/[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 699 by ph10, Tue Sep 20 10:46:54 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 1068  for (;;) Line 1070  for (;;)
1070        if (pcre_callout != NULL)        if (pcre_callout != NULL)
1071          {          {
1072          pcre_callout_block cb;          pcre_callout_block cb;
1073          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1074          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1075          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1076          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 1080  for (;;) Line 1082  for (;;)
1082          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1083          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1084          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1085            cb.mark             = markptr;
1086          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1087          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1088          }          }
# Line 1238  for (;;) Line 1241  for (;;)
1241    
1242      else      else
1243        {        {
1244        md->match_function_type = MATCH_CONDASSERT;        md->match_function_type = MATCH_CONDASSERT;
1245        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1246        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1247          {          {
1248            if (md->end_offset_top > offset_top)
1249              offset_top = md->end_offset_top;  /* Captures may have happened */
1250          condition = TRUE;          condition = TRUE;
1251          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1252          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
# Line 1259  for (;;) Line 1264  for (;;)
1264        }        }
1265    
1266      /* We are now at the branch that is to be obeyed. As there is only one,      /* We are now at the branch that is to be obeyed. As there is only one,
1267      we 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
1268      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
1269      alternative doesn't exist, we can just plough on. */      strategy no longer works because of the possibilty of (*THEN) being
1270        encountered in the branch. A recursive call to match() is always required,
1271        unless the second alternative doesn't exist, in which case we can just
1272        plough on. */
1273    
1274      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1275        {        {
1276        ecode += 1 + LINK_SIZE;        if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1277        if (op == OP_SCOND)        /* Possibly empty group */        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1278          {  
1279          md->match_function_type = MATCH_CBEGROUP;        /* If the result is THEN from within the "true" branch of the condition,
1280          RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);        md->start_match_ptr will point to the original OP_COND, not to the start
1281          RRETURN(rrc);        of the branch, so we have do work to see if it matches. If THEN comes
1282          }        from the "false" branch, md->start_match_ptr does point to OP_ALT. */
1283        else goto TAIL_RECURSE;  
1284          if (rrc == MATCH_THEN)
1285            {
1286            if (*ecode != OP_ALT)
1287              {
1288              do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
1289              ecode -= GET(ecode, 1);
1290              }
1291            if (md->start_match_ptr == ecode) rrc = MATCH_NOMATCH;
1292            }
1293          RRETURN(rrc);
1294        }        }
1295      else                         /* Condition false & no alternative */      else                         /* Condition false & no alternative */
1296        {        {
# Line 1305  for (;;) Line 1323  for (;;)
1323      break;      break;
1324    
1325    
1326      /* 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. */  
1327    
     case OP_ACCEPT:  
1328      case OP_END:      case OP_END:
1329      if (md->recursive != NULL)      case OP_ACCEPT:
1330        {      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. */  
1331    
1332      else if (eptr == mstart &&      /* If we have matched an empty string, fail if not in an assertion and not
1333          (md->notempty ||      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1334            (md->notempty_atstart &&      is set and we have matched at the start of the subject. In both cases,
1335              mstart == md->start_subject + md->start_offset)))      backtracking will then try other alternatives, if any. */
1336    
1337        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1338             md->recursive == NULL &&
1339             (md->notempty ||
1340               (md->notempty_atstart &&
1341                 mstart == md->start_subject + md->start_offset)))
1342        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1343    
1344      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
1345    
1346      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1347      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1348      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 1357  for (;;)
1357      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,
1358      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
1359      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
1360      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
1361      of a condition, we want to return immediately afterwards. The caller of      of a condition, we want to return immediately afterwards. The caller of
1362      this incarnation of the match() function will have set MATCH_CONDASSERT in      this incarnation of the match() function will have set MATCH_CONDASSERT in
1363      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
1364      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
1365      match() to remember this case. */      match() to remember this case. */
1366    
1367      case OP_ASSERT:      case OP_ASSERT:
# Line 1366  for (;;) Line 1371  for (;;)
1371        condassert = TRUE;        condassert = TRUE;
1372        md->match_function_type = 0;        md->match_function_type = 0;
1373        }        }
1374      else condassert = FALSE;      else condassert = FALSE;
1375    
1376      do      do
1377        {        {
1378        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1379        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1380          {          {
1381          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
1382            markptr = md->mark;
1383          break;          break;
1384          }          }
1385        if (rrc != MATCH_NOMATCH &&        if (rrc != MATCH_NOMATCH &&
# Line 1382  for (;;) Line 1388  for (;;)
1388        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1389        }        }
1390      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1391    
1392      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1393    
1394      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
# Line 1408  for (;;) Line 1414  for (;;)
1414        condassert = TRUE;        condassert = TRUE;
1415        md->match_function_type = 0;        md->match_function_type = 0;
1416        }        }
1417      else condassert = FALSE;      else condassert = FALSE;
1418    
1419      do      do
1420        {        {
# Line 1427  for (;;) Line 1433  for (;;)
1433      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1434    
1435      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1436    
1437      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1438      continue;      continue;
1439    
# Line 1472  for (;;) Line 1478  for (;;)
1478      if (pcre_callout != NULL)      if (pcre_callout != NULL)
1479        {        {
1480        pcre_callout_block cb;        pcre_callout_block cb;
1481        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1482        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1483        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1484        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
# Line 1484  for (;;) Line 1490  for (;;)
1490        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1491        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1492        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1493          cb.mark             = markptr;
1494        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1495        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1496        }        }
# Line 1494  for (;;) Line 1501  for (;;)
1501      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
1502      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1503    
1504      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1505      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
1506      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
1507      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
1508      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
1509      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
1510      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.  
1511    
1512      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
1513      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
1514      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1515        a lot, so he is not to blame for the current way it works. */
1516    
1517      case OP_RECURSE:      case OP_RECURSE:
1518        {        {
1519          recursion_info *ri;
1520          int recno;
1521    
1522        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1523        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1524          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1525    
1526          /* Check for repeating a recursion without advancing the subject pointer.
1527          This should catch convoluted mutual recursions. (Some simple cases are
1528          caught at compile time.) */
1529    
1530          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1531            if (recno == ri->group_num && eptr == ri->subject_position)
1532              RRETURN(PCRE_ERROR_RECURSELOOP);
1533    
1534        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1535    
1536          new_recursive.group_num = recno;
1537          new_recursive.subject_position = eptr;
1538        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1539        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1540    
1541        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1542    
1543        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1544    
1545        /* Now save the offset data. */        /* Now save the offset data */
1546    
1547        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1548        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1536  for (;;) Line 1553  for (;;)
1553            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1554          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1555          }          }
   
1556        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1557              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1558        new_recursive.save_offset_top = offset_top;  
1559          /* OK, now we can do the recursion. After processing each alternative,
1560        /* OK, now we can do the recursion. For each top-level alternative we        restore the offset data. If there were nested recursions, md->recursive
1561        restore the offset and recursion data. */        might be changed, so reset it before looping. */
1562    
1563        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1564        cbegroup = (*callpat >= OP_SBRA);        cbegroup = (*callpat >= OP_SBRA);
# Line 1551  for (;;) Line 1567  for (;;)
1567          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1568          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1569            md, eptrb, RM6);            md, eptrb, RM6);
1570            memcpy(md->offset_vector, new_recursive.offset_save,
1571                new_recursive.saved_max * sizeof(int));
1572            md->recursive = new_recursive.prevrec;
1573          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1574            {            {
1575            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1576            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1577              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1578            MRRETURN(MATCH_MATCH);  
1579              /* Set where we got to in the subject, and reset the start in case
1580              it was changed by \K. This *is* propagated back out of a recursion,
1581              for Perl compatibility. */
1582    
1583              eptr = md->end_match_ptr;
1584              mstart = md->start_match_ptr;
1585              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1586            }            }
1587          else if (rrc != MATCH_NOMATCH &&          else if (rrc != MATCH_NOMATCH &&
1588                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
# Line 1569  for (;;) Line 1594  for (;;)
1594            }            }
1595    
1596          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1597          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1598          }          }
1599        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1581  for (;;) Line 1604  for (;;)
1604          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1605        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1606        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer, but resetting  
     the start-of-match value in case it was changed by \K. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);  
       if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */  
         {  
         mstart = md->start_match_ptr;  
         break;  
         }  
       if (rrc != MATCH_NOMATCH &&  
           (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
         RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
1607    
1608      if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);      RECURSION_MATCHED:
1609        break;
     /* 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 */  
1610    
1611      /* 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
1612      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1667  for (;;) Line 1620  for (;;)
1620      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
1621      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
1622      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1623    
1624      case OP_BRAZERO:      case OP_BRAZERO:
1625      next = ecode + 1;      next = ecode + 1;
1626      RMATCH(eptr, next, offset_top, md, eptrb, RM10);      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
# Line 1675  for (;;) Line 1628  for (;;)
1628      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
1629      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1630      break;      break;
1631    
1632      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1633      next = ecode + 1;      next = ecode + 1;
1634      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
# Line 1689  for (;;) Line 1642  for (;;)
1642      do next += GET(next,1); while (*next == OP_ALT);      do next += GET(next,1); while (*next == OP_ALT);
1643      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1644      break;      break;
1645    
1646      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1647      here; just jump to the group, with allow_zero set TRUE. */      here; just jump to the group, with allow_zero set TRUE. */
1648    
1649      case OP_BRAPOSZERO:      case OP_BRAPOSZERO:
1650      op = *(++ecode);      op = *(++ecode);
1651      allow_zero = TRUE;      allow_zero = TRUE;
1652      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1653        goto POSSESSIVE_NON_CAPTURE;        goto POSSESSIVE_NON_CAPTURE;
# Line 1704  for (;;) Line 1657  for (;;)
1657      case OP_KET:      case OP_KET:
1658      case OP_KETRMIN:      case OP_KETRMIN:
1659      case OP_KETRMAX:      case OP_KETRMAX:
1660      case OP_KETRPOS:      case OP_KETRPOS:
1661      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1662    
1663      /* 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
1664      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1665      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1666    
1667      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1668        {        {
1669        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1670        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1671        }        }
1672      else saved_eptr = NULL;      else saved_eptr = NULL;
1673    
1674      /* 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
1675      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
1676      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
1677      it was changed by \K. */      by \K. */
1678    
1679      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1680          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1681        {        {
1682        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1683        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1684        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1685        MRRETURN(MATCH_MATCH);        MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1686        }        }
1687    
1688      /* 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
1689      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1690      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
1691      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
1692      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
1693        the current subject position and start match pointer and give a MATCH
1694        return. */
1695    
1696      if (*prev == OP_CBRA || *prev == OP_SCBRA ||      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1697          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
# Line 1750  for (;;) Line 1704  for (;;)
1704        printf("\n");        printf("\n");
1705  #endif  #endif
1706    
1707          /* Handle a recursively called group. */
1708    
1709          if (md->recursive != NULL && md->recursive->group_num == number)
1710            {
1711            md->end_match_ptr = eptr;
1712            md->start_match_ptr = mstart;
1713            RRETURN(MATCH_MATCH);
1714            }
1715    
1716          /* Deal with capturing */
1717    
1718        md->capture_last = number;        md->capture_last = number;
1719        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1720          {          {
1721            /* If offset is greater than offset_top, it means that we are
1722            "skipping" a capturing group, and that group's offsets must be marked
1723            unset. In earlier versions of PCRE, all the offsets were unset at the
1724            start of matching, but this doesn't work because atomic groups and
1725            assertions can cause a value to be set that should later be unset.
1726            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1727            part of the atomic group, but this is not on the final matching path,
1728            so must be unset when 2 is set. (If there is no group 2, there is no
1729            problem, because offset_top will then be 2, indicating no capture.) */
1730    
1731            if (offset > offset_top)
1732              {
1733              register int *iptr = md->offset_vector + offset_top;
1734              register int *iend = md->offset_vector + offset;
1735              while (iptr < iend) *iptr++ = -1;
1736              }
1737    
1738            /* Now make the extraction */
1739    
1740          md->offset_vector[offset] =          md->offset_vector[offset] =
1741            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1742          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1743          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1744          }          }
   
       /* 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;  
         }  
1745        }        }
1746    
1747      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1748      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
1749      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1750      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
1751      course of events. */      processing the rest of the pattern at a lower level. If this results in a
1752        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1753        bypassing intermediate backup points, but resetting any captures that
1754        happened along the way. */
1755    
1756      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1757        {        {
1758        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1759            {
1760            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1761            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1762            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1763            RRETURN(MATCH_ONCE);
1764            }
1765          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1766        break;        break;
1767        }        }
1768    
1769      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1770      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
1771      at a time from the outer level, thus saving stack. */      at a time from the outer level, thus saving stack. */
1772    
1773      if (*ecode == OP_KETRPOS)      if (*ecode == OP_KETRPOS)
1774        {        {
1775        md->end_match_ptr = eptr;        md->end_match_ptr = eptr;
1776        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1777        RRETURN(MATCH_KETRPOS);        RRETURN(MATCH_KETRPOS);
1778        }        }
1779    
1780      /* 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
1781      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
1782      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
1783      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
1784        string. */
1785    
1786      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1787        {        {
1788        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1789        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1790          if (*prev == OP_ONCE)
1791            {
1792            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1793            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1794            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1795            RRETURN(MATCH_ONCE);
1796            }
1797        if (*prev >= OP_SBRA)    /* Could match an empty string */        if (*prev >= OP_SBRA)    /* Could match an empty string */
1798          {          {
1799          md->match_function_type = MATCH_CBEGROUP;          md->match_function_type = MATCH_CBEGROUP;
1800          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1801          RRETURN(rrc);          RRETURN(rrc);
1802          }          }
# Line 1818  for (;;) Line 1805  for (;;)
1805        }        }
1806      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1807        {        {
1808        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1809        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1810          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1811        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1812          if (*prev == OP_ONCE)
1813            {
1814            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1815            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1816            md->once_target = prev;
1817            RRETURN(MATCH_ONCE);
1818            }
1819        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1820        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1821        }        }
# Line 1830  for (;;) Line 1825  for (;;)
1825    
1826      case OP_CIRC:      case OP_CIRC:
1827      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1828    
1829      /* Start of subject assertion */      /* Start of subject assertion */
1830    
1831      case OP_SOD:      case OP_SOD:
1832      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1833      ecode++;      ecode++;
1834      break;      break;
1835    
1836      /* Multiline mode: start of subject unless notbol, or after any newline. */      /* Multiline mode: start of subject unless notbol, or after any newline. */
1837    
1838      case OP_CIRCM:      case OP_CIRCM:
# Line 1876  for (;;) Line 1871  for (;;)
1871      ecode++;      ecode++;
1872      break;      break;
1873    
1874      /* Not multiline mode: assert before a terminating newline or before end of      /* Not multiline mode: assert before a terminating newline or before end of
1875      subject unless noteol is set. */      subject unless noteol is set. */
1876    
1877      case OP_DOLL:      case OP_DOLL:
# Line 2032  for (;;) Line 2027  for (;;)
2027      /* Fall through */      /* Fall through */
2028    
2029      case OP_ALLANY:      case OP_ALLANY:
2030      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2031        {        {                            /* not be updated before SCHECK_PARTIAL. */
2032        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2033        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2034        }        }
2035        eptr++;
2036      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2037      ecode++;      ecode++;
2038      break;      break;
# Line 2045  for (;;) Line 2041  for (;;)
2041      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2042    
2043      case OP_ANYBYTE:      case OP_ANYBYTE:
2044      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2045        {        {                            /* not be updated before SCHECK_PARTIAL. */
2046        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2047        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2048        }        }
2049        eptr++;
2050      ecode++;      ecode++;
2051      break;      break;
2052    
# Line 2165  for (;;) Line 2162  for (;;)
2162      switch(c)      switch(c)
2163        {        {
2164        default: MRRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2165    
2166        case 0x000d:        case 0x000d:
2167        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2168        break;        break;
# Line 2389  for (;;) Line 2386  for (;;)
2386        MRRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2387        }        }
2388      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2389        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2390        while (eptr < md->end_subject)
2391        {        {
2392        int category = UCD_CATEGORY(c);        int len = 1;
2393        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2394        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2395          {        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;  
         }  
2396        }        }
2397      ecode++;      ecode++;
2398      break;      break;
# Line 2418  for (;;) Line 2408  for (;;)
2408      loops). */      loops). */
2409    
2410      case OP_REF:      case OP_REF:
2411      case OP_REFI:      case OP_REFI:
2412      caseless = op == OP_REFI;      caseless = op == OP_REFI;
2413      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2414      ecode += 3;      ecode += 3;
2415    
# Line 2487  for (;;) Line 2477  for (;;)
2477    
2478      for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2479        {        {
2480        int slength;        int slength;
2481        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2482          {          {
2483          CHECK_PARTIAL();          CHECK_PARTIAL();
# Line 2507  for (;;) Line 2497  for (;;)
2497        {        {
2498        for (fi = min;; fi++)        for (fi = min;; fi++)
2499          {          {
2500          int slength;          int slength;
2501          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2502          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2503          if (fi >= max) MRRETURN(MATCH_NOMATCH);          if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 2528  for (;;) Line 2518  for (;;)
2518        pp = eptr;        pp = eptr;
2519        for (i = min; i < max; i++)        for (i = min; i < max; i++)
2520          {          {
2521          int slength;          int slength;
2522          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2523            {            {
2524            CHECK_PARTIAL();            CHECK_PARTIAL();
# Line 3261  for (;;) Line 3251  for (;;)
3251      checking can be multibyte. */      checking can be multibyte. */
3252    
3253      case OP_NOT:      case OP_NOT:
3254      case OP_NOTI:      case OP_NOTI:
3255      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3256        {        {
3257        SCHECK_PARTIAL();        SCHECK_PARTIAL();
# Line 3756  for (;;) Line 3746  for (;;)
3746            case PT_LAMP:            case PT_LAMP:
3747            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3748              {              {
3749                int chartype;
3750              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3751                {                {
3752                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3753                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3754                }                }
3755              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3756              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
3757              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
3758                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
3759                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
3760                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3761              }              }
3762            break;            break;
# Line 3779  for (;;) Line 3770  for (;;)
3770                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3771                }                }
3772              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3773              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
3774                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3775              }              }
3776            break;            break;
# Line 3794  for (;;) Line 3784  for (;;)
3784                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3785                }                }
3786              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3787              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
3788                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3789              }              }
3790            break;            break;
# Line 3809  for (;;) Line 3798  for (;;)
3798                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3799                }                }
3800              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3801              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
3802                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3803              }              }
3804            break;            break;
# Line 3818  for (;;) Line 3806  for (;;)
3806            case PT_ALNUM:            case PT_ALNUM:
3807            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3808              {              {
3809                int category;
3810              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3811                {                {
3812                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3813                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3814                }                }
3815              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3816              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3817              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
3818                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3819              }              }
3820            break;            break;
# Line 3840  for (;;) Line 3828  for (;;)
3828                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3829                }                }
3830              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3831              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 ||  
3832                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
3833                     == prop_fail_result)                     == prop_fail_result)
3834                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3857  for (;;) Line 3844  for (;;)
3844                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3845                }                }
3846              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3847              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||  
3848                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
3849                     == prop_fail_result)                     == prop_fail_result)
3850                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 3868  for (;;) Line 3854  for (;;)
3854            case PT_WORD:            case PT_WORD:
3855            for (i = 1; i <= min; i++)            for (i = 1; i <= min; i++)
3856              {              {
3857                int category;
3858              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
3859                {                {
3860                SCHECK_PARTIAL();                SCHECK_PARTIAL();
3861                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3862                }                }
3863              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
3864              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
3865              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N || c == CHAR_UNDERSCORE)
                  c == CHAR_UNDERSCORE)  
3866                     == prop_fail_result)                     == prop_fail_result)
3867                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
3868              }              }
# Line 3902  for (;;) Line 3888  for (;;)
3888              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3889              }              }
3890            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3891            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
3892            while (eptr < md->end_subject)            while (eptr < md->end_subject)
3893              {              {
3894              int len = 1;              int len = 1;
3895              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
3896                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
3897              eptr += len;              eptr += len;
3898              }              }
3899            }            }
# Line 3968  for (;;) Line 3951  for (;;)
3951            switch(c)            switch(c)
3952              {              {
3953              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
3954    
3955              case 0x000d:              case 0x000d:
3956              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3957              break;              break;
# Line 4245  for (;;) Line 4228  for (;;)
4228            switch(*eptr++)            switch(*eptr++)
4229              {              {
4230              default: MRRETURN(MATCH_NOMATCH);              default: MRRETURN(MATCH_NOMATCH);
4231    
4232              case 0x000d:              case 0x000d:
4233              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
4234              break;              break;
4235    
4236              case 0x000a:              case 0x000a:
4237              break;              break;
4238    
# Line 4455  for (;;) Line 4438  for (;;)
4438            case PT_LAMP:            case PT_LAMP:
4439            for (fi = min;; fi++)            for (fi = min;; fi++)
4440              {              {
4441                int chartype;
4442              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM37);
4443              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4444              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4464  for (;;) Line 4448  for (;;)
4448                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4449                }                }
4450              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4451              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4452              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4453                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4454                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4455                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4456              }              }
4457            /* Control never gets here */            /* Control never gets here */
# Line 4484  for (;;) Line 4468  for (;;)
4468                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4469                }                }
4470              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4471              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result)
             if ((prop_category == prop_value) == prop_fail_result)  
4472                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4473              }              }
4474            /* Control never gets here */            /* Control never gets here */
# Line 4502  for (;;) Line 4485  for (;;)
4485                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4486                }                }
4487              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4488              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result)
             if ((prop_chartype == prop_value) == prop_fail_result)  
4489                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4490              }              }
4491            /* Control never gets here */            /* Control never gets here */
# Line 4520  for (;;) Line 4502  for (;;)
4502                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4503                }                }
4504              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4505              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result)
             if ((prop_script == prop_value) == prop_fail_result)  
4506                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4507              }              }
4508            /* Control never gets here */            /* Control never gets here */
# Line 4529  for (;;) Line 4510  for (;;)
4510            case PT_ALNUM:            case PT_ALNUM:
4511            for (fi = min;; fi++)            for (fi = min;; fi++)
4512              {              {
4513                int category;
4514              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM59);
4515              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4516              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4538  for (;;) Line 4520  for (;;)
4520                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4521                }                }
4522              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4523              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4524              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                    == prop_fail_result)  
4525                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4526              }              }
4527            /* Control never gets here */            /* Control never gets here */
# Line 4557  for (;;) Line 4538  for (;;)
4538                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4539                }                }
4540              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4541              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 ||  
4542                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
4543                     == prop_fail_result)                     == prop_fail_result)
4544                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4577  for (;;) Line 4557  for (;;)
4557                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4558                }                }
4559              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4560              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 ||  
4561                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
4562                     == prop_fail_result)                     == prop_fail_result)
4563                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4588  for (;;) Line 4567  for (;;)
4567            case PT_WORD:            case PT_WORD:
4568            for (fi = min;; fi++)            for (fi = min;; fi++)
4569              {              {
4570                int category;
4571              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM62);
4572              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4573              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
# Line 4597  for (;;) Line 4577  for (;;)
4577                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
4578                }                }
4579              GETCHARINCTEST(c, eptr);              GETCHARINCTEST(c, eptr);
4580              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
4581              if ((prop_category == ucp_L ||              if ((category == ucp_L ||
4582                   prop_category == ucp_N ||                   category == ucp_N ||
4583                   c == CHAR_UNDERSCORE)                   c == CHAR_UNDERSCORE)
4584                     == prop_fail_result)                     == prop_fail_result)
4585                MRRETURN(MATCH_NOMATCH);                MRRETURN(MATCH_NOMATCH);
# Line 4629  for (;;) Line 4609  for (;;)
4609              MRRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
4610              }              }
4611            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
4612            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);  
4613            while (eptr < md->end_subject)            while (eptr < md->end_subject)
4614              {              {
4615              int len = 1;              int len = 1;
4616              if (!utf8) c = *eptr;              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
4617                else { GETCHARLEN(c, eptr, len); }              if (UCD_CATEGORY(c) != ucp_M) break;
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
4618              eptr += len;              eptr += len;
4619              }              }
4620            }            }
4621          }          }
   
4622        else        else
4623  #endif     /* SUPPORT_UCP */  #endif     /* SUPPORT_UCP */
4624    
# Line 4963  for (;;) Line 4939  for (;;)
4939            case PT_LAMP:            case PT_LAMP:
4940            for (i = min; i < max; i++)            for (i = min; i < max; i++)
4941              {              {
4942                int chartype;
4943              int len = 1;              int len = 1;
4944              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
4945                {                {
# Line 4970  for (;;) Line 4947  for (;;)
4947                break;                break;
4948                }                }
4949              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4950              prop_chartype = UCD_CHARTYPE(c);              chartype = UCD_CHARTYPE(c);
4951              if ((prop_chartype == ucp_Lu ||              if ((chartype == ucp_Lu ||
4952                   prop_chartype == ucp_Ll ||                   chartype == ucp_Ll ||
4953                   prop_chartype == ucp_Lt) == prop_fail_result)                   chartype == ucp_Lt) == prop_fail_result)
4954                break;                break;
4955              eptr+= len;              eptr+= len;
4956              }              }
# Line 4989  for (;;) Line 4966  for (;;)
4966                break;                break;
4967                }                }
4968              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4969              prop_category = UCD_CATEGORY(c);              if ((UCD_CATEGORY(c) == prop_value) == prop_fail_result) break;
             if ((prop_category == prop_value) == prop_fail_result)  
               break;  
4970              eptr+= len;              eptr+= len;
4971              }              }
4972            break;            break;
# Line 5006  for (;;) Line 4981  for (;;)
4981                break;                break;
4982                }                }
4983              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4984              prop_chartype = UCD_CHARTYPE(c);              if ((UCD_CHARTYPE(c) == prop_value) == prop_fail_result) break;
             if ((prop_chartype == prop_value) == prop_fail_result)  
               break;  
4985              eptr+= len;              eptr+= len;
4986              }              }
4987            break;            break;
# Line 5023  for (;;) Line 4996  for (;;)
4996                break;                break;
4997                }                }
4998              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
4999              prop_script = UCD_SCRIPT(c);              if ((UCD_SCRIPT(c) == prop_value) == prop_fail_result) break;
             if ((prop_script == prop_value) == prop_fail_result)  
               break;  
5000              eptr+= len;              eptr+= len;
5001              }              }
5002            break;            break;
# Line 5033  for (;;) Line 5004  for (;;)
5004            case PT_ALNUM:            case PT_ALNUM:
5005            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5006              {              {
5007                int category;
5008              int len = 1;              int len = 1;
5009              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5010                {                {
# Line 5040  for (;;) Line 5012  for (;;)
5012                break;                break;
5013                }                }
5014              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5015              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5016              if ((prop_category == ucp_L || prop_category == ucp_N)              if ((category == ucp_L || category == ucp_N) == prop_fail_result)
                  == prop_fail_result)  
5017                break;                break;
5018              eptr+= len;              eptr+= len;
5019              }              }
# Line 5058  for (;;) Line 5029  for (;;)
5029                break;                break;
5030                }                }
5031              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5032              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 ||  
5033                   c == CHAR_FF || c == CHAR_CR)                   c == CHAR_FF || c == CHAR_CR)
5034                   == prop_fail_result)                   == prop_fail_result)
5035                break;                break;
# Line 5077  for (;;) Line 5047  for (;;)
5047                break;                break;
5048                }                }
5049              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5050              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 ||  
5051                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)                   c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
5052                   == prop_fail_result)                   == prop_fail_result)
5053                break;                break;
# Line 5089  for (;;) Line 5058  for (;;)
5058            case PT_WORD:            case PT_WORD:
5059            for (i = min; i < max; i++)            for (i = min; i < max; i++)
5060              {              {
5061                int category;
5062              int len = 1;              int len = 1;
5063              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
5064                {                {
# Line 5096  for (;;) Line 5066  for (;;)
5066                break;                break;
5067                }                }
5068              GETCHARLENTEST(c, eptr, len);              GETCHARLENTEST(c, eptr, len);
5069              prop_category = UCD_CATEGORY(c);              category = UCD_CATEGORY(c);
5070              if ((prop_category == ucp_L || prop_category == ucp_N ||              if ((category == ucp_L || category == ucp_N ||
5071                   c == CHAR_UNDERSCORE) == prop_fail_result)                   c == CHAR_UNDERSCORE) == prop_fail_result)
5072                break;                break;
5073              eptr+= len;              eptr+= len;
# Line 5127  for (;;) Line 5097  for (;;)
5097          {          {
5098          for (i = min; i < max; i++)          for (i = min; i < max; i++)
5099            {            {
5100              int len = 1;
5101            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
5102              {              {
5103              SCHECK_PARTIAL();              SCHECK_PARTIAL();
5104              break;              break;
5105              }              }
5106            GETCHARINCTEST(c, eptr);            if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5107            prop_category = UCD_CATEGORY(c);            if (UCD_CATEGORY(c) == ucp_M) break;
5108            if (prop_category == ucp_M) break;            eptr += len;
5109            while (eptr < md->end_subject)            while (eptr < md->end_subject)
5110              {              {
5111              int len = 1;              len = 1;
5112              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
5113                {              if (UCD_CATEGORY(c) != ucp_M) break;
               GETCHARLEN(c, eptr, len);  
               }  
             prop_category = UCD_CATEGORY(c);  
             if (prop_category != ucp_M) break;  
5114              eptr += len;              eptr += len;
5115              }              }
5116            }            }
# Line 5159  for (;;) Line 5126  for (;;)
5126            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5127            for (;;)                        /* Move back over one extended */            for (;;)                        /* Move back over one extended */
5128              {              {
             int len = 1;  
5129              if (!utf8) c = *eptr; else              if (!utf8) c = *eptr; else
5130                {                {
5131                BACKCHAR(eptr);                BACKCHAR(eptr);
5132                GETCHARLEN(c, eptr, len);                GETCHAR(c, eptr);
5133                }                }
5134              prop_category = UCD_CATEGORY(c);              if (UCD_CATEGORY(c) != ucp_M) break;
             if (prop_category != ucp_M) break;  
5135              eptr--;              eptr--;
5136              }              }
5137            }            }
# Line 5230  for (;;) Line 5195  for (;;)
5195                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;                while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
5196                }                }
5197              }              }
5198            else eptr = md->end_subject;   /* Unlimited UTF-8 repeat */            else
5199                {
5200                eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
5201                SCHECK_PARTIAL();
5202                }
5203            break;            break;
5204    
5205            /* The byte case is the same as non-UTF8 */            /* The byte case is the same as non-UTF8 */
# Line 5451  for (;;) Line 5420  for (;;)
5420            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5421            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
5422            BACKCHAR(eptr);            BACKCHAR(eptr);
5423            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5424                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5425            }            }
5426          }          }
# Line 5663  for (;;) Line 5632  for (;;)
5632            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM47);
5633            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
5634            eptr--;            eptr--;
5635            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&            if (ctype == OP_ANYNL && eptr > pp  && *eptr == '\n' &&
5636                eptr[-1] == '\r') eptr--;                eptr[-1] == '\r') eptr--;
5637            }            }
5638          }          }
# Line 5703  switch (frame->Xwhere) Line 5672  switch (frame->Xwhere)
5672    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)
5673    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)
5674    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)
5675    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)
5676  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5677    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)
5678    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 5774  pcre_exec(const pcre *argument_re, const
5774    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
5775    int offsetcount)    int offsetcount)
5776  {  {
5777  int rc, resetcount, ocount;  int rc, ocount, arg_offset_max;
5778  int first_byte = -1;  int first_byte = -1;
5779  int req_byte = -1;  int req_byte = -1;
5780  int req_byte2 = -1;  int req_byte2 = -1;
# Line 5841  if (re == NULL || subject == NULL || Line 5810  if (re == NULL || subject == NULL ||
5810  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;  if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
5811  if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;  if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
5812    
5813  /* This information is for finding all the numbers associated with a given  /* These two settings are used in the code for checking a UTF-8 string that
5814  name, for condition testing. */  follows immediately afterwards. Other values in the md block are used only
5815    during "normal" pcre_exec() processing, not when the JIT support is in use,
5816    so they are set up later. */
5817    
5818    utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
5819    md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
5820                  ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
5821    
5822    /* Check a UTF-8 string if required. Pass back the character offset and error
5823    code for an invalid string if a results vector is available. */
5824    
5825    #ifdef SUPPORT_UTF8
5826    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
5827      {
5828      int erroroffset;
5829      int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);
5830      if (errorcode != 0)
5831        {
5832        if (offsetcount >= 2)
5833          {
5834          offsets[0] = erroroffset;
5835          offsets[1] = errorcode;
5836          }
5837        return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?
5838          PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
5839        }
5840    
5841      /* Check that a start_offset points to the start of a UTF-8 character. */
5842      if (start_offset > 0 && start_offset < length &&
5843          (((USPTR)subject)[start_offset] & 0xc0) == 0x80)
5844        return PCRE_ERROR_BADUTF8_OFFSET;
5845      }
5846    #endif
5847    
5848    /* If the pattern was successfully studied with JIT support, run the JIT
5849    executable instead of the rest of this function. Most options must be set at
5850    compile time for the JIT code to be usable. Fallback to the normal code path if
5851    an unsupported flag is set. In particular, JIT does not support partial
5852    matching. */
5853    
5854    #ifdef SUPPORT_JIT
5855    if (extra_data != NULL
5856        && (extra_data->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0
5857        && extra_data->executable_jit != NULL
5858        && (options & ~(PCRE_NO_UTF8_CHECK | PCRE_NOTBOL | PCRE_NOTEOL |
5859                        PCRE_NOTEMPTY | PCRE_NOTEMPTY_ATSTART)) == 0)
5860      return _pcre_jit_exec(re, extra_data->executable_jit, subject, length,
5861        start_offset, options, ((extra_data->flags & PCRE_EXTRA_MATCH_LIMIT) == 0)
5862        ? MATCH_LIMIT : extra_data->match_limit, offsets, offsetcount);
5863    #endif
5864    
5865    /* Carry on with non-JIT matching. This information is for finding all the
5866    numbers associated with a given name, for condition testing. */
5867    
5868  md->name_table = (uschar *)re + re->name_table_offset;  md->name_table = (uschar *)re + re->name_table_offset;
5869  md->name_count = re->name_count;  md->name_count = re->name_count;
# Line 5909  md->end_subject = md->start_subject + le Line 5930  md->end_subject = md->start_subject + le
5930  end_subject = md->end_subject;  end_subject = md->end_subject;
5931    
5932  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
 utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;  
5933  md->use_ucp = (re->options & PCRE_UCP) != 0;  md->use_ucp = (re->options & PCRE_UCP) != 0;
5934  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;  md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
5935    
5936    /* Some options are unpacked into BOOL variables in the hope that testing
5937    them will be faster than individual option bits. */
5938    
5939  md->notbol = (options & PCRE_NOTBOL) != 0;  md->notbol = (options & PCRE_NOTBOL) != 0;
5940  md->noteol = (options & PCRE_NOTEOL) != 0;  md->noteol = (options & PCRE_NOTEOL) != 0;
5941  md->notempty = (options & PCRE_NOTEMPTY) != 0;  md->notempty = (options & PCRE_NOTEMPTY) != 0;
5942  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;  md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
5943  md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :  
               ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;  
5944  md->hitend = FALSE;  md->hitend = FALSE;
5945  md->mark = NULL;                        /* In case never set */  md->mark = NULL;                        /* In case never set */
5946    
# Line 6000  defined (though never set). So there's n Line 6022  defined (though never set). So there's n
6022  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)  if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
6023    return PCRE_ERROR_BADPARTIAL;    return PCRE_ERROR_BADPARTIAL;
6024    
 /* Check a UTF-8 string if required. Pass back the character offset and error  
 code for an invalid string if a results vector is available. */  
   
 #ifdef SUPPORT_UTF8  
 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)  
   {  
   int erroroffset;  
   int errorcode = _pcre_valid_utf8((USPTR)subject, length, &erroroffset);  
   if (errorcode != 0)  
     {  
     if (offsetcount >= 2)  
       {  
       offsets[0] = erroroffset;  
       offsets[1] = errorcode;  
       }  
     return (errorcode <= PCRE_UTF8_ERR5 && md->partial > 1)?  
       PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;  
     }  
   
   /* Check that a start_offset points to the start of a UTF-8 character. */  
   
   if (start_offset > 0 && start_offset < length &&  
       (((USPTR)subject)[start_offset] & 0xc0) == 0x80)  
     return PCRE_ERROR_BADUTF8_OFFSET;  
   }  
 #endif  
   
6025  /* If the expression has got more back references than the offsets supplied can  /* If the expression has got more back references than the offsets supplied can
6026  hold, we get a temporary chunk of working store to use during the matching.  hold, we get a temporary chunk of working store to use during the matching.
6027  Otherwise, we can use the vector supplied, rounding down its size to a multiple  Otherwise, we can use the vector supplied, rounding down its size to a multiple
6028  of 3. */  of 3. */
6029    
6030  ocount = offsetcount - (offsetcount % 3);  ocount = offsetcount - (offsetcount % 3);
6031    arg_offset_max = (2*ocount)/3;
6032    
6033  if (re->top_backref > 0 && re->top_backref >= ocount/3)  if (re->top_backref > 0 && re->top_backref >= ocount/3)
6034    {    {
# Line 6049  md->offset_max = (2*ocount)/3; Line 6045  md->offset_max = (2*ocount)/3;
6045  md->offset_overflow = FALSE;  md->offset_overflow = FALSE;
6046  md->capture_last = -1;  md->capture_last = -1;
6047    
 /* 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;  
   
6048  /* Reset the working variable associated with each extraction. These should  /* Reset the working variable associated with each extraction. These should
6049  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
6050  initialize them to avoid reading uninitialized locations. */  initialize them to avoid reading uninitialized locations. Also, unset the
6051    offsets for the matched string. This is really just for tidiness with callouts,
6052    in case they inspect these fields. */
6053    
6054  if (md->offset_vector != NULL)  if (md->offset_vector != NULL)
6055    {    {
6056    register int *iptr = md->offset_vector + ocount;    register int *iptr = md->offset_vector + ocount;
6057    register int *iend = iptr - resetcount/2 + 1;    register int *iend = iptr - re->top_bracket;
6058      if (iend < md->offset_vector + 2) iend = md->offset_vector + 2;
6059    while (--iptr >= iend) *iptr = -1;    while (--iptr >= iend) *iptr = -1;
6060      md->offset_vector[0] = md->offset_vector[1] = -1;
6061    }    }
6062    
6063  /* 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 6091  if ((re->flags & PCRE_REQCHSET) != 0)
6091    }    }
6092    
6093    
6094    
6095    
6096  /* ==========================================================================*/  /* ==========================================================================*/
6097    
6098  /* Loop for handling unanchored repeated matching attempts; for anchored regexs  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
# Line 6108  for(;;) Line 6103  for(;;)
6103    USPTR save_end_subject = end_subject;    USPTR save_end_subject = end_subject;
6104    USPTR new_start_match;    USPTR new_start_match;
6105    
   /* 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;  
     }  
   
6106    /* 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
6107    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
6108    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 6291  for(;;)
6291    md->start_match_ptr = start_match;    md->start_match_ptr = start_match;
6292    md->start_used_ptr = start_match;    md->start_used_ptr = start_match;
6293    md->match_call_count = 0;    md->match_call_count = 0;
6294    md->match_function_type = 0;    md->match_function_type = 0;
6295      md->end_offset_top = 0;
6296    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);
6297    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;    if (md->hitend && start_partial == NULL) start_partial = md->start_used_ptr;
6298    
# Line 6416  if (rc == MATCH_MATCH || rc == MATCH_ACC Line 6403  if (rc == MATCH_MATCH || rc == MATCH_ACC
6403    {    {
6404    if (using_temporary_offsets)    if (using_temporary_offsets)
6405      {      {
6406      if (offsetcount >= 4)      if (arg_offset_max >= 4)
6407        {        {
6408        memcpy(offsets + 2, md->offset_vector + 2,        memcpy(offsets + 2, md->offset_vector + 2,
6409          (offsetcount - 2) * sizeof(int));          (arg_offset_max - 2) * sizeof(int));
6410        DPRINTF(("Copied offsets from temporary memory\n"));        DPRINTF(("Copied offsets from temporary memory\n"));
6411        }        }
6412      if (md->end_offset_top > offsetcount) md->offset_overflow = TRUE;      if (md->end_offset_top > arg_offset_max) md->offset_overflow = TRUE;
6413      DPRINTF(("Freeing temporary memory\n"));      DPRINTF(("Freeing temporary memory\n"));
6414      (pcre_free)(md->offset_vector);      (pcre_free)(md->offset_vector);
6415      }      }
6416    
6417    /* Set the return code to the number of captured strings, or 0 if there are    /* Set the return code to the number of captured strings, or 0 if there were
6418    too many to fit into the vector. */    too many to fit into the vector. */
6419    
6420    rc = md->offset_overflow? 0 : md->end_offset_top/2;    rc = (md->offset_overflow && md->end_offset_top >= arg_offset_max)?
6421        0 : md->end_offset_top/2;
6422    
6423      /* If there is space in the offset vector, set any unused pairs at the end of
6424      the pattern to -1 for backwards compatibility. It is documented that this
6425      happens. In earlier versions, the whole set of potential capturing offsets
6426      was set to -1 each time round the loop, but this is handled differently now.
6427      "Gaps" are set to -1 dynamically instead (this fixes a bug). Thus, it is only
6428      those at the end that need unsetting here. We can't just unset them all at
6429      the start of the whole thing because they may get set in one branch that is
6430      not the final matching branch. */
6431    
6432      if (md->end_offset_top/2 <= re->top_bracket && offsets != NULL)
6433        {
6434        register int *iptr, *iend;
6435        int resetcount = 2 + re->top_bracket * 2;
6436        if (resetcount > offsetcount) resetcount = ocount;
6437        iptr = offsets + md->end_offset_top;
6438        iend = offsets + resetcount;
6439        while (iptr < iend) *iptr++ = -1;
6440        }
6441    
6442    /* 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
6443    md->start_match_ptr might be modified if \K was encountered on the success    md->start_match_ptr might be modified if \K was encountered on the success

Legend:
Removed from v.608  
changed lines
  Added in v.699

  ViewVC Help
Powered by ViewVC 1.1.5