/[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 609 by ph10, Wed Jun 15 18:09:23 2011 UTC revision 1411 by ph10, Fri Dec 6 17:11:44 2013 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) 1997-2011 University of Cambridge             Copyright (c) 1997-2013 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 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  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
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
# Line 57  possible. There are also some static sup Line 56  possible. There are also some static sup
56  #undef min  #undef min
57  #undef max  #undef max
58    
59  /* Values for setting in md->match_function_type to indicate two special types  /* The md->capture_last field uses the lower 16 bits for the last captured
60  of call to match(). We do it this way to save on using another stack variable,  substring (which can never be greater than 65535) and a bit in the top half
61    to mean "capture vector overflowed". This odd way of doing things was
62    implemented when it was realized that preserving and restoring the overflow bit
63    whenever the last capture number was saved/restored made for a neater
64    interface, and doing it this way saved on (a) another variable, which would
65    have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66    separate set of save/restore instructions. The following defines are used in
67    implementing this. */
68    
69    #define CAPLMASK    0x0000ffff    /* The bits used for last_capture */
70    #define OVFLMASK    0xffff0000    /* The bits used for the overflow flag */
71    #define OVFLBIT     0x00010000    /* The bit that is set for overflow */
72    
73    /* Values for setting in md->match_function_type to indicate two special types
74    of call to match(). We do it this way to save on using another stack variable,
75  as stack usage is to be discouraged. */  as stack usage is to be discouraged. */
76    
77  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
# Line 74  defined PCRE_ERROR_xxx codes, which are Line 87  defined PCRE_ERROR_xxx codes, which are
87  negative to avoid the external error codes. */  negative to avoid the external error codes. */
88    
89  #define MATCH_ACCEPT       (-999)  #define MATCH_ACCEPT       (-999)
90  #define MATCH_COMMIT       (-998)  #define MATCH_KETRPOS      (-998)
91  #define MATCH_KETRPOS      (-997)  #define MATCH_ONCE         (-997)
92  #define MATCH_PRUNE        (-996)  /* The next 5 must be kept together and in sequence so that a test that checks
93  #define MATCH_SKIP         (-995)  for any one of them can use a range. */
94  #define MATCH_SKIP_ARG     (-994)  #define MATCH_COMMIT       (-996)
95  #define MATCH_THEN         (-993)  #define MATCH_PRUNE        (-995)
96    #define MATCH_SKIP         (-994)
97  /* This is a convenience macro for code that occurs many times. */  #define MATCH_SKIP_ARG     (-993)
98    #define MATCH_THEN         (-992)
99  #define MRRETURN(ra) \  #define MATCH_BACKTRACK_MAX MATCH_THEN
100    { \  #define MATCH_BACKTRACK_MIN MATCH_COMMIT
   md->mark = markptr; \  
   RRETURN(ra); \  
   }  
101    
102  /* 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.
103  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,
# Line 97  because the offset vector is always a mu Line 107  because the offset vector is always a mu
107    
108  /* Min and max values for the common repeats; for the maxima, 0 => infinity */  /* Min and max values for the common repeats; for the maxima, 0 => infinity */
109    
110  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, };
111  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, };
   
   
112    
113  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
114  /*************************************************  /*************************************************
# Line 120  Returns:     nothing Line 128  Returns:     nothing
128  */  */
129    
130  static void  static void
131  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)
132  {  {
133  unsigned int c;  pcre_uint32 c;
134    BOOL utf = md->utf;
135  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;
136  while (length-- > 0)  while (length-- > 0)
137    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138  }  }
139  #endif  #endif
140    
# Line 137  while (length-- > 0) Line 146  while (length-- > 0)
146    
147  /* Normally, if a back reference hasn't been set, the length that is passed is  /* Normally, if a back reference hasn't been set, the length that is passed is
148  negative, so the match always fails. However, in JavaScript compatibility mode,  negative, so the match always fails. However, in JavaScript compatibility mode,
149  the length passed is zero. Note that in caseless UTF-8 mode, the number of  the length passed is zero. Note that in caseless UTF-8 mode, the number of
150  subject bytes matched may be different to the number of reference bytes.  subject bytes matched may be different to the number of reference bytes.
151    
152  Arguments:  Arguments:
# Line 147  Arguments: Line 156  Arguments:
156    md          points to match data block    md          points to match data block
157    caseless    TRUE if caseless    caseless    TRUE if caseless
158    
159  Returns:      < 0 if not matched, otherwise the number of subject bytes matched  Returns:      >= 0 the number of subject bytes matched
160                  -1 no match
161                  -2 partial match; always given if at end subject
162  */  */
163    
164  static int  static int
165  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,
166    BOOL caseless)    BOOL caseless)
167  {  {
168  USPTR eptr_start = eptr;  PCRE_PUCHAR eptr_start = eptr;
169  register USPTR p = md->start_subject + md->offset_vector[offset];  register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170    #if defined SUPPORT_UTF && defined SUPPORT_UCP
171    BOOL utf = md->utf;
172    #endif
173    
174  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
175  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
# Line 170  pchars(p, length, FALSE, md); Line 184  pchars(p, length, FALSE, md);
184  printf("\n");  printf("\n");
185  #endif  #endif
186    
187  /* Always fail if reference not set (and not JavaScript compatible). */  /* Always fail if reference not set (and not JavaScript compatible - in that
188    case the length is passed as zero). */
189    
190  if (length < 0) return -1;  if (length < 0) return -1;
191    
# Line 180  ASCII characters. */ Line 195  ASCII characters. */
195    
196  if (caseless)  if (caseless)
197    {    {
198  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined SUPPORT_UCP
199  #ifdef SUPPORT_UCP    if (utf)
   if (md->utf8)  
200      {      {
201      /* Match characters up to the end of the reference. NOTE: the number of      /* Match characters up to the end of the reference. NOTE: the number of
202      bytes matched may differ, because there are some characters whose upper and      data units matched may differ, because in UTF-8 there are some characters
203      lower case versions code as different numbers of bytes. For example, U+023A      whose upper and lower case versions code have different numbers of bytes.
204      (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);      For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
205      a sequence of 3 of the former uses 6 bytes, as does a sequence of two of      (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
206      the latter. It is important, therefore, to check the length along the      sequence of two of the latter. It is important, therefore, to check the
207      reference, not along the subject (earlier code did this wrong). */      length along the reference, not along the subject (earlier code did this
208        wrong). */
209      USPTR endptr = p + length;  
210        PCRE_PUCHAR endptr = p + length;
211      while (p < endptr)      while (p < endptr)
212        {        {
213        int c, d;        pcre_uint32 c, d;
214        if (eptr >= md->end_subject) return -1;        const ucd_record *ur;
215          if (eptr >= md->end_subject) return -2;   /* Partial match */
216        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
217        GETCHARINC(d, p);        GETCHARINC(d, p);
218        if (c != d && c != UCD_OTHERCASE(d)) return -1;        ur = GET_UCD(d);
219          if (c != d && c != d + ur->other_case)
220            {
221            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
222            for (;;)
223              {
224              if (c < *pp) return -1;
225              if (c == *pp++) break;
226              }
227            }
228        }        }
229      }      }
230    else    else
231  #endif  #endif
 #endif  
232    
233    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
234    is no UCP support. */    is no UCP support. */
235      {      {
     if (eptr + length > md->end_subject) return -1;  
236      while (length-- > 0)      while (length-- > 0)
237        { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }        {
238      }        pcre_uint32 cc, cp;
239          if (eptr >= md->end_subject) return -2;   /* Partial match */
240          cc = RAWUCHARTEST(eptr);
241          cp = RAWUCHARTEST(p);
242          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
243          p++;
244          eptr++;
245          }
246        }
247    }    }
248    
249  /* In the caseful case, we can just compare the bytes, whether or not we  /* In the caseful case, we can just compare the bytes, whether or not we
250  are in UTF-8 mode. */  are in UTF-8 mode. */
251    
252  else  else
253    {    {
254    if (eptr + length > md->end_subject) return -1;    while (length-- > 0)
255    while (length-- > 0) if (*p++ != *eptr++) return -1;      {
256        if (eptr >= md->end_subject) return -2;   /* Partial match */
257        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
258        }
259    }    }
260    
261  return eptr - eptr_start;  return (int)(eptr - eptr_start);
262  }  }
263    
264    
# Line 276  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 310  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
310         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
311         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
312         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
313         RM61,  RM62, RM63};         RM61,  RM62, RM63, RM64, RM65, RM66, RM67 };
314    
315  /* 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
316  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
# Line 289  actually used in this definition. */ Line 323  actually used in this definition. */
323  #define RMATCH(ra,rb,rc,rd,re,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
324    { \    { \
325    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
326    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
327    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
328    }    }
329  #define RRETURN(ra) \  #define RRETURN(ra) \
330    { \    { \
331    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
332    return ra; \    return ra; \
333    }    }
334  #else  #else
335  #define RMATCH(ra,rb,rc,rd,re,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
336    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
337  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
338  #endif  #endif
339    
# Line 314  argument of match(), which never changes Line 348  argument of match(), which never changes
348    
349  #define RMATCH(ra,rb,rc,rd,re,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
350    {\    {\
351    heapframe *newframe = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
352    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\    if (newframe == NULL)\
353    frame->Xwhere = rw; \      {\
354        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
355        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
356        newframe->Xnextframe = NULL;\
357        frame->Xnextframe = newframe;\
358        }\
359      frame->Xwhere = rw;\
360    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
361    newframe->Xecode = rb;\    newframe->Xecode = rb;\
362    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
   newframe->Xmarkptr = markptr;\  
363    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
364    newframe->Xeptrb = re;\    newframe->Xeptrb = re;\
365    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
# Line 336  argument of match(), which never changes Line 375  argument of match(), which never changes
375    {\    {\
376    heapframe *oldframe = frame;\    heapframe *oldframe = frame;\
377    frame = oldframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(oldframe);\  
378    if (frame != NULL)\    if (frame != NULL)\
379      {\      {\
380      rrc = ra;\      rrc = ra;\
# Line 350  argument of match(), which never changes Line 388  argument of match(), which never changes
388    
389  typedef struct heapframe {  typedef struct heapframe {
390    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
391      struct heapframe *Xnextframe;
392    
393    /* Function arguments that may change */    /* Function arguments that may change */
394    
395    USPTR Xeptr;    PCRE_PUCHAR Xeptr;
396    const uschar *Xecode;    const pcre_uchar *Xecode;
397    USPTR Xmstart;    PCRE_PUCHAR Xmstart;
   USPTR Xmarkptr;  
398    int Xoffset_top;    int Xoffset_top;
399    eptrblock *Xeptrb;    eptrblock *Xeptrb;
400    unsigned int Xrdepth;    unsigned int Xrdepth;
401    
402    /* Function local variables */    /* Function local variables */
403    
404    USPTR Xcallpat;    PCRE_PUCHAR Xcallpat;
405  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
406    USPTR Xcharptr;    PCRE_PUCHAR Xcharptr;
407  #endif  #endif
408    USPTR Xdata;    PCRE_PUCHAR Xdata;
409    USPTR Xnext;    PCRE_PUCHAR Xnext;
410    USPTR Xpp;    PCRE_PUCHAR Xpp;
411    USPTR Xprev;    PCRE_PUCHAR Xprev;
412    USPTR Xsaved_eptr;    PCRE_PUCHAR Xsaved_eptr;
413    
414    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
415    
# Line 381  typedef struct heapframe { Line 419  typedef struct heapframe {
419    
420  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
421    int Xprop_type;    int Xprop_type;
422    int Xprop_value;    unsigned int Xprop_value;
423    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
424    int Xoclength;    int Xoclength;
425    uschar Xocchars[8];    pcre_uchar Xocchars[6];
426  #endif  #endif
427    
428    int Xcodelink;    int Xcodelink;
# Line 397  typedef struct heapframe { Line 432  typedef struct heapframe {
432    int Xlength;    int Xlength;
433    int Xmax;    int Xmax;
434    int Xmin;    int Xmin;
435    int Xnumber;    unsigned int Xnumber;
436    int Xoffset;    int Xoffset;
437    int Xop;    unsigned int Xop;
438    int Xsave_capture_last;    pcre_int32 Xsave_capture_last;
439    int Xsave_offset1, Xsave_offset2, Xsave_offset3;    int Xsave_offset1, Xsave_offset2, Xsave_offset3;
440    int Xstacksave[REC_STACK_SAVE_MAX];    int Xstacksave[REC_STACK_SAVE_MAX];
441    
# Line 429  returns a negative (error) response, the Line 464  returns a negative (error) response, the
464  same response. */  same response. */
465    
466  /* These macros pack up tests that are used for partial matching, and which  /* These macros pack up tests that are used for partial matching, and which
467  appears several times in the code. We set the "hit end" flag if the pointer is  appear several times in the code. We set the "hit end" flag if the pointer is
468  at the end of the subject and also past the start of the subject (i.e.  at the end of the subject and also past the start of the subject (i.e.
469  something has been matched). For hard partial matching, we then return  something has been matched). For hard partial matching, we then return
470  immediately. The second one is used when we already know we are past the end of  immediately. The second one is used when we already know we are past the end of
# Line 440  the subject. */ Line 475  the subject. */
475        eptr > md->start_used_ptr) \        eptr > md->start_used_ptr) \
476      { \      { \
477      md->hitend = TRUE; \      md->hitend = TRUE; \
478      if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \      if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
479      }      }
480    
481  #define SCHECK_PARTIAL()\  #define SCHECK_PARTIAL()\
482    if (md->partial != 0 && eptr > md->start_used_ptr) \    if (md->partial != 0 && eptr > md->start_used_ptr) \
483      { \      { \
484      md->hitend = TRUE; \      md->hitend = TRUE; \
485      if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \      if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
486      }      }
487    
488    
489  /* Performance note: It might be tempting to extract commonly used fields from  /* Performance note: It might be tempting to extract commonly used fields from
490  the md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf, end_subject) into individual variables to improve
491  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
492  made performance worse.  made performance worse.
493    
# Line 461  Arguments: Line 496  Arguments:
496     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
497     mstart      pointer to the current match start position (can be modified     mstart      pointer to the current match start position (can be modified
498                   by encountering \K)                   by encountering \K)
    markptr     pointer to the most recent MARK name, or NULL  
499     offset_top  current top pointer     offset_top  current top pointer
500     md          pointer to "static" info for the match     md          pointer to "static" info for the match
501     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
# Line 476  Returns:       MATCH_MATCH if matched Line 510  Returns:       MATCH_MATCH if matched
510  */  */
511    
512  static int  static int
513  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
514    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
515    unsigned int rdepth)    unsigned int rdepth)
516  {  {
517  /* 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,
# Line 486  so they can be ordinary variables in all Line 520  so they can be ordinary variables in all
520    
521  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
522  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
523  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
524  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
525    
526  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
527  BOOL caseless;  BOOL caseless;
528  int condcode;  int condcode;
529    
530  /* 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
531  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
532  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
533  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
534    the top-level on the stack rather than malloc-ing them all gives a performance
535    boost in many cases where there is not much "recursion". */
536    
537  #ifdef NO_RECURSE  #ifdef NO_RECURSE
538  heapframe *frame = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);  
 frame->Xprevframe = NULL;            /* Marks the top level */  
539    
540  /* Copy in the original argument variables */  /* Copy in the original argument variables */
541    
542  frame->Xeptr = eptr;  frame->Xeptr = eptr;
543  frame->Xecode = ecode;  frame->Xecode = ecode;
544  frame->Xmstart = mstart;  frame->Xmstart = mstart;
 frame->Xmarkptr = markptr;  
545  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
546  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
547  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
# Line 522  HEAP_RECURSE: Line 555  HEAP_RECURSE:
555  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
556  #define ecode              frame->Xecode  #define ecode              frame->Xecode
557  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
 #define markptr            frame->Xmarkptr  
558  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
559  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
560  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
561    
562  /* Ditto for the local variables */  /* Ditto for the local variables */
563    
564  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
565  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
566  #endif  #endif
567  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
# Line 550  HEAP_RECURSE: Line 582  HEAP_RECURSE:
582  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
583  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
584  #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  
585  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
586  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
587  #endif  #endif
# Line 590  declarations can be cut out in a block. Line 619  declarations can be cut out in a block.
619  below are for variables that do not have to be preserved over a recursive call  below are for variables that do not have to be preserved over a recursive call
620  to RMATCH(). */  to RMATCH(). */
621    
622  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
623  const uschar *charptr;  const pcre_uchar *charptr;
624  #endif  #endif
625  const uschar *callpat;  const pcre_uchar *callpat;
626  const uschar *data;  const pcre_uchar *data;
627  const uschar *next;  const pcre_uchar *next;
628  USPTR         pp;  PCRE_PUCHAR       pp;
629  const uschar *prev;  const pcre_uchar *prev;
630  USPTR         saved_eptr;  PCRE_PUCHAR       saved_eptr;
631    
632  recursion_info new_recursive;  recursion_info new_recursive;
633    
634  BOOL cur_is_word;  BOOL cur_is_word;
635  BOOL condition;  BOOL condition;
636  BOOL prev_is_word;  BOOL prev_is_word;
637    
638  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
639  int prop_type;  int prop_type;
640  int prop_value;  unsigned int prop_value;
641  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
642  int oclength;  int oclength;
643  uschar occhars[8];  pcre_uchar occhars[6];
644  #endif  #endif
645    
646  int codelink;  int codelink;
# Line 622  int ctype; Line 648  int ctype;
648  int length;  int length;
649  int max;  int max;
650  int min;  int min;
651  int number;  unsigned int number;
652  int offset;  int offset;
653  int op;  unsigned int op;
654  int save_capture_last;  pcre_int32 save_capture_last;
655  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
656  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
657    
658  eptrblock newptrb;  eptrblock newptrb;
659    
660    /* There is a special fudge for calling match() in a way that causes it to
661    measure the size of its basic stack frame when the stack is being used for
662    recursion. The second argument (ecode) being NULL triggers this behaviour. It
663    cannot normally ever be NULL. The return is the negated value of the frame
664    size. */
665    
666    if (ecode == NULL)
667      {
668      if (rdepth == 0)
669        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
670      else
671        {
672        int len = (char *)&rdepth - (char *)eptr;
673        return (len > 0)? -len : len;
674        }
675      }
676  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
677    
678  /* To save space on the stack and in the heap frame, I have doubled up on some  /* To save space on the stack and in the heap frame, I have doubled up on some
679  of the local variables that are used only in localised parts of the code, but  of the local variables that are used only in localised parts of the code, but
680  still need to be preserved over recursive calls of match(). These macros define  still need to be preserved over recursive calls of match(). These macros define
681  the alternative names that are used. */  the alternative names that are used. */
682    
683  #define allow_zero    cur_is_word  #define allow_zero    cur_is_word
# Line 642  the alternative names that are used. */ Line 685  the alternative names that are used. */
685  #define code_offset   codelink  #define code_offset   codelink
686  #define condassert    condition  #define condassert    condition
687  #define matched_once  prev_is_word  #define matched_once  prev_is_word
688    #define foc           number
689    #define save_mark     data
690    
691  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
692  variables. */  variables. */
# Line 667  defined). However, RMATCH isn't like a f Line 712  defined). However, RMATCH isn't like a f
712  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,
713  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
714    
715  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
716  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
717  #else  #else
718  utf8 = FALSE;  utf = FALSE;
719  #endif  #endif
720    
721  /* 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
# Line 680  if (md->match_call_count++ >= md->match_ Line 725  if (md->match_call_count++ >= md->match_
725  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
726    
727  /* 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
728  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
729  done this way to save having to use another function argument, which would take  done this way to save having to use another function argument, which would take
730  up space on the stack. See also MATCH_CONDASSERT below.  up space on the stack. See also MATCH_CONDASSERT below.
731    
732  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
# Line 705  for (;;) Line 750  for (;;)
750    {    {
751    minimize = possessive = FALSE;    minimize = possessive = FALSE;
752    op = *ecode;    op = *ecode;
753    
754    switch(op)    switch(op)
755      {      {
756      case OP_MARK:      case OP_MARK:
757      markptr = ecode + 2;      md->nomatch_mark = ecode + 2;
758      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      md->mark = NULL;    /* In case previously set by assertion */
759        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
760        eptrb, RM55);        eptrb, RM55);
761        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
762             md->mark == NULL) md->mark = ecode + 2;
763    
764      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
765      argument, and we must check whether that argument matches this MARK's      argument, and we must check whether that argument matches this MARK's
# Line 720  for (;;) Line 768  for (;;)
768      position and return MATCH_SKIP. Otherwise, pass back the return code      position and return MATCH_SKIP. Otherwise, pass back the return code
769      unaltered. */      unaltered. */
770    
771      if (rrc == MATCH_SKIP_ARG &&      else if (rrc == MATCH_SKIP_ARG &&
772          strcmp((char *)markptr, (char *)(md->start_match_ptr)) == 0)          STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
773        {        {
774        md->start_match_ptr = eptr;        md->start_match_ptr = eptr;
775        RRETURN(MATCH_SKIP);        RRETURN(MATCH_SKIP);
776        }        }
   
     if (md->mark == NULL) md->mark = markptr;  
777      RRETURN(rrc);      RRETURN(rrc);
778    
779      case OP_FAIL:      case OP_FAIL:
780      MRRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
   
     /* COMMIT overrides PRUNE, SKIP, and THEN */  
781    
782      case OP_COMMIT:      case OP_COMMIT:
783      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
784        eptrb, RM52);        eptrb, RM52);
785      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
786          rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&      RRETURN(MATCH_COMMIT);
         rrc != MATCH_THEN)  
       RRETURN(rrc);  
     MRRETURN(MATCH_COMMIT);  
   
     /* PRUNE overrides THEN */  
787    
788      case OP_PRUNE:      case OP_PRUNE:
789      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
790        eptrb, RM51);        eptrb, RM51);
791      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
792      MRRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
793    
794      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
795      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      md->nomatch_mark = ecode + 2;
796        md->mark = NULL;    /* In case previously set by assertion */
797        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
798        eptrb, RM56);        eptrb, RM56);
799      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);      if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
800      md->mark = ecode + 2;           md->mark == NULL) md->mark = ecode + 2;
801        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
802      RRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
803    
     /* SKIP overrides PRUNE and THEN */  
   
804      case OP_SKIP:      case OP_SKIP:
805      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
806        eptrb, RM53);        eptrb, RM53);
807      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
       RRETURN(rrc);  
808      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
809      MRRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
810    
811        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
812        nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
813        not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
814        that failed and any that precede it (either they also failed, or were not
815        triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
816        SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
817        set to the count of the one that failed. */
818    
819      case OP_SKIP_ARG:      case OP_SKIP_ARG:
820      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      md->skip_arg_count++;
821        if (md->skip_arg_count <= md->ignore_skip_arg)
822          {
823          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
824          break;
825          }
826        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
827        eptrb, RM57);        eptrb, RM57);
828      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
       RRETURN(rrc);  
829    
830      /* Pass back the current skip name by overloading md->start_match_ptr and      /* Pass back the current skip name by overloading md->start_match_ptr and
831      returning the special MATCH_SKIP_ARG return code. This will either be      returning the special MATCH_SKIP_ARG return code. This will either be
832      caught by a matching MARK, or get to the top, where it is treated the same      caught by a matching MARK, or get to the top, where it causes a rematch
833      as PRUNE. */      with md->ignore_skip_arg set to the value of md->skip_arg_count. */
834    
835      md->start_match_ptr = ecode + 2;      md->start_match_ptr = ecode + 2;
836      RRETURN(MATCH_SKIP_ARG);      RRETURN(MATCH_SKIP_ARG);
837    
838      /* For THEN (and THEN_ARG) we pass back the address of the bracket or      /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
839      the alt that is at the start of the current branch. This makes it possible      the branch in which it occurs can be determined. Overload the start of
840      to skip back past alternatives that precede the THEN within the current      match pointer to do this. */
     branch. */  
841    
842      case OP_THEN:      case OP_THEN:
843      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
844        eptrb, RM54);        eptrb, RM54);
845      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
846      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode;
847      MRRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
848    
849      case OP_THEN_ARG:      case OP_THEN_ARG:
850      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],      md->nomatch_mark = ecode + 2;
851        offset_top, md, eptrb, RM58);      md->mark = NULL;    /* In case previously set by assertion */
852        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
853          md, eptrb, RM58);
854        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
855             md->mark == NULL) md->mark = ecode + 2;
856      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
857      md->start_match_ptr = ecode - GET(ecode, 1);      md->start_match_ptr = ecode;
     md->mark = ecode + LINK_SIZE + 2;  
858      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
859    
860        /* Handle an atomic group that does not contain any capturing parentheses.
861        This can be handled like an assertion. Prior to 8.13, all atomic groups
862        were handled this way. In 8.13, the code was changed as below for ONCE, so
863        that backups pass through the group and thereby reset captured values.
864        However, this uses a lot more stack, so in 8.20, atomic groups that do not
865        contain any captures generate OP_ONCE_NC, which can be handled in the old,
866        less stack intensive way.
867    
868        Check the alternative branches in turn - the matching won't pass the KET
869        for this kind of subpattern. If any one branch matches, we carry on as at
870        the end of a normal bracket, leaving the subject pointer, but resetting
871        the start-of-match value in case it was changed by \K. */
872    
873        case OP_ONCE_NC:
874        prev = ecode;
875        saved_eptr = eptr;
876        save_mark = md->mark;
877        do
878          {
879          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
880          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
881            {
882            mstart = md->start_match_ptr;
883            break;
884            }
885          if (rrc == MATCH_THEN)
886            {
887            next = ecode + GET(ecode,1);
888            if (md->start_match_ptr < next &&
889                (*ecode == OP_ALT || *next == OP_ALT))
890              rrc = MATCH_NOMATCH;
891            }
892    
893          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
894          ecode += GET(ecode,1);
895          md->mark = save_mark;
896          }
897        while (*ecode == OP_ALT);
898    
899        /* If hit the end of the group (which could be repeated), fail */
900    
901        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
902    
903        /* Continue as from after the group, updating the offsets high water
904        mark, since extracts may have been taken. */
905    
906        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
907    
908        offset_top = md->end_offset_top;
909        eptr = md->end_match_ptr;
910    
911        /* For a non-repeating ket, just continue at this level. This also
912        happens for a repeating ket if no characters were matched in the group.
913        This is the forcible breaking of infinite loops as implemented in Perl
914        5.005. */
915    
916        if (*ecode == OP_KET || eptr == saved_eptr)
917          {
918          ecode += 1+LINK_SIZE;
919          break;
920          }
921    
922        /* The repeating kets try the rest of the pattern or restart from the
923        preceding bracket, in the appropriate order. The second "call" of match()
924        uses tail recursion, to avoid using another stack frame. */
925    
926        if (*ecode == OP_KETRMIN)
927          {
928          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
929          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
930          ecode = prev;
931          goto TAIL_RECURSE;
932          }
933        else  /* OP_KETRMAX */
934          {
935          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
936          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
937          ecode += 1 + LINK_SIZE;
938          goto TAIL_RECURSE;
939          }
940        /* Control never gets here */
941    
942      /* Handle a capturing bracket, other than those that are possessive with an      /* Handle a capturing bracket, other than those that are possessive with an
943      unlimited repeat. If there is space in the offset vector, save the current      unlimited repeat. If there is space in the offset vector, save the current
944      subject position in the working slot at the top of the vector. We mustn't      subject position in the working slot at the top of the vector. We mustn't
945      change the current values of the data slot, because they may be set from a      change the current values of the data slot, because they may be set from a
946      previous iteration of this group, and be referred to by a reference inside      previous iteration of this group, and be referred to by a reference inside
947      the group. If we fail to match, we need to restore this value and also the      the group. A failure to match might occur after the group has succeeded,
948      values of the final offsets, in case they were set by a previous iteration      if something later on doesn't match. For this reason, we need to restore
949      of the same bracket.      the working value and also the values of the final offsets, in case they
950        were set by a previous iteration of the same bracket.
951    
952      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
953      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 820  for (;;) Line 957  for (;;)
957      case OP_SCBRA:      case OP_SCBRA:
958      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
959      offset = number << 1;      offset = number << 1;
960    
961  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
962      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
963      printf("subject=");      printf("subject=");
# Line 834  for (;;) Line 971  for (;;)
971        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
972        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
973        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
974          save_mark = md->mark;
975    
976        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
977        md->offset_vector[md->offset_end - number] =        md->offset_vector[md->offset_end - number] =
# Line 841  for (;;) Line 979  for (;;)
979    
980        for (;;)        for (;;)
981          {          {
982          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
983          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
984            eptrb, RM1);            eptrb, RM1);
985          if (rrc != MATCH_NOMATCH &&          if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
986              (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
987            RRETURN(rrc);          /* If we backed up to a THEN, check whether it is within the current
988            branch by comparing the address of the THEN that is passed back with
989            the end of the branch. If it is within the current branch, and the
990            branch is one of two or more alternatives (it either starts or ends
991            with OP_ALT), we have reached the limit of THEN's action, so convert
992            the return code to NOMATCH, which will cause normal backtracking to
993            happen from now on. Otherwise, THEN is passed back to an outer
994            alternative. This implements Perl's treatment of parenthesized groups,
995            where a group not containing | does not affect the current alternative,
996            that is, (X) is NOT the same as (X|(*F)). */
997    
998            if (rrc == MATCH_THEN)
999              {
1000              next = ecode + GET(ecode,1);
1001              if (md->start_match_ptr < next &&
1002                  (*ecode == OP_ALT || *next == OP_ALT))
1003                rrc = MATCH_NOMATCH;
1004              }
1005    
1006            /* Anything other than NOMATCH is passed back. */
1007    
1008            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1009          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1010          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1011          if (*ecode != OP_ALT) break;          md->mark = save_mark;
1012            if (*ecode != OP_ALT) break;
1013          }          }
1014    
1015        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1016        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1017        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1018        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1019    
1020        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1021        RRETURN(MATCH_NOMATCH);  
1022          RRETURN(rrc);
1023        }        }
1024    
1025      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 873  for (;;) Line 1033  for (;;)
1033      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1034      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1035    
1036      /* Non-capturing bracket, except for possessive with unlimited repeat. Loop      /* Non-capturing or atomic group, except for possessive with unlimited
1037      for all the alternatives. When we get to the final alternative within the      repeat and ONCE group with no captures. Loop for all the alternatives.
     brackets, we used to return the result of a recursive call to match()  
     whatever happened so it was possible to reduce stack usage by turning this  
     into a tail recursion, except in the case of a possibly empty group.  
     However, now that there is the possiblity of (*THEN) occurring in the final  
     alternative, this optimization is no longer possible. */  
1038    
1039        When we get to the final alternative within the brackets, we used to return
1040        the result of a recursive call to match() whatever happened so it was
1041        possible to reduce stack usage by turning this into a tail recursion,
1042        except in the case of a possibly empty group. However, now that there is
1043        the possiblity of (*THEN) occurring in the final alternative, this
1044        optimization is no longer always possible.
1045    
1046        We can optimize if we know there are no (*THEN)s in the pattern; at present
1047        this is the best that can be done.
1048    
1049        MATCH_ONCE is returned when the end of an atomic group is successfully
1050        reached, but subsequent matching fails. It passes back up the tree (causing
1051        captured values to be reset) until the original atomic group level is
1052        reached. This is tested by comparing md->once_target with the start of the
1053        group. At this point, the return is converted into MATCH_NOMATCH so that
1054        previous backup points can be taken. */
1055    
1056        case OP_ONCE:
1057      case OP_BRA:      case OP_BRA:
1058      case OP_SBRA:      case OP_SBRA:
1059      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1060    
1061      for (;;)      for (;;)
1062        {        {
1063        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (op >= OP_SBRA || op == OP_ONCE)
1064        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,          md->match_function_type = MATCH_CBEGROUP;
1065    
1066          /* If this is not a possibly empty group, and there are no (*THEN)s in
1067          the pattern, and this is the final alternative, optimize as described
1068          above. */
1069    
1070          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1071            {
1072            ecode += PRIV(OP_lengths)[*ecode];
1073            goto TAIL_RECURSE;
1074            }
1075    
1076          /* In all other cases, we have to make another call to match(). */
1077    
1078          save_mark = md->mark;
1079          save_capture_last = md->capture_last;
1080          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1081          RM2);          RM2);
1082        if (rrc != MATCH_NOMATCH &&  
1083            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* See comment in the code for capturing groups above about handling
1084          THEN. */
1085    
1086          if (rrc == MATCH_THEN)
1087            {
1088            next = ecode + GET(ecode,1);
1089            if (md->start_match_ptr < next &&
1090                (*ecode == OP_ALT || *next == OP_ALT))
1091              rrc = MATCH_NOMATCH;
1092            }
1093    
1094          if (rrc != MATCH_NOMATCH)
1095            {
1096            if (rrc == MATCH_ONCE)
1097              {
1098              const pcre_uchar *scode = ecode;
1099              if (*scode != OP_ONCE)           /* If not at start, find it */
1100                {
1101                while (*scode == OP_ALT) scode += GET(scode, 1);
1102                scode -= GET(scode, 1);
1103                }
1104              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1105              }
1106          RRETURN(rrc);          RRETURN(rrc);
1107            }
1108        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1109        if (*ecode != OP_ALT) break;        md->mark = save_mark;
1110          if (*ecode != OP_ALT) break;
1111          md->capture_last = save_capture_last;
1112        }        }
1113    
     if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;  
1114      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1115    
1116      /* Handle possessive capturing brackets with an unlimited repeat. We come      /* Handle possessive capturing brackets with an unlimited repeat. We come
1117      here from BRAZERO with allow_zero set TRUE. The offset_vector values are      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1118      handled similarly to the normal case above. However, the matching is      handled similarly to the normal case above. However, the matching is
1119      different. The end of these brackets will always be OP_KETRPOS, which      different. The end of these brackets will always be OP_KETRPOS, which
1120      returns MATCH_KETRPOS without going further in the pattern. By this means      returns MATCH_KETRPOS without going further in the pattern. By this means
1121      we can handle the group by iteration rather than recursion, thereby      we can handle the group by iteration rather than recursion, thereby
1122      reducing the amount of stack needed. */      reducing the amount of stack needed. */
1123    
1124      case OP_CBRAPOS:      case OP_CBRAPOS:
1125      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1126      allow_zero = FALSE;      allow_zero = FALSE;
1127    
1128      POSSESSIVE_CAPTURE:      POSSESSIVE_CAPTURE:
1129      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
1130      offset = number << 1;      offset = number << 1;
# Line 925  for (;;) Line 1139  for (;;)
1139      if (offset < md->offset_max)      if (offset < md->offset_max)
1140        {        {
1141        matched_once = FALSE;        matched_once = FALSE;
1142        code_offset = ecode - md->start_code;        code_offset = (int)(ecode - md->start_code);
1143    
1144        save_offset1 = md->offset_vector[offset];        save_offset1 = md->offset_vector[offset];
1145        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
# Line 933  for (;;) Line 1147  for (;;)
1147        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
1148    
1149        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1150    
1151        /* Each time round the loop, save the current subject position for use        /* Each time round the loop, save the current subject position for use
1152        when the group matches. For MATCH_MATCH, the group has matched, so we        when the group matches. For MATCH_MATCH, the group has matched, so we
1153        restart it with a new subject starting position, remembering that we had        restart it with a new subject starting position, remembering that we had
1154        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1155        usual. If we haven't matched any alternatives in any iteration, check to        usual. If we haven't matched any alternatives in any iteration, check to
1156        see if a previous iteration matched. If so, the group has matched;        see if a previous iteration matched. If so, the group has matched;
1157        continue from afterwards. Otherwise it has failed; restore the previous        continue from afterwards. Otherwise it has failed; restore the previous
1158        capture values before returning NOMATCH. */        capture values before returning NOMATCH. */
1159    
1160        for (;;)        for (;;)
1161          {          {
1162          md->offset_vector[md->offset_end - number] =          md->offset_vector[md->offset_end - number] =
1163            (int)(eptr - md->start_subject);            (int)(eptr - md->start_subject);
1164          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1165          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1166            eptrb, RM63);            eptrb, RM63);
1167          if (rrc == MATCH_KETRPOS)          if (rrc == MATCH_KETRPOS)
1168            {            {
1169            offset_top = md->end_offset_top;            offset_top = md->end_offset_top;
1170            eptr = md->end_match_ptr;            eptr = md->end_match_ptr;
1171            ecode = md->start_code + code_offset;            ecode = md->start_code + code_offset;
1172            save_capture_last = md->capture_last;            save_capture_last = md->capture_last;
1173            matched_once = TRUE;            matched_once = TRUE;
1174            continue;            mstart = md->start_match_ptr;    /* In case \K changed it */
1175            }            continue;
1176          if (rrc != MATCH_NOMATCH &&            }
1177              (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
1178            RRETURN(rrc);          /* See comment in the code for capturing groups above about handling
1179            THEN. */
1180    
1181            if (rrc == MATCH_THEN)
1182              {
1183              next = ecode + GET(ecode,1);
1184              if (md->start_match_ptr < next &&
1185                  (*ecode == OP_ALT || *next == OP_ALT))
1186                rrc = MATCH_NOMATCH;
1187              }
1188    
1189            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1190          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1191          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1192          if (*ecode != OP_ALT) break;          if (*ecode != OP_ALT) break;
1193          }          }
1194    
1195        if (!matched_once)        if (!matched_once)
1196          {          {
1197          md->offset_vector[offset] = save_offset1;          md->offset_vector[offset] = save_offset1;
1198          md->offset_vector[offset+1] = save_offset2;          md->offset_vector[offset+1] = save_offset2;
1199          md->offset_vector[md->offset_end - number] = save_offset3;          md->offset_vector[md->offset_end - number] = save_offset3;
1200          }          }
1201    
       if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;  
1202        if (allow_zero || matched_once)        if (allow_zero || matched_once)
1203          {          {
1204          ecode += 1 + LINK_SIZE;          ecode += 1 + LINK_SIZE;
1205          break;          break;
1206          }          }
1207    
1208        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1209        }        }
1210    
1211      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1212      as a non-capturing bracket. */      as a non-capturing bracket. */
1213    
# Line 995  for (;;) Line 1219  for (;;)
1219      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1220      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221    
1222      /* Non-capturing possessive bracket with unlimited repeat. We come here      /* Non-capturing possessive bracket with unlimited repeat. We come here
1223      from BRAZERO with allow_zero = TRUE. The code is similar to the above,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1224      without the capturing complication. It is written out separately for speed      without the capturing complication. It is written out separately for speed
1225      and cleanliness. */      and cleanliness. */
1226    
1227      case OP_BRAPOS:      case OP_BRAPOS:
1228      case OP_SBRAPOS:      case OP_SBRAPOS:
1229      allow_zero = FALSE;      allow_zero = FALSE;
1230    
1231      POSSESSIVE_NON_CAPTURE:      POSSESSIVE_NON_CAPTURE:
1232      matched_once = FALSE;      matched_once = FALSE;
1233      code_offset = ecode - md->start_code;      code_offset = (int)(ecode - md->start_code);
1234        save_capture_last = md->capture_last;
1235    
1236      for (;;)      for (;;)
1237        {        {
1238        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1239        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1240          eptrb, RM48);          eptrb, RM48);
1241        if (rrc == MATCH_KETRPOS)        if (rrc == MATCH_KETRPOS)
1242          {          {
1243            offset_top = md->end_offset_top;
1244          eptr = md->end_match_ptr;          eptr = md->end_match_ptr;
1245          ecode = md->start_code + code_offset;          ecode = md->start_code + code_offset;
1246          matched_once = TRUE;          matched_once = TRUE;
1247          continue;          mstart = md->start_match_ptr;   /* In case \K reset it */
1248          }          continue;
1249        if (rrc != MATCH_NOMATCH &&          }
1250            (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
1251          RRETURN(rrc);        /* See comment in the code for capturing groups above about handling
1252          THEN. */
1253    
1254          if (rrc == MATCH_THEN)
1255            {
1256            next = ecode + GET(ecode,1);
1257            if (md->start_match_ptr < next &&
1258                (*ecode == OP_ALT || *next == OP_ALT))
1259              rrc = MATCH_NOMATCH;
1260            }
1261    
1262          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1263        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1264        if (*ecode != OP_ALT) break;        if (*ecode != OP_ALT) break;
1265          md->capture_last = save_capture_last;
1266        }        }
1267    
1268      if (matched_once || allow_zero)      if (matched_once || allow_zero)
1269        {        {
1270        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1271        break;        break;
1272        }        }
1273      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1274    
1275      /* Control never reaches here. */      /* Control never reaches here. */
1276    
1277      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than two
1278      two branches. If the condition is false, skipping the first branch takes us      branches. If the condition is false, skipping the first branch takes us
1279      past the end if there is only one branch, but that's OK because that is      past the end of the item if there is only one branch, but that's exactly
1280      exactly what going to the ket would do. */      what we want. */
1281    
1282      case OP_COND:      case OP_COND:
1283      case OP_SCOND:      case OP_SCOND:
1284      codelink = GET(ecode, 1);  
1285        /* The variable codelink will be added to ecode when the condition is
1286        false, to get to the second branch. Setting it to the offset to the ALT
1287        or KET, then incrementing ecode achieves this effect. We now have ecode
1288        pointing to the condition or callout. */
1289    
1290        codelink = GET(ecode, 1);   /* Offset to the second branch */
1291        ecode += 1 + LINK_SIZE;     /* From this opcode */
1292    
1293      /* Because of the way auto-callout works during compile, a callout item is      /* Because of the way auto-callout works during compile, a callout item is
1294      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
1295    
1296      if (ecode[LINK_SIZE+1] == OP_CALLOUT)      if (*ecode == OP_CALLOUT)
1297        {        {
1298        if (pcre_callout != NULL)        if (PUBL(callout) != NULL)
1299          {          {
1300          pcre_callout_block cb;          PUBL(callout_block) cb;
1301          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1302          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[1];
1303          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1304    #if defined COMPILE_PCRE8
1305          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
1306    #elif defined COMPILE_PCRE16
1307            cb.subject          = (PCRE_SPTR16)md->start_subject;
1308    #elif defined COMPILE_PCRE32
1309            cb.subject          = (PCRE_SPTR32)md->start_subject;
1310    #endif
1311          cb.subject_length   = (int)(md->end_subject - md->start_subject);          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1312          cb.start_match      = (int)(mstart - md->start_subject);          cb.start_match      = (int)(mstart - md->start_subject);
1313          cb.current_position = (int)(eptr - md->start_subject);          cb.current_position = (int)(eptr - md->start_subject);
1314          cb.pattern_position = GET(ecode, LINK_SIZE + 3);          cb.pattern_position = GET(ecode, 2);
1315          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);          cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1316          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1317          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last & CAPLMASK;
1318            /* Internal change requires this for API compatibility. */
1319            if (cb.capture_last == 0) cb.capture_last = -1;
1320          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1321          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);          cb.mark             = md->nomatch_mark;
1322            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1323          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1324          }          }
       ecode += _pcre_OP_lengths[OP_CALLOUT];  
       }  
1325    
1326      condcode = ecode[LINK_SIZE+1];        /* Advance ecode past the callout, so it now points to the condition. We
1327          must adjust codelink so that the value of ecode+codelink is unchanged. */
1328    
1329      /* Now see what the actual condition is */        ecode += PRIV(OP_lengths)[OP_CALLOUT];
1330          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1331          }
1332    
1333      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */      /* Test the various possible conditions */
1334    
1335        condition = FALSE;
1336        switch(condcode = *ecode)
1337        {        {
1338        if (md->recursive == NULL)                /* Not recursing => FALSE */        case OP_RREF:         /* Numbered group recursion test */
1339          if (md->recursive != NULL)     /* Not recursing => FALSE */
1340          {          {
1341          condition = FALSE;          unsigned int recno = GET2(ecode, 1);   /* Recursion group number*/
1342          ecode += GET(ecode, 1);          condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1343          }          }
1344        else        break;
         {  
         int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/  
         condition =  (recno == RREF_ANY || recno == md->recursive->group_num);  
   
         /* If the test is for recursion into a specific subpattern, and it is  
         false, but the test was set up by name, scan the table to see if the  
         name refers to any other numbers, and test them. The condition is true  
         if any one is set. */  
   
         if (!condition && condcode == OP_NRREF && recno != RREF_ANY)  
           {  
           uschar *slotA = md->name_table;  
           for (i = 0; i < md->name_count; i++)  
             {  
             if (GET2(slotA, 0) == recno) break;  
             slotA += md->name_entry_size;  
             }  
   
           /* Found a name for the number - there can be only one; duplicate  
           names for different numbers are allowed, but not vice versa. First  
           scan down for duplicates. */  
   
           if (i < md->name_count)  
             {  
             uschar *slotB = slotA;  
             while (slotB > md->name_table)  
               {  
               slotB -= md->name_entry_size;  
               if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)  
                 {  
                 condition = GET2(slotB, 0) == md->recursive->group_num;  
                 if (condition) break;  
                 }  
               else break;  
               }  
   
             /* Scan up for duplicates */  
1345    
1346              if (!condition)        case OP_DNRREF:       /* Duplicate named group recursion test */
1347                {        if (md->recursive != NULL)
1348                slotB = slotA;          {
1349                for (i++; i < md->name_count; i++)          int count = GET2(ecode, 1 + IMM2_SIZE);
1350                  {          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1351                  slotB += md->name_entry_size;          while (count-- > 0)
1352                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)            {
1353                    {            unsigned int recno = GET2(slot, 0);
1354                    condition = GET2(slotB, 0) == md->recursive->group_num;            condition = recno == md->recursive->group_num;
1355                    if (condition) break;            if (condition) break;
1356                    }            slot += md->name_entry_size;
                 else break;  
                 }  
               }  
             }  
1357            }            }
   
         /* Chose branch according to the condition */  
   
         ecode += condition? 3 : GET(ecode, 1);  
1358          }          }
1359        }        break;
1360    
1361      else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */        case OP_CREF:         /* Numbered group used test */
1362        {        offset = GET2(ecode, 1) << 1;  /* Doubled ref number */
       offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */  
1363        condition = offset < offset_top && md->offset_vector[offset] >= 0;        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1364          break;
1365    
1366        /* If the numbered capture is unset, but the reference was by name,        case OP_DNCREF:      /* Duplicate named group used test */
       scan the table to see if the name refers to any other numbers, and test  
       them. The condition is true if any one is set. This is tediously similar  
       to the code above, but not close enough to try to amalgamate. */  
   
       if (!condition && condcode == OP_NCREF)  
1367          {          {
1368          int refno = offset >> 1;          int count = GET2(ecode, 1 + IMM2_SIZE);
1369          uschar *slotA = md->name_table;          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1370            while (count-- > 0)
1371          for (i = 0; i < md->name_count; i++)            {
1372            {            offset = GET2(slot, 0) << 1;
1373            if (GET2(slotA, 0) == refno) break;            condition = offset < offset_top && md->offset_vector[offset] >= 0;
1374            slotA += md->name_entry_size;            if (condition) break;
1375            }            slot += md->name_entry_size;
   
         /* Found a name for the number - there can be only one; duplicate names  
         for different numbers are allowed, but not vice versa. First scan down  
         for duplicates. */  
   
         if (i < md->name_count)  
           {  
           uschar *slotB = slotA;  
           while (slotB > md->name_table)  
             {  
             slotB -= md->name_entry_size;  
             if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)  
               {  
               offset = GET2(slotB, 0) << 1;  
               condition = offset < offset_top &&  
                 md->offset_vector[offset] >= 0;  
               if (condition) break;  
               }  
             else break;  
             }  
   
           /* Scan up for duplicates */  
   
           if (!condition)  
             {  
             slotB = slotA;  
             for (i++; i < md->name_count; i++)  
               {  
               slotB += md->name_entry_size;  
               if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)  
                 {  
                 offset = GET2(slotB, 0) << 1;  
                 condition = offset < offset_top &&  
                   md->offset_vector[offset] >= 0;  
                 if (condition) break;  
                 }  
               else break;  
               }  
             }  
1376            }            }
1377          }          }
1378          break;
1379    
1380        /* Chose branch according to the condition */        case OP_DEF:     /* DEFINE - always false */
1381          break;
       ecode += condition? 3 : GET(ecode, 1);  
       }  
   
     else if (condcode == OP_DEF)     /* DEFINE - always false */  
       {  
       condition = FALSE;  
       ecode += GET(ecode, 1);  
       }  
   
     /* The condition is an assertion. Call match() to evaluate it - setting  
     md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of  
     an assertion. */  
1382    
1383      else        /* The condition is an assertion. Call match() to evaluate it - setting
1384        {        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end
1385        md->match_function_type = MATCH_CONDASSERT;        of an assertion. */
1386        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);  
1387          default:
1388          md->match_function_type = MATCH_CONDASSERT;
1389          RMATCH(eptr, ecode, offset_top, md, NULL, RM3);
1390        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1391          {          {
1392            if (md->end_offset_top > offset_top)
1393              offset_top = md->end_offset_top;  /* Captures may have happened */
1394          condition = TRUE;          condition = TRUE;
1395          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);  
1396            /* Advance ecode past the assertion to the start of the first branch,
1397            but adjust it so that the general choosing code below works. */
1398    
1399            ecode += GET(ecode, 1);
1400          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1401            ecode += 1 + LINK_SIZE - PRIV(OP_lengths)[condcode];
1402          }          }
1403        else if (rrc != MATCH_NOMATCH &&  
1404                (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1405          assertion; it is therefore treated as NOMATCH. Any other return is an
1406          error. */
1407    
1408          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1409          {          {
1410          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1411          }          }
1412        else        break;
         {  
         condition = FALSE;  
         ecode += codelink;  
         }  
1413        }        }
1414    
1415      /* We are now at the branch that is to be obeyed. As there is only one,      /* Choose branch according to the condition */
1416      we used to use tail recursion to avoid using another stack frame, except  
1417      when there was unlimited repeat of a possibly empty group. However, that      ecode += condition? PRIV(OP_lengths)[condcode] : codelink;
     strategy no longer works because of the possibilty of (*THEN) being  
     encountered in the branch. A recursive call to match() is always required,  
     unless the second alternative doesn't exist, in which case we can just  
     plough on. */  
1418    
1419      if (condition || *ecode == OP_ALT)      /* We are now at the branch that is to be obeyed. As there is only one, we
1420        can use tail recursion to avoid using another stack frame, except when
1421        there is unlimited repeat of a possibly empty group. In the latter case, a
1422        recursive call to match() is always required, unless the second alternative
1423        doesn't exist, in which case we can just plough on. Note that, for
1424        compatibility with Perl, the | in a conditional group is NOT treated as
1425        creating two alternatives. If a THEN is encountered in the branch, it
1426        propagates out to the enclosing alternative (unless nested in a deeper set
1427        of alternatives, of course). */
1428    
1429        if (condition || ecode[-(1+LINK_SIZE)] == OP_ALT)
1430        {        {
1431        if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;        if (op != OP_SCOND)
1432        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);          {
1433        if (rrc == MATCH_THEN && md->start_match_ptr == ecode)          goto TAIL_RECURSE;
1434          rrc = MATCH_NOMATCH;          }
1435    
1436          md->match_function_type = MATCH_CBEGROUP;
1437          RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);
1438        RRETURN(rrc);        RRETURN(rrc);
1439        }        }
1440      else                         /* Condition false & no alternative */  
1441         /* Condition false & no alternative; continue after the group. */
1442    
1443        else
1444        {        {
       ecode += 1 + LINK_SIZE;  
1445        }        }
1446      break;      break;
1447    
# Line 1270  for (;;) Line 1450  for (;;)
1450      to close any currently open capturing brackets. */      to close any currently open capturing brackets. */
1451    
1452      case OP_CLOSE:      case OP_CLOSE:
1453      number = GET2(ecode, 1);      number = GET2(ecode, 1);   /* Must be less than 65536 */
1454      offset = number << 1;      offset = number << 1;
1455    
1456  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 1278  for (;;) Line 1458  for (;;)
1458        printf("\n");        printf("\n");
1459  #endif  #endif
1460    
1461      md->capture_last = number;      md->capture_last = (md->capture_last & OVFLMASK) | number;
1462      if (offset >= md->offset_max) md->offset_overflow = TRUE; else      if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1463        {        {
1464        md->offset_vector[offset] =        md->offset_vector[offset] =
1465          md->offset_vector[md->offset_end - number];          md->offset_vector[md->offset_end - number];
1466        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1467        if (offset_top <= offset) offset_top = offset + 2;        if (offset_top <= offset) offset_top = offset + 2;
1468        }        }
1469      ecode += 3;      ecode += 1 + IMM2_SIZE;
1470      break;      break;
1471    
1472    
1473      /* End of the pattern, either real or forced. If we are in a recursion, we      /* End of the pattern, either real or forced. */
     should restore the offsets appropriately, and if it's a top-level  
     recursion, continue from after the call. */  
1474    
     case OP_ACCEPT:  
1475      case OP_END:      case OP_END:
1476      if (md->recursive != NULL)      case OP_ACCEPT:
1477        {      case OP_ASSERT_ACCEPT:
1478        recursion_info *rec = md->recursive;  
1479        md->recursive = rec->prevrec;      /* If we have matched an empty string, fail if not in an assertion and not
1480        memmove(md->offset_vector, rec->offset_save,      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1481          rec->saved_max * sizeof(int));      is set and we have matched at the start of the subject. In both cases,
1482        offset_top = rec->save_offset_top;      backtracking will then try other alternatives, if any. */
1483        if (rec->group_num == 0)  
1484          {      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1485          ecode = rec->after_call;           md->recursive == NULL &&
1486          break;           (md->notempty ||
1487          }             (md->notempty_atstart &&
1488        }               mstart == md->start_subject + md->start_offset)))
1489          RRETURN(MATCH_NOMATCH);
     /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is  
     set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of  
     the subject. In both cases, backtracking will then try other alternatives,  
     if any. */  
   
     else if (eptr == mstart &&  
         (md->notempty ||  
           (md->notempty_atstart &&  
             mstart == md->start_subject + md->start_offset)))  
       MRRETURN(MATCH_NOMATCH);  
1490    
1491      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
1492    
1493      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1494      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1495      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1496    
1497      /* For some reason, the macros don't work properly if an expression is      /* For some reason, the macros don't work properly if an expression is
1498      given as the argument to MRRETURN when the heap is in use. */      given as the argument to RRETURN when the heap is in use. */
1499    
1500      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1501      MRRETURN(rrc);      RRETURN(rrc);
1502    
1503      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1504      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1505      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1506      start of each branch to move the current point backwards, so the code at      start of each branch to move the current point backwards, so the code at
1507      this level is identical to the lookahead case. When the assertion is part      this level is identical to the lookahead case. When the assertion is part
1508      of a condition, we want to return immediately afterwards. The caller of      of a condition, we want to return immediately afterwards. The caller of
1509      this incarnation of the match() function will have set MATCH_CONDASSERT in      this incarnation of the match() function will have set MATCH_CONDASSERT in
1510      md->match_function type, and one of these opcodes will be the first opcode      md->match_function type, and one of these opcodes will be the first opcode
1511      that is processed. We use a local variable that is preserved over calls to      that is processed. We use a local variable that is preserved over calls to
1512      match() to remember this case. */      match() to remember this case. */
1513    
1514      case OP_ASSERT:      case OP_ASSERT:
1515      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1516        save_mark = md->mark;
1517      if (md->match_function_type == MATCH_CONDASSERT)      if (md->match_function_type == MATCH_CONDASSERT)
1518        {        {
1519        condassert = TRUE;        condassert = TRUE;
1520        md->match_function_type = 0;        md->match_function_type = 0;
1521        }        }
1522      else condassert = FALSE;      else condassert = FALSE;
1523    
1524        /* Loop for each branch */
1525    
1526      do      do
1527        {        {
1528        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1529    
1530          /* A match means that the assertion is true; break out of the loop
1531          that matches its alternatives. */
1532    
1533        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1534          {          {
1535          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
1536          break;          break;
1537          }          }
1538        if (rrc != MATCH_NOMATCH &&  
1539            (rrc != MATCH_THEN || md->start_match_ptr != ecode))        /* If not matched, restore the previous mark setting. */
1540          RRETURN(rrc);  
1541          md->mark = save_mark;
1542    
1543          /* See comment in the code for capturing groups above about handling
1544          THEN. */
1545    
1546          if (rrc == MATCH_THEN)
1547            {
1548            next = ecode + GET(ecode,1);
1549            if (md->start_match_ptr < next &&
1550                (*ecode == OP_ALT || *next == OP_ALT))
1551              rrc = MATCH_NOMATCH;
1552            }
1553    
1554          /* Anything other than NOMATCH causes the entire assertion to fail,
1555          passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1556          uncaptured THEN, which means they take their normal effect. This
1557          consistent approach does not always have exactly the same effect as in
1558          Perl. */
1559    
1560          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1561        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1562        }        }
1563      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);   /* Continue for next alternative */
1564    
1565      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);      /* If we have tried all the alternative branches, the assertion has
1566        failed. If not, we broke out after a match. */
1567    
1568        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1569    
1570      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1571    
1572      if (condassert) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1573    
1574      /* Continue from after the assertion, updating the offsets high water      /* Continue from after a successful assertion, updating the offsets high
1575      mark, since extracts may have been taken during the assertion. */      water mark, since extracts may have been taken during the assertion. */
1576    
1577      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1578      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1579      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1580      continue;      continue;
1581    
1582      /* Negative assertion: all branches must fail to match. Encountering SKIP,      /* Negative assertion: all branches must fail to match for the assertion to
1583      PRUNE, or COMMIT means we must assume failure without checking subsequent      succeed. */
     branches. */  
1584    
1585      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1586      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1587        save_mark = md->mark;
1588      if (md->match_function_type == MATCH_CONDASSERT)      if (md->match_function_type == MATCH_CONDASSERT)
1589        {        {
1590        condassert = TRUE;        condassert = TRUE;
1591        md->match_function_type = 0;        md->match_function_type = 0;
1592        }        }
1593      else condassert = FALSE;      else condassert = FALSE;
1594    
1595        /* Loop for each alternative branch. */
1596    
1597      do      do
1598        {        {
1599        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1600        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);        md->mark = save_mark;   /* Always restore the mark setting */
1601        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)  
1602          switch(rrc)
1603          {          {
1604          do ecode += GET(ecode,1); while (*ecode == OP_ALT);          case MATCH_MATCH:            /* A successful match means */
1605            case MATCH_ACCEPT:           /* the assertion has failed. */
1606            RRETURN(MATCH_NOMATCH);
1607    
1608            case MATCH_NOMATCH:          /* Carry on with next branch */
1609          break;          break;
1610          }  
1611        if (rrc != MATCH_NOMATCH &&          /* See comment in the code for capturing groups above about handling
1612            (rrc != MATCH_THEN || md->start_match_ptr != ecode))          THEN. */
1613    
1614            case MATCH_THEN:
1615            next = ecode + GET(ecode,1);
1616            if (md->start_match_ptr < next &&
1617                (*ecode == OP_ALT || *next == OP_ALT))
1618              {
1619              rrc = MATCH_NOMATCH;
1620              break;
1621              }
1622            /* Otherwise fall through. */
1623    
1624            /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1625            assertion to fail to match, without considering any more alternatives.
1626            Failing to match means the assertion is true. This is a consistent
1627            approach, but does not always have the same effect as in Perl. */
1628    
1629            case MATCH_COMMIT:
1630            case MATCH_SKIP:
1631            case MATCH_SKIP_ARG:
1632            case MATCH_PRUNE:
1633            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1634            goto NEG_ASSERT_TRUE;   /* Break out of alternation loop */
1635    
1636            /* Anything else is an error */
1637    
1638            default:
1639          RRETURN(rrc);          RRETURN(rrc);
1640            }
1641    
1642          /* Continue with next branch */
1643    
1644        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1645        }        }
1646      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1647    
1648        /* All branches in the assertion failed to match. */
1649    
1650        NEG_ASSERT_TRUE:
1651      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1652        ecode += 1 + LINK_SIZE;                /* Continue with current branch */
     ecode += 1 + LINK_SIZE;  
1653      continue;      continue;
1654    
1655      /* 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 1422  for (;;) Line 1658  for (;;)
1658      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1659    
1660      case OP_REVERSE:      case OP_REVERSE:
1661  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1662      if (utf8)      if (utf)
1663        {        {
1664        i = GET(ecode, 1);        i = GET(ecode, 1);
1665        while (i-- > 0)        while (i-- > 0)
1666          {          {
1667          eptr--;          eptr--;
1668          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1669          BACKCHAR(eptr);          BACKCHAR(eptr);
1670          }          }
1671        }        }
# Line 1440  for (;;) Line 1676  for (;;)
1676    
1677        {        {
1678        eptr -= GET(ecode, 1);        eptr -= GET(ecode, 1);
1679        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1680        }        }
1681    
1682      /* Save the earliest consulted character, then skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
# Line 1454  for (;;) Line 1690  for (;;)
1690      function is able to force a failure. */      function is able to force a failure. */
1691    
1692      case OP_CALLOUT:      case OP_CALLOUT:
1693      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1694        {        {
1695        pcre_callout_block cb;        PUBL(callout_block) cb;
1696        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1697        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1698        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1699    #if defined COMPILE_PCRE8
1700        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1701    #elif defined COMPILE_PCRE16
1702          cb.subject          = (PCRE_SPTR16)md->start_subject;
1703    #elif defined COMPILE_PCRE32
1704          cb.subject          = (PCRE_SPTR32)md->start_subject;
1705    #endif
1706        cb.subject_length   = (int)(md->end_subject - md->start_subject);        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1707        cb.start_match      = (int)(mstart - md->start_subject);        cb.start_match      = (int)(mstart - md->start_subject);
1708        cb.current_position = (int)(eptr - md->start_subject);        cb.current_position = (int)(eptr - md->start_subject);
1709        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1710        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1711        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1712        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1713          /* Internal change requires this for API compatibility. */
1714          if (cb.capture_last == 0) cb.capture_last = -1;
1715        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1716        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1717          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1718        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1719        }        }
1720      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 1479  for (;;) Line 1724  for (;;)
1724      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
1725      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1726    
1727      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1728      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
1729      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
1730      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
1731      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
1732      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
1733      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.  
1734    
1735      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
1736      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
1737      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1738        a lot, so he is not to blame for the current way it works. */
1739    
1740      case OP_RECURSE:      case OP_RECURSE:
1741        {        {
1742          recursion_info *ri;
1743          unsigned int recno;
1744    
1745        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1746        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1747          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1748    
1749          /* Check for repeating a recursion without advancing the subject pointer.
1750          This should catch convoluted mutual recursions. (Some simple cases are
1751          caught at compile time.) */
1752    
1753          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1754            if (recno == ri->group_num && eptr == ri->subject_position)
1755              RRETURN(PCRE_ERROR_RECURSELOOP);
1756    
1757        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1758    
1759          new_recursive.group_num = recno;
1760          new_recursive.saved_capture_last = md->capture_last;
1761          new_recursive.subject_position = eptr;
1762        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1763        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1764    
1765        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1766    
1767        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1768    
1769        /* Now save the offset data. */        /* Now save the offset data */
1770    
1771        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1772        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1518  for (;;) Line 1774  for (;;)
1774        else        else
1775          {          {
1776          new_recursive.offset_save =          new_recursive.offset_save =
1777            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1778          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1779          }          }
   
1780        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1781              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1782        new_recursive.save_offset_top = offset_top;  
1783          /* OK, now we can do the recursion. After processing each alternative,
1784        /* OK, now we can do the recursion. For each top-level alternative we        restore the offset data and the last captured value. If there were nested
1785        restore the offset and recursion data. */        recursions, md->recursive might be changed, so reset it before looping.
1786          */
1787    
1788        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1789        cbegroup = (*callpat >= OP_SBRA);        cbegroup = (*callpat >= OP_SBRA);
1790        do        do
1791          {          {
1792          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1793          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1794            md, eptrb, RM6);            md, eptrb, RM6);
1795            memcpy(md->offset_vector, new_recursive.offset_save,
1796                new_recursive.saved_max * sizeof(int));
1797            md->capture_last = new_recursive.saved_capture_last;
1798            md->recursive = new_recursive.prevrec;
1799          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1800            {            {
1801            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1802            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1803              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1804            MRRETURN(MATCH_MATCH);  
1805              /* Set where we got to in the subject, and reset the start in case
1806              it was changed by \K. This *is* propagated back out of a recursion,
1807              for Perl compatibility. */
1808    
1809              eptr = md->end_match_ptr;
1810              mstart = md->start_match_ptr;
1811              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1812            }            }
1813          else if (rrc != MATCH_NOMATCH &&  
1814                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))          /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1815            recursion; they cause a NOMATCH for the entire recursion. These codes
1816            are defined in a range that can be tested for. */
1817    
1818            if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1819              RRETURN(MATCH_NOMATCH);
1820    
1821            /* Any return code other than NOMATCH is an error. */
1822    
1823            if (rrc != MATCH_NOMATCH)
1824            {            {
1825            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1826            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1827              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1828            RRETURN(rrc);            RRETURN(rrc);
1829            }            }
1830    
1831          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1832          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1833          }          }
1834        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1563  for (;;) Line 1836  for (;;)
1836        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1837        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1838        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1839          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1840        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
       }  
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer, but resetting  
     the start-of-match value in case it was changed by \K. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);  
       if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */  
         {  
         mstart = md->start_match_ptr;  
         break;  
         }  
       if (rrc != MATCH_NOMATCH &&  
           (rrc != MATCH_THEN || md->start_match_ptr != ecode))  
         RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
1841        }        }
1842    
1843      /* The repeating kets try the rest of the pattern or restart from the      RECURSION_MATCHED:
1844      preceding bracket, in the appropriate order. The second "call" of match()      break;
     uses tail recursion, to avoid using another stack frame. */  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       md->match_function_type = MATCH_CBEGROUP;  
       RMATCH(eptr, prev, offset_top, md, eptrb, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1845    
1846      /* 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
1847      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1652  for (;;) Line 1855  for (;;)
1855      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1856      with fixed upper repeat limits are compiled as a number of copies, with the      with fixed upper repeat limits are compiled as a number of copies, with the
1857      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1858    
1859      case OP_BRAZERO:      case OP_BRAZERO:
1860      next = ecode + 1;      next = ecode + 1;
1861      RMATCH(eptr, next, offset_top, md, eptrb, RM10);      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
# Line 1660  for (;;) Line 1863  for (;;)
1863      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
1864      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1865      break;      break;
1866    
1867      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1868      next = ecode + 1;      next = ecode + 1;
1869      do next += GET(next, 1); while (*next == OP_ALT);      do next += GET(next, 1); while (*next == OP_ALT);
# Line 1674  for (;;) Line 1877  for (;;)
1877      do next += GET(next,1); while (*next == OP_ALT);      do next += GET(next,1); while (*next == OP_ALT);
1878      ecode = next + 1 + LINK_SIZE;      ecode = next + 1 + LINK_SIZE;
1879      break;      break;
1880    
1881      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1882      here; just jump to the group, with allow_zero set TRUE. */      here; just jump to the group, with allow_zero set TRUE. */
1883    
1884      case OP_BRAPOSZERO:      case OP_BRAPOSZERO:
1885      op = *(++ecode);      op = *(++ecode);
1886      allow_zero = TRUE;      allow_zero = TRUE;
1887      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1888        goto POSSESSIVE_NON_CAPTURE;        goto POSSESSIVE_NON_CAPTURE;
# Line 1689  for (;;) Line 1892  for (;;)
1892      case OP_KET:      case OP_KET:
1893      case OP_KETRMIN:      case OP_KETRMIN:
1894      case OP_KETRMAX:      case OP_KETRMAX:
1895      case OP_KETRPOS:      case OP_KETRPOS:
1896      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1897    
1898      /* 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
1899      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1900      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1901    
1902      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1903        {        {
1904        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1905        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1906        }        }
1907      else saved_eptr = NULL;      else saved_eptr = NULL;
1908    
1909      /* If we are at the end of an assertion group or an atomic group, stop      /* If we are at the end of an assertion group or a non-capturing atomic
1910      matching and return MATCH_MATCH, but record the current high water mark for      group, stop matching and return MATCH_MATCH, but record the current high
1911      use by positive assertions. We also need to record the match start in case      water mark for use by positive assertions. We also need to record the match
1912      it was changed by \K. */      start in case it was changed by \K. */
1913    
1914      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1915          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||           *prev == OP_ONCE_NC)
         *prev == OP_ONCE)  
1916        {        {
1917        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1918        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1919        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1920        MRRETURN(MATCH_MATCH);        RRETURN(MATCH_MATCH);         /* Sets md->mark */
1921        }        }
1922    
1923      /* 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
1924      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1925      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
1926      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
1927      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
1928        the current subject position and start match pointer and give a MATCH
1929        return. */
1930    
1931      if (*prev == OP_CBRA || *prev == OP_SCBRA ||      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1932          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
# Line 1735  for (;;) Line 1939  for (;;)
1939        printf("\n");        printf("\n");
1940  #endif  #endif
1941    
1942        md->capture_last = number;        /* Handle a recursively called group. */
1943        if (offset >= md->offset_max) md->offset_overflow = TRUE; else  
1944          if (md->recursive != NULL && md->recursive->group_num == number)
1945          {          {
1946          md->offset_vector[offset] =          md->end_match_ptr = eptr;
1947            md->offset_vector[md->offset_end - number];          md->start_match_ptr = mstart;
1948          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);          RRETURN(MATCH_MATCH);
         if (offset_top <= offset) offset_top = offset + 2;  
1949          }          }
1950    
1951        /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
       appropriately and continue from after the call. */  
1952    
1953        if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
1954          if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1955          {          {
1956          recursion_info *rec = md->recursive;          /* If offset is greater than offset_top, it means that we are
1957          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          "skipping" a capturing group, and that group's offsets must be marked
1958          md->recursive = rec->prevrec;          unset. In earlier versions of PCRE, all the offsets were unset at the
1959          memcpy(md->offset_vector, rec->offset_save,          start of matching, but this doesn't work because atomic groups and
1960            rec->saved_max * sizeof(int));          assertions can cause a value to be set that should later be unset.
1961          offset_top = rec->save_offset_top;          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1962          ecode = rec->after_call;          part of the atomic group, but this is not on the final matching path,
1963          break;          so must be unset when 2 is set. (If there is no group 2, there is no
1964            problem, because offset_top will then be 2, indicating no capture.) */
1965    
1966            if (offset > offset_top)
1967              {
1968              register int *iptr = md->offset_vector + offset_top;
1969              register int *iend = md->offset_vector + offset;
1970              while (iptr < iend) *iptr++ = -1;
1971              }
1972    
1973            /* Now make the extraction */
1974    
1975            md->offset_vector[offset] =
1976              md->offset_vector[md->offset_end - number];
1977            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1978            if (offset_top <= offset) offset_top = offset + 2;
1979          }          }
1980        }        }
1981    
1982      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1983      happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1984      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1985      5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group that includes captures,
1986      course of events. */      establish a backup point by processing the rest of the pattern at a lower
1987        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1988        original OP_ONCE level, thereby bypassing intermediate backup points, but
1989        resetting any captures that happened along the way. */
1990    
1991      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1992        {        {
1993        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1994            {
1995            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1996            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1997            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1998            RRETURN(MATCH_ONCE);
1999            }
2000          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2001        break;        break;
2002        }        }
2003    
2004      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2005      and return the MATCH_KETRPOS. This makes it possible to do the repeats one      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2006      at a time from the outer level, thus saving stack. */      at a time from the outer level, thus saving stack. */
2007    
2008      if (*ecode == OP_KETRPOS)      if (*ecode == OP_KETRPOS)
2009        {        {
2010          md->start_match_ptr = mstart;    /* In case \K reset it */
2011        md->end_match_ptr = eptr;        md->end_match_ptr = eptr;
2012        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
2013        RRETURN(MATCH_KETRPOS);        RRETURN(MATCH_KETRPOS);
2014        }        }
2015    
2016      /* The normal repeating kets try the rest of the pattern or restart from      /* The normal repeating kets try the rest of the pattern or restart from
2017      the preceding bracket, in the appropriate order. In the second case, we can      the preceding bracket, in the appropriate order. In the second case, we can
2018      use tail recursion to avoid using another stack frame, unless we have an      use tail recursion to avoid using another stack frame, unless we have an
2019      unlimited repeat of a group that can match an empty string. */      an atomic group or an unlimited repeat of a group that can match an empty
2020        string. */
2021    
2022      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2023        {        {
2024        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2025        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2026          if (*prev == OP_ONCE)
2027            {
2028            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2029            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2030            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2031            RRETURN(MATCH_ONCE);
2032            }
2033        if (*prev >= OP_SBRA)    /* Could match an empty string */        if (*prev >= OP_SBRA)    /* Could match an empty string */
2034          {          {
         md->match_function_type = MATCH_CBEGROUP;  
2035          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2036          RRETURN(rrc);          RRETURN(rrc);
2037          }          }
# Line 1803  for (;;) Line 2040  for (;;)
2040        }        }
2041      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2042        {        {
       if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;  
2043        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2044          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2045        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2046          if (*prev == OP_ONCE)
2047            {
2048            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2049            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2050            md->once_target = prev;
2051            RRETURN(MATCH_ONCE);
2052            }
2053        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
2054        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2055        }        }
# Line 1814  for (;;) Line 2058  for (;;)
2058      /* Not multiline mode: start of subject assertion, unless notbol. */      /* Not multiline mode: start of subject assertion, unless notbol. */
2059    
2060      case OP_CIRC:      case OP_CIRC:
2061      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2062    
2063      /* Start of subject assertion */      /* Start of subject assertion */
2064    
2065      case OP_SOD:      case OP_SOD:
2066      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
2067      ecode++;      ecode++;
2068      break;      break;
2069    
2070      /* Multiline mode: start of subject unless notbol, or after any newline. */      /* Multiline mode: start of subject unless notbol, or after any newline. */
2071    
2072      case OP_CIRCM:      case OP_CIRCM:
2073      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2074      if (eptr != md->start_subject &&      if (eptr != md->start_subject &&
2075          (eptr == md->end_subject || !WAS_NEWLINE(eptr)))          (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2076        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2077      ecode++;      ecode++;
2078      break;      break;
2079    
2080      /* Start of match assertion */      /* Start of match assertion */
2081    
2082      case OP_SOM:      case OP_SOM:
2083      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
2084      ecode++;      ecode++;
2085      break;      break;
2086    
# Line 1852  for (;;) Line 2096  for (;;)
2096    
2097      case OP_DOLLM:      case OP_DOLLM:
2098      if (eptr < md->end_subject)      if (eptr < md->end_subject)
2099        { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }        {
2100          if (!IS_NEWLINE(eptr))
2101            {
2102            if (md->partial != 0 &&
2103                eptr + 1 >= md->end_subject &&
2104                NLBLOCK->nltype == NLTYPE_FIXED &&
2105                NLBLOCK->nllen == 2 &&
2106                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2107              {
2108              md->hitend = TRUE;
2109              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2110              }
2111            RRETURN(MATCH_NOMATCH);
2112            }
2113          }
2114      else      else
2115        {        {
2116        if (md->noteol) MRRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2117        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2118        }        }
2119      ecode++;      ecode++;
2120      break;      break;
2121    
2122      /* Not multiline mode: assert before a terminating newline or before end of      /* Not multiline mode: assert before a terminating newline or before end of
2123      subject unless noteol is set. */      subject unless noteol is set. */
2124    
2125      case OP_DOLL:      case OP_DOLL:
2126      if (md->noteol) MRRETURN(MATCH_NOMATCH);      if (md->noteol) RRETURN(MATCH_NOMATCH);
2127      if (!md->endonly) goto ASSERT_NL_OR_EOS;      if (!md->endonly) goto ASSERT_NL_OR_EOS;
2128    
2129      /* ... else fall through for endonly */      /* ... else fall through for endonly */
# Line 1873  for (;;) Line 2131  for (;;)
2131      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2132    
2133      case OP_EOD:      case OP_EOD:
2134      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2135      SCHECK_PARTIAL();      SCHECK_PARTIAL();
2136      ecode++;      ecode++;
2137      break;      break;
# Line 1884  for (;;) Line 2142  for (;;)
2142      ASSERT_NL_OR_EOS:      ASSERT_NL_OR_EOS:
2143      if (eptr < md->end_subject &&      if (eptr < md->end_subject &&
2144          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2145        MRRETURN(MATCH_NOMATCH);        {
2146          if (md->partial != 0 &&
2147              eptr + 1 >= md->end_subject &&
2148              NLBLOCK->nltype == NLTYPE_FIXED &&
2149              NLBLOCK->nllen == 2 &&
2150              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2151            {
2152            md->hitend = TRUE;
2153            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2154            }
2155          RRETURN(MATCH_NOMATCH);
2156          }
2157    
2158      /* Either at end of string or \n before end. */      /* Either at end of string or \n before end. */
2159    
# Line 1903  for (;;) Line 2172  for (;;)
2172        be "non-word" characters. Remember the earliest consulted character for        be "non-word" characters. Remember the earliest consulted character for
2173        partial matching. */        partial matching. */
2174    
2175  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2176        if (utf8)        if (utf)
2177          {          {
2178          /* Get status of previous character */          /* Get status of previous character */
2179    
2180          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2181            {            {
2182            USPTR lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2183            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2184            if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;            if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2185            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2186  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 1976  for (;;) Line 2245  for (;;)
2245              }              }
2246            else            else
2247  #endif  #endif
2248            prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);            prev_is_word = MAX_255(eptr[-1])
2249                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2250            }            }
2251    
2252          /* Get status of next character */          /* Get status of next character */
# Line 1999  for (;;) Line 2269  for (;;)
2269            }            }
2270          else          else
2271  #endif  #endif
2272          cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);          cur_is_word = MAX_255(*eptr)
2273              && ((md->ctypes[*eptr] & ctype_word) != 0);
2274          }          }
2275    
2276        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
2277    
2278        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
2279             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2280          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2281        }        }
2282      break;      break;
2283    
2284      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2285        CRLF newlines and partial matching. */
2286    
2287      case OP_ANY:      case OP_ANY:
2288      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2289        if (md->partial != 0 &&
2290            eptr + 1 >= md->end_subject &&
2291            NLBLOCK->nltype == NLTYPE_FIXED &&
2292            NLBLOCK->nllen == 2 &&
2293            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2294          {
2295          md->hitend = TRUE;
2296          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2297          }
2298    
2299      /* Fall through */      /* Fall through */
2300    
2301        /* Match any single character whatsoever. */
2302    
2303      case OP_ALLANY:      case OP_ALLANY:
2304      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2305        {        {                            /* not be updated before SCHECK_PARTIAL. */
2306        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2307        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2308        }        }
2309      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      eptr++;
2310    #ifdef SUPPORT_UTF
2311        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2312    #endif
2313      ecode++;      ecode++;
2314      break;      break;
2315    
# Line 2030  for (;;) Line 2317  for (;;)
2317      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2318    
2319      case OP_ANYBYTE:      case OP_ANYBYTE:
2320      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2321        {        {                            /* not be updated before SCHECK_PARTIAL. */
2322        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2323        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2324        }        }
2325        eptr++;
2326      ecode++;      ecode++;
2327      break;      break;
2328    
# Line 2042  for (;;) Line 2330  for (;;)
2330      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2331        {        {
2332        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2333        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2334        }        }
2335      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2336      if (      if (
2337  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2338         c < 256 &&         c < 256 &&
2339  #endif  #endif
2340         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2341         )         )
2342        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2343      ecode++;      ecode++;
2344      break;      break;
2345    
# Line 2059  for (;;) Line 2347  for (;;)
2347      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2348        {        {
2349        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2350        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2351        }        }
2352      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2353      if (      if (
2354  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2355         c >= 256 ||         c > 255 ||
2356  #endif  #endif
2357         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2358         )         )
2359        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2360      ecode++;      ecode++;
2361      break;      break;
2362    
# Line 2076  for (;;) Line 2364  for (;;)
2364      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2365        {        {
2366        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2367        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2368        }        }
2369      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2370      if (      if (
2371  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2372         c < 256 &&         c < 256 &&
2373  #endif  #endif
2374         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2375         )         )
2376        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2377      ecode++;      ecode++;
2378      break;      break;
2379    
# Line 2093  for (;;) Line 2381  for (;;)
2381      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2382        {        {
2383        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2384        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2385        }        }
2386      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2387      if (      if (
2388  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2389         c >= 256 ||         c > 255 ||
2390  #endif  #endif
2391         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2392         )         )
2393        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2394      ecode++;      ecode++;
2395      break;      break;
2396    
# Line 2110  for (;;) Line 2398  for (;;)
2398      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2399        {        {
2400        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2401        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2402        }        }
2403      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2404      if (      if (
2405  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2406         c < 256 &&         c < 256 &&
2407  #endif  #endif
2408         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2409         )         )
2410        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2411      ecode++;      ecode++;
2412      break;      break;
2413    
# Line 2127  for (;;) Line 2415  for (;;)
2415      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2416        {        {
2417        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2418        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2419        }        }
2420      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2421      if (      if (
2422  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2423         c >= 256 ||         c > 255 ||
2424  #endif  #endif
2425         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2426         )         )
2427        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2428      ecode++;      ecode++;
2429      break;      break;
2430    
# Line 2144  for (;;) Line 2432  for (;;)
2432      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2433        {        {
2434        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2435        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2436        }        }
2437      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2438      switch(c)      switch(c)
2439        {        {
2440        default: MRRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2441    
2442        case 0x000d:        case CHAR_CR:
2443        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr >= md->end_subject)
2444            {
2445            SCHECK_PARTIAL();
2446            }
2447          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2448        break;        break;
2449    
2450        case 0x000a:        case CHAR_LF:
2451        break;        break;
2452    
2453        case 0x000b:        case CHAR_VT:
2454        case 0x000c:        case CHAR_FF:
2455        case 0x0085:        case CHAR_NEL:
2456    #ifndef EBCDIC
2457        case 0x2028:        case 0x2028:
2458        case 0x2029:        case 0x2029:
2459        if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);  #endif  /* Not EBCDIC */
2460          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2461        break;        break;
2462        }        }
2463      ecode++;      ecode++;
# Line 2173  for (;;) Line 2467  for (;;)
2467      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2468        {        {
2469        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2470        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2471        }        }
2472      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2473      switch(c)      switch(c)
2474        {        {
2475          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2476        default: break;        default: break;
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       MRRETURN(MATCH_NOMATCH);  
2477        }        }
2478      ecode++;      ecode++;
2479      break;      break;
# Line 2207  for (;;) Line 2482  for (;;)
2482      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2483        {        {
2484        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2485        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2486        }        }
2487      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2488      switch(c)      switch(c)
2489        {        {
2490        default: MRRETURN(MATCH_NOMATCH);        HSPACE_CASES: break;  /* Byte and multibyte cases */
2491        case 0x09:      /* HT */        default: RRETURN(MATCH_NOMATCH);
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       break;  
2492        }        }
2493      ecode++;      ecode++;
2494      break;      break;
# Line 2241  for (;;) Line 2497  for (;;)
2497      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2498        {        {
2499        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2500        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2501        }        }
2502      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2503      switch(c)      switch(c)
2504        {        {
2505          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2506        default: break;        default: break;
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       MRRETURN(MATCH_NOMATCH);  
2507        }        }
2508      ecode++;      ecode++;
2509      break;      break;
# Line 2263  for (;;) Line 2512  for (;;)
2512      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2513        {        {
2514        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2515        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2516        }        }
2517      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2518      switch(c)      switch(c)
2519        {        {
2520        default: MRRETURN(MATCH_NOMATCH);        VSPACE_CASES: break;
2521        case 0x0a:      /* LF */        default: RRETURN(MATCH_NOMATCH);
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       break;  
2522        }        }
2523      ecode++;      ecode++;
2524      break;      break;
# Line 2290  for (;;) Line 2532  for (;;)
2532      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2533        {        {
2534        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2535        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2536        }        }
2537      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2538        {        {
2539          const pcre_uint32 *cp;
2540        const ucd_record *prop = GET_UCD(c);        const ucd_record *prop = GET_UCD(c);
2541    
2542        switch(ecode[1])        switch(ecode[1])
2543          {          {
2544          case PT_ANY:          case PT_ANY:
2545          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2546          break;          break;
2547    
2548          case PT_LAMP:          case PT_LAMP:
2549          if ((prop->chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2550               prop->chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2551               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2552            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2553          break;          break;
2554    
2555          case PT_GC:          case PT_GC:
2556          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2557            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2558          break;          break;
2559    
2560          case PT_PC:          case PT_PC:
2561          if ((ecode[2] != prop->chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2562            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2563          break;          break;
2564    
2565          case PT_SC:          case PT_SC:
2566          if ((ecode[2] != prop->script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2567            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2568          break;          break;
2569    
2570          /* These are specials */          /* These are specials */
2571    
2572          case PT_ALNUM:          case PT_ALNUM:
2573          if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||          if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2574               _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))               PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2575            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2576          break;          break;
2577    
2578          case PT_SPACE:    /* Perl space */          /* Perl space used to exclude VT, but from Perl 5.18 it is included,
2579          if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||          which means that Perl space and POSIX space are now identical. PCRE
2580               c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)          was changed at release 8.34. */
                == (op == OP_NOTPROP))  
           MRRETURN(MATCH_NOMATCH);  
         break;  
2581    
2582            case PT_SPACE:    /* Perl space */
2583          case PT_PXSPACE:  /* POSIX space */          case PT_PXSPACE:  /* POSIX space */
2584          if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||          switch(c)
2585               c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||            {
2586               c == CHAR_FF || c == CHAR_CR)            HSPACE_CASES:
2587                 == (op == OP_NOTPROP))            VSPACE_CASES:
2588            MRRETURN(MATCH_NOMATCH);            if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2589              break;
2590    
2591              default:
2592              if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z) ==
2593                (op == OP_NOTPROP)) RRETURN(MATCH_NOMATCH);
2594              break;
2595              }
2596          break;          break;
2597    
2598          case PT_WORD:          case PT_WORD:
2599          if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||          if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2600               _pcre_ucp_gentype[prop->chartype] == ucp_N ||               PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2601               c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))               c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2602            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2603            break;
2604    
2605            case PT_CLIST:
2606            cp = PRIV(ucd_caseless_sets) + ecode[2];
2607            for (;;)
2608              {
2609              if (c < *cp)
2610                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2611              if (c == *cp++)
2612                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2613              }
2614            break;
2615    
2616            case PT_UCNC:
2617            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2618                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2619                 c >= 0xe000) == (op == OP_NOTPROP))
2620              RRETURN(MATCH_NOMATCH);
2621          break;          break;
2622    
2623          /* This should never occur */          /* This should never occur */
# Line 2371  for (;;) Line 2637  for (;;)
2637      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2638        {        {
2639        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2640        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2641        }        }
2642      GETCHARINCTEST(c, eptr);      else
2643        {        {
2644        int category = UCD_CATEGORY(c);        int lgb, rgb;
2645        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        GETCHARINCTEST(c, eptr);
2646          lgb = UCD_GRAPHBREAK(c);
2647        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2648          {          {
2649          int len = 1;          int len = 1;
2650          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2651            {          rgb = UCD_GRAPHBREAK(c);
2652            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2653            }          lgb = rgb;
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
2654          eptr += len;          eptr += len;
2655          }          }
2656        }        }
2657        CHECK_PARTIAL();
2658      ecode++;      ecode++;
2659      break;      break;
2660  #endif  #endif  /* SUPPORT_UCP */
2661    
2662    
2663      /* Match a back reference, possibly repeatedly. Look past the end of the      /* Match a back reference, possibly repeatedly. Look past the end of the
# Line 2400  for (;;) Line 2666  for (;;)
2666      similar code to character type repeats - written out again for speed.      similar code to character type repeats - written out again for speed.
2667      However, if the referenced string is the empty string, always treat      However, if the referenced string is the empty string, always treat
2668      it as matched, any number of times (otherwise there could be infinite      it as matched, any number of times (otherwise there could be infinite
2669      loops). */      loops). If the reference is unset, there are two possibilities:
   
     case OP_REF:  
     case OP_REFI:  
     caseless = op == OP_REFI;  
     offset = GET2(ecode, 1) << 1;               /* Doubled ref number */  
     ecode += 3;  
   
     /* If the reference is unset, there are two possibilities:  
2670    
2671      (a) In the default, Perl-compatible state, set the length negative;      (a) In the default, Perl-compatible state, set the length negative;
2672      this ensures that every attempt at a match fails. We can't just fail      this ensures that every attempt at a match fails. We can't just fail
# Line 2418  for (;;) Line 2676  for (;;)
2676      so that the back reference matches an empty string.      so that the back reference matches an empty string.
2677    
2678      Otherwise, set the length to the length of what was matched by the      Otherwise, set the length to the length of what was matched by the
2679      referenced subpattern. */      referenced subpattern.
2680    
2681        The OP_REF and OP_REFI opcodes are used for a reference to a numbered group
2682        or to a non-duplicated named group. For a duplicated named group, OP_DNREF
2683        and OP_DNREFI are used. In this case we must scan the list of groups to
2684        which the name refers, and use the first one that is set. */
2685    
2686        case OP_DNREF:
2687        case OP_DNREFI:
2688        caseless = op == OP_DNREFI;
2689          {
2690          int count = GET2(ecode, 1+IMM2_SIZE);
2691          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
2692          ecode += 1 + 2*IMM2_SIZE;
2693    
2694          while (count-- > 0)
2695            {
2696            offset = GET2(slot, 0) << 1;
2697            if (offset < offset_top && md->offset_vector[offset] >= 0) break;
2698            slot += md->name_entry_size;
2699            }
2700          if (count < 0)
2701            length = (md->jscript_compat)? 0 : -1;
2702          else
2703            length = md->offset_vector[offset+1] - md->offset_vector[offset];
2704          }
2705        goto REF_REPEAT;
2706    
2707        case OP_REF:
2708        case OP_REFI:
2709        caseless = op == OP_REFI;
2710        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2711        ecode += 1 + IMM2_SIZE;
2712      if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2713        length = (md->jscript_compat)? 0 : -1;        length = (md->jscript_compat)? 0 : -1;
2714      else      else
# Line 2427  for (;;) Line 2716  for (;;)
2716    
2717      /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2718    
2719        REF_REPEAT:
2720      switch (*ecode)      switch (*ecode)
2721        {        {
2722        case OP_CRSTAR:        case OP_CRSTAR:
# Line 2446  for (;;) Line 2736  for (;;)
2736        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2737        minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2738        min = GET2(ecode, 1);        min = GET2(ecode, 1);
2739        max = GET2(ecode, 3);        max = GET2(ecode, 1 + IMM2_SIZE);
2740        if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2741        ecode += 5;        ecode += 1 + 2 * IMM2_SIZE;
2742        break;        break;
2743    
2744        default:               /* No repeat follows */        default:               /* No repeat follows */
2745        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2746          {          {
2747            if (length == -2) eptr = md->end_subject;   /* Partial match */
2748          CHECK_PARTIAL();          CHECK_PARTIAL();
2749          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2750          }          }
2751        eptr += length;        eptr += length;
2752        continue;              /* With the main loop */        continue;              /* With the main loop */
2753        }        }
2754    
2755      /* Handle repeated back references. If the length of the reference is      /* Handle repeated back references. If the length of the reference is
2756      zero, just continue with the main loop. */      zero, just continue with the main loop. If the length is negative, it
2757        means the reference is unset in non-Java-compatible mode. If the minimum is
2758        zero, we can continue at the same level without recursion. For any other
2759        minimum, carrying on will result in NOMATCH. */
2760    
2761      if (length == 0) continue;      if (length == 0) continue;
2762        if (length < 0 && min == 0) continue;
2763    
2764      /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2765      the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
# Line 2472  for (;;) Line 2767  for (;;)
2767    
2768      for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2769        {        {
2770        int slength;        int slength;
2771        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)        if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2772          {          {
2773            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2774          CHECK_PARTIAL();          CHECK_PARTIAL();
2775          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2776          }          }
2777        eptr += slength;        eptr += slength;
2778        }        }
# Line 2492  for (;;) Line 2788  for (;;)
2788        {        {
2789        for (fi = min;; fi++)        for (fi = min;; fi++)
2790          {          {
2791          int slength;          int slength;
2792          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);          RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2793          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2794          if (fi >= max) MRRETURN(MATCH_NOMATCH);          if (fi >= max) RRETURN(MATCH_NOMATCH);
2795          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2796            {            {
2797              if (slength == -2) eptr = md->end_subject;   /* Partial match */
2798            CHECK_PARTIAL();            CHECK_PARTIAL();
2799            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2800            }            }
2801          eptr += slength;          eptr += slength;
2802          }          }
# Line 2513  for (;;) Line 2810  for (;;)
2810        pp = eptr;        pp = eptr;
2811        for (i = min; i < max; i++)        for (i = min; i < max; i++)
2812          {          {
2813          int slength;          int slength;
2814          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2815            {            {
2816            CHECK_PARTIAL();            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2817              the soft partial matching case. */
2818    
2819              if (slength == -2 && md->partial != 0 &&
2820                  md->end_subject > md->start_used_ptr)
2821                {
2822                md->hitend = TRUE;
2823                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2824                }
2825            break;            break;
2826            }            }
2827          eptr += slength;          eptr += slength;
2828          }          }
2829    
2830        while (eptr >= pp)        while (eptr >= pp)
2831          {          {
2832          RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);          RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2833          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2834          eptr -= length;          eptr -= length;
2835          }          }
2836        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2837        }        }
2838      /* Control never gets here */      /* Control never gets here */
2839    
# Line 2545  for (;;) Line 2851  for (;;)
2851      case OP_NCLASS:      case OP_NCLASS:
2852      case OP_CLASS:      case OP_CLASS:
2853        {        {
2854          /* The data variable is saved across frames, so the byte map needs to
2855          be stored there. */
2856    #define BYTE_MAP ((pcre_uint8 *)data)
2857        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2858        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2859    
2860        switch (*ecode)        switch (*ecode)
2861          {          {
# Line 2556  for (;;) Line 2865  for (;;)
2865          case OP_CRMINPLUS:          case OP_CRMINPLUS:
2866          case OP_CRQUERY:          case OP_CRQUERY:
2867          case OP_CRMINQUERY:          case OP_CRMINQUERY:
2868            case OP_CRPOSSTAR:
2869            case OP_CRPOSPLUS:
2870            case OP_CRPOSQUERY:
2871          c = *ecode++ - OP_CRSTAR;          c = *ecode++ - OP_CRSTAR;
2872          minimize = (c & 1) != 0;          if (c < OP_CRPOSSTAR - OP_CRSTAR) minimize = (c & 1) != 0;
2873            else possessive = TRUE;
2874          min = rep_min[c];                 /* Pick up values from tables; */          min = rep_min[c];                 /* Pick up values from tables; */
2875          max = rep_max[c];                 /* zero for max => infinity */          max = rep_max[c];                 /* zero for max => infinity */
2876          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
# Line 2565  for (;;) Line 2878  for (;;)
2878    
2879          case OP_CRRANGE:          case OP_CRRANGE:
2880          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2881            case OP_CRPOSRANGE:
2882          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2883            possessive = (*ecode == OP_CRPOSRANGE);
2884          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2885          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2886          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2887          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2888          break;          break;
2889    
2890          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2579  for (;;) Line 2894  for (;;)
2894    
2895        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2896    
2897  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2898        /* UTF-8 mode */        if (utf)
       if (utf8)  
2899          {          {
2900          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2901            {            {
2902            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
2903              {              {
2904              SCHECK_PARTIAL();              SCHECK_PARTIAL();
2905              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
2906              }              }
2907            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2908            if (c > 255)            if (c > 255)
2909              {              {
2910              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2911              }              }
2912            else            else
2913              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  
             }  
2914            }            }
2915          }          }
2916        else        else
2917  #endif  #endif
2918        /* Not UTF-8 mode */        /* Not UTF mode */
2919          {          {
2920          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2921            {            {
2922            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
2923              {              {
2924              SCHECK_PARTIAL();              SCHECK_PARTIAL();
2925              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
2926              }              }
2927            c = *eptr++;            c = *eptr++;
2928            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2929              if (c > 255)
2930                {
2931                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2932                }
2933              else
2934    #endif
2935                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2936            }            }
2937          }          }
2938    
# Line 2627  for (;;) Line 2946  for (;;)
2946    
2947        if (minimize)        if (minimize)
2948          {          {
2949  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2950          /* UTF-8 mode */          if (utf)
         if (utf8)  
2951            {            {
2952            for (fi = min;; fi++)            for (fi = min;; fi++)
2953              {              {
2954              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2955              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2956              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2957              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
2958                {                {
2959                SCHECK_PARTIAL();                SCHECK_PARTIAL();
2960                MRRETURN(MATCH_NOMATCH);                RRETURN(MATCH_NOMATCH);
2961                }                }
2962              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2963              if (c > 255)              if (c > 255)
2964                {                {
2965                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2966                }                }
2967              else              else
2968                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  
               }  
2969              }              }
2970            }            }
2971          else          else
2972  #endif  #endif
2973          /* Not UTF-8 mode */          /* Not UTF mode */
2974            {            {
2975            for (fi = min;; fi++)            for (fi = min;; fi++)
2976              {              {
2977              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2978              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2979              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2980              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
2981                {                {
2982                SCHECK_PARTIAL();                SCHECK_PARTIAL();
2983                MRRETURN(MATCH_NOMATCH);                RRETURN(MATCH_NOMATCH);
2984                }                }
2985              c = *eptr++;              c = *eptr++;
2986              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2987                if (c > 255)
2988                  {
2989                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2990                  }
2991                else
2992    #endif
2993                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2994              }              }
2995            }            }
2996          /* Control never gets here */          /* Control never gets here */
# Line 2679  for (;;) Line 3002  for (;;)
3002          {          {
3003          pp = eptr;          pp = eptr;
3004    
3005  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3006          /* UTF-8 mode */          if (utf)
         if (utf8)  
3007            {            {
3008            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3009              {              {
# Line 2697  for (;;) Line 3019  for (;;)
3019                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3020                }                }
3021              else              else
3022                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3023              eptr += len;              eptr += len;
3024              }              }
3025    
3026              if (possessive) continue;    /* No backtracking */
3027    
3028            for (;;)            for (;;)
3029              {              {
3030              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
# Line 2712  for (;;) Line 3035  for (;;)
3035            }            }
3036          else          else
3037  #endif  #endif
3038            /* Not UTF-8 mode */            /* Not UTF mode */
3039            {            {
3040            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3041              {              {
# Line 2722  for (;;) Line 3045  for (;;)
3045                break;                break;
3046                }                }
3047              c = *eptr;              c = *eptr;
3048              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3049                if (c > 255)
3050                  {
3051                  if (op == OP_CLASS) break;
3052                  }
3053                else
3054    #endif
3055                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3056              eptr++;              eptr++;
3057              }              }
3058    
3059              if (possessive) continue;    /* No backtracking */
3060    
3061            while (eptr >= pp)            while (eptr >= pp)
3062              {              {
3063              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
# Line 2733  for (;;) Line 3066  for (;;)
3066              }              }
3067            }            }
3068    
3069          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3070          }          }
3071    #undef BYTE_MAP
3072        }        }
3073      /* Control never gets here */      /* Control never gets here */
3074    
3075    
3076      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. In the 8-bit library, this opcode is
3077      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8      encountered only when UTF-8 mode mode is supported. In the 16-bit and
3078      mode, because Unicode properties are supported in non-UTF-8 mode. */      32-bit libraries, codepoints greater than 255 may be encountered even when
3079        UTF is not supported. */
3080    
3081  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3082      case OP_XCLASS:      case OP_XCLASS:
3083        {        {
3084        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2757  for (;;) Line 3092  for (;;)
3092          case OP_CRMINPLUS:          case OP_CRMINPLUS:
3093          case OP_CRQUERY:          case OP_CRQUERY:
3094          case OP_CRMINQUERY:          case OP_CRMINQUERY:
3095            case OP_CRPOSSTAR:
3096            case OP_CRPOSPLUS:
3097            case OP_CRPOSQUERY:
3098          c = *ecode++ - OP_CRSTAR;          c = *ecode++ - OP_CRSTAR;
3099          minimize = (c & 1) != 0;          if (c < OP_CRPOSSTAR - OP_CRSTAR) minimize = (c & 1) != 0;
3100            else possessive = TRUE;
3101          min = rep_min[c];                 /* Pick up values from tables; */          min = rep_min[c];                 /* Pick up values from tables; */
3102          max = rep_max[c];                 /* zero for max => infinity */          max = rep_max[c];                 /* zero for max => infinity */
3103          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
# Line 2766  for (;;) Line 3105  for (;;)
3105    
3106          case OP_CRRANGE:          case OP_CRRANGE:
3107          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3108            case OP_CRPOSRANGE:
3109          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3110            possessive = (*ecode == OP_CRPOSRANGE);
3111          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3112          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3113          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3114          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3115          break;          break;
3116    
3117          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2785  for (;;) Line 3126  for (;;)
3126          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
3127            {            {
3128            SCHECK_PARTIAL();            SCHECK_PARTIAL();
3129            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3130            }            }
3131          GETCHARINCTEST(c, eptr);          GETCHARINCTEST(c, eptr);
3132          if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);          if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3133          }          }
3134    
3135        /* 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 2805  for (;;) Line 3146  for (;;)
3146            {            {
3147            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3148            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3149            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3150            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3151              {              {
3152              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3153              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3154              }              }
3155            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3156            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3157            }            }
3158          /* Control never gets here */          /* Control never gets here */
3159          }          }
# Line 2830  for (;;) Line 3171  for (;;)
3171              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3172              break;              break;
3173              }              }
3174    #ifdef SUPPORT_UTF
3175            GETCHARLENTEST(c, eptr, len);            GETCHARLENTEST(c, eptr, len);
3176            if (!_pcre_xclass(c, data)) break;  #else
3177              c = *eptr;
3178    #endif
3179              if (!PRIV(xclass)(c, data, utf)) break;
3180            eptr += len;            eptr += len;
3181            }            }
3182    
3183            if (possessive) continue;    /* No backtracking */
3184    
3185          for(;;)          for(;;)
3186            {            {
3187            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3188            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3189            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3190            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3191              if (utf) BACKCHAR(eptr);
3192    #endif
3193            }            }
3194          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3195          }          }
3196    
3197        /* Control never gets here */        /* Control never gets here */
# Line 2851  for (;;) Line 3201  for (;;)
3201      /* Match a single character, casefully */      /* Match a single character, casefully */
3202    
3203      case OP_CHAR:      case OP_CHAR:
3204  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3205      if (utf8)      if (utf)
3206        {        {
3207        length = 1;        length = 1;
3208        ecode++;        ecode++;
# Line 2860  for (;;) Line 3210  for (;;)
3210        if (length > md->end_subject - eptr)        if (length > md->end_subject - eptr)
3211          {          {
3212          CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */          CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3213          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3214          }          }
3215        while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3216        }        }
3217      else      else
3218  #endif  #endif
3219        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3220        {        {
3221        if (md->end_subject - eptr < 1)        if (md->end_subject - eptr < 1)
3222          {          {
3223          SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */          SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3224          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3225          }          }
3226        if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3227        ecode += 2;        ecode += 2;
3228        }        }
3229      break;      break;
3230    
3231      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3232        subject, give up immediately. */
3233    
3234      case OP_CHARI:      case OP_CHARI:
3235  #ifdef SUPPORT_UTF8      if (eptr >= md->end_subject)
3236      if (utf8)        {
3237          SCHECK_PARTIAL();
3238          RRETURN(MATCH_NOMATCH);
3239          }
3240    
3241    #ifdef SUPPORT_UTF
3242        if (utf)
3243        {        {
3244        length = 1;        length = 1;
3245        ecode++;        ecode++;
3246        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3247    
       if (length > md->end_subject - eptr)  
         {  
         CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */  
         MRRETURN(MATCH_NOMATCH);  
         }  
   
3248        /* 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
3249        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3250          fast lookup table. We know that there is at least one byte left in the
3251          subject. */
3252    
3253        if (fc < 128)        if (fc < 128)
3254          {          {
3255          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3256            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3257            ecode++;
3258            eptr++;
3259          }          }
3260    
3261        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3262          use the value of "length" to check for sufficient bytes left, because the
3263          other case of the character may have more or fewer bytes.  */
3264    
3265        else        else
3266          {          {
3267          unsigned int dc;          pcre_uint32 dc;
3268          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3269          ecode += length;          ecode += length;
3270    
# Line 2919  for (;;) Line 3276  for (;;)
3276  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3277            if (dc != UCD_OTHERCASE(fc))            if (dc != UCD_OTHERCASE(fc))
3278  #endif  #endif
3279              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3280            }            }
3281          }          }
3282        }        }
3283      else      else
3284  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3285    
3286      /* Non-UTF-8 mode */      /* Not UTF mode */
3287        {        {
3288        if (md->end_subject - eptr < 1)        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3289          {            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3290          SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */        eptr++;
         MRRETURN(MATCH_NOMATCH);  
         }  
       if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);  
3291        ecode += 2;        ecode += 2;
3292        }        }
3293      break;      break;
# Line 2943  for (;;) Line 3297  for (;;)
3297      case OP_EXACT:      case OP_EXACT:
3298      case OP_EXACTI:      case OP_EXACTI:
3299      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3300      ecode += 3;      ecode += 1 + IMM2_SIZE;
3301      goto REPEATCHAR;      goto REPEATCHAR;
3302    
3303      case OP_POSUPTO:      case OP_POSUPTO:
# Line 2958  for (;;) Line 3312  for (;;)
3312      min = 0;      min = 0;
3313      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3314      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3315      ecode += 3;      ecode += 1 + IMM2_SIZE;
3316      goto REPEATCHAR;      goto REPEATCHAR;
3317    
3318      case OP_POSSTAR:      case OP_POSSTAR:
# Line 3003  for (;;) Line 3357  for (;;)
3357      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3358      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3359    
3360      /* Common code for all repeated single-character matches. */      /* Common code for all repeated single-character matches. We first check
3361        for the minimum number of characters. If the minimum equals the maximum, we
3362        are done. Otherwise, if minimizing, check the rest of the pattern for a
3363        match; if there isn't one, advance up to the maximum, one character at a
3364        time.
3365    
3366        If maximizing, advance up to the maximum number of matching characters,
3367        until eptr is past the end of the maximum run. If possessive, we are
3368        then done (no backing up). Otherwise, match at this position; anything
3369        other than no match is immediately returned. For nomatch, back up one
3370        character, unless we are matching \R and the last thing matched was
3371        \r\n, in which case, back up two bytes. When we reach the first optional
3372        character position, we can save stack by doing a tail recurse.
3373    
3374        The various UTF/non-UTF and caseful/caseless cases are handled separately,
3375        for speed. */
3376    
3377      REPEATCHAR:      REPEATCHAR:
3378  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3379      if (utf8)      if (utf)
3380        {        {
3381        length = 1;        length = 1;
3382        charptr = ecode;        charptr = ecode;
# Line 3020  for (;;) Line 3389  for (;;)
3389        if (length > 1)        if (length > 1)
3390          {          {
3391  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3392          unsigned int othercase;          pcre_uint32 othercase;
3393          if (op >= OP_STARI &&     /* Caseless */          if (op >= OP_STARI &&     /* Caseless */
3394              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3395            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3396          else oclength = 0;          else oclength = 0;
3397  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3398    
3399          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3400            {            {
3401            if (eptr <= md->end_subject - length &&            if (eptr <= md->end_subject - length &&
3402              memcmp(eptr, charptr, length) == 0) eptr += length;              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3403  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3404            else if (oclength > 0 &&            else if (oclength > 0 &&
3405                     eptr <= md->end_subject - oclength &&                     eptr <= md->end_subject - oclength &&
3406                     memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                     memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3407  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3408            else            else
3409              {              {
3410              CHECK_PARTIAL();              CHECK_PARTIAL();
3411              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3412              }              }
3413            }            }
3414    
# Line 3051  for (;;) Line 3420  for (;;)
3420              {              {
3421              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3422              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3423              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3424              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
3425                memcmp(eptr, charptr, length) == 0) eptr += length;                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3426  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3427              else if (oclength > 0 &&              else if (oclength > 0 &&
3428                       eptr <= md->end_subject - oclength &&                       eptr <= md->end_subject - oclength &&
3429                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3430  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3431              else              else
3432                {                {
3433                CHECK_PARTIAL();                CHECK_PARTIAL();
3434                MRRETURN(MATCH_NOMATCH);                RRETURN(MATCH_NOMATCH);
3435                }                }
3436              }              }
3437            /* Control never gets here */            /* Control never gets here */
# Line 3074  for (;;) Line 3443  for (;;)
3443            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3444              {              {
3445              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
3446                  memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3447  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3448              else if (oclength > 0 &&              else if (oclength > 0 &&
3449                       eptr <= md->end_subject - oclength &&                       eptr <= md->end_subject - oclength &&
3450                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3451  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3452              else              else
3453                {                {
# Line 3087  for (;;) Line 3456  for (;;)
3456                }                }
3457              }              }
3458    
3459            if (possessive) continue;            if (possessive) continue;    /* No backtracking */
   
3460            for(;;)            for(;;)
3461              {              {
3462                if (eptr == pp) goto TAIL_RECURSE;
3463              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3464              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
             if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }  
3465  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3466              eptr--;              eptr--;
3467              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 3110  for (;;) Line 3478  for (;;)
3478        value of fc will always be < 128. */        value of fc will always be < 128. */
3479        }        }
3480      else      else
3481  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3482          /* When not in UTF-8 mode, load a single-byte character. */
3483      /* When not in UTF-8 mode, load a single-byte character. */        fc = *ecode++;
3484    
3485      fc = *ecode++;      /* The value of fc at this point is always one character, though we may
3486        or may not be in UTF mode. The code is duplicated for the caseless and
     /* The value of fc at this point is always less than 256, though we may or  
     may not be in UTF-8 mode. The code is duplicated for the caseless and  
3487      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3488      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3489      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 3126  for (;;) Line 3492  for (;;)
3492      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3493    
3494      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3495        max, eptr));        max, (char *)eptr));
3496    
3497      if (op >= OP_STARI)  /* Caseless */      if (op >= OP_STARI)  /* Caseless */
3498        {        {
3499        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3500          /* fc must be < 128 if UTF is enabled. */
3501          foc = md->fcc[fc];
3502    #else
3503    #ifdef SUPPORT_UTF
3504    #ifdef SUPPORT_UCP
3505          if (utf && fc > 127)
3506            foc = UCD_OTHERCASE(fc);
3507    #else
3508          if (utf && fc > 127)
3509            foc = fc;
3510    #endif /* SUPPORT_UCP */
3511          else
3512    #endif /* SUPPORT_UTF */
3513            foc = TABLE_GET(fc, md->fcc, fc);
3514    #endif /* COMPILE_PCRE8 */
3515    
3516        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3517          {          {
3518            pcre_uint32 cc;                 /* Faster than pcre_uchar */
3519          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
3520            {            {
3521            SCHECK_PARTIAL();            SCHECK_PARTIAL();
3522            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3523            }            }
3524          if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);          cc = RAWUCHARTEST(eptr);
3525            if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3526            eptr++;
3527          }          }
3528        if (min == max) continue;        if (min == max) continue;
3529        if (minimize)        if (minimize)
3530          {          {
3531          for (fi = min;; fi++)          for (fi = min;; fi++)
3532            {            {
3533              pcre_uint32 cc;               /* Faster than pcre_uchar */
3534            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3535            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3536            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3537            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3538              {              {
3539              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3540              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3541              }              }
3542            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);            cc = RAWUCHARTEST(eptr);
3543              if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3544              eptr++;
3545            }            }
3546          /* Control never gets here */          /* Control never gets here */
3547          }          }
# Line 3162  for (;;) Line 3550  for (;;)
3550          pp = eptr;          pp = eptr;
3551          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3552            {            {
3553              pcre_uint32 cc;               /* Faster than pcre_uchar */
3554            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3555              {              {
3556              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3557              break;              break;
3558              }              }
3559            if (fc != md->lcc[*eptr]) break;            cc = RAWUCHARTEST(eptr);
3560              if (fc != cc && foc != cc) break;
3561            eptr++;            eptr++;
3562            }            }
3563            if (possessive) continue;       /* No backtracking */
3564          if (possessive) continue;          for (;;)
   
         while (eptr >= pp)  
3565            {            {
3566              if (eptr == pp) goto TAIL_RECURSE;
3567            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3568            eptr--;            eptr--;
3569            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3570            }            }
3571          MRRETURN(MATCH_NOMATCH);          /* Control never gets here */
3572          }          }
       /* Control never gets here */  
3573        }        }
3574    
3575      /* Caseful comparisons (includes all multi-byte characters) */      /* Caseful comparisons (includes all multi-byte characters) */
# Line 3193  for (;;) Line 3581  for (;;)
3581          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
3582            {            {
3583            SCHECK_PARTIAL();            SCHECK_PARTIAL();
3584            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3585            }            }
3586          if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);          if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3587          }          }
3588    
3589        if (min == max) continue;        if (min == max) continue;
# Line 3206  for (;;) Line 3594  for (;;)
3594            {            {
3595            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3596            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3597            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3598            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3599              {              {
3600              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3601              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3602              }              }
3603            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);            if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3604            }            }
3605          /* Control never gets here */          /* Control never gets here */
3606          }          }
# Line 3226  for (;;) Line 3614  for (;;)
3614              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3615              break;              break;
3616              }              }
3617            if (fc != *eptr) break;            if (fc != RAWUCHARTEST(eptr)) break;
3618            eptr++;            eptr++;
3619            }            }
3620          if (possessive) continue;          if (possessive) continue;    /* No backtracking */
3621            for (;;)
         while (eptr >= pp)  
3622            {            {
3623              if (eptr == pp) goto TAIL_RECURSE;
3624            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3625            eptr--;            eptr--;
3626            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3627            }            }
3628          MRRETURN(MATCH_NOMATCH);          /* Control never gets here */
3629          }          }
3630        }        }
3631      /* Control never gets here */      /* Control never gets here */
# Line 3246  for (;;) Line 3634  for (;;)
3634      checking can be multibyte. */      checking can be multibyte. */
3635    
3636      case OP_NOT:      case OP_NOT:
3637      case OP_NOTI:      case OP_NOTI:
3638      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3639        {        {
3640        SCHECK_PARTIAL();        SCHECK_PARTIAL();
3641        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
3642        }        }
3643      ecode++;  #ifdef SUPPORT_UTF
3644      GETCHARINCTEST(c, eptr);      if (utf)
     if (op == OP_NOTI)         /* The caseless case */  
3645        {        {
3646  #ifdef SUPPORT_UTF8        register pcre_uint32 ch, och;
3647        if (c < 256)  
3648  #endif        ecode++;
3649        c = md->lcc[c];        GETCHARINC(ch, ecode);
3650        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);        GETCHARINC(c, eptr);
3651    
3652          if (op == OP_NOT)
3653            {
3654            if (ch == c) RRETURN(MATCH_NOMATCH);
3655            }
3656          else
3657            {
3658    #ifdef SUPPORT_UCP
3659            if (ch > 127)
3660              och = UCD_OTHERCASE(ch);
3661    #else
3662            if (ch > 127)
3663              och = ch;
3664    #endif /* SUPPORT_UCP */
3665            else
3666              och = TABLE_GET(ch, md->fcc, ch);
3667            if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3668            }
3669        }        }
3670      else    /* Caseful */      else
3671    #endif
3672        {        {
3673        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);        register pcre_uint32 ch = ecode[1];
3674          c = *eptr++;
3675          if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3676            RRETURN(MATCH_NOMATCH);
3677          ecode += 2;
3678        }        }
3679      break;      break;
3680    
# Line 3278  for (;;) Line 3688  for (;;)
3688      case OP_NOTEXACT:      case OP_NOTEXACT:
3689      case OP_NOTEXACTI:      case OP_NOTEXACTI:
3690      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3691      ecode += 3;      ecode += 1 + IMM2_SIZE;
3692      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3693    
3694      case OP_NOTUPTO:      case OP_NOTUPTO:
# Line 3288  for (;;) Line 3698  for (;;)
3698      min = 0;      min = 0;
3699      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3700      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3701      ecode += 3;      ecode += 1 + IMM2_SIZE;
3702      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3703    
3704      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
# Line 3320  for (;;) Line 3730  for (;;)
3730      possessive = TRUE;      possessive = TRUE;
3731      min = 0;      min = 0;
3732      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3733      ecode += 3;      ecode += 1 + IMM2_SIZE;
3734      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3735    
3736      case OP_NOTSTAR:      case OP_NOTSTAR:
# Line 3344  for (;;) Line 3754  for (;;)
3754      /* Common code for all repeated single-byte matches. */      /* Common code for all repeated single-byte matches. */
3755    
3756      REPEATNOTCHAR:      REPEATNOTCHAR:
3757      fc = *ecode++;      GETCHARINCTEST(fc, ecode);
3758    
3759      /* The code is duplicated for the caseless and caseful cases, for speed,      /* The code is duplicated for the caseless and caseful cases, for speed,
3760      since matching characters is likely to be quite common. First, ensure the      since matching characters is likely to be quite common. First, ensure the
# Line 3355  for (;;) Line 3765  for (;;)
3765      characters and work backwards. */      characters and work backwards. */
3766    
3767      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3768        max, eptr));        max, (char *)eptr));
3769    
3770      if (op >= OP_NOTSTARI)     /* Caseless */      if (op >= OP_NOTSTARI)     /* Caseless */
3771        {        {
3772        fc = md->lcc[fc];  #ifdef SUPPORT_UTF
3773    #ifdef SUPPORT_UCP
3774          if (utf && fc > 127)
3775            foc = UCD_OTHERCASE(fc);
3776    #else
3777          if (utf && fc > 127)
3778            foc = fc;
3779    #endif /* SUPPORT_UCP */
3780          else
3781    #endif /* SUPPORT_UTF */
3782            foc = TABLE_GET(fc, md->fcc, fc);
3783    
3784  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3785        /* UTF-8 mode */        if (utf)
       if (utf8)  
3786          {          {
3787          register unsigned int d;          register pcre_uint32 d;
3788          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3789            {            {
3790            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3791              {              {
3792              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3793              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3794              }              }
3795            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3796            if (d < 256) d = md->lcc[d];            if (fc == d || (unsigned int)foc == d) RRETURN(MATCH_NOMATCH);
           if (fc == d) MRRETURN(MATCH_NOMATCH);  
3797            }            }
3798          }          }
3799        else        else