/tags/harbour-2.0.0/external/pcre/pcreexec.c

/*************************************************
*      Perl-Compatible Regular Expressions       *
*************************************************/

/* PCRE is a library of functions to support regular expressions whose syntax
and semantics are as close as possible to those of the Perl 5 language.

                       Written by Philip Hazel
           Copyright (c) 1997-2009 University of Cambridge

-----------------------------------------------------------------------------
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

    * Redistributions of source code must retain the above copyright notice,
      this list of conditions and the following disclaimer.

    * Redistributions in binary form must reproduce the above copyright
      notice, this list of conditions and the following disclaimer in the
      documentation and/or other materials provided with the distribution.

    * Neither the name of the University of Cambridge nor the names of its
      contributors may be used to endorse or promote products derived from
      this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
-----------------------------------------------------------------------------
*/


/* This module contains pcre_exec(), the externally visible function that does
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 "pcreinal.h"

/* Undefine some potentially clashing cpp symbols */

#undef min
#undef max

/* Flag bits for the match() function */

#define match_condassert     0x01  /* Called to check a condition assertion */
#define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */

/* Non-error returns from the match() function. Error returns are externally
defined PCRE_ERROR_xxx codes, which are all negative. */

#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. */

#define REC_STACK_SAVE_MAX 30

/* Min and max values for the common repeats; for the maxima, 0 => infinity */

static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };



#ifdef DEBUG
/*************************************************
*        Debugging function to print chars       *
*************************************************/

/* Print a sequence of chars in printable format, stopping at the end of the
subject if the requested.

Arguments:
  p           points to characters
  length      number to print
  is_subject  TRUE if printing from within md->start_subject
  md          pointer to matching data block, if is_subject is TRUE

Returns:     nothing
*/

static void
pchars(const uschar *p, int length, BOOL is_subject, match_data *md)
{
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);
}
#endif



/*************************************************
*          Match a back-reference                *
*************************************************/

/* If a back reference hasn't been set, the length that is passed is greater
than the number of characters left in the string, so the match fails.

Arguments:
  offset      index into the offset vector
  eptr        points into the subject
  length      length to be matched
  md          points to match data block
  ims         the ims flags

Returns:      TRUE if matched
*/

static BOOL
match_ref(int offset, register USPTR eptr, int length, match_data *md,
  unsigned long int ims)
{
USPTR p = md->start_subject + md->offset_vector[offset];

#ifdef DEBUG
if (eptr >= md->end_subject)
  printf("matching subject <null>");
else
  {
  printf("matching subject ");
  pchars(eptr, length, TRUE, md);
  }
printf(" against backref ");
pchars(p, length, FALSE, md);
printf("\n");
#endif

/* Always fail if not enough characters left */

if (length > md->end_subject - eptr) return FALSE;

/* 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; }
  }

/* 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; }

return TRUE;
}



/***************************************************************************
****************************************************************************
                   RECURSION IN THE match() FUNCTION

The match() function is highly recursive, though not every recursive call
increases the recursive depth. Nevertheless, some regular expressions can cause
it to recurse to a great depth. I was writing for Unix, so I just let it call
itself recursively. This uses the stack for saving everything that has to be
saved for a recursive call. On Unix, the stack can be large, and this works
fine.

It turns out that on some non-Unix-like systems there are problems with
programs that use a lot of stack. (This despite the fact that every last chip
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
achieve this so that the actual code doesn't look very different to what it
always used to.

The original heap-recursive code used longjmp(). However, it seems that this
can be very slow on some operating systems. Following a suggestion from Stan
Switzer, the use of longjmp() has been abolished, at the cost of having to
provide a unique number for each call to RMATCH. There is no way of generating
a sequence of numbers at compile time in C. I have given them names, to make
them stand out more clearly.

Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
tests. Furthermore, not using longjmp() means that local dynamic variables
don't have indeterminate values; this has meant that the frame size can be
reduced because the result can be "passed back" by straight setting of the
variable instead of being passed in the frame.
****************************************************************************
***************************************************************************/

/* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
below must be updated in sync.  */

enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
       RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
       RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
       RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
       RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
       RM51,  RM52, RM53, RM54 };

/* These versions of the macros use the stack, as normal. There are debugging
versions and production versions. Note that the "rw" argument of RMATCH isn't
actuall used in this definition. */

#ifndef NO_RECURSE
#define REGISTER register

#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 that
the "rd" argument of RMATCH isn't actually used in this definition. It's the md
argument of match(), which never changes. */

#define REGISTER

#define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\
  {\
  heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\
  frame->Xwhere = rw; \
  newframe->Xeptr = ra;\
  newframe->Xecode = rb;\
  newframe->Xmstart = mstart;\
  newframe->Xoffset_top = rc;\
  newframe->Xims = re;\
  newframe->Xeptrb = rf;\
  newframe->Xflags = rg;\
  newframe->Xrdepth = frame->Xrdepth + 1;\
  newframe->Xprevframe = frame;\
  frame = newframe;\
  DPRINTF(("restarting from line %d\n", __LINE__));\
  goto HEAP_RECURSE;\
  L_##rw:\
  DPRINTF(("jumped back to line %d\n", __LINE__));\
  }

#define RRETURN(ra)\
  {\
  heapframe *newframe = frame;\
  frame = newframe->Xprevframe;\
  (pcre_stack_free)(newframe);\
  if (frame != NULL)\
    {\
    rrc = ra;\
    goto HEAP_RETURN;\
    }\
  return ra;\
  }


/* Structure for remembering the local variables in a private frame */

typedef struct heapframe {
  struct heapframe *Xprevframe;

  /* Function arguments that may change */

  USPTR Xeptr;
  const uschar *Xecode;
  USPTR Xmstart;
  int Xoffset_top;
  long int Xims;
  eptrblock *Xeptrb;
  int Xflags;
  unsigned int Xrdepth;

  /* Function local variables */

  USPTR Xcallpat;
#ifdef SUPPORT_UTF8
  USPTR Xcharptr;
#endif
  USPTR Xdata;
  USPTR Xnext;
  USPTR Xpp;
  USPTR Xprev;
  USPTR Xsaved_eptr;

  recursion_info Xnew_recursive;

  BOOL Xcur_is_word;
  BOOL Xcondition;
  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_script;
  int Xoclength;
  uschar Xocchars[8];
#endif

  int Xcodelink;
  int Xctype;
  unsigned int Xfc;
  int Xfi;
  int Xlength;
  int Xmax;
  int Xmin;
  int Xnumber;
  int Xoffset;
  int Xop;
  int Xsave_capture_last;
  int Xsave_offset1, Xsave_offset2, Xsave_offset3;
  int Xstacksave[REC_STACK_SAVE_MAX];

  eptrblock Xnewptrb;

  /* Where to jump back to */

  int Xwhere;

} heapframe;

#endif


/***************************************************************************
***************************************************************************/



/*************************************************
*         Match from current position            *
*************************************************/

/* This function is called recursively in many circumstances. Whenever it
returns a negative (error) response, the outer incarnation must also return the
same response. */

/* These macros pack up tests that are used for partial matching, and which
appears several times in the code. We set the "hit end" flag if the pointer is
at the end of the subject and also past the start of the subject (i.e.
something has been matched). For hard partial matching, we then return
immediately. The second one is used when we already know we are past the end of
the subject. */

#define CHECK_PARTIAL()\
  if (md->partial != 0 && eptr >= md->end_subject && eptr > mstart)\
    {\
    md->hitend = TRUE;\
    if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);\
    }

#define SCHECK_PARTIAL()\
  if (md->partial != 0 && eptr > mstart)\
    {\
    md->hitend = TRUE;\
    if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);\
    }


/* Performance note: It might be tempting to extract commonly used fields from
the md structure (e.g. utf8, end_subject) into individual variables to improve
performance. Tests using gcc on a SPARC disproved this; in the first case, it
made performance worse.

Arguments:
   eptr        pointer to current character in subject
   ecode       pointer to current position in compiled code
   mstart      pointer to the current match start position (can be modified
                 by encountering \K)
   offset_top  current top pointer
   md          pointer to "static" info for the match
   ims         current /i, /m, and /s options
   eptrb       pointer to chain of blocks containing eptr at start of
                 brackets - for testing for empty matches
   flags       can contain
                 match_condassert - this is an assertion condition
                 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 repeated call or recursion limit)
*/

static int
match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
  int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,
  int flags, unsigned int rdepth)
{
/* These variables do not need to be preserved over recursion in this function,
so they can be ordinary variables in all cases. Mark some of them with
"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 */

BOOL minimize, possessive; /* Quantifier options */
int condcode;

/* When recursion is not being used, all "local" variables that have to be
preserved over calls to RMATCH() are part of a "frame" which is obtained from
heap storage. Set up the top-level frame here; others are obtained from the
heap whenever RMATCH() does a "recursion". See the macro definitions above. */

#ifdef NO_RECURSE
heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));
frame->Xprevframe = NULL;            /* Marks the top level */

/* Copy in the original argument variables */

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" */

HEAP_RECURSE:

/* Macros make the argument variables come from the current frame */

#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 */

#ifdef SUPPORT_UTF8
#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
#define prev               frame->Xprev
#define saved_eptr         frame->Xsaved_eptr

#define new_recursive      frame->Xnew_recursive

#define cur_is_word        frame->Xcur_is_word
#define condition          frame->Xcondition
#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_script        frame->Xprop_script
#define oclength           frame->Xoclength
#define occhars            frame->Xocchars
#endif

#define ctype              frame->Xctype
#define fc                 frame->Xfc
#define fi                 frame->Xfi
#define length             frame->Xlength
#define max                frame->Xmax
#define min                frame->Xmin
#define number             frame->Xnumber
#define offset             frame->Xoffset
#define op                 frame->Xop
#define save_capture_last  frame->Xsave_capture_last
#define save_offset1       frame->Xsave_offset1
#define save_offset2       frame->Xsave_offset2
#define save_offset3       frame->Xsave_offset3
#define stacksave          frame->Xstacksave

#define newptrb            frame->Xnewptrb

/* When recursion is being used, local variables are allocated on the stack and
get preserved during recursion in the normal way. In this environment, fi and
i, and fc and c, can be the same variables. */

#else         /* NO_RECURSE not defined */
#define fi i
#define fc c


#ifdef SUPPORT_UTF8                /* Many of these variables are used only  */
const uschar *charptr;             /* in small blocks of the code. My normal */
#endif                             /* style of coding would have declared    */
const uschar *callpat;             /* them within each of those blocks.      */
const uschar *data;                /* However, in order to accommodate the   */
const uschar *next;                /* version of this code that uses an      */
USPTR         pp;                  /* external "stack" implemented on the    */
const uschar *prev;                /* heap, it is easier to declare them all */
USPTR         saved_eptr;          /* here, so the declarations can be cut   */
                                   /* out in a block. The only declarations  */
recursion_info new_recursive;      /* within blocks below are for variables  */
                                   /* that do not have to be preserved over  */
BOOL cur_is_word;                  /* a recursive call to RMATCH().          */
BOOL condition;
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_script;
int oclength;
uschar occhars[8];
#endif

int codelink;
int ctype;
int length;
int max;
int min;
int number;
int offset;
int op;
int save_capture_last;
int save_offset1, save_offset2, save_offset3;
int stacksave[REC_STACK_SAVE_MAX];

eptrblock newptrb;
#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;
#endif


/* This label is used for tail recursion, which is used in a few cases even
when NO_RECURSE is not defined, in order to reduce the amount of stack that is
used. Thanks to Ian Taylor for noticing this possibility and sending the
original patch. */

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 ')' */

/* At the start of a group with an unlimited repeat that may match an empty
string, the match_cbegroup flag is set. When this is the case, add the current
subject pointer to the chain of such remembered pointers, to be checked when we
hit the closing ket, in order to break infinite loops that match no characters.
When match() is called in other circumstances, don't add to the chain. 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_cbegroup) != 0)
  {
  newptrb.epb_saved_eptr = eptr;
  newptrb.epb_prev = eptrb;
  eptrb = &newptrb;
  }

/* Now start processing the opcodes. */

for (;;)
  {
  minimize = possessive = FALSE;
  op = *ecode;

  switch(op)
    {
    case OP_FAIL:
    RRETURN(MATCH_NOMATCH);

    case OP_PRUNE:
    RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
      ims, eptrb, flags, RM51);
    if (rrc != MATCH_NOMATCH) RRETURN(rrc);
    RRETURN(MATCH_PRUNE);

    case OP_COMMIT:
    RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
      ims, eptrb, flags, RM52);
    if (rrc != MATCH_NOMATCH) RRETURN(rrc);
    RRETURN(MATCH_COMMIT);

    case OP_SKIP:
    RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
      ims, eptrb, flags, RM53);
    if (rrc != MATCH_NOMATCH) RRETURN(rrc);
    md->start_match_ptr = eptr;   /* Pass back current position */
    RRETURN(MATCH_SKIP);

    case OP_THEN:
    RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
      ims, eptrb, flags, RM54);
    if (rrc != MATCH_NOMATCH) RRETURN(rrc);
    RRETURN(MATCH_THEN);

    /* Handle a capturing bracket. If there is space in the offset vector, save
    the current subject position in the working slot at the top of the vector.
    We mustn't change the current values of the data slot, because they may be
    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\n", number);
    printf("subject=");
    pchars(eptr, 16, TRUE, md);
    printf("\n");
#endif

    if (offset < md->offset_max)
      {
      save_offset1 = md->offset_vector[offset];
      save_offset2 = md->offset_vector[offset+1];
      save_offset3 = md->offset_vector[md->offset_end - number];
      save_capture_last = md->capture_last;

      DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
      md->offset_vector[md->offset_end - number] = eptr - md->start_subject;

      flags = (op == OP_SCBRA)? match_cbegroup : 0;
      do
        {
        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
          ims, eptrb, flags, RM1);
        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
        md->capture_last = save_capture_last;
        ecode += GET(ecode, 1);
        }
      while (*ecode == OP_ALT);

      DPRINTF(("bracket %d failed\n", number));

      md->offset_vector[offset] = save_offset1;
      md->offset_vector[offset+1] = save_offset2;
      md->offset_vector[md->offset_end - number] = save_offset3;

      RRETURN(MATCH_NOMATCH);
      }

    /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
    as a non-capturing bracket. */

    /* VVVVVVVVVVVVVVVVVVVVVVVVV */
    /* VVVVVVVVVVVVVVVVVVVVVVVVV */

    DPRINTF(("insufficient capture room: treat as non-capturing\n"));

    /* VVVVVVVVVVVVVVVVVVVVVVVVV */
    /* VVVVVVVVVVVVVVVVVVVVVVVVV */

    /* Non-capturing bracket. Loop for all the alternatives. When we get to the
    final alternative within the brackets, we would return the result of a
    recursive call to match() whatever happened. We can reduce stack usage by
    turning this into a tail recursion, 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 (;;)
      {
      if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */
        {
        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);
      }
    /* Control never reaches here. */

    /* 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. As there is only one branch to be
    obeyed, we can use tail recursion to avoid using another stack frame. */

    case OP_COND:
    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 || condcode == OP_NRREF)    /* Recursion test */
      {
      if (md->recursive == NULL)                /* Not recursing => FALSE */
        {
        condition = FALSE;
        ecode += GET(ecode, 1);
        }
      else
        {
        int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
        condition =  (recno == RREF_ANY || recno == md->recursive->group_num);

        /* If the test is for recursion into a specific subpattern, and it is
        false, but the test was set up by name, scan the table to see if the
        name refers to any other numbers, and test them. The condition is true
        if any one is set. */

        if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
          {
          uschar *slotA = md->name_table;
          for (i = 0; i < md->name_count; i++)
            {
            if (GET2(slotA, 0) == recno) break;
            slotA += md->name_entry_size;
            }

          /* Found a name for the number - there can be only one; duplicate
          names for different numbers are allowed, but not vice versa. First
          scan down for duplicates. */

          if (i < md->name_count)
            {
            uschar *slotB = slotA;
            while (slotB > md->name_table)
              {
              slotB -= md->name_entry_size;
              if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
                {
                condition = GET2(slotB, 0) == md->recursive->group_num;
                if (condition) break;
                }
              else break;
              }

            /* Scan up for duplicates */

            if (!condition)
              {
              slotB = slotA;
              for (i++; i < md->name_count; i++)
                {
                slotB += md->name_entry_size;
                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
                  {
                  condition = GET2(slotB, 0) == md->recursive->group_num;
                  if (condition) break;
                  }
                else break;
                }
              }
            }
          }

        /* Chose branch according to the condition */

        ecode += condition? 3 : GET(ecode, 1);
        }
      }

    else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
      {
      offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
      condition = offset < offset_top && md->offset_vector[offset] >= 0;

      /* If the numbered capture is unset, but the reference was by name,
      scan the table to see if the name refers to any other numbers, and test
      them. The condition is true if any one is set. This is tediously similar
      to the code above, but not close enough to try to amalgamate. */

      if (!condition && condcode == OP_NCREF)
        {
        int refno = offset >> 1;
        uschar *slotA = md->name_table;

        for (i = 0; i < md->name_count; i++)
          {
          if (GET2(slotA, 0) == refno) break;
          slotA += md->name_entry_size;
          }

        /* Found a name for the number - there can be only one; duplicate names
        for different numbers are allowed, but not vice versa. First scan down
        for duplicates. */

        if (i < md->name_count)
          {
          uschar *slotB = slotA;
          while (slotB > md->name_table)
            {
            slotB -= md->name_entry_size;
            if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
              {
              offset = GET2(slotB, 0) << 1;
              condition = offset < offset_top &&
                md->offset_vector[offset] >= 0;
              if (condition) break;
              }
            else break;
            }

          /* Scan up for duplicates */

          if (!condition)
            {
            slotB = slotA;
            for (i++; i < md->name_count; i++)
              {
              slotB += md->name_entry_size;
              if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
                {
                offset = GET2(slotB, 0) << 1;
                condition = offset < offset_top &&
                  md->offset_vector[offset] >= 0;
                if (condition) break;
                }
              else break;
              }
            }
          }
        }

      /* Chose branch according to the condition */

      ecode += condition? 3 : GET(ecode, 1);
      }

    else if (condcode == OP_DEF)     /* DEFINE - always false */
      {
      condition = FALSE;
      ecode += GET(ecode, 1);
      }

    /* The condition is an assertion. Call match() to evaluate it - setting
    the final argument match_condassert causes it to stop at the end of an
    assertion. */

    else
      {
      RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
          match_condassert, RM3);
      if (rrc == MATCH_MATCH)
        {
        condition = TRUE;
        ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
        while (*ecode == OP_ALT) ecode += GET(ecode, 1);
        }
      else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
        {
        RRETURN(rrc);         /* Need braces because of following else */
        }
      else
        {
        condition = FALSE;
        ecode += codelink;
        }
      }

    /* We are now at the branch that is to be obeyed. As there is only one,
    we can use tail recursion to avoid using another stack frame, 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. */

    if (condition || *ecode == OP_ALT)
      {
      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;


    /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
    to close any currently open capturing brackets. */

    case OP_CLOSE:
    number = GET2(ecode, 1);
    offset = number << 1;

#ifdef DEBUG
      printf("end bracket %d at *ACCEPT", number);
      printf("\n");
#endif

    md->capture_last = number;
    if (offset >= md->offset_max) md->offset_overflow = TRUE; else
      {
      md->offset_vector[offset] =
        md->offset_vector[md->offset_end - number];
      md->offset_vector[offset+1] = eptr - md->start_subject;
      if (offset_top <= offset) offset_top = offset + 2;
      }
    ecode += 3;
    break;


    /* End of the pattern, either real or forced. If we are in a top-level
    recursion, we should restore the offsets appropriately and continue from
    after the call. */

    case OP_ACCEPT:
    case OP_END:
    if (md->recursive != NULL && md->recursive->group_num == 0)
      {
      recursion_info *rec = md->recursive;
      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));
      offset_top = rec->save_offset_top;
      mstart = rec->save_start;
      ims = original_ims;
      ecode = rec->after_call;
      break;
      }

    /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is
    set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of
    the subject. In both cases, backtracking will then try other alternatives,
    if any. */

    if (eptr == mstart &&
        (md->notempty ||
          (md->notempty_atstart &&
            mstart == md->start_subject + md->start_offset)))
      RRETURN(MATCH_NOMATCH);

    /* Otherwise, we have a match. */

    md->end_match_ptr = eptr;           /* Record where we ended */
    md->end_offset_top = offset_top;    /* and how many extracts were taken */
    md->start_match_ptr = mstart;       /* and the start (\K can modify) */
    RRETURN(MATCH_MATCH);

    /* Change option settings */

    case OP_OPT:
    ims = ecode[1];
    ecode += 2;
    DPRINTF(("ims set to %02lx\n", ims));
    break;

    /* Assertion brackets. Check the alternative branches in turn - the
    matching won't pass the KET for an assertion. If any one branch matches,
    the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
    start of each branch to move the current point backwards, so the code at
    this level is identical to the lookahead case. */

    case OP_ASSERT:
    case OP_ASSERTBACK:
    do
      {
      RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
        RM4);
      if (rrc == MATCH_MATCH) break;
      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
      ecode += GET(ecode, 1);
      }
    while (*ecode == OP_ALT);
    if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);

    /* If checking an assertion for a condition, return MATCH_MATCH. */

    if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);

    /* Continue from after the assertion, updating the offsets high water
    mark, since extracts may have been taken during the assertion. */

    do ecode += GET(ecode,1); while (*ecode == OP_ALT);
    ecode += 1 + LINK_SIZE;
    offset_top = md->end_offset_top;
    continue;

    /* Negative assertion: all branches must fail to match */

    case OP_ASSERT_NOT:
    case OP_ASSERTBACK_NOT:
    do
      {
      RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
        RM5);
      if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);
      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
      ecode += GET(ecode,1);
      }
    while (*ecode == OP_ALT);

    if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);

    ecode += 1 + LINK_SIZE;
    continue;

    /* Move the subject pointer back. This occurs only at the start of
    each branch of a lookbehind assertion. If we are too close to the start to
    move back, this match function fails. When working with UTF-8 we move
    back a number of characters, not bytes. */

    case OP_REVERSE:
#ifdef SUPPORT_UTF8
    if (utf8)
      {
      i = GET(ecode, 1);
      while (i-- > 0)
        {
        eptr--;
        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
        BACKCHAR(eptr);
        }
      }
    else
#endif

    /* No UTF-8 support, or not in UTF-8 mode: count is byte count */

      {
      eptr -= GET(ecode, 1);
      if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
      }

    /* Save the earliest consulted character, then skip to next op code */

    if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
    ecode += 1 + LINK_SIZE;
    break;

    /* The callout item calls an external function, if one is provided, passing
    details of the match so far. This is mainly for debugging, though the
    function is able to force a failure. */

    case OP_CALLOUT:
    if (pcre_callout != NULL)
      {
      pcre_callout_block cb;
      cb.version          = 1;   /* Version 1 of the callout block */
      cb.callout_number   = ecode[1];
      cb.offset_vector    = md->offset_vector;
      cb.subject          = (PCRE_SPTR)md->start_subject;
      cb.subject_length   = md->end_subject - md->start_subject;
      cb.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);
      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 += 2 + 2*LINK_SIZE;
    break;

    /* Recursion either matches the current regex, or some subexpression. The
    offset data is the offset to the starting bracket from the start of the
    whole pattern. (This is so that it works from duplicated subpatterns.)

    If there are any capturing brackets started but not finished, we have to
    save their starting points and reinstate them after the recursion. However,
    we don't know how many such there are (offset_top records the completed
    total) so we just have to save all the potential data. There may be up to
    65535 such values, which is too large to put on the stack, but using malloc
    for small numbers seems expensive. As a compromise, the stack is used when
    there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc
    is used. A problem is what to do if the malloc fails ... there is no way of
    returning to the top level with an error. Save the top REC_STACK_SAVE_MAX
    values on the stack, and accept that the rest may be wrong.

    There are also other values that have to be saved. We use a chained
    sequence of blocks that actually live on the stack. Thanks to Robin Houston
    for the original version of this logic. */

    case OP_RECURSE:
      {
      callpat = md->start_code + GET(ecode, 1);
      new_recursive.group_num = (callpat == md->start_code)? 0 :
        GET2(callpat, 1 + LINK_SIZE);

      /* Add to "recursing stack" */

      new_recursive.prevrec = md->recursive;
      md->recursive = &new_recursive;

      /* Find where to continue from afterwards */

      ecode += 1 + LINK_SIZE;
      new_recursive.after_call = ecode;

      /* Now save the offset data. */

      new_recursive.saved_max = md->offset_end;
      if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
        new_recursive.offset_save = stacksave;
      else
        {
        new_recursive.offset_save =
          (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
        if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
        }

      memcpy(new_recursive.offset_save, md->offset_vector,
            new_recursive.saved_max * sizeof(int));
      new_recursive.save_start = mstart;
      new_recursive.save_offset_top = offset_top;
      mstart = eptr;

      /* OK, now we can do the recursion. For each top-level alternative we
      restore the offset and recursion data. */

      DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
      flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;
      do
        {
        RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
          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 && 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,
            new_recursive.saved_max * sizeof(int));
        callpat += GET(callpat, 1);
        }
      while (*callpat == OP_ALT);

      DPRINTF(("Recursion didn't match\n"));
      md->recursive = new_recursive.prevrec;
      if (new_recursive.offset_save != stacksave)
        (pcre_free)(new_recursive.offset_save);
      RRETURN(MATCH_NOMATCH);
      }
    /* Control never reaches here */

    /* "Once" brackets are like assertion brackets except that after a match,
    the point in the subject string is not moved back. Thus there can never be
    a move back into the brackets. Friedl calls these "atomic" subpatterns.
    Check the alternative branches in turn - the matching won't pass the KET
    for this kind of subpattern. If any one branch matches, we carry on as at
    the end of a normal bracket, leaving the subject pointer. */

    case OP_ONCE:
    prev = ecode;
    saved_eptr = eptr;

    do
      {
      RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);
      if (rrc == MATCH_MATCH) break;
      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
      ecode += GET(ecode,1);
      }
    while (*ecode == OP_ALT);

    /* If hit the end of the group (which could be repeated), fail */

    if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);

    /* Continue as from after the assertion, updating the offsets high water
    mark, since extracts may have been taken. */

    do ecode += GET(ecode, 1); while (*ecode == OP_ALT);

    offset_top = md->end_offset_top;
    eptr = md->end_match_ptr;

    /* For a non-repeating ket, just continue at this level. This also
    happens for a repeating ket if no characters were matched in the group.
    This is the forcible breaking of infinite loops as implemented in Perl
    5.005. If there is an options reset, it will get obeyed in the normal
    course of events. */

    if (*ecode == OP_KET || eptr == saved_eptr)
      {
      ecode += 1+LINK_SIZE;
      break;
      }

    /* The repeating kets try the rest of the pattern or restart from the
    preceding bracket, in the appropriate order. The second "call" of match()
    uses tail recursion, to avoid using another stack frame. We need to reset
    any options that changed within the bracket before re-running it, so
    check the next opcode. */

    if (ecode[1+LINK_SIZE] == OP_OPT)
      {
      ims = (ims & ~PCRE_IMS) | ecode[4];
      DPRINTF(("ims set to %02lx at group repeat\n", ims));
      }

    if (*ecode == OP_KETRMIN)
      {
      RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);
      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
      ecode = prev;
      flags = 0;
      goto TAIL_RECURSE;
      }
    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. */

    case OP_ALT:
    do ecode += GET(ecode,1); while (*ecode == OP_ALT);
    break;

    /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
    indicating that it may occur zero times. It may repeat infinitely, or not
    at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
    with fixed upper repeat limits are compiled as a number of copies, with the
    optional ones preceded by BRAZERO or BRAMINZERO. */

    case OP_BRAZERO:
      {
      next = ecode+1;
      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;
      }
    break;

    case OP_BRAMINZERO:
      {
      next = ecode+1;
      do next += GET(next, 1); while (*next == OP_ALT);
      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);
      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
      ecode++;
      }
    break;

    case OP_SKIPZERO:
      {
      next = ecode+1;
      do next += GET(next,1); while (*next == OP_ALT);
      ecode = next + 1 + LINK_SIZE;
      }
    break;

    /* End of a group, repeated or non-repeating. */

    case OP_KET:
    case OP_KETRMIN:
    case OP_KETRMAX:
    prev = ecode - GET(ecode, 1);

    /* 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. */

    if (*prev >= OP_SBRA)
      {
      saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
      eptrb = eptrb->epb_prev;              /* Backup to previous group */
      }
    else saved_eptr = NULL;

    /* If we are at the end of an assertion group, stop matching and return
    MATCH_MATCH, but record the current high water mark for use by positive
    assertions. Do this also for the "once" (atomic) groups. */

    if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
        *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
        *prev == OP_ONCE)
      {
      md->end_match_ptr = eptr;      /* For ONCE */
      md->end_offset_top = offset_top;
      RRETURN(MATCH_MATCH);
      }

    /* For capturing groups we have to check the group number back 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 (*prev == OP_CBRA || *prev == OP_SCBRA)
      {
      number = GET2(prev, 1+LINK_SIZE);
      offset = number << 1;

#ifdef DEBUG
      printf("end bracket %d", number);
      printf("\n");
#endif

      md->capture_last = number;
      if (offset >= md->offset_max) md->offset_overflow = TRUE; else
        {
        md->offset_vector[offset] =
          md->offset_vector[md->offset_end - number];
        md->offset_vector[offset+1] = eptr - md->start_subject;
        if (offset_top <= offset) offset_top = offset + 2;
        }

      /* Handle a recursively called group. Restore the offsets
      appropriately and continue from after the call. */

      if (md->recursive != NULL && md->recursive->group_num == number)
        {
        recursion_info *rec = md->recursive;
        DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
        md->recursive = rec->prevrec;
        mstart = rec->save_start;
        memcpy(md->offset_vector, rec->offset_save,
          rec->saved_max * sizeof(int));
        offset_top = rec->save_offset_top;
        ecode = rec->after_call;
        ims = original_ims;
        break;
        }
      }

    /* For both capturing and non-capturing groups, reset the value of the ims
    flags, in case they got changed during the group. */

    ims = original_ims;
    DPRINTF(("ims reset to %02lx\n", ims));

    /* For a non-repeating ket, just continue at this level. This also
    happens for a repeating ket if no characters were matched in the group.
    This is the forcible breaking of infinite loops as implemented in Perl
    5.005. If there is an options reset, it will get obeyed in the normal
    course of events. */

    if (*ecode == OP_KET || eptr == saved_eptr)
      {
      ecode += 1 + LINK_SIZE;
      break;
      }

    /* The repeating kets try the rest of the pattern or restart from the
    preceding bracket, in the appropriate order. 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. */

    flags = (*prev >= OP_SBRA)? match_cbegroup : 0;

    if (*ecode == OP_KETRMIN)
      {
      RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);
      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
      if (flags != 0)    /* Could match an empty string */
        {
        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);
        RRETURN(rrc);
        }
      ecode = prev;
      goto TAIL_RECURSE;
      }
    else  /* OP_KETRMAX */
      {
      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 */

    case OP_CIRC:
    if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
    if ((ims & PCRE_MULTILINE) != 0)
      {
      if (eptr != md->start_subject &&
          (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
        RRETURN(MATCH_NOMATCH);
      ecode++;
      break;
      }
    /* ... else fall through */

    /* Start of subject assertion */

    case OP_SOD:
    if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    /* Start of match assertion */

    case OP_SOM:
    if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
    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. */

    case OP_DOLL:
    if ((ims & PCRE_MULTILINE) != 0)
      {
      if (eptr < md->end_subject)
        { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
      else
        { if (md->noteol) RRETURN(MATCH_NOMATCH); }
      ecode++;
      break;
      }
    else
      {
      if (md->noteol) RRETURN(MATCH_NOMATCH);
      if (!md->endonly)
        {
        if (eptr != md->end_subject &&
            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
          RRETURN(MATCH_NOMATCH);
        ecode++;
        break;
        }
      }
    /* ... else fall through for endonly */

    /* End of subject assertion (\z) */

    case OP_EOD:
    if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    /* End of subject or ending \n assertion (\Z) */

    case OP_EODN:
    if (eptr != md->end_subject &&
        (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
      RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    /* Word boundary assertions */

    case OP_NOT_WORD_BOUNDARY:
    case OP_WORD_BOUNDARY:
      {

      /* Find out if the previous and current characters are "word" characters.
      It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
      be "non-word" characters. Remember the earliest consulted character for
      partial matching. */

#ifdef SUPPORT_UTF8
      if (utf8)
        {
        if (eptr == md->start_subject) prev_is_word = FALSE; else
          {
          USPTR lastptr = eptr - 1;
          while((*lastptr & 0xc0) == 0x80) lastptr--;
          if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
          GETCHAR(c, lastptr);
          prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
          }
        if (eptr >= md->end_subject)
          {
          SCHECK_PARTIAL();
          cur_is_word = FALSE;
          }
        else
          {
          GETCHAR(c, eptr);
          cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
          }
        }
      else
#endif

      /* Not in UTF-8 mode */

        {
        if (eptr == md->start_subject) prev_is_word = FALSE; else
          {
          if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
          prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
          }
        if (eptr >= md->end_subject)
          {
          SCHECK_PARTIAL();
          cur_is_word = FALSE;
          }
        else cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
        }

      /* Now see if the situation is what we want */

      if ((*ecode++ == OP_WORD_BOUNDARY)?
           cur_is_word == prev_is_word : cur_is_word != prev_is_word)
        RRETURN(MATCH_NOMATCH);
      }
    break;

    /* Match a single character type; inline for speed */

    case OP_ANY:
    if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
    /* Fall through */

    case OP_ALLANY:
    if (eptr++ >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
    ecode++;
    break;

    /* Match a single byte, even in UTF-8 mode. This opcode really does match
    any byte, even newline, independent of the setting of PCRE_DOTALL. */

    case OP_ANYBYTE:
    if (eptr++ >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    ecode++;
    break;

    case OP_NOT_DIGIT:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    if (
#ifdef SUPPORT_UTF8
       c < 256 &&
#endif
       (md->ctypes[c] & ctype_digit) != 0
       )
      RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    case OP_DIGIT:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    if (
#ifdef SUPPORT_UTF8
       c >= 256 ||
#endif
       (md->ctypes[c] & ctype_digit) == 0
       )
      RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    case OP_NOT_WHITESPACE:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    if (
#ifdef SUPPORT_UTF8
       c < 256 &&
#endif
       (md->ctypes[c] & ctype_space) != 0
       )
      RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    case OP_WHITESPACE:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    if (
#ifdef SUPPORT_UTF8
       c >= 256 ||
#endif
       (md->ctypes[c] & ctype_space) == 0
       )
      RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    case OP_NOT_WORDCHAR:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    if (
#ifdef SUPPORT_UTF8
       c < 256 &&
#endif
       (md->ctypes[c] & ctype_word) != 0
       )
      RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    case OP_WORDCHAR:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    if (
#ifdef SUPPORT_UTF8
       c >= 256 ||
#endif
       (md->ctypes[c] & ctype_word) == 0
       )
      RRETURN(MATCH_NOMATCH);
    ecode++;
    break;

    case OP_ANYNL:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      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)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    switch(c)
      {
      default: break;
      case 0x09:      /* HT */
      case 0x20:      /* SPACE */
      case 0xa0:      /* NBSP */
      case 0x1680:    /* OGHAM SPACE MARK */
      case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
      case 0x2000:    /* EN QUAD */
      case 0x2001:    /* EM QUAD */
      case 0x2002:    /* EN SPACE */
      case 0x2003:    /* EM SPACE */
      case 0x2004:    /* THREE-PER-EM SPACE */
      case 0x2005:    /* FOUR-PER-EM SPACE */
      case 0x2006:    /* SIX-PER-EM SPACE */
      case 0x2007:    /* FIGURE SPACE */
      case 0x2008:    /* PUNCTUATION SPACE */
      case 0x2009:    /* THIN SPACE */
      case 0x200A:    /* HAIR SPACE */
      case 0x202f:    /* NARROW NO-BREAK SPACE */
      case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
      case 0x3000:    /* IDEOGRAPHIC SPACE */
      RRETURN(MATCH_NOMATCH);
      }
    ecode++;
    break;

    case OP_HSPACE:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    switch(c)
      {
      default: RRETURN(MATCH_NOMATCH);
      case 0x09:      /* HT */
      case 0x20:      /* SPACE */
      case 0xa0:      /* NBSP */
      case 0x1680:    /* OGHAM SPACE MARK */
      case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
      case 0x2000:    /* EN QUAD */
      case 0x2001:    /* EM QUAD */
      case 0x2002:    /* EN SPACE */
      case 0x2003:    /* EM SPACE */
      case 0x2004:    /* THREE-PER-EM SPACE */
      case 0x2005:    /* FOUR-PER-EM SPACE */
      case 0x2006:    /* SIX-PER-EM SPACE */
      case 0x2007:    /* FIGURE SPACE */
      case 0x2008:    /* PUNCTUATION SPACE */
      case 0x2009:    /* THIN SPACE */
      case 0x200A:    /* HAIR SPACE */
      case 0x202f:    /* NARROW NO-BREAK SPACE */
      case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
      case 0x3000:    /* IDEOGRAPHIC SPACE */
      break;
      }
    ecode++;
    break;

    case OP_NOT_VSPACE:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    switch(c)
      {
      default: break;
      case 0x0a:      /* LF */
      case 0x0b:      /* VT */
      case 0x0c:      /* FF */
      case 0x0d:      /* CR */
      case 0x85:      /* NEL */
      case 0x2028:    /* LINE SEPARATOR */
      case 0x2029:    /* PARAGRAPH SEPARATOR */
      RRETURN(MATCH_NOMATCH);
      }
    ecode++;
    break;

    case OP_VSPACE:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
    switch(c)
      {
      default: RRETURN(MATCH_NOMATCH);
      case 0x0a:      /* LF */
      case 0x0b:      /* VT */
      case 0x0c:      /* FF */
      case 0x0d:      /* CR */
      case 0x85:      /* NEL */
      case 0x2028:    /* LINE SEPARATOR */
      case 0x2029:    /* PARAGRAPH SEPARATOR */
      break;
      }
    ecode++;
    break;

#ifdef SUPPORT_UCP
    /* Check the next character by Unicode property. We will get here only
    if the support is in the binary; otherwise a compile-time error occurs. */

    case OP_PROP:
    case OP_NOTPROP:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
      {
      const ucd_record *prop = GET_UCD(c);

      switch(ecode[1])
        {
        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;

        case PT_GC:
        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;

    /* Match an extended Unicode sequence. We will get here only if the support
    is in the binary; otherwise a compile-time error occurs. */

    case OP_EXTUNI:
    if (eptr >= md->end_subject)
      {
      SCHECK_PARTIAL();
      RRETURN(MATCH_NOMATCH);
      }
    GETCHARINCTEST(c, eptr);
      {
      int category = UCD_CATEGORY(c);
      if (category == ucp_M) RRETURN(MATCH_NOMATCH);
      while (eptr < md->end_subject)
        {
        int len = 1;
        if (!utf8) c = *eptr; else
          {
          GETCHARLEN(c, eptr, len);
          }
        category = UCD_CATEGORY(c);
        if (category != ucp_M) break;
        eptr += len;
        }
      }
    ecode++;
    break;
#endif


    /* Match a back reference, possibly repeatedly. Look past the end of the
    item to see if there is repeat information following. The code is similar
    to that for character classes, but repeated for efficiency. Then obey
    similar code to character type repeats - written out again for speed.
    However, if the referenced string is the empty string, always treat
    it as matched, any number of times (otherwise there could be infinite
    loops). */

    case OP_REF:
      {
      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
      ecode += 3;

      /* If the reference is unset, there are two possibilities:

      (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.

      (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:
        case OP_CRPLUS:
        case OP_CRMINPLUS:
        case OP_CRQUERY:
        case OP_CRMINQUERY:
        c