/[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 150 by ph10, Tue Apr 17 08:22:40 2007 UTC revision 893 by ph10, Thu Jan 19 17:15:11 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) 1997-2007 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 */
# Line 53  possible. There are also some static sup Line 57  possible. There are also some static sup
57  #undef min  #undef min
58  #undef max  #undef max
59    
60  /* The chain of eptrblocks for tail recursions uses memory in stack workspace,  /* Values for setting in md->match_function_type to indicate two special types
61  obtained at top level, the size of which is defined by EPTR_WORK_SIZE. */  of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
 #define EPTR_WORK_SIZE (1000)  
   
 /* Flag bits for the match() function */  
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 70  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 83  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 101  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 116  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 148  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 188  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 210  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);\    if (oldframe != &frame_zero) (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 271  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 295  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;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
381    int Xoclength;    int Xoclength;
382    uschar Xocchars[8];    pcre_uchar Xocchars[6];
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 323  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 344  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 384  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". We set up the top-level
489  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
490  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
491    the top-level on the stack rather than malloc-ing them all gives a performance
492    boost in many cases where there is not much "recursion". */
493    
494  #ifdef NO_RECURSE  #ifdef NO_RECURSE
495  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe frame_zero;
496    heapframe *frame = &frame_zero;
497  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
498    
499  /* Copy in the original argument variables */  /* Copy in the original argument variables */
500    
501  frame->Xeptr = eptr;  frame->Xeptr = eptr;
502  frame->Xecode = ecode;  frame->Xecode = ecode;
503    frame->Xmstart = mstart;
504  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
505  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
506  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
507    
508  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 415  HEAP_RECURSE: Line 513  HEAP_RECURSE:
513    
514  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
515  #define ecode              frame->Xecode  #define ecode              frame->Xecode
516    #define mstart             frame->Xmstart
517  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
518  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
519  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
520    
521  /* Ditto for the local variables */  /* Ditto for the local variables */
522    
523  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
524  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
525  #endif  #endif
526  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
527    #define codelink           frame->Xcodelink
528  #define data               frame->Xdata  #define data               frame->Xdata
529  #define next               frame->Xnext  #define next               frame->Xnext
530  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 439  HEAP_RECURSE: Line 537  HEAP_RECURSE:
537  #define condition          frame->Xcondition  #define condition          frame->Xcondition
538  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
539    
 #define original_ims       frame->Xoriginal_ims  
   
540  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
541  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
542  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
543  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
544  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
545  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
546  #endif  #endif
# Line 477  i, and fc and c, can be the same variabl Line 570  i, and fc and c, can be the same variabl
570  #define fi i  #define fi i
571  #define fc c  #define fc c
572    
573    /* Many of the following variables are used only in small blocks of the code.
574    My normal style of coding would have declared them within each of those blocks.
575    However, in order to accommodate the version of this code that uses an external
576    "stack" implemented on the heap, it is easier to declare them all here, so the
577    declarations can be cut out in a block. The only declarations within blocks
578    below are for variables that do not have to be preserved over a recursive call
579    to RMATCH(). */
580    
581    #ifdef SUPPORT_UTF
582    const pcre_uchar *charptr;
583    #endif
584    const pcre_uchar *callpat;
585    const pcre_uchar *data;
586    const pcre_uchar *next;
587    PCRE_PUCHAR       pp;
588    const pcre_uchar *prev;
589    PCRE_PUCHAR       saved_eptr;
590    
591    recursion_info new_recursive;
592    
593  #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().          */  
594  BOOL condition;  BOOL condition;
595  BOOL prev_is_word;  BOOL prev_is_word;
596    
 unsigned long int original_ims;  
   
597  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
598  int prop_type;  int prop_type;
599  int prop_value;  int prop_value;
600  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
601  int oclength;  int oclength;
602  uschar occhars[8];  pcre_uchar occhars[6];
603  #endif  #endif
604    
605    int codelink;
606  int ctype;  int ctype;
607  int length;  int length;
608  int max;  int max;
# Line 519  int save_offset1, save_offset2, save_off Line 615  int save_offset1, save_offset2, save_off
615  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
616    
617  eptrblock newptrb;  eptrblock newptrb;
618    
619    /* There is a special fudge for calling match() in a way that causes it to
620    measure the size of its basic stack frame when the stack is being used for
621    recursion. The first argument (eptr) points to a pointer that is used
622    "statically" for doing the calculation. The second argument (ecode) being NULL
623    triggers this behaviour. It cannot normally every be NULL. The return is the
624    negated value of the frame size. */
625    
626    if (ecode == NULL)
627      {
628      char **aptr = (char **)eptr;
629      if (rdepth == 0)
630        {
631        *aptr = (char *)&rdepth;
632        return match(eptr, NULL, NULL, 0, NULL, NULL, 1);
633        }
634      else
635        {
636        int len = (char *)&rdepth - *aptr;
637        return (len > 0)? -len : len;
638        }
639      }
640  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
641    
642    /* To save space on the stack and in the heap frame, I have doubled up on some
643    of the local variables that are used only in localised parts of the code, but
644    still need to be preserved over recursive calls of match(). These macros define
645    the alternative names that are used. */
646    
647    #define allow_zero    cur_is_word
648    #define cbegroup      condition
649    #define code_offset   codelink
650    #define condassert    condition
651    #define matched_once  prev_is_word
652    #define foc           number
653    #define save_mark     data
654    
655  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
656  variables. */  variables. */
657    
# Line 540  TAIL_RECURSE: Line 671  TAIL_RECURSE:
671  /* 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
672  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
673  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()
674  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
675  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
676  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,
677  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
678    
679    #ifdef SUPPORT_UTF
680    utf = md->utf;       /* Local copy of the flag */
681    #else
682    utf = FALSE;
683    #endif
684    
685  /* 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
686  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
687    
688  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
689  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
690    
 original_ims = ims;    /* Save for resetting on ')' */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
691  /* 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
692  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
693  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
694  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
695  When match() is called in other circumstances, don't add to the chain. If this  
696  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
697  already used. */  such remembered pointers, to be checked when we hit the closing ket, in order
698    to break infinite loops that match no characters. When match() is called in
699    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
700    NOT be used with tail recursion, because the memory block that is used is on
701    the stack, so a new one may be required for each match(). */
702    
703  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
704    {    {
705    eptrblock *p;    newptrb.epb_saved_eptr = eptr;
706    if ((flags & match_tail_recursed) != 0)    newptrb.epb_prev = eptrb;
707      {    eptrb = &newptrb;
708      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;  
709    }    }
710    
711  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 588  for (;;) Line 715  for (;;)
715    minimize = possessive = FALSE;    minimize = possessive = FALSE;
716    op = *ecode;    op = *ecode;
717    
   /* 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;  
   
718    switch(op)    switch(op)
719      {      {
720      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
721      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
722      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
723      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,
724      reference inside the group.        eptrb, RM55);
725        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
726      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
727      values of the final offsets, in case they were set by a previous iteration  
728      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
729        argument, and we must check whether that argument matches this MARK's
730        argument. It is passed back in md->start_match_ptr (an overloading of that
731        variable). If it does match, we reset that variable to the current subject
732        position and return MATCH_SKIP. Otherwise, pass back the return code
733        unaltered. */
734    
735        else if (rrc == MATCH_SKIP_ARG &&
736            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
737          {
738          md->start_match_ptr = eptr;
739          RRETURN(MATCH_SKIP);
740          }
741        RRETURN(rrc);
742    
743        case OP_FAIL:
744        RRETURN(MATCH_NOMATCH);
745    
746        /* COMMIT overrides PRUNE, SKIP, and THEN */
747    
748        case OP_COMMIT:
749        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
750          eptrb, RM52);
751        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
752            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
753            rrc != MATCH_THEN)
754          RRETURN(rrc);
755        RRETURN(MATCH_COMMIT);
756    
757        /* PRUNE overrides THEN */
758    
759        case OP_PRUNE:
760        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
761          eptrb, RM51);
762        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
763        RRETURN(MATCH_PRUNE);
764    
765        case OP_PRUNE_ARG:
766        md->nomatch_mark = ecode + 2;
767        md->mark = NULL;    /* In case previously set by assertion */
768        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
769          eptrb, RM56);
770        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
771             md->mark == NULL) md->mark = ecode + 2;
772        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
773        RRETURN(MATCH_PRUNE);
774    
775        /* SKIP overrides PRUNE and THEN */
776    
777        case OP_SKIP:
778        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
779          eptrb, RM53);
780        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
781          RRETURN(rrc);
782        md->start_match_ptr = eptr;   /* Pass back current position */
783        RRETURN(MATCH_SKIP);
784    
785        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
786        nomatch_mark. There is a flag that disables this opcode when re-matching a
787        pattern that ended with a SKIP for which there was not a matching MARK. */
788    
789        case OP_SKIP_ARG:
790        if (md->ignore_skip_arg)
791          {
792          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
793          break;
794          }
795        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
796          eptrb, RM57);
797        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
798          RRETURN(rrc);
799    
800        /* Pass back the current skip name by overloading md->start_match_ptr and
801        returning the special MATCH_SKIP_ARG return code. This will either be
802        caught by a matching MARK, or get to the top, where it causes a rematch
803        with the md->ignore_skip_arg flag set. */
804    
805        md->start_match_ptr = ecode + 2;
806        RRETURN(MATCH_SKIP_ARG);
807    
808        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
809        the branch in which it occurs can be determined. Overload the start of
810        match pointer to do this. */
811    
812        case OP_THEN:
813        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
814          eptrb, RM54);
815        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
816        md->start_match_ptr = ecode;
817        RRETURN(MATCH_THEN);
818    
819        case OP_THEN_ARG:
820        md->nomatch_mark = ecode + 2;
821        md->mark = NULL;    /* In case previously set by assertion */
822        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
823          md, eptrb, RM58);
824        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
825             md->mark == NULL) md->mark = ecode + 2;
826        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
827        md->start_match_ptr = ecode;
828        RRETURN(MATCH_THEN);
829    
830        /* Handle an atomic group that does not contain any capturing parentheses.
831        This can be handled like an assertion. Prior to 8.13, all atomic groups
832        were handled this way. In 8.13, the code was changed as below for ONCE, so
833        that backups pass through the group and thereby reset captured values.
834        However, this uses a lot more stack, so in 8.20, atomic groups that do not
835        contain any captures generate OP_ONCE_NC, which can be handled in the old,
836        less stack intensive way.
837    
838        Check the alternative branches in turn - the matching won't pass the KET
839        for this kind of subpattern. If any one branch matches, we carry on as at
840        the end of a normal bracket, leaving the subject pointer, but resetting
841        the start-of-match value in case it was changed by \K. */
842    
843        case OP_ONCE_NC:
844        prev = ecode;
845        saved_eptr = eptr;
846        save_mark = md->mark;
847        do
848          {
849          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
850          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
851            {
852            mstart = md->start_match_ptr;
853            break;
854            }
855          if (rrc == MATCH_THEN)
856            {
857            next = ecode + GET(ecode,1);
858            if (md->start_match_ptr < next &&
859                (*ecode == OP_ALT || *next == OP_ALT))
860              rrc = MATCH_NOMATCH;
861            }
862    
863          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
864          ecode += GET(ecode,1);
865          md->mark = save_mark;
866          }
867        while (*ecode == OP_ALT);
868    
869        /* If hit the end of the group (which could be repeated), fail */
870    
871        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
872    
873        /* Continue as from after the group, updating the offsets high water
874        mark, since extracts may have been taken. */
875    
876        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
877    
878        offset_top = md->end_offset_top;
879        eptr = md->end_match_ptr;
880    
881        /* For a non-repeating ket, just continue at this level. This also
882        happens for a repeating ket if no characters were matched in the group.
883        This is the forcible breaking of infinite loops as implemented in Perl
884        5.005. */
885    
886        if (*ecode == OP_KET || eptr == saved_eptr)
887          {
888          ecode += 1+LINK_SIZE;
889          break;
890          }
891    
892        /* The repeating kets try the rest of the pattern or restart from the
893        preceding bracket, in the appropriate order. The second "call" of match()
894        uses tail recursion, to avoid using another stack frame. */
895    
896        if (*ecode == OP_KETRMIN)
897          {
898          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
899          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
900          ecode = prev;
901          goto TAIL_RECURSE;
902          }
903        else  /* OP_KETRMAX */
904          {
905          md->match_function_type = MATCH_CBEGROUP;
906          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
907          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
908          ecode += 1 + LINK_SIZE;
909          goto TAIL_RECURSE;
910          }
911        /* Control never gets here */
912    
913        /* Handle a capturing bracket, other than those that are possessive with an
914        unlimited repeat. If there is space in the offset vector, save the current
915        subject position in the working slot at the top of the vector. We mustn't
916        change the current values of the data slot, because they may be set from a
917        previous iteration of this group, and be referred to by a reference inside
918        the group. A failure to match might occur after the group has succeeded,
919        if something later on doesn't match. For this reason, we need to restore
920        the working value and also the values of the final offsets, in case they
921        were set by a previous iteration of the same bracket.
922    
923      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
924      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 617  for (;;) Line 929  for (;;)
929      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
930      offset = number << 1;      offset = number << 1;
931    
932  #ifdef DEBUG  #ifdef PCRE_DEBUG
933      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
934      printf("subject=");      printf("subject=");
935      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 630  for (;;) Line 942  for (;;)
942        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
943        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
944        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
945          save_mark = md->mark;
946    
947        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
948        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
949            (int)(eptr - md->start_subject);
950    
951        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
952          {          {
953          RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
954            ims, eptrb, flags);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
955              eptrb, RM1);
956            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
957    
958            /* If we backed up to a THEN, check whether it is within the current
959            branch by comparing the address of the THEN that is passed back with
960            the end of the branch. If it is within the current branch, and the
961            branch is one of two or more alternatives (it either starts or ends
962            with OP_ALT), we have reached the limit of THEN's action, so convert
963            the return code to NOMATCH, which will cause normal backtracking to
964            happen from now on. Otherwise, THEN is passed back to an outer
965            alternative. This implements Perl's treatment of parenthesized groups,
966            where a group not containing | does not affect the current alternative,
967            that is, (X) is NOT the same as (X|(*F)). */
968    
969            if (rrc == MATCH_THEN)
970              {
971              next = ecode + GET(ecode,1);
972              if (md->start_match_ptr < next &&
973                  (*ecode == OP_ALT || *next == OP_ALT))
974                rrc = MATCH_NOMATCH;
975              }
976    
977            /* Anything other than NOMATCH is passed back. */
978    
979          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
980          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
981          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
982            md->mark = save_mark;
983            if (*ecode != OP_ALT) break;
984          }          }
       while (*ecode == OP_ALT);  
985    
986        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
987        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
988        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
989        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
990    
991        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
992    
993          RRETURN(rrc);
994        }        }
995    
996      /* Insufficient room for saving captured contents. Treat as a non-capturing      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
997      bracket. */      as a non-capturing bracket. */
998    
999        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1000        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1001    
1002      DPRINTF(("insufficient capture room: treat as non-capturing\n"));      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1003    
1004      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1005      final alternative within the brackets, we would return the result of a      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1006      recursive call to match() whatever happened. We can reduce stack usage by  
1007      turning this into a tail recursion. */      /* Non-capturing or atomic group, except for possessive with unlimited
1008        repeat and ONCE group with no captures. Loop for all the alternatives.
1009    
1010        When we get to the final alternative within the brackets, we used to return
1011        the result of a recursive call to match() whatever happened so it was
1012        possible to reduce stack usage by turning this into a tail recursion,
1013        except in the case of a possibly empty group. However, now that there is
1014        the possiblity of (*THEN) occurring in the final alternative, this
1015        optimization is no longer always possible.
1016    
1017        We can optimize if we know there are no (*THEN)s in the pattern; at present
1018        this is the best that can be done.
1019    
1020        MATCH_ONCE is returned when the end of an atomic group is successfully
1021        reached, but subsequent matching fails. It passes back up the tree (causing
1022        captured values to be reset) until the original atomic group level is
1023        reached. This is tested by comparing md->once_target with the start of the
1024        group. At this point, the return is converted into MATCH_NOMATCH so that
1025        previous backup points can be taken. */
1026    
1027        case OP_ONCE:
1028      case OP_BRA:      case OP_BRA:
1029      case OP_SBRA:      case OP_SBRA:
1030      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1031      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1032      for (;;)      for (;;)
1033        {        {
1034        if (ecode[GET(ecode, 1)] != OP_ALT)        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1035    
1036          /* If this is not a possibly empty group, and there are no (*THEN)s in
1037          the pattern, and this is the final alternative, optimize as described
1038          above. */
1039    
1040          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1041          {          {
1042          ecode += _pcre_OP_lengths[*ecode];          ecode += PRIV(OP_lengths)[*ecode];
         flags |= match_tail_recursed;  
         DPRINTF(("bracket 0 tail recursion\n"));  
1043          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1044          }          }
1045    
1046        /* 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. */  
1047    
1048        RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1049          eptrb, flags);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1050        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RM2);
1051    
1052          /* See comment in the code for capturing groups above about handling
1053          THEN. */
1054    
1055          if (rrc == MATCH_THEN)
1056            {
1057            next = ecode + GET(ecode,1);
1058            if (md->start_match_ptr < next &&
1059                (*ecode == OP_ALT || *next == OP_ALT))
1060              rrc = MATCH_NOMATCH;
1061            }
1062    
1063          if (rrc != MATCH_NOMATCH)
1064            {
1065            if (rrc == MATCH_ONCE)
1066              {
1067              const pcre_uchar *scode = ecode;
1068              if (*scode != OP_ONCE)           /* If not at start, find it */
1069                {
1070                while (*scode == OP_ALT) scode += GET(scode, 1);
1071                scode -= GET(scode, 1);
1072                }
1073              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1074              }
1075            RRETURN(rrc);
1076            }
1077        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1078          md->mark = save_mark;
1079          if (*ecode != OP_ALT) break;
1080        }        }
     /* Control never reaches here. */  
1081    
1082      /* 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. */  
1083    
1084      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1085      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1086      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1087        {      different. The end of these brackets will always be OP_KETRPOS, which
1088        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1089        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1090          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1091    
1092      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      case OP_CBRAPOS:
1093        {      case OP_SCBRAPOS:
1094        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);  
       }  
1095    
1096      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      POSSESSIVE_CAPTURE:
1097        {      number = GET2(ecode, 1+LINK_SIZE);
1098        condition = FALSE;      offset = number << 1;
       ecode += GET(ecode, 1);  
       }  
1099    
1100      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1101      the final argument match_condassert causes it to stop at the end of an      printf("start possessive bracket %d\n", number);
1102      assertion. */      printf("subject=");
1103        pchars(eptr, 16, TRUE, md);
1104        printf("\n");
1105    #endif
1106    
1107      else      if (offset < md->offset_max)
1108        {        {
1109        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1110            match_condassert);        code_offset = (int)(ecode - md->start_code);
1111        if (rrc == MATCH_MATCH)  
1112          {        save_offset1 = md->offset_vector[offset];
1113          condition = TRUE;        save_offset2 = md->offset_vector[offset+1];
1114          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        save_offset3 = md->offset_vector[md->offset_end - number];
1115          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
1116    
1117          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1118    
1119          /* Each time round the loop, save the current subject position for use
1120          when the group matches. For MATCH_MATCH, the group has matched, so we
1121          restart it with a new subject starting position, remembering that we had
1122          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1123          usual. If we haven't matched any alternatives in any iteration, check to
1124          see if a previous iteration matched. If so, the group has matched;
1125          continue from afterwards. Otherwise it has failed; restore the previous
1126          capture values before returning NOMATCH. */
1127    
1128          for (;;)
1129            {
1130            md->offset_vector[md->offset_end - number] =
1131              (int)(eptr - md->start_subject);
1132            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1133            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1134              eptrb, RM63);
1135            if (rrc == MATCH_KETRPOS)
1136              {
1137              offset_top = md->end_offset_top;
1138              eptr = md->end_match_ptr;
1139              ecode = md->start_code + code_offset;
1140              save_capture_last = md->capture_last;
1141              matched_once = TRUE;
1142              continue;
1143              }
1144    
1145            /* See comment in the code for capturing groups above about handling
1146            THEN. */
1147    
1148            if (rrc == MATCH_THEN)
1149              {
1150              next = ecode + GET(ecode,1);
1151              if (md->start_match_ptr < next &&
1152                  (*ecode == OP_ALT || *next == OP_ALT))
1153                rrc = MATCH_NOMATCH;
1154              }
1155    
1156            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1157            md->capture_last = save_capture_last;
1158            ecode += GET(ecode, 1);
1159            if (*ecode != OP_ALT) break;
1160          }          }
1161        else if (rrc != MATCH_NOMATCH)  
1162          if (!matched_once)
1163          {          {
1164          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1165            md->offset_vector[offset+1] = save_offset2;
1166            md->offset_vector[md->offset_end - number] = save_offset3;
1167          }          }
1168        else  
1169          if (allow_zero || matched_once)
1170          {          {
1171          condition = FALSE;          ecode += 1 + LINK_SIZE;
1172          ecode += GET(ecode, 1);          break;
1173          }          }
       }  
1174    
1175      /* 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;  
1176        }        }
     break;  
   
1177    
1178      /* End of the pattern. If we are in a top-level recursion, we should      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1179      restore the offsets appropriately and continue from after the call. */      as a non-capturing bracket. */
1180    
1181      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1182      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;  
       }  
1183    
1184      /* 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. */  
1185    
1186      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1187      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;  
1188    
1189      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1190      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,
1191      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1192      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1193      this level is identical to the lookahead case. */  
1194        case OP_BRAPOS:
1195        case OP_SBRAPOS:
1196        allow_zero = FALSE;
1197    
1198        POSSESSIVE_NON_CAPTURE:
1199        matched_once = FALSE;
1200        code_offset = (int)(ecode - md->start_code);
1201    
1202      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1203        {        {
1204        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1205        if (rrc == MATCH_MATCH) break;        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1206            eptrb, RM48);
1207          if (rrc == MATCH_KETRPOS)
1208            {
1209            offset_top = md->end_offset_top;
1210            eptr = md->end_match_ptr;
1211            ecode = md->start_code + code_offset;
1212            matched_once = TRUE;
1213            continue;
1214            }
1215    
1216          /* See comment in the code for capturing groups above about handling
1217          THEN. */
1218    
1219          if (rrc == MATCH_THEN)
1220            {
1221            next = ecode + GET(ecode,1);
1222            if (md->start_match_ptr < next &&
1223                (*ecode == OP_ALT || *next == OP_ALT))
1224              rrc = MATCH_NOMATCH;
1225            }
1226    
1227        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1228        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1229          if (*ecode != OP_ALT) break;
1230        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1231    
1232      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1233          {
1234          ecode += 1 + LINK_SIZE;
1235          break;
1236          }
1237        RRETURN(MATCH_NOMATCH);
1238    
1239      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1240    
1241      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1242      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1243        past the end if there is only one branch, but that's OK because that is
1244        exactly what going to the ket would do. */
1245    
1246      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1247      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1248      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1249    
1250      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1251        inserted between OP_COND and an assertion condition. */
1252    
1253      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1254        {        {
1255        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (PUBL(callout) != NULL)
1256        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          {
1257        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          PUBL(callout_block) cb;
1258        ecode += GET(ecode,1);          cb.version          = 2;   /* Version 1 of the callout block */
1259            cb.callout_number   = ecode[LINK_SIZE+2];
1260            cb.offset_vector    = md->offset_vector;
1261    #ifdef COMPILE_PCRE8
1262            cb.subject          = (PCRE_SPTR)md->start_subject;
1263    #else
1264            cb.subject          = (PCRE_SPTR16)md->start_subject;
1265    #endif
1266            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1267            cb.start_match      = (int)(mstart - md->start_subject);
1268            cb.current_position = (int)(eptr - md->start_subject);
1269            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1270            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1271            cb.capture_top      = offset_top/2;
1272            cb.capture_last     = md->capture_last;
1273            cb.callout_data     = md->callout_data;
1274            cb.mark             = md->nomatch_mark;
1275            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1276            if (rrc < 0) RRETURN(rrc);
1277            }
1278          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1279        }        }
     while (*ecode == OP_ALT);  
1280    
1281      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      condcode = ecode[LINK_SIZE+1];
1282    
1283      ecode += 1 + LINK_SIZE;      /* Now see what the actual condition is */
1284    
1285        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1286          {
1287          if (md->recursive == NULL)                /* Not recursing => FALSE */
1288            {
1289            condition = FALSE;
1290            ecode += GET(ecode, 1);
1291            }
1292          else
1293            {
1294            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1295            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1296    
1297            /* If the test is for recursion into a specific subpattern, and it is
1298            false, but the test was set up by name, scan the table to see if the
1299            name refers to any other numbers, and test them. The condition is true
1300            if any one is set. */
1301    
1302            if (!condition && condcode == OP_NRREF)
1303              {
1304              pcre_uchar *slotA = md->name_table;
1305              for (i = 0; i < md->name_count; i++)
1306                {
1307                if (GET2(slotA, 0) == recno) break;
1308                slotA += md->name_entry_size;
1309                }
1310    
1311              /* Found a name for the number - there can be only one; duplicate
1312              names for different numbers are allowed, but not vice versa. First
1313              scan down for duplicates. */
1314    
1315              if (i < md->name_count)
1316                {
1317                pcre_uchar *slotB = slotA;
1318                while (slotB > md->name_table)
1319                  {
1320                  slotB -= md->name_entry_size;
1321                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1322                    {
1323                    condition = GET2(slotB, 0) == md->recursive->group_num;
1324                    if (condition) break;
1325                    }
1326                  else break;
1327                  }
1328    
1329                /* Scan up for duplicates */
1330    
1331                if (!condition)
1332                  {
1333                  slotB = slotA;
1334                  for (i++; i < md->name_count; i++)
1335                    {
1336                    slotB += md->name_entry_size;
1337                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1338                      {
1339                      condition = GET2(slotB, 0) == md->recursive->group_num;
1340                      if (condition) break;
1341                      }
1342                    else break;
1343                    }
1344                  }
1345                }
1346              }
1347    
1348            /* Chose branch according to the condition */
1349    
1350            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1351            }
1352          }
1353    
1354        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1355          {
1356          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1357          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1358    
1359          /* If the numbered capture is unset, but the reference was by name,
1360          scan the table to see if the name refers to any other numbers, and test
1361          them. The condition is true if any one is set. This is tediously similar
1362          to the code above, but not close enough to try to amalgamate. */
1363    
1364          if (!condition && condcode == OP_NCREF)
1365            {
1366            int refno = offset >> 1;
1367            pcre_uchar *slotA = md->name_table;
1368    
1369            for (i = 0; i < md->name_count; i++)
1370              {
1371              if (GET2(slotA, 0) == refno) break;
1372              slotA += md->name_entry_size;
1373              }
1374    
1375            /* Found a name for the number - there can be only one; duplicate names
1376            for different numbers are allowed, but not vice versa. First scan down
1377            for duplicates. */
1378    
1379            if (i < md->name_count)
1380              {
1381              pcre_uchar *slotB = slotA;
1382              while (slotB > md->name_table)
1383                {
1384                slotB -= md->name_entry_size;
1385                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1386                  {
1387                  offset = GET2(slotB, 0) << 1;
1388                  condition = offset < offset_top &&
1389                    md->offset_vector[offset] >= 0;
1390                  if (condition) break;
1391                  }
1392                else break;
1393                }
1394    
1395              /* Scan up for duplicates */
1396    
1397              if (!condition)
1398                {
1399                slotB = slotA;
1400                for (i++; i < md->name_count; i++)
1401                  {
1402                  slotB += md->name_entry_size;
1403                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1404                    {
1405                    offset = GET2(slotB, 0) << 1;
1406                    condition = offset < offset_top &&
1407                      md->offset_vector[offset] >= 0;
1408                    if (condition) break;
1409                    }
1410                  else break;
1411                  }
1412                }
1413              }
1414            }
1415    
1416          /* Chose branch according to the condition */
1417    
1418          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1419          }
1420    
1421        else if (condcode == OP_DEF)     /* DEFINE - always false */
1422          {
1423          condition = FALSE;
1424          ecode += GET(ecode, 1);
1425          }
1426    
1427        /* The condition is an assertion. Call match() to evaluate it - setting
1428        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1429        an assertion. */
1430    
1431        else
1432          {
1433          md->match_function_type = MATCH_CONDASSERT;
1434          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1435          if (rrc == MATCH_MATCH)
1436            {
1437            if (md->end_offset_top > offset_top)
1438              offset_top = md->end_offset_top;  /* Captures may have happened */
1439            condition = TRUE;
1440            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1441            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1442            }
1443    
1444          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1445          assertion; it is therefore treated as NOMATCH. */
1446    
1447          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1448            {
1449            RRETURN(rrc);         /* Need braces because of following else */
1450            }
1451          else
1452            {
1453            condition = FALSE;
1454            ecode += codelink;
1455            }
1456          }
1457    
1458        /* We are now at the branch that is to be obeyed. As there is only one, can
1459        use tail recursion to avoid using another stack frame, except when there is
1460        unlimited repeat of a possibly empty group. In the latter case, a recursive
1461        call to match() is always required, unless the second alternative doesn't
1462        exist, in which case we can just plough on. Note that, for compatibility
1463        with Perl, the | in a conditional group is NOT treated as creating two
1464        alternatives. If a THEN is encountered in the branch, it propagates out to
1465        the enclosing alternative (unless nested in a deeper set of alternatives,
1466        of course). */
1467    
1468        if (condition || *ecode == OP_ALT)
1469          {
1470          if (op != OP_SCOND)
1471            {
1472            ecode += 1 + LINK_SIZE;
1473            goto TAIL_RECURSE;
1474            }
1475    
1476          md->match_function_type = MATCH_CBEGROUP;
1477          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1478          RRETURN(rrc);
1479          }
1480    
1481         /* Condition false & no alternative; continue after the group. */
1482    
1483        else
1484          {
1485          ecode += 1 + LINK_SIZE;
1486          }
1487        break;
1488    
1489    
1490        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1491        to close any currently open capturing brackets. */
1492    
1493        case OP_CLOSE:
1494        number = GET2(ecode, 1);
1495        offset = number << 1;
1496    
1497    #ifdef PCRE_DEBUG
1498          printf("end bracket %d at *ACCEPT", number);
1499          printf("\n");
1500    #endif
1501    
1502        md->capture_last = number;
1503        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1504          {
1505          md->offset_vector[offset] =
1506            md->offset_vector[md->offset_end - number];
1507          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1508          if (offset_top <= offset) offset_top = offset + 2;
1509          }
1510        ecode += 1 + IMM2_SIZE;
1511        break;
1512    
1513    
1514        /* End of the pattern, either real or forced. */
1515    
1516        case OP_END:
1517        case OP_ACCEPT:
1518        case OP_ASSERT_ACCEPT:
1519    
1520        /* If we have matched an empty string, fail if not in an assertion and not
1521        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1522        is set and we have matched at the start of the subject. In both cases,
1523        backtracking will then try other alternatives, if any. */
1524    
1525        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1526             md->recursive == NULL &&
1527             (md->notempty ||
1528               (md->notempty_atstart &&
1529                 mstart == md->start_subject + md->start_offset)))
1530          RRETURN(MATCH_NOMATCH);
1531    
1532        /* Otherwise, we have a match. */
1533    
1534        md->end_match_ptr = eptr;           /* Record where we ended */
1535        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1536        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1537    
1538        /* For some reason, the macros don't work properly if an expression is
1539        given as the argument to RRETURN when the heap is in use. */
1540    
1541        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1542        RRETURN(rrc);
1543    
1544        /* Assertion brackets. Check the alternative branches in turn - the
1545        matching won't pass the KET for an assertion. If any one branch matches,
1546        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1547        start of each branch to move the current point backwards, so the code at
1548        this level is identical to the lookahead case. When the assertion is part
1549        of a condition, we want to return immediately afterwards. The caller of
1550        this incarnation of the match() function will have set MATCH_CONDASSERT in
1551        md->match_function type, and one of these opcodes will be the first opcode
1552        that is processed. We use a local variable that is preserved over calls to
1553        match() to remember this case. */
1554    
1555        case OP_ASSERT:
1556        case OP_ASSERTBACK:
1557        save_mark = md->mark;
1558        if (md->match_function_type == MATCH_CONDASSERT)
1559          {
1560          condassert = TRUE;
1561          md->match_function_type = 0;
1562          }
1563        else condassert = FALSE;
1564    
1565        do
1566          {
1567          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1568          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1569            {
1570            mstart = md->start_match_ptr;   /* In case \K reset it */
1571            break;
1572            }
1573    
1574          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1575          as NOMATCH. */
1576    
1577          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1578          ecode += GET(ecode, 1);
1579          md->mark = save_mark;
1580          }
1581        while (*ecode == OP_ALT);
1582    
1583        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1584    
1585        /* If checking an assertion for a condition, return MATCH_MATCH. */
1586    
1587        if (condassert) RRETURN(MATCH_MATCH);
1588    
1589        /* Continue from after the assertion, updating the offsets high water
1590        mark, since extracts may have been taken during the assertion. */
1591    
1592        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1593        ecode += 1 + LINK_SIZE;
1594        offset_top = md->end_offset_top;
1595        continue;
1596    
1597        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1598        PRUNE, or COMMIT means we must assume failure without checking subsequent
1599        branches. */
1600    
1601        case OP_ASSERT_NOT:
1602        case OP_ASSERTBACK_NOT:
1603        save_mark = md->mark;
1604        if (md->match_function_type == MATCH_CONDASSERT)
1605          {
1606          condassert = TRUE;
1607          md->match_function_type = 0;
1608          }
1609        else condassert = FALSE;
1610    
1611        do
1612          {
1613          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1614          md->mark = save_mark;
1615          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1616          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1617            {
1618            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1619            break;
1620            }
1621    
1622          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1623          as NOMATCH. */
1624    
1625          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1626          ecode += GET(ecode,1);
1627          }
1628        while (*ecode == OP_ALT);
1629    
1630        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1631    
1632        ecode += 1 + LINK_SIZE;
1633      continue;      continue;
1634    
1635      /* Move the subject pointer back. This occurs only at the start of      /* Move the subject pointer back. This occurs only at the start of
# Line 846  for (;;) Line 1638  for (;;)
1638      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1639    
1640      case OP_REVERSE:      case OP_REVERSE:
1641  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1642      if (utf8)      if (utf)
1643        {        {
1644        i = GET(ecode, 1);        i = GET(ecode, 1);
1645        while (i-- > 0)        while (i-- > 0)
1646          {          {
1647          eptr--;          eptr--;
1648          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1649          BACKCHAR(eptr)          BACKCHAR(eptr);
1650          }          }
1651        }        }
1652      else      else
# Line 867  for (;;) Line 1659  for (;;)
1659        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1660        }        }
1661    
1662      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1663    
1664        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1665      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1666      break;      break;
1667    
# Line 877  for (;;) Line 1670  for (;;)
1670      function is able to force a failure. */      function is able to force a failure. */
1671    
1672      case OP_CALLOUT:      case OP_CALLOUT:
1673      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1674        {        {
1675        pcre_callout_block cb;        PUBL(callout_block) cb;
1676        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1677        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1678        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1679    #ifdef COMPILE_PCRE8
1680        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1681        cb.subject_length   = md->end_subject - md->start_subject;  #else
1682        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1683        cb.current_position = eptr - md->start_subject;  #endif
1684          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1685          cb.start_match      = (int)(mstart - md->start_subject);
1686          cb.current_position = (int)(eptr - md->start_subject);
1687        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1688        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1689        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1690        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1691        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1692        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1693          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1694        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1695        }        }
1696      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 902  for (;;) Line 1700  for (;;)
1700      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
1701      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1702    
1703      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1704      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
1705      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
1706      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
1707      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
1708      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
1709      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.  
1710    
1711      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
1712      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
1713      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1714        a lot, so he is not to blame for the current way it works. */
1715    
1716      case OP_RECURSE:      case OP_RECURSE:
1717        {        {
1718          recursion_info *ri;
1719          int recno;
1720    
1721        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1722        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1723          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1724    
1725          /* Check for repeating a recursion without advancing the subject pointer.
1726          This should catch convoluted mutual recursions. (Some simple cases are
1727          caught at compile time.) */
1728    
1729          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1730            if (recno == ri->group_num && eptr == ri->subject_position)
1731              RRETURN(PCRE_ERROR_RECURSELOOP);
1732    
1733        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1734    
1735          new_recursive.group_num = recno;
1736          new_recursive.subject_position = eptr;
1737        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1738        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1739    
1740        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1741    
1742        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1743    
1744        /* Now save the offset data. */        /* Now save the offset data */
1745    
1746        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1747        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 941  for (;;) Line 1749  for (;;)
1749        else        else
1750          {          {
1751          new_recursive.offset_save =          new_recursive.offset_save =
1752            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1753          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1754          }          }
   
1755        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1756              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1757    
1758        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1759        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1760          might be changed, so reset it before looping. */
1761    
1762        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1763        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1764        do        do
1765          {          {
1766          RMATCH(rrc, eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1767            md, ims, eptrb, flags);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1768          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1769            memcpy(md->offset_vector, new_recursive.offset_save,
1770                new_recursive.saved_max * sizeof(int));
1771            md->recursive = new_recursive.prevrec;
1772            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1773            {            {
1774            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1775            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1776              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1777            RRETURN(MATCH_MATCH);  
1778              /* Set where we got to in the subject, and reset the start in case
1779              it was changed by \K. This *is* propagated back out of a recursion,
1780              for Perl compatibility. */
1781    
1782              eptr = md->end_match_ptr;
1783              mstart = md->start_match_ptr;
1784              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1785            }            }
1786          else if (rrc != MATCH_NOMATCH)  
1787            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1788            as NOMATCH. */
1789    
1790            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1791            {            {
1792            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1793              if (new_recursive.offset_save != stacksave)
1794                (PUBL(free))(new_recursive.offset_save);
1795            RRETURN(rrc);            RRETURN(rrc);
1796            }            }
1797    
1798          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1799          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1800          }          }
1801        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 983  for (;;) Line 1803  for (;;)
1803        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1804        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1805        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1806          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1807        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1808        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(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);  
1809    
1810      offset_top = md->end_offset_top;      RECURSION_MATCHED:
1811      eptr = md->end_match_ptr;      break;
   
     /* 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 */  
1812    
1813      /* 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
1814      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1070  for (;;) Line 1817  for (;;)
1817      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1818      break;      break;
1819    
1820      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1821      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
1822      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
1823      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
1824      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1825    
1826      case OP_BRAZERO:      case OP_BRAZERO:
1827        {      next = ecode + 1;
1828        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1829        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1830        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1831        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1832      break;      break;
1833    
1834      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1835        {      next = ecode + 1;
1836        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1837        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1838        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1839        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1840      break;      break;
1841    
1842        case OP_SKIPZERO:
1843        next = ecode+1;
1844        do next += GET(next,1); while (*next == OP_ALT);
1845        ecode = next + 1 + LINK_SIZE;
1846        break;
1847    
1848        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1849        here; just jump to the group, with allow_zero set TRUE. */
1850    
1851        case OP_BRAPOSZERO:
1852        op = *(++ecode);
1853        allow_zero = TRUE;
1854        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1855          goto POSSESSIVE_NON_CAPTURE;
1856    
1857      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1858    
1859      case OP_KET:      case OP_KET:
1860      case OP_KETRMIN:      case OP_KETRMIN:
1861      case OP_KETRMAX:      case OP_KETRMAX:
1862        case OP_KETRPOS:
1863      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1864    
1865      /* 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
1866      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1867      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1868    
1869      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1870        {        {
1871        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1872        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1873        }        }
1874      else saved_eptr = NULL;      else saved_eptr = NULL;
1875    
1876      /* 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
1877      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
1878      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1879        start in case it was changed by \K. */
1880      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1881          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1882          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1883        {        {
1884        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1885        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1886        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1887          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1888        }        }
1889    
1890      /* 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
1891      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1892      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
1893      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
1894      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
1895        the current subject position and start match pointer and give a MATCH
1896        return. */
1897    
1898      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1899            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1900        {        {
1901        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1902        offset = number << 1;        offset = number << 1;
1903    
1904  #ifdef DEBUG  #ifdef PCRE_DEBUG
1905        printf("end bracket %d", number);        printf("end bracket %d", number);
1906        printf("\n");        printf("\n");
1907  #endif  #endif
1908    
1909          /* Handle a recursively called group. */
1910    
1911          if (md->recursive != NULL && md->recursive->group_num == number)
1912            {
1913            md->end_match_ptr = eptr;
1914            md->start_match_ptr = mstart;
1915            RRETURN(MATCH_MATCH);
1916            }
1917    
1918          /* Deal with capturing */
1919    
1920        md->capture_last = number;        md->capture_last = number;
1921        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1922          {          {
1923            /* If offset is greater than offset_top, it means that we are
1924            "skipping" a capturing group, and that group's offsets must be marked
1925            unset. In earlier versions of PCRE, all the offsets were unset at the
1926            start of matching, but this doesn't work because atomic groups and
1927            assertions can cause a value to be set that should later be unset.
1928            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1929            part of the atomic group, but this is not on the final matching path,
1930            so must be unset when 2 is set. (If there is no group 2, there is no
1931            problem, because offset_top will then be 2, indicating no capture.) */
1932    
1933            if (offset > offset_top)
1934              {
1935              register int *iptr = md->offset_vector + offset_top;
1936              register int *iend = md->offset_vector + offset;
1937              while (iptr < iend) *iptr++ = -1;
1938              }
1939    
1940            /* Now make the extraction */
1941    
1942          md->offset_vector[offset] =          md->offset_vector[offset] =
1943            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1944          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1945          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1946          }          }
1947          }
1948    
1949        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1950        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1951        group. This is the forcible breaking of infinite loops as implemented in
1952        Perl 5.005. For a non-repeating atomic group that includes captures,
1953        establish a backup point by processing the rest of the pattern at a lower
1954        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1955        original OP_ONCE level, thereby bypassing intermediate backup points, but
1956        resetting any captures that happened along the way. */
1957    
1958        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1959          {
1960          if (*prev == OP_ONCE)
1961          {          {
1962          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1963          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1964          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1965          md->start_match = rec->save_start;          RRETURN(MATCH_ONCE);
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
1966          }          }
1967          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1968          break;
1969        }        }
1970    
1971      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1972      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1973        at a time from the outer level, thus saving stack. */
     ims = original_ims;  
     DPRINTF(("ims reset to %02lx\n", ims));  
1974    
1975      /* For a non-repeating ket, just continue at this level. This also      if (*ecode == OP_KETRPOS)
     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)  
1976        {        {
1977        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1978        break;        md->end_offset_top = offset_top;
1979          RRETURN(MATCH_KETRPOS);
1980        }        }
1981    
1982      /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
1983      preceding bracket, in the appropriate order. In the second case, we can use      the preceding bracket, in the appropriate order. In the second case, we can
1984      tail recursion to avoid using another stack frame. */      use tail recursion to avoid using another stack frame, unless we have an
1985        an atomic group or an unlimited repeat of a group that can match an empty
1986      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      string. */
1987    
1988      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1989        {        {
1990        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1991        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1992          if (*prev == OP_ONCE)
1993            {
1994            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1995            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1996            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1997            RRETURN(MATCH_ONCE);
1998            }
1999          if (*prev >= OP_SBRA)    /* Could match an empty string */
2000            {
2001            md->match_function_type = MATCH_CBEGROUP;
2002            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2003            RRETURN(rrc);
2004            }
2005        ecode = prev;        ecode = prev;
       flags |= match_tail_recursed;  
2006        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2007        }        }
2008      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2009        {        {
2010        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, flags);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
2011          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2012          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2013        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2014          if (*prev == OP_ONCE)
2015            {
2016            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2017            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2018            md->once_target = prev;
2019            RRETURN(MATCH_ONCE);
2020            }
2021        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_tail_recursed;  
2022        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2023        }        }
2024      /* Control never gets here */      /* Control never gets here */
2025    
2026      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2027    
2028      case OP_CIRC:      case OP_CIRC:
2029      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 */  
2030    
2031      /* Start of subject assertion */      /* Start of subject assertion */
2032    
# Line 1232  for (;;) Line 2035  for (;;)
2035      ecode++;      ecode++;
2036      break;      break;
2037    
2038        /* Multiline mode: start of subject unless notbol, or after any newline. */
2039    
2040        case OP_CIRCM:
2041        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2042        if (eptr != md->start_subject &&
2043            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2044          RRETURN(MATCH_NOMATCH);
2045        ecode++;
2046        break;
2047    
2048      /* Start of match assertion */      /* Start of match assertion */
2049    
2050      case OP_SOM:      case OP_SOM:
# Line 1239  for (;;) Line 2052  for (;;)
2052      ecode++;      ecode++;
2053      break;      break;
2054    
2055      /* 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. */  
2056    
2057      case OP_DOLL:      case OP_SET_SOM:
2058      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2059        {      ecode++;
2060        if (eptr < md->end_subject)      break;
2061          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
2062        else      /* Multiline mode: assert before any newline, or before end of subject
2063          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
2064        ecode++;  
2065        break;      case OP_DOLLM:
2066        }      if (eptr < md->end_subject)
2067          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
2068      else      else
2069        {        {
2070        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2071        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2072        }        }
2073        ecode++;
2074        break;
2075    
2076        /* Not multiline mode: assert before a terminating newline or before end of
2077        subject unless noteol is set. */
2078    
2079        case OP_DOLL:
2080        if (md->noteol) RRETURN(MATCH_NOMATCH);
2081        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2082    
2083      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2084    
2085      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2086    
2087      case OP_EOD:      case OP_EOD:
2088      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2089        SCHECK_PARTIAL();
2090      ecode++;      ecode++;
2091      break;      break;
2092    
2093      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2094    
2095      case OP_EODN:      case OP_EODN:
2096      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2097        if (eptr < md->end_subject &&
2098          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2099        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2100    
2101        /* Either at end of string or \n before end. */
2102    
2103        SCHECK_PARTIAL();
2104      ecode++;      ecode++;
2105      break;      break;
2106    
# Line 1290  for (;;) Line 2112  for (;;)
2112    
2113        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2114        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
2115        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2116          partial matching. */
2117    
2118  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2119        if (utf8)        if (utf)
2120          {          {
2121            /* Get status of previous character */
2122    
2123          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2124            {            {
2125            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2126            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2127              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2128            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2129    #ifdef SUPPORT_UCP
2130              if (md->use_ucp)
2131                {
2132                if (c == '_') prev_is_word = TRUE; else
2133                  {
2134                  int cat = UCD_CATEGORY(c);
2135                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2136                  }
2137                }
2138              else
2139    #endif
2140            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2141            }            }
2142          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2143            /* Get status of next character */
2144    
2145            if (eptr >= md->end_subject)
2146              {
2147              SCHECK_PARTIAL();
2148              cur_is_word = FALSE;
2149              }
2150            else
2151            {            {
2152            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2153    #ifdef SUPPORT_UCP
2154              if (md->use_ucp)
2155                {
2156                if (c == '_') cur_is_word = TRUE; else
2157                  {
2158                  int cat = UCD_CATEGORY(c);
2159                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2160                  }
2161                }
2162              else
2163    #endif
2164            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2165            }            }
2166          }          }
2167        else        else
2168  #endif  #endif
2169    
2170        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2171          consistency with the behaviour of \w we do use it in this case. */
2172    
2173          {          {
2174          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2175            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2176          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2177            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2178              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2179    #ifdef SUPPORT_UCP
2180              if (md->use_ucp)
2181                {
2182                c = eptr[-1];
2183                if (c == '_') prev_is_word = TRUE; else
2184                  {
2185                  int cat = UCD_CATEGORY(c);
2186                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2187                  }
2188                }
2189              else
2190    #endif
2191              prev_is_word = MAX_255(eptr[-1])
2192                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2193              }
2194    
2195            /* Get status of next character */
2196    
2197            if (eptr >= md->end_subject)
2198              {
2199              SCHECK_PARTIAL();
2200              cur_is_word = FALSE;
2201              }
2202            else
2203    #ifdef SUPPORT_UCP
2204            if (md->use_ucp)
2205              {
2206              c = *eptr;
2207              if (c == '_') cur_is_word = TRUE; else
2208                {
2209                int cat = UCD_CATEGORY(c);
2210                cur_is_word = (cat == ucp_L || cat == ucp_N);
2211                }
2212              }
2213            else
2214    #endif
2215            cur_is_word = MAX_255(*eptr)
2216              && ((md->ctypes[*eptr] & ctype_word) != 0);
2217          }          }
2218    
2219        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1331  for (;;) Line 2227  for (;;)
2227      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2228    
2229      case OP_ANY:      case OP_ANY:
2230      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2231        {      /* Fall through */
2232        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);  
2233        case OP_ALLANY:
2234        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2235          {                            /* not be updated before SCHECK_PARTIAL. */
2236          SCHECK_PARTIAL();
2237          RRETURN(MATCH_NOMATCH);
2238        }        }
2239      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2240      if (utf8)  #ifdef SUPPORT_UTF
2241        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2242    #endif
2243      ecode++;      ecode++;
2244      break;      break;
2245    
# Line 1345  for (;;) Line 2247  for (;;)
2247      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2248    
2249      case OP_ANYBYTE:      case OP_ANYBYTE:
2250      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2251          {                            /* not be updated before SCHECK_PARTIAL. */
2252          SCHECK_PARTIAL();
2253          RRETURN(MATCH_NOMATCH);
2254          }
2255        eptr++;
2256      ecode++;      ecode++;
2257      break;      break;
2258    
2259      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2260      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2261          {
2262          SCHECK_PARTIAL();
2263          RRETURN(MATCH_NOMATCH);
2264          }
2265      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2266      if (      if (
2267  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2268         c < 256 &&         c < 256 &&
2269  #endif  #endif
2270         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1363  for (;;) Line 2274  for (;;)
2274      break;      break;
2275    
2276      case OP_DIGIT:      case OP_DIGIT:
2277      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2278          {
2279          SCHECK_PARTIAL();
2280          RRETURN(MATCH_NOMATCH);
2281          }
2282      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2283      if (      if (
2284  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2285         c >= 256 ||         c > 255 ||
2286  #endif  #endif
2287         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2288         )         )
# Line 1376  for (;;) Line 2291  for (;;)
2291      break;      break;
2292    
2293      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2294      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2295          {
2296          SCHECK_PARTIAL();
2297          RRETURN(MATCH_NOMATCH);
2298          }
2299      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2300      if (      if (
2301  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2302         c < 256 &&         c < 256 &&
2303  #endif  #endif
2304         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1389  for (;;) Line 2308  for (;;)
2308      break;      break;
2309    
2310      case OP_WHITESPACE:      case OP_WHITESPACE:
2311      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2312          {
2313          SCHECK_PARTIAL();
2314          RRETURN(MATCH_NOMATCH);
2315          }
2316      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2317      if (      if (
2318  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2319         c >= 256 ||         c > 255 ||
2320  #endif  #endif
2321         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2322         )         )
# Line 1402  for (;;) Line 2325  for (;;)
2325      break;      break;
2326    
2327      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2328      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2329          {
2330          SCHECK_PARTIAL();
2331          RRETURN(MATCH_NOMATCH);
2332          }
2333      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2334      if (      if (
2335  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2336         c < 256 &&         c < 256 &&
2337  #endif  #endif
2338         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1415  for (;;) Line 2342  for (;;)
2342      break;      break;
2343    
2344      case OP_WORDCHAR:      case OP_WORDCHAR:
2345      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2346          {
2347          SCHECK_PARTIAL();
2348          RRETURN(MATCH_NOMATCH);
2349          }
2350      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2351      if (      if (
2352  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2353         c >= 256 ||         c > 255 ||
2354  #endif  #endif
2355         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2356         )         )
# Line 1428  for (;;) Line 2359  for (;;)
2359      break;      break;
2360    
2361      case OP_ANYNL:      case OP_ANYNL:
2362      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2363          {
2364          SCHECK_PARTIAL();
2365          RRETURN(MATCH_NOMATCH);
2366          }
2367      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2368      switch(c)      switch(c)
2369        {        {
2370        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2371    
2372        case 0x000d:        case 0x000d:
2373        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2374        break;        break;
2375    
2376        case 0x000a:        case 0x000a:
2377          break;
2378    
2379        case 0x000b:        case 0x000b:
2380        case 0x000c:        case 0x000c:
2381        case 0x0085:        case 0x0085:
2382        case 0x2028:        case 0x2028:
2383        case 0x2029:        case 0x2029:
2384          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2385          break;
2386          }
2387        ecode++;
2388        break;
2389    
2390        case OP_NOT_HSPACE:
2391        if (eptr >= md->end_subject)
2392          {
2393          SCHECK_PARTIAL();
2394          RRETURN(MATCH_NOMATCH);
2395          }
2396        GETCHARINCTEST(c, eptr);
2397        switch(c)
2398          {
2399          default: break;
2400          case 0x09:      /* HT */
2401          case 0x20:      /* SPACE */
2402          case 0xa0:      /* NBSP */
2403          case 0x1680:    /* OGHAM SPACE MARK */
2404          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2405          case 0x2000:    /* EN QUAD */
2406          case 0x2001:    /* EM QUAD */
2407          case 0x2002:    /* EN SPACE */
2408          case 0x2003:    /* EM SPACE */
2409          case 0x2004:    /* THREE-PER-EM SPACE */
2410          case 0x2005:    /* FOUR-PER-EM SPACE */
2411          case 0x2006:    /* SIX-PER-EM SPACE */
2412          case 0x2007:    /* FIGURE SPACE */
2413          case 0x2008:    /* PUNCTUATION SPACE */
2414          case 0x2009:    /* THIN SPACE */
2415          case 0x200A:    /* HAIR SPACE */
2416          case 0x202f:    /* NARROW NO-BREAK SPACE */
2417          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2418          case 0x3000:    /* IDEOGRAPHIC SPACE */
2419          RRETURN(MATCH_NOMATCH);
2420          }
2421        ecode++;
2422        break;
2423    
2424        case OP_HSPACE:
2425        if (eptr >= md->end_subject)
2426          {
2427          SCHECK_PARTIAL();
2428          RRETURN(MATCH_NOMATCH);
2429          }
2430        GETCHARINCTEST(c, eptr);
2431        switch(c)
2432          {
2433          default: RRETURN(MATCH_NOMATCH);
2434          case 0x09:      /* HT */
2435          case 0x20:      /* SPACE */
2436          case 0xa0:      /* NBSP */
2437          case 0x1680:    /* OGHAM SPACE MARK */
2438          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2439          case 0x2000:    /* EN QUAD */
2440          case 0x2001:    /* EM QUAD */
2441          case 0x2002:    /* EN SPACE */
2442          case 0x2003:    /* EM SPACE */
2443          case 0x2004:    /* THREE-PER-EM SPACE */
2444          case 0x2005:    /* FOUR-PER-EM SPACE */
2445          case 0x2006:    /* SIX-PER-EM SPACE */
2446          case 0x2007:    /* FIGURE SPACE */
2447          case 0x2008:    /* PUNCTUATION SPACE */
2448          case 0x2009:    /* THIN SPACE */
2449          case 0x200A:    /* HAIR SPACE */
2450          case 0x202f:    /* NARROW NO-BREAK SPACE */
2451          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2452          case 0x3000:    /* IDEOGRAPHIC SPACE */
2453          break;
2454          }
2455        ecode++;
2456        break;
2457    
2458        case OP_NOT_VSPACE:
2459        if (eptr >= md->end_subject)
2460          {
2461          SCHECK_PARTIAL();
2462          RRETURN(MATCH_NOMATCH);
2463          }
2464        GETCHARINCTEST(c, eptr);
2465        switch(c)
2466          {
2467          default: break;
2468          case 0x0a:      /* LF */
2469          case 0x0b:      /* VT */
2470          case 0x0c:      /* FF */
2471          case 0x0d:      /* CR */
2472          case 0x85:      /* NEL */
2473          case 0x2028:    /* LINE SEPARATOR */
2474          case 0x2029:    /* PARAGRAPH SEPARATOR */
2475          RRETURN(MATCH_NOMATCH);
2476          }
2477        ecode++;
2478        break;
2479    
2480        case OP_VSPACE:
2481        if (eptr >= md->end_subject)
2482          {
2483          SCHECK_PARTIAL();
2484          RRETURN(MATCH_NOMATCH);
2485          }
2486        GETCHARINCTEST(c, eptr);
2487        switch(c)
2488          {
2489          default: RRETURN(MATCH_NOMATCH);
2490          case 0x0a:      /* LF */
2491          case 0x0b:      /* VT */
2492          case 0x0c:      /* FF */
2493          case 0x0d:      /* CR */
2494          case 0x85:      /* NEL */
2495          case 0x2028:    /* LINE SEPARATOR */
2496          case 0x2029:    /* PARAGRAPH SEPARATOR */
2497        break;        break;
2498        }        }
2499      ecode++;      ecode++;
# Line 1453  for (;;) Line 2505  for (;;)
2505    
2506      case OP_PROP:      case OP_PROP:
2507      case OP_NOTPROP:      case OP_NOTPROP:
2508      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2509          {
2510          SCHECK_PARTIAL();
2511          RRETURN(MATCH_NOMATCH);
2512          }
2513      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2514        {        {
2515        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2516    
2517        switch(ecode[1])        switch(ecode[1])
2518          {          {
# Line 1466  for (;;) Line 2521  for (;;)
2521          break;          break;
2522    
2523          case PT_LAMP:          case PT_LAMP:
2524          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2525               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2526               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2527            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2528           break;          break;
2529    
2530          case PT_GC:          case PT_GC:
2531          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2532            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2533          break;          break;
2534    
2535          case PT_PC:          case PT_PC:
2536          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2537            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2538          break;          break;
2539    
2540          case PT_SC:          case PT_SC:
2541          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2542              RRETURN(MATCH_NOMATCH);
2543            break;
2544    
2545            /* These are specials */
2546    
2547            case PT_ALNUM:
2548            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2549                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2550              RRETURN(MATCH_NOMATCH);
2551            break;
2552    
2553            case PT_SPACE:    /* Perl space */
2554            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2555                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2556                   == (op == OP_NOTPROP))
2557            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2558          break;          break;
2559    
2560            case PT_PXSPACE:  /* POSIX space */
2561            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2562                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2563                 c == CHAR_FF || c == CHAR_CR)
2564                   == (op == OP_NOTPROP))
2565              RRETURN(MATCH_NOMATCH);
2566            break;
2567    
2568            case PT_WORD:
2569            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2570                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2571                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2572              RRETURN(MATCH_NOMATCH);
2573            break;
2574    
2575            /* This should never occur */
2576    
2577          default:          default:
2578          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2579          }          }
# Line 1499  for (;;) Line 2586  for (;;)
2586      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2587    
2588      case OP_EXTUNI:      case OP_EXTUNI:
2589      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2590          {
2591          SCHECK_PARTIAL();
2592          RRETURN(MATCH_NOMATCH);
2593          }
2594      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2595        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2596        while (eptr < md->end_subject)
2597        {        {
2598        int chartype, script;        int len = 1;
2599        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2600        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2601        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;  
         }  
2602        }        }
2603      ecode++;      ecode++;
2604      break;      break;
# Line 1531  for (;;) Line 2614  for (;;)
2614      loops). */      loops). */
2615    
2616      case OP_REF:      case OP_REF:
2617        {      case OP_REFI:
2618        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2619        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2620        ecode += 1 + IMM2_SIZE;
2621    
2622        /* 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];  
2623    
2624        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2625        this ensures that every attempt at a match fails. We can't just fail
2626        here, because of the possibility of quantifiers with zero minima.
2627    
2628        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2629          {      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;  
2630    
2631          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2632          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;  
2633    
2634          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2635          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2636          eptr += length;      else
2637          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2638    
2639        /* Set up for repetition, or handle the non-repeated case */
2640    
2641        switch (*ecode)
2642          {
2643          case OP_CRSTAR:
2644          case OP_CRMINSTAR:
2645          case OP_CRPLUS:
2646          case OP_CRMINPLUS:
2647          case OP_CRQUERY:
2648          case OP_CRMINQUERY:
2649          c = *ecode++ - OP_CRSTAR;
2650          minimize = (c & 1) != 0;
2651          min = rep_min[c];                 /* Pick up values from tables; */
2652          max = rep_max[c];                 /* zero for max => infinity */
2653          if (max == 0) max = INT_MAX;
2654          break;
2655    
2656          case OP_CRRANGE:
2657          case OP_CRMINRANGE:
2658          minimize = (*ecode == OP_CRMINRANGE);
2659          min = GET2(ecode, 1);
2660          max = GET2(ecode, 1 + IMM2_SIZE);
2661          if (max == 0) max = INT_MAX;
2662          ecode += 1 + 2 * IMM2_SIZE;
2663          break;
2664    
2665          default:               /* No repeat follows */
2666          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2667            {
2668            CHECK_PARTIAL();
2669            RRETURN(MATCH_NOMATCH);
2670          }          }
2671          eptr += length;
2672          continue;              /* With the main loop */
2673          }
2674    
2675        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2676        main loop. */      zero, just continue with the main loop. If the length is negative, it
2677        means the reference is unset in non-Java-compatible mode. If the minimum is
2678        zero, we can continue at the same level without recursion. For any other
2679        minimum, carrying on will result in NOMATCH. */
2680    
2681        if (length == 0) continue;      if (length == 0) continue;
2682        if (length < 0 && min == 0) continue;
2683    
2684        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2685        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2686        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2687    
2688        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2689          {
2690          int slength;
2691          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2692          {          {
2693          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2694          eptr += length;          RRETURN(MATCH_NOMATCH);
2695          }          }
2696          eptr += slength;
2697          }
2698    
2699        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2700        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2701    
2702        if (min == max) continue;      if (min == max) continue;
2703    
2704        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2705    
2706        if (minimize)      if (minimize)
2707          {
2708          for (fi = min;; fi++)
2709          {          {
2710          for (fi = min;; fi++)          int slength;
2711            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2712            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2713            if (fi >= max) RRETURN(MATCH_NOMATCH);
2714            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2715            {            {
2716            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2717            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            RRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2718            }            }
2719          /* Control never gets here */          eptr += slength;
2720          }          }
2721          /* Control never gets here */
2722          }
2723    
2724        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2725    
2726        else      else
2727          {
2728          pp = eptr;
2729          for (i = min; i < max; i++)
2730          {          {
2731          pp = eptr;          int slength;
2732          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)  
2733            {            {
2734            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2735            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2736            }            }
2737          RRETURN(MATCH_NOMATCH);          eptr += slength;
2738          }          }
2739          while (eptr >= pp)
2740            {
2741            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2742            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2743            eptr -= length;
2744            }
2745          RRETURN(MATCH_NOMATCH);
2746        }        }
2747      /* Control never gets here */      /* Control never gets here */
2748    
   
   
2749      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2750      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,
2751      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 1648  for (;;) Line 2760  for (;;)
2760      case OP_NCLASS:      case OP_NCLASS:
2761      case OP_CLASS:      case OP_CLASS:
2762        {        {
2763          /* The data variable is saved across frames, so the byte map needs to
2764          be stored there. */
2765    #define BYTE_MAP ((pcre_uint8 *)data)
2766        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2767        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2768    
2769        switch (*ecode)        switch (*ecode)
2770          {          {
# Line 1670  for (;;) Line 2785  for (;;)
2785          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2786          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2787          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2788          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2789          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2790          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2791          break;          break;
2792    
2793          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1682  for (;;) Line 2797  for (;;)
2797    
2798        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2799    
2800  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2801        /* UTF-8 mode */        if (utf)
       if (utf8)  
2802          {          {
2803          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2804            {            {
2805            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2806                {
2807                SCHECK_PARTIAL();
2808                RRETURN(MATCH_NOMATCH);
2809                }
2810            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2811            if (c > 255)            if (c > 255)
2812              {              {
2813              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2814              }              }
2815            else            else
2816              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2817            }            }
2818          }          }
2819        else        else
2820  #endif  #endif
2821        /* Not UTF-8 mode */        /* Not UTF mode */
2822          {          {
2823          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2824            {            {
2825            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2826                {
2827                SCHECK_PARTIAL();
2828                RRETURN(MATCH_NOMATCH);
2829                }
2830            c = *eptr++;            c = *eptr++;
2831            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2832              if (c > 255)
2833                {
2834                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2835                }
2836              else
2837    #endif
2838                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2839            }            }
2840          }          }
2841    
# Line 1722  for (;;) Line 2849  for (;;)
2849    
2850        if (minimize)        if (minimize)
2851          {          {
2852  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2853          /* UTF-8 mode */          if (utf)
         if (utf8)  
2854            {            {
2855            for (fi = min;; fi++)            for (fi = min;; fi++)
2856              {              {
2857              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2858              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2859              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2860                if (eptr >= md->end_subject)
2861                  {
2862                  SCHECK_PARTIAL();
2863                  RRETURN(MATCH_NOMATCH);
2864                  }
2865              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2866              if (c > 255)              if (c > 255)
2867                {                {
2868                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2869                }                }
2870              else              else
2871                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2872              }              }
2873            }            }
2874          else          else
2875  #endif  #endif
2876          /* Not UTF-8 mode */          /* Not UTF mode */
2877            {            {
2878            for (fi = min;; fi++)            for (fi = min;; fi++)
2879              {              {
2880              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2881              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2882              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2883                if (eptr >= md->end_subject)
2884                  {
2885                  SCHECK_PARTIAL();
2886                  RRETURN(MATCH_NOMATCH);
2887                  }
2888              c = *eptr++;              c = *eptr++;
2889              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2890                if (c > 255)
2891                  {
2892                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2893                  }
2894                else
2895    #endif
2896                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2897              }              }
2898            }            }
2899          /* Control never gets here */          /* Control never gets here */
# Line 1764  for (;;) Line 2905  for (;;)
2905          {          {
2906          pp = eptr;          pp = eptr;
2907    
2908  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2909          /* UTF-8 mode */          if (utf)
         if (utf8)  
2910            {            {
2911            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2912              {              {
2913              int len = 1;              int len = 1;
2914              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2915                  {
2916                  SCHECK_PARTIAL();
2917                  break;
2918                  }
2919              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2920              if (c > 255)              if (c > 255)
2921                {                {
2922                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2923                }                }
2924              else              else
2925                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2926              eptr += len;              eptr += len;
2927              }              }
2928            for (;;)            for (;;)
2929              {              {
2930              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2931              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2932              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2933              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1793  for (;;) Line 2935  for (;;)
2935            }            }
2936          else          else
2937  #endif  #endif
2938            /* Not UTF-8 mode */            /* Not UTF mode */
2939            {            {
2940            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2941              {              {
2942              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2943                  {
2944                  SCHECK_PARTIAL();
2945                  break;
2946                  }
2947              c = *eptr;              c = *eptr;
2948              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2949                if (c > 255)
2950                  {
2951                  if (op == OP_CLASS) break;
2952                  }
2953                else
2954    #endif
2955                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2956              eptr++;              eptr++;
2957              }              }
2958            while (eptr >= pp)            while (eptr >= pp)
2959              {              {
2960              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2961              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2962              eptr--;              eptr--;
2963              }              }
# Line 1812  for (;;) Line 2965  for (;;)
2965    
2966          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2967          }          }
2968    #undef BYTE_MAP
2969        }        }
2970      /* Control never gets here */      /* Control never gets here */
2971    
2972    
2973      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2974      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
2975        mode, because Unicode properties are supported in non-UTF-8 mode. */
2976    
2977  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2978      case OP_XCLASS:      case OP_XCLASS:
2979        {        {
2980        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1844  for (;;) Line 2999  for (;;)
2999          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3000          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3001          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3002          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3003          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3004          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3005          break;          break;
3006    
3007          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1858  for (;;) Line 3013  for (;;)
3013    
3014        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3015          {          {
3016          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3017          GETCHARINC(c, eptr);            {
3018          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3019              RRETURN(MATCH_NOMATCH);
3020              }
3021            GETCHARINCTEST(c, eptr);
3022            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3023          }          }
3024    
3025        /* 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 1875  for (;;) Line 3034  for (;;)
3034          {          {
3035          for (fi = min;; fi++)          for (fi = min;; fi++)
3036            {            {
3037            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3038            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3039            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3040            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3041            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3042                SCHECK_PARTIAL();
3043                RRETURN(MATCH_NOMATCH);
3044                }
3045              GETCHARINCTEST(c, eptr);
3046              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3047            }            }
3048          /* Control never gets here */          /* Control never gets here */
3049          }          }
# Line 1892  for (;;) Line 3056  for (;;)
3056          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3057            {            {
3058            int len = 1;            int len = 1;
3059            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3060            GETCHARLEN(c, eptr, len);              {
3061            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3062                break;
3063                }
3064    #ifdef SUPPORT_UTF
3065              GETCHARLENTEST(c, eptr, len);
3066    #else
3067              c = *eptr;
3068    #endif
3069              if (!PRIV(xclass)(c, data, utf)) break;
3070            eptr += len;            eptr += len;
3071            }            }
3072          for(;;)          for(;;)
3073            {            {
3074            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3075            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3076            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3077            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3078              if (utf) BACKCHAR(eptr);
3079    #endif
3080            }            }
3081          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3082          }          }
# Line 1914  for (;;) Line 3088  for (;;)
3088      /* Match a single character, casefully */      /* Match a single character, casefully */
3089    
3090      case OP_CHAR:      case OP_CHAR:
3091  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3092      if (utf8)      if (utf)
3093        {        {
3094        length = 1;        length = 1;
3095        ecode++;        ecode++;
3096        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3097        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3098            {
3099            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3100            RRETURN(MATCH_NOMATCH);
3101            }
3102        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3103        }        }
3104      else      else
3105  #endif  #endif
3106        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3107        {        {
3108        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3109            {
3110            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3111            RRETURN(MATCH_NOMATCH);
3112            }
3113        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3114        ecode += 2;        ecode += 2;
3115        }        }
3116      break;      break;
3117    
3118      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3119        subject, give up immediately. */
3120    
3121        case OP_CHARI:
3122        if (eptr >= md->end_subject)
3123          {
3124          SCHECK_PARTIAL();
3125          RRETURN(MATCH_NOMATCH);
3126          }
3127    
3128      case OP_CHARNC:  #ifdef SUPPORT_UTF
3129  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3130        {        {
3131        length = 1;        length = 1;
3132        ecode++;        ecode++;
3133        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3134    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3135        /* 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
3136        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3137          fast lookup table. We know that there is at least one byte left in the
3138          subject. */
3139    
3140        if (fc < 128)        if (fc < 128)
3141          {          {
3142          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3143                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3144            ecode++;
3145            eptr++;
3146          }          }
3147    
3148        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3149          use the value of "length" to check for sufficient bytes left, because the
3150          other case of the character may have more or fewer bytes.  */
3151    
3152        else        else
3153          {          {
# Line 1968  for (;;) Line 3161  for (;;)
3161          if (fc != dc)          if (fc != dc)
3162            {            {
3163  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3164            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3165  #endif  #endif
3166              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3167            }            }
3168          }          }
3169        }        }
3170      else      else
3171  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3172    
3173      /* Non-UTF-8 mode */      /* Not UTF mode */
3174        {        {
3175        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3176        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3177          eptr++;
3178        ecode += 2;        ecode += 2;
3179        }        }
3180      break;      break;
# Line 1988  for (;;) Line 3182  for (;;)
3182      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3183    
3184      case OP_EXACT:      case OP_EXACT:
3185        case OP_EXACTI:
3186      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3187      ecode += 3;      ecode += 1 + IMM2_SIZE;
3188      goto REPEATCHAR;      goto REPEATCHAR;
3189    
3190      case OP_POSUPTO:      case OP_POSUPTO:
3191        case OP_POSUPTOI:
3192      possessive = TRUE;      possessive = TRUE;
3193      /* Fall through */      /* Fall through */
3194    
3195      case OP_UPTO:      case OP_UPTO:
3196        case OP_UPTOI:
3197      case OP_MINUPTO:      case OP_MINUPTO:
3198        case OP_MINUPTOI:
3199      min = 0;      min = 0;
3200      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3201      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3202      ecode += 3;      ecode += 1 + IMM2_SIZE;
3203      goto REPEATCHAR;      goto REPEATCHAR;
3204    
3205      case OP_POSSTAR:      case OP_POSSTAR:
3206        case OP_POSSTARI:
3207      possessive = TRUE;      possessive = TRUE;
3208      min = 0;      min = 0;
3209      max = INT_MAX;      max = INT_MAX;
# Line 2012  for (;;) Line 3211  for (;;)
3211      goto REPEATCHAR;      goto REPEATCHAR;
3212    
3213      case OP_POSPLUS:      case OP_POSPLUS:
3214        case OP_POSPLUSI:
3215      possessive = TRUE;      possessive = TRUE;
3216      min = 1;      min = 1;
3217      max = INT_MAX;      max = INT_MAX;
# Line 2019  for (;;) Line 3219  for (;;)
3219      goto REPEATCHAR;      goto REPEATCHAR;
3220    
3221      case OP_POSQUERY:      case OP_POSQUERY:
3222        case OP_POSQUERYI:
3223      possessive = TRUE;      possessive = TRUE;
3224      min = 0;      min = 0;
3225      max = 1;      max = 1;
# Line 2026  for (;;) Line 3227  for (;;)
3227      goto REPEATCHAR;      goto REPEATCHAR;
3228    
3229      case OP_STAR:      case OP_STAR:
3230        case OP_STARI:
3231      case OP_MINSTAR:      case OP_MINSTAR:
3232        case OP_MINSTARI:
3233      case OP_PLUS:      case OP_PLUS:
3234        case OP_PLUSI:
3235      case OP_MINPLUS:      case OP_MINPLUS:
3236        case OP_MINPLUSI:
3237      case OP_QUERY:      case OP_QUERY:
3238        case OP_QUERYI:
3239      case OP_MINQUERY:      case OP_MINQUERY:
3240      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3241        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3242      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3243      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3244      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3245      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3246    
3247      /* 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. */  
3248    
3249      REPEATCHAR:      REPEATCHAR:
3250  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3251      if (utf8)      if (utf)
3252        {        {
3253        length = 1;        length = 1;
3254        charptr = ecode;        charptr = ecode;
3255        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3256        ecode += length;        ecode += length;
3257    
3258        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2058  for (;;) Line 3262  for (;;)
3262          {          {
3263  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3264          unsigned int othercase;          unsigned int othercase;
3265          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3266              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3267            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3268          else oclength = 0;          else oclength = 0;
3269  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3270    
3271          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3272            {            {
3273            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3274                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3275  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3276            /* Need braces because of following else */            else if (oclength > 0 &&
3277            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3278                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3279    #endif  /* SUPPORT_UCP */
3280            else            else
3281              {              {
3282              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3283              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3284              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3285            }            }
3286    
3287          if (min == max) continue;          if (min == max) continue;
# Line 2086  for (;;) Line 3290  for (;;)
3290            {            {
3291            for (fi = min;; fi++)            for (fi = min;; fi++)
3292              {              {
3293              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3294              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3295              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3296              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3297                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3298  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3299              /* Need braces because of following else */              else if (oclength > 0 &&
3300              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3301                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3302    #endif  /* SUPPORT_UCP */
3303              else              else
3304                {                {
3305                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3306                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3307                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3308              }              }
3309            /* Control never gets here */            /* Control never gets here */
3310            }            }
# Line 2110  for (;;) Line 3314  for (;;)
3314            pp = eptr;            pp = eptr;
3315            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3316              {              {
3317              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3318              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3319  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3320              else if (oclength == 0) break;              else if (oclength > 0 &&
3321                         eptr <= md->end_subject - oclength &&
3322                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3323    #endif  /* SUPPORT_UCP */
3324              else              else
3325                {                {
3326                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3327                eptr += oclength;                break;
3328                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3329              }              }
3330    
3331            if (possessive) continue;            if (possessive) continue;
3332    
3333            for(;;)            for(;;)
3334             {              {
3335             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3336             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3337             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3338  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3339             eptr--;              eptr--;
3340             BACKCHAR(eptr);              BACKCHAR(eptr);
3341  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3342             eptr -= length;              eptr -= length;
3343  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3344             }              }
3345            }            }
3346          /* Control never gets here */          /* Control never gets here */
3347          }          }
# Line 2146  for (;;) Line 3351  for (;;)
3351        value of fc will always be < 128. */        value of fc will always be < 128. */
3352        }        }
3353      else      else
3354  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3355          /* 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);  
3356        fc = *ecode++;        fc = *ecode++;
       }  
3357    
3358      /* 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
3359      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
3360      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3361      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3362      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2166  for (;;) Line 3367  for (;;)
3367      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3368        max, eptr));        max, eptr));
3369    
3370      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3371        {        {
3372        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3373          /* fc must be < 128 if UTF is enabled. */
3374          foc = md->fcc[fc];
3375    #else
3376    #ifdef SUPPORT_UTF
3377    #ifdef SUPPORT_UCP
3378          if (utf && fc > 127)
3379            foc = UCD_OTHERCASE(fc);
3380    #else
3381          if (utf && fc > 127)
3382            foc = fc;
3383    #endif /* SUPPORT_UCP */
3384          else
3385    #endif /* SUPPORT_UTF */
3386