/[pcre]/code/trunk/pcre_exec.c

Diff of /code/trunk/pcre_exec.c

Parent Directory | Revision Log | View Patch Patch

-revision 123 by ph10,
Mon Mar 12 15:19:06 2007 UTC
+revision 534 by ph10,
Thu Jun  3 18:26:05 2010 UTC
 Line 6
  and semantics are as close as possible to those of the Perl 5 language.
                         Written by Philip Hazel
-            Copyright (c) 1997-2007 University of Cambridge
+            Copyright (c) 1997-2010 University of Cambridge
  -----------------------------------------------------------------------------
  Redistribution and use in source and binary forms, with or without
 Line 42 
 POSSIBILITY OF SUCH DAMAGE.
  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
  possible. There are also some static supporting functions. */
+ #ifdef HAVE_CONFIG_H
+ #include "config.h"
+ #endif
  #define NLBLOCK md             /* Block containing newline information */
  #define PSSTART start_subject  /* Field containing processed string start */
  #define PSEND   end_subject    /* Field containing processed string end */
  #include "pcre_internal.h"
- /* The chain of eptrblocks for tail recursions uses memory in stack workspace,
+ /* Undefine some potentially clashing cpp symbols */
- obtained at top level, the size of which is defined by EPTR_WORK_SIZE. */
- #define EPTR_WORK_SIZE (1000)
+ #undef min
+ #undef max
  /* Flag bits for the match() function */
  #define match_condassert     0x01  /* Called to check a condition assertion */
  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */
- #define match_tail_recursed  0x04  /* Tail recursive call */
  /* Non-error returns from the match() function. Error returns are externally
  defined PCRE_ERROR_xxx codes, which are all negative. */
-Line 65 
 defined PCRE_ERROR_xxx codes, which are
+Line 68 
 defined PCRE_ERROR_xxx codes, which are
  #define MATCH_MATCH        1
  #define MATCH_NOMATCH      0
+ /* Special internal returns from the match() function. Make them sufficiently
+ negative to avoid the external error codes. */
+ #define MATCH_ACCEPT       (-999)
+ #define MATCH_COMMIT       (-998)
+ #define MATCH_PRUNE        (-997)
+ #define MATCH_SKIP         (-996)
+ #define MATCH_SKIP_ARG     (-995)
+ #define MATCH_THEN         (-994)
+ /* This is a convenience macro for code that occurs many times. */
+ #define MRRETURN(ra) \
+   { \
+   md->mark = markptr; \
+   RRETURN(ra); \
+   }
  /* Maximum number of ints of offset to save on the stack for recursive calls.
  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
  because the offset vector is always a multiple of 3 long. */
-Line 78 
 static const char rep_max[] = { 0, 0, 0,
+Line 99 
 static const char rep_max[] = { 0, 0, 0,
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
  /*************************************************
  *        Debugging function to print chars       *
  *************************************************/
-Line 130 
 match_ref(int offset, register USPTR ept
+Line 151 
 match_ref(int offset, register USPTR ept
  {
  USPTR p = md->start_subject + md->offset_vector[offset];
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
  if (eptr >= md->end_subject)
    printf("matching subject <null>");
  else
-Line 147 
 printf("\n");
+Line 168 
 printf("\n");
  if (length > md->end_subject - eptr) return FALSE;
- /* Separate the caselesss case for speed */
+ /* Separate the caseless case for speed. In UTF-8 mode we can only do this
+ properly if Unicode properties are supported. Otherwise, we can check only
+ ASCII characters. */
  if ((ims & PCRE_CASELESS) != 0)
    {
+ #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UCP
+   if (md->utf8)
+     {
+     USPTR endptr = eptr + length;
+     while (eptr < endptr)
+       {
+       int c, d;
+       GETCHARINC(c, eptr);
+       GETCHARINC(d, p);
+       if (c != d && c != UCD_OTHERCASE(d)) return FALSE;
+       }
+     }
+   else
+ #endif
+ #endif
+   /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
+   is no UCP support. */
    while (length-- > 0)
-     if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;
+     { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }
    }
+ /* In the caseful case, we can just compare the bytes, whether or not we
+ are in UTF-8 mode. */
  else
    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }
-Line 183 
 calls by keeping local variables that ne
+Line 230 
 calls by keeping local variables that ne
  obtained from malloc() instead instead of on the stack. Macros are used to
  achieve this so that the actual code doesn't look very different to what it
  always used to.
+ The original heap-recursive code used longjmp(). However, it seems that this
+ can be very slow on some operating systems. Following a suggestion from Stan
+ Switzer, the use of longjmp() has been abolished, at the cost of having to
+ provide a unique number for each call to RMATCH. There is no way of generating
+ a sequence of numbers at compile time in C. I have given them names, to make
+ them stand out more clearly.
+ Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
+ FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
+ tests. Furthermore, not using longjmp() means that local dynamic variables
+ don't have indeterminate values; this has meant that the frame size can be
+ reduced because the result can be "passed back" by straight setting of the
+ variable instead of being passed in the frame.
  ****************************************************************************
  ***************************************************************************/
+ /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
+ below must be updated in sync.  */
+ enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
+        RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
+        RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
+        RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
+        RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
+        RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
+        RM61,  RM62 };
  /* These versions of the macros use the stack, as normal. There are debugging
- versions and production versions. */
+ versions and production versions. Note that the "rw" argument of RMATCH isn't
+ actually used in this definition. */
  #ifndef NO_RECURSE
  #define REGISTER register
- #ifdef DEBUG
- #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \
+ #ifdef PCRE_DEBUG
+ #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \
    { \
    printf("match() called in line %d\n", __LINE__); \
-   rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \
+   rrc = match(ra,rb,mstart,markptr,rc,rd,re,rf,rg,rdepth+1); \
    printf("to line %d\n", __LINE__); \
    }
  #define RRETURN(ra) \
-Line 205 
 versions and production versions. */
+Line 278 
 versions and production versions. */
    return ra; \
    }
  #else
- #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \
+ #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \
-   rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)
+   rrc = match(ra,rb,mstart,markptr,rc,rd,re,rf,rg,rdepth+1)
  #define RRETURN(ra) return ra
  #endif
  #else
- /* These versions of the macros manage a private stack on the heap. Note
+ /* These versions of the macros manage a private stack on the heap. Note that
- that the rd argument of RMATCH isn't actually used. It's the md argument of
+ the "rd" argument of RMATCH isn't actually used in this definition. It's the md
- match(), which never changes. */
+ argument of match(), which never changes. */
  #define REGISTER
- #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\
+ #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\
    {\
    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\
-   if (setjmp(frame->Xwhere) == 0)\
+   if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
-     {\
+   frame->Xwhere = rw; \
-     newframe->Xeptr = ra;\
+   newframe->Xeptr = ra;\
-     newframe->Xecode = rb;\
+   newframe->Xecode = rb;\
-     newframe->Xoffset_top = rc;\
+   newframe->Xmstart = mstart;\
-     newframe->Xims = re;\
+   newframe->Xmarkptr = markptr;\
-     newframe->Xeptrb = rf;\
+   newframe->Xoffset_top = rc;\
-     newframe->Xflags = rg;\
+   newframe->Xims = re;\
-     newframe->Xrdepth = frame->Xrdepth + 1;\
+   newframe->Xeptrb = rf;\
-     newframe->Xprevframe = frame;\
+   newframe->Xflags = rg;\
-     frame = newframe;\
+   newframe->Xrdepth = frame->Xrdepth + 1;\
-     DPRINTF(("restarting from line %d\n", __LINE__));\
+   newframe->Xprevframe = frame;\
-     goto HEAP_RECURSE;\
+   frame = newframe;\
-     }\
+   DPRINTF(("restarting from line %d\n", __LINE__));\
-   else\
+   goto HEAP_RECURSE;\
-     {\
+   L_##rw:\
-     DPRINTF(("longjumped back to line %d\n", __LINE__));\
+   DPRINTF(("jumped back to line %d\n", __LINE__));\
-     frame = md->thisframe;\
-     rx = frame->Xresult;\
-     }\
    }
  #define RRETURN(ra)\
    {\
-   heapframe *newframe = frame;\
+   heapframe *oldframe = frame;\
-   frame = newframe->Xprevframe;\
+   frame = oldframe->Xprevframe;\
-   (pcre_stack_free)(newframe);\
+   (pcre_stack_free)(oldframe);\
    if (frame != NULL)\
      {\
-     frame->Xresult = ra;\
+     rrc = ra;\
-     md->thisframe = frame;\
+     goto HEAP_RETURN;\
-     longjmp(frame->Xwhere, 1);\
      }\
    return ra;\
    }
-Line 266 
 typedef struct heapframe {
+Line 335 
 typedef struct heapframe {
    /* Function arguments that may change */
-   const uschar *Xeptr;
+   USPTR Xeptr;
    const uschar *Xecode;
+   USPTR Xmstart;
+   USPTR Xmarkptr;
    int Xoffset_top;
    long int Xims;
    eptrblock *Xeptrb;
-Line 276 
 typedef struct heapframe {
+Line 347 
 typedef struct heapframe {
    /* Function local variables */
-   const uschar *Xcallpat;
+   USPTR Xcallpat;
-   const uschar *Xcharptr;
+ #ifdef SUPPORT_UTF8
-   const uschar *Xdata;
+   USPTR Xcharptr;
-   const uschar *Xnext;
+ #endif
-   const uschar *Xpp;
+   USPTR Xdata;
-   const uschar *Xprev;
+   USPTR Xnext;
-   const uschar *Xsaved_eptr;
+   USPTR Xpp;
+   USPTR Xprev;
+   USPTR Xsaved_eptr;
    recursion_info Xnew_recursive;
-Line 303 
 typedef struct heapframe {
+Line 376 
 typedef struct heapframe {
    uschar Xocchars[8];
  #endif
+   int Xcodelink;
    int Xctype;
    unsigned int Xfc;
    int Xfi;
-Line 318 
 typedef struct heapframe {
+Line 392 
 typedef struct heapframe {
    eptrblock Xnewptrb;
-   /* Place to pass back result, and where to jump back to */
+   /* Where to jump back to */
-   int  Xresult;
+   int Xwhere;
-   jmp_buf Xwhere;
  } heapframe;
-Line 339 
 typedef struct heapframe {
+Line 412 
 typedef struct heapframe {
  /* This function is called recursively in many circumstances. Whenever it
  returns a negative (error) response, the outer incarnation must also return the
- same response.
+ same response. */
- Performance note: It might be tempting to extract commonly used fields from the
+ /* These macros pack up tests that are used for partial matching, and which
- md structure (e.g. utf8, end_subject) into individual variables to improve
+ appears several times in the code. We set the "hit end" flag if the pointer is
+ at the end of the subject and also past the start of the subject (i.e.
+ something has been matched). For hard partial matching, we then return
+ immediately. The second one is used when we already know we are past the end of
+ the subject. */
+ #define CHECK_PARTIAL()\
+   if (md->partial != 0 && eptr >= md->end_subject && eptr > mstart)\
+     {\
+     md->hitend = TRUE;\
+     if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL);\
+     }
+ #define SCHECK_PARTIAL()\
+   if (md->partial != 0 && eptr > mstart)\
+     {\
+     md->hitend = TRUE;\
+     if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL);\
+     }
+ /* Performance note: It might be tempting to extract commonly used fields from
+ the md structure (e.g. utf8, end_subject) into individual variables to improve
  performance. Tests using gcc on a SPARC disproved this; in the first case, it
  made performance worse.
  Arguments:
     eptr        pointer to current character in subject
     ecode       pointer to current position in compiled code
+    mstart      pointer to the current match start position (can be modified
+                  by encountering \K)
+    markptr     pointer to the most recent MARK name, or NULL
     offset_top  current top pointer
     md          pointer to "static" info for the match
     ims         current /i, /m, and /s options
-Line 358 
 Arguments:
+Line 456 
 Arguments:
                   match_condassert - this is an assertion condition
                   match_cbegroup - this is the start of an unlimited repeat
                     group that can match an empty string
-                  match_tail_recursed - this is a tail_recursed group
     rdepth      the recursion depth
  Returns:       MATCH_MATCH if matched            )  these values are >= 0
                 MATCH_NOMATCH if failed to match  )
+                a negative MATCH_xxx value for PRUNE, SKIP, etc
                 a negative PCRE_ERROR_xxx value if aborted by an error condition
                   (e.g. stopped by repeated call or recursion limit)
  */
  static int
- match(REGISTER USPTR eptr, REGISTER const uschar *ecode,
+ match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
-   int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,
+   const uschar *markptr, int offset_top, match_data *md, unsigned long int ims,
-   int flags, unsigned int rdepth)
+   eptrblock *eptrb, int flags, unsigned int rdepth)
  {
  /* These variables do not need to be preserved over recursion in this function,
  so they can be ordinary variables in all cases. Mark some of them with
-Line 382 
 register unsigned int c;   /* Character
+Line 480 
 register unsigned int c;   /* Character
  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
  BOOL minimize, possessive; /* Quantifier options */
+ int condcode;
  /* When recursion is not being used, all "local" variables that have to be
  preserved over calls to RMATCH() are part of a "frame" which is obtained from
-Line 390 
 heap whenever RMATCH() does a "recursion
+Line 489 
 heap whenever RMATCH() does a "recursion
  #ifdef NO_RECURSE
  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));
+ if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
  frame->Xprevframe = NULL;            /* Marks the top level */
  /* Copy in the original argument variables */
  frame->Xeptr = eptr;
  frame->Xecode = ecode;
+ frame->Xmstart = mstart;
+ frame->Xmarkptr = markptr;
  frame->Xoffset_top = offset_top;
  frame->Xims = ims;
  frame->Xeptrb = eptrb;
-Line 410 
 HEAP_RECURSE:
+Line 512 
 HEAP_RECURSE:
  #define eptr               frame->Xeptr
  #define ecode              frame->Xecode
+ #define mstart             frame->Xmstart
+ #define markptr            frame->Xmarkptr
  #define offset_top         frame->Xoffset_top
  #define ims                frame->Xims
  #define eptrb              frame->Xeptrb
-Line 422 
 HEAP_RECURSE:
+Line 526 
 HEAP_RECURSE:
  #define charptr            frame->Xcharptr
  #endif
  #define callpat            frame->Xcallpat
+ #define codelink           frame->Xcodelink
  #define data               frame->Xdata
  #define next               frame->Xnext
  #define pp                 frame->Xpp
-Line 502 
 int oclength;
+Line 607 
 int oclength;
  uschar occhars[8];
  #endif
+ int codelink;
  int ctype;
  int length;
  int max;
-Line 535 
 TAIL_RECURSE:
+Line 641 
 TAIL_RECURSE:
  /* OK, now we can get on with the real code of the function. Recursive calls
  are specified by the macro RMATCH and RRETURN is used to return. When
  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
- and a "return", respectively (possibly with some debugging if DEBUG is
+ and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
  defined). However, RMATCH isn't like a function call because it's quite a
  complicated macro. It has to be used in one particular way. This shouldn't,
  however, impact performance when true recursion is being used. */
+ #ifdef SUPPORT_UTF8
+ utf8 = md->utf8;       /* Local copy of the flag */
+ #else
+ utf8 = FALSE;
+ #endif
  /* First check that we haven't called match() too many times, or that we
  haven't exceeded the recursive call limit. */
-Line 548 
 if (rdepth >= md->match_limit_recursion)
+Line 660 
 if (rdepth >= md->match_limit_recursion)
  original_ims = ims;    /* Save for resetting on ')' */
- #ifdef SUPPORT_UTF8
- utf8 = md->utf8;       /* Local copy of the flag */
- #else
- utf8 = FALSE;
- #endif
  /* At the start of a group with an unlimited repeat that may match an empty
  string, the match_cbegroup flag is set. When this is the case, add the current
  subject pointer to the chain of such remembered pointers, to be checked when we
  hit the closing ket, in order to break infinite loops that match no characters.
- When match() is called in other circumstances, don't add to the chain. If this
+ When match() is called in other circumstances, don't add to the chain. The
- is a tail recursion, use a block from the workspace, as the one on the stack is
+ match_cbegroup flag must NOT be used with tail recursion, because the memory
- already used. */
+ block that is used is on the stack, so a new one may be required for each
+ match(). */
  if ((flags & match_cbegroup) != 0)
    {
-   eptrblock *p;
+   newptrb.epb_saved_eptr = eptr;
-   if ((flags & match_tail_recursed) != 0)
+   newptrb.epb_prev = eptrb;
-     {
+   eptrb = &newptrb;
-     if (md->eptrn >= EPTR_WORK_SIZE) RRETURN(PCRE_ERROR_NULLWSLIMIT);
-     p = md->eptrchain + md->eptrn++;
-     }
-   else p = &newptrb;
-   p->epb_saved_eptr = eptr;
-   p->epb_prev = eptrb;
-   eptrb = p;
    }
  /* Now start processing the opcodes. */
-Line 583 
 for (;;)
+Line 683 
 for (;;)
    minimize = possessive = FALSE;
    op = *ecode;
-   /* For partial matching, remember if we ever hit the end of the subject after
-   matching at least one subject character. */
-   if (md->partial &&
-       eptr >= md->end_subject &&
-       eptr > md->start_match)
-     md->hitend = TRUE;
    switch(op)
      {
+     case OP_MARK:
+     markptr = ecode + 2;
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
+       ims, eptrb, flags, RM55);
+     /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
+     argument, and we must check whether that argument matches this MARK's
+     argument. It is passed back in md->start_match_ptr (an overloading of that
+     variable). If it does match, we reset that variable to the current subject
+     position and return MATCH_SKIP. Otherwise, pass back the return code
+     unaltered. */
+     if (rrc == MATCH_SKIP_ARG &&
+         strcmp((char *)markptr, (char *)(md->start_match_ptr)) == 0)
+       {
+       md->start_match_ptr = eptr;
+       RRETURN(MATCH_SKIP);
+       }
+     if (md->mark == NULL) md->mark = markptr;
+     RRETURN(rrc);
+     case OP_FAIL:
+     MRRETURN(MATCH_NOMATCH);
+     case OP_COMMIT:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+       ims, eptrb, flags, RM52);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     MRRETURN(MATCH_COMMIT);
+     case OP_PRUNE:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+       ims, eptrb, flags, RM51);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     MRRETURN(MATCH_PRUNE);
+     case OP_PRUNE_ARG:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
+       ims, eptrb, flags, RM56);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     md->mark = ecode + 2;
+     RRETURN(MATCH_PRUNE);
+     case OP_SKIP:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+       ims, eptrb, flags, RM53);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     md->start_match_ptr = eptr;   /* Pass back current position */
+     MRRETURN(MATCH_SKIP);
+     case OP_SKIP_ARG:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
+       ims, eptrb, flags, RM57);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     /* Pass back the current skip name by overloading md->start_match_ptr and
+     returning the special MATCH_SKIP_ARG return code. This will either be
+     caught by a matching MARK, or get to the top, where it is treated the same
+     as PRUNE. */
+     md->start_match_ptr = ecode + 2;
+     RRETURN(MATCH_SKIP_ARG);
+     case OP_THEN:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+       ims, eptrb, flags, RM54);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     MRRETURN(MATCH_THEN);
+     case OP_THEN_ARG:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
+       ims, eptrb, flags, RM58);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     md->mark = ecode + 2;
+     RRETURN(MATCH_THEN);
      /* Handle a capturing bracket. If there is space in the offset vector, save
      the current subject position in the working slot at the top of the vector.
      We mustn't change the current values of the data slot, because they may be
-Line 612 
 for (;;)
+Line 781 
 for (;;)
      number = GET2(ecode, 1+LINK_SIZE);
      offset = number << 1;
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
      printf("start bracket %d\n", number);
      printf("subject=");
      pchars(eptr, 16, TRUE, md);
-Line 627 
 for (;;)
+Line 796 
 for (;;)
        save_capture_last = md->capture_last;
        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
-       md->offset_vector[md->offset_end - number] = eptr - md->start_subject;
+       md->offset_vector[md->offset_end - number] =
+         (int)(eptr - md->start_subject);
        flags = (op == OP_SCBRA)? match_cbegroup : 0;
        do
          {
-         RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+         RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
-           ims, eptrb, flags);
+           ims, eptrb, flags, RM1);
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+         if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
          md->capture_last = save_capture_last;
          ecode += GET(ecode, 1);
          }
-Line 646 
 for (;;)
+Line 816 
 for (;;)
        md->offset_vector[offset+1] = save_offset2;
        md->offset_vector[md->offset_end - number] = save_offset3;
+       if (rrc != MATCH_THEN) md->mark = markptr;
        RRETURN(MATCH_NOMATCH);
        }
-     /* Insufficient room for saving captured contents. Treat as a non-capturing
+     /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
-     bracket. */
+     as a non-capturing bracket. */
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
      /* Non-capturing bracket. Loop for all the alternatives. When we get to the
      final alternative within the brackets, we would return the result of a
      recursive call to match() whatever happened. We can reduce stack usage by
-     turning this into a tail recursion. */
+     turning this into a tail recursion, except in the case when match_cbegroup
+     is set.*/
      case OP_BRA:
      case OP_SBRA:
-Line 665 
 for (;;)
+Line 843 
 for (;;)
      flags = (op >= OP_SBRA)? match_cbegroup : 0;
      for (;;)
        {
-       if (ecode[GET(ecode, 1)] != OP_ALT)
+       if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */
          {
-         ecode += _pcre_OP_lengths[*ecode];
+         if (flags == 0)    /* Not a possibly empty group */
-         flags |= match_tail_recursed;
+           {
-         DPRINTF(("bracket 0 tail recursion\n"));
+           ecode += _pcre_OP_lengths[*ecode];
-         goto TAIL_RECURSE;
+           DPRINTF(("bracket 0 tail recursion\n"));
+           goto TAIL_RECURSE;
+           }
+         /* Possibly empty group; can't use tail recursion. */
+         RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
+           eptrb, flags, RM48);
+         if (rrc == MATCH_NOMATCH) md->mark = markptr;
+         RRETURN(rrc);
          }
        /* For non-final alternatives, continue the loop for a NOMATCH result;
        otherwise return. */
-       RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
+       RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
-         eptrb, flags);
+         eptrb, flags, RM2);
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode, 1);
        }
      /* Control never reaches here. */
-Line 691 
 for (;;)
+Line 878 
 for (;;)
      case OP_COND:
      case OP_SCOND:
-     if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */
+     codelink= GET(ecode, 1);
+     /* Because of the way auto-callout works during compile, a callout item is
+     inserted between OP_COND and an assertion condition. */
+     if (ecode[LINK_SIZE+1] == OP_CALLOUT)
        {
-       offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/
+       if (pcre_callout != NULL)
-       condition = md->recursive != NULL &&
+         {
-         (offset == RREF_ANY || offset == md->recursive->group_num);
+         pcre_callout_block cb;
-       ecode += condition? 3 : GET(ecode, 1);
+         cb.version          = 1;   /* Version 1 of the callout block */
+         cb.callout_number   = ecode[LINK_SIZE+2];
+         cb.offset_vector    = md->offset_vector;
+         cb.subject          = (PCRE_SPTR)md->start_subject;
+         cb.subject_length   = (int)(md->end_subject - md->start_subject);
+         cb.start_match      = (int)(mstart - md->start_subject);
+         cb.current_position = (int)(eptr - md->start_subject);
+         cb.pattern_position = GET(ecode, LINK_SIZE + 3);
+         cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
+         cb.capture_top      = offset_top/2;
+         cb.capture_last     = md->capture_last;
+         cb.callout_data     = md->callout_data;
+         if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
+         if (rrc < 0) RRETURN(rrc);
+         }
+       ecode += _pcre_OP_lengths[OP_CALLOUT];
+       }
+     condcode = ecode[LINK_SIZE+1];
+     /* Now see what the actual condition is */
+     if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
+       {
+       if (md->recursive == NULL)                /* Not recursing => FALSE */
+         {
+         condition = FALSE;
+         ecode += GET(ecode, 1);
+         }
+       else
+         {
+         int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
+         condition =  (recno == RREF_ANY || recno == md->recursive->group_num);
+         /* If the test is for recursion into a specific subpattern, and it is
+         false, but the test was set up by name, scan the table to see if the
+         name refers to any other numbers, and test them. The condition is true
+         if any one is set. */
+         if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
+           {
+           uschar *slotA = md->name_table;
+           for (i = 0; i < md->name_count; i++)
+             {
+             if (GET2(slotA, 0) == recno) break;
+             slotA += md->name_entry_size;
+             }
+           /* Found a name for the number - there can be only one; duplicate
+           names for different numbers are allowed, but not vice versa. First
+           scan down for duplicates. */
+           if (i < md->name_count)
+             {
+             uschar *slotB = slotA;
+             while (slotB > md->name_table)
+               {
+               slotB -= md->name_entry_size;
+               if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
+                 {
+                 condition = GET2(slotB, 0) == md->recursive->group_num;
+                 if (condition) break;
+                 }
+               else break;
+               }
+             /* Scan up for duplicates */
+             if (!condition)
+               {
+               slotB = slotA;
+               for (i++; i < md->name_count; i++)
+                 {
+                 slotB += md->name_entry_size;
+                 if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
+                   {
+                   condition = GET2(slotB, 0) == md->recursive->group_num;
+                   if (condition) break;
+                   }
+                 else break;
+                 }
+               }
+             }
+           }
+         /* Chose branch according to the condition */
+         ecode += condition? 3 : GET(ecode, 1);
+         }
        }
-     else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */
+     else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
        {
        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
        condition = offset < offset_top && md->offset_vector[offset] >= 0;
+       /* If the numbered capture is unset, but the reference was by name,
+       scan the table to see if the name refers to any other numbers, and test
+       them. The condition is true if any one is set. This is tediously similar
+       to the code above, but not close enough to try to amalgamate. */
+       if (!condition && condcode == OP_NCREF)
+         {
+         int refno = offset >> 1;
+         uschar *slotA = md->name_table;
+         for (i = 0; i < md->name_count; i++)
+           {
+           if (GET2(slotA, 0) == refno) break;
+           slotA += md->name_entry_size;
+           }
+         /* Found a name for the number - there can be only one; duplicate names
+         for different numbers are allowed, but not vice versa. First scan down
+         for duplicates. */
+         if (i < md->name_count)
+           {
+           uschar *slotB = slotA;
+           while (slotB > md->name_table)
+             {
+             slotB -= md->name_entry_size;
+             if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
+               {
+               offset = GET2(slotB, 0) << 1;
+               condition = offset < offset_top &&
+                 md->offset_vector[offset] >= 0;
+               if (condition) break;
+               }
+             else break;
+             }
+           /* Scan up for duplicates */
+           if (!condition)
+             {
+             slotB = slotA;
+             for (i++; i < md->name_count; i++)
+               {
+               slotB += md->name_entry_size;
+               if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
+                 {
+                 offset = GET2(slotB, 0) << 1;
+                 condition = offset < offset_top &&
+                   md->offset_vector[offset] >= 0;
+                 if (condition) break;
+                 }
+               else break;
+               }
+             }
+           }
+         }
+       /* Chose branch according to the condition */
        ecode += condition? 3 : GET(ecode, 1);
        }
-     else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */
+     else if (condcode == OP_DEF)     /* DEFINE - always false */
        {
        condition = FALSE;
        ecode += GET(ecode, 1);
-Line 718 
 for (;;)
+Line 1058 
 for (;;)
      else
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
-           match_condassert);
+           match_condassert, RM3);
        if (rrc == MATCH_MATCH)
          {
          condition = TRUE;
          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
          }
-       else if (rrc != MATCH_NOMATCH)
+       else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
          {
          RRETURN(rrc);         /* Need braces because of following else */
          }
        else
          {
          condition = FALSE;
-         ecode += GET(ecode, 1);
+         ecode += codelink;
          }
        }
      /* We are now at the branch that is to be obeyed. As there is only one,
-     we can use tail recursion to avoid using another stack frame. If the second
+     we can use tail recursion to avoid using another stack frame, except when
-     alternative doesn't exist, we can just plough on. */
+     match_cbegroup is required for an unlimited repeat of a possibly empty
+     group. If the second alternative doesn't exist, we can just plough on. */
      if (condition || *ecode == OP_ALT)
        {
        ecode += 1 + LINK_SIZE;
-       flags = match_tail_recursed | ((op == OP_SCOND)? match_cbegroup : 0);
+       if (op == OP_SCOND)        /* Possibly empty group */
-       goto TAIL_RECURSE;
+         {
+         RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);
+         RRETURN(rrc);
+         }
+       else                       /* Group must match something */
+         {
+         flags = 0;
+         goto TAIL_RECURSE;
+         }
        }
-     else
+     else                         /* Condition false & no alternative */
        {
        ecode += 1 + LINK_SIZE;
        }
      break;
-     /* End of the pattern. If we are in a top-level recursion, we should
+     /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
-     restore the offsets appropriately and continue from after the call. */
+     to close any currently open capturing brackets. */
+     case OP_CLOSE:
+     number = GET2(ecode, 1);
+     offset = number << 1;
+ #ifdef PCRE_DEBUG
+       printf("end bracket %d at *ACCEPT", number);
+       printf("\n");
+ #endif
+     md->capture_last = number;
+     if (offset >= md->offset_max) md->offset_overflow = TRUE; else
+       {
+       md->offset_vector[offset] =
+         md->offset_vector[md->offset_end - number];
+       md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
+       if (offset_top <= offset) offset_top = offset + 2;
+       }
+     ecode += 3;
+     break;
+     /* End of the pattern, either real or forced. If we are in a top-level
+     recursion, we should restore the offsets appropriately and continue from
+     after the call. */
+     case OP_ACCEPT:
      case OP_END:
      if (md->recursive != NULL && md->recursive->group_num == 0)
        {
-Line 765 
 for (;;)
+Line 1140 
 for (;;)
        md->recursive = rec->prevrec;
        memmove(md->offset_vector, rec->offset_save,
          rec->saved_max * sizeof(int));
-       md->start_match = rec->save_start;
+       offset_top = rec->save_offset_top;
        ims = original_ims;
        ecode = rec->after_call;
        break;
        }
-     /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty
+     /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is
-     string - backtracking will then try other alternatives, if any. */
+     set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of
+     the subject. In both cases, backtracking will then try other alternatives,
+     if any. */
+     if (eptr == mstart &&
+         (md->notempty ||
+           (md->notempty_atstart &&
+             mstart == md->start_subject + md->start_offset)))
+       MRRETURN(MATCH_NOMATCH);
+     /* Otherwise, we have a match. */
+     md->end_match_ptr = eptr;           /* Record where we ended */
+     md->end_offset_top = offset_top;    /* and how many extracts were taken */
+     md->start_match_ptr = mstart;       /* and the start (\K can modify) */
-     if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);
+     /* For some reason, the macros don't work properly if an expression is
-     md->end_match_ptr = eptr;          /* Record where we ended */
+     given as the argument to MRRETURN when the heap is in use. */
-     md->end_offset_top = offset_top;   /* and how many extracts were taken */
-     RRETURN(MATCH_MATCH);
+     rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
+     MRRETURN(rrc);
      /* Change option settings */
-Line 797 
 for (;;)
+Line 1187 
 for (;;)
      case OP_ASSERTBACK:
      do
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
-       if (rrc == MATCH_MATCH) break;
+         RM4);
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
+         {
+         mstart = md->start_match_ptr;   /* In case \K reset it */
+         break;
+         }
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode, 1);
        }
      while (*ecode == OP_ALT);
-     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
+     if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
      /* If checking an assertion for a condition, return MATCH_MATCH. */
-Line 817 
 for (;;)
+Line 1212 
 for (;;)
      offset_top = md->end_offset_top;
      continue;
-     /* Negative assertion: all branches must fail to match */
+     /* Negative assertion: all branches must fail to match. Encountering SKIP,
+     PRUNE, or COMMIT means we must assume failure without checking subsequent
+     branches. */
      case OP_ASSERT_NOT:
      case OP_ASSERTBACK_NOT:
      do
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
-       if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);
+         RM5);
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
+       if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
+         {
+         do ecode += GET(ecode,1); while (*ecode == OP_ALT);
+         break;
+         }
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode,1);
        }
      while (*ecode == OP_ALT);
-Line 848 
 for (;;)
+Line 1251 
 for (;;)
        while (i-- > 0)
          {
          eptr--;
-         if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
+         if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
-         BACKCHAR(eptr)
+         BACKCHAR(eptr);
          }
        }
      else
-Line 859 
 for (;;)
+Line 1262 
 for (;;)
        {
        eptr -= GET(ecode, 1);
-       if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
+       if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
        }
-     /* Skip to next op code */
+     /* Save the earliest consulted character, then skip to next op code */
+     if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
      ecode += 1 + LINK_SIZE;
      break;
-Line 879 
 for (;;)
+Line 1283 
 for (;;)
        cb.callout_number   = ecode[1];
        cb.offset_vector    = md->offset_vector;
        cb.subject          = (PCRE_SPTR)md->start_subject;
-       cb.subject_length   = md->end_subject - md->start_subject;
+       cb.subject_length   = (int)(md->end_subject - md->start_subject);
-       cb.start_match      = md->start_match - md->start_subject;
+       cb.start_match      = (int)(mstart - md->start_subject);
-       cb.current_position = eptr - md->start_subject;
+       cb.current_position = (int)(eptr - md->start_subject);
        cb.pattern_position = GET(ecode, 2);
        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
        cb.capture_top      = offset_top/2;
        cb.capture_last     = md->capture_last;
        cb.callout_data     = md->callout_data;
-       if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);
+       if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
        if (rrc < 0) RRETURN(rrc);
        }
      ecode += 2 + 2*LINK_SIZE;
-Line 942 
 for (;;)
+Line 1346 
 for (;;)
        memcpy(new_recursive.offset_save, md->offset_vector,
              new_recursive.saved_max * sizeof(int));
-       new_recursive.save_start = md->start_match;
+       new_recursive.save_offset_top = offset_top;
-       md->start_match = eptr;
        /* OK, now we can do the recursion. For each top-level alternative we
        restore the offset and recursion data. */
-Line 952 
 for (;;)
+Line 1355 
 for (;;)
        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;
        do
          {
-         RMATCH(rrc, eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
+         RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
-           md, ims, eptrb, flags);
+           md, ims, eptrb, flags, RM6);
-         if (rrc == MATCH_MATCH)
+         if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
            {
            DPRINTF(("Recursion matched\n"));
            md->recursive = new_recursive.prevrec;
            if (new_recursive.offset_save != stacksave)
              (pcre_free)(new_recursive.offset_save);
-           RRETURN(MATCH_MATCH);
+           MRRETURN(MATCH_MATCH);
            }
-         else if (rrc != MATCH_NOMATCH)
+         else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
            {
            DPRINTF(("Recursion gave error %d\n", rrc));
+           if (new_recursive.offset_save != stacksave)
+             (pcre_free)(new_recursive.offset_save);
            RRETURN(rrc);
            }
-Line 979 
 for (;;)
+Line 1384 
 for (;;)
        md->recursive = new_recursive.prevrec;
        if (new_recursive.offset_save != stacksave)
          (pcre_free)(new_recursive.offset_save);
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
        }
      /* Control never reaches here */
-Line 988 
 for (;;)
+Line 1393 
 for (;;)
      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. */
+     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;
-Line 996 
 for (;;)
+Line 1402 
 for (;;)
      do
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);
-         eptrb, 0);
+       if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
-       if (rrc == MATCH_MATCH) break;
+         {
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+         mstart = md->start_match_ptr;
+         break;
+         }
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode,1);
        }
      while (*ecode == OP_ALT);
-Line 1042 
 for (;;)
+Line 1451 
 for (;;)
      if (*ecode == OP_KETRMIN)
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);
        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
        ecode = prev;
-       flags = match_tail_recursed;
+       flags = 0;
        goto TAIL_RECURSE;
        }
      else  /* OP_KETRMAX */
        {
-       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_cbegroup);
+       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);
        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
        ecode += 1 + LINK_SIZE;
-       flags = match_tail_recursed;
+       flags = 0;
        goto TAIL_RECURSE;
        }
      /* Control never gets here */
-Line 1065 
 for (;;)
+Line 1474 
 for (;;)
      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
      break;
-     /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating
+     /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
-     that it may occur zero times. It may repeat infinitely, or not at all -
+     indicating that it may occur zero times. It may repeat infinitely, or not
-     i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper
+     at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
-     repeat limits are compiled as a number of copies, with the optional ones
+     with fixed upper repeat limits are compiled as a number of copies, with the
-     preceded by BRAZERO or BRAMINZERO. */
+     optional ones preceded by BRAZERO or BRAMINZERO. */
      case OP_BRAZERO:
        {
        next = ecode+1;
-       RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);
+       RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);
        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
        do next += GET(next,1); while (*next == OP_ALT);
        ecode = next + 1 + LINK_SIZE;
-Line 1085 
 for (;;)
+Line 1494 
 for (;;)
        {
        next = ecode+1;
        do next += GET(next, 1); while (*next == OP_ALT);
-       RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
+       RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);
        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
        ecode++;
        }
      break;
+     case OP_SKIPZERO:
+       {
+       next = ecode+1;
+       do next += GET(next,1); while (*next == OP_ALT);
+       ecode = next + 1 + LINK_SIZE;
+       }
+     break;
      /* End of a group, repeated or non-repeating. */
      case OP_KET:
-Line 1109 
 for (;;)
+Line 1526 
 for (;;)
        }
      else saved_eptr = NULL;
-     /* If we are at the end of an assertion group, stop matching and return
+     /* If we are at the end of an assertion group or an atomic group, stop
-     MATCH_MATCH, but record the current high water mark for use by positive
+     matching and return MATCH_MATCH, but record the current high water mark for
-     assertions. Do this also for the "once" (atomic) groups. */
+     use by positive assertions. We also need to record the match start in case
+     it was changed by \K. */
      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
-Line 1119 
 for (;;)
+Line 1537 
 for (;;)
        {
        md->end_match_ptr = eptr;      /* For ONCE */
        md->end_offset_top = offset_top;
-       RRETURN(MATCH_MATCH);
+       md->start_match_ptr = mstart;
+       MRRETURN(MATCH_MATCH);
        }
      /* For capturing groups we have to check the group number back at the start
-Line 1133 
 for (;;)
+Line 1552 
 for (;;)
        number = GET2(prev, 1+LINK_SIZE);
        offset = number << 1;
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
        printf("end bracket %d", number);
        printf("\n");
  #endif
-Line 1143 
 for (;;)
+Line 1562 
 for (;;)
          {
          md->offset_vector[offset] =
            md->offset_vector[md->offset_end - number];
-         md->offset_vector[offset+1] = eptr - md->start_subject;
+         md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
          if (offset_top <= offset) offset_top = offset + 2;
          }
-Line 1155 
 for (;;)
+Line 1574 
 for (;;)
          recursion_info *rec = md->recursive;
          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
          md->recursive = rec->prevrec;
-         md->start_match = rec->save_start;
          memcpy(md->offset_vector, rec->offset_save,
            rec->saved_max * sizeof(int));
+         offset_top = rec->save_offset_top;
          ecode = rec->after_call;
          ims = original_ims;
          break;
-Line 1184 
 for (;;)
+Line 1603 
 for (;;)
      /* The repeating kets try the rest of the pattern or restart from the
      preceding bracket, in the appropriate order. In the second case, we can use
-     tail recursion to avoid using another stack frame. */
+     tail recursion to avoid using another stack frame, unless we have an
+     unlimited repeat of a group that can match an empty string. */
      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;
      if (*ecode == OP_KETRMIN)
        {
-       RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);
        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (flags != 0)    /* Could match an empty string */
+         {
+         RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);
+         RRETURN(rrc);
+         }
        ecode = prev;
-       flags |= match_tail_recursed;
        goto TAIL_RECURSE;
        }
      else  /* OP_KETRMAX */
        {
-       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, flags);
+       RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);
        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
        ecode += 1 + LINK_SIZE;
-       flags = match_tail_recursed;
+       flags = 0;
        goto TAIL_RECURSE;
        }
      /* Control never gets here */
-Line 1209 
 for (;;)
+Line 1633 
 for (;;)
      /* Start of subject unless notbol, or after internal newline if multiline */
      case OP_CIRC:
-     if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
+     if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
      if ((ims & PCRE_MULTILINE) != 0)
        {
        if (eptr != md->start_subject &&
            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
        ecode++;
        break;
        }
-Line 1223 
 for (;;)
+Line 1647 
 for (;;)
      /* Start of subject assertion */
      case OP_SOD:
-     if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      /* Start of match assertion */
      case OP_SOM:
-     if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
+     if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
+     ecode++;
+     break;
+     /* Reset the start of match point */
+     case OP_SET_SOM:
+     mstart = eptr;
      ecode++;
      break;
-Line 1241 
 for (;;)
+Line 1672 
 for (;;)
      if ((ims & PCRE_MULTILINE) != 0)
        {
        if (eptr < md->end_subject)
-         { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
+         { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
        else
-         { if (md->noteol) RRETURN(MATCH_NOMATCH); }
+         { if (md->noteol) MRRETURN(MATCH_NOMATCH); }
        ecode++;
        break;
        }
      else
        {
-       if (md->noteol) RRETURN(MATCH_NOMATCH);
+       if (md->noteol) MRRETURN(MATCH_NOMATCH);
        if (!md->endonly)
          {
          if (eptr != md->end_subject &&
              (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
-           RRETURN(MATCH_NOMATCH);
+           MRRETURN(MATCH_NOMATCH);
          ecode++;
          break;
          }
-Line 1264 
 for (;;)
+Line 1695 
 for (;;)
      /* End of subject assertion (\z) */
      case OP_EOD:
-     if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
-Line 1273 
 for (;;)
+Line 1704 
 for (;;)
      case OP_EODN:
      if (eptr != md->end_subject &&
          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
-Line 1285 
 for (;;)
+Line 1716 
 for (;;)
        /* Find out if the previous and current characters are "word" characters.
        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
-       be "non-word" characters. */
+       be "non-word" characters. Remember the earliest consulted character for
+       partial matching. */
  #ifdef SUPPORT_UTF8
        if (utf8)
          {
+         /* Get status of previous character */
          if (eptr == md->start_subject) prev_is_word = FALSE; else
            {
-           const uschar *lastptr = eptr - 1;
+           USPTR lastptr = eptr - 1;
            while((*lastptr & 0xc0) == 0x80) lastptr--;
+           if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
            GETCHAR(c, lastptr);
+ #ifdef SUPPORT_UCP
+           if (md->use_ucp)
+             {
+             if (c == '_') prev_is_word = TRUE; else
+               {
+               int cat = UCD_CATEGORY(c);
+               prev_is_word = (cat == ucp_L || cat == ucp_N);
+               }
+             }
+           else
+ #endif
            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
            }
-         if (eptr >= md->end_subject) cur_is_word = FALSE; else
+         /* Get status of next character */
+         if (eptr >= md->end_subject)
+           {
+           SCHECK_PARTIAL();
+           cur_is_word = FALSE;
+           }
+         else
            {
            GETCHAR(c, eptr);
+ #ifdef SUPPORT_UCP
+           if (md->use_ucp)
+             {
+             if (c == '_') cur_is_word = TRUE; else
+               {
+               int cat = UCD_CATEGORY(c);
+               cur_is_word = (cat == ucp_L || cat == ucp_N);
+               }
+             }
+           else
+ #endif
            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
            }
          }
        else
  #endif
-       /* More streamlined when not in UTF-8 mode */
+       /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
+       consistency with the behaviour of \w we do use it in this case. */
          {
-         prev_is_word = (eptr != md->start_subject) &&
+         /* Get status of previous character */
-           ((md->ctypes[eptr[-1]] & ctype_word) != 0);
-         cur_is_word = (eptr < md->end_subject) &&
+         if (eptr == md->start_subject) prev_is_word = FALSE; else
-           ((md->ctypes[*eptr] & ctype_word) != 0);
+           {
+           if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
+ #ifdef SUPPORT_UCP
+           if (md->use_ucp)
+             {
+             c = eptr[-1];
+             if (c == '_') prev_is_word = TRUE; else
+               {
+               int cat = UCD_CATEGORY(c);
+               prev_is_word = (cat == ucp_L || cat == ucp_N);
+               }
+             }
+           else
+ #endif
+           prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
+           }
+         /* Get status of next character */
+         if (eptr >= md->end_subject)
+           {
+           SCHECK_PARTIAL();
+           cur_is_word = FALSE;
+           }
+         else
+ #ifdef SUPPORT_UCP
+         if (md->use_ucp)
+           {
+           c = *eptr;
+           if (c == '_') cur_is_word = TRUE; else
+             {
+             int cat = UCD_CATEGORY(c);
+             cur_is_word = (cat == ucp_L || cat == ucp_N);
+             }
+           }
+         else
+ #endif
+         cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
          }
        /* Now see if the situation is what we want */
        if ((*ecode++ == OP_WORD_BOUNDARY)?
             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
        }
      break;
      /* Match a single character type; inline for speed */
      case OP_ANY:
-     if ((ims & PCRE_DOTALL) == 0)
+     if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
+     /* Fall through */
+     case OP_ALLANY:
+     if (eptr++ >= md->end_subject)
        {
-       if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
        }
-     if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-     if (utf8)
-       while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
      ecode++;
      break;
-Line 1340 
 for (;;)
+Line 1846 
 for (;;)
      any byte, even newline, independent of the setting of PCRE_DOTALL. */
      case OP_ANYBYTE:
-     if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr++ >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      ecode++;
      break;
      case OP_NOT_DIGIT:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1353 
 for (;;)
+Line 1867 
 for (;;)
  #endif
         (md->ctypes[c] & ctype_digit) != 0
         )
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      case OP_DIGIT:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1366 
 for (;;)
+Line 1884 
 for (;;)
  #endif
         (md->ctypes[c] & ctype_digit) == 0
         )
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      case OP_NOT_WHITESPACE:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1379 
 for (;;)
+Line 1901 
 for (;;)
  #endif
         (md->ctypes[c] & ctype_space) != 0
         )
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      case OP_WHITESPACE:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1392 
 for (;;)
+Line 1918 
 for (;;)
  #endif
         (md->ctypes[c] & ctype_space) == 0
         )
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      case OP_NOT_WORDCHAR:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1405 
 for (;;)
+Line 1935 
 for (;;)
  #endif
         (md->ctypes[c] & ctype_word) != 0
         )
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      case OP_WORDCHAR:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1418 
 for (;;)
+Line 1952 
 for (;;)
  #endif
         (md->ctypes[c] & ctype_word) == 0
         )
-       RRETURN(MATCH_NOMATCH);
+       MRRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      case OP_ANYNL:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      switch(c)
        {
-       default: RRETURN(MATCH_NOMATCH);
+       default: MRRETURN(MATCH_NOMATCH);
        case 0x000d:
        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
        break;
        case 0x000a:
+       break;
        case 0x000b:
        case 0x000c:
        case 0x0085:
        case 0x2028:
        case 0x2029:
+       if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
        break;
        }
      ecode++;
      break;
- #ifdef SUPPORT_UCP
+     case OP_NOT_HSPACE:
-     /* Check the next character by Unicode property. We will get here only
+     if (eptr >= md->end_subject)
-     if the support is in the binary; otherwise a compile-time error occurs. */
+       {
+       SCHECK_PARTIAL();
-     case OP_PROP:
+       MRRETURN(MATCH_NOMATCH);
+       }
+     GETCHARINCTEST(c, eptr);
+     switch(c)
+       {
+       default: break;
+       case 0x09:      /* HT */
+       case 0x20:      /* SPACE */
+       case 0xa0:      /* NBSP */
+       case 0x1680:    /* OGHAM SPACE MARK */
+       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
+       case 0x2000:    /* EN QUAD */
+       case 0x2001:    /* EM QUAD */
+       case 0x2002:    /* EN SPACE */
+       case 0x2003:    /* EM SPACE */
+       case 0x2004:    /* THREE-PER-EM SPACE */
+       case 0x2005:    /* FOUR-PER-EM SPACE */
+       case 0x2006:    /* SIX-PER-EM SPACE */
+       case 0x2007:    /* FIGURE SPACE */
+       case 0x2008:    /* PUNCTUATION SPACE */
+       case 0x2009:    /* THIN SPACE */
+       case 0x200A:    /* HAIR SPACE */
+       case 0x202f:    /* NARROW NO-BREAK SPACE */
+       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
+       case 0x3000:    /* IDEOGRAPHIC SPACE */
+       MRRETURN(MATCH_NOMATCH);
+       }
+     ecode++;
+     break;
+     case OP_HSPACE:
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
+     GETCHARINCTEST(c, eptr);
+     switch(c)
+       {
+       default: MRRETURN(MATCH_NOMATCH);
+       case 0x09:      /* HT */
+       case 0x20:      /* SPACE */
+       case 0xa0:      /* NBSP */
+       case 0x1680:    /* OGHAM SPACE MARK */
+       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
+       case 0x2000:    /* EN QUAD */
+       case 0x2001:    /* EM QUAD */
+       case 0x2002:    /* EN SPACE */
+       case 0x2003:    /* EM SPACE */
+       case 0x2004:    /* THREE-PER-EM SPACE */
+       case 0x2005:    /* FOUR-PER-EM SPACE */
+       case 0x2006:    /* SIX-PER-EM SPACE */
+       case 0x2007:    /* FIGURE SPACE */
+       case 0x2008:    /* PUNCTUATION SPACE */
+       case 0x2009:    /* THIN SPACE */
+       case 0x200A:    /* HAIR SPACE */
+       case 0x202f:    /* NARROW NO-BREAK SPACE */
+       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
+       case 0x3000:    /* IDEOGRAPHIC SPACE */
+       break;
+       }
+     ecode++;
+     break;
+     case OP_NOT_VSPACE:
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
+     GETCHARINCTEST(c, eptr);
+     switch(c)
+       {
+       default: break;
+       case 0x0a:      /* LF */
+       case 0x0b:      /* VT */
+       case 0x0c:      /* FF */
+       case 0x0d:      /* CR */
+       case 0x85:      /* NEL */
+       case 0x2028:    /* LINE SEPARATOR */
+       case 0x2029:    /* PARAGRAPH SEPARATOR */
+       MRRETURN(MATCH_NOMATCH);
+       }
+     ecode++;
+     break;
+     case OP_VSPACE:
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
+     GETCHARINCTEST(c, eptr);
+     switch(c)
+       {
+       default: MRRETURN(MATCH_NOMATCH);
+       case 0x0a:      /* LF */
+       case 0x0b:      /* VT */
+       case 0x0c:      /* FF */
+       case 0x0d:      /* CR */
+       case 0x85:      /* NEL */
+       case 0x2028:    /* LINE SEPARATOR */
+       case 0x2029:    /* PARAGRAPH SEPARATOR */
+       break;
+       }
+     ecode++;
+     break;
+ #ifdef SUPPORT_UCP
+     /* Check the next character by Unicode property. We will get here only
+     if the support is in the binary; otherwise a compile-time error occurs. */
+     case OP_PROP:
      case OP_NOTPROP:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
        {
-       int chartype, script;
+       const ucd_record *prop = GET_UCD(c);
-       int category = _pcre_ucp_findprop(c, &chartype, &script);
        switch(ecode[1])
          {
          case PT_ANY:
-         if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
+         if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
          break;
          case PT_LAMP:
-         if ((chartype == ucp_Lu ||
+         if ((prop->chartype == ucp_Lu ||
-              chartype == ucp_Ll ||
+              prop->chartype == ucp_Ll ||
-              chartype == ucp_Lt) == (op == OP_NOTPROP))
+              prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
-           RRETURN(MATCH_NOMATCH);
+           MRRETURN(MATCH_NOMATCH);
-          break;
+         break;
          case PT_GC:
-         if ((ecode[2] != category) == (op == OP_PROP))
+         if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
-           RRETURN(MATCH_NOMATCH);
+           MRRETURN(MATCH_NOMATCH);
          break;
          case PT_PC:
-         if ((ecode[2] != chartype) == (op == OP_PROP))
+         if ((ecode[2] != prop->chartype) == (op == OP_PROP))
-           RRETURN(MATCH_NOMATCH);
+           MRRETURN(MATCH_NOMATCH);
          break;
          case PT_SC:
-         if ((ecode[2] != script) == (op == OP_PROP))
+         if ((ecode[2] != prop->script) == (op == OP_PROP))
-           RRETURN(MATCH_NOMATCH);
+           MRRETURN(MATCH_NOMATCH);
+         break;
+         /* These are specials */
+         case PT_ALNUM:
+         if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
+              _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
+           MRRETURN(MATCH_NOMATCH);
+         break;
+         case PT_SPACE:    /* Perl space */
+         if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
+              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
+                == (op == OP_NOTPROP))
+           MRRETURN(MATCH_NOMATCH);
+         break;
+         case PT_PXSPACE:  /* POSIX space */
+         if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
+              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
+              c == CHAR_FF || c == CHAR_CR)
+                == (op == OP_NOTPROP))
+           MRRETURN(MATCH_NOMATCH);
+         break;
+         case PT_WORD:
+         if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
+              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
+              c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
+           MRRETURN(MATCH_NOMATCH);
          break;
+         /* This should never occur */
          default:
          RRETURN(PCRE_ERROR_INTERNAL);
          }
-Line 1494 
 for (;;)
+Line 2183 
 for (;;)
      is in the binary; otherwise a compile-time error occurs. */
      case OP_EXTUNI:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
        {
-       int chartype, script;
+       int category = UCD_CATEGORY(c);
-       int category = _pcre_ucp_findprop(c, &chartype, &script);
+       if (category == ucp_M) MRRETURN(MATCH_NOMATCH);
-       if (category == ucp_M) RRETURN(MATCH_NOMATCH);
        while (eptr < md->end_subject)
          {
          int len = 1;
-Line 1507 
 for (;;)
+Line 2199 
 for (;;)
            {
            GETCHARLEN(c, eptr, len);
            }
-         category = _pcre_ucp_findprop(c, &chartype, &script);
+         category = UCD_CATEGORY(c);
          if (category != ucp_M) break;
          eptr += len;
          }
-Line 1528 
 for (;;)
+Line 2220 
 for (;;)
      case OP_REF:
        {
        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
-       ecode += 3;                                 /* Advance past item */
+       ecode += 3;
+       /* If the reference is unset, there are two possibilities:
-       /* If the reference is unset, set the length to be longer than the amount
+       (a) In the default, Perl-compatible state, set the length to be longer
-       of subject left; this ensures that every attempt at a match fails. We
+       than the amount of subject left; this ensures that every attempt at a
-       can't just fail here, because of the possibility of quantifiers with zero
+       match fails. We can't just fail here, because of the possibility of
-       minima. */
+       quantifiers with zero minima.
-       length = (offset >= offset_top || md->offset_vector[offset] < 0)?
+       (b) If the JavaScript compatibility flag is set, set the length to zero
-         md->end_subject - eptr + 1 :
+       so that the back reference matches an empty string.
-         md->offset_vector[offset+1] - md->offset_vector[offset];
+       Otherwise, set the length to the length of what was matched by the
+       referenced subpattern. */
+       if (offset >= offset_top || md->offset_vector[offset] < 0)
+         length = (md->jscript_compat)? 0 : (int)(md->end_subject - eptr + 1);
+       else
+         length = md->offset_vector[offset+1] - md->offset_vector[offset];
        /* Set up for repetition, or handle the non-repeated case */
-Line 1566 
 for (;;)
+Line 2267 
 for (;;)
          break;
          default:               /* No repeat follows */
-         if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);
+         if (!match_ref(offset, eptr, length, md, ims))
+           {
+           CHECK_PARTIAL();
+           MRRETURN(MATCH_NOMATCH);
+           }
          eptr += length;
          continue;              /* With the main loop */
          }
-Line 1582 
 for (;;)
+Line 2287 
 for (;;)
        for (i = 1; i <= min; i++)
          {
-         if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);
+         if (!match_ref(offset, eptr, length, md, ims))
+           {
+           CHECK_PARTIAL();
+           MRRETURN(MATCH_NOMATCH);
+           }
          eptr += length;
          }
-Line 1597 
 for (;;)
+Line 2306 
 for (;;)
          {
          for (fi = min;; fi++)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if (fi >= max || !match_ref(offset, eptr, length, md, ims))
+           if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             RRETURN(MATCH_NOMATCH);
+           if (!match_ref(offset, eptr, length, md, ims))
+             {
+             CHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            eptr += length;
            }
          /* Control never gets here */
-Line 1613 
 for (;;)
+Line 2326 
 for (;;)
          pp = eptr;
          for (i = min; i < max; i++)
            {
-           if (!match_ref(offset, eptr, length, md, ims)) break;
+           if (!match_ref(offset, eptr, length, md, ims))
+             {
+             CHECK_PARTIAL();
+             break;
+             }
            eptr += length;
            }
          while (eptr >= pp)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            eptr -= length;
            }
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
          }
        }
      /* Control never gets here */
      /* Match a bit-mapped character class, possibly repeatedly. This op code is
      used when all the characters in the class have values in the range 0-255,
      and either the matching is caseful, or the characters are in the range
-Line 1683 
 for (;;)
+Line 2398 
 for (;;)
          {
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(c, eptr);
            if (c > 255)
              {
-             if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
+             if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
              }
            else
              {
-             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
+             if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
              }
            }
          }
-Line 1701 
 for (;;)
+Line 2420 
 for (;;)
          {
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            c = *eptr++;
-           if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
+           if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
            }
          }
-Line 1723 
 for (;;)
+Line 2446 
 for (;;)
            {
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(c, eptr);
              if (c > 255)
                {
-               if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
+               if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
                }
              else
                {
-               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
+               if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
                }
              }
            }
-Line 1743 
 for (;;)
+Line 2471 
 for (;;)
            {
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
              c = *eptr++;
-             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
+             if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
              }
            }
          /* Control never gets here */
-Line 1766 
 for (;;)
+Line 2499 
 for (;;)
            for (i = min; i < max; i++)
              {
              int len = 1;
-             if (eptr >= md->end_subject) break;
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               break;
+               }
              GETCHARLEN(c, eptr, len);
              if (c > 255)
                {
-Line 1780 
 for (;;)
+Line 2517 
 for (;;)
              }
            for (;;)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              if (eptr-- == pp) break;        /* Stop if tried at original pos */
              BACKCHAR(eptr);
-Line 1792 
 for (;;)
+Line 2529 
 for (;;)
            {
            for (i = min; i < max; i++)
              {
-             if (eptr >= md->end_subject) break;
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               break;
+               }
              c = *eptr;
              if ((data[c/8] & (1 << (c&7))) == 0) break;
              eptr++;
              }
            while (eptr >= pp)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              eptr--;
              }
            }
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
          }
        }
      /* Control never gets here */
      /* Match an extended character class. This opcode is encountered only
-     in UTF-8 mode, because that's the only time it is compiled. */
+     when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
+     mode, because Unicode properties are supported in non-UTF-8 mode. */
  #ifdef SUPPORT_UTF8
      case OP_XCLASS:
-Line 1853 
 for (;;)
+Line 2595 
 for (;;)
        for (i = 1; i <= min; i++)
          {
-         if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+         if (eptr >= md->end_subject)
-         GETCHARINC(c, eptr);
+           {
-         if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
+           SCHECK_PARTIAL();
+           MRRETURN(MATCH_NOMATCH);
+           }
+         GETCHARINCTEST(c, eptr);
+         if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
          }
        /* If max == min we can continue with the main loop without the
-Line 1870 
 for (;;)
+Line 2616 
 for (;;)
          {
          for (fi = min;; fi++)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (fi >= max) MRRETURN(MATCH_NOMATCH);
-           GETCHARINC(c, eptr);
+           if (eptr >= md->end_subject)
-           if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINCTEST(c, eptr);
+           if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
          }
-Line 1887 
 for (;;)
+Line 2638 
 for (;;)
          for (i = min; i < max; i++)
            {
            int len = 1;
-           if (eptr >= md->end_subject) break;
+           if (eptr >= md->end_subject)
-           GETCHARLEN(c, eptr, len);
+             {
+             SCHECK_PARTIAL();
+             break;
+             }
+           GETCHARLENTEST(c, eptr, len);
            if (!_pcre_xclass(c, data)) break;
            eptr += len;
            }
          for(;;)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            if (eptr-- == pp) break;        /* Stop if tried at original pos */
-           BACKCHAR(eptr)
+           if (utf8) BACKCHAR(eptr);
            }
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
          }
        /* Control never gets here */
-Line 1915 
 for (;;)
+Line 2670 
 for (;;)
        length = 1;
        ecode++;
        GETCHARLEN(fc, ecode, length);
-       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
+       if (length > md->end_subject - eptr)
-       while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
+         {
+         CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
+         MRRETURN(MATCH_NOMATCH);
+         }
+       while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
        }
      else
  #endif
      /* Non-UTF-8 mode */
        {
-       if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);
+       if (md->end_subject - eptr < 1)
-       if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
+         {
+         SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
+         MRRETURN(MATCH_NOMATCH);
+         }
+       if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
        ecode += 2;
        }
      break;
-Line 1939 
 for (;;)
+Line 2702 
 for (;;)
        ecode++;
        GETCHARLEN(fc, ecode, length);
-       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
+       if (length > md->end_subject - eptr)
+         {
+         CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
+         MRRETURN(MATCH_NOMATCH);
+         }
        /* If the pattern character's value is < 128, we have only one byte, and
        can use the fast lookup table. */
        if (fc < 128)
          {
-         if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
+         if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
          }
        /* Otherwise we must pick up the subject character */
-Line 1963 
 for (;;)
+Line 2730 
 for (;;)
          if (fc != dc)
            {
  #ifdef SUPPORT_UCP
-           if (dc != _pcre_ucp_othercase(fc))
+           if (dc != UCD_OTHERCASE(fc))
  #endif
-             RRETURN(MATCH_NOMATCH);
+             MRRETURN(MATCH_NOMATCH);
            }
          }
        }
-Line 1974 
 for (;;)
+Line 2741 
 for (;;)
      /* Non-UTF-8 mode */
        {
-       if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);
+       if (md->end_subject - eptr < 1)
-       if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
+         {
+         SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
+         MRRETURN(MATCH_NOMATCH);
+         }
+       if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
        ecode += 2;
        }
      break;
-Line 2028 
 for (;;)
+Line 2799 
 for (;;)
      case OP_MINQUERY:
      c = *ecode++ - OP_STAR;
      minimize = (c & 1) != 0;
      min = rep_min[c];                 /* Pick up values from tables; */
      max = rep_max[c];                 /* zero for max => infinity */
      if (max == 0) max = INT_MAX;
-     /* Common code for all repeated single-character matches. We can give
+     /* Common code for all repeated single-character matches. */
-     up quickly if there are fewer than the minimum number of characters left in
-     the subject. */
      REPEATCHAR:
  #ifdef SUPPORT_UTF8
-Line 2043 
 for (;;)
+Line 2813 
 for (;;)
        length = 1;
        charptr = ecode;
        GETCHARLEN(fc, ecode, length);
-       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
        ecode += length;
        /* Handle multibyte character matching specially here. There is
-Line 2054 
 for (;;)
+Line 2823 
 for (;;)
  #ifdef SUPPORT_UCP
          unsigned int othercase;
          if ((ims & PCRE_CASELESS) != 0 &&
-             (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)
+             (othercase = UCD_OTHERCASE(fc)) != fc)
            oclength = _pcre_ord2utf8(othercase, occhars);
          else oclength = 0;
  #endif  /* SUPPORT_UCP */
          for (i = 1; i <= min; i++)
            {
-           if (memcmp(eptr, charptr, length) == 0) eptr += length;
+           if (eptr <= md->end_subject - length &&
+             memcmp(eptr, charptr, length) == 0) eptr += length;
  #ifdef SUPPORT_UCP
-           /* Need braces because of following else */
+           else if (oclength > 0 &&
-           else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }
+                    eptr <= md->end_subject - oclength &&
+                    memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
+ #endif  /* SUPPORT_UCP */
            else
              {
-             if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);
+             CHECK_PARTIAL();
-             eptr += oclength;
+             MRRETURN(MATCH_NOMATCH);
              }
- #else   /* without SUPPORT_UCP */
-           else { RRETURN(MATCH_NOMATCH); }
- #endif  /* SUPPORT_UCP */
            }
          if (min == max) continue;
-Line 2081 
 for (;;)
+Line 2850 
 for (;;)
            {
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             if (memcmp(eptr, charptr, length) == 0) eptr += length;
+             if (eptr <= md->end_subject - length &&
+               memcmp(eptr, charptr, length) == 0) eptr += length;
  #ifdef SUPPORT_UCP
-             /* Need braces because of following else */
+             else if (oclength > 0 &&
-             else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }
+                      eptr <= md->end_subject - oclength &&
+                      memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
+ #endif  /* SUPPORT_UCP */
              else
                {
-               if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);
+               CHECK_PARTIAL();
-               eptr += oclength;
+               MRRETURN(MATCH_NOMATCH);
                }
- #else   /* without SUPPORT_UCP */
-             else { RRETURN (MATCH_NOMATCH); }
- #endif  /* SUPPORT_UCP */
              }
            /* Control never gets here */
            }
-Line 2105 
 for (;;)
+Line 2874 
 for (;;)
            pp = eptr;
            for (i = min; i < max; i++)
              {
-             if (eptr > md->end_subject - length) break;
+             if (eptr <= md->end_subject - length &&
-             if (memcmp(eptr, charptr, length) == 0) eptr += length;
+                 memcmp(eptr, charptr, length) == 0) eptr += length;
  #ifdef SUPPORT_UCP
-             else if (oclength == 0) break;
+             else if (oclength > 0 &&
+                      eptr <= md->end_subject - oclength &&
+                      memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
+ #endif  /* SUPPORT_UCP */
              else
                {
-               if (memcmp(eptr, occhars, oclength) != 0) break;
+               CHECK_PARTIAL();
-               eptr += oclength;
+               break;
                }
- #else   /* without SUPPORT_UCP */
-             else break;
- #endif  /* SUPPORT_UCP */
              }
            if (possessive) continue;
            for(;;)
-            {
+             {
-            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);
-            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+             if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-            if (eptr == pp) RRETURN(MATCH_NOMATCH);
+             if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
  #ifdef SUPPORT_UCP
-            eptr--;
+             eptr--;
-            BACKCHAR(eptr);
+             BACKCHAR(eptr);
  #else   /* without SUPPORT_UCP */
-            eptr -= length;
+             eptr -= length;
  #endif  /* SUPPORT_UCP */
-            }
+             }
            }
          /* Control never gets here */
          }
-Line 2144 
 for (;;)
+Line 2914 
 for (;;)
  #endif  /* SUPPORT_UTF8 */
      /* When not in UTF-8 mode, load a single-byte character. */
-       {
-       if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
+     fc = *ecode++;
-       fc = *ecode++;
-       }
      /* The value of fc at this point is always less than 256, though we may or
      may not be in UTF-8 mode. The code is duplicated for the caseless and
-Line 2165 
 for (;;)
+Line 2933 
 for (;;)
        {
        fc = md->lcc[fc];
        for (i = 1; i <= min; i++)
-         if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
+         {
+         if (eptr >= md->end_subject)
+           {
+           SCHECK_PARTIAL();
+           MRRETURN(MATCH_NOMATCH);
+           }
+         if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
+         }
        if (min == max) continue;
        if (minimize)
          {
          for (fi = min;; fi++)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if (fi >= max || eptr >= md->end_subject ||
+           if (fi >= max) MRRETURN(MATCH_NOMATCH);
-               fc != md->lcc[*eptr++])
+           if (eptr >= md->end_subject)
-             RRETURN(MATCH_NOMATCH);
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
          }
-Line 2184 
 for (;;)
+Line 2963 
 for (;;)
          pp = eptr;
          for (i = min; i < max; i++)
            {
-           if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             break;
+             }
+           if (fc != md->lcc[*eptr]) break;
            eptr++;
            }
          if (possessive) continue;
          while (eptr >= pp)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);
            eptr--;
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            }
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
          }
        /* Control never gets here */
        }
-Line 2203 
 for (;;)
+Line 2989 
 for (;;)
      else
        {
-       for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
+       for (i = 1; i <= min; i++)
+         {
+         if (eptr >= md->end_subject)
+           {
+           SCHECK_PARTIAL();
+           MRRETURN(MATCH_NOMATCH);
+           }
+         if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
+         }
        if (min == max) continue;
        if (minimize)
          {
          for (fi = min;; fi++)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if (fi >= max || eptr >= md->end_subject || fc != *eptr++)
+           if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
          }
-Line 2221 
 for (;;)
+Line 3022 
 for (;;)
          pp = eptr;
          for (i = min; i < max; i++)
            {
-           if (eptr >= md->end_subject || fc != *eptr) break;
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             break;
+             }
+           if (fc != *eptr) break;
            eptr++;
            }
          if (possessive) continue;
          while (eptr >= pp)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);
            eptr--;
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            }
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
          }
        }
      /* Control never gets here */
-Line 2240 
 for (;;)
+Line 3047 
 for (;;)
      checking can be multibyte. */
      case OP_NOT:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       MRRETURN(MATCH_NOMATCH);
+       }
      ecode++;
      GETCHARINCTEST(c, eptr);
      if ((ims & PCRE_CASELESS) != 0)
-Line 2249 
 for (;;)
+Line 3060 
 for (;;)
        if (c < 256)
  #endif
        c = md->lcc[c];
-       if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);
+       if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
        }
      else
        {
-       if (*ecode++ == c) RRETURN(MATCH_NOMATCH);
+       if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
        }
      break;
-Line 2317 
 for (;;)
+Line 3128 
 for (;;)
      max = rep_max[c];                 /* zero for max => infinity */
      if (max == 0) max = INT_MAX;
-     /* Common code for all repeated single-byte matches. We can give up quickly
+     /* Common code for all repeated single-byte matches. */
-     if there are fewer than the minimum number of bytes left in the
-     subject. */
      REPEATNOTCHAR:
-     if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
      fc = *ecode++;
      /* The code is duplicated for the caseless and caseful cases, for speed,
-Line 2347 
 for (;;)
+Line 3155 
 for (;;)
          register unsigned int d;
          for (i = 1; i <= min; i++)
            {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(d, eptr);
            if (d < 256) d = md->lcc[d];
-           if (fc == d) RRETURN(MATCH_NOMATCH);
+           if (fc == d) MRRETURN(MATCH_NOMATCH);
            }
          }
        else
-Line 2358 
 for (;;)
+Line 3171 
 for (;;)
        /* Not UTF-8 mode */
          {
          for (i = 1; i <= min; i++)
-           if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
+           }
          }
        if (min == max) continue;
-Line 2372 
 for (;;)
+Line 3192 
 for (;;)
            register unsigned int d;
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM28);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(d, eptr);
              if (d < 256) d = md->lcc[d];
-             if (fi >= max || eptr >= md->end_subject || fc == d)
+             if (fc == d) MRRETURN(MATCH_NOMATCH);
-               RRETURN(MATCH_NOMATCH);
              }
            }
          else
-Line 2386 
 for (;;)
+Line 3211 
 for (;;)
            {
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM29);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject || fc == md->lcc[*eptr++])
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-               RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
              }
            }
          /* Control never gets here */
-Line 2409 
 for (;;)
+Line 3239 
 for (;;)
            for (i = min; i < max; i++)
              {
              int len = 1;
-             if (eptr >= md->end_subject) break;
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               break;
+               }
              GETCHARLEN(d, eptr, len);
              if (d < 256) d = md->lcc[d];
              if (fc == d) break;
-Line 2418 
 for (;;)
+Line 3252 
 for (;;)
          if (possessive) continue;
          for(;;)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM30);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              if (eptr-- == pp) break;        /* Stop if tried at original pos */
              BACKCHAR(eptr);
-Line 2430 
 for (;;)
+Line 3264 
 for (;;)
            {
            for (i = min; i < max; i++)
              {
-             if (eptr >= md->end_subject || fc == md->lcc[*eptr]) break;
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               break;
+               }
+             if (fc == md->lcc[*eptr]) break;
              eptr++;
              }
            if (possessive) continue;
            while (eptr >= pp)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM31);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              eptr--;
              }
            }
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
          }
        /* Control never gets here */
        }
-Line 2458 
 for (;;)
+Line 3297 
 for (;;)
          register unsigned int d;
          for (i = 1; i <= min; i++)
            {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(d, eptr);
-           if (fc == d) RRETURN(MATCH_NOMATCH);
+           if (fc == d) MRRETURN(MATCH_NOMATCH);
            }
          }
        else
-Line 2467 
 for (;;)
+Line 3311 
 for (;;)
        /* Not UTF-8 mode */
          {
          for (i = 1; i <= min; i++)
-           if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
+           }
          }
        if (min == max) continue;
-Line 2481 
 for (;;)
+Line 3332 
 for (;;)
            register unsigned int d;
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM32);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(d, eptr);
-             if (fi >= max || eptr >= md->end_subject || fc == d)
+             if (fc == d) MRRETURN(MATCH_NOMATCH);
-               RRETURN(MATCH_NOMATCH);
              }
            }
          else
-Line 2494 
 for (;;)
+Line 3350 
 for (;;)
            {
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM33);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject || fc == *eptr++)
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-               RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
              }
            }
          /* Control never gets here */
-Line 2517 
 for (;;)
+Line 3378 
 for (;;)
            for (i = min; i < max; i++)
              {
              int len = 1;
-             if (eptr >= md->end_subject) break;
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               break;
+               }
              GETCHARLEN(d, eptr, len);
              if (fc == d) break;
              eptr += len;
-Line 2525 
 for (;;)
+Line 3390 
 for (;;)
            if (possessive) continue;
            for(;;)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM34);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              if (eptr-- == pp) break;        /* Stop if tried at original pos */
              BACKCHAR(eptr);
-Line 2537 
 for (;;)
+Line 3402 
 for (;;)
            {
            for (i = min; i < max; i++)
              {
-             if (eptr >= md->end_subject || fc == *eptr) break;
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               break;
+               }
+             if (fc == *eptr) break;
              eptr++;
              }
            if (possessive) continue;
            while (eptr >= pp)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM35);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              eptr--;
              }
            }
-         RRETURN(MATCH_NOMATCH);
+         MRRETURN(MATCH_NOMATCH);
          }
        }
      /* Control never gets here */
-Line 2631 
 for (;;)
+Line 3501 
 for (;;)
      /* First, ensure the minimum number of matches are present. Use inline
      code for maximizing the speed, and do the type test once at the start
-     (i.e. keep it out of the loop). Also we can test that there are at least
+     (i.e. keep it out of the loop). Separate the UTF-8 code completely as that
-     the minimum number of bytes before we start. This isn't as effective in
-     UTF-8 mode, but it does no harm. Separate the UTF-8 code completely as that
      is tidier. Also separate the UCP code, which can be the same for both UTF-8
      and single-bytes. */
-     if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
      if (min > 0)
        {
  #ifdef SUPPORT_UCP
-Line 2646 
 for (;;)
+Line 3513 
 for (;;)
          switch(prop_type)
            {
            case PT_ANY:
-           if (prop_fail_result) RRETURN(MATCH_NOMATCH);
+           if (prop_fail_result) MRRETURN(MATCH_NOMATCH);
            for (i = 1; i <= min; i++)
              {
-             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
-             GETCHARINC(c, eptr);
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
              }
            break;
            case PT_LAMP:
            for (i = 1; i <= min; i++)
              {
-             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
-             GETCHARINC(c, eptr);
+               {
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == ucp_Lu ||
                   prop_chartype == ucp_Ll ||
                   prop_chartype == ucp_Lt) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
              }
            break;
            case PT_GC:
            for (i = 1; i <= min; i++)
              {
-             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
-             GETCHARINC(c, eptr);
+               {
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
              if ((prop_category == prop_value) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
              }
            break;
            case PT_PC:
            for (i = 1; i <= min; i++)
              {
-             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
-             GETCHARINC(c, eptr);
+               {
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == prop_value) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
              }
            break;
            case PT_SC:
            for (i = 1; i <= min; i++)
              {
-             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
-             GETCHARINC(c, eptr);
+               {
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_script = UCD_SCRIPT(c);
              if ((prop_script == prop_value) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
+             }
+           break;
+           case PT_ALNUM:
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == ucp_L || prop_category == ucp_N)
+                    == prop_fail_result)
+               MRRETURN(MATCH_NOMATCH);
+             }
+           break;
+           case PT_SPACE:    /* Perl space */
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
+                  c == CHAR_FF || c == CHAR_CR)
+                    == prop_fail_result)
+               MRRETURN(MATCH_NOMATCH);
+             }
+           break;
+           case PT_PXSPACE:  /* POSIX space */
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
+                  c == CHAR_VT || c == CHAR_FF || c == CHAR_CR)
+                    == prop_fail_result)
+               MRRETURN(MATCH_NOMATCH);
+             }
+           break;
+           case PT_WORD:
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == ucp_L || prop_category == ucp_N ||
+                  c == CHAR_UNDERSCORE)
+                    == prop_fail_result)
+               MRRETURN(MATCH_NOMATCH);
              }
            break;
+           /* This should not occur */
            default:
            RRETURN(PCRE_ERROR_INTERNAL);
            }
-Line 2712 
 for (;;)
+Line 3668 
 for (;;)
          {
          for (i = 1; i <= min; i++)
            {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            GETCHARINCTEST(c, eptr);
-           prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+           prop_category = UCD_CATEGORY(c);
-           if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH);
+           if (prop_category == ucp_M) MRRETURN(MATCH_NOMATCH);
            while (eptr < md->end_subject)
              {
              int len = 1;
-             if (!utf8) c = *eptr; else
+             if (!utf8) c = *eptr;
-               {
+               else { GETCHARLEN(c, eptr, len); }
-               GETCHARLEN(c, eptr, len);
+             prop_category = UCD_CATEGORY(c);
-               }
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
              if (prop_category != ucp_M) break;
              eptr += len;
              }
-Line 2740 
 for (;;)
+Line 3699 
 for (;;)
          case OP_ANY:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject ||
+           if (eptr >= md->end_subject)
-                ((ims & PCRE_DOTALL) == 0 && IS_NEWLINE(eptr)))
+             {
-             RRETURN(MATCH_NOMATCH);
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
+           eptr++;
+           while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
+           }
+         break;
+         case OP_ALLANY:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            eptr++;
            while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
            }
          break;
          case OP_ANYBYTE:
+         if (eptr > md->end_subject - min) MRRETURN(MATCH_NOMATCH);
          eptr += min;
          break;
          case OP_ANYNL:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(c, eptr);
            switch(c)
              {
-             default: RRETURN(MATCH_NOMATCH);
+             default: MRRETURN(MATCH_NOMATCH);
              case 0x000d:
              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
              break;
              case 0x000a:
+             break;
              case 0x000b:
              case 0x000c:
              case 0x0085:
              case 0x2028:
              case 0x2029:
+             if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
              break;
              }
            }
          break;
-         case OP_NOT_DIGIT:
+         case OP_NOT_HSPACE:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINC(c, eptr);
+           switch(c)
+             {
+             default: break;
+             case 0x09:      /* HT */
+             case 0x20:      /* SPACE */
+             case 0xa0:      /* NBSP */
+             case 0x1680:    /* OGHAM SPACE MARK */
+             case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
+             case 0x2000:    /* EN QUAD */
+             case 0x2001:    /* EM QUAD */
+             case 0x2002:    /* EN SPACE */
+             case 0x2003:    /* EM SPACE */
+             case 0x2004:    /* THREE-PER-EM SPACE */
+             case 0x2005:    /* FOUR-PER-EM SPACE */
+             case 0x2006:    /* SIX-PER-EM SPACE */
+             case 0x2007:    /* FIGURE SPACE */
+             case 0x2008:    /* PUNCTUATION SPACE */
+             case 0x2009:    /* THIN SPACE */
+             case 0x200A:    /* HAIR SPACE */
+             case 0x202f:    /* NARROW NO-BREAK SPACE */
+             case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
+             case 0x3000:    /* IDEOGRAPHIC SPACE */
+             MRRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINC(c, eptr);
+           switch(c)
+             {
+             default: MRRETURN(MATCH_NOMATCH);
+             case 0x09:      /* HT */
+             case 0x20:      /* SPACE */
+             case 0xa0:      /* NBSP */
+             case 0x1680:    /* OGHAM SPACE MARK */
+             case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
+             case 0x2000:    /* EN QUAD */
+             case 0x2001:    /* EM QUAD */
+             case 0x2002:    /* EN SPACE */
+             case 0x2003:    /* EM SPACE */
+             case 0x2004:    /* THREE-PER-EM SPACE */
+             case 0x2005:    /* FOUR-PER-EM SPACE */
+             case 0x2006:    /* SIX-PER-EM SPACE */
+             case 0x2007:    /* FIGURE SPACE */
+             case 0x2008:    /* PUNCTUATION SPACE */
+             case 0x2009:    /* THIN SPACE */
+             case 0x200A:    /* HAIR SPACE */
+             case 0x202f:    /* NARROW NO-BREAK SPACE */
+             case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
+             case 0x3000:    /* IDEOGRAPHIC SPACE */
+             break;
+             }
+           }
+         break;
+         case OP_NOT_VSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINC(c, eptr);
+           switch(c)
+             {
+             default: break;
+             case 0x0a:      /* LF */
+             case 0x0b:      /* VT */
+             case 0x0c:      /* FF */
+             case 0x0d:      /* CR */
+             case 0x85:      /* NEL */
+             case 0x2028:    /* LINE SEPARATOR */
+             case 0x2029:    /* PARAGRAPH SEPARATOR */
+             MRRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_VSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINC(c, eptr);
+           switch(c)
+             {
+             default: MRRETURN(MATCH_NOMATCH);
+             case 0x0a:      /* LF */
+             case 0x0b:      /* VT */
+             case 0x0c:      /* FF */
+             case 0x0d:      /* CR */
+             case 0x85:      /* NEL */
+             case 0x2028:    /* LINE SEPARATOR */
+             case 0x2029:    /* PARAGRAPH SEPARATOR */
+             break;
+             }
+           }
+         break;
+         case OP_NOT_DIGIT:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(c, eptr);
            if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
-             RRETURN(MATCH_NOMATCH);
+             MRRETURN(MATCH_NOMATCH);
            }
          break;
          case OP_DIGIT:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject ||
+           if (eptr >= md->end_subject)
-              *eptr >= 128 || (md->ctypes[*eptr++] & ctype_digit) == 0)
+             {
-             RRETURN(MATCH_NOMATCH);
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_digit) == 0)
+             MRRETURN(MATCH_NOMATCH);
            /* No need to skip more bytes - we know it's a 1-byte character */
            }
          break;
-Line 2797 
 for (;;)
+Line 3909 
 for (;;)
          case OP_NOT_WHITESPACE:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject ||
+           if (eptr >= md->end_subject)
-              (*eptr < 128 && (md->ctypes[*eptr++] & ctype_space) != 0))
+             {
-             RRETURN(MATCH_NOMATCH);
+             SCHECK_PARTIAL();
-           while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr < 128 && (md->ctypes[*eptr] & ctype_space) != 0)
+             MRRETURN(MATCH_NOMATCH);
+           while (++eptr < md->end_subject && (*eptr & 0xc0) == 0x80);
            }
          break;
          case OP_WHITESPACE:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject ||
+           if (eptr >= md->end_subject)
-              *eptr >= 128 || (md->ctypes[*eptr++] & ctype_space) == 0)
+             {
-             RRETURN(MATCH_NOMATCH);
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_space) == 0)
+             MRRETURN(MATCH_NOMATCH);
            /* No need to skip more bytes - we know it's a 1-byte character */
            }
          break;
-Line 2817 
 for (;;)
+Line 3937 
 for (;;)
          case OP_NOT_WORDCHAR:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject ||
+           if (eptr >= md->end_subject)
-              (*eptr < 128 && (md->ctypes[*eptr++] & ctype_word) != 0))
+             {
-             RRETURN(MATCH_NOMATCH);
+             SCHECK_PARTIAL();
-           while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr < 128 && (md->ctypes[*eptr] & ctype_word) != 0)
+             MRRETURN(MATCH_NOMATCH);
+           while (++eptr < md->end_subject && (*eptr & 0xc0) == 0x80);
            }
          break;
          case OP_WORDCHAR:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject ||
+           if (eptr >= md->end_subject)
-              *eptr >= 128 || (md->ctypes[*eptr++] & ctype_word) == 0)
+             {
-             RRETURN(MATCH_NOMATCH);
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_word) == 0)
+             MRRETURN(MATCH_NOMATCH);
            /* No need to skip more bytes - we know it's a 1-byte character */
            }
          break;
-Line 2842 
 for (;;)
+Line 3970 
 for (;;)
  #endif     /* SUPPORT_UTF8 */
        /* Code for the non-UTF-8 case for minimum matching of operators other
-       than OP_PROP and OP_NOTPROP. We can assume that there are the minimum
+       than OP_PROP and OP_NOTPROP. */
-       number of bytes present, as this was tested above. */
        switch(ctype)
          {
          case OP_ANY:
-         if ((ims & PCRE_DOTALL) == 0)
+         for (i = 1; i <= min; i++)
            {
-           for (i = 1; i <= min; i++)
+           if (eptr >= md->end_subject)
              {
-             if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
+             SCHECK_PARTIAL();
-             eptr++;
+             MRRETURN(MATCH_NOMATCH);
              }
+           if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
+           eptr++;
            }
-         else eptr += min;
          break;
-         case OP_ANYBYTE:
+         case OP_ALLANY:
+         if (eptr > md->end_subject - min)
+           {
+           SCHECK_PARTIAL();
+           MRRETURN(MATCH_NOMATCH);
+           }
          eptr += min;
          break;
-         /* Because of the CRLF case, we can't assume the minimum number of
+         case OP_ANYBYTE:
-         bytes are present in this case. */
+         if (eptr > md->end_subject - min)
+           {
+           SCHECK_PARTIAL();
+           MRRETURN(MATCH_NOMATCH);
+           }
+         eptr += min;
+         break;
          case OP_ANYNL:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            switch(*eptr++)
              {
-             default: RRETURN(MATCH_NOMATCH);
+             default: MRRETURN(MATCH_NOMATCH);
              case 0x000d:
              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
              break;
              case 0x000a:
+             break;
              case 0x000b:
              case 0x000c:
              case 0x0085:
+             if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
+             break;
+             }
+           }
+         break;
+         case OP_NOT_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: break;
+             case 0x09:      /* HT */
+             case 0x20:      /* SPACE */
+             case 0xa0:      /* NBSP */
+             MRRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: MRRETURN(MATCH_NOMATCH);
+             case 0x09:      /* HT */
+             case 0x20:      /* SPACE */
+             case 0xa0:      /* NBSP */
+             break;
+             }
+           }
+         break;
+         case OP_NOT_VSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: break;
+             case 0x0a:      /* LF */
+             case 0x0b:      /* VT */
+             case 0x0c:      /* FF */
+             case 0x0d:      /* CR */
+             case 0x85:      /* NEL */
+             MRRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_VSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: MRRETURN(MATCH_NOMATCH);
+             case 0x0a:      /* LF */
+             case 0x0b:      /* VT */
+             case 0x0c:      /* FF */
+             case 0x0d:      /* CR */
+             case 0x85:      /* NEL */
              break;
              }
            }
-Line 2887 
 for (;;)
+Line 4113 
 for (;;)
          case OP_NOT_DIGIT:
          for (i = 1; i <= min; i++)
-           if ((md->ctypes[*eptr++] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if ((md->ctypes[*eptr++] & ctype_digit) != 0) MRRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_DIGIT:
          for (i = 1; i <= min; i++)
-           if ((md->ctypes[*eptr++] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if ((md->ctypes[*eptr++] & ctype_digit) == 0) MRRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_NOT_WHITESPACE:
          for (i = 1; i <= min; i++)
-           if ((md->ctypes[*eptr++] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if ((md->ctypes[*eptr++] & ctype_space) != 0) MRRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_WHITESPACE:
          for (i = 1; i <= min; i++)
-           if ((md->ctypes[*eptr++] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
+           if ((md->ctypes[*eptr++] & ctype_space) == 0) MRRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_NOT_WORDCHAR:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_word) != 0)
-             RRETURN(MATCH_NOMATCH);
+             MRRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_WORDCHAR:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             MRRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_word) == 0)
-             RRETURN(MATCH_NOMATCH);
+             MRRETURN(MATCH_NOMATCH);
+           }
          break;
          default:
-Line 2940 
 for (;;)
+Line 4208 
 for (;;)
            case PT_ANY:
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM36);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             GETCHARINC(c, eptr);
+             if (eptr >= md->end_subject)
-             if (prop_fail_result) RRETURN(MATCH_NOMATCH);
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             if (prop_fail_result) MRRETURN(MATCH_NOMATCH);
              }
            /* Control never gets here */
            case PT_LAMP:
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM37);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             GETCHARINC(c, eptr);
+             if (eptr >= md->end_subject)
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == ucp_Lu ||
                   prop_chartype == ucp_Ll ||
                   prop_chartype == ucp_Lt) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
              }
            /* Control never gets here */
            case PT_GC:
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM38);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             GETCHARINC(c, eptr);
+             if (eptr >= md->end_subject)
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
              if ((prop_category == prop_value) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
              }
            /* Control never gets here */
            case PT_PC:
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM39);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             GETCHARINC(c, eptr);
+             if (eptr >= md->end_subject)
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == prop_value) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
              }
            /* Control never gets here */
            case PT_SC:
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM40);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
-             GETCHARINC(c, eptr);
+             if (eptr >= md->end_subject)
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_script = UCD_SCRIPT(c);
              if ((prop_script == prop_value) == prop_fail_result)
-               RRETURN(MATCH_NOMATCH);
+               MRRETURN(MATCH_NOMATCH);
+             }
+           /* Control never gets here */
+           case PT_ALNUM:
+           for (fi = min;; fi++)
+             {
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM59);
+             if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == ucp_L || prop_category == ucp_N)
+                    == prop_fail_result)
+               MRRETURN(MATCH_NOMATCH);
+             }
+           /* Control never gets here */
+           case PT_SPACE:    /* Perl space */
+           for (fi = min;; fi++)
+             {
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM60);
+             if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+             if (fi >= max) MRRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               MRRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == ucp_Z || c == CHAR_HT || c == CHAR_NL ||
+                  c == CHAR_FF || c == CHAR_CR)
+                    == prop_fail_result)
+               MRRETURN(MATCH_NOMATCH);
              }
            /* Control never gets here */
 <