/[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 446 by ph10,
Tue Sep 15 10:49:50 2009 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) 1997-2009 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_COMMIT       (-999)
+ #define MATCH_PRUNE        (-998)
+ #define MATCH_SKIP         (-997)
+ #define MATCH_THEN         (-996)
  /* 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 147 
 printf("\n");
+Line 158 
 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 220 
 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 };
  /* 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
+ actuall used in this definition. */
  #ifndef NO_RECURSE
  #define REGISTER register
  #ifdef DEBUG
- #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \
+ #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,rc,rd,re,rf,rg,rdepth+1); \
    printf("to line %d\n", __LINE__); \
    }
  #define RRETURN(ra) \
-Line 205 
 versions and production versions. */
+Line 267 
 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,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)\
+   frame->Xwhere = rw; \
-     {\
+   newframe->Xeptr = ra;\
-     newframe->Xeptr = ra;\
+   newframe->Xecode = rb;\
-     newframe->Xecode = rb;\
+   newframe->Xmstart = mstart;\
-     newframe->Xoffset_top = rc;\
+   newframe->Xoffset_top = rc;\
-     newframe->Xims = re;\
+   newframe->Xims = re;\
-     newframe->Xeptrb = rf;\
+   newframe->Xeptrb = rf;\
-     newframe->Xflags = rg;\
+   newframe->Xflags = rg;\
-     newframe->Xrdepth = frame->Xrdepth + 1;\
+   newframe->Xrdepth = frame->Xrdepth + 1;\
-     newframe->Xprevframe = frame;\
+   newframe->Xprevframe = frame;\
-     frame = newframe;\
+   frame = newframe;\
-     DPRINTF(("restarting from line %d\n", __LINE__));\
+   DPRINTF(("restarting from line %d\n", __LINE__));\
-     goto HEAP_RECURSE;\
+   goto HEAP_RECURSE;\
-     }\
+   L_##rw:\
-   else\
+   DPRINTF(("jumped back to line %d\n", __LINE__));\
-     {\
-     DPRINTF(("longjumped back to line %d\n", __LINE__));\
-     frame = md->thisframe;\
-     rx = frame->Xresult;\
-     }\
    }
  #define RRETURN(ra)\
-Line 251 
 match(), which never changes. */
+Line 308 
 match(), which never changes. */
    (pcre_stack_free)(newframe);\
    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 322 
 typedef struct heapframe {
    /* Function arguments that may change */
-   const uschar *Xeptr;
+   USPTR Xeptr;
    const uschar *Xecode;
+   USPTR Xmstart;
    int Xoffset_top;
    long int Xims;
    eptrblock *Xeptrb;
-Line 276 
 typedef struct heapframe {
+Line 333 
 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 362 
 typedef struct heapframe {
    uschar Xocchars[8];
  #endif
+   int Xcodelink;
    int Xctype;
    unsigned int Xfc;
    int Xfi;
-Line 318 
 typedef struct heapframe {
+Line 378 
 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 398 
 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. */
+ /* These macros pack up tests that are used for partial matching, and which
+ 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) RRETURN(PCRE_ERROR_PARTIAL);\
+     }
- Performance note: It might be tempting to extract commonly used fields from the
+ #define SCHECK_PARTIAL()\
- md structure (e.g. utf8, end_subject) into individual variables to improve
+   if (md->partial && eptr > mstart)\
+     {\
+     md->hitend = TRUE;\
+     if (md->partial > 1) RRETURN(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)
     offset_top  current top pointer
     md          pointer to "static" info for the match
     ims         current /i, /m, and /s options
-Line 358 
 Arguments:
+Line 441 
 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
-Line 368 
 Returns:       MATCH_MATCH if matched
+Line 450 
 Returns:       MATCH_MATCH if matched
  */
  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,
    int flags, unsigned int rdepth)
  {
-Line 382 
 register unsigned int c;   /* Character
+Line 464 
 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 396 
 frame->Xprevframe = NULL;            /*
+Line 479 
 frame->Xprevframe = NULL;            /*
  frame->Xeptr = eptr;
  frame->Xecode = ecode;
+ frame->Xmstart = mstart;
  frame->Xoffset_top = offset_top;
  frame->Xims = ims;
  frame->Xeptrb = eptrb;
-Line 410 
 HEAP_RECURSE:
+Line 494 
 HEAP_RECURSE:
  #define eptr               frame->Xeptr
  #define ecode              frame->Xecode
+ #define mstart             frame->Xmstart
  #define offset_top         frame->Xoffset_top
  #define ims                frame->Xims
  #define eptrb              frame->Xeptrb
-Line 422 
 HEAP_RECURSE:
+Line 507 
 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 588 
 int oclength;
  uschar occhars[8];
  #endif
+ int codelink;
  int ctype;
  int length;
  int max;
-Line 540 
 defined). However, RMATCH isn't like a f
+Line 627 
 defined). However, RMATCH isn't like a f
  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 641 
 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 664 
 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_FAIL:
+     RRETURN(MATCH_NOMATCH);
+     case OP_PRUNE:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+       ims, eptrb, flags, RM51);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     RRETURN(MATCH_PRUNE);
+     case OP_COMMIT:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+       ims, eptrb, flags, RM52);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     RRETURN(MATCH_COMMIT);
+     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 */
+     RRETURN(MATCH_SKIP);
+     case OP_THEN:
+     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
+       ims, eptrb, flags, RM54);
+     if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+     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 632 
 for (;;)
+Line 733 
 for (;;)
        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 649 
 for (;;)
+Line 750 
 for (;;)
        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 773 
 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);
+         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 807 
 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)
+       {
+       if (pcre_callout != NULL)
+         {
+         pcre_callout_block cb;
+         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   = md->end_subject - md->start_subject;
+         cb.start_match      = mstart - md->start_subject;
+         cb.current_position = 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) RRETURN(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)         /* Recursion test */
        {
        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/
        condition = md->recursive != NULL &&
-Line 699 
 for (;;)
+Line 847 
 for (;;)
        ecode += condition? 3 : GET(ecode, 1);
        }
-     else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */
+     else if (condcode == OP_CREF)    /* Group used test */
        {
        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
        condition = offset < offset_top && md->offset_vector[offset] >= 0;
        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 866 
 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
+     /* End of the pattern, either real or forced. If we are in a top-level
-     restore the offsets appropriately and continue from after the call. */
+     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 924 
 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->offset_top;
+       mstart = rec->save_start;
        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)))
+       RRETURN(MATCH_NOMATCH);
+     /* Otherwise, we have a match. */
-     if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);
+     md->end_match_ptr = eptr;           /* Record where we ended */
-     md->end_match_ptr = eptr;          /* Record where we ended */
+     md->end_offset_top = offset_top;    /* and how many extracts were taken */
-     md->end_offset_top = offset_top;   /* and how many extracts were taken */
+     md->start_match_ptr = mstart;       /* and the start (\K can modify) */
      RRETURN(MATCH_MATCH);
      /* Change option settings */
-Line 797 
 for (;;)
+Line 967 
 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,
+         RM4);
        if (rrc == MATCH_MATCH) break;
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode, 1);
        }
      while (*ecode == OP_ALT);
-Line 823 
 for (;;)
+Line 994 
 for (;;)
      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,
+         RM5);
        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode,1);
        }
      while (*ecode == OP_ALT);
-Line 849 
 for (;;)
+Line 1021 
 for (;;)
          {
          eptr--;
          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
-         BACKCHAR(eptr)
+         BACKCHAR(eptr);
          }
        }
      else
-Line 862 
 for (;;)
+Line 1034 
 for (;;)
        if (eptr < md->start_subject) RRETURN(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 880 
 for (;;)
+Line 1053 
 for (;;)
        cb.offset_vector    = md->offset_vector;
        cb.subject          = (PCRE_SPTR)md->start_subject;
        cb.subject_length   = md->end_subject - md->start_subject;
-       cb.start_match      = md->start_match - md->start_subject;
+       cb.start_match      = mstart - md->start_subject;
        cb.current_position = eptr - md->start_subject;
        cb.pattern_position = GET(ecode, 2);
        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
-Line 942 
 for (;;)
+Line 1115 
 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_start = mstart;
-       md->start_match = eptr;
+       new_recursive.offset_top = offset_top;
+       mstart = eptr;
        /* OK, now we can do the recursion. For each top-level alternative we
        restore the offset and recursion data. */
-Line 952 
 for (;;)
+Line 1126 
 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)
            {
            DPRINTF(("Recursion matched\n"));
-Line 962 
 for (;;)
+Line 1136 
 for (;;)
              (pcre_free)(new_recursive.offset_save);
            RRETURN(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 996 
 for (;;)
+Line 1172 
 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) break;
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode,1);
        }
      while (*ecode == OP_ALT);
-Line 1042 
 for (;;)
+Line 1217 
 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 1240 
 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 1260 
 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 1132 
 for (;;)
+Line 1315 
 for (;;)
        {
        number = GET2(prev, 1+LINK_SIZE);
        offset = number << 1;
  #ifdef DEBUG
        printf("end bracket %d", number);
        printf("\n");
-Line 1155 
 for (;;)
+Line 1338 
 for (;;)
          recursion_info *rec = md->recursive;
          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
          md->recursive = rec->prevrec;
-         md->start_match = rec->save_start;
+         mstart = rec->save_start;
          memcpy(md->offset_vector, rec->offset_save,
            rec->saved_max * sizeof(int));
+         offset_top = rec->offset_top;
          ecode = rec->after_call;
          ims = original_ims;
          break;
-Line 1184 
 for (;;)
+Line 1368 
 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 1234 
 for (;;)
+Line 1423 
 for (;;)
      ecode++;
      break;
+     /* Reset the start of match point */
+     case OP_SET_SOM:
+     mstart = eptr;
+     ecode++;
+     break;
      /* Assert before internal newline if multiline, or before a terminating
      newline unless endonly is set, else end of subject unless noteol is set. */
-Line 1285 
 for (;;)
+Line 1481 
 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)
          {
          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);
            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
            }
-         if (eptr >= md->end_subject) cur_is_word = FALSE; else
+         if (eptr >= md->end_subject)
+           {
+           SCHECK_PARTIAL();
+           cur_is_word = FALSE;
+           }
+         else
            {
            GETCHAR(c, eptr);
            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
-Line 1306 
 for (;;)
+Line 1509 
 for (;;)
        else
  #endif
-       /* More streamlined when not in UTF-8 mode */
+       /* Not in UTF-8 mode */
          {
-         prev_is_word = (eptr != md->start_subject) &&
+         if (eptr == md->start_subject) prev_is_word = FALSE; else
-           ((md->ctypes[eptr[-1]] & ctype_word) != 0);
+           {
-         cur_is_word = (eptr < md->end_subject) &&
+           if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
-           ((md->ctypes[*eptr] & ctype_word) != 0);
+           prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
+           }
+         if (eptr >= md->end_subject)
+           {
+           SCHECK_PARTIAL();
+           cur_is_word = FALSE;
+           }
+         else cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
          }
        /* Now see if the situation is what we want */
-Line 1326 
 for (;;)
+Line 1536 
 for (;;)
      /* Match a single character type; inline for speed */
      case OP_ANY:
-     if ((ims & PCRE_DOTALL) == 0)
+     if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
+     /* Fall through */
+     case OP_ALLANY:
+     if (eptr++ >= md->end_subject)
        {
-       if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
+       SCHECK_PARTIAL();
+       RRETURN(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 1553 
 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();
+       RRETURN(MATCH_NOMATCH);
+       }
      ecode++;
      break;
      case OP_NOT_DIGIT:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1358 
 for (;;)
+Line 1579 
 for (;;)
      break;
      case OP_DIGIT:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1371 
 for (;;)
+Line 1596 
 for (;;)
      break;
      case OP_NOT_WHITESPACE:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1384 
 for (;;)
+Line 1613 
 for (;;)
      break;
      case OP_WHITESPACE:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1397 
 for (;;)
+Line 1630 
 for (;;)
      break;
      case OP_NOT_WORDCHAR:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1410 
 for (;;)
+Line 1647 
 for (;;)
      break;
      case OP_WORDCHAR:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      if (
  #ifdef SUPPORT_UTF8
-Line 1423 
 for (;;)
+Line 1664 
 for (;;)
      break;
      case OP_ANYNL:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      GETCHARINCTEST(c, eptr);
      switch(c)
        {
-Line 1431 
 for (;;)
+Line 1676 
 for (;;)
        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) RRETURN(MATCH_NOMATCH);
+       break;
+       }
+     ecode++;
+     break;
+     case OP_NOT_HSPACE:
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(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 */
+       RRETURN(MATCH_NOMATCH);
+       }
+     ecode++;
+     break;
+     case OP_HSPACE:
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
+     GETCHARINCTEST(c, eptr);
+     switch(c)
+       {
+       default: RRETURN(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();
+       RRETURN(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 */
+       RRETURN(MATCH_NOMATCH);
+       }
+     ecode++;
+     break;
+     case OP_VSPACE:
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
+     GETCHARINCTEST(c, eptr);
+     switch(c)
+       {
+       default: RRETURN(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++;
-Line 1448 
 for (;;)
+Line 1809 
 for (;;)
      case OP_PROP:
      case OP_NOTPROP:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(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])
          {
-Line 1461 
 for (;;)
+Line 1825 
 for (;;)
          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);
           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);
          break;
          case PT_PC:
-         if ((ecode[2] != chartype) == (op == OP_PROP))
+         if ((ecode[2] != prop->chartype) == (op == OP_PROP))
            RRETURN(MATCH_NOMATCH);
          break;
          case PT_SC:
-         if ((ecode[2] != script) == (op == OP_PROP))
+         if ((ecode[2] != prop->script) == (op == OP_PROP))
            RRETURN(MATCH_NOMATCH);
          break;
-Line 1494 
 for (;;)
+Line 1858 
 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();
+       RRETURN(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) RRETURN(MATCH_NOMATCH);
        while (eptr < md->end_subject)
          {
-Line 1507 
 for (;;)
+Line 1874 
 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 1895 
 for (;;)
      case OP_REF:
        {
        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
-       ecode += 3;                                 /* Advance past item */
+       ecode += 3;
-       /* If the reference is unset, set the length to be longer than the amount
+       /* If the reference is unset, there are two possibilities:
-       of subject left; this ensures that every attempt at a match fails. We
-       can't just fail here, because of the possibility of quantifiers with zero
+       (a) In the default, Perl-compatible state, set the length to be longer
-       minima. */
+       than the amount of subject left; this ensures that every attempt at a
+       match fails. We can't just fail here, because of the possibility of
-       length = (offset >= offset_top || md->offset_vector[offset] < 0)?
+       quantifiers with zero minima.
-         md->end_subject - eptr + 1 :
-         md->offset_vector[offset+1] - md->offset_vector[offset];
+       (b) If the JavaScript compatibility flag is set, set the length to zero
+       so that the back reference matches an empty string.
+       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 : 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 1942 
 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();
+           RRETURN(MATCH_NOMATCH);
+           }
          eptr += length;
          continue;              /* With the main loop */
          }
-Line 1582 
 for (;;)
+Line 1962 
 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();
+           RRETURN(MATCH_NOMATCH);
+           }
          eptr += length;
          }
-Line 1597 
 for (;;)
+Line 1981 
 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) RRETURN(MATCH_NOMATCH);
+           if (!match_ref(offset, eptr, length, md, ims))
+             {
+             CHECK_PARTIAL();
              RRETURN(MATCH_NOMATCH);
+             }
            eptr += length;
            }
          /* Control never gets here */
-Line 1618 
 for (;;)
+Line 2006 
 for (;;)
            }
          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;
            }
-Line 1627 
 for (;;)
+Line 2015 
 for (;;)
        }
      /* 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 2069 
 for (;;)
          {
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(c, eptr);
            if (c > 255)
              {
-Line 1701 
 for (;;)
+Line 2091 
 for (;;)
          {
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            c = *eptr++;
            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
            }
-Line 1723 
 for (;;)
+Line 2117 
 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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(c, eptr);
              if (c > 255)
                {
-Line 1743 
 for (;;)
+Line 2142 
 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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              c = *eptr++;
              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
              }
-Line 1780 
 for (;;)
+Line 2184 
 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 1799 
 for (;;)
+Line 2203 
 for (;;)
              }
            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--;
              }
-Line 1812 
 for (;;)
+Line 2216 
 for (;;)
      /* 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 2258 
 for (;;)
        for (i = 1; i <= min; i++)
          {
-         if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+         if (eptr >= md->end_subject)
-         GETCHARINC(c, eptr);
+           {
+           SCHECK_PARTIAL();
+           RRETURN(MATCH_NOMATCH);
+           }
+         GETCHARINCTEST(c, eptr);
          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
          }
-Line 1870 
 for (;;)
+Line 2279 
 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) RRETURN(MATCH_NOMATCH);
-           GETCHARINC(c, eptr);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINCTEST(c, eptr);
            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
-Line 1888 
 for (;;)
+Line 2302 
 for (;;)
            {
            int len = 1;
            if (eptr >= md->end_subject) break;
-           GETCHARLEN(c, eptr, len);
+           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);
          }
-Line 1915 
 for (;;)
+Line 2329 
 for (;;)
        length = 1;
        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() */
+         RRETURN(MATCH_NOMATCH);
+         }
        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
        }
      else
-Line 1923 
 for (;;)
+Line 2341 
 for (;;)
      /* Non-UTF-8 mode */
        {
-       if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);
+       if (md->end_subject - eptr < 1)
+         {
+         SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
+         RRETURN(MATCH_NOMATCH);
+         }
        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
        ecode += 2;
        }
-Line 1939 
 for (;;)
+Line 2361 
 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() */
+         RRETURN(MATCH_NOMATCH);
+         }
        /* If the pattern character's value is < 128, we have only one byte, and
        can use the fast lookup table. */
-Line 1963 
 for (;;)
+Line 2389 
 for (;;)
          if (fc != dc)
            {
  #ifdef SUPPORT_UCP
-           if (dc != _pcre_ucp_othercase(fc))
+           if (dc != UCD_OTHERCASE(fc))
  #endif
              RRETURN(MATCH_NOMATCH);
            }
-Line 1974 
 for (;;)
+Line 2400 
 for (;;)
      /* Non-UTF-8 mode */
        {
-       if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);
+       if (md->end_subject - eptr < 1)
+         {
+         SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
+         RRETURN(MATCH_NOMATCH);
+         }
        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
        ecode += 2;
        }
-Line 2028 
 for (;;)
+Line 2458 
 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 2472 
 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 2482 
 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;
+             RRETURN(MATCH_NOMATCH);
              }
- #else   /* without SUPPORT_UCP */
-           else { RRETURN(MATCH_NOMATCH); }
- #endif  /* SUPPORT_UCP */
            }
          if (min == max) continue;
-Line 2081 
 for (;;)
+Line 2509 
 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) RRETURN(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;
+               RRETURN(MATCH_NOMATCH);
                }
- #else   /* without SUPPORT_UCP */
-             else { RRETURN (MATCH_NOMATCH); }
- #endif  /* SUPPORT_UCP */
              }
            /* Control never gets here */
            }
-Line 2105 
 for (;;)
+Line 2533 
 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 &&
-             else
+                      eptr <= md->end_subject - oclength &&
-               {
+                      memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
-               if (memcmp(eptr, occhars, oclength) != 0) break;
-               eptr += oclength;
-               }
- #else   /* without SUPPORT_UCP */
-             else break;
  #endif  /* SUPPORT_UCP */
+             else break;
              }
            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) { RRETURN(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 2569 
 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 2588 
 for (;;)
        {
        fc = md->lcc[fc];
        for (i = 1; i <= min; i++)
+         {
+         if (eptr >= md->end_subject)
+           {
+           SCHECK_PARTIAL();
+           RRETURN(MATCH_NOMATCH);
+           }
          if (fc != md->lcc[*eptr++]) RRETURN(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) RRETURN(MATCH_NOMATCH);
-               fc != md->lcc[*eptr++])
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
              RRETURN(MATCH_NOMATCH);
+             }
+           if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
          }
-Line 2187 
 for (;;)
+Line 2621 
 for (;;)
            if (eptr >= md->end_subject || 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);
            }
-Line 2203 
 for (;;)
+Line 2639 
 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();
+           RRETURN(MATCH_NOMATCH);
+           }
+         if (fc != *eptr++) RRETURN(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) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
              RRETURN(MATCH_NOMATCH);
+             }
+           if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
          }
-Line 2225 
 for (;;)
+Line 2676 
 for (;;)
            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);
            }
-Line 2240 
 for (;;)
+Line 2692 
 for (;;)
      checking can be multibyte. */
      case OP_NOT:
-     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     if (eptr >= md->end_subject)
+       {
+       SCHECK_PARTIAL();
+       RRETURN(MATCH_NOMATCH);
+       }
      ecode++;
      GETCHARINCTEST(c, eptr);
      if ((ims & PCRE_CASELESS) != 0)
-Line 2317 
 for (;;)
+Line 2773 
 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 2800 
 for (;;)
          register unsigned int d;
          for (i = 1; i <= min; i++)
            {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(d, eptr);
            if (d < 256) d = md->lcc[d];
            if (fc == d) RRETURN(MATCH_NOMATCH);
-Line 2358 
 for (;;)
+Line 2816 
 for (;;)
        /* Not UTF-8 mode */
          {
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
+           }
          }
        if (min == max) continue;
-Line 2372 
 for (;;)
+Line 2837 
 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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(d, eptr);
              if (d < 256) d = md->lcc[d];
-             if (fi >= max || eptr >= md->end_subject || fc == d)
+             if (fc == d) RRETURN(MATCH_NOMATCH);
-               RRETURN(MATCH_NOMATCH);
              }
            }
          else
-Line 2386 
 for (;;)
+Line 2856 
 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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
                RRETURN(MATCH_NOMATCH);
+               }
+             if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);
              }
            }
          /* Control never gets here */
-Line 2418 
 for (;;)
+Line 2893 
 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 2436 
 for (;;)
+Line 2911 
 for (;;)
            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--;
              }
-Line 2458 
 for (;;)
+Line 2933 
 for (;;)
          register unsigned int d;
          for (i = 1; i <= min; i++)
            {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(d, eptr);
            if (fc == d) RRETURN(MATCH_NOMATCH);
            }
-Line 2467 
 for (;;)
+Line 2947 
 for (;;)
        /* Not UTF-8 mode */
          {
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
+           }
          }
        if (min == max) continue;
-Line 2481 
 for (;;)
+Line 2968 
 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);
-             GETCHARINC(d, eptr);
+             if (fi >= max) RRETURN(MATCH_NOMATCH);
-             if (fi >= max || eptr >= md->end_subject || fc == d)
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
                RRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINC(d, eptr);
+             if (fc == d) RRETURN(MATCH_NOMATCH);
              }
            }
          else
-Line 2494 
 for (;;)
+Line 2986 
 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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
                RRETURN(MATCH_NOMATCH);
+               }
+             if (fc == *eptr++) RRETURN(MATCH_NOMATCH);
              }
            }
          /* Control never gets here */
-Line 2525 
 for (;;)
+Line 3022 
 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 2543 
 for (;;)
+Line 3040 
 for (;;)
            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--;
              }
-Line 2631 
 for (;;)
+Line 3128 
 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 2649 
 for (;;)
+Line 3143 
 for (;;)
            if (prop_fail_result) RRETURN(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();
+               RRETURN(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();
+               RRETURN(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)
-Line 2670 
 for (;;)
+Line 3172 
 for (;;)
            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();
+               RRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
              if ((prop_category == prop_value) == prop_fail_result)
                RRETURN(MATCH_NOMATCH);
              }
-Line 2681 
 for (;;)
+Line 3187 
 for (;;)
            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();
+               RRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == prop_value) == prop_fail_result)
                RRETURN(MATCH_NOMATCH);
              }
-Line 2692 
 for (;;)
+Line 3202 
 for (;;)
            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();
+               RRETURN(MATCH_NOMATCH);
+               }
+             GETCHARINCTEST(c, eptr);
+             prop_script = UCD_SCRIPT(c);
              if ((prop_script == prop_value) == prop_fail_result)
                RRETURN(MATCH_NOMATCH);
              }
-Line 2712 
 for (;;)
+Line 3226 
 for (;;)
          {
          for (i = 1; i <= min; i++)
            {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(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);
            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 3257 
 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)))
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           if (IS_NEWLINE(eptr)) RRETURN(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();
              RRETURN(MATCH_NOMATCH);
+             }
            eptr++;
            while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
            }
          break;
          case OP_ANYBYTE:
+         if (eptr > md->end_subject - min) RRETURN(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();
+             RRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(c, eptr);
            switch(c)
              {
-Line 2763 
 for (;;)
+Line 3301 
 for (;;)
              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) RRETURN(MATCH_NOMATCH);
+             break;
+             }
+           }
+         break;
+         case OP_NOT_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(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 */
+             RRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINC(c, eptr);
+           switch(c)
+             {
+             default: RRETURN(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();
+             RRETURN(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 */
+             RRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_VSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           GETCHARINC(c, eptr);
+           switch(c)
+             {
+             default: RRETURN(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;
              }
            }
-Line 2777 
 for (;;)
+Line 3439 
 for (;;)
          case OP_NOT_DIGIT:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            GETCHARINC(c, eptr);
            if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
              RRETURN(MATCH_NOMATCH);
-Line 2787 
 for (;;)
+Line 3453 
 for (;;)
          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)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_digit) == 0)
              RRETURN(MATCH_NOMATCH);
            /* No need to skip more bytes - we know it's a 1-byte character */
            }
-Line 2797 
 for (;;)
+Line 3467 
 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))
+             {
+             SCHECK_PARTIAL();
              RRETURN(MATCH_NOMATCH);
-           while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
+             }
+           if (*eptr < 128 && (md->ctypes[*eptr] & ctype_space) != 0)
+             RRETURN(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)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_space) == 0)
              RRETURN(MATCH_NOMATCH);
            /* No need to skip more bytes - we know it's a 1-byte character */
            }
-Line 2818 
 for (;;)
+Line 3496 
 for (;;)
          for (i = 1; i <= min; i++)
            {
            if (eptr >= md->end_subject ||
-              (*eptr < 128 && (md->ctypes[*eptr++] & ctype_word) != 0))
+              (*eptr < 128 && (md->ctypes[*eptr] & ctype_word) != 0))
              RRETURN(MATCH_NOMATCH);
-           while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
+           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)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           if (*eptr >= 128 || (md->ctypes[*eptr++] & ctype_word) == 0)
              RRETURN(MATCH_NOMATCH);
            /* No need to skip more bytes - we know it's a 1-byte character */
            }
-Line 2842 
 for (;;)
+Line 3524 
 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++;
+             RRETURN(MATCH_NOMATCH);
              }
+           if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
+           eptr++;
            }
-         else eptr += min;
          break;
-         case OP_ANYBYTE:
+         case OP_ALLANY:
+         if (eptr > md->end_subject - min)
+           {
+           SCHECK_PARTIAL();
+           RRETURN(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();
+           RRETURN(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();
+             RRETURN(MATCH_NOMATCH);
+             }
            switch(*eptr++)
              {
              default: RRETURN(MATCH_NOMATCH);
-Line 2877 
 for (;;)
+Line 3574 
 for (;;)
              if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
              break;
              case 0x000a:
+             break;
              case 0x000b:
              case 0x000c:
              case 0x0085:
+             if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
+             break;
+             }
+           }
+         break;
+         case OP_NOT_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: break;
+             case 0x09:      /* HT */
+             case 0x20:      /* SPACE */
+             case 0xa0:      /* NBSP */
+             RRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: RRETURN(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();
+             RRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: break;
+             case 0x0a:      /* LF */
+             case 0x0b:      /* VT */
+             case 0x0c:      /* FF */
+             case 0x0d:      /* CR */
+             case 0x85:      /* NEL */
+             RRETURN(MATCH_NOMATCH);
+             }
+           }
+         break;
+         case OP_VSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           switch(*eptr++)
+             {
+             default: RRETURN(MATCH_NOMATCH);
+             case 0x0a:      /* LF */
+             case 0x0b:      /* VT */
+             case 0x0c:      /* FF */
+             case 0x0d:      /* CR */
+             case 0x85:      /* NEL */
              break;
              }
            }
-Line 2887 
 for (;;)
+Line 3667 
 for (;;)
          case OP_NOT_DIGIT:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_DIGIT:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_digit) == 0) RRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_NOT_WHITESPACE:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_space) != 0) RRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_WHITESPACE:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_space) == 0) RRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_NOT_WORDCHAR:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_word) != 0)
              RRETURN(MATCH_NOMATCH);
+           }
          break;
          case OP_WORDCHAR:
          for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
            if ((md->ctypes[*eptr++] & ctype_word) == 0)
              RRETURN(MATCH_NOMATCH);
+           }
          break;
          default:
-Line 2940 
 for (;;)
+Line 3762 
 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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(c, eptr);
              if (prop_fail_result) RRETURN(MATCH_NOMATCH);
              }
-Line 2951 
 for (;;)
+Line 3778 
 for (;;)
            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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(c, eptr);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == ucp_Lu ||
                   prop_chartype == ucp_Ll ||
                   prop_chartype == ucp_Lt) == prop_fail_result)
-Line 2966 
 for (;;)
+Line 3798 
 for (;;)
            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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(c, eptr);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_category = UCD_CATEGORY(c);
              if ((prop_category == prop_value) == prop_fail_result)
                RRETURN(MATCH_NOMATCH);
              }
-Line 2979 
 for (;;)
+Line 3816 
 for (;;)
            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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(c, eptr);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == prop_value) == prop_fail_result)
                RRETURN(MATCH_NOMATCH);
              }
-Line 2992 
 for (;;)
+Line 3834 
 for (;;)
            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) RRETURN(MATCH_NOMATCH);
+             if (eptr >= md->end_subject)
+               {
+               SCHECK_PARTIAL();
+               RRETURN(MATCH_NOMATCH);
+               }
              GETCHARINC(c, eptr);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_script = UCD_SCRIPT(c);
              if ((prop_script == prop_value) == prop_fail_result)
                RRETURN(MATCH_NOMATCH);
              }
-Line 3014 
 for (;;)
+Line 3861 
 for (;;)
          {
          for (fi = min;; fi++)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM41);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           if (fi >= max) RRETURN(MATCH_NOMATCH);
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(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);
            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 3043 
 for (;;)
+Line 3893 
 for (;;)
          {
          for (fi = min;; fi++)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM42);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if (fi >= max || eptr >= md->end_subject ||
+           if (fi >= max) RRETURN(MATCH_NOMATCH);
-                (ctype == OP_ANY && (ims & PCRE_DOTALL) == 0 &&
+           if (eptr >= md->end_subject)
-                 IS_NEWLINE(eptr)))
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           if (ctype == OP_ANY && IS_NEWLINE(eptr))
              RRETURN(MATCH_NOMATCH);
            GETCHARINC(c, eptr);
            switch(ctype)
              {
-             case OP_ANY:        /* This is the DOTALL case */
+             case OP_ANY:        /* This is the non-NL case */
-             break;
+             case OP_ALLANY:
              case OP_ANYBYTE:
              break;
-Line 3067 
 for (;;)
+Line 3919 
 for (;;)
                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) RRETURN(MATCH_NOMATCH);
+               break;
+               }
+             break;
+             case OP_NOT_HSPACE:
+             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 */
+               RRETURN(MATCH_NOMATCH);
+               }
+             break;
+             case OP_HSPACE:
+             switch(c)
+               {
+               default: RRETURN(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:
+             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 */
+               RRETURN(MATCH_NOMATCH);
+               }
+             break;
+             case OP_VSPACE:
+             switch(c)
+               {
+               default: RRETURN(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;
-Line 3117 
 for (;;)
+Line 4056 
 for (;;)
          {
          for (fi = min;; fi++)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM43);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if (fi >= max || eptr >= md->end_subject ||
+           if (fi >= max) RRETURN(MATCH_NOMATCH);
-                ((ims & PCRE_DOTALL) == 0 && IS_NEWLINE(eptr)))
+           if (eptr >= md->end_subject)
+             {
+             SCHECK_PARTIAL();
+             RRETURN(MATCH_NOMATCH);
+             }
+           if (ctype == OP_ANY && IS_NEWLINE(eptr))
              RRETURN(MATCH_NOMATCH);
            c = *eptr++;
            switch(ctype)
              {
-             case OP_ANY:   /* This is the DOTALL case */
+             case OP_ANY:     /* This is the non-NL case */
-             break;
+             case OP_ALLANY:
              case OP_ANYBYTE:
              break;
-Line 3139 
 for (;;)
+Line 4081 
 for (;;)
                case 0x000d:
                if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
                break;
                case 0x000a:
+               break;
                case 0x000b:
                case 0x000c:
                case 0x0085:
+               if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
+               break;
+               }
+             break;
+             case OP_NOT_HSPACE:
+             switch(c)
+               {
+               default: break;
+               case 0x09:      /* HT */
+               case 0x20:      /* SPACE */
+               case 0xa0:      /* NBSP */
+               RRETURN(MATCH_NOMATCH);
+               }
+             break;
+             case OP_HSPACE:
+             switch(c)
+               {
+               default: RRETURN(MATCH_NOMATCH);
+               case 0x09:      /* HT */
+               case 0x20:      /* SPACE */
+               case 0xa0:      /* NBSP */
+               break;
+               }
+             break;
+             case OP_NOT_VSPACE:
+             switch(c)
+               {
+               default: break;
+               case 0x0a:      /* LF */
+               case 0x0b:      /* VT */
+               case 0x0c:      /* FF */
+               case 0x0d:      /* CR */
+               case 0x85:      /* NEL */
+               RRETURN(MATCH_NOMATCH);
+               }
+             break;
+             case OP_VSPACE:
+             switch(c)
+               {
+               default: RRETURN(MATCH_NOMATCH);
+               case 0x0a:      /* LF */
+               case 0x0b:      /* VT */
+               case 0x0c:      /* FF */
+               case 0x0d:      /* CR */
+               case 0x85:      /* NEL */
                break;
                }
              break;
-Line 3209 
 for (;;)
+Line 4203 
 for (;;)
              int len = 1;
              if (eptr >= md->end_subject) break;
              GETCHARLEN(c, eptr, len);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == ucp_Lu ||
                   prop_chartype == ucp_Ll ||
                   prop_chartype == ucp_Lt) == prop_fail_result)
-Line 3224 
 for (;;)
+Line 4218 
 for (;;)
              int len = 1;
              if (eptr >= md->end_subject) break;
              GETCHARLEN(c, eptr, len);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_category = UCD_CATEGORY(c);
              if ((prop_category == prop_value) == prop_fail_result)
                break;
              eptr+= len;
-Line 3237 
 for (;;)
+Line 4231 
 for (;;)
              int len = 1;
              if (eptr >= md->end_subject) break;
              GETCHARLEN(c, eptr, len);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_chartype = UCD_CHARTYPE(c);
              if ((prop_chartype == prop_value) == prop_fail_result)
                break;
              eptr+= len;
-Line 3250 
 for (;;)
+Line 4244 
 for (;;)
              int len = 1;
              if (eptr >= md->end_subject) break;
              GETCHARLEN(c, eptr, len);
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_script = UCD_SCRIPT(c);
              if ((prop_script == prop_value) == prop_fail_result)
                break;
              eptr+= len;
-Line 3263 
 for (;;)
+Line 4257 
 for (;;)
          if (possessive) continue;
          for(;;)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM44);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            if (eptr-- == pp) break;        /* Stop if tried at original pos */
-           BACKCHAR(eptr);
+           if (utf8) BACKCHAR(eptr);
            }
          }
-Line 3279 
 for (;;)
+Line 4273 
 for (;;)
            {
            if (eptr >= md->end_subject) break;
            GETCHARINCTEST(c, eptr);
-           prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+           prop_category = UCD_CATEGORY(c);
            if (prop_category == ucp_M) break;
            while (eptr < md->end_subject)
              {
-Line 3288 
 for (;;)
+Line 4282 
 for (;;)
                {
                GETCHARLEN(c, eptr, len);
                }
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_category = UCD_CATEGORY(c);
              if (prop_category != ucp_M) break;
              eptr += len;
              }
-Line 3299 
 for (;;)
+Line 4293 
 for (;;)
          if (possessive) continue;
          for(;;)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM45);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            if (eptr-- == pp) break;        /* Stop if tried at original pos */
            for (;;)                        /* Move back over one extended */
              {
              int len = 1;
-             BACKCHAR(eptr);
              if (!utf8) c = *eptr; else
                {
+               BACKCHAR(eptr);
                GETCHARLEN(c, eptr, len);
                }
-             prop_category = _pcre_ucp_findprop(c, &prop_chartype, &prop_script);
+             prop_category = UCD_CATEGORY(c);
              if (prop_category != ucp_M) break;
              eptr--;
              }
-Line 3328 
 for (;;)
+Line 4322 
 for (;;)
          switch(ctype)
            {
            case OP_ANY:
-           /* Special code is required for UTF8, but when the maximum is
-           unlimited we don't need it, so we repeat the non-UTF8 code. This is
-           probably worth it, because .* is quite a common idiom. */
            if (max < INT_MAX)
              {
-             if ((ims & PCRE_DOTALL) == 0)
+             for (i = min; i < max; i++)
-               {
-               for (i = min; i < max; i++)
-                 {
-                 if (eptr >= md->end_subject || IS_NEWLINE(eptr)) break;
-                 eptr++;
-                 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-                 }
-               }
-             else
                {
-               for (i = min; i < max; i++)
+               if (eptr >= md->end_subject || IS_NEWLINE(eptr)) break;
-                 {
+               eptr++;
-                 if (eptr >= md->end_subject) break;
+               while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-                 eptr++;
-                 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-                 }
                }
              }
-Line 3359 
 for (;;)
+Line 4336 
 for (;;)
            else
              {
-             if ((ims & PCRE_DOTALL) == 0)
+             for (i = min; i < max; i++)
                {
-               for (i = min; i < max; i++)
+               if (eptr >= md->end_subject || IS_NEWLINE(eptr)) break;
-                 {
+               eptr++;
-                 if (eptr >= md->end_subject || IS_NEWLINE(eptr)) break;
+               while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-                 eptr++;
-                 }
-               break;
                }
-             else
+             }
+           break;
+           case OP_ALLANY:
+           if (max < INT_MAX)
+             {
+             for (i = min; i < max; i++)
                {
-               c = max - min;
+               if (eptr >= md->end_subject) break;
-               if (c > (unsigned int)(md->end_subject - eptr))
+               eptr++;
-                 c = md->end_subject - eptr;
+               while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-               eptr += c;
                }
              }
+           else eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
            break;
            /* The byte case is the same as non-UTF8 */
-Line 3400 
 for (;;)
+Line 4380 
 for (;;)
                }
              else
                {
-               if (c != 0x000a && c != 0x000b && c != 0x000c &&
+               if (c != 0x000a &&
-                   c != 0x0085 && c != 0x2028 && c != 0x2029)
+                   (md->bsr_anycrlf ||
+                    (c != 0x000b && c != 0x000c &&
+                     c != 0x0085 && c != 0x2028 && c != 0x2029)))
                  break;
                eptr += len;
                }
              }
            break;
+           case OP_NOT_HSPACE:
+           case OP_HSPACE:
+           for (i = min; i < max; i++)
+             {
+             BOOL gotspace;
+             int len = 1;
+             if (eptr >= md->end_subject) break;
+             GETCHARLEN(c, eptr, len);
+             switch(c)
+               {
+               default: gotspace = FALSE; 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 */
+               gotspace = TRUE;
+               break;
+               }
+             if (gotspace == (ctype == OP_NOT_HSPACE)) break;
+             eptr += len;
+             }
+           break;
+           case OP_NOT_VSPACE:
+           case OP_VSPACE:
+           for (i = min; i < max; i++)
+             {
+             BOOL gotspace;
+             int len = 1;
+             if (eptr >= md->end_subject) break;
+             GETCHARLEN(c, eptr, len);
+             switch(c)
+               {
+               default: gotspace = FALSE; break;
+               case 0x0a:      /* LF */
+               case 0x0b:      /* VT */
+               case 0x0c:      /* FF */
+               case 0x0d:      /* CR */
+               case 0x85:      /* NEL */
+               case 0x2028:    /* LINE SEPARATOR */
+               case 0x2029:    /* PARAGRAPH SEPARATOR */
+               gotspace = TRUE;
+               break;
+               }
+             if (gotspace == (ctype == OP_NOT_VSPACE)) break;
+             eptr += len;
+             }
+           break;
            case OP_NOT_DIGIT:
            for (i = min; i < max; i++)
              {
-Line 3483 
 for (;;)
+Line 4529 
 for (;;)
          if (possessive) continue;
          for(;;)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM46);
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            if (eptr-- == pp) break;        /* Stop if tried at original pos */
            BACKCHAR(eptr);
            }
          }
        else
- #endif
+ #endif  /* SUPPORT_UTF8 */
        /* Not UTF-8 mode */
          {
          switch(ctype)
            {
            case OP_ANY:
-           if ((ims & PCRE_DOTALL) == 0)
+           for (i = min; i < max; i++)
              {
-             for (i = min; i < max; i++)
+             if (eptr >= md->end_subject || IS_NEWLINE(eptr)) break;
-               {
+             eptr++;
-               if (eptr >= md->end_subject || IS_NEWLINE(eptr)) break;
-               eptr++;
-               }
-             break;
              }
-           /* For DOTALL case, fall through and treat as \C */
+           break;
+           case OP_ALLANY:
            case OP_ANYBYTE:
            c = max - min;
            if (c > (unsigned int)(md->end_subject - eptr))
-Line 3527 
 for (;;)
+Line 4570 
 for (;;)
                }
              else
                {
-               if (c != 0x000a && c != 0x000b && c != 0x000c && c != 0x0085)
+               if (c != 0x000a &&
+                   (md->bsr_anycrlf ||
+                     (c != 0x000b && c != 0x000c && c != 0x0085)))
                  break;
                eptr++;
                }
              }
            break;
+           case OP_NOT_HSPACE:
+           for (i = min; i < max; i++)
+             {
+             if (eptr >= md->end_subject) break;
+             c = *eptr;
+             if (c == 0x09 || c == 0x20 || c == 0xa0) break;
+             eptr++;
+             }
+           break;
+           case OP_HSPACE:
+           for (i = min; i < max; i++)
+             {
+             if (eptr >= md->end_subject) break;
+             c = *eptr;
+             if (c != 0x09 && c != 0x20 && c != 0xa0) break;
+             eptr++;
+             }
+           break;
+           case OP_NOT_VSPACE:
+           for (i = min; i < max; i++)
+             {
+             if (eptr >= md->end_subject) break;
+             c = *eptr;
+             if (c == 0x0a || c == 0x0b || c == 0x0c || c == 0x0d || c == 0x85)
+               break;
+             eptr++;
+             }
+           break;
+           case OP_VSPACE:
+           for (i = min; i < max; i++)
+             {
+             if (eptr >= md->end_subject) break;
+             c = *eptr;
+             if (c != 0x0a && c != 0x0b && c != 0x0c && c != 0x0d && c != 0x85)
+               break;
+             eptr++;
+             }
+           break;
            case OP_NOT_DIGIT:
            for (i = min; i < max; i++)
              {
-Line 3597 
 for (;;)
+Line 4684 
 for (;;)
          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, RM47);
            eptr--;
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            }
-Line 3623 
 for (;;)
+Line 4710 
 for (;;)
    }             /* End of main loop */
  /* Control never reaches here */
+ /* When compiling to use the heap rather than the stack for recursive calls to
+ match(), the RRETURN() macro jumps here. The number that is saved in
+ frame->Xwhere indicates which label we actually want to return to. */
+ #ifdef NO_RECURSE
+ #define LBL(val) case val: goto L_RM##val;
+ HEAP_RETURN:
+ switch (frame->Xwhere)
+   {
+   LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
+   LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
+   LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
+   LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
+   LBL(53) LBL(54)
+ #ifdef SUPPORT_UTF8
+   LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)
+   LBL(32) LBL(34) LBL(42) LBL(46)
+ #ifdef SUPPORT_UCP
+   LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
+ #endif  /* SUPPORT_UCP */
+ #endif  /* SUPPORT_UTF8 */
+   default:
+   DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
+   return PCRE_ERROR_INTERNAL;
+   }
+ #undef LBL
+ #endif  /* NO_RECURSE */
  }
-Line 3635 
 Undefine all the macros that were define
+Line 4751 
 Undefine all the macros that were define
  #ifdef NO_RECURSE
  #undef eptr
  #undef ecode
+ #undef mstart
  #undef offset_top
  #undef ims
  #undef eptrb
-Line 3707 
 Returns:          > 0 => success; value
+Line 4824 
 Returns:          > 0 => success; value
                   < -1 => some kind of unexpected problem
  */
- PCRE_DATA_SCOPE int
+ PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
  pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
    PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
    int offsetcount)
-Line 3731 
 const uschar *tables;
+Line 4848 
 const uschar *tables;
  const uschar *start_bits = NULL;
  USPTR start_match = (USPTR)subject + start_offset;
  USPTR end_subject;
+ USPTR start_partial = NULL;
  USPTR req_byte_ptr = start_match - 1;
- eptrblock eptrchain[EPTR_WORK_SIZE];
  pcre_study_data internal_study;
  const pcre_study_data *study;
-Line 3795 
 if (re->magic_number != MAGIC_NUMBER)
+Line 4912 
 if (re->magic_number != MAGIC_NUMBER)
  /* Set up other data */
  anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
- startline = (re->options & PCRE_STARTLINE) != 0;
+ startline = (re->flags & PCRE_STARTLINE) != 0;
  firstline = (re->options & PCRE_FIRSTLINE) != 0;
  /* The code starts after the real_pcre block and the capture name table. */
-Line 3810 
 end_subject = md->end_subject;
+Line 4927 
 end_subject = md->end_subject;
  md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
  utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
+ md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
  md->notbol = (options & PCRE_NOTBOL) != 0;
  md->noteol = (options & PCRE_NOTEOL) != 0;
  md->notempty = (options & PCRE_NOTEMPTY) != 0;
- md->partial = (options & PCRE_PARTIAL) != 0;
+ md->notempty_atstart = (options & PCRE_NOTEMPTY_ATSTART) != 0;
+ md->partial = ((options & PCRE_PARTIAL_HARD) != 0)? 2 :
+               ((options & PCRE_PARTIAL_SOFT) != 0)? 1 : 0;
  md->hitend = FALSE;
  md->recursive = NULL;                   /* No recursion at top level */
- md->eptrchain = eptrchain;              /* Make workspace generally available */
  md->lcc = tables + lcc_offset;
  md->ctypes = tables + ctypes_offset;
+ /* Handle different \R options. */
+ switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
+   {
+   case 0:
+   if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
+     md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
+   else
+ #ifdef BSR_ANYCRLF
+   md->bsr_anycrlf = TRUE;
+ #else
+   md->bsr_anycrlf = FALSE;
+ #endif
+   break;
+   case PCRE_BSR_ANYCRLF:
+   md->bsr_anycrlf = TRUE;
+   break;
+   case PCRE_BSR_UNICODE:
+   md->bsr_anycrlf = FALSE;
+   break;
+   default: return PCRE_ERROR_BADNEWLINE;
+   }
  /* Handle different types of newline. The three bits give eight cases. If
  nothing is set at run time, whatever was used at compile time applies. */
- switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options : options) &
+ switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
-        PCRE_NEWLINE_BITS)
+         (pcre_uint32)options) & PCRE_NEWLINE_BITS)
    {
    case 0: newline = NEWLINE; break;   /* Compile-time default */
-   case PCRE_NEWLINE_CR: newline = '\r'; break;
+   case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
-   case PCRE_NEWLINE_LF: newline = '\n'; break;
+   case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
    case PCRE_NEWLINE_CR+
-        PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;
+        PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
    case PCRE_NEWLINE_ANY: newline = -1; break;
+   case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
    default: return PCRE_ERROR_BADNEWLINE;
    }
- if (newline < 0)
+ if (newline == -2)
+   {
+   md->nltype = NLTYPE_ANYCRLF;
+   }
+ else if (newline < 0)
    {
    md->nltype = NLTYPE_ANY;
    }
-Line 3858 
 else
+Line 5008 
 else
      }
    }
- /* Partial matching is supported only for a restricted set of regexes at the
+ /* Partial matching was originally supported only for a restricted set of
- moment. */
+ regexes; from release 8.00 there are no restrictions, but the bits are still
+ defined (though never set). So there's no harm in leaving this code. */
- if (md->partial && (re->options & PCRE_NOPARTIAL) != 0)
+ if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
    return PCRE_ERROR_BADPARTIAL;
  /* Check a UTF-8 string if required. Unfortunately there's no way of passing
-Line 3870 
 back the character offset. */
+Line 5021 
 back the character offset. */
  #ifdef SUPPORT_UTF8
  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
    {
-   if (_pcre_valid_utf8((uschar *)subject, length) >= 0)
+   if (_pcre_valid_utf8((USPTR)subject, length) >= 0)
      return PCRE_ERROR_BADUTF8;
    if (start_offset > 0 && start_offset < length)
      {
-     int tb = ((uschar *)subject)[start_offset];
+     int tb = ((USPTR)subject)[start_offset];
      if (tb > 127)
        {
        tb &= 0xc0;
-Line 3938 
 studied, there may be a bitmap of possib
+Line 5089 
 studied, there may be a bitmap of possib
  if (!anchored)
    {
-   if ((re->options & PCRE_FIRSTSET) != 0)
+   if ((re->flags & PCRE_FIRSTSET) != 0)
      {
      first_byte = re->first_byte & 255;
      if ((first_byte_caseless = ((re->first_byte & REQ_CASELESS) != 0)) == TRUE)
-Line 3953 
 if (!anchored)
+Line 5104 
 if (!anchored)
  /* For anchored or unanchored matches, there may be a "last known required
  character" set. */
- if ((re->options & PCRE_REQCHSET) != 0)
+ if ((re->flags & PCRE_REQCHSET) != 0)
    {
    req_byte = re->req_byte & 255;
    req_byte_caseless = (re->req_byte & REQ_CASELESS) != 0;
-Line 3969 
 the loop runs just once. */
+Line 5120 
 the loop runs just once. */
  for(;;)
    {
    USPTR save_end_subject = end_subject;
+   USPTR new_start_match;
    /* Reset the maximum number of extractions we might see. */
-Line 3979 
 for(;;)
+Line 5131 
 for(;;)
      while (iptr < iend) *iptr++ = -1;
      }
-   /* Advance to a unique first char if possible. If firstline is TRUE, the
+   /* If firstline is TRUE, the start of the match is constrained to the first
-   start of the match is constrained to the first line of a multiline string.
+   line of a multiline string. That is, the match must be before or at the first
-   That is, the match must be before or at the first newline. Implement this by
+   newline. Implement this by temporarily adjusting end_subject so that we stop
-   temporarily adjusting end_subject so that we stop scanning at a newline. If
+   scanning at a newline. If the match fails at the newline, later code breaks
-   the match fails at the newline, later code breaks this loop. */
+   this loop. */
    if (firstline)
      {
      USPTR t = start_match;
+ #ifdef SUPPORT_UTF8
+     if (utf8)
+       {
+       while (t < md->end_subject && !IS_NEWLINE(t))
+         {
+         t++;
+         while (t < end_subject && (*t & 0xc0) == 0x80) t++;
+         }
+       }
+     else
+ #endif
      while (t < md->end_subject && !IS_NEWLINE(t)) t++;
      end_subject = t;
      }
-   /* Now test for a unique first byte */
+   /* There are some optimizations that avoid running the match if a known
+   starting point is not found, or if a known later character is not present.
+   However, there is an option that disables these, for testing and for ensuring
+   that all callouts do actually occur. */
-   if (first_byte >= 0)
+   if ((options & PCRE_NO_START_OPTIMIZE) == 0)
      {
-     if (first_byte_caseless)
+     /* Advance to a unique first byte if there is one. */
-       while (start_match < end_subject &&
-              md->lcc[*start_match] != first_byte)
+     if (first_byte >= 0)
-         start_match++;
+       {
-     else
+       if (first_byte_caseless)
-       while (start_match < end_subject && *start_match != first_byte)
+         while (start_match < end_subject && md->lcc[*start_match] != first_byte)
-         start_match++;
+           start_match++;
-     }
+       else
+         while (start_match < end_subject && *start_match != first_byte)
+           start_match++;
+       }
-   /* Or to just after a linebreak for a multiline match if possible */
+     /* Or to just after a linebreak for a multiline match */
-   else if (startline)
+     else if (startline)
-     {
-     if (start_match > md->start_subject + start_offset)
        {
-       while (start_match <= end_subject && !WAS_NEWLINE(start_match))
+       if (start_match > md->start_subject + start_offset)
-         start_match++;
+         {
+ #ifdef SUPPORT_UTF8
+         if (utf8)
+           {
+           while (start_match < end_subject && !WAS_NEWLINE(start_match))
+             {
+             start_match++;
+             while(start_match < end_subject && (*start_match & 0xc0) == 0x80)
+               start_match++;
+             }
+           }
+         else
+ #endif
+         while (start_match < end_subject && !WAS_NEWLINE(start_match))
+           start_match++;
+         /* If we have just passed a CR and the newline option is ANY or ANYCRLF,
+         and we are now at a LF, advance the match position by one more character.
+         */
+         if (start_match[-1] == CHAR_CR &&
+              (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&
+              start_match < end_subject &&
+              *start_match == CHAR_NL)
+           start_match++;
+         }
        }
-     }
-   /* Or to a non-unique first char after study */
+     /* Or to a non-unique first byte after study */
-   else if (start_bits != NULL)
+     else if (start_bits != NULL)
-     {
-     while (start_match < end_subject)
        {
-       register unsigned int c = *start_match;
+       while (start_match < end_subject)
-       if ((start_bits[c/8] & (1 << (c&7))) == 0) start_match++; else break;
+         {
+         register unsigned int c = *start_match;
+         if ((start_bits[c/8] & (1 << (c&7))) == 0) start_match++;
+           else break;
+         }
        }
-     }
+     }   /* Starting optimizations */
    /* Restore fudged end_subject */
-Line 4037 
 for(;;)
+Line 5230 
 for(;;)
    printf("\n");
  #endif
-   /* If req_byte is set, we know that that character must appear in the subject
+   /* If req_byte is set, we know that that character must appear in the
-   for the match to succeed. If the first character is set, req_byte must be
+   subject for the match to succeed. If the first character is set, req_byte
-   later in the subject; otherwise the test starts at the match point. This
+   must be later in the subject; otherwise the test starts at the match point.
-   optimization can save a huge amount of backtracking in patterns with nested
+   This optimization can save a huge amount of backtracking in patterns with
-   unlimited repeats that aren't going to match. Writing separate code for
+   nested unlimited repeats that aren't going to match. Writing separate code
-   cased/caseless versions makes it go faster, as does using an autoincrement
+   for cased/caseless versions makes it go faster, as does using an
-   and backing off on a match.
+   autoincrement and backing off on a match.
-   HOWEVER: when the subject string is very, very long, searching to its end can
+   HOWEVER: when the subject string is very, very long, searching to its end
-   take a long time, and give bad performance on quite ordinary patterns. This
+   can take a long time, and give bad performance on quite ordinary patterns.
-   showed up when somebody was matching something like /^\d+C/ on a 32-megabyte
+   This showed up when somebody was matching something like /^\d+C/ on a
-   string... so we don't do this when the string is sufficiently long.
+-megabyte string... so we don't do this when the string is sufficiently
+   long.
-   ALSO: this processing is disabled when partial matching is requested.
+   ALSO: this processing is disabled when partial matching is requested, or if
-   */
+   disabling is explicitly requested. */
-   if (req_byte >= 0 &&
+   if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&
+       req_byte >= 0 &&
        end_subject - start_match < REQ_BYTE_MAX &&
        !md->partial)
      {
-Line 4097 
 for(;;)
+Line 5292 
 for(;;)
        }
      }