/[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 531 by ph10, Tue Jun 1 13:53:26 2010 UTC revision 721 by ph10, Fri Oct 7 15:51:39 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 (heapframe == 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;\
322    newframe->Xecode = rb;\    newframe->Xecode = rb;\
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    recursion_info new_recursive;
599    
600  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  BOOL cur_is_word;
 const uschar *charptr;             /* in small blocks of the code. My normal */  
 #endif                             /* style of coding would have declared    */  
 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 749  for (;;) Line 775  for (;;)
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 opcode, so that
779        the branch in which it occurs can be determined. Overload the start of
780        match pointer to do this. */
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;
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], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top,
791        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;
794      md->mark = ecode + 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 799  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 && rrc != MATCH_THEN) RRETURN(rrc);          if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
840    
841            /* 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 831  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            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 (flags == 0)    /* Not a possibly empty group */          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          if (!matched_once)
1045            {
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        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (md->mark == NULL) md->mark = markptr;
1052        otherwise return. */        if (allow_zero || matched_once)
1053            {
1054            ecode += 1 + LINK_SIZE;
1055            break;
1056            }
1057    
1058        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        RRETURN(MATCH_NOMATCH);
1059          eptrb, flags, RM2);        }
1060        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
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 888  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 900  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 1053  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 && rrc != MATCH_THEN)  
1323          /* 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 1077  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 */  
1350          {          {
1351          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);          ecode += 1 + LINK_SIZE;
         RRETURN(rrc);  
         }  
       else                       /* Group must match something */  
         {  
         flags = 0;  
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 1127  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 1169  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    
1453          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1454          as NOMATCH. */
1455    
1456        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        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 1218  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    
1498          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1499          as NOMATCH. */
1500    
1501        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        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 1279  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 1291  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 1301  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 1343  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 && rrc != MATCH_THEN)  
1659            /* 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 1374  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 1386  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) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     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. */  
1681    
1682      if (ecode[1+LINK_SIZE] == OP_OPT)      RECURSION_MATCHED:
1683        {      break;
       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 1481  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_ASSERTBACK_NOT)
         *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||  
         *prev == OP_ONCE)  
1754        {        {
1755        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1756        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1757        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1758        MRRETURN(MATCH_MATCH);        MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1759        }        }
1760    
1761      /* 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
1762      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1763      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
1764      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
1765      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
1766        the current subject position and start match pointer and give a MATCH
1767        return. */
1768    
1769      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1770            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1771        {        {
1772        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1773        offset = number << 1;        offset = number << 1;
# Line 1557  for (;;) Line 1777  for (;;)
1777        printf("\n");        printf("\n");
1778  #endif  #endif
1779    
1780          /* Handle a recursively called group. */
1781    
1782          if (md->recursive != NULL && md->recursive->group_num == number)
1783            {
1784            md->end_match_ptr = eptr;
1785            md->start_match_ptr = mstart;
1786            RRETURN(MATCH_MATCH);
1787            }
1788    
1789          /* Deal with capturing */
1790    
1791        md->capture_last = number;        md->capture_last = number;
1792        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1793          {          {
1794            /* If offset is greater than offset_top, it means that we are
1795            "skipping" a capturing group, and that group's offsets must be marked
1796            unset. In earlier versions of PCRE, all the offsets were unset at the
1797            start of matching, but this doesn't work because atomic groups and
1798            assertions can cause a value to be set that should later be unset.
1799            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1800            part of the atomic group, but this is not on the final matching path,
1801            so must be unset when 2 is set. (If there is no group 2, there is no
1802            problem, because offset_top will then be 2, indicating no capture.) */
1803    
1804            if (offset > offset_top)
1805              {
1806              register int *iptr = md->offset_vector + offset_top;
1807              register int *iend = md->offset_vector + offset;
1808              while (iptr < iend) *iptr++ = -1;
1809              }
1810    
1811            /* Now make the extraction */
1812    
1813          md->offset_vector[offset] =          md->offset_vector[offset] =
1814            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1815          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1816          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1817          }          }
1818          }
1819    
1820        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1821        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1822        group. This is the forcible breaking of infinite loops as implemented in
1823        Perl 5.005. For a non-repeating atomic group, establish a backup point by
1824        processing the rest of the pattern at a lower level. If this results in a
1825        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1826        bypassing intermediate backup points, but resetting any captures that
1827        happened along the way. */
1828    
1829        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1830          {
1831          if (*prev == OP_ONCE)
1832          {          {
1833          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1834          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1835          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1836          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;  
1837          }          }
1838          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1839          break;
1840        }        }
1841    
1842      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1843      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1844        at a time from the outer level, thus saving stack. */
1845    
1846      ims = original_ims;      if (*ecode == OP_KETRPOS)
     DPRINTF(("ims reset to %02lx\n", ims));  
   
     /* 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)  
1847        {        {
1848        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1849        break;        md->end_offset_top = offset_top;
1850          RRETURN(MATCH_KETRPOS);
1851        }        }
1852    
1853      /* 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
1854      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
1855      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
1856      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
1857        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
1858    
1859      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1860        {        {
1861        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1862        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1863        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
1864          {          {
1865          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1866            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1867            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1868            RRETURN(MATCH_ONCE);
1869            }
1870          if (*prev >= OP_SBRA)    /* Could match an empty string */
1871            {
1872            md->match_function_type = MATCH_CBEGROUP;
1873            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1874          RRETURN(rrc);          RRETURN(rrc);
1875          }          }
1876        ecode = prev;        ecode = prev;
# Line 1622  for (;;) Line 1878  for (;;)
1878        }        }
1879      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1880        {        {
1881        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1882          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1883          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1884        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1885          if (*prev == OP_ONCE)
1886            {
1887            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1888            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1889            md->once_target = prev;
1890            RRETURN(MATCH_ONCE);
1891            }
1892        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1893        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1894        }        }
1895      /* Control never gets here */      /* Control never gets here */
1896    
1897      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1898    
1899      case OP_CIRC:      case OP_CIRC:
1900      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 */  
1901    
1902      /* Start of subject assertion */      /* Start of subject assertion */
1903    
# Line 1651  for (;;) Line 1906  for (;;)
1906      ecode++;      ecode++;
1907      break;      break;
1908    
1909        /* Multiline mode: start of subject unless notbol, or after any newline. */
1910    
1911        case OP_CIRCM:
1912        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1913        if (eptr != md->start_subject &&
1914            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1915          MRRETURN(MATCH_NOMATCH);
1916        ecode++;
1917        break;
1918    
1919      /* Start of match assertion */      /* Start of match assertion */
1920    
1921      case OP_SOM:      case OP_SOM:
# Line 1665  for (;;) Line 1930  for (;;)
1930      ecode++;      ecode++;
1931      break;      break;
1932    
1933      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
1934      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
1935    
1936      case OP_DOLL:      case OP_DOLLM:
1937      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
1938        {        { 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;  
       }  
1939      else      else
1940        {        {
1941        if (md->noteol) MRRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1942        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           MRRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
1943        }        }
1944        ecode++;
1945        break;
1946    
1947        /* Not multiline mode: assert before a terminating newline or before end of
1948        subject unless noteol is set. */
1949    
1950        case OP_DOLL:
1951        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1952        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1953    
1954      /* ... else fall through for endonly */      /* ... else fall through for endonly */
1955    
1956      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
1957    
1958      case OP_EOD:      case OP_EOD:
1959      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
1960        SCHECK_PARTIAL();
1961      ecode++;      ecode++;
1962      break;      break;
1963    
1964      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
1965    
1966      case OP_EODN:      case OP_EODN:
1967      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
1968        if (eptr < md->end_subject &&
1969          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1970        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1971    
1972        /* Either at end of string or \n before end. */
1973    
1974        SCHECK_PARTIAL();
1975      ecode++;      ecode++;
1976      break;      break;
1977    
# Line 1833  for (;;) Line 2100  for (;;)
2100      /* Fall through */      /* Fall through */
2101    
2102      case OP_ALLANY:      case OP_ALLANY:
2103      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2104        {        {                            /* not be updated before SCHECK_PARTIAL. */
2105        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2106        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2107        }        }
2108        eptr++;
2109      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2110      ecode++;      ecode++;
2111      break;      break;
# Line 1846  for (;;) Line 2114  for (;;)
2114      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2115    
2116      case OP_ANYBYTE:      case OP_ANYBYTE:
2117      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2118        {        {                            /* not be updated before SCHECK_PARTIAL. */
2119        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2120        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2121        }        }
2122        eptr++;
2123      ecode++;      ecode++;
2124      break;      break;
2125    
# Line 1966  for (;;) Line 2235  for (;;)
2235      switch(c)      switch(c)
2236        {        {
2237        default: MRRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2238    
2239        case 0x000d:        case 0x000d:
2240        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2241        break;        break;
# Line 2189  for (;;) Line 2459  for (;;)
2459        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2460        }        }
2461      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2462        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2463        while (eptr < md->end_subject)
2464        {        {
2465        int category = UCD_CATEGORY(c);        int len = 1;
2466        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2467        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2468          {        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;  
         }  
2469        }        }
2470      ecode++;      ecode++;
2471      break;      break;
# Line 2218  for (;;) Line 2481  for (;;)
2481      loops). */      loops). */
2482    
2483      case OP_REF:      case OP_REF:
2484        {      case OP_REFI:
2485        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2486        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2487        ecode += 3;
2488    
2489        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2490    
2491        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2492        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
2493        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.  
2494    
2495        (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
2496        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2497    
2498        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
2499        referenced subpattern. */      referenced subpattern. */
2500    
2501        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2502          length = (md->jscript_compat)? 0 : (int)(md->end_subject - eptr + 1);        length = (md->jscript_compat)? 0 : -1;
2503        else      else
2504          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2505    
2506        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2507    
2508        switch (*ecode)      switch (*ecode)
2509          {        {
2510          case OP_CRSTAR:        case OP_CRSTAR:
2511          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2512          case OP_CRPLUS:        case OP_CRPLUS:
2513          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2514          case OP_CRQUERY:        case OP_CRQUERY:
2515          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2516          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2517          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2518          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2519          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2520          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2521          break;        break;
2522    
2523          case OP_CRRANGE:        case OP_CRRANGE:
2524          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2525          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2526          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2527          max = GET2(ecode, 3);        max = GET2(ecode, 3);
2528          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2529          ecode += 5;        ecode += 5;
2530          break;        break;
2531    
2532          default:               /* No repeat follows */        default:               /* No repeat follows */
2533          if (!match_ref(offset, eptr, length, md, ims))        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2534            {          {
2535            CHECK_PARTIAL();          CHECK_PARTIAL();
2536            MRRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
           }  
         eptr += length;  
         continue;              /* With the main loop */  
2537          }          }
2538          eptr += length;
2539          continue;              /* With the main loop */
2540          }
2541    
2542        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2543        main loop. */      zero, just continue with the main loop. */
2544    
2545        if (length == 0) continue;      if (length == 0) continue;
2546    
2547        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2548        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2549        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2550    
2551        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2552          {
2553          int slength;
2554          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2555          {          {
2556          if (!match_ref(offset, eptr, length, md, ims))          CHECK_PARTIAL();
2557            {          MRRETURN(MATCH_NOMATCH);
           CHECK_PARTIAL();  
           MRRETURN(MATCH_NOMATCH);  
           }  
         eptr += length;  
2558          }          }
2559          eptr += slength;
2560          }
2561    
2562        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2563        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2564    
2565        if (min == max) continue;      if (min == max) continue;
2566    
2567        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2568    
2569        if (minimize)      if (minimize)
2570          {
2571          for (fi = min;; fi++)
2572          {          {
2573          for (fi = min;; fi++)          int slength;
2574            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2575            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2576            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2577            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2578            {            {
2579            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            CHECK_PARTIAL();
2580            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;  
2581            }            }
2582          /* Control never gets here */          eptr += slength;
2583          }          }
2584          /* Control never gets here */
2585          }
2586    
2587        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2588    
2589        else      else
2590          {
2591          pp = eptr;
2592          for (i = min; i < max; i++)
2593          {          {
2594          pp = eptr;          int slength;
2595          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims))  
             {  
             CHECK_PARTIAL();  
             break;  
             }  
           eptr += length;  
           }  
         while (eptr >= pp)  
2596            {            {
2597            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            CHECK_PARTIAL();
2598            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2599            }            }
2600          MRRETURN(MATCH_NOMATCH);          eptr += slength;
2601            }
2602          while (eptr >= pp)
2603            {
2604            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2605            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2606            eptr -= length;
2607          }          }
2608          MRRETURN(MATCH_NOMATCH);
2609        }        }
2610      /* Control never gets here */      /* Control never gets here */
2611    
# Line 2446  for (;;) Line 2711  for (;;)
2711            {            {
2712            for (fi = min;; fi++)            for (fi = min;; fi++)
2713              {              {
2714              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2715              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2716              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2717              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2471  for (;;) Line 2736  for (;;)
2736            {            {
2737            for (fi = min;; fi++)            for (fi = min;; fi++)
2738              {              {
2739              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2740              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2741              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2742              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2517  for (;;) Line 2782  for (;;)
2782              }              }
2783            for (;;)            for (;;)
2784              {              {
2785              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2786              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2787              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2788              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2540  for (;;) Line 2805  for (;;)
2805              }              }
2806            while (eptr >= pp)            while (eptr >= pp)
2807              {              {
2808              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2809              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2810              eptr--;              eptr--;
2811              }              }
# Line 2616  for (;;) Line 2881  for (;;)
2881          {          {
2882          for (fi = min;; fi++)          for (fi = min;; fi++)
2883            {            {
2884            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2885            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2886            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2887            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 2649  for (;;) Line 2914  for (;;)
2914            }            }
2915          for(;;)          for(;;)
2916            {            {
2917            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2918            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2919            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2920            if (utf8) BACKCHAR(eptr);            if (utf8) BACKCHAR(eptr);
# Line 2694  for (;;) Line 2959  for (;;)
2959    
2960      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2961    
2962      case OP_CHARNC:      case OP_CHARI:
2963  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2964      if (utf8)      if (utf8)
2965        {        {
# Line 2754  for (;;) Line 3019  for (;;)
3019      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3020    
3021      case OP_EXACT:      case OP_EXACT:
3022        case OP_EXACTI:
3023      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3024      ecode += 3;      ecode += 3;
3025      goto REPEATCHAR;      goto REPEATCHAR;
3026    
3027      case OP_POSUPTO:      case OP_POSUPTO:
3028        case OP_POSUPTOI:
3029      possessive = TRUE;      possessive = TRUE;
3030      /* Fall through */      /* Fall through */
3031    
3032      case OP_UPTO:      case OP_UPTO:
3033        case OP_UPTOI:
3034      case OP_MINUPTO:      case OP_MINUPTO:
3035        case OP_MINUPTOI:
3036      min = 0;      min = 0;
3037      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3038      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3039      ecode += 3;      ecode += 3;
3040      goto REPEATCHAR;      goto REPEATCHAR;
3041    
3042      case OP_POSSTAR:      case OP_POSSTAR:
3043        case OP_POSSTARI:
3044      possessive = TRUE;      possessive = TRUE;
3045      min = 0;      min = 0;
3046      max = INT_MAX;      max = INT_MAX;
# Line 2778  for (;;) Line 3048  for (;;)
3048      goto REPEATCHAR;      goto REPEATCHAR;
3049    
3050      case OP_POSPLUS:      case OP_POSPLUS:
3051        case OP_POSPLUSI:
3052      possessive = TRUE;      possessive = TRUE;
3053      min = 1;      min = 1;
3054      max = INT_MAX;      max = INT_MAX;
# Line 2785  for (;;) Line 3056  for (;;)
3056      goto REPEATCHAR;      goto REPEATCHAR;
3057    
3058      case OP_POSQUERY:      case OP_POSQUERY:
3059        case OP_POSQUERYI:
3060      possessive = TRUE;      possessive = TRUE;
3061      min = 0;      min = 0;
3062      max = 1;      max = 1;
# Line 2792  for (;;) Line 3064  for (;;)
3064      goto REPEATCHAR;      goto REPEATCHAR;
3065    
3066      case OP_STAR:      case OP_STAR:
3067        case OP_STARI:
3068      case OP_MINSTAR:      case OP_MINSTAR:
3069        case OP_MINSTARI:
3070      case OP_PLUS:      case OP_PLUS:
3071        case OP_PLUSI:
3072      case OP_MINPLUS:      case OP_MINPLUS:
3073        case OP_MINPLUSI:
3074      case OP_QUERY:      case OP_QUERY:
3075        case OP_QUERYI:
3076      case OP_MINQUERY:      case OP_MINQUERY:
3077      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3078        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3079      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
   
3080      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3081      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3082      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
# Line 2822  for (;;) Line 3099  for (;;)
3099          {          {
3100  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3101          unsigned int othercase;          unsigned int othercase;
3102          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3103              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3104            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3105          else oclength = 0;          else oclength = 0;
# Line 2850  for (;;) Line 3127  for (;;)
3127            {            {
3128            for (fi = min;; fi++)            for (fi = min;; fi++)
3129              {              {
3130              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3131              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3132              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3133              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
# Line 2892  for (;;) Line 3169  for (;;)
3169    
3170            for(;;)            for(;;)
3171              {              {
3172              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3173              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3174              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3175  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2929  for (;;) Line 3206  for (;;)
3206      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3207        max, eptr));        max, eptr));
3208    
3209      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3210        {        {
3211        fc = md->lcc[fc];        fc = md->lcc[fc];
3212        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
# Line 2946  for (;;) Line 3223  for (;;)
3223          {          {
3224          for (fi = min;; fi++)          for (fi = min;; fi++)
3225            {            {
3226            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3227            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3228            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3229            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 2976  for (;;) Line 3253  for (;;)
3253    
3254          while (eptr >= pp)          while (eptr >= pp)
3255            {            {
3256            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3257            eptr--;            eptr--;
3258            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3259            }            }
# Line 3005  for (;;) Line 3282  for (;;)
3282          {          {
3283          for (fi = min;; fi++)          for (fi = min;; fi++)
3284            {            {
3285            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3286            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3287            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3288            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 3034  for (;;) Line 3311  for (;;)
3311    
3312          while (eptr >= pp)          while (eptr >= pp)
3313            {            {
3314            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3315            eptr--;            eptr--;
3316            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3317            }            }
# Line 3047  for (;;) Line 3324  for (;;)
3324      checking can be multibyte. */      checking can be multibyte. */
3325    
3326      case OP_NOT:      case OP_NOT:
3327        case OP_NOTI:
3328      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3329        {        {
3330        SCHECK_PARTIAL();        SCHECK_PARTIAL();
# Line 3054  for (;;) Line 3332  for (;;)
3332        }        }
3333      ecode++;      ecode++;
3334      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3335      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3336        {        {
3337  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3338        if (c < 256)        if (c < 256)
# Line 3062  for (;;) Line 3340  for (;;)
3340        c = md->lcc[c];        c = md->lcc[c];
3341        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3342        }        }
3343      else      else    /* Caseful */
3344        {        {
3345        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3346        }        }
# Line 3076  for (;;) Line 3354  for (;;)
3354      about... */      about... */
3355    
3356      case OP_NOTEXACT:      case OP_NOTEXACT:
3357        case OP_NOTEXACTI:
3358      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3359      ecode += 3;      ecode += 3;
3360      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3361    
3362      case OP_NOTUPTO:      case OP_NOTUPTO:
3363        case OP_NOTUPTOI:
3364      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3365        case OP_NOTMINUPTOI:
3366      min = 0;      min = 0;
3367      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3368      minimize = *ecode == OP_NOTMINUPTO;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3369      ecode += 3;      ecode += 3;
3370      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3371    
3372      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
3373        case OP_NOTPOSSTARI:
3374      possessive = TRUE;      possessive = TRUE;
3375      min = 0;      min = 0;
3376      max = INT_MAX;      max = INT_MAX;
# Line 3096  for (;;) Line 3378  for (;;)
3378      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3379    
3380      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
3381        case OP_NOTPOSPLUSI:
3382      possessive = TRUE;      possessive = TRUE;
3383      min = 1;      min = 1;
3384      max = INT_MAX;      max = INT_MAX;
# Line 3103  for (;;) Line 3386  for (;;)
3386      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3387    
3388      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
3389        case OP_NOTPOSQUERYI:
3390      possessive = TRUE;      possessive = TRUE;
3391      min = 0;      min = 0;
3392      max = 1;      max = 1;
# Line 3110  for (;;) Line 3394  for (;;)
3394      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3395    
3396      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
3397        case OP_NOTPOSUPTOI:
3398      possessive = TRUE;      possessive = TRUE;
3399      min = 0;      min = 0;
3400      max = GET2(ecode, 1);      max = GET2(ecode, 1);
# Line 3117  for (;;) Line 3402  for (;;)
3402      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3403    
3404      case OP_NOTSTAR:      case OP_NOTSTAR:
3405        case OP_NOTSTARI:
3406      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3407        case OP_NOTMINSTARI:
3408      case OP_NOTPLUS:      case OP_NOTPLUS:
3409        case OP_NOTPLUSI:
3410      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3411        case OP_NOTMINPLUSI:
3412      case OP_NOTQUERY:      case OP_NOTQUERY:
3413        case OP_NOTQUERYI:
3414      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3415      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3416        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3417      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3418      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3419      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
# Line 3144  for (;;) Line 3435  for (;;)
3435      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,
3436        max, eptr));        max, eptr));
3437    
3438      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3439        {        {
3440        fc = md->lcc[fc];        fc = md->lcc[fc];
3441    
# Line 3192  for (;;) Line 3483  for (;;)
3483            register unsigned int d;            register unsigned int d;
3484            for (fi = min;; fi++)            for (fi = min;; fi++)
3485              {              {
3486              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM28);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3487              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3488              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3489              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3211  for (;;) Line 3502  for (;;)
3502            {            {
3503            for (fi = min;; fi++)            for (fi = min;; fi++)
3504              {              {
3505              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM29);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3506              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3507              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3508              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3252  for (;;) Line 3543  for (;;)
3543          if (possessive) continue;          if (possessive) continue;
3544          for(;;)          for(;;)
3545              {              {
3546              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM30);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3547              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3548              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3549              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 3275  for (;;) Line 3566  for (;;)
3566            if (possessive) continue;            if (possessive) continue;
3567            while (eptr >= pp)            while (eptr >= pp)
3568              {              {
3569              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM31);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3570              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3571              eptr--;              eptr--;
3572              }              }
# Line 3332  for (;;) Line 3623  for (;;)
3623            register unsigned int d;            register unsigned int d;
3624            for (fi = min;; fi++)            for (fi = min;; fi++)
3625              {              {
3626              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM32);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3627              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3628              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3629              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3350  for (;;) Line 3641  for (;;)
3641            {            {
3642            for (fi = min;; fi++)            for (fi = min;; fi++)
3643              {              {
3644              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM33);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
3645              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3646              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3647              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3390  for (;;) Line 3681  for (;;)
3681            if (possessive) continue;            if (possessive) continue;
3682            for(;;)            for(;;)
3683              {              {
3684              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM34);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
3685              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3686              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3687              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 3413  for (;;) Line 3704  for (;;)
3704            if (possessive) continue;            if (possessive) continue;
3705            while (eptr >= pp)            while (eptr >= pp)
3706              {              {
3707              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM35);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
3708              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3709              eptr--;              eptr--;
3710              }              }
# Line 3528  for (;;) Line 3819  for (;;)
3819            case PT_LAMP:            case PT_LAMP:
3820            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3821              {              {
3822                int chartype;
3823              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3824                {                {
3825                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3826                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3827                }                }
3828              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3829              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
3830              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
3831                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
3832                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
3833                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3834              }              }
3835            break;            break;
# Line 3551  for (;;) Line 3843  for (;;)
3843                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3844                }                }
3845              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3846              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
3847                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3848              }              }
3849            break;            break;
# Line 3566  for (;;) Line 3857  for (;;)
3857                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3858                }                }
3859              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3860              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
3861                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3862              }              }
3863            break;            break;
# Line 3581  for (;;) Line 3871  for (;;)
3871                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3872                }                }
3873              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3874              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
3875                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3876              }              }
3877            break;            break;
# Line 3590  for (;;) Line 3879  for (;;)
3879            case PT_ALNUM:            case PT_ALNUM:
3880            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3881              {              {
3882                int category;
3883              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3884                {                {
3885                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3886                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3887                }                }
3888              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3889              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3890              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
3891                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3892              }              }
3893            break;            break;
# Line 3612  for (;;) Line 3901  for (;;)
3901                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3902                }                }
3903              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3904              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 ||  
3905                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
3906                     == prop_fail_result)                     == prop_fail_result)
3907                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3629  for (;;) Line 3917  for (;;)
3917                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3918                }                }
3919              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3920              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 ||  
3921                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
3922                     == prop_fail_result)                     == prop_fail_result)
3923                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3640  for (;;) Line 3927  for (;;)
3927            case PT_WORD:            case PT_WORD:
3928            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3929              {              {
3930                int category;
3931              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3932                {                {
3933                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3934                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3935                }                }
3936              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3937              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3938              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
                  c == CHAR_UNDERSCORE)  
3939                     == prop_fail_result)                     == prop_fail_result)
3940                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3941              }              }
# Line 3674  for (;;) Line 3961  for (;;)
3961              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3962              }              }
3963            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3964            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
3965            while (eptr < md->end_subject)            while (eptr < md->end_subject)
3966              {              {
3967              int len = 1;              int len = 1;
3968              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
3969                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
3970              eptr += len;              eptr += len;
3971              }              }
3972            }            }
# Line 3740  for (;;) Line 4024  for (;;)
4024            switch(c)            switch(c)
4025              {              {
4026              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4027    
4028              case 0x000d:              case 0x000d:
4029              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4030              break;              break;
# Line 4016  for (;;) Line 4301  for (;;)
4301            switch(*eptr++)            switch(*eptr++)
4302              {              {
4303              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4304    
4305              case 0x000d:              case 0x000d:
4306              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4307              break;              break;
4308    
4309              case 0x000a:              case 0x000a:
4310              break;              break;
4311    
# Line 4208  for (;;) Line 4495  for (;;)
4495            case PT_ANY:            case PT_ANY:
4496            for (fi = min;; fi++)            for (fi = min;; fi++)
4497              {              {
4498              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM36);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4499              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4500              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4501              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4224  for (;;) Line 4511  for (;;)
4511            case PT_LAMP:            case PT_LAMP:
4512            for (fi = min;; fi++)            for (fi = min;; fi++)
4513              {              {
4514              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM37);              int chartype;
4515                RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4516              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4517              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4518              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4233  for (;;) Line 4521  for (;;)
4521                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4522                }                }
4523              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4524              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4525              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4526                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4527                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4528                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4529              }              }
4530            /* Control never gets here */            /* Control never gets here */
# Line 4244  for (;;) Line 4532  for (;;)
4532            case PT_GC:            case PT_GC:
4533            for (fi = min;; fi++)            for (fi = min;; fi++)
4534              {              {
4535              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM38);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4536              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4537              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4538              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4253  for (;;) Line 4541  for (;;)
4541                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4542                }                }
4543              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4544              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
4545                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4546              }              }
4547            /* Control never gets here */            /* Control never gets here */
# Line 4262  for (;;) Line 4549  for (;;)
4549            case PT_PC:            case PT_PC:
4550            for (fi = min;; fi++)            for (fi = min;; fi++)
4551              {              {
4552              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM39);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4553              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4554              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4555              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4271  for (;;) Line 4558  for (;;)
4558                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4559                }                }
4560              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4561              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
4562                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4563              }              }
4564            /* Control never gets here */            /* Control never gets here */
# Line 4280  for (;;) Line 4566  for (;;)
4566            case PT_SC:            case PT_SC:
4567            for (fi = min;; fi++)            for (fi = min;; fi++)
4568              {              {
4569              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM40);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
4570              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4571              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4572              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4289  for (;;) Line 4575  for (;;)
4575                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4576                }                }
4577              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4578              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
4579                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4580              }              }
4581            /* Control never gets here */            /* Control never gets here */
# Line 4298  for (;;) Line 4583  for (;;)
4583            case PT_ALNUM:            case PT_ALNUM:
4584            for (fi = min;; fi++)            for (fi = min;; fi++)
4585              {              {
4586              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM59);              int category;
4587                RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4588              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4589              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4590              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4307  for (;;) Line 4593  for (;;)
4593                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4594                }                }
4595              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4596              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4597              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
4598                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4599              }              }
4600            /* Control never gets here */            /* Control never gets here */
# Line 4317  for (;;) Line 4602  for (;;)
4602            case PT_SPACE:    /* Perl space */            case PT_SPACE:    /* Perl space */
4603            for (fi = min;; fi++)            for (fi = min;; fi++)
4604              {              {
4605              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM60);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM60);
4606              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4607              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4608              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4326  for (;;) Line 4611  for (;;)
4611                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4612                }                }
4613              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4614              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 ||  
4615                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
4616                     == prop_fail_result)                     == prop_fail_result)
4617                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4337  for (;;) Line 4621  for (;;)
4621            case PT_PXSPACE:  /* POSIX space */            case PT_PXSPACE:  /* POSIX space */
4622            for (fi = min;; fi++)            for (fi = min;; fi++)
4623              {              {
4624              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM61);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
4625              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4626              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4627              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4346  for (;;) Line 4630  for (;;)
4630                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4631                }                }
4632              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4633              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 ||  
4634                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4635                     == prop_fail_result)                     == prop_fail_result)
4636                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4357  for (;;) Line 4640  for (;;)
4640            case PT_WORD:            case PT_WORD:
4641            for (fi = min;; fi++)            for (fi = min;; fi++)
4642              {              {
4643              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM62);              int category;
4644                RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4645              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4646              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4647              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4366  for (;;) Line 4650  for (;;)
4650                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4651                }                }
4652              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4653              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4654              if ((prop_category == ucp_L ||              if ((category == ucp_L ||
4655                   prop_category == ucp_N ||                   category == ucp_N ||
4656                   c == CHAR_UNDERSCORE)                   c == CHAR_UNDERSCORE)
4657                     == prop_fail_result)                     == prop_fail_result)
4658                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4389  for (;;) Line 4673  for (;;)
4673          {          {
4674          for (fi = min;; fi++)          for (fi = min;; fi++)
4675            {            {
4676            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM41);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
4677            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4678            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4679            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4398  for (;;) Line 4682  for (;;)
4682              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
4683              }              }
4684            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
4685            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
4686            while (eptr < md->end_subject)            while (eptr < md->end_subject)
4687              {              {
4688              int len = 1;              int len = 1;
4689              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4690                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
4691              eptr += len;              eptr += len;
4692              }              }
4693            }            }
4694          }          }
   
4695        else        else
4696  #endif     /* SUPPORT_UCP */  #endif     /* SUPPORT_UCP */
4697    
# Line 4421  for (;;) Line 4701  for (;;)
4701          {          {
4702          for (fi = min;; fi++)          for (fi = min;; fi++)
4703            {            {
4704            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM42);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
4705            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4706            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4707            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4584  for (;;) Line 4864  for (;;)
4864          {          {
4865          for (fi = min;; fi++)          for (fi = min;; fi++)
4866            {            {
4867            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM43);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
4868            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4869            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4870            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4732  for (;;) Line 5012  for (;;)
5012            case PT_LAMP:            case PT_LAMP:
5013            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5014              {              {
5015                int chartype;
5016              int len = 1;              int len = 1;
5017              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5018                {                {
# Line 4739  for (;;) Line 5020  for (;;)
5020                break;                break;
5021                }                }
5022              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5023              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
5024              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
5025                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
5026                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
5027                break;                break;
5028              eptr+= len;              eptr+= len;
5029              }              }
# Line 4758  for (;;) Line 5039  for (;;)
5039                break;                break;
5040                }                }
5041              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5042              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
             if ((prop_category == prop_value) == prop_fail_result)  
               break;  
5043              eptr+= len;              eptr+= len;
5044              }              }
5045            break;            break;
# Line 4775  for (;;) Line 5054  for (;;)
5054                break;                break;
5055                }                }
5056              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5057              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
             if ((prop_chartype == prop_value) == prop_fail_result)  
               break;  
5058              eptr+= len;              eptr+= len;
5059              }              }
5060            break;            break;
# Line 4792  for (;;) Line 5069  for (;;)
5069                break;                break;
5070                }                }
5071              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5072              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
             if ((prop_script == prop_value) == prop_fail_result)  
               break;  
5073              eptr+= len;              eptr+= len;
5074              }              }
5075            break;            break;
# Line 4802  for (;;) Line 5077  for (;;)
5077            case PT_ALNUM:            case PT_ALNUM:
5078            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5079              {              {
5080                int category;
5081              int len = 1;              int len = 1;
5082              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5083                {                {
# Line 4809  for (;;) Line 5085  for (;;)
5085                break;                break;
5086                }                }
5087              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5088              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5089              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                  == prop_fail_result)  
5090                break;                break;
5091              eptr+= len;              eptr+= len;
5092              }              }
# Line 4827  for (;;) Line 5102  for (;;)
5102                break;                break;
5103                }                }
5104              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5105              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 ||  
5106                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
5107                   == prop_fail_result)                   == prop_fail_result)
5108                break;                break;
# Line 4846  for (;;) Line 5120  for (;;)
5120                break;                break;
5121                }                }
5122              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5123              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 ||  
5124                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5125                   == prop_fail_result)                   == prop_fail_result)
5126                break;                break;
# Line 4858  for (;;) Line 5131  for (;;)
5131            case PT_WORD:            case PT_WORD:
5132            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5133              {              {
5134                int category;
5135              int len = 1;              int len = 1;
5136              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5137                {                {
# Line 4865  for (;;) Line 5139  for (;;)
5139                break;                break;
5140                }                }
5141              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5142              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5143              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N ||
5144                   c == CHAR_UNDERSCORE) == prop_fail_result)                   c == CHAR_UNDERSCORE) == prop_fail_result)
5145                break;                break;
5146              eptr+= len;              eptr+= len;
# Line 4882  for (;;) Line 5156  for (;;)
5156          if (possessive) continue;          if (possessive) continue;
5157          for(;;)          for(;;)
5158            {            {
5159            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM44);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5160            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5161            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5162            if (utf8) BACKCHAR(eptr);            if (utf8) BACKCHAR(eptr);
# Line 4896  for (;;) Line 5170  for (;;)
5170          {          {
5171          for (i = min; i < max; i++)          for (i = min; i < max; i++)
5172            {            {
5173              int len = 1;
5174            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
5175              {              {
5176              SCHECK_PARTIAL();              SCHECK_PARTIAL();
5177              break;              break;
5178              }              }
5179            GETCHARINCTEST(c, eptr);            if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5180            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) break;
5181            if (prop_category == ucp_M) break;            eptr += len;
5182            while (eptr < md->end_subject)            while (eptr < md->end_subject)
5183              {              {
5184              int len = 1;              len = 1;
5185              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5186                {              if (UCD_CATEGORY(c) != ucp_M) break;
               GETCHARLEN(c, eptr, len);  
               }  
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
5187              eptr += len;              eptr += len;
5188              }              }
5189            }            }
# Line 4923  for (;;) Line 5194  for (;;)
5194    
5195          for(;;)          for(;;)
5196            {            {
5197            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM45);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5198            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5199            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5200            for (;;)                        /* Move back over one extended */            for (;;)                        /* Move back over one extended */
5201              {              {
             int len = 1;  
5202              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else
5203                {                {
5204                BACKCHAR(eptr);                BACKCHAR(eptr);
5205                GETCHARLEN(c, eptr, len);                GETCHAR(c, eptr);
5206                }                }
5207              prop_category = UCD_CATEGORY(c);              if (UCD_CATEGORY(c) != ucp_M) break;
             if (prop_category != ucp_M) break;  
5208              eptr--;              eptr--;
5209              }              }
5210            }            }
# Line 4999  for (;;) Line 5268  for (;;)
5268                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
5269                }                }
5270              }              }
5271            else eptr = md->end_subject;   /* Unlimited UTF-8 repeat */            else
5272                {
5273                eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
5274                SCHECK_PARTIAL();
5275                }
5276            break;            break;
5277    
5278            /* The byte case is the same as non-UTF8 */            /* The byte case is the same as non-UTF8 */
# Line 5207  for (;;) Line 5480  for (;;)
5480            RRETURN(PCRE_ERROR_INTERNAL);            RRETURN(PCRE_ERROR_INTERNAL);
5481            }            }
5482    
5483          /* eptr is now past the end of the maximum run */          /* eptr is now past the end of the maximum run. If possessive, we are
5484            done (no backing up). Otherwise, match at this position; anything other
5485            than no match is immediately returned. For nomatch, back up one
5486            character, unless we are matching \R and the last thing matched was
5487            \r\n, in which case, back up two bytes. */
5488    
5489          if (possessive) continue;          if (possessive) continue;
5490          for(;;)          for(;;)
5491            {            {
5492            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM46);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
5493            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5494            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5495            BACKCHAR(eptr);            BACKCHAR(eptr);
5496              if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5497                  eptr[-1] == '\r') eptr--;
5498            }            }
5499          }          }
5500        else        else
# Line 5414  for (;;) Line 5693  for (;;)
5693            RRETURN(PCRE_ERROR_INTERNAL);            RRETURN(PCRE_ERROR_INTERNAL);
5694            }            }
5695    
5696          /* eptr is now past the end of the maximum run */          /* eptr is now past the end of the maximum run. If possessive, we are
5697            done (no backing up). Otherwise, match at this position; anything other
5698            than no match is immediately returned. For nomatch, back up one
5699            character (byte), unless we are matching \R and the last thing matched
5700            was \r\n, in which case, back up two bytes. */
5701    
5702          if (possessive) continue;          if (possessive) continue;
5703          while (eptr >= pp)          while (eptr >= pp)
5704            {            {
5705            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM47);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
           eptr--;  
5706            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5707              eptr--;
5708              if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5709                  eptr[-1] == '\r') eptr--;
5710            }            }
5711          }          }
5712    
# Line 5460  switch (frame->Xwhere) Line 5745  switch (frame->Xwhere)
5745    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)
5746    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)
5747    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)
5748    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)
5749  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5750    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)
5751    LBL(32) LBL(34) LBL(42) LBL(46)    LBL(32) LBL(34) LBL(42) LBL(46)
# Line 5489  Undefine all the macros that were define Line 5774  Undefine all the macros that were define
5774  #undef ecode  #undef ecode
5775  #undef mstart  #undef mstart
5776  #undef offset_top  #undef offset_top
 #undef ims  
5777  #undef eptrb  #undef eptrb
5778  #undef flags  #undef flags
5779    
# Line 5507  Undefine all the macros that were define Line 5791  Undefine all the macros that were define
5791  #undef condition  #undef condition
5792  #undef prev_is_word  #undef prev_is_word
5793    
 #undef original_ims  
   
5794  #undef ctype  #undef ctype
5795  #undef length  #undef length
5796  #undef max  #undef max
# Line 5565  pcre_exec(const pcre *argument_re, const Line 5847  pcre_exec(const pcre *argument_re, const
5847    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5848    int offsetcount)    int offsetcount)
5849  {  {
5850  int rc, resetcount, ocount;  int rc, ocount, arg_offset_max;
5851  int first_byte = -1;  int first_byte = -1;
5852  int req_byte = -1;  int req_byte = -1;
5853  int req_byte2 = -1;  int req_byte2 = -1;
5854  int newline;  int newline;
 unsigned long int ims;  
5855  BOOL using_temporary_offsets = FALSE;  BOOL using_temporary_offsets = FALSE;
5856  BOOL anchored;  BOOL anchored;
5857  BOOL startline;  BOOL startline;
# Line 5600  if ((options & ~PUBLIC_EXEC_OPTIONS) != Line 5881  if ((options & ~PUBLIC_EXEC_OPTIONS) !=
5881  if (re == NULL || subject == NULL ||  if (re == NULL || subject == NULL ||
5882     (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL;     (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL;
5883  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
5884    if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
5885    
5886    /* These two settings are used in the code for checking a UTF-8 string that
5887    follows immediately afterwards. Other values in the md block are used only
5888    during "normal" pcre_exec() processing, not when the JIT support is in use,
5889    so they are set up later. */
5890    
5891    utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
5892    md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5893                  ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5894    
5895    /* Check a UTF-8 string if required. Pass back the character offset and error
5896    code for an invalid string if a results vector is available. */
5897    
5898    #ifdef SUPPORT_UTF8
5899    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5900      {
5901      int erroroffset;
5902      int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);
5903      if (errorcode != 0)
5904        {
5905        if (offsetcount >= 2)
5906          {
5907          offsets[0] = erroroffset;
5908          offsets[1] = errorcode;
5909          }
5910        return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
5911          PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
5912        }
5913    
5914      /* Check that a start_offset points to the start of a UTF-8 character. */
5915      if (start_offset > 0 && start_offset < length &&
5916          (((USPTR)subject)[start_offset] & 0xc0) == 0x80)
5917        return PCRE_ERROR_BADUTF8_OFFSET;
5918      }
5919    #endif
5920    
5921    /* If the pattern was successfully studied with JIT support, run the JIT
5922    executable instead of the rest of this function. Most options must be set at
5923    compile time for the JIT code to be usable. Fallback to the normal code path if
5924    an unsupported flag is set. In particular, JIT does not support partial
5925    matching. */
5926    
5927    #ifdef SUPPORT_JIT
5928    if (extra_data != NULL
5929        && (extra_data->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0
5930        && extra_data->executable_jit != NULL
5931        && (options & ~(PCRE_NO_UTF8_CHECK | PCRE_NOTBOL | PCRE_NOTEOL |
5932                        PCRE_NOTEMPTY | PCRE_NOTEMPTY_ATSTART)) == 0)
5933      return _pcre_jit_exec(re, extra_data->executable_jit, subject, length,
5934        start_offset, options, ((extra_data->flags & PCRE_EXTRA_MATCH_LIMIT) == 0)
5935        ? MATCH_LIMIT : extra_data->match_limit, offsets, offsetcount);
5936    #endif
5937    
5938  /* This information is for finding all the numbers associated with a given  /* Carry on with non-JIT matching. This information is for finding all the
5939  name, for condition testing. */  numbers associated with a given name, for condition testing. */
5940    
5941  md->name_table = (uschar *)re + re->name_table_offset;  md->name_table = (uschar *)re + re->name_table_offset;
5942  md->name_count = re->name_count;  md->name_count = re->name_count;
# Line 5669  md->end_subject = md->start_subject + le Line 6003  md->end_subject = md->start_subject + le
6003  end_subject = md->end_subject;  end_subject = md->end_subject;
6004    
6005  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
 utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;  
6006  md->use_ucp = (re->options & PCRE_UCP) != 0;  md->use_ucp = (re->options & PCRE_UCP) != 0;
6007  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
6008    
6009    /* Some options are unpacked into BOOL variables in the hope that testing
6010    them will be faster than individual option bits. */
6011    
6012  md->notbol = (options & PCRE_NOTBOL) != 0;  md->notbol = (options & PCRE_NOTBOL) != 0;
6013  md->noteol = (options & PCRE_NOTEOL) != 0;  md->noteol = (options & PCRE_NOTEOL) != 0;
6014  md->notempty = (options & PCRE_NOTEMPTY) != 0;  md->notempty = (options & PCRE_NOTEMPTY) != 0;
6015  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
6016  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :  
               ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;  
6017  md->hitend = FALSE;  md->hitend = FALSE;
6018  md->mark = NULL;                        /* In case never set */  md->mark = NULL;                        /* In case never set */
6019    
6020  md->recursive = NULL;                   /* No recursion at top level */  md->recursive = NULL;                   /* No recursion at top level */
6021    md->hasthen = (re->flags & PCRE_HASTHEN) != 0;
6022    
6023  md->lcc = tables + lcc_offset;  md->lcc = tables + lcc_offset;
6024  md->ctypes = tables + ctypes_offset;  md->ctypes = tables + ctypes_offset;
# Line 5760  defined (though never set). So there's n Line 6096  defined (though never set). So there's n
6096  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6097    return PCRE_ERROR_BADPARTIAL;    return PCRE_ERROR_BADPARTIAL;
6098    
 /* 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);  
   
6099  /* 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
6100  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.
6101  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
6102  of 3. */  of 3. */
6103    
6104  ocount = offsetcount - (offsetcount % 3);  ocount = offsetcount - (offsetcount % 3);
6105    arg_offset_max = (2*ocount)/3;
6106    
6107  if (re->top_backref > 0 && re->top_backref >= ocount/3)  if (re->top_backref > 0 && re->top_backref >= ocount/3)
6108    {    {
# Line 5808  md->offset_max = (2*ocount)/3; Line 6119  md->offset_max = (2*ocount)/3;
6119  md->offset_overflow = FALSE;  md->offset_overflow = FALSE;
6120  md->capture_last = -1;  md->capture_last = -1;
6121    
 /* 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;  
   
6122  /* Reset the working variable associated with each extraction. These should  /* Reset the working variable associated with each extraction. These should
6123  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
6124  initialize them to avoid reading uninitialized locations. */  initialize them to avoid reading uninitialized locations. Also, unset the
6125    offsets for the matched string. This is really just for tidiness with callouts,
6126    in case they inspect these fields. */
6127    
6128  if (md->offset_vector != NULL)  if (md->offset_vector != NULL)
6129    {    {
6130    register int *iptr = md->offset_vector + ocount;    register int *iptr = md->offset_vector + ocount;
6131    register int *iend = iptr - resetcount/2 + 1;    register int *iend = iptr - re->top_bracket;
6132      if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6133    while (--iptr >= iend) *iptr = -1;    while (--iptr >= iend) *iptr = -1;
6134      md->offset_vector[0] = md->offset_vector[1] = -1;
6135    }    }
6136    
6137  /* 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 5857  if ((re->flags & PCRE_REQCHSET) != 0) Line 6165  if ((re->flags & PCRE_REQCHSET) != 0)
6165    }    }
6166    
6167    
6168    
6169    
6170  /* ==========================================================================*/  /* ==========================================================================*/
6171    
6172  /* Loop for handling unanchored repeated matching attempts; for anchored regexs  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
# Line 5867  for(;;) Line 6177  for(;;)
6177    USPTR save_end_subject = end_subject;    USPTR save_end_subject = end_subject;
6178    USPTR new_start_match;    USPTR new_start_match;
6179    
   /* 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;  
     }  
   
6180    /* 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
6181    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
6182    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 5903  for(;;) Line 6204  for(;;)
6204    /* There are some optimizations that avoid running the match if a known    /* There are some optimizations that avoid running the match if a known
6205    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.
6206    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
6207    that all callouts do actually occur. */    that all callouts do actually occur. The option can be set in the regex by
6208      (*NO_START_OPT) or passed in match-time options. */
6209    
6210    if ((options & PCRE_NO_START_OPTIMIZE) == 0)    if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0)
6211      {      {
6212      /* Advance to a unique first byte if there is one. */      /* Advance to a unique first byte if there is one. */
6213    
# Line 5959  for(;;) Line 6261  for(;;)
6261        while (start_match < end_subject)        while (start_match < end_subject)
6262          {          {
6263          register unsigned int c = *start_match;          register unsigned int c = *start_match;
6264          if ((start_bits[c/8] & (1 << (c&7))) == 0) start_match++;          if ((start_bits[c/8] & (1 << (c&7))) == 0)
6265            else break;            {
6266              start_match++;
6267    #ifdef SUPPORT_UTF8
6268              if (utf8)
6269                while(start_match < end_subject && (*start_match & 0xc0) == 0x80)
6270                  start_match++;
6271    #endif
6272              }
6273            else break;
6274          }          }
6275        }        }
6276      }   /* Starting optimizations */      }   /* Starting optimizations */
# Line 6055  for(;;) Line 6365  for(;;)
6365    md->start_match_ptr = start_match;    md->start_match_ptr = start_match;
6366    md->start_used_ptr = start_match;    md->start_used_ptr = start_match;
6367    md->match_call_count = 0;    md->match_call_count = 0;
6368    rc = match(start_match, md->start_code, start_match, NULL, 2, md, ims, NULL,    md->match_function_type = 0;
6369      0, 0);    md->end_offset_top = 0;
6370      rc = match(start_match, md->start_code, start_match, NULL, 2, md, NULL, 0);
6371    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
6372    
6373    switch(rc)    switch(rc)
6374      {      {
6375      /* NOMATCH and PRUNE advance by one character. If MATCH_SKIP_ARG reaches      /* SKIP passes back the next starting point explicitly, but if it is the
6376      this level it means that a MARK that matched the SKIP's arg was not found.      same as the match we have just done, treat it as NOMATCH. */
6377      We treat this as NOMATCH. THEN at this level acts exactly like PRUNE. */  
6378        case MATCH_SKIP:
6379        if (md->start_match_ptr != start_match)
6380          {
6381          new_start_match = md->start_match_ptr;
6382          break;
6383          }
6384        /* Fall through */
6385    
6386        /* If MATCH_SKIP_ARG reaches this level it means that a MARK that matched
6387        the SKIP's arg was not found. We also treat this as NOMATCH. */
6388    
6389        case MATCH_SKIP_ARG:
6390        /* Fall through */
6391    
6392        /* NOMATCH and PRUNE advance by one character. THEN at this level acts
6393        exactly like PRUNE. */
6394    
6395      case MATCH_NOMATCH:      case MATCH_NOMATCH:
6396      case MATCH_PRUNE:      case MATCH_PRUNE:
     case MATCH_SKIP_ARG:  
6397      case MATCH_THEN:      case MATCH_THEN:
6398      new_start_match = start_match + 1;      new_start_match = start_match + 1;
6399  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 6077  for(;;) Line 6403  for(;;)
6403  #endif  #endif
6404      break;      break;
6405    
     /* SKIP passes back the next starting point explicitly. */  
   
     case MATCH_SKIP:  
     new_start_match = md->start_match_ptr;  
     break;  
   
6406      /* COMMIT disables the bumpalong, but otherwise behaves as NOMATCH. */      /* COMMIT disables the bumpalong, but otherwise behaves as NOMATCH. */
6407    
6408      case MATCH_COMMIT:      case MATCH_COMMIT:
# Line 6157  if (rc == MATCH_MATCH || rc == MATCH_ACC Line 6477  if (rc == MATCH_MATCH || rc == MATCH_ACC
6477    {    {
6478    if (using_temporary_offsets)    if (using_temporary_offsets)
6479      {      {
6480      if (offsetcount >= 4)      if (arg_offset_max >= 4)
6481        {        {
6482        memcpy(offsets + 2, md->offset_vector + 2,        memcpy(offsets + 2, md->offset_vector + 2,
6483          (offsetcount - 2) * sizeof(int));          (arg_offset_max - 2) * sizeof(int));
6484        DPRINTF(("Copied offsets from temporary memory\n"));        DPRINTF(("Copied offsets from temporary memory\n"));
6485        }        }
6486      if (md->end_offset_top > offsetcount) md->offset_overflow = TRUE;      if (md->end_offset_top > arg_offset_max) md->offset_overflow = TRUE;
6487      DPRINTF(("Freeing temporary memory\n"));      DPRINTF(("Freeing temporary memory\n"));
6488      (pcre_free)(md->offset_vector);      (pcre_free)(md->offset_vector);
6489      }      }
6490    
6491    /* 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
6492    too many to fit into the vector. */    too many to fit into the vector. */
6493    
6494    rc = md->offset_overflow? 0 : md->end_offset_top/2;    rc = (md->offset_overflow && md->end_offset_top >= arg_offset_max)?
6495        0 : md->end_offset_top/2;
6496    
6497      /* If there is space in the offset vector, set any unused pairs at the end of
6498      the pattern to -1 for backwards compatibility. It is documented that this
6499      happens. In earlier versions, the whole set of potential capturing offsets
6500      was set to -1 each time round the loop, but this is handled differently now.
6501      "Gaps" are set to -1 dynamically instead (this fixes a bug). Thus, it is only
6502      those at the end that need unsetting here. We can't just unset them all at
6503      the start of the whole thing because they may get set in one branch that is
6504      not the final matching branch. */
6505    
6506      if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
6507        {
6508        register int *iptr, *iend;
6509        int resetcount = 2 + re->top_bracket * 2;
6510        if (resetcount > offsetcount) resetcount = ocount;
6511        iptr = offsets + md->end_offset_top;
6512        iend = offsets + resetcount;
6513        while (iptr < iend) *iptr++ = -1;
6514        }
6515    
6516    /* 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
6517    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.531  
changed lines
  Added in v.721

  ViewVC Help
Powered by ViewVC 1.1.5