/[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 569 by ph10, Sun Nov 7 16:14:50 2010 UTC revision 619 by ph10, Sun Jul 17 13:23:14 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, RM64, RM65, RM66 };
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 = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\
319    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
# Line 302  argument of match(), which never changes Line 323  argument of match(), which never changes
323    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
324    newframe->Xmarkptr = markptr;\    newframe->Xmarkptr = markptr;\
325    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
326    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
327    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
328    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
329    frame = newframe;\    frame = newframe;\
# Line 340  typedef struct heapframe { Line 359  typedef struct heapframe {
359    USPTR Xmstart;    USPTR Xmstart;
360    USPTR Xmarkptr;    USPTR Xmarkptr;
361    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
362    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
363    unsigned int Xrdepth;    unsigned int Xrdepth;
364    
365    /* Function local variables */    /* Function local variables */
# Line 363  typedef struct heapframe { Line 380  typedef struct heapframe {
380    BOOL Xcondition;    BOOL Xcondition;
381    BOOL Xprev_is_word;    BOOL Xprev_is_word;
382    
   unsigned long int Xoriginal_ims;  
   
383  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
384    int Xprop_type;    int Xprop_type;
385    int Xprop_value;    int Xprop_value;
# Line 450  Arguments: Line 465  Arguments:
465     markptr     pointer to the most recent MARK name, or NULL     markptr     pointer to the most recent MARK name, or NULL
466     offset_top  current top pointer     offset_top  current top pointer
467     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
468     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
469                   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  
470     rdepth      the recursion depth     rdepth      the recursion depth
471    
472  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
# Line 468  Returns: MATCH_MATCH if matched Line 478  Returns: MATCH_MATCH if matched
478    
479  static int  static int
480  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
481    const uschar *markptr, int offset_top, match_data *md, unsigned long int ims,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
482    eptrblock *eptrb, int flags, unsigned int rdepth)    unsigned int rdepth)
483  {  {
484  /* 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,
485  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 481  register unsigned int c; /* Character Line 491  register unsigned int c; /* Character
491  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
492    
493  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
494    BOOL caseless;
495  int condcode;  int condcode;
496    
497  /* 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 500  frame->Xecode = ecode; Line 511  frame->Xecode = ecode;
511  frame->Xmstart = mstart;  frame->Xmstart = mstart;
512  frame->Xmarkptr = markptr;  frame->Xmarkptr = markptr;
513  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
514  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
515  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
516    
517  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 516  HEAP_RECURSE: Line 525  HEAP_RECURSE:
525  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
526  #define markptr            frame->Xmarkptr  #define markptr            frame->Xmarkptr
527  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
528  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
529  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
530    
531  /* Ditto for the local variables */  /* Ditto for the local variables */
# Line 540  HEAP_RECURSE: Line 547  HEAP_RECURSE:
547  #define condition          frame->Xcondition  #define condition          frame->Xcondition
548  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
549    
 #define original_ims       frame->Xoriginal_ims  
   
550  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
551  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
552  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
# Line 578  i, and fc and c, can be the same variabl Line 583  i, and fc and c, can be the same variabl
583  #define fi i  #define fi i
584  #define fc c  #define fc c
585    
586    /* Many of the following variables are used only in small blocks of the code.
587  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  My normal style of coding would have declared them within each of those blocks.
588  const uschar *charptr;             /* in small blocks of the code. My normal */  However, in order to accommodate the version of this code that uses an external
589  #endif                             /* style of coding would have declared    */  "stack" implemented on the heap, it is easier to declare them all here, so the
590  const uschar *callpat;             /* them within each of those blocks.      */  declarations can be cut out in a block. The only declarations within blocks
591  const uschar *data;                /* However, in order to accommodate the   */  below are for variables that do not have to be preserved over a recursive call
592  const uschar *next;                /* version of this code that uses an      */  to RMATCH(). */
593  USPTR         pp;                  /* external "stack" implemented on the    */  
594  const uschar *prev;                /* heap, it is easier to declare them all */  #ifdef SUPPORT_UTF8
595  USPTR         saved_eptr;          /* here, so the declarations can be cut   */  const uschar *charptr;
596                                     /* out in a block. The only declarations  */  #endif
597  recursion_info new_recursive;      /* within blocks below are for variables  */  const uschar *callpat;
598                                     /* that do not have to be preserved over  */  const uschar *data;
599  BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  const uschar *next;
600    USPTR         pp;
601    const uschar *prev;
602    USPTR         saved_eptr;
603    
604    recursion_info new_recursive;
605    
606    BOOL cur_is_word;
607  BOOL condition;  BOOL condition;
608  BOOL prev_is_word;  BOOL prev_is_word;
609    
 unsigned long int original_ims;  
   
610  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
611  int prop_type;  int prop_type;
612  int prop_value;  int prop_value;
# Line 623  int stacksave[REC_STACK_SAVE_MAX]; Line 633  int stacksave[REC_STACK_SAVE_MAX];
633  eptrblock newptrb;  eptrblock newptrb;
634  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
635    
636    /* To save space on the stack and in the heap frame, I have doubled up on some
637    of the local variables that are used only in localised parts of the code, but
638    still need to be preserved over recursive calls of match(). These macros define
639    the alternative names that are used. */
640    
641    #define allow_zero    cur_is_word
642    #define cbegroup      condition
643    #define code_offset   codelink
644    #define condassert    condition
645    #define matched_once  prev_is_word
646    
647  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
648  variables. */  variables. */
649    
# Line 659  haven't exceeded the recursive call limi Line 680  haven't exceeded the recursive call limi
680  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
681  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
682    
 original_ims = ims;    /* Save for resetting on ')' */  
   
683  /* 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
684  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
685  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
686  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
687  When match() is called in other circumstances, don't add to the chain. The  
688  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
689  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
690  match(). */  to break infinite loops that match no characters. When match() is called in
691    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
692    NOT be used with tail recursion, because the memory block that is used is on
693    the stack, so a new one may be required for each match(). */
694    
695  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
696    {    {
697    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
698    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
699    eptrb = &newptrb;    eptrb = &newptrb;
700      md->match_function_type = 0;
701    }    }
702    
703  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 683  for (;;) Line 706  for (;;)
706    {    {
707    minimize = possessive = FALSE;    minimize = possessive = FALSE;
708    op = *ecode;    op = *ecode;
709    
710    switch(op)    switch(op)
711      {      {
712      case OP_MARK:      case OP_MARK:
713      markptr = ecode + 2;      markptr = ecode + 2;
714      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
715        ims, eptrb, flags, RM55);        eptrb, RM55);
716    
717      /* 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
718      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 715  for (;;) Line 738  for (;;)
738    
739      case OP_COMMIT:      case OP_COMMIT:
740      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
741        ims, eptrb, flags, RM52);        eptrb, RM52);
742      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
743          rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&          rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
744          rrc != MATCH_THEN)          rrc != MATCH_THEN)
# Line 726  for (;;) Line 749  for (;;)
749    
750      case OP_PRUNE:      case OP_PRUNE:
751      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
752        ims, eptrb, flags, RM51);        eptrb, RM51);
753      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
754      MRRETURN(MATCH_PRUNE);      MRRETURN(MATCH_PRUNE);
755    
756      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
757      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
758        ims, eptrb, flags, RM56);        eptrb, RM56);
759      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
760      md->mark = ecode + 2;      md->mark = ecode + 2;
761      RRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
# Line 741  for (;;) Line 764  for (;;)
764    
765      case OP_SKIP:      case OP_SKIP:
766      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
767        ims, eptrb, flags, RM53);        eptrb, RM53);
768      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
769        RRETURN(rrc);        RRETURN(rrc);
770      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
# Line 749  for (;;) Line 772  for (;;)
772    
773      case OP_SKIP_ARG:      case OP_SKIP_ARG:
774      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
775        ims, eptrb, flags, RM57);        eptrb, RM57);
776      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
777        RRETURN(rrc);        RRETURN(rrc);
778    
# Line 768  for (;;) Line 791  for (;;)
791    
792      case OP_THEN:      case OP_THEN:
793      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
794        ims, eptrb, flags, RM54);        eptrb, RM54);
795      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
796      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode - GET(ecode, 1);
797      MRRETURN(MATCH_THEN);      MRRETURN(MATCH_THEN);
798    
799      case OP_THEN_ARG:      case OP_THEN_ARG:
800      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],
801        offset_top, md, ims, eptrb, flags, RM58);        offset_top, md, eptrb, RM58);
802      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
803      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode - GET(ecode, 1);
804      md->mark = ecode + LINK_SIZE + 2;      md->mark = ecode + LINK_SIZE + 2;
805      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
806    
807      /* 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
808      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
809      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
810      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
811      reference inside the group.      previous iteration of this group, and be referred to by a reference inside
812        the group. A failure to match might occur after the group has succeeded,
813      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
814      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
815      of the same bracket.      were set by a previous iteration of the same bracket.
816    
817      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
818      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 822  for (;;)
822      case OP_SCBRA:      case OP_SCBRA:
823      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
824      offset = number << 1;      offset = number << 1;
825    
826  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
827      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
828      printf("subject=");      printf("subject=");
# Line 818  for (;;) Line 841  for (;;)
841        md->offset_vector[md->offset_end - number] =        md->offset_vector[md->offset_end - number] =
842          (int)(eptr - md->start_subject);          (int)(eptr - md->start_subject);
843    
844        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
845          {          {
846          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
847            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
848              eptrb, RM1);
849            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
850          if (rrc != MATCH_NOMATCH &&          if (rrc != MATCH_NOMATCH &&
851              (rrc != MATCH_THEN || md->start_match_ptr != ecode))              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
852            RRETURN(rrc);            RRETURN(rrc);
853          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
854          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
855            if (*ecode != OP_ALT) break;
856          }          }
       while (*ecode == OP_ALT);  
857    
858        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
859        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
860        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
861        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
862    
863          /* At this point, rrc will be one of MATCH_ONCE, MATCH_NOMATCH, or
864          MATCH_THEN. */
865    
866        if (rrc != MATCH_THEN) md->mark = markptr;        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
867        RRETURN(MATCH_NOMATCH);        RRETURN(((rrc == MATCH_ONCE)? MATCH_ONCE:MATCH_NOMATCH));
868        }        }
869    
870      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 852  for (;;) Line 878  for (;;)
878      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
879      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
880    
881      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
882      final alternative within the brackets, we would return the result of a      repeat. Loop for all the alternatives. When we get to the final alternative
883      recursive call to match() whatever happened. We can reduce stack usage by      within the brackets, we used to return the result of a recursive call to
884      turning this into a tail recursion, except in the case when match_cbegroup      match() whatever happened so it was possible to reduce stack usage by
885      is set.*/      turning this into a tail recursion, except in the case of a possibly empty
886        group. However, now that there is the possiblity of (*THEN) occurring in
887        the final alternative, this optimization is no longer possible.
888    
889        MATCH_ONCE is returned when the end of an atomic group is successfully
890        reached, but subsequent matching fails. It passes back up the tree (causing
891        captured values to be reset) until the original atomic group level is
892        reached. This is tested by comparing md->once_target with the start of the
893        group. At this point, the return is converted into MATCH_NOMATCH so that
894        previous backup points can be taken. */
895    
896        case OP_ONCE:
897      case OP_BRA:      case OP_BRA:
898      case OP_SBRA:      case OP_SBRA:
899      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
900      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
901      for (;;)      for (;;)
902        {        {
903        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
904          {        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
905          if (flags == 0)    /* Not a possibly empty group */          RM2);
906          if (rrc != MATCH_NOMATCH &&
907              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
908            {
909            if (rrc == MATCH_ONCE)
910            {            {
911            ecode += _pcre_OP_lengths[*ecode];            const uschar *scode = ecode;
912            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
913            goto TAIL_RECURSE;              {
914            }              while (*scode == OP_ALT) scode += GET(scode, 1);
915                scode -= GET(scode, 1);
916                }
917              if (md->once_target == scode) rrc = MATCH_NOMATCH;
918              }
919            RRETURN(rrc);
920            }
921          ecode += GET(ecode, 1);
922          if (*ecode != OP_ALT) break;
923          }
924        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
925        RRETURN(MATCH_NOMATCH);
926    
927          /* Possibly empty group; can't use tail recursion. */      /* Handle possessive capturing brackets with an unlimited repeat. We come
928        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
929        handled similarly to the normal case above. However, the matching is
930        different. The end of these brackets will always be OP_KETRPOS, which
931        returns MATCH_KETRPOS without going further in the pattern. By this means
932        we can handle the group by iteration rather than recursion, thereby
933        reducing the amount of stack needed. */
934    
935        case OP_CBRAPOS:
936        case OP_SCBRAPOS:
937        allow_zero = FALSE;
938    
939        POSSESSIVE_CAPTURE:
940        number = GET2(ecode, 1+LINK_SIZE);
941        offset = number << 1;
942    
943          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,  #ifdef PCRE_DEBUG
944            eptrb, flags, RM48);      printf("start possessive bracket %d\n", number);
945          if (rrc == MATCH_NOMATCH) md->mark = markptr;      printf("subject=");
946          RRETURN(rrc);      pchars(eptr, 16, TRUE, md);
947        printf("\n");
948    #endif
949    
950        if (offset < md->offset_max)
951          {
952          matched_once = FALSE;
953          code_offset = ecode - md->start_code;
954    
955          save_offset1 = md->offset_vector[offset];
956          save_offset2 = md->offset_vector[offset+1];
957          save_offset3 = md->offset_vector[md->offset_end - number];
958          save_capture_last = md->capture_last;
959    
960          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
961    
962          /* Each time round the loop, save the current subject position for use
963          when the group matches. For MATCH_MATCH, the group has matched, so we
964          restart it with a new subject starting position, remembering that we had
965          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
966          usual. If we haven't matched any alternatives in any iteration, check to
967          see if a previous iteration matched. If so, the group has matched;
968          continue from afterwards. Otherwise it has failed; restore the previous
969          capture values before returning NOMATCH. */
970    
971          for (;;)
972            {
973            md->offset_vector[md->offset_end - number] =
974              (int)(eptr - md->start_subject);
975            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
976            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
977              eptrb, RM63);
978            if (rrc == MATCH_KETRPOS)
979              {
980              offset_top = md->end_offset_top;
981              eptr = md->end_match_ptr;
982              ecode = md->start_code + code_offset;
983              save_capture_last = md->capture_last;
984              matched_once = TRUE;
985              continue;
986              }
987            if (rrc != MATCH_NOMATCH &&
988                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
989              RRETURN(rrc);
990            md->capture_last = save_capture_last;
991            ecode += GET(ecode, 1);
992            if (*ecode != OP_ALT) break;
993          }          }
994    
995        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (!matched_once)
996        otherwise return. */          {
997            md->offset_vector[offset] = save_offset1;
998            md->offset_vector[offset+1] = save_offset2;
999            md->offset_vector[md->offset_end - number] = save_offset3;
1000            }
1001    
1002          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
1003          if (allow_zero || matched_once)
1004            {
1005            ecode += 1 + LINK_SIZE;
1006            break;
1007            }
1008    
1009          RRETURN(MATCH_NOMATCH);
1010          }
1011    
1012        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1013        as a non-capturing bracket. */
1014    
1015        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1016        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1017    
1018        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1019    
1020        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1021        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1022    
1023        /* Non-capturing possessive bracket with unlimited repeat. We come here
1024        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1025        without the capturing complication. It is written out separately for speed
1026        and cleanliness. */
1027    
1028        case OP_BRAPOS:
1029        case OP_SBRAPOS:
1030        allow_zero = FALSE;
1031    
1032        POSSESSIVE_NON_CAPTURE:
1033        matched_once = FALSE;
1034        code_offset = ecode - md->start_code;
1035    
1036        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      for (;;)
1037          eptrb, flags, RM2);        {
1038          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1039          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1040            eptrb, RM48);
1041          if (rrc == MATCH_KETRPOS)
1042            {
1043            offset_top = md->end_offset_top;
1044            eptr = md->end_match_ptr;
1045            ecode = md->start_code + code_offset;
1046            matched_once = TRUE;
1047            continue;
1048            }
1049        if (rrc != MATCH_NOMATCH &&        if (rrc != MATCH_NOMATCH &&
1050            (rrc != MATCH_THEN || md->start_match_ptr != ecode))            (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1051          RRETURN(rrc);          RRETURN(rrc);
1052        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1053          if (*ecode != OP_ALT) break;
1054        }        }
1055    
1056        if (matched_once || allow_zero)
1057          {
1058          ecode += 1 + LINK_SIZE;
1059          break;
1060          }
1061        RRETURN(MATCH_NOMATCH);
1062    
1063      /* Control never reaches here. */      /* Control never reaches here. */
1064    
1065      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1066      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
1067      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
1068      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. */  
1069    
1070      case OP_COND:      case OP_COND:
1071      case OP_SCOND:      case OP_SCOND:
1072      codelink= GET(ecode, 1);      codelink = GET(ecode, 1);
1073    
1074      /* 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
1075      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
# Line 1076  for (;;) Line 1244  for (;;)
1244        }        }
1245    
1246      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1247      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
1248      assertion. */      an assertion. */
1249    
1250      else      else
1251        {        {
1252        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1253            match_condassert, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1254        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1255          {          {
1256            if (md->end_offset_top > offset_top)
1257              offset_top = md->end_offset_top;  /* Captures may have happened */
1258          condition = TRUE;          condition = TRUE;
1259          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1260          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
# Line 1102  for (;;) Line 1272  for (;;)
1272        }        }
1273    
1274      /* 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,
1275      we can use tail recursion to avoid using another stack frame, except when      we used to use tail recursion to avoid using another stack frame, except
1276      match_cbegroup is required for an unlimited repeat of a possibly empty      when there was unlimited repeat of a possibly empty group. However, that
1277      group. If the second alternative doesn't exist, we can just plough on. */      strategy no longer works because of the possibilty of (*THEN) being
1278        encountered in the branch. A recursive call to match() is always required,
1279        unless the second alternative doesn't exist, in which case we can just
1280        plough on. */
1281    
1282      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1283        {        {
1284        ecode += 1 + LINK_SIZE;        if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1285        if (op == OP_SCOND)        /* Possibly empty group */        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1286          {        if (rrc == MATCH_THEN && md->start_match_ptr == ecode)
1287          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);          rrc = MATCH_NOMATCH;
1288          RRETURN(rrc);        RRETURN(rrc);
         }  
       else                       /* Group must match something */  
         {  
         flags = 0;  
         goto TAIL_RECURSE;  
         }  
1289        }        }
1290      else                         /* Condition false & no alternative */      else                         /* Condition false & no alternative */
1291        {        {
# Line 1151  for (;;) Line 1318  for (;;)
1318      break;      break;
1319    
1320    
1321      /* 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. */  
1322    
     case OP_ACCEPT:  
1323      case OP_END:      case OP_END:
1324      if (md->recursive != NULL && md->recursive->group_num == 0)      case OP_ACCEPT:
1325        {      case OP_ASSERT_ACCEPT:
1326        recursion_info *rec = md->recursive;  
1327        DPRINTF(("End of pattern in a (?0) recursion\n"));      /* If we have matched an empty string, fail if not in an assertion and not
1328        md->recursive = rec->prevrec;      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1329        memmove(md->offset_vector, rec->offset_save,      is set and we have matched at the start of the subject. In both cases,
1330          rec->saved_max * sizeof(int));      backtracking will then try other alternatives, if any. */
1331        offset_top = rec->save_offset_top;  
1332        ims = original_ims;      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1333        ecode = rec->after_call;           md->recursive == NULL &&
1334        break;           (md->notempty ||
1335        }             (md->notempty_atstart &&
1336                 mstart == md->start_subject + md->start_offset)))
     /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is  
     set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of  
     the subject. In both cases, backtracking will then try other alternatives,  
     if any. */  
   
     if (eptr == mstart &&  
         (md->notempty ||  
           (md->notempty_atstart &&  
             mstart == md->start_subject + md->start_offset)))  
1337        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1338    
1339      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
1340    
1341      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1342      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1343      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
# Line 1193  for (;;) Line 1348  for (;;)
1348      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1349      MRRETURN(rrc);      MRRETURN(rrc);
1350    
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
   
1351      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1352      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,
1353      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
1354      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
1355      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1356        of a condition, we want to return immediately afterwards. The caller of
1357        this incarnation of the match() function will have set MATCH_CONDASSERT in
1358        md->match_function type, and one of these opcodes will be the first opcode
1359        that is processed. We use a local variable that is preserved over calls to
1360        match() to remember this case. */
1361    
1362      case OP_ASSERT:      case OP_ASSERT:
1363      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1364        if (md->match_function_type == MATCH_CONDASSERT)
1365          {
1366          condassert = TRUE;
1367          md->match_function_type = 0;
1368          }
1369        else condassert = FALSE;
1370    
1371      do      do
1372        {        {
1373        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
         RM4);  
1374        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1375          {          {
1376          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
# Line 1224  for (;;) Line 1382  for (;;)
1382        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1383        }        }
1384      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1385    
1386      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1387    
1388      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1389    
1390      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1391    
1392      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1393      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 1244  for (;;) Line 1403  for (;;)
1403    
1404      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1405      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1406        if (md->match_function_type == MATCH_CONDASSERT)
1407          {
1408          condassert = TRUE;
1409          md->match_function_type = 0;
1410          }
1411        else condassert = FALSE;
1412    
1413      do      do
1414        {        {
1415        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
         RM5);  
1416        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1417        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1418          {          {
# Line 1261  for (;;) Line 1426  for (;;)
1426        }        }
1427      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1428    
1429      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1430    
1431      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1432      continue;      continue;
1433    
# Line 1328  for (;;) Line 1493  for (;;)
1493      /* Recursion either matches the current regex, or some subexpression. The      /* Recursion either matches the current regex, or some subexpression. The
1494      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
1495      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1496    
1497      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1498      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
1499      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
1500      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
1501      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
1502      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
1503      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.  
1504    
1505      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
1506      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
1507      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1508        a lot, so he is not to blame for the current way it works. */
1509    
1510      case OP_RECURSE:      case OP_RECURSE:
1511        {        {
# Line 1355  for (;;) Line 1518  for (;;)
1518        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1519        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1520    
1521        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1522    
1523        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1524    
1525        /* Now save the offset data. */        /* Now save the offset data */
1526    
1527        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1528        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1371  for (;;) Line 1533  for (;;)
1533            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1534          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1535          }          }
   
1536        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1537              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1538        new_recursive.save_offset_top = offset_top;  
1539          /* OK, now we can do the recursion. After processing each alternative,
1540        /* OK, now we can do the recursion. For each top-level alternative we        restore the offset data. If there were nested recursions, md->recursive
1541        restore the offset and recursion data. */        might be changed, so reset it before looping. */
1542    
1543        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1544        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1545        do        do
1546          {          {
1547            if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1548          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1549            md, ims, eptrb, flags, RM6);            md, eptrb, RM6);
1550            memcpy(md->offset_vector, new_recursive.offset_save,
1551                new_recursive.saved_max * sizeof(int));
1552          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1553            {            {
1554            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
1555            md->recursive = new_recursive.prevrec;            md->recursive = new_recursive.prevrec;
1556            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1557              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1558            MRRETURN(MATCH_MATCH);  
1559              /* Set where we got to in the subject, and reset the start in case
1560              it was changed by \K. This *is* propagated back out of a recursion,
1561              for Perl compatibility. */
1562    
1563              eptr = md->end_match_ptr;
1564              mstart = md->start_match_ptr;
1565              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1566            }            }
1567          else if (rrc != MATCH_NOMATCH &&          else if (rrc != MATCH_NOMATCH &&
1568                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
# Line 1403  for (;;) Line 1574  for (;;)
1574            }            }
1575    
1576          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1577          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1578          }          }
1579        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1415  for (;;) Line 1584  for (;;)
1584          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1585        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1586        }        }
1587      /* Control never reaches here */  
1588        RECURSION_MATCHED:
1589      /* "Once" brackets are like assertion brackets except that after a match,      break;
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer, but resetting  
     the start-of-match value in case it was changed by \K. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */  
         {  
         mstart = md->start_match_ptr;  
         break;  
         }  
       if (rrc != MATCH_NOMATCH &&  
           (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
         RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* 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. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1590    
1591      /* 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
1592      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1510  for (;;) Line 1600  for (;;)
1600      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1601      with fixed upper repeat limits are compiled as a number of copies, with the      with fixed upper repeat limits are compiled as a number of copies, with the
1602      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1603    
1604      case OP_BRAZERO:      case OP_BRAZERO:
1605        {      next = ecode + 1;
1606        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1607        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1608        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1609        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1610      break;      break;
1611    
1612      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1613        {      next = ecode + 1;
1614        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1615        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1616        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1617        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1618      break;      break;
1619    
1620      case OP_SKIPZERO:      case OP_SKIPZERO:
1621        {      next = ecode+1;
1622        next = ecode+1;      do next += GET(next,1); while (*next == OP_ALT);
1623        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1624      break;      break;
1625    
1626        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1627        here; just jump to the group, with allow_zero set TRUE. */
1628    
1629        case OP_BRAPOSZERO:
1630        op = *(++ecode);
1631        allow_zero = TRUE;
1632        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1633          goto POSSESSIVE_NON_CAPTURE;
1634    
1635      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1636    
1637      case OP_KET:      case OP_KET:
1638      case OP_KETRMIN:      case OP_KETRMIN:
1639      case OP_KETRMAX:      case OP_KETRMAX:
1640        case OP_KETRPOS:
1641      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1642    
1643      /* 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
1644      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1645      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1646    
1647      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1648        {        {
1649        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1650        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1651        }        }
1652      else saved_eptr = NULL;      else saved_eptr = NULL;
1653    
1654      /* 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
1655      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
1656      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
1657      it was changed by \K. */      by \K. */
1658    
1659      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1660          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1661        {        {
1662        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1663        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
# Line 1574  for (;;) Line 1667  for (;;)
1667    
1668      /* 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
1669      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1670      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
1671      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
1672      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
1673        the current subject position and start match pointer and give a MATCH
1674        return. */
1675    
1676      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1677            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1678        {        {
1679        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1680        offset = number << 1;        offset = number << 1;
# Line 1588  for (;;) Line 1684  for (;;)
1684        printf("\n");        printf("\n");
1685  #endif  #endif
1686    
1687          /* Handle a recursively called group. */
1688    
1689          if (md->recursive != NULL && md->recursive->group_num == number)
1690            {
1691            md->end_match_ptr = eptr;
1692            md->start_match_ptr = mstart;
1693            RRETURN(MATCH_MATCH);
1694            }
1695    
1696          /* Deal with capturing */
1697    
1698        md->capture_last = number;        md->capture_last = number;
1699        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1700          {          {
1701            /* If offset is greater than offset_top, it means that we are
1702            "skipping" a capturing group, and that group's offsets must be marked
1703            unset. In earlier versions of PCRE, all the offsets were unset at the
1704            start of matching, but this doesn't work because atomic groups and
1705            assertions can cause a value to be set that should later be unset.
1706            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1707            part of the atomic group, but this is not on the final matching path,
1708            so must be unset when 2 is set. (If there is no group 2, there is no
1709            problem, because offset_top will then be 2, indicating no capture.) */
1710    
1711            if (offset > offset_top)
1712              {
1713              register int *iptr = md->offset_vector + offset_top;
1714              register int *iend = md->offset_vector + offset;
1715              while (iptr < iend) *iptr++ = -1;
1716              }
1717    
1718            /* Now make the extraction */
1719    
1720          md->offset_vector[offset] =          md->offset_vector[offset] =
1721            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1722          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1723          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1724          }          }
   
       /* Handle a recursively called group. Restore the offsets  
       appropriately and continue from after the call. */  
   
       if (md->recursive != NULL && md->recursive->group_num == number)  
         {  
         recursion_info *rec = md->recursive;  
         DPRINTF(("Recursion (%d) succeeded - continuing\n", number));  
         md->recursive = rec->prevrec;  
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         offset_top = rec->save_offset_top;  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
         }  
1725        }        }
1726    
1727      /* For both capturing and non-capturing groups, reset the value of the ims      /* For an ordinary non-repeating ket, just continue at this level. This
1728      flags, in case they got changed during the group. */      also happens for a repeating ket if no characters were matched in the
1729        group. This is the forcible breaking of infinite loops as implemented in
1730      ims = original_ims;      Perl 5.005. For a non-repeating atomic group, establish a backup point by
1731      DPRINTF(("ims reset to %02lx\n", ims));      processing the rest of the pattern at a lower level. If this results in a
1732        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1733      /* For a non-repeating ket, just continue at this level. This also      bypassing intermediate backup points, but resetting any captures that
1734      happens for a repeating ket if no characters were matched in the group.      happened along the way. */
     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. */  
1735    
1736      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1737        {        {
1738        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1739            {
1740            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1741            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1742            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1743            RRETURN(MATCH_ONCE);
1744            }
1745          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1746        break;        break;
1747        }        }
1748    
1749      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1750      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1751      tail recursion to avoid using another stack frame, unless we have an      at a time from the outer level, thus saving stack. */
1752      unlimited repeat of a group that can match an empty string. */  
1753        if (*ecode == OP_KETRPOS)
1754      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;        {
1755          md->end_match_ptr = eptr;
1756          md->end_offset_top = offset_top;
1757          RRETURN(MATCH_KETRPOS);
1758          }
1759    
1760        /* The normal repeating kets try the rest of the pattern or restart from
1761        the preceding bracket, in the appropriate order. In the second case, we can
1762        use tail recursion to avoid using another stack frame, unless we have an
1763        an atomic group or an unlimited repeat of a group that can match an empty
1764        string. */
1765    
1766      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1767        {        {
1768        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
1769        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1770        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
1771          {          {
1772          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
1773            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1774            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1775            RRETURN(MATCH_ONCE);
1776            }
1777          if (*prev >= OP_SBRA)    /* Could match an empty string */
1778            {
1779            md->match_function_type = MATCH_CBEGROUP;
1780            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1781          RRETURN(rrc);          RRETURN(rrc);
1782          }          }
1783        ecode = prev;        ecode = prev;
# Line 1653  for (;;) Line 1785  for (;;)
1785        }        }
1786      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1787        {        {
1788        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1789          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1790          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1791        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1792          if (*prev == OP_ONCE)
1793            {
1794            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
1795            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1796            md->once_target = prev;
1797            RRETURN(MATCH_ONCE);
1798            }
1799        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1800        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1801        }        }
1802      /* Control never gets here */      /* Control never gets here */
1803    
1804      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1805    
1806      case OP_CIRC:      case OP_CIRC:
1807      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1808      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 */  
   
1809      /* Start of subject assertion */      /* Start of subject assertion */
1810    
1811      case OP_SOD:      case OP_SOD:
1812      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1813      ecode++;      ecode++;
1814      break;      break;
1815    
1816        /* Multiline mode: start of subject unless notbol, or after any newline. */
1817    
1818        case OP_CIRCM:
1819        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1820        if (eptr != md->start_subject &&
1821            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1822          MRRETURN(MATCH_NOMATCH);
1823        ecode++;
1824        break;
1825    
1826      /* Start of match assertion */      /* Start of match assertion */
1827    
# Line 1696  for (;;) Line 1837  for (;;)
1837      ecode++;      ecode++;
1838      break;      break;
1839    
1840      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
1841      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
1842    
1843      case OP_DOLL:      case OP_DOLLM:
1844      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
1845        {        { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
1846        if (eptr < md->end_subject)      else
         { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }  
       else  
         {  
         if (md->noteol) MRRETURN(MATCH_NOMATCH);  
         SCHECK_PARTIAL();  
         }  
       ecode++;  
       break;  
       }  
     else  /* Not multiline */  
1847        {        {
1848        if (md->noteol) MRRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1849        if (!md->endonly) goto ASSERT_NL_OR_EOS;        SCHECK_PARTIAL();
1850        }        }
1851        ecode++;
1852        break;
1853    
1854        /* Not multiline mode: assert before a terminating newline or before end of
1855        subject unless noteol is set. */
1856    
1857        case OP_DOLL:
1858        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1859        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1860    
1861      /* ... else fall through for endonly */      /* ... else fall through for endonly */
1862    
1863      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
# Line 1735  for (;;) Line 1875  for (;;)
1875      if (eptr < md->end_subject &&      if (eptr < md->end_subject &&
1876          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1877        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1878    
1879      /* Either at end of string or \n before end. */      /* Either at end of string or \n before end. */
1880    
1881      SCHECK_PARTIAL();      SCHECK_PARTIAL();
1882      ecode++;      ecode++;
1883      break;      break;
# Line 2000  for (;;) Line 2140  for (;;)
2140      switch(c)      switch(c)
2141        {        {
2142        default: MRRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2143    
2144        case 0x000d:        case 0x000d:
2145        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2146        break;        break;
# Line 2252  for (;;) Line 2393  for (;;)
2393      loops). */      loops). */
2394    
2395      case OP_REF:      case OP_REF:
2396        {      case OP_REFI:
2397        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2398        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2399        ecode += 3;
2400    
2401        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2402    
2403        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2404        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
2405        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.  
2406    
2407        (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
2408        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2409    
2410        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
2411        referenced subpattern. */      referenced subpattern. */
2412    
2413        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2414          length = (md->jscript_compat)? 0 : (int)(md->end_subject - eptr + 1);        length = (md->jscript_compat)? 0 : -1;
2415        else      else
2416          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2417    
2418        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2419    
2420        switch (*ecode)      switch (*ecode)
2421          {        {
2422          case OP_CRSTAR:        case OP_CRSTAR:
2423          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2424          case OP_CRPLUS:        case OP_CRPLUS:
2425          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2426          case OP_CRQUERY:        case OP_CRQUERY:
2427          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2428          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2429          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2430          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2431          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2432          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2433          break;        break;
2434    
2435          case OP_CRRANGE:        case OP_CRRANGE:
2436          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2437          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2438          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2439          max = GET2(ecode, 3);        max = GET2(ecode, 3);
2440          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2441          ecode += 5;        ecode += 5;
2442          break;        break;
2443    
2444          default:               /* No repeat follows */        default:               /* No repeat follows */
2445          if (!match_ref(offset, eptr, length, md, ims))        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2446            {          {
2447            CHECK_PARTIAL();          CHECK_PARTIAL();
2448            MRRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
           }  
         eptr += length;  
         continue;              /* With the main loop */  
2449          }          }
2450          eptr += length;
2451          continue;              /* With the main loop */
2452          }
2453    
2454        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2455        main loop. */      zero, just continue with the main loop. */
2456    
2457        if (length == 0) continue;      if (length == 0) continue;
2458    
2459        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2460        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2461        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2462    
2463        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2464          {
2465          int slength;
2466          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2467          {          {
2468          if (!match_ref(offset, eptr, length, md, ims))          CHECK_PARTIAL();
2469            {          MRRETURN(MATCH_NOMATCH);
           CHECK_PARTIAL();  
           MRRETURN(MATCH_NOMATCH);  
           }  
         eptr += length;  
2470          }          }
2471          eptr += slength;
2472          }
2473    
2474        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2475        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2476    
2477        if (min == max) continue;      if (min == max) continue;
2478    
2479        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2480    
2481        if (minimize)      if (minimize)
2482          {
2483          for (fi = min;; fi++)
2484          {          {
2485          for (fi = min;; fi++)          int slength;
2486            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2487            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2488            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2489            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2490            {            {
2491            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            CHECK_PARTIAL();
2492            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;  
2493            }            }
2494          /* Control never gets here */          eptr += slength;
2495          }          }
2496          /* Control never gets here */
2497          }
2498    
2499        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2500    
2501        else      else
2502          {
2503          pp = eptr;
2504          for (i = min; i < max; i++)
2505          {          {
2506          pp = eptr;          int slength;
2507          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2508            {            {
2509            if (!match_ref(offset, eptr, length, md, ims))            CHECK_PARTIAL();
2510              {            break;
             CHECK_PARTIAL();  
             break;  
             }  
           eptr += length;  
           }  
         while (eptr >= pp)  
           {  
           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);  
           if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
           eptr -= length;  
2511            }            }
2512          MRRETURN(MATCH_NOMATCH);          eptr += slength;
2513            }
2514          while (eptr >= pp)
2515            {
2516            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2517            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2518            eptr -= length;
2519          }          }
2520          MRRETURN(MATCH_NOMATCH);
2521        }        }
2522      /* Control never gets here */      /* Control never gets here */
2523    
# Line 2480  for (;;) Line 2623  for (;;)
2623            {            {
2624            for (fi = min;; fi++)            for (fi = min;; fi++)
2625              {              {
2626              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2627              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2628              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2629              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2505  for (;;) Line 2648  for (;;)
2648            {            {
2649            for (fi = min;; fi++)            for (fi = min;; fi++)
2650              {              {
2651              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2652              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2653              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2654              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2551  for (;;) Line 2694  for (;;)
2694              }              }
2695            for (;;)            for (;;)
2696              {              {
2697              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2698              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2699              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2700              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2574  for (;;) Line 2717  for (;;)
2717              }              }
2718            while (eptr >= pp)            while (eptr >= pp)
2719              {              {
2720              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2721              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2722              eptr--;              eptr--;
2723              }              }
# Line 2650  for (;;) Line 2793  for (;;)
2793          {          {
2794          for (fi = min;; fi++)          for (fi = min;; fi++)
2795            {            {
2796            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2797            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2798            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2799            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 2683  for (;;) Line 2826  for (;;)
2826            }            }
2827          for(;;)          for(;;)
2828            {            {
2829            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2830            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2831            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2832            if (utf8) BACKCHAR(eptr);            if (utf8) BACKCHAR(eptr);
# Line 2728  for (;;) Line 2871  for (;;)
2871    
2872      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2873    
2874      case OP_CHARNC:      case OP_CHARI:
2875  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2876      if (utf8)      if (utf8)
2877        {        {
# Line 2788  for (;;) Line 2931  for (;;)
2931      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
2932    
2933      case OP_EXACT:      case OP_EXACT:
2934        case OP_EXACTI:
2935      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
2936      ecode += 3;      ecode += 3;
2937      goto REPEATCHAR;      goto REPEATCHAR;
2938    
2939      case OP_POSUPTO:      case OP_POSUPTO:
2940        case OP_POSUPTOI:
2941      possessive = TRUE;      possessive = TRUE;
2942      /* Fall through */      /* Fall through */
2943    
2944      case OP_UPTO:      case OP_UPTO:
2945        case OP_UPTOI:
2946      case OP_MINUPTO:      case OP_MINUPTO:
2947        case OP_MINUPTOI:
2948      min = 0;      min = 0;
2949      max = GET2(ecode, 1);      max = GET2(ecode, 1);
2950      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
2951      ecode += 3;      ecode += 3;
2952      goto REPEATCHAR;      goto REPEATCHAR;
2953    
2954      case OP_POSSTAR:      case OP_POSSTAR:
2955        case OP_POSSTARI:
2956      possessive = TRUE;      possessive = TRUE;
2957      min = 0;      min = 0;
2958      max = INT_MAX;      max = INT_MAX;
# Line 2812  for (;;) Line 2960  for (;;)
2960      goto REPEATCHAR;      goto REPEATCHAR;
2961    
2962      case OP_POSPLUS:      case OP_POSPLUS:
2963        case OP_POSPLUSI:
2964      possessive = TRUE;      possessive = TRUE;
2965      min = 1;      min = 1;
2966      max = INT_MAX;      max = INT_MAX;
# Line 2819  for (;;) Line 2968  for (;;)
2968      goto REPEATCHAR;      goto REPEATCHAR;
2969    
2970      case OP_POSQUERY:      case OP_POSQUERY:
2971        case OP_POSQUERYI:
2972      possessive = TRUE;      possessive = TRUE;
2973      min = 0;      min = 0;
2974      max = 1;      max = 1;
# Line 2826  for (;;) Line 2976  for (;;)
2976      goto REPEATCHAR;      goto REPEATCHAR;
2977    
2978      case OP_STAR:      case OP_STAR:
2979        case OP_STARI:
2980      case OP_MINSTAR:      case OP_MINSTAR:
2981        case OP_MINSTARI:
2982      case OP_PLUS:      case OP_PLUS:
2983        case OP_PLUSI:
2984      case OP_MINPLUS:      case OP_MINPLUS:
2985        case OP_MINPLUSI:
2986      case OP_QUERY:      case OP_QUERY:
2987        case OP_QUERYI:
2988      case OP_MINQUERY:      case OP_MINQUERY:
2989      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
2990        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
2991      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
   
2992      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
2993      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
2994      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
# Line 2856  for (;;) Line 3011  for (;;)
3011          {          {
3012  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3013          unsigned int othercase;          unsigned int othercase;
3014          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3015              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3016            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3017          else oclength = 0;          else oclength = 0;
# Line 2884  for (;;) Line 3039  for (;;)
3039            {            {
3040            for (fi = min;; fi++)            for (fi = min;; fi++)
3041              {              {
3042              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3043              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3044              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3045              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
# Line 2926  for (;;) Line 3081  for (;;)
3081    
3082            for(;;)            for(;;)
3083              {              {
3084              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3085              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3086              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3087  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2963  for (;;) Line 3118  for (;;)
3118      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3119        max, eptr));        max, eptr));
3120    
3121      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3122        {        {
3123        fc = md->lcc[fc];        fc = md->lcc[fc];
3124        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
# Line 2980  for (;;) Line 3135  for (;;)
3135          {          {
3136          for (fi = min;; fi++)          for (fi = min;; fi++)
3137            {            {
3138            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3139            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3140            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3141            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 3010  for (;;) Line 3165  for (;;)
3165    
3166          while (eptr >= pp)          while (eptr >= pp)
3167            {            {
3168            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3169            eptr--;            eptr--;
3170            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3171            }            }
# Line 3039  for (;;) Line 3194  for (;;)
3194          {          {
3195          for (fi = min;; fi++)          for (fi = min;; fi++)
3196            {            {
3197            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3198            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3199            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3200            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 3068  for (;;) Line 3223  for (;;)
3223    
3224          while (eptr >= pp)          while (eptr >= pp)
3225            {            {
3226            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3227            eptr--;            eptr--;
3228            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3229            }            }
# Line 3081  for (;;) Line 3236  for (;;)
3236      checking can be multibyte. */      checking can be multibyte. */
3237    
3238      case OP_NOT:      case OP_NOT:
3239        case OP_NOTI:
3240      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3241        {        {
3242        SCHECK_PARTIAL();        SCHECK_PARTIAL();
# Line 3088  for (;;) Line 3244  for (;;)
3244        }        }
3245      ecode++;      ecode++;
3246      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3247      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3248        {        {
3249  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3250        if (c < 256)        if (c < 256)
# Line 3096  for (;;) Line 3252  for (;;)
3252        c = md->lcc[c];        c = md->lcc[c];
3253        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3254        }        }
3255      else      else    /* Caseful */
3256        {        {
3257        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3258        }        }
# Line 3110  for (;;) Line 3266  for (;;)
3266      about... */      about... */
3267    
3268      case OP_NOTEXACT:      case OP_NOTEXACT:
3269        case OP_NOTEXACTI:
3270      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3271      ecode += 3;      ecode += 3;
3272      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3273    
3274      case OP_NOTUPTO:      case OP_NOTUPTO:
3275        case OP_NOTUPTOI:
3276      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3277        case OP_NOTMINUPTOI:
3278      min = 0;      min = 0;
3279      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3280      minimize = *ecode == OP_NOTMINUPTO;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3281      ecode += 3;      ecode += 3;
3282      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3283    
3284      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
3285        case OP_NOTPOSSTARI:
3286      possessive = TRUE;      possessive = TRUE;
3287      min = 0;      min = 0;
3288      max = INT_MAX;      max = INT_MAX;
# Line 3130  for (;;) Line 3290  for (;;)
3290      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3291    
3292      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
3293        case OP_NOTPOSPLUSI:
3294      possessive = TRUE;      possessive = TRUE;
3295      min = 1;      min = 1;
3296      max = INT_MAX;      max = INT_MAX;
# Line 3137  for (;;) Line 3298  for (;;)
3298      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3299    
3300      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
3301        case OP_NOTPOSQUERYI:
3302      possessive = TRUE;      possessive = TRUE;
3303      min = 0;      min = 0;
3304      max = 1;      max = 1;
# Line 3144  for (;;) Line 3306  for (;;)
3306      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3307    
3308      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
3309        case OP_NOTPOSUPTOI:
3310      possessive = TRUE;      possessive = TRUE;
3311      min = 0;      min = 0;
3312      max = GET2(ecode, 1);      max = GET2(ecode, 1);
# Line 3151  for (;;) Line 3314  for (;;)
3314      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3315    
3316      case OP_NOTSTAR:      case OP_NOTSTAR:
3317        case OP_NOTSTARI:
3318      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3319        case OP_NOTMINSTARI:
3320      case OP_NOTPLUS:      case OP_NOTPLUS:
3321        case OP_NOTPLUSI:
3322      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3323        case OP_NOTMINPLUSI:
3324      case OP_NOTQUERY:      case OP_NOTQUERY:
3325        case OP_NOTQUERYI:
3326      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3327      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3328        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3329      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3330      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3331      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
# Line 3178  for (;;) Line 3347  for (;;)
3347      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,
3348        max, eptr));        max, eptr));
3349    
3350      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3351        {        {
3352        fc = md->lcc[fc];        fc = md->lcc[fc];
3353    
# Line 3226  for (;;) Line 3395  for (;;)
3395            register unsigned int d;            register unsigned int d;
3396            for (fi = min;; fi++)            for (fi = min;; fi++)
3397              {              {
3398              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM28);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3399              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3400              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3401              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3245  for (;;) Line 3414  for (;;)
3414            {            {
3415            for (fi = min;; fi++)            for (fi = min;; fi++)
3416              {              {
3417              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM29);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3418              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3419              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3420              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3286  for (;;) Line 3455  for (;;)
3455          if (possessive) continue;          if (possessive) continue;
3456          for(;;)          for(;;)
3457              {              {
3458              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM30);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3459              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3460              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3461              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 3309  for (;;) Line 3478  for (;;)
3478            if (possessive) continue;            if (possessive) continue;
3479            while (eptr >= pp)            while (eptr >= pp)
3480              {              {
3481              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM31);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3482              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3483              eptr--;              eptr--;
3484              }              }
# Line 3366  for (;;) Line 3535  for (;;)
3535            register unsigned int d;            register unsigned int d;
3536            for (fi = min;; fi++)            for (fi = min;; fi++)
3537              {              {
3538              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM32);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3539              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3540              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3541              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3384  for (;;) Line 3553  for (;;)
3553            {            {
3554            for (fi = min;; fi++)            for (fi = min;; fi++)
3555              {              {
3556              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM33);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
3557              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3558              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3559              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 3424  for (;;) Line 3593  for (;;)
3593            if (possessive) continue;            if (possessive) continue;
3594            for(;;)            for(;;)
3595              {              {
3596              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM34);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
3597              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3598              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3599              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 3447  for (;;) Line 3616  for (;;)
3616            if (possessive) continue;            if (possessive) continue;
3617            while (eptr >= pp)            while (eptr >= pp)
3618              {              {
3619              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM35);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
3620              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3621              eptr--;              eptr--;
3622              }              }
# Line 3774  for (;;) Line 3943  for (;;)
3943            switch(c)            switch(c)
3944              {              {
3945              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
3946    
3947              case 0x000d:              case 0x000d:
3948              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3949              break;              break;
# Line 4050  for (;;) Line 4220  for (;;)
4220            switch(*eptr++)            switch(*eptr++)
4221              {              {
4222              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4223    
4224              case 0x000d:              case 0x000d:
4225              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4226              break;              break;
4227    
4228              case 0x000a:              case 0x000a:
4229              break;              break;
4230    
# Line 4242  for (;;) Line 4414  for (;;)
4414            case PT_ANY:            case PT_ANY:
4415            for (fi = min;; fi++)            for (fi = min;; fi++)
4416              {              {
4417              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM36);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM36);
4418              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4419              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4420              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4258  for (;;) Line 4430  for (;;)
4430            case PT_LAMP:            case PT_LAMP:
4431            for (fi = min;; fi++)            for (fi = min;; fi++)
4432              {              {
4433              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM37);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4434              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4435              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4436              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4278  for (;;) Line 4450  for (;;)
4450            case PT_GC:            case PT_GC:
4451            for (fi = min;; fi++)            for (fi = min;; fi++)
4452              {              {
4453              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM38);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM38);
4454              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4455              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4456              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4296  for (;;) Line 4468  for (;;)
4468            case PT_PC:            case PT_PC:
4469            for (fi = min;; fi++)            for (fi = min;; fi++)
4470              {              {
4471              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM39);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM39);
4472              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4473              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4474              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4314  for (;;) Line 4486  for (;;)
4486            case PT_SC:            case PT_SC:
4487            for (fi = min;; fi++)            for (fi = min;; fi++)
4488              {              {
4489              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM40);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM40);
4490              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4491              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4492              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4332  for (;;) Line 4504  for (;;)
4504            case PT_ALNUM:            case PT_ALNUM:
4505            for (fi = min;; fi++)            for (fi = min;; fi++)
4506              {              {
4507              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM59);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4508              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4509              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4510              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4351  for (;;) Line 4523  for (;;)
4523            case PT_SPACE:    /* Perl space */            case PT_SPACE:    /* Perl space */
4524            for (fi = min;; fi++)            for (fi = min;; fi++)
4525              {              {
4526              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM60);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM60);
4527              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4528              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4529              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4371  for (;;) Line 4543  for (;;)
4543            case PT_PXSPACE:  /* POSIX space */            case PT_PXSPACE:  /* POSIX space */
4544            for (fi = min;; fi++)            for (fi = min;; fi++)
4545              {              {
4546              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM61);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM61);
4547              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4548              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4549              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4391  for (;;) Line 4563  for (;;)
4563            case PT_WORD:            case PT_WORD:
4564            for (fi = min;; fi++)            for (fi = min;; fi++)
4565              {              {
4566              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM62);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4567              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4568              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
4569              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 4423  for (;;) Line 4595  for (;;)
4595          {          {
4596          for (fi = min;; fi++)          for (fi = min;; fi++)
4597            {            {
4598            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM41);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM41);
4599            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4600            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4601            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4455  for (;;) Line 4627  for (;;)
4627          {          {
4628          for (fi = min;; fi++)          for (fi = min;; fi++)
4629            {            {
4630            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM42);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM42);
4631            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4632            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4633            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4618  for (;;) Line 4790  for (;;)
4790          {          {
4791          for (fi = min;; fi++)          for (fi = min;; fi++)
4792            {            {
4793            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM43);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM43);
4794            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4795            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
4796            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 4916  for (;;) Line 5088  for (;;)
5088          if (possessive) continue;          if (possessive) continue;
5089          for(;;)          for(;;)
5090            {            {
5091            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM44);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM44);
5092            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5093            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5094            if (utf8) BACKCHAR(eptr);            if (utf8) BACKCHAR(eptr);
# Line 4957  for (;;) Line 5129  for (;;)
5129    
5130          for(;;)          for(;;)
5131            {            {
5132            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM45);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM45);
5133            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5134            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5135            for (;;)                        /* Move back over one extended */            for (;;)                        /* Move back over one extended */
# Line 5241  for (;;) Line 5413  for (;;)
5413            RRETURN(PCRE_ERROR_INTERNAL);            RRETURN(PCRE_ERROR_INTERNAL);
5414            }            }
5415    
5416          /* eptr is now past the end of the maximum run */          /* eptr is now past the end of the maximum run. If possessive, we are
5417            done (no backing up). Otherwise, match at this position; anything other
5418            than no match is immediately returned. For nomatch, back up one
5419            character, unless we are matching \R and the last thing matched was
5420            \r\n, in which case, back up two bytes. */
5421    
5422          if (possessive) continue;          if (possessive) continue;
5423          for(;;)          for(;;)
5424            {            {
5425            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM46);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM46);
5426            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5427            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5428            BACKCHAR(eptr);            BACKCHAR(eptr);
5429              if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5430                  eptr[-1] == '\r') eptr--;
5431            }            }
5432          }          }
5433        else        else
# Line 5448  for (;;) Line 5626  for (;;)
5626            RRETURN(PCRE_ERROR_INTERNAL);            RRETURN(PCRE_ERROR_INTERNAL);
5627            }            }
5628    
5629          /* eptr is now past the end of the maximum run */          /* eptr is now past the end of the maximum run. If possessive, we are
5630            done (no backing up). Otherwise, match at this position; anything other
5631            than no match is immediately returned. For nomatch, back up one
5632            character (byte), unless we are matching \R and the last thing matched
5633            was \r\n, in which case, back up two bytes. */
5634    
5635          if (possessive) continue;          if (possessive) continue;
5636          while (eptr >= pp)          while (eptr >= pp)
5637            {            {
5638            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM47);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
           eptr--;  
5639            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5640              eptr--;
5641              if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5642                  eptr[-1] == '\r') eptr--;
5643            }            }
5644          }          }
5645    
# Line 5494  switch (frame->Xwhere) Line 5678  switch (frame->Xwhere)
5678    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)
5679    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)
5680    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)
5681    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) LBL(64)
5682      LBL(65) LBL(66)
5683  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5684    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)
5685    LBL(32) LBL(34) LBL(42) LBL(46)    LBL(32) LBL(34) LBL(42) LBL(46)
# Line 5523  Undefine all the macros that were define Line 5708  Undefine all the macros that were define
5708  #undef ecode  #undef ecode
5709  #undef mstart  #undef mstart
5710  #undef offset_top  #undef offset_top
 #undef ims  
5711  #undef eptrb  #undef eptrb
5712  #undef flags  #undef flags
5713    
# Line 5541  Undefine all the macros that were define Line 5725  Undefine all the macros that were define
5725  #undef condition  #undef condition
5726  #undef prev_is_word  #undef prev_is_word
5727    
 #undef original_ims  
   
5728  #undef ctype  #undef ctype
5729  #undef length  #undef length
5730  #undef max  #undef max
# Line 5599  pcre_exec(const pcre *argument_re, const Line 5781  pcre_exec(const pcre *argument_re, const
5781    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5782    int offsetcount)    int offsetcount)
5783  {  {
5784  int rc, resetcount, ocount;  int rc, ocount;
5785  int first_byte = -1;  int first_byte = -1;
5786  int req_byte = -1;  int req_byte = -1;
5787  int req_byte2 = -1;  int req_byte2 = -1;
5788  int newline;  int newline;
 unsigned long int ims;  
5789  BOOL using_temporary_offsets = FALSE;  BOOL using_temporary_offsets = FALSE;
5790  BOOL anchored;  BOOL anchored;
5791  BOOL startline;  BOOL startline;
# Line 5708  utf8 = md->utf8 = (re->options & PCRE_UT Line 5889  utf8 = md->utf8 = (re->options & PCRE_UT
5889  md->use_ucp = (re->options & PCRE_UCP) != 0;  md->use_ucp = (re->options & PCRE_UCP) != 0;
5890  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
5891    
5892    /* Some options are unpacked into BOOL variables in the hope that testing
5893    them will be faster than individual option bits. */
5894    
5895  md->notbol = (options & PCRE_NOTBOL) != 0;  md->notbol = (options & PCRE_NOTBOL) != 0;
5896  md->noteol = (options & PCRE_NOTEOL) != 0;  md->noteol = (options & PCRE_NOTEOL) != 0;
5897  md->notempty = (options & PCRE_NOTEMPTY) != 0;  md->notempty = (options & PCRE_NOTEMPTY) != 0;
5898  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
5899  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5900                ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;                ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5901    
5902    
5903  md->hitend = FALSE;  md->hitend = FALSE;
5904  md->mark = NULL;                        /* In case never set */  md->mark = NULL;                        /* In case never set */
5905    
# Line 5795  defined (though never set). So there's n Line 5981  defined (though never set). So there's n
5981  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
5982    return PCRE_ERROR_BADPARTIAL;    return PCRE_ERROR_BADPARTIAL;
5983    
5984  /* Check a UTF-8 string if required. Unfortunately there's no way of passing  /* Check a UTF-8 string if required. Pass back the character offset and error
5985  back the character offset. */  code for an invalid string if a results vector is available. */
5986    
5987  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5988  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5989    {    {
5990    int tb;    int erroroffset;
5991    if ((tb = _pcre_valid_utf8((USPTR)subject, length)) >= 0)    int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);
5992      return (tb == length && md->partial > 1)?    if (errorcode != 0)
       PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;  
   if (start_offset > 0 && start_offset < length)  
5993      {      {
5994      tb = ((USPTR)subject)[start_offset] & 0xc0;      if (offsetcount >= 2)
5995      if (tb == 0x80) return PCRE_ERROR_BADUTF8_OFFSET;        {
5996      }        offsets[0] = erroroffset;
5997          offsets[1] = errorcode;
5998          }
5999        return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
6000          PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
6001        }
6002    
6003      /* Check that a start_offset points to the start of a UTF-8 character. */
6004    
6005      if (start_offset > 0 && start_offset < length &&
6006          (((USPTR)subject)[start_offset] & 0xc0) == 0x80)
6007        return PCRE_ERROR_BADUTF8_OFFSET;
6008    }    }
6009  #endif  #endif
6010    
 /* 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);  
   
6011  /* 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
6012  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.
6013  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
# Line 5841  md->offset_max = (2*ocount)/3; Line 6030  md->offset_max = (2*ocount)/3;
6030  md->offset_overflow = FALSE;  md->offset_overflow = FALSE;
6031  md->capture_last = -1;  md->capture_last = -1;
6032    
 /* 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;  
   
6033  /* Reset the working variable associated with each extraction. These should  /* Reset the working variable associated with each extraction. These should
6034  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
6035  initialize them to avoid reading uninitialized locations. */  initialize them to avoid reading uninitialized locations. Also, unset the
6036    offsets for the matched string. This is really just for tidiness with callouts,
6037    in case they inspect these fields. */
6038    
6039  if (md->offset_vector != NULL)  if (md->offset_vector != NULL)
6040    {    {
6041    register int *iptr = md->offset_vector + ocount;    register int *iptr = md->offset_vector + ocount;
6042    register int *iend = iptr - resetcount/2 + 1;    register int *iend = iptr - re->top_bracket;
6043      if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6044    while (--iptr >= iend) *iptr = -1;    while (--iptr >= iend) *iptr = -1;
6045      md->offset_vector[0] = md->offset_vector[1] = -1;
6046    }    }
6047    
6048  /* 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 5890  if ((re->flags & PCRE_REQCHSET) != 0) Line 6076  if ((re->flags & PCRE_REQCHSET) != 0)
6076    }    }
6077    
6078    
6079    
6080    
6081  /* ==========================================================================*/  /* ==========================================================================*/
6082    
6083  /* Loop for handling unanchored repeated matching attempts; for anchored regexs  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
# Line 5900  for(;;) Line 6088  for(;;)
6088    USPTR save_end_subject = end_subject;    USPTR save_end_subject = end_subject;
6089    USPTR new_start_match;    USPTR new_start_match;
6090    
   /* 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;  
     }  
   
6091    /* 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
6092    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
6093    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 5936  for(;;) Line 6115  for(;;)
6115    /* There are some optimizations that avoid running the match if a known    /* There are some optimizations that avoid running the match if a known
6116    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.
6117    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
6118    that all callouts do actually occur. */    that all callouts do actually occur. The option can be set in the regex by
6119      (*NO_START_OPT) or passed in match-time options. */
6120    
6121    if ((options & PCRE_NO_START_OPTIMIZE) == 0)    if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0)
6122      {      {
6123      /* Advance to a unique first byte if there is one. */      /* Advance to a unique first byte if there is one. */
6124    
# Line 6096  for(;;) Line 6276  for(;;)
6276    md->start_match_ptr = start_match;    md->start_match_ptr = start_match;
6277    md->start_used_ptr = start_match;    md->start_used_ptr = start_match;
6278    md->match_call_count = 0;    md->match_call_count = 0;
6279    rc = match(start_match, md->start_code, start_match, NULL, 2, md, ims, NULL,    md->match_function_type = 0;
6280      0, 0);    md->end_offset_top = 0;
6281      rc = match(start_match, md->start_code, start_match, NULL, 2, md, NULL, 0);
6282    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
6283    
6284    switch(rc)    switch(rc)
# Line 6220  if (rc == MATCH_MATCH || rc == MATCH_ACC Line 6401  if (rc == MATCH_MATCH || rc == MATCH_ACC
6401    
6402    /* 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 are
6403    too many to fit into the vector. */    too many to fit into the vector. */
6404    
6405    rc = md->offset_overflow? 0 : md->end_offset_top/2;    rc = md->offset_overflow? 0 : md->end_offset_top/2;
6406    
6407    /* 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

Legend:
Removed from v.569  
changed lines
  Added in v.619

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12