/[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 93 by nigel, Sat Feb 24 21:41:42 2007 UTC revision 882 by ph10, Sun Jan 15 18:45:27 2012 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) 1997-2006 University of Cambridge             Copyright (c) 1997-2012 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 */  #define NLBLOCK md             /* Block containing newline information */
50  #define PSSTART start_subject  /* Field containing processed string start */  #define PSSTART start_subject  /* Field containing processed string start */
51  #define PSEND   end_subject    /* Field containing processed string end */  #define PSEND   end_subject    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55  /* The chain of eptrblocks for tail recursions uses memory in stack workspace,  /* Undefine some potentially clashing cpp symbols */
 obtained at top level, the size of which is defined by EPTR_WORK_SIZE. */  
56    
57  #define EPTR_WORK_SIZE (1000)  #undef min
58    #undef max
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_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
 #define match_tail_recursed  0x04  /* Tail recursive call */  
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 65  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_ONCE         (-996)
80    #define MATCH_PRUNE        (-995)
81    #define MATCH_SKIP         (-994)
82    #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85  /* 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.
86  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,
87  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 78  static const char rep_max[] = { 0, 0, 0, Line 95  static const char rep_max[] = { 0, 0, 0,
95    
96    
97    
98  #ifdef DEBUG  #ifdef PCRE_DEBUG
99  /*************************************************  /*************************************************
100  *        Debugging function to print chars       *  *        Debugging function to print chars       *
101  *************************************************/  *************************************************/
# Line 96  Returns:     nothing Line 113  Returns:     nothing
113  */  */
114    
115  static void  static void
116  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
117  {  {
118  unsigned int c;  unsigned int c;
119  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;
# Line 111  while (length-- > 0) Line 128  while (length-- > 0)
128  *          Match a back-reference                *  *          Match a back-reference                *
129  *************************************************/  *************************************************/
130    
131  /* 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
132  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
133    the length passed is zero. Note that in caseless UTF-8 mode, the number of
134    subject bytes matched may be different to the number of reference bytes.
135    
136  Arguments:  Arguments:
137    offset      index into the offset vector    offset      index into the offset vector
138    eptr        points into the subject    eptr        pointer into the subject
139    length      length to be matched    length      length of reference to be matched (number of bytes)
140    md          points to match data block    md          points to match data block
141    ims         the ims flags    caseless    TRUE if caseless
142    
143  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
144  */  */
145    
146  static BOOL  static int
147  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
148    unsigned long int ims)    BOOL caseless)
149  {  {
150  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
151    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
152    
153  #ifdef DEBUG  #ifdef PCRE_DEBUG
154  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
155    printf("matching subject <null>");    printf("matching subject <null>");
156  else  else
# Line 143  pchars(p, length, FALSE, md); Line 163  pchars(p, length, FALSE, md);
163  printf("\n");  printf("\n");
164  #endif  #endif
165    
166  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
167    
168  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
169    
170  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
171    properly if Unicode properties are supported. Otherwise, we can check only
172    ASCII characters. */
173    
174  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
175    {    {
176    while (length-- > 0)  #ifdef SUPPORT_UTF
177      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
178      if (md->utf)
179        {
180        /* Match characters up to the end of the reference. NOTE: the number of
181        bytes matched may differ, because there are some characters whose upper and
182        lower case versions code as different numbers of bytes. For example, U+023A
183        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
184        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
185        the latter. It is important, therefore, to check the length along the
186        reference, not along the subject (earlier code did this wrong). */
187    
188        PCRE_PUCHAR endptr = p + length;
189        while (p < endptr)
190          {
191          int c, d;
192          if (eptr >= md->end_subject) return -1;
193          GETCHARINC(c, eptr);
194          GETCHARINC(d, p);
195          if (c != d && c != UCD_OTHERCASE(d)) return -1;
196          }
197        }
198      else
199    #endif
200    #endif
201    
202      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
203      is no UCP support. */
204        {
205        if (eptr + length > md->end_subject) return -1;
206        while (length-- > 0)
207          {
208          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
209          p++;
210          eptr++;
211          }
212        }
213    }    }
214    
215    /* In the caseful case, we can just compare the bytes, whether or not we
216    are in UTF-8 mode. */
217    
218  else  else
219    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
220      if (eptr + length > md->end_subject) return -1;
221      while (length-- > 0) if (*p++ != *eptr++) return -1;
222      }
223    
224  return TRUE;  return (int)(eptr - eptr_start);
225  }  }
226    
227    
# Line 183  calls by keeping local variables that ne Line 247  calls by keeping local variables that ne
247  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
248  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
249  always used to.  always used to.
250    
251    The original heap-recursive code used longjmp(). However, it seems that this
252    can be very slow on some operating systems. Following a suggestion from Stan
253    Switzer, the use of longjmp() has been abolished, at the cost of having to
254    provide a unique number for each call to RMATCH. There is no way of generating
255    a sequence of numbers at compile time in C. I have given them names, to make
256    them stand out more clearly.
257    
258    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
259    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
260    tests. Furthermore, not using longjmp() means that local dynamic variables
261    don't have indeterminate values; this has meant that the frame size can be
262    reduced because the result can be "passed back" by straight setting of the
263    variable instead of being passed in the frame.
264  ****************************************************************************  ****************************************************************************
265  ***************************************************************************/  ***************************************************************************/
266    
267    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
268    below must be updated in sync.  */
269    
270    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
271           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
272           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
273           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
274           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
275           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
276           RM61,  RM62, RM63, RM64, RM65, RM66 };
277    
278  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
279  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
280    actually used in this definition. */
281    
282  #ifndef NO_RECURSE  #ifndef NO_RECURSE
283  #define REGISTER register  #define REGISTER register
284  #ifdef DEBUG  
285  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
286    #define RMATCH(ra,rb,rc,rd,re,rw) \
287    { \    { \
288    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
289    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
290    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
291    }    }
292  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 205  versions and production versions. */ Line 295  versions and production versions. */
295    return ra; \    return ra; \
296    }    }
297  #else  #else
298  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
299    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
300  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
301  #endif  #endif
302    
303  #else  #else
304    
305    
306  /* 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
307  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
308  match(), which never changes. */  argument of match(), which never changes. */
309    
310  #define REGISTER  #define REGISTER
311    
312  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
313    {\    {\
314    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
315    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
316      {\    frame->Xwhere = rw; \
317      newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
318      newframe->Xecode = rb;\    newframe->Xecode = rb;\
319      newframe->Xoffset_top = rc;\    newframe->Xmstart = mstart;\
320      newframe->Xims = re;\    newframe->Xoffset_top = rc;\
321      newframe->Xeptrb = rf;\    newframe->Xeptrb = re;\
322      newframe->Xflags = rg;\    newframe->Xrdepth = frame->Xrdepth + 1;\
323      newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xprevframe = frame;\
324      newframe->Xprevframe = frame;\    frame = newframe;\
325      frame = newframe;\    DPRINTF(("restarting from line %d\n", __LINE__));\
326      DPRINTF(("restarting from line %d\n", __LINE__));\    goto HEAP_RECURSE;\
327      goto HEAP_RECURSE;\    L_##rw:\
328      }\    DPRINTF(("jumped back to line %d\n", __LINE__));\
   else\  
     {\  
     DPRINTF(("longjumped back to line %d\n", __LINE__));\  
     frame = md->thisframe;\  
     rx = frame->Xresult;\  
     }\  
329    }    }
330    
331  #define RRETURN(ra)\  #define RRETURN(ra)\
332    {\    {\
333    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
334    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
335    (pcre_stack_free)(newframe);\    (PUBL(stack_free))(oldframe);\
336    if (frame != NULL)\    if (frame != NULL)\
337      {\      {\
338      frame->Xresult = ra;\      rrc = ra;\
339      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
340      }\      }\
341    return ra;\    return ra;\
342    }    }
# Line 266  typedef struct heapframe { Line 349  typedef struct heapframe {
349    
350    /* Function arguments that may change */    /* Function arguments that may change */
351    
352    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
353    const uschar *Xecode;    const pcre_uchar *Xecode;
354      PCRE_PUCHAR Xmstart;
355    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
356    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
357    unsigned int Xrdepth;    unsigned int Xrdepth;
358    
359    /* Function local variables */    /* Function local variables */
360    
361    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
362    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
363    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
364    const uschar *Xnext;  #endif
365    const uschar *Xpp;    PCRE_PUCHAR Xdata;
366    const uschar *Xprev;    PCRE_PUCHAR Xnext;
367    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
368      PCRE_PUCHAR Xprev;
369      PCRE_PUCHAR Xsaved_eptr;
370    
371    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
372    
# Line 290  typedef struct heapframe { Line 374  typedef struct heapframe {
374    BOOL Xcondition;    BOOL Xcondition;
375    BOOL Xprev_is_word;    BOOL Xprev_is_word;
376    
   unsigned long int Xoriginal_ims;  
   
377  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
378    int Xprop_type;    int Xprop_type;
379    int Xprop_value;    int Xprop_value;
380    int Xprop_fail_result;    int Xprop_fail_result;
381    int Xprop_category;    int Xoclength;
382    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_script;  
383  #endif  #endif
384    
385      int Xcodelink;
386    int Xctype;    int Xctype;
387    unsigned int Xfc;    unsigned int Xfc;
388    int Xfi;    int Xfi;
# Line 316  typedef struct heapframe { Line 398  typedef struct heapframe {
398    
399    eptrblock Xnewptrb;    eptrblock Xnewptrb;
400    
401    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
402    
403    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
404    
405  } heapframe;  } heapframe;
406    
# Line 337  typedef struct heapframe { Line 418  typedef struct heapframe {
418    
419  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
420  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
421  same response.  same response. */
422    
423  Performance note: It might be tempting to extract commonly used fields from the  /* These macros pack up tests that are used for partial matching, and which
424  md structure (e.g. utf8, end_subject) into individual variables to improve  appear several times in the code. We set the "hit end" flag if the pointer is
425    at the end of the subject and also past the start of the subject (i.e.
426    something has been matched). For hard partial matching, we then return
427    immediately. The second one is used when we already know we are past the end of
428    the subject. */
429    
430    #define CHECK_PARTIAL()\
431      if (md->partial != 0 && eptr >= md->end_subject && \
432          eptr > md->start_used_ptr) \
433        { \
434        md->hitend = TRUE; \
435        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
436        }
437    
438    #define SCHECK_PARTIAL()\
439      if (md->partial != 0 && eptr > md->start_used_ptr) \
440        { \
441        md->hitend = TRUE; \
442        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
443        }
444    
445    
446    /* Performance note: It might be tempting to extract commonly used fields from
447    the md structure (e.g. utf, end_subject) into individual variables to improve
448  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
449  made performance worse.  made performance worse.
450    
451  Arguments:  Arguments:
452     eptr        pointer to current character in subject     eptr        pointer to current character in subject
453     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
454       mstart      pointer to the current match start position (can be modified
455                     by encountering \K)
456     offset_top  current top pointer     offset_top  current top pointer
457     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
458     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
459                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
                  match_tail_recursed - this is a tail_recursed group  
460     rdepth      the recursion depth     rdepth      the recursion depth
461    
462  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
463                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
464                   a negative MATCH_xxx value for PRUNE, SKIP, etc
465                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
466                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
467  */  */
468    
469  static int  static int
470  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
471    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
472    int flags, unsigned int rdepth)    unsigned int rdepth)
473  {  {
474  /* 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,
475  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 377  so they can be ordinary variables in all Line 478  so they can be ordinary variables in all
478  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
479  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
480  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
481  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
482    
483  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
484    BOOL caseless;
485    int condcode;
486    
487  /* 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
488  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 387  heap storage. Set up the top-level frame Line 490  heap storage. Set up the top-level frame
490  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
491    
492  #ifdef NO_RECURSE  #ifdef NO_RECURSE
493  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));
494    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
495  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
496    
497  /* Copy in the original argument variables */  /* Copy in the original argument variables */
498    
499  frame->Xeptr = eptr;  frame->Xeptr = eptr;
500  frame->Xecode = ecode;  frame->Xecode = ecode;
501    frame->Xmstart = mstart;
502  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
503  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
504  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
505    
506  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 408  HEAP_RECURSE: Line 511  HEAP_RECURSE:
511    
512  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
513  #define ecode              frame->Xecode  #define ecode              frame->Xecode
514    #define mstart             frame->Xmstart
515  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
516  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
517  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
518    
519  /* Ditto for the local variables */  /* Ditto for the local variables */
520    
521  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
522  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
523  #endif  #endif
524  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
525    #define codelink           frame->Xcodelink
526  #define data               frame->Xdata  #define data               frame->Xdata
527  #define next               frame->Xnext  #define next               frame->Xnext
528  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 432  HEAP_RECURSE: Line 535  HEAP_RECURSE:
535  #define condition          frame->Xcondition  #define condition          frame->Xcondition
536  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
537    
 #define original_ims       frame->Xoriginal_ims  
   
538  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
539  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
540  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
541  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
542  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
543  #define prop_chartype      frame->Xprop_chartype  #define occhars            frame->Xocchars
 #define prop_script        frame->Xprop_script  
544  #endif  #endif
545    
546  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 468  i, and fc and c, can be the same variabl Line 568  i, and fc and c, can be the same variabl
568  #define fi i  #define fi i
569  #define fc c  #define fc c
570    
571    /* Many of the following variables are used only in small blocks of the code.
572    My normal style of coding would have declared them within each of those blocks.
573    However, in order to accommodate the version of this code that uses an external
574    "stack" implemented on the heap, it is easier to declare them all here, so the
575    declarations can be cut out in a block. The only declarations within blocks
576    below are for variables that do not have to be preserved over a recursive call
577    to RMATCH(). */
578    
579    #ifdef SUPPORT_UTF
580    const pcre_uchar *charptr;
581    #endif
582    const pcre_uchar *callpat;
583    const pcre_uchar *data;
584    const pcre_uchar *next;
585    PCRE_PUCHAR       pp;
586    const pcre_uchar *prev;
587    PCRE_PUCHAR       saved_eptr;
588    
589    recursion_info new_recursive;
590    
591  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  BOOL cur_is_word;
 const uschar *charptr;             /* in small blocks of the code. My normal */  
 #endif                             /* style of coding would have declared    */  
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
592  BOOL condition;  BOOL condition;
593  BOOL prev_is_word;  BOOL prev_is_word;
594    
 unsigned long int original_ims;  
   
595  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
596  int prop_type;  int prop_type;
597  int prop_value;  int prop_value;
598  int prop_fail_result;  int prop_fail_result;
599  int prop_category;  int oclength;
600  int prop_chartype;  pcre_uchar occhars[6];
 int prop_script;  
601  #endif  #endif
602    
603    int codelink;
604  int ctype;  int ctype;
605  int length;  int length;
606  int max;  int max;
# Line 510  int stacksave[REC_STACK_SAVE_MAX]; Line 615  int stacksave[REC_STACK_SAVE_MAX];
615  eptrblock newptrb;  eptrblock newptrb;
616  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
617    
618    /* To save space on the stack and in the heap frame, I have doubled up on some
619    of the local variables that are used only in localised parts of the code, but
620    still need to be preserved over recursive calls of match(). These macros define
621    the alternative names that are used. */
622    
623    #define allow_zero    cur_is_word
624    #define cbegroup      condition
625    #define code_offset   codelink
626    #define condassert    condition
627    #define matched_once  prev_is_word
628    #define foc           number
629    #define save_mark     data
630    
631  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
632  variables. */  variables. */
633    
# Line 529  TAIL_RECURSE: Line 647  TAIL_RECURSE:
647  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
648  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
649  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
650  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
651  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
652  complicated macro. It has to be used in one particular way. This shouldn't,  complicated macro. It has to be used in one particular way. This shouldn't,
653  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
654    
655    #ifdef SUPPORT_UTF
656    utf = md->utf;       /* Local copy of the flag */
657    #else
658    utf = FALSE;
659    #endif
660    
661  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
662  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
663    
664  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
665  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
666    
 original_ims = ims;    /* Save for resetting on ')' */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
667  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
668  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
669  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
670  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
671  When match() is called in other circumstances, don't add to the chain. If this  
672  is a tail recursion, use a block from the workspace, as the one on the stack is  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
673  already used. */  such remembered pointers, to be checked when we hit the closing ket, in order
674    to break infinite loops that match no characters. When match() is called in
675    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
676    NOT be used with tail recursion, because the memory block that is used is on
677    the stack, so a new one may be required for each match(). */
678    
679  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
680    {    {
681    eptrblock *p;    newptrb.epb_saved_eptr = eptr;
682    if ((flags & match_tail_recursed) != 0)    newptrb.epb_prev = eptrb;
683      {    eptrb = &newptrb;
684      if (md->eptrn >= EPTR_WORK_SIZE) RRETURN(PCRE_ERROR_NULLWSLIMIT);    md->match_function_type = 0;
     p = md->eptrchain + md->eptrn++;  
     }  
   else p = &newptrb;  
   p->epb_saved_eptr = eptr;  
   p->epb_prev = eptrb;  
   eptrb = p;  
685    }    }
686    
687  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 577  for (;;) Line 691  for (;;)
691    minimize = possessive = FALSE;    minimize = possessive = FALSE;
692    op = *ecode;    op = *ecode;
693    
   /* For partial matching, remember if we ever hit the end of the subject after  
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > md->start_match)  
     md->hitend = TRUE;  
   
694    switch(op)    switch(op)
695      {      {
696      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
697      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
698      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
699      set from a previous iteration of this group, and be referred to by a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
700      reference inside the group.        eptrb, RM55);
701        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
702      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
703      values of the final offsets, in case they were set by a previous iteration  
704      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
705        argument, and we must check whether that argument matches this MARK's
706        argument. It is passed back in md->start_match_ptr (an overloading of that
707        variable). If it does match, we reset that variable to the current subject
708        position and return MATCH_SKIP. Otherwise, pass back the return code
709        unaltered. */
710    
711        else if (rrc == MATCH_SKIP_ARG &&
712            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
713          {
714          md->start_match_ptr = eptr;
715          RRETURN(MATCH_SKIP);
716          }
717        RRETURN(rrc);
718    
719        case OP_FAIL:
720        RRETURN(MATCH_NOMATCH);
721    
722        /* COMMIT overrides PRUNE, SKIP, and THEN */
723    
724        case OP_COMMIT:
725        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
726          eptrb, RM52);
727        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
728            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
729            rrc != MATCH_THEN)
730          RRETURN(rrc);
731        RRETURN(MATCH_COMMIT);
732    
733        /* PRUNE overrides THEN */
734    
735        case OP_PRUNE:
736        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
737          eptrb, RM51);
738        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
739        RRETURN(MATCH_PRUNE);
740    
741        case OP_PRUNE_ARG:
742        md->nomatch_mark = ecode + 2;
743        md->mark = NULL;    /* In case previously set by assertion */
744        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
745          eptrb, RM56);
746        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
747             md->mark == NULL) md->mark = ecode + 2;
748        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
749        RRETURN(MATCH_PRUNE);
750    
751        /* SKIP overrides PRUNE and THEN */
752    
753        case OP_SKIP:
754        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
755          eptrb, RM53);
756        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
757          RRETURN(rrc);
758        md->start_match_ptr = eptr;   /* Pass back current position */
759        RRETURN(MATCH_SKIP);
760    
761        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
762        nomatch_mark. There is a flag that disables this opcode when re-matching a
763        pattern that ended with a SKIP for which there was not a matching MARK. */
764    
765        case OP_SKIP_ARG:
766        if (md->ignore_skip_arg)
767          {
768          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
769          break;
770          }
771        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
772          eptrb, RM57);
773        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
774          RRETURN(rrc);
775    
776        /* Pass back the current skip name by overloading md->start_match_ptr and
777        returning the special MATCH_SKIP_ARG return code. This will either be
778        caught by a matching MARK, or get to the top, where it causes a rematch
779        with the md->ignore_skip_arg flag set. */
780    
781        md->start_match_ptr = ecode + 2;
782        RRETURN(MATCH_SKIP_ARG);
783    
784        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
785        the branch in which it occurs can be determined. Overload the start of
786        match pointer to do this. */
787    
788        case OP_THEN:
789        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
790          eptrb, RM54);
791        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
792        md->start_match_ptr = ecode;
793        RRETURN(MATCH_THEN);
794    
795        case OP_THEN_ARG:
796        md->nomatch_mark = ecode + 2;
797        md->mark = NULL;    /* In case previously set by assertion */
798        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
799          md, eptrb, RM58);
800        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
801             md->mark == NULL) md->mark = ecode + 2;
802        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
803        md->start_match_ptr = ecode;
804        RRETURN(MATCH_THEN);
805    
806        /* Handle an atomic group that does not contain any capturing parentheses.
807        This can be handled like an assertion. Prior to 8.13, all atomic groups
808        were handled this way. In 8.13, the code was changed as below for ONCE, so
809        that backups pass through the group and thereby reset captured values.
810        However, this uses a lot more stack, so in 8.20, atomic groups that do not
811        contain any captures generate OP_ONCE_NC, which can be handled in the old,
812        less stack intensive way.
813    
814        Check the alternative branches in turn - the matching won't pass the KET
815        for this kind of subpattern. If any one branch matches, we carry on as at
816        the end of a normal bracket, leaving the subject pointer, but resetting
817        the start-of-match value in case it was changed by \K. */
818    
819        case OP_ONCE_NC:
820        prev = ecode;
821        saved_eptr = eptr;
822        save_mark = md->mark;
823        do
824          {
825          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
826          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
827            {
828            mstart = md->start_match_ptr;
829            break;
830            }
831          if (rrc == MATCH_THEN)
832            {
833            next = ecode + GET(ecode,1);
834            if (md->start_match_ptr < next &&
835                (*ecode == OP_ALT || *next == OP_ALT))
836              rrc = MATCH_NOMATCH;
837            }
838    
839          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
840          ecode += GET(ecode,1);
841          md->mark = save_mark;
842          }
843        while (*ecode == OP_ALT);
844    
845        /* If hit the end of the group (which could be repeated), fail */
846    
847        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
848    
849        /* Continue as from after the group, updating the offsets high water
850        mark, since extracts may have been taken. */
851    
852        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
853    
854        offset_top = md->end_offset_top;
855        eptr = md->end_match_ptr;
856    
857        /* For a non-repeating ket, just continue at this level. This also
858        happens for a repeating ket if no characters were matched in the group.
859        This is the forcible breaking of infinite loops as implemented in Perl
860        5.005. */
861    
862        if (*ecode == OP_KET || eptr == saved_eptr)
863          {
864          ecode += 1+LINK_SIZE;
865          break;
866          }
867    
868        /* The repeating kets try the rest of the pattern or restart from the
869        preceding bracket, in the appropriate order. The second "call" of match()
870        uses tail recursion, to avoid using another stack frame. */
871    
872        if (*ecode == OP_KETRMIN)
873          {
874          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
875          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
876          ecode = prev;
877          goto TAIL_RECURSE;
878          }
879        else  /* OP_KETRMAX */
880          {
881          md->match_function_type = MATCH_CBEGROUP;
882          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
883          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
884          ecode += 1 + LINK_SIZE;
885          goto TAIL_RECURSE;
886          }
887        /* Control never gets here */
888    
889        /* Handle a capturing bracket, other than those that are possessive with an
890        unlimited repeat. If there is space in the offset vector, save the current
891        subject position in the working slot at the top of the vector. We mustn't
892        change the current values of the data slot, because they may be set from a
893        previous iteration of this group, and be referred to by a reference inside
894        the group. A failure to match might occur after the group has succeeded,
895        if something later on doesn't match. For this reason, we need to restore
896        the working value and also the values of the final offsets, in case they
897        were set by a previous iteration of the same bracket.
898    
899      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
900      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 606  for (;;) Line 905  for (;;)
905      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
906      offset = number << 1;      offset = number << 1;
907    
908  #ifdef DEBUG  #ifdef PCRE_DEBUG
909      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
910      printf("subject=");      printf("subject=");
911      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 619  for (;;) Line 918  for (;;)
918        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
919        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
920        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
921          save_mark = md->mark;
922    
923        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
924        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
925            (int)(eptr - md->start_subject);
926    
927        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
928          {          {
929          RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
930            ims, eptrb, flags);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
931              eptrb, RM1);
932            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
933    
934            /* If we backed up to a THEN, check whether it is within the current
935            branch by comparing the address of the THEN that is passed back with
936            the end of the branch. If it is within the current branch, and the
937            branch is one of two or more alternatives (it either starts or ends
938            with OP_ALT), we have reached the limit of THEN's action, so convert
939            the return code to NOMATCH, which will cause normal backtracking to
940            happen from now on. Otherwise, THEN is passed back to an outer
941            alternative. This implements Perl's treatment of parenthesized groups,
942            where a group not containing | does not affect the current alternative,
943            that is, (X) is NOT the same as (X|(*F)). */
944    
945            if (rrc == MATCH_THEN)
946              {
947              next = ecode + GET(ecode,1);
948              if (md->start_match_ptr < next &&
949                  (*ecode == OP_ALT || *next == OP_ALT))
950                rrc = MATCH_NOMATCH;
951              }
952    
953            /* Anything other than NOMATCH is passed back. */
954    
955          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
956          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
957          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
958            md->mark = save_mark;
959            if (*ecode != OP_ALT) break;
960          }          }
       while (*ecode == OP_ALT);  
961    
962        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
963        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
964        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
965        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
966    
967        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
968    
969          RRETURN(rrc);
970        }        }
971    
972      /* Insufficient room for saving captured contents. Treat as a non-capturing      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
973      bracket. */      as a non-capturing bracket. */
974    
975        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
976        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
977    
978      DPRINTF(("insufficient capture room: treat as non-capturing\n"));      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
979    
980      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
981      final alternative within the brackets, we would return the result of a      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
982      recursive call to match() whatever happened. We can reduce stack usage by  
983      turning this into a tail recursion. */      /* Non-capturing or atomic group, except for possessive with unlimited
984        repeat and ONCE group with no captures. Loop for all the alternatives.
985    
986        When we get to the final alternative within the brackets, we used to return
987        the result of a recursive call to match() whatever happened so it was
988        possible to reduce stack usage by turning this into a tail recursion,
989        except in the case of a possibly empty group. However, now that there is
990        the possiblity of (*THEN) occurring in the final alternative, this
991        optimization is no longer always possible.
992    
993        We can optimize if we know there are no (*THEN)s in the pattern; at present
994        this is the best that can be done.
995    
996        MATCH_ONCE is returned when the end of an atomic group is successfully
997        reached, but subsequent matching fails. It passes back up the tree (causing
998        captured values to be reset) until the original atomic group level is
999        reached. This is tested by comparing md->once_target with the start of the
1000        group. At this point, the return is converted into MATCH_NOMATCH so that
1001        previous backup points can be taken. */
1002    
1003        case OP_ONCE:
1004      case OP_BRA:      case OP_BRA:
1005      case OP_SBRA:      case OP_SBRA:
1006      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1007      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1008      for (;;)      for (;;)
1009        {        {
1010        if (ecode[GET(ecode, 1)] != OP_ALT)        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1011    
1012          /* If this is not a possibly empty group, and there are no (*THEN)s in
1013          the pattern, and this is the final alternative, optimize as described
1014          above. */
1015    
1016          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1017          {          {
1018          ecode += _pcre_OP_lengths[*ecode];          ecode += PRIV(OP_lengths)[*ecode];
         flags |= match_tail_recursed;  
         DPRINTF(("bracket 0 tail recursion\n"));  
1019          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1020          }          }
1021    
1022        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* In all other cases, we have to make another call to match(). */
       otherwise return. */  
1023    
1024        RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1025          eptrb, flags);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1026        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RM2);
1027    
1028          /* See comment in the code for capturing groups above about handling
1029          THEN. */
1030    
1031          if (rrc == MATCH_THEN)
1032            {
1033            next = ecode + GET(ecode,1);
1034            if (md->start_match_ptr < next &&
1035                (*ecode == OP_ALT || *next == OP_ALT))
1036              rrc = MATCH_NOMATCH;
1037            }
1038    
1039          if (rrc != MATCH_NOMATCH)
1040            {
1041            if (rrc == MATCH_ONCE)
1042              {
1043              const pcre_uchar *scode = ecode;
1044              if (*scode != OP_ONCE)           /* If not at start, find it */
1045                {
1046                while (*scode == OP_ALT) scode += GET(scode, 1);
1047                scode -= GET(scode, 1);
1048                }
1049              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1050              }
1051            RRETURN(rrc);
1052            }
1053        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1054          md->mark = save_mark;
1055          if (*ecode != OP_ALT) break;
1056        }        }
     /* Control never reaches here. */  
1057    
1058      /* Conditional group: compilation checked that there are no more than      RRETURN(MATCH_NOMATCH);
     two branches. If the condition is false, skipping the first branch takes us  
     past the end if there is only one branch, but that's OK because that is  
     exactly what going to the ket would do. As there is only one branch to be  
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1059    
1060      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1061      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1062      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1063        {      different. The end of these brackets will always be OP_KETRPOS, which
1064        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1065        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1066          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1067    
1068      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      case OP_CBRAPOS:
1069        {      case OP_SCBRAPOS:
1070        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      allow_zero = FALSE;
       condition = offset < offset_top && md->offset_vector[offset] >= 0;  
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1071    
1072      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      POSSESSIVE_CAPTURE:
1073        {      number = GET2(ecode, 1+LINK_SIZE);
1074        condition = FALSE;      offset = number << 1;
       ecode += GET(ecode, 1);  
       }  
1075    
1076      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1077      the final argument match_condassert causes it to stop at the end of an      printf("start possessive bracket %d\n", number);
1078      assertion. */      printf("subject=");
1079        pchars(eptr, 16, TRUE, md);
1080        printf("\n");
1081    #endif
1082    
1083      else      if (offset < md->offset_max)
1084        {        {
1085        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1086            match_condassert);        code_offset = (int)(ecode - md->start_code);
1087        if (rrc == MATCH_MATCH)  
1088          {        save_offset1 = md->offset_vector[offset];
1089          condition = TRUE;        save_offset2 = md->offset_vector[offset+1];
1090          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        save_offset3 = md->offset_vector[md->offset_end - number];
1091          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
1092    
1093          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1094    
1095          /* Each time round the loop, save the current subject position for use
1096          when the group matches. For MATCH_MATCH, the group has matched, so we
1097          restart it with a new subject starting position, remembering that we had
1098          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1099          usual. If we haven't matched any alternatives in any iteration, check to
1100          see if a previous iteration matched. If so, the group has matched;
1101          continue from afterwards. Otherwise it has failed; restore the previous
1102          capture values before returning NOMATCH. */
1103    
1104          for (;;)
1105            {
1106            md->offset_vector[md->offset_end - number] =
1107              (int)(eptr - md->start_subject);
1108            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1109            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1110              eptrb, RM63);
1111            if (rrc == MATCH_KETRPOS)
1112              {
1113              offset_top = md->end_offset_top;
1114              eptr = md->end_match_ptr;
1115              ecode = md->start_code + code_offset;
1116              save_capture_last = md->capture_last;
1117              matched_once = TRUE;
1118              continue;
1119              }
1120    
1121            /* See comment in the code for capturing groups above about handling
1122            THEN. */
1123    
1124            if (rrc == MATCH_THEN)
1125              {
1126              next = ecode + GET(ecode,1);
1127              if (md->start_match_ptr < next &&
1128                  (*ecode == OP_ALT || *next == OP_ALT))
1129                rrc = MATCH_NOMATCH;
1130              }
1131    
1132            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1133            md->capture_last = save_capture_last;
1134            ecode += GET(ecode, 1);
1135            if (*ecode != OP_ALT) break;
1136          }          }
1137        else if (rrc != MATCH_NOMATCH)  
1138          if (!matched_once)
1139          {          {
1140          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1141            md->offset_vector[offset+1] = save_offset2;
1142            md->offset_vector[md->offset_end - number] = save_offset3;
1143          }          }
1144        else  
1145          if (allow_zero || matched_once)
1146          {          {
1147          condition = FALSE;          ecode += 1 + LINK_SIZE;
1148          ecode += GET(ecode, 1);          break;
1149          }          }
       }  
1150    
1151      /* We are now at the branch that is to be obeyed. As there is only one,        RRETURN(MATCH_NOMATCH);
     we can use tail recursion to avoid using another stack frame. If the second  
     alternative doesn't exist, we can just plough on. */  
   
     if (condition || *ecode == OP_ALT)  
       {  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed | ((op == OP_SCOND)? match_cbegroup : 0);  
       goto TAIL_RECURSE;  
       }  
     else  
       {  
       ecode += 1 + LINK_SIZE;  
1152        }        }
     break;  
   
1153    
1154      /* End of the pattern. If we are in a top-level recursion, we should      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1155      restore the offsets appropriately and continue from after the call. */      as a non-capturing bracket. */
1156    
1157      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1158      if (md->recursive != NULL && md->recursive->group_num == 0)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       md->start_match = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1159    
1160      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     string - backtracking will then try other alternatives, if any. */  
1161    
1162      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1163      md->end_match_ptr = eptr;          /* Record where we ended */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     md->end_offset_top = offset_top;   /* and how many extracts were taken */  
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1164    
1165      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1166      matching won't pass the KET for an assertion. If any one branch matches,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1167      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1168      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1169      this level is identical to the lookahead case. */  
1170        case OP_BRAPOS:
1171        case OP_SBRAPOS:
1172        allow_zero = FALSE;
1173    
1174        POSSESSIVE_NON_CAPTURE:
1175        matched_once = FALSE;
1176        code_offset = (int)(ecode - md->start_code);
1177    
1178      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1179        {        {
1180        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1181        if (rrc == MATCH_MATCH) break;        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1182            eptrb, RM48);
1183          if (rrc == MATCH_KETRPOS)
1184            {
1185            offset_top = md->end_offset_top;
1186            eptr = md->end_match_ptr;
1187            ecode = md->start_code + code_offset;
1188            matched_once = TRUE;
1189            continue;
1190            }
1191    
1192          /* See comment in the code for capturing groups above about handling
1193          THEN. */
1194    
1195          if (rrc == MATCH_THEN)
1196            {
1197            next = ecode + GET(ecode,1);
1198            if (md->start_match_ptr < next &&
1199                (*ecode == OP_ALT || *next == OP_ALT))
1200              rrc = MATCH_NOMATCH;
1201            }
1202    
1203        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1204        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1205          if (*ecode != OP_ALT) break;
1206        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1207    
1208      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1209          {
1210          ecode += 1 + LINK_SIZE;
1211          break;
1212          }
1213        RRETURN(MATCH_NOMATCH);
1214    
1215      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1216    
1217      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1218      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1219        past the end if there is only one branch, but that's OK because that is
1220        exactly what going to the ket would do. */
1221    
1222      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1223      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1224      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1225    
1226      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1227        inserted between OP_COND and an assertion condition. */
1228    
1229      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1230        {        {
1231        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (PUBL(callout) != NULL)
1232        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          {
1233        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          PUBL(callout_block) cb;
1234        ecode += GET(ecode,1);          cb.version          = 2;   /* Version 1 of the callout block */
1235            cb.callout_number   = ecode[LINK_SIZE+2];
1236            cb.offset_vector    = md->offset_vector;
1237    #ifdef COMPILE_PCRE8
1238            cb.subject          = (PCRE_SPTR)md->start_subject;
1239    #else
1240            cb.subject          = (PCRE_SPTR16)md->start_subject;
1241    #endif
1242            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1243            cb.start_match      = (int)(mstart - md->start_subject);
1244            cb.current_position = (int)(eptr - md->start_subject);
1245            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1246            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1247            cb.capture_top      = offset_top/2;
1248            cb.capture_last     = md->capture_last;
1249            cb.callout_data     = md->callout_data;
1250            cb.mark             = md->nomatch_mark;
1251            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1252            if (rrc < 0) RRETURN(rrc);
1253            }
1254          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1255        }        }
     while (*ecode == OP_ALT);  
   
     if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);  
1256    
1257      ecode += 1 + LINK_SIZE;      condcode = ecode[LINK_SIZE+1];
     continue;  
1258    
1259      /* Move the subject pointer back. This occurs only at the start of      /* Now see what the actual condition is */
     each branch of a lookbehind assertion. If we are too close to the start to  
     move back, this match function fails. When working with UTF-8 we move  
     back a number of characters, not bytes. */  
1260    
1261      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1262        {        {
1263        i = GET(ecode, 1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
1264            {
1265            condition = FALSE;
1266            ecode += GET(ecode, 1);
1267            }
1268          else
1269            {
1270            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1271            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1272    
1273            /* If the test is for recursion into a specific subpattern, and it is
1274            false, but the test was set up by name, scan the table to see if the
1275            name refers to any other numbers, and test them. The condition is true
1276            if any one is set. */
1277    
1278            if (!condition && condcode == OP_NRREF)
1279              {
1280              pcre_uchar *slotA = md->name_table;
1281              for (i = 0; i < md->name_count; i++)
1282                {
1283                if (GET2(slotA, 0) == recno) break;
1284                slotA += md->name_entry_size;
1285                }
1286    
1287              /* Found a name for the number - there can be only one; duplicate
1288              names for different numbers are allowed, but not vice versa. First
1289              scan down for duplicates. */
1290    
1291              if (i < md->name_count)
1292                {
1293                pcre_uchar *slotB = slotA;
1294                while (slotB > md->name_table)
1295                  {
1296                  slotB -= md->name_entry_size;
1297                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1298                    {
1299                    condition = GET2(slotB, 0) == md->recursive->group_num;
1300                    if (condition) break;
1301                    }
1302                  else break;
1303                  }
1304    
1305                /* Scan up for duplicates */
1306    
1307                if (!condition)
1308                  {
1309                  slotB = slotA;
1310                  for (i++; i < md->name_count; i++)
1311                    {
1312                    slotB += md->name_entry_size;
1313                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1314                      {
1315                      condition = GET2(slotB, 0) == md->recursive->group_num;
1316                      if (condition) break;
1317                      }
1318                    else break;
1319                    }
1320                  }
1321                }
1322              }
1323    
1324            /* Chose branch according to the condition */
1325    
1326            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1327            }
1328          }
1329    
1330        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1331          {
1332          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1333          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1334    
1335          /* If the numbered capture is unset, but the reference was by name,
1336          scan the table to see if the name refers to any other numbers, and test
1337          them. The condition is true if any one is set. This is tediously similar
1338          to the code above, but not close enough to try to amalgamate. */
1339    
1340          if (!condition && condcode == OP_NCREF)
1341            {
1342            int refno = offset >> 1;
1343            pcre_uchar *slotA = md->name_table;
1344    
1345            for (i = 0; i < md->name_count; i++)
1346              {
1347              if (GET2(slotA, 0) == refno) break;
1348              slotA += md->name_entry_size;
1349              }
1350    
1351            /* Found a name for the number - there can be only one; duplicate names
1352            for different numbers are allowed, but not vice versa. First scan down
1353            for duplicates. */
1354    
1355            if (i < md->name_count)
1356              {
1357              pcre_uchar *slotB = slotA;
1358              while (slotB > md->name_table)
1359                {
1360                slotB -= md->name_entry_size;
1361                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1362                  {
1363                  offset = GET2(slotB, 0) << 1;
1364                  condition = offset < offset_top &&
1365                    md->offset_vector[offset] >= 0;
1366                  if (condition) break;
1367                  }
1368                else break;
1369                }
1370    
1371              /* Scan up for duplicates */
1372    
1373              if (!condition)
1374                {
1375                slotB = slotA;
1376                for (i++; i < md->name_count; i++)
1377                  {
1378                  slotB += md->name_entry_size;
1379                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1380                    {
1381                    offset = GET2(slotB, 0) << 1;
1382                    condition = offset < offset_top &&
1383                      md->offset_vector[offset] >= 0;
1384                    if (condition) break;
1385                    }
1386                  else break;
1387                  }
1388                }
1389              }
1390            }
1391    
1392          /* Chose branch according to the condition */
1393    
1394          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1395          }
1396    
1397        else if (condcode == OP_DEF)     /* DEFINE - always false */
1398          {
1399          condition = FALSE;
1400          ecode += GET(ecode, 1);
1401          }
1402    
1403        /* The condition is an assertion. Call match() to evaluate it - setting
1404        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1405        an assertion. */
1406    
1407        else
1408          {
1409          md->match_function_type = MATCH_CONDASSERT;
1410          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1411          if (rrc == MATCH_MATCH)
1412            {
1413            if (md->end_offset_top > offset_top)
1414              offset_top = md->end_offset_top;  /* Captures may have happened */
1415            condition = TRUE;
1416            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1417            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1418            }
1419    
1420          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1421          assertion; it is therefore treated as NOMATCH. */
1422    
1423          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1424            {
1425            RRETURN(rrc);         /* Need braces because of following else */
1426            }
1427          else
1428            {
1429            condition = FALSE;
1430            ecode += codelink;
1431            }
1432          }
1433    
1434        /* We are now at the branch that is to be obeyed. As there is only one, can
1435        use tail recursion to avoid using another stack frame, except when there is
1436        unlimited repeat of a possibly empty group. In the latter case, a recursive
1437        call to match() is always required, unless the second alternative doesn't
1438        exist, in which case we can just plough on. Note that, for compatibility
1439        with Perl, the | in a conditional group is NOT treated as creating two
1440        alternatives. If a THEN is encountered in the branch, it propagates out to
1441        the enclosing alternative (unless nested in a deeper set of alternatives,
1442        of course). */
1443    
1444        if (condition || *ecode == OP_ALT)
1445          {
1446          if (op != OP_SCOND)
1447            {
1448            ecode += 1 + LINK_SIZE;
1449            goto TAIL_RECURSE;
1450            }
1451    
1452          md->match_function_type = MATCH_CBEGROUP;
1453          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1454          RRETURN(rrc);
1455          }
1456    
1457         /* Condition false & no alternative; continue after the group. */
1458    
1459        else
1460          {
1461          ecode += 1 + LINK_SIZE;
1462          }
1463        break;
1464    
1465    
1466        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1467        to close any currently open capturing brackets. */
1468    
1469        case OP_CLOSE:
1470        number = GET2(ecode, 1);
1471        offset = number << 1;
1472    
1473    #ifdef PCRE_DEBUG
1474          printf("end bracket %d at *ACCEPT", number);
1475          printf("\n");
1476    #endif
1477    
1478        md->capture_last = number;
1479        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1480          {
1481          md->offset_vector[offset] =
1482            md->offset_vector[md->offset_end - number];
1483          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1484          if (offset_top <= offset) offset_top = offset + 2;
1485          }
1486        ecode += 1 + IMM2_SIZE;
1487        break;
1488    
1489    
1490        /* End of the pattern, either real or forced. */
1491    
1492        case OP_END:
1493        case OP_ACCEPT:
1494        case OP_ASSERT_ACCEPT:
1495    
1496        /* If we have matched an empty string, fail if not in an assertion and not
1497        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1498        is set and we have matched at the start of the subject. In both cases,
1499        backtracking will then try other alternatives, if any. */
1500    
1501        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1502             md->recursive == NULL &&
1503             (md->notempty ||
1504               (md->notempty_atstart &&
1505                 mstart == md->start_subject + md->start_offset)))
1506          RRETURN(MATCH_NOMATCH);
1507    
1508        /* Otherwise, we have a match. */
1509    
1510        md->end_match_ptr = eptr;           /* Record where we ended */
1511        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1512        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1513    
1514        /* For some reason, the macros don't work properly if an expression is
1515        given as the argument to RRETURN when the heap is in use. */
1516    
1517        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1518        RRETURN(rrc);
1519    
1520        /* Assertion brackets. Check the alternative branches in turn - the
1521        matching won't pass the KET for an assertion. If any one branch matches,
1522        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1523        start of each branch to move the current point backwards, so the code at
1524        this level is identical to the lookahead case. When the assertion is part
1525        of a condition, we want to return immediately afterwards. The caller of
1526        this incarnation of the match() function will have set MATCH_CONDASSERT in
1527        md->match_function type, and one of these opcodes will be the first opcode
1528        that is processed. We use a local variable that is preserved over calls to
1529        match() to remember this case. */
1530    
1531        case OP_ASSERT:
1532        case OP_ASSERTBACK:
1533        if (md->match_function_type == MATCH_CONDASSERT)
1534          {
1535          condassert = TRUE;
1536          md->match_function_type = 0;
1537          }
1538        else condassert = FALSE;
1539    
1540        do
1541          {
1542          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1543          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1544            {
1545            mstart = md->start_match_ptr;   /* In case \K reset it */
1546            break;
1547            }
1548    
1549          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1550          as NOMATCH. */
1551    
1552          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1553          ecode += GET(ecode, 1);
1554          }
1555        while (*ecode == OP_ALT);
1556    
1557        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1558    
1559        /* If checking an assertion for a condition, return MATCH_MATCH. */
1560    
1561        if (condassert) RRETURN(MATCH_MATCH);
1562    
1563        /* Continue from after the assertion, updating the offsets high water
1564        mark, since extracts may have been taken during the assertion. */
1565    
1566        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1567        ecode += 1 + LINK_SIZE;
1568        offset_top = md->end_offset_top;
1569        continue;
1570    
1571        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1572        PRUNE, or COMMIT means we must assume failure without checking subsequent
1573        branches. */
1574    
1575        case OP_ASSERT_NOT:
1576        case OP_ASSERTBACK_NOT:
1577        if (md->match_function_type == MATCH_CONDASSERT)
1578          {
1579          condassert = TRUE;
1580          md->match_function_type = 0;
1581          }
1582        else condassert = FALSE;
1583    
1584        do
1585          {
1586          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1587          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1588          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1589            {
1590            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1591            break;
1592            }
1593    
1594          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1595          as NOMATCH. */
1596    
1597          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1598          ecode += GET(ecode,1);
1599          }
1600        while (*ecode == OP_ALT);
1601    
1602        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1603    
1604        ecode += 1 + LINK_SIZE;
1605        continue;
1606    
1607        /* Move the subject pointer back. This occurs only at the start of
1608        each branch of a lookbehind assertion. If we are too close to the start to
1609        move back, this match function fails. When working with UTF-8 we move
1610        back a number of characters, not bytes. */
1611    
1612        case OP_REVERSE:
1613    #ifdef SUPPORT_UTF
1614        if (utf)
1615          {
1616          i = GET(ecode, 1);
1617        while (i-- > 0)        while (i-- > 0)
1618          {          {
1619          eptr--;          eptr--;
1620          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1621          BACKCHAR(eptr)          BACKCHAR(eptr);
1622          }          }
1623        }        }
1624      else      else
# Line 856  for (;;) Line 1631  for (;;)
1631        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1632        }        }
1633    
1634      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1635    
1636        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1637      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1638      break;      break;
1639    
# Line 866  for (;;) Line 1642  for (;;)
1642      function is able to force a failure. */      function is able to force a failure. */
1643    
1644      case OP_CALLOUT:      case OP_CALLOUT:
1645      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1646        {        {
1647        pcre_callout_block cb;        PUBL(callout_block) cb;
1648        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1649        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1650        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1651    #ifdef COMPILE_PCRE8
1652        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1653        cb.subject_length   = md->end_subject - md->start_subject;  #else
1654        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1655        cb.current_position = eptr - md->start_subject;  #endif
1656          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1657          cb.start_match      = (int)(mstart - md->start_subject);
1658          cb.current_position = (int)(eptr - md->start_subject);
1659        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1660        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1661        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1662        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1663        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1664        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1665          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1666        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1667        }        }
1668      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 891  for (;;) Line 1672  for (;;)
1672      offset data is the offset to the starting bracket from the start of the      offset data is the offset to the starting bracket from the start of the
1673      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1674    
1675      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1676      save their starting points and reinstate them after the recursion. However,      re-instated afterwards. We don't know how many are started and not yet
1677      we don't know how many such there are (offset_top records the completed      finished (offset_top records the completed total) so we just have to save
1678      total) so we just have to save all the potential data. There may be up to      all the potential data. There may be up to 65535 such values, which is too
1679      65535 such values, which is too large to put on the stack, but using malloc      large to put on the stack, but using malloc for small numbers seems
1680      for small numbers seems expensive. As a compromise, the stack is used when      expensive. As a compromise, the stack is used when there are no more than
1681      there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc      REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
     is used. A problem is what to do if the malloc fails ... there is no way of  
     returning to the top level with an error. Save the top REC_STACK_SAVE_MAX  
     values on the stack, and accept that the rest may be wrong.  
1682    
1683      There are also other values that have to be saved. We use a chained      There are also other values that have to be saved. We use a chained
1684      sequence of blocks that actually live on the stack. Thanks to Robin Houston      sequence of blocks that actually live on the stack. Thanks to Robin Houston
1685      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1686        a lot, so he is not to blame for the current way it works. */
1687    
1688      case OP_RECURSE:      case OP_RECURSE:
1689        {        {
1690          recursion_info *ri;
1691          int recno;
1692    
1693        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1694        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1695          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1696    
1697          /* Check for repeating a recursion without advancing the subject pointer.
1698          This should catch convoluted mutual recursions. (Some simple cases are
1699          caught at compile time.) */
1700    
1701          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1702            if (recno == ri->group_num && eptr == ri->subject_position)
1703              RRETURN(PCRE_ERROR_RECURSELOOP);
1704    
1705        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1706    
1707          new_recursive.group_num = recno;
1708          new_recursive.subject_position = eptr;
1709        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1710        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1711    
1712        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1713    
1714        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1715    
1716        /* Now save the offset data. */        /* Now save the offset data */
1717    
1718        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1719        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 930  for (;;) Line 1721  for (;;)
1721        else        else
1722          {          {
1723          new_recursive.offset_save =          new_recursive.offset_save =
1724            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1725          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1726          }          }
   
1727        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1728              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1729    
1730        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1731        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1732          might be changed, so reset it before looping. */
1733    
1734        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1735        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1736        do        do
1737          {          {
1738          RMATCH(rrc, eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1739            md, ims, eptrb, flags);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1740          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1741            memcpy(md->offset_vector, new_recursive.offset_save,
1742                new_recursive.saved_max * sizeof(int));
1743            md->recursive = new_recursive.prevrec;
1744            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1745            {            {
1746            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1747            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1748              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1749            RRETURN(MATCH_MATCH);  
1750              /* Set where we got to in the subject, and reset the start in case
1751              it was changed by \K. This *is* propagated back out of a recursion,
1752              for Perl compatibility. */
1753    
1754              eptr = md->end_match_ptr;
1755              mstart = md->start_match_ptr;
1756              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1757            }            }
1758          else if (rrc != MATCH_NOMATCH)  
1759            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1760            as NOMATCH. */
1761    
1762            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1763            {            {
1764            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1765              if (new_recursive.offset_save != stacksave)
1766                (PUBL(free))(new_recursive.offset_save);
1767            RRETURN(rrc);            RRETURN(rrc);
1768            }            }
1769    
1770          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1771          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1772          }          }
1773        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 972  for (;;) Line 1775  for (;;)
1775        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1776        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1777        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1778          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1779        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1780        }        }
     /* Control never reaches here */  
1781    
1782      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1783      the point in the subject string is not moved back. Thus there can never be      break;
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
         eptrb, 0);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_cbegroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1784    
1785      /* 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
1786      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1059  for (;;) Line 1789  for (;;)
1789      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1790      break;      break;
1791    
1792      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1793      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
1794      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
1795      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
1796      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1797    
1798      case OP_BRAZERO:      case OP_BRAZERO:
1799        {      next = ecode + 1;
1800        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1801        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1802        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1803        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1804      break;      break;
1805    
1806      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1807        {      next = ecode + 1;
1808        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1809        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1810        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1811        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
1812        ecode++;      break;
1813        }  
1814        case OP_SKIPZERO:
1815        next = ecode+1;
1816        do next += GET(next,1); while (*next == OP_ALT);
1817        ecode = next + 1 + LINK_SIZE;
1818      break;      break;
1819    
1820        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1821        here; just jump to the group, with allow_zero set TRUE. */
1822    
1823        case OP_BRAPOSZERO:
1824        op = *(++ecode);
1825        allow_zero = TRUE;
1826        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1827          goto POSSESSIVE_NON_CAPTURE;
1828    
1829      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1830    
1831      case OP_KET:      case OP_KET:
1832      case OP_KETRMIN:      case OP_KETRMIN:
1833      case OP_KETRMAX:      case OP_KETRMAX:
1834        case OP_KETRPOS:
1835      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1836    
1837      /* If this was a group that remembered the subject start, in order to break      /* If this was a group that remembered the subject start, in order to break
1838      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1839      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1840    
1841      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1842        {        {
1843        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1844        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1845        }        }
1846      else saved_eptr = NULL;      else saved_eptr = NULL;
1847    
1848      /* If we are at the end of an assertion group, stop matching and return      /* If we are at the end of an assertion group or a non-capturing atomic
1849      MATCH_MATCH, but record the current high water mark for use by positive      group, stop matching and return MATCH_MATCH, but record the current high
1850      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1851        start in case it was changed by \K. */
1852      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1853          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1854          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1855        {        {
1856        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1857        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1858        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1859          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1860        }        }
1861    
1862      /* For capturing groups we have to check the group number back at the start      /* For capturing groups we have to check the group number back at the start
1863      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1864      bumping the high water mark. Note that whole-pattern recursion is coded as      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1865      a recurse into group 0, so it won't be picked up here. Instead, we catch it      into group 0, so it won't be picked up here. Instead, we catch it when the
1866      when the OP_END is reached. Other recursion is handled here. */      OP_END is reached. Other recursion is handled here. We just have to record
1867        the current subject position and start match pointer and give a MATCH
1868        return. */
1869    
1870      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1871            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1872        {        {
1873        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1874        offset = number << 1;        offset = number << 1;
1875    
1876  #ifdef DEBUG  #ifdef PCRE_DEBUG
1877        printf("end bracket %d", number);        printf("end bracket %d", number);
1878        printf("\n");        printf("\n");
1879  #endif  #endif
1880    
1881          /* Handle a recursively called group. */
1882    
1883          if (md->recursive != NULL && md->recursive->group_num == number)
1884            {
1885            md->end_match_ptr = eptr;
1886            md->start_match_ptr = mstart;
1887            RRETURN(MATCH_MATCH);
1888            }
1889    
1890          /* Deal with capturing */
1891    
1892        md->capture_last = number;        md->capture_last = number;
1893        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1894          {          {
1895            /* If offset is greater than offset_top, it means that we are
1896            "skipping" a capturing group, and that group's offsets must be marked
1897            unset. In earlier versions of PCRE, all the offsets were unset at the
1898            start of matching, but this doesn't work because atomic groups and
1899            assertions can cause a value to be set that should later be unset.
1900            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1901            part of the atomic group, but this is not on the final matching path,
1902            so must be unset when 2 is set. (If there is no group 2, there is no
1903            problem, because offset_top will then be 2, indicating no capture.) */
1904    
1905            if (offset > offset_top)
1906              {
1907              register int *iptr = md->offset_vector + offset_top;
1908              register int *iend = md->offset_vector + offset;
1909              while (iptr < iend) *iptr++ = -1;
1910              }
1911    
1912            /* Now make the extraction */
1913    
1914          md->offset_vector[offset] =          md->offset_vector[offset] =
1915            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1916          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1917          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1918          }          }
   
       /* 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;  
         }  
1919        }        }
1920    
1921      /* For both capturing and non-capturing groups, reset the value of the ims      /* For an ordinary non-repeating ket, just continue at this level. This
1922      flags, in case they got changed during the group. */      also happens for a repeating ket if no characters were matched in the
1923        group. This is the forcible breaking of infinite loops as implemented in
1924      ims = original_ims;      Perl 5.005. For a non-repeating atomic group that includes captures,
1925      DPRINTF(("ims reset to %02lx\n", ims));      establish a backup point by processing the rest of the pattern at a lower
1926        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1927      /* For a non-repeating ket, just continue at this level. This also      original OP_ONCE level, thereby bypassing intermediate backup points, but
1928      happens for a repeating ket if no characters were matched in the group.      resetting any captures that happened along the way. */
     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. */  
1929    
1930      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1931        {        {
1932        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1933            {
1934            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1935            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1936            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1937            RRETURN(MATCH_ONCE);
1938            }
1939          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1940        break;        break;
1941        }        }
1942    
1943      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1944      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1945      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
1946    
1947        if (*ecode == OP_KETRPOS)
1948          {
1949          md->end_match_ptr = eptr;
1950          md->end_offset_top = offset_top;
1951          RRETURN(MATCH_KETRPOS);
1952          }
1953    
1954      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      /* The normal repeating kets try the rest of the pattern or restart from
1955        the preceding bracket, in the appropriate order. In the second case, we can
1956        use tail recursion to avoid using another stack frame, unless we have an
1957        an atomic group or an unlimited repeat of a group that can match an empty
1958        string. */
1959    
1960      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1961        {        {
1962        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1963        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1964          if (*prev == OP_ONCE)
1965            {
1966            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1967            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1968            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1969            RRETURN(MATCH_ONCE);
1970            }
1971          if (*prev >= OP_SBRA)    /* Could match an empty string */
1972            {
1973            md->match_function_type = MATCH_CBEGROUP;
1974            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1975            RRETURN(rrc);
1976            }
1977        ecode = prev;        ecode = prev;
       flags |= match_tail_recursed;  
1978        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1979        }        }
1980      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1981        {        {
1982        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, flags);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1983          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1984          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1985        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1986          if (*prev == OP_ONCE)
1987            {
1988            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1989            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1990            md->once_target = prev;
1991            RRETURN(MATCH_ONCE);
1992            }
1993        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_tail_recursed;  
1994        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1995        }        }
1996      /* Control never gets here */      /* Control never gets here */
1997    
1998      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1999    
2000      case OP_CIRC:      case OP_CIRC:
2001      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2002    
2003      /* Start of subject assertion */      /* Start of subject assertion */
2004    
# Line 1221  for (;;) Line 2007  for (;;)
2007      ecode++;      ecode++;
2008      break;      break;
2009    
2010        /* Multiline mode: start of subject unless notbol, or after any newline. */
2011    
2012        case OP_CIRCM:
2013        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2014        if (eptr != md->start_subject &&
2015            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2016          RRETURN(MATCH_NOMATCH);
2017        ecode++;
2018        break;
2019    
2020      /* Start of match assertion */      /* Start of match assertion */
2021    
2022      case OP_SOM:      case OP_SOM:
# Line 1228  for (;;) Line 2024  for (;;)
2024      ecode++;      ecode++;
2025      break;      break;
2026    
2027      /* 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. */  
2028    
2029      case OP_DOLL:      case OP_SET_SOM:
2030      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2031        {      ecode++;
2032        if (eptr < md->end_subject)      break;
2033          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
2034        else      /* Multiline mode: assert before any newline, or before end of subject
2035          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
2036        ecode++;  
2037        break;      case OP_DOLLM:
2038        }      if (eptr < md->end_subject)
2039          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
2040      else      else
2041        {        {
2042        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2043        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2044        }        }
2045        ecode++;
2046        break;
2047    
2048        /* Not multiline mode: assert before a terminating newline or before end of
2049        subject unless noteol is set. */
2050    
2051        case OP_DOLL:
2052        if (md->noteol) RRETURN(MATCH_NOMATCH);
2053        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2054    
2055      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2056    
2057      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2058    
2059      case OP_EOD:      case OP_EOD:
2060      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2061        SCHECK_PARTIAL();
2062      ecode++;      ecode++;
2063      break;      break;
2064    
2065      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2066    
2067      case OP_EODN:      case OP_EODN:
2068      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2069        if (eptr < md->end_subject &&
2070          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2071        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2072    
2073        /* Either at end of string or \n before end. */
2074    
2075        SCHECK_PARTIAL();
2076      ecode++;      ecode++;
2077      break;      break;
2078    
# Line 1279  for (;;) Line 2084  for (;;)
2084    
2085        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2086        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
2087        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2088          partial matching. */
2089    
2090  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2091        if (utf8)        if (utf)
2092          {          {
2093            /* Get status of previous character */
2094    
2095          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2096            {            {
2097            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2098            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2099              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2100            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2101    #ifdef SUPPORT_UCP
2102              if (md->use_ucp)
2103                {
2104                if (c == '_') prev_is_word = TRUE; else
2105                  {
2106                  int cat = UCD_CATEGORY(c);
2107                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2108                  }
2109                }
2110              else
2111    #endif
2112            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2113            }            }
2114          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2115            /* Get status of next character */
2116    
2117            if (eptr >= md->end_subject)
2118              {
2119              SCHECK_PARTIAL();
2120              cur_is_word = FALSE;
2121              }
2122            else
2123            {            {
2124            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2125    #ifdef SUPPORT_UCP
2126              if (md->use_ucp)
2127                {
2128                if (c == '_') cur_is_word = TRUE; else
2129                  {
2130                  int cat = UCD_CATEGORY(c);
2131                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2132                  }
2133                }
2134              else
2135    #endif
2136            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2137            }            }
2138          }          }
2139        else        else
2140  #endif  #endif
2141    
2142        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2143          consistency with the behaviour of \w we do use it in this case. */
2144    
2145          {          {
2146          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2147            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2148          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2149            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2150              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2151    #ifdef SUPPORT_UCP
2152              if (md->use_ucp)
2153                {
2154                c = eptr[-1];
2155                if (c == '_') prev_is_word = TRUE; else
2156                  {
2157                  int cat = UCD_CATEGORY(c);
2158                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2159                  }
2160                }
2161              else
2162    #endif
2163              prev_is_word = MAX_255(eptr[-1])
2164                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2165              }
2166    
2167            /* Get status of next character */
2168    
2169            if (eptr >= md->end_subject)
2170              {
2171              SCHECK_PARTIAL();
2172              cur_is_word = FALSE;
2173              }
2174            else
2175    #ifdef SUPPORT_UCP
2176            if (md->use_ucp)
2177              {
2178              c = *eptr;
2179              if (c == '_') cur_is_word = TRUE; else
2180                {
2181                int cat = UCD_CATEGORY(c);
2182                cur_is_word = (cat == ucp_L || cat == ucp_N);
2183                }
2184              }
2185            else
2186    #endif
2187            cur_is_word = MAX_255(*eptr)
2188              && ((md->ctypes[*eptr] & ctype_word) != 0);
2189          }          }
2190    
2191        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1320  for (;;) Line 2199  for (;;)
2199      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2200    
2201      case OP_ANY:      case OP_ANY:
2202      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2203        {      /* Fall through */
2204        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);  
2205        case OP_ALLANY:
2206        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2207          {                            /* not be updated before SCHECK_PARTIAL. */
2208          SCHECK_PARTIAL();
2209          RRETURN(MATCH_NOMATCH);
2210        }        }
2211      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2212      if (utf8)  #ifdef SUPPORT_UTF
2213        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2214    #endif
2215      ecode++;      ecode++;
2216      break;      break;
2217    
# Line 1334  for (;;) Line 2219  for (;;)
2219      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2220    
2221      case OP_ANYBYTE:      case OP_ANYBYTE:
2222      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2223          {                            /* not be updated before SCHECK_PARTIAL. */
2224          SCHECK_PARTIAL();
2225          RRETURN(MATCH_NOMATCH);
2226          }
2227        eptr++;
2228      ecode++;      ecode++;
2229      break;      break;
2230    
2231      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2232      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2233          {
2234          SCHECK_PARTIAL();
2235          RRETURN(MATCH_NOMATCH);
2236          }
2237      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2238      if (      if (
2239  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2240         c < 256 &&         c < 256 &&
2241  #endif  #endif
2242         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1352  for (;;) Line 2246  for (;;)
2246      break;      break;
2247    
2248      case OP_DIGIT:      case OP_DIGIT:
2249      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2250          {
2251          SCHECK_PARTIAL();
2252          RRETURN(MATCH_NOMATCH);
2253          }
2254      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2255      if (      if (
2256  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2257         c >= 256 ||         c > 255 ||
2258  #endif  #endif
2259         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2260         )         )
# Line 1365  for (;;) Line 2263  for (;;)
2263      break;      break;
2264    
2265      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2266      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2267          {
2268          SCHECK_PARTIAL();
2269          RRETURN(MATCH_NOMATCH);
2270          }
2271      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2272      if (      if (
2273  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2274         c < 256 &&         c < 256 &&
2275  #endif  #endif
2276         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1378  for (;;) Line 2280  for (;;)
2280      break;      break;
2281    
2282      case OP_WHITESPACE:      case OP_WHITESPACE:
2283      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2284          {
2285          SCHECK_PARTIAL();
2286          RRETURN(MATCH_NOMATCH);
2287          }
2288      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2289      if (      if (
2290  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2291         c >= 256 ||         c > 255 ||
2292  #endif  #endif
2293         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2294         )         )
# Line 1391  for (;;) Line 2297  for (;;)
2297      break;      break;
2298    
2299      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2300      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2301          {
2302          SCHECK_PARTIAL();
2303          RRETURN(MATCH_NOMATCH);
2304          }
2305      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2306      if (      if (
2307  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2308         c < 256 &&         c < 256 &&
2309  #endif  #endif
2310         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1404  for (;;) Line 2314  for (;;)
2314      break;      break;
2315    
2316      case OP_WORDCHAR:      case OP_WORDCHAR:
2317      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2318          {
2319          SCHECK_PARTIAL();
2320          RRETURN(MATCH_NOMATCH);
2321          }
2322      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2323      if (      if (
2324  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2325         c >= 256 ||         c > 255 ||
2326  #endif  #endif
2327         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2328         )         )
# Line 1417  for (;;) Line 2331  for (;;)
2331      break;      break;
2332    
2333      case OP_ANYNL:      case OP_ANYNL:
2334      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2335          {
2336          SCHECK_PARTIAL();
2337          RRETURN(MATCH_NOMATCH);
2338          }
2339      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2340      switch(c)      switch(c)
2341        {        {
2342        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2343    
2344        case 0x000d:        case 0x000d:
2345        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2346        break;        break;
2347    
2348        case 0x000a:        case 0x000a:
2349          break;
2350    
2351        case 0x000b:        case 0x000b:
2352        case 0x000c:        case 0x000c:
2353        case 0x0085:        case 0x0085:
2354        case 0x2028:        case 0x2028:
2355        case 0x2029:        case 0x2029:
2356          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2357          break;
2358          }
2359        ecode++;
2360        break;
2361    
2362        case OP_NOT_HSPACE:
2363        if (eptr >= md->end_subject)
2364          {
2365          SCHECK_PARTIAL();
2366          RRETURN(MATCH_NOMATCH);
2367          }
2368        GETCHARINCTEST(c, eptr);
2369        switch(c)
2370          {
2371          default: break;
2372          case 0x09:      /* HT */
2373          case 0x20:      /* SPACE */
2374          case 0xa0:      /* NBSP */
2375          case 0x1680:    /* OGHAM SPACE MARK */
2376          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2377          case 0x2000:    /* EN QUAD */
2378          case 0x2001:    /* EM QUAD */
2379          case 0x2002:    /* EN SPACE */
2380          case 0x2003:    /* EM SPACE */
2381          case 0x2004:    /* THREE-PER-EM SPACE */
2382          case 0x2005:    /* FOUR-PER-EM SPACE */
2383          case 0x2006:    /* SIX-PER-EM SPACE */
2384          case 0x2007:    /* FIGURE SPACE */
2385          case 0x2008:    /* PUNCTUATION SPACE */
2386          case 0x2009:    /* THIN SPACE */
2387          case 0x200A:    /* HAIR SPACE */
2388          case 0x202f:    /* NARROW NO-BREAK SPACE */
2389          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2390          case 0x3000:    /* IDEOGRAPHIC SPACE */
2391          RRETURN(MATCH_NOMATCH);
2392          }
2393        ecode++;
2394        break;
2395    
2396        case OP_HSPACE:
2397        if (eptr >= md->end_subject)
2398          {
2399          SCHECK_PARTIAL();
2400          RRETURN(MATCH_NOMATCH);
2401          }
2402        GETCHARINCTEST(c, eptr);
2403        switch(c)
2404          {
2405          default: RRETURN(MATCH_NOMATCH);
2406          case 0x09:      /* HT */
2407          case 0x20:      /* SPACE */
2408          case 0xa0:      /* NBSP */
2409          case 0x1680:    /* OGHAM SPACE MARK */
2410          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2411          case 0x2000:    /* EN QUAD */
2412          case 0x2001:    /* EM QUAD */
2413          case 0x2002:    /* EN SPACE */
2414          case 0x2003:    /* EM SPACE */
2415          case 0x2004:    /* THREE-PER-EM SPACE */
2416          case 0x2005:    /* FOUR-PER-EM SPACE */
2417          case 0x2006:    /* SIX-PER-EM SPACE */
2418          case 0x2007:    /* FIGURE SPACE */
2419          case 0x2008:    /* PUNCTUATION SPACE */
2420          case 0x2009:    /* THIN SPACE */
2421          case 0x200A:    /* HAIR SPACE */
2422          case 0x202f:    /* NARROW NO-BREAK SPACE */
2423          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2424          case 0x3000:    /* IDEOGRAPHIC SPACE */
2425          break;
2426          }
2427        ecode++;
2428        break;
2429    
2430        case OP_NOT_VSPACE:
2431        if (eptr >= md->end_subject)
2432          {
2433          SCHECK_PARTIAL();
2434          RRETURN(MATCH_NOMATCH);
2435          }
2436        GETCHARINCTEST(c, eptr);
2437        switch(c)
2438          {
2439          default: break;
2440          case 0x0a:      /* LF */
2441          case 0x0b:      /* VT */
2442          case 0x0c:      /* FF */
2443          case 0x0d:      /* CR */
2444          case 0x85:      /* NEL */
2445          case 0x2028:    /* LINE SEPARATOR */
2446          case 0x2029:    /* PARAGRAPH SEPARATOR */
2447          RRETURN(MATCH_NOMATCH);
2448          }
2449        ecode++;
2450        break;
2451    
2452        case OP_VSPACE:
2453        if (eptr >= md->end_subject)
2454          {
2455          SCHECK_PARTIAL();
2456          RRETURN(MATCH_NOMATCH);
2457          }
2458        GETCHARINCTEST(c, eptr);
2459        switch(c)
2460          {
2461          default: RRETURN(MATCH_NOMATCH);
2462          case 0x0a:      /* LF */
2463          case 0x0b:      /* VT */
2464          case 0x0c:      /* FF */
2465          case 0x0d:      /* CR */
2466          case 0x85:      /* NEL */
2467          case 0x2028:    /* LINE SEPARATOR */
2468          case 0x2029:    /* PARAGRAPH SEPARATOR */
2469        break;        break;
2470        }        }
2471      ecode++;      ecode++;
# Line 1442  for (;;) Line 2477  for (;;)
2477    
2478      case OP_PROP:      case OP_PROP:
2479      case OP_NOTPROP:      case OP_NOTPROP:
2480      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2481          {
2482          SCHECK_PARTIAL();
2483          RRETURN(MATCH_NOMATCH);
2484          }
2485      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2486        {        {
2487        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2488    
2489        switch(ecode[1])        switch(ecode[1])
2490          {          {
# Line 1455  for (;;) Line 2493  for (;;)
2493          break;          break;
2494    
2495          case PT_LAMP:          case PT_LAMP:
2496          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2497               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2498               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2499            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2500           break;          break;
2501    
2502          case PT_GC:          case PT_GC:
2503          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2504            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2505          break;          break;
2506    
2507          case PT_PC:          case PT_PC:
2508          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2509            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2510          break;          break;
2511    
2512          case PT_SC:          case PT_SC:
2513          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2514              RRETURN(MATCH_NOMATCH);
2515            break;
2516    
2517            /* These are specials */
2518    
2519            case PT_ALNUM:
2520            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2521                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2522              RRETURN(MATCH_NOMATCH);
2523            break;
2524    
2525            case PT_SPACE:    /* Perl space */
2526            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2527                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2528                   == (op == OP_NOTPROP))
2529              RRETURN(MATCH_NOMATCH);
2530            break;
2531    
2532            case PT_PXSPACE:  /* POSIX space */
2533            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2534                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2535                 c == CHAR_FF || c == CHAR_CR)
2536                   == (op == OP_NOTPROP))
2537              RRETURN(MATCH_NOMATCH);
2538            break;
2539    
2540            case PT_WORD:
2541            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2542                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2543                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2544            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2545          break;          break;
2546    
2547            /* This should never occur */
2548    
2549          default:          default:
2550          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2551          }          }
# Line 1488  for (;;) Line 2558  for (;;)
2558      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2559    
2560      case OP_EXTUNI:      case OP_EXTUNI:
2561      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2562          {
2563          SCHECK_PARTIAL();
2564          RRETURN(MATCH_NOMATCH);
2565          }
2566      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2567        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2568        while (eptr < md->end_subject)
2569        {        {
2570        int chartype, script;        int len = 1;
2571        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2572        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2573        while (eptr < md->end_subject)        eptr += len;
         {  
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2574        }        }
2575      ecode++;      ecode++;
2576      break;      break;
# Line 1520  for (;;) Line 2586  for (;;)
2586      loops). */      loops). */
2587    
2588      case OP_REF:      case OP_REF:
2589        {      case OP_REFI:
2590        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2591        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2592        ecode += 1 + IMM2_SIZE;
2593    
2594        /* If the reference is unset, set the length to be longer than the amount      /* If the reference is unset, there are two possibilities:
       of subject left; this ensures that every attempt at a match fails. We  
       can't just fail here, because of the possibility of quantifiers with zero  
       minima. */  
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2595    
2596        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2597        this ensures that every attempt at a match fails. We can't just fail
2598        here, because of the possibility of quantifiers with zero minima.
2599    
2600        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2601          {      so that the back reference matches an empty string.
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         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;  
2602    
2603          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2604          case OP_CRMINRANGE:      referenced subpattern. */
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2605    
2606          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2607          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2608          eptr += length;      else
2609          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2610    
2611        /* Set up for repetition, or handle the non-repeated case */
2612    
2613        switch (*ecode)
2614          {
2615          case OP_CRSTAR:
2616          case OP_CRMINSTAR:
2617          case OP_CRPLUS:
2618          case OP_CRMINPLUS:
2619          case OP_CRQUERY:
2620          case OP_CRMINQUERY:
2621          c = *ecode++ - OP_CRSTAR;
2622          minimize = (c & 1) != 0;
2623          min = rep_min[c];                 /* Pick up values from tables; */
2624          max = rep_max[c];                 /* zero for max => infinity */
2625          if (max == 0) max = INT_MAX;
2626          break;
2627    
2628          case OP_CRRANGE:
2629          case OP_CRMINRANGE:
2630          minimize = (*ecode == OP_CRMINRANGE);
2631          min = GET2(ecode, 1);
2632          max = GET2(ecode, 1 + IMM2_SIZE);
2633          if (max == 0) max = INT_MAX;
2634          ecode += 1 + 2 * IMM2_SIZE;
2635          break;
2636    
2637          default:               /* No repeat follows */
2638          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2639            {
2640            CHECK_PARTIAL();
2641            RRETURN(MATCH_NOMATCH);
2642          }          }
2643          eptr += length;
2644          continue;              /* With the main loop */
2645          }
2646    
2647        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2648        main loop. */      zero, just continue with the main loop. If the length is negative, it
2649        means the reference is unset in non-Java-compatible mode. If the minimum is
2650        zero, we can continue at the same level without recursion. For any other
2651        minimum, carrying on will result in NOMATCH. */
2652    
2653        if (length == 0) continue;      if (length == 0) continue;
2654        if (length < 0 && min == 0) continue;
2655    
2656        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2657        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2658        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2659    
2660        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2661          {
2662          int slength;
2663          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2664          {          {
2665          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2666          eptr += length;          RRETURN(MATCH_NOMATCH);
2667          }          }
2668          eptr += slength;
2669          }
2670    
2671        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2672        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2673    
2674        if (min == max) continue;      if (min == max) continue;
2675    
2676        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2677    
2678        if (minimize)      if (minimize)
2679          {
2680          for (fi = min;; fi++)
2681          {          {
2682          for (fi = min;; fi++)          int slength;
2683            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2684            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2685            if (fi >= max) RRETURN(MATCH_NOMATCH);
2686            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2687            {            {
2688            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2689            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            RRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2690            }            }
2691          /* Control never gets here */          eptr += slength;
2692          }          }
2693          /* Control never gets here */
2694          }
2695    
2696        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2697    
2698        else      else
2699          {
2700          pp = eptr;
2701          for (i = min; i < max; i++)
2702          {          {
2703          pp = eptr;          int slength;
2704          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)  
2705            {            {
2706            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2707            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2708            }            }
2709          RRETURN(MATCH_NOMATCH);          eptr += slength;
2710          }          }
2711          while (eptr >= pp)
2712            {
2713            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2714            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2715            eptr -= length;
2716            }
2717          RRETURN(MATCH_NOMATCH);
2718        }        }
2719      /* Control never gets here */      /* Control never gets here */
2720    
   
   
2721      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2722      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,
2723      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 1637  for (;;) Line 2732  for (;;)
2732      case OP_NCLASS:      case OP_NCLASS:
2733      case OP_CLASS:      case OP_CLASS:
2734        {        {
2735          /* The data variable is saved across frames, so the byte map needs to
2736          be stored there. */
2737    #define BYTE_MAP ((pcre_uint8 *)data)
2738        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2739        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2740    
2741        switch (*ecode)        switch (*ecode)
2742          {          {
# Line 1659  for (;;) Line 2757  for (;;)
2757          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2758          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2759          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2760          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2761          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2762          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2763          break;          break;
2764    
2765          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1671  for (;;) Line 2769  for (;;)
2769    
2770        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2771    
2772  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2773        /* UTF-8 mode */        if (utf)
       if (utf8)  
2774          {          {
2775          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2776            {            {
2777            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2778                {
2779                SCHECK_PARTIAL();
2780                RRETURN(MATCH_NOMATCH);
2781                }
2782            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2783            if (c > 255)            if (c > 255)
2784              {              {
2785              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2786              }              }
2787            else            else
2788              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2789            }            }
2790          }          }
2791        else        else
2792  #endif  #endif
2793        /* Not UTF-8 mode */        /* Not UTF mode */
2794          {          {
2795          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2796            {            {
2797            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2798                {
2799                SCHECK_PARTIAL();
2800                RRETURN(MATCH_NOMATCH);
2801                }
2802            c = *eptr++;            c = *eptr++;
2803            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2804              if (c > 255)
2805                {
2806                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2807                }
2808              else
2809    #endif
2810                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2811            }            }
2812          }          }
2813    
# Line 1711  for (;;) Line 2821  for (;;)
2821    
2822        if (minimize)        if (minimize)
2823          {          {
2824  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2825          /* UTF-8 mode */          if (utf)
         if (utf8)  
2826            {            {
2827            for (fi = min;; fi++)            for (fi = min;; fi++)
2828              {              {
2829              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2830              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2831              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2832                if (eptr >= md->end_subject)
2833                  {
2834                  SCHECK_PARTIAL();
2835                  RRETURN(MATCH_NOMATCH);
2836                  }
2837              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2838              if (c > 255)              if (c > 255)
2839                {                {
2840                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2841                }                }
2842              else              else
2843                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2844              }              }
2845            }            }
2846          else          else
2847  #endif  #endif
2848          /* Not UTF-8 mode */          /* Not UTF mode */
2849            {            {
2850            for (fi = min;; fi++)            for (fi = min;; fi++)
2851              {              {
2852              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2853              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2854              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2855                if (eptr >= md->end_subject)
2856                  {
2857                  SCHECK_PARTIAL();
2858                  RRETURN(MATCH_NOMATCH);
2859                  }
2860              c = *eptr++;              c = *eptr++;
2861              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2862                if (c > 255)
2863                  {
2864                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2865                  }
2866                else
2867    #endif
2868                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2869              }              }
2870            }            }
2871          /* Control never gets here */          /* Control never gets here */
# Line 1753  for (;;) Line 2877  for (;;)
2877          {          {
2878          pp = eptr;          pp = eptr;
2879    
2880  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2881          /* UTF-8 mode */          if (utf)
         if (utf8)  
2882            {            {
2883            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2884              {              {
2885              int len = 1;              int len = 1;
2886              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2887                  {
2888                  SCHECK_PARTIAL();
2889                  break;
2890                  }
2891              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2892              if (c > 255)              if (c > 255)
2893                {                {
2894                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2895                }                }
2896              else              else
2897                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2898              eptr += len;              eptr += len;
2899              }              }
2900            for (;;)            for (;;)
2901              {              {
2902              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2903              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2904              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2905              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1782  for (;;) Line 2907  for (;;)
2907            }            }
2908          else          else
2909  #endif  #endif
2910            /* Not UTF-8 mode */            /* Not UTF mode */
2911            {            {
2912            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2913              {              {
2914              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2915                  {
2916                  SCHECK_PARTIAL();
2917                  break;
2918                  }
2919              c = *eptr;              c = *eptr;
2920              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2921                if (c > 255)
2922                  {
2923                  if (op == OP_CLASS) break;
2924                  }
2925                else
2926    #endif
2927                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2928              eptr++;              eptr++;
2929              }              }
2930            while (eptr >= pp)            while (eptr >= pp)
2931              {              {
2932              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2933              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2934              eptr--;              eptr--;
2935              }              }
# Line 1801  for (;;) Line 2937  for (;;)
2937    
2938          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2939          }          }
2940    #undef BYTE_MAP
2941        }        }
2942      /* Control never gets here */      /* Control never gets here */
2943    
2944    
2945      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2946      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
2947        mode, because Unicode properties are supported in non-UTF-8 mode. */
2948    
2949  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2950      case OP_XCLASS:      case OP_XCLASS:
2951        {        {
2952        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1833  for (;;) Line 2971  for (;;)
2971          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2972          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2973          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2974          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2975          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2976          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2977          break;          break;
2978    
2979          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1847  for (;;) Line 2985  for (;;)
2985    
2986        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2987          {          {
2988          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2989          GETCHARINC(c, eptr);            {
2990          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2991              RRETURN(MATCH_NOMATCH);
2992              }
2993            GETCHARINCTEST(c, eptr);
2994            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
2995          }          }
2996    
2997        /* 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 1864  for (;;) Line 3006  for (;;)
3006          {          {
3007          for (fi = min;; fi++)          for (fi = min;; fi++)
3008            {            {
3009            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3010            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3011            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3012            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3013            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3014                SCHECK_PARTIAL();
3015                RRETURN(MATCH_NOMATCH);
3016                }
3017              GETCHARINCTEST(c, eptr);
3018              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3019            }            }
3020          /* Control never gets here */          /* Control never gets here */
3021          }          }
# Line 1881  for (;;) Line 3028  for (;;)
3028          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3029            {            {
3030            int len = 1;            int len = 1;
3031            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3032            GETCHARLEN(c, eptr, len);              {
3033            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3034                break;
3035                }
3036    #ifdef SUPPORT_UTF
3037              GETCHARLENTEST(c, eptr, len);
3038    #else
3039              c = *eptr;
3040    #endif
3041              if (!PRIV(xclass)(c, data, utf)) break;
3042            eptr += len;            eptr += len;
3043            }            }
3044          for(;;)          for(;;)
3045            {            {
3046            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3047            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3048            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3049            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3050              if (utf) BACKCHAR(eptr);
3051    #endif
3052            }            }
3053          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3054          }          }
# Line 1903  for (;;) Line 3060  for (;;)
3060      /* Match a single character, casefully */      /* Match a single character, casefully */
3061    
3062      case OP_CHAR:      case OP_CHAR:
3063  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3064      if (utf8)      if (utf)
3065        {        {
3066        length = 1;        length = 1;
3067        ecode++;        ecode++;
3068        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3069        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3070            {
3071            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3072            RRETURN(MATCH_NOMATCH);
3073            }
3074        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3075        }        }
3076      else      else
3077  #endif  #endif
3078        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3079        {        {
3080        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3081            {
3082            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3083            RRETURN(MATCH_NOMATCH);
3084            }
3085        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3086        ecode += 2;        ecode += 2;
3087        }        }
3088      break;      break;
3089    
3090      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3091        subject, give up immediately. */
3092    
3093        case OP_CHARI:
3094        if (eptr >= md->end_subject)
3095          {
3096          SCHECK_PARTIAL();
3097          RRETURN(MATCH_NOMATCH);
3098          }
3099    
3100      case OP_CHARNC:  #ifdef SUPPORT_UTF
3101  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3102        {        {
3103        length = 1;        length = 1;
3104        ecode++;        ecode++;
3105        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3106    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3107        /* 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
3108        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3109          fast lookup table. We know that there is at least one byte left in the
3110          subject. */
3111    
3112        if (fc < 128)        if (fc < 128)
3113          {          {
3114          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3115                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3116            ecode++;
3117            eptr++;
3118          }          }
3119    
3120        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3121          use the value of "length" to check for sufficient bytes left, because the
3122          other case of the character may have more or fewer bytes.  */
3123    
3124        else        else
3125          {          {
# Line 1957  for (;;) Line 3133  for (;;)
3133          if (fc != dc)          if (fc != dc)
3134            {            {
3135  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3136            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3137  #endif  #endif
3138              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3139            }            }
3140          }          }
3141        }        }
3142      else      else
3143  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3144    
3145      /* Non-UTF-8 mode */      /* Not UTF mode */
3146        {        {
3147        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3148        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3149          eptr++;
3150        ecode += 2;        ecode += 2;
3151        }        }
3152      break;      break;
# Line 1977  for (;;) Line 3154  for (;;)
3154      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3155    
3156      case OP_EXACT:      case OP_EXACT:
3157        case OP_EXACTI:
3158      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3159      ecode += 3;      ecode += 1 + IMM2_SIZE;
3160      goto REPEATCHAR;      goto REPEATCHAR;
3161    
3162      case OP_POSUPTO:      case OP_POSUPTO:
3163        case OP_POSUPTOI:
3164      possessive = TRUE;      possessive = TRUE;
3165      /* Fall through */      /* Fall through */
3166    
3167      case OP_UPTO:      case OP_UPTO:
3168        case OP_UPTOI:
3169      case OP_MINUPTO:      case OP_MINUPTO:
3170        case OP_MINUPTOI:
3171      min = 0;      min = 0;
3172      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3173      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3174      ecode += 3;      ecode += 1 + IMM2_SIZE;
3175      goto REPEATCHAR;      goto REPEATCHAR;
3176    
3177      case OP_POSSTAR:      case OP_POSSTAR:
3178        case OP_POSSTARI:
3179      possessive = TRUE;      possessive = TRUE;
3180      min = 0;      min = 0;
3181      max = INT_MAX;      max = INT_MAX;
# Line 2001  for (;;) Line 3183  for (;;)
3183      goto REPEATCHAR;      goto REPEATCHAR;
3184    
3185      case OP_POSPLUS:      case OP_POSPLUS:
3186        case OP_POSPLUSI:
3187      possessive = TRUE;      possessive = TRUE;
3188      min = 1;      min = 1;
3189      max = INT_MAX;      max = INT_MAX;
# Line 2008  for (;;) Line 3191  for (;;)
3191      goto REPEATCHAR;      goto REPEATCHAR;
3192    
3193      case OP_POSQUERY:      case OP_POSQUERY:
3194        case OP_POSQUERYI:
3195      possessive = TRUE;      possessive = TRUE;
3196      min = 0;      min = 0;
3197      max = 1;      max = 1;
# Line 2015  for (;;) Line 3199  for (;;)
3199      goto REPEATCHAR;      goto REPEATCHAR;
3200    
3201      case OP_STAR:      case OP_STAR:
3202        case OP_STARI:
3203      case OP_MINSTAR:      case OP_MINSTAR:
3204        case OP_MINSTARI:
3205      case OP_PLUS:      case OP_PLUS:
3206        case OP_PLUSI:
3207      case OP_MINPLUS:      case OP_MINPLUS:
3208        case OP_MINPLUSI:
3209      case OP_QUERY:      case OP_QUERY:
3210        case OP_QUERYI:
3211      case OP_MINQUERY:      case OP_MINQUERY:
3212      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3213        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3214      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3215      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3216      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3217      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3218    
3219      /* 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. */  
3220    
3221      REPEATCHAR:      REPEATCHAR:
3222  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3223      if (utf8)      if (utf)
3224        {        {
3225        length = 1;        length = 1;
3226        charptr = ecode;        charptr = ecode;
3227        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3228        ecode += length;        ecode += length;
3229    
3230        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2045  for (;;) Line 3232  for (;;)
3232    
3233        if (length > 1)        if (length > 1)
3234          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3235  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3236          unsigned int othercase;          unsigned int othercase;
3237          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3238              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3239            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3240            else oclength = 0;
3241  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3242    
3243          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3244            {            {
3245            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3246            /* Need braces because of following else */              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3247            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3248              else if (oclength > 0 &&
3249                       eptr <= md->end_subject - oclength &&
3250                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3251    #endif  /* SUPPORT_UCP */
3252            else            else
3253              {              {
3254              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3255              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3256              }              }
3257            }            }
3258    
# Line 2073  for (;;) Line 3262  for (;;)
3262            {            {
3263            for (fi = min;; fi++)            for (fi = min;; fi++)
3264              {              {
3265              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3266              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3267              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3268              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3269              /* Need braces because of following else */                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3270              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3271                else if (oclength > 0 &&
3272                         eptr <= md->end_subject - oclength &&
3273                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3274    #endif  /* SUPPORT_UCP */
3275              else              else
3276                {                {
3277                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3278                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3279                }                }
3280              }              }
3281            /* Control never gets here */            /* Control never gets here */
# Line 2093  for (;;) Line 3286  for (;;)
3286            pp = eptr;            pp = eptr;
3287            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3288              {              {
3289              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3290              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3291              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3292                else if (oclength > 0 &&
3293                         eptr <= md->end_subject - oclength &&
3294                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3295    #endif  /* SUPPORT_UCP */
3296              else              else
3297                {                {
3298                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3299                eptr += oclength;                break;
3300                }                }
3301              }              }
3302    
3303            if (possessive) continue;            if (possessive) continue;
3304            while (eptr >= pp)  
3305             {            for(;;)
3306             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              {
3307             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3308             eptr -= length;              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3309             }              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3310            RRETURN(MATCH_NOMATCH);  #ifdef SUPPORT_UCP
3311                eptr--;
3312                BACKCHAR(eptr);
3313    #else   /* without SUPPORT_UCP */
3314                eptr -= length;
3315    #endif  /* SUPPORT_UCP */
3316                }
3317            }            }
3318          /* Control never gets here */          /* Control never gets here */
3319          }          }
# Line 2120  for (;;) Line 3323  for (;;)
3323        value of fc will always be < 128. */        value of fc will always be < 128. */
3324        }        }
3325      else      else
3326  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3327          /* When not in UTF-8 mode, load a single-byte character. */
     /* When not in UTF-8 mode, load a single-byte character. */  
       {  
       if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3328        fc = *ecode++;        fc = *ecode++;
       }  
3329    
3330      /* 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 one character, though we may
3331      may not be in UTF-8 mode. The code is duplicated for the caseless and      or may not be in UTF mode. The code is duplicated for the caseless and
3332      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3333      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3334      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2140  for (;;) Line 3339  for (;;)
3339      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3340        max, eptr));        max, eptr));
3341    
3342      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3343        {        {
3344        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3345          /* fc must be < 128 if UTF is enabled. */
3346          foc = md->fcc[fc];
3347    #else
3348    #ifdef SUPPORT_UTF
3349    #ifdef SUPPORT_UCP
3350          if (utf && fc > 127)
3351            foc = UCD_OTHERCASE(fc);
3352    #else
3353          if (utf && fc > 127)
3354            foc = fc;
3355    #endif /* SUPPORT_UCP */
3356          else
3357    #endif /* SUPPORT_UTF */
3358            foc = TABLE_GET(fc, md->fcc, fc);
3359    #endif /* COMPILE_PCRE8 */
3360    
3361        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3362          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3363            if (eptr >= md->end_subject)
3364              {
3365              SCHECK_PARTIAL();
3366              RRETURN(MATCH_NOMATCH);
3367              }
3368            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3369            eptr++;
3370            }
3371        if (min == max) continue;        if (min == max) continue;
3372        if (minimize)        if (minimize)
3373          {          {
3374          for (fi = min;; fi++)          for (fi = min;; fi++)
3375            {            {
3376            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3377            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3378            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3379                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3380                {
3381                SCHECK_PARTIAL();
3382              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3383                }
3384              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3385              eptr++;
3386            }            }