/[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 614 by ph10, Sat Jul 9 10:48:16 2011 UTC revision 667 by ph10, Mon Aug 22 14:57:32 2011 UTC
# 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  /* Values for setting in md->match_function_type to indicate two special types  /* 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,  of call to match(). We do it this way to save on using another stack variable,
62  as stack usage is to be discouraged. */  as stack usage is to be discouraged. */
63    
64  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
# Line 76  negative to avoid the external error cod Line 76  negative to avoid the external error cod
76  #define MATCH_ACCEPT       (-999)  #define MATCH_ACCEPT       (-999)
77  #define MATCH_COMMIT       (-998)  #define MATCH_COMMIT       (-998)
78  #define MATCH_KETRPOS      (-997)  #define MATCH_KETRPOS      (-997)
79  #define MATCH_PRUNE        (-996)  #define MATCH_ONCE         (-996)
80  #define MATCH_SKIP         (-995)  #define MATCH_PRUNE        (-995)
81  #define MATCH_SKIP_ARG     (-994)  #define MATCH_SKIP         (-994)
82  #define MATCH_THEN         (-993)  #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 137  while (length-- > 0) Line 138  while (length-- > 0)
138    
139  /* Normally, if a back reference hasn't been set, the length that is passed is  /* Normally, if a back reference hasn't been set, the length that is passed is
140  negative, so the match always fails. However, in JavaScript compatibility mode,  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  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.  subject bytes matched may be different to the number of reference bytes.
143    
144  Arguments:  Arguments:
# Line 184  if (caseless) Line 185  if (caseless)
185  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
186    if (md->utf8)    if (md->utf8)
187      {      {
188      /* Match characters up to the end of the reference. NOTE: the number of      /* Match characters up to the end of the reference. NOTE: the number of
189      bytes matched may differ, because there are some characters whose upper and      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      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);      (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      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      the latter. It is important, therefore, to check the length along the
194      reference, not along the subject (earlier code did this wrong). */      reference, not along the subject (earlier code did this wrong). */
195    
196      USPTR endptr = p + length;      USPTR endptr = p + length;
197      while (p < endptr)      while (p < endptr)
198        {        {
# Line 209  if (caseless) Line 210  if (caseless)
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      if (eptr + length > md->end_subject) return -1;      if (eptr + length > md->end_subject) return -1;
214      while (length-- > 0)      while (length-- > 0)
215        { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }        { 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    {    {
224    if (eptr + length > md->end_subject) return -1;    if (eptr + length > md->end_subject) return -1;
225    while (length-- > 0) if (*p++ != *eptr++) return -1;    while (length-- > 0) if (*p++ != *eptr++) return -1;
226    }    }
227    
228  return eptr - eptr_start;  return eptr - eptr_start;
# Line 276  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, RM63};         RM61,  RM62, RM63 };
281    
282  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
283  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
# Line 383  typedef struct heapframe { Line 384  typedef struct heapframe {
384    int Xprop_type;    int Xprop_type;
385    int Xprop_value;    int Xprop_value;
386    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
387    int Xoclength;    int Xoclength;
388    uschar Xocchars[8];    uschar Xocchars[8];
389  #endif  #endif
# Line 477  Returns:       MATCH_MATCH if matched Line 475  Returns:       MATCH_MATCH if matched
475    
476  static int  static int
477  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
478    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
479    unsigned int rdepth)    unsigned int rdepth)
480  {  {
481  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
# Line 550  HEAP_RECURSE: Line 548  HEAP_RECURSE:
548  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
549  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
550  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
551  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
552  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
553  #endif  #endif
# Line 590  declarations can be cut out in a block. Line 585  declarations can be cut out in a block.
585  below are for variables that do not have to be preserved over a recursive call  below are for variables that do not have to be preserved over a recursive call
586  to RMATCH(). */  to RMATCH(). */
587    
588  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
589  const uschar *charptr;  const uschar *charptr;
590  #endif  #endif
591  const uschar *callpat;  const uschar *callpat;
592  const uschar *data;  const uschar *data;
593  const uschar *next;  const uschar *next;
594  USPTR         pp;  USPTR         pp;
595  const uschar *prev;  const uschar *prev;
596  USPTR         saved_eptr;  USPTR         saved_eptr;
597    
598  recursion_info new_recursive;  recursion_info new_recursive;
599    
600  BOOL cur_is_word;  BOOL cur_is_word;
601  BOOL condition;  BOOL condition;
602  BOOL prev_is_word;  BOOL prev_is_word;
603    
# Line 610  BOOL prev_is_word; Line 605  BOOL prev_is_word;
605  int prop_type;  int prop_type;
606  int prop_value;  int prop_value;
607  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
608  int oclength;  int oclength;
609  uschar occhars[8];  uschar occhars[8];
610  #endif  #endif
# Line 632  int stacksave[REC_STACK_SAVE_MAX]; Line 624  int stacksave[REC_STACK_SAVE_MAX];
624  eptrblock newptrb;  eptrblock newptrb;
625  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
626    
627  /* To save space on the stack and in the heap frame, I have doubled up on some  /* To save space on the stack and in the heap frame, I have doubled up on some
628  of the local variables that are used only in localised parts of the code, but  of the local variables that are used only in localised parts of the code, but
629  still need to be preserved over recursive calls of match(). These macros define  still need to be preserved over recursive calls of match(). These macros define
630  the alternative names that are used. */  the alternative names that are used. */
631    
632  #define allow_zero    cur_is_word  #define allow_zero    cur_is_word
# Line 680  if (md->match_call_count++ >= md->match_ Line 672  if (md->match_call_count++ >= md->match_
672  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
673    
674  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
675  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
676  done this way to save having to use another function argument, which would take  done this way to save having to use another function argument, which would take
677  up space on the stack. See also MATCH_CONDASSERT below.  up space on the stack. See also MATCH_CONDASSERT below.
678    
679  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
# Line 705  for (;;) Line 697  for (;;)
697    {    {
698    minimize = possessive = FALSE;    minimize = possessive = FALSE;
699    op = *ecode;    op = *ecode;
700    
701    switch(op)    switch(op)
702      {      {
703      case OP_MARK:      case OP_MARK:
# Line 808  for (;;) Line 800  for (;;)
800      subject position in the working slot at the top of the vector. We mustn't      subject position in the working slot at the top of the vector. We mustn't
801      change the current values of the data slot, because they may be set from a      change the current values of the data slot, because they may be set from a
802      previous iteration of this group, and be referred to by a reference inside      previous iteration of this group, and be referred to by a reference inside
803      the group. If we fail to match, we need to restore this value and also the      the group. A failure to match might occur after the group has succeeded,
804      values of the final offsets, in case they were set by a previous iteration      if something later on doesn't match. For this reason, we need to restore
805      of the same bracket.      the working value and also the values of the final offsets, in case they
806        were set by a previous iteration of the same bracket.
807    
808      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
809      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 820  for (;;) Line 813  for (;;)
813      case OP_SCBRA:      case OP_SCBRA:
814      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
815      offset = number << 1;      offset = number << 1;
816    
817  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
818      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
819      printf("subject=");      printf("subject=");
# Line 841  for (;;) Line 834  for (;;)
834    
835        for (;;)        for (;;)
836          {          {
837          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
838          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
839            eptrb, RM1);            eptrb, RM1);
840            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
841          if (rrc != MATCH_NOMATCH &&          if (rrc != MATCH_NOMATCH &&
842              (rrc != MATCH_THEN || md->start_match_ptr != ecode))              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
843            RRETURN(rrc);            RRETURN(rrc);
   
         /* If md->end_offset_top is greater than offset_top, it means that the  
         branch we have just failed to match did manage to match some capturing  
         parentheses within an atomic group or an assertion. Although offset_top  
         reverts to its original value at this level, we must unset the captured  
         values in case a later match sets a higher capturing number. Example:  
         matching /((?>(a))b|(a)c)/ against "ac". This captures 3, but we need  
         to ensure that 2 - which was captured in the atomic matching - is  
         unset. */  
   
         if (md->end_offset_top > offset_top)  
           {  
           register int *iptr = md->offset_vector + offset_top;  
           register int *iend = md->offset_vector + md->end_offset_top;  
           while (iptr < iend) *iptr++ = -1;  
           }  
   
844          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
845          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
846          if (*ecode != OP_ALT) break;          if (*ecode != OP_ALT) break;
847          }          }
848    
849        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
850        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
851        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
852        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
853    
854          /* At this point, rrc will be one of MATCH_ONCE, MATCH_NOMATCH, or
855          MATCH_THEN. */
856    
857        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
858        RRETURN(MATCH_NOMATCH);        RRETURN(((rrc == MATCH_ONCE)? MATCH_ONCE:MATCH_NOMATCH));
859        }        }
860    
861      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 890  for (;;) Line 869  for (;;)
869      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
870      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
871    
872      /* Non-capturing bracket, except for possessive with unlimited repeat. Loop      /* Non-capturing or atomic group, except for possessive with unlimited
873      for all the alternatives. When we get to the final alternative within the      repeat. Loop for all the alternatives. When we get to the final alternative
874      brackets, we used to return the result of a recursive call to match()      within the brackets, we used to return the result of a recursive call to
875      whatever happened so it was possible to reduce stack usage by turning this      match() whatever happened so it was possible to reduce stack usage by
876      into a tail recursion, except in the case of a possibly empty group.      turning this into a tail recursion, except in the case of a possibly empty
877      However, now that there is the possiblity of (*THEN) occurring in the final      group. However, now that there is the possiblity of (*THEN) occurring in
878      alternative, this optimization is no longer possible. */      the final alternative, this optimization is no longer possible.
879    
880        MATCH_ONCE is returned when the end of an atomic group is successfully
881        reached, but subsequent matching fails. It passes back up the tree (causing
882        captured values to be reset) until the original atomic group level is
883        reached. This is tested by comparing md->once_target with the start of the
884        group. At this point, the return is converted into MATCH_NOMATCH so that
885        previous backup points can be taken. */
886    
887        case OP_ONCE:
888      case OP_BRA:      case OP_BRA:
889      case OP_SBRA:      case OP_SBRA:
890      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
891    
892      for (;;)      for (;;)
893        {        {
894        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
895        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
896          RM2);          RM2);
897        if (rrc != MATCH_NOMATCH &&        if (rrc != MATCH_NOMATCH &&
898            (rrc != MATCH_THEN || md->start_match_ptr != ecode))            (rrc != MATCH_THEN || md->start_match_ptr != ecode))
899            {
900            if (rrc == MATCH_ONCE)
901              {
902              const uschar *scode = ecode;
903              if (*scode != OP_ONCE)           /* If not at start, find it */
904                {
905                while (*scode == OP_ALT) scode += GET(scode, 1);
906                scode -= GET(scode, 1);
907                }
908              if (md->once_target == scode) rrc = MATCH_NOMATCH;
909              }
910          RRETURN(rrc);          RRETURN(rrc);
   
       /* See explanatory comment above under OP_CBRA. */  
   
       if (md->end_offset_top > offset_top)  
         {  
         register int *iptr = md->offset_vector + offset_top;  
         register int *iend = md->offset_vector + md->end_offset_top;  
         while (iptr < iend) *iptr++ = -1;  
911          }          }
   
912        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
913        if (*ecode != OP_ALT) break;        if (*ecode != OP_ALT) break;
914        }        }
   
915      if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;      if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
916      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
917    
918      /* Handle possessive capturing brackets with an unlimited repeat. We come      /* Handle possessive capturing brackets with an unlimited repeat. We come
919      here from BRAZERO with allow_zero set TRUE. The offset_vector values are      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
920      handled similarly to the normal case above. However, the matching is      handled similarly to the normal case above. However, the matching is
921      different. The end of these brackets will always be OP_KETRPOS, which      different. The end of these brackets will always be OP_KETRPOS, which
922      returns MATCH_KETRPOS without going further in the pattern. By this means      returns MATCH_KETRPOS without going further in the pattern. By this means
923      we can handle the group by iteration rather than recursion, thereby      we can handle the group by iteration rather than recursion, thereby
924      reducing the amount of stack needed. */      reducing the amount of stack needed. */
925    
926      case OP_CBRAPOS:      case OP_CBRAPOS:
927      case OP_SCBRAPOS:      case OP_SCBRAPOS:
928      allow_zero = FALSE;      allow_zero = FALSE;
929    
930      POSSESSIVE_CAPTURE:      POSSESSIVE_CAPTURE:
931      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
932      offset = number << 1;      offset = number << 1;
# Line 952  for (;;) Line 941  for (;;)
941      if (offset < md->offset_max)      if (offset < md->offset_max)
942        {        {
943        matched_once = FALSE;        matched_once = FALSE;
944        code_offset = ecode - md->start_code;        code_offset = ecode - md->start_code;
945    
946        save_offset1 = md->offset_vector[offset];        save_offset1 = md->offset_vector[offset];
947        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
# Line 960  for (;;) Line 949  for (;;)
949        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
950    
951        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
952    
953        /* Each time round the loop, save the current subject position for use        /* Each time round the loop, save the current subject position for use
954        when the group matches. For MATCH_MATCH, the group has matched, so we        when the group matches. For MATCH_MATCH, the group has matched, so we
955        restart it with a new subject starting position, remembering that we had        restart it with a new subject starting position, remembering that we had
956        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
957        usual. If we haven't matched any alternatives in any iteration, check to        usual. If we haven't matched any alternatives in any iteration, check to
958        see if a previous iteration matched. If so, the group has matched;        see if a previous iteration matched. If so, the group has matched;
959        continue from afterwards. Otherwise it has failed; restore the previous        continue from afterwards. Otherwise it has failed; restore the previous
960        capture values before returning NOMATCH. */        capture values before returning NOMATCH. */
961    
962        for (;;)        for (;;)
963          {          {
964          md->offset_vector[md->offset_end - number] =          md->offset_vector[md->offset_end - number] =
965            (int)(eptr - md->start_subject);            (int)(eptr - md->start_subject);
966          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
967          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
968            eptrb, RM63);            eptrb, RM63);
969          if (rrc == MATCH_KETRPOS)          if (rrc == MATCH_KETRPOS)
970            {            {
971            offset_top = md->end_offset_top;            offset_top = md->end_offset_top;
972            eptr = md->end_match_ptr;            eptr = md->end_match_ptr;
973            ecode = md->start_code + code_offset;            ecode = md->start_code + code_offset;
974            save_capture_last = md->capture_last;            save_capture_last = md->capture_last;
975            matched_once = TRUE;            matched_once = TRUE;
976            continue;            continue;
977            }            }
978          if (rrc != MATCH_NOMATCH &&          if (rrc != MATCH_NOMATCH &&
979              (rrc != MATCH_THEN || md->start_match_ptr != ecode))              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
980            RRETURN(rrc);            RRETURN(rrc);
   
         /* See explanatory comment above under OP_CBRA. */  
   
         if (md->end_offset_top > offset_top)  
           {  
           register int *iptr = md->offset_vector + offset_top;  
           register int *iend = md->offset_vector + md->end_offset_top;  
           while (iptr < iend) *iptr++ = -1;  
           }  
   
981          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
982          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
983          if (*ecode != OP_ALT) break;          if (*ecode != OP_ALT) break;
984          }          }
985    
986        if (!matched_once)        if (!matched_once)
987          {          {
988          md->offset_vector[offset] = save_offset1;          md->offset_vector[offset] = save_offset1;
989          md->offset_vector[offset+1] = save_offset2;          md->offset_vector[offset+1] = save_offset2;
990          md->offset_vector[md->offset_end - number] = save_offset3;          md->offset_vector[md->offset_end - number] = save_offset3;
991          }          }
992    
993        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
994        if (allow_zero || matched_once)        if (allow_zero || matched_once)
995          {          {
996          ecode += 1 + LINK_SIZE;          ecode += 1 + LINK_SIZE;
997          break;          break;
998          }          }
999    
1000        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1001        }        }
1002    
1003      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1004      as a non-capturing bracket. */      as a non-capturing bracket. */
1005    
# Line 1032  for (;;) Line 1011  for (;;)
1011      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1012      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1013    
1014      /* Non-capturing possessive bracket with unlimited repeat. We come here      /* Non-capturing possessive bracket with unlimited repeat. We come here
1015      from BRAZERO with allow_zero = TRUE. The code is similar to the above,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1016      without the capturing complication. It is written out separately for speed      without the capturing complication. It is written out separately for speed
1017      and cleanliness. */      and cleanliness. */
1018    
1019      case OP_BRAPOS:      case OP_BRAPOS:
1020      case OP_SBRAPOS:      case OP_SBRAPOS:
1021      allow_zero = FALSE;      allow_zero = FALSE;
1022    
1023      POSSESSIVE_NON_CAPTURE:      POSSESSIVE_NON_CAPTURE:
1024      matched_once = FALSE;      matched_once = FALSE;
1025      code_offset = ecode - md->start_code;      code_offset = ecode - md->start_code;
1026    
1027      for (;;)      for (;;)
1028        {        {
1029        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1030        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1031          eptrb, RM48);          eptrb, RM48);
1032        if (rrc == MATCH_KETRPOS)        if (rrc == MATCH_KETRPOS)
1033          {          {
1034          offset_top = md->end_offset_top;          offset_top = md->end_offset_top;
1035          eptr = md->end_match_ptr;          eptr = md->end_match_ptr;
1036          ecode = md->start_code + code_offset;          ecode = md->start_code + code_offset;
1037          matched_once = TRUE;          matched_once = TRUE;
1038          continue;          continue;
1039          }          }
1040        if (rrc != MATCH_NOMATCH &&        if (rrc != MATCH_NOMATCH &&
1041            (rrc != MATCH_THEN || md->start_match_ptr != ecode))            (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1042          RRETURN(rrc);          RRETURN(rrc);
   
       /* See explanatory comment above under OP_CBRA. */  
   
       if (md->end_offset_top > offset_top)  
         {  
         register int *iptr = md->offset_vector + offset_top;  
         register int *iend = md->offset_vector + md->end_offset_top;  
         while (iptr < iend) *iptr++ = -1;  
         }  
   
1043        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1044        if (*ecode != OP_ALT) break;        if (*ecode != OP_ALT) break;
1045        }        }
1046    
1047      if (matched_once || allow_zero)      if (matched_once || allow_zero)
1048        {        {
1049        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1050        break;        break;
1051        }        }
1052      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1053    
1054      /* Control never reaches here. */      /* Control never reaches here. */
# Line 1101  for (;;) Line 1070  for (;;)
1070        if (pcre_callout != NULL)        if (pcre_callout != NULL)
1071          {          {
1072          pcre_callout_block cb;          pcre_callout_block cb;
1073          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1074          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1075          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1076          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 1113  for (;;) Line 1082  for (;;)
1082          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1083          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1084          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1085            cb.mark             = markptr;
1086          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1087          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1088          }          }
# Line 1271  for (;;) Line 1241  for (;;)
1241    
1242      else      else
1243        {        {
1244        md->match_function_type = MATCH_CONDASSERT;        md->match_function_type = MATCH_CONDASSERT;
1245        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1246        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1247          {          {
1248            if (md->end_offset_top > offset_top)
1249              offset_top = md->end_offset_top;  /* Captures may have happened */
1250          condition = TRUE;          condition = TRUE;
1251          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1252          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
# Line 1293  for (;;) Line 1265  for (;;)
1265    
1266      /* 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,
1267      we used to use tail recursion to avoid using another stack frame, except      we used to use tail recursion to avoid using another stack frame, except
1268      when there was unlimited repeat of a possibly empty group. However, that      when there was unlimited repeat of a possibly empty group. However, that
1269      strategy no longer works because of the possibilty of (*THEN) being      strategy no longer works because of the possibilty of (*THEN) being
1270      encountered in the branch. A recursive call to match() is always required,      encountered in the branch. A recursive call to match() is always required,
1271      unless the second alternative doesn't exist, in which case we can just      unless the second alternative doesn't exist, in which case we can just
1272      plough on. */      plough on. */
1273    
1274      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1275        {        {
1276        if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;        if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1277        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1278        if (rrc == MATCH_THEN && md->start_match_ptr == ecode)        if (rrc == MATCH_THEN && md->start_match_ptr == ecode)
1279          rrc = MATCH_NOMATCH;          rrc = MATCH_NOMATCH;
1280        RRETURN(rrc);        RRETURN(rrc);
1281        }        }
# Line 1338  for (;;) Line 1310  for (;;)
1310      break;      break;
1311    
1312    
1313      /* End of the pattern, either real or forced. If we are in a recursion, we      /* End of the pattern, either real or forced. */
     should restore the offsets appropriately, and if it's a top-level  
     recursion, continue from after the call. */  
1314    
     case OP_ACCEPT:  
     case OP_ASSERT_ACCEPT:  
1315      case OP_END:      case OP_END:
1316      if (md->recursive != NULL)      case OP_ACCEPT:
1317        {      case OP_ASSERT_ACCEPT:
       recursion_info *rec = md->recursive;  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       offset_top = rec->save_offset_top;  
       if (rec->group_num == 0)  
         {  
         ecode = rec->after_call;  
         break;  
         }  
       }  
1318    
1319      /* Otherwise, if we have matched an empty string, fail if not in an      /* If we have matched an empty string, fail if not in an assertion and not
1320      assertion and if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1321      is set and we have matched at the start of the subject. In both cases,      is set and we have matched at the start of the subject. In both cases,
1322      backtracking will then try other alternatives, if any. */      backtracking will then try other alternatives, if any. */
1323    
1324      else if (eptr == mstart && op != OP_ASSERT_ACCEPT &&      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1325          (md->notempty ||           md->recursive == NULL &&
1326            (md->notempty_atstart &&           (md->notempty ||
1327              mstart == md->start_subject + md->start_offset)))             (md->notempty_atstart &&
1328                 mstart == md->start_subject + md->start_offset)))
1329        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1330    
1331      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
1332    
1333      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1334      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1335      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
# Line 1386  for (;;) Line 1344  for (;;)
1344      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,
1345      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
1346      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
1347      this level is identical to the lookahead case. When the assertion is part      this level is identical to the lookahead case. When the assertion is part
1348      of a condition, we want to return immediately afterwards. The caller of      of a condition, we want to return immediately afterwards. The caller of
1349      this incarnation of the match() function will have set MATCH_CONDASSERT in      this incarnation of the match() function will have set MATCH_CONDASSERT in
1350      md->match_function type, and one of these opcodes will be the first opcode      md->match_function type, and one of these opcodes will be the first opcode
1351      that is processed. We use a local variable that is preserved over calls to      that is processed. We use a local variable that is preserved over calls to
1352      match() to remember this case. */      match() to remember this case. */
1353    
1354      case OP_ASSERT:      case OP_ASSERT:
# Line 1400  for (;;) Line 1358  for (;;)
1358        condassert = TRUE;        condassert = TRUE;
1359        md->match_function_type = 0;        md->match_function_type = 0;
1360        }        }
1361      else condassert = FALSE;      else condassert = FALSE;
1362    
1363      do      do
1364        {        {
1365        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1366        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1367          {          {
1368          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
1369            markptr = md->mark;
1370          break;          break;
1371          }          }
1372        if (rrc != MATCH_NOMATCH &&        if (rrc != MATCH_NOMATCH &&
1373            (rrc != MATCH_THEN || md->start_match_ptr != ecode))            (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1374          RRETURN(rrc);          RRETURN(rrc);
   
       /* See explanatory comment above under OP_CBRA. */  
   
       if (md->end_offset_top > offset_top)  
         {  
         register int *iptr = md->offset_vector + offset_top;  
         register int *iend = md->offset_vector + md->end_offset_top;  
         while (iptr < iend) *iptr++ = -1;  
         }  
   
1375        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1376        }        }
1377      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1378    
1379      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1380    
1381      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
# Line 1452  for (;;) Line 1401  for (;;)
1401        condassert = TRUE;        condassert = TRUE;
1402        md->match_function_type = 0;        md->match_function_type = 0;
1403        }        }
1404      else condassert = FALSE;      else condassert = FALSE;
1405    
1406      do      do
1407        {        {
# Line 1471  for (;;) Line 1420  for (;;)
1420      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1421    
1422      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1423    
1424      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1425      continue;      continue;
1426    
# Line 1516  for (;;) Line 1465  for (;;)
1465      if (pcre_callout != NULL)      if (pcre_callout != NULL)
1466        {        {
1467        pcre_callout_block cb;        pcre_callout_block cb;
1468        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1469        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1470        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1471        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 1528  for (;;) Line 1477  for (;;)
1477        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1478        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1479        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1480          cb.mark             = markptr;
1481        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1482        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1483        }        }
# Line 1538  for (;;) Line 1488  for (;;)
1488      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
1489      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1490    
1491      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1492      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
1493      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
1494      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
1495      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
1496      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
1497      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.  
1498    
1499      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
1500      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
1501      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1502        a lot, so he is not to blame for the current way it works. */
1503    
1504      case OP_RECURSE:      case OP_RECURSE:
1505        {        {
1506          recursion_info *ri;
1507          int recno;
1508    
1509        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1510        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1511          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1512    
1513          /* Check for repeating a recursion without advancing the subject pointer.
1514          This should catch convoluted mutual recursions. (Some simple cases are
1515          caught at compile time.) */
1516    
1517          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1518            if (recno == ri->group_num && eptr == ri->subject_position)
1519              RRETURN(PCRE_ERROR_RECURSELOOP);
1520    
1521        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1522    
1523          new_recursive.group_num = recno;
1524          new_recursive.subject_position = eptr;
1525        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1526        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1527    
1528        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1529    
1530        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1531    
1532        /* Now save the offset data. */        /* Now save the offset data */
1533    
1534        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1535        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1580  for (;;) Line 1540  for (;;)
1540            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1541          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1542          }          }
   
1543        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1544              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1545        new_recursive.save_offset_top = offset_top;  
1546          /* OK, now we can do the recursion. After processing each alternative,
1547        /* OK, now we can do the recursion. For each top-level alternative we        restore the offset data. If there were nested recursions, md->recursive
1548        restore the offset and recursion data. */        might be changed, so reset it before looping. */
1549    
1550        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1551        cbegroup = (*callpat >= OP_SBRA);        cbegroup = (*callpat >= OP_SBRA);
# Line 1595  for (;;) Line 1554  for (;;)
1554          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1555          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1556            md, eptrb, RM6);            md, eptrb, RM6);
1557            memcpy(md->offset_vector, new_recursive.offset_save,
1558                new_recursive.saved_max * sizeof(int));
1559          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1560            {            {
1561            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
1562            md->recursive = new_recursive.prevrec;            md->recursive = new_recursive.prevrec;
1563            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1564              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1565            MRRETURN(MATCH_MATCH);  
1566              /* Set where we got to in the subject, and reset the start in case
1567              it was changed by \K. This *is* propagated back out of a recursion,
1568              for Perl compatibility. */
1569    
1570              eptr = md->end_match_ptr;
1571              mstart = md->start_match_ptr;
1572              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1573            }            }
1574          else if (rrc != MATCH_NOMATCH &&          else if (rrc != MATCH_NOMATCH &&
1575                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
# Line 1613  for (;;) Line 1581  for (;;)
1581            }            }
1582    
1583          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1584          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1585          }          }
1586        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1625  for (;;) Line 1591  for (;;)
1591          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1592        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1593        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer, but resetting  
     the start-of-match value in case it was changed by \K. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, 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);  
   
       /* See explanatory comment above under OP_CBRA. */  
   
       if (md->end_offset_top > offset_top)  
         {  
         register int *iptr = md->offset_vector + offset_top;  
         register int *iend = md->offset_vector + md->end_offset_top;  
         while (iptr < iend) *iptr++ = -1;  
         }  
   
       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 after the group, 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. */  
1594    
1595      if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1596        {      break;
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       md->match_function_type = MATCH_CBEGROUP;  
       RMATCH(eptr, prev, offset_top, md, eptrb, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1597    
1598      /* 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
1599      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1721  for (;;) Line 1607  for (;;)
1607      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
1608      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
1609      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1610    
1611      case OP_BRAZERO:      case OP_BRAZERO:
1612      next = ecode + 1;      next = ecode + 1;
1613      RMATCH(eptr, next, offset_top, md, eptrb, RM10);      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
# Line 1729  for (;;) Line 1615  for (;;)
1615      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
1616      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1617      break;      break;
1618    
1619      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1620      next = ecode + 1;      next = ecode + 1;
1621      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
# Line 1743  for (;;) Line 1629  for (;;)
1629      do next += GET(next,1); while (*next == OP_ALT);      do next += GET(next,1); while (*next == OP_ALT);
1630      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1631      break;      break;
1632    
1633      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1634      here; just jump to the group, with allow_zero set TRUE. */      here; just jump to the group, with allow_zero set TRUE. */
1635    
1636      case OP_BRAPOSZERO:      case OP_BRAPOSZERO:
1637      op = *(++ecode);      op = *(++ecode);
1638      allow_zero = TRUE;      allow_zero = TRUE;
1639      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1640        goto POSSESSIVE_NON_CAPTURE;        goto POSSESSIVE_NON_CAPTURE;
# Line 1758  for (;;) Line 1644  for (;;)
1644      case OP_KET:      case OP_KET:
1645      case OP_KETRMIN:      case OP_KETRMIN:
1646      case OP_KETRMAX:      case OP_KETRMAX:
1647      case OP_KETRPOS:      case OP_KETRPOS:
1648      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1649    
1650      /* 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
1651      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1652      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1653    
1654      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1655        {        {
1656        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1657        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1658        }        }
1659      else saved_eptr = NULL;      else saved_eptr = NULL;
1660    
1661      /* 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
1662      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
1663      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
1664      it was changed by \K. */      by \K. */
1665    
1666      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1667          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1668        {        {
1669        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1670        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1671        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1672        MRRETURN(MATCH_MATCH);        MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1673        }        }
1674    
1675      /* 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
1676      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1677      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
1678      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
1679      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
1680        the current subject position and start match pointer and give a MATCH
1681        return. */
1682    
1683      if (*prev == OP_CBRA || *prev == OP_SCBRA ||      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1684          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
# Line 1804  for (;;) Line 1691  for (;;)
1691        printf("\n");        printf("\n");
1692  #endif  #endif
1693    
1694          /* Handle a recursively called group. */
1695    
1696          if (md->recursive != NULL && md->recursive->group_num == number)
1697            {
1698            md->end_match_ptr = eptr;
1699            md->start_match_ptr = mstart;
1700            RRETURN(MATCH_MATCH);
1701            }
1702    
1703          /* Deal with capturing */
1704    
1705        md->capture_last = number;        md->capture_last = number;
1706        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1707          {          {
1708            /* If offset is greater than offset_top, it means that we are
1709            "skipping" a capturing group, and that group's offsets must be marked
1710            unset. In earlier versions of PCRE, all the offsets were unset at the
1711            start of matching, but this doesn't work because atomic groups and
1712            assertions can cause a value to be set that should later be unset.
1713            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1714            part of the atomic group, but this is not on the final matching path,
1715            so must be unset when 2 is set. (If there is no group 2, there is no
1716            problem, because offset_top will then be 2, indicating no capture.) */
1717    
1718            if (offset > offset_top)
1719              {
1720              register int *iptr = md->offset_vector + offset_top;
1721              register int *iend = md->offset_vector + offset;
1722              while (iptr < iend) *iptr++ = -1;
1723              }
1724    
1725            /* Now make the extraction */
1726    
1727          md->offset_vector[offset] =          md->offset_vector[offset] =
1728            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1729          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1730          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1731          }          }
   
       /* 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;  
         break;  
         }  
1732        }        }
1733    
1734      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1735      happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1736      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1737      5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group, establish a backup point by
1738      course of events. */      processing the rest of the pattern at a lower level. If this results in a
1739        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1740        bypassing intermediate backup points, but resetting any captures that
1741        happened along the way. */
1742    
1743      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1744        {        {
1745        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1746            {
1747            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1748            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1749            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1750            RRETURN(MATCH_ONCE);
1751            }
1752          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1753        break;        break;
1754        }        }
1755    
1756      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1757      and return the MATCH_KETRPOS. This makes it possible to do the repeats one      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1758      at a time from the outer level, thus saving stack. */      at a time from the outer level, thus saving stack. */
1759    
1760      if (*ecode == OP_KETRPOS)      if (*ecode == OP_KETRPOS)
1761        {        {
1762        md->end_match_ptr = eptr;        md->end_match_ptr = eptr;
1763        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1764        RRETURN(MATCH_KETRPOS);        RRETURN(MATCH_KETRPOS);
1765        }        }
1766    
1767      /* The normal repeating kets try the rest of the pattern or restart from      /* The normal repeating kets try the rest of the pattern or restart from
1768      the preceding bracket, in the appropriate order. In the second case, we can      the preceding bracket, in the appropriate order. In the second case, we can
1769      use tail recursion to avoid using another stack frame, unless we have an      use tail recursion to avoid using another stack frame, unless we have an
1770      unlimited repeat of a group that can match an empty string. */      an atomic group or an unlimited repeat of a group that can match an empty
1771        string. */
1772    
1773      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1774        {        {
1775        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1776        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1777          if (*prev == OP_ONCE)
1778            {
1779            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1780            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1781            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1782            RRETURN(MATCH_ONCE);
1783            }
1784        if (*prev >= OP_SBRA)    /* Could match an empty string */        if (*prev >= OP_SBRA)    /* Could match an empty string */
1785          {          {
1786          md->match_function_type = MATCH_CBEGROUP;          md->match_function_type = MATCH_CBEGROUP;
1787          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1788          RRETURN(rrc);          RRETURN(rrc);
1789          }          }
# Line 1872  for (;;) Line 1792  for (;;)
1792        }        }
1793      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1794        {        {
1795        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1796        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1797          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1798        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1799          if (*prev == OP_ONCE)
1800            {
1801            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1802            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1803            md->once_target = prev;
1804            RRETURN(MATCH_ONCE);
1805            }
1806        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1807        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1808        }        }
# Line 1884  for (;;) Line 1812  for (;;)
1812    
1813      case OP_CIRC:      case OP_CIRC:
1814      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1815    
1816      /* Start of subject assertion */      /* Start of subject assertion */
1817    
1818      case OP_SOD:      case OP_SOD:
1819      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1820      ecode++;      ecode++;
1821      break;      break;
1822    
1823      /* Multiline mode: start of subject unless notbol, or after any newline. */      /* Multiline mode: start of subject unless notbol, or after any newline. */
1824    
1825      case OP_CIRCM:      case OP_CIRCM:
# Line 1930  for (;;) Line 1858  for (;;)
1858      ecode++;      ecode++;
1859      break;      break;
1860    
1861      /* Not multiline mode: assert before a terminating newline or before end of      /* Not multiline mode: assert before a terminating newline or before end of
1862      subject unless noteol is set. */      subject unless noteol is set. */
1863    
1864      case OP_DOLL:      case OP_DOLL:
# Line 2086  for (;;) Line 2014  for (;;)
2014      /* Fall through */      /* Fall through */
2015    
2016      case OP_ALLANY:      case OP_ALLANY:
2017      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2018        {        {                            /* not be updated before SCHECK_PARTIAL. */
2019        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2020        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2021        }        }
2022        eptr++;
2023      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2024      ecode++;      ecode++;
2025      break;      break;
# Line 2099  for (;;) Line 2028  for (;;)
2028      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2029    
2030      case OP_ANYBYTE:      case OP_ANYBYTE:
2031      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2032        {        {                            /* not be updated before SCHECK_PARTIAL. */
2033        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2034        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2035        }        }
2036        eptr++;
2037      ecode++;      ecode++;
2038      break;      break;
2039    
# Line 2219  for (;;) Line 2149  for (;;)
2149      switch(c)      switch(c)
2150        {        {
2151        default: MRRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2152    
2153        case 0x000d:        case 0x000d:
2154        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2155        break;        break;
# Line 2443  for (;;) Line 2373  for (;;)
2373        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2374        }        }
2375      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2376        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2377        while (eptr < md->end_subject)
2378        {        {
2379        int category = UCD_CATEGORY(c);        int len = 1;
2380        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2381        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2382          {        eptr += len;
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2383        }        }
2384      ecode++;      ecode++;
2385      break;      break;
# Line 2472  for (;;) Line 2395  for (;;)
2395      loops). */      loops). */
2396    
2397      case OP_REF:      case OP_REF:
2398      case OP_REFI:      case OP_REFI:
2399      caseless = op == OP_REFI;      caseless = op == OP_REFI;
2400      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2401      ecode += 3;      ecode += 3;
2402    
# Line 2541  for (;;) Line 2464  for (;;)
2464    
2465      for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2466        {        {
2467        int slength;        int slength;
2468        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2469          {          {
2470          CHECK_PARTIAL();          CHECK_PARTIAL();
# Line 2561  for (;;) Line 2484  for (;;)
2484        {        {
2485        for (fi = min;; fi++)        for (fi = min;; fi++)
2486          {          {
2487          int slength;          int slength;
2488          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2489          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2490          if (fi >= max) MRRETURN(MATCH_NOMATCH);          if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 2582  for (;;) Line 2505  for (;;)
2505        pp = eptr;        pp = eptr;
2506        for (i = min; i < max; i++)        for (i = min; i < max; i++)
2507          {          {
2508          int slength;          int slength;
2509          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2510            {            {
2511            CHECK_PARTIAL();            CHECK_PARTIAL();
# Line 3315  for (;;) Line 3238  for (;;)
3238      checking can be multibyte. */      checking can be multibyte. */
3239    
3240      case OP_NOT:      case OP_NOT:
3241      case OP_NOTI:      case OP_NOTI:
3242      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3243        {        {
3244        SCHECK_PARTIAL();        SCHECK_PARTIAL();
# Line 3810  for (;;) Line 3733  for (;;)
3733            case PT_LAMP:            case PT_LAMP:
3734            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3735              {              {
3736                int chartype;
3737              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3738                {                {
3739                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3740                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3741                }                }
3742              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3743              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
3744              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
3745                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
3746                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
3747                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3748              }              }
3749            break;            break;
# Line 3833  for (;;) Line 3757  for (;;)
3757                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3758                }                }
3759              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3760              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
3761                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3762              }              }
3763            break;            break;
# Line 3848  for (;;) Line 3771  for (;;)
3771                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3772                }                }
3773              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3774              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
3775                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3776              }              }
3777            break;            break;
# Line 3863  for (;;) Line 3785  for (;;)
3785                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3786                }                }
3787              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3788              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
3789                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3790              }              }
3791            break;            break;
# Line 3872  for (;;) Line 3793  for (;;)
3793            case PT_ALNUM:            case PT_ALNUM:
3794            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3795              {              {
3796                int category;
3797              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3798                {                {
3799                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3800                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3801                }                }
3802              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3803              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3804              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
3805                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3806              }              }
3807            break;            break;
# Line 3894  for (;;) Line 3815  for (;;)
3815                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3816                }                }
3817              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3818              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
3819                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
3820                     == prop_fail_result)                     == prop_fail_result)
3821                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3911  for (;;) Line 3831  for (;;)
3831                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3832                }                }
3833              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3834              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
3835                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
3836                     == prop_fail_result)                     == prop_fail_result)
3837                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3922  for (;;) Line 3841  for (;;)
3841            case PT_WORD:            case PT_WORD:
3842            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3843              {              {
3844                int category;
3845              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3846                {                {
3847                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3848                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3849                }                }
3850              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3851              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3852              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
                  c == CHAR_UNDERSCORE)  
3853                     == prop_fail_result)                     == prop_fail_result)
3854                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3855              }              }
# Line 3956  for (;;) Line 3875  for (;;)
3875              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3876              }              }
3877            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3878            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
3879            while (eptr < md->end_subject)            while (eptr < md->end_subject)
3880              {              {
3881              int len = 1;              int len = 1;
3882              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
3883                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
3884              eptr += len;              eptr += len;
3885              }              }
3886            }            }
# Line 4022  for (;;) Line 3938  for (;;)
3938            switch(c)            switch(c)
3939              {              {
3940              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
3941    
3942              case 0x000d:              case 0x000d:
3943              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3944              break;              break;
# Line 4299  for (;;) Line 4215  for (;;)
4215            switch(*eptr++)            switch(*eptr++)
4216              {              {
4217              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4218    
4219              case 0x000d:              case 0x000d:
4220              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4221              break;              break;
4222    
4223              case 0x000a:              case 0x000a:
4224              break;              break;
4225    
# Line 4509  for (;;) Line 4425  for (;;)
4425            case PT_LAMP:            case PT_LAMP:
4426            for (fi = min;; fi++)            for (fi = min;; fi++)
4427              {              {
4428                int chartype;
4429              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4430              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4431              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4518  for (;;) Line 4435  for (;;)
4435                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4436                }                }
4437              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4438              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4439              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4440                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4441                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4442                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4443              }              }
4444            /* Control never gets here */            /* Control never gets here */
# Line 4538  for (;;) Line 4455  for (;;)
4455                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4456                }                }
4457              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4458              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
4459                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4460              }              }
4461            /* Control never gets here */            /* Control never gets here */
# Line 4556  for (;;) Line 4472  for (;;)
4472                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4473                }                }
4474              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4475              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
4476                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4477              }              }
4478            /* Control never gets here */            /* Control never gets here */
# Line 4574  for (;;) Line 4489  for (;;)
4489                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4490                }                }
4491              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4492              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
4493                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4494              }              }
4495            /* Control never gets here */            /* Control never gets here */
# Line 4583  for (;;) Line 4497  for (;;)
4497            case PT_ALNUM:            case PT_ALNUM:
4498            for (fi = min;; fi++)            for (fi = min;; fi++)
4499              {              {
4500                int category;
4501              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4502              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4503              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4592  for (;;) Line 4507  for (;;)
4507                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4508                }                }
4509              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4510              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4511              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
4512                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4513              }              }
4514            /* Control never gets here */            /* Control never gets here */
# Line 4611  for (;;) Line 4525  for (;;)
4525                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4526                }                }
4527              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4528              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
4529                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
4530                     == prop_fail_result)                     == prop_fail_result)
4531                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4631  for (;;) Line 4544  for (;;)
4544                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4545                }                }
4546              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4547              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
4548                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4549                     == prop_fail_result)                     == prop_fail_result)
4550                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4642  for (;;) Line 4554  for (;;)
4554            case PT_WORD:            case PT_WORD:
4555            for (fi = min;; fi++)            for (fi = min;; fi++)
4556              {              {
4557                int category;
4558              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4559              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4560              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4651  for (;;) Line 4564  for (;;)
4564                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4565                }                }
4566              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4567              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4568              if ((prop_category == ucp_L ||              if ((category == ucp_L ||
4569                   prop_category == ucp_N ||                   category == ucp_N ||
4570                   c == CHAR_UNDERSCORE)                   c == CHAR_UNDERSCORE)
4571                     == prop_fail_result)                     == prop_fail_result)
4572                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4683  for (;;) Line 4596  for (;;)
4596              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
4597              }              }
4598            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
4599            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
4600            while (eptr < md->end_subject)            while (eptr < md->end_subject)
4601              {              {
4602              int len = 1;              int len = 1;
4603              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4604                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
4605              eptr += len;              eptr += len;
4606              }              }
4607            }            }
4608          }          }
   
4609        else        else
4610  #endif     /* SUPPORT_UCP */  #endif     /* SUPPORT_UCP */
4611    
# Line 5017  for (;;) Line 4926  for (;;)
4926            case PT_LAMP:            case PT_LAMP:
4927            for (i = min; i < max; i++)            for (i = min; i < max; i++)
4928              {              {
4929                int chartype;
4930              int len = 1;              int len = 1;
4931              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
4932                {                {
# Line 5024  for (;;) Line 4934  for (;;)
4934                break;                break;
4935                }                }
4936              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4937              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4938              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4939                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4940                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4941                break;                break;
4942              eptr+= len;              eptr+= len;
4943              }              }
# Line 5043  for (;;) Line 4953  for (;;)
4953                break;                break;
4954                }                }
4955              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4956              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
             if ((prop_category == prop_value) == prop_fail_result)  
               break;  
4957              eptr+= len;              eptr+= len;
4958              }              }
4959            break;            break;
# Line 5060  for (;;) Line 4968  for (;;)
4968                break;                break;
4969                }                }
4970              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4971              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
             if ((prop_chartype == prop_value) == prop_fail_result)  
               break;  
4972              eptr+= len;              eptr+= len;
4973              }              }
4974            break;            break;
# Line 5077  for (;;) Line 4983  for (;;)
4983                break;                break;
4984                }                }
4985              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4986              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
             if ((prop_script == prop_value) == prop_fail_result)  
               break;  
4987              eptr+= len;              eptr+= len;
4988              }              }
4989            break;            break;
# Line 5087  for (;;) Line 4991  for (;;)
4991            case PT_ALNUM:            case PT_ALNUM:
4992            for (i = min; i < max; i++)            for (i = min; i < max; i++)
4993              {              {
4994                int category;
4995              int len = 1;              int len = 1;
4996              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
4997                {                {
# Line 5094  for (;;) Line 4999  for (;;)
4999                break;                break;
5000                }                }
5001              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5002              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5003              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                  == prop_fail_result)  
5004                break;                break;
5005              eptr+= len;              eptr+= len;
5006              }              }
# Line 5112  for (;;) Line 5016  for (;;)
5016                break;                break;
5017                }                }
5018              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5019              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
5020                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
5021                   == prop_fail_result)                   == prop_fail_result)
5022                break;                break;
# Line 5131  for (;;) Line 5034  for (;;)
5034                break;                break;
5035                }                }
5036              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5037              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
5038                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5039                   == prop_fail_result)                   == prop_fail_result)
5040                break;                break;
# Line 5143  for (;;) Line 5045  for (;;)
5045            case PT_WORD:            case PT_WORD:
5046            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5047              {              {
5048                int category;
5049              int len = 1;              int len = 1;
5050              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5051                {                {
# Line 5150  for (;;) Line 5053  for (;;)
5053                break;                break;
5054                }                }
5055              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5056              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5057              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N ||
5058                   c == CHAR_UNDERSCORE) == prop_fail_result)                   c == CHAR_UNDERSCORE) == prop_fail_result)
5059                break;                break;
5060              eptr+= len;              eptr+= len;
# Line 5181  for (;;) Line 5084  for (;;)
5084          {          {
5085          for (i = min; i < max; i++)          for (i = min; i < max; i++)
5086            {            {
5087              int len = 1;
5088            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
5089              {              {
5090              SCHECK_PARTIAL();              SCHECK_PARTIAL();
5091              break;              break;
5092              }              }
5093            GETCHARINCTEST(c, eptr);            if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5094            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) break;
5095            if (prop_category == ucp_M) break;            eptr += len;
5096            while (eptr < md->end_subject)            while (eptr < md->end_subject)
5097              {              {
5098              int len = 1;              len = 1;
5099              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5100                {              if (UCD_CATEGORY(c) != ucp_M) break;
               GETCHARLEN(c, eptr, len);  
               }  
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
5101              eptr += len;              eptr += len;
5102              }              }
5103            }            }
# Line 5213  for (;;) Line 5113  for (;;)
5113            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5114            for (;;)                        /* Move back over one extended */            for (;;)                        /* Move back over one extended */
5115              {              {
             int len = 1;  
5116              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else
5117                {                {
5118                BACKCHAR(eptr);                BACKCHAR(eptr);
5119                GETCHARLEN(c, eptr, len);                GETCHAR(c, eptr);
5120                }                }
5121              prop_category = UCD_CATEGORY(c);              if (UCD_CATEGORY(c) != ucp_M) break;
             if (prop_category != ucp_M) break;  
5122              eptr--;              eptr--;
5123              }              }
5124            }            }
# Line 5284  for (;;) Line 5182  for (;;)
5182                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
5183                }                }
5184              }              }
5185            else eptr = md->end_subject;   /* Unlimited UTF-8 repeat */            else
5186                {
5187                eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
5188                SCHECK_PARTIAL();
5189                }
5190            break;            break;
5191    
5192            /* The byte case is the same as non-UTF8 */            /* The byte case is the same as non-UTF8 */
# Line 5505  for (;;) Line 5407  for (;;)
5407            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5408            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5409            BACKCHAR(eptr);            BACKCHAR(eptr);
5410            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5411                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5412            }            }
5413          }          }
# Line 5717  for (;;) Line 5619  for (;;)
5619            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
5620            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5621            eptr--;            eptr--;
5622            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5623                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5624            }            }
5625          }          }
# Line 5859  pcre_exec(const pcre *argument_re, const Line 5761  pcre_exec(const pcre *argument_re, const
5761    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5762    int offsetcount)    int offsetcount)
5763  {  {
5764  int rc, resetcount, ocount;  int rc, ocount, arg_offset_max;
5765  int first_byte = -1;  int first_byte = -1;
5766  int req_byte = -1;  int req_byte = -1;
5767  int req_byte2 = -1;  int req_byte2 = -1;
# Line 5895  if (re == NULL || subject == NULL || Line 5797  if (re == NULL || subject == NULL ||
5797  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
5798  if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;  if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
5799    
5800  /* This information is for finding all the numbers associated with a given  /* These two settings are used in the code for checking a UTF-8 string that
5801  name, for condition testing. */  follows immediately afterwards. Other values in the md block are used only
5802    during "normal" pcre_exec() processing, not when the JIT support is in use,
5803    so they are set up later. */
5804    
5805    utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
5806    md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5807                  ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5808    
5809    /* Check a UTF-8 string if required. Pass back the character offset and error
5810    code for an invalid string if a results vector is available. */
5811    
5812    #ifdef SUPPORT_UTF8
5813    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5814      {
5815      int erroroffset;
5816      int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);
5817      if (errorcode != 0)
5818        {
5819        if (offsetcount >= 2)
5820          {
5821          offsets[0] = erroroffset;
5822          offsets[1] = errorcode;
5823          }
5824        return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
5825          PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
5826        }
5827    
5828      /* Check that a start_offset points to the start of a UTF-8 character. */
5829      if (start_offset > 0 && start_offset < length &&
5830          (((USPTR)subject)[start_offset] & 0xc0) == 0x80)
5831        return PCRE_ERROR_BADUTF8_OFFSET;
5832      }
5833    #endif
5834    
5835    /* If the pattern was successfully studied with JIT support, run the JIT
5836    executable instead of the rest of this function. Most options must be set at
5837    compile time for the JIT code to be usable. Fallback to the normal code path if
5838    an unsupported flag is set. In particular, JIT does not support partial
5839    matching. */
5840    
5841    #ifdef SUPPORT_JIT
5842    if (extra_data != NULL
5843        && (extra_data->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0
5844        && extra_data->executable_jit != NULL
5845        && (options & ~(PCRE_NO_UTF8_CHECK | PCRE_NOTBOL | PCRE_NOTEOL |
5846                        PCRE_NOTEMPTY | PCRE_NOTEMPTY_ATSTART)) == 0)
5847      return _pcre_jit_exec(re, extra_data->executable_jit, subject, length,
5848        start_offset, options, offsets, offsetcount);
5849    #endif
5850    
5851    /* Carry on with non-JIT matching. This information is for finding all the
5852    numbers associated with a given name, for condition testing. */
5853    
5854  md->name_table = (uschar *)re + re->name_table_offset;  md->name_table = (uschar *)re + re->name_table_offset;
5855  md->name_count = re->name_count;  md->name_count = re->name_count;
# Line 5963  md->end_subject = md->start_subject + le Line 5916  md->end_subject = md->start_subject + le
5916  end_subject = md->end_subject;  end_subject = md->end_subject;
5917    
5918  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
 utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;  
5919  md->use_ucp = (re->options & PCRE_UCP) != 0;  md->use_ucp = (re->options & PCRE_UCP) != 0;
5920  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
5921    
# Line 5974  md->notbol = (options & PCRE_NOTBOL) != Line 5926  md->notbol = (options & PCRE_NOTBOL) !=
5926  md->noteol = (options & PCRE_NOTEOL) != 0;  md->noteol = (options & PCRE_NOTEOL) != 0;
5927  md->notempty = (options & PCRE_NOTEMPTY) != 0;  md->notempty = (options & PCRE_NOTEMPTY) != 0;
5928  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
5929  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :  
               ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;  
   
   
5930  md->hitend = FALSE;  md->hitend = FALSE;
5931  md->mark = NULL;                        /* In case never set */  md->mark = NULL;                        /* In case never set */
5932    
# Line 6059  defined (though never set). So there's n Line 6008  defined (though never set). So there's n
6008  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6009    return PCRE_ERROR_BADPARTIAL;    return PCRE_ERROR_BADPARTIAL;
6010    
 /* Check a UTF-8 string if required. Pass back the character offset and error  
 code for an invalid string if a results vector is available. */  
   
 #ifdef SUPPORT_UTF8  
 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)  
   {  
   int erroroffset;  
   int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);  
   if (errorcode != 0)  
     {  
     if (offsetcount >= 2)  
       {  
       offsets[0] = erroroffset;  
       offsets[1] = errorcode;  
       }  
     return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?  
       PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;  
     }  
   
   /* Check that a start_offset points to the start of a UTF-8 character. */  
   
   if (start_offset > 0 && start_offset < length &&  
       (((USPTR)subject)[start_offset] & 0xc0) == 0x80)  
     return PCRE_ERROR_BADUTF8_OFFSET;  
   }  
 #endif  
   
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
6014  of 3. */  of 3. */
6015    
6016  ocount = offsetcount - (offsetcount % 3);  ocount = offsetcount - (offsetcount % 3);
6017    arg_offset_max = (2*ocount)/3;
6018    
6019  if (re->top_backref > 0 && re->top_backref >= ocount/3)  if (re->top_backref > 0 && re->top_backref >= ocount/3)
6020    {    {
# Line 6108  md->offset_max = (2*ocount)/3; Line 6031  md->offset_max = (2*ocount)/3;
6031  md->offset_overflow = FALSE;  md->offset_overflow = FALSE;
6032  md->capture_last = -1;  md->capture_last = -1;
6033    
 /* 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;  
   
6034  /* Reset the working variable associated with each extraction. These should  /* Reset the working variable associated with each extraction. These should
6035  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
6036  initialize them to avoid reading uninitialized locations. */  initialize them to avoid reading uninitialized locations. Also, unset the
6037    offsets for the matched string. This is really just for tidiness with callouts,
6038    in case they inspect these fields. */
6039    
6040  if (md->offset_vector != NULL)  if (md->offset_vector != NULL)
6041    {    {
6042    register int *iptr = md->offset_vector + ocount;    register int *iptr = md->offset_vector + ocount;
6043    register int *iend = iptr - resetcount/2 + 1;    register int *iend = iptr - re->top_bracket;
6044      if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6045    while (--iptr >= iend) *iptr = -1;    while (--iptr >= iend) *iptr = -1;
6046      md->offset_vector[0] = md->offset_vector[1] = -1;
6047    }    }
6048    
6049  /* 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 6157  if ((re->flags & PCRE_REQCHSET) != 0) Line 6077  if ((re->flags & PCRE_REQCHSET) != 0)
6077    }    }
6078    
6079    
6080    
6081    
6082  /* ==========================================================================*/  /* ==========================================================================*/
6083    
6084  /* Loop for handling unanchored repeated matching attempts; for anchored regexs  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
# Line 6167  for(;;) Line 6089  for(;;)
6089    USPTR save_end_subject = end_subject;    USPTR save_end_subject = end_subject;
6090    USPTR new_start_match;    USPTR new_start_match;
6091    
   /* 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;  
     }  
   
6092    /* 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
6093    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
6094    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 6364  for(;;) Line 6277  for(;;)
6277    md->start_match_ptr = start_match;    md->start_match_ptr = start_match;
6278    md->start_used_ptr = start_match;    md->start_used_ptr = start_match;
6279    md->match_call_count = 0;    md->match_call_count = 0;
6280    md->match_function_type = 0;    md->match_function_type = 0;
6281    md->end_offset_top = 0;    md->end_offset_top = 0;
6282    rc = match(start_match, md->start_code, start_match, NULL, 2, md, NULL, 0);    rc = match(start_match, md->start_code, start_match, NULL, 2, md, NULL, 0);
6283    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
# Line 6476  if (rc == MATCH_MATCH || rc == MATCH_ACC Line 6389  if (rc == MATCH_MATCH || rc == MATCH_ACC
6389    {    {
6390    if (using_temporary_offsets)    if (using_temporary_offsets)
6391      {      {
6392      if (offsetcount >= 4)      if (arg_offset_max >= 4)
6393        {        {
6394        memcpy(offsets + 2, md->offset_vector + 2,        memcpy(offsets + 2, md->offset_vector + 2,
6395          (offsetcount - 2) * sizeof(int));          (arg_offset_max - 2) * sizeof(int));
6396        DPRINTF(("Copied offsets from temporary memory\n"));        DPRINTF(("Copied offsets from temporary memory\n"));
6397        }        }
6398      if (md->end_offset_top > offsetcount) md->offset_overflow = TRUE;      if (md->end_offset_top > arg_offset_max) md->offset_overflow = TRUE;
6399      DPRINTF(("Freeing temporary memory\n"));      DPRINTF(("Freeing temporary memory\n"));
6400      (pcre_free)(md->offset_vector);      (pcre_free)(md->offset_vector);
6401      }      }
6402    
6403    /* Set the return code to the number of captured strings, or 0 if there are    /* Set the return code to the number of captured strings, or 0 if there were
6404    too many to fit into the vector. */    too many to fit into the vector. */
6405    
6406    rc = md->offset_overflow? 0 : md->end_offset_top/2;    rc = (md->offset_overflow && md->end_offset_top >= arg_offset_max)?
6407        0 : md->end_offset_top/2;
6408    
6409      /* If there is space in the offset vector, set any unused pairs at the end of
6410      the pattern to -1 for backwards compatibility. It is documented that this
6411      happens. In earlier versions, the whole set of potential capturing offsets
6412      was set to -1 each time round the loop, but this is handled differently now.
6413      "Gaps" are set to -1 dynamically instead (this fixes a bug). Thus, it is only
6414      those at the end that need unsetting here. We can't just unset them all at
6415      the start of the whole thing because they may get set in one branch that is
6416      not the final matching branch. */
6417    
6418      if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
6419        {
6420        register int *iptr, *iend;
6421        int resetcount = 2 + re->top_bracket * 2;
6422        if (resetcount > offsetcount) resetcount = ocount;
6423        iptr = offsets + md->end_offset_top;
6424        iend = offsets + resetcount;
6425        while (iptr < iend) *iptr++ = -1;
6426        }
6427    
6428    /* 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
6429    md->start_match_ptr might be modified if \K was encountered on the success    md->start_match_ptr might be modified if \K was encountered on the success

Legend:
Removed from v.614  
changed lines
  Added in v.667

  ViewVC Help
Powered by ViewVC 1.1.5