_UTF
2974         if (utf)
2975           {
2976           for (fi = min;; fi++)
2977             {
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)
2982               {
2983               SCHECK_PARTIAL();
2984               RRETURN(MATCH_NOMATCH);
2985               }
2986             GETCHARINC(c, eptr);
2987             if (c > 255)
2988               {
2989               if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2990               }
2991             else
2992               if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2993             }
2994           }
2995         else
2996 #endif
2997         /* Not UTF mode */
2998           {
2999           for (fi = min;; fi++)
3000             {
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)
3005               {
3006               SCHECK_PARTIAL();
3007               RRETURN(MATCH_NOMATCH);
3008               }
3009             c = *eptr++;
3010 #ifndef COMPILE_PCRE8
3011             if (c > 255)
3012               {
3013               if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3014               }
3015             else
3016 #endif
3017               if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3018             }
3019           }
3020         /* Control never gets here */
3021         }
3022 
3023       /* If maximizing, find the longest possible run, then work backwards. */
3024 
3025       else
3026         {
3027         pp = eptr;
3028 
3029 #ifdef SUPPORT_UTF
3030         if (utf)
3031           {
3032           for (i = min; i < max; i++)
3033             {
3034             int len = 1;
3035             if (eptr >= md->end_subject)
3036               {
3037               SCHECK_PARTIAL();
3038               break;
3039               }
3040             GETCHARLEN(c, eptr, len);
3041             if (c > 255)
3042               {
3043               if (op == OP_CLASS) break;
3044               }
3045             else
3046               if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3047             eptr += len;
3048             }
3049 
3050           if (possessive) continue;    /* No backtracking */
3051 
3052           for (;;)
3053             {
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);
3058             }
3059           }
3060         else
3061 #endif
3062           /* Not UTF mode */
3063           {
3064           for (i = min; i < max; i++)
3065             {
3066             if (eptr >= md->end_subject)
3067               {
3068               SCHECK_PARTIAL();
3069               break;
3070               }
3071             c = *eptr;
3072 #ifndef COMPILE_PCRE8
3073             if (c > 255)
3074               {
3075               if (op == OP_CLASS) break;
3076               }
3077             else
3078 #endif
3079               if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3080             eptr++;
3081             }
3082 
3083           if (possessive) continue;    /* No backtracking */
3084 
3085           while (eptr >= pp)
3086             {
3087             RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3088             if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3089             eptr--;
3090             }
3091           }
3092 
3093         RRETURN(MATCH_NOMATCH);
3094         }
3095 #undef BYTE_MAP
3096       }
3097     /* Control never gets here */
3098 
3099 
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. */
3104 
3105 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3106     case OP_XCLASS:
3107       {
3108       data = ecode + 1 + LINK_SIZE;                /* Save for matching */
3109       ecode += GET(ecode, 1);                      /* Advance past the item */
3110 
3111       switch (*ecode)
3112         {
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;
3129 
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;
3140 
3141         default:               /* No repeat follows */
3142         min = max = 1;
3143         break;
3144         }
3145 
3146       /* First, ensure the minimum number of matches are present. */
3147 
3148       for (i = 1; i <= min; i++)
3149         {
3150         if (eptr >= md->end_subject)
3151           {
3152           SCHECK_PARTIAL();
3153           RRETURN(MATCH_NOMATCH);
3154           }
3155         GETCHARINCTEST(c, eptr);
3156         if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3157         }
3158 
3159       /* If max == min we can continue with the main loop without the
3160       need to recurse. */
3161 
3162       if (min == max) continue;
3163 
3164       /* If minimizing, keep testing the rest of the expression and advancing
3165       the pointer while it matches the class. */
3166 
3167       if (minimize)
3168         {
3169         for (fi = min;; fi++)
3170           {
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)
3175             {
3176             SCHECK_PARTIAL();
3177             RRETURN(MATCH_NOMATCH);
3178             }
3179           GETCHARINCTEST(c, eptr);
3180           if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3181           }
3182         /* Control never gets here */
3183         }
3184 
3185       /* If maximizing, find the longest possible run, then work backwards. */
3186 
3187       else
3188         {
3189         pp = eptr;
3190         for (i = min; i < max; i++)
3191           {
3192           int len = 1;
3193           if (eptr >= md->end_subject)
3194             {
3195             SCHECK_PARTIAL();
3196             break;
3197             }
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;
3205           }
3206 
3207         if (possessive) continue;    /* No backtracking */
3208 
3209         for(;;)
3210           {
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
3217           }
3218         RRETURN(MATCH_NOMATCH);
3219         }
3220 
3221       /* Control never gets here */
3222       }
3223 #endif    /* End of XCLASS */
3224 
3225     /* Match a single character, casefully */
3226 
3227     case OP_CHAR:
3228 #ifdef SUPPORT_UTF
3229     if (utf)
3230       {
3231       length = 1;
3232       ecode++;
3233       GETCHARLEN(fc, ecode, length);
3234       if (length > md->end_subject - eptr)
3235         {
3236         CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3237         RRETURN(MATCH_NOMATCH);
3238         }
3239       while (length-- > 0) if (*ecode++ != UCHAR21INC(eptr)) RRETURN(MATCH_NOMATCH);
3240       }
3241     else
3242 #endif
3243     /* Not UTF mode */
3244       {
3245       if (md->end_subject - eptr < 1)
3246         {
3247         SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3248         RRETURN(MATCH_NOMATCH);
3249         }
3250       if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3251       ecode += 2;
3252       }
3253     break;
3254 
3255     /* Match a single character, caselessly. If we are at the end of the
3256     subject, give up immediately. */
3257 
3258     case OP_CHARI:
3259     if (eptr >= md->end_subject)
3260       {
3261       SCHECK_PARTIAL();
3262       RRETURN(MATCH_NOMATCH);
3263       }
3264 
3265 #ifdef SUPPORT_UTF
3266     if (utf)
3267       {
3268       length = 1;
3269       ecode++;
3270       GETCHARLEN(fc, ecode, length);
3271 
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. */
3276 
3277       if (fc < 128)
3278         {
3279         pcre_uint32 cc = UCHAR21(eptr);
3280         if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3281         ecode++;
3282         eptr++;
3283         }
3284 
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.  */
3288 
3289       else
3290         {
3291         pcre_uint32 dc;
3292         GETCHARINC(dc, eptr);
3293         ecode += length;
3294 
3295         /* If we have Unicode property support, we can use it to test the other
3296         case of the character, if there is one. */
3297 
3298         if (fc != dc)
3299           {
3300 #ifdef SUPPORT_UCP
3301           if (dc != UCD_OTHERCASE(fc))
3302 #endif
3303             RRETURN(MATCH_NOMATCH);
3304           }
3305         }
3306       }
3307     else
3308 #endif   /* SUPPORT_UTF */
3309 
3310     /* Not UTF mode */
3311       {
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;
3316       }
3317     break;
3318 
3319     /* Match a single character repeatedly. */
3320 
3321     case OP_EXACT:
3322     case OP_EXACTI:
3323     min = max = GET2(ecode, 1);
3324     ecode += 1 + IMM2_SIZE;
3325     goto REPEATCHAR;
3326 
3327     case OP_POSUPTO:
3328     case OP_POSUPTOI:
3329     possessive = TRUE;
3330     /* Fall through */
3331 
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;
3341 
3342     case OP_POSSTAR:
3343     case OP_POSSTARI:
3344     possessive = TRUE;
3345     min = 0;
3346     max = INT_MAX;
3347     ecode++;
3348     goto REPEATCHAR;
3349 
3350     case OP_POSPLUS:
3351     case OP_POSPLUSI:
3352     possessive = TRUE;
3353     min = 1;
3354     max = INT_MAX;
3355     ecode++;
3356     goto REPEATCHAR;
3357 
3358     case OP_POSQUERY:
3359     case OP_POSQUERYI:
3360     possessive = TRUE;
3361     min = 0;
3362     max = 1;
3363     ecode++;
3364     goto REPEATCHAR;
3365 
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;
3383 
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.
3389 
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.
3397 
3398     The various UTF/non-UTF and caseful/caseless cases are handled separately,
3399     for speed. */
3400 
3401     REPEATCHAR:
3402 #ifdef SUPPORT_UTF
3403     if (utf)
3404       {
3405       length = 1;
3406       charptr = ecode;
3407       GETCHARLEN(fc, ecode, length);
3408       ecode += length;
3409 
3410       /* Handle multibyte character matching specially here. There is
3411       support for caseless matching if UCP support is present. */
3412 
3413       if (length > 1)
3414         {
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 */
3422 
3423         for (i = 1; i <= min; i++)
3424           {
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
3433             {
3434             CHECK_PARTIAL();
3435             RRETURN(MATCH_NOMATCH);
3436             }
3437           }
3438 
3439         if (min == max) continue;
3440 
3441         if (minimize)
3442           {
3443           for (fi = min;; fi++)
3444             {
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
3456               {
3457               CHECK_PARTIAL();
3458               RRETURN(MATCH_NOMATCH);
3459               }
3460             }
3461           /* Control never gets here */
3462           }
3463 
3464         else  /* Maximize */
3465           {
3466           pp = eptr;
3467           for (i = min; i < max; i++)
3468             {
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
3477               {
3478               CHECK_PARTIAL();
3479               break;
3480               }
3481             }
3482 
3483           if (possessive) continue;    /* No backtracking */
3484           for(;;)
3485             {
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 */
3495             }
3496           }
3497         /* Control never gets here */
3498         }
3499 
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. */
3503       }
3504     else
3505 #endif  /* SUPPORT_UTF */
3506       /* When not in UTF-8 mode, load a single-byte character. */
3507       fc = *ecode++;
3508 
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. */
3517 
3518     DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3519       max, (char *)eptr));
3520 
3521     if (op >= OP_STARI)  /* Caseless */
3522       {
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 */
3539 
3540       for (i = 1; i <= min; i++)
3541         {
3542         pcre_uint32 cc;                 /* Faster than pcre_uchar */
3543         if (eptr >= md->end_subject)
3544           {
3545           SCHECK_PARTIAL();
3546           RRETURN(MATCH_NOMATCH);
3547           }
3548         cc = UCHAR21TEST(eptr);
3549         if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3550         eptr++;
3551         }
3552       if (min == max) continue;
3553       if (minimize)
3554         {
3555         for (fi = min;; fi++)
3556           {
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)
3562             {
3563             SCHECK_PARTIAL();
3564             RRETURN(MATCH_NOMATCH);
3565             }
3566           cc = UCHAR21TEST(eptr);
3567           if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3568           eptr++;
3569           }
3570         /* Control never gets here */
3571         }
3572       else  /* Maximize */
3573         {
3574         pp = eptr;
3575         for (i = min; i < max; i++)
3576           {
3577           pcre_uint32 cc;               /* Faster than pcre_uchar */
3578           if (eptr >= md->end_subject)
3579             {
3580             SCHECK_PARTIAL();
3581             break;
3582             }
3583           cc = UCHAR21TEST(eptr);
3584           if (fc != cc && foc != cc) break;
3585           eptr++;
3586           }
3587         if (possessive) continue;       /* No backtracking */
3588         for (;;)
3589           {
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);
3594           }
3595         /* Control never gets here */
3596         }
3597       }
3598 
3599     /* Caseful comparisons (includes all multi-byte characters) */
3600 
3601     else
3602       {
3603       for (i = 1; i <= min; i++)
3604         {
3605         if (eptr >= md->end_subject)
3606           {
3607           SCHECK_PARTIAL();
3608           RRETURN(MATCH_NOMATCH);
3609           }
3610         if (fc != UCHAR21INCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3611         }
3612 
3613       if (min == max) continue;
3614 
3615       if (minimize)
3616         {
3617         for (fi = min;; fi++)
3618           {
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)
3623             {
3624             SCHECK_PARTIAL();
3625             RRETURN(MATCH_NOMATCH);
3626             }
3627           if (fc != UCHAR21INCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3628           }
3629         /* Control never gets here */
3630         }
3631       else  /* Maximize */
3632         {
3633         pp = eptr;
3634         for (i = min; i < max; i++)
3635           {
3636           if (eptr >= md->end_subject)
3637             {
3638             SCHECK_PARTIAL();
3639             break;
3640             }
3641           if (fc != UCHAR21TEST(eptr)) break;
3642           eptr++;
3643           }
3644         if (possessive) continue;    /* No backtracking */
3645         for (;;)
3646           {
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);
3651           }
3652         /* Control never gets here */
3653         }
3654       }
3655     /* Control never gets here */
3656 
3657     /* Match a negated single one-byte character. The character we are
3658     checking can be multibyte. */
3659 
3660     case OP_NOT:
3661     case OP_NOTI:
3662     if (eptr >= md->end_subject)
3663       {
3664       SCHECK_PARTIAL();
3665       RRETURN(MATCH_NOMATCH);
3666       }
3667 #ifdef SUPPORT_UTF
3668     if (utf)
3669       {
3670       register pcre_uint32 ch, och;
3671 
3672       ecode++;
3673       GETCHARINC(ch, ecode);
3674       GETCHARINC(c, eptr);
3675 
3676       if (op == OP_NOT)
3677         {
3678         if (ch == c) RRETURN(MATCH_NOMATCH);
3679         }
3680       else
3681         {
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);
3692         }
3693       }
3694     else
3695 #endif
3696       {
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;
3702       }
3703     break;
3704 
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... */
3711 
3712     case OP_NOTEXACT:
3713     case OP_NOTEXACTI:
3714     min = max = GET2(ecode, 1);
3715     ecode += 1 + IMM2_SIZE;
3716     goto REPEATNOTCHAR;
3717 
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;
3727 
3728     case OP_NOTPOSSTAR:
3729     case OP_NOTPOSSTARI:
3730     possessive = TRUE;
3731     min = 0;
3732     max = INT_MAX;
3733     ecode++;
3734     goto REPEATNOTCHAR;
3735 
3736     case OP_NOTPOSPLUS:
3737     case OP_NOTPOSPLUSI:
3738     possessive = TRUE;
3739     min = 1;
3740     max = INT_MAX;
3741     ecode++;
3742     goto REPEATNOTCHAR;
3743 
3744     case OP_NOTPOSQUERY:
3745     case OP_NOTPOSQUERYI:
3746     possessive = TRUE;
3747     min = 0;
3748     max = 1;
3749     ecode++;
3750     goto REPEATNOTCHAR;
3751 
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;
3759 
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;
3777 
3778     /* Common code for all repeated single-byte matches. */
3779 
3780     REPEATNOTCHAR:
3781     GETCHARINCTEST(fc, ecode);
3782 
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. */
3790 
3791     DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3792       max, (char *)eptr));
3793 
3794     if (op >= OP_NOTSTARI)     /* Caseless */
3795       {
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);
3807 
3808 #ifdef SUPPORT_UTF
3809       if (utf)
3810         {
3811         register pcre_uint32 d;
3812         for (i = 1; i <= min; i++)
3813           {
3814           if (eptr >= md->end_subject)
3815             {
3816             SCHECK_PARTIAL();
3817             RRETURN(MATCH_NOMATCH);
3818             }
3819           GETCHARINC(d, eptr);
3820           if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3821           }
3822         }
3823       else
3824 #endif  /* SUPPORT_UTF */
3825       /* Not UTF mode */
3826         {
3827         for (i = 1; i <= min; i++)
3828           {
3829           if (eptr >= md->end_subject)
3830             {
3831             SCHECK_PARTIAL();
3832             RRETURN(MATCH_NOMATCH);
3833             }
3834           if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3835           eptr++;
3836           }
3837         }
3838 
3839       if (min == max) continue;
3840 
3841       if (minimize)
3842         {
3843 #ifdef SUPPORT_UTF
3844         if (utf)
3845           {
3846           register pcre_uint32 d;
3847           for (fi = min;; fi++)
3848             {
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)
3853               {
3854               SCHECK_PARTIAL();
3855               RRETURN(MATCH_NOMATCH);
3856               }
3857             GETCHARINC(d, eptr);
3858             if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
3859             }
3860           }
3861         else
3862 #endif  /*SUPPORT_UTF */
3863         /* Not UTF mode */
3864           {
3865           for (fi = min;; fi++)
3866             {
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)
3871               {
3872               SCHECK_PARTIAL();
3873               RRETURN(MATCH_NOMATCH);
3874               }
3875             if (fc == *eptr || foc == *eptr) RRETURN(MATCH_NOMATCH);
3876             eptr++;
3877             }
3878           }
3879         /* Control never gets here */
3880         }
3881 
3882       /* Maximize case */
3883 
3884       else
3885         {
3886         pp = eptr;
3887 
3888 #ifdef SUPPORT_UTF
3889         if (utf)
3890           {
3891           register pcre_uint32 d;
3892           for (i = min; i < max; i++)
3893             {
3894             int len = 1;
3895             if (eptr >= md->end_subject)
3896               {
3897               SCHECK_PARTIAL();
3898               break;
3899               }
3900             GETCHARLEN(d, eptr, len);
3901             if (fc == d || (unsigned int)foc == d) break;
3902             eptr += len;
3903             }
3904           if (possessive) continue;    /* No backtracking */
3905           for(;;)
3906             {
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);
3912             }
3913           }
3914         else
3915 #endif  /* SUPPORT_UTF */
3916         /* Not UTF mode */
3917           {
3918           for (i = min; i < max; i++)
3919             {
3920             if (eptr >= md->end_subject)
3921               {
3922               SCHECK_PARTIAL();
3923               break;
3924               }
3925             if (fc == *eptr || foc == *eptr) break;
3926             eptr++;
3927             }
3928           if (possessive) continue;    /* No backtracking */
3929           for (;;)
3930             {
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--;
3935             }
3936           }
3937         /* Control never gets here */
3938         }
3939       }
3940 
3941     /* Caseful comparisons */
3942 
3943     else
3944       {
3945 #ifdef SUPPORT_UTF
3946       if (utf)
3947         {
3948         register pcre_uint32 d;
3949         for (i = 1; i <= min; i++)
3950           {
3951           if (eptr >= md->end_subject)
3952             {
3953             SCHECK_PARTIAL();
3954             RRETURN(MATCH_NOMATCH);
3955             }
3956           GETCHARINC(d, eptr);
3957           if (fc == d) RRETURN(MATCH_NOMATCH);
3958           }
3959         }
3960       else
3961 #endif
3962       /* Not UTF mode */
3963         {
3964         for (i = 1; i <= min; i++)
3965           {
3966           if (eptr >= md->end_subject)
3967             {
3968             SCHECK_PARTIAL();
3969             RRETURN(MATCH_NOMATCH);
3970             }
3971           if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
3972           }
3973         }
3974 
3975       if (min == max) continue;
3976 
3977       if (minimize)
3978         {
3979 #ifdef SUPPORT_UTF
3980         if (utf)
3981           {
3982           register pcre_uint32 d;
3983           for (fi = min;; fi++)
3984             {
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)
3989               {
3990               SCHECK_PARTIAL();
3991               RRETURN(MATCH_NOMATCH);
3992               }
3993             GETCHARINC(d, eptr);
3994             if (fc == d) RRETURN(MATCH_NOMATCH);
3995             }
3996           }
3997         else
3998 #endif
3999         /* Not UTF mode */
4000           {
4001           for (fi = min;; fi++)
4002             {
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)
4007               {
4008               SCHECK_PARTIAL();
4009               RRETURN(MATCH_NOMATCH);
4010               }
4011             if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
4012             }
4013           }
4014         /* Control never gets here */
4015         }
4016 
4017       /* Maximize case */
4018 
4019       else
4020         {
4021         pp = eptr;
4022 
4023 #ifdef SUPPORT_UTF
4024         if (utf)
4025           {
4026           register pcre_uint32 d;
4027           for (i = min; i < max; i++)
4028             {
4029             int len = 1;
4030             if (eptr >= md->end_subject)
4031               {
4032               SCHECK_PARTIAL();
4033               break;
4034               }
4035             GETCHARLEN(d, eptr, len);
4036             if (fc == d) break;
4037             eptr += len;
4038             }
4039           if (possessive) continue;    /* No backtracking */
4040           for(;;)
4041             {
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);
4047             }
4048           }
4049         else
4050 #endif
4051         /* Not UTF mode */
4052           {
4053           for (i = min; i < max; i++)
4054             {
4055             if (eptr >= md->end_subject)
4056               {
4057               SCHECK_PARTIAL();
4058               break;
4059               }
4060             if (fc == *eptr) break;
4061             eptr++;
4062             }
4063           if (possessive) continue;    /* No backtracking */
4064           for (;;)
4065             {
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--;
4070             }
4071           }
4072         /* Control never gets here */
4073         }
4074       }
4075     /* Control never gets here */
4076 
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. */
4080 
4081     case OP_TYPEEXACT:
4082     min = max = GET2(ecode, 1);
4083     minimize = TRUE;
4084     ecode += 1 + IMM2_SIZE;
4085     goto REPEATTYPE;
4086 
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;
4094 
4095     case OP_TYPEPOSSTAR:
4096     possessive = TRUE;
4097     min = 0;
4098     max = INT_MAX;
4099     ecode++;
4100     goto REPEATTYPE;
4101 
4102     case OP_TYPEPOSPLUS:
4103     possessive = TRUE;
4104     min = 1;
4105     max = INT_MAX;
4106     ecode++;
4107     goto REPEATTYPE;
4108 
4109     case OP_TYPEPOSQUERY:
4110     possessive = TRUE;
4111     min = 0;
4112     max = 1;
4113     ecode++;
4114     goto REPEATTYPE;
4115 
4116     case OP_TYPEPOSUPTO:
4117     possessive = TRUE;
4118     min = 0;
4119     max = GET2(ecode, 1);
4120     ecode += 1 + IMM2_SIZE;
4121     goto REPEATTYPE;
4122 
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;
4134 
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. */
4138 
4139     REPEATTYPE:
4140     ctype = *ecode++;      /* Code for the character type */
4141 
4142 #ifdef SUPPORT_UCP
4143     if (ctype == OP_PROP || ctype == OP_NOTPROP)
4144       {
4145       prop_fail_result = ctype == OP_NOTPROP;
4146       prop_type = *ecode++;
4147       prop_value = *ecode++;
4148       }
4149     else prop_type = -1;
4150 #endif
4151 
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. */
4157 
4158     if (min > 0)
4159       {
4160 #ifdef SUPPORT_UCP
4161       if (prop_type >= 0)
4162         {
4163         switch(prop_type)
4164           {
4165           case PT_ANY:
4166           if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4167           for (i = 1; i <= min; i++)
4168             {
4169             if (eptr >= md->end_subject)
4170               {
4171               SCHECK_PARTIAL();
4172               RRETURN(MATCH_NOMATCH);
4173               }
4174             GETCHARINCTEST(c, eptr);
4175             }
4176           break;
4177 
4178           case PT_LAMP:
4179           for (i = 1; i <= min; i++)
4180             {
4181             int chartype;
4182             if (eptr >= md->end_subject)
4183               {
4184               SCHECK_PARTIAL();
4185               RRETURN(MATCH_NOMATCH);
4186               }
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);
4193             }
4194           break;
4195 
4196           case PT_GC:
4197           for (i = 1; i <= min; i++)
4198             {
4199             if (eptr >= md->end_subject)
4200               {
4201               SCHECK_PARTIAL();
4202               RRETURN(MATCH_NOMATCH);
4203               }
4204             GETCHARINCTEST(c, eptr);
4205             if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4206               RRETURN(MATCH_NOMATCH);
4207             }
4208           break;
4209 
4210           case PT_PC:
4211           for (i = 1; i <= min; i++)
4212             {
4213             if (eptr >= md->end_subject)
4214               {
4215               SCHECK_PARTIAL();
4216               RRETURN(MATCH_NOMATCH);
4217               }
4218             GETCHARINCTEST(c, eptr);
4219             if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4220               RRETURN(MATCH_NOMATCH);
4221             }
4222           break;
4223 
4224           case PT_SC:
4225           for (i = 1; i <= min; i++)
4226             {
4227             if (eptr >= md->end_subject)
4228               {
4229               SCHECK_PARTIAL();
4230               RRETURN(MATCH_NOMATCH);
4231               }
4232             GETCHARINCTEST(c, eptr);
4233             if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4234               RRETURN(MATCH_NOMATCH);
4235             }
4236           break;
4237 
4238           case PT_ALNUM:
4239           for (i = 1; i <= min; i++)
4240             {
4241             int category;
4242             if (eptr >= md->end_subject)
4243               {
4244               SCHECK_PARTIAL();
4245               RRETURN(MATCH_NOMATCH);
4246               }
4247             GETCHARINCTEST(c, eptr);
4248             category = UCD_CATEGORY(c);
4249             if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4250               RRETURN(MATCH_NOMATCH);
4251             }
4252           break;
4253 
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. */
4257 
4258           case PT_SPACE:    /* Perl space */
4259           case PT_PXSPACE:  /* POSIX space */
4260           for (i = 1; i <= min; i++)
4261             {
4262             if (eptr >= md->end_subject)
4263               {
4264               SCHECK_PARTIAL();
4265               RRETURN(MATCH_NOMATCH);
4266               }
4267             GETCHARINCTEST(c, eptr);
4268             switch(c)
4269               {
4270               HSPACE_CASES:
4271               VSPACE_CASES:
4272               if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4273               break;
4274 
4275               default:
4276               if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
4277                 RRETURN(MATCH_NOMATCH);
4278               break;
4279               }
4280             }
4281           break;
4282 
4283           case PT_WORD:
4284           for (i = 1; i <= min; i++)
4285             {
4286             int category;
4287             if (eptr >= md->end_subject)
4288               {
4289               SCHECK_PARTIAL();
4290               RRETURN(MATCH_NOMATCH);
4291               }
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);
4297             }
4298           break;
4299 
4300           case PT_CLIST:
4301           for (i = 1; i <= min; i++)
4302             {
4303             const pcre_uint32 *cp;
4304             if (eptr >= md->end_subject)
4305               {
4306               SCHECK_PARTIAL();
4307               RRETURN(MATCH_NOMATCH);
4308               }
4309             GETCHARINCTEST(c, eptr);
4310             cp = PRIV(ucd_caseless_sets) + prop_value;
4311             for (;;)
4312               {
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; }
4317               }
4318             }
4319           break;
4320 
4321           case PT_UCNC:
4322           for (i = 1; i <= min; i++)
4323             {
4324             if (eptr >= md->end_subject)
4325               {
4326               SCHECK_PARTIAL();
4327               RRETURN(MATCH_NOMATCH);
4328               }
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);
4334             }
4335           break;
4336 
4337           /* This should not occur */
4338 
4339           default:
4340           RRETURN(PCRE_ERROR_INTERNAL);
4341           }
4342         }
4343 
4344       /* Match extended Unicode sequences. We will get here only if the
4345       support is in the binary; otherwise a compile-time error occurs. */
4346 
4347       else if (ctype == OP_EXTUNI)
4348         {
4349         for (i = 1; i <= min; i++)
4350           {
4351           if (eptr >= md->end_subject)
4352             {
4353             SCHECK_PARTIAL();
4354             RRETURN(MATCH_NOMATCH);
4355             }
4356           else
4357             {
4358             int lgb, rgb;
4359             GETCHARINCTEST(c, eptr);
4360             lgb = UCD_GRAPHBREAK(c);
4361            while (eptr < md->end_subject)
4362               {
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;
4369               }
4370             }
4371           CHECK_PARTIAL();
4372           }
4373         }
4374 
4375       else
4376 #endif     /* SUPPORT_UCP */
4377 
4378 /* Handle all other cases when the coding is UTF-8 */
4379 
4380 #ifdef SUPPORT_UTF
4381       if (utf) switch(ctype)
4382         {
4383         case OP_ANY:
4384         for (i = 1; i <= min; i++)
4385           {
4386           if (eptr >= md->end_subject)
4387             {
4388             SCHECK_PARTIAL();
4389             RRETURN(MATCH_NOMATCH);
4390             }
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])
4397             {
4398             md->hitend = TRUE;
4399             if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4400             }
4401           eptr++;
4402           ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4403           }
4404         break;
4405 
4406         case OP_ALLANY:
4407         for (i = 1; i <= min; i++)
4408           {
4409           if (eptr >= md->end_subject)
4410             {
4411             SCHECK_PARTIAL();
4412             RRETURN(MATCH_NOMATCH);
4413             }
4414           eptr++;
4415           ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
4416           }
4417         break;
4418 
4419         case OP_ANYBYTE:
4420         if (eptr > md->end_subject - min) RRETURN(MATCH_NOMATCH);
4421         eptr += min;
4422         break;
4423 
4424         case OP_ANYNL:
4425         for (i = 1; i <= min; i++)
4426           {
4427           if (eptr >= md->end_subject)
4428             {
4429             SCHECK_PARTIAL();
4430             RRETURN(MATCH_NOMATCH);
4431             }
4432           GETCHARINC(c, eptr);
4433           switch(c)
4434             {
4435             default: RRETURN(MATCH_NOMATCH);
4436 
4437             case CHAR_CR:
4438             if (eptr < md->end_subject && UCHAR21(eptr) == CHAR_LF) eptr++;
4439             break;
4440 
4441             case CHAR_LF:
4442             break;
4443 
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;
4453             }
4454           }
4455         break;
4456 
4457         case OP_NOT_HSPACE:
4458         for (i = 1; i <= min; i++)
4459           {
4460           if (eptr >= md->end_subject)
4461             {
4462             SCHECK_PARTIAL();
4463             RRETURN(MATCH_NOMATCH);
4464             }
4465           GETCHARINC(c, eptr);
4466           switch(c)
4467             {
4468             HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
4469             default: break;
4470             }
4471           }
4472         break;
4473 
4474         case OP_HSPACE:
4475         for (i = 1; i <= min; i++)
4476           {
4477           if (eptr >= md->end_subject)
4478             {
4479             SCHECK_PARTIAL();
4480             RRETURN(MATCH_NOMATCH);
4481             }
4482           GETCHARINC(c, eptr);
4483           switch(c)
4484             {
4485             HSPACE_CASES: break;  /* Byte and multibyte cases */
4486             default: RRETURN(MATCH_NOMATCH);
4487             }
4488           }
4489         break;
4490 
4491         case OP_NOT_VSPACE:
4492         for (i = 1; i <= min; i++)
4493           {
4494           if (eptr >= md->end_subject)
4495             {
4496             SCHECK_PARTIAL();
4497             RRETURN(MATCH_NOMATCH);
4498             }
4499           GETCHARINC(c, eptr);
4500           switch(c)
4501             {
4502             VSPACE_CASES: RRETURN(MATCH_NOMATCH);
4503             default: break;
4504             }
4505           }
4506         break;
4507 
4508         case OP_VSPACE:
4509         for (i = 1; i <= min; i++)
4510           {
4511           if (eptr >= md->end_subject)
4512             {
4513             SCHECK_PARTIAL();
4514             RRETURN(MATCH_NOMATCH);
4515             }
4516           GETCHARINC(c, eptr);
4517           switch(c)
4518             {
4519             VSPACE_CASES: break;
4520             default: RRETURN(MATCH_NOMATCH);
4521             }
4522           }
4523         break;
4524 
4525         case OP_NOT_DIGIT:
4526         for (i = 1; i <= min; i++)
4527           {
4528           if (eptr >= md->end_subject)
4529             {
4530             SCHECK_PARTIAL();
4531             RRETURN(MATCH_NOMATCH);
4532             }
4533           GETCHARINC(c, eptr);
4534           if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
4535             RRETURN(MATCH_NOMATCH);
4536           }
4537         break;
4538 
4539         case OP_DIGIT:
4540         for (i = 1; i <= min; i++)
4541           {
4542           pcre_uint32 cc;
4543           if (eptr >= md->end_subject)
4544             {
4545             SCHECK_PARTIAL();
4546             RRETURN(MATCH_NOMATCH);
4547             }
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 */
4553           }
4554         break;
4555 
4556         case OP_NOT_WHITESPACE:
4557         for (i = 1; i <= min; i++)
4558           {
4559           pcre_uint32 cc;
4560           if (eptr >= md->end_subject)
4561             {
4562             SCHECK_PARTIAL();
4563             RRETURN(MATCH_NOMATCH);
4564             }
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++);
4570           }
4571         break;
4572 
4573         case OP_WHITESPACE:
4574         for (i = 1; i <= min; i++)
4575           {
4576           pcre_uint32 cc;
4577           if (eptr >= md->end_subject)
4578             {
4579             SCHECK_PARTIAL();
4580             RRETURN(MATCH_NOMATCH);
4581             }
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 */
4587           }
4588         break;
4589 
4590         case OP_NOT_WORDCHAR:
4591         for (i = 1; i <= min; i++)
4592           {
4593           pcre_uint32 cc;
4594           if (eptr >= md->end_subject)
4595             {
4596             SCHECK_PARTIAL();
4597             RRETURN(MATCH_NOMATCH);
4598             }
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++);
4604           }
4605         break;
4606 
4607         case OP_WORDCHAR:
4608         for (i = 1; i <= min; i++)
4609           {
4610           pcre_uint32 cc;
4611           if (eptr >= md->end_subject)
4612             {
4613             SCHECK_PARTIAL();
4614             RRETURN(MATCH_NOMATCH);
4615             }
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 */
4621           }
4622         break;
4623 
4624         default:
4625         RRETURN(PCRE_ERROR_INTERNAL);
4626         }  /* End switch(ctype) */
4627 
4628       else
4629 #endif     /* SUPPORT_UTF */
4630 
4631       /* Code for the non-UTF-8 case for minimum matching of operators other
4632       than OP_PROP and OP_NOTPROP. */
4633 
4634       switch(ctype)
4635         {
4636         case OP_ANY:
4637         for (i = 1; i <= min; i++)
4638           {
4639           if (eptr >= md->end_subject)
4640             {
4641             SCHECK_PARTIAL();
4642             RRETURN(MATCH_NOMATCH);
4643             }
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])
4650             {
4651             md->hitend = TRUE;
4652             if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
4653             }
4654           eptr++;
4655           }
4656         break;
4657 
4658         case OP_ALLANY:
4659         if (eptr > md->end_subject - min)
4660           {
4661           SCHECK_PARTIAL();
4662           RRETURN(MATCH_NOMATCH);
4663           }
4664         eptr += min;
4665         break;
4666 
4667         case OP_ANYBYTE:
4668         if (eptr > md->end_subject - min)
4669           {
4670           SCHECK_PARTIAL();
4671           RRETURN(MATCH_NOMATCH);
4672           }
4673         eptr += min;
4674         break;
4675 
4676         case OP_ANYNL:
4677         for (i = 1; i <= min; i++)
4678           {
4679           if (eptr >= md->end_subject)
4680             {
4681             SCHECK_PARTIAL();
4682             RRETURN(MATCH_NOMATCH);
4683             }
4684           switch(*eptr++)
4685             {
4686             default: RRETURN(MATCH_NOMATCH);
4687 
4688             case CHAR_CR:
4689             if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
4690             break;
4691 
4692             case CHAR_LF:
4693             break;
4694 
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;
4704             }
4705           }
4706         break;
4707 
4708         case OP_NOT_HSPACE:
4709         for (i = 1; i <= min; i++)
4710           {
4711           if (eptr >= md->end_subject)
4712             {
4713             SCHECK_PARTIAL();
4714             RRETURN(MATCH_NOMATCH);
4715             }
4716           switch(*eptr++)
4717             {
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);
4724             }
4725           }
4726         break;
4727 
4728         case OP_HSPACE:
4729         for (i = 1; i <= min; i++)
4730           {
4731           if (eptr >= md->end_subject)
4732             {
4733             SCHECK_PARTIAL();
4734             RRETURN(MATCH_NOMATCH);
4735             }
4736           switch(*eptr++)
4737             {
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;
4744             }
4745           }
4746         break;
4747 
4748         case OP_NOT_VSPACE:
4749         for (i = 1; i <= min; i++)
4750           {
4751           if (eptr >= md->end_subject)
4752             {
4753             SCHECK_PARTIAL();
4754             RRETURN(MATCH_NOMATCH);
4755             }
4756           switch(*eptr++)
4757             {
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;
4764             }
4765           }
4766         break;
4767 
4768         case OP_VSPACE:
4769         for (i = 1; i <= min; i++)
4770           {
4771           if (eptr >= md->end_subject)
4772             {
4773             SCHECK_PARTIAL();
4774             RRETURN(MATCH_NOMATCH);
4775             }
4776           switch(*eptr++)
4777             {
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;
4784             }
4785           }
4786         break;
4787 
4788         case OP_NOT_DIGIT:
4789         for (i = 1; i <= min; i++)
4790           {
4791           if (eptr >= md->end_subject)
4792             {
4793             SCHECK_PARTIAL();
4794             RRETURN(MATCH_NOMATCH);
4795             }
4796           if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0)
4797             RRETURN(MATCH_NOMATCH);
4798           eptr++;
4799           }
4800         break;
4801 
4802         case OP_DIGIT:
4803         for (i = 1; i <= min; i++)
4804           {
4805           if (eptr >= md->end_subject)
4806             {
4807             SCHECK_PARTIAL();
4808             RRETURN(MATCH_NOMATCH);
4809             }
4810           if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0)
4811             RRETURN(MATCH_NOMATCH);
4812           eptr++;
4813           }
4814         break;
4815 
4816         case OP_NOT_WHITESPACE:
4817         for (i = 1; i <= min; i++)
4818           {
4819           if (eptr >= md->end_subject)
4820             {
4821             SCHECK_PARTIAL();
4822             RRETURN(MATCH_NOMATCH);
4823             }
4824           if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0)
4825             RRETURN(MATCH_NOMATCH);
4826           eptr++;
4827           }
4828         break;
4829 
4830         case OP_WHITESPACE:
4831         for (i = 1; i <= min; i++)
4832           {
4833           if (eptr >= md->end_subject)
4834             {
4835             SCHECK_PARTIAL();
4836             RRETURN(MATCH_NOMATCH);
4837             }
4838           if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0)
4839             RRETURN(MATCH_NOMATCH);
4840           eptr++;
4841           }
4842         break;
4843 
4844         case OP_NOT_WORDCHAR:
4845         for (i = 1; i <= min; i++)
4846           {
4847           if (eptr >= md->end_subject)
4848             {
4849             SCHECK_PARTIAL();
4850             RRETURN(MATCH_NOMATCH);
4851             }
4852           if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0)
4853             RRETURN(MATCH_NOMATCH);
4854           eptr++;
4855           }
4856         break;
4857 
4858         case OP_WORDCHAR:
4859         for (i = 1; i <= min; i++)
4860           {
4861           if (eptr >= md->end_subject)
4862             {
4863             SCHECK_PARTIAL();
4864             RRETURN(MATCH_NOMATCH);
4865             }
4866           if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0)
4867             RRETURN(MATCH_NOMATCH);
4868           eptr++;
4869           }
4870         break;
4871 
4872         default:
4873         RRETURN(PCRE_ERROR_INTERNAL);
4874         }
4875       }
4876 
4877     /* If min = max, continue at the same level without recursing */
4878 
4879     if (min == max) continue;
4880 
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. */
4884 
4885     if (minimize)
4886       {
4887 #ifdef SUPPORT_UCP
4888       if (prop_type >= 0)
4889         {
4890         switch(prop_type)
4891           {
4892           case PT_ANY:
4893           for (fi = min;; fi++)
4894             {
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)
4899               {
4900               SCHECK_PARTIAL();
4901               RRETURN(MATCH_NOMATCH);
4902               }
4903             GETCHARINCTEST(c, eptr);
4904             if (prop_fail_result) RRETURN(MATCH_NOMATCH);
4905             }
4906           /* Control never gets here */
4907 
4908           case PT_LAMP:
4909           for (fi = min;; fi++)
4910             {
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)
4916               {
4917               SCHECK_PARTIAL();
4918               RRETURN(MATCH_NOMATCH);
4919               }
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);
4926             }
4927           /* Control never gets here */
4928 
4929           case PT_GC:
4930           for (fi = min;; fi++)
4931             {
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)
4936               {
4937               SCHECK_PARTIAL();
4938               RRETURN(MATCH_NOMATCH);
4939               }
4940             GETCHARINCTEST(c, eptr);
4941             if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
4942               RRETURN(MATCH_NOMATCH);
4943             }
4944           /* Control never gets here */
4945 
4946           case PT_PC:
4947           for (fi = min;; fi++)
4948             {
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)
4953               {
4954               SCHECK_PARTIAL();
4955               RRETURN(MATCH_NOMATCH);
4956               }
4957             GETCHARINCTEST(c, eptr);
4958             if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
4959               RRETURN(MATCH_NOMATCH);
4960             }
4961           /* Control never gets here */
4962 
4963           case PT_SC:
4964           for (fi = min;; fi++)
4965             {
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)
4970               {
4971               SCHECK_PARTIAL();
4972               RRETURN(MATCH_NOMATCH);
4973               }
4974             GETCHARINCTEST(c, eptr);
4975             if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
4976               RRETURN(MATCH_NOMATCH);
4977             }
4978           /* Control never gets here */
4979 
4980           case PT_ALNUM:
4981           for (fi = min;; fi++)
4982             {
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)
4988               {
4989               SCHECK_PARTIAL();
4990               RRETURN(MATCH_NOMATCH);
4991               }
4992             GETCHARINCTEST(c, eptr);
4993             category = UCD_CATEGORY(c);
4994             if ((category == ucp_L || category == ucp_N) == prop_fail_result)
4995               RRETURN(MATCH_NOMATCH);
4996             }
4997           /* Control never gets here */
4998 
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. */
5002 
5003           case PT_SPACE:    /* Perl space */
5004           case PT_PXSPACE:  /* POSIX space */
5005           for (fi = min;; fi++)
5006             {
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)
5011               {
5012               SCHECK_PARTIAL();
5013               RRETURN(MATCH_NOMATCH);
5014               }
5015             GETCHARINCTEST(c, eptr);
5016             switch(c)
5017               {
5018               HSPACE_CASES:
5019               VSPACE_CASES:
5020               if (prop_fail_result) RRETURN(MATCH_NOMATCH);
5021               break;
5022 
5023               default:
5024               if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
5025                 RRETURN(MATCH_NOMATCH);
5026               break;
5027               }
5028             }
5029           /* Control never gets here */
5030 
5031           case PT_WORD:
5032           for (fi = min;; fi++)
5033             {
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)
5039               {
5040               SCHECK_PARTIAL();
5041               RRETURN(MATCH_NOMATCH);
5042               }
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);
5050             }
5051           /* Control never gets here */
5052 
5053           case PT_CLIST:
5054           for (fi = min;; fi++)
5055             {
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)
5061               {
5062               SCHECK_PARTIAL();
5063               RRETURN(MATCH_NOMATCH);
5064               }
5065             GETCHARINCTEST(c, eptr);
5066             cp = PRIV(ucd_caseless_sets) + prop_value;
5067             for (;;)
5068               {
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; }
5073               }
5074             }
5075           /* Control never gets here */
5076 
5077           case PT_UCNC:
5078           for (fi = min;; fi++)
5079             {
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)
5084               {
5085               SCHECK_PARTIAL();
5086               RRETURN(MATCH_NOMATCH);
5087               }
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);
5093             }
5094           /* Control never gets here */
5095 
5096           /* This should never occur */
5097           default:
5098           RRETURN(PCRE_ERROR_INTERNAL);
5099           }
5100         }
5101 
5102       /* Match extended Unicode sequences. We will get here only if the
5103       support is in the binary; otherwise a compile-time error occurs. */
5104 
5105       else if (ctype == OP_EXTUNI)
5106         {
5107         for (fi = min;; fi++)
5108           {
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)
5113             {
5114             SCHECK_PARTIAL();
5115             RRETURN(MATCH_NOMATCH);
5116             }
5117           else
5118             {
5119             int lgb, rgb;
5120             GETCHARINCTEST(c, eptr);
5121             lgb = UCD_GRAPHBREAK(c);
5122             while (eptr < md->end_subject)
5123               {
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;
5130               }
5131             }
5132           CHECK_PARTIAL();
5133           }
5134         }
5135       else
5136 #endif     /* SUPPORT_UCP */
5137 
5138 #ifdef SUPPORT_UTF
5139       if (utf)
5140         {
5141         for (fi = min;; fi++)
5142           {
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)
5147             {
5148             SCHECK_PARTIAL();
5149             RRETURN(MATCH_NOMATCH);
5150             }
5151           if (ctype == OP_ANY && IS_NEWLINE(eptr))
5152             RRETURN(MATCH_NOMATCH);
5153           GETCHARINC(c, eptr);
5154           switch(ctype)
5155             {
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])
5162               {
5163               md->hitend = TRUE;
5164               if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5165               }
5166             break;
5167 
5168             case OP_ALLANY:
5169             case OP_ANYBYTE:
5170             break;
5171 
5172             case OP_ANYNL:
5173             switch(c)
5174               {
5175               default: RRETURN(MATCH_NOMATCH);
5176               case CHAR_CR:
5177               if (eptr < md->end_subject && UCHAR21(eptr) == CHAR_LF) eptr++;
5178               break;
5179 
5180               case CHAR_LF:
5181               break;
5182 
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;
5192               }
5193             break;
5194 
5195             case OP_NOT_HSPACE:
5196             switch(c)
5197               {
5198               HSPACE_CASES: RRETURN(MATCH_NOMATCH);
5199               default: break;
5200               }
5201             break;
5202 
5203             case OP_HSPACE:
5204             switch(c)
5205               {
5206               HSPACE_CASES: break;
5207               default: RRETURN(MATCH_NOMATCH);
5208               }
5209             break;
5210 
5211             case OP_NOT_VSPACE:
5212             switch(c)
5213               {
5214               VSPACE_CASES: RRETURN(MATCH_NOMATCH);
5215               default: break;
5216               }
5217             break;
5218 
5219             case OP_VSPACE:
5220             switch(c)
5221               {
5222               VSPACE_CASES: break;
5223               default: RRETURN(MATCH_NOMATCH);
5224               }
5225             break;
5226 
5227             case OP_NOT_DIGIT:
5228             if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
5229               RRETURN(MATCH_NOMATCH);
5230             break;
5231 
5232             case OP_DIGIT:
5233             if (c >= 256 || (md->ctypes[c] & ctype_digit) == 0)
5234               RRETURN(MATCH_NOMATCH);
5235             break;
5236 
5237             case OP_NOT_WHITESPACE:
5238             if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
5239               RRETURN(MATCH_NOMATCH);
5240             break;
5241 
5242             case OP_WHITESPACE:
5243             if (c >= 256 || (md->ctypes[c] & ctype_space) == 0)
5244               RRETURN(MATCH_NOMATCH);
5245             break;
5246 
5247             case OP_NOT_WORDCHAR:
5248             if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
5249               RRETURN(MATCH_NOMATCH);
5250             break;
5251 
5252             case OP_WORDCHAR:
5253             if (c >= 256 || (md->ctypes[c] & ctype_word) == 0)
5254               RRETURN(MATCH_NOMATCH);
5255             break;
5256 
5257             default:
5258             RRETURN(PCRE_ERROR_INTERNAL);
5259             }
5260           }
5261         }
5262       else
5263 #endif
5264       /* Not UTF mode */
5265         {
5266         for (fi = min;; fi++)
5267           {
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)
5272             {
5273             SCHECK_PARTIAL();
5274             RRETURN(MATCH_NOMATCH);
5275             }
5276           if (ctype == OP_ANY && IS_NEWLINE(eptr))
5277             RRETURN(MATCH_NOMATCH);
5278           c = *eptr++;
5279           switch(ctype)
5280             {
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])
5287               {
5288               md->hitend = TRUE;
5289               if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5290               }
5291             break;
5292 
5293             case OP_ALLANY:
5294             case OP_ANYBYTE:
5295             break;
5296 
5297             case OP_ANYNL:
5298             switch(c)
5299               {
5300               default: RRETURN(MATCH_NOMATCH);
5301               case CHAR_CR:
5302               if (eptr < md->end_subject && *eptr == CHAR_LF) eptr++;
5303               break;
5304 
5305               case CHAR_LF:
5306               break;
5307 
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;
5317               }
5318             break;
5319 
5320             case OP_NOT_HSPACE:
5321             switch(c)
5322               {
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);
5329               }
5330             break;
5331 
5332             case OP_HSPACE:
5333             switch(c)
5334               {
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;
5341               }
5342             break;
5343 
5344             case OP_NOT_VSPACE:
5345             switch(c)
5346               {
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);
5353               }
5354             break;
5355 
5356             case OP_VSPACE:
5357             switch(c)
5358               {
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;
5365               }
5366             break;
5367 
5368             case OP_NOT_DIGIT:
5369             if (MAX_255(c) && (md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
5370             break;
5371 
5372             case OP_DIGIT:
5373             if (!MAX_255(c) || (md->ctypes[c] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
5374             break;
5375 
5376             case OP_NOT_WHITESPACE:
5377             if (MAX_255(c) && (md->ctypes[c] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
5378             break;
5379 
5380             case OP_WHITESPACE:
5381             if (!MAX_255(c) || (md->ctypes[c] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
5382             break;
5383 
5384             case OP_NOT_WORDCHAR:
5385             if (MAX_255(c) && (md->ctypes[c] & ctype_word) != 0) RRETURN(MATCH_NOMATCH);
5386             break;
5387 
5388             case OP_WORDCHAR:
5389             if (!MAX_255(c) || (md->ctypes[c] & ctype_word) == 0) RRETURN(MATCH_NOMATCH);
5390             break;
5391 
5392             default:
5393             RRETURN(PCRE_ERROR_INTERNAL);
5394             }
5395           }
5396         }
5397       /* Control never gets here */
5398       }
5399 
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. */
5403 
5404     else
5405       {
5406       pp = eptr;  /* Remember where we started */
5407 
5408 #ifdef SUPPORT_UCP
5409       if (prop_type >= 0)
5410         {
5411         switch(prop_type)
5412           {
5413           case PT_ANY:
5414           for (i = min; i < max; i++)
5415             {
5416             int len = 1;
5417             if (eptr >= md->end_subject)
5418               {
5419               SCHECK_PARTIAL();
5420               break;
5421               }
5422             GETCHARLENTEST(c, eptr, len);
5423             if (prop_fail_result) break;
5424             eptr+= len;
5425             }
5426           break;
5427 
5428           case PT_LAMP:
5429           for (i = min; i < max; i++)
5430             {
5431             int chartype;
5432             int len = 1;
5433             if (eptr >= md->end_subject)
5434               {
5435               SCHECK_PARTIAL();
5436               break;
5437               }
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;
5445             }
5446           break;
5447 
5448           case PT_GC:
5449           for (i = min; i < max; i++)
5450             {
5451             int len = 1;
5452             if (eptr >= md->end_subject)
5453               {
5454               SCHECK_PARTIAL();
5455               break;
5456               }
5457             GETCHARLENTEST(c, eptr, len);
5458             if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
5459             eptr+= len;
5460             }
5461           break;
5462 
5463           case PT_PC:
5464           for (i = min; i < max; i++)
5465             {
5466             int len = 1;
5467             if (eptr >= md->end_subject)
5468               {
5469               SCHECK_PARTIAL();
5470               break;
5471               }
5472             GETCHARLENTEST(c, eptr, len);
5473             if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
5474             eptr+= len;
5475             }
5476           break;
5477 
5478           case PT_SC:
5479           for (i = min; i < max; i++)
5480             {
5481             int len = 1;
5482             if (eptr >= md->end_subject)
5483               {
5484               SCHECK_PARTIAL();
5485               break;
5486               }
5487             GETCHARLENTEST(c, eptr, len);
5488             if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
5489             eptr+= len;
5490             }
5491           break;
5492 
5493           case PT_ALNUM:
5494           for (i = min; i < max; i++)
5495             {
5496             int category;
5497             int len = 1;
5498             if (eptr >= md->end_subject)
5499               {
5500               SCHECK_PARTIAL();
5501               break;
5502               }
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;
5508             }
5509           break;
5510 
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. */
5514 
5515           case PT_SPACE:    /* Perl space */
5516           case PT_PXSPACE:  /* POSIX space */
5517           for (i = min; i < max; i++)
5518             {
5519             int len = 1;
5520             if (eptr >= md->end_subject)
5521               {
5522               SCHECK_PARTIAL();
5523               break;
5524               }
5525             GETCHARLENTEST(c, eptr, len);
5526             switch(c)
5527               {
5528               HSPACE_CASES:
5529               VSPACE_CASES:
5530               if (prop_fail_result) goto ENDLOOP99;  /* Break the loop */
5531               break;
5532 
5533               default:
5534               if ((UCD_CATEGORY(c) == ucp_Z) == prop_fail_result)
5535                 goto ENDLOOP99;   /* Break the loop */
5536               break;
5537               }
5538             eptr+= len;
5539             }
5540           ENDLOOP99:
5541           break;
5542 
5543           case PT_WORD:
5544           for (i = min; i < max; i++)
5545             {
5546             int category;
5547             int len = 1;
5548             if (eptr >= md->end_subject)
5549               {
5550               SCHECK_PARTIAL();
5551               break;
5552               }
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;
5559             }
5560           break;
5561 
5562           case PT_CLIST:
5563           for (i = min; i < max; i++)
5564             {
5565             const pcre_uint32 *cp;
5566             int len = 1;
5567             if (eptr >= md->end_subject)
5568               {
5569               SCHECK_PARTIAL();
5570               break;
5571               }
5572             GETCHARLENTEST(c, eptr, len);
5573             cp = PRIV(ucd_caseless_sets) + prop_value;
5574             for (;;)
5575               {
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; }
5580               }
5581             eptr += len;
5582             }
5583           GOT_MAX:
5584           break;
5585 
5586           case PT_UCNC:
5587           for (i = min; i < max; i++)
5588             {
5589             int len = 1;
5590             if (eptr >= md->end_subject)
5591               {
5592               SCHECK_PARTIAL();
5593               break;
5594               }
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;
5601             }
5602           break;
5603 
5604           default:
5605           RRETURN(PCRE_ERROR_INTERNAL);
5606           }
5607 
5608         /* eptr is now past the end of the maximum run */
5609 
5610         if (possessive) continue;    /* No backtracking */
5611         for(;;)
5612           {
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);
5618           }
5619         }
5620 
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. */
5623 
5624       else if (ctype == OP_EXTUNI)
5625         {
5626         for (i = min; i < max; i++)
5627           {
5628           if (eptr >= md->end_subject)
5629             {
5630             SCHECK_PARTIAL();
5631             break;
5632             }
5633           else
5634             {
5635             int lgb, rgb;
5636             GETCHARINCTEST(c, eptr);
5637             lgb = UCD_GRAPHBREAK(c);
5638             while (eptr < md->end_subject)
5639               {
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;
5646               }
5647             }
5648           CHECK_PARTIAL();
5649           }
5650 
5651         /* eptr is now past the end of the maximum run */
5652 
5653         if (possessive) continue;    /* No backtracking */
5654 
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. */
5659 
5660         for(;;)
5661           {
5662           int lgb, rgb;
5663           PCRE_PUCHAR fptr;
5664 
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);
5668 
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. */
5672 
5673           eptr--;
5674           if (!utf) c = *eptr; else
5675             {
5676             BACKCHAR(eptr);
5677             GETCHAR(c, eptr);
5678             }
5679           rgb = UCD_GRAPHBREAK(c);
5680 
5681           for (;;)
5682             {
5683             if (eptr <= pp) goto TAIL_RECURSE;   /* At start of char run */
5684             fptr = eptr - 1;
5685             if (!utf) c = *fptr; else
5686               {
5687               BACKCHAR(fptr);
5688               GETCHAR(c, fptr);
5689               }
5690             lgb = UCD_GRAPHBREAK(c);
5691             if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
5692             eptr = fptr;
5693             rgb = lgb;
5694             }
5695           }
5696         }
5697 
5698       else
5699 #endif   /* SUPPORT_UCP */
5700 
5701 #ifdef SUPPORT_UTF
5702       if (utf)
5703         {
5704         switch(ctype)
5705           {
5706           case OP_ANY:
5707           for (i = min; i < max; i++)
5708             {
5709             if (eptr >= md->end_subject)
5710               {
5711               SCHECK_PARTIAL();
5712               break;
5713               }
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])
5720               {
5721               md->hitend = TRUE;
5722               if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5723               }
5724             eptr++;
5725             ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5726             }
5727           break;
5728 
5729           case OP_ALLANY:
5730           if (max < INT_MAX)
5731             {
5732             for (i = min; i < max; i++)
5733               {
5734               if (eptr >= md->end_subject)
5735                 {
5736                 SCHECK_PARTIAL();
5737                 break;
5738                 }
5739               eptr++;
5740               ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
5741               }
5742             }
5743           else
5744             {
5745             eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
5746             SCHECK_PARTIAL();
5747             }
5748           break;
5749 
5750           /* The byte case is the same as non-UTF8 */
5751 
5752           case OP_ANYBYTE:
5753           c = max - min;
5754           if (c > (unsigned int)(md->end_subject - eptr))
5755             {
5756             eptr = md->end_subject;
5757             SCHECK_PARTIAL();
5758             }
5759           else eptr += c;
5760           break;
5761 
5762           case OP_ANYNL:
5763           for (i = min; i < max; i++)
5764             {
5765             int len = 1;
5766             if (eptr >= md->end_subject)
5767               {
5768               SCHECK_PARTIAL();
5769               break;
5770               }
5771             GETCHARLEN(c, eptr, len);
5772             if (c == CHAR_CR)
5773               {
5774               if (++eptr >= md->end_subject) break;
5775               if (UCHAR21(eptr) == CHAR_LF) eptr++;
5776               }
5777             else
5778               {
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 */
5785                     )))
5786                 break;
5787               eptr += len;
5788               }
5789             }
5790           break;
5791 
5792           case OP_NOT_HSPACE:
5793           case OP_HSPACE:
5794           for (i = min; i < max; i++)
5795             {
5796             BOOL gotspace;
5797             int len = 1;
5798             if (eptr >= md->end_subject)
5799               {
5800               SCHECK_PARTIAL();
5801               break;
5802               }
5803             GETCHARLEN(c, eptr, len);
5804             switch(c)
5805               {
5806               HSPACE_CASES: gotspace = TRUE; break;
5807               default: gotspace = FALSE; break;
5808               }
5809             if (gotspace == (ctype == OP_NOT_HSPACE)) break;
5810             eptr += len;
5811             }
5812           break;
5813 
5814           case OP_NOT_VSPACE:
5815           case OP_VSPACE:
5816           for (i = min; i < max; i++)
5817             {
5818             BOOL gotspace;
5819             int len = 1;
5820             if (eptr >= md->end_subject)
5821               {
5822               SCHECK_PARTIAL();
5823               break;
5824               }
5825             GETCHARLEN(c, eptr, len);
5826             switch(c)
5827               {
5828               VSPACE_CASES: gotspace = TRUE; break;
5829               default: gotspace = FALSE; break;
5830               }
5831             if (gotspace == (ctype == OP_NOT_VSPACE)) break;
5832             eptr += len;
5833             }
5834           break;
5835 
5836           case OP_NOT_DIGIT:
5837           for (i = min; i < max; i++)
5838             {
5839             int len = 1;
5840             if (eptr >= md->end_subject)
5841               {
5842               SCHECK_PARTIAL();
5843               break;
5844               }
5845             GETCHARLEN(c, eptr, len);
5846             if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
5847             eptr+= len;
5848             }
5849           break;
5850 
5851           case OP_DIGIT:
5852           for (i = min; i < max; i++)
5853             {
5854             int len = 1;
5855             if (eptr >= md->end_subject)
5856               {
5857               SCHECK_PARTIAL();
5858               break;
5859               }
5860             GETCHARLEN(c, eptr, len);
5861             if (c >= 256 ||(md->ctypes[c] & ctype_digit) == 0) break;
5862             eptr+= len;
5863             }
5864           break;
5865 
5866           case OP_NOT_WHITESPACE:
5867           for (i = min; i < max; i++)
5868             {
5869             int len = 1;
5870             if (eptr >= md->end_subject)
5871               {
5872               SCHECK_PARTIAL();
5873               break;
5874               }
5875             GETCHARLEN(c, eptr, len);
5876             if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
5877             eptr+= len;
5878             }
5879           break;
5880 
5881           case OP_WHITESPACE:
5882           for (i = min; i < max; i++)
5883             {
5884             int len = 1;
5885             if (eptr >= md->end_subject)
5886               {
5887               SCHECK_PARTIAL();
5888               break;
5889               }
5890             GETCHARLEN(c, eptr, len);
5891             if (c >= 256 ||(md->ctypes[c] & ctype_space) == 0) break;
5892             eptr+= len;
5893             }
5894           break;
5895 
5896           case OP_NOT_WORDCHAR:
5897           for (i = min; i < max; i++)
5898             {
5899             int len = 1;
5900             if (eptr >= md->end_subject)
5901               {
5902               SCHECK_PARTIAL();
5903               break;
5904               }
5905             GETCHARLEN(c, eptr, len);
5906             if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
5907             eptr+= len;
5908             }
5909           break;
5910 
5911           case OP_WORDCHAR:
5912           for (i = min; i < max; i++)
5913             {
5914             int len = 1;
5915             if (eptr >= md->end_subject)
5916               {
5917               SCHECK_PARTIAL();
5918               break;
5919               }
5920             GETCHARLEN(c, eptr, len);
5921             if (c >= 256 || (md->ctypes[c] & ctype_word) == 0) break;
5922             eptr+= len;
5923             }
5924           break;
5925 
5926           default:
5927           RRETURN(PCRE_ERROR_INTERNAL);
5928           }
5929 
5930         if (possessive) continue;    /* No backtracking */
5931         for(;;)
5932           {
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--;
5940           }
5941         }
5942       else
5943 #endif  /* SUPPORT_UTF */
5944       /* Not UTF mode */
5945         {
5946         switch(ctype)
5947           {
5948           case OP_ANY:
5949           for (i = min; i < max; i++)
5950             {
5951             if (eptr >= md->end_subject)
5952               {
5953               SCHECK_PARTIAL();
5954               break;
5955               }
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])
5962               {
5963               md->hitend = TRUE;
5964               if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
5965               }
5966             eptr++;
5967             }
5968           break;
5969 
5970           case OP_ALLANY:
5971           case OP_ANYBYTE:
5972           c = max - min;
5973           if (c > (unsigned int)(md->end_subject - eptr))
5974             {
5975             eptr = md->end_subject;
5976             SCHECK_PARTIAL();
5977             }
5978           else eptr += c;
5979           break;
5980 
5981           case OP_ANYNL:
5982           for (i = min; i < max; i++)
5983             {
5984             if (eptr >= md->end_subject)
5985               {
5986               SCHECK_PARTIAL();
5987               break;
5988               }
5989             c = *eptr;
5990             if (c == CHAR_CR)
5991               {
5992               if (++eptr >= md->end_subject) break;
5993               if (*eptr == CHAR_LF) eptr++;
5994               }
5995             else
5996               {
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++;
6004               }
6005             }
6006           break;
6007 
6008           case OP_NOT_HSPACE:
6009           for (i = min; i < max; i++)
6010             {
6011             if (eptr >= md->end_subject)
6012               {
6013               SCHECK_PARTIAL();
6014               break;
6015               }
6016             switch(*eptr)
6017               {
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;
6024               }
6025             }
6026           ENDLOOP00:
6027           break;
6028 
6029           case OP_HSPACE:
6030           for (i = min; i < max; i++)
6031             {
6032             if (eptr >= md->end_subject)
6033               {
6034               SCHECK_PARTIAL();
6035               break;
6036               }
6037             switch(*eptr)
6038               {
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;
6045               }
6046             }
6047           ENDLOOP01:
6048           break;
6049 
6050           case OP_NOT_VSPACE:
6051           for (i = min; i < max; i++)
6052             {
6053             if (eptr >= md->end_subject)
6054               {
6055               SCHECK_PARTIAL();
6056               break;
6057               }
6058             switch(*eptr)
6059               {
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;
6066               }
6067             }
6068           ENDLOOP02:
6069           break;
6070 
6071           case OP_VSPACE:
6072           for (i = min; i < max; i++)
6073             {
6074             if (eptr >= md->end_subject)
6075               {
6076               SCHECK_PARTIAL();
6077               break;
6078               }
6079             switch(*eptr)
6080               {
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;
6087               }
6088             }
6089           ENDLOOP03:
6090           break;
6091 
6092           case OP_NOT_DIGIT:
6093           for (i = min; i < max; i++)
6094             {
6095             if (eptr >= md->end_subject)
6096               {
6097               SCHECK_PARTIAL();
6098               break;
6099               }
6100             if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_digit) != 0) break;
6101             eptr++;
6102             }
6103           break;
6104 
6105           case OP_DIGIT:
6106           for (i = min; i < max; i++)
6107             {
6108             if (eptr >= md->end_subject)
6109               {
6110               SCHECK_PARTIAL();
6111               break;
6112               }
6113             if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_digit) == 0) break;
6114             eptr++;
6115             }
6116           break;
6117 
6118           case OP_NOT_WHITESPACE:
6119           for (i = min; i < max; i++)
6120             {
6121             if (eptr >= md->end_subject)
6122               {
6123               SCHECK_PARTIAL();
6124               break;
6125               }
6126             if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_space) != 0) break;
6127             eptr++;
6128             }
6129           break;
6130 
6131           case OP_WHITESPACE:
6132           for (i = min; i < max; i++)
6133             {
6134             if (eptr >= md->end_subject)
6135               {
6136               SCHECK_PARTIAL();
6137               break;
6138               }
6139             if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_space) == 0) break;
6140             eptr++;
6141             }
6142           break;
6143 
6144           case OP_NOT_WORDCHAR:
6145           for (i = min; i < max; i++)
6146             {
6147             if (eptr >= md->end_subject)
6148               {
6149               SCHECK_PARTIAL();
6150               break;
6151               }
6152             if (MAX_255(*eptr) && (md->ctypes[*eptr] & ctype_word) != 0) break;
6153             eptr++;
6154             }
6155           break;
6156 
6157           case OP_WORDCHAR:
6158           for (i = min; i < max; i++)
6159             {
6160             if (eptr >= md->end_subject)
6161               {
6162               SCHECK_PARTIAL();
6163               break;
6164               }
6165             if (!MAX_255(*eptr) || (md->ctypes[*eptr] & ctype_word) == 0) break;
6166             eptr++;
6167             }
6168           break;
6169 
6170           default:
6171           RRETURN(PCRE_ERROR_INTERNAL);
6172           }
6173 
6174         if (possessive) continue;    /* No backtracking */
6175         for (;;)
6176           {
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--;
6183           }
6184         }
6185 
6186       /* Control never gets here */
6187       }
6188 
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. */
6191 
6192     default:
6193     DPRINTF(("Unknown opcode %d\n", *ecode));
6194     RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
6195     }
6196 
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. */
6200 
6201   }             /* End of main loop */
6202 /* Control never reaches here */
6203 
6204 
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. */
6208 
6209 #ifdef NO_RECURSE
6210 #define LBL(val) case val: goto L_RM##val;
6211 HEAP_RETURN:
6212 switch (frame->Xwhere)
6213   {
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;
6235   }
6236 #undef LBL
6237 #endif  /* NO_RECURSE */
6238 }
6239 
6240 
6241 /***************************************************************************
6242 ****************************************************************************
6243                    RECURSION IN THE match() FUNCTION
6244 
6245 Undefine all the macros that were defined above to handle this. */
6246 
6247 #ifdef NO_RECURSE
6248 #undef eptr
6249 #undef ecode
6250 #undef mstart
6251 #undef offset_top
6252 #undef eptrb
6253 #undef flags
6254 
6255 #undef callpat
6256 #undef charptr
6257 #undef data
6258 #undef next
6259 #undef pp
6260 #undef prev
6261 #undef saved_eptr
6262 
6263 #undef new_recursive
6264 
6265 #undef cur_is_word
6266 #undef condition
6267 #undef prev_is_word
6268 
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
6281 
6282 #undef newptrb
6283 
6284 #endif
6285 
6286 /* These two are defined as macros in both cases */
6287 
6288 #undef fc
6289 #undef fi
6290 
6291 /***************************************************************************
6292 ***************************************************************************/
6293 
6294 
6295 #ifdef NO_RECURSE
6296 /*************************************************
6297 *          Release allocated heap frames         *
6298 *************************************************/
6299 
6300 /* This function releases all the allocated frames. The base frame is on the
6301 machine stack, and so must not be freed.
6302 
6303 Argument: the address of the base frame
6304 Returns:  nothing
6305 */
6306 
6307 static void
release_match_heapframes(heapframe * frame_base)6308 release_match_heapframes (heapframe *frame_base)
6309 {
6310 heapframe *nextframe = frame_base->Xnextframe;
6311 while (nextframe != NULL)
6312   {
6313   heapframe *oldframe = nextframe;
6314   nextframe = nextframe->Xnextframe;
6315   (PUBL(stack_free))(oldframe);
6316   }
6317 }
6318 #endif
6319 
6320 
6321 /*************************************************
6322 *         Execute a Regular Expression           *
6323 *************************************************/
6324 
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.
6328 
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
6338 
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
6343 */
6344 
6345 #if defined COMPILE_PCRE8
6346 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
pcre_exec(const pcre * argument_re,const pcre_extra * extra_data,PCRE_SPTR subject,int length,int start_offset,int options,int * offsets,int offsetcount)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
6361 {
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;
6384 
6385 const pcre_study_data *study;
6386 const REAL_PCRE *re = (const REAL_PCRE *)argument_re;
6387 
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
6394 
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. */
6399 
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
6407 
6408 /* Plausibility checks */
6409 
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;
6416 
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. */
6421 
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;
6426 
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. */
6431 
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;
6436 
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. */
6439 
6440 #ifdef SUPPORT_UTF
6441 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0)
6442   {
6443   int erroroffset;
6444   int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset);
6445   if (errorcode != 0)
6446     {
6447     if (offsetcount >= 2)
6448       {
6449       offsets[0] = erroroffset;
6450       offsets[1] = errorcode;
6451       }
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
6461     }
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
6468   }
6469 #endif
6470 
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. */
6475 
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)
6482   {
6483   rc = PRIV(jit_exec)(extra_data, (const pcre_uchar *)subject, length,
6484        start_offset, options, offsets, offsetcount);
6485 
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. */
6488 
6489   if (rc != PCRE_ERROR_JIT_BADOPTION) return rc;
6490   }
6491 #endif
6492 
6493 /* Carry on with non-JIT matching. This information is for finding all the
6494 numbers associated with a given name, for condition testing. */
6495 
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;
6499 
6500 /* Fish out the optional data from the extra_data structure, first setting
6501 the default values. */
6502 
6503 study = NULL;
6504 md->match_limit = MATCH_LIMIT;
6505 md->match_limit_recursion = MATCH_LIMIT_RECURSION;
6506 md->callout_data = NULL;
6507 
6508 /* The table pointer is always in native byte order. */
6509 
6510 tables = re->tables;
6511 
6512 /* The two limit values override the defaults, whatever their value. */
6513 
6514 if (extra_data != NULL)
6515   {
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;
6526   }
6527 
6528 /* Limits in the regex override only if they are smaller. */
6529 
6530 if ((re->flags & PCRE_MLSET) != 0 && re->limit_match < md->match_limit)
6531   md->match_limit = re->limit_match;
6532 
6533 if ((re->flags & PCRE_RLSET) != 0 &&
6534     re->limit_recursion < md->match_limit_recursion)
6535   md->match_limit_recursion = re->limit_recursion;
6536 
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. */
6540 
6541 if (tables == NULL) tables = PRIV(default_tables);
6542 
6543 /* Set up other data */
6544 
6545 anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
6546 startline = (re->flags & PCRE_STARTLINE) != 0;
6547 firstline = (re->options & PCRE_FIRSTLINE) != 0;
6548 
6549 /* The code starts after the real_pcre block and the capture name table. */
6550 
6551 md->start_code = (const pcre_uchar *)re + re->name_table_offset +
6552   re->name_count * re->name_entry_size;
6553 
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;
6558 
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;
6563 
6564 /* Some options are unpacked into BOOL variables in the hope that testing
6565 them will be faster than individual option bits. */
6566 
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;
6571 
6572 md->hitend = FALSE;
6573 md->mark = md->nomatch_mark = NULL;     /* In case never set */
6574 
6575 md->recursive = NULL;                   /* No recursion at top level */
6576 md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6577 
6578 md->lcc = tables + lcc_offset;
6579 md->fcc = tables + fcc_offset;
6580 md->ctypes = tables + ctypes_offset;
6581 
6582 /* Handle different \R options. */
6583 
6584 switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6585   {
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;
6596 
6597   case PCRE_BSR_ANYCRLF:
6598   md->bsr_anycrlf = TRUE;
6599   break;
6600 
6601   case PCRE_BSR_UNICODE:
6602   md->bsr_anycrlf = FALSE;
6603   break;
6604 
6605   default: return PCRE_ERROR_BADNEWLINE;
6606   }
6607 
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. */
6610 
6611 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
6612         (pcre_uint32)options) & PCRE_NEWLINE_BITS)
6613   {
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;
6622   }
6623 
6624 if (newline == -2)
6625   {
6626   md->nltype = NLTYPE_ANYCRLF;
6627   }
6628 else if (newline < 0)
6629   {
6630   md->nltype = NLTYPE_ANY;
6631   }
6632 else
6633   {
6634   md->nltype = NLTYPE_FIXED;
6635   if (newline > 255)
6636     {
6637     md->nllen = 2;
6638     md->nl[0] = (newline >> 8) & 255;
6639     md->nl[1] = newline & 255;
6640     }
6641   else
6642     {
6643     md->nllen = 1;
6644     md->nl[0] = newline;
6645     }
6646   }
6647 
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. */
6651 
6652 if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6653   return PCRE_ERROR_BADPARTIAL;
6654 
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. */
6659 
6660 ocount = offsetcount - (offsetcount % 3);
6661 arg_offset_max = (2*ocount)/3;
6662 
6663 if (re->top_backref > 0 && re->top_backref >= ocount/3)
6664   {
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"));
6670   }
6671 else md->offset_vector = offsets;
6672 md->offset_end = ocount;
6673 md->offset_max = (2*ocount)/3;
6674 md->capture_last = 0;
6675 
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. */
6681 
6682 if (md->offset_vector != NULL)
6683   {
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;
6690   }
6691 
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. */
6697 
6698 if (!anchored)
6699   {
6700   if ((re->flags & PCRE_FIRSTSET) != 0)
6701     {
6702     has_first_char = TRUE;
6703     first_char = first_char2 = (pcre_uchar)(re->first_char);
6704     if ((re->flags & PCRE_FCH_CASELESS) != 0)
6705       {
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
6711       }
6712     }
6713   else
6714     if (!startline && study != NULL &&
6715       (study->flags & PCRE_STUDY_MAPPED) != 0)
6716         start_bits = study->start_bits;
6717   }
6718 
6719 /* For anchored or unanchored matches, there may be a "last known required
6720 character" set. */
6721 
6722 if ((re->flags & PCRE_REQCHSET) != 0)
6723   {
6724   has_req_char = TRUE;
6725   req_char = req_char2 = (pcre_uchar)(re->req_char);
6726   if ((re->flags & PCRE_RCH_CASELESS) != 0)
6727     {
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
6733     }
6734   }
6735 
6736 
6737 /* ==========================================================================*/
6738 
6739 /* Loop for handling unanchored repeated matching attempts; for anchored regexs
6740 the loop runs just once. */
6741 
6742 for(;;)
6743   {
6744   PCRE_PUCHAR save_end_subject = end_subject;
6745   PCRE_PUCHAR new_start_match;
6746 
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. */
6752 
6753   if (firstline)
6754     {
6755     PCRE_PUCHAR t = start_match;
6756 #ifdef SUPPORT_UTF
6757     if (utf)
6758       {
6759       while (t < md->end_subject && !IS_NEWLINE(t))
6760         {
6761         t++;
6762         ACROSSCHAR(t < end_subject, *t, t++);
6763         }
6764       }
6765     else
6766 #endif
6767     while (t < md->end_subject && !IS_NEWLINE(t)) t++;
6768     end_subject = t;
6769     }
6770 
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. */
6776 
6777   if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0)
6778     {
6779     /* Advance to a unique first char if there is one. */
6780 
6781     if (has_first_char)
6782       {
6783       pcre_uchar smc;
6784 
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++;
6792       }
6793 
6794     /* Or to just after a linebreak for a multiline match */
6795 
6796     else if (startline)
6797       {
6798       if (start_match > md->start_subject + start_offset)
6799         {
6800 #ifdef SUPPORT_UTF
6801         if (utf)
6802           {
6803           while (start_match < end_subject && !WAS_NEWLINE(start_match))
6804             {
6805             start_match++;
6806             ACROSSCHAR(start_match < end_subject, *start_match,
6807               start_match++);
6808             }
6809           }
6810         else
6811 #endif
6812         while (start_match < end_subject && !WAS_NEWLINE(start_match))
6813           start_match++;
6814 
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.
6817         */
6818 
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++;
6824         }
6825       }
6826 
6827     /* Or to a non-unique first byte after study */
6828 
6829     else if (start_bits != NULL)
6830       {
6831       while (start_match < end_subject)
6832         {
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++;
6839         }
6840       }
6841     }   /* Starting optimizations */
6842 
6843   /* Restore fudged end_subject */
6844 
6845   end_subject = save_end_subject;
6846 
6847   /* The following two optimizations are disabled for partial matching or if
6848   disabling is explicitly requested. */
6849 
6850   if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0 && !md->partial)
6851     {
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. */
6856 
6857     if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
6858         (pcre_uint32)(end_subject - start_match) < study->minlength)
6859       {
6860       rc = MATCH_NOMATCH;
6861       break;
6862       }
6863 
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.
6871 
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. */
6877 
6878     if (has_req_char && end_subject - start_match < REQ_BYTE_MAX)
6879       {
6880       register PCRE_PUCHAR p = start_match + (has_first_char? 1:0);
6881 
6882       /* We don't need to repeat the search if we haven't yet reached the
6883       place we found it at last time. */
6884 
6885       if (p > req_char_ptr)
6886         {
6887         if (req_char != req_char2)
6888           {
6889           while (p < end_subject)
6890             {
6891             register pcre_uint32 pp = UCHAR21INCTEST(p);
6892             if (pp == req_char || pp == req_char2) { p--; break; }
6893             }
6894           }
6895         else
6896           {
6897           while (p < end_subject)
6898             {
6899             if (UCHAR21INCTEST(p) == req_char) { p--; break; }
6900             }
6901           }
6902 
6903         /* If we can't find the required character, break the matching loop,
6904         forcing a match failure. */
6905 
6906         if (p >= end_subject)
6907           {
6908           rc = MATCH_NOMATCH;
6909           break;
6910           }
6911 
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. */
6915 
6916         req_char_ptr = p;
6917         }
6918       }
6919     }
6920 
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
6926 
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. */
6929 
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)
6938     {
6939     start_partial = md->start_used_ptr;
6940     match_partial = start_match;
6941     }
6942 
6943   switch(rc)
6944     {
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. */
6951 
6952     case MATCH_SKIP_ARG:
6953     new_start_match = start_match;
6954     md->ignore_skip_arg = md->skip_arg_count;
6955     break;
6956 
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. */
6959 
6960     case MATCH_SKIP:
6961     if (md->start_match_ptr > start_match)
6962       {
6963       new_start_match = md->start_match_ptr;
6964       break;
6965       }
6966     /* Fall through */
6967 
6968     /* NOMATCH and PRUNE advance by one character. THEN at this level acts
6969     exactly like PRUNE. Unset ignore SKIP-with-argument. */
6970 
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;
6982 
6983     /* COMMIT disables the bumpalong, but otherwise behaves as NOMATCH. */
6984 
6985     case MATCH_COMMIT:
6986     rc = MATCH_NOMATCH;
6987     goto ENDLOOP;
6988 
6989     /* Any other return is either a match, or some kind of error. */
6990 
6991     default:
6992     goto ENDLOOP;
6993     }
6994 
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. */
6997 
6998   rc = MATCH_NOMATCH;
6999 
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. */
7003 
7004   if (firstline && IS_NEWLINE(start_match)) break;
7005 
7006   /* Advance to new matching position */
7007 
7008   start_match = new_start_match;
7009 
7010   /* Break the loop if the pattern is anchored or if we have passed the end of
7011   the subject. */
7012 
7013   if (anchored || start_match > end_subject) break;
7014 
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. */
7021 
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++;
7031 
7032   md->mark = NULL;   /* Reset for start of next match attempt */
7033   }                  /* End of for(;;) "bumpalong" loop */
7034 
7035 /* ==========================================================================*/
7036 
7037 /* We reach here when rc is not MATCH_NOMATCH, or if one of the stopping
7038 conditions is true:
7039 
7040 (1) The pattern is anchored or the match was failed by (*COMMIT);
7041 
7042 (2) We are past the end of the subject;
7043 
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.
7047 
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. */
7054 
7055 ENDLOOP:
7056 
7057 if (rc == MATCH_MATCH || rc == MATCH_ACCEPT)
7058   {
7059   if (using_temporary_offsets)
7060     {
7061     if (arg_offset_max >= 4)
7062       {
7063       memcpy(offsets + 2, md->offset_vector + 2,
7064         (arg_offset_max - 2) * sizeof(int));
7065       DPRINTF(("Copied offsets from temporary memory\n"));
7066       }
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);
7070     }
7071 
7072   /* Set the return code to the number of captured strings, or 0 if there were
7073   too many to fit into the vector. */
7074 
7075   rc = ((md->capture_last & OVFLBIT) != 0 &&
7076          md->end_offset_top >= arg_offset_max)?
7077     0 : md->end_offset_top/2;
7078 
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. */
7087 
7088   if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
7089     {
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;
7096     }
7097 
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. */
7101 
7102   if (offsetcount < 2) rc = 0; else
7103     {
7104     offsets[0] = (int)(md->start_match_ptr - md->start_subject);
7105     offsets[1] = (int)(md->end_match_ptr - md->start_subject);
7106     }
7107 
7108   /* Return MARK data if requested */
7109 
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;
7117   }
7118 
7119 /* Control gets here if there has been an error, or if the overall match
7120 attempt has failed at all permitted starting positions. */
7121 
7122 if (using_temporary_offsets)
7123   {
7124   DPRINTF(("Freeing temporary memory\n"));
7125   (PUBL(free))(md->offset_vector);
7126   }
7127 
7128 /* For anything other than nomatch or partial match, just return the code. */
7129 
7130 if (rc != MATCH_NOMATCH && rc != PCRE_ERROR_PARTIAL)
7131   {
7132   DPRINTF((">>>> error: returning %d\n", rc));
7133 #ifdef NO_RECURSE
7134   release_match_heapframes(&frame_zero);
7135 #endif
7136   return rc;
7137   }
7138 
7139 /* Handle partial matches - disable any mark data */
7140 
7141 if (match_partial != NULL)
7142   {
7143   DPRINTF((">>>> returning PCRE_ERROR_PARTIAL\n"));
7144   md->mark = NULL;
7145   if (offsetcount > 1)
7146     {
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);
7151     }
7152   rc = PCRE_ERROR_PARTIAL;
7153   }
7154 
7155 /* This is the classic nomatch case */
7156 
7157 else
7158   {
7159   DPRINTF((">>>> returning PCRE_ERROR_NOMATCH\n"));
7160   rc = PCRE_ERROR_NOMATCH;
7161   }
7162 
7163 /* Return the MARK data if it has been requested. */
7164 
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;
7171 }
7172 
7173 /* End of pcre_exec.c */
7174