/[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 608 by ph10, Sun Jun 12 16:25:55 2011 UTC revision 630 by ph10, Fri Jul 22 10:00:10 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, RM64 };         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);
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        if (rrc != MATCH_THEN) md->mark = markptr;        /* At this point, rrc will be one of MATCH_ONCE, MATCH_NOMATCH, or
855        RRETURN(MATCH_NOMATCH);        MATCH_THEN. */
856    
857          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
858          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 873  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 would 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. We can reduce stack usage by turning this into a tail      match() whatever happened so it was possible to reduce stack usage by
876      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        group. However, now that there is the possiblity of (*THEN) occurring in
878        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 (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        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,
         if (op >= OP_SBRA)   /* Possibly empty group */  
           {  
           md->match_function_type = MATCH_CBEGROUP;  
           RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,  
             RM48);  
           if (rrc == MATCH_NOMATCH) md->mark = markptr;  
           RRETURN(rrc);  
           }  
         /* Not a possibly empty group; use tail recursion */  
         ecode += _pcre_OP_lengths[*ecode];  
         DPRINTF(("bracket 0 tail recursion\n"));  
         goto TAIL_RECURSE;  
         }  
   
       /* For non-final alternatives, continue the loop for a NOMATCH result;  
       otherwise return. */  
   
       if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;  
       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);
911            }
912        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
913          if (*ecode != OP_ALT) break;
914        }        }
915      /* Control never reaches here. */      if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
916        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 939  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 947  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);
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 = 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 1009  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, RM64);          eptrb, RM48);
1032        if (rrc == MATCH_KETRPOS)        if (rrc == MATCH_KETRPOS)
1033          {          {
1034            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);
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 1053  for (;;) Line 1056  for (;;)
1056      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1057      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1058      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1059      exactly what going to the ket would do. As there is only one branch to be      exactly what going to the ket would do. */
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1060    
1061      case OP_COND:      case OP_COND:
1062      case OP_SCOND:      case OP_SCOND:
# Line 1238  for (;;) Line 1240  for (;;)
1240    
1241      else      else
1242        {        {
1243        md->match_function_type = MATCH_CONDASSERT;        md->match_function_type = MATCH_CONDASSERT;
1244        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1245        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1246          {          {
1247            if (md->end_offset_top > offset_top)
1248              offset_top = md->end_offset_top;  /* Captures may have happened */
1249          condition = TRUE;          condition = TRUE;
1250          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1251          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
# Line 1259  for (;;) Line 1263  for (;;)
1263        }        }
1264    
1265      /* 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,
1266      we can use tail recursion to avoid using another stack frame, except when      we used to use tail recursion to avoid using another stack frame, except
1267      we have an unlimited repeat of a possibly empty group. If the second      when there was unlimited repeat of a possibly empty group. However, that
1268      alternative doesn't exist, we can just plough on. */      strategy no longer works because of the possibilty of (*THEN) being
1269        encountered in the branch. A recursive call to match() is always required,
1270        unless the second alternative doesn't exist, in which case we can just
1271        plough on. */
1272    
1273      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1274        {        {
1275        ecode += 1 + LINK_SIZE;        if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1276        if (op == OP_SCOND)        /* Possibly empty group */        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1277          {        if (rrc == MATCH_THEN && md->start_match_ptr == ecode)
1278          md->match_function_type = MATCH_CBEGROUP;          rrc = MATCH_NOMATCH;
1279          RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);        RRETURN(rrc);
         RRETURN(rrc);  
         }  
       else goto TAIL_RECURSE;  
1280        }        }
1281      else                         /* Condition false & no alternative */      else                         /* Condition false & no alternative */
1282        {        {
# Line 1305  for (;;) Line 1309  for (;;)
1309      break;      break;
1310    
1311    
1312      /* 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. */  
1313    
     case OP_ACCEPT:  
1314      case OP_END:      case OP_END:
1315      if (md->recursive != NULL)      case OP_ACCEPT:
1316        {      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;  
         }  
       }  
   
     /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is  
     set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of  
     the subject. In both cases, backtracking will then try other alternatives,  
     if any. */  
1317    
1318      else if (eptr == mstart &&      /* If we have matched an empty string, fail if not in an assertion and not
1319          (md->notempty ||      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1320            (md->notempty_atstart &&      is set and we have matched at the start of the subject. In both cases,
1321              mstart == md->start_subject + md->start_offset)))      backtracking will then try other alternatives, if any. */
1322    
1323        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1324             md->recursive == NULL &&
1325             (md->notempty ||
1326               (md->notempty_atstart &&
1327                 mstart == md->start_subject + md->start_offset)))
1328        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1329    
1330      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
1331    
1332      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1333      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1334      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
# Line 1352  for (;;) Line 1343  for (;;)
1343      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,
1344      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
1345      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
1346      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
1347      of a condition, we want to return immediately afterwards. The caller of      of a condition, we want to return immediately afterwards. The caller of
1348      this incarnation of the match() function will have set MATCH_CONDASSERT in      this incarnation of the match() function will have set MATCH_CONDASSERT in
1349      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
1350      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
1351      match() to remember this case. */      match() to remember this case. */
1352    
1353      case OP_ASSERT:      case OP_ASSERT:
# Line 1366  for (;;) Line 1357  for (;;)
1357        condassert = TRUE;        condassert = TRUE;
1358        md->match_function_type = 0;        md->match_function_type = 0;
1359        }        }
1360      else condassert = FALSE;      else condassert = FALSE;
1361    
1362      do      do
1363        {        {
1364        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1365        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1366          {          {
1367          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
1368            markptr = md->mark;
1369          break;          break;
1370          }          }
1371        if (rrc != MATCH_NOMATCH &&        if (rrc != MATCH_NOMATCH &&
# Line 1382  for (;;) Line 1374  for (;;)
1374        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1375        }        }
1376      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1377    
1378      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1379    
1380      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
# Line 1408  for (;;) Line 1400  for (;;)
1400        condassert = TRUE;        condassert = TRUE;
1401        md->match_function_type = 0;        md->match_function_type = 0;
1402        }        }
1403      else condassert = FALSE;      else condassert = FALSE;
1404    
1405      do      do
1406        {        {
# Line 1427  for (;;) Line 1419  for (;;)
1419      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1420    
1421      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1422    
1423      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1424      continue;      continue;
1425    
# Line 1494  for (;;) Line 1486  for (;;)
1486      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
1487      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1488    
1489      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1490      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
1491      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
1492      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
1493      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
1494      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
1495      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.  
1496    
1497      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
1498      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
1499      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1500        a lot, so he is not to blame for the current way it works. */
1501    
1502      case OP_RECURSE:      case OP_RECURSE:
1503        {        {
# Line 1520  for (;;) Line 1510  for (;;)
1510        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1511        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1512    
1513        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1514    
1515        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1516    
1517        /* Now save the offset data. */        /* Now save the offset data */
1518    
1519        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1520        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1536  for (;;) Line 1525  for (;;)
1525            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1526          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1527          }          }
   
1528        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1529              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1530        new_recursive.save_offset_top = offset_top;  
1531          /* OK, now we can do the recursion. After processing each alternative,
1532        /* OK, now we can do the recursion. For each top-level alternative we        restore the offset data. If there were nested recursions, md->recursive
1533        restore the offset and recursion data. */        might be changed, so reset it before looping. */
1534    
1535        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1536        cbegroup = (*callpat >= OP_SBRA);        cbegroup = (*callpat >= OP_SBRA);
# Line 1551  for (;;) Line 1539  for (;;)
1539          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1540          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1541            md, eptrb, RM6);            md, eptrb, RM6);
1542            memcpy(md->offset_vector, new_recursive.offset_save,
1543                new_recursive.saved_max * sizeof(int));
1544          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1545            {            {
1546            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
1547            md->recursive = new_recursive.prevrec;            md->recursive = new_recursive.prevrec;
1548            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1549              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1550            MRRETURN(MATCH_MATCH);  
1551              /* Set where we got to in the subject, and reset the start in case
1552              it was changed by \K. This *is* propagated back out of a recursion,
1553              for Perl compatibility. */
1554    
1555              eptr = md->end_match_ptr;
1556              mstart = md->start_match_ptr;
1557              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1558            }            }
1559          else if (rrc != MATCH_NOMATCH &&          else if (rrc != MATCH_NOMATCH &&
1560                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
# Line 1569  for (;;) Line 1566  for (;;)
1566            }            }
1567    
1568          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1569          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1570          }          }
1571        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1581  for (;;) Line 1576  for (;;)
1576          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1577        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1578        }        }
     /* 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. */  
1579    
1580      case OP_ONCE:      RECURSION_MATCHED:
1581      prev = ecode;      break;
     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);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. */  
   
     if (*ecode == OP_KETRMIN)  
       {  
       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 */  
1582    
1583      /* 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
1584      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1667  for (;;) Line 1592  for (;;)
1592      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
1593      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
1594      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1595    
1596      case OP_BRAZERO:      case OP_BRAZERO:
1597      next = ecode + 1;      next = ecode + 1;
1598      RMATCH(eptr, next, offset_top, md, eptrb, RM10);      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
# Line 1675  for (;;) Line 1600  for (;;)
1600      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
1601      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1602      break;      break;
1603    
1604      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1605      next = ecode + 1;      next = ecode + 1;
1606      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
# Line 1689  for (;;) Line 1614  for (;;)
1614      do next += GET(next,1); while (*next == OP_ALT);      do next += GET(next,1); while (*next == OP_ALT);
1615      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1616      break;      break;
1617    
1618      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1619      here; just jump to the group, with allow_zero set TRUE. */      here; just jump to the group, with allow_zero set TRUE. */
1620    
1621      case OP_BRAPOSZERO:      case OP_BRAPOSZERO:
1622      op = *(++ecode);      op = *(++ecode);
1623      allow_zero = TRUE;      allow_zero = TRUE;
1624      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1625        goto POSSESSIVE_NON_CAPTURE;        goto POSSESSIVE_NON_CAPTURE;
# Line 1704  for (;;) Line 1629  for (;;)
1629      case OP_KET:      case OP_KET:
1630      case OP_KETRMIN:      case OP_KETRMIN:
1631      case OP_KETRMAX:      case OP_KETRMAX:
1632      case OP_KETRPOS:      case OP_KETRPOS:
1633      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1634    
1635      /* 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
1636      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1637      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1638    
1639      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1640        {        {
1641        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1642        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1643        }        }
1644      else saved_eptr = NULL;      else saved_eptr = NULL;
1645    
1646      /* 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
1647      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
1648      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
1649      it was changed by \K. */      by \K. */
1650    
1651      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1652          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1653        {        {
1654        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1655        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1656        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1657        MRRETURN(MATCH_MATCH);        MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1658        }        }
1659    
1660      /* 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
1661      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1662      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
1663      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
1664      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
1665        the current subject position and start match pointer and give a MATCH
1666        return. */
1667    
1668      if (*prev == OP_CBRA || *prev == OP_SCBRA ||      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1669          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
# Line 1750  for (;;) Line 1676  for (;;)
1676        printf("\n");        printf("\n");
1677  #endif  #endif
1678    
1679          /* Handle a recursively called group. */
1680    
1681          if (md->recursive != NULL && md->recursive->group_num == number)
1682            {
1683            md->end_match_ptr = eptr;
1684            md->start_match_ptr = mstart;
1685            RRETURN(MATCH_MATCH);
1686            }
1687    
1688          /* Deal with capturing */
1689    
1690        md->capture_last = number;        md->capture_last = number;
1691        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1692          {          {
1693            /* If offset is greater than offset_top, it means that we are
1694            "skipping" a capturing group, and that group's offsets must be marked
1695            unset. In earlier versions of PCRE, all the offsets were unset at the
1696            start of matching, but this doesn't work because atomic groups and
1697            assertions can cause a value to be set that should later be unset.
1698            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1699            part of the atomic group, but this is not on the final matching path,
1700            so must be unset when 2 is set. (If there is no group 2, there is no
1701            problem, because offset_top will then be 2, indicating no capture.) */
1702    
1703            if (offset > offset_top)
1704              {
1705              register int *iptr = md->offset_vector + offset_top;
1706              register int *iend = md->offset_vector + offset;
1707              while (iptr < iend) *iptr++ = -1;
1708              }
1709    
1710            /* Now make the extraction */
1711    
1712          md->offset_vector[offset] =          md->offset_vector[offset] =
1713            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1714          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1715          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1716          }          }
   
       /* 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;  
         }  
1717        }        }
1718    
1719      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1720      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
1721      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1722      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
1723      course of events. */      processing the rest of the pattern at a lower level. If this results in a
1724        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1725        bypassing intermediate backup points, but resetting any captures that
1726        happened along the way. */
1727    
1728      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1729        {        {
1730        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1731            {
1732            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1733            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1734            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1735            RRETURN(MATCH_ONCE);
1736            }
1737          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1738        break;        break;
1739        }        }
1740    
1741      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1742      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
1743      at a time from the outer level, thus saving stack. */      at a time from the outer level, thus saving stack. */
1744    
1745      if (*ecode == OP_KETRPOS)      if (*ecode == OP_KETRPOS)
1746        {        {
1747        md->end_match_ptr = eptr;        md->end_match_ptr = eptr;
1748        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1749        RRETURN(MATCH_KETRPOS);        RRETURN(MATCH_KETRPOS);
1750        }        }
1751    
1752      /* 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
1753      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
1754      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
1755      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
1756        string. */
1757    
1758      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1759        {        {
1760        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1761        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1762          if (*prev == OP_ONCE)
1763            {
1764            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1765            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1766            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1767            RRETURN(MATCH_ONCE);
1768            }
1769        if (*prev >= OP_SBRA)    /* Could match an empty string */        if (*prev >= OP_SBRA)    /* Could match an empty string */
1770          {          {
1771          md->match_function_type = MATCH_CBEGROUP;          md->match_function_type = MATCH_CBEGROUP;
1772          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1773          RRETURN(rrc);          RRETURN(rrc);
1774          }          }
# Line 1818  for (;;) Line 1777  for (;;)
1777        }        }
1778      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1779        {        {
1780        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1781        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1782          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1783        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1784          if (*prev == OP_ONCE)
1785            {
1786            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1787            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1788            md->once_target = prev;
1789            RRETURN(MATCH_ONCE);
1790            }
1791        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1792        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1793        }        }
# Line 1830  for (;;) Line 1797  for (;;)
1797    
1798      case OP_CIRC:      case OP_CIRC:
1799      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1800    
1801      /* Start of subject assertion */      /* Start of subject assertion */
1802    
1803      case OP_SOD:      case OP_SOD:
1804      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1805      ecode++;      ecode++;
1806      break;      break;
1807    
1808      /* Multiline mode: start of subject unless notbol, or after any newline. */      /* Multiline mode: start of subject unless notbol, or after any newline. */
1809    
1810      case OP_CIRCM:      case OP_CIRCM:
# Line 1876  for (;;) Line 1843  for (;;)
1843      ecode++;      ecode++;
1844      break;      break;
1845    
1846      /* Not multiline mode: assert before a terminating newline or before end of      /* Not multiline mode: assert before a terminating newline or before end of
1847      subject unless noteol is set. */      subject unless noteol is set. */
1848    
1849      case OP_DOLL:      case OP_DOLL:
# Line 2165  for (;;) Line 2132  for (;;)
2132      switch(c)      switch(c)
2133        {        {
2134        default: MRRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2135    
2136        case 0x000d:        case 0x000d:
2137        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2138        break;        break;
# Line 2389  for (;;) Line 2356  for (;;)
2356        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2357        }        }
2358      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2359        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2360        while (eptr < md->end_subject)
2361        {        {
2362        int category = UCD_CATEGORY(c);        int len = 1;
2363        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2364        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2365          {        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;  
         }  
2366        }        }
2367      ecode++;      ecode++;
2368      break;      break;
# Line 2418  for (;;) Line 2378  for (;;)
2378      loops). */      loops). */
2379    
2380      case OP_REF:      case OP_REF:
2381      case OP_REFI:      case OP_REFI:
2382      caseless = op == OP_REFI;      caseless = op == OP_REFI;
2383      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2384      ecode += 3;      ecode += 3;
2385    
# Line 2487  for (;;) Line 2447  for (;;)
2447    
2448      for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2449        {        {
2450        int slength;        int slength;
2451        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2452          {          {
2453          CHECK_PARTIAL();          CHECK_PARTIAL();
# Line 2507  for (;;) Line 2467  for (;;)
2467        {        {
2468        for (fi = min;; fi++)        for (fi = min;; fi++)
2469          {          {
2470          int slength;          int slength;
2471          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2472          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2473          if (fi >= max) MRRETURN(MATCH_NOMATCH);          if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 2528  for (;;) Line 2488  for (;;)
2488        pp = eptr;        pp = eptr;
2489        for (i = min; i < max; i++)        for (i = min; i < max; i++)
2490          {          {
2491          int slength;          int slength;
2492          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2493            {            {
2494            CHECK_PARTIAL();            CHECK_PARTIAL();
# Line 3261  for (;;) Line 3221  for (;;)
3221      checking can be multibyte. */      checking can be multibyte. */
3222    
3223      case OP_NOT:      case OP_NOT:
3224      case OP_NOTI:      case OP_NOTI:
3225      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3226        {        {
3227        SCHECK_PARTIAL();        SCHECK_PARTIAL();
# Line 3756  for (;;) Line 3716  for (;;)
3716            case PT_LAMP:            case PT_LAMP:
3717            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3718              {              {
3719                int chartype;
3720              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3721                {                {
3722                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3723                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3724                }                }
3725              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3726              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
3727              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
3728                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
3729                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
3730                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3731              }              }
3732            break;            break;
# Line 3779  for (;;) Line 3740  for (;;)
3740                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3741                }                }
3742              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3743              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
3744                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3745              }              }
3746            break;            break;
# Line 3794  for (;;) Line 3754  for (;;)
3754                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3755                }                }
3756              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3757              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
3758                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3759              }              }
3760            break;            break;
# Line 3809  for (;;) Line 3768  for (;;)
3768                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3769                }                }
3770              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3771              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
3772                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3773              }              }
3774            break;            break;
# Line 3818  for (;;) Line 3776  for (;;)
3776            case PT_ALNUM:            case PT_ALNUM:
3777            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3778              {              {
3779                int category;
3780              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3781                {                {
3782                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3783                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3784                }                }
3785              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3786              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3787              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
3788                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3789              }              }
3790            break;            break;
# Line 3840  for (;;) Line 3798  for (;;)
3798                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3799                }                }
3800              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3801              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 ||  
3802                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
3803                     == prop_fail_result)                     == prop_fail_result)
3804                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3857  for (;;) Line 3814  for (;;)
3814                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3815                }                }
3816              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3817              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 ||  
3818                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
3819                     == prop_fail_result)                     == prop_fail_result)
3820                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3868  for (;;) Line 3824  for (;;)
3824            case PT_WORD:            case PT_WORD:
3825            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3826              {              {
3827                int category;
3828              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3829                {                {
3830                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3831                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3832                }                }
3833              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3834              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3835              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
                  c == CHAR_UNDERSCORE)  
3836                     == prop_fail_result)                     == prop_fail_result)
3837                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3838              }              }
# Line 3902  for (;;) Line 3858  for (;;)
3858              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3859              }              }
3860            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3861            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
3862            while (eptr < md->end_subject)            while (eptr < md->end_subject)
3863              {              {
3864              int len = 1;              int len = 1;
3865              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
3866                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
3867              eptr += len;              eptr += len;
3868              }              }
3869            }            }
# Line 3968  for (;;) Line 3921  for (;;)
3921            switch(c)            switch(c)
3922              {              {
3923              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
3924    
3925              case 0x000d:              case 0x000d:
3926              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3927              break;              break;
# Line 4245  for (;;) Line 4198  for (;;)
4198            switch(*eptr++)            switch(*eptr++)
4199              {              {
4200              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4201    
4202              case 0x000d:              case 0x000d:
4203              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4204              break;              break;
4205    
4206              case 0x000a:              case 0x000a:
4207              break;              break;
4208    
# Line 4455  for (;;) Line 4408  for (;;)
4408            case PT_LAMP:            case PT_LAMP:
4409            for (fi = min;; fi++)            for (fi = min;; fi++)
4410              {              {
4411                int chartype;
4412              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4413              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4414              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4464  for (;;) Line 4418  for (;;)
4418                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4419                }                }
4420              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4421              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4422              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4423                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4424                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4425                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4426              }              }
4427            /* Control never gets here */            /* Control never gets here */
# Line 4484  for (;;) Line 4438  for (;;)
4438                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4439                }                }
4440              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4441              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
4442                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4443              }              }
4444            /* Control never gets here */            /* Control never gets here */
# Line 4502  for (;;) Line 4455  for (;;)
4455                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4456                }                }
4457              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4458              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
4459                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4460              }              }
4461            /* Control never gets here */            /* Control never gets here */
# Line 4520  for (;;) Line 4472  for (;;)
4472                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4473                }                }
4474              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4475              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
4476                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4477              }              }
4478            /* Control never gets here */            /* Control never gets here */
# Line 4529  for (;;) Line 4480  for (;;)
4480            case PT_ALNUM:            case PT_ALNUM:
4481            for (fi = min;; fi++)            for (fi = min;; fi++)
4482              {              {
4483                int category;
4484              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4485              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4486              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4538  for (;;) Line 4490  for (;;)
4490                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4491                }                }
4492              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4493              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4494              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
4495                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4496              }              }
4497            /* Control never gets here */            /* Control never gets here */
# Line 4557  for (;;) Line 4508  for (;;)
4508                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4509                }                }
4510              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4511              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 ||  
4512                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
4513                     == prop_fail_result)                     == prop_fail_result)
4514                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4577  for (;;) Line 4527  for (;;)
4527                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4528                }                }
4529              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4530              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 ||  
4531                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4532                     == prop_fail_result)                     == prop_fail_result)
4533                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4588  for (;;) Line 4537  for (;;)
4537            case PT_WORD:            case PT_WORD:
4538            for (fi = min;; fi++)            for (fi = min;; fi++)
4539              {              {
4540                int category;
4541              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4542              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4543              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4597  for (;;) Line 4547  for (;;)
4547                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4548                }                }
4549              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4550              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4551              if ((prop_category == ucp_L ||              if ((category == ucp_L ||
4552                   prop_category == ucp_N ||                   category == ucp_N ||
4553                   c == CHAR_UNDERSCORE)                   c == CHAR_UNDERSCORE)
4554                     == prop_fail_result)                     == prop_fail_result)
4555                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4629  for (;;) Line 4579  for (;;)
4579              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
4580              }              }
4581            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
4582            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
4583            while (eptr < md->end_subject)            while (eptr < md->end_subject)
4584              {              {
4585              int len = 1;              int len = 1;
4586              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4587                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
4588              eptr += len;              eptr += len;
4589              }              }
4590            }            }
4591          }          }
   
4592        else        else
4593  #endif     /* SUPPORT_UCP */  #endif     /* SUPPORT_UCP */
4594    
# Line 4963  for (;;) Line 4909  for (;;)
4909            case PT_LAMP:            case PT_LAMP:
4910            for (i = min; i < max; i++)            for (i = min; i < max; i++)
4911              {              {
4912                int chartype;
4913              int len = 1;              int len = 1;
4914              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
4915                {                {
# Line 4970  for (;;) Line 4917  for (;;)
4917                break;                break;
4918                }                }
4919              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4920              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4921              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4922                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4923                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4924                break;                break;
4925              eptr+= len;              eptr+= len;
4926              }              }
# Line 4989  for (;;) Line 4936  for (;;)
4936                break;                break;
4937                }                }
4938              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4939              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
             if ((prop_category == prop_value) == prop_fail_result)  
               break;  
4940              eptr+= len;              eptr+= len;
4941              }              }
4942            break;            break;
# Line 5006  for (;;) Line 4951  for (;;)
4951                break;                break;
4952                }                }
4953              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4954              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
             if ((prop_chartype == prop_value) == prop_fail_result)  
               break;  
4955              eptr+= len;              eptr+= len;
4956              }              }
4957            break;            break;
# Line 5023  for (;;) Line 4966  for (;;)
4966                break;                break;
4967                }                }
4968              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4969              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
             if ((prop_script == prop_value) == prop_fail_result)  
               break;  
4970              eptr+= len;              eptr+= len;
4971              }              }
4972            break;            break;
# Line 5033  for (;;) Line 4974  for (;;)
4974            case PT_ALNUM:            case PT_ALNUM:
4975            for (i = min; i < max; i++)            for (i = min; i < max; i++)
4976              {              {
4977                int category;
4978              int len = 1;              int len = 1;
4979              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
4980                {                {
# Line 5040  for (;;) Line 4982  for (;;)
4982                break;                break;
4983                }                }
4984              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4985              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4986              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                  == prop_fail_result)  
4987                break;                break;
4988              eptr+= len;              eptr+= len;
4989              }              }
# Line 5058  for (;;) Line 4999  for (;;)
4999                break;                break;
5000                }                }
5001              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5002              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 ||  
5003                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
5004                   == prop_fail_result)                   == prop_fail_result)
5005                break;                break;
# Line 5077  for (;;) Line 5017  for (;;)
5017                break;                break;
5018                }                }
5019              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5020              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 ||  
5021                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5022                   == prop_fail_result)                   == prop_fail_result)
5023                break;                break;
# Line 5089  for (;;) Line 5028  for (;;)
5028            case PT_WORD:            case PT_WORD:
5029            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5030              {              {
5031                int category;
5032              int len = 1;              int len = 1;
5033              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5034                {                {
# Line 5096  for (;;) Line 5036  for (;;)
5036                break;                break;
5037                }                }
5038              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5039              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5040              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N ||
5041                   c == CHAR_UNDERSCORE) == prop_fail_result)                   c == CHAR_UNDERSCORE) == prop_fail_result)
5042                break;                break;
5043              eptr+= len;              eptr+= len;
# Line 5127  for (;;) Line 5067  for (;;)
5067          {          {
5068          for (i = min; i < max; i++)          for (i = min; i < max; i++)
5069            {            {
5070              int len = 1;
5071            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
5072              {              {
5073              SCHECK_PARTIAL();              SCHECK_PARTIAL();
5074              break;              break;
5075              }              }
5076            GETCHARINCTEST(c, eptr);            if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5077            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) break;
5078            if (prop_category == ucp_M) break;            eptr += len;
5079            while (eptr < md->end_subject)            while (eptr < md->end_subject)
5080              {              {
5081              int len = 1;              len = 1;
5082              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5083                {              if (UCD_CATEGORY(c) != ucp_M) break;
               GETCHARLEN(c, eptr, len);  
               }  
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
5084              eptr += len;              eptr += len;
5085              }              }
5086            }            }
# Line 5159  for (;;) Line 5096  for (;;)
5096            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5097            for (;;)                        /* Move back over one extended */            for (;;)                        /* Move back over one extended */
5098              {              {
             int len = 1;  
5099              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else
5100                {                {
5101                BACKCHAR(eptr);                BACKCHAR(eptr);
5102                GETCHARLEN(c, eptr, len);                GETCHAR(c, eptr);
5103                }                }
5104              prop_category = UCD_CATEGORY(c);              if (UCD_CATEGORY(c) != ucp_M) break;
             if (prop_category != ucp_M) break;  
5105              eptr--;              eptr--;
5106              }              }
5107            }            }
# Line 5451  for (;;) Line 5386  for (;;)
5386            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5387            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5388            BACKCHAR(eptr);            BACKCHAR(eptr);
5389            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5390                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5391            }            }
5392          }          }
# Line 5663  for (;;) Line 5598  for (;;)
5598            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
5599            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5600            eptr--;            eptr--;
5601            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5602                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5603            }            }
5604          }          }
# Line 5703  switch (frame->Xwhere) Line 5638  switch (frame->Xwhere)
5638    LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)    LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
5639    LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)    LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
5640    LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)    LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
5641    LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63) LBL(64)    LBL(53) LBL(54) LBL(55) LBL(56) LBL(57) LBL(58) LBL(63)
5642  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5643    LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)    LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)
5644    LBL(32) LBL(34) LBL(42) LBL(46)    LBL(32) LBL(34) LBL(42) LBL(46)
# Line 5805  pcre_exec(const pcre *argument_re, const Line 5740  pcre_exec(const pcre *argument_re, const
5740    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5741    int offsetcount)    int offsetcount)
5742  {  {
5743  int rc, resetcount, ocount;  int rc, ocount;
5744  int first_byte = -1;  int first_byte = -1;
5745  int req_byte = -1;  int req_byte = -1;
5746  int req_byte2 = -1;  int req_byte2 = -1;
# Line 5913  utf8 = md->utf8 = (re->options & PCRE_UT Line 5848  utf8 = md->utf8 = (re->options & PCRE_UT
5848  md->use_ucp = (re->options & PCRE_UCP) != 0;  md->use_ucp = (re->options & PCRE_UCP) != 0;
5849  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
5850    
5851    /* Some options are unpacked into BOOL variables in the hope that testing
5852    them will be faster than individual option bits. */
5853    
5854  md->notbol = (options & PCRE_NOTBOL) != 0;  md->notbol = (options & PCRE_NOTBOL) != 0;
5855  md->noteol = (options & PCRE_NOTEOL) != 0;  md->noteol = (options & PCRE_NOTEOL) != 0;
5856  md->notempty = (options & PCRE_NOTEMPTY) != 0;  md->notempty = (options & PCRE_NOTEMPTY) != 0;
5857  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
5858  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5859                ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;                ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5860    
5861    
5862  md->hitend = FALSE;  md->hitend = FALSE;
5863  md->mark = NULL;                        /* In case never set */  md->mark = NULL;                        /* In case never set */
5864    
# Line 6000  defined (though never set). So there's n Line 5940  defined (though never set). So there's n
5940  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
5941    return PCRE_ERROR_BADPARTIAL;    return PCRE_ERROR_BADPARTIAL;
5942    
5943  /* Check a UTF-8 string if required. Pass back the character offset and error  /* Check a UTF-8 string if required. Pass back the character offset and error
5944  code for an invalid string if a results vector is available. */  code for an invalid string if a results vector is available. */
5945    
5946  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5947  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5948    {    {
5949    int erroroffset;    int erroroffset;
5950    int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);    int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);
5951    if (errorcode != 0)    if (errorcode != 0)
5952      {      {
# Line 6014  if (utf8 && (options & PCRE_NO_UTF8_CHEC Line 5954  if (utf8 && (options & PCRE_NO_UTF8_CHEC
5954        {        {
5955        offsets[0] = erroroffset;        offsets[0] = erroroffset;
5956        offsets[1] = errorcode;        offsets[1] = errorcode;
5957        }        }
5958      return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?      return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
5959        PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;        PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
5960      }      }
5961    
5962    /* Check that a start_offset points to the start of a UTF-8 character. */    /* Check that a start_offset points to the start of a UTF-8 character. */
5963    
5964    if (start_offset > 0 && start_offset < length &&    if (start_offset > 0 && start_offset < length &&
5965        (((USPTR)subject)[start_offset] & 0xc0) == 0x80)        (((USPTR)subject)[start_offset] & 0xc0) == 0x80)
5966      return PCRE_ERROR_BADUTF8_OFFSET;      return PCRE_ERROR_BADUTF8_OFFSET;
5967    }    }
5968  #endif  #endif
# Line 6049  md->offset_max = (2*ocount)/3; Line 5989  md->offset_max = (2*ocount)/3;
5989  md->offset_overflow = FALSE;  md->offset_overflow = FALSE;
5990  md->capture_last = -1;  md->capture_last = -1;
5991    
 /* 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;  
   
5992  /* Reset the working variable associated with each extraction. These should  /* Reset the working variable associated with each extraction. These should
5993  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
5994  initialize them to avoid reading uninitialized locations. */  initialize them to avoid reading uninitialized locations. Also, unset the
5995    offsets for the matched string. This is really just for tidiness with callouts,
5996    in case they inspect these fields. */
5997    
5998  if (md->offset_vector != NULL)  if (md->offset_vector != NULL)
5999    {    {
6000    register int *iptr = md->offset_vector + ocount;    register int *iptr = md->offset_vector + ocount;
6001    register int *iend = iptr - resetcount/2 + 1;    register int *iend = iptr - re->top_bracket;
6002      if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6003    while (--iptr >= iend) *iptr = -1;    while (--iptr >= iend) *iptr = -1;
6004      md->offset_vector[0] = md->offset_vector[1] = -1;
6005    }    }
6006    
6007  /* 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 6098  if ((re->flags & PCRE_REQCHSET) != 0) Line 6035  if ((re->flags & PCRE_REQCHSET) != 0)
6035    }    }
6036    
6037    
6038    
6039    
6040  /* ==========================================================================*/  /* ==========================================================================*/
6041    
6042  /* Loop for handling unanchored repeated matching attempts; for anchored regexs  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
# Line 6108  for(;;) Line 6047  for(;;)
6047    USPTR save_end_subject = end_subject;    USPTR save_end_subject = end_subject;
6048    USPTR new_start_match;    USPTR new_start_match;
6049    
   /* 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;  
     }  
   
6050    /* 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
6051    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
6052    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 6305  for(;;) Line 6235  for(;;)
6235    md->start_match_ptr = start_match;    md->start_match_ptr = start_match;
6236    md->start_used_ptr = start_match;    md->start_used_ptr = start_match;
6237    md->match_call_count = 0;    md->match_call_count = 0;
6238    md->match_function_type = 0;    md->match_function_type = 0;
6239      md->end_offset_top = 0;
6240    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);
6241    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
6242    
# Line 6429  if (rc == MATCH_MATCH || rc == MATCH_ACC Line 6360  if (rc == MATCH_MATCH || rc == MATCH_ACC
6360    
6361    /* Set the return code to the number of captured strings, or 0 if there are    /* Set the return code to the number of captured strings, or 0 if there are
6362    too many to fit into the vector. */    too many to fit into the vector. */
6363    
6364    rc = md->offset_overflow? 0 : md->end_offset_top/2;    rc = md->offset_overflow? 0 : md->end_offset_top/2;
6365    
6366      /* If there is space in the offset vector, set any unused pairs at the end of
6367      the pattern to -1 for backwards compatibility. It is documented that this
6368      happens. In earlier versions, the whole set of potential capturing offsets
6369      was set to -1 each time round the loop, but this is handled differently now.
6370      "Gaps" are set to -1 dynamically instead (this fixes a bug). Thus, it is only
6371      those at the end that need unsetting here. We can't just unset them all at
6372      the start of the whole thing because they may get set in one branch that is
6373      not the final matching branch. */
6374    
6375      if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
6376        {
6377        register int *iptr, *iend;
6378        int resetcount = 2 + re->top_bracket * 2;
6379        if (resetcount > offsetcount) resetcount = ocount;
6380        iptr = offsets + md->end_offset_top;
6381        iend = offsets + resetcount;
6382        while (iptr < iend) *iptr++ = -1;
6383        }
6384    
6385    /* 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
6386    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
6387    matching path. */    matching path. */

Legend:
Removed from v.608  
changed lines
  Added in v.630

  ViewVC Help
Powered by ViewVC 1.1.5