/[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 366 by ph10, Mon Jul 14 15:45:32 2008 UTC revision 1251 by ph10, Wed Feb 20 17:42:03 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) 1997-2008 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  /* Flag bits for the match() function */  /* The md->capture_last field uses the lower 16 bits for the last captured
60    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. */
76    
77  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
78  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
79    
80  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
81  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 71  defined PCRE_ERROR_xxx codes, which are Line 86  defined PCRE_ERROR_xxx codes, which are
86  /* Special internal returns from the match() function. Make them sufficiently  /* Special internal returns from the match() function. Make them sufficiently
87  negative to avoid the external error codes. */  negative to avoid the external error codes. */
88    
89  #define MATCH_COMMIT       (-999)  #define MATCH_ACCEPT       (-999)
90  #define MATCH_PRUNE        (-998)  #define MATCH_COMMIT       (-998)
91  #define MATCH_SKIP         (-997)  #define MATCH_KETRPOS      (-997)
92  #define MATCH_THEN         (-996)  #define MATCH_ONCE         (-996)
93    #define MATCH_PRUNE        (-995)
94    #define MATCH_SKIP         (-994)
95    #define MATCH_SKIP_ARG     (-993)
96    #define MATCH_THEN         (-992)
97    
98  /* 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.
99  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 87  because the offset vector is always a mu Line 106  because the offset vector is always a mu
106  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
107  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
108    
109    #ifdef PCRE_DEBUG
   
 #ifdef DEBUG  
110  /*************************************************  /*************************************************
111  *        Debugging function to print chars       *  *        Debugging function to print chars       *
112  *************************************************/  *************************************************/
# Line 107  Returns:     nothing Line 124  Returns:     nothing
124  */  */
125    
126  static void  static void
127  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)
128  {  {
129  unsigned int c;  pcre_uint32 c;
130    BOOL utf = md->utf;
131  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;
132  while (length-- > 0)  while (length-- > 0)
133    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);
134  }  }
135  #endif  #endif
136    
# Line 122  while (length-- > 0) Line 140  while (length-- > 0)
140  *          Match a back-reference                *  *          Match a back-reference                *
141  *************************************************/  *************************************************/
142    
143  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
144  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
145    the length passed is zero. Note that in caseless UTF-8 mode, the number of
146    subject bytes matched may be different to the number of reference bytes.
147    
148  Arguments:  Arguments:
149    offset      index into the offset vector    offset      index into the offset vector
150    eptr        points into the subject    eptr        pointer into the subject
151    length      length to be matched    length      length of reference to be matched (number of bytes)
152    md          points to match data block    md          points to match data block
153    ims         the ims flags    caseless    TRUE if caseless
154    
155  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
156                  -1 no match
157                  -2 partial match; always given if at end subject
158  */  */
159    
160  static BOOL  static int
161  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,
162    unsigned long int ims)    BOOL caseless)
163  {  {
164  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
165    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
166    #ifdef SUPPORT_UTF
167    BOOL utf = md->utf;
168    #endif
169    
170  #ifdef DEBUG  #ifdef PCRE_DEBUG
171  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
172    printf("matching subject <null>");    printf("matching subject <null>");
173  else  else
# Line 154  pchars(p, length, FALSE, md); Line 180  pchars(p, length, FALSE, md);
180  printf("\n");  printf("\n");
181  #endif  #endif
182    
183  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
184    case the length is passed as zero). */
185    
186  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
187    
188  /* Separate the caseless case for speed. In UTF-8 mode we can only do this  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
189  properly if Unicode properties are supported. Otherwise, we can check only  properly if Unicode properties are supported. Otherwise, we can check only
190  ASCII characters. */  ASCII characters. */
191    
192  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
193    {    {
194  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
195  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
196    if (md->utf8)    if (utf)
197      {      {
198      USPTR endptr = eptr + length;      /* Match characters up to the end of the reference. NOTE: the number of
199      while (eptr < endptr)      data units matched may differ, because in UTF-8 there are some characters
200        {      whose upper and lower case versions code have different numbers of bytes.
201        int c, d;      For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
202        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
203        sequence of two of the latter. It is important, therefore, to check the
204        length along the reference, not along the subject (earlier code did this
205        wrong). */
206    
207        PCRE_PUCHAR endptr = p + length;
208        while (p < endptr)
209          {
210          pcre_uint32 c, d;
211          const ucd_record *ur;
212          if (eptr >= md->end_subject) return -2;   /* Partial match */
213        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
214        GETCHARINC(d, p);        GETCHARINC(d, p);
215        if (c != d && c != UCD_OTHERCASE(d)) return FALSE;        ur = GET_UCD(d);
216          if (c != d && c != d + ur->other_case)
217            {
218            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
219            for (;;)
220              {
221              if (c < *pp) return -1;
222              if (c == *pp++) break;
223              }
224            }
225        }        }
226      }      }
227    else    else
# Line 183  if ((ims & PCRE_CASELESS) != 0) Line 230  if ((ims & PCRE_CASELESS) != 0)
230    
231    /* 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
232    is no UCP support. */    is no UCP support. */
233        {
234    while (length-- > 0)      while (length-- > 0)
235      { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }        {
236          pcre_uchar cc, cp;
237          if (eptr >= md->end_subject) return -2;   /* Partial match */
238          cc = RAWUCHARTEST(eptr);
239          cp = RAWUCHARTEST(p);
240          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
241          p++;
242          eptr++;
243          }
244        }
245    }    }
246    
247  /* 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
248  are in UTF-8 mode. */  are in UTF-8 mode. */
249    
250  else  else
251    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
252      while (length-- > 0)
253        {
254        if (eptr >= md->end_subject) return -2;   /* Partial match */
255        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
256        }
257      }
258    
259  return TRUE;  return (int)(eptr - eptr_start);
260  }  }
261    
262    
# Line 245  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 307  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
307         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
308         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
309         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
310         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
311           RM61,  RM62, RM63, RM64, RM65, RM66, RM67 };
312    
313  /* 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
314  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
315  actuall used in this definition. */  actually used in this definition. */
316    
317  #ifndef NO_RECURSE  #ifndef NO_RECURSE
318  #define REGISTER register  #define REGISTER register
319    
320  #ifdef DEBUG  #ifdef PCRE_DEBUG
321  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
322    { \    { \
323    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
324    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
325    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
326    }    }
327  #define RRETURN(ra) \  #define RRETURN(ra) \
328    { \    { \
329    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
330    return ra; \    return ra; \
331    }    }
332  #else  #else
333  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
334    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
335  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
336  #endif  #endif
337    
# Line 281  argument of match(), which never changes Line 344  argument of match(), which never changes
344    
345  #define REGISTER  #define REGISTER
346    
347  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
348    {\    {\
349    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
350    frame->Xwhere = rw; \    if (newframe == NULL)\
351        {\
352        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
353        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
354        newframe->Xnextframe = NULL;\
355        frame->Xnextframe = newframe;\
356        }\
357      frame->Xwhere = rw;\
358    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
359    newframe->Xecode = rb;\    newframe->Xecode = rb;\
360    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
361    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
362    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
363    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
364    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
365    frame = newframe;\    frame = newframe;\
# Line 303  argument of match(), which never changes Line 371  argument of match(), which never changes
371    
372  #define RRETURN(ra)\  #define RRETURN(ra)\
373    {\    {\
374    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
375    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
376    if (frame != NULL)\    if (frame != NULL)\
377      {\      {\
378      rrc = ra;\      rrc = ra;\
# Line 319  argument of match(), which never changes Line 386  argument of match(), which never changes
386    
387  typedef struct heapframe {  typedef struct heapframe {
388    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
389      struct heapframe *Xnextframe;
390    
391    /* Function arguments that may change */    /* Function arguments that may change */
392    
393    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
394    const uschar *Xecode;    const pcre_uchar *Xecode;
395    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
396    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
397    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
398    unsigned int Xrdepth;    unsigned int Xrdepth;
399    
400    /* Function local variables */    /* Function local variables */
401    
402    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
403    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
404    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
405    const uschar *Xnext;  #endif
406    const uschar *Xpp;    PCRE_PUCHAR Xdata;
407    const uschar *Xprev;    PCRE_PUCHAR Xnext;
408    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
409      PCRE_PUCHAR Xprev;
410      PCRE_PUCHAR Xsaved_eptr;
411    
412    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
413    
# Line 347  typedef struct heapframe { Line 415  typedef struct heapframe {
415    BOOL Xcondition;    BOOL Xcondition;
416    BOOL Xprev_is_word;    BOOL Xprev_is_word;
417    
   unsigned long int Xoriginal_ims;  
   
418  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
419    int Xprop_type;    int Xprop_type;
420    int Xprop_value;    unsigned int Xprop_value;
421    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
422    int Xoclength;    int Xoclength;
423    uschar Xocchars[8];    pcre_uchar Xocchars[6];
424  #endif  #endif
425    
426      int Xcodelink;
427    int Xctype;    int Xctype;
428    unsigned int Xfc;    unsigned int Xfc;
429    int Xfi;    int Xfi;
430    int Xlength;    int Xlength;
431    int Xmax;    int Xmax;
432    int Xmin;    int Xmin;
433    int Xnumber;    unsigned int Xnumber;
434    int Xoffset;    int Xoffset;
435    int Xop;    unsigned int Xop;
436    int Xsave_capture_last;    pcre_int32 Xsave_capture_last;
437    int Xsave_offset1, Xsave_offset2, Xsave_offset3;    int Xsave_offset1, Xsave_offset2, Xsave_offset3;
438    int Xstacksave[REC_STACK_SAVE_MAX];    int Xstacksave[REC_STACK_SAVE_MAX];
439    
# Line 395  typedef struct heapframe { Line 459  typedef struct heapframe {
459    
460  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
461  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
462  same response.  same response. */
463    
464  Performance note: It might be tempting to extract commonly used fields from the  /* These macros pack up tests that are used for partial matching, and which
465  md structure (e.g. utf8, end_subject) into individual variables to improve  appear several times in the code. We set the "hit end" flag if the pointer is
466    at the end of the subject and also past the start of the subject (i.e.
467    something has been matched). For hard partial matching, we then return
468    immediately. The second one is used when we already know we are past the end of
469    the subject. */
470    
471    #define CHECK_PARTIAL()\
472      if (md->partial != 0 && eptr >= md->end_subject && \
473          eptr > md->start_used_ptr) \
474        { \
475        md->hitend = TRUE; \
476        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
477        }
478    
479    #define SCHECK_PARTIAL()\
480      if (md->partial != 0 && eptr > md->start_used_ptr) \
481        { \
482        md->hitend = TRUE; \
483        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
484        }
485    
486    
487    /* Performance note: It might be tempting to extract commonly used fields from
488    the md structure (e.g. utf, end_subject) into individual variables to improve
489  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
490  made performance worse.  made performance worse.
491    
# Line 409  Arguments: Line 496  Arguments:
496                   by encountering \K)                   by encountering \K)
497     offset_top  current top pointer     offset_top  current top pointer
498     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
499     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
500                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
501     rdepth      the recursion depth     rdepth      the recursion depth
502    
503  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
504                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
505                   a negative MATCH_xxx value for PRUNE, SKIP, etc
506                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
507                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
508  */  */
509    
510  static int  static int
511  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
512    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
513    int flags, unsigned int rdepth)    unsigned int rdepth)
514  {  {
515  /* 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,
516  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 435  so they can be ordinary variables in all Line 518  so they can be ordinary variables in all
518    
519  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
520  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
521  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
522  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
523    
524  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
525    BOOL caseless;
526    int condcode;
527    
528  /* 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
529  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
530  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
531  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
532    the top-level on the stack rather than malloc-ing them all gives a performance
533    boost in many cases where there is not much "recursion". */
534    
535  #ifdef NO_RECURSE  #ifdef NO_RECURSE
536  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
537    
538  /* Copy in the original argument variables */  /* Copy in the original argument variables */
539    
# Line 455  frame->Xeptr = eptr; Line 541  frame->Xeptr = eptr;
541  frame->Xecode = ecode;  frame->Xecode = ecode;
542  frame->Xmstart = mstart;  frame->Xmstart = mstart;
543  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
544  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
545  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
546    
547  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 470  HEAP_RECURSE: Line 554  HEAP_RECURSE:
554  #define ecode              frame->Xecode  #define ecode              frame->Xecode
555  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
556  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
557  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
558  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
559    
560  /* Ditto for the local variables */  /* Ditto for the local variables */
561    
562  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
563  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
564  #endif  #endif
565  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
566    #define codelink           frame->Xcodelink
567  #define data               frame->Xdata  #define data               frame->Xdata
568  #define next               frame->Xnext  #define next               frame->Xnext
569  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 493  HEAP_RECURSE: Line 576  HEAP_RECURSE:
576  #define condition          frame->Xcondition  #define condition          frame->Xcondition
577  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
578    
 #define original_ims       frame->Xoriginal_ims  
   
579  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
580  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
581  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
582  #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  
583  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
584  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
585  #endif  #endif
# Line 531  i, and fc and c, can be the same variabl Line 609  i, and fc and c, can be the same variabl
609  #define fi i  #define fi i
610  #define fc c  #define fc c
611    
612    /* Many of the following variables are used only in small blocks of the code.
613    My normal style of coding would have declared them within each of those blocks.
614    However, in order to accommodate the version of this code that uses an external
615    "stack" implemented on the heap, it is easier to declare them all here, so the
616    declarations can be cut out in a block. The only declarations within blocks
617    below are for variables that do not have to be preserved over a recursive call
618    to RMATCH(). */
619    
620    #ifdef SUPPORT_UTF
621    const pcre_uchar *charptr;
622    #endif
623    const pcre_uchar *callpat;
624    const pcre_uchar *data;
625    const pcre_uchar *next;
626    PCRE_PUCHAR       pp;
627    const pcre_uchar *prev;
628    PCRE_PUCHAR       saved_eptr;
629    
630  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
631  const uschar *charptr;             /* in small blocks of the code. My normal */  
632  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
633  BOOL condition;  BOOL condition;
634  BOOL prev_is_word;  BOOL prev_is_word;
635    
 unsigned long int original_ims;  
   
636  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
637  int prop_type;  int prop_type;
638  int prop_value;  unsigned int prop_value;
639  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
640  int oclength;  int oclength;
641  uschar occhars[8];  pcre_uchar occhars[6];
642  #endif  #endif
643    
644    int codelink;
645  int ctype;  int ctype;
646  int length;  int length;
647  int max;  int max;
648  int min;  int min;
649  int number;  unsigned int number;
650  int offset;  int offset;
651  int op;  unsigned int op;
652  int save_capture_last;  pcre_int32 save_capture_last;
653  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
654  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
655    
656  eptrblock newptrb;  eptrblock newptrb;
657    
658    /* There is a special fudge for calling match() in a way that causes it to
659    measure the size of its basic stack frame when the stack is being used for
660    recursion. The second argument (ecode) being NULL triggers this behaviour. It
661    cannot normally ever be NULL. The return is the negated value of the frame
662    size. */
663    
664    if (ecode == NULL)
665      {
666      if (rdepth == 0)
667        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
668      else
669        {
670        int len = (char *)&rdepth - (char *)eptr;
671        return (len > 0)? -len : len;
672        }
673      }
674  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
675    
676    /* To save space on the stack and in the heap frame, I have doubled up on some
677    of the local variables that are used only in localised parts of the code, but
678    still need to be preserved over recursive calls of match(). These macros define
679    the alternative names that are used. */
680    
681    #define allow_zero    cur_is_word
682    #define cbegroup      condition
683    #define code_offset   codelink
684    #define condassert    condition
685    #define matched_once  prev_is_word
686    #define foc           number
687    #define save_mark     data
688    
689  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
690  variables. */  variables. */
691    
# Line 594  TAIL_RECURSE: Line 705  TAIL_RECURSE:
705  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
706  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
707  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
708  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
709  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
710  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,
711  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
712    
713  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
714  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
715  #else  #else
716  utf8 = FALSE;  utf = FALSE;
717  #endif  #endif
718    
719  /* 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 611  haven't exceeded the recursive call limi Line 722  haven't exceeded the recursive call limi
722  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
723  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
724    
 original_ims = ims;    /* Save for resetting on ')' */  
   
725  /* 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
726  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
727  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
728  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
729  When match() is called in other circumstances, don't add to the chain. The  
730  match_cbegroup flag must NOT be used with tail recursion, because the memory  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
731  block that is used is on the stack, so a new one may be required for each  such remembered pointers, to be checked when we hit the closing ket, in order
732  match(). */  to break infinite loops that match no characters. When match() is called in
733    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
734    NOT be used with tail recursion, because the memory block that is used is on
735    the stack, so a new one may be required for each match(). */
736    
737  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
738    {    {
739    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
740    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
741    eptrb = &newptrb;    eptrb = &newptrb;
742      md->match_function_type = 0;
743    }    }
744    
745  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 636  for (;;) Line 749  for (;;)
749    minimize = possessive = FALSE;    minimize = possessive = FALSE;
750    op = *ecode;    op = *ecode;
751    
   /* For partial matching, remember if we ever hit the end of the subject after  
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > mstart)  
     md->hitend = TRUE;  
   
752    switch(op)    switch(op)
753      {      {
754        case OP_MARK:
755        md->nomatch_mark = ecode + 2;
756        md->mark = NULL;    /* In case previously set by assertion */
757        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
758          eptrb, RM55);
759        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
760             md->mark == NULL) md->mark = ecode + 2;
761    
762        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
763        argument, and we must check whether that argument matches this MARK's
764        argument. It is passed back in md->start_match_ptr (an overloading of that
765        variable). If it does match, we reset that variable to the current subject
766        position and return MATCH_SKIP. Otherwise, pass back the return code
767        unaltered. */
768    
769        else if (rrc == MATCH_SKIP_ARG &&
770            STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
771          {
772          md->start_match_ptr = eptr;
773          RRETURN(MATCH_SKIP);
774          }
775        RRETURN(rrc);
776    
777      case OP_FAIL:      case OP_FAIL:
778      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
779    
780      case OP_PRUNE:      /* COMMIT overrides PRUNE, SKIP, and THEN */
     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,  
       ims, eptrb, flags, RM51);  
     if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
     RRETURN(MATCH_PRUNE);  
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        ims, eptrb, flags, RM52);        eptrb, RM52);
785      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
786            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
787            rrc != MATCH_THEN)
788          RRETURN(rrc);
789      RRETURN(MATCH_COMMIT);      RRETURN(MATCH_COMMIT);
790    
791        /* PRUNE overrides THEN */
792    
793        case OP_PRUNE:
794        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
795          eptrb, RM51);
796        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
797        RRETURN(MATCH_PRUNE);
798    
799        case OP_PRUNE_ARG:
800        md->nomatch_mark = ecode + 2;
801        md->mark = NULL;    /* In case previously set by assertion */
802        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
803          eptrb, RM56);
804        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
805             md->mark == NULL) md->mark = ecode + 2;
806        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
807        RRETURN(MATCH_PRUNE);
808    
809        /* SKIP overrides PRUNE and THEN */
810    
811      case OP_SKIP:      case OP_SKIP:
812      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
813        ims, eptrb, flags, RM53);        eptrb, RM53);
814      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
815          RRETURN(rrc);
816      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
817      RRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
818    
819        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
820        nomatch_mark. There is a flag that disables this opcode when re-matching a
821        pattern that ended with a SKIP for which there was not a matching MARK. */
822    
823        case OP_SKIP_ARG:
824        if (md->ignore_skip_arg)
825          {
826          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
827          break;
828          }
829        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
830          eptrb, RM57);
831        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
832          RRETURN(rrc);
833    
834        /* Pass back the current skip name by overloading md->start_match_ptr and
835        returning the special MATCH_SKIP_ARG return code. This will either be
836        caught by a matching MARK, or get to the top, where it causes a rematch
837        with the md->ignore_skip_arg flag set. */
838    
839        md->start_match_ptr = ecode + 2;
840        RRETURN(MATCH_SKIP_ARG);
841    
842        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
843        the branch in which it occurs can be determined. Overload the start of
844        match pointer to do this. */
845    
846      case OP_THEN:      case OP_THEN:
847      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
848        ims, eptrb, flags, RM54);        eptrb, RM54);
849      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
850        md->start_match_ptr = ecode;
851      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
852    
853      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_THEN_ARG:
854      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
855      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
856      set from a previous iteration of this group, and be referred to by a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
857      reference inside the group.        md, eptrb, RM58);
858        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
859      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
860      values of the final offsets, in case they were set by a previous iteration      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
861      of the same bracket.      md->start_match_ptr = ecode;
862        RRETURN(MATCH_THEN);
863    
864        /* Handle an atomic group that does not contain any capturing parentheses.
865        This can be handled like an assertion. Prior to 8.13, all atomic groups
866        were handled this way. In 8.13, the code was changed as below for ONCE, so
867        that backups pass through the group and thereby reset captured values.
868        However, this uses a lot more stack, so in 8.20, atomic groups that do not
869        contain any captures generate OP_ONCE_NC, which can be handled in the old,
870        less stack intensive way.
871    
872        Check the alternative branches in turn - the matching won't pass the KET
873        for this kind of subpattern. If any one branch matches, we carry on as at
874        the end of a normal bracket, leaving the subject pointer, but resetting
875        the start-of-match value in case it was changed by \K. */
876    
877        case OP_ONCE_NC:
878        prev = ecode;
879        saved_eptr = eptr;
880        save_mark = md->mark;
881        do
882          {
883          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
884          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
885            {
886            mstart = md->start_match_ptr;
887            break;
888            }
889          if (rrc == MATCH_THEN)
890            {
891            next = ecode + GET(ecode,1);
892            if (md->start_match_ptr < next &&
893                (*ecode == OP_ALT || *next == OP_ALT))
894              rrc = MATCH_NOMATCH;
895            }
896    
897          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
898          ecode += GET(ecode,1);
899          md->mark = save_mark;
900          }
901        while (*ecode == OP_ALT);
902    
903        /* If hit the end of the group (which could be repeated), fail */
904    
905        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
906    
907        /* Continue as from after the group, updating the offsets high water
908        mark, since extracts may have been taken. */
909    
910        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
911    
912        offset_top = md->end_offset_top;
913        eptr = md->end_match_ptr;
914    
915        /* For a non-repeating ket, just continue at this level. This also
916        happens for a repeating ket if no characters were matched in the group.
917        This is the forcible breaking of infinite loops as implemented in Perl
918        5.005. */
919    
920        if (*ecode == OP_KET || eptr == saved_eptr)
921          {
922          ecode += 1+LINK_SIZE;
923          break;
924          }
925    
926        /* The repeating kets try the rest of the pattern or restart from the
927        preceding bracket, in the appropriate order. The second "call" of match()
928        uses tail recursion, to avoid using another stack frame. */
929    
930        if (*ecode == OP_KETRMIN)
931          {
932          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
933          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
934          ecode = prev;
935          goto TAIL_RECURSE;
936          }
937        else  /* OP_KETRMAX */
938          {
939          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
940          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
941          ecode += 1 + LINK_SIZE;
942          goto TAIL_RECURSE;
943          }
944        /* Control never gets here */
945    
946        /* Handle a capturing bracket, other than those that are possessive with an
947        unlimited repeat. If there is space in the offset vector, save the current
948        subject position in the working slot at the top of the vector. We mustn't
949        change the current values of the data slot, because they may be set from a
950        previous iteration of this group, and be referred to by a reference inside
951        the group. A failure to match might occur after the group has succeeded,
952        if something later on doesn't match. For this reason, we need to restore
953        the working value and also the values of the final offsets, in case they
954        were set by a previous iteration of the same bracket.
955    
956      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
957      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 693  for (;;) Line 962  for (;;)
962      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
963      offset = number << 1;      offset = number << 1;
964    
965  #ifdef DEBUG  #ifdef PCRE_DEBUG
966      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
967      printf("subject=");      printf("subject=");
968      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 706  for (;;) Line 975  for (;;)
975        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
976        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
977        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
978          save_mark = md->mark;
979    
980        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
981        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
982            (int)(eptr - md->start_subject);
983    
984        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
985          {          {
986          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
987            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
988          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
989            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
990    
991            /* If we backed up to a THEN, check whether it is within the current
992            branch by comparing the address of the THEN that is passed back with
993            the end of the branch. If it is within the current branch, and the
994            branch is one of two or more alternatives (it either starts or ends
995            with OP_ALT), we have reached the limit of THEN's action, so convert
996            the return code to NOMATCH, which will cause normal backtracking to
997            happen from now on. Otherwise, THEN is passed back to an outer
998            alternative. This implements Perl's treatment of parenthesized groups,
999            where a group not containing | does not affect the current alternative,
1000            that is, (X) is NOT the same as (X|(*F)). */
1001    
1002            if (rrc == MATCH_THEN)
1003              {
1004              next = ecode + GET(ecode,1);
1005              if (md->start_match_ptr < next &&
1006                  (*ecode == OP_ALT || *next == OP_ALT))
1007                rrc = MATCH_NOMATCH;
1008              }
1009    
1010            /* Anything other than NOMATCH is passed back. */
1011    
1012            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1013          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1014          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1015            md->mark = save_mark;
1016            if (*ecode != OP_ALT) break;
1017          }          }
       while (*ecode == OP_ALT);  
1018    
1019        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1020        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1021        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1022        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1023    
1024        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1025    
1026          RRETURN(rrc);
1027        }        }
1028    
1029      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 741  for (;;) Line 1037  for (;;)
1037      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1038      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1039    
1040      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1041      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion, except in the case when match_cbegroup  
     is set.*/  
1042    
1043        When we get to the final alternative within the brackets, we used to return
1044        the result of a recursive call to match() whatever happened so it was
1045        possible to reduce stack usage by turning this into a tail recursion,
1046        except in the case of a possibly empty group. However, now that there is
1047        the possiblity of (*THEN) occurring in the final alternative, this
1048        optimization is no longer always possible.
1049    
1050        We can optimize if we know there are no (*THEN)s in the pattern; at present
1051        this is the best that can be done.
1052    
1053        MATCH_ONCE is returned when the end of an atomic group is successfully
1054        reached, but subsequent matching fails. It passes back up the tree (causing
1055        captured values to be reset) until the original atomic group level is
1056        reached. This is tested by comparing md->once_target with the start of the
1057        group. At this point, the return is converted into MATCH_NOMATCH so that
1058        previous backup points can be taken. */
1059    
1060        case OP_ONCE:
1061      case OP_BRA:      case OP_BRA:
1062      case OP_SBRA:      case OP_SBRA:
1063      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1064      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1065      for (;;)      for (;;)
1066        {        {
1067        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE)
1068          {          md->match_function_type = MATCH_CBEGROUP;
         if (flags == 0)    /* Not a possibly empty group */  
           {  
           ecode += _pcre_OP_lengths[*ecode];  
           DPRINTF(("bracket 0 tail recursion\n"));  
           goto TAIL_RECURSE;  
           }  
1069    
1070          /* Possibly empty group; can't use tail recursion. */        /* If this is not a possibly empty group, and there are no (*THEN)s in
1071          the pattern, and this is the final alternative, optimize as described
1072          above. */
1073    
1074          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1075            eptrb, flags, RM48);          {
1076          RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1077            goto TAIL_RECURSE;
1078          }          }
1079    
1080        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* In all other cases, we have to make another call to match(). */
       otherwise return. */  
1081    
1082        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1083          eptrb, flags, RM2);        save_capture_last = md->capture_last;
1084        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1085            RM2);
1086    
1087          /* See comment in the code for capturing groups above about handling
1088          THEN. */
1089    
1090          if (rrc == MATCH_THEN)
1091            {
1092            next = ecode + GET(ecode,1);
1093            if (md->start_match_ptr < next &&
1094                (*ecode == OP_ALT || *next == OP_ALT))
1095              rrc = MATCH_NOMATCH;
1096            }
1097    
1098          if (rrc != MATCH_NOMATCH)
1099            {
1100            if (rrc == MATCH_ONCE)
1101              {
1102              const pcre_uchar *scode = ecode;
1103              if (*scode != OP_ONCE)           /* If not at start, find it */
1104                {
1105                while (*scode == OP_ALT) scode += GET(scode, 1);
1106                scode -= GET(scode, 1);
1107                }
1108              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1109              }
1110            RRETURN(rrc);
1111            }
1112        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1113          md->mark = save_mark;
1114          if (*ecode != OP_ALT) break;
1115          md->capture_last = save_capture_last;
1116        }        }
     /* Control never reaches here. */  
1117    
1118      /* Conditional group: compilation checked that there are no more than      RRETURN(MATCH_NOMATCH);
     two branches. If the condition is false, skipping the first branch takes us  
     past the end if there is only one branch, but that's OK because that is  
     exactly what going to the ket would do. As there is only one branch to be  
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1119    
1120      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1121      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1122      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1123        {      different. The end of these brackets will always be OP_KETRPOS, which
1124        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1125        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1126          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
1127        ecode += condition? 3 : GET(ecode, 1);  
1128        }      case OP_CBRAPOS:
1129        case OP_SCBRAPOS:
1130        allow_zero = FALSE;
1131    
1132      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      POSSESSIVE_CAPTURE:
1133        {      number = GET2(ecode, 1+LINK_SIZE);
1134        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      offset = number << 1;
       condition = offset < offset_top && md->offset_vector[offset] >= 0;  
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1135    
1136      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */  #ifdef PCRE_DEBUG
1137        printf("start possessive bracket %d\n", number);
1138        printf("subject=");
1139        pchars(eptr, 16, TRUE, md);
1140        printf("\n");
1141    #endif
1142    
1143        if (offset < md->offset_max)
1144        {        {
1145        condition = FALSE;        matched_once = FALSE;
1146        ecode += GET(ecode, 1);        code_offset = (int)(ecode - md->start_code);
       }  
1147    
1148      /* The condition is an assertion. Call match() to evaluate it - setting        save_offset1 = md->offset_vector[offset];
1149      the final argument match_condassert causes it to stop at the end of an        save_offset2 = md->offset_vector[offset+1];
1150      assertion. */        save_offset3 = md->offset_vector[md->offset_end - number];
1151          save_capture_last = md->capture_last;
1152    
1153      else        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1154        {  
1155        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        /* Each time round the loop, save the current subject position for use
1156            match_condassert, RM3);        when the group matches. For MATCH_MATCH, the group has matched, so we
1157        if (rrc == MATCH_MATCH)        restart it with a new subject starting position, remembering that we had
1158          {        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1159          condition = TRUE;        usual. If we haven't matched any alternatives in any iteration, check to
1160          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        see if a previous iteration matched. If so, the group has matched;
1161          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        continue from afterwards. Otherwise it has failed; restore the previous
1162          capture values before returning NOMATCH. */
1163    
1164          for (;;)
1165            {
1166            md->offset_vector[md->offset_end - number] =
1167              (int)(eptr - md->start_subject);
1168            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1169            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1170              eptrb, RM63);
1171            if (rrc == MATCH_KETRPOS)
1172              {
1173              offset_top = md->end_offset_top;
1174              eptr = md->end_match_ptr;
1175              ecode = md->start_code + code_offset;
1176              save_capture_last = md->capture_last;
1177              matched_once = TRUE;
1178              continue;
1179              }
1180    
1181            /* See comment in the code for capturing groups above about handling
1182            THEN. */
1183    
1184            if (rrc == MATCH_THEN)
1185              {
1186              next = ecode + GET(ecode,1);
1187              if (md->start_match_ptr < next &&
1188                  (*ecode == OP_ALT || *next == OP_ALT))
1189                rrc = MATCH_NOMATCH;
1190              }
1191    
1192            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1193            md->capture_last = save_capture_last;
1194            ecode += GET(ecode, 1);
1195            if (*ecode != OP_ALT) break;
1196          }          }
1197        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1198          if (!matched_once)
1199          {          {
1200          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1201            md->offset_vector[offset+1] = save_offset2;
1202            md->offset_vector[md->offset_end - number] = save_offset3;
1203          }          }
1204        else  
1205          if (allow_zero || matched_once)
1206          {          {
1207          condition = FALSE;          ecode += 1 + LINK_SIZE;
1208          ecode += GET(ecode, 1);          break;
1209          }          }
1210    
1211          RRETURN(MATCH_NOMATCH);
1212        }        }
1213    
1214      /* We are now at the branch that is to be obeyed. As there is only one,      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1215      we can use tail recursion to avoid using another stack frame, except when      as a non-capturing bracket. */
     match_cbegroup is required for an unlimited repeat of a possibly empty  
     group. If the second alternative doesn't exist, we can just plough on. */  
1216    
1217      if (condition || *ecode == OP_ALT)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1218        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1219    
1220        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1221    
1222        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1223        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1224    
1225        /* Non-capturing possessive bracket with unlimited repeat. We come here
1226        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1227        without the capturing complication. It is written out separately for speed
1228        and cleanliness. */
1229    
1230        case OP_BRAPOS:
1231        case OP_SBRAPOS:
1232        allow_zero = FALSE;
1233    
1234        POSSESSIVE_NON_CAPTURE:
1235        matched_once = FALSE;
1236        code_offset = (int)(ecode - md->start_code);
1237        save_capture_last = md->capture_last;
1238    
1239        for (;;)
1240        {        {
1241        ecode += 1 + LINK_SIZE;        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1242        if (op == OP_SCOND)        /* Possibly empty group */        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1243          {          eptrb, RM48);
1244          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);        if (rrc == MATCH_KETRPOS)
1245          RRETURN(rrc);          {
1246          }          offset_top = md->end_offset_top;
1247        else                       /* Group must match something */          eptr = md->end_match_ptr;
1248          {          ecode = md->start_code + code_offset;
1249          flags = 0;          matched_once = TRUE;
1250          goto TAIL_RECURSE;          continue;
1251            }
1252    
1253          /* See comment in the code for capturing groups above about handling
1254          THEN. */
1255    
1256          if (rrc == MATCH_THEN)
1257            {
1258            next = ecode + GET(ecode,1);
1259            if (md->start_match_ptr < next &&
1260                (*ecode == OP_ALT || *next == OP_ALT))
1261              rrc = MATCH_NOMATCH;
1262          }          }
1263    
1264          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1265          ecode += GET(ecode, 1);
1266          if (*ecode != OP_ALT) break;
1267          md->capture_last = save_capture_last;
1268        }        }
1269      else                         /* Condition false & no 2nd alternative */  
1270        if (matched_once || allow_zero)
1271        {        {
1272        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1273          break;
1274        }        }
1275      break;      RRETURN(MATCH_NOMATCH);
1276    
1277        /* Control never reaches here. */
1278    
1279        /* Conditional group: compilation checked that there are no more than
1280        two branches. If the condition is false, skipping the first branch takes us
1281        past the end if there is only one branch, but that's OK because that is
1282        exactly what going to the ket would do. */
1283    
1284      /* End of the pattern, either real or forced. If we are in a top-level      case OP_COND:
1285      recursion, we should restore the offsets appropriately and continue from      case OP_SCOND:
1286      after the call. */      codelink = GET(ecode, 1);
1287    
1288      case OP_ACCEPT:      /* Because of the way auto-callout works during compile, a callout item is
1289      case OP_END:      inserted between OP_COND and an assertion condition. */
1290      if (md->recursive != NULL && md->recursive->group_num == 0)  
1291        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1292        {        {
1293        recursion_info *rec = md->recursive;        if (PUBL(callout) != NULL)
1294        DPRINTF(("End of pattern in a (?0) recursion\n"));          {
1295        md->recursive = rec->prevrec;          PUBL(callout_block) cb;
1296        memmove(md->offset_vector, rec->offset_save,          cb.version          = 2;   /* Version 1 of the callout block */
1297          rec->saved_max * sizeof(int));          cb.callout_number   = ecode[LINK_SIZE+2];
1298        mstart = rec->save_start;          cb.offset_vector    = md->offset_vector;
1299        ims = original_ims;  #if defined COMPILE_PCRE8
1300        ecode = rec->after_call;          cb.subject          = (PCRE_SPTR)md->start_subject;
1301        break;  #elif defined COMPILE_PCRE16
1302            cb.subject          = (PCRE_SPTR16)md->start_subject;
1303    #elif defined COMPILE_PCRE32
1304            cb.subject          = (PCRE_SPTR32)md->start_subject;
1305    #endif
1306            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1307            cb.start_match      = (int)(mstart - md->start_subject);
1308            cb.current_position = (int)(eptr - md->start_subject);
1309            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1310            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1311            cb.capture_top      = offset_top/2;
1312            cb.capture_last     = md->capture_last & CAPLMASK;
1313            /* Internal change requires this for API compatibility. */
1314            if (cb.capture_last == 0) cb.capture_last = -1;
1315            cb.callout_data     = md->callout_data;
1316            cb.mark             = md->nomatch_mark;
1317            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1318            if (rrc < 0) RRETURN(rrc);
1319            }
1320          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1321        }        }
1322    
1323      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      condcode = ecode[LINK_SIZE+1];
     string - backtracking will then try other alternatives, if any. */  
1324    
1325      if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);      /* Now see what the actual condition is */
     md->end_match_ptr = eptr;           /* Record where we ended */  
     md->end_offset_top = offset_top;    /* and how many extracts were taken */  
     md->start_match_ptr = mstart;       /* and the start (\K can modify) */  
     RRETURN(MATCH_MATCH);  
1326    
1327      /* Change option settings */      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1328          {
1329          if (md->recursive == NULL)                /* Not recursing => FALSE */
1330            {
1331            condition = FALSE;
1332            ecode += GET(ecode, 1);
1333            }
1334          else
1335            {
1336            unsigned int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1337            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1338    
1339      case OP_OPT:          /* If the test is for recursion into a specific subpattern, and it is
1340      ims = ecode[1];          false, but the test was set up by name, scan the table to see if the
1341      ecode += 2;          name refers to any other numbers, and test them. The condition is true
1342      DPRINTF(("ims set to %02lx\n", ims));          if any one is set. */
     break;  
1343    
1344      /* Assertion brackets. Check the alternative branches in turn - the          if (!condition && condcode == OP_NRREF)
1345      matching won't pass the KET for an assertion. If any one branch matches,            {
1346              pcre_uchar *slotA = md->name_table;
1347              for (i = 0; i < md->name_count; i++)
1348                {
1349                if (GET2(slotA, 0) == recno) break;
1350                slotA += md->name_entry_size;
1351                }
1352    
1353              /* Found a name for the number - there can be only one; duplicate
1354              names for different numbers are allowed, but not vice versa. First
1355              scan down for duplicates. */
1356    
1357              if (i < md->name_count)
1358                {
1359                pcre_uchar *slotB = slotA;
1360                while (slotB > md->name_table)
1361                  {
1362                  slotB -= md->name_entry_size;
1363                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1364                    {
1365                    condition = GET2(slotB, 0) == md->recursive->group_num;
1366                    if (condition) break;
1367                    }
1368                  else break;
1369                  }
1370    
1371                /* Scan up for duplicates */
1372    
1373                if (!condition)
1374                  {
1375                  slotB = slotA;
1376                  for (i++; i < md->name_count; i++)
1377                    {
1378                    slotB += md->name_entry_size;
1379                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1380                      {
1381                      condition = GET2(slotB, 0) == md->recursive->group_num;
1382                      if (condition) break;
1383                      }
1384                    else break;
1385                    }
1386                  }
1387                }
1388              }
1389    
1390            /* Chose branch according to the condition */
1391    
1392            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1393            }
1394          }
1395    
1396        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1397          {
1398          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1399          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1400    
1401          /* If the numbered capture is unset, but the reference was by name,
1402          scan the table to see if the name refers to any other numbers, and test
1403          them. The condition is true if any one is set. This is tediously similar
1404          to the code above, but not close enough to try to amalgamate. */
1405    
1406          if (!condition && condcode == OP_NCREF)
1407            {
1408            unsigned int refno = offset >> 1;
1409            pcre_uchar *slotA = md->name_table;
1410    
1411            for (i = 0; i < md->name_count; i++)
1412              {
1413              if (GET2(slotA, 0) == refno) break;
1414              slotA += md->name_entry_size;
1415              }
1416    
1417            /* Found a name for the number - there can be only one; duplicate names
1418            for different numbers are allowed, but not vice versa. First scan down
1419            for duplicates. */
1420    
1421            if (i < md->name_count)
1422              {
1423              pcre_uchar *slotB = slotA;
1424              while (slotB > md->name_table)
1425                {
1426                slotB -= md->name_entry_size;
1427                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1428                  {
1429                  offset = GET2(slotB, 0) << 1;
1430                  condition = offset < offset_top &&
1431                    md->offset_vector[offset] >= 0;
1432                  if (condition) break;
1433                  }
1434                else break;
1435                }
1436    
1437              /* Scan up for duplicates */
1438    
1439              if (!condition)
1440                {
1441                slotB = slotA;
1442                for (i++; i < md->name_count; i++)
1443                  {
1444                  slotB += md->name_entry_size;
1445                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1446                    {
1447                    offset = GET2(slotB, 0) << 1;
1448                    condition = offset < offset_top &&
1449                      md->offset_vector[offset] >= 0;
1450                    if (condition) break;
1451                    }
1452                  else break;
1453                  }
1454                }
1455              }
1456            }
1457    
1458          /* Chose branch according to the condition */
1459    
1460          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1461          }
1462    
1463        else if (condcode == OP_DEF)     /* DEFINE - always false */
1464          {
1465          condition = FALSE;
1466          ecode += GET(ecode, 1);
1467          }
1468    
1469        /* The condition is an assertion. Call match() to evaluate it - setting
1470        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1471        an assertion. */
1472    
1473        else
1474          {
1475          md->match_function_type = MATCH_CONDASSERT;
1476          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1477          if (rrc == MATCH_MATCH)
1478            {
1479            if (md->end_offset_top > offset_top)
1480              offset_top = md->end_offset_top;  /* Captures may have happened */
1481            condition = TRUE;
1482            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1483            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1484            }
1485    
1486          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1487          assertion; it is therefore treated as NOMATCH. */
1488    
1489          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1490            {
1491            RRETURN(rrc);         /* Need braces because of following else */
1492            }
1493          else
1494            {
1495            condition = FALSE;
1496            ecode += codelink;
1497            }
1498          }
1499    
1500        /* We are now at the branch that is to be obeyed. As there is only one, can
1501        use tail recursion to avoid using another stack frame, except when there is
1502        unlimited repeat of a possibly empty group. In the latter case, a recursive
1503        call to match() is always required, unless the second alternative doesn't
1504        exist, in which case we can just plough on. Note that, for compatibility
1505        with Perl, the | in a conditional group is NOT treated as creating two
1506        alternatives. If a THEN is encountered in the branch, it propagates out to
1507        the enclosing alternative (unless nested in a deeper set of alternatives,
1508        of course). */
1509    
1510        if (condition || *ecode == OP_ALT)
1511          {
1512          if (op != OP_SCOND)
1513            {
1514            ecode += 1 + LINK_SIZE;
1515            goto TAIL_RECURSE;
1516            }
1517    
1518          md->match_function_type = MATCH_CBEGROUP;
1519          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1520          RRETURN(rrc);
1521          }
1522    
1523         /* Condition false & no alternative; continue after the group. */
1524    
1525        else
1526          {
1527          ecode += 1 + LINK_SIZE;
1528          }
1529        break;
1530    
1531    
1532        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1533        to close any currently open capturing brackets. */
1534    
1535        case OP_CLOSE:
1536        number = GET2(ecode, 1);   /* Must be less than 65536 */
1537        offset = number << 1;
1538    
1539    #ifdef PCRE_DEBUG
1540          printf("end bracket %d at *ACCEPT", number);
1541          printf("\n");
1542    #endif
1543    
1544        md->capture_last = (md->capture_last & OVFLMASK) | number;
1545        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1546          {
1547          md->offset_vector[offset] =
1548            md->offset_vector[md->offset_end - number];
1549          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1550          if (offset_top <= offset) offset_top = offset + 2;
1551          }
1552        ecode += 1 + IMM2_SIZE;
1553        break;
1554    
1555    
1556        /* End of the pattern, either real or forced. */
1557    
1558        case OP_END:
1559        case OP_ACCEPT:
1560        case OP_ASSERT_ACCEPT:
1561    
1562        /* If we have matched an empty string, fail if not in an assertion and not
1563        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1564        is set and we have matched at the start of the subject. In both cases,
1565        backtracking will then try other alternatives, if any. */
1566    
1567        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1568             md->recursive == NULL &&
1569             (md->notempty ||
1570               (md->notempty_atstart &&
1571                 mstart == md->start_subject + md->start_offset)))
1572          RRETURN(MATCH_NOMATCH);
1573    
1574        /* Otherwise, we have a match. */
1575    
1576        md->end_match_ptr = eptr;           /* Record where we ended */
1577        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1578        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1579    
1580        /* For some reason, the macros don't work properly if an expression is
1581        given as the argument to RRETURN when the heap is in use. */
1582    
1583        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1584        RRETURN(rrc);
1585    
1586        /* Assertion brackets. Check the alternative branches in turn - the
1587        matching won't pass the KET for an assertion. If any one branch matches,
1588      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
1589      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
1590      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1591        of a condition, we want to return immediately afterwards. The caller of
1592        this incarnation of the match() function will have set MATCH_CONDASSERT in
1593        md->match_function type, and one of these opcodes will be the first opcode
1594        that is processed. We use a local variable that is preserved over calls to
1595        match() to remember this case. */
1596    
1597      case OP_ASSERT:      case OP_ASSERT:
1598      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1599        save_mark = md->mark;
1600        if (md->match_function_type == MATCH_CONDASSERT)
1601          {
1602          condassert = TRUE;
1603          md->match_function_type = 0;
1604          }
1605        else condassert = FALSE;
1606    
1607      do      do
1608        {        {
1609        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1610          RM4);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1611        if (rrc == MATCH_MATCH) break;          {
1612            mstart = md->start_match_ptr;   /* In case \K reset it */
1613            break;
1614            }
1615          md->mark = save_mark;
1616    
1617          /* A COMMIT failure must fail the entire assertion, without trying any
1618          subsequent branches. */
1619    
1620          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1621    
1622          /* PCRE does not allow THEN to escape beyond an assertion; it
1623          is treated as NOMATCH. */
1624    
1625        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1626        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1627        }        }
1628      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1629    
1630      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1631    
1632      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1633    
1634      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1635    
1636      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1637      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 926  for (;;) Line 1641  for (;;)
1641      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1642      continue;      continue;
1643    
1644      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1645        PRUNE, or COMMIT means we must assume failure without checking subsequent
1646        branches. */
1647    
1648      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1649      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1650        save_mark = md->mark;
1651        if (md->match_function_type == MATCH_CONDASSERT)
1652          {
1653          condassert = TRUE;
1654          md->match_function_type = 0;
1655          }
1656        else condassert = FALSE;
1657    
1658      do      do
1659        {        {
1660        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1661          RM5);        md->mark = save_mark;
1662        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1663          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1664            {
1665            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1666            break;
1667            }
1668    
1669          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1670          as NOMATCH. */
1671    
1672        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1673        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1674        }        }
1675      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1676    
1677      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1678    
1679      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1680      continue;      continue;
# Line 951  for (;;) Line 1685  for (;;)
1685      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1686    
1687      case OP_REVERSE:      case OP_REVERSE:
1688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1689      if (utf8)      if (utf)
1690        {        {
1691        i = GET(ecode, 1);        i = GET(ecode, 1);
1692        while (i-- > 0)        while (i-- > 0)
# Line 972  for (;;) Line 1706  for (;;)
1706        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1707        }        }
1708    
1709      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1710    
1711        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1712      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1713      break;      break;
1714    
# Line 982  for (;;) Line 1717  for (;;)
1717      function is able to force a failure. */      function is able to force a failure. */
1718    
1719      case OP_CALLOUT:      case OP_CALLOUT:
1720      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1721        {        {
1722        pcre_callout_block cb;        PUBL(callout_block) cb;
1723        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1724        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1725        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1726    #if defined COMPILE_PCRE8
1727        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1728        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1729        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1730        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1731          cb.subject          = (PCRE_SPTR32)md->start_subject;
1732    #endif
1733          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1734          cb.start_match      = (int)(mstart - md->start_subject);
1735          cb.current_position = (int)(eptr - md->start_subject);
1736        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1737        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1738        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1739        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1740          /* Internal change requires this for API compatibility. */
1741          if (cb.capture_last == 0) cb.capture_last = -1;
1742        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1743        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1744          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1745        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1746        }        }
1747      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 1007  for (;;) Line 1751  for (;;)
1751      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
1752      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1753    
1754      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1755      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
1756      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
1757      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
1758      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
1759      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
1760      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.  
1761    
1762      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
1763      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
1764      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1765        a lot, so he is not to blame for the current way it works. */
1766    
1767      case OP_RECURSE:      case OP_RECURSE:
1768        {        {
1769          recursion_info *ri;
1770          unsigned int recno;
1771    
1772        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1773        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1774          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1775    
1776          /* Check for repeating a recursion without advancing the subject pointer.
1777          This should catch convoluted mutual recursions. (Some simple cases are
1778          caught at compile time.) */
1779    
1780          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1781            if (recno == ri->group_num && eptr == ri->subject_position)
1782              RRETURN(PCRE_ERROR_RECURSELOOP);
1783    
1784        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1785    
1786          new_recursive.group_num = recno;
1787          new_recursive.saved_capture_last = md->capture_last;
1788          new_recursive.subject_position = eptr;
1789        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1790        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1791    
1792        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1793    
1794        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1795    
1796        /* Now save the offset data. */        /* Now save the offset data */
1797    
1798        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1799        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1046  for (;;) Line 1801  for (;;)
1801        else        else
1802          {          {
1803          new_recursive.offset_save =          new_recursive.offset_save =
1804            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1805          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1806          }          }
   
1807        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1808              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1809    
1810        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1811        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1812          recursions, md->recursive might be changed, so reset it before looping.
1813          */
1814    
1815        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1816        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1817        do        do
1818          {          {
1819          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1820            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1821          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1822            memcpy(md->offset_vector, new_recursive.offset_save,
1823                new_recursive.saved_max * sizeof(int));
1824            md->capture_last = new_recursive.saved_capture_last;
1825            md->recursive = new_recursive.prevrec;
1826            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1827            {            {
1828            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1829            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1830              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1831            RRETURN(MATCH_MATCH);  
1832              /* Set where we got to in the subject, and reset the start in case
1833              it was changed by \K. This *is* propagated back out of a recursion,
1834              for Perl compatibility. */
1835    
1836              eptr = md->end_match_ptr;
1837              mstart = md->start_match_ptr;
1838              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1839            }            }
1840          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1841            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1842            is treated as NOMATCH. */
1843    
1844            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1845                     rrc != MATCH_COMMIT)
1846            {            {
1847            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1848              if (new_recursive.offset_save != stacksave)
1849                (PUBL(free))(new_recursive.offset_save);
1850            RRETURN(rrc);            RRETURN(rrc);
1851            }            }
1852    
1853          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1854          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1855          }          }
1856        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1088  for (;;) Line 1858  for (;;)
1858        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1859        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1860        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1861          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1862        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1863        }        }
     /* Control never reaches here */  
1864    
1865      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1866      the point in the subject string is not moved back. Thus there can never be      break;
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1867    
1868      /* 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
1869      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1181  for (;;) Line 1879  for (;;)
1879      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1880    
1881      case OP_BRAZERO:      case OP_BRAZERO:
1882        {      next = ecode + 1;
1883        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1884        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1885        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1886        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1887      break;      break;
1888    
1889      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1890        {      next = ecode + 1;
1891        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1892        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1893        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1894        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1895      break;      break;
1896    
1897      case OP_SKIPZERO:      case OP_SKIPZERO:
1898        {      next = ecode+1;
1899        next = ecode+1;      do next += GET(next,1); while (*next == OP_ALT);
1900        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1901      break;      break;
1902    
1903        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1904        here; just jump to the group, with allow_zero set TRUE. */
1905    
1906        case OP_BRAPOSZERO:
1907        op = *(++ecode);
1908        allow_zero = TRUE;
1909        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1910          goto POSSESSIVE_NON_CAPTURE;
1911    
1912      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1913    
1914      case OP_KET:      case OP_KET:
1915      case OP_KETRMIN:      case OP_KETRMIN:
1916      case OP_KETRMAX:      case OP_KETRMAX:
1917        case OP_KETRPOS:
1918      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1919    
1920      /* 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
1921      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1922      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1923    
1924      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1925        {        {
1926        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1927        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1928        }        }
1929      else saved_eptr = NULL;      else saved_eptr = NULL;
1930    
1931      /* If we are at the end of an assertion group, stop matching and return      /* If we are at the end of an assertion group or a non-capturing atomic
1932      MATCH_MATCH, but record the current high water mark for use by positive      group, stop matching and return MATCH_MATCH, but record the current high
1933      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1934        start in case it was changed by \K. */
1935      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1936          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1937          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1938        {        {
1939        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1940        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1941        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1942          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1943        }        }
1944    
1945      /* 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
1946      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1947      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
1948      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
1949      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
1950        the current subject position and start match pointer and give a MATCH
1951        return. */
1952    
1953      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1954            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1955        {        {
1956        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1957        offset = number << 1;        offset = number << 1;
1958    
1959  #ifdef DEBUG  #ifdef PCRE_DEBUG
1960        printf("end bracket %d", number);        printf("end bracket %d", number);
1961        printf("\n");        printf("\n");
1962  #endif  #endif
1963    
1964        md->capture_last = number;        /* Handle a recursively called group. */
1965        if (offset >= md->offset_max) md->offset_overflow = TRUE; else  
1966          if (md->recursive != NULL && md->recursive->group_num == number)
1967          {          {
1968          md->offset_vector[offset] =          md->end_match_ptr = eptr;
1969            md->offset_vector[md->offset_end - number];          md->start_match_ptr = mstart;
1970          md->offset_vector[offset+1] = eptr - md->start_subject;          RRETURN(MATCH_MATCH);
         if (offset_top <= offset) offset_top = offset + 2;  
1971          }          }
1972    
1973        /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
       appropriately and continue from after the call. */  
1974    
1975        if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
1976          if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1977          {          {
1978          recursion_info *rec = md->recursive;          /* If offset is greater than offset_top, it means that we are
1979          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          "skipping" a capturing group, and that group's offsets must be marked
1980          md->recursive = rec->prevrec;          unset. In earlier versions of PCRE, all the offsets were unset at the
1981          mstart = rec->save_start;          start of matching, but this doesn't work because atomic groups and
1982          memcpy(md->offset_vector, rec->offset_save,          assertions can cause a value to be set that should later be unset.
1983            rec->saved_max * sizeof(int));          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1984          ecode = rec->after_call;          part of the atomic group, but this is not on the final matching path,
1985          ims = original_ims;          so must be unset when 2 is set. (If there is no group 2, there is no
1986          break;          problem, because offset_top will then be 2, indicating no capture.) */
         }  
       }  
1987    
1988      /* For both capturing and non-capturing groups, reset the value of the ims          if (offset > offset_top)
1989      flags, in case they got changed during the group. */            {
1990              register int *iptr = md->offset_vector + offset_top;
1991              register int *iend = md->offset_vector + offset;
1992              while (iptr < iend) *iptr++ = -1;
1993              }
1994    
1995      ims = original_ims;          /* Now make the extraction */
     DPRINTF(("ims reset to %02lx\n", ims));  
1996    
1997      /* For a non-repeating ket, just continue at this level. This also          md->offset_vector[offset] =
1998      happens for a repeating ket if no characters were matched in the group.            md->offset_vector[md->offset_end - number];
1999      This is the forcible breaking of infinite loops as implemented in Perl          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2000      5.005. If there is an options reset, it will get obeyed in the normal          if (offset_top <= offset) offset_top = offset + 2;
2001      course of events. */          }
2002          }
2003    
2004        /* For an ordinary non-repeating ket, just continue at this level. This
2005        also happens for a repeating ket if no characters were matched in the
2006        group. This is the forcible breaking of infinite loops as implemented in
2007        Perl 5.005. For a non-repeating atomic group that includes captures,
2008        establish a backup point by processing the rest of the pattern at a lower
2009        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2010        original OP_ONCE level, thereby bypassing intermediate backup points, but
2011        resetting any captures that happened along the way. */
2012    
2013      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
2014        {        {
2015        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
2016            {
2017            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2018            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2019            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2020            RRETURN(MATCH_ONCE);
2021            }
2022          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2023        break;        break;
2024        }        }
2025    
2026      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2027      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2028      tail recursion to avoid using another stack frame, unless we have an      at a time from the outer level, thus saving stack. */
     unlimited repeat of a group that can match an empty string. */  
2029    
2030      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      if (*ecode == OP_KETRPOS)
2031          {
2032          md->end_match_ptr = eptr;
2033          md->end_offset_top = offset_top;
2034          RRETURN(MATCH_KETRPOS);
2035          }
2036    
2037        /* The normal repeating kets try the rest of the pattern or restart from
2038        the preceding bracket, in the appropriate order. In the second case, we can
2039        use tail recursion to avoid using another stack frame, unless we have an
2040        an atomic group or an unlimited repeat of a group that can match an empty
2041        string. */
2042    
2043      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2044        {        {
2045        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2046        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2047        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2048            {
2049            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2050            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2051            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2052            RRETURN(MATCH_ONCE);
2053            }
2054          if (*prev >= OP_SBRA)    /* Could match an empty string */
2055          {          {
2056          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2057          RRETURN(rrc);          RRETURN(rrc);
2058          }          }
2059        ecode = prev;        ecode = prev;
# Line 1320  for (;;) Line 2061  for (;;)
2061        }        }
2062      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2063        {        {
2064        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2065          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2066        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2067          if (*prev == OP_ONCE)
2068            {
2069            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2070            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2071            md->once_target = prev;
2072            RRETURN(MATCH_ONCE);
2073            }
2074        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2075        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2076        }        }
2077      /* Control never gets here */      /* Control never gets here */
2078    
2079      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2080    
2081      case OP_CIRC:      case OP_CIRC:
2082      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2083    
2084      /* Start of subject assertion */      /* Start of subject assertion */
2085    
# Line 1349  for (;;) Line 2088  for (;;)
2088      ecode++;      ecode++;
2089      break;      break;
2090    
2091        /* Multiline mode: start of subject unless notbol, or after any newline. */
2092    
2093        case OP_CIRCM:
2094        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2095        if (eptr != md->start_subject &&
2096            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2097          RRETURN(MATCH_NOMATCH);
2098        ecode++;
2099        break;
2100    
2101      /* Start of match assertion */      /* Start of match assertion */
2102    
2103      case OP_SOM:      case OP_SOM:
# Line 1363  for (;;) Line 2112  for (;;)
2112      ecode++;      ecode++;
2113      break;      break;
2114    
2115      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2116      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2117    
2118      case OP_DOLL:      case OP_DOLLM:
2119      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2120        {        {
2121        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2122          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2123        else          if (md->partial != 0 &&
2124          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2125        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2126        break;              NLBLOCK->nllen == 2 &&
2127                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2128              {
2129              md->hitend = TRUE;
2130              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2131              }
2132            RRETURN(MATCH_NOMATCH);
2133            }
2134        }        }
2135      else      else
2136        {        {
2137        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2138        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2139        }        }
2140        ecode++;
2141        break;
2142    
2143        /* Not multiline mode: assert before a terminating newline or before end of
2144        subject unless noteol is set. */
2145    
2146        case OP_DOLL:
2147        if (md->noteol) RRETURN(MATCH_NOMATCH);
2148        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2149    
2150      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2151    
2152      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2153    
2154      case OP_EOD:      case OP_EOD:
2155      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2156        SCHECK_PARTIAL();
2157      ecode++;      ecode++;
2158      break;      break;
2159    
2160      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2161    
2162      case OP_EODN:      case OP_EODN:
2163      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2164        if (eptr < md->end_subject &&
2165          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2166          {
2167          if (md->partial != 0 &&
2168              eptr + 1 >= md->end_subject &&
2169              NLBLOCK->nltype == NLTYPE_FIXED &&
2170              NLBLOCK->nllen == 2 &&
2171              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2172            {
2173            md->hitend = TRUE;
2174            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2175            }
2176        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2177          }
2178    
2179        /* Either at end of string or \n before end. */
2180    
2181        SCHECK_PARTIAL();
2182      ecode++;      ecode++;
2183      break;      break;
2184    
# Line 1414  for (;;) Line 2190  for (;;)
2190    
2191        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2192        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2193        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2194          partial matching. */
2195    
2196  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2197        if (utf8)        if (utf)
2198          {          {
2199            /* Get status of previous character */
2200    
2201          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2202            {            {
2203            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2204            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2205              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2206            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2207    #ifdef SUPPORT_UCP
2208              if (md->use_ucp)
2209                {
2210                if (c == '_') prev_is_word = TRUE; else
2211                  {
2212                  int cat = UCD_CATEGORY(c);
2213                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2214                  }
2215                }
2216              else
2217    #endif
2218            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2219            }            }
2220          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2221            /* Get status of next character */
2222    
2223            if (eptr >= md->end_subject)
2224              {
2225              SCHECK_PARTIAL();
2226              cur_is_word = FALSE;
2227              }
2228            else
2229            {            {
2230            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2231    #ifdef SUPPORT_UCP
2232              if (md->use_ucp)
2233                {
2234                if (c == '_') cur_is_word = TRUE; else
2235                  {
2236                  int cat = UCD_CATEGORY(c);
2237                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2238                  }
2239                }
2240              else
2241    #endif
2242            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2243            }            }
2244          }          }
2245        else        else
2246  #endif  #endif
2247    
2248        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2249          consistency with the behaviour of \w we do use it in this case. */
2250    
2251          {          {
2252          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2253            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2254          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2255            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2256              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2257    #ifdef SUPPORT_UCP
2258              if (md->use_ucp)
2259                {
2260                c = eptr[-1];
2261                if (c == '_') prev_is_word = TRUE; else
2262                  {
2263                  int cat = UCD_CATEGORY(c);
2264                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2265                  }
2266                }
2267              else
2268    #endif
2269              prev_is_word = MAX_255(eptr[-1])
2270                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2271              }
2272    
2273            /* Get status of next character */
2274    
2275            if (eptr >= md->end_subject)
2276              {
2277              SCHECK_PARTIAL();
2278              cur_is_word = FALSE;
2279              }
2280            else
2281    #ifdef SUPPORT_UCP
2282            if (md->use_ucp)
2283              {
2284              c = *eptr;
2285              if (c == '_') cur_is_word = TRUE; else
2286                {
2287                int cat = UCD_CATEGORY(c);
2288                cur_is_word = (cat == ucp_L || cat == ucp_N);
2289                }
2290              }
2291            else
2292    #endif
2293            cur_is_word = MAX_255(*eptr)
2294              && ((md->ctypes[*eptr] & ctype_word) != 0);
2295          }          }
2296    
2297        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1452  for (;;) Line 2302  for (;;)
2302        }        }
2303      break;      break;
2304    
2305      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2306        CRLF newlines and partial matching. */
2307    
2308      case OP_ANY:      case OP_ANY:
2309      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2310        if (md->partial != 0 &&
2311            eptr + 1 >= md->end_subject &&
2312            NLBLOCK->nltype == NLTYPE_FIXED &&
2313            NLBLOCK->nllen == 2 &&
2314            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2315          {
2316          md->hitend = TRUE;
2317          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2318          }
2319    
2320      /* Fall through */      /* Fall through */
2321    
2322        /* Match any single character whatsoever. */
2323    
2324      case OP_ALLANY:      case OP_ALLANY:
2325      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2326      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;        {                            /* not be updated before SCHECK_PARTIAL. */
2327          SCHECK_PARTIAL();
2328          RRETURN(MATCH_NOMATCH);
2329          }
2330        eptr++;
2331    #ifdef SUPPORT_UTF
2332        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2333    #endif
2334      ecode++;      ecode++;
2335      break;      break;
2336    
# Line 1468  for (;;) Line 2338  for (;;)
2338      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2339    
2340      case OP_ANYBYTE:      case OP_ANYBYTE:
2341      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2342          {                            /* not be updated before SCHECK_PARTIAL. */
2343          SCHECK_PARTIAL();
2344          RRETURN(MATCH_NOMATCH);
2345          }
2346        eptr++;
2347      ecode++;      ecode++;
2348      break;      break;
2349    
2350      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2351      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2352          {
2353          SCHECK_PARTIAL();
2354          RRETURN(MATCH_NOMATCH);
2355          }
2356      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2357      if (      if (
2358  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2359         c < 256 &&         c < 256 &&
2360  #endif  #endif
2361         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1486  for (;;) Line 2365  for (;;)
2365      break;      break;
2366    
2367      case OP_DIGIT:      case OP_DIGIT:
2368      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2369          {
2370          SCHECK_PARTIAL();
2371          RRETURN(MATCH_NOMATCH);
2372          }
2373      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2374      if (      if (
2375  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2376         c >= 256 ||         c > 255 ||
2377  #endif  #endif
2378         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2379         )         )
# Line 1499  for (;;) Line 2382  for (;;)
2382      break;      break;
2383    
2384      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2385      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2386          {
2387          SCHECK_PARTIAL();
2388          RRETURN(MATCH_NOMATCH);
2389          }
2390      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2391      if (      if (
2392  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2393         c < 256 &&         c < 256 &&
2394  #endif  #endif
2395         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1512  for (;;) Line 2399  for (;;)
2399      break;      break;
2400    
2401      case OP_WHITESPACE:      case OP_WHITESPACE:
2402      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2403          {
2404          SCHECK_PARTIAL();
2405          RRETURN(MATCH_NOMATCH);
2406          }
2407      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2408      if (      if (
2409  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2410         c >= 256 ||         c > 255 ||
2411  #endif  #endif
2412         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2413         )         )
# Line 1525  for (;;) Line 2416  for (;;)
2416      break;      break;
2417    
2418      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2419      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2420          {
2421          SCHECK_PARTIAL();
2422          RRETURN(MATCH_NOMATCH);
2423          }
2424      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2425      if (      if (
2426  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2427         c < 256 &&         c < 256 &&
2428  #endif  #endif
2429         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1538  for (;;) Line 2433  for (;;)
2433      break;      break;
2434    
2435      case OP_WORDCHAR:      case OP_WORDCHAR:
2436      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2437          {
2438          SCHECK_PARTIAL();
2439          RRETURN(MATCH_NOMATCH);
2440          }
2441      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2442      if (      if (
2443  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2444         c >= 256 ||         c > 255 ||
2445  #endif  #endif
2446         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2447         )         )
# Line 1551  for (;;) Line 2450  for (;;)
2450      break;      break;
2451    
2452      case OP_ANYNL:      case OP_ANYNL:
2453      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2454          {
2455          SCHECK_PARTIAL();
2456          RRETURN(MATCH_NOMATCH);
2457          }
2458      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2459      switch(c)      switch(c)
2460        {        {
2461        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2462        case 0x000d:  
2463        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2464          if (eptr >= md->end_subject)
2465            {
2466            SCHECK_PARTIAL();
2467            }
2468          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2469        break;        break;
2470    
2471        case 0x000a:        case CHAR_LF:
2472        break;        break;
2473    
2474        case 0x000b:        case CHAR_VT:
2475        case 0x000c:        case CHAR_FF:
2476        case 0x0085:        case CHAR_NEL:
2477    #ifndef EBCDIC
2478        case 0x2028:        case 0x2028:
2479        case 0x2029:        case 0x2029:
2480    #endif  /* Not EBCDIC */
2481        if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);        if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2482        break;        break;
2483        }        }
# Line 1575  for (;;) Line 2485  for (;;)
2485      break;      break;
2486    
2487      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2488      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2489          {
2490          SCHECK_PARTIAL();
2491          RRETURN(MATCH_NOMATCH);
2492          }
2493      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2494      switch(c)      switch(c)
2495        {        {
2496          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2497        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 */  
       RRETURN(MATCH_NOMATCH);  
2498        }        }
2499      ecode++;      ecode++;
2500      break;      break;
2501    
2502      case OP_HSPACE:      case OP_HSPACE:
2503      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2504          {
2505          SCHECK_PARTIAL();
2506          RRETURN(MATCH_NOMATCH);
2507          }
2508      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2509      switch(c)      switch(c)
2510        {        {
2511          HSPACE_CASES: break;  /* Byte and multibyte cases */
2512        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       break;  
2513        }        }
2514      ecode++;      ecode++;
2515      break;      break;
2516    
2517      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2518      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2519          {
2520          SCHECK_PARTIAL();
2521          RRETURN(MATCH_NOMATCH);
2522          }
2523      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2524      switch(c)      switch(c)
2525        {        {
2526          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2527        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 */  
       RRETURN(MATCH_NOMATCH);  
2528        }        }
2529      ecode++;      ecode++;
2530      break;      break;
2531    
2532      case OP_VSPACE:      case OP_VSPACE:
2533      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2534          {
2535          SCHECK_PARTIAL();
2536          RRETURN(MATCH_NOMATCH);
2537          }
2538      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2539      switch(c)      switch(c)
2540        {        {
2541          VSPACE_CASES: break;
2542        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       break;  
2543        }        }
2544      ecode++;      ecode++;
2545      break;      break;
# Line 1676  for (;;) Line 2550  for (;;)
2550    
2551      case OP_PROP:      case OP_PROP:
2552      case OP_NOTPROP:      case OP_NOTPROP:
2553      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2554          {
2555          SCHECK_PARTIAL();
2556          RRETURN(MATCH_NOMATCH);
2557          }
2558      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2559        {        {
2560        const ucd_record * prop = GET_UCD(c);        const pcre_uint32 *cp;
2561          const ucd_record *prop = GET_UCD(c);
2562    
2563        switch(ecode[1])        switch(ecode[1])
2564          {          {
# Line 1692  for (;;) Line 2571  for (;;)
2571               prop->chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2572               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2573            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2574           break;          break;
2575    
2576          case PT_GC:          case PT_GC:
2577          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2578            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2579          break;          break;
2580    
# Line 1709  for (;;) Line 2588  for (;;)
2588            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2589          break;          break;
2590    
2591            /* These are specials */
2592    
2593            case PT_ALNUM:
2594            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2595                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2596              RRETURN(MATCH_NOMATCH);
2597            break;
2598    
2599            case PT_SPACE:    /* Perl space */
2600            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2601                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2602                   == (op == OP_NOTPROP))
2603              RRETURN(MATCH_NOMATCH);
2604            break;
2605    
2606            case PT_PXSPACE:  /* POSIX space */
2607            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2608                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2609                 c == CHAR_FF || c == CHAR_CR)
2610                   == (op == OP_NOTPROP))
2611              RRETURN(MATCH_NOMATCH);
2612            break;
2613    
2614            case PT_WORD:
2615            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2616                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2617                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2618              RRETURN(MATCH_NOMATCH);
2619            break;
2620    
2621            case PT_CLIST:
2622            cp = PRIV(ucd_caseless_sets) + ecode[2];
2623            for (;;)
2624              {
2625              if (c < *cp)
2626                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2627              if (c == *cp++)
2628                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2629              }
2630            break;
2631    
2632            /* This should never occur */
2633    
2634          default:          default:
2635          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2636          }          }
# Line 1721  for (;;) Line 2643  for (;;)
2643      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2644    
2645      case OP_EXTUNI:      case OP_EXTUNI:
2646      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2647      GETCHARINCTEST(c, eptr);        {
2648          SCHECK_PARTIAL();
2649          RRETURN(MATCH_NOMATCH);
2650          }
2651        else
2652        {        {
2653        int category = UCD_CATEGORY(c);        int lgb, rgb;
2654        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        GETCHARINCTEST(c, eptr);
2655          lgb = UCD_GRAPHBREAK(c);
2656        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2657          {          {
2658          int len = 1;          int len = 1;
2659          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2660            {          rgb = UCD_GRAPHBREAK(c);
2661            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2662            }          lgb = rgb;
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
2663          eptr += len;          eptr += len;
2664          }          }
2665        }        }
2666        CHECK_PARTIAL();
2667      ecode++;      ecode++;
2668      break;      break;
2669  #endif  #endif  /* SUPPORT_UCP */
2670    
2671    
2672      /* 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 1752  for (;;) Line 2678  for (;;)
2678      loops). */      loops). */
2679    
2680      case OP_REF:      case OP_REF:
2681        {      case OP_REFI:
2682        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2683        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2684        ecode += 1 + IMM2_SIZE;
2685    
2686        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2687    
2688        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2689        than the amount of subject left; this ensures that every attempt at a      this ensures that every attempt at a match fails. We can't just fail
2690        match fails. We can't just fail here, because of the possibility of      here, because of the possibility of quantifiers with zero minima.
       quantifiers with zero minima.  
2691    
2692        (b) If the JavaScript compatibility flag is set, set the length to zero      (b) If the JavaScript compatibility flag is set, set the length to zero
2693        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2694    
2695        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
2696        referenced subpattern. */      referenced subpattern. */
2697    
2698        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2699          length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;        length = (md->jscript_compat)? 0 : -1;
2700        else      else
2701          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2702    
2703        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2704    
2705        switch (*ecode)      switch (*ecode)
2706          {        {
2707          case OP_CRSTAR:        case OP_CRSTAR:
2708          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2709          case OP_CRPLUS:        case OP_CRPLUS:
2710          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2711          case OP_CRQUERY:        case OP_CRQUERY:
2712          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2713          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2714          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2715          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2716          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2717          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2718          break;        break;
2719    
2720          case OP_CRRANGE:        case OP_CRRANGE:
2721          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2722          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2723          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2724          max = GET2(ecode, 3);        max = GET2(ecode, 1 + IMM2_SIZE);
2725          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2726          ecode += 5;        ecode += 1 + 2 * IMM2_SIZE;
2727          break;        break;
2728    
2729          default:               /* No repeat follows */        default:               /* No repeat follows */
2730          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2731          eptr += length;          {
2732          continue;              /* With the main loop */          if (length == -2) eptr = md->end_subject;   /* Partial match */
2733            CHECK_PARTIAL();
2734            RRETURN(MATCH_NOMATCH);
2735          }          }
2736          eptr += length;
2737          continue;              /* With the main loop */
2738          }
2739    
2740        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2741        main loop. */      zero, just continue with the main loop. If the length is negative, it
2742        means the reference is unset in non-Java-compatible mode. If the minimum is
2743        zero, we can continue at the same level without recursion. For any other
2744        minimum, carrying on will result in NOMATCH. */
2745    
2746        if (length == 0) continue;      if (length == 0) continue;
2747        if (length < 0 && min == 0) continue;
2748    
2749        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2750        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2751        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2752    
2753        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2754          {
2755          int slength;
2756          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2757          {          {
2758          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2759          eptr += length;          CHECK_PARTIAL();
2760            RRETURN(MATCH_NOMATCH);
2761          }          }
2762          eptr += slength;
2763          }
2764    
2765        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2766        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2767    
2768        if (min == max) continue;      if (min == max) continue;
2769    
2770        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2771    
2772        if (minimize)      if (minimize)
2773          {
2774          for (fi = min;; fi++)
2775          {          {
2776          for (fi = min;; fi++)          int slength;
2777            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2778            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2779            if (fi >= max) RRETURN(MATCH_NOMATCH);
2780            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2781            {            {
2782            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2783            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2784            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2785            }            }
2786          /* Control never gets here */          eptr += slength;
2787          }          }
2788          /* Control never gets here */
2789          }
2790    
2791        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2792    
2793        else      else
2794          {
2795          pp = eptr;
2796          for (i = min; i < max; i++)
2797          {          {
2798          pp = eptr;          int slength;
2799          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
2800            {            {
2801            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2802            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2803            eptr -= length;  
2804              if (slength == -2 && md->partial != 0 &&
2805                  md->end_subject > md->start_used_ptr)
2806                {
2807                md->hitend = TRUE;
2808                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2809                }
2810              break;
2811            }            }
2812          RRETURN(MATCH_NOMATCH);          eptr += slength;
2813            }
2814    
2815          while (eptr >= pp)
2816            {
2817            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2818            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2819            eptr -= length;
2820          }          }
2821          RRETURN(MATCH_NOMATCH);
2822        }        }
2823      /* Control never gets here */      /* Control never gets here */
2824    
   
   
2825      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2826      used when all the characters in the class have values in the range 0-255,      used when all the characters in the class have values in the range 0-255,
2827      and either the matching is caseful, or the characters are in the range      and either the matching is caseful, or the characters are in the range
# Line 1878  for (;;) Line 2836  for (;;)
2836      case OP_NCLASS:      case OP_NCLASS:
2837      case OP_CLASS:      case OP_CLASS:
2838        {        {
2839          /* The data variable is saved across frames, so the byte map needs to
2840          be stored there. */
2841    #define BYTE_MAP ((pcre_uint8 *)data)
2842        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2843        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2844    
2845        switch (*ecode)        switch (*ecode)
2846          {          {
# Line 1900  for (;;) Line 2861  for (;;)
2861          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2862          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2863          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2864          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2865          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2866          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2867          break;          break;
2868    
2869          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1912  for (;;) Line 2873  for (;;)
2873    
2874        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2875    
2876  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2877        /* UTF-8 mode */        if (utf)
       if (utf8)  
2878          {          {
2879          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2880            {            {
2881            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2882                {
2883                SCHECK_PARTIAL();
2884                RRETURN(MATCH_NOMATCH);
2885                }
2886            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2887            if (c > 255)            if (c > 255)
2888              {              {
2889              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2890              }              }
2891            else            else
2892              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2893            }            }
2894          }          }
2895        else        else
2896  #endif  #endif
2897        /* Not UTF-8 mode */        /* Not UTF mode */
2898          {          {
2899          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2900            {            {
2901            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2902                {
2903                SCHECK_PARTIAL();
2904                RRETURN(MATCH_NOMATCH);
2905                }
2906            c = *eptr++;            c = *eptr++;
2907            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2908              if (c > 255)
2909                {
2910                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2911                }
2912              else
2913    #endif
2914                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2915            }            }
2916          }          }
2917    
# Line 1952  for (;;) Line 2925  for (;;)
2925    
2926        if (minimize)        if (minimize)
2927          {          {
2928  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2929          /* UTF-8 mode */          if (utf)
         if (utf8)  
2930            {            {
2931            for (fi = min;; fi++)            for (fi = min;; fi++)
2932              {              {
2933              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2934              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2935              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2936                if (eptr >= md->end_subject)
2937                  {
2938                  SCHECK_PARTIAL();
2939                  RRETURN(MATCH_NOMATCH);
2940                  }
2941              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2942              if (c > 255)              if (c > 255)
2943                {                {
2944                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2945                }                }
2946              else              else
2947                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2948              }              }
2949            }            }
2950          else          else
2951  #endif  #endif
2952          /* Not UTF-8 mode */          /* Not UTF mode */
2953            {            {
2954            for (fi = min;; fi++)            for (fi = min;; fi++)
2955              {              {
2956              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2957              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2958              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2959                if (eptr >= md->end_subject)
2960                  {
2961                  SCHECK_PARTIAL();
2962                  RRETURN(MATCH_NOMATCH);
2963                  }
2964              c = *eptr++;              c = *eptr++;
2965              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2966                if (c > 255)
2967                  {
2968                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2969                  }
2970                else
2971    #endif
2972                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2973              }              }
2974            }            }
2975          /* Control never gets here */          /* Control never gets here */
# Line 1994  for (;;) Line 2981  for (;;)
2981          {          {
2982          pp = eptr;          pp = eptr;
2983    
2984  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2985          /* UTF-8 mode */          if (utf)
         if (utf8)  
2986            {            {
2987            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2988              {              {
2989              int len = 1;              int len = 1;
2990              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2991                  {
2992                  SCHECK_PARTIAL();
2993                  break;
2994                  }
2995              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2996              if (c > 255)              if (c > 255)
2997                {                {
2998                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2999                }                }
3000              else              else
3001                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3002              eptr += len;              eptr += len;
3003              }              }
3004            for (;;)            for (;;)
3005              {              {
3006              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3007              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3008              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3009              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2023  for (;;) Line 3011  for (;;)
3011            }            }
3012          else          else
3013  #endif  #endif
3014            /* Not UTF-8 mode */            /* Not UTF mode */
3015            {            {
3016            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3017              {              {
3018              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3019                  {
3020                  SCHECK_PARTIAL();
3021                  break;
3022                  }
3023              c = *eptr;              c = *eptr;
3024              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3025                if (c > 255)
3026                  {
3027                  if (op == OP_CLASS) break;
3028                  }
3029                else
3030    #endif
3031                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3032              eptr++;              eptr++;
3033              }              }
3034            while (eptr >= pp)            while (eptr >= pp)
3035              {              {
3036              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3037              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3038              eptr--;              eptr--;
3039              }              }
# Line 2042  for (;;) Line 3041  for (;;)
3041    
3042          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3043          }          }
3044    #undef BYTE_MAP
3045        }        }
3046      /* Control never gets here */      /* Control never gets here */
3047    
3048    
3049      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3050      in UTF-8 mode, because that's the only time it is compiled. */      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
3051        mode, because Unicode properties are supported in non-UTF-8 mode. */
3052    
3053  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3054      case OP_XCLASS:      case OP_XCLASS:
3055        {        {
3056        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2074  for (;;) Line 3075  for (;;)
3075          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3076          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3077          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3078          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3079          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3080          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3081          break;          break;
3082    
3083          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2088  for (;;) Line 3089  for (;;)
3089    
3090        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3091          {          {
3092          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3093          GETCHARINC(c, eptr);            {
3094          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3095              RRETURN(MATCH_NOMATCH);
3096              }
3097            GETCHARINCTEST(c, eptr);
3098            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3099          }          }
3100    
3101        /* 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 2105  for (;;) Line 3110  for (;;)
3110          {          {
3111          for (fi = min;; fi++)          for (fi = min;; fi++)
3112            {            {
3113            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3114            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3115            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3116            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3117            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3118                SCHECK_PARTIAL();
3119                RRETURN(MATCH_NOMATCH);
3120                }
3121              GETCHARINCTEST(c, eptr);
3122              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3123            }            }
3124          /* Control never gets here */          /* Control never gets here */
3125          }          }
# Line 2122  for (;;) Line 3132  for (;;)
3132          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3133            {            {
3134            int len = 1;            int len = 1;
3135            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3136            GETCHARLEN(c, eptr, len);              {
3137            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3138                break;
3139                }
3140    #ifdef SUPPORT_UTF
3141              GETCHARLENTEST(c, eptr, len);
3142    #else
3143              c = *eptr;
3144    #endif
3145              if (!PRIV(xclass)(c, data, utf)) break;
3146            eptr += len;            eptr += len;
3147            }            }
3148          for(;;)          for(;;)
3149            {            {
3150            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3151            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3152            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3153            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3154              if (utf) BACKCHAR(eptr);
3155    #endif
3156            }            }
3157          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3158          }          }
# Line 2144  for (;;) Line 3164  for (;;)
3164      /* Match a single character, casefully */      /* Match a single character, casefully */
3165    
3166      case OP_CHAR:      case OP_CHAR:
3167  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3168      if (utf8)      if (utf)
3169        {        {
3170        length = 1;        length = 1;
3171        ecode++;        ecode++;
3172        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3173        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3174        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3175            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3176            RRETURN(MATCH_NOMATCH);
3177            }
3178          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3179        }        }
3180      else      else
3181  #endif  #endif
3182        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3183        {        {
3184        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3185            {
3186            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3187            RRETURN(MATCH_NOMATCH);
3188            }
3189        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3190        ecode += 2;        ecode += 2;
3191        }        }
3192      break;      break;
3193    
3194      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3195        subject, give up immediately. */
3196    
3197        case OP_CHARI:
3198        if (eptr >= md->end_subject)
3199          {
3200          SCHECK_PARTIAL();
3201          RRETURN(MATCH_NOMATCH);
3202          }
3203    
3204      case OP_CHARNC:  #ifdef SUPPORT_UTF
3205  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3206        {        {
3207        length = 1;        length = 1;
3208        ecode++;        ecode++;
3209        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3210    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3211        /* 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
3212        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3213          fast lookup table. We know that there is at least one byte left in the
3214          subject. */
3215    
3216        if (fc < 128)        if (fc < 128)
3217          {          {
3218          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uchar cc = RAWUCHAR(eptr);
3219            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3220            ecode++;
3221            eptr++;
3222          }          }
3223    
3224        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3225          use the value of "length" to check for sufficient bytes left, because the
3226          other case of the character may have more or fewer bytes.  */
3227    
3228        else        else
3229          {          {
3230          unsigned int dc;          pcre_uint32 dc;
3231          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3232          ecode += length;          ecode += length;
3233    
# Line 2205  for (;;) Line 3244  for (;;)
3244          }          }
3245        }        }
3246      else      else
3247  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3248    
3249      /* Non-UTF-8 mode */      /* Not UTF mode */
3250        {        {
3251        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3252        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3253          eptr++;
3254        ecode += 2;        ecode += 2;
3255        }        }
3256      break;      break;
# Line 2218  for (;;) Line 3258  for (;;)
3258      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3259    
3260      case OP_EXACT:      case OP_EXACT:
3261        case OP_EXACTI:
3262      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3263      ecode += 3;      ecode += 1 + IMM2_SIZE;
3264      goto REPEATCHAR;      goto REPEATCHAR;
3265    
3266      case OP_POSUPTO:      case OP_POSUPTO:
3267        case OP_POSUPTOI:
3268      possessive = TRUE;      possessive = TRUE;
3269      /* Fall through */      /* Fall through */
3270    
3271      case OP_UPTO:      case OP_UPTO:
3272        case OP_UPTOI:
3273      case OP_MINUPTO:      case OP_MINUPTO:
3274        case OP_MINUPTOI:
3275      min = 0;      min = 0;
3276      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3277      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3278      ecode += 3;      ecode += 1 + IMM2_SIZE;
3279      goto REPEATCHAR;      goto REPEATCHAR;
3280    
3281      case OP_POSSTAR:      case OP_POSSTAR:
3282        case OP_POSSTARI:
3283      possessive = TRUE;      possessive = TRUE;
3284      min = 0;      min = 0;
3285      max = INT_MAX;      max = INT_MAX;
# Line 2242  for (;;) Line 3287  for (;;)
3287      goto REPEATCHAR;      goto REPEATCHAR;
3288    
3289      case OP_POSPLUS:      case OP_POSPLUS:
3290        case OP_POSPLUSI:
3291      possessive = TRUE;      possessive = TRUE;
3292      min = 1;      min = 1;
3293      max = INT_MAX;      max = INT_MAX;
# Line 2249  for (;;) Line 3295  for (;;)
3295      goto REPEATCHAR;      goto REPEATCHAR;
3296    
3297      case OP_POSQUERY:      case OP_POSQUERY:
3298        case OP_POSQUERYI:
3299      possessive = TRUE;      possessive = TRUE;
3300      min = 0;      min = 0;
3301      max = 1;      max = 1;
# Line 2256  for (;;) Line 3303  for (;;)
3303      goto REPEATCHAR;      goto REPEATCHAR;
3304    
3305      case OP_STAR:      case OP_STAR:
3306        case OP_STARI:
3307      case OP_MINSTAR:      case OP_MINSTAR:
3308        case OP_MINSTARI:
3309      case OP_PLUS:      case OP_PLUS:
3310        case OP_PLUSI:
3311      case OP_MINPLUS:      case OP_MINPLUS:
3312        case OP_MINPLUSI:
3313      case OP_QUERY:      case OP_QUERY:
3314        case OP_QUERYI:
3315      case OP_MINQUERY:      case OP_MINQUERY:
3316      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3317        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3318      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3319      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3320      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3321      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3322    
3323      /* Common code for all repeated single-character matches. We can give      /* Common code for all repeated single-character matches. */
     up quickly if there are fewer than the minimum number of characters left in  
     the subject. */  
3324    
3325      REPEATCHAR:      REPEATCHAR:
3326  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3327      if (utf8)      if (utf)
3328        {        {
3329        length = 1;        length = 1;
3330        charptr = ecode;        charptr = ecode;
3331        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3332        ecode += length;        ecode += length;
3333    
3334        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2287  for (;;) Line 3337  for (;;)
3337        if (length > 1)        if (length > 1)
3338          {          {
3339  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3340          unsigned int othercase;          pcre_uint32 othercase;
3341          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3342              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3343            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3344          else oclength = 0;          else oclength = 0;
3345  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3346    
3347          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3348            {            {
3349            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3350                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3351  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3352            /* Need braces because of following else */            else if (oclength > 0 &&
3353            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3354                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3355    #endif  /* SUPPORT_UCP */
3356            else            else
3357              {              {
3358              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3359              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3360              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3361            }            }
3362    
3363          if (min == max) continue;          if (min == max) continue;
# Line 2316  for (;;) Line 3366  for (;;)
3366            {            {
3367            for (fi = min;; fi++)            for (fi = min;; fi++)
3368              {              {
3369              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3370              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3371              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3372              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3373                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3374  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3375              /* Need braces because of following else */              else if (oclength > 0 &&
3376              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3377                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3378    #endif  /* SUPPORT_UCP */
3379              else              else
3380                {                {
3381                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3382                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3383                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3384              }              }
3385            /* Control never gets here */            /* Control never gets here */
3386            }            }
# Line 2340  for (;;) Line 3390  for (;;)
3390            pp = eptr;            pp = eptr;
3391            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3392              {              {
3393              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3394              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3395  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3396              else if (oclength == 0) break;              else if (oclength > 0 &&
3397                         eptr <= md->end_subject - oclength &&
3398                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3399    #endif  /* SUPPORT_UCP */
3400              else              else
3401                {                {
3402                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3403                eptr += oclength;                break;
3404                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3405              }              }
3406    
3407            if (possessive) continue;            if (possessive) continue;
3408    
3409            for(;;)            for(;;)
3410             {              {
3411             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3412             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3413             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3414  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3415             eptr--;              eptr--;
3416             BACKCHAR(eptr);              BACKCHAR(eptr);
3417  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3418             eptr -= length;              eptr -= length;
3419  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3420             }              }
3421            }            }
3422          /* Control never gets here */          /* Control never gets here */
3423          }          }
# Line 2376  for (;;) Line 3427  for (;;)
3427        value of fc will always be < 128. */        value of fc will always be < 128. */
3428        }        }
3429      else      else
3430  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3431          /* When not in UTF-8 mode, load a single-byte character. */
     /* When not in UTF-8 mode, load a single-byte character. */  
       {  
       if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3432        fc = *ecode++;        fc = *ecode++;
       }  
3433