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

Diff of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 550 by ph10, Sun Oct 10 16:24:11 2010 UTC revision 716 by ph10, Tue Oct 4 16:38:05 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 57  possible. There are also some static sup Line 57  possible. There are also some static sup
57  #undef min  #undef min
58  #undef max  #undef max
59    
60  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
61    of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
63    
64  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 73  negative to avoid the external error cod Line 75  negative to avoid the external error cod
75    
76  #define MATCH_ACCEPT       (-999)  #define MATCH_ACCEPT       (-999)
77  #define MATCH_COMMIT       (-998)  #define MATCH_COMMIT       (-998)
78  #define MATCH_PRUNE        (-997)  #define MATCH_KETRPOS      (-997)
79  #define MATCH_SKIP         (-996)  #define MATCH_ONCE         (-996)
80  #define MATCH_SKIP_ARG     (-995)  #define MATCH_PRUNE        (-995)
81  #define MATCH_THEN         (-994)  #define MATCH_SKIP         (-994)
82    #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85  /* This is a convenience macro for code that occurs many times. */  /* This is a convenience macro for code that occurs many times. */
86    
# Line 132  while (length-- > 0) Line 136  while (length-- > 0)
136  *          Match a back-reference                *  *          Match a back-reference                *
137  *************************************************/  *************************************************/
138    
139  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
140  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
141    the length passed is zero. Note that in caseless UTF-8 mode, the number of
142    subject bytes matched may be different to the number of reference bytes.
143    
144  Arguments:  Arguments:
145    offset      index into the offset vector    offset      index into the offset vector
146    eptr        points into the subject    eptr        pointer into the subject
147    length      length to be matched    length      length of reference to be matched (number of bytes)
148    md          points to match data block    md          points to match data block
149    ims         the ims flags    caseless    TRUE if caseless
150    
151  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
152  */  */
153    
154  static BOOL  static int
155  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register USPTR eptr, int length, match_data *md,
156    unsigned long int ims)    BOOL caseless)
157  {  {
158  USPTR p = md->start_subject + md->offset_vector[offset];  USPTR eptr_start = eptr;
159    register USPTR p = md->start_subject + md->offset_vector[offset];
160    
161  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
162  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
# Line 164  pchars(p, length, FALSE, md); Line 171  pchars(p, length, FALSE, md);
171  printf("\n");  printf("\n");
172  #endif  #endif
173    
174  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
175    
176  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
177    
178  /* Separate the caseless case for speed. In UTF-8 mode we can only do this  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
179  properly if Unicode properties are supported. Otherwise, we can check only  properly if Unicode properties are supported. Otherwise, we can check only
180  ASCII characters. */  ASCII characters. */
181    
182  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
183    {    {
184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
185  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
186    if (md->utf8)    if (md->utf8)
187      {      {
188      USPTR endptr = eptr + length;      /* Match characters up to the end of the reference. NOTE: the number of
189      while (eptr < endptr)      bytes matched may differ, because there are some characters whose upper and
190        lower case versions code as different numbers of bytes. For example, U+023A
191        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
192        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
193        the latter. It is important, therefore, to check the length along the
194        reference, not along the subject (earlier code did this wrong). */
195    
196        USPTR endptr = p + length;
197        while (p < endptr)
198        {        {
199        int c, d;        int c, d;
200          if (eptr >= md->end_subject) return -1;
201        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
202        GETCHARINC(d, p);        GETCHARINC(d, p);
203        if (c != d && c != UCD_OTHERCASE(d)) return FALSE;        if (c != d && c != UCD_OTHERCASE(d)) return -1;
204        }        }
205      }      }
206    else    else
# Line 193  if ((ims & PCRE_CASELESS) != 0) Line 209  if ((ims & PCRE_CASELESS) != 0)
209    
210    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
211    is no UCP support. */    is no UCP support. */
212        {
213    while (length-- > 0)      if (eptr + length > md->end_subject) return -1;
214      { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }      while (length-- > 0)
215          { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }
216        }
217    }    }
218    
219  /* In the caseful case, we can just compare the bytes, whether or not we  /* In the caseful case, we can just compare the bytes, whether or not we
220  are in UTF-8 mode. */  are in UTF-8 mode. */
221    
222  else  else
223    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
224      if (eptr + length > md->end_subject) return -1;
225      while (length-- > 0) if (*p++ != *eptr++) return -1;
226      }
227    
228  return TRUE;  return eptr - eptr_start;
229  }  }
230    
231    
# Line 256  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 277  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
277         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
278         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
279         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
280         RM61,  RM62 };         RM61,  RM62, RM63 };
281    
282  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
283  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
# Line 266  actually used in this definition. */ Line 287  actually used in this definition. */
287  #define REGISTER register  #define REGISTER register
288    
289  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
290  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
291    { \    { \
292    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
293    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1); \
294    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
295    }    }
296  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 278  actually used in this definition. */ Line 299  actually used in this definition. */
299    return ra; \    return ra; \
300    }    }
301  #else  #else
302  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
303    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1)
304  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
305  #endif  #endif
306    
# Line 292  argument of match(), which never changes Line 313  argument of match(), which never changes
313    
314  #define REGISTER  #define REGISTER
315    
316  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
317    {\    {\
318    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\
319    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
320    frame->Xwhere = rw; \    frame->Xwhere = rw; \
321    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
# Line 302  argument of match(), which never changes Line 323  argument of match(), which never changes
323    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
324    newframe->Xmarkptr = markptr;\    newframe->Xmarkptr = markptr;\
325    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
326    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
327    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
328    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
329    frame = newframe;\    frame = newframe;\
# Line 340  typedef struct heapframe { Line 359  typedef struct heapframe {
359    USPTR Xmstart;    USPTR Xmstart;
360    USPTR Xmarkptr;    USPTR Xmarkptr;
361    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
362    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
363    unsigned int Xrdepth;    unsigned int Xrdepth;
364    
365    /* Function local variables */    /* Function local variables */
# Line 363  typedef struct heapframe { Line 380  typedef struct heapframe {
380    BOOL Xcondition;    BOOL Xcondition;
381    BOOL Xprev_is_word;    BOOL Xprev_is_word;
382    
   unsigned long int Xoriginal_ims;  
   
383  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
384    int Xprop_type;    int Xprop_type;
385    int Xprop_value;    int Xprop_value;
386    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
387    int Xoclength;    int Xoclength;
388    uschar Xocchars[8];    uschar Xocchars[8];
389  #endif  #endif
# Line 422  immediately. The second one is used when Line 434  immediately. The second one is used when
434  the subject. */  the subject. */
435    
436  #define CHECK_PARTIAL()\  #define CHECK_PARTIAL()\
437    if (md->partial != 0 && eptr >= md->end_subject && eptr > mstart)\    if (md->partial != 0 && eptr >= md->end_subject && \
438      {\        eptr > md->start_used_ptr) \
439      md->hitend = TRUE;\      { \
440      if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL);\      md->hitend = TRUE; \
441        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
442      }      }
443    
444  #define SCHECK_PARTIAL()\  #define SCHECK_PARTIAL()\
445    if (md->partial != 0 && eptr > mstart)\    if (md->partial != 0 && eptr > md->start_used_ptr) \
446      {\      { \
447      md->hitend = TRUE;\      md->hitend = TRUE; \
448      if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL);\      if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
449      }      }
450    
451    
# Line 449  Arguments: Line 462  Arguments:
462     markptr     pointer to the most recent MARK name, or NULL     markptr     pointer to the most recent MARK name, or NULL
463     offset_top  current top pointer     offset_top  current top pointer
464     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
465     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
466                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
467     rdepth      the recursion depth     rdepth      the recursion depth
468    
469  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
# Line 467  Returns:       MATCH_MATCH if matched Line 475  Returns:       MATCH_MATCH if matched
475    
476  static int  static int
477  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
478    const uschar *markptr, int offset_top, match_data *md, unsigned long int ims,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
479    eptrblock *eptrb, int flags, unsigned int rdepth)    unsigned int rdepth)
480  {  {
481  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
482  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 480  register unsigned int c;   /* Character Line 488  register unsigned int c;   /* Character
488  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
489    
490  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
491    BOOL caseless;
492  int condcode;  int condcode;
493    
494  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
# Line 488  heap storage. Set up the top-level frame Line 497  heap storage. Set up the top-level frame
497  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
498    
499  #ifdef NO_RECURSE  #ifdef NO_RECURSE
500  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));
501  if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);  if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
502  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
503    
# Line 499  frame->Xecode = ecode; Line 508  frame->Xecode = ecode;
508  frame->Xmstart = mstart;  frame->Xmstart = mstart;
509  frame->Xmarkptr = markptr;  frame->Xmarkptr = markptr;
510  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
511  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
512  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
513    
514  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 515  HEAP_RECURSE: Line 522  HEAP_RECURSE:
522  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
523  #define markptr            frame->Xmarkptr  #define markptr            frame->Xmarkptr
524  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
525  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
526  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
527    
528  /* Ditto for the local variables */  /* Ditto for the local variables */
# Line 539  HEAP_RECURSE: Line 544  HEAP_RECURSE:
544  #define condition          frame->Xcondition  #define condition          frame->Xcondition
545  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
546    
 #define original_ims       frame->Xoriginal_ims  
   
547  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
548  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
549  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
550  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
551  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
552  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
553  #endif  #endif
# Line 577  i, and fc and c, can be the same variabl Line 577  i, and fc and c, can be the same variabl
577  #define fi i  #define fi i
578  #define fc c  #define fc c
579    
580    /* Many of the following variables are used only in small blocks of the code.
581    My normal style of coding would have declared them within each of those blocks.
582    However, in order to accommodate the version of this code that uses an external
583    "stack" implemented on the heap, it is easier to declare them all here, so the
584    declarations can be cut out in a block. The only declarations within blocks
585    below are for variables that do not have to be preserved over a recursive call
586    to RMATCH(). */
587    
588    #ifdef SUPPORT_UTF8
589    const uschar *charptr;
590    #endif
591    const uschar *callpat;
592    const uschar *data;
593    const uschar *next;
594    USPTR         pp;
595    const uschar *prev;
596    USPTR         saved_eptr;
597    
598  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
599  const uschar *charptr;             /* in small blocks of the code. My normal */  
600  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
601  BOOL condition;  BOOL condition;
602  BOOL prev_is_word;  BOOL prev_is_word;
603    
 unsigned long int original_ims;  
   
604  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
605  int prop_type;  int prop_type;
606  int prop_value;  int prop_value;
607  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
608  int oclength;  int oclength;
609  uschar occhars[8];  uschar occhars[8];
610  #endif  #endif
# Line 622  int stacksave[REC_STACK_SAVE_MAX]; Line 624  int stacksave[REC_STACK_SAVE_MAX];
624  eptrblock newptrb;  eptrblock newptrb;
625  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
626    
627    /* To save space on the stack and in the heap frame, I have doubled up on some
628    of the local variables that are used only in localised parts of the code, but
629    still need to be preserved over recursive calls of match(). These macros define
630    the alternative names that are used. */
631    
632    #define allow_zero    cur_is_word
633    #define cbegroup      condition
634    #define code_offset   codelink
635    #define condassert    condition
636    #define matched_once  prev_is_word
637    
638  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
639  variables. */  variables. */
640    
# Line 658  haven't exceeded the recursive call limi Line 671  haven't exceeded the recursive call limi
671  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
672  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
673    
 original_ims = ims;    /* Save for resetting on ')' */  
   
674  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
675  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
676  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
677  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
678  When match() is called in other circumstances, don't add to the chain. The  
679  match_cbegroup flag must NOT be used with tail recursion, because the memory  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
680  block that is used is on the stack, so a new one may be required for each  such remembered pointers, to be checked when we hit the closing ket, in order
681  match(). */  to break infinite loops that match no characters. When match() is called in
682    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
683    NOT be used with tail recursion, because the memory block that is used is on
684    the stack, so a new one may be required for each match(). */
685    
686  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
687    {    {
688    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
689    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
690    eptrb = &newptrb;    eptrb = &newptrb;
691      md->match_function_type = 0;
692    }    }
693    
694  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 688  for (;;) Line 703  for (;;)
703      case OP_MARK:      case OP_MARK:
704      markptr = ecode + 2;      markptr = ecode + 2;
705      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
706        ims, eptrb, flags, RM55);        eptrb, RM55);
707    
708      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
709      argument, and we must check whether that argument matches this MARK's      argument, and we must check whether that argument matches this MARK's
# Line 710  for (;;) Line 725  for (;;)
725      case OP_FAIL:      case OP_FAIL:
726      MRRETURN(MATCH_NOMATCH);      MRRETURN(MATCH_NOMATCH);
727    
728        /* COMMIT overrides PRUNE, SKIP, and THEN */
729    
730      case OP_COMMIT:      case OP_COMMIT:
731      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
732        ims, eptrb, flags, RM52);        eptrb, RM52);
733      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
734            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
735            rrc != MATCH_THEN)
736          RRETURN(rrc);
737      MRRETURN(MATCH_COMMIT);      MRRETURN(MATCH_COMMIT);
738    
739        /* PRUNE overrides THEN */
740    
741      case OP_PRUNE:      case OP_PRUNE:
742      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
743        ims, eptrb, flags, RM51);        eptrb, RM51);
744      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
745      MRRETURN(MATCH_PRUNE);      MRRETURN(MATCH_PRUNE);
746    
747      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
748      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
749        ims, eptrb, flags, RM56);        eptrb, RM56);
750      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
751      md->mark = ecode + 2;      md->mark = ecode + 2;
752      RRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
753    
754        /* SKIP overrides PRUNE and THEN */
755    
756      case OP_SKIP:      case OP_SKIP:
757      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
758        ims, eptrb, flags, RM53);        eptrb, RM53);
759      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
760          RRETURN(rrc);
761      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
762      MRRETURN(MATCH_SKIP);      MRRETURN(MATCH_SKIP);
763    
764      case OP_SKIP_ARG:      case OP_SKIP_ARG:
765      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
766        ims, eptrb, flags, RM57);        eptrb, RM57);
767      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
768          RRETURN(rrc);
769    
770      /* Pass back the current skip name by overloading md->start_match_ptr and      /* Pass back the current skip name by overloading md->start_match_ptr and
771      returning the special MATCH_SKIP_ARG return code. This will either be      returning the special MATCH_SKIP_ARG return code. This will either be
# Line 748  for (;;) Line 774  for (;;)
774    
775      md->start_match_ptr = ecode + 2;      md->start_match_ptr = ecode + 2;
776      RRETURN(MATCH_SKIP_ARG);      RRETURN(MATCH_SKIP_ARG);
777    
778      /* For THEN (and THEN_ARG) we pass back the address of the bracket or      /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
779      the alt that is at the start of the current branch. This makes it possible      the branch in which it occurs can be determined. Overload the start of
780      to skip back past alternatives that precede the THEN within the current      match pointer to do this. */
     branch. */  
781    
782      case OP_THEN:      case OP_THEN:
783      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
784        ims, eptrb, flags, RM54);        eptrb, RM54);
785      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
786      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode;
787      MRRETURN(MATCH_THEN);      MRRETURN(MATCH_THEN);
788    
789      case OP_THEN_ARG:      case OP_THEN_ARG:
790      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top,
791        offset_top, md, ims, eptrb, flags, RM58);        md, eptrb, RM58);
792      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
793      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode;
794      md->mark = ecode + LINK_SIZE + 2;      md->mark = ecode + 2;
795      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
796    
797      /* Handle a capturing bracket. If there is space in the offset vector, save      /* Handle a capturing bracket, other than those that are possessive with an
798      the current subject position in the working slot at the top of the vector.      unlimited repeat. If there is space in the offset vector, save the current
799      We mustn't change the current values of the data slot, because they may be      subject position in the working slot at the top of the vector. We mustn't
800      set from a previous iteration of this group, and be referred to by a      change the current values of the data slot, because they may be set from a
801      reference inside the group.      previous iteration of this group, and be referred to by a reference inside
802        the group. A failure to match might occur after the group has succeeded,
803      If the bracket fails to match, we need to restore this value and also the      if something later on doesn't match. For this reason, we need to restore
804      values of the final offsets, in case they were set by a previous iteration      the working value and also the values of the final offsets, in case they
805      of the same bracket.      were set by a previous iteration of the same bracket.
806    
807      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
808      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 806  for (;;) Line 831  for (;;)
831        md->offset_vector[md->offset_end - number] =        md->offset_vector[md->offset_end - number] =
832          (int)(eptr - md->start_subject);          (int)(eptr - md->start_subject);
833    
834        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
835          {          {
836            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
837          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
838            ims, eptrb, flags, RM1);            eptrb, RM1);
839          if (rrc != MATCH_NOMATCH &&          if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
840              (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
841            RRETURN(rrc);          /* If we backed up to a THEN, check whether it is within the current
842            branch by comparing the address of the THEN that is passed back with
843            the end of the branch. If it is within the current branch, and the
844            branch is one of two or more alternatives (it either starts or ends
845            with OP_ALT), we have reached the limit of THEN's action, so convert
846            the return code to NOMATCH, which will cause normal backtracking to
847            happen from now on. Otherwise, THEN is passed back to an outer
848            alternative. This implements Perl's treatment of parenthesized groups,
849            where a group not containing | does not affect the current alternative,
850            that is, (X) is NOT the same as (X|(*F)). */
851    
852            if (rrc == MATCH_THEN)
853              {
854              next = ecode + GET(ecode,1);
855              if (md->start_match_ptr < next &&
856                  (*ecode == OP_ALT || *next == OP_ALT))
857                rrc = MATCH_NOMATCH;
858              }
859    
860            /* Anything other than NOMATCH is passed back. */
861    
862            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
863          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
864          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
865            if (*ecode != OP_ALT) break;
866          }          }
       while (*ecode == OP_ALT);  
867    
868        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
869        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
870        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
871        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
872    
873        if (rrc != MATCH_THEN) md->mark = markptr;        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
874        RRETURN(MATCH_NOMATCH);  
875          if (md->mark == NULL) md->mark = markptr;
876          RRETURN(rrc);
877        }        }
878    
879      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 840  for (;;) Line 887  for (;;)
887      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
888      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
889    
890      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
891      final alternative within the brackets, we would return the result of a      repeat. Loop for all the alternatives.
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion, except in the case when match_cbegroup  
     is set.*/  
892    
893        When we get to the final alternative within the brackets, we used to return
894        the result of a recursive call to match() whatever happened so it was
895        possible to reduce stack usage by turning this into a tail recursion,
896        except in the case of a possibly empty group. However, now that there is
897        the possiblity of (*THEN) occurring in the final alternative, this
898        optimization is no longer always possible.
899    
900        We can optimize if we know there are no (*THEN)s in the pattern; at present
901        this is the best that can be done.
902    
903        MATCH_ONCE is returned when the end of an atomic group is successfully
904        reached, but subsequent matching fails. It passes back up the tree (causing
905        captured values to be reset) until the original atomic group level is
906        reached. This is tested by comparing md->once_target with the start of the
907        group. At this point, the return is converted into MATCH_NOMATCH so that
908        previous backup points can be taken. */
909    
910        case OP_ONCE:
911      case OP_BRA:      case OP_BRA:
912      case OP_SBRA:      case OP_SBRA:
913      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
914      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
915      for (;;)      for (;;)
916        {        {
917        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
918    
919          /* If this is not a possibly empty group, and there are no (*THEN)s in
920          the pattern, and this is the final alternative, optimize as described
921          above. */
922    
923          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
924          {          {
925          if (flags == 0)    /* Not a possibly empty group */          ecode += _pcre_OP_lengths[*ecode];
926            goto TAIL_RECURSE;
927            }
928    
929          /* In all other cases, we have to make another call to match(). */
930    
931          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
932            RM2);
933    
934          /* See comment in the code for capturing groups above about handling
935          THEN. */
936    
937          if (rrc == MATCH_THEN)
938            {
939            next = ecode + GET(ecode,1);
940            if (md->start_match_ptr < next &&
941                (*ecode == OP_ALT || *next == OP_ALT))
942              rrc = MATCH_NOMATCH;
943            }
944    
945          if (rrc != MATCH_NOMATCH)
946            {
947            if (rrc == MATCH_ONCE)
948            {            {
949            ecode += _pcre_OP_lengths[*ecode];            const uschar *scode = ecode;
950            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
951            goto TAIL_RECURSE;              {
952                while (*scode == OP_ALT) scode += GET(scode, 1);
953                scode -= GET(scode, 1);
954                }
955              if (md->once_target == scode) rrc = MATCH_NOMATCH;
956            }            }
957            RRETURN(rrc);
958            }
959          ecode += GET(ecode, 1);
960          if (*ecode != OP_ALT) break;
961          }
962    
963        if (md->mark == NULL) md->mark = markptr;
964        RRETURN(MATCH_NOMATCH);
965    
966          /* Possibly empty group; can't use tail recursion. */      /* Handle possessive capturing brackets with an unlimited repeat. We come
967        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
968        handled similarly to the normal case above. However, the matching is
969        different. The end of these brackets will always be OP_KETRPOS, which
970        returns MATCH_KETRPOS without going further in the pattern. By this means
971        we can handle the group by iteration rather than recursion, thereby
972        reducing the amount of stack needed. */
973    
974        case OP_CBRAPOS:
975        case OP_SCBRAPOS:
976        allow_zero = FALSE;
977    
978          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      POSSESSIVE_CAPTURE:
979            eptrb, flags, RM48);      number = GET2(ecode, 1+LINK_SIZE);
980          if (rrc == MATCH_NOMATCH) md->mark = markptr;      offset = number << 1;
981          RRETURN(rrc);  
982    #ifdef PCRE_DEBUG
983        printf("start possessive bracket %d\n", number);
984        printf("subject=");
985        pchars(eptr, 16, TRUE, md);
986        printf("\n");
987    #endif
988    
989        if (offset < md->offset_max)
990          {
991          matched_once = FALSE;
992          code_offset = ecode - md->start_code;
993    
994          save_offset1 = md->offset_vector[offset];
995          save_offset2 = md->offset_vector[offset+1];
996          save_offset3 = md->offset_vector[md->offset_end - number];
997          save_capture_last = md->capture_last;
998    
999          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1000    
1001          /* Each time round the loop, save the current subject position for use
1002          when the group matches. For MATCH_MATCH, the group has matched, so we
1003          restart it with a new subject starting position, remembering that we had
1004          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1005          usual. If we haven't matched any alternatives in any iteration, check to
1006          see if a previous iteration matched. If so, the group has matched;
1007          continue from afterwards. Otherwise it has failed; restore the previous
1008          capture values before returning NOMATCH. */
1009    
1010          for (;;)
1011            {
1012            md->offset_vector[md->offset_end - number] =
1013              (int)(eptr - md->start_subject);
1014            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1015            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1016              eptrb, RM63);
1017            if (rrc == MATCH_KETRPOS)
1018              {
1019              offset_top = md->end_offset_top;
1020              eptr = md->end_match_ptr;
1021              ecode = md->start_code + code_offset;
1022              save_capture_last = md->capture_last;
1023              matched_once = TRUE;
1024              continue;
1025              }
1026    
1027            /* See comment in the code for capturing groups above about handling
1028            THEN. */
1029    
1030            if (rrc == MATCH_THEN)
1031              {
1032              next = ecode + GET(ecode,1);
1033              if (md->start_match_ptr < next &&
1034                  (*ecode == OP_ALT || *next == OP_ALT))
1035                rrc = MATCH_NOMATCH;
1036              }
1037    
1038            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1039            md->capture_last = save_capture_last;
1040            ecode += GET(ecode, 1);
1041            if (*ecode != OP_ALT) break;
1042          }          }
1043    
1044        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (!matched_once)
1045        otherwise return. */          {
1046            md->offset_vector[offset] = save_offset1;
1047            md->offset_vector[offset+1] = save_offset2;
1048            md->offset_vector[md->offset_end - number] = save_offset3;
1049            }
1050    
1051        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        if (md->mark == NULL) md->mark = markptr;
1052          eptrb, flags, RM2);        if (allow_zero || matched_once)
1053        if (rrc != MATCH_NOMATCH &&          {
1054            (rrc != MATCH_THEN || md->start_match_ptr != ecode))          ecode += 1 + LINK_SIZE;
1055          RRETURN(rrc);          break;
1056            }
1057    
1058          RRETURN(MATCH_NOMATCH);
1059          }
1060    
1061        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1062        as a non-capturing bracket. */
1063    
1064        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1065        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1066    
1067        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1068    
1069        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1070        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1071    
1072        /* Non-capturing possessive bracket with unlimited repeat. We come here
1073        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1074        without the capturing complication. It is written out separately for speed
1075        and cleanliness. */
1076    
1077        case OP_BRAPOS:
1078        case OP_SBRAPOS:
1079        allow_zero = FALSE;
1080    
1081        POSSESSIVE_NON_CAPTURE:
1082        matched_once = FALSE;
1083        code_offset = ecode - md->start_code;
1084    
1085        for (;;)
1086          {
1087          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1088          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1089            eptrb, RM48);
1090          if (rrc == MATCH_KETRPOS)
1091            {
1092            offset_top = md->end_offset_top;
1093            eptr = md->end_match_ptr;
1094            ecode = md->start_code + code_offset;
1095            matched_once = TRUE;
1096            continue;
1097            }
1098    
1099          /* See comment in the code for capturing groups above about handling
1100          THEN. */
1101    
1102          if (rrc == MATCH_THEN)
1103            {
1104            next = ecode + GET(ecode,1);
1105            if (md->start_match_ptr < next &&
1106                (*ecode == OP_ALT || *next == OP_ALT))
1107              rrc = MATCH_NOMATCH;
1108            }
1109    
1110          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1111        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1112          if (*ecode != OP_ALT) break;
1113          }
1114    
1115        if (matched_once || allow_zero)
1116          {
1117          ecode += 1 + LINK_SIZE;
1118          break;
1119        }        }
1120        RRETURN(MATCH_NOMATCH);
1121    
1122      /* Control never reaches here. */      /* Control never reaches here. */
1123    
1124      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1125      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1126      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1127      exactly what going to the ket would do. As there is only one branch to be      exactly what going to the ket would do. */
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1128    
1129      case OP_COND:      case OP_COND:
1130      case OP_SCOND:      case OP_SCOND:
1131      codelink= GET(ecode, 1);      codelink = GET(ecode, 1);
1132    
1133      /* Because of the way auto-callout works during compile, a callout item is      /* Because of the way auto-callout works during compile, a callout item is
1134      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
# Line 899  for (;;) Line 1138  for (;;)
1138        if (pcre_callout != NULL)        if (pcre_callout != NULL)
1139          {          {
1140          pcre_callout_block cb;          pcre_callout_block cb;
1141          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1142          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1143          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1144          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 911  for (;;) Line 1150  for (;;)
1150          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1151          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1152          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1153            cb.mark             = markptr;
1154          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1155          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1156          }          }
# Line 1064  for (;;) Line 1304  for (;;)
1304        }        }
1305    
1306      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1307      the final argument match_condassert causes it to stop at the end of an      md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1308      assertion. */      an assertion. */
1309    
1310      else      else
1311        {        {
1312        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1313            match_condassert, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1314        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1315          {          {
1316            if (md->end_offset_top > offset_top)
1317              offset_top = md->end_offset_top;  /* Captures may have happened */
1318          condition = TRUE;          condition = TRUE;
1319          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1320          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1321          }          }
1322        else if (rrc != MATCH_NOMATCH &&  
1323                (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1324          assertion; it is therefore treated as NOMATCH. */
1325    
1326          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1327          {          {
1328          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1329          }          }
# Line 1089  for (;;) Line 1334  for (;;)
1334          }          }
1335        }        }
1336    
1337      /* We are now at the branch that is to be obeyed. As there is only one,      /* We are now at the branch that is to be obeyed. As there is only one, can
1338      we can use tail recursion to avoid using another stack frame, except when      use tail recursion to avoid using another stack frame, except when there is
1339      match_cbegroup is required for an unlimited repeat of a possibly empty      unlimited repeat of a possibly empty group. In the latter case, a recursive
1340      group. If the second alternative doesn't exist, we can just plough on. */      call to match() is always required, unless the second alternative doesn't
1341        exist, in which case we can just plough on. Note that, for compatibility
1342        with Perl, the | in a conditional group is NOT treated as creating two
1343        alternatives. If a THEN is encountered in the branch, it propagates out to
1344        the enclosing alternative (unless nested in a deeper set of alternatives,
1345        of course). */
1346    
1347      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1348        {        {
1349        ecode += 1 + LINK_SIZE;        if (op != OP_SCOND)
       if (op == OP_SCOND)        /* Possibly empty group */  
         {  
         RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);  
         RRETURN(rrc);  
         }  
       else                       /* Group must match something */  
1350          {          {
1351          flags = 0;          ecode += 1 + LINK_SIZE;
1352          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1353          }          }
1354    
1355          md->match_function_type = MATCH_CBEGROUP;
1356          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1357          RRETURN(rrc);
1358        }        }
1359      else                         /* Condition false & no alternative */  
1360         /* Condition false & no alternative; continue after the group. */
1361    
1362        else
1363        {        {
1364        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1365        }        }
# Line 1139  for (;;) Line 1390  for (;;)
1390      break;      break;
1391    
1392    
1393      /* End of the pattern, either real or forced. If we are in a top-level      /* End of the pattern, either real or forced. */
     recursion, we should restore the offsets appropriately and continue from  
     after the call. */  
1394    
     case OP_ACCEPT:  
1395      case OP_END:      case OP_END:
1396      if (md->recursive != NULL && md->recursive->group_num == 0)      case OP_ACCEPT:
1397        {      case OP_ASSERT_ACCEPT:
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       offset_top = rec->save_offset_top;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1398    
1399      /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is      /* If we have matched an empty string, fail if not in an assertion and not
1400      set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1401      the subject. In both cases, backtracking will then try other alternatives,      is set and we have matched at the start of the subject. In both cases,
1402      if any. */      backtracking will then try other alternatives, if any. */
1403    
1404      if (eptr == mstart &&      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1405          (md->notempty ||           md->recursive == NULL &&
1406            (md->notempty_atstart &&           (md->notempty ||
1407              mstart == md->start_subject + md->start_offset)))             (md->notempty_atstart &&
1408                 mstart == md->start_subject + md->start_offset)))
1409        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1410    
1411      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
# Line 1181  for (;;) Line 1420  for (;;)
1420      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1421      MRRETURN(rrc);      MRRETURN(rrc);
1422    
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
   
1423      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1424      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1425      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1426      start of each branch to move the current point backwards, so the code at      start of each branch to move the current point backwards, so the code at
1427      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1428        of a condition, we want to return immediately afterwards. The caller of
1429        this incarnation of the match() function will have set MATCH_CONDASSERT in
1430        md->match_function type, and one of these opcodes will be the first opcode
1431        that is processed. We use a local variable that is preserved over calls to
1432        match() to remember this case. */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
1435      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1436        if (md->match_function_type == MATCH_CONDASSERT)
1437          {
1438          condassert = TRUE;
1439          md->match_function_type = 0;
1440          }
1441        else condassert = FALSE;
1442    
1443      do      do
1444        {        {
1445        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
         RM4);  
1446        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1447          {          {
1448          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
1449            markptr = md->mark;
1450          break;          break;
1451          }          }
1452        if (rrc != MATCH_NOMATCH &&  
1453            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1454          RRETURN(rrc);        as NOMATCH. */
1455    
1456          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1457        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1458        }        }
1459      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1460    
1461      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1462    
1463      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1464    
1465      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1466    
1467      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1468      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 1232  for (;;) Line 1478  for (;;)
1478    
1479      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1480      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1481        if (md->match_function_type == MATCH_CONDASSERT)
1482          {
1483          condassert = TRUE;
1484          md->match_function_type = 0;
1485          }
1486        else condassert = FALSE;
1487    
1488      do      do
1489        {        {
1490        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
         RM5);  
1491        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1492        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1493          {          {
1494          do ecode += GET(ecode,1); while (*ecode == OP_ALT);          do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1495          break;          break;
1496          }          }
1497        if (rrc != MATCH_NOMATCH &&  
1498            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1499          RRETURN(rrc);        as NOMATCH. */
1500    
1501          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1502        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1503        }        }
1504      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1505    
1506      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1507    
1508      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1509      continue;      continue;
# Line 1295  for (;;) Line 1549  for (;;)
1549      if (pcre_callout != NULL)      if (pcre_callout != NULL)
1550        {        {
1551        pcre_callout_block cb;        pcre_callout_block cb;
1552        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1553        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1554        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1555        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 1307  for (;;) Line 1561  for (;;)
1561        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1562        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1563        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1564          cb.mark             = markptr;
1565        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1566        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1567        }        }
# Line 1317  for (;;) Line 1572  for (;;)
1572      offset data is the offset to the starting bracket from the start of the      offset data is the offset to the starting bracket from the start of the
1573      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1574    
1575      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1576      save their starting points and reinstate them after the recursion. However,      re-instated afterwards. We don't know how many are started and not yet
1577      we don't know how many such there are (offset_top records the completed      finished (offset_top records the completed total) so we just have to save
1578      total) so we just have to save all the potential data. There may be up to      all the potential data. There may be up to 65535 such values, which is too
1579      65535 such values, which is too large to put on the stack, but using malloc      large to put on the stack, but using malloc for small numbers seems
1580      for small numbers seems expensive. As a compromise, the stack is used when      expensive. As a compromise, the stack is used when there are no more than
1581      there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc      REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
     is used. A problem is what to do if the malloc fails ... there is no way of  
     returning to the top level with an error. Save the top REC_STACK_SAVE_MAX  
     values on the stack, and accept that the rest may be wrong.  
1582    
1583      There are also other values that have to be saved. We use a chained      There are also other values that have to be saved. We use a chained
1584      sequence of blocks that actually live on the stack. Thanks to Robin Houston      sequence of blocks that actually live on the stack. Thanks to Robin Houston
1585      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1586        a lot, so he is not to blame for the current way it works. */
1587    
1588      case OP_RECURSE:      case OP_RECURSE:
1589        {        {
1590          recursion_info *ri;
1591          int recno;
1592    
1593        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1594        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1595          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1596    
1597          /* Check for repeating a recursion without advancing the subject pointer.
1598          This should catch convoluted mutual recursions. (Some simple cases are
1599          caught at compile time.) */
1600    
1601          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1602            if (recno == ri->group_num && eptr == ri->subject_position)
1603              RRETURN(PCRE_ERROR_RECURSELOOP);
1604    
1605        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1606    
1607          new_recursive.group_num = recno;
1608          new_recursive.subject_position = eptr;
1609        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1610        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1611    
1612        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1613    
1614        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1615    
1616        /* Now save the offset data. */        /* Now save the offset data */
1617    
1618        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1619        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1359  for (;;) Line 1624  for (;;)
1624            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1625          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1626          }          }
   
1627        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1628              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_offset_top = offset_top;  
1629    
1630        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1631        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1632          might be changed, so reset it before looping. */
1633    
1634        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1635        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1636        do        do
1637          {          {
1638            if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1639          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1640            md, ims, eptrb, flags, RM6);            md, eptrb, RM6);
1641            memcpy(md->offset_vector, new_recursive.offset_save,
1642                new_recursive.saved_max * sizeof(int));
1643            md->recursive = new_recursive.prevrec;
1644          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1645            {            {
1646            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1647            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1648              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1649            MRRETURN(MATCH_MATCH);  
1650              /* Set where we got to in the subject, and reset the start in case
1651              it was changed by \K. This *is* propagated back out of a recursion,
1652              for Perl compatibility. */
1653    
1654              eptr = md->end_match_ptr;
1655              mstart = md->start_match_ptr;
1656              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1657            }            }
1658          else if (rrc != MATCH_NOMATCH &&  
1659                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))          /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1660            as NOMATCH. */
1661    
1662            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1663            {            {
1664            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1665            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
# Line 1391  for (;;) Line 1668  for (;;)
1668            }            }
1669    
1670          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1671          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1672          }          }
1673        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1403  for (;;) Line 1678  for (;;)
1678          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1679        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1680        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer, but resetting  
     the start-of-match value in case it was changed by \K. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */  
         {  
         mstart = md->start_match_ptr;  
         break;  
         }  
       if (rrc != MATCH_NOMATCH &&  
           (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
         RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
1681    
1682      /* If hit the end of the group (which could be repeated), fail */      RECURSION_MATCHED:
1683        break;
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1684    
1685      /* An alternation is the end of a branch; scan along to find the end of the      /* An alternation is the end of a branch; scan along to find the end of the
1686      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1500  for (;;) Line 1696  for (;;)
1696      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1697    
1698      case OP_BRAZERO:      case OP_BRAZERO:
1699        {      next = ecode + 1;
1700        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1701        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1702        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1703        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1704      break;      break;
1705    
1706      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1707        {      next = ecode + 1;
1708        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1709        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1710        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1711        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1712      break;      break;
1713    
1714      case OP_SKIPZERO:      case OP_SKIPZERO:
1715        {      next = ecode+1;
1716        next = ecode+1;      do next += GET(next,1); while (*next == OP_ALT);
1717        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1718      break;      break;
1719    
1720        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1721        here; just jump to the group, with allow_zero set TRUE. */
1722    
1723        case OP_BRAPOSZERO:
1724        op = *(++ecode);
1725        allow_zero = TRUE;
1726        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1727          goto POSSESSIVE_NON_CAPTURE;
1728    
1729      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1730    
1731      case OP_KET:      case OP_KET:
1732      case OP_KETRMIN:      case OP_KETRMIN:
1733      case OP_KETRMAX:      case OP_KETRMAX:
1734        case OP_KETRPOS:
1735      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1736    
1737      /* If this was a group that remembered the subject start, in order to break      /* If this was a group that remembered the subject start, in order to break
1738      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1739      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1740    
1741      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1742        {        {
1743        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1744        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1745        }        }
1746      else saved_eptr = NULL;      else saved_eptr = NULL;
1747    
1748      /* If we are at the end of an assertion group or an atomic group, stop      /* If we are at the end of an assertion group, stop matching and return
1749      matching and return MATCH_MATCH, but record the current high water mark for      MATCH_MATCH, but record the current high water mark for use by positive
1750      use by positive assertions. We also need to record the match start in case      assertions. We also need to record the match start in case it was changed
1751      it was changed by \K. */      by \K. */
1752    
1753      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1754          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1755        {        {
1756        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1757        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1758        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1759        MRRETURN(MATCH_MATCH);        MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1760        }        }
1761    
1762      /* For capturing groups we have to check the group number back at the start      /* For capturing groups we have to check the group number back at the start
1763      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1764      bumping the high water mark. Note that whole-pattern recursion is coded as      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1765      a recurse into group 0, so it won't be picked up here. Instead, we catch it      into group 0, so it won't be picked up here. Instead, we catch it when the
1766      when the OP_END is reached. Other recursion is handled here. */      OP_END is reached. Other recursion is handled here. We just have to record
1767        the current subject position and start match pointer and give a MATCH
1768        return. */
1769    
1770      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1771            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1772        {        {
1773        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1774        offset = number << 1;        offset = number << 1;
# Line 1576  for (;;) Line 1778  for (;;)
1778        printf("\n");        printf("\n");
1779  #endif  #endif
1780    
1781          /* Handle a recursively called group. */
1782    
1783          if (md->recursive != NULL && md->recursive->group_num == number)
1784            {
1785            md->end_match_ptr = eptr;
1786            md->start_match_ptr = mstart;
1787            RRETURN(MATCH_MATCH);
1788            }
1789    
1790          /* Deal with capturing */
1791    
1792        md->capture_last = number;        md->capture_last = number;
1793        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1794          {          {
1795            /* If offset is greater than offset_top, it means that we are
1796            "skipping" a capturing group, and that group's offsets must be marked
1797            unset. In earlier versions of PCRE, all the offsets were unset at the
1798            start of matching, but this doesn't work because atomic groups and
1799            assertions can cause a value to be set that should later be unset.
1800            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1801            part of the atomic group, but this is not on the final matching path,
1802            so must be unset when 2 is set. (If there is no group 2, there is no
1803            problem, because offset_top will then be 2, indicating no capture.) */
1804    
1805            if (offset > offset_top)
1806              {
1807              register int *iptr = md->offset_vector + offset_top;
1808              register int *iend = md->offset_vector + offset;
1809              while (iptr < iend) *iptr++ = -1;
1810              }
1811    
1812            /* Now make the extraction */
1813    
1814          md->offset_vector[offset] =          md->offset_vector[offset] =
1815            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1816          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1817          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1818          }          }
1819          }
1820    
1821        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1822        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1823        group. This is the forcible breaking of infinite loops as implemented in
1824        Perl 5.005. For a non-repeating atomic group, establish a backup point by
1825        processing the rest of the pattern at a lower level. If this results in a
1826        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1827        bypassing intermediate backup points, but resetting any captures that
1828        happened along the way. */
1829    
1830        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1831          {
1832          if (*prev == OP_ONCE)
1833          {          {
1834          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1835          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1836          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1837          memcpy(md->offset_vector, rec->offset_save,          RRETURN(MATCH_ONCE);
           rec->saved_max * sizeof(int));  
         offset_top = rec->save_offset_top;  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
1838          }          }
1839          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1840          break;
1841        }        }
1842    
1843      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1844      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1845        at a time from the outer level, thus saving stack. */
     ims = original_ims;  
     DPRINTF(("ims reset to %02lx\n", ims));  
1846    
1847      /* For a non-repeating ket, just continue at this level. This also      if (*ecode == OP_KETRPOS)
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
1848        {        {
1849        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1850        break;        md->end_offset_top = offset_top;
1851          RRETURN(MATCH_KETRPOS);
1852        }        }
1853    
1854      /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
1855      preceding bracket, in the appropriate order. In the second case, we can use      the preceding bracket, in the appropriate order. In the second case, we can
1856      tail recursion to avoid using another stack frame, unless we have an      use tail recursion to avoid using another stack frame, unless we have an
1857      unlimited repeat of a group that can match an empty string. */      an atomic group or an unlimited repeat of a group that can match an empty
1858        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
1859    
1860      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1861        {        {
1862        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1863        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1864        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
1865          {          {
1866          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1867            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1868            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1869            RRETURN(MATCH_ONCE);
1870            }
1871          if (*prev >= OP_SBRA)    /* Could match an empty string */
1872            {
1873            md->match_function_type = MATCH_CBEGROUP;
1874            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1875          RRETURN(rrc);          RRETURN(rrc);
1876          }          }
1877        ecode = prev;        ecode = prev;
# Line 1641  for (;;) Line 1879  for (;;)
1879        }        }
1880      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1881        {        {
1882        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1883          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1884          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1885        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1886          if (*prev == OP_ONCE)
1887            {
1888            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1889            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1890            md->once_target = prev;
1891            RRETURN(MATCH_ONCE);
1892            }
1893        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1894        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1895        }        }
1896      /* Control never gets here */      /* Control never gets here */
1897    
1898      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1899    
1900      case OP_CIRC:      case OP_CIRC:
1901      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))  
         MRRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
1902    
1903      /* Start of subject assertion */      /* Start of subject assertion */
1904    
# Line 1670  for (;;) Line 1907  for (;;)
1907      ecode++;      ecode++;
1908      break;      break;
1909    
1910        /* Multiline mode: start of subject unless notbol, or after any newline. */
1911    
1912        case OP_CIRCM:
1913        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1914        if (eptr != md->start_subject &&
1915            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1916          MRRETURN(MATCH_NOMATCH);
1917        ecode++;
1918        break;
1919    
1920      /* Start of match assertion */      /* Start of match assertion */
1921    
1922      case OP_SOM:      case OP_SOM:
# Line 1684  for (;;) Line 1931  for (;;)
1931      ecode++;      ecode++;
1932      break;      break;
1933    
1934      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
1935      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
1936    
1937      case OP_DOLL:      case OP_DOLLM:
1938      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
1939        {        { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
       if (eptr < md->end_subject)  
         { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }  
       else  
         { if (md->noteol) MRRETURN(MATCH_NOMATCH); }  
       ecode++;  
       break;  
       }  
1940      else      else
1941        {        {
1942        if (md->noteol) MRRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1943        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           MRRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
1944        }        }
1945        ecode++;
1946        break;
1947    
1948        /* Not multiline mode: assert before a terminating newline or before end of
1949        subject unless noteol is set. */
1950    
1951        case OP_DOLL:
1952        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1953        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1954    
1955      /* ... else fall through for endonly */      /* ... else fall through for endonly */
1956    
1957      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
1958    
1959      case OP_EOD:      case OP_EOD:
1960      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
1961        SCHECK_PARTIAL();
1962      ecode++;      ecode++;
1963      break;      break;
1964    
1965      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
1966    
1967      case OP_EODN:      case OP_EODN:
1968      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
1969        if (eptr < md->end_subject &&
1970          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1971        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1972    
1973        /* Either at end of string or \n before end. */
1974    
1975        SCHECK_PARTIAL();
1976      ecode++;      ecode++;
1977      break;      break;
1978    
# Line 1852  for (;;) Line 2101  for (;;)
2101      /* Fall through */      /* Fall through */
2102    
2103      case OP_ALLANY:      case OP_ALLANY:
2104      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2105        {        {                            /* not be updated before SCHECK_PARTIAL. */
2106        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2107        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2108        }        }
2109        eptr++;
2110      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2111      ecode++;      ecode++;
2112      break;      break;
# Line 1865  for (;;) Line 2115  for (;;)
2115      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2116    
2117      case OP_ANYBYTE:      case OP_ANYBYTE:
2118      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2119        {        {                            /* not be updated before SCHECK_PARTIAL. */
2120        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2121        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2122        }        }
2123        eptr++;
2124      ecode++;      ecode++;
2125      break;      break;
2126    
# Line 1985  for (;;) Line 2236  for (;;)
2236      switch(c)      switch(c)
2237        {        {
2238        default: MRRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2239    
2240        case 0x000d:        case 0x000d:
2241        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2242        break;        break;
# Line 2208  for (;;) Line 2460  for (;;)
2460        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2461        }        }
2462      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2463        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2464        while (eptr < md->end_subject)
2465        {        {
2466        int category = UCD_CATEGORY(c);        int len = 1;
2467        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2468        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2469          {        eptr += len;
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2470        }        }
2471      ecode++;      ecode++;
2472      break;      break;
# Line 2237  for (;;) Line 2482  for (;;)
2482      loops). */      loops). */
2483    
2484      case OP_REF:      case OP_REF:
2485        {      case OP_REFI:
2486        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2487        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2488        ecode += 3;
2489    
2490        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2491    
2492        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2493        than the amount of subject left; this ensures that every attempt at a      this ensures that every attempt at a match fails. We can't just fail
2494        match fails. We can't just fail here, because of the possibility of      here, because of the possibility of quantifiers with zero minima.
       quantifiers with zero minima.  
2495    
2496        (b) If the JavaScript compatibility flag is set, set the length to zero      (b) If the JavaScript compatibility flag is set, set the length to zero
2497        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2498    
2499        Otherwise, set the length to the length of what was matched by the      Otherwise, set the length to the length of what was matched by the
2500        referenced subpattern. */      referenced subpattern. */
2501    
2502        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2503          length = (md->jscript_compat)? 0 : (int)(md->end_subject - eptr + 1);        length = (md->jscript_compat)? 0 : -1;
2504        else      else
2505          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2506    
2507        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2508    
2509        switch (*ecode)      switch (*ecode)
2510          {        {
2511          case OP_CRSTAR:        case OP_CRSTAR:
2512          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2513          case OP_CRPLUS:        case OP_CRPLUS:
2514          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2515          case OP_CRQUERY:        case OP_CRQUERY:
2516          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2517          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2518          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2519          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2520          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2521          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2522          break;        break;
2523    
2524          case OP_CRRANGE:        case OP_CRRANGE:
2525          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2526          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2527          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2528          max = GET2(ecode, 3);        max = GET2(ecode, 3);
2529          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2530          ecode += 5;        ecode += 5;
2531          break;        break;
2532    
2533          default:               /* No repeat follows */        default:               /* No repeat follows */
2534          if (!match_ref(offset, eptr, length, md, ims))        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2535            {          {
2536            CHECK_PARTIAL();          CHECK_PARTIAL();
2537            MRRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
           }  
         eptr += length;  
         continue;              /* With the main loop */  
2538          }          }
2539          eptr += length;
2540          continue;              /* With the main loop */
2541          }
2542    
2543        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2544        main loop. */      zero, just continue with the main loop. */
2545    
2546        if (length == 0) continue;      if (length == 0) continue;
2547    
2548        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2549        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2550        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2551    
2552        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2553          {
2554          int slength;
2555          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2556          {          {
2557          if (!match_ref(offset, eptr, length, md, ims))          CHECK_PARTIAL();
2558            {          MRRETURN(MATCH_NOMATCH);
           CHECK_PARTIAL();  
           MRRETURN(MATCH_NOMATCH);  
           }  
         eptr += length;  
2559          }          }
2560          eptr += slength;
2561          }
2562    
2563        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2564        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2565    
2566        if (min == max) continue;      if (min == max) continue;
2567    
2568        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2569    
2570        if (minimize)      if (minimize)
2571          {
2572          for (fi = min;; fi++)
2573          {          {
2574          for (fi = min;; fi++)          int slength;
2575            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2576            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2577            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2578            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2579            {            {
2580            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            CHECK_PARTIAL();
2581            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max) MRRETURN(MATCH_NOMATCH);  
           if (!match_ref(offset, eptr, length, md, ims))  
             {  
             CHECK_PARTIAL();  
             MRRETURN(MATCH_NOMATCH);  
             }  
           eptr += length;  
2582            }            }
2583          /* Control never gets here */          eptr += slength;
2584          }          }
2585          /* Control never gets here */
2586          }
2587    
2588        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2589    
2590        else      else
2591          {
2592          pp = eptr;
2593          for (i = min; i < max; i++)
2594          {          {
2595          pp = eptr;          int slength;
2596          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2597            {            {
2598            if (!match_ref(offset, eptr, length, md, ims))            CHECK_PARTIAL();
2599              {            break;
             CHECK_PARTIAL();  
             break;  
             }  
           eptr += length;  
           }  
         while (eptr >= pp)  
           {  
           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);  
           if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
           eptr -= length;  
2600            }            }
2601          MRRETURN(MATCH_NOMATCH);          eptr += slength;
2602            }
2603          while (eptr >= pp)
2604            {
2605            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2606            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2607            eptr -= length;
2608          }          }
2609          MRRETURN(MATCH_NOMATCH);
2610        }        }
2611      /* Control never gets here */      /* Control never gets here */
2612    
# Line 2465  for (;;) Line 2712  for (;;)
2712            {            {
2713            for (fi = min;; fi++)            for (fi = min;; fi++)
2714              {              {
2715              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2716              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2717              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2718              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2490  for (;;) Line 2737  for (;;)
2737            {            {
2738            for (fi = min;; fi++)            for (fi = min;; fi++)
2739              {              {
2740              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2741              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2742              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2743              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2536  for (;;) Line 2783  for (;;)
2783              }              }
2784            for (;;)            for (;;)
2785              {              {
2786              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2787              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2788              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2789              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2559  for (;;) Line 2806  for (;;)
2806              }              }
2807            while (eptr >= pp)            while (eptr >= pp)
2808              {              {
2809              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2810              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2811              eptr--;              eptr--;
2812              }              }
# Line 2635  for (;;) Line 2882  for (;;)
2882          {          {
2883          for (fi = min;; fi++)          for (fi = min;; fi++)
2884            {            {
2885            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2886            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2887            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2888            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 2668  for (;;) Line 2915  for (;;)
2915            }            }
2916          for(;;)          for(;;)
2917            {            {
2918            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2919            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2920            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2921            if (utf8) BACKCHAR(eptr);            if (utf8) BACKCHAR(eptr);
# Line 2713  for (;;) Line 2960  for (;;)
2960    
2961      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2962    
2963      case OP_CHARNC:      case OP_CHARI:
2964  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2965      if (utf8)      if (utf8)
2966        {        {
# Line 2773  for (;;) Line 3020  for (;;)
3020      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3021    
3022      case OP_EXACT:      case OP_EXACT:
3023        case OP_EXACTI:
3024      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3025      ecode += 3;      ecode += 3;
3026      goto REPEATCHAR;      goto REPEATCHAR;
3027    
3028      case OP_POSUPTO:      case OP_POSUPTO:
3029        case OP_POSUPTOI:
3030      possessive = TRUE;      possessive = TRUE;
3031      /* Fall through */      /* Fall through */
3032    
3033      case OP_UPTO:      case OP_UPTO:
3034        case OP_UPTOI:
3035      case OP_MINUPTO:      case OP_MINUPTO:
3036        case OP_MINUPTOI:
3037      min = 0;      min = 0;
3038      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3039      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3040      ecode += 3;      ecode += 3;
3041      goto REPEATCHAR;      goto REPEATCHAR;
3042    
3043      case OP_POSSTAR:      case OP_POSSTAR:
3044        case OP_POSSTARI:
3045      possessive = TRUE;      possessive = TRUE;
3046      min = 0;      min = 0;
3047      max = INT_MAX;      max = INT_MAX;
# Line 2797  for (;;) Line 3049  for (;;)
3049      goto REPEATCHAR;      goto REPEATCHAR;
3050    
3051      case OP_POSPLUS:      case OP_POSPLUS:
3052        case OP_POSPLUSI:
3053      possessive = TRUE;      possessive = TRUE;
3054      min = 1;      min = 1;
3055      max = INT_MAX;      max = INT_MAX;
# Line 2804  for (;;) Line 3057  for (;;)
3057      goto REPEATCHAR;      goto REPEATCHAR;
3058    
3059      case OP_POSQUERY:      case OP_POSQUERY:
3060        case OP_POSQUERYI:
3061      possessive = TRUE;      possessive = TRUE;
3062      min = 0;      min = 0;
3063      max = 1;      max = 1;
# Line 2811  for (;;) Line 3065  for (;;)
3065      goto REPEATCHAR;      goto REPEATCHAR;
3066    
3067      case OP_STAR:      case OP_STAR:
3068        case OP_STARI:
3069      case OP_MINSTAR:      case OP_MINSTAR:
3070        case OP_MINSTARI:
3071      case OP_PLUS:      case OP_PLUS:
3072        case OP_PLUSI:
3073      case OP_MINPLUS:      case OP_MINPLUS:
3074        case OP_MINPLUSI:
3075      case OP_QUERY:      case OP_QUERY:
3076        case OP_QUERYI:
3077      case OP_MINQUERY:      case OP_MINQUERY:
3078      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3079        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3080      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
   
3081      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3082      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3083      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
# Line 2841  for (;;) Line 3100  for (;;)
3100          {          {
3101  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3102          unsigned int othercase;          unsigned int othercase;
3103          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3104              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3105            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3106          else oclength = 0;          else oclength = 0;
# Line 2869  for (;;) Line 3128  for (;;)
3128            {            {
3129            for (fi = min;; fi++)            for (fi = min;; fi++)
3130              {              {
3131              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3132              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3133              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3134              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
# Line 2911  for (;;) Line 3170  for (;;)
3170    
3171            for(;;)            for(;;)
3172              {              {
3173              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3174              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3175              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3176  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2948  for (;;) Line 3207  for (;;)
3207      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3208        max, eptr));        max, eptr));
3209    
3210      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3211        {        {
3212        fc = md->lcc[fc];        fc = md->lcc[fc];
3213        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
# Line 2965  for (;;) Line 3224  for (;;)
3224          {          {
3225          for (fi = min;; fi++)          for (fi = min;; fi++)
3226            {            {
3227            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3228            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3229            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3230            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 2995  for (;;) Line 3254  for (;;)
3254    
3255          while (eptr >= pp)          while (eptr >= pp)
3256            {            {
3257            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3258            eptr--;            eptr--;
3259            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3260            }            }
# Line 3024  for (;;) Line 3283  for (;;)
3283          {          {
3284          for (fi = min;; fi++)          for (fi = min;; fi++)
3285            {            {
3286            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3287            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3288            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3289            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 3053  for (;;) Line 3312  for (;;)
3312    
3313          while (eptr >= pp)          while (eptr >= pp)
3314            {            {
3315            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3316            eptr--;            eptr--;
3317            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3318            }            }
# Line 3066  for (;;) Line 3325  for (;;)
3325      checking can be multibyte. */      checking can be multibyte. */
3326    
3327      case OP_NOT:      case OP_NOT:
3328        case OP_NOTI:
3329      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3330        {        {
3331        SCHECK_PARTIAL();        SCHECK_PARTIAL();
# Line 3073  for (;;) Line 3333  for (;;)
3333        }        }
3334      ecode++;      ecode++;
3335      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3336      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3337        {        {
3338  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3339        if (c < 256)        if (c < 256)
# Line 3081  for (;;) Line 3341  for (;;)
3341        c = md->lcc[c];        c = md->lcc[c];
3342        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3343        }        }
3344      else      else    /* Caseful */
3345        {        {
3346        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3347        }        }
# Line 3095  for (;;) Line 3355  for (;;)
3355      about... */      about... */
3356    
3357      case OP_NOTEXACT:      case OP_NOTEXACT:
3358        case OP_NOTEXACTI:
3359      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3360      ecode += 3;      ecode += 3;
3361      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3362    
3363      case OP_NOTUPTO:      case OP_NOTUPTO:
3364        case OP_NOTUPTOI:
3365      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3366        case OP_NOTMINUPTOI:
3367      min = 0;      min = 0;
3368      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3369      minimize = *ecode == OP_NOTMINUPTO;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3370      ecode += 3;      ecode += 3;
3371      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3372    
3373      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
3374        case OP_NOTPOSSTARI:
3375      possessive = TRUE;      possessive = TRUE;
3376      min = 0;      min = 0;
3377      max = INT_MAX;      max = INT_MAX;
# Line 3115  for (;;) Line 3379  for (;;)
3379      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3380    
3381      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
3382        case OP_NOTPOSPLUSI:
3383      possessive = TRUE;      possessive = TRUE;
3384      min = 1;      min = 1;
3385      max = INT_MAX;      max = INT_MAX;
# Line 3122  for (;;) Line 3387  for (;;)
3387      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3388    
3389      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
3390        case OP_NOTPOSQUERYI:
3391      possessive = TRUE;      possessive = TRUE;
3392      min = 0;      min = 0;
3393      max = 1;      max = 1;
# Line 3129  for (;;) Line 3395  for (;;)
3395      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3396    
3397      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
3398        case OP_NOTPOSUPTOI:
3399      possessive = TRUE;      possessive = TRUE;
3400      min = 0;      min = 0;
3401      max = GET2(ecode, 1);      max = GET2(ecode, 1);
# Line 3136  for (;;) Line 3403  for (;;)
3403      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3404    
3405      case OP_NOTSTAR:      case OP_NOTSTAR:
3406        case OP_NOTSTARI:
3407      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3408        case OP_NOTMINSTARI:
3409      case OP_NOTPLUS:      case OP_NOTPLUS:
3410        case OP_NOTPLUSI:
3411      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3412        case OP_NOTMINPLUSI:
3413      case OP_NOTQUERY:      case OP_NOTQUERY:
3414        case OP_NOTQUERYI:
3415      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3416      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3417        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3418      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3419      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3420      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
# Line 3163  for (;;) Line 3436  for (;;)
3436      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3437        max, eptr));        max, eptr));
3438    
3439      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3440        {        {
3441        fc = md->lcc[fc];        fc = md->lcc[fc];
3442    
# Line 3211  for (;;) Line 3484  for (;;)
3484            register unsigned int d;            register unsigned int d;
3485            for (fi = min;; fi++)            for (fi = min;; fi++)
3486              {              {
3487              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM28);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3488              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3489              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3490              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3230  for (;;) Line 3503  for (;;)
3503            {            {
3504            for (fi = min;; fi++)            for (fi = min;; fi++)
3505              {              {
3506              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM29);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3507              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3508              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3509              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3271  for (;;) Line 3544  for (;;)
3544          if (possessive) continue;          if (possessive) continue;
3545          for(;;)          for(;;)
3546              {              {
3547              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM30);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3548              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3549              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3550              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 3294  for (;;) Line 3567  for (;;)
3567            if (possessive) continue;            if (possessive) continue;
3568            while (eptr >= pp)            while (eptr >= pp)
3569              {              {
3570              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM31);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3571              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3572              eptr--;              eptr--;
3573              }              }
# Line 3351  for (;;) Line 3624  for (;;)
3624            register unsigned int d;            register unsigned int d;
3625            for (fi = min;; fi++)            for (fi = min;; fi++)
3626              {              {
3627              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM32);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3628              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3629              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3630              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3369  for (;;) Line 3642  for (;;)
3642            {            {
3643            for (fi = min;; fi++)            for (fi = min;; fi++)
3644              {              {
3645              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM33);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
3646              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3647              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3648              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3409  for (;;) Line 3682  for (;;)
3682            if (possessive) continue;            if (possessive) continue;
3683            for(;;)            for(;;)
3684              {              {
3685              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM34);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
3686              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3687              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3688              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 3432  for (;;) Line 3705  for (;;)
3705            if (possessive) continue;            if (possessive) continue;
3706            while (eptr >= pp)            while (eptr >= pp)
3707              {              {
3708              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM35);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
3709              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3710              eptr--;              eptr--;
3711              }              }
# Line 3547  for (;;) Line 3820  for (;;)
3820            case PT_LAMP:            case PT_LAMP:
3821            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3822              {              {
3823                int chartype;
3824              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3825                {                {
3826                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3827                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3828                }                }
3829              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3830              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
3831              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
3832                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
3833                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
3834                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3835              }              }
3836            break;            break;
# Line 3570  for (;;) Line 3844  for (;;)
3844                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3845                }                }
3846              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3847              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
3848                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3849              }              }
3850            break;            break;
# Line 3585  for (;;) Line 3858  for (;;)
3858                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3859                }                }
3860              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3861              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
3862                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3863              }              }
3864            break;            break;
# Line 3600  for (;;) Line 3872  for (;;)
3872                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3873                }                }
3874              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3875              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
3876                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3877              }              }
3878            break;            break;
# Line 3609  for (;;) Line 3880  for (;;)
3880            case PT_ALNUM:            case PT_ALNUM:
3881            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3882              {              {
3883                int category;
3884              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3885                {                {
3886                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3887                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3888                }                }
3889              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3890              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3891              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
3892                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3893              }              }
3894            break;            break;
# Line 3631  for (;;) Line 3902  for (;;)
3902                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3903                }                }
3904              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3905              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
3906                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
3907                     == prop_fail_result)                     == prop_fail_result)
3908                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3648  for (;;) Line 3918  for (;;)
3918                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3919                }                }
3920              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3921              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
3922                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
3923                     == prop_fail_result)                     == prop_fail_result)
3924                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3659  for (;;) Line 3928  for (;;)
3928            case PT_WORD:            case PT_WORD:
3929            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3930              {              {
3931                int category;
3932              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3933                {                {
3934                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3935                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3936                }                }
3937              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3938              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3939              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
                  c == CHAR_UNDERSCORE)  
3940                     == prop_fail_result)                     == prop_fail_result)
3941                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3942              }              }
# Line 3693  for (;;) Line 3962  for (;;)
3962              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3963              }              }
3964            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3965            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
3966            while (eptr < md->end_subject)            while (eptr < md->end_subject)
3967              {              {
3968              int len = 1;              int len = 1;
3969              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
3970                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
3971              eptr += len;              eptr += len;
3972              }              }
3973            }            }
# Line 3759  for (;;) Line 4025  for (;;)
4025            switch(c)            switch(c)
4026              {              {
4027              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4028    
4029              case 0x000d:              case 0x000d:
4030              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4031              break;              break;
# Line 4035  for (;;) Line 4302  for (;;)
4302            switch(*eptr++)            switch(*eptr++)
4303              {              {
4304              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4305    
4306              case 0x000d:              case 0x000d:
4307              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4308              break;              break;
4309    
4310              case 0x000a:              case 0x000a:
4311              break;              break;
4312    
# Line 4227  for (;;) Line 4496  for (;;)
4496            case PT_ANY:            case PT_ANY:
4497            for (fi = min;; fi++)            for (fi = min;; fi++)
4498              {              {
4499              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM36);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4500              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4501              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4502              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4243  for (;;) Line 4512  for (;;)
4512            case PT_LAMP:            case PT_LAMP:
4513            for (fi = min;; fi++)            for (fi = min;; fi++)
4514              {              {
4515              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM37);              int chartype;
4516                RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4517              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4518              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4519              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4252  for (;;) Line 4522  for (;;)
4522                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4523                }                }
4524              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4525              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4526              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4527                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4528                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4529                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4530              }              }
4531            /* Control never gets here */            /* Control never gets here */
# Line 4263  for (;;) Line 4533  for (;;)
4533            case PT_GC:            case PT_GC:
4534            for (fi = min;; fi++)            for (fi = min;; fi++)
4535              {              {
4536              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM38);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4537              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4538              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4539              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4272  for (;;) Line 4542  for (;;)
4542                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4543                }                }
4544              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4545              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
4546                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4547              }              }
4548            /* Control never gets here */            /* Control never gets here */
# Line 4281  for (;;) Line 4550  for (;;)
4550            case PT_PC:            case PT_PC:
4551            for (fi = min;; fi++)            for (fi = min;; fi++)
4552              {              {
4553              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM39);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4554              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4555              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4556              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4290  for (;;) Line 4559  for (;;)
4559                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4560                }                }
4561              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4562              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
4563                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4564              }              }
4565            /* Control never gets here */            /* Control never gets here */
# Line 4299  for (;;) Line 4567  for (;;)
4567            case PT_SC:            case PT_SC:
4568            for (fi = min;; fi++)            for (fi = min;; fi++)
4569              {              {
4570              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM40);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
4571              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4572              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4573              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4308  for (;;) Line 4576  for (;;)
4576                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4577                }                }
4578              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4579              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
4580                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4581              }              }
4582            /* Control never gets here */            /* Control never gets here */
# Line 4317  for (;;) Line 4584  for (;;)
4584            case PT_ALNUM:            case PT_ALNUM:
4585            for (fi = min;; fi++)            for (fi = min;; fi++)
4586              {              {
4587              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM59);              int category;
4588                RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4589              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4590              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4591              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4326  for (;;) Line 4594  for (;;)
4594                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4595                }                }
4596              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4597              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4598              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
4599                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4600              }              }
4601            /* Control never gets here */            /* Control never gets here */
# Line 4336  for (;;) Line 4603  for (;;)
4603            case PT_SPACE:    /* Perl space */            case PT_SPACE:    /* Perl space */
4604            for (fi = min;; fi++)            for (fi = min;; fi++)
4605              {              {
4606              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM60);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM60);
4607              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4608              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4609              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4345  for (;;) Line 4612  for (;;)
4612                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4613                }                }
4614              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4615              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
4616                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
4617                     == prop_fail_result)                     == prop_fail_result)
4618                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4356  for (;;) Line 4622  for (;;)
4622            case PT_PXSPACE:  /* POSIX space */            case PT_PXSPACE:  /* POSIX space */
4623            for (fi = min;; fi++)            for (fi = min;; fi++)
4624              {              {
4625              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM61);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
4626              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4627              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4628              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4365  for (;;) Line 4631  for (;;)
4631                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4632                }                }
4633              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4634              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
4635                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4636                     == prop_fail_result)                     == prop_fail_result)
4637                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4376  for (;;) Line 4641  for (;;)
4641            case PT_WORD:            case PT_WORD:
4642            for (fi = min;; fi++)            for (fi = min;; fi++)
4643              {              {
4644              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM62);              int category;
4645                RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4646              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4647              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4648              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4385  for (;;) Line 4651  for (;;)
4651                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4652                }                }
4653              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4654              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4655              if ((prop_category == ucp_L ||              if ((category == ucp_L ||
4656                   prop_category == ucp_N ||                   category == ucp_N ||
4657                   c == CHAR_UNDERSCORE)                   c == CHAR_UNDERSCORE)
4658                     == prop_fail_result)                     == prop_fail_result)
4659                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4408  for (;;) Line 4674  for (;;)
4674          {          {
4675          for (fi = min;; fi++)          for (fi = min;; fi++)
4676            {            {
4677            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM41);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
4678            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4679            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4680            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4417  for (;;) Line 4683  for (;;)
4683              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
4684              }              }
4685            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
4686            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
4687            while (eptr < md->end_subject)            while (eptr < md->end_subject)
4688              {              {
4689              int len = 1;              int len = 1;
4690              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4691                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
4692              eptr += len;              eptr += len;
4693              }              }
4694            }            }
4695          }          }
   
4696        else        else
4697  #endif     /* SUPPORT_UCP */  #endif     /* SUPPORT_UCP */
4698    
# Line 4440  for (;;) Line 4702  for (;;)
4702          {          {
4703          for (fi = min;; fi++)          for (fi = min;; fi++)
4704            {            {
4705            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM42);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
4706            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4707            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4708            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4603  for (;;) Line 4865  for (;;)
4865          {          {
4866          for (fi = min;; fi++)          for (fi = min;; fi++)
4867            {            {
4868            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM43);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
4869            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4870            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4871            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4751  for (;;) Line 5013  for (;;)
5013            case PT_LAMP:            case PT_LAMP:
5014            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5015              {              {
5016                int chartype;
5017              int len = 1;              int len = 1;
5018              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5019                {                {
# Line 4758  for (;;) Line 5021  for (;;)
5021                break;                break;
5022                }                }
5023              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5024              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
5025              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
5026                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
5027                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
5028                break;                break;
5029              eptr+= len;              eptr+= len;
5030              }              }
# Line 4777  for (;;) Line 5040  for (;;)
5040                break;                break;
5041                }                }
5042              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5043              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
             if ((prop_category == prop_value) == prop_fail_result)  
               break;  
5044              eptr+= len;              eptr+= len;
5045              }              }
5046            break;            break;
# Line 4794  for (;;) Line 5055  for (;;)
5055                break;                break;
5056                }                }
5057              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5058              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
             if ((prop_chartype == prop_value) == prop_fail_result)  
               break;  
5059              eptr+= len;              eptr+= len;
5060              }              }
5061            break;            break;
# Line 4811  for (;;) Line 5070  for (;;)
5070                break;                break;
5071                }                }
5072              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5073              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
             if ((prop_script == prop_value) == prop_fail_result)  
               break;  
5074              eptr+= len;              eptr+= len;
5075              }              }
5076            break;            break;
# Line 4821  for (;;) Line 5078  for (;;)
5078            case PT_ALNUM:            case PT_ALNUM:
5079            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5080              {              {
5081                int category;
5082              int len = 1;              int len = 1;
5083              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5084                {                {
# Line 4828  for (;;) Line 5086  for (;;)
5086                break;                break;
5087                }                }
5088              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5089              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5090              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                  == prop_fail_result)  
5091                break;                break;
5092              eptr+= len;              eptr+= len;
5093              }              }
# Line 4846  for (;;) Line 5103  for (;;)
5103                break;                break;
5104                }                }
5105              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5106              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
5107                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
5108                   == prop_fail_result)                   == prop_fail_result)
5109                break;                break;
# Line 4865  for (;;) Line 5121  for (;;)
5121                break;                break;
5122                }                }
5123              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5124              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
5125                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5126                   == prop_fail_result)                   == prop_fail_result)
5127                break;                break;
# Line 4877  for (;;) Line 5132  for (;;)
5132            case PT_WORD:            case PT_WORD:
5133            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5134              {              {
5135                int category;
5136              int len = 1;              int len = 1;
5137              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5138                {                {
# Line 4884  for (;;) Line 5140  for (;;)
5140                break;                break;
5141                }                }
5142              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5143              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5144              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N ||
5145                   c == CHAR_UNDERSCORE) == prop_fail_result)                   c == CHAR_UNDERSCORE) == prop_fail_result)
5146                break;                break;
5147              eptr+= len;              eptr+= len;
# Line 4901  for (;;) Line 5157  for (;;)
5157          if (possessive) continue;          if (possessive) continue;
5158          for(;;)          for(;;)
5159            {            {
5160            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM44);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5161            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5162            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5163            if (utf8) BACKCHAR(eptr);            if (utf8) BACKCHAR(eptr);
# Line 4915  for (;;) Line 5171  for (;;)
5171          {          {
5172          for (i = min; i < max; i++)          for (i = min; i < max; i++)
5173            {            {
5174              int len = 1;
5175            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
5176              {              {
5177              SCHECK_PARTIAL();              SCHECK_PARTIAL();
5178              break;              break;
5179              }              }
5180            GETCHARINCTEST(c, eptr);            if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5181            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) break;
5182            if (prop_category == ucp_M) break;            eptr += len;
5183            while (eptr < md->end_subject)            while (eptr < md->end_subject)
5184              {              {
5185              int len = 1;              len = 1;
5186              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5187                {              if (UCD_CATEGORY(c) != ucp_M) break;
               GETCHARLEN(c, eptr, len);  
               }  
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
5188              eptr += len;              eptr += len;
5189              }              }
5190            }            }
# Line 4942  for (;;) Line 5195  for (;;)
5195    
5196          for(;;)          for(;;)
5197            {            {
5198            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM45);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5199            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5200            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5201            for (;;)                        /* Move back over one extended */            for (;;)                        /* Move back over one extended */
5202              {              {
             int len = 1;  
5203              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else
5204                {                {
5205                BACKCHAR(eptr);                BACKCHAR(eptr);
5206                GETCHARLEN(c, eptr, len);                GETCHAR(c, eptr);
5207                }                }
5208              prop_category = UCD_CATEGORY(c);              if (UCD_CATEGORY(c) != ucp_M) break;
             if (prop_category != ucp_M) break;  
5209              eptr--;              eptr--;
5210              }              }
5211            }            }
# Line 5018  for (;;) Line 5269  for (;;)
5269                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
5270                }                }
5271              }              }
5272            else eptr = md->end_subject;   /* Unlimited UTF-8 repeat */            else
5273                {
5274                eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
5275                SCHECK_PARTIAL();
5276                }
5277            break;            break;
5278    
5279            /* The byte case is the same as non-UTF8 */            /* The byte case is the same as non-UTF8 */
# Line 5226  for (;;) Line 5481  for (;;)
5481            RRETURN(PCRE_ERROR_INTERNAL);            RRETURN(PCRE_ERROR_INTERNAL);
5482            }            }
5483    
5484          /* eptr is now past the end of the maximum run */          /* eptr is now past the end of the maximum run. If possessive, we are
5485            done (no backing up). Otherwise, match at this position; anything other
5486            than no match is immediately returned. For nomatch, back up one
5487            character, unless we are matching \R and the last thing matched was
5488            \r\n, in which case, back up two bytes. */
5489    
5490          if (possessive) continue;          if (possessive) continue;
5491          for(;;)          for(;;)
5492            {            {
5493            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM46);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
5494            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5495            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5496            BACKCHAR(eptr);            BACKCHAR(eptr);
5497              if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5498                  eptr[-1] == '\r') eptr--;
5499            }            }
5500          }          }
5501        else        else
# Line 5433  for (;;) Line 5694  for (;;)
5694            RRETURN(PCRE_ERROR_INTERNAL);            RRETURN(PCRE_ERROR_INTERNAL);
5695            }            }
5696    
5697          /* eptr is now past the end of the maximum run */          /* eptr is now past the end of the maximum run. If possessive, we are
5698            done (no backing up). Otherwise, match at this position; anything other
5699            than no match is immediately returned. For nomatch, back up one
5700            character (byte), unless we are matching \R and the last thing matched
5701            was \r\n, in which case, back up two bytes. */
5702    
5703          if (possessive) continue;          if (possessive) continue;
5704          while (eptr >= pp)          while (eptr >= pp)
5705            {            {
5706            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM47);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
           eptr--;  
5707            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5708              eptr--;
5709              if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5710                  eptr[-1] == '\r') eptr--;
5711            }            }
5712          }          }
5713    
# Line 5479  switch (frame->Xwhere) Line 5746  switch (frame->Xwhere)
5746    LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)    LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
5747    LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)    LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
5748    LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)    LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
5749    LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58)    LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63)
5750  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5751    LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)    LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)
5752    LBL(32) LBL(34) LBL(42) LBL(46)    LBL(32) LBL(34) LBL(42) LBL(46)
# Line 5508  Undefine all the macros that were define Line 5775  Undefine all the macros that were define
5775  #undef ecode  #undef ecode
5776  #undef mstart  #undef mstart
5777  #undef offset_top  #undef offset_top
 #undef ims  
5778  #undef eptrb  #undef eptrb
5779  #undef flags  #undef flags
5780    
# Line 5526  Undefine all the macros that were define Line 5792  Undefine all the macros that were define
5792  #undef condition  #undef condition
5793  #undef prev_is_word  #undef prev_is_word
5794    
 #undef original_ims  
   
5795  #undef ctype  #undef ctype
5796  #undef length  #undef length
5797  #undef max  #undef max
# Line 5584  pcre_exec(const pcre *argument_re, const Line 5848  pcre_exec(const pcre *argument_re, const
5848    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5849    int offsetcount)    int offsetcount)
5850  {  {
5851  int rc, resetcount, ocount;  int rc, ocount, arg_offset_max;
5852  int first_byte = -1;  int first_byte = -1;
5853  int req_byte = -1;  int req_byte = -1;
5854  int req_byte2 = -1;  int req_byte2 = -1;
5855  int newline;  int newline;
 unsigned long int ims;  
5856  BOOL using_temporary_offsets = FALSE;  BOOL using_temporary_offsets = FALSE;
5857  BOOL anchored;  BOOL anchored;
5858  BOOL startline;  BOOL startline;
# Line 5619  if ((options & ~PUBLIC_EXEC_OPTIONS) != Line 5882  if ((options & ~PUBLIC_EXEC_OPTIONS) !=
5882  if (re == NULL || subject == NULL ||  if (re == NULL || subject == NULL ||
5883     (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL;     (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL;
5884  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
5885    if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
5886    
5887    /* These two settings are used in the code for checking a UTF-8 string that
5888    follows immediately afterwards. Other values in the md block are used only
5889    during "normal" pcre_exec() processing, not when the JIT support is in use,
5890    so they are set up later. */
5891    
5892    utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
5893    md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5894                  ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5895    
5896  /* This information is for finding all the numbers associated with a given  /* Check a UTF-8 string if required. Pass back the character offset and error
5897  name, for condition testing. */  code for an invalid string if a results vector is available. */
5898    
5899    #ifdef SUPPORT_UTF8
5900    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5901      {
5902      int erroroffset;
5903      int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);
5904      if (errorcode != 0)
5905        {
5906        if (offsetcount >= 2)
5907          {
5908          offsets[0] = erroroffset;
5909          offsets[1] = errorcode;
5910          }
5911        return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
5912          PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
5913        }
5914    
5915      /* Check that a start_offset points to the start of a UTF-8 character. */
5916      if (start_offset > 0 && start_offset < length &&
5917          (((USPTR)subject)[start_offset] & 0xc0) == 0x80)
5918        return PCRE_ERROR_BADUTF8_OFFSET;
5919      }
5920    #endif
5921    
5922    /* If the pattern was successfully studied with JIT support, run the JIT
5923    executable instead of the rest of this function. Most options must be set at
5924    compile time for the JIT code to be usable. Fallback to the normal code path if
5925    an unsupported flag is set. In particular, JIT does not support partial
5926    matching. */
5927    
5928    #ifdef SUPPORT_JIT
5929    if (extra_data != NULL
5930        && (extra_data->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0
5931        && extra_data->executable_jit != NULL
5932        && (options & ~(PCRE_NO_UTF8_CHECK | PCRE_NOTBOL | PCRE_NOTEOL |
5933                        PCRE_NOTEMPTY | PCRE_NOTEMPTY_ATSTART)) == 0)
5934      return _pcre_jit_exec(re, extra_data->executable_jit, subject, length,
5935        start_offset, options, ((extra_data->flags & PCRE_EXTRA_MATCH_LIMIT) == 0)
5936        ? MATCH_LIMIT : extra_data->match_limit, offsets, offsetcount);
5937    #endif
5938    
5939    /* Carry on with non-JIT matching. This information is for finding all the
5940    numbers associated with a given name, for condition testing. */
5941    
5942  md->name_table = (uschar *)re + re->name_table_offset;  md->name_table = (uschar *)re + re->name_table_offset;
5943  md->name_count = re->name_count;  md->name_count = re->name_count;
# Line 5688  md->end_subject = md->start_subject + le Line 6004  md->end_subject = md->start_subject + le
6004  end_subject = md->end_subject;  end_subject = md->end_subject;
6005    
6006  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
 utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;  
6007  md->use_ucp = (re->options & PCRE_UCP) != 0;  md->use_ucp = (re->options & PCRE_UCP) != 0;
6008  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6009    
6010    /* Some options are unpacked into BOOL variables in the hope that testing
6011    them will be faster than individual option bits. */
6012    
6013  md->notbol = (options & PCRE_NOTBOL) != 0;  md->notbol = (options & PCRE_NOTBOL) != 0;
6014  md->noteol = (options & PCRE_NOTEOL) != 0;  md->noteol = (options & PCRE_NOTEOL) != 0;
6015  md->notempty = (options & PCRE_NOTEMPTY) != 0;  md->notempty = (options & PCRE_NOTEMPTY) != 0;
6016  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6017  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :  
               ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;  
6018  md->hitend = FALSE;  md->hitend = FALSE;
6019  md->mark = NULL;                        /* In case never set */  md->mark = NULL;                        /* In case never set */
6020    
6021  md->recursive = NULL;                   /* No recursion at top level */  md->recursive = NULL;                   /* No recursion at top level */
6022    md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6023    
6024  md->lcc = tables + lcc_offset;  md->lcc = tables + lcc_offset;
6025  md->ctypes = tables + ctypes_offset;  md->ctypes = tables + ctypes_offset;
# Line 5779  defined (though never set). So there's n Line 6097  defined (though never set). So there's n
6097  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6098    return PCRE_ERROR_BADPARTIAL;    return PCRE_ERROR_BADPARTIAL;
6099    
 /* Check a UTF-8 string if required. Unfortunately there's no way of passing  
 back the character offset. */  
   
 #ifdef SUPPORT_UTF8  
 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)  
   {  
   if (_pcre_valid_utf8((USPTR)subject, length) >= 0)  
     return PCRE_ERROR_BADUTF8;  
   if (start_offset > 0 && start_offset < length)  
     {  
     int tb = ((USPTR)subject)[start_offset];  
     if (tb > 127)  
       {  
       tb &= 0xc0;  
       if (tb != 0 && tb != 0xc0) return PCRE_ERROR_BADUTF8_OFFSET;  
       }  
     }  
   }  
 #endif  
   
 /* The ims options can vary during the matching as a result of the presence  
 of (?ims) items in the pattern. They are kept in a local variable so that  
 restoring at the exit of a group is easy. */  
   
 ims = re->options & (PCRE_CASELESS|PCRE_MULTILINE|PCRE_DOTALL);  
   
6100  /* If the expression has got more back references than the offsets supplied can  /* If the expression has got more back references than the offsets supplied can
6101  hold, we get a temporary chunk of working store to use during the matching.  hold, we get a temporary chunk of working store to use during the matching.
6102  Otherwise, we can use the vector supplied, rounding down its size to a multiple  Otherwise, we can use the vector supplied, rounding down its size to a multiple
6103  of 3. */  of 3. */
6104    
6105  ocount = offsetcount - (offsetcount % 3);  ocount = offsetcount - (offsetcount % 3);
6106    arg_offset_max = (2*ocount)/3;
6107    
6108  if (re->top_backref > 0 && re->top_backref >= ocount/3)  if (re->top_backref > 0 && re->top_backref >= ocount/3)
6109    {    {
# Line 5827  md->offset_max = (2*ocount)/3; Line 6120  md->offset_max = (2*ocount)/3;
6120  md->offset_overflow = FALSE;  md->offset_overflow = FALSE;
6121  md->capture_last = -1;  md->capture_last = -1;
6122    
 /* Compute the minimum number of offsets that we need to reset each time. Doing  
 this makes a huge difference to execution time when there aren't many brackets  
 in the pattern. */  
   
 resetcount = 2 + re->top_bracket * 2;  
 if (resetcount > offsetcount) resetcount = ocount;  
   
6123  /* Reset the working variable associated with each extraction. These should  /* Reset the working variable associated with each extraction. These should
6124  never be used unless previously set, but they get saved and restored, and so we  never be used unless previously set, but they get saved and restored, and so we
6125  initialize them to avoid reading uninitialized locations. */  initialize them to avoid reading uninitialized locations. Also, unset the
6126    offsets for the matched string. This is really just for tidiness with callouts,
6127    in case they inspect these fields. */
6128    
6129  if (md->offset_vector != NULL)  if (md->offset_vector != NULL)
6130    {    {
6131    register int *iptr = md->offset_vector + ocount;    register int *iptr = md->offset_vector + ocount;
6132    register int *iend = iptr - resetcount/2 + 1;    register int *iend = iptr - re->top_bracket;
6133      if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6134    while (--iptr >= iend) *iptr = -1;    while (--iptr >= iend) *iptr = -1;
6135      md->offset_vector[0] = md->offset_vector[1] = -1;
6136    }    }
6137    
6138  /* Set up the first character to match, if available. The first_byte value is  /* Set up the first character to match, if available. The first_byte value is
# Line 5876  if ((re->flags & PCRE_REQCHSET) != 0) Line 6166  if ((re->flags & PCRE_REQCHSET) != 0)
6166    }    }
6167    
6168    
6169    
6170    
6171  /* ==========================================================================*/  /* ==========================================================================*/
6172    
6173  /* Loop for handling unanchored repeated matching attempts; for anchored regexs  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
# Line 5886  for(;;) Line 6178  for(;;)
6178    USPTR save_end_subject = end_subject;    USPTR save_end_subject = end_subject;
6179    USPTR new_start_match;    USPTR new_start_match;
6180    
   /* Reset the maximum number of extractions we might see. */  
   
   if (md->offset_vector != NULL)  
     {  
     register int *iptr = md->offset_vector;  
     register int *iend = iptr + resetcount;  
     while (iptr < iend) *iptr++ = -1;  
     }  
   
6181    /* If firstline is TRUE, the start of the match is constrained to the first    /* If firstline is TRUE, the start of the match is constrained to the first
6182    line of a multiline string. That is, the match must be before or at the first    line of a multiline string. That is, the match must be before or at the first
6183    newline. Implement this by temporarily adjusting end_subject so that we stop    newline. Implement this by temporarily adjusting end_subject so that we stop
# Line 5922  for(;;) Line 6205  for(;;)
6205    /* There are some optimizations that avoid running the match if a known    /* There are some optimizations that avoid running the match if a known
6206    starting point is not found, or if a known later character is not present.    starting point is not found, or if a known later character is not present.
6207    However, there is an option that disables these, for testing and for ensuring    However, there is an option that disables these, for testing and for ensuring
6208    that all callouts do actually occur. */    that all callouts do actually occur. The option can be set in the regex by
6209      (*NO_START_OPT) or passed in match-time options. */
6210    
6211    if ((options & PCRE_NO_START_OPTIMIZE) == 0)    if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0)
6212      {      {
6213      /* Advance to a unique first byte if there is one. */      /* Advance to a unique first byte if there is one. */
6214    
# Line 6082  for(;;) Line 6366  for(;;)
6366    md->start_match_ptr = start_match;    md->start_match_ptr = start_match;
6367    md->start_used_ptr = start_match;    md->start_used_ptr = start_match;
6368    md->match_call_count = 0;    md->match_call_count = 0;
6369    rc = match(start_match, md->start_code, start_match, NULL, 2, md, ims, NULL,    md->match_function_type = 0;
6370      0, 0);    md->end_offset_top = 0;
6371      rc = match(start_match, md->start_code, start_match, NULL, 2, md, NULL, 0);
6372    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
6373    
6374    switch(rc)    switch(rc)
# Line 6193  if (rc == MATCH_MATCH || rc == MATCH_ACC Line 6478  if (rc == MATCH_MATCH || rc == MATCH_ACC
6478    {    {
6479    if (using_temporary_offsets)    if (using_temporary_offsets)
6480      {      {
6481      if (offsetcount >= 4)      if (arg_offset_max >= 4)
6482        {        {
6483        memcpy(offsets + 2, md->offset_vector + 2,        memcpy(offsets + 2, md->offset_vector + 2,
6484          (offsetcount - 2) * sizeof(int));          (arg_offset_max - 2) * sizeof(int));
6485        DPRINTF(("Copied offsets from temporary memory\n"));        DPRINTF(("Copied offsets from temporary memory\n"));
6486        }        }
6487      if (md->end_offset_top > offsetcount) md->offset_overflow = TRUE;      if (md->end_offset_top > arg_offset_max) md->offset_overflow = TRUE;
6488      DPRINTF(("Freeing temporary memory\n"));      DPRINTF(("Freeing temporary memory\n"));
6489      (pcre_free)(md->offset_vector);      (pcre_free)(md->offset_vector);
6490      }      }
6491    
6492    /* Set the return code to the number of captured strings, or 0 if there are    /* Set the return code to the number of captured strings, or 0 if there were
6493    too many to fit into the vector. */    too many to fit into the vector. */
6494    
6495    rc = md->offset_overflow? 0 : md->end_offset_top/2;    rc = (md->offset_overflow && md->end_offset_top >= arg_offset_max)?
6496        0 : md->end_offset_top/2;
6497    
6498      /* If there is space in the offset vector, set any unused pairs at the end of
6499      the pattern to -1 for backwards compatibility. It is documented that this
6500      happens. In earlier versions, the whole set of potential capturing offsets
6501      was set to -1 each time round the loop, but this is handled differently now.
6502      "Gaps" are set to -1 dynamically instead (this fixes a bug). Thus, it is only
6503      those at the end that need unsetting here. We can't just unset them all at
6504      the start of the whole thing because they may get set in one branch that is
6505      not the final matching branch. */
6506    
6507      if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
6508        {
6509        register int *iptr, *iend;
6510        int resetcount = 2 + re->top_bracket * 2;
6511        if (resetcount > offsetcount) resetcount = ocount;
6512        iptr = offsets + md->end_offset_top;
6513        iend = offsets + resetcount;
6514        while (iptr < iend) *iptr++ = -1;
6515        }
6516    
6517    /* If there is space, set up the whole thing as substring 0. The value of    /* If there is space, set up the whole thing as substring 0. The value of
6518    md->start_match_ptr might be modified if \K was encountered on the success    md->start_match_ptr might be modified if \K was encountered on the success

Legend:
Removed from v.550  
changed lines
  Added in v.716

  ViewVC Help
Powered by ViewVC 1.1.5