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

Diff of /code/trunk/pcre_exec.c

Parent Directory | Revision Log | View Patch Patch

-revision 77 by nigel,
Sat Feb 24 21:40:45 2007 UTC
+revision 403 by ph10,
Sat Mar 21 17:33:11 2009 UTC
 Line 6
  and semantics are as close as possible to those of the Perl 5 language.
                         Written by Philip Hazel
-            Copyright (c) 1997-2005 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"
+ /* Undefine some potentially clashing cpp symbols */
- /* Structure for building a chain of data that actually lives on the
+ #undef min
- stack, for holding the values of the subject pointer at the start of each
+ #undef max
- subpattern, so as to detect when an empty string has been matched by a
- subpattern - to break infinite loops. When NO_RECURSE is set, these blocks
- are on the heap, not on the stack. */
- typedef struct eptrblock {
-   struct eptrblock *epb_prev;
-   const uschar *epb_saved_eptr;
- } eptrblock;
  /* Flag bits for the match() function */
- #define match_condassert   0x01    /* Called to check a condition assertion */
+ #define match_condassert     0x01  /* Called to check a condition assertion */
- #define match_isgroup      0x02    /* Set if start of bracketed group */
+ #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */
  /* Non-error returns from the match() function. Error returns are externally
  defined PCRE_ERROR_xxx codes, which are all negative. */
 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 101 
 Returns:     nothing
+Line 109 
 Returns:     nothing
  static void
  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)
  {
- int c;
+ unsigned int c;
  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
  while (length-- > 0)
    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);
-Line 128 
 Returns:      TRUE if matched
+Line 136 
 Returns:      TRUE if matched
  */
  static BOOL
- match_ref(int offset, register const uschar *eptr, int length, match_data *md,
+ match_ref(int offset, register USPTR eptr, int length, match_data *md,
    unsigned long int ims)
  {
- const uschar *p = md->start_subject + md->offset_vector[offset];
+ USPTR p = md->start_subject + md->offset_vector[offset];
  #ifdef DEBUG
  if (eptr >= md->end_subject)
-Line 150 
 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 169 
 return TRUE;
+Line 203 
 return TRUE;
  ****************************************************************************
                     RECURSION IN THE match() FUNCTION
- The match() function is highly recursive. Some regular expressions can cause
+ The match() function is highly recursive, though not every recursive call
- it to recurse thousands of times. I was writing for Unix, so I just let it
+ increases the recursive depth. Nevertheless, some regular expressions can cause
- call itself recursively. This uses the stack for saving everything that has
+ it to recurse to a great depth. I was writing for Unix, so I just let it call
- to be saved for a recursive call. On Unix, the stack can be large, and this
+ itself recursively. This uses the stack for saving everything that has to be
- works fine.
+ saved for a recursive call. On Unix, the stack can be large, and this works
+ fine.
- It turns out that on non-Unix systems there are problems with programs that
- use a lot of stack. (This despite the fact that every last chip has oodles
+ It turns out that on some non-Unix-like systems there are problems with
- of memory these days, and techniques for extending the stack have been known
+ programs that use a lot of stack. (This despite the fact that every last chip
- for decades.) So....
+ has oodles of memory these days, and techniques for extending the stack have
+ been known for decades.) So....
  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
  calls by keeping local variables that need to be preserved in blocks of memory
- obtained from malloc instead instead of on the stack. Macros are used to
+ 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.  */
- /* These versions of the macros use the stack, as normal */
+ 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. Note that the "rw" argument of RMATCH isn't
+ actuall used in this definition. */
  #ifndef NO_RECURSE
  #define REGISTER register
- #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg)
+ #ifdef DEBUG
+ #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \
+   { \
+   printf("match() called in line %d\n", __LINE__); \
+   rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \
+   printf("to line %d\n", __LINE__); \
+   }
+ #define RRETURN(ra) \
+   { \
+   printf("match() returned %d from line %d ", ra, __LINE__); \
+   return ra; \
+   }
+ #else
+ #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \
+   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->Xprevframe = frame;\
+   newframe->Xrdepth = frame->Xrdepth + 1;\
-     frame = newframe;\
+   newframe->Xprevframe = frame;\
-     DPRINTF(("restarting from line %d\n", __LINE__));\
+   frame = newframe;\
-     goto HEAP_RECURSE;\
+   DPRINTF(("restarting from line %d\n", __LINE__));\
-     }\
+   goto HEAP_RECURSE;\
-   else\
+   L_##rw:\
-     {\
+   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 235 
 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 252 
 typedef struct heapframe {
+Line 324 
 typedef struct heapframe {
    const uschar *Xeptr;
    const uschar *Xecode;
+   const uschar *Xmstart;
    int Xoffset_top;
    long int Xims;
    eptrblock *Xeptrb;
    int Xflags;
+   unsigned int Xrdepth;
    /* Function local variables */
    const uschar *Xcallpat;
+ #ifdef SUPPORT_UTF8
    const uschar *Xcharptr;
+ #endif
    const uschar *Xdata;
    const uschar *Xnext;
    const uschar *Xpp;
-Line 271 
 typedef struct heapframe {
+Line 347 
 typedef struct heapframe {
    BOOL Xcur_is_word;
    BOOL Xcondition;
-   BOOL Xminimize;
    BOOL Xprev_is_word;
    unsigned long int Xoriginal_ims;
  #ifdef SUPPORT_UCP
    int Xprop_type;
+   int Xprop_value;
    int Xprop_fail_result;
    int Xprop_category;
    int Xprop_chartype;
-   int Xprop_othercase;
+   int Xprop_script;
-   int Xprop_test_against;
+   int Xoclength;
-   int *Xprop_test_variable;
+   uschar Xocchars[8];
  #endif
+   int Xcodelink;
    int Xctype;
-   int Xfc;
+   unsigned int Xfc;
    int Xfi;
    int Xlength;
    int Xmax;
-Line 301 
 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 320 
 typedef struct heapframe {
+Line 396 
 typedef struct heapframe {
  *         Match from current position            *
  *************************************************/
- /* On entry ecode points to the first opcode, and eptr to the first character
+ /* This function is called recursively in many circumstances. Whenever it
- in the subject string, while eptrb holds the value of eptr at the start of the
- last bracketed group - used for breaking infinite loops matching zero-length
- strings. This function is called recursively in many circumstances. Whenever it
  returns a negative (error) response, the outer incarnation must also return the
  same response.
-Line 333 
 performance. Tests using gcc on a SPARC
+Line 406 
 performance. Tests using gcc on a SPARC
  made performance worse.
  Arguments:
-    eptr        pointer in subject
+    eptr        pointer to current character in subject
-    ecode       position in code
+    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 342 
 Arguments:
+Line 417 
 Arguments:
                   brackets - for testing for empty matches
     flags       can contain
                   match_condassert - this is an assertion condition
-                  match_isgroup - this is the start of a bracketed group
+                  match_cbegroup - this is the start of an unlimited repeat
+                    group that can match an empty string
+    rdepth      the recursion depth
  Returns:       MATCH_MATCH if matched            )  these values are >= 0
                 MATCH_NOMATCH if failed to match  )
                 a negative PCRE_ERROR_xxx value if aborted by an error condition
-                  (e.g. stopped by recursion limit)
+                  (e.g. stopped by repeated call or recursion limit)
  */
  static int
- match(REGISTER const uschar *eptr, REGISTER const uschar *ecode,
+ match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,
    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,
-   int flags)
+   int flags, unsigned int rdepth)
  {
  /* These variables do not need to be preserved over recursion in this function,
- so they can be ordinary variables in all cases. Mark them with "register"
+ so they can be ordinary variables in all cases. Mark some of them with
- because they are used a lot in loops. */
+ "register" because they are used a lot in loops. */
+ register int  rrc;         /* Returns from recursive calls */
+ register int  i;           /* Used for loops not involving calls to RMATCH() */
+ register unsigned int c;   /* Character values not kept over RMATCH() calls */
+ register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
- register int  rrc;    /* Returns from recursive calls */
+ BOOL minimize, possessive; /* Quantifier options */
- register int  i;      /* Used for loops not involving calls to RMATCH() */
+ int condcode;
- register int  c;      /* Character values not kept over RMATCH() calls */
- register BOOL utf8;   /* Local copy of UTF-8 flag for speed */
  /* 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 377 
 frame->Xprevframe = NULL;            /*
+Line 457 
 frame->Xprevframe = NULL;            /*
  frame->Xeptr = eptr;
  frame->Xecode = ecode;
+ frame->Xmstart = mstart;
  frame->Xoffset_top = offset_top;
  frame->Xims = ims;
  frame->Xeptrb = eptrb;
  frame->Xflags = flags;
+ frame->Xrdepth = rdepth;
  /* This is where control jumps back to to effect "recursion" */
-Line 390 
 HEAP_RECURSE:
+Line 472 
 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
  #define flags              frame->Xflags
+ #define rdepth             frame->Xrdepth
  /* Ditto for the local variables */
-Line 401 
 HEAP_RECURSE:
+Line 485 
 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 411 
 HEAP_RECURSE:
+Line 496 
 HEAP_RECURSE:
  #define cur_is_word        frame->Xcur_is_word
  #define condition          frame->Xcondition
- #define minimize           frame->Xminimize
  #define prev_is_word       frame->Xprev_is_word
  #define original_ims       frame->Xoriginal_ims
  #ifdef SUPPORT_UCP
  #define prop_type          frame->Xprop_type
+ #define prop_value         frame->Xprop_value
  #define prop_fail_result   frame->Xprop_fail_result
  #define prop_category      frame->Xprop_category
  #define prop_chartype      frame->Xprop_chartype
- #define prop_othercase     frame->Xprop_othercase
+ #define prop_script        frame->Xprop_script
- #define prop_test_against  frame->Xprop_test_against
+ #define oclength           frame->Xoclength
- #define prop_test_variable frame->Xprop_test_variable
+ #define occhars            frame->Xocchars
  #endif
  #define ctype              frame->Xctype
-Line 447 
 HEAP_RECURSE:
+Line 532 
 HEAP_RECURSE:
  get preserved during recursion in the normal way. In this environment, fi and
  i, and fc and c, can be the same variables. */
- #else
+ #else         /* NO_RECURSE not defined */
  #define fi i
  #define fc c
- #ifdef SUPPORT_UTF8                /* Many of these variables are used ony */
+ #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */
- const uschar *charptr;             /* small blocks of the code. My normal  */
+ const uschar *charptr;             /* in small blocks of the code. My normal */
- #endif                             /* style of coding would have declared  */
+ #endif                             /* style of coding would have declared    */
- const uschar *callpat;             /* them within each of those blocks.    */
+ const uschar *callpat;             /* them within each of those blocks.      */
- const uschar *data;                /* However, in order to accommodate the */
+ const uschar *data;                /* However, in order to accommodate the   */
- const uschar *next;                /* version of this code that uses an    */
+ const uschar *next;                /* version of this code that uses an      */
- const uschar *pp;                  /* external "stack" implemented on the  */
+ USPTR         pp;                  /* external "stack" implemented on the    */
- const uschar *prev;                /* heap, it is easier to declare them   */
+ const uschar *prev;                /* heap, it is easier to declare them all */
- const uschar *saved_eptr;          /* all here, so the declarations can    */
+ USPTR         saved_eptr;          /* here, so the declarations can be cut   */
-                                    /* be cut out in a block. The only      */
+                                    /* out in a block. The only declarations  */
- recursion_info new_recursive;      /* declarations within blocks below are */
+ recursion_info new_recursive;      /* within blocks below are for variables  */
-                                    /* for variables that do not have to    */
+                                    /* that do not have to be preserved over  */
- BOOL cur_is_word;                  /* be preserved over a recursive call   */
+ BOOL cur_is_word;                  /* a recursive call to RMATCH().          */
- BOOL condition;                    /* to RMATCH().                         */
+ BOOL condition;
- BOOL minimize;
  BOOL prev_is_word;
  unsigned long int original_ims;
  #ifdef SUPPORT_UCP
  int prop_type;
+ int prop_value;
  int prop_fail_result;
  int prop_category;
  int prop_chartype;
- int prop_othercase;
+ int prop_script;
- int prop_test_against;
+ int oclength;
- int *prop_test_variable;
+ uschar occhars[8];
  #endif
+ int codelink;
  int ctype;
  int length;
  int max;
-Line 493 
 int save_offset1, save_offset2, save_off
+Line 579 
 int save_offset1, save_offset2, save_off
  int stacksave[REC_STACK_SAVE_MAX];
  eptrblock newptrb;
- #endif
+ #endif     /* NO_RECURSE */
  /* These statements are here to stop the compiler complaining about unitialized
  variables. */
  #ifdef SUPPORT_UCP
+ prop_value = 0;
  prop_fail_result = 0;
- prop_test_against = 0;
- prop_test_variable = NULL;
  #endif
- /* OK, now we can get on with the real code of the function. Recursion is
- specified by the macros RMATCH and RRETURN. When NO_RECURSE is *not* defined,
+ /* This label is used for tail recursion, which is used in a few cases even
- these just turn into a recursive call to match() and a "return", respectively.
+ when NO_RECURSE is not defined, in order to reduce the amount of stack that is
- However, RMATCH isn't like a function call because it's quite a complicated
+ used. Thanks to Ian Taylor for noticing this possibility and sending the
- macro. It has to be used in one particular way. This shouldn't, however, impact
+ original patch. */
- performance when true recursion is being used. */
+ TAIL_RECURSE:
+ /* OK, now we can get on with the real code of the function. Recursive calls
+ are specified by the macro RMATCH and RRETURN is used to return. When
+ NO_RECURSE is *not* defined, these just turn into a recursive call to match()
+ and a "return", respectively (possibly with some debugging if DEBUG is
+ defined). However, RMATCH isn't like a function call because it's quite a
+ complicated macro. It has to be used in one particular way. This shouldn't,
+ however, impact performance when true recursion is being used. */
+ #ifdef SUPPORT_UTF8
+ utf8 = md->utf8;       /* Local copy of the flag */
+ #else
+ utf8 = FALSE;
+ #endif
+ /* First check that we haven't called match() too many times, or that we
+ haven't exceeded the recursive call limit. */
  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
+ if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
  original_ims = ims;    /* Save for resetting on ')' */
- utf8 = md->utf8;       /* Local copy of the flag */
- /* At the start of a bracketed group, add the current subject pointer to the
+ /* At the start of a group with an unlimited repeat that may match an empty
- stack of such pointers, to be re-instated at the end of the group when we hit
+ string, the match_cbegroup flag is set. When this is the case, add the current
- the closing ket. When match() is called in other circumstances, we don't add to
+ subject pointer to the chain of such remembered pointers, to be checked when we
- this stack. */
+ 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. The
+ match_cbegroup flag must NOT be used with tail recursion, because the memory
+ block that is used is on the stack, so a new one may be required for each
+ match(). */
- if ((flags & match_isgroup) != 0)
+ if ((flags & match_cbegroup) != 0)
    {
-   newptrb.epb_prev = eptrb;
    newptrb.epb_saved_eptr = eptr;
+   newptrb.epb_prev = eptrb;
    eptrb = &newptrb;
    }
- /* Now start processing the operations. */
+ /* Now start processing the opcodes. */
  for (;;)
    {
+   minimize = possessive = FALSE;
    op = *ecode;
-   minimize = FALSE;
    /* 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)
+       eptr > mstart)
      md->hitend = TRUE;
-   /* Opening capturing bracket. If there is space in the offset vector, save
+   switch(op)
-   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 set
-   from a previous iteration of this group, and be referred to by a reference
-   inside the group.
-   If the bracket fails to match, we need to restore this value and also the
-   values of the final offsets, in case they were set by a previous iteration of
-   the same bracket.
-   If there isn't enough space in the offset vector, treat this as if it were a
-   non-capturing bracket. Don't worry about setting the flag for the error case
-   here; that is handled in the code for KET. */
-   if (op > OP_BRA)
      {
-     number = op - OP_BRA;
+     case OP_FAIL:
+     RRETURN(MATCH_NOMATCH);
-     /* For extended extraction brackets (large number), we have to fish out the
-     number from a dummy opcode at the start. */
-     if (number > EXTRACT_BASIC_MAX)
+     case OP_PRUNE:
-       number = GET2(ecode, 2+LINK_SIZE);
+     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
+     set from a previous iteration of this group, and be referred to by a
+     reference inside the group.
+     If the bracket fails to match, we need to restore this value and also the
+     values of the final offsets, in case they were set by a previous iteration
+     of the same bracket.
+     If there isn't enough space in the offset vector, treat this as if it were
+     a non-capturing bracket. Don't worry about setting the flag for the error
+     case here; that is handled in the code for KET. */
+     case OP_CBRA:
+     case OP_SCBRA:
+     number = GET2(ecode, 1+LINK_SIZE);
      offset = number << 1;
  #ifdef DEBUG
-     printf("start bracket %d subject=", number);
+     printf("start bracket %d\n", number);
+     printf("subject=");
      pchars(eptr, 16, TRUE, md);
      printf("\n");
  #endif
-Line 584 
 for (;;)
+Line 716 
 for (;;)
        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;
+       flags = (op == OP_SCBRA)? match_cbegroup : 0;
        do
          {
-         RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,
+         RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
-           match_isgroup);
+           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 603 
 for (;;)
+Line 736 
 for (;;)
        RRETURN(MATCH_NOMATCH);
        }
-     /* Insufficient room for saving captured contents */
+     /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
+     as a non-capturing bracket. */
-     else op = OP_BRA;
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
-     }
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
-   /* Other types of node can be handled by a switch */
+     DPRINTF(("insufficient capture room: treat as non-capturing\n"));
-   switch(op)
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
-     {
+     /* VVVVVVVVVVVVVVVVVVVVVVVVV */
-     case OP_BRA:     /* Non-capturing bracket: optimized */
-     DPRINTF(("start bracket 0\n"));
+     /* Non-capturing bracket. Loop for all the alternatives. When we get to the
-     do
+     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, except in the case when match_cbegroup
+     is set.*/
+     case OP_BRA:
+     case OP_SBRA:
+     DPRINTF(("start non-capturing bracket\n"));
+     flags = (op >= OP_SBRA)? match_cbegroup : 0;
+     for (;;)
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,
+       if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */
-         match_isgroup);
+         {
-       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+         if (flags == 0)    /* Not a possibly empty group */
+           {
+           ecode += _pcre_OP_lengths[*ecode];
+           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(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
+         eptrb, flags, RM2);
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        ecode += GET(ecode, 1);
        }
-     while (*ecode == OP_ALT);
+     /* Control never reaches here. */
-     DPRINTF(("bracket 0 failed\n"));
-     RRETURN(MATCH_NOMATCH);
      /* Conditional group: compilation checked that there are no more than
      two branches. If the condition is false, skipping the first branch takes us
      past the end if there is only one branch, but that's OK because that is
-     exactly what going to the ket would do. */
+     exactly what going to the ket would do. As there is only one branch to be
+     obeyed, we can use tail recursion to avoid using another stack frame. */
      case OP_COND:
-     if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */
+     case OP_SCOND:
+     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 &&
+         (offset == RREF_ANY || offset == md->recursive->group_num);
+       ecode += condition? 3 : GET(ecode, 1);
+       }
+     else if (condcode == OP_CREF)    /* Group used test */
        {
        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
-       condition = (offset == CREF_RECURSE * 2)?
+       condition = offset < offset_top && md->offset_vector[offset] >= 0;
-         (md->recursive != NULL) :
+       ecode += condition? 3 : GET(ecode, 1);
-         (offset < offset_top && md->offset_vector[offset] >= 0);
+       }
-       RMATCH(rrc, eptr, ecode + (condition?
-         (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1))),
+     else if (condcode == OP_DEF)     /* DEFINE - always false */
-         offset_top, md, ims, eptrb, match_isgroup);
+       {
-       RRETURN(rrc);
+       condition = FALSE;
+       ecode += GET(ecode, 1);
        }
      /* The condition is an assertion. Call match() to evaluate it - setting
-     the final argument TRUE causes it to stop at the end of an assertion. */
+     the final argument match_condassert causes it to stop at the end of an
+     assertion. */
      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_isgroup);
+           match_condassert, RM3);
        if (rrc == MATCH_MATCH)
          {
-         ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);
+         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 ecode += GET(ecode, 1);
+       else
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,
+         {
-         match_isgroup);
+         condition = FALSE;
-       RRETURN(rrc);
+         ecode += codelink;
+         }
        }
-     /* Control never reaches here */
-     /* Skip over conditional reference or large extraction number data if
+     /* We are now at the branch that is to be obeyed. As there is only one,
-     encountered. */
+     we can use tail recursion to avoid using another stack frame, except when
+     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. */
-     case OP_CREF:
+     if (condition || *ecode == OP_ALT)
-     case OP_BRANUMBER:
+       {
-     ecode += 3;
+       ecode += 1 + LINK_SIZE;
+       if (op == OP_SCOND)        /* Possibly empty group */
+         {
+         RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);
+         RRETURN(rrc);
+         }
+       else                       /* Group must match something */
+         {
+         flags = 0;
+         goto TAIL_RECURSE;
+         }
+       }
+     else                         /* Condition false & no alternative */
+       {
+       ecode += 1 + LINK_SIZE;
+       }
      break;
-     /* End of the pattern. If we are in a recursion, we should restore the
-     offsets appropriately and continue from after the call. */
+     /* End of the pattern, either real or forced. If we are in a top-level
+     recursion, we should restore the offsets appropriately and continue from
+     after the call. */
+     case OP_ACCEPT:
      case OP_END:
      if (md->recursive != NULL && md->recursive->group_num == 0)
        {
        recursion_info *rec = md->recursive;
-       DPRINTF(("Hit the end in a (?0) recursion\n"));
+       DPRINTF(("End of pattern in a (?0) recursion\n"));
        md->recursive = rec->prevrec;
        memmove(md->offset_vector, rec->offset_save,
          rec->saved_max * sizeof(int));
-       md->start_match = rec->save_start;
+       mstart = rec->save_start;
        ims = original_ims;
        ecode = rec->after_call;
        break;
-Line 694 
 for (;;)
+Line 919 
 for (;;)
      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty
      string - backtracking will then try other alternatives, if any. */
-     if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);
+     if (md->notempty && eptr == mstart) 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 717 
 for (;;)
+Line 943 
 for (;;)
      case OP_ASSERTBACK:
      do
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
-         match_isgroup);
+         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 744 
 for (;;)
+Line 970 
 for (;;)
      case OP_ASSERTBACK_NOT:
      do
        {
-       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
-         match_isgroup);
+         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 766 
 for (;;)
+Line 992 
 for (;;)
  #ifdef SUPPORT_UTF8
      if (utf8)
        {
-       c = GET(ecode,1);
+       i = GET(ecode, 1);
-       for (i = 0; i < c; i++)
+       while (i-- > 0)
          {
          eptr--;
          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
-         BACKCHAR(eptr)
+         BACKCHAR(eptr);
          }
        }
      else
-Line 780 
 for (;;)
+Line 1006 
 for (;;)
      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
        {
-       eptr -= GET(ecode,1);
+       eptr -= GET(ecode, 1);
        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
        }
-Line 800 
 for (;;)
+Line 1026 
 for (;;)
        cb.version          = 1;   /* Version 1 of the callout block */
        cb.callout_number   = ecode[1];
        cb.offset_vector    = md->offset_vector;
-       cb.subject          = (const char *)md->start_subject;
+       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 837 
 for (;;)
+Line 1063 
 for (;;)
      case OP_RECURSE:
        {
        callpat = md->start_code + GET(ecode, 1);
-       new_recursive.group_num = *callpat - OP_BRA;
+       new_recursive.group_num = (callpat == md->start_code)? 0 :
+         GET2(callpat, 1 + LINK_SIZE);
-       /* For extended extraction brackets (large number), we have to fish out
-       the number from a dummy opcode at the start. */
-       if (new_recursive.group_num > EXTRACT_BASIC_MAX)
-         new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);
        /* Add to "recursing stack" */
-Line 869 
 for (;;)
+Line 1090 
 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;
+       mstart = eptr;
        /* OK, now we can do the recursion. For each top-level alternative we
        restore the offset and recursion data. */
        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
+       flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;
        do
          {
-         RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,
+         RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
-             eptrb, match_isgroup);
+           md, ims, eptrb, flags, RM6);
          if (rrc == MATCH_MATCH)
            {
+           DPRINTF(("Recursion matched\n"));
            md->recursive = new_recursive.prevrec;
            if (new_recursive.offset_save != stacksave)
              (pcre_free)(new_recursive.offset_save);
            RRETURN(MATCH_MATCH);
            }
-         else if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+         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);
+           }
          md->recursive = &new_recursive;
          memcpy(md->offset_vector, new_recursive.offset_save,
-Line 912 
 for (;;)
+Line 1141 
 for (;;)
      the end of a normal bracket, leaving the subject pointer. */
      case OP_ONCE:
-       {
+     prev = ecode;
-       prev = ecode;
+     saved_eptr = eptr;
-       saved_eptr = eptr;
-       do
+     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, match_isgroup);
+       if (rrc == MATCH_MATCH) break;
-         if (rrc == MATCH_MATCH) break;
+       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       ecode += GET(ecode,1);
-         ecode += GET(ecode,1);
+       }
-         }
+     while (*ecode == OP_ALT);
-       while (*ecode == OP_ALT);
-       /* If hit the end of the group (which could be repeated), fail */
+     /* If hit the end of the group (which could be repeated), fail */
-       if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
+     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
-       /* Continue as from after the assertion, updating the offsets high water
+     /* Continue as from after the assertion, updating the offsets high water
-       mark, since extracts may have been taken. */
+     mark, since extracts may have been taken. */
-       do ecode += GET(ecode,1); while (*ecode == OP_ALT);
+     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
-       offset_top = md->end_offset_top;
+     offset_top = md->end_offset_top;
-       eptr = md->end_match_ptr;
+     eptr = md->end_match_ptr;
-       /* For a non-repeating ket, just continue at this level. This also
+     /* For a non-repeating ket, just continue at this level. This also
-       happens for a repeating ket if no characters were matched in the group.
+     happens for a repeating ket if no characters were matched in the group.
-       This is the forcible breaking of infinite loops as implemented in Perl
+     This is the forcible breaking of infinite loops as implemented in Perl
 .005. If there is an options reset, it will get obeyed in the normal
-       course of events. */
+     course of events. */
-       if (*ecode == OP_KET || eptr == saved_eptr)
+     if (*ecode == OP_KET || eptr == saved_eptr)
-         {
+       {
-         ecode += 1+LINK_SIZE;
+       ecode += 1+LINK_SIZE;
-         break;
+       break;
-         }
+       }
-       /* The repeating kets try the rest of the pattern or restart from the
+     /* The repeating kets try the rest of the pattern or restart from the
-       preceding bracket, in the appropriate order. We need to reset any options
+     preceding bracket, in the appropriate order. The second "call" of match()
-       that changed within the bracket before re-running it, so check the next
+     uses tail recursion, to avoid using another stack frame. We need to reset
-       opcode. */
+     any options that changed within the bracket before re-running it, so
+     check the next opcode. */
-       if (ecode[1+LINK_SIZE] == OP_OPT)
+     if (ecode[1+LINK_SIZE] == OP_OPT)
-         {
+       {
-         ims = (ims & ~PCRE_IMS) | ecode[4];
+       ims = (ims & ~PCRE_IMS) | ecode[4];
-         DPRINTF(("ims set to %02lx at group repeat\n", ims));
+       DPRINTF(("ims set to %02lx at group repeat\n", ims));
-         }
+       }
-       if (*ecode == OP_KETRMIN)
+     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);
+       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
+       ecode = prev;
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       flags = 0;
-         }
+       goto TAIL_RECURSE;
-       else  /* OP_KETRMAX */
-         {
-         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-         RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-         }
        }
-     RRETURN(MATCH_NOMATCH);
+     else  /* OP_KETRMAX */
+       {
+       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);
+       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       ecode += 1 + LINK_SIZE;
+       flags = 0;
+       goto TAIL_RECURSE;
+       }
+     /* Control never gets here */
      /* An alternation is the end of a branch; scan along to find the end of the
      bracketed group and go to there. */
-Line 985 
 for (;;)
+Line 1214 
 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, match_isgroup);
+       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;
+       ecode = next + 1 + LINK_SIZE;
        }
      break;
      case OP_BRAMINZERO:
        {
        next = ecode+1;
-       do next += GET(next,1); while (*next == OP_ALT);
+       do next += GET(next, 1); while (*next == OP_ALT);
-       RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,
+       RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);
-         match_isgroup);
        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
        ecode++;
        }
      break;
-     /* End of a group, repeated or non-repeating. If we are at the end of
+     case OP_SKIPZERO:
-     an assertion "group", stop matching and return MATCH_MATCH, but record the
+       {
-     current high water mark for use by positive assertions. Do this also
+       next = ecode+1;
-     for the "once" (not-backup up) groups. */
+       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:
      case OP_KETRMIN:
      case OP_KETRMAX:
-       {
+     prev = ecode - GET(ecode, 1);
-       prev = ecode - GET(ecode, 1);
-       saved_eptr = eptrb->epb_saved_eptr;
-       /* Back up the stack of bracket start pointers. */
+     /* If this was a group that remembered the subject start, in order to break
+     infinite repeats of empty string matches, retrieve the subject start from
+     the chain. Otherwise, set it NULL. */
-       eptrb = eptrb->epb_prev;
+     if (*prev >= OP_SBRA)
+       {
-       if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
+       saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
-           *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
+       eptrb = eptrb->epb_prev;              /* Backup to previous group */
-           *prev == OP_ONCE)
+       }
-         {
+     else saved_eptr = NULL;
-         md->end_match_ptr = eptr;      /* For ONCE */
-         md->end_offset_top = offset_top;
-         RRETURN(MATCH_MATCH);
-         }
-       /* In all other cases except a conditional group we have to check the
+     /* If we are at the end of an assertion group, stop matching and return
-       group number back at the start and if necessary complete handling an
+     MATCH_MATCH, but record the current high water mark for use by positive
-       extraction by setting the offsets and bumping the high water mark. */
+     assertions. Do this also for the "once" (atomic) groups. */
-       if (*prev != OP_COND)
+     if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
-         {
+         *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
-         number = *prev - OP_BRA;
+         *prev == OP_ONCE)
+       {
+       md->end_match_ptr = eptr;      /* For ONCE */
+       md->end_offset_top = offset_top;
+       RRETURN(MATCH_MATCH);
+       }
-         /* For extended extraction brackets (large number), we have to fish out
+     /* For capturing groups we have to check the group number back at the start
-         the number from a dummy opcode at the start. */
+     and if necessary complete handling an extraction by setting the offsets and
+     bumping the high water mark. Note that whole-pattern recursion is coded as
+     a recurse into group 0, so it won't be picked up here. Instead, we catch it
+     when the OP_END is reached. Other recursion is handled here. */
-         if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);
+     if (*prev == OP_CBRA || *prev == OP_SCBRA)
-         offset = number << 1;
+       {
+       number = GET2(prev, 1+LINK_SIZE);
+       offset = number << 1;
  #ifdef DEBUG
-         printf("end bracket %d", number);
+       printf("end bracket %d", number);
-         printf("\n");
+       printf("\n");
  #endif
-         /* Test for a numbered group. This includes groups called as a result
+       md->capture_last = number;
-         of recursion. Note that whole-pattern recursion is coded as a recurse
+       if (offset >= md->offset_max) md->offset_overflow = TRUE; else
-         into group 0, so it won't be picked up here. Instead, we catch it when
+         {
-         the OP_END is reached. */
+         md->offset_vector[offset] =
+           md->offset_vector[md->offset_end - number];
-         if (number > 0)
+         md->offset_vector[offset+1] = eptr - md->start_subject;
-           {
+         if (offset_top <= offset) offset_top = offset + 2;
-           md->capture_last = number;
+         }
-           if (offset >= md->offset_max) md->offset_overflow = TRUE; else
-             {
+       /* Handle a recursively called group. Restore the offsets
-             md->offset_vector[offset] =
+       appropriately and continue from after the call. */
-               md->offset_vector[md->offset_end - number];
-             md->offset_vector[offset+1] = eptr - md->start_subject;
+       if (md->recursive != NULL && md->recursive->group_num == number)
-             if (offset_top <= offset) offset_top = offset + 2;
+         {
-             }
+         recursion_info *rec = md->recursive;
+         DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
-           /* Handle a recursively called group. Restore the offsets
+         md->recursive = rec->prevrec;
-           appropriately and continue from after the call. */
+         mstart = rec->save_start;
+         memcpy(md->offset_vector, rec->offset_save,
-           if (md->recursive != NULL && md->recursive->group_num == number)
+           rec->saved_max * sizeof(int));
-             {
+         ecode = rec->after_call;
-             recursion_info *rec = md->recursive;
+         ims = original_ims;
-             DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
+         break;
-             md->recursive = rec->prevrec;
-             md->start_match = rec->save_start;
-             memcpy(md->offset_vector, rec->offset_save,
-               rec->saved_max * sizeof(int));
-             ecode = rec->after_call;
-             ims = original_ims;
-             break;
-             }
-           }
          }
+       }
-       /* Reset the value of the ims flags, in case they got changed during
+     /* For both capturing and non-capturing groups, reset the value of the ims
-       the group. */
+     flags, in case they got changed during the group. */
-       ims = original_ims;
+     ims = original_ims;
-       DPRINTF(("ims reset to %02lx\n", ims));
+     DPRINTF(("ims reset to %02lx\n", ims));
-       /* For a non-repeating ket, just continue at this level. This also
+     /* For a non-repeating ket, just continue at this level. This also
-       happens for a repeating ket if no characters were matched in the group.
+     happens for a repeating ket if no characters were matched in the group.
-       This is the forcible breaking of infinite loops as implemented in Perl
+     This is the forcible breaking of infinite loops as implemented in Perl
 .005. If there is an options reset, it will get obeyed in the normal
-       course of events. */
+     course of events. */
-       if (*ecode == OP_KET || eptr == saved_eptr)
+     if (*ecode == OP_KET || eptr == saved_eptr)
-         {
+       {
-         ecode += 1 + LINK_SIZE;
+       ecode += 1 + LINK_SIZE;
-         break;
+       break;
-         }
+       }
-       /* The repeating kets try the rest of the pattern or restart from the
+     /* The repeating kets try the rest of the pattern or restart from the
-       preceding bracket, in the appropriate order. */
+     preceding bracket, in the appropriate order. In the second case, we can use
+     tail recursion to avoid using another stack frame, unless we have an
+     unlimited repeat of a group that can match an empty string. */
-       if (*ecode == OP_KETRMIN)
+     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;
-         {
-         RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
+     if (*ecode == OP_KETRMIN)
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       {
-         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
+       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-         }
+       if (flags != 0)    /* Could match an empty string */
-       else  /* OP_KETRMAX */
          {
-         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
+         RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+         RRETURN(rrc);
-         RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
-         if (rrc != MATCH_NOMATCH) RRETURN(rrc);
          }
+       ecode = prev;
+       goto TAIL_RECURSE;
        }
+     else  /* OP_KETRMAX */
-     RRETURN(MATCH_NOMATCH);
+       {
+       RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);
+       if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+       ecode += 1 + LINK_SIZE;
+       flags = 0;
+       goto TAIL_RECURSE;
+       }
+     /* Control never gets here */
      /* Start of subject unless notbol, or after internal newline if multiline */
-Line 1135 
 for (;;)
+Line 1374 
 for (;;)
      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
      if ((ims & PCRE_MULTILINE) != 0)
        {
-       if (eptr != md->start_subject && eptr[-1] != NEWLINE)
+       if (eptr != md->start_subject &&
+           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
          RRETURN(MATCH_NOMATCH);
        ecode++;
        break;
-Line 1156 
 for (;;)
+Line 1396 
 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 1163 
 for (;;)
+Line 1410 
 for (;;)
      if ((ims & PCRE_MULTILINE) != 0)
        {
        if (eptr < md->end_subject)
-         { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }
+         { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
        else
          { if (md->noteol) RRETURN(MATCH_NOMATCH); }
        ecode++;
-Line 1174 
 for (;;)
+Line 1421 
 for (;;)
        if (md->noteol) RRETURN(MATCH_NOMATCH);
        if (!md->endonly)
          {
-         if (eptr < md->end_subject - 1 ||
+         if (eptr != md->end_subject &&
-            (eptr == md->end_subject - 1 && *eptr != NEWLINE))
+             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
            RRETURN(MATCH_NOMATCH);
          ecode++;
          break;
          }
        }
-     /* ... else fall through */
+     /* ... else fall through for endonly */
      /* End of subject assertion (\z) */
-Line 1193 
 for (;;)
+Line 1440 
 for (;;)
      /* End of subject or ending \n assertion (\Z) */
      case OP_EODN:
-     if (eptr < md->end_subject - 1 ||
+     if (eptr != md->end_subject &&
-        (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);
+         (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
+       RRETURN(MATCH_NOMATCH);
      ecode++;
      break;
-Line 1247 
 for (;;)
+Line 1495 
 for (;;)
      /* Match a single character type; inline for speed */
      case OP_ANY:
-     if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)
+     if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
-       RRETURN(MATCH_NOMATCH);
+     /* Fall through */
+     case OP_ALLANY:
      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);
- #ifdef SUPPORT_UTF8
+     if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-     if (utf8)
-       while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
- #endif
      ecode++;
      break;
-Line 1343 
 for (;;)
+Line 1590 
 for (;;)
      ecode++;
      break;
+     case OP_ANYNL:
+     if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+     GETCHARINCTEST(c, eptr);
+     switch(c)
+       {
+       default: RRETURN(MATCH_NOMATCH);
+       case 0x000d:
+       if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
+       break;
+       case 0x000a:
+       break;
+       case 0x000b:
+       case 0x000c:
+       case 0x0085:
+       case 0x2028:
+       case 0x2029:
+       if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
+       break;
+       }
+     ecode++;
+     break;
+     case OP_NOT_HSPACE:
+     if (eptr >= md->end_subject) 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) 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) 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) 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++;
+     break;
  #ifdef SUPPORT_UCP
      /* Check the next character by Unicode property. We will get here only
      if the support is in the binary; otherwise a compile-time error occurs. */
-Line 1352 
 for (;;)
+Line 1719 
 for (;;)
      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
      GETCHARINCTEST(c, eptr);
        {
-       int chartype, rqdtype;
+       const ucd_record *prop = GET_UCD(c);
-       int othercase;
-       int category = ucp_findchar(c, &chartype, &othercase);
-       rqdtype = *(++ecode);
-       ecode++;
-       if (rqdtype >= 128)
+       switch(ecode[1])
          {
-         if ((rqdtype - 128 != category) == (op == OP_PROP))
+         case PT_ANY:
+         if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
+         break;
+         case PT_LAMP:
+         if ((prop->chartype == ucp_Lu ||
+              prop->chartype == ucp_Ll ||
+              prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
            RRETURN(MATCH_NOMATCH);
-         }
+          break;
-       else
-         {
+         case PT_GC:
-         if ((rqdtype != chartype) == (op == OP_PROP))
+         if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
+           RRETURN(MATCH_NOMATCH);
+         break;
+         case PT_PC:
+         if ((ecode[2] != prop->chartype) == (op == OP_PROP))
+           RRETURN(MATCH_NOMATCH);
+         break;
+         case PT_SC:
+         if ((ecode[2] != prop->script) == (op == OP_PROP))
            RRETURN(MATCH_NOMATCH);
+         break;
+         default:
+         RRETURN(PCRE_ERROR_INTERNAL);
          }
+       ecode += 3;
        }
      break;
-Line 1379 
 for (;;)
+Line 1764 
 for (;;)
      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
      GETCHARINCTEST(c, eptr);
        {
-       int chartype;
+       int category = UCD_CATEGORY(c);
-       int othercase;
-       int category = ucp_findchar(c, &chartype, &othercase);
        if (category == ucp_M) RRETURN(MATCH_NOMATCH);
        while (eptr < md->end_subject)
          {
-Line 1390 
 for (;;)
+Line 1773 
 for (;;)
            {
            GETCHARLEN(c, eptr, len);
            }
-         category = ucp_findchar(c, &chartype, &othercase);
+         category = UCD_CATEGORY(c);
          if (category != ucp_M) break;
          eptr += len;
          }
-Line 1411 
 for (;;)
+Line 1794 
 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
-       minima. */
-       length = (offset >= offset_top || md->offset_vector[offset] < 0)?
-         md->end_subject - eptr + 1 :
-         md->offset_vector[offset+1] - md->offset_vector[offset];
-       /* Set up for repetition, or handle the non-repeated case */
+       (a) In the default, Perl-compatible state, set the length to be longer
+       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
+       quantifiers with zero minima.
-       switch (*ecode)
+       (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 */
+       switch (*ecode)
          {
          case OP_CRSTAR:
          case OP_CRMINSTAR:
-Line 1480 
 for (;;)
+Line 1872 
 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))
              RRETURN(MATCH_NOMATCH);
-Line 1501 
 for (;;)
+Line 1893 
 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 1606 
 for (;;)
+Line 1998 
 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);
              GETCHARINC(c, eptr);
-Line 1626 
 for (;;)
+Line 2018 
 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);
              c = *eptr++;
-Line 1663 
 for (;;)
+Line 2055 
 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 1682 
 for (;;)
+Line 2074 
 for (;;)
              }
            while (eptr >= pp)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);
-             eptr--;
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+             eptr--;
              }
            }
-Line 1695 
 for (;;)
+Line 2087 
 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 1737 
 for (;;)
+Line 2130 
 for (;;)
        for (i = 1; i <= min; i++)
          {
          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
-         GETCHARINC(c, eptr);
+         GETCHARINCTEST(c, eptr);
          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
          }
-Line 1753 
 for (;;)
+Line 2146 
 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);
-           GETCHARINC(c, eptr);
+           GETCHARINCTEST(c, eptr);
            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
-Line 1771 
 for (;;)
+Line 2164 
 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 1836 
 for (;;)
+Line 2229 
 for (;;)
        else
          {
-         int dc;
+         unsigned int dc;
          GETCHARINC(dc, eptr);
          ecode += length;
          /* If we have Unicode property support, we can use it to test the other
-         case of the character, if there is one. The result of ucp_findchar() is
+         case of the character, if there is one. */
-         < 0 if the char isn't found, and othercase is returned as zero if there
-         isn't one. */
          if (fc != dc)
            {
  #ifdef SUPPORT_UCP
-           int chartype;
+           if (dc != UCD_OTHERCASE(fc))
-           int othercase;
-           if (ucp_findchar(fc, &chartype, &othercase) < 0 || dc != othercase)
  #endif
              RRETURN(MATCH_NOMATCH);
            }
-Line 1867 
 for (;;)
+Line 2256 
 for (;;)
        }
      break;
-     /* Match a single character repeatedly; different opcodes share code. */
+     /* Match a single character repeatedly. */
      case OP_EXACT:
      min = max = GET2(ecode, 1);
      ecode += 3;
      goto REPEATCHAR;
+     case OP_POSUPTO:
+     possessive = TRUE;
+     /* Fall through */
      case OP_UPTO:
      case OP_MINUPTO:
      min = 0;
-Line 1882 
 for (;;)
+Line 2275 
 for (;;)
      ecode += 3;
      goto REPEATCHAR;
+     case OP_POSSTAR:
+     possessive = TRUE;
+     min = 0;
+     max = INT_MAX;
+     ecode++;
+     goto REPEATCHAR;
+     case OP_POSPLUS:
+     possessive = TRUE;
+     min = 1;
+     max = INT_MAX;
+     ecode++;
+     goto REPEATCHAR;
+     case OP_POSQUERY:
+     possessive = TRUE;
+     min = 0;
+     max = 1;
+     ecode++;
+     goto REPEATCHAR;
      case OP_STAR:
      case OP_MINSTAR:
      case OP_PLUS:
-Line 1913 
 for (;;)
+Line 2327 
 for (;;)
        if (length > 1)
          {
-         int oclength = 0;
-         uschar occhars[8];
  #ifdef SUPPORT_UCP
-         int othercase;
+         unsigned int othercase;
-         int chartype;
          if ((ims & PCRE_CASELESS) != 0 &&
-              ucp_findchar(fc, &chartype, &othercase) >= 0 &&
+             (othercase = UCD_OTHERCASE(fc)) != fc)
-              othercase > 0)
            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;
+ #ifdef SUPPORT_UCP
            /* Need braces because of following else */
            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }
            else
-Line 1935 
 for (;;)
+Line 2346 
 for (;;)
              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);
              eptr += oclength;
              }
+ #else   /* without SUPPORT_UCP */
+           else { RRETURN(MATCH_NOMATCH); }
+ #endif  /* SUPPORT_UCP */
            }
          if (min == max) continue;
-Line 1943 
 for (;;)
+Line 2357 
 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 (memcmp(eptr, charptr, length) == 0) eptr += length;
+ #ifdef SUPPORT_UCP
              /* Need braces because of following else */
              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }
              else
-Line 1954 
 for (;;)
+Line 2369 
 for (;;)
                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);
                eptr += oclength;
                }
+ #else   /* without SUPPORT_UCP */
+             else { RRETURN (MATCH_NOMATCH); }
+ #endif  /* SUPPORT_UCP */
              }
            /* Control never gets here */
            }
-         else
+         else  /* Maximize */
            {
            pp = eptr;
            for (i = min; i < max; i++)
              {
              if (eptr > md->end_subject - length) break;
              if (memcmp(eptr, charptr, length) == 0) eptr += length;
+ #ifdef SUPPORT_UCP
              else if (oclength == 0) break;
              else
                {
                if (memcmp(eptr, occhars, oclength) != 0) break;
                eptr += oclength;
                }
+ #else   /* without SUPPORT_UCP */
+             else break;
+ #endif  /* SUPPORT_UCP */
              }
-           while (eptr >= pp)
+           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 (eptr == pp) RRETURN(MATCH_NOMATCH);
+ #ifdef SUPPORT_UCP
+            eptr--;
+            BACKCHAR(eptr);
+ #else   /* without SUPPORT_UCP */
             eptr -= length;
+ #endif  /* SUPPORT_UCP */
             }
-           RRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
          }
-Line 2017 
 for (;;)
+Line 2447 
 for (;;)
          {
          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 ||
                fc != md->lcc[*eptr++])
-Line 2025 
 for (;;)
+Line 2455 
 for (;;)
            }
          /* Control never gets here */
          }
-       else
+       else  /* Maximize */
          {
          pp = eptr;
          for (i = min; i < max; i++)
-Line 2033 
 for (;;)
+Line 2463 
 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 2054 
 for (;;)
+Line 2485 
 for (;;)
          {
          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++)
              RRETURN(MATCH_NOMATCH);
            }
          /* Control never gets here */
          }
-       else
+       else  /* Maximize */
          {
          pp = eptr;
          for (i = min; i < max; i++)
-Line 2069 
 for (;;)
+Line 2500 
 for (;;)
            if (eptr >= md->end_subject || fc != *eptr) break;
            eptr++;
            }
+         if (possessive) continue;
          while (eptr >= pp)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);
            eptr--;
            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
            }
-Line 2121 
 for (;;)
+Line 2553 
 for (;;)
      ecode += 3;
      goto REPEATNOTCHAR;
+     case OP_NOTPOSSTAR:
+     possessive = TRUE;
+     min = 0;
+     max = INT_MAX;
+     ecode++;
+     goto REPEATNOTCHAR;
+     case OP_NOTPOSPLUS:
+     possessive = TRUE;
+     min = 1;
+     max = INT_MAX;
+     ecode++;
+     goto REPEATNOTCHAR;
+     case OP_NOTPOSQUERY:
+     possessive = TRUE;
+     min = 0;
+     max = 1;
+     ecode++;
+     goto REPEATNOTCHAR;
+     case OP_NOTPOSUPTO:
+     possessive = TRUE;
+     min = 0;
+     max = GET2(ecode, 1);
+     ecode += 3;
+     goto REPEATNOTCHAR;
      case OP_NOTSTAR:
      case OP_NOTMINSTAR:
      case OP_NOTPLUS:
-Line 2160 
 for (;;)
+Line 2620 
 for (;;)
        /* UTF-8 mode */
        if (utf8)
          {
-         register int d;
+         register unsigned int d;
          for (i = 1; i <= min; i++)
            {
            GETCHARINC(d, eptr);
-Line 2185 
 for (;;)
+Line 2645 
 for (;;)
          /* UTF-8 mode */
          if (utf8)
            {
-           register int d;
+           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 || eptr >= md->end_subject) 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 2202 
 for (;;)
+Line 2663 
 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++])
                RRETURN(MATCH_NOMATCH);
-Line 2221 
 for (;;)
+Line 2682 
 for (;;)
          /* UTF-8 mode */
          if (utf8)
            {
-           register int d;
+           register unsigned int d;
            for (i = min; i < max; i++)
              {
              int len = 1;
-Line 2231 
 for (;;)
+Line 2692 
 for (;;)
              if (fc == d) break;
              eptr += len;
              }
-           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 2248 
 for (;;)
+Line 2710 
 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, RM31);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              eptr--;
              }
-Line 2269 
 for (;;)
+Line 2732 
 for (;;)
        /* UTF-8 mode */
        if (utf8)
          {
-         register int d;
+         register unsigned int d;
          for (i = 1; i <= min; i++)
            {
            GETCHARINC(d, eptr);
-Line 2292 
 for (;;)
+Line 2755 
 for (;;)
          /* UTF-8 mode */
          if (utf8)
            {
-           register int d;
+           register unsigned int d;
            for (fi = min;; fi++)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM32);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
              GETCHARINC(d, eptr);
-             if (fi >= max || eptr >= md->end_subject || fc == d)
+             if (fc == d) RRETURN(MATCH_NOMATCH);
-               RRETURN(MATCH_NOMATCH);
              }
            }
          else
-Line 2308 
 for (;;)
+Line 2771 
 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++)
                RRETURN(MATCH_NOMATCH);
-Line 2327 
 for (;;)
+Line 2790 
 for (;;)
          /* UTF-8 mode */
          if (utf8)
            {
-           register int d;
+           register unsigned int d;
            for (i = min; i < max; i++)
              {
              int len = 1;
-Line 2336 
 for (;;)
+Line 2799 
 for (;;)
              if (fc == d) break;
              eptr += len;
              }
+           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 2353 
 for (;;)
+Line 2817 
 for (;;)
              if (eptr >= md->end_subject || fc == *eptr) break;
              eptr++;
              }
+           if (possessive) continue;
            while (eptr >= pp)
              {
-             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM35);
              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
              eptr--;
              }
-Line 2384 
 for (;;)
+Line 2849 
 for (;;)
      ecode += 3;
      goto REPEATTYPE;
+     case OP_TYPEPOSSTAR:
+     possessive = TRUE;
+     min = 0;
+     max = INT_MAX;
+     ecode++;
+     goto REPEATTYPE;
+     case OP_TYPEPOSPLUS:
+     possessive = TRUE;
+     min = 1;
+     max = INT_MAX;
+     ecode++;
+     goto REPEATTYPE;
+     case OP_TYPEPOSQUERY:
+     possessive = TRUE;
+     min = 0;
+     max = 1;
+     ecode++;
+     goto REPEATTYPE;
+     case OP_TYPEPOSUPTO:
+     possessive = TRUE;
+     min = 0;
+     max = GET2(ecode, 1);
+     ecode += 3;
+     goto REPEATTYPE;
      case OP_TYPESTAR:
      case OP_TYPEMINSTAR:
      case OP_TYPEPLUS:
-Line 2408 
 for (;;)
+Line 2901 
 for (;;)
        {
        prop_fail_result = ctype == OP_NOTPROP;
        prop_type = *ecode++;
-       if (prop_type >= 128)
+       prop_value = *ecode++;
-         {
-         prop_test_against = prop_type - 128;
-         prop_test_variable = &prop_category;
-         }
-       else
-         {
-         prop_test_against = prop_type;
-         prop_test_variable = &prop_chartype;
-         }
        }
      else prop_type = -1;
  #endif
-Line 2434 
 for (;;)
+Line 2918 
 for (;;)
      if (min > 0)
        {
  #ifdef SUPPORT_UCP
-       if (prop_type > 0)
+       if (prop_type >= 0)
          {
-         for (i = 1; i <= min; i++)
+         switch(prop_type)
            {
-           GETCHARINC(c, eptr);
+           case PT_ANY:
-           prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+           if (prop_fail_result) RRETURN(MATCH_NOMATCH);
-           if ((*prop_test_variable == prop_test_against) == prop_fail_result)
+           for (i = 1; i <= min; i++)
-             RRETURN(MATCH_NOMATCH);
+             {
+             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             GETCHARINCTEST(c, eptr);
+             }
+           break;
+           case PT_LAMP:
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject) 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)
+               RRETURN(MATCH_NOMATCH);
+             }
+           break;
+           case PT_GC:
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             GETCHARINCTEST(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == prop_value) == prop_fail_result)
+               RRETURN(MATCH_NOMATCH);
+             }
+           break;
+           case PT_PC:
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             GETCHARINCTEST(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
+             if ((prop_chartype == prop_value) == prop_fail_result)
+               RRETURN(MATCH_NOMATCH);
+             }
+           break;
+           case PT_SC:
+           for (i = 1; i <= min; i++)
+             {
+             if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             GETCHARINCTEST(c, eptr);
+             prop_script = UCD_SCRIPT(c);
+             if ((prop_script == prop_value) == prop_fail_result)
+               RRETURN(MATCH_NOMATCH);
+             }
+           break;
+           default:
+           RRETURN(PCRE_ERROR_INTERNAL);
            }
          }
-Line 2453 
 for (;;)
+Line 2990 
 for (;;)
          for (i = 1; i <= min; i++)
            {
            GETCHARINCTEST(c, eptr);
-           prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+           prop_category = UCD_CATEGORY(c);
            if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH);
            while (eptr < md->end_subject)
              {
-Line 2462 
 for (;;)
+Line 2999 
 for (;;)
                {
                GETCHARLEN(c, eptr, len);
                }
-             prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+             prop_category = UCD_CATEGORY(c);
              if (prop_category != ucp_M) break;
              eptr += len;
              }
-Line 2480 
 for (;;)
+Line 3017 
 for (;;)
          case OP_ANY:
          for (i = 1; i <= min; i++)
            {
-           if (eptr >= md->end_subject ||
+           if (eptr >= md->end_subject || IS_NEWLINE(eptr))
-              (*eptr++ == NEWLINE && (ims & PCRE_DOTALL) == 0))
              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) RRETURN(MATCH_NOMATCH);
+           eptr++;
            while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
            }
          break;
-Line 2491 
 for (;;)
+Line 3037 
 for (;;)
          eptr += min;
          break;
+         case OP_ANYNL:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           GETCHARINC(c, eptr);
+           switch(c)
+             {
+             default: RRETURN(MATCH_NOMATCH);
+             case 0x000d:
+             if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
+             break;
+             case 0x000a:
+             break;
+             case 0x000b:
+             case 0x000c:
+             case 0x0085:
+             case 0x2028:
+             case 0x2029:
+             if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
+             break;
+             }
+           }
+         break;
+         case OP_NOT_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject) 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) 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) 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) 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;
+             }
+           }
+         break;
          case OP_NOT_DIGIT:
          for (i = 1; i <= min; i++)
            {
-Line 2515 
 for (;;)
+Line 3191 
 for (;;)
          for (i = 1; i <= min; i++)
            {
            if (eptr >= md->end_subject ||
-              (*eptr < 128 && (md->ctypes[*eptr++] & ctype_space) != 0))
+              (*eptr < 128 && (md->ctypes[*eptr] & ctype_space) != 0))
              RRETURN(MATCH_NOMATCH);
-           while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
+           while (++eptr < md->end_subject && (*eptr & 0xc0) == 0x80);
            }
          break;
-Line 2535 
 for (;;)
+Line 3211 
 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;
-Line 2559 
 for (;;)
+Line 3235 
 for (;;)
  #endif     /* SUPPORT_UTF8 */
        /* Code for the non-UTF-8 case for minimum matching of operators other
-       than OP_PROP and OP_NOTPROP. */
+       than OP_PROP and OP_NOTPROP. We can assume that there are the minimum
+       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 (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
-             if (*eptr++ == NEWLINE) RRETURN(MATCH_NOMATCH);
+           eptr++;
            }
-         else eptr += min;
+         break;
+         case OP_ALLANY:
+         eptr += min;
          break;
          case OP_ANYBYTE:
          eptr += min;
          break;
+         /* Because of the CRLF case, we can't assume the minimum number of
+         bytes are present in this case. */
+         case OP_ANYNL:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+           switch(*eptr++)
+             {
+             default: RRETURN(MATCH_NOMATCH);
+             case 0x000d:
+             if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
+             break;
+             case 0x000a:
+             break;
+             case 0x000b:
+             case 0x000c:
+             case 0x0085:
+             if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
+             break;
+             }
+           }
+         break;
+         case OP_NOT_HSPACE:
+         for (i = 1; i <= min; i++)
+           {
+           if (eptr >= md->end_subject) 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) 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) 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) 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;
+             }
+           }
+         break;
          case OP_NOT_DIGIT:
          for (i = 1; i <= min; i++)
            if ((md->ctypes[*eptr++] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
-Line 2624 
 for (;;)
+Line 3393 
 for (;;)
      if (minimize)
        {
  #ifdef SUPPORT_UCP
-       if (prop_type > 0)
+       if (prop_type >= 0)
          {
-         for (fi = min;; fi++)
+         switch(prop_type)
            {
-           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
+           case PT_ANY:
-           if (rrc != MATCH_NOMATCH) RRETURN(rrc);
+           for (fi = min;; fi++)
-           if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
+             {
-           GETCHARINC(c, eptr);
+             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM36);
-           prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+             if (rrc != MATCH_NOMATCH) RRETURN(rrc);
-           if ((*prop_test_variable == prop_test_against) == prop_fail_result)
+             if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
-             RRETURN(MATCH_NOMATCH);
+             GETCHARINC(c, eptr);
-           }
+             if (prop_fail_result) RRETURN(MATCH_NOMATCH);
-         }
+             }
+           /* Control never gets here */
-       /* Match extended Unicode sequences. We will get here only if the
+           case PT_LAMP:
-       support is in the binary; otherwise a compile-time error occurs. */
+           for (fi = min;; fi++)
+             {
+             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);
+             GETCHARINC(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
+             if ((prop_chartype == ucp_Lu ||
+                  prop_chartype == ucp_Ll ||
+                  prop_chartype == ucp_Lt) == prop_fail_result)
+               RRETURN(MATCH_NOMATCH);
+             }
+           /* Control never gets here */
+           case PT_GC:
+           for (fi = min;; fi++)
+             {
+             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);
+             GETCHARINC(c, eptr);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == prop_value) == prop_fail_result)
+               RRETURN(MATCH_NOMATCH);
+             }
+           /* Control never gets here */
+           case PT_PC:
+           for (fi = min;; fi++)
+             {
+             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);
+             GETCHARINC(c, eptr);
+             prop_chartype = UCD_CHARTYPE(c);
+             if ((prop_chartype == prop_value) == prop_fail_result)
+               RRETURN(MATCH_NOMATCH);
+             }
+           /* Control never gets here */
+           case PT_SC:
+           for (fi = min;; fi++)
+             {
+             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);
+             GETCHARINC(c, eptr);
+             prop_script = UCD_SCRIPT(c);
+             if ((prop_script == prop_value) == prop_fail_result)
+               RRETURN(MATCH_NOMATCH);
+             }
+           /* Control never gets here */
+           default:
+           RRETURN(PCRE_ERROR_INTERNAL);
+           }
+         }
+       /* Match extended Unicode sequences. We will get here only if the
+       support is in the binary; otherwise a compile-time error occurs. */
        else if (ctype == OP_EXTUNI)
          {
          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);
            GETCHARINCTEST(c, eptr);
-           prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+           prop_category = UCD_CATEGORY(c);
            if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH);
            while (eptr < md->end_subject)
              {
-Line 2658 
 for (;;)
+Line 3487 
 for (;;)
                {
                GETCHARLEN(c, eptr, len);
                }
-             prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+             prop_category = UCD_CATEGORY(c);
              if (prop_category != ucp_M) break;
              eptr += len;
              }
-Line 2674 
 for (;;)
+Line 3503 
 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) RRETURN(MATCH_NOMATCH);
+           if (fi >= max || eptr >= md->end_subject ||
+                (ctype == OP_ANY && IS_NEWLINE(eptr)))
+             RRETURN(MATCH_NOMATCH);
            GETCHARINC(c, eptr);
            switch(ctype)
              {
-             case OP_ANY:
+             case OP_ANY:        /* This is the non-NL case */
-             if ((ims & PCRE_DOTALL) == 0 && c == NEWLINE) RRETURN(MATCH_NOMATCH);
+             case OP_ALLANY:
+             case OP_ANYBYTE:
              break;
-             case OP_ANYBYTE:
+             case OP_ANYNL:
+             switch(c)
+               {
+               default: RRETURN(MATCH_NOMATCH);
+               case 0x000d:
+               if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
+               break;
+               case 0x000a:
+               break;
+               case 0x000b:
+               case 0x000c:
+               case 0x0085:
+               case 0x2028:
+               case 0x2029:
+               if (md->bsr_anycrlf) 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;
              case OP_NOT_DIGIT:
-Line 2729 
 for (;;)
+Line 3662 
 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) RRETURN(MATCH_NOMATCH);
+           if (fi >= max || eptr >= md->end_subject ||
+                (ctype == OP_ANY && IS_NEWLINE(eptr)))
+             RRETURN(MATCH_NOMATCH);
            c = *eptr++;
            switch(ctype)
              {
-             case OP_ANY:
+             case OP_ANY:     /* This is the non-NL case */
-             if ((ims & PCRE_DOTALL) == 0 && c == NEWLINE) RRETURN(MATCH_NOMATCH);
+             case OP_ALLANY:
+             case OP_ANYBYTE:
              break;
-             case OP_ANYBYTE:
+             case OP_ANYNL:
+             switch(c)
+               {
+               default: RRETURN(MATCH_NOMATCH);
+               case 0x000d:
+               if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
+               break;
+               case 0x000a:
+               break;
+               case 0x000b:
+               case 0x000c:
+               case 0x0085:
+               if (md->bsr_anycrlf) 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;
              case OP_NOT_DIGIT:
-Line 2774 
 for (;;)
+Line 3775 
 for (;;)
        /* Control never gets here */
        }
-     /* If maximizing it is worth using inline code for speed, doing the type
+     /* If maximizing, it is worth using inline code for speed, doing the type
      test once at the start (i.e. keep it out of the loop). Again, keep the
      UTF-8 and UCP stuff separate. */
-Line 2783 
 for (;;)
+Line 3784 
 for (;;)
        pp = eptr;  /* Remember where we started */
  #ifdef SUPPORT_UCP
-       if (prop_type > 0)
+       if (prop_type >= 0)
          {
-         for (i = min; i < max; i++)
+         switch(prop_type)
            {
-           int len = 1;
+           case PT_ANY:
-           if (eptr >= md->end_subject) break;
+           for (i = min; i < max; i++)
-           GETCHARLEN(c, eptr, len);
+             {
-           prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+             int len = 1;
-           if ((*prop_test_variable == prop_test_against) == prop_fail_result)
+             if (eptr >= md->end_subject) break;
-             break;
+             GETCHARLEN(c, eptr, len);
-           eptr+= len;
+             if (prop_fail_result) break;
+             eptr+= len;
+             }
+           break;
+           case PT_LAMP:
+           for (i = min; i < max; i++)
+             {
+             int len = 1;
+             if (eptr >= md->end_subject) break;
+             GETCHARLEN(c, eptr, len);
+             prop_chartype = UCD_CHARTYPE(c);
+             if ((prop_chartype == ucp_Lu ||
+                  prop_chartype == ucp_Ll ||
+                  prop_chartype == ucp_Lt) == prop_fail_result)
+               break;
+             eptr+= len;
+             }
+           break;
+           case PT_GC:
+           for (i = min; i < max; i++)
+             {
+             int len = 1;
+             if (eptr >= md->end_subject) break;
+             GETCHARLEN(c, eptr, len);
+             prop_category = UCD_CATEGORY(c);
+             if ((prop_category == prop_value) == prop_fail_result)
+               break;
+             eptr+= len;
+             }
+           break;
+           case PT_PC:
+           for (i = min; i < max; i++)
+             {
+             int len = 1;
+             if (eptr >= md->end_subject) break;
+             GETCHARLEN(c, eptr, len);
+             prop_chartype = UCD_CHARTYPE(c);
+             if ((prop_chartype == prop_value) == prop_fail_result)
+               break;
+             eptr+= len;
+             }
+           break;
+           case PT_SC:
+           for (i = min; i < max; i++)
+             {
+             int len = 1;
+             if (eptr >= md->end_subject) break;
+             GETCHARLEN(c, eptr, len);
+             prop_script = UCD_SCRIPT(c);
+             if ((prop_script == prop_value) == prop_fail_result)
+               break;
+             eptr+= len;
+             }
+           break;
            }
          /* eptr is now past the end of the maximum run */
+         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 2816 
 for (;;)
+Line 3875 
 for (;;)
            {
            if (eptr >= md->end_subject) break;
            GETCHARINCTEST(c, eptr);
-           prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+           prop_category = UCD_CATEGORY(c);
            if (prop_category == ucp_M) break;
            while (eptr < md->end_subject)
              {
-Line 2825 
 for (;;)
+Line 3884 
 for (;;)
                {
                GETCHARLEN(c, eptr, len);
                }
-             prop_category = ucp_findchar(c, &prop_chartype, &prop_othercase);
+             prop_category = UCD_CATEGORY(c);
              if (prop_category != ucp_M) break;
              eptr += len;
              }
-Line 2833 
 for (;;)
+Line 3892 
 for (;;)
          /* eptr is now past the end of the maximum run */
+         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 = ucp_findchar(c, &prop_chartype, &prop_othercase);
+             prop_category = UCD_CATEGORY(c);
              if (prop_category != ucp_M) break;
              eptr--;
              }
-Line 2864 
 for (;;)
+Line 3924 
 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 || *eptr == NEWLINE) 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++;
-                 eptr++;
+               while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-                 while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
-                 }
                }
              }
-Line 2894 
 for (;;)
+Line 3938 
 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 || *eptr == NEWLINE) 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 > md->end_subject - eptr) c = md->end_subject - eptr;
+               eptr++;
-               eptr += c;
+               while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
                }
              }
+           else eptr = md->end_subject;   /* Unlimited UTF-8 repeat */
            break;
            /* The byte case is the same as non-UTF8 */
            case OP_ANYBYTE:
            c = max - min;
-           if (c > md->end_subject - eptr) c = md->end_subject - eptr;
+           if (c > (unsigned int)(md->end_subject - eptr))
+             c = md->end_subject - eptr;
            eptr += c;
            break;
+           case OP_ANYNL:
+           for (i = min; i < max; i++)
+             {
+             int len = 1;
+             if (eptr >= md->end_subject) break;
+             GETCHARLEN(c, eptr, len);
+             if (c == 0x000d)
+               {
+               if (++eptr >= md->end_subject) break;
+               if (*eptr == 0x000a) eptr++;
+               }
+             else
+               {
+               if (c != 0x000a &&
+                   (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 2992 
 for (;;)
+Line 4128 
 for (;;)
          /* eptr is now past the end of the maximum run */
+         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 || *eptr == NEWLINE) break;
-               eptr++;
-               }
-             break;
              }
-           /* For DOTALL case, fall through and treat as \C */
+           break;
+           case OP_ALLANY:
            case OP_ANYBYTE:
            c = max - min;
-           if (c > md->end_subject - eptr) c = md->end_subject - eptr;
+           if (c > (unsigned int)(md->end_subject - eptr))
+             c = md->end_subject - eptr;
            eptr += c;
            break;
+           case OP_ANYNL:
+           for (i = min; i < max; i++)
+             {
+             if (eptr >= md->end_subject) break;
+             c = *eptr;
+             if (c == 0x000d)
+               {
+               if (++eptr >= md->end_subject) break;
+               if (*eptr == 0x000a) eptr++;
+               }
+             else
+               {
+               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 3085 
 for (;;)
+Line 4283 
 for (;;)
          /* eptr is now past the end of the maximum run */
+         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 3099 
 for (;;)
+Line 4298 
 for (;;)
        }
      /* Control never gets here */
-     /* There's been some horrible disaster. Since all codes > OP_BRA are
+     /* There's been some horrible disaster. Arrival here can only mean there is
-     for capturing brackets, and there shouldn't be any gaps between 0 and
+     something seriously wrong in the code above or the OP_xxx definitions. */
-     OP_BRA, arrival here can only mean there is something seriously wrong
-     in the code above or the OP_xxx definitions. */
      default:
      DPRINTF(("Unknown opcode %d\n", *ecode));
-     RRETURN(PCRE_ERROR_UNKNOWN_NODE);
+     RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
      }
    /* Do not stick any code in here without much thought; it is assumed
-Line 3115 
 for (;;)
+Line 4312 
 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 3127 
 Undefine all the macros that were define
+Line 4353 
 Undefine all the macros that were define
  #ifdef NO_RECURSE
  #undef eptr
  #undef ecode
+ #undef mstart
  #undef offset_top
  #undef ims
  #undef eptrb
-Line 3144 
 Undefine all the macros that were define
+Line 4371 
 Undefine all the macros that were define
  #undef cur_is_word
  #undef condition
- #undef minimize
  #undef prev_is_word
  #undef original_ims
-Line 3200 
 Returns:          > 0 => success; value
+Line 4426 
 Returns:          > 0 => success; value
                   < -1 => some kind of unexpected problem
  */
- EXPORT int
+ PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
  pcre_exec(const pcre *argument_re, const pcre_extra *extra_data,
-   const char *subject, int length, int start_offset, int options, int *offsets,
+   PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
    int offsetcount)
  {
  int rc, resetcount, ocount;
  int first_byte = -1;
  int req_byte = -1;
  int req_byte2 = -1;
- unsigned long int ims = 0;
+ int newline;
+ unsigned long int ims;
  BOOL using_temporary_offsets = FALSE;
  BOOL anchored;
  BOOL startline;
  BOOL firstline;
  BOOL first_byte_caseless = FALSE;
  BOOL req_byte_caseless = FALSE;
+ BOOL utf8;
  match_data match_block;
+ match_data *md = &match_block;
  const uschar *tables;
  const uschar *start_bits = NULL;
- const uschar *start_match = (const uschar *)subject + start_offset;
+ USPTR start_match = (USPTR)subject + start_offset;
- const uschar *end_subject;
+ USPTR end_subject;
- const uschar *req_byte_ptr = start_match - 1;
+ USPTR req_byte_ptr = start_match - 1;
  pcre_study_data internal_study;
  const pcre_study_data *study;
-Line 3241 
 if (offsetcount < 0) return PCRE_ERROR_B
+Line 4470 
 if (offsetcount < 0) return PCRE_ERROR_B
  the default values. */
  study = NULL;
- match_block.match_limit = MATCH_LIMIT;
+ md->match_limit = MATCH_LIMIT;
- match_block.callout_data = NULL;
+ md->match_limit_recursion = MATCH_LIMIT_RECURSION;
+ md->callout_data = NULL;
  /* The table pointer is always in native byte order. */
-Line 3254 
 if (extra_data != NULL)
+Line 4484 
 if (extra_data != NULL)
    if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
      study = (const pcre_study_data *)extra_data->study_data;
    if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
-     match_block.match_limit = extra_data->match_limit;
+     md->match_limit = extra_data->match_limit;
+   if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
+     md->match_limit_recursion = extra_data->match_limit_recursion;
    if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
-     match_block.callout_data = extra_data->callout_data;
+     md->callout_data = extra_data->callout_data;
    if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
    }
-Line 3281 
 if (re->magic_number != MAGIC_NUMBER)
+Line 4513 
 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. */
- match_block.start_code = (const uschar *)external_re + re->name_table_offset +
+ md->start_code = (const uschar *)external_re + re->name_table_offset +
    re->name_count * re->name_entry_size;
- match_block.start_subject = (const uschar *)subject;
+ md->start_subject = (USPTR)subject;
- match_block.start_offset = start_offset;
+ md->start_offset = start_offset;
- match_block.end_subject = match_block.start_subject + length;
+ md->end_subject = md->start_subject + length;
- end_subject = match_block.end_subject;
+ end_subject = md->end_subject;
- match_block.endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
+ md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
- match_block.utf8 = (re->options & PCRE_UTF8) != 0;
+ utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
+ md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
- match_block.notbol = (options & PCRE_NOTBOL) != 0;
- match_block.noteol = (options & PCRE_NOTEOL) != 0;
+ md->notbol = (options & PCRE_NOTBOL) != 0;
- match_block.notempty = (options & PCRE_NOTEMPTY) != 0;
+ md->noteol = (options & PCRE_NOTEOL) != 0;
- match_block.partial = (options & PCRE_PARTIAL) != 0;
+ md->notempty = (options & PCRE_NOTEMPTY) != 0;
- match_block.hitend = FALSE;
+ md->partial = (options & PCRE_PARTIAL) != 0;
+ md->hitend = FALSE;
+ md->recursive = NULL;                   /* No recursion at top level */
+ 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;
- match_block.recursive = NULL;                   /* No recursion at top level */
+   case PCRE_BSR_ANYCRLF:
+   md->bsr_anycrlf = TRUE;
+   break;
- match_block.lcc = tables + lcc_offset;
+   case PCRE_BSR_UNICODE:
- match_block.ctypes = tables + ctypes_offset;
+   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 :
+         (pcre_uint32)options) & PCRE_NEWLINE_BITS)
+   {
+   case 0: newline = NEWLINE; break;   /* Compile-time default */
+   case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
+   case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
+   case PCRE_NEWLINE_CR+
+        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 == -2)
+   {
+   md->nltype = NLTYPE_ANYCRLF;
+   }
+ else if (newline < 0)
+   {
+   md->nltype = NLTYPE_ANY;
+   }
+ else
+   {
+   md->nltype = NLTYPE_FIXED;
+   if (newline > 255)
+     {
+     md->nllen = 2;
+     md->nl[0] = (newline >> 8) & 255;
+     md->nl[1] = newline & 255;
+     }
+   else
+     {
+     md->nllen = 1;
+     md->nl[0] = newline;
+     }
+   }
  /* Partial matching is supported only for a restricted set of regexes at the
  moment. */
- if (match_block.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
  back the character offset. */
  #ifdef SUPPORT_UTF8
- if (match_block.utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
+ if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
    {
    if (_pcre_valid_utf8((uschar *)subject, length) >= 0)
      return PCRE_ERROR_BADUTF8;
-Line 3350 
 ocount = offsetcount - (offsetcount % 3)
+Line 4649 
 ocount = offsetcount - (offsetcount % 3)
  if (re->top_backref > 0 && re->top_backref >= ocount/3)
    {
    ocount = re->top_backref * 3 + 3;
-   match_block.offset_vector = (int *)(pcre_malloc)(ocount * sizeof(int));
+   md->offset_vector = (int *)(pcre_malloc)(ocount * sizeof(int));
-   if (match_block.offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
+   if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
    using_temporary_offsets = TRUE;
    DPRINTF(("Got memory to hold back references\n"));
    }
- else match_block.offset_vector = offsets;
+ else md->offset_vector = offsets;
- match_block.offset_end = ocount;
+ md->offset_end = ocount;
- match_block.offset_max = (2*ocount)/3;
+ md->offset_max = (2*ocount)/3;
- match_block.offset_overflow = FALSE;
+ md->offset_overflow = FALSE;
- match_block.capture_last = -1;
+ md->capture_last = -1;
  /* Compute the minimum number of offsets that we need to reset each time. Doing
  this makes a huge difference to execution time when there aren't many brackets
-Line 3373 
 if (resetcount > offsetcount) resetcount
+Line 4672 
 if (resetcount > offsetcount) resetcount
  never be used unless previously set, but they get saved and restored, and so we
  initialize them to avoid reading uninitialized locations. */
- if (match_block.offset_vector != NULL)
+ if (md->offset_vector != NULL)
    {
-   register int *iptr = match_block.offset_vector + ocount;
+   register int *iptr = md->offset_vector + ocount;
    register int *iend = iptr - resetcount/2 + 1;
    while (--iptr >= iend) *iptr = -1;
    }
-Line 3388 
 studied, there may be a bitmap of possib
+Line 4687 
 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)
-       first_byte = match_block.lcc[first_byte];
+       first_byte = md->lcc[first_byte];
      }
    else
      if (!startline && study != NULL &&
-Line 3403 
 if (!anchored)
+Line 4702 
 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;
    req_byte2 = (tables + fcc_offset)[req_byte];  /* case flipped */
    }
+ /* ==========================================================================*/
  /* Loop for handling unanchored repeated matching attempts; for anchored regexs
  the loop runs just once. */
- do
+ for(;;)
    {
-   const uschar *save_end_subject = end_subject;
+   USPTR save_end_subject = end_subject;
+   USPTR new_start_match;
    /* Reset the maximum number of extractions we might see. */
-   if (match_block.offset_vector != NULL)
+   if (md->offset_vector != NULL)
      {
-     register int *iptr = match_block.offset_vector;
+     register int *iptr = md->offset_vector;
      register int *iend = iptr + resetcount;
      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
-   Implement this by temporarily adjusting end_subject so that we stop scanning
+   newline. Implement this by temporarily adjusting end_subject so that we stop
-   at a newline. If the match fails at the newline, later code breaks this loop.
+   scanning at a newline. If the match fails at the newline, later code breaks
-   */
+   this loop. */
    if (firstline)
      {
-     const uschar *t = start_match;
+     USPTR t = start_match;
-     while (t < save_end_subject && *t != '\n') t++;
+ #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 &&
-              match_block.lcc[*start_match] != first_byte)
-         start_match++;
-     else
-       while (start_match < end_subject && *start_match != first_byte)
-         start_match++;
-     }
-   /* Or to just after \n for a multiline match if possible */
+     if (first_byte >= 0)
+       {
+       if (first_byte_caseless)
+         while (start_match < end_subject && md->lcc[*start_match] != first_byte)
+           start_match++;
+       else
+         while (start_match < end_subject && *start_match != first_byte)
+           start_match++;
+       }
-   else if (startline)
+     /* Or to just after a linebreak for a multiline match */
-     {
-     if (start_match > match_block.start_subject + start_offset)
+     else if (startline)
        {
-       while (start_match < end_subject && start_match[-1] != NEWLINE)
+       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 3480 
 do
+Line 4824 
 do
  #ifdef DEBUG  /* Sigh. Some compilers never learn. */
    printf(">>>> Match against: ");
-   pchars(start_match, end_subject - start_match, TRUE, &match_block);
+   pchars(start_match, end_subject - start_match, TRUE, md);
    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 /^C/ on a 32-megabyte string... so we
+   This showed up when somebody was matching something like /^\d+C/ on a
-   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 &&
-       !match_block.partial)
+       !md->partial)
      {
-     register const uschar *p = start_match + ((first_byte >= 0)? 1 : 0);
+     register USPTR p = start_match + ((first_byte >= 0)? 1 : 0);
      /* We don't need to repeat the search if we haven't yet reached the
      place we found it at last time. */
-Line 3527 
 do
+Line 4873 
 do
            }
          }
-       /* If we can't find the required character, break the matching loop */
+       /* If we can't find the required character, break the matching loop,
+       forcing a match failure. */
-       if (p >= end_subject) break;
+       if (p >= end_subject)
+         {
+         rc = MATCH_NOMATCH;
+         break;
+         }
        /* If we have found the required character, save the point where we
        found it, so that we don't search again next time round the loop if
-Line 3539 
 do
+Line 4890 
 do
        }
      }
-   /* When a match occurs, substrings will be set for all internal extractions;
+   /* OK, we can now run the match. */
-   we just need to set up the whole thing as substring 0 before returning. If
-   there were too many extractions, set the return code to zero. In the case
+   md->start_match_ptr = start_match;
-   where we had to get some local store to hold offsets for backreferences, copy
+   md->match_call_count = 0;
-   those back references that we can. In this case there need not be overflow
+   rc = match(start_match, md->start_code, start_match, 2, md, ims, NULL, 0, 0);
-   if certain parts of the pattern were not used. */
-   match_block.start_match = start_match;
-   match_block.match_call_count = 0;
-   rc = match(start_match, match_block.start_code, 2, &match_block, ims, NULL,
-     match_isgroup);
-   /* When the result is no match, if the subject's first character was a
-   newline and the PCRE_FIRSTLINE option is set, break (which will return
-   PCRE_ERROR_NOMATCH). The option requests that a match occur before the first
-   newline in the subject. Otherwise, advance the pointer to the next character
-   and continue - but the continuation will actually happen only when the
-   pattern is not anchored. */
-   if (rc == MATCH_NOMATCH)
+   switch(rc)