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

Diff of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 77 by nigel, Sat Feb 24 21:40:45 2007 UTC revision 615 by ph10, Mon Jul 11 14:23:06 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
43  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK md             /* Block containing newline information */
50    #define PSSTART start_subject  /* Field containing processed string start */
51    #define PSEND   end_subject    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    /* Undefine some potentially clashing cpp symbols */
56    
57  /* Structure for building a chain of data that actually lives on the  #undef min
58  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;  
59    
60  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
61    of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
63    
64  #define match_condassert   0x01    /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_isgroup      0x02    /* Set if start of bracketed group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 68  defined PCRE_ERROR_xxx codes, which are Line 70  defined PCRE_ERROR_xxx codes, which are
70  #define MATCH_MATCH        1  #define MATCH_MATCH        1
71  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
72    
73    /* Special internal returns from the match() function. Make them sufficiently
74    negative to avoid the external error codes. */
75    
76    #define MATCH_ACCEPT       (-999)
77    #define MATCH_COMMIT       (-998)
78    #define MATCH_KETRPOS      (-997)
79    #define MATCH_PRUNE        (-996)
80    #define MATCH_SKIP         (-995)
81    #define MATCH_SKIP_ARG     (-994)
82    #define MATCH_THEN         (-993)
83    
84    /* This is a convenience macro for code that occurs many times. */
85    
86    #define MRRETURN(ra) \
87      { \
88      md->mark = markptr; \
89      RRETURN(ra); \
90      }
91    
92  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
93  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
94  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 81  static const char rep_max[] = { 0, 0, 0, Line 102  static const char rep_max[] = { 0, 0, 0,
102    
103    
104    
105  #ifdef DEBUG  #ifdef PCRE_DEBUG
106  /*************************************************  /*************************************************
107  *        Debugging function to print chars       *  *        Debugging function to print chars       *
108  *************************************************/  *************************************************/
# Line 101  Returns:     nothing Line 122  Returns:     nothing
122  static void  static void
123  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)
124  {  {
125  int c;  unsigned int c;
126  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
127  while (length-- > 0)  while (length-- > 0)
128    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);
# Line 114  while (length-- > 0) Line 135  while (length-- > 0)
135  *          Match a back-reference                *  *          Match a back-reference                *
136  *************************************************/  *************************************************/
137    
138  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
139  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
140    the length passed is zero. Note that in caseless UTF-8 mode, the number of
141    subject bytes matched may be different to the number of reference bytes.
142    
143  Arguments:  Arguments:
144    offset      index into the offset vector    offset      index into the offset vector
145    eptr        points into the subject    eptr        pointer into the subject
146    length      length to be matched    length      length of reference to be matched (number of bytes)
147    md          points to match data block    md          points to match data block
148    ims         the ims flags    caseless    TRUE if caseless
149    
150  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
151  */  */
152    
153  static BOOL  static int
154  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,
155    unsigned long int ims)    BOOL caseless)
156  {  {
157  const uschar *p = md->start_subject + md->offset_vector[offset];  USPTR eptr_start = eptr;
158    register USPTR p = md->start_subject + md->offset_vector[offset];
159    
160  #ifdef DEBUG  #ifdef PCRE_DEBUG
161  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
162    printf("matching subject <null>");    printf("matching subject <null>");
163  else  else
# Line 146  pchars(p, length, FALSE, md); Line 170  pchars(p, length, FALSE, md);
170  printf("\n");  printf("\n");
171  #endif  #endif
172    
173  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
174    
175  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
176    
177  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
178    properly if Unicode properties are supported. Otherwise, we can check only
179    ASCII characters. */
180    
181  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
182    {    {
183    while (length-- > 0)  #ifdef SUPPORT_UTF8
184      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
185      if (md->utf8)
186        {
187        /* Match characters up to the end of the reference. NOTE: the number of
188        bytes matched may differ, because there are some characters whose upper and
189        lower case versions code as different numbers of bytes. For example, U+023A
190        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
191        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
192        the latter. It is important, therefore, to check the length along the
193        reference, not along the subject (earlier code did this wrong). */
194    
195        USPTR endptr = p + length;
196        while (p < endptr)
197          {
198          int c, d;
199          if (eptr >= md->end_subject) return -1;
200          GETCHARINC(c, eptr);
201          GETCHARINC(d, p);
202          if (c != d && c != UCD_OTHERCASE(d)) return -1;
203          }
204        }
205      else
206    #endif
207    #endif
208    
209      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
210      is no UCP support. */
211        {
212        if (eptr + length > md->end_subject) return -1;
213        while (length-- > 0)
214          { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }
215        }
216    }    }
217    
218    /* In the caseful case, we can just compare the bytes, whether or not we
219    are in UTF-8 mode. */
220    
221  else  else
222    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
223      if (eptr + length > md->end_subject) return -1;
224      while (length-- > 0) if (*p++ != *eptr++) return -1;
225      }
226    
227  return TRUE;  return eptr - eptr_start;
228  }  }
229    
230    
# Line 169  return TRUE; Line 233  return TRUE;
233  ****************************************************************************  ****************************************************************************
234                     RECURSION IN THE match() FUNCTION                     RECURSION IN THE match() FUNCTION
235    
236  The match() function is highly recursive. Some regular expressions can cause  The match() function is highly recursive, though not every recursive call
237  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
238  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
239  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
240  works fine.  saved for a recursive call. On Unix, the stack can be large, and this works
241    fine.
242  It turns out that on non-Unix systems there are problems with programs that  
243  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
244  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
245  for decades.) So....  has oodles of memory these days, and techniques for extending the stack have
246    been known for decades.) So....
247    
248  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
249  calls by keeping local variables that need to be preserved in blocks of memory  calls by keeping local variables that need to be preserved in blocks of memory
250  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
251  achieve this so that the actual code doesn't look very different to what it  achieve this so that the actual code doesn't look very different to what it
252  always used to.  always used to.
253    
254    The original heap-recursive code used longjmp(). However, it seems that this
255    can be very slow on some operating systems. Following a suggestion from Stan
256    Switzer, the use of longjmp() has been abolished, at the cost of having to
257    provide a unique number for each call to RMATCH. There is no way of generating
258    a sequence of numbers at compile time in C. I have given them names, to make
259    them stand out more clearly.
260    
261    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
262    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
263    tests. Furthermore, not using longjmp() means that local dynamic variables
264    don't have indeterminate values; this has meant that the frame size can be
265    reduced because the result can be "passed back" by straight setting of the
266    variable instead of being passed in the frame.
267  ****************************************************************************  ****************************************************************************
268  ***************************************************************************/  ***************************************************************************/
269    
270    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
271    below must be updated in sync.  */
272    
273  /* These versions of the macros use the stack, as normal */  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
274           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
275           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
276           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
277           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
278           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
279           RM61,  RM62, RM63};
280    
281    /* These versions of the macros use the stack, as normal. There are debugging
282    versions and production versions. Note that the "rw" argument of RMATCH isn't
283    actually used in this definition. */
284    
285  #ifndef NO_RECURSE  #ifndef NO_RECURSE
286  #define REGISTER register  #define REGISTER register
287  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg)  
288    #ifdef PCRE_DEBUG
289    #define RMATCH(ra,rb,rc,rd,re,rw) \
290      { \
291      printf("match() called in line %d\n", __LINE__); \
292      rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1); \
293      printf("to line %d\n", __LINE__); \
294      }
295    #define RRETURN(ra) \
296      { \
297      printf("match() returned %d from line %d ", ra, __LINE__); \
298      return ra; \
299      }
300    #else
301    #define RMATCH(ra,rb,rc,rd,re,rw) \
302      rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1)
303  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
304    #endif
305    
306  #else  #else
307    
308    
309  /* 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
310  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
311  match(), which never changes. */  argument of match(), which never changes. */
312    
313  #define REGISTER  #define REGISTER
314    
315  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
316    {\    {\
317    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\
318    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
319      {\    frame->Xwhere = rw; \
320      newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
321      newframe->Xecode = rb;\    newframe->Xecode = rb;\
322      newframe->Xoffset_top = rc;\    newframe->Xmstart = mstart;\
323      newframe->Xims = re;\    newframe->Xmarkptr = markptr;\
324      newframe->Xeptrb = rf;\    newframe->Xoffset_top = rc;\
325      newframe->Xflags = rg;\    newframe->Xeptrb = re;\
326      newframe->Xprevframe = frame;\    newframe->Xrdepth = frame->Xrdepth + 1;\
327      frame = newframe;\    newframe->Xprevframe = frame;\
328      DPRINTF(("restarting from line %d\n", __LINE__));\    frame = newframe;\
329      goto HEAP_RECURSE;\    DPRINTF(("restarting from line %d\n", __LINE__));\
330      }\    goto HEAP_RECURSE;\
331    else\    L_##rw:\
332      {\    DPRINTF(("jumped back to line %d\n", __LINE__));\
     DPRINTF(("longjumped back to line %d\n", __LINE__));\  
     frame = md->thisframe;\  
     rx = frame->Xresult;\  
     }\  
333    }    }
334    
335  #define RRETURN(ra)\  #define RRETURN(ra)\
336    {\    {\
337    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
338    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
339    (pcre_stack_free)(newframe);\    (pcre_stack_free)(oldframe);\
340    if (frame != NULL)\    if (frame != NULL)\
341      {\      {\
342      frame->Xresult = ra;\      rrc = ra;\
343      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
344      }\      }\
345    return ra;\    return ra;\
346    }    }
# Line 250  typedef struct heapframe { Line 353  typedef struct heapframe {
353    
354    /* Function arguments that may change */    /* Function arguments that may change */
355    
356    const uschar *Xeptr;    USPTR Xeptr;
357    const uschar *Xecode;    const uschar *Xecode;
358      USPTR Xmstart;
359      USPTR Xmarkptr;
360    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
361    eptrblock *Xeptrb;    eptrblock *Xeptrb;
362    int Xflags;    unsigned int Xrdepth;
363    
364    /* Function local variables */    /* Function local variables */
365    
366    const uschar *Xcallpat;    USPTR Xcallpat;
367    const uschar *Xcharptr;  #ifdef SUPPORT_UTF8
368    const uschar *Xdata;    USPTR Xcharptr;
369    const uschar *Xnext;  #endif
370    const uschar *Xpp;    USPTR Xdata;
371    const uschar *Xprev;    USPTR Xnext;
372    const uschar *Xsaved_eptr;    USPTR Xpp;
373      USPTR Xprev;
374      USPTR Xsaved_eptr;
375    
376    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
377    
378    BOOL Xcur_is_word;    BOOL Xcur_is_word;
379    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
380    BOOL Xprev_is_word;    BOOL Xprev_is_word;
381    
   unsigned long int Xoriginal_ims;  
   
382  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
383    int Xprop_type;    int Xprop_type;
384      int Xprop_value;
385    int Xprop_fail_result;    int Xprop_fail_result;
386    int Xprop_category;    int Xprop_category;
387    int Xprop_chartype;    int Xprop_chartype;
388    int Xprop_othercase;    int Xprop_script;
389    int Xprop_test_against;    int Xoclength;
390    int *Xprop_test_variable;    uschar Xocchars[8];
391  #endif  #endif
392    
393      int Xcodelink;
394    int Xctype;    int Xctype;
395    int Xfc;    unsigned int Xfc;
396    int Xfi;    int Xfi;
397    int Xlength;    int Xlength;
398    int Xmax;    int Xmax;
# Line 301  typedef struct heapframe { Line 406  typedef struct heapframe {
406    
407    eptrblock Xnewptrb;    eptrblock Xnewptrb;
408    
409    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
410    
411    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
412    
413  } heapframe;  } heapframe;
414    
# Line 320  typedef struct heapframe { Line 424  typedef struct heapframe {
424  *         Match from current position            *  *         Match from current position            *
425  *************************************************/  *************************************************/
426    
427  /* 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  
428  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
429  same response.  same response. */
430    
431    /* These macros pack up tests that are used for partial matching, and which
432    appears several times in the code. We set the "hit end" flag if the pointer is
433    at the end of the subject and also past the start of the subject (i.e.
434    something has been matched). For hard partial matching, we then return
435    immediately. The second one is used when we already know we are past the end of
436    the subject. */
437    
438    #define CHECK_PARTIAL()\
439      if (md->partial != 0 && eptr >= md->end_subject && \
440          eptr > md->start_used_ptr) \
441        { \
442        md->hitend = TRUE; \
443        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
444        }
445    
446    #define SCHECK_PARTIAL()\
447      if (md->partial != 0 && eptr > md->start_used_ptr) \
448        { \
449        md->hitend = TRUE; \
450        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
451        }
452    
453    
454  Performance note: It might be tempting to extract commonly used fields from the  /* Performance note: It might be tempting to extract commonly used fields from
455  md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf8, end_subject) into individual variables to improve
456  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
457  made performance worse.  made performance worse.
458    
459  Arguments:  Arguments:
460     eptr        pointer in subject     eptr        pointer to current character in subject
461     ecode       position in code     ecode       pointer to current position in compiled code
462       mstart      pointer to the current match start position (can be modified
463                     by encountering \K)
464       markptr     pointer to the most recent MARK name, or NULL
465     offset_top  current top pointer     offset_top  current top pointer
466     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
467     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
468                   brackets - for testing for empty matches                   brackets - for testing for empty matches
469     flags       can contain     rdepth      the recursion depth
                  match_condassert - this is an assertion condition  
                  match_isgroup - this is the start of a bracketed group  
470    
471  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
472                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
473                   a negative MATCH_xxx value for PRUNE, SKIP, etc
474                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
475                   (e.g. stopped by recursion limit)                   (e.g. stopped by repeated call or recursion limit)
476  */  */
477    
478  static int  static int
479  match(REGISTER const uschar *eptr, REGISTER const uschar *ecode,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
480    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
481    int flags)    unsigned int rdepth)
482  {  {
483  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
484  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
485  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
486    
487  register int  rrc;    /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
488  register int  i;      /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
489  register int  c;      /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
490  register BOOL utf8;   /* Local copy of UTF-8 flag for speed */  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
491    
492    BOOL minimize, possessive; /* Quantifier options */
493    BOOL caseless;
494    int condcode;
495    
496  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
497  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame" which is obtained from
# Line 370  heap storage. Set up the top-level frame Line 499  heap storage. Set up the top-level frame
499  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
500    
501  #ifdef NO_RECURSE  #ifdef NO_RECURSE
502  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));
503    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
504  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
505    
506  /* Copy in the original argument variables */  /* Copy in the original argument variables */
507    
508  frame->Xeptr = eptr;  frame->Xeptr = eptr;
509  frame->Xecode = ecode;  frame->Xecode = ecode;
510    frame->Xmstart = mstart;
511    frame->Xmarkptr = markptr;
512  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
513  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
514  frame->Xflags = flags;  frame->Xrdepth = rdepth;
515    
516  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
517    
# Line 390  HEAP_RECURSE: Line 521  HEAP_RECURSE:
521    
522  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
523  #define ecode              frame->Xecode  #define ecode              frame->Xecode
524    #define mstart             frame->Xmstart
525    #define markptr            frame->Xmarkptr
526  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
527  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
528  #define flags              frame->Xflags  #define rdepth             frame->Xrdepth
529    
530  /* Ditto for the local variables */  /* Ditto for the local variables */
531    
# Line 401  HEAP_RECURSE: Line 533  HEAP_RECURSE:
533  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
534  #endif  #endif
535  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
536    #define codelink           frame->Xcodelink
537  #define data               frame->Xdata  #define data               frame->Xdata
538  #define next               frame->Xnext  #define next               frame->Xnext
539  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 411  HEAP_RECURSE: Line 544  HEAP_RECURSE:
544    
545  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
546  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
547  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
548    
 #define original_ims       frame->Xoriginal_ims  
   
549  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
550  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
551    #define prop_value         frame->Xprop_value
552  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
553  #define prop_category      frame->Xprop_category  #define prop_category      frame->Xprop_category
554  #define prop_chartype      frame->Xprop_chartype  #define prop_chartype      frame->Xprop_chartype
555  #define prop_othercase     frame->Xprop_othercase  #define prop_script        frame->Xprop_script
556  #define prop_test_against  frame->Xprop_test_against  #define oclength           frame->Xoclength
557  #define prop_test_variable frame->Xprop_test_variable  #define occhars            frame->Xocchars
558  #endif  #endif
559    
560  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 447  HEAP_RECURSE: Line 578  HEAP_RECURSE:
578  get preserved during recursion in the normal way. In this environment, fi and  get preserved during recursion in the normal way. In this environment, fi and
579  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
580    
581  #else  #else         /* NO_RECURSE not defined */
582  #define fi i  #define fi i
583  #define fc c  #define fc c
584    
585    /* Many of the following variables are used only in small blocks of the code.
586  #ifdef SUPPORT_UTF8                /* Many of these variables are used ony */  My normal style of coding would have declared them within each of those blocks.
587  const uschar *charptr;             /* small blocks of the code. My normal  */  However, in order to accommodate the version of this code that uses an external
588  #endif                             /* style of coding would have declared  */  "stack" implemented on the heap, it is easier to declare them all here, so the
589  const uschar *callpat;             /* them within each of those blocks.    */  declarations can be cut out in a block. The only declarations within blocks
590  const uschar *data;                /* However, in order to accommodate the */  below are for variables that do not have to be preserved over a recursive call
591  const uschar *next;                /* version of this code that uses an    */  to RMATCH(). */
592  const uschar *pp;                  /* external "stack" implemented on the  */  
593  const uschar *prev;                /* heap, it is easier to declare them   */  #ifdef SUPPORT_UTF8
594  const uschar *saved_eptr;          /* all here, so the declarations can    */  const uschar *charptr;
595                                     /* be cut out in a block. The only      */  #endif
596  recursion_info new_recursive;      /* declarations within blocks below are */  const uschar *callpat;
597                                     /* for variables that do not have to    */  const uschar *data;
598  BOOL cur_is_word;                  /* be preserved over a recursive call   */  const uschar *next;
599  BOOL condition;                    /* to RMATCH().                         */  USPTR         pp;
600  BOOL minimize;  const uschar *prev;
601    USPTR         saved_eptr;
602    
603    recursion_info new_recursive;
604    
605    BOOL cur_is_word;
606    BOOL condition;
607  BOOL prev_is_word;  BOOL prev_is_word;
608    
 unsigned long int original_ims;  
   
609  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
610  int prop_type;  int prop_type;
611    int prop_value;
612  int prop_fail_result;  int prop_fail_result;
613  int prop_category;  int prop_category;
614  int prop_chartype;  int prop_chartype;
615  int prop_othercase;  int prop_script;
616  int prop_test_against;  int oclength;
617  int *prop_test_variable;  uschar occhars[8];
618  #endif  #endif
619    
620    int codelink;
621  int ctype;  int ctype;
622  int length;  int length;
623  int max;  int max;
# Line 493  int save_offset1, save_offset2, save_off Line 630  int save_offset1, save_offset2, save_off
630  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
631    
632  eptrblock newptrb;  eptrblock newptrb;
633  #endif  #endif     /* NO_RECURSE */
634    
635    /* To save space on the stack and in the heap frame, I have doubled up on some
636    of the local variables that are used only in localised parts of the code, but
637    still need to be preserved over recursive calls of match(). These macros define
638    the alternative names that are used. */
639    
640    #define allow_zero    cur_is_word
641    #define cbegroup      condition
642    #define code_offset   codelink
643    #define condassert    condition
644    #define matched_once  prev_is_word
645    
646  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
647  variables. */  variables. */
648    
649  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
650    prop_value = 0;
651  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_against = 0;  
 prop_test_variable = NULL;  
652  #endif  #endif
653    
 /* 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,  
 these just turn into a recursive call to match() and a "return", respectively.  
 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. */  
654    
655  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  /* This label is used for tail recursion, which is used in a few cases even
656    when NO_RECURSE is not defined, in order to reduce the amount of stack that is
657    used. Thanks to Ian Taylor for noticing this possibility and sending the
658    original patch. */
659    
660    TAIL_RECURSE:
661    
662    /* OK, now we can get on with the real code of the function. Recursive calls
663    are specified by the macro RMATCH and RRETURN is used to return. When
664    NO_RECURSE is *not* defined, these just turn into a recursive call to match()
665    and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
666    defined). However, RMATCH isn't like a function call because it's quite a
667    complicated macro. It has to be used in one particular way. This shouldn't,
668    however, impact performance when true recursion is being used. */
669    
670  original_ims = ims;    /* Save for resetting on ')' */  #ifdef SUPPORT_UTF8
671  utf8 = md->utf8;       /* Local copy of the flag */  utf8 = md->utf8;       /* Local copy of the flag */
672    #else
673    utf8 = FALSE;
674    #endif
675    
676    /* First check that we haven't called match() too many times, or that we
677    haven't exceeded the recursive call limit. */
678    
679    if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
680    if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
681    
682  /* 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
683  stack of such pointers, to be re-instated at the end of the group when we hit  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
684  the closing ket. When match() is called in other circumstances, we don't add to  done this way to save having to use another function argument, which would take
685  this stack. */  up space on the stack. See also MATCH_CONDASSERT below.
686    
687    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
688    such remembered pointers, to be checked when we hit the closing ket, in order
689    to break infinite loops that match no characters. When match() is called in
690    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
691    NOT be used with tail recursion, because the memory block that is used is on
692    the stack, so a new one may be required for each match(). */
693    
694  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
695    {    {
   newptrb.epb_prev = eptrb;  
696    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
697      newptrb.epb_prev = eptrb;
698    eptrb = &newptrb;    eptrb = &newptrb;
699      md->match_function_type = 0;
700    }    }
701    
702  /* Now start processing the operations. */  /* Now start processing the opcodes. */
703    
704  for (;;)  for (;;)
705    {    {
706      minimize = possessive = FALSE;
707    op = *ecode;    op = *ecode;
708    minimize = FALSE;  
709      switch(op)
710    /* For partial matching, remember if we ever hit the end of the subject after      {
711    matching at least one subject character. */      case OP_MARK:
712        markptr = ecode + 2;
713        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
714          eptrb, RM55);
715    
716        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
717        argument, and we must check whether that argument matches this MARK's
718        argument. It is passed back in md->start_match_ptr (an overloading of that
719        variable). If it does match, we reset that variable to the current subject
720        position and return MATCH_SKIP. Otherwise, pass back the return code
721        unaltered. */
722    
723        if (rrc == MATCH_SKIP_ARG &&
724            strcmp((char *)markptr, (char *)(md->start_match_ptr)) == 0)
725          {
726          md->start_match_ptr = eptr;
727          RRETURN(MATCH_SKIP);
728          }
729    
730        if (md->mark == NULL) md->mark = markptr;
731        RRETURN(rrc);
732    
733        case OP_FAIL:
734        MRRETURN(MATCH_NOMATCH);
735    
736        /* COMMIT overrides PRUNE, SKIP, and THEN */
737    
738        case OP_COMMIT:
739        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
740          eptrb, RM52);
741        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
742            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
743            rrc != MATCH_THEN)
744          RRETURN(rrc);
745        MRRETURN(MATCH_COMMIT);
746    
747    if (md->partial &&      /* PRUNE overrides THEN */
       eptr >= md->end_subject &&  
       eptr > md->start_match)  
     md->hitend = TRUE;  
   
   /* Opening 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. */  
748    
749    if (op > OP_BRA)      case OP_PRUNE:
750      {      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
751      number = op - OP_BRA;        eptrb, RM51);
752        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
753        MRRETURN(MATCH_PRUNE);
754    
755        case OP_PRUNE_ARG:
756        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
757          eptrb, RM56);
758        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
759        md->mark = ecode + 2;
760        RRETURN(MATCH_PRUNE);
761    
762        /* SKIP overrides PRUNE and THEN */
763    
764        case OP_SKIP:
765        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
766          eptrb, RM53);
767        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
768          RRETURN(rrc);
769        md->start_match_ptr = eptr;   /* Pass back current position */
770        MRRETURN(MATCH_SKIP);
771    
772      /* For extended extraction brackets (large number), we have to fish out the      case OP_SKIP_ARG:
773      number from a dummy opcode at the start. */      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
774          eptrb, RM57);
775        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
776          RRETURN(rrc);
777    
778      if (number > EXTRACT_BASIC_MAX)      /* Pass back the current skip name by overloading md->start_match_ptr and
779        number = GET2(ecode, 2+LINK_SIZE);      returning the special MATCH_SKIP_ARG return code. This will either be
780        caught by a matching MARK, or get to the top, where it is treated the same
781        as PRUNE. */
782    
783        md->start_match_ptr = ecode + 2;
784        RRETURN(MATCH_SKIP_ARG);
785    
786        /* For THEN (and THEN_ARG) we pass back the address of the bracket or
787        the alt that is at the start of the current branch. This makes it possible
788        to skip back past alternatives that precede the THEN within the current
789        branch. */
790    
791        case OP_THEN:
792        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
793          eptrb, RM54);
794        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
795        md->start_match_ptr = ecode - GET(ecode, 1);
796        MRRETURN(MATCH_THEN);
797    
798        case OP_THEN_ARG:
799        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],
800          offset_top, md, eptrb, RM58);
801        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
802        md->start_match_ptr = ecode - GET(ecode, 1);
803        md->mark = ecode + LINK_SIZE + 2;
804        RRETURN(MATCH_THEN);
805    
806        /* Handle a capturing bracket, other than those that are possessive with an
807        unlimited repeat. If there is space in the offset vector, save the current
808        subject position in the working slot at the top of the vector. We mustn't
809        change the current values of the data slot, because they may be set from a
810        previous iteration of this group, and be referred to by a reference inside
811        the group. If we fail to match, we need to restore this value and also the
812        values of the final offsets, in case they were set by a previous iteration
813        of the same bracket.
814    
815        If there isn't enough space in the offset vector, treat this as if it were
816        a non-capturing bracket. Don't worry about setting the flag for the error
817        case here; that is handled in the code for KET. */
818    
819        case OP_CBRA:
820        case OP_SCBRA:
821        number = GET2(ecode, 1+LINK_SIZE);
822      offset = number << 1;      offset = number << 1;
823    
824  #ifdef DEBUG  #ifdef PCRE_DEBUG
825      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
826        printf("subject=");
827      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
828      printf("\n");      printf("\n");
829  #endif  #endif
# Line 582  for (;;) Line 836  for (;;)
836        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
837    
838        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
839        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
840            (int)(eptr - md->start_subject);
841    
842        do        for (;;)
843          {          {
844          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
845            match_isgroup);          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
846          if (rrc != MATCH_NOMATCH) RRETURN(rrc);            eptrb, RM1);
847            if (rrc != MATCH_NOMATCH &&
848                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
849              RRETURN(rrc);
850          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
851          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
852            if (*ecode != OP_ALT) break;
853          }          }
       while (*ecode == OP_ALT);  
854    
855        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
856    
# Line 600  for (;;) Line 858  for (;;)
858        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
859        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
860    
861          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
862        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
863        }        }
864    
865      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
866        as a non-capturing bracket. */
867    
868      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
869      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
870    
871    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
872    
873    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
874      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
875      case OP_BRA:     /* Non-capturing bracket: optimized */  
876      DPRINTF(("start bracket 0\n"));      /* Non-capturing bracket, except for possessive with unlimited repeat. Loop
877      do      for all the alternatives. When we get to the final alternative within the
878        brackets, we used to return the result of a recursive call to match()
879        whatever happened so it was possible to reduce stack usage by turning this
880        into a tail recursion, except in the case of a possibly empty group.
881        However, now that there is the possiblity of (*THEN) occurring in the final
882        alternative, this optimization is no longer possible. */
883    
884        case OP_BRA:
885        case OP_SBRA:
886        DPRINTF(("start non-capturing bracket\n"));
887        for (;;)
888          {
889          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
890          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
891            RM2);
892          if (rrc != MATCH_NOMATCH &&
893              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
894            RRETURN(rrc);
895          ecode += GET(ecode, 1);
896          if (*ecode != OP_ALT) break;
897          }
898    
899        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
900        RRETURN(MATCH_NOMATCH);
901    
902        /* Handle possessive capturing brackets with an unlimited repeat. We come
903        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
904        handled similarly to the normal case above. However, the matching is
905        different. The end of these brackets will always be OP_KETRPOS, which
906        returns MATCH_KETRPOS without going further in the pattern. By this means
907        we can handle the group by iteration rather than recursion, thereby
908        reducing the amount of stack needed. */
909    
910        case OP_CBRAPOS:
911        case OP_SCBRAPOS:
912        allow_zero = FALSE;
913    
914        POSSESSIVE_CAPTURE:
915        number = GET2(ecode, 1+LINK_SIZE);
916        offset = number << 1;
917    
918    #ifdef PCRE_DEBUG
919        printf("start possessive bracket %d\n", number);
920        printf("subject=");
921        pchars(eptr, 16, TRUE, md);
922        printf("\n");
923    #endif
924    
925        if (offset < md->offset_max)
926        {        {
927        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        matched_once = FALSE;
928          match_isgroup);        code_offset = ecode - md->start_code;
929        if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
930          save_offset1 = md->offset_vector[offset];
931          save_offset2 = md->offset_vector[offset+1];
932          save_offset3 = md->offset_vector[md->offset_end - number];
933          save_capture_last = md->capture_last;
934    
935          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
936    
937          /* Each time round the loop, save the current subject position for use
938          when the group matches. For MATCH_MATCH, the group has matched, so we
939          restart it with a new subject starting position, remembering that we had
940          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
941          usual. If we haven't matched any alternatives in any iteration, check to
942          see if a previous iteration matched. If so, the group has matched;
943          continue from afterwards. Otherwise it has failed; restore the previous
944          capture values before returning NOMATCH. */
945    
946          for (;;)
947            {
948            md->offset_vector[md->offset_end - number] =
949              (int)(eptr - md->start_subject);
950            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
951            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
952              eptrb, RM63);
953            if (rrc == MATCH_KETRPOS)
954              {
955              offset_top = md->end_offset_top;
956              eptr = md->end_match_ptr;
957              ecode = md->start_code + code_offset;
958              save_capture_last = md->capture_last;
959              matched_once = TRUE;
960              continue;
961              }
962            if (rrc != MATCH_NOMATCH &&
963                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
964              RRETURN(rrc);
965            md->capture_last = save_capture_last;
966            ecode += GET(ecode, 1);
967            if (*ecode != OP_ALT) break;
968            }
969    
970          if (!matched_once)
971            {
972            md->offset_vector[offset] = save_offset1;
973            md->offset_vector[offset+1] = save_offset2;
974            md->offset_vector[md->offset_end - number] = save_offset3;
975            }
976    
977          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
978          if (allow_zero || matched_once)
979            {
980            ecode += 1 + LINK_SIZE;
981            break;
982            }
983    
984          RRETURN(MATCH_NOMATCH);
985          }
986    
987        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
988        as a non-capturing bracket. */
989    
990        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
991        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
992    
993        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
994    
995        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
996        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
997    
998        /* Non-capturing possessive bracket with unlimited repeat. We come here
999        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1000        without the capturing complication. It is written out separately for speed
1001        and cleanliness. */
1002    
1003        case OP_BRAPOS:
1004        case OP_SBRAPOS:
1005        allow_zero = FALSE;
1006    
1007        POSSESSIVE_NON_CAPTURE:
1008        matched_once = FALSE;
1009        code_offset = ecode - md->start_code;
1010    
1011        for (;;)
1012          {
1013          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1014          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1015            eptrb, RM48);
1016          if (rrc == MATCH_KETRPOS)
1017            {
1018            offset_top = md->end_offset_top;
1019            eptr = md->end_match_ptr;
1020            ecode = md->start_code + code_offset;
1021            matched_once = TRUE;
1022            continue;
1023            }
1024          if (rrc != MATCH_NOMATCH &&
1025              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1026            RRETURN(rrc);
1027        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1028          if (*ecode != OP_ALT) break;
1029        }        }
1030      while (*ecode == OP_ALT);  
1031      DPRINTF(("bracket 0 failed\n"));      if (matched_once || allow_zero)
1032          {
1033          ecode += 1 + LINK_SIZE;
1034          break;
1035          }
1036      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1037    
1038        /* Control never reaches here. */
1039    
1040      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1041      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1042      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1043      exactly what going to the ket would do. */      exactly what going to the ket would do. */
1044    
1045      case OP_COND:      case OP_COND:
1046      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      case OP_SCOND:
1047        codelink = GET(ecode, 1);
1048    
1049        /* Because of the way auto-callout works during compile, a callout item is
1050        inserted between OP_COND and an assertion condition. */
1051    
1052        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1053          {
1054          if (pcre_callout != NULL)
1055            {
1056            pcre_callout_block cb;
1057            cb.version          = 1;   /* Version 1 of the callout block */
1058            cb.callout_number   = ecode[LINK_SIZE+2];
1059            cb.offset_vector    = md->offset_vector;
1060            cb.subject          = (PCRE_SPTR)md->start_subject;
1061            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1062            cb.start_match      = (int)(mstart - md->start_subject);
1063            cb.current_position = (int)(eptr - md->start_subject);
1064            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1065            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1066            cb.capture_top      = offset_top/2;
1067            cb.capture_last     = md->capture_last;
1068            cb.callout_data     = md->callout_data;
1069            if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1070            if (rrc < 0) RRETURN(rrc);
1071            }
1072          ecode += _pcre_OP_lengths[OP_CALLOUT];
1073          }
1074    
1075        condcode = ecode[LINK_SIZE+1];
1076    
1077        /* Now see what the actual condition is */
1078    
1079        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1080          {
1081          if (md->recursive == NULL)                /* Not recursing => FALSE */
1082            {
1083            condition = FALSE;
1084            ecode += GET(ecode, 1);
1085            }
1086          else
1087            {
1088            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1089            condition =  (recno == RREF_ANY || recno == md->recursive->group_num);
1090    
1091            /* If the test is for recursion into a specific subpattern, and it is
1092            false, but the test was set up by name, scan the table to see if the
1093            name refers to any other numbers, and test them. The condition is true
1094            if any one is set. */
1095    
1096            if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
1097              {
1098              uschar *slotA = md->name_table;
1099              for (i = 0; i < md->name_count; i++)
1100                {
1101                if (GET2(slotA, 0) == recno) break;
1102                slotA += md->name_entry_size;
1103                }
1104    
1105              /* Found a name for the number - there can be only one; duplicate
1106              names for different numbers are allowed, but not vice versa. First
1107              scan down for duplicates. */
1108    
1109              if (i < md->name_count)
1110                {
1111                uschar *slotB = slotA;
1112                while (slotB > md->name_table)
1113                  {
1114                  slotB -= md->name_entry_size;
1115                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1116                    {
1117                    condition = GET2(slotB, 0) == md->recursive->group_num;
1118                    if (condition) break;
1119                    }
1120                  else break;
1121                  }
1122    
1123                /* Scan up for duplicates */
1124    
1125                if (!condition)
1126                  {
1127                  slotB = slotA;
1128                  for (i++; i < md->name_count; i++)
1129                    {
1130                    slotB += md->name_entry_size;
1131                    if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1132                      {
1133                      condition = GET2(slotB, 0) == md->recursive->group_num;
1134                      if (condition) break;
1135                      }
1136                    else break;
1137                    }
1138                  }
1139                }
1140              }
1141    
1142            /* Chose branch according to the condition */
1143    
1144            ecode += condition? 3 : GET(ecode, 1);
1145            }
1146          }
1147    
1148        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1149        {        {
1150        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1151        condition = (offset == CREF_RECURSE * 2)?        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1152          (md->recursive != NULL) :  
1153          (offset < offset_top && md->offset_vector[offset] >= 0);        /* If the numbered capture is unset, but the reference was by name,
1154        RMATCH(rrc, eptr, ecode + (condition?        scan the table to see if the name refers to any other numbers, and test
1155          (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1))),        them. The condition is true if any one is set. This is tediously similar
1156          offset_top, md, ims, eptrb, match_isgroup);        to the code above, but not close enough to try to amalgamate. */
1157        RRETURN(rrc);  
1158          if (!condition && condcode == OP_NCREF)
1159            {
1160            int refno = offset >> 1;
1161            uschar *slotA = md->name_table;
1162    
1163            for (i = 0; i < md->name_count; i++)
1164              {
1165              if (GET2(slotA, 0) == refno) break;
1166              slotA += md->name_entry_size;
1167              }
1168    
1169            /* Found a name for the number - there can be only one; duplicate names
1170            for different numbers are allowed, but not vice versa. First scan down
1171            for duplicates. */
1172    
1173            if (i < md->name_count)
1174              {
1175              uschar *slotB = slotA;
1176              while (slotB > md->name_table)
1177                {
1178                slotB -= md->name_entry_size;
1179                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1180                  {
1181                  offset = GET2(slotB, 0) << 1;
1182                  condition = offset < offset_top &&
1183                    md->offset_vector[offset] >= 0;
1184                  if (condition) break;
1185                  }
1186                else break;
1187                }
1188    
1189              /* Scan up for duplicates */
1190    
1191              if (!condition)
1192                {
1193                slotB = slotA;
1194                for (i++; i < md->name_count; i++)
1195                  {
1196                  slotB += md->name_entry_size;
1197                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1198                    {
1199                    offset = GET2(slotB, 0) << 1;
1200                    condition = offset < offset_top &&
1201                      md->offset_vector[offset] >= 0;
1202                    if (condition) break;
1203                    }
1204                  else break;
1205                  }
1206                }
1207              }
1208            }
1209    
1210          /* Chose branch according to the condition */
1211    
1212          ecode += condition? 3 : GET(ecode, 1);
1213          }
1214    
1215        else if (condcode == OP_DEF)     /* DEFINE - always false */
1216          {
1217          condition = FALSE;
1218          ecode += GET(ecode, 1);
1219        }        }
1220    
1221      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1222      the final argument TRUE causes it to stop at the end of an assertion. */      md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1223        an assertion. */
1224    
1225      else      else
1226        {        {
1227        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1228            match_condassert | match_isgroup);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1229        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1230          {          {
1231          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          condition = TRUE;
1232            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1233          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1234          }          }
1235        else if (rrc != MATCH_NOMATCH)        else if (rrc != MATCH_NOMATCH &&
1236                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1237          {          {
1238          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1239          }          }
1240        else ecode += GET(ecode, 1);        else
1241        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          {
1242          match_isgroup);          condition = FALSE;
1243            ecode += codelink;
1244            }
1245          }
1246    
1247        /* We are now at the branch that is to be obeyed. As there is only one,
1248        we used to use tail recursion to avoid using another stack frame, except
1249        when there was unlimited repeat of a possibly empty group. However, that
1250        strategy no longer works because of the possibilty of (*THEN) being
1251        encountered in the branch. A recursive call to match() is always required,
1252        unless the second alternative doesn't exist, in which case we can just
1253        plough on. */
1254    
1255        if (condition || *ecode == OP_ALT)
1256          {
1257          if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1258          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1259          if (rrc == MATCH_THEN && md->start_match_ptr == ecode)
1260            rrc = MATCH_NOMATCH;
1261        RRETURN(rrc);        RRETURN(rrc);
1262        }        }
1263      /* Control never reaches here */      else                         /* Condition false & no alternative */
1264          {
1265          ecode += 1 + LINK_SIZE;
1266          }
1267        break;
1268    
     /* Skip over conditional reference or large extraction number data if  
     encountered. */  
1269    
1270      case OP_CREF:      /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1271      case OP_BRANUMBER:      to close any currently open capturing brackets. */
1272    
1273        case OP_CLOSE:
1274        number = GET2(ecode, 1);
1275        offset = number << 1;
1276    
1277    #ifdef PCRE_DEBUG
1278          printf("end bracket %d at *ACCEPT", number);
1279          printf("\n");
1280    #endif
1281    
1282        md->capture_last = number;
1283        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1284          {
1285          md->offset_vector[offset] =
1286            md->offset_vector[md->offset_end - number];
1287          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1288          if (offset_top <= offset) offset_top = offset + 2;
1289          }
1290      ecode += 3;      ecode += 3;
1291      break;      break;
1292    
     /* End of the pattern. If we are in a recursion, we should restore the  
     offsets appropriately and continue from after the call. */  
1293    
1294        /* End of the pattern, either real or forced. If we are in a recursion, we
1295        should restore the offsets appropriately, and if it's a top-level
1296        recursion, continue from after the call. */
1297    
1298        case OP_ACCEPT:
1299        case OP_ASSERT_ACCEPT:
1300      case OP_END:      case OP_END:
1301      if (md->recursive != NULL && md->recursive->group_num == 0)      if (md->recursive != NULL)
1302        {        {
1303        recursion_info *rec = md->recursive;        recursion_info *rec = md->recursive;
       DPRINTF(("Hit the end in a (?0) recursion\n"));  
1304        md->recursive = rec->prevrec;        md->recursive = rec->prevrec;
1305        memmove(md->offset_vector, rec->offset_save,        memmove(md->offset_vector, rec->offset_save,
1306          rec->saved_max * sizeof(int));          rec->saved_max * sizeof(int));
1307        md->start_match = rec->save_start;        offset_top = rec->save_offset_top;
1308        ims = original_ims;        if (rec->group_num == 0)
1309        ecode = rec->after_call;          {
1310        break;          ecode = rec->after_call;
1311            break;
1312            }
1313        }        }
1314    
1315      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Otherwise, if we have matched an empty string, fail if not in an
1316      string - backtracking will then try other alternatives, if any. */      assertion and if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1317        is set and we have matched at the start of the subject. In both cases,
1318        backtracking will then try other alternatives, if any. */
1319    
1320        else if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1321            (md->notempty ||
1322              (md->notempty_atstart &&
1323                mstart == md->start_subject + md->start_offset)))
1324          MRRETURN(MATCH_NOMATCH);
1325    
1326        /* Otherwise, we have a match. */
1327    
1328        md->end_match_ptr = eptr;           /* Record where we ended */
1329        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1330        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1331    
1332      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* For some reason, the macros don't work properly if an expression is
1333      md->end_match_ptr = eptr;          /* Record where we ended */      given as the argument to MRRETURN when the heap is in use. */
1334      md->end_offset_top = offset_top;   /* and how many extracts were taken */  
1335      RRETURN(MATCH_MATCH);      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1336        MRRETURN(rrc);
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1337    
1338      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1339      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1340      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1341      start of each branch to move the current point backwards, so the code at      start of each branch to move the current point backwards, so the code at
1342      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1343        of a condition, we want to return immediately afterwards. The caller of
1344        this incarnation of the match() function will have set MATCH_CONDASSERT in
1345        md->match_function type, and one of these opcodes will be the first opcode
1346        that is processed. We use a local variable that is preserved over calls to
1347        match() to remember this case. */
1348    
1349      case OP_ASSERT:      case OP_ASSERT:
1350      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1351        if (md->match_function_type == MATCH_CONDASSERT)
1352          {
1353          condassert = TRUE;
1354          md->match_function_type = 0;
1355          }
1356        else condassert = FALSE;
1357    
1358      do      do
1359        {        {
1360        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1361          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1362        if (rrc == MATCH_MATCH) break;          {
1363        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          mstart = md->start_match_ptr;   /* In case \K reset it */
1364            break;
1365            }
1366          if (rrc != MATCH_NOMATCH &&
1367              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1368            RRETURN(rrc);
1369        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1370        }        }
1371      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1372      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1373        if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1374    
1375      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1376    
1377      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1378    
1379      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1380      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 738  for (;;) Line 1384  for (;;)
1384      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1385      continue;      continue;
1386    
1387      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1388        PRUNE, or COMMIT means we must assume failure without checking subsequent
1389        branches. */
1390    
1391      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1392      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1393        if (md->match_function_type == MATCH_CONDASSERT)
1394          {
1395          condassert = TRUE;
1396          md->match_function_type = 0;
1397          }
1398        else condassert = FALSE;
1399    
1400      do      do
1401        {        {
1402        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1403          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1404        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1405        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          {
1406            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1407            break;
1408            }
1409          if (rrc != MATCH_NOMATCH &&
1410              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1411            RRETURN(rrc);
1412        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1413        }        }
1414      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1415    
1416      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1417    
1418      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1419      continue;      continue;
1420    
# Line 766  for (;;) Line 1427  for (;;)
1427  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1428      if (utf8)      if (utf8)
1429        {        {
1430        c = GET(ecode,1);        i = GET(ecode, 1);
1431        for (i = 0; i < c; i++)        while (i-- > 0)
1432          {          {
1433          eptr--;          eptr--;
1434          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1435          BACKCHAR(eptr)          BACKCHAR(eptr);
1436          }          }
1437        }        }
1438      else      else
# Line 780  for (;;) Line 1441  for (;;)
1441      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1442    
1443        {        {
1444        eptr -= GET(ecode,1);        eptr -= GET(ecode, 1);
1445        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1446        }        }
1447    
1448      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1449    
1450        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1451      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1452      break;      break;
1453    
# Line 800  for (;;) Line 1462  for (;;)
1462        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 1;   /* Version 1 of the callout block */
1463        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1464        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1465        cb.subject          = (const char *)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1466        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1467        cb.start_match      = md->start_match - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1468        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1469        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1470        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1471        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1472        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1473        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1474        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1475        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1476        }        }
1477      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 837  for (;;) Line 1499  for (;;)
1499      case OP_RECURSE:      case OP_RECURSE:
1500        {        {
1501        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1502        new_recursive.group_num = *callpat - OP_BRA;        new_recursive.group_num = (callpat == md->start_code)? 0 :
1503            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);  
1504    
1505        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1506    
# Line 869  for (;;) Line 1526  for (;;)
1526    
1527        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1528              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1529        new_recursive.save_start = md->start_match;        new_recursive.save_offset_top = offset_top;
1530        md->start_match = eptr;  
   
1531        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. For each top-level alternative we
1532        restore the offset and recursion data. */        restore the offset and recursion data. */
1533    
1534        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1535          cbegroup = (*callpat >= OP_SBRA);
1536        do        do
1537          {          {
1538          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1539              eptrb, match_isgroup);          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1540          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1541            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1542            {            {
1543              DPRINTF(("Recursion matched\n"));
1544            md->recursive = new_recursive.prevrec;            md->recursive = new_recursive.prevrec;
1545            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1546              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1547            RRETURN(MATCH_MATCH);            MRRETURN(MATCH_MATCH);
1548              }
1549            else if (rrc != MATCH_NOMATCH &&
1550                    (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1551              {
1552              DPRINTF(("Recursion gave error %d\n", rrc));
1553              if (new_recursive.offset_save != stacksave)
1554                (pcre_free)(new_recursive.offset_save);
1555              RRETURN(rrc);
1556            }            }
         else if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1557    
1558          md->recursive = &new_recursive;          md->recursive = &new_recursive;
1559          memcpy(md->offset_vector, new_recursive.offset_save,          memcpy(md->offset_vector, new_recursive.offset_save,
# Line 900  for (;;) Line 1566  for (;;)
1566        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1567        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1568          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1569        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1570        }        }
1571      /* Control never reaches here */      /* Control never reaches here */
1572    
# Line 909  for (;;) Line 1575  for (;;)
1575      a move back into the brackets. Friedl calls these "atomic" subpatterns.      a move back into the brackets. Friedl calls these "atomic" subpatterns.
1576      Check the alternative branches in turn - the matching won't pass the KET      Check the alternative branches in turn - the matching won't pass the KET
1577      for this kind of subpattern. If any one branch matches, we carry on as at      for this kind of subpattern. If any one branch matches, we carry on as at
1578      the end of a normal bracket, leaving the subject pointer. */      the end of a normal bracket, leaving the subject pointer, but resetting
1579        the start-of-match value in case it was changed by \K. */
1580    
1581      case OP_ONCE:      case OP_ONCE:
1582        {      prev = ecode;
1583        prev = ecode;      saved_eptr = eptr;
       saved_eptr = eptr;  
1584    
1585        do      do
1586          {
1587          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1588          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
1589          {          {
1590          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,          mstart = md->start_match_ptr;
1591            eptrb, match_isgroup);          break;
         if (rrc == MATCH_MATCH) break;  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         ecode += GET(ecode,1);  
1592          }          }
1593        while (*ecode == OP_ALT);        if (rrc != MATCH_NOMATCH &&
1594              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1595        /* If hit the end of the group (which could be repeated), fail */          RRETURN(rrc);
1596          ecode += GET(ecode,1);
1597          }
1598        while (*ecode == OP_ALT);
1599    
1600        if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);      /* If hit the end of the group (which could be repeated), fail */
1601    
1602        /* Continue as from after the assertion, updating the offsets high water      if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
       mark, since extracts may have been taken. */  
1603    
1604        do ecode += GET(ecode,1); while (*ecode == OP_ALT);      /* Continue after the group, updating the offsets high water mark, since
1605        extracts may have been taken. */
1606    
1607        offset_top = md->end_offset_top;      do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
       eptr = md->end_match_ptr;  
1608    
1609        /* For a non-repeating ket, just continue at this level. This also      offset_top = md->end_offset_top;
1610        happens for a repeating ket if no characters were matched in the group.      eptr = md->end_match_ptr;
       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. */  
1611    
1612        if (*ecode == OP_KET || eptr == saved_eptr)      /* For a non-repeating ket, just continue at this level. This also
1613          {      happens for a repeating ket if no characters were matched in the group.
1614          ecode += 1+LINK_SIZE;      This is the forcible breaking of infinite loops as implemented in Perl
1615          break;      5.005. If there is an options reset, it will get obeyed in the normal
1616          }      course of events. */
1617    
1618        /* The repeating kets try the rest of the pattern or restart from the      if (*ecode == OP_KET || eptr == saved_eptr)
1619        preceding bracket, in the appropriate order. We need to reset any options        {
1620        that changed within the bracket before re-running it, so check the next        ecode += 1+LINK_SIZE;
1621        opcode. */        break;
1622          }
1623    
1624        if (ecode[1+LINK_SIZE] == OP_OPT)      /* The repeating kets try the rest of the pattern or restart from the
1625          {      preceding bracket, in the appropriate order. The second "call" of match()
1626          ims = (ims & ~PCRE_IMS) | ecode[4];      uses tail recursion, to avoid using another stack frame. */
         DPRINTF(("ims set to %02lx at group repeat\n", ims));  
         }  
1627    
1628        if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1629          {        {
1630          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM8);
1631          if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1632          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        ecode = prev;
1633          if (rrc != MATCH_NOMATCH) RRETURN(rrc);        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);  
         }  
1634        }        }
1635      RRETURN(MATCH_NOMATCH);      else  /* OP_KETRMAX */
1636          {
1637          md->match_function_type = MATCH_CBEGROUP;
1638          RMATCH(eptr, prev, offset_top, md, eptrb, RM9);
1639          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1640          ecode += 1 + LINK_SIZE;
1641          goto TAIL_RECURSE;
1642          }
1643        /* Control never gets here */
1644    
1645      /* An alternation is the end of a branch; scan along to find the end of the      /* An alternation is the end of a branch; scan along to find the end of the
1646      bracketed group and go to there. */      bracketed group and go to there. */
# Line 985  for (;;) Line 1649  for (;;)
1649      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1650      break;      break;
1651    
1652      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1653      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
1654      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
1655      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
1656      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1657    
1658      case OP_BRAZERO:      case OP_BRAZERO:
1659        {      next = ecode + 1;
1660        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1661        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1662        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1663        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1664      break;      break;
1665    
1666      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1667        {      next = ecode + 1;
1668        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1669        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1670        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1671          match_isgroup);      ecode++;
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode++;  
       }  
1672      break;      break;
1673    
1674      /* End of a group, repeated or non-repeating. If we are at the end of      case OP_SKIPZERO:
1675      an assertion "group", stop matching and return MATCH_MATCH, but record the      next = ecode+1;
1676      current high water mark for use by positive assertions. Do this also      do next += GET(next,1); while (*next == OP_ALT);
1677      for the "once" (not-backup up) groups. */      ecode = next + 1 + LINK_SIZE;
1678        break;
1679    
1680        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1681        here; just jump to the group, with allow_zero set TRUE. */
1682    
1683        case OP_BRAPOSZERO:
1684        op = *(++ecode);
1685        allow_zero = TRUE;
1686        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1687          goto POSSESSIVE_NON_CAPTURE;
1688    
1689        /* End of a group, repeated or non-repeating. */
1690    
1691      case OP_KET:      case OP_KET:
1692      case OP_KETRMIN:      case OP_KETRMIN:
1693      case OP_KETRMAX:      case OP_KETRMAX:
1694        {      case OP_KETRPOS:
1695        prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
       saved_eptr = eptrb->epb_saved_eptr;  
   
       /* Back up the stack of bracket start pointers. */  
1696    
1697        eptrb = eptrb->epb_prev;      /* If this was a group that remembered the subject start, in order to break
1698        infinite repeats of empty string matches, retrieve the subject start from
1699        the chain. Otherwise, set it NULL. */
1700    
1701        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev >= OP_SBRA)
1702            *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||        {
1703            *prev == OP_ONCE)        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1704          {        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1705          md->end_match_ptr = eptr;      /* For ONCE */        }
1706          md->end_offset_top = offset_top;      else saved_eptr = NULL;
         RRETURN(MATCH_MATCH);  
         }  
1707    
1708        /* In all other cases except a conditional group we have to check the      /* If we are at the end of an assertion group or an atomic group, stop
1709        group number back at the start and if necessary complete handling an      matching and return MATCH_MATCH, but record the current high water mark for
1710        extraction by setting the offsets and bumping the high water mark. */      use by positive assertions. We also need to record the match start in case
1711        it was changed by \K. */
1712    
1713        if (*prev != OP_COND)      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1714          {          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
1715          number = *prev - OP_BRA;          *prev == OP_ONCE)
1716          {
1717          md->end_match_ptr = eptr;      /* For ONCE */
1718          md->end_offset_top = offset_top;
1719          md->start_match_ptr = mstart;
1720          MRRETURN(MATCH_MATCH);
1721          }
1722    
1723          /* 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
1724          the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1725        bumping the high water mark. Note that whole-pattern recursion is coded as
1726        a recurse into group 0, so it won't be picked up here. Instead, we catch it
1727        when the OP_END is reached. Other recursion is handled here. */
1728    
1729          if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1730          offset = number << 1;          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1731          {
1732          number = GET2(prev, 1+LINK_SIZE);
1733          offset = number << 1;
1734    
1735  #ifdef DEBUG  #ifdef PCRE_DEBUG
1736          printf("end bracket %d", number);        printf("end bracket %d", number);
1737          printf("\n");        printf("\n");
1738  #endif  #endif
1739    
1740          /* Test for a numbered group. This includes groups called as a result        md->capture_last = number;
1741          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. */  
   
         if (number > 0)  
           {  
           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;  
             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  
       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)  
1742          {          {
1743          ecode += 1 + LINK_SIZE;          /* If offset is greater than offset_top, it means that we are
1744            "skipping" a capturing group, and that group's offsets must be marked
1745            unset. In earlier versions of PCRE, all the offsets were unset at the
1746            start of matching, but this doesn't work because atomic groups and
1747            assertions can cause a value to be set that should later be unset.
1748            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1749            part of the atomic group, but this is not on the final matching path,
1750            so must be unset when 2 is set. (If there is no group 2, there is no
1751            problem, because offset_top will then be 2, indicating no capture.) */
1752    
1753            if (offset > offset_top)
1754              {
1755              register int *iptr = md->offset_vector + offset_top;
1756              register int *iend = md->offset_vector + offset;
1757              while (iptr < iend) *iptr++ = -1;
1758              }
1759    
1760            /* Now make the extraction */
1761    
1762            md->offset_vector[offset] =
1763              md->offset_vector[md->offset_end - number];
1764            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1765            if (offset_top <= offset) offset_top = offset + 2;
1766            }
1767    
1768          /* Handle a recursively called group. Restore the offsets
1769          appropriately and continue from after the call. */
1770    
1771          if (md->recursive != NULL && md->recursive->group_num == number)
1772            {
1773            recursion_info *rec = md->recursive;
1774            DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
1775            md->recursive = rec->prevrec;
1776            memcpy(md->offset_vector, rec->offset_save,
1777              rec->saved_max * sizeof(int));
1778            offset_top = rec->save_offset_top;
1779            ecode = rec->after_call;
1780          break;          break;
1781          }          }
1782          }
1783    
1784        /* The repeating kets try the rest of the pattern or restart from the      /* For a non-repeating ket, just continue at this level. This also
1785        preceding bracket, in the appropriate order. */      happens for a repeating ket if no characters were matched in the group.
1786        This is the forcible breaking of infinite loops as implemented in Perl
1787        5.005. If there is an options reset, it will get obeyed in the normal
1788        course of events. */
1789    
1790        if (*ecode == OP_KETRMIN)      if (*ecode == OP_KET || eptr == saved_eptr)
1791          {        {
1792          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        ecode += 1 + LINK_SIZE;
1793          if (rrc != MATCH_NOMATCH) RRETURN(rrc);        break;
1794          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        }
1795          if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1796          }      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1797        else  /* OP_KETRMAX */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1798        at a time from the outer level, thus saving stack. */
1799    
1800        if (*ecode == OP_KETRPOS)
1801          {
1802          md->end_match_ptr = eptr;
1803          md->end_offset_top = offset_top;
1804          RRETURN(MATCH_KETRPOS);
1805          }
1806    
1807        /* The normal repeating kets try the rest of the pattern or restart from
1808        the preceding bracket, in the appropriate order. In the second case, we can
1809        use tail recursion to avoid using another stack frame, unless we have an
1810        unlimited repeat of a group that can match an empty string. */
1811    
1812        if (*ecode == OP_KETRMIN)
1813          {
1814          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1815          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1816          if (*prev >= OP_SBRA)    /* Could match an empty string */
1817          {          {
1818          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          md->match_function_type = MATCH_CBEGROUP;
1819          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1820          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          RRETURN(rrc);
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1821          }          }
1822          ecode = prev;
1823          goto TAIL_RECURSE;
1824        }        }
1825        else  /* OP_KETRMAX */
     RRETURN(MATCH_NOMATCH);  
   
     /* 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)  
1826        {        {
1827        if (eptr != md->start_subject && eptr[-1] != NEWLINE)        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1828          RRETURN(MATCH_NOMATCH);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1829        ecode++;        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1830        break;        ecode += 1 + LINK_SIZE;
1831          goto TAIL_RECURSE;
1832        }        }
1833      /* ... else fall through */      /* Control never gets here */
1834    
1835        /* Not multiline mode: start of subject assertion, unless notbol. */
1836    
1837        case OP_CIRC:
1838        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1839    
1840      /* Start of subject assertion */      /* Start of subject assertion */
1841    
1842      case OP_SOD:      case OP_SOD:
1843      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1844        ecode++;
1845        break;
1846    
1847        /* Multiline mode: start of subject unless notbol, or after any newline. */
1848    
1849        case OP_CIRCM:
1850        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1851        if (eptr != md->start_subject &&
1852            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1853          MRRETURN(MATCH_NOMATCH);
1854      ecode++;      ecode++;
1855      break;      break;
1856    
1857      /* Start of match assertion */      /* Start of match assertion */
1858    
1859      case OP_SOM:      case OP_SOM:
1860      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
1861      ecode++;      ecode++;
1862      break;      break;
1863    
1864      /* Assert before internal newline if multiline, or before a terminating      /* Reset the start of match point */
     newline unless endonly is set, else end of subject unless noteol is set. */  
1865    
1866      case OP_DOLL:      case OP_SET_SOM:
1867      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
1868        {      ecode++;
1869        if (eptr < md->end_subject)      break;
1870          { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }  
1871        else      /* Multiline mode: assert before any newline, or before end of subject
1872          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
1873        ecode++;  
1874        break;      case OP_DOLLM:
1875        }      if (eptr < md->end_subject)
1876          { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
1877      else      else
1878        {        {
1879        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1880        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr < md->end_subject - 1 ||  
            (eptr == md->end_subject - 1 && *eptr != NEWLINE))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
1881        }        }
1882      /* ... else fall through */      ecode++;
1883        break;
1884    
1885        /* Not multiline mode: assert before a terminating newline or before end of
1886        subject unless noteol is set. */
1887    
1888        case OP_DOLL:
1889        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1890        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1891    
1892        /* ... else fall through for endonly */
1893    
1894      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
1895    
1896      case OP_EOD:      case OP_EOD:
1897      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
1898        SCHECK_PARTIAL();
1899      ecode++;      ecode++;
1900      break;      break;
1901    
1902      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
1903    
1904      case OP_EODN:      case OP_EODN:
1905      if (eptr < md->end_subject - 1 ||      ASSERT_NL_OR_EOS:
1906         (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject &&
1907            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1908          MRRETURN(MATCH_NOMATCH);
1909    
1910        /* Either at end of string or \n before end. */
1911    
1912        SCHECK_PARTIAL();
1913      ecode++;      ecode++;
1914      break;      break;
1915    
# Line 1206  for (;;) Line 1921  for (;;)
1921    
1922        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
1923        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
1924        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
1925          partial matching. */
1926    
1927  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1928        if (utf8)        if (utf8)
1929          {          {
1930            /* Get status of previous character */
1931    
1932          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
1933            {            {
1934            const uschar *lastptr = eptr - 1;            USPTR lastptr = eptr - 1;
1935            while((*lastptr & 0xc0) == 0x80) lastptr--;            while((*lastptr & 0xc0) == 0x80) lastptr--;
1936              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
1937            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
1938    #ifdef SUPPORT_UCP
1939              if (md->use_ucp)
1940                {
1941                if (c == '_') prev_is_word = TRUE; else
1942                  {
1943                  int cat = UCD_CATEGORY(c);
1944                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1945                  }
1946                }
1947              else
1948    #endif
1949            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1950            }            }
1951          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
1952            /* Get status of next character */
1953    
1954            if (eptr >= md->end_subject)
1955              {
1956              SCHECK_PARTIAL();
1957              cur_is_word = FALSE;
1958              }
1959            else
1960            {            {
1961            GETCHAR(c, eptr);            GETCHAR(c, eptr);
1962    #ifdef SUPPORT_UCP
1963              if (md->use_ucp)
1964                {
1965                if (c == '_') cur_is_word = TRUE; else
1966                  {
1967                  int cat = UCD_CATEGORY(c);
1968                  cur_is_word = (cat == ucp_L || cat == ucp_N);
1969                  }
1970                }
1971              else
1972    #endif
1973            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1974            }            }
1975          }          }
1976        else        else
1977  #endif  #endif
1978    
1979        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
1980          consistency with the behaviour of \w we do use it in this case. */
1981    
1982          {          {
1983          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
1984            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
1985          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
1986            ((md->ctypes[*eptr] & ctype_word) != 0);            {
1987              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
1988    #ifdef SUPPORT_UCP
1989              if (md->use_ucp)
1990                {
1991                c = eptr[-1];
1992                if (c == '_') prev_is_word = TRUE; else
1993                  {
1994                  int cat = UCD_CATEGORY(c);
1995                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1996                  }
1997                }
1998              else
1999    #endif
2000              prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2001              }
2002    
2003            /* Get status of next character */
2004    
2005            if (eptr >= md->end_subject)
2006              {
2007              SCHECK_PARTIAL();
2008              cur_is_word = FALSE;
2009              }
2010            else
2011    #ifdef SUPPORT_UCP
2012            if (md->use_ucp)
2013              {
2014              c = *eptr;
2015              if (c == '_') cur_is_word = TRUE; else
2016                {
2017                int cat = UCD_CATEGORY(c);
2018                cur_is_word = (cat == ucp_L || cat == ucp_N);
2019                }
2020              }
2021            else
2022    #endif
2023            cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
2024          }          }
2025    
2026        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
2027    
2028        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
2029             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2030          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2031        }        }
2032      break;      break;
2033    
2034      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2035    
2036      case OP_ANY:      case OP_ANY:
2037      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
2038        RRETURN(MATCH_NOMATCH);      /* Fall through */
2039      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);  
2040  #ifdef SUPPORT_UTF8      case OP_ALLANY:
2041      if (utf8)      if (eptr++ >= md->end_subject)
2042        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;        {
2043  #endif        SCHECK_PARTIAL();
2044          MRRETURN(MATCH_NOMATCH);
2045          }
2046        if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2047      ecode++;      ecode++;
2048      break;      break;
2049    
# Line 1261  for (;;) Line 2051  for (;;)
2051      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2052    
2053      case OP_ANYBYTE:      case OP_ANYBYTE:
2054      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr++ >= md->end_subject)
2055          {
2056          SCHECK_PARTIAL();
2057          MRRETURN(MATCH_NOMATCH);
2058          }
2059      ecode++;      ecode++;
2060      break;      break;
2061    
2062      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2063      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2064          {
2065          SCHECK_PARTIAL();
2066          MRRETURN(MATCH_NOMATCH);
2067          }
2068      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2069      if (      if (
2070  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1274  for (;;) Line 2072  for (;;)
2072  #endif  #endif
2073         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2074         )         )
2075        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2076      ecode++;      ecode++;
2077      break;      break;
2078    
2079      case OP_DIGIT:      case OP_DIGIT:
2080      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2081          {
2082          SCHECK_PARTIAL();
2083          MRRETURN(MATCH_NOMATCH);
2084          }
2085      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2086      if (      if (
2087  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1287  for (;;) Line 2089  for (;;)
2089  #endif  #endif
2090         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2091         )         )
2092        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2093      ecode++;      ecode++;
2094      break;      break;
2095    
2096      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2097      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2098          {
2099          SCHECK_PARTIAL();
2100          MRRETURN(MATCH_NOMATCH);
2101          }
2102      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2103      if (      if (
2104  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1300  for (;;) Line 2106  for (;;)
2106  #endif  #endif
2107         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2108         )         )
2109        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2110      ecode++;      ecode++;
2111      break;      break;
2112    
2113      case OP_WHITESPACE:      case OP_WHITESPACE:
2114      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2115          {
2116          SCHECK_PARTIAL();
2117          MRRETURN(MATCH_NOMATCH);
2118          }
2119      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2120      if (      if (
2121  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1313  for (;;) Line 2123  for (;;)
2123  #endif  #endif
2124         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2125         )         )
2126        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2127      ecode++;      ecode++;
2128      break;      break;
2129    
2130      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2131      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2132          {
2133          SCHECK_PARTIAL();
2134          MRRETURN(MATCH_NOMATCH);
2135          }
2136      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2137      if (      if (
2138  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1326  for (;;) Line 2140  for (;;)
2140  #endif  #endif
2141         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2142         )         )
2143        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2144      ecode++;      ecode++;
2145      break;      break;
2146    
2147      case OP_WORDCHAR:      case OP_WORDCHAR:
2148      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2149          {
2150          SCHECK_PARTIAL();
2151          MRRETURN(MATCH_NOMATCH);
2152          }
2153      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2154      if (      if (
2155  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1339  for (;;) Line 2157  for (;;)
2157  #endif  #endif
2158         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2159         )         )
2160        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2161        ecode++;
2162        break;
2163    
2164        case OP_ANYNL:
2165        if (eptr >= md->end_subject)
2166          {
2167          SCHECK_PARTIAL();
2168          MRRETURN(MATCH_NOMATCH);
2169          }
2170        GETCHARINCTEST(c, eptr);
2171        switch(c)
2172          {
2173          default: MRRETURN(MATCH_NOMATCH);
2174    
2175          case 0x000d:
2176          if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2177          break;
2178    
2179          case 0x000a:
2180          break;
2181    
2182          case 0x000b:
2183          case 0x000c:
2184          case 0x0085:
2185          case 0x2028:
2186          case 0x2029:
2187          if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
2188          break;
2189          }
2190        ecode++;
2191        break;
2192    
2193        case OP_NOT_HSPACE:
2194        if (eptr >= md->end_subject)
2195          {
2196          SCHECK_PARTIAL();
2197          MRRETURN(MATCH_NOMATCH);
2198          }
2199        GETCHARINCTEST(c, eptr);
2200        switch(c)
2201          {
2202          default: break;
2203          case 0x09:      /* HT */
2204          case 0x20:      /* SPACE */
2205          case 0xa0:      /* NBSP */
2206          case 0x1680:    /* OGHAM SPACE MARK */
2207          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2208          case 0x2000:    /* EN QUAD */
2209          case 0x2001:    /* EM QUAD */
2210          case 0x2002:    /* EN SPACE */
2211          case 0x2003:    /* EM SPACE */
2212          case 0x2004:    /* THREE-PER-EM SPACE */
2213          case 0x2005:    /* FOUR-PER-EM SPACE */
2214          case 0x2006:    /* SIX-PER-EM SPACE */
2215          case 0x2007:    /* FIGURE SPACE */
2216          case 0x2008:    /* PUNCTUATION SPACE */
2217          case 0x2009:    /* THIN SPACE */
2218          case 0x200A:    /* HAIR SPACE */
2219          case 0x202f:    /* NARROW NO-BREAK SPACE */
2220          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2221          case 0x3000:    /* IDEOGRAPHIC SPACE */
2222          MRRETURN(MATCH_NOMATCH);
2223          }
2224        ecode++;
2225        break;
2226    
2227        case OP_HSPACE:
2228        if (eptr >= md->end_subject)
2229          {
2230          SCHECK_PARTIAL();
2231          MRRETURN(MATCH_NOMATCH);
2232          }
2233        GETCHARINCTEST(c, eptr);
2234        switch(c)
2235          {
2236          default: MRRETURN(MATCH_NOMATCH);
2237          case 0x09:      /* HT */
2238          case 0x20:      /* SPACE */
2239          case 0xa0:      /* NBSP */
2240          case 0x1680:    /* OGHAM SPACE MARK */
2241          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2242          case 0x2000:    /* EN QUAD */
2243          case 0x2001:    /* EM QUAD */
2244          case 0x2002:    /* EN SPACE */
2245          case 0x2003:    /* EM SPACE */
2246          case 0x2004:    /* THREE-PER-EM SPACE */
2247          case 0x2005:    /* FOUR-PER-EM SPACE */
2248          case 0x2006:    /* SIX-PER-EM SPACE */
2249          case 0x2007:    /* FIGURE SPACE */
2250          case 0x2008:    /* PUNCTUATION SPACE */
2251          case 0x2009:    /* THIN SPACE */
2252          case 0x200A:    /* HAIR SPACE */
2253          case 0x202f:    /* NARROW NO-BREAK SPACE */
2254          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2255          case 0x3000:    /* IDEOGRAPHIC SPACE */
2256          break;
2257          }
2258        ecode++;
2259        break;
2260    
2261        case OP_NOT_VSPACE:
2262        if (eptr >= md->end_subject)
2263          {
2264          SCHECK_PARTIAL();
2265          MRRETURN(MATCH_NOMATCH);
2266          }
2267        GETCHARINCTEST(c, eptr);
2268        switch(c)
2269          {
2270          default: break;
2271          case 0x0a:      /* LF */
2272          case 0x0b:      /* VT */
2273          case 0x0c:      /* FF */
2274          case 0x0d:      /* CR */
2275          case 0x85:      /* NEL */
2276          case 0x2028:    /* LINE SEPARATOR */
2277          case 0x2029:    /* PARAGRAPH SEPARATOR */
2278          MRRETURN(MATCH_NOMATCH);
2279          }
2280        ecode++;
2281        break;
2282    
2283        case OP_VSPACE:
2284        if (eptr >= md->end_subject)
2285          {
2286          SCHECK_PARTIAL();
2287          MRRETURN(MATCH_NOMATCH);
2288          }
2289        GETCHARINCTEST(c, eptr);
2290        switch(c)
2291          {
2292          default: MRRETURN(MATCH_NOMATCH);
2293          case 0x0a:      /* LF */
2294          case 0x0b:      /* VT */
2295          case 0x0c:      /* FF */
2296          case 0x0d:      /* CR */
2297          case 0x85:      /* NEL */
2298          case 0x2028:    /* LINE SEPARATOR */
2299          case 0x2029:    /* PARAGRAPH SEPARATOR */
2300          break;
2301          }
2302      ecode++;      ecode++;
2303      break;      break;
2304    
# Line 1349  for (;;) Line 2308  for (;;)
2308    
2309      case OP_PROP:      case OP_PROP:
2310      case OP_NOTPROP:      case OP_NOTPROP:
2311      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2312          {
2313          SCHECK_PARTIAL();
2314          MRRETURN(MATCH_NOMATCH);
2315          }
2316      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2317        {        {
2318        int chartype, rqdtype;        const ucd_record *prop = GET_UCD(c);
       int othercase;  
       int category = ucp_findchar(c, &chartype, &othercase);  
   
       rqdtype = *(++ecode);  
       ecode++;  
2319    
2320        if (rqdtype >= 128)        switch(ecode[1])
         {  
         if ((rqdtype - 128 != category) == (op == OP_PROP))  
           RRETURN(MATCH_NOMATCH);  
         }  
       else  
2321          {          {
2322          if ((rqdtype != chartype) == (op == OP_PROP))          case PT_ANY:
2323            RRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
2324            break;
2325    
2326            case PT_LAMP:
2327            if ((prop->chartype == ucp_Lu ||
2328                 prop->chartype == ucp_Ll ||
2329                 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2330              MRRETURN(MATCH_NOMATCH);
2331            break;
2332    
2333            case PT_GC:
2334            if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
2335              MRRETURN(MATCH_NOMATCH);
2336            break;
2337    
2338            case PT_PC:
2339            if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2340              MRRETURN(MATCH_NOMATCH);
2341            break;
2342    
2343            case PT_SC:
2344            if ((ecode[2] != prop->script) == (op == OP_PROP))
2345              MRRETURN(MATCH_NOMATCH);
2346            break;
2347    
2348            /* These are specials */
2349    
2350            case PT_ALNUM:
2351            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2352                 _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2353              MRRETURN(MATCH_NOMATCH);
2354            break;
2355    
2356            case PT_SPACE:    /* Perl space */
2357            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2358                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2359                   == (op == OP_NOTPROP))
2360              MRRETURN(MATCH_NOMATCH);
2361            break;
2362    
2363            case PT_PXSPACE:  /* POSIX space */
2364            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2365                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2366                 c == CHAR_FF || c == CHAR_CR)
2367                   == (op == OP_NOTPROP))
2368              MRRETURN(MATCH_NOMATCH);
2369            break;
2370    
2371            case PT_WORD:
2372            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2373                 _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2374                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2375              MRRETURN(MATCH_NOMATCH);
2376            break;
2377    
2378            /* This should never occur */
2379    
2380            default:
2381            RRETURN(PCRE_ERROR_INTERNAL);
2382          }          }
2383    
2384          ecode += 3;
2385        }        }
2386      break;      break;
2387    
# Line 1376  for (;;) Line 2389  for (;;)
2389      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2390    
2391      case OP_EXTUNI:      case OP_EXTUNI:
2392      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2393          {
2394          SCHECK_PARTIAL();
2395          MRRETURN(MATCH_NOMATCH);
2396          }
2397      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2398        {        {
2399        int chartype;        int category = UCD_CATEGORY(c);
2400        int othercase;        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);
       int category = ucp_findchar(c, &chartype, &othercase);  
       if (category == ucp_M) RRETURN(MATCH_NOMATCH);  
2401        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2402          {          {
2403          int len = 1;          int len = 1;
# Line 1390  for (;;) Line 2405  for (;;)
2405            {            {
2406            GETCHARLEN(c, eptr, len);            GETCHARLEN(c, eptr, len);
2407            }            }
2408          category = ucp_findchar(c, &chartype, &othercase);          category = UCD_CATEGORY(c);
2409          if (category != ucp_M) break;          if (category != ucp_M) break;
2410          eptr += len;          eptr += len;
2411          }          }
# Line 1409  for (;;) Line 2424  for (;;)
2424      loops). */      loops). */
2425    
2426      case OP_REF:      case OP_REF:
2427        case OP_REFI:
2428        caseless = op == OP_REFI;
2429        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2430        ecode += 3;
2431    
2432        /* If the reference is unset, there are two possibilities:
2433    
2434        (a) In the default, Perl-compatible state, set the length negative;
2435        this ensures that every attempt at a match fails. We can't just fail
2436        here, because of the possibility of quantifiers with zero minima.
2437    
2438        (b) If the JavaScript compatibility flag is set, set the length to zero
2439        so that the back reference matches an empty string.
2440    
2441        Otherwise, set the length to the length of what was matched by the
2442        referenced subpattern. */
2443    
2444        if (offset >= offset_top || md->offset_vector[offset] < 0)
2445          length = (md->jscript_compat)? 0 : -1;
2446        else
2447          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2448    
2449        /* Set up for repetition, or handle the non-repeated case */
2450    
2451        switch (*ecode)
2452        {        {
2453        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */        case OP_CRSTAR:
2454        ecode += 3;                                 /* Advance past item */        case OP_CRMINSTAR:
2455          case OP_CRPLUS:
2456          case OP_CRMINPLUS:
2457          case OP_CRQUERY:
2458          case OP_CRMINQUERY:
2459          c = *ecode++ - OP_CRSTAR;
2460          minimize = (c & 1) != 0;
2461          min = rep_min[c];                 /* Pick up values from tables; */
2462          max = rep_max[c];                 /* zero for max => infinity */
2463          if (max == 0) max = INT_MAX;
2464          break;
2465    
2466          case OP_CRRANGE:
2467          case OP_CRMINRANGE:
2468          minimize = (*ecode == OP_CRMINRANGE);
2469          min = GET2(ecode, 1);
2470          max = GET2(ecode, 3);
2471          if (max == 0) max = INT_MAX;
2472          ecode += 5;
2473          break;
2474    
2475          default:               /* No repeat follows */
2476          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2477            {
2478            CHECK_PARTIAL();
2479            MRRETURN(MATCH_NOMATCH);
2480            }
2481          eptr += length;
2482          continue;              /* With the main loop */
2483          }
2484    
2485        /* If the reference is unset, set the length to be longer than the amount      /* Handle repeated back references. If the length of the reference is
2486        of subject left; this ensures that every attempt at a match fails. We      zero, just continue with the main loop. */
       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];  
2487    
2488        /* Set up for repetition, or handle the non-repeated case */      if (length == 0) continue;
2489    
2490        switch (*ecode)      /* First, ensure the minimum number of matches are present. We get back
2491        the length of the reference string explicitly rather than passing the
2492        address of eptr, so that eptr can be a register variable. */
2493    
2494        for (i = 1; i <= min; i++)
2495          {
2496          int slength;
2497          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2498          {          {
2499          case OP_CRSTAR:          CHECK_PARTIAL();
2500          case OP_CRMINSTAR:          MRRETURN(MATCH_NOMATCH);
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
   
         case OP_CRRANGE:  
         case OP_CRMINRANGE:  
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
   
         default:               /* No repeat follows */  
         if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);  
         eptr += length;  
         continue;              /* With the main loop */  
         }  
   
       /* If the length of the reference is zero, just continue with the  
       main loop. */  
   
       if (length == 0) continue;  
   
       /* First, ensure the minimum number of matches are present. We get back  
       the length of the reference string explicitly rather than passing the  
       address of eptr, so that eptr can be a register variable. */  
   
       for (i = 1; i <= min; i++)  
         {  
         if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);  
         eptr += length;  
2501          }          }
2502          eptr += slength;
2503          }
2504    
2505        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2506        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2507    
2508        if (min == max) continue;      if (min == max) continue;
2509    
2510        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2511    
2512        if (minimize)      if (minimize)
2513          {
2514          for (fi = min;; fi++)
2515          {          {
2516          for (fi = min;; fi++)          int slength;
2517            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2518            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2519            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2520            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2521            {            {
2522            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2523            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2524            }            }
2525          /* Control never gets here */          eptr += slength;
2526          }          }
2527          /* Control never gets here */
2528          }
2529    
2530        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2531    
2532        else      else
2533          {
2534          pp = eptr;
2535          for (i = min; i < max; i++)
2536          {          {
2537          pp = eptr;          int slength;
2538          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
2539            {            {
2540            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2541            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2542            }            }
2543          RRETURN(MATCH_NOMATCH);          eptr += slength;
2544            }
2545          while (eptr >= pp)
2546            {
2547            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2548            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2549            eptr -= length;
2550          }          }
2551          MRRETURN(MATCH_NOMATCH);
2552        }        }
2553      /* Control never gets here */      /* Control never gets here */
2554    
   
   
2555      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2556      used when all the characters in the class have values in the range 0-255,      used when all the characters in the class have values in the range 0-255,
2557      and either the matching is caseful, or the characters are in the range      and either the matching is caseful, or the characters are in the range
# Line 1566  for (;;) Line 2606  for (;;)
2606          {          {
2607          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2608            {            {
2609            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2610                {
2611                SCHECK_PARTIAL();
2612                MRRETURN(MATCH_NOMATCH);
2613                }
2614            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2615            if (c > 255)            if (c > 255)
2616              {              {
2617              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2618              }              }
2619            else            else
2620              {              {
2621              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2622              }              }
2623            }            }
2624          }          }
# Line 1584  for (;;) Line 2628  for (;;)
2628          {          {
2629          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2630            {            {
2631            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2632                {
2633                SCHECK_PARTIAL();
2634                MRRETURN(MATCH_NOMATCH);
2635                }
2636            c = *eptr++;            c = *eptr++;
2637            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2638            }            }
2639          }          }
2640    
# Line 1606  for (;;) Line 2654  for (;;)
2654            {            {
2655            for (fi = min;; fi++)            for (fi = min;; fi++)
2656              {              {
2657              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2658              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2659              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2660                if (eptr >= md->end_subject)
2661                  {
2662                  SCHECK_PARTIAL();
2663                  MRRETURN(MATCH_NOMATCH);
2664                  }
2665              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2666              if (c > 255)              if (c > 255)
2667                {                {
2668                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2669                }                }
2670              else              else
2671                {                {
2672                if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);                if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2673                }                }
2674              }              }
2675            }            }
# Line 1626  for (;;) Line 2679  for (;;)
2679            {            {
2680            for (fi = min;; fi++)            for (fi = min;; fi++)
2681              {              {
2682              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2683              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2684              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2685                if (eptr >= md->end_subject)
2686                  {
2687                  SCHECK_PARTIAL();
2688                  MRRETURN(MATCH_NOMATCH);
2689                  }
2690              c = *eptr++;              c = *eptr++;
2691              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2692              }              }
2693            }            }
2694          /* Control never gets here */          /* Control never gets here */
# Line 1649  for (;;) Line 2707  for (;;)
2707            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2708              {              {
2709              int len = 1;              int len = 1;
2710              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2711                  {
2712                  SCHECK_PARTIAL();
2713                  break;
2714                  }
2715              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2716              if (c > 255)              if (c > 255)
2717                {                {
# Line 1663  for (;;) Line 2725  for (;;)
2725              }              }
2726            for (;;)            for (;;)
2727              {              {
2728              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2729              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2730              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2731              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1675  for (;;) Line 2737  for (;;)
2737            {            {
2738            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2739              {              {
2740              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2741                  {
2742                  SCHECK_PARTIAL();
2743                  break;
2744                  }
2745              c = *eptr;              c = *eptr;
2746              if ((data[c/8] & (1 << (c&7))) == 0) break;              if ((data[c/8] & (1 << (c&7))) == 0) break;
2747              eptr++;              eptr++;
2748              }              }
2749            while (eptr >= pp)            while (eptr >= pp)
2750              {              {
2751              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
             eptr--;  
2752              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2753                eptr--;
2754              }              }
2755            }            }
2756    
2757          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2758          }          }
2759        }        }
2760      /* Control never gets here */      /* Control never gets here */
2761    
2762    
2763      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2764      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
2765        mode, because Unicode properties are supported in non-UTF-8 mode. */
2766    
2767  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2768      case OP_XCLASS:      case OP_XCLASS:
# Line 1736  for (;;) Line 2803  for (;;)
2803    
2804        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2805          {          {
2806          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2807          GETCHARINC(c, eptr);            {
2808          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2809              MRRETURN(MATCH_NOMATCH);
2810              }
2811            GETCHARINCTEST(c, eptr);
2812            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2813          }          }
2814    
2815        /* If max == min we can continue with the main loop without the        /* If max == min we can continue with the main loop without the
# Line 1753  for (;;) Line 2824  for (;;)
2824          {          {
2825          for (fi = min;; fi++)          for (fi = min;; fi++)
2826            {            {
2827            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2828            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2829            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2830            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
2831            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
2832                SCHECK_PARTIAL();
2833                MRRETURN(MATCH_NOMATCH);
2834                }
2835              GETCHARINCTEST(c, eptr);
2836              if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2837            }            }
2838          /* Control never gets here */          /* Control never gets here */
2839          }          }
# Line 1770  for (;;) Line 2846  for (;;)
2846          for (i = min; i < max; i++)          for (i = min; i < max; i++)
2847            {            {
2848            int len = 1;            int len = 1;
2849            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
2850            GETCHARLEN(c, eptr, len);              {
2851                SCHECK_PARTIAL();
2852                break;
2853                }
2854              GETCHARLENTEST(c, eptr, len);
2855            if (!_pcre_xclass(c, data)) break;            if (!_pcre_xclass(c, data)) break;
2856            eptr += len;            eptr += len;
2857            }            }
2858          for(;;)          for(;;)
2859            {            {
2860            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2861            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2862            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2863            BACKCHAR(eptr)            if (utf8) BACKCHAR(eptr);
2864            }            }
2865          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2866          }          }
2867    
2868        /* Control never gets here */        /* Control never gets here */
# Line 1798  for (;;) Line 2878  for (;;)
2878        length = 1;        length = 1;
2879        ecode++;        ecode++;
2880        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2881        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2882        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
2883            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2884            MRRETURN(MATCH_NOMATCH);
2885            }
2886          while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
2887        }        }
2888      else      else
2889  #endif  #endif
2890    
2891      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2892        {        {
2893        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2894        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);          {
2895            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2896            MRRETURN(MATCH_NOMATCH);
2897            }
2898          if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
2899        ecode += 2;        ecode += 2;
2900        }        }
2901      break;      break;
2902    
2903      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2904    
2905      case OP_CHARNC:      case OP_CHARI:
2906  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2907      if (utf8)      if (utf8)
2908        {        {
# Line 1822  for (;;) Line 2910  for (;;)
2910        ecode++;        ecode++;
2911        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2912    
2913        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2914            {
2915            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2916            MRRETURN(MATCH_NOMATCH);
2917            }
2918    
2919        /* If the pattern character's value is < 128, we have only one byte, and        /* If the pattern character's value is < 128, we have only one byte, and
2920        can use the fast lookup table. */        can use the fast lookup table. */
2921    
2922        if (fc < 128)        if (fc < 128)
2923          {          {
2924          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2925          }          }
2926    
2927        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character */
2928    
2929        else        else
2930          {          {
2931          int dc;          unsigned int dc;
2932          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
2933          ecode += length;          ecode += length;
2934    
2935          /* If we have Unicode property support, we can use it to test the other          /* If we have Unicode property support, we can use it to test the other
2936          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. */  
2937    
2938          if (fc != dc)          if (fc != dc)
2939            {            {
2940  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2941            int chartype;            if (dc != UCD_OTHERCASE(fc))
           int othercase;  
           if (ucp_findchar(fc, &chartype, &othercase) < 0 || dc != othercase)  
2942  #endif  #endif
2943              RRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
2944            }            }
2945          }          }
2946        }        }
# Line 1861  for (;;) Line 2949  for (;;)
2949    
2950      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2951        {        {
2952        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2953        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
2954            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2955            MRRETURN(MATCH_NOMATCH);
2956            }
2957          if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2958        ecode += 2;        ecode += 2;
2959        }        }
2960      break;      break;
2961    
2962      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
2963    
2964      case OP_EXACT:      case OP_EXACT:
2965        case OP_EXACTI:
2966      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
2967      ecode += 3;      ecode += 3;
2968      goto REPEATCHAR;      goto REPEATCHAR;
2969    
2970        case OP_POSUPTO:
2971        case OP_POSUPTOI:
2972        possessive = TRUE;
2973        /* Fall through */
2974    
2975      case OP_UPTO:      case OP_UPTO:
2976        case OP_UPTOI:
2977      case OP_MINUPTO:      case OP_MINUPTO:
2978        case OP_MINUPTOI:
2979      min = 0;      min = 0;
2980      max = GET2(ecode, 1);      max = GET2(ecode, 1);
2981      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
2982      ecode += 3;      ecode += 3;
2983      goto REPEATCHAR;      goto REPEATCHAR;
2984    
2985        case OP_POSSTAR:
2986        case OP_POSSTARI:
2987        possessive = TRUE;
2988        min = 0;
2989        max = INT_MAX;
2990        ecode++;
2991        goto REPEATCHAR;
2992    
2993        case OP_POSPLUS:
2994        case OP_POSPLUSI:
2995        possessive = TRUE;
2996        min = 1;
2997        max = INT_MAX;
2998        ecode++;
2999        goto REPEATCHAR;
3000    
3001        case OP_POSQUERY:
3002        case OP_POSQUERYI:
3003        possessive = TRUE;
3004        min = 0;
3005        max = 1;
3006        ecode++;
3007        goto REPEATCHAR;
3008    
3009      case OP_STAR:      case OP_STAR:
3010        case OP_STARI:
3011      case OP_MINSTAR:      case OP_MINSTAR:
3012        case OP_MINSTARI:
3013      case OP_PLUS:      case OP_PLUS:
3014        case OP_PLUSI:
3015      case OP_MINPLUS:      case OP_MINPLUS:
3016        case OP_MINPLUSI:
3017      case OP_QUERY:      case OP_QUERY:
3018        case OP_QUERYI:
3019      case OP_MINQUERY:      case OP_MINQUERY:
3020      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3021        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3022      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3023      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3024      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3025      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3026    
3027      /* Common code for all repeated single-character matches. We can give      /* Common code for all repeated single-character matches. */
     up quickly if there are fewer than the minimum number of characters left in  
     the subject. */  
3028    
3029      REPEATCHAR:      REPEATCHAR:
3030  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1905  for (;;) Line 3033  for (;;)
3033        length = 1;        length = 1;
3034        charptr = ecode;        charptr = ecode;
3035        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3036        ecode += length;        ecode += length;
3037    
3038        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1913  for (;;) Line 3040  for (;;)
3040    
3041        if (length > 1)        if (length > 1)
3042          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3043  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3044          int othercase;          unsigned int othercase;
3045          int chartype;          if (op >= OP_STARI &&     /* Caseless */
3046          if ((ims & PCRE_CASELESS) != 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
              ucp_findchar(fc, &chartype, &othercase) >= 0 &&  
              othercase > 0)  
3047            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3048            else oclength = 0;
3049  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3050    
3051          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3052            {            {
3053            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3054            /* Need braces because of following else */              memcmp(eptr, charptr, length) == 0) eptr += length;
3055            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3056              else if (oclength > 0 &&
3057                       eptr <= md->end_subject - oclength &&
3058                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3059    #endif  /* SUPPORT_UCP */
3060            else            else
3061              {              {
3062              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3063              eptr += oclength;              MRRETURN(MATCH_NOMATCH);
3064              }              }
3065            }            }
3066    
# Line 1943  for (;;) Line 3070  for (;;)
3070            {            {
3071            for (fi = min;; fi++)            for (fi = min;; fi++)
3072              {              {
3073              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3074              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3075              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3076              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3077              /* Need braces because of following else */                memcmp(eptr, charptr, length) == 0) eptr += length;
3078              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3079                else if (oclength > 0 &&
3080                         eptr <= md->end_subject - oclength &&
3081                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3082    #endif  /* SUPPORT_UCP */
3083              else              else
3084                {                {
3085                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3086                eptr += oclength;                MRRETURN(MATCH_NOMATCH);
3087                }                }
3088              }              }
3089            /* Control never gets here */            /* Control never gets here */
3090            }            }
3091          else  
3092            else  /* Maximize */
3093            {            {
3094            pp = eptr;            pp = eptr;
3095            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3096              {              {
3097              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3098              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, length) == 0) eptr += length;
3099              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3100                else if (oclength > 0 &&
3101                         eptr <= md->end_subject - oclength &&
3102                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3103    #endif  /* SUPPORT_UCP */
3104              else              else
3105                {                {
3106                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3107                eptr += oclength;                break;
3108                }                }
3109              }              }
3110            while (eptr >= pp)  
3111             {            if (possessive) continue;
3112             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3113             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3114             eptr -= length;              {
3115             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3116            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3117                if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3118    #ifdef SUPPORT_UCP
3119                eptr--;
3120                BACKCHAR(eptr);
3121    #else   /* without SUPPORT_UCP */
3122                eptr -= length;
3123    #endif  /* SUPPORT_UCP */
3124                }
3125            }            }
3126          /* Control never gets here */          /* Control never gets here */
3127          }          }
# Line 1990  for (;;) Line 3134  for (;;)
3134  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3135    
3136      /* When not in UTF-8 mode, load a single-byte character. */      /* When not in UTF-8 mode, load a single-byte character. */
3137        {  
3138        if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);      fc = *ecode++;
       fc = *ecode++;  
       }  
3139    
3140      /* The value of fc at this point is always less than 256, though we may or      /* The value of fc at this point is always less than 256, though we may or
3141      may not be in UTF-8 mode. The code is duplicated for the caseless and      may not be in UTF-8 mode. The code is duplicated for the caseless and
# Line 2007  for (;;) Line 3149  for (;;)
3149      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3150        max, eptr));        max, eptr));
3151    
3152      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3153        {        {
3154        fc = md->lcc[fc];        fc = md->lcc[fc];
3155        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3156          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3157            if (eptr >= md->end_subject)
3158              {
3159              SCHECK_PARTIAL();
3160              MRRETURN(MATCH_NOMATCH);
3161              }
3162            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3163            }
3164        if (min == max) continue;        if (min == max) continue;
3165        if (minimize)        if (minimize)
3166          {          {
3167          for (fi = min;; fi++)          for (fi = min;; fi++)
3168            {            {
3169            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3170            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3171            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3172                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3173              RRETURN(MATCH_NOMATCH);              {
3174                SCHECK_PARTIAL();
3175                MRRETURN(MATCH_NOMATCH);
3176                }
3177              if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3178            }            }
3179          /* Control never gets here */          /* Control never gets here */
3180          }          }
3181        else        else  /* Maximize */
3182          {          {
3183          pp = eptr;          pp = eptr;
3184          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3185            {            {
3186            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3187                {
3188                SCHECK_PARTIAL();
3189                break;
3190                }
3191              if (fc != md->lcc[*eptr]) break;
3192            eptr++;            eptr++;
3193            }            }
3194    
3195            if (possessive) continue;
3196    
3197          while (eptr >= pp)          while (eptr >= pp)
3198            {            {
3199            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3200            eptr--;            eptr--;
3201            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3202            }            }
3203          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3204          }          }
3205        /* Control never gets here */        /* Control never gets here */
3206        }        }
# Line 2048  for (;;) Line 3209  for (;;)
3209    
3210      else      else
3211        {        {
3212        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3213            {
3214            if (eptr >= md->end_subject)
3215              {
3216              SCHECK_PARTIAL();
3217              MRRETURN(MATCH_NOMATCH);
3218              }
3219            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3220            }
3221    
3222        if (min == max) continue;        if (min == max) continue;
3223    
3224        if (minimize)        if (minimize)
3225          {          {
3226          for (fi = min;; fi++)          for (fi = min;; fi++)
3227            {            {
3228            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3229            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3230            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3231              RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
3232                {
3233                SCHECK_PARTIAL();
3234                MRRETURN(MATCH_NOMATCH);
3235                }
3236              if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3237            }            }
3238          /* Control never gets here */          /* Control never gets here */
3239          }          }
3240        else        else  /* Maximize */
3241          {          {
3242          pp = eptr;          pp = eptr;
3243          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3244            {            {
3245            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3246                {
3247                SCHECK_PARTIAL();
3248                break;
3249                }
3250              if (fc != *eptr) break;
3251            eptr++;            eptr++;
3252            }            }
3253            if (possessive) continue;
3254    
3255          while (eptr >= pp)          while (eptr >= pp)
3256            {            {
3257            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3258            eptr--;            eptr--;
3259            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3260            }            }
3261          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3262          }          }
3263        }        }
3264      /* Control never gets here */      /* Control never gets here */
# Line 2084  for (;;) Line 3267  for (;;)
3267      checking can be multibyte. */      checking can be multibyte. */
3268    
3269      case OP_NOT:      case OP_NOT:
3270      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3271        if (eptr >= md->end_subject)
3272          {
3273          SCHECK_PARTIAL();
3274          MRRETURN(MATCH_NOMATCH);
3275          }
3276      ecode++;      ecode++;
3277      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3278      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3279        {        {
3280  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3281        if (c < 256)        if (c < 256)
3282  #endif  #endif
3283        c = md->lcc[c];        c = md->lcc[c];
3284        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3285        }        }
3286      else      else    /* Caseful */
3287        {        {
3288        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3289        }        }
3290      break;      break;
3291    
# Line 2109  for (;;) Line 3297  for (;;)
3297      about... */      about... */
3298    
3299      case OP_NOTEXACT:      case OP_NOTEXACT:
3300        case OP_NOTEXACTI:
3301      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3302      ecode += 3;      ecode += 3;
3303      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3304    
3305      case OP_NOTUPTO:      case OP_NOTUPTO:
3306        case OP_NOTUPTOI:
3307      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3308        case OP_NOTMINUPTOI:
3309        min = 0;
3310        max = GET2(ecode, 1);
3311        minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3312        ecode += 3;
3313        goto REPEATNOTCHAR;
3314    
3315        case OP_NOTPOSSTAR:
3316        case OP_NOTPOSSTARI:
3317        possessive = TRUE;
3318        min = 0;
3319        max = INT_MAX;
3320        ecode++;
3321        goto REPEATNOTCHAR;
3322    
3323        case OP_NOTPOSPLUS:
3324        case OP_NOTPOSPLUSI:
3325        possessive = TRUE;
3326        min = 1;
3327        max = INT_MAX;
3328        ecode++;
3329        goto REPEATNOTCHAR;
3330    
3331        case OP_NOTPOSQUERY:
3332        case OP_NOTPOSQUERYI:
3333        possessive = TRUE;
3334        min = 0;
3335        max = 1;
3336        ecode++;
3337        goto REPEATNOTCHAR;
3338    
3339        case OP_NOTPOSUPTO:
3340        case OP_NOTPOSUPTOI:
3341        possessive = TRUE;
3342      min = 0;      min = 0;
3343      max = GET2(ecode, 1);      max = GET2(ecode, 1);
     minimize = *ecode == OP_NOTMINUPTO;  
3344      ecode += 3;      ecode += 3;
3345      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3346    
3347      case OP_NOTSTAR:      case OP_NOTSTAR:
3348        case OP_NOTSTARI:
3349      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3350        case OP_NOTMINSTARI:
3351      case OP_NOTPLUS:      case OP_NOTPLUS:
3352        case OP_NOTPLUSI:
3353      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3354        case OP_NOTMINPLUSI:
3355      case OP_NOTQUERY:      case OP_NOTQUERY:
3356        case OP_NOTQUERYI:
3357      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3358      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3359        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3360      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3361      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3362      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3363      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3364    
3365      /* Common code for all repeated single-byte matches. We can give up quickly      /* Common code for all repeated single-byte matches. */
     if there are fewer than the minimum number of bytes left in the  
     subject. */  
3366    
3367      REPEATNOTCHAR:      REPEATNOTCHAR:
     if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3368      fc = *ecode++;      fc = *ecode++;
3369    
3370      /* The code is duplicated for the caseless and caseful cases, for speed,      /* The code is duplicated for the caseless and caseful cases, for speed,
# Line 2152  for (;;) Line 3378  for (;;)
3378      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3379        max, eptr));        max, eptr));
3380    
3381      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3382        {        {
3383        fc = md->lcc[fc];        fc = md->lcc[fc];
3384    
# Line 2160  for (;;) Line 3386  for (;;)
3386        /* UTF-8 mode */        /* UTF-8 mode */
3387        if (utf8)        if (utf8)
3388          {          {
3389          register int d;          register unsigned int d;
3390          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3391            {            {
3392              if (eptr >= md->end_subject)
3393                {
3394                SCHECK_PARTIAL();
3395                MRRETURN(MATCH_NOMATCH);
3396                }
3397            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3398            if (d < 256) d = md->lcc[d];            if (d < 256) d = md->lcc[d];
3399            if (fc == d) RRETURN(MATCH_NOMATCH);            if (fc == d) MRRETURN(MATCH_NOMATCH);
3400            }            }
3401          }          }
3402        else        else
# Line 2174  for (;;) Line 3405  for (;;)
3405        /* Not UTF-8 mode */        /* Not UTF-8 mode */
3406          {          {
3407          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3408            if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            {
3409              if (eptr >= md->end_subject)
3410                {
3411                SCHECK_PARTIAL();
3412                MRRETURN(MATCH_NOMATCH);
3413                }
3414              if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3415              }
3416          }          }
3417    
3418        if (min == max) continue;        if (min == max) continue;
# Line 2185  for (;;) Line 3423  for (;;)
3423          /* UTF-8 mode */          /* UTF-8 mode */
3424          if (utf8)          if (utf8)
3425            {            {
3426            register int d;            register unsigned int d;
3427            for (fi = min;; fi++)            for (fi = min;; fi++)
3428              {              {
3429              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3430              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3431                if (fi >= max) MRRETURN(MATCH_NOMATCH);
3432                if (eptr >= md->end_subject)
3433                  {
3434                  SCHECK_PARTIAL();
3435                  MRRETURN(MATCH_NOMATCH);
3436                  }
3437              GETCHARINC(d, eptr);              GETCHARINC(d, eptr);
3438              if (d < 256) d = md->lcc[d];              if (d < 256) d = md->lcc[d];
3439              if (fi >= max || eptr >= md->end_subject || fc == d)              if (fc == d) MRRETURN(MATCH_NOMATCH);
               RRETURN(MATCH_NOMATCH);  
3440              }              }
3441            }            }
3442          else          else
# Line 2202  for (;;) Line 3445  for (;;)
3445            {            {
3446            for (fi = min;; fi++)            for (fi = min;; fi++)
3447              {              {
3448              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3449              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3450              if (fi >= max || eptr >= md->end_subject || fc == md->lcc[*eptr++])              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3451                RRETURN(MATCH_NOMATCH);              if (eptr >= md->end_subject)
3452                  {
3453                  SCHECK_PARTIAL();
3454                  MRRETURN(MATCH_NOMATCH);
3455                  }
3456                if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3457              }              }
3458            }            }
3459          /* Control never gets here */          /* Control never gets here */
# Line 2221  for (;;) Line 3469  for (;;)
3469          /* UTF-8 mode */          /* UTF-8 mode */
3470          if (utf8)          if (utf8)
3471            {            {
3472            register int d;            register unsigned int d;
3473            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3474              {              {
3475              int len = 1;              int len = 1;
3476              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3477                  {
3478                  SCHECK_PARTIAL();
3479                  break;
3480                  }
3481              GETCHARLEN(d, eptr, len);              GETCHARLEN(d, eptr, len);
3482              if (d < 256) d = md->lcc[d];              if (d < 256) d = md->lcc[d];
3483              if (fc == d) break;              if (fc == d) break;
3484              eptr += len;              eptr += len;
3485              }              }
3486            for(;;)          if (possessive) continue;
3487            for(;;)
3488              {              {
3489              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3490              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3491              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3492              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2245  for (;;) Line 3498  for (;;)
3498            {            {
3499            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3500              {              {
3501              if (eptr >= md->end_subject || fc == md->lcc[*eptr]) break;              if (eptr >= md->end_subject)
3502                  {
3503                  SCHECK_PARTIAL();
3504                  break;
3505                  }
3506                if (fc == md->lcc[*eptr]) break;
3507              eptr++;              eptr++;
3508              }              }
3509              if (possessive) continue;
3510            while (eptr >= pp)            while (eptr >= pp)
3511              {              {
3512              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3513              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3514              eptr--;              eptr--;
3515              }              }
3516            }            }
3517    
3518          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3519          }          }
3520        /* Control never gets here */        /* Control never gets here */
3521        }        }
# Line 2269  for (;;) Line 3528  for (;;)
3528        /* UTF-8 mode */        /* UTF-8 mode */
3529        if (utf8)        if (utf8)
3530          {          {
3531          register int d;          register unsigned int d;
3532          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3533            {            {
3534              if (eptr >= md->end_subject)
3535                {
3536                SCHECK_PARTIAL();
3537                MRRETURN(MATCH_NOMATCH);
3538                }
3539            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3540            if (fc == d) RRETURN(MATCH_NOMATCH);            if (fc == d) MRRETURN(MATCH_NOMATCH);
3541            }            }
3542          }          }
3543        else        else
# Line 2281  for (;;) Line 3545  for (;;)
3545        /* Not UTF-8 mode */        /* Not UTF-8 mode */
3546          {          {
3547          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3548            if (fc == *eptr++) RRETURN(MATCH_NOMATCH);            {
3549              if (eptr >= md->end_subject)
3550                {
3551                SCHECK_PARTIAL();
3552                MRRETURN(MATCH_NOMATCH);
3553                }
3554              if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
3555              }
3556          }          }
3557    
3558        if (min == max) continue;        if (min == max) continue;
# Line 2292  for (;;) Line 3563  for (;;)
3563          /* UTF-8 mode */          /* UTF-8 mode */
3564          if (utf8)          if (utf8)
3565            {            {
3566            register int d;            register unsigned int d;
3567            for (fi = min;; fi++)            for (fi = min;; fi++)
3568              {              {
3569              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3570              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3571                if (fi >= max) MRRETURN(MATCH_NOMATCH);
3572                if (eptr >= md->end_subject)
3573                  {
3574                  SCHECK_PARTIAL();
3575                  MRRETURN(MATCH_NOMATCH);
3576                  }
3577              GETCHARINC(d, eptr);              GETCHARINC(d, eptr);
3578              if (fi >= max || eptr >= md->end_subject || fc == d)              if (fc == d) MRRETURN(MATCH_NOMATCH);
               RRETURN(MATCH_NOMATCH);  
3579              }              }
3580            }            }
3581          else          else
# Line 2308  for (;;) Line 3584  for (;;)
3584            {            {
3585            for (fi = min;; fi++)            for (fi = min;; fi++)
3586              {              {
3587              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
3588              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3589              if (fi >= max || eptr >= md->end_subject || fc == *eptr++)              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3590                RRETURN(MATCH_NOMATCH);              if (eptr >= md->end_subject)
3591                  {
3592                  SCHECK_PARTIAL();
3593                  MRRETURN(MATCH_NOMATCH);
3594                  }
3595                if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
3596              }              }
3597            }            }
3598          /* Control never gets here */          /* Control never gets here */
# Line 2327  for (;;) Line 3608  for (;;)
3608          /* UTF-8 mode */          /* UTF-8 mode */
3609          if (utf8)          if (utf8)
3610            {            {
3611            register int d;            register unsigned int d;
3612            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3613              {              {
3614              int len = 1;              int len = 1;
3615              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3616                  {
3617                  SCHECK_PARTIAL();
3618                  break;
3619                  }
3620              GETCHARLEN(d, eptr, len);              GETCHARLEN(d, eptr, len);
3621              if (fc == d) break;              if (fc == d) break;
3622              eptr += len;              eptr += len;
3623              }              }
3624              if (possessive) continue;
3625            for(;;)            for(;;)
3626              {              {
3627              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
3628              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3629              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3630              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2350  for (;;) Line 3636  for (;;)
3636            {            {
3637            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3638              {              {
3639              if (eptr >= md->end_subject || fc == *eptr) break;              if (eptr >= md->end_subject)
3640                  {
3641                  SCHECK_PARTIAL();
3642                  break;
3643                  }
3644                if (fc == *eptr) break;
3645              eptr++;              eptr++;
3646              }              }
3647              if (possessive) continue;
3648            while (eptr >= pp)            while (eptr >= pp)
3649              {              {
3650              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM35);
3651              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3652              eptr--;              eptr--;
3653              }              }
3654            }            }
3655    
3656          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3657          }          }
3658        }        }
3659      /* Control never gets here */      /* Control never gets here */
# Line 2384  for (;;) Line 3676  for (;;)
3676      ecode += 3;      ecode += 3;
3677      goto REPEATTYPE;      goto REPEATTYPE;
3678    
3679        case OP_TYPEPOSSTAR:
3680        possessive = TRUE;
3681        min = 0;
3682        max = INT_MAX;
3683        ecode++;
3684        goto REPEATTYPE;
3685    
3686        case OP_TYPEPOSPLUS:
3687        possessive = TRUE;
3688        min = 1;
3689        max = INT_MAX;
3690        ecode++;
3691        goto REPEATTYPE;