/[pcre]/code/branches/pcre16/pcre_exec.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_exec.c

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

code/trunk/pcre_exec.c revision 406 by ph10, Mon Mar 23 12:05:43 2009 UTC code/branches/pcre16/pcre_exec.c revision 806 by zherczeg, Thu Dec 15 11:57:39 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2009 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 57  possible. There are also some static sup Line 57  possible. There are also some static sup
57  #undef min  #undef min
58  #undef max  #undef max
59    
60  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
61    of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
63    
64  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 71  defined PCRE_ERROR_xxx codes, which are Line 73  defined PCRE_ERROR_xxx codes, which are
73  /* Special internal returns from the match() function. Make them sufficiently  /* Special internal returns from the match() function. Make them sufficiently
74  negative to avoid the external error codes. */  negative to avoid the external error codes. */
75    
76  #define MATCH_COMMIT       (-999)  #define MATCH_ACCEPT       (-999)
77  #define MATCH_PRUNE        (-998)  #define MATCH_COMMIT       (-998)
78  #define MATCH_SKIP         (-997)  #define MATCH_KETRPOS      (-997)
79  #define MATCH_THEN         (-996)  #define MATCH_ONCE         (-996)
80    #define MATCH_PRUNE        (-995)
81    #define MATCH_SKIP         (-994)
82    #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
86  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
# Line 89  static const char rep_max[] = { 0, 0, 0, Line 95  static const char rep_max[] = { 0, 0, 0,
95    
96    
97    
98  #ifdef DEBUG  #ifdef PCRE_DEBUG
99  /*************************************************  /*************************************************
100  *        Debugging function to print chars       *  *        Debugging function to print chars       *
101  *************************************************/  *************************************************/
# Line 107  Returns:     nothing Line 113  Returns:     nothing
113  */  */
114    
115  static void  static void
116  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
117  {  {
118  unsigned int c;  unsigned int c;
119  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
# Line 122  while (length-- > 0) Line 128  while (length-- > 0)
128  *          Match a back-reference                *  *          Match a back-reference                *
129  *************************************************/  *************************************************/
130    
131  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
132  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
133    the length passed is zero. Note that in caseless UTF-8 mode, the number of
134    subject bytes matched may be different to the number of reference bytes.
135    
136  Arguments:  Arguments:
137    offset      index into the offset vector    offset      index into the offset vector
138    eptr        points into the subject    eptr        pointer into the subject
139    length      length to be matched    length      length of reference to be matched (number of bytes)
140    md          points to match data block    md          points to match data block
141    ims         the ims flags    caseless    TRUE if caseless
142    
143  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
144  */  */
145    
146  static BOOL  static int
147  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
148    unsigned long int ims)    BOOL caseless)
149  {  {
150  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
151    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
152    
153  #ifdef DEBUG  #ifdef PCRE_DEBUG
154  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
155    printf("matching subject <null>");    printf("matching subject <null>");
156  else  else
# Line 154  pchars(p, length, FALSE, md); Line 163  pchars(p, length, FALSE, md);
163  printf("\n");  printf("\n");
164  #endif  #endif
165    
166  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
167    
168  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
169    
170  /* Separate the 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
171  properly if Unicode properties are supported. Otherwise, we can check only  properly if Unicode properties are supported. Otherwise, we can check only
172  ASCII characters. */  ASCII characters. */
173    
174  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
175    {    {
176  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
177  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
178    if (md->utf8)    if (md->utf)
179      {      {
180      USPTR endptr = eptr + length;      /* Match characters up to the end of the reference. NOTE: the number of
181      while (eptr < endptr)      bytes matched may differ, because there are some characters whose upper and
182        lower case versions code as different numbers of bytes. For example, U+023A
183        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
184        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
185        the latter. It is important, therefore, to check the length along the
186        reference, not along the subject (earlier code did this wrong). */
187    
188        PCRE_PUCHAR endptr = p + length;
189        while (p < endptr)
190        {        {
191        int c, d;        int c, d;
192          if (eptr >= md->end_subject) return -1;
193        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
194        GETCHARINC(d, p);        GETCHARINC(d, p);
195        if (c != d && c != UCD_OTHERCASE(d)) return FALSE;        if (c != d && c != UCD_OTHERCASE(d)) return -1;
196        }        }
197      }      }
198    else    else
# Line 183  if ((ims & PCRE_CASELESS) != 0) Line 201  if ((ims & PCRE_CASELESS) != 0)
201    
202    /* 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
203    is no UCP support. */    is no UCP support. */
204        {
205    while (length-- > 0)      if (eptr + length > md->end_subject) return -1;
206      { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }      while (length-- > 0)
207          {
208          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
209          p++;
210          eptr++;
211          }
212        }
213    }    }
214    
215  /* In the caseful case, we can just compare the bytes, whether or not we  /* In the caseful case, we can just compare the bytes, whether or not we
216  are in UTF-8 mode. */  are in UTF-8 mode. */
217    
218  else  else
219    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
220      if (eptr + length > md->end_subject) return -1;
221      while (length-- > 0) if (*p++ != *eptr++) return -1;
222      }
223    
224  return TRUE;  return (int)(eptr - eptr_start);
225  }  }
226    
227    
# Line 245  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 272  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
272         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
273         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
274         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
275         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
276           RM61,  RM62, RM63, RM64, RM65, RM66 };
277    
278  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
279  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
280  actuall used in this definition. */  actually used in this definition. */
281    
282  #ifndef NO_RECURSE  #ifndef NO_RECURSE
283  #define REGISTER register  #define REGISTER register
284    
285  #ifdef DEBUG  #ifdef PCRE_DEBUG
286  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
287    { \    { \
288    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
289    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
290    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
291    }    }
292  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 267  actuall used in this definition. */ Line 295  actuall used in this definition. */
295    return ra; \    return ra; \
296    }    }
297  #else  #else
298  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
299    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
300  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
301  #endif  #endif
302    
# Line 281  argument of match(), which never changes Line 309  argument of match(), which never changes
309    
310  #define REGISTER  #define REGISTER
311    
312  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
313    {\    {\
314    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
315      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
316    frame->Xwhere = rw; \    frame->Xwhere = rw; \
317    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
318    newframe->Xecode = rb;\    newframe->Xecode = rb;\
319    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
320    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
321    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
322    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
323    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
324    frame = newframe;\    frame = newframe;\
# Line 303  argument of match(), which never changes Line 330  argument of match(), which never changes
330    
331  #define RRETURN(ra)\  #define RRETURN(ra)\
332    {\    {\
333    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
334    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
335    (pcre_stack_free)(newframe);\    (PUBL(stack_free))(oldframe);\
336    if (frame != NULL)\    if (frame != NULL)\
337      {\      {\
338      rrc = ra;\      rrc = ra;\
# Line 322  typedef struct heapframe { Line 349  typedef struct heapframe {
349    
350    /* Function arguments that may change */    /* Function arguments that may change */
351    
352    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
353    const uschar *Xecode;    const pcre_uchar *Xecode;
354    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
355    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
356    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
357    unsigned int Xrdepth;    unsigned int Xrdepth;
358    
359    /* Function local variables */    /* Function local variables */
360    
361    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
362  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
363    const uschar *Xcharptr;    PCRE_PUCHAR Xcharptr;
364  #endif  #endif
365    const uschar *Xdata;    PCRE_PUCHAR Xdata;
366    const uschar *Xnext;    PCRE_PUCHAR Xnext;
367    const uschar *Xpp;    PCRE_PUCHAR Xpp;
368    const uschar *Xprev;    PCRE_PUCHAR Xprev;
369    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xsaved_eptr;
370    
371    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
372    
# Line 349  typedef struct heapframe { Line 374  typedef struct heapframe {
374    BOOL Xcondition;    BOOL Xcondition;
375    BOOL Xprev_is_word;    BOOL Xprev_is_word;
376    
   unsigned long int Xoriginal_ims;  
   
377  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
378    int Xprop_type;    int Xprop_type;
379    int Xprop_value;    int Xprop_value;
380    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
381    int Xoclength;    int Xoclength;
382    uschar Xocchars[8];    pcre_uchar Xocchars[6];
383  #endif  #endif
384    
385    int Xcodelink;    int Xcodelink;
# Line 398  typedef struct heapframe { Line 418  typedef struct heapframe {
418    
419  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
420  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
421  same response.  same response. */
422    
423    /* These macros pack up tests that are used for partial matching, and which
424    appear several times in the code. We set the "hit end" flag if the pointer is
425    at the end of the subject and also past the start of the subject (i.e.
426    something has been matched). For hard partial matching, we then return
427    immediately. The second one is used when we already know we are past the end of
428    the subject. */
429    
430    #define CHECK_PARTIAL()\
431      if (md->partial != 0 && eptr >= md->end_subject && \
432          eptr > md->start_used_ptr) \
433        { \
434        md->hitend = TRUE; \
435        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
436        }
437    
438  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
439  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
440        { \
441        md->hitend = TRUE; \
442        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
443        }
444    
445    
446    /* Performance note: It might be tempting to extract commonly used fields from
447    the md structure (e.g. utf, end_subject) into individual variables to improve
448  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
449  made performance worse.  made performance worse.
450    
# Line 412  Arguments: Line 455  Arguments:
455                   by encountering \K)                   by encountering \K)
456     offset_top  current top pointer     offset_top  current top pointer
457     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
458     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
459                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
460     rdepth      the recursion depth     rdepth      the recursion depth
461    
462  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
463                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
464                   a negative MATCH_xxx value for PRUNE, SKIP, etc
465                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
466                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
467  */  */
468    
469  static int  static int
470  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
471    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
472    int flags, unsigned int rdepth)    unsigned int rdepth)
473  {  {
474  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
475  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 439  so they can be ordinary variables in all Line 478  so they can be ordinary variables in all
478  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
479  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
480  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
481  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
482    
483  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
484    BOOL caseless;
485  int condcode;  int condcode;
486    
487  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
# Line 450  heap storage. Set up the top-level frame Line 490  heap storage. Set up the top-level frame
490  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
491    
492  #ifdef NO_RECURSE  #ifdef NO_RECURSE
493  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));
494    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
495  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
496    
497  /* Copy in the original argument variables */  /* Copy in the original argument variables */
# Line 459  frame->Xeptr = eptr; Line 500  frame->Xeptr = eptr;
500  frame->Xecode = ecode;  frame->Xecode = ecode;
501  frame->Xmstart = mstart;  frame->Xmstart = mstart;
502  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
503  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
504  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
505    
506  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 474  HEAP_RECURSE: Line 513  HEAP_RECURSE:
513  #define ecode              frame->Xecode  #define ecode              frame->Xecode
514  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
515  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
516  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
517  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
518    
519  /* Ditto for the local variables */  /* Ditto for the local variables */
520    
521  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
522  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
523  #endif  #endif
524  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
# Line 498  HEAP_RECURSE: Line 535  HEAP_RECURSE:
535  #define condition          frame->Xcondition  #define condition          frame->Xcondition
536  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
537    
 #define original_ims       frame->Xoriginal_ims  
   
538  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
539  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
540  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
541  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
542  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
543  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
544  #endif  #endif
# Line 536  i, and fc and c, can be the same variabl Line 568  i, and fc and c, can be the same variabl
568  #define fi i  #define fi i
569  #define fc c  #define fc c
570    
571    /* Many of the following variables are used only in small blocks of the code.
572    My normal style of coding would have declared them within each of those blocks.
573    However, in order to accommodate the version of this code that uses an external
574    "stack" implemented on the heap, it is easier to declare them all here, so the
575    declarations can be cut out in a block. The only declarations within blocks
576    below are for variables that do not have to be preserved over a recursive call
577    to RMATCH(). */
578    
579    #ifdef SUPPORT_UTF
580    const pcre_uchar *charptr;
581    #endif
582    const pcre_uchar *callpat;
583    const pcre_uchar *data;
584    const pcre_uchar *next;
585    PCRE_PUCHAR       pp;
586    const pcre_uchar *prev;
587    PCRE_PUCHAR       saved_eptr;
588    
589  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
590  const uschar *charptr;             /* in small blocks of the code. My normal */  
591  #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().          */  
592  BOOL condition;  BOOL condition;
593  BOOL prev_is_word;  BOOL prev_is_word;
594    
 unsigned long int original_ims;  
   
595  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
596  int prop_type;  int prop_type;
597  int prop_value;  int prop_value;
598  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
599  int oclength;  int oclength;
600  uschar occhars[8];  pcre_uchar occhars[6];
601  #endif  #endif
602    
603  int codelink;  int codelink;
# Line 581  int stacksave[REC_STACK_SAVE_MAX]; Line 615  int stacksave[REC_STACK_SAVE_MAX];
615  eptrblock newptrb;  eptrblock newptrb;
616  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
617    
618    /* To save space on the stack and in the heap frame, I have doubled up on some
619    of the local variables that are used only in localised parts of the code, but
620    still need to be preserved over recursive calls of match(). These macros define
621    the alternative names that are used. */
622    
623    #define allow_zero    cur_is_word
624    #define cbegroup      condition
625    #define code_offset   codelink
626    #define condassert    condition
627    #define matched_once  prev_is_word
628    #define foc           number
629    
630  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
631  variables. */  variables. */
632    
# Line 600  TAIL_RECURSE: Line 646  TAIL_RECURSE:
646  /* 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
647  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
648  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()
649  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
650  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
651  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,
652  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
653    
654  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
655  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
656  #else  #else
657  utf8 = FALSE;  utf = FALSE;
658  #endif  #endif
659    
660  /* 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 617  haven't exceeded the recursive call limi Line 663  haven't exceeded the recursive call limi
663  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
664  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
665    
 original_ims = ims;    /* Save for resetting on ')' */  
   
666  /* 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
667  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
668  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
669  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
670  When match() is called in other circumstances, don't add to the chain. The  
671  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
672  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
673  match(). */  to break infinite loops that match no characters. When match() is called in
674    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
675    NOT be used with tail recursion, because the memory block that is used is on
676    the stack, so a new one may be required for each match(). */
677    
678  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
679    {    {
680    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
681    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
682    eptrb = &newptrb;    eptrb = &newptrb;
683      md->match_function_type = 0;
684    }    }
685    
686  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 642  for (;;) Line 690  for (;;)
690    minimize = possessive = FALSE;    minimize = possessive = FALSE;
691    op = *ecode;    op = *ecode;
692    
   /* 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;  
   
693    switch(op)    switch(op)
694      {      {
695        case OP_MARK:
696        md->nomatch_mark = ecode + 2;
697        md->mark = NULL;    /* In case previously set by assertion */
698        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
699          eptrb, RM55);
700        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
701             md->mark == NULL) md->mark = ecode + 2;
702    
703        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
704        argument, and we must check whether that argument matches this MARK's
705        argument. It is passed back in md->start_match_ptr (an overloading of that
706        variable). If it does match, we reset that variable to the current subject
707        position and return MATCH_SKIP. Otherwise, pass back the return code
708        unaltered. */
709    
710        else if (rrc == MATCH_SKIP_ARG &&
711            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
712          {
713          md->start_match_ptr = eptr;
714          RRETURN(MATCH_SKIP);
715          }
716        RRETURN(rrc);
717    
718      case OP_FAIL:      case OP_FAIL:
719      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
720    
721      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);  
722    
723      case OP_COMMIT:      case OP_COMMIT:
724      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
725        ims, eptrb, flags, RM52);        eptrb, RM52);
726      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
727            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
728            rrc != MATCH_THEN)
729          RRETURN(rrc);
730      RRETURN(MATCH_COMMIT);      RRETURN(MATCH_COMMIT);
731    
732        /* PRUNE overrides THEN */
733    
734        case OP_PRUNE:
735        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
736          eptrb, RM51);
737        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
738        RRETURN(MATCH_PRUNE);
739    
740        case OP_PRUNE_ARG:
741        md->nomatch_mark = ecode + 2;
742        md->mark = NULL;    /* In case previously set by assertion */
743        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
744          eptrb, RM56);
745        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
746             md->mark == NULL) md->mark = ecode + 2;
747        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
748        RRETURN(MATCH_PRUNE);
749    
750        /* SKIP overrides PRUNE and THEN */
751    
752      case OP_SKIP:      case OP_SKIP:
753      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
754        ims, eptrb, flags, RM53);        eptrb, RM53);
755      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
756          RRETURN(rrc);
757      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
758      RRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
759    
760        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
761        nomatch_mark. There is a flag that disables this opcode when re-matching a
762        pattern that ended with a SKIP for which there was not a matching MARK. */
763    
764        case OP_SKIP_ARG:
765        if (md->ignore_skip_arg)
766          {
767          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
768          break;
769          }
770        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
771          eptrb, RM57);
772        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
773          RRETURN(rrc);
774    
775        /* Pass back the current skip name by overloading md->start_match_ptr and
776        returning the special MATCH_SKIP_ARG return code. This will either be
777        caught by a matching MARK, or get to the top, where it causes a rematch
778        with the md->ignore_skip_arg flag set. */
779    
780        md->start_match_ptr = ecode + 2;
781        RRETURN(MATCH_SKIP_ARG);
782    
783        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
784        the branch in which it occurs can be determined. Overload the start of
785        match pointer to do this. */
786    
787      case OP_THEN:      case OP_THEN:
788      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
789        ims, eptrb, flags, RM54);        eptrb, RM54);
790        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
791        md->start_match_ptr = ecode;
792        RRETURN(MATCH_THEN);
793    
794        case OP_THEN_ARG:
795        md->nomatch_mark = ecode + 2;
796        md->mark = NULL;    /* In case previously set by assertion */
797        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
798          md, eptrb, RM58);
799        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
800             md->mark == NULL) md->mark = ecode + 2;
801      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
802        md->start_match_ptr = ecode;
803      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
804    
805      /* Handle a capturing bracket. If there is space in the offset vector, save      /* Handle an atomic group that does not contain any capturing parentheses.
806      the current subject position in the working slot at the top of the vector.      This can be handled like an assertion. Prior to 8.13, all atomic groups
807      We mustn't change the current values of the data slot, because they may be      were handled this way. In 8.13, the code was changed as below for ONCE, so
808      set from a previous iteration of this group, and be referred to by a      that backups pass through the group and thereby reset captured values.
809      reference inside the group.      However, this uses a lot more stack, so in 8.20, atomic groups that do not
810        contain any captures generate OP_ONCE_NC, which can be handled in the old,
811      If the bracket fails to match, we need to restore this value and also the      less stack intensive way.
812      values of the final offsets, in case they were set by a previous iteration  
813      of the same bracket.      Check the alternative branches in turn - the matching won't pass the KET
814        for this kind of subpattern. If any one branch matches, we carry on as at
815        the end of a normal bracket, leaving the subject pointer, but resetting
816        the start-of-match value in case it was changed by \K. */
817    
818        case OP_ONCE_NC:
819        prev = ecode;
820        saved_eptr = eptr;
821        do
822          {
823          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
824          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
825            {
826            mstart = md->start_match_ptr;
827            break;
828            }
829          if (rrc == MATCH_THEN)
830            {
831            next = ecode + GET(ecode,1);
832            if (md->start_match_ptr < next &&
833                (*ecode == OP_ALT || *next == OP_ALT))
834              rrc = MATCH_NOMATCH;
835            }
836    
837          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
838          ecode += GET(ecode,1);
839          }
840        while (*ecode == OP_ALT);
841    
842        /* If hit the end of the group (which could be repeated), fail */
843    
844        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
845    
846        /* Continue as from after the group, updating the offsets high water
847        mark, since extracts may have been taken. */
848    
849        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
850    
851        offset_top = md->end_offset_top;
852        eptr = md->end_match_ptr;
853    
854        /* For a non-repeating ket, just continue at this level. This also
855        happens for a repeating ket if no characters were matched in the group.
856        This is the forcible breaking of infinite loops as implemented in Perl
857        5.005. */
858    
859        if (*ecode == OP_KET || eptr == saved_eptr)
860          {
861          ecode += 1+LINK_SIZE;
862          break;
863          }
864    
865        /* The repeating kets try the rest of the pattern or restart from the
866        preceding bracket, in the appropriate order. The second "call" of match()
867        uses tail recursion, to avoid using another stack frame. */
868    
869        if (*ecode == OP_KETRMIN)
870          {
871          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
872          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
873          ecode = prev;
874          goto TAIL_RECURSE;
875          }
876        else  /* OP_KETRMAX */
877          {
878          md->match_function_type = MATCH_CBEGROUP;
879          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
880          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
881          ecode += 1 + LINK_SIZE;
882          goto TAIL_RECURSE;
883          }
884        /* Control never gets here */
885    
886        /* Handle a capturing bracket, other than those that are possessive with an
887        unlimited repeat. If there is space in the offset vector, save the current
888        subject position in the working slot at the top of the vector. We mustn't
889        change the current values of the data slot, because they may be set from a
890        previous iteration of this group, and be referred to by a reference inside
891        the group. A failure to match might occur after the group has succeeded,
892        if something later on doesn't match. For this reason, we need to restore
893        the working value and also the values of the final offsets, in case they
894        were set by a previous iteration of the same bracket.
895    
896      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
897      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 699  for (;;) Line 902  for (;;)
902      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
903      offset = number << 1;      offset = number << 1;
904    
905  #ifdef DEBUG  #ifdef PCRE_DEBUG
906      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
907      printf("subject=");      printf("subject=");
908      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 714  for (;;) Line 917  for (;;)
917        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
918    
919        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
920        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
921            (int)(eptr - md->start_subject);
922    
923        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
924          {          {
925          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
926            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
927          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
928            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
929    
930            /* If we backed up to a THEN, check whether it is within the current
931            branch by comparing the address of the THEN that is passed back with
932            the end of the branch. If it is within the current branch, and the
933            branch is one of two or more alternatives (it either starts or ends
934            with OP_ALT), we have reached the limit of THEN's action, so convert
935            the return code to NOMATCH, which will cause normal backtracking to
936            happen from now on. Otherwise, THEN is passed back to an outer
937            alternative. This implements Perl's treatment of parenthesized groups,
938            where a group not containing | does not affect the current alternative,
939            that is, (X) is NOT the same as (X|(*F)). */
940    
941            if (rrc == MATCH_THEN)
942              {
943              next = ecode + GET(ecode,1);
944              if (md->start_match_ptr < next &&
945                  (*ecode == OP_ALT || *next == OP_ALT))
946                rrc = MATCH_NOMATCH;
947              }
948    
949            /* Anything other than NOMATCH is passed back. */
950    
951            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
952          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
953          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
954            if (*ecode != OP_ALT) break;
955          }          }
       while (*ecode == OP_ALT);  
956    
957        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
958        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
959        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
960        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
961    
962        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
963    
964          RRETURN(rrc);
965        }        }
966    
967      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 747  for (;;) Line 975  for (;;)
975      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
976      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
977    
978      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
979      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
980      recursive call to match() whatever happened. We can reduce stack usage by  
981      turning this into a tail recursion, except in the case when match_cbegroup      When we get to the final alternative within the brackets, we used to return
982      is set.*/      the result of a recursive call to match() whatever happened so it was
983        possible to reduce stack usage by turning this into a tail recursion,
984        except in the case of a possibly empty group. However, now that there is
985        the possiblity of (*THEN) occurring in the final alternative, this
986        optimization is no longer always possible.
987    
988        We can optimize if we know there are no (*THEN)s in the pattern; at present
989        this is the best that can be done.
990    
991        MATCH_ONCE is returned when the end of an atomic group is successfully
992        reached, but subsequent matching fails. It passes back up the tree (causing
993        captured values to be reset) until the original atomic group level is
994        reached. This is tested by comparing md->once_target with the start of the
995        group. At this point, the return is converted into MATCH_NOMATCH so that
996        previous backup points can be taken. */
997    
998        case OP_ONCE:
999      case OP_BRA:      case OP_BRA:
1000      case OP_SBRA:      case OP_SBRA:
1001      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1002      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1003      for (;;)      for (;;)
1004        {        {
1005        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1006    
1007          /* If this is not a possibly empty group, and there are no (*THEN)s in
1008          the pattern, and this is the final alternative, optimize as described
1009          above. */
1010    
1011          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1012            {
1013            ecode += PRIV(OP_lengths)[*ecode];
1014            goto TAIL_RECURSE;
1015            }
1016    
1017          /* In all other cases, we have to make another call to match(). */
1018    
1019          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1020            RM2);
1021    
1022          /* See comment in the code for capturing groups above about handling
1023          THEN. */
1024    
1025          if (rrc == MATCH_THEN)
1026            {
1027            next = ecode + GET(ecode,1);
1028            if (md->start_match_ptr < next &&
1029                (*ecode == OP_ALT || *next == OP_ALT))
1030              rrc = MATCH_NOMATCH;
1031            }
1032    
1033          if (rrc != MATCH_NOMATCH)
1034          {          {
1035          if (flags == 0)    /* Not a possibly empty group */          if (rrc == MATCH_ONCE)
1036            {            {
1037            ecode += _pcre_OP_lengths[*ecode];            const pcre_uchar *scode = ecode;
1038            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
1039            goto TAIL_RECURSE;              {
1040                while (*scode == OP_ALT) scode += GET(scode, 1);
1041                scode -= GET(scode, 1);
1042                }
1043              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1044              }
1045            RRETURN(rrc);
1046            }
1047          ecode += GET(ecode, 1);
1048          if (*ecode != OP_ALT) break;
1049          }
1050    
1051        RRETURN(MATCH_NOMATCH);
1052    
1053        /* Handle possessive capturing brackets with an unlimited repeat. We come
1054        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1055        handled similarly to the normal case above. However, the matching is
1056        different. The end of these brackets will always be OP_KETRPOS, which
1057        returns MATCH_KETRPOS without going further in the pattern. By this means
1058        we can handle the group by iteration rather than recursion, thereby
1059        reducing the amount of stack needed. */
1060    
1061        case OP_CBRAPOS:
1062        case OP_SCBRAPOS:
1063        allow_zero = FALSE;
1064    
1065        POSSESSIVE_CAPTURE:
1066        number = GET2(ecode, 1+LINK_SIZE);
1067        offset = number << 1;
1068    
1069    #ifdef PCRE_DEBUG
1070        printf("start possessive bracket %d\n", number);
1071        printf("subject=");
1072        pchars(eptr, 16, TRUE, md);
1073        printf("\n");
1074    #endif
1075    
1076        if (offset < md->offset_max)
1077          {
1078          matched_once = FALSE;
1079          code_offset = (int)(ecode - md->start_code);
1080    
1081          save_offset1 = md->offset_vector[offset];
1082          save_offset2 = md->offset_vector[offset+1];
1083          save_offset3 = md->offset_vector[md->offset_end - number];
1084          save_capture_last = md->capture_last;
1085    
1086          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1087    
1088          /* Each time round the loop, save the current subject position for use
1089          when the group matches. For MATCH_MATCH, the group has matched, so we
1090          restart it with a new subject starting position, remembering that we had
1091          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1092          usual. If we haven't matched any alternatives in any iteration, check to
1093          see if a previous iteration matched. If so, the group has matched;
1094          continue from afterwards. Otherwise it has failed; restore the previous
1095          capture values before returning NOMATCH. */
1096    
1097          for (;;)
1098            {
1099            md->offset_vector[md->offset_end - number] =
1100              (int)(eptr - md->start_subject);
1101            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1102            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1103              eptrb, RM63);
1104            if (rrc == MATCH_KETRPOS)
1105              {
1106              offset_top = md->end_offset_top;
1107              eptr = md->end_match_ptr;
1108              ecode = md->start_code + code_offset;
1109              save_capture_last = md->capture_last;
1110              matched_once = TRUE;
1111              continue;
1112              }
1113    
1114            /* See comment in the code for capturing groups above about handling
1115            THEN. */
1116    
1117            if (rrc == MATCH_THEN)
1118              {
1119              next = ecode + GET(ecode,1);
1120              if (md->start_match_ptr < next &&
1121                  (*ecode == OP_ALT || *next == OP_ALT))
1122                rrc = MATCH_NOMATCH;
1123            }            }
1124    
1125          /* Possibly empty group; can't use tail recursion. */          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1126            md->capture_last = save_capture_last;
1127            ecode += GET(ecode, 1);
1128            if (*ecode != OP_ALT) break;
1129            }
1130    
1131          if (!matched_once)
1132            {
1133            md->offset_vector[offset] = save_offset1;
1134            md->offset_vector[offset+1] = save_offset2;
1135            md->offset_vector[md->offset_end - number] = save_offset3;
1136            }
1137    
1138          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        if (allow_zero || matched_once)
1139            eptrb, flags, RM48);          {
1140          RRETURN(rrc);          ecode += 1 + LINK_SIZE;
1141            break;
1142          }          }
1143    
1144        /* For non-final alternatives, continue the loop for a NOMATCH result;        RRETURN(MATCH_NOMATCH);
1145        otherwise return. */        }
1146    
1147        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1148        as a non-capturing bracket. */
1149    
1150        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1151          eptrb, flags, RM2);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1152        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
1153        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1154    
1155        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1156        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1157    
1158        /* Non-capturing possessive bracket with unlimited repeat. We come here
1159        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1160        without the capturing complication. It is written out separately for speed
1161        and cleanliness. */
1162    
1163        case OP_BRAPOS:
1164        case OP_SBRAPOS:
1165        allow_zero = FALSE;
1166    
1167        POSSESSIVE_NON_CAPTURE:
1168        matched_once = FALSE;
1169        code_offset = (int)(ecode - md->start_code);
1170    
1171        for (;;)
1172          {
1173          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1174          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1175            eptrb, RM48);
1176          if (rrc == MATCH_KETRPOS)
1177            {
1178            offset_top = md->end_offset_top;
1179            eptr = md->end_match_ptr;
1180            ecode = md->start_code + code_offset;
1181            matched_once = TRUE;
1182            continue;
1183            }
1184    
1185          /* See comment in the code for capturing groups above about handling
1186          THEN. */
1187    
1188          if (rrc == MATCH_THEN)
1189            {
1190            next = ecode + GET(ecode,1);
1191            if (md->start_match_ptr < next &&
1192                (*ecode == OP_ALT || *next == OP_ALT))
1193              rrc = MATCH_NOMATCH;
1194            }
1195    
1196          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1197        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1198          if (*ecode != OP_ALT) break;
1199          }
1200    
1201        if (matched_once || allow_zero)
1202          {
1203          ecode += 1 + LINK_SIZE;
1204          break;
1205        }        }
1206        RRETURN(MATCH_NOMATCH);
1207    
1208      /* Control never reaches here. */      /* Control never reaches here. */
1209    
1210      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1211      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1212      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1213      exactly what going to the ket would do. As there is only one branch to be      exactly what going to the ket would do. */
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1214    
1215      case OP_COND:      case OP_COND:
1216      case OP_SCOND:      case OP_SCOND:
1217      codelink= GET(ecode, 1);      codelink = GET(ecode, 1);
1218    
1219      /* Because of the way auto-callout works during compile, a callout item is      /* Because of the way auto-callout works during compile, a callout item is
1220      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
1221    
1222      if (ecode[LINK_SIZE+1] == OP_CALLOUT)      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1223        {        {
1224        if (pcre_callout != NULL)        if (PUBL(callout) != NULL)
1225          {          {
1226          pcre_callout_block cb;          pcre_callout_block cb;
1227          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1228          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1229          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1230          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
1231          cb.subject_length   = md->end_subject - md->start_subject;          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1232          cb.start_match      = mstart - md->start_subject;          cb.start_match      = (int)(mstart - md->start_subject);
1233          cb.current_position = eptr - md->start_subject;          cb.current_position = (int)(eptr - md->start_subject);
1234          cb.pattern_position = GET(ecode, LINK_SIZE + 3);          cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1235          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1236          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1237          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1238          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1239          if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);          cb.mark             = md->nomatch_mark;
1240            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1241          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1242          }          }
1243        ecode += _pcre_OP_lengths[OP_CALLOUT];        ecode += PRIV(OP_lengths)[OP_CALLOUT];
1244        }        }
1245    
1246      condcode = ecode[LINK_SIZE+1];      condcode = ecode[LINK_SIZE+1];
1247    
1248      /* Now see what the actual condition is */      /* Now see what the actual condition is */
1249    
1250      if (condcode == OP_RREF)         /* Recursion test */      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1251        {        {
1252        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/        if (md->recursive == NULL)                /* Not recursing => FALSE */
1253        condition = md->recursive != NULL &&          {
1254          (offset == RREF_ANY || offset == md->recursive->group_num);          condition = FALSE;
1255        ecode += condition? 3 : GET(ecode, 1);          ecode += GET(ecode, 1);
1256            }
1257          else
1258            {
1259            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1260            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1261    
1262            /* If the test is for recursion into a specific subpattern, and it is
1263            false, but the test was set up by name, scan the table to see if the
1264            name refers to any other numbers, and test them. The condition is true
1265            if any one is set. */
1266    
1267            if (!condition && condcode == OP_NRREF)
1268              {
1269              pcre_uchar *slotA = md->name_table;
1270              for (i = 0; i < md->name_count; i++)
1271                {
1272                if (GET2(slotA, 0) == recno) break;
1273                slotA += md->name_entry_size;
1274                }
1275    
1276              /* Found a name for the number - there can be only one; duplicate
1277              names for different numbers are allowed, but not vice versa. First
1278              scan down for duplicates. */
1279    
1280              if (i < md->name_count)
1281                {
1282                pcre_uchar *slotB = slotA;
1283                while (slotB > md->name_table)
1284                  {
1285                  slotB -= md->name_entry_size;
1286                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1287                    {
1288                    condition = GET2(slotB, 0) == md->recursive->group_num;
1289                    if (condition) break;
1290                    }
1291                  else break;
1292                  }
1293    
1294                /* Scan up for duplicates */
1295    
1296                if (!condition)
1297                  {
1298                  slotB = slotA;
1299                  for (i++; i < md->name_count; i++)
1300                    {
1301                    slotB += md->name_entry_size;
1302                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1303                      {
1304                      condition = GET2(slotB, 0) == md->recursive->group_num;
1305                      if (condition) break;
1306                      }
1307                    else break;
1308                    }
1309                  }
1310                }
1311              }
1312    
1313            /* Chose branch according to the condition */
1314    
1315            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1316            }
1317        }        }
1318    
1319      else if (condcode == OP_CREF)    /* Group used test */      else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1320        {        {
1321        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1322        condition = offset < offset_top && md->offset_vector[offset] >= 0;        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1323        ecode += condition? 3 : GET(ecode, 1);  
1324          /* If the numbered capture is unset, but the reference was by name,
1325          scan the table to see if the name refers to any other numbers, and test
1326          them. The condition is true if any one is set. This is tediously similar
1327          to the code above, but not close enough to try to amalgamate. */
1328    
1329          if (!condition && condcode == OP_NCREF)
1330            {
1331            int refno = offset >> 1;
1332            pcre_uchar *slotA = md->name_table;
1333    
1334            for (i = 0; i < md->name_count; i++)
1335              {
1336              if (GET2(slotA, 0) == refno) break;
1337              slotA += md->name_entry_size;
1338              }
1339    
1340            /* Found a name for the number - there can be only one; duplicate names
1341            for different numbers are allowed, but not vice versa. First scan down
1342            for duplicates. */
1343    
1344            if (i < md->name_count)
1345              {
1346              pcre_uchar *slotB = slotA;
1347              while (slotB > md->name_table)
1348                {
1349                slotB -= md->name_entry_size;
1350                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1351                  {
1352                  offset = GET2(slotB, 0) << 1;
1353                  condition = offset < offset_top &&
1354                    md->offset_vector[offset] >= 0;
1355                  if (condition) break;
1356                  }
1357                else break;
1358                }
1359    
1360              /* Scan up for duplicates */
1361    
1362              if (!condition)
1363                {
1364                slotB = slotA;
1365                for (i++; i < md->name_count; i++)
1366                  {
1367                  slotB += md->name_entry_size;
1368                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1369                    {
1370                    offset = GET2(slotB, 0) << 1;
1371                    condition = offset < offset_top &&
1372                      md->offset_vector[offset] >= 0;
1373                    if (condition) break;
1374                    }
1375                  else break;
1376                  }
1377                }
1378              }
1379            }
1380    
1381          /* Chose branch according to the condition */
1382    
1383          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1384        }        }
1385    
1386      else if (condcode == OP_DEF)     /* DEFINE - always false */      else if (condcode == OP_DEF)     /* DEFINE - always false */
# Line 847  for (;;) Line 1390  for (;;)
1390        }        }
1391    
1392      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1393      the final argument match_condassert causes it to stop at the end of an      md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1394      assertion. */      an assertion. */
1395    
1396      else      else
1397        {        {
1398        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1399            match_condassert, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1400        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1401          {          {
1402            if (md->end_offset_top > offset_top)
1403              offset_top = md->end_offset_top;  /* Captures may have happened */
1404          condition = TRUE;          condition = TRUE;
1405          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1406          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1407          }          }
1408    
1409          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1410          assertion; it is therefore treated as NOMATCH. */
1411    
1412        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1413          {          {
1414          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
# Line 871  for (;;) Line 1420  for (;;)
1420          }          }
1421        }        }
1422    
1423      /* We are now at the branch that is to be obeyed. As there is only one,      /* We are now at the branch that is to be obeyed. As there is only one, can
1424      we can use tail recursion to avoid using another stack frame, except when      use tail recursion to avoid using another stack frame, except when there is
1425      match_cbegroup is required for an unlimited repeat of a possibly empty      unlimited repeat of a possibly empty group. In the latter case, a recursive
1426      group. If the second alternative doesn't exist, we can just plough on. */      call to match() is always required, unless the second alternative doesn't
1427        exist, in which case we can just plough on. Note that, for compatibility
1428        with Perl, the | in a conditional group is NOT treated as creating two
1429        alternatives. If a THEN is encountered in the branch, it propagates out to
1430        the enclosing alternative (unless nested in a deeper set of alternatives,
1431        of course). */
1432    
1433      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1434        {        {
1435        ecode += 1 + LINK_SIZE;        if (op != OP_SCOND)
       if (op == OP_SCOND)        /* Possibly empty group */  
         {  
         RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);  
         RRETURN(rrc);  
         }  
       else                       /* Group must match something */  
1436          {          {
1437          flags = 0;          ecode += 1 + LINK_SIZE;
1438          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1439          }          }
1440    
1441          md->match_function_type = MATCH_CBEGROUP;
1442          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1443          RRETURN(rrc);
1444        }        }
1445      else                         /* Condition false & no alternative */  
1446         /* Condition false & no alternative; continue after the group. */
1447    
1448        else
1449        {        {
1450        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1451        }        }
1452      break;      break;
1453    
1454    
1455      /* End of the pattern, either real or forced. If we are in a top-level      /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1456      recursion, we should restore the offsets appropriately and continue from      to close any currently open capturing brackets. */
     after the call. */  
1457    
1458      case OP_ACCEPT:      case OP_CLOSE:
1459      case OP_END:      number = GET2(ecode, 1);
1460      if (md->recursive != NULL && md->recursive->group_num == 0)      offset = number << 1;
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       mstart = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1461    
1462      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty  #ifdef PCRE_DEBUG
1463      string - backtracking will then try other alternatives, if any. */        printf("end bracket %d at *ACCEPT", number);
1464          printf("\n");
1465    #endif
1466    
1467        md->capture_last = number;
1468        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1469          {
1470          md->offset_vector[offset] =
1471            md->offset_vector[md->offset_end - number];
1472          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1473          if (offset_top <= offset) offset_top = offset + 2;
1474          }
1475        ecode += 1 + IMM2_SIZE;
1476        break;
1477    
1478    
1479        /* End of the pattern, either real or forced. */
1480    
1481        case OP_END:
1482        case OP_ACCEPT:
1483        case OP_ASSERT_ACCEPT:
1484    
1485        /* If we have matched an empty string, fail if not in an assertion and not
1486        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1487        is set and we have matched at the start of the subject. In both cases,
1488        backtracking will then try other alternatives, if any. */
1489    
1490        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1491             md->recursive == NULL &&
1492             (md->notempty ||
1493               (md->notempty_atstart &&
1494                 mstart == md->start_subject + md->start_offset)))
1495          RRETURN(MATCH_NOMATCH);
1496    
1497        /* Otherwise, we have a match. */
1498    
     if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);  
1499      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1500      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1501      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
     RRETURN(MATCH_MATCH);  
1502    
1503      /* Change option settings */      /* For some reason, the macros don't work properly if an expression is
1504        given as the argument to RRETURN when the heap is in use. */
1505    
1506      case OP_OPT:      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1507      ims = ecode[1];      RRETURN(rrc);
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1508    
1509      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1510      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1511      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
1512      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
1513      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1514        of a condition, we want to return immediately afterwards. The caller of
1515        this incarnation of the match() function will have set MATCH_CONDASSERT in
1516        md->match_function type, and one of these opcodes will be the first opcode
1517        that is processed. We use a local variable that is preserved over calls to
1518        match() to remember this case. */
1519    
1520      case OP_ASSERT:      case OP_ASSERT:
1521      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1522        if (md->match_function_type == MATCH_CONDASSERT)
1523          {
1524          condassert = TRUE;
1525          md->match_function_type = 0;
1526          }
1527        else condassert = FALSE;
1528    
1529      do      do
1530        {        {
1531        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1532          RM4);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1533        if (rrc == MATCH_MATCH) break;          {
1534            mstart = md->start_match_ptr;   /* In case \K reset it */
1535            break;
1536            }
1537    
1538          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1539          as NOMATCH. */
1540    
1541        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1542        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1543        }        }
1544      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1545    
1546      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1547    
1548      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1549    
1550      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1551    
1552      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1553      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 964  for (;;) Line 1557  for (;;)
1557      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1558      continue;      continue;
1559    
1560      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1561        PRUNE, or COMMIT means we must assume failure without checking subsequent
1562        branches. */
1563    
1564      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1565      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1566        if (md->match_function_type == MATCH_CONDASSERT)
1567          {
1568          condassert = TRUE;
1569          md->match_function_type = 0;
1570          }
1571        else condassert = FALSE;
1572    
1573      do      do
1574        {        {
1575        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1576          RM5);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1577        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1578            {
1579            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1580            break;
1581            }
1582    
1583          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1584          as NOMATCH. */
1585    
1586        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1587        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1588        }        }
1589      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1590    
1591      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1592    
1593      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1594      continue;      continue;
# Line 989  for (;;) Line 1599  for (;;)
1599      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1600    
1601      case OP_REVERSE:      case OP_REVERSE:
1602  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1603      if (utf8)      if (utf)
1604        {        {
1605        i = GET(ecode, 1);        i = GET(ecode, 1);
1606        while (i-- > 0)        while (i-- > 0)
# Line 1010  for (;;) Line 1620  for (;;)
1620        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1621        }        }
1622    
1623      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1624    
1625        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1626      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1627      break;      break;
1628    
# Line 1020  for (;;) Line 1631  for (;;)
1631      function is able to force a failure. */      function is able to force a failure. */
1632    
1633      case OP_CALLOUT:      case OP_CALLOUT:
1634      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1635        {        {
1636        pcre_callout_block cb;        pcre_callout_block cb;
1637        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1638        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1639        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1640        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1641        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1642        cb.start_match      = mstart - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1643        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1644        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1645        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1646        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1647        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1648        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1649        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1650          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1651        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1652        }        }
1653      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 1045  for (;;) Line 1657  for (;;)
1657      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
1658      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1659    
1660      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1661      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
1662      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
1663      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
1664      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
1665      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
1666      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.  
1667    
1668      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
1669      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
1670      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1671        a lot, so he is not to blame for the current way it works. */
1672    
1673      case OP_RECURSE:      case OP_RECURSE:
1674        {        {
1675          recursion_info *ri;
1676          int recno;
1677    
1678        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1679        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1680          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1681    
1682          /* Check for repeating a recursion without advancing the subject pointer.
1683          This should catch convoluted mutual recursions. (Some simple cases are
1684          caught at compile time.) */
1685    
1686          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1687            if (recno == ri->group_num && eptr == ri->subject_position)
1688              RRETURN(PCRE_ERROR_RECURSELOOP);
1689    
1690        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1691    
1692          new_recursive.group_num = recno;
1693          new_recursive.subject_position = eptr;
1694        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1695        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1696    
1697        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1698    
1699        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1700    
1701        /* Now save the offset data. */        /* Now save the offset data */
1702    
1703        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1704        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1084  for (;;) Line 1706  for (;;)
1706        else        else
1707          {          {
1708          new_recursive.offset_save =          new_recursive.offset_save =
1709            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1710          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1711          }          }
   
1712        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1713              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1714    
1715        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1716        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1717          might be changed, so reset it before looping. */
1718    
1719        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1720        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1721        do        do
1722          {          {
1723          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1724            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1725          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1726            memcpy(md->offset_vector, new_recursive.offset_save,
1727                new_recursive.saved_max * sizeof(int));
1728            md->recursive = new_recursive.prevrec;
1729            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1730            {            {
1731            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1732            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1733              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1734            RRETURN(MATCH_MATCH);  
1735              /* Set where we got to in the subject, and reset the start in case
1736              it was changed by \K. This *is* propagated back out of a recursion,
1737              for Perl compatibility. */
1738    
1739              eptr = md->end_match_ptr;
1740              mstart = md->start_match_ptr;
1741              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1742            }            }
1743    
1744            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1745            as NOMATCH. */
1746    
1747          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1748            {            {
1749            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1750            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1751              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1752            RRETURN(rrc);            RRETURN(rrc);
1753            }            }
1754    
1755          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1756          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1757          }          }
1758        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1128  for (;;) Line 1760  for (;;)
1760        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1761        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1762        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1763          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1764        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1765        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(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));  
       }  
1766    
1767      if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1768        {      break;
       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 */  
1769    
1770      /* 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
1771      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1221  for (;;) Line 1781  for (;;)
1781      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1782    
1783      case OP_BRAZERO:      case OP_BRAZERO:
1784        {      next = ecode + 1;
1785        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1786        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1787        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1788        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1789      break;      break;
1790    
1791      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1792        {      next = ecode + 1;
1793        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1794        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1795        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1796        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1797      break;      break;
1798    
1799      case OP_SKIPZERO:      case OP_SKIPZERO:
1800        {      next = ecode+1;
1801        next = ecode+1;      do next += GET(next,1); while (*next == OP_ALT);
1802        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1803      break;      break;
1804    
1805        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1806        here; just jump to the group, with allow_zero set TRUE. */
1807    
1808        case OP_BRAPOSZERO:
1809        op = *(++ecode);
1810        allow_zero = TRUE;
1811        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1812          goto POSSESSIVE_NON_CAPTURE;
1813    
1814      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1815    
1816      case OP_KET:      case OP_KET:
1817      case OP_KETRMIN:      case OP_KETRMIN:
1818      case OP_KETRMAX:      case OP_KETRMAX:
1819        case OP_KETRPOS:
1820      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1821    
1822      /* 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
1823      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1824      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1825    
1826      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1827        {        {
1828        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1829        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1830        }        }
1831      else saved_eptr = NULL;      else saved_eptr = NULL;
1832    
1833      /* 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
1834      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
1835      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1836        start in case it was changed by \K. */
1837      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1838          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1839          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1840        {        {
1841        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1842        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1843        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1844          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1845        }        }
1846    
1847      /* 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
1848      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1849      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
1850      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
1851      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
1852        the current subject position and start match pointer and give a MATCH
1853        return. */
1854    
1855      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1856            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1857        {        {
1858        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1859        offset = number << 1;        offset = number << 1;
1860    
1861  #ifdef DEBUG  #ifdef PCRE_DEBUG
1862        printf("end bracket %d", number);        printf("end bracket %d", number);
1863        printf("\n");        printf("\n");
1864  #endif  #endif
1865    
1866          /* Handle a recursively called group. */
1867    
1868          if (md->recursive != NULL && md->recursive->group_num == number)
1869            {
1870            md->end_match_ptr = eptr;
1871            md->start_match_ptr = mstart;
1872            RRETURN(MATCH_MATCH);
1873            }
1874    
1875          /* Deal with capturing */
1876    
1877        md->capture_last = number;        md->capture_last = number;
1878        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1879          {          {
1880            /* If offset is greater than offset_top, it means that we are
1881            "skipping" a capturing group, and that group's offsets must be marked
1882            unset. In earlier versions of PCRE, all the offsets were unset at the
1883            start of matching, but this doesn't work because atomic groups and
1884            assertions can cause a value to be set that should later be unset.
1885            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1886            part of the atomic group, but this is not on the final matching path,
1887            so must be unset when 2 is set. (If there is no group 2, there is no
1888            problem, because offset_top will then be 2, indicating no capture.) */
1889    
1890            if (offset > offset_top)
1891              {
1892              register int *iptr = md->offset_vector + offset_top;
1893              register int *iend = md->offset_vector + offset;
1894              while (iptr < iend) *iptr++ = -1;
1895              }
1896    
1897            /* Now make the extraction */
1898    
1899          md->offset_vector[offset] =          md->offset_vector[offset] =
1900            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1901          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1902          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1903          }          }
1904          }
1905    
1906        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1907        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1908        group. This is the forcible breaking of infinite loops as implemented in
1909        Perl 5.005. For a non-repeating atomic group that includes captures,
1910        establish a backup point by processing the rest of the pattern at a lower
1911        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1912        original OP_ONCE level, thereby bypassing intermediate backup points, but
1913        resetting any captures that happened along the way. */
1914    
1915        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1916          {
1917          if (*prev == OP_ONCE)
1918          {          {
1919          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1920          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1921          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1922          mstart = rec->save_start;          RRETURN(MATCH_ONCE);
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
1923          }          }
1924          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1925          break;
1926        }        }
1927    
1928      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1929      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1930        at a time from the outer level, thus saving stack. */
     ims = original_ims;  
     DPRINTF(("ims reset to %02lx\n", ims));  
1931    
1932      /* For a non-repeating ket, just continue at this level. This also      if (*ecode == OP_KETRPOS)
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
1933        {        {
1934        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1935        break;        md->end_offset_top = offset_top;
1936          RRETURN(MATCH_KETRPOS);
1937        }        }
1938    
1939      /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
1940      preceding bracket, in the appropriate order. In the second case, we can use      the preceding bracket, in the appropriate order. In the second case, we can
1941      tail recursion to avoid using another stack frame, unless we have an      use tail recursion to avoid using another stack frame, unless we have an
1942      unlimited repeat of a group that can match an empty string. */      an atomic group or an unlimited repeat of a group that can match an empty
1943        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
1944    
1945      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1946        {        {
1947        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1948        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1949        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
1950            {
1951            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1952            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1953            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1954            RRETURN(MATCH_ONCE);
1955            }
1956          if (*prev >= OP_SBRA)    /* Could match an empty string */
1957          {          {
1958          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          md->match_function_type = MATCH_CBEGROUP;
1959            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1960          RRETURN(rrc);          RRETURN(rrc);
1961          }          }
1962        ecode = prev;        ecode = prev;
# Line 1360  for (;;) Line 1964  for (;;)
1964        }        }
1965      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1966        {        {
1967        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1968          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1969          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1970        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1971          if (*prev == OP_ONCE)
1972            {
1973            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1974            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1975            md->once_target = prev;
1976            RRETURN(MATCH_ONCE);
1977            }
1978        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1979        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1980        }        }
1981      /* Control never gets here */      /* Control never gets here */
1982    
1983      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1984    
1985      case OP_CIRC:      case OP_CIRC:
1986      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 */  
1987    
1988      /* Start of subject assertion */      /* Start of subject assertion */
1989    
# Line 1389  for (;;) Line 1992  for (;;)
1992      ecode++;      ecode++;
1993      break;      break;
1994    
1995        /* Multiline mode: start of subject unless notbol, or after any newline. */
1996    
1997        case OP_CIRCM:
1998        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
1999        if (eptr != md->start_subject &&
2000            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2001          RRETURN(MATCH_NOMATCH);
2002        ecode++;
2003        break;
2004    
2005      /* Start of match assertion */      /* Start of match assertion */
2006    
2007      case OP_SOM:      case OP_SOM:
# Line 1403  for (;;) Line 2016  for (;;)
2016      ecode++;      ecode++;
2017      break;      break;
2018    
2019      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2020      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2021    
2022      case OP_DOLL:      case OP_DOLLM:
2023      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2024        {        { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
       if (eptr < md->end_subject)  
         { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
       else  
         { if (md->noteol) RRETURN(MATCH_NOMATCH); }  
       ecode++;  
       break;  
       }  
2025      else      else
2026        {        {
2027        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2028        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2029        }        }
2030        ecode++;
2031        break;
2032    
2033        /* Not multiline mode: assert before a terminating newline or before end of
2034        subject unless noteol is set. */
2035    
2036        case OP_DOLL:
2037        if (md->noteol) RRETURN(MATCH_NOMATCH);
2038        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2039    
2040      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2041    
2042      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2043    
2044      case OP_EOD:      case OP_EOD:
2045      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2046        SCHECK_PARTIAL();
2047      ecode++;      ecode++;
2048      break;      break;
2049    
2050      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2051    
2052      case OP_EODN:      case OP_EODN:
2053      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2054        if (eptr < md->end_subject &&
2055          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2056        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2057    
2058        /* Either at end of string or \n before end. */
2059    
2060        SCHECK_PARTIAL();
2061      ecode++;      ecode++;
2062      break;      break;
2063    
# Line 1454  for (;;) Line 2069  for (;;)
2069    
2070        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2071        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
2072        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2073          partial matching. */
2074    
2075  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2076        if (utf8)        if (utf)
2077          {          {
2078            /* Get status of previous character */
2079    
2080          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2081            {            {
2082            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2083            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2084              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2085            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2086    #ifdef SUPPORT_UCP
2087              if (md->use_ucp)
2088                {
2089                if (c == '_') prev_is_word = TRUE; else
2090                  {
2091                  int cat = UCD_CATEGORY(c);
2092                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2093                  }
2094                }
2095              else
2096    #endif
2097            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2098            }            }
2099          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2100            /* Get status of next character */
2101    
2102            if (eptr >= md->end_subject)
2103              {
2104              SCHECK_PARTIAL();
2105              cur_is_word = FALSE;
2106              }
2107            else
2108            {            {
2109            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2110    #ifdef SUPPORT_UCP
2111              if (md->use_ucp)
2112                {
2113                if (c == '_') cur_is_word = TRUE; else
2114                  {
2115                  int cat = UCD_CATEGORY(c);
2116                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2117                  }
2118                }
2119              else
2120    #endif
2121            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2122            }            }
2123          }          }
2124        else        else
2125  #endif  #endif
2126    
2127        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2128          consistency with the behaviour of \w we do use it in this case. */
2129    
2130          {          {
2131          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2132            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2133          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2134            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2135              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2136    #ifdef SUPPORT_UCP
2137              if (md->use_ucp)
2138                {
2139                c = eptr[-1];
2140                if (c == '_') prev_is_word = TRUE; else
2141                  {
2142                  int cat = UCD_CATEGORY(c);
2143                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2144                  }
2145                }
2146              else
2147    #endif
2148              prev_is_word = MAX_255(eptr[-1])
2149                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2150              }
2151    
2152            /* Get status of next character */
2153    
2154            if (eptr >= md->end_subject)
2155              {
2156              SCHECK_PARTIAL();
2157              cur_is_word = FALSE;
2158              }
2159            else
2160    #ifdef SUPPORT_UCP
2161            if (md->use_ucp)
2162              {
2163              c = *eptr;
2164              if (c == '_') cur_is_word = TRUE; else
2165                {
2166                int cat = UCD_CATEGORY(c);
2167                cur_is_word = (cat == ucp_L || cat == ucp_N);
2168                }
2169              }
2170            else
2171    #endif
2172            cur_is_word = MAX_255(*eptr)
2173              && ((md->ctypes[*eptr] & ctype_word) != 0);
2174          }          }
2175    
2176        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1499  for (;;) Line 2188  for (;;)
2188      /* Fall through */      /* Fall through */
2189    
2190      case OP_ALLANY:      case OP_ALLANY:
2191      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2192      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;        {                            /* not be updated before SCHECK_PARTIAL. */
2193          SCHECK_PARTIAL();
2194          RRETURN(MATCH_NOMATCH);
2195          }
2196        eptr++;
2197    #ifdef SUPPORT_UTF
2198        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2199    #endif
2200      ecode++;      ecode++;
2201      break;      break;
2202    
# Line 1508  for (;;) Line 2204  for (;;)
2204      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2205    
2206      case OP_ANYBYTE:      case OP_ANYBYTE:
2207      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2208          {                            /* not be updated before SCHECK_PARTIAL. */
2209          SCHECK_PARTIAL();
2210          RRETURN(MATCH_NOMATCH);
2211          }
2212        eptr++;
2213      ecode++;      ecode++;
2214      break;      break;
2215    
2216      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2217      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2218          {
2219          SCHECK_PARTIAL();
2220          RRETURN(MATCH_NOMATCH);
2221          }
2222      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2223      if (      if (
2224  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2225         c < 256 &&         c < 256 &&
2226  #endif  #endif
2227         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1526  for (;;) Line 2231  for (;;)
2231      break;      break;
2232    
2233      case OP_DIGIT:      case OP_DIGIT:
2234      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2235          {
2236          SCHECK_PARTIAL();
2237          RRETURN(MATCH_NOMATCH);
2238          }
2239      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2240      if (      if (
2241  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2242         c >= 256 ||         c > 255 ||
2243  #endif  #endif
2244         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2245         )         )
# Line 1539  for (;;) Line 2248  for (;;)
2248      break;      break;
2249    
2250      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2251      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2252          {
2253          SCHECK_PARTIAL();
2254          RRETURN(MATCH_NOMATCH);
2255          }
2256      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2257      if (      if (
2258  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2259         c < 256 &&         c < 256 &&
2260  #endif  #endif
2261         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1552  for (;;) Line 2265  for (;;)
2265      break;      break;
2266    
2267      case OP_WHITESPACE:      case OP_WHITESPACE:
2268      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2269          {
2270          SCHECK_PARTIAL();
2271          RRETURN(MATCH_NOMATCH);
2272          }
2273      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2274      if (      if (
2275  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2276         c >= 256 ||         c > 255 ||
2277  #endif  #endif
2278         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2279         )         )
# Line 1565  for (;;) Line 2282  for (;;)
2282      break;      break;
2283    
2284      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2285      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2286          {
2287          SCHECK_PARTIAL();
2288          RRETURN(MATCH_NOMATCH);
2289          }
2290      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2291      if (      if (
2292  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2293         c < 256 &&         c < 256 &&
2294  #endif  #endif
2295         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1578  for (;;) Line 2299  for (;;)
2299      break;      break;
2300    
2301      case OP_WORDCHAR:      case OP_WORDCHAR:
2302      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2303          {
2304          SCHECK_PARTIAL();
2305          RRETURN(MATCH_NOMATCH);
2306          }
2307      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2308      if (      if (
2309  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2310         c >= 256 ||         c > 255 ||
2311  #endif  #endif
2312         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2313         )         )
# Line 1591  for (;;) Line 2316  for (;;)
2316      break;      break;
2317    
2318      case OP_ANYNL:      case OP_ANYNL:
2319      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2320          {
2321          SCHECK_PARTIAL();
2322          RRETURN(MATCH_NOMATCH);
2323          }
2324      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2325      switch(c)      switch(c)
2326        {        {
2327        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2328    
2329        case 0x000d:        case 0x000d:
2330        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2331        break;        break;
# Line 1615  for (;;) Line 2345  for (;;)
2345      break;      break;
2346    
2347      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2348      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2349          {
2350          SCHECK_PARTIAL();
2351          RRETURN(MATCH_NOMATCH);
2352          }
2353      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2354      switch(c)      switch(c)
2355        {        {
# Line 1645  for (;;) Line 2379  for (;;)
2379      break;      break;
2380    
2381      case OP_HSPACE:      case OP_HSPACE:
2382      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2383          {
2384          SCHECK_PARTIAL();
2385          RRETURN(MATCH_NOMATCH);
2386          }
2387      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2388      switch(c)      switch(c)
2389        {        {
# Line 1675  for (;;) Line 2413  for (;;)
2413      break;      break;
2414    
2415      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2416      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2417          {
2418          SCHECK_PARTIAL();
2419          RRETURN(MATCH_NOMATCH);
2420          }
2421      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2422      switch(c)      switch(c)
2423        {        {
# Line 1693  for (;;) Line 2435  for (;;)
2435      break;      break;
2436    
2437      case OP_VSPACE:      case OP_VSPACE:
2438      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2439          {
2440          SCHECK_PARTIAL();
2441          RRETURN(MATCH_NOMATCH);
2442          }
2443      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2444      switch(c)      switch(c)
2445        {        {
# Line 1716  for (;;) Line 2462  for (;;)
2462    
2463      case OP_PROP:      case OP_PROP:
2464      case OP_NOTPROP:      case OP_NOTPROP:
2465      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2466          {
2467          SCHECK_PARTIAL();
2468          RRETURN(MATCH_NOMATCH);
2469          }
2470      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2471        {        {
2472        const ucd_record *prop = GET_UCD(c);        const ucd_record *prop = GET_UCD(c);
# Line 1732  for (;;) Line 2482  for (;;)
2482               prop->chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2483               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2484            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2485           break;          break;
2486    
2487          case PT_GC:          case PT_GC:
2488          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2489            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2490          break;          break;
2491    
# Line 1749  for (;;) Line 2499  for (;;)
2499            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2500          break;          break;
2501    
2502            /* These are specials */
2503    
2504            case PT_ALNUM:
2505            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2506                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2507              RRETURN(MATCH_NOMATCH);
2508            break;
2509    
2510            case PT_SPACE:    /* Perl space */
2511            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2512                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2513                   == (op == OP_NOTPROP))
2514              RRETURN(MATCH_NOMATCH);
2515            break;
2516    
2517            case PT_PXSPACE:  /* POSIX space */
2518            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2519                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2520                 c == CHAR_FF || c == CHAR_CR)
2521                   == (op == OP_NOTPROP))
2522              RRETURN(MATCH_NOMATCH);
2523            break;
2524    
2525            case PT_WORD:
2526            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2527                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2528                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2529              RRETURN(MATCH_NOMATCH);
2530            break;
2531    
2532            /* This should never occur */
2533    
2534          default:          default:
2535          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2536          }          }
# Line 1761  for (;;) Line 2543  for (;;)
2543      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2544    
2545      case OP_EXTUNI:      case OP_EXTUNI:
2546      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2547          {
2548          SCHECK_PARTIAL();
2549          RRETURN(MATCH_NOMATCH);
2550          }
2551      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2552        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2553        while (eptr < md->end_subject)
2554        {        {
2555        int category = UCD_CATEGORY(c);        int len = 1;
2556        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2557        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2558          {        eptr += len;
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2559        }        }
2560      ecode++;      ecode++;
2561      break;      break;
# Line 1792  for (;;) Line 2571  for (;;)
2571      loops). */      loops). */
2572    
2573      case OP_REF:      case OP_REF:
2574        {      case OP_REFI:
2575        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2576        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2577        ecode += 1 + IMM2_SIZE;
2578    
2579        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2580    
2581        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2582        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
2583        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.  
2584    
2585        (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
2586        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2587    
2588        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
2589        referenced subpattern. */      referenced subpattern. */
2590    
2591        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2592          length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;        length = (md->jscript_compat)? 0 : -1;
2593        else      else
2594          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2595    
2596        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2597    
2598        switch (*ecode)      switch (*ecode)
2599          {        {
2600          case OP_CRSTAR:        case OP_CRSTAR:
2601          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2602          case OP_CRPLUS:        case OP_CRPLUS:
2603          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2604          case OP_CRQUERY:        case OP_CRQUERY:
2605          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2606          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2607          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2608          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2609          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2610          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2611          break;        break;
2612    
2613          case OP_CRRANGE:        case OP_CRRANGE:
2614          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2615          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2616          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2617          max = GET2(ecode, 3);        max = GET2(ecode, 1 + IMM2_SIZE);
2618          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2619          ecode += 5;        ecode += 1 + 2 * IMM2_SIZE;
2620          break;        break;
2621    
2622          default:               /* No repeat follows */        default:               /* No repeat follows */
2623          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2624          eptr += length;          {
2625          continue;              /* With the main loop */          CHECK_PARTIAL();
2626            RRETURN(MATCH_NOMATCH);
2627          }          }
2628          eptr += length;
2629          continue;              /* With the main loop */
2630          }
2631    
2632        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2633        main loop. */      zero, just continue with the main loop. */
2634    
2635        if (length == 0) continue;      if (length == 0) continue;
2636    
2637        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2638        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2639        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2640    
2641        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2642          {
2643          int slength;
2644          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2645          {          {
2646          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2647          eptr += length;          RRETURN(MATCH_NOMATCH);
2648          }          }
2649          eptr += slength;
2650          }
2651    
2652        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2653        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2654    
2655        if (min == max) continue;      if (min == max) continue;
2656    
2657        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2658    
2659        if (minimize)      if (minimize)
2660          {
2661          for (fi = min;; fi++)
2662          {          {
2663          for (fi = min;; fi++)          int slength;
2664            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2665            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2666            if (fi >= max) RRETURN(MATCH_NOMATCH);
2667            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2668            {            {
2669            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            CHECK_PARTIAL();
2670            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            RRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2671            }            }
2672          /* Control never gets here */          eptr += slength;
2673          }          }
2674          /* Control never gets here */
2675          }
2676    
2677        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2678    
2679        else      else
2680          {
2681          pp = eptr;
2682          for (i = min; i < max; i++)
2683          {          {
2684          pp = eptr;          int slength;
2685          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)  
2686            {            {
2687            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            CHECK_PARTIAL();
2688            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2689            }            }
2690          RRETURN(MATCH_NOMATCH);          eptr += slength;
2691          }          }
2692          while (eptr >= pp)
2693            {
2694            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2695            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2696            eptr -= length;
2697            }
2698          RRETURN(MATCH_NOMATCH);
2699        }        }
2700      /* Control never gets here */      /* Control never gets here */
2701    
   
   
2702      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2703      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,
2704      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 1918  for (;;) Line 2713  for (;;)
2713      case OP_NCLASS:      case OP_NCLASS:
2714      case OP_CLASS:      case OP_CLASS:
2715        {        {
2716          /* The data variable is saved across frames, so the byte map needs to
2717          be stored there. */
2718    #define BYTE_MAP ((pcre_uint8 *)data)
2719        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2720        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2721    
2722        switch (*ecode)        switch (*ecode)
2723          {          {
# Line 1940  for (;;) Line 2738  for (;;)
2738          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2739          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2740          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2741          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2742          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2743          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2744          break;          break;
2745    
2746          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1952  for (;;) Line 2750  for (;;)
2750    
2751        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2752    
2753  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2754        /* UTF-8 mode */        if (utf)
       if (utf8)  
2755          {          {
2756          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2757            {            {
2758            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2759                {
2760                SCHECK_PARTIAL();
2761                RRETURN(MATCH_NOMATCH);
2762                }
2763            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2764            if (c > 255)            if (c > 255)
2765              {              {
2766              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2767              }              }
2768            else            else
2769              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2770            }            }
2771          }          }
2772        else        else
2773  #endif  #endif
2774        /* Not UTF-8 mode */        /* Not UTF mode */
2775          {          {
2776          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2777            {            {
2778            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2779                {
2780                SCHECK_PARTIAL();
2781                RRETURN(MATCH_NOMATCH);
2782                }
2783            c = *eptr++;            c = *eptr++;
2784            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2785              if (c > 255)
2786                {
2787                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2788                }
2789              else
2790    #endif
2791                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2792            }            }
2793          }          }
2794    
# Line 1992  for (;;) Line 2802  for (;;)
2802    
2803        if (minimize)        if (minimize)
2804          {          {
2805  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2806          /* UTF-8 mode */          if (utf)
         if (utf8)  
2807            {            {
2808            for (fi = min;; fi++)            for (fi = min;; fi++)
2809              {              {
2810              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2811              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2812              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2813                if (eptr >= md->end_subject)
2814                  {
2815                  SCHECK_PARTIAL();
2816                  RRETURN(MATCH_NOMATCH);
2817                  }
2818              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2819              if (c > 255)              if (c > 255)
2820                {                {
2821                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2822                }                }
2823              else              else
2824                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2825              }              }
2826            }            }
2827          else          else
2828  #endif  #endif
2829          /* Not UTF-8 mode */          /* Not UTF mode */
2830            {            {
2831            for (fi = min;; fi++)            for (fi = min;; fi++)
2832              {              {
2833              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2834              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2835              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2836                if (eptr >= md->end_subject)
2837                  {
2838                  SCHECK_PARTIAL();
2839                  RRETURN(MATCH_NOMATCH);
2840                  }
2841              c = *eptr++;              c = *eptr++;
2842              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2843                if (c > 255)
2844                  {
2845                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2846                  }
2847                else
2848    #endif
2849                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2850              }              }
2851            }            }
2852          /* Control never gets here */          /* Control never gets here */
# Line 2034  for (;;) Line 2858  for (;;)
2858          {          {
2859          pp = eptr;          pp = eptr;
2860    
2861  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2862          /* UTF-8 mode */          if (utf)
         if (utf8)  
2863            {            {
2864            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2865              {              {
2866              int len = 1;              int len = 1;
2867              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2868                  {
2869                  SCHECK_PARTIAL();
2870                  break;
2871                  }
2872              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2873              if (c > 255)              if (c > 255)
2874                {                {
2875                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2876                }                }
2877              else              else
2878                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2879              eptr += len;              eptr += len;
2880              }              }
2881            for (;;)            for (;;)
2882              {              {
2883              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2884              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2885              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2886              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2063  for (;;) Line 2888  for (;;)
2888            }            }
2889          else          else
2890  #endif  #endif
2891            /* Not UTF-8 mode */            /* Not UTF mode */
2892            {            {
2893            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2894              {              {
2895              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2896                  {
2897                  SCHECK_PARTIAL();
2898                  break;
2899                  }
2900              c = *eptr;              c = *eptr;
2901              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2902                if (c > 255)
2903                  {
2904                  if (op == OP_CLASS) break;
2905                  }
2906                else
2907    #endif
2908                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2909              eptr++;              eptr++;
2910              }              }
2911            while (eptr >= pp)            while (eptr >= pp)
2912              {              {
2913              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2914              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2915              eptr--;              eptr--;
2916              }              }
# Line 2082  for (;;) Line 2918  for (;;)
2918    
2919          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2920          }          }
2921    #undef BYTE_MAP
2922        }        }
2923      /* Control never gets here */      /* Control never gets here */
2924    
# Line 2090  for (;;) Line 2927  for (;;)
2927      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
2928      mode, because Unicode properties are supported in non-UTF-8 mode. */      mode, because Unicode properties are supported in non-UTF-8 mode. */
2929    
2930  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2931      case OP_XCLASS:      case OP_XCLASS:
2932        {        {
2933        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2115  for (;;) Line 2952  for (;;)
2952          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2953          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2954          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2955          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2956          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2957          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2958          break;          break;
2959    
2960          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2129  for (;;) Line 2966  for (;;)
2966    
2967        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2968          {          {
2969          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2970              {
2971              SCHECK_PARTIAL();
2972              RRETURN(MATCH_NOMATCH);
2973              }
2974          GETCHARINCTEST(c, eptr);          GETCHARINCTEST(c, eptr);
2975          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);          if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
2976          }          }
2977    
2978        /* 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 2146  for (;;) Line 2987  for (;;)
2987          {          {
2988          for (fi = min;; fi++)          for (fi = min;; fi++)
2989            {            {
2990            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2991            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2992            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
2993              if (eptr >= md->end_subject)
2994                {
2995                SCHECK_PARTIAL();
2996                RRETURN(MATCH_NOMATCH);
2997                }
2998            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
2999            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3000            }            }
3001          /* Control never gets here */          /* Control never gets here */
3002          }          }
# Line 2163  for (;;) Line 3009  for (;;)
3009          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3010            {            {
3011            int len = 1;            int len = 1;
3012            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3013                {
3014                SCHECK_PARTIAL();
3015                break;
3016                }
3017    #ifdef SUPPORT_UTF
3018            GETCHARLENTEST(c, eptr, len);            GETCHARLENTEST(c, eptr, len);
3019            if (!_pcre_xclass(c, data)) break;  #else
3020              c = *eptr;
3021    #endif
3022              if (!PRIV(xclass)(c, data, utf)) break;
3023            eptr += len;            eptr += len;
3024            }            }
3025          for(;;)          for(;;)
3026            {            {
3027            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3028            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3029            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3030            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3031              if (utf) BACKCHAR(eptr);
3032    #endif
3033            }            }
3034          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3035          }          }
# Line 2185  for (;;) Line 3041  for (;;)
3041      /* Match a single character, casefully */      /* Match a single character, casefully */
3042    
3043      case OP_CHAR:      case OP_CHAR:
3044  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3045      if (utf8)      if (utf)
3046        {        {
3047        length = 1;        length = 1;
3048        ecode++;        ecode++;
3049        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3050        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3051            {
3052            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3053            RRETURN(MATCH_NOMATCH);
3054            }
3055        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3056        }        }
3057      else      else
3058  #endif  #endif
3059        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3060        {        {
3061        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3062            {
3063            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3064            RRETURN(MATCH_NOMATCH);
3065            }
3066        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3067        ecode += 2;        ecode += 2;
3068        }        }
3069      break;      break;
3070    
3071      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3072        subject, give up immediately. */
3073    
3074        case OP_CHARI:
3075        if (eptr >= md->end_subject)
3076          {
3077          SCHECK_PARTIAL();
3078          RRETURN(MATCH_NOMATCH);
3079          }
3080    
3081      case OP_CHARNC:  #ifdef SUPPORT_UTF
3082  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3083        {        {
3084        length = 1;        length = 1;
3085        ecode++;        ecode++;
3086        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3087    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3088        /* 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
3089        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3090          fast lookup table. We know that there is at least one byte left in the
3091          subject. */
3092    
3093        if (fc < 128)        if (fc < 128)
3094          {          {
3095          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3096                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3097            ecode++;
3098            eptr++;
3099          }          }
3100    
3101        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3102          use the value of "length" to check for sufficient bytes left, because the
3103          other case of the character may have more or fewer bytes.  */
3104    
3105        else        else
3106          {          {
# Line 2246  for (;;) Line 3121  for (;;)
3121          }          }
3122        }        }
3123      else      else
3124  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3125    
3126      /* Non-UTF-8 mode */      /* Not UTF mode */
3127        {        {
3128        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3129        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3130          eptr++;
3131        ecode += 2;        ecode += 2;
3132        }        }
3133      break;      break;
# Line 2259  for (;;) Line 3135  for (;;)
3135      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3136    
3137      case OP_EXACT:      case OP_EXACT:
3138        case OP_EXACTI:
3139      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3140      ecode += 3;      ecode += 1 + IMM2_SIZE;
3141      goto REPEATCHAR;      goto REPEATCHAR;
3142    
3143      case OP_POSUPTO:      case OP_POSUPTO:
3144        case OP_POSUPTOI:
3145      possessive = TRUE;      possessive = TRUE;
3146      /* Fall through */      /* Fall through */
3147    
3148      case OP_UPTO:      case OP_UPTO:
3149        case OP_UPTOI:
3150      case OP_MINUPTO:      case OP_MINUPTO:
3151        case OP_MINUPTOI:
3152      min = 0;      min = 0;
3153      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3154      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3155      ecode += 3;      ecode += 1 + IMM2_SIZE;
3156      goto REPEATCHAR;      goto REPEATCHAR;
3157    
3158      case OP_POSSTAR:      case OP_POSSTAR:
3159        case OP_POSSTARI:
3160      possessive = TRUE;      possessive = TRUE;
3161      min = 0;      min = 0;
3162      max = INT_MAX;      max = INT_MAX;
# Line 2283  for (;;) Line 3164  for (;;)
3164      goto REPEATCHAR;      goto REPEATCHAR;
3165    
3166      case OP_POSPLUS:      case OP_POSPLUS:
3167        case OP_POSPLUSI:
3168      possessive = TRUE;      possessive = TRUE;
3169      min = 1;      min = 1;
3170      max = INT_MAX;      max = INT_MAX;
# Line 2290  for (;;) Line 3172  for (;;)
3172      goto REPEATCHAR;      goto REPEATCHAR;
3173    
3174      case OP_POSQUERY:      case OP_POSQUERY:
3175        case OP_POSQUERYI:
3176      possessive = TRUE;      possessive = TRUE;
3177      min = 0;      min = 0;
3178      max = 1;      max = 1;
# Line 2297  for (;;) Line 3180  for (;;)
3180      goto REPEATCHAR;      goto REPEATCHAR;
3181    
3182      case OP_STAR:      case OP_STAR:
3183        case OP_STARI:
3184      case OP_MINSTAR:      case OP_MINSTAR:
3185        case OP_MINSTARI:
3186      case OP_PLUS:      case OP_PLUS:
3187        case OP_PLUSI:
3188      case OP_MINPLUS:      case OP_MINPLUS:
3189        case OP_MINPLUSI:
3190      case OP_QUERY:      case OP_QUERY:
3191        case OP_QUERYI:
3192      case OP_MINQUERY:      case OP_MINQUERY:
3193      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3194        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3195      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3196      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3197      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3198      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3199    
3200      /* 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. */  
3201    
3202      REPEATCHAR:      REPEATCHAR:
3203  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3204      if (utf8)      if (utf)
3205        {        {
3206        length = 1;        length = 1;
3207        charptr = ecode;        charptr = ecode;
3208        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3209        ecode += length;        ecode += length;
3210    
3211        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2329  for (;;) Line 3215  for (;;)
3215          {          {
3216  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3217          unsigned int othercase;          unsigned int othercase;
3218          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3219              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3220            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3221          else oclength = 0;          else oclength = 0;
3222  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3223    
3224          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3225            {            {
3226            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3227                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3228  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3229            /* Need braces because of following else */            else if (oclength > 0 &&
3230            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3231                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3232    #endif  /* SUPPORT_UCP */
3233            else            else
3234              {              {
3235              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3236              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3237              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3238            }            }
3239    
3240          if (min == max) continue;          if (min == max) continue;
# Line 2357  for (;;) Line 3243  for (;;)
3243            {            {
3244            for (fi = min;; fi++)            for (fi = min;; fi++)
3245              {              {
3246              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3247              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3248              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3249              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3250                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3251  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3252              /* Need braces because of following else */              else if (oclength > 0 &&
3253              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3254                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3255    #endif  /* SUPPORT_UCP */
3256              else              else
3257                {                {
3258                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3259                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3260                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3261              }              }
3262            /* Control never gets here */            /* Control never gets here */
3263            }            }
# Line 2381  for (;;) Line 3267  for (;;)
3267            pp = eptr;            pp = eptr;
3268            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3269              {              {
3270              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3271              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3272  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3273              else if (oclength == 0) break;              else if (oclength > 0 &&
3274                         eptr <= md->end_subject - oclength &&
3275                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3276    #endif  /* SUPPORT_UCP */
3277              else              else
3278                {                {
3279                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3280                eptr += oclength;                break;
3281                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3282              }              }
3283    
3284            if (possessive) continue;            if (possessive) continue;
3285    
3286            for(;;)            for(;;)
3287             {              {
3288             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3289             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3290             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3291  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3292             eptr--;              eptr--;
3293             BACKCHAR(eptr);              BACKCHAR(eptr);
3294  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3295             eptr -= length;              eptr -= length;
3296  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3297             }              }
3298            }            }
3299          /* Control never gets here */          /* Control never gets here */
3300          }          }
# Line 2417  for (;;) Line 3304  for (;;)
3304        value of fc will always be < 128. */        value of fc will always be < 128. */
3305        }        }
3306      else      else
3307  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3308          /* 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);  
3309        fc = *ecode++;        fc = *ecode++;
       }  
3310    
3311      /* The value of fc at this point is always less than 256, though we may or      /* The value of fc at this point is always one character, though we may
3312      may not be in UTF-8 mode. The code is duplicated for the caseless and      or may not be in UTF mode. The code is duplicated for the caseless and
3313      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3314      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3315      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2437  for (;;) Line 3320  for (;;)
3320      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3321        max, eptr));        max, eptr));
3322    
3323      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3324        {        {
3325        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3326          /* fc must be < 128 if UTF is enabled. */
3327          foc = md->fcc[fc];
3328    #else
3329    #ifdef SUPPORT_UTF
3330    #ifdef SUPPORT_UCP
3331          if (utf && fc > 127)
3332            foc = UCD_OTHERCASE(fc);
3333    #else
3334          if (utf && fc > 127)
3335            foc = fc;
3336    #endif /* SUPPORT_UCP */
3337          else
3338    #endif /* SUPPORT_UTF */
3339            foc = TABLE_GET(fc, md->fcc, fc);
3340    #endif /* COMPILE_PCRE8 */
3341    
3342        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3343          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3344            if (eptr >= md->end_subject)
3345              {
3346              SCHECK_PARTIAL();
3347              RRETURN(MATCH_NOMATCH);
3348              }
3349            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3350            eptr++;
3351            }
3352        if (min == max) continue;        if (min == max) continue;
3353        if (minimize)        if (minimize)
3354          {          {
3355          for (fi = min;; fi++)          for (fi = min;; fi++)
3356            {            {
3357            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3358            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3359            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3360                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3361                {
3362                SCHECK_PARTIAL();
3363              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3364                }
3365              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3366              eptr++;
3367            }            }
3368          /* Control never gets here */          /* Control never gets here */
3369          }          }
# Line 2460  for (;;) Line 3372  for (;;)
3372          pp = eptr;          pp = eptr;
3373          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3374            {            {
3375            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3376                {
3377                SCHECK_PARTIAL();
3378                break;
3379                }
3380              if (fc != *eptr && foc != *eptr) break;
3381            eptr++;            eptr++;
3382            }            }
3383    
3384          if (possessive) continue;          if (possessive) continue;
3385    
3386          while (eptr >= pp)          while (eptr >= pp)
3387            {            {
3388            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3389            eptr--;            eptr--;
3390            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3391            }            }
# Line 2479  for (;;) Line 3398  for (;;)
3398    
3399      else      else
3400        {        {
3401        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3402            {
3403            if (eptr >= md->end_subject)
3404              {
3405              SCHECK_PARTIAL();
3406              RRETURN(MATCH_NOMATCH);
3407              }
3408            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3409            }
3410    
3411        if (min == max) continue;        if (min == max) continue;
3412    
3413        if (minimize)        if (minimize)
3414          {          {
3415          for (fi = min;; fi++)          for (fi = min;; fi++)
3416            {            {
3417            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3418            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3419            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3420              if (eptr >= md->end_subject)
3421                {
3422                SCHECK_PARTIAL();
3423              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3424                }
3425              if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3426            }            }
3427          /* Control never gets here */          /* Control never gets here */
3428          }          }
# Line 2497  for (;;) Line 3431  for (;;)
3431          pp = eptr;          pp = eptr;
3432          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3433            {            {
3434            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3435                {
3436                SCHECK_PARTIAL();
3437                break;
3438                }
3439              if (fc != *eptr) break;
3440            eptr++;            eptr++;
3441            }            }
3442          if (possessive) continue;          if (possessive) continue;
3443    
3444          while (eptr >= pp)          while (eptr >= pp)
3445            {            {
3446            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3447            eptr--;            eptr--;
3448            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3449            }            }
# Line 2516  for (;;) Line 3456  for (;;)
3456      checking can be multibyte. */      checking can be multibyte. */
3457    
3458      case OP_NOT:      case OP_NOT:
3459      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3460        if (eptr >= md->end_subject)
3461          {
3462          SCHECK_PARTIAL();
3463          RRETURN(MATCH_NOMATCH);
3464          }
3465      ecode++;      ecode++;
3466      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3467      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3468        {        {
3469  #ifdef SUPPORT_UTF8        register int ch, och;
3470        if (c < 256)        ch = *ecode++;
3471  #endif  #ifdef COMPILE_PCRE8
3472        c = md->lcc[c];        /* ch must be < 128 if UTF is enabled. */
3473        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        och = md->fcc[ch];
3474    #else
3475    #ifdef SUPPORT_UTF
3476    #ifdef SUPPORT_UCP
3477          if (utf && ch > 127)
3478            och = UCD_OTHERCASE(ch);
3479    #else
3480          if (utf && ch > 127)
3481            och = ch;
3482    #endif /* SUPPORT_UCP */
3483          else
3484    #endif /* SUPPORT_UTF */
3485            och = TABLE_GET(ch, md->fcc, ch);
3486    #endif /* COMPILE_PCRE8 */
3487          if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3488        }        }
3489      else      else    /* Caseful */
3490        {        {
3491        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);
3492        }        }
# Line 2541  for (;;) Line 3500  for (;;)
3500      about... */      about... */
3501    
3502      case OP_NOTEXACT:      case OP_NOTEXACT:
3503        case OP_NOTEXACTI:
3504      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3505      ecode += 3;      ecode += 1 + IMM2_SIZE;
3506      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3507    
3508      case OP_NOTUPTO:      case OP_NOTUPTO:
3509        case OP_NOTUPTOI:
3510      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3511        case OP_NOTMINUPTOI:
3512      min = 0;      min = 0;
3513      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3514      minimize = *ecode == OP_NOTMINUPTO;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3515      ecode += 3;      ecode += 1 + IMM2_SIZE;
3516      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3517    
3518      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
3519        case OP_NOTPOSSTARI:
3520      possessive = TRUE;      possessive = TRUE;
3521      min = 0;      min = 0;
3522      max = INT_MAX;      max = INT_MAX;
# Line 2561  for (;;) Line 3524  for (;;)
3524      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3525    
3526      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
3527        case OP_NOTPOSPLUSI:
3528      possessive = TRUE;      possessive = TRUE;
3529      min = 1;      min = 1;
3530      max = INT_MAX;      max = INT_MAX;
# Line 2568  for (;;) Line 3532  for (;;)
3532      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3533    
3534      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
3535        case OP_NOTPOSQUERYI:
3536      possessive = TRUE;      possessive = TRUE;
3537      min = 0;      min = 0;
3538      max = 1;      max = 1;
# Line 2575  for (;;) Line 3540  for (;;)
3540      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3541    
3542      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
3543        case OP_NOTPOSUPTOI:
3544      possessive = TRUE;      possessive = TRUE;
3545      min = 0;      min = 0;
3546      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3547      ecode += 3;      ecode += 1 + IMM2_SIZE;
3548      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3549    
3550      case OP_NOTSTAR:      case OP_NOTSTAR:
3551        case OP_NOTSTARI:
3552      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3553        case OP_NOTMINSTARI:
3554      case OP_NOTPLUS:      case OP_NOTPLUS:
3555        case OP_NOTPLUSI:
3556      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3557        case OP_NOTMINPLUSI:
3558      case OP_NOTQUERY:      case OP_NOTQUERY:
3559        case OP_NOTQUERYI:
3560      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3561      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3562        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3563      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3564      min = rep_min[c];