/[pcre]/code/trunk/pcre_exec.c
ViewVC logotype

Diff of /code/trunk/pcre_exec.c

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

revision 409 by ph10, Sat Mar 28 17:10:56 2009 UTC revision 937 by ph10, Sun Feb 26 15:58:56 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2009 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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 the number of subject bytes matched
144                  -1 no match
145                  -2 partial match; always given if at end subject
146  */  */
147    
148  static BOOL  static int
149  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,
150    unsigned long int ims)    BOOL caseless)
151  {  {
152  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
153    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
154    
155  #ifdef DEBUG  #ifdef PCRE_DEBUG
156  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
157    printf("matching subject <null>");    printf("matching subject <null>");
158  else  else
# Line 154  pchars(p, length, FALSE, md); Line 165  pchars(p, length, FALSE, md);
165  printf("\n");  printf("\n");
166  #endif  #endif
167    
168  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
169    case the length is passed as zero). */
170    
171  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
172    
173  /* 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
174  properly if Unicode properties are supported. Otherwise, we can check only  properly if Unicode properties are supported. Otherwise, we can check only
175  ASCII characters. */  ASCII characters. */
176    
177  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
178    {    {
179  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
180  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
181    if (md->utf8)    if (md->utf)
182      {      {
183      USPTR endptr = eptr + length;      /* Match characters up to the end of the reference. NOTE: the number of
184      while (eptr < endptr)      bytes matched may differ, because there are some characters whose upper and
185        lower case versions code as different numbers of bytes. For example, U+023A
186        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
187        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
188        the latter. It is important, therefore, to check the length along the
189        reference, not along the subject (earlier code did this wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193        {        {
194        int c, d;        int c, d;
195          if (eptr >= md->end_subject) return -2;   /* Partial match */
196        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
197        GETCHARINC(d, p);        GETCHARINC(d, p);
198        if (c != d && c != UCD_OTHERCASE(d)) return FALSE;        if (c != d && c != UCD_OTHERCASE(d)) return -1;
199        }        }
200      }      }
201    else    else
# Line 183  if ((ims & PCRE_CASELESS) != 0) Line 204  if ((ims & PCRE_CASELESS) != 0)
204    
205    /* 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
206    is no UCP support. */    is no UCP support. */
207        {
208    while (length-- > 0)      while (length-- > 0)
209      { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }        {
210          if (eptr >= md->end_subject) return -2;   /* Partial match */
211          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
212          p++;
213          eptr++;
214          }
215        }
216    }    }
217    
218  /* 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
219  are in UTF-8 mode. */  are in UTF-8 mode. */
220    
221  else  else
222    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
223      while (length-- > 0)
224        {
225        if (eptr >= md->end_subject) return -2;   /* Partial match */
226        if (*p++ != *eptr++) return -1;
227        }
228      }
229    
230  return TRUE;  return (int)(eptr - eptr_start);
231  }  }
232    
233    
# Line 245  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 278  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
278         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
279         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
280         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
281         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
282           RM61,  RM62, RM63, RM64, RM65, RM66 };
283    
284  /* 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
285  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
286  actuall used in this definition. */  actually used in this definition. */
287    
288  #ifndef NO_RECURSE  #ifndef NO_RECURSE
289  #define REGISTER register  #define REGISTER register
290    
291  #ifdef DEBUG  #ifdef PCRE_DEBUG
292  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
293    { \    { \
294    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
295    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
296    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
297    }    }
298  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 267  actuall used in this definition. */ Line 301  actuall used in this definition. */
301    return ra; \    return ra; \
302    }    }
303  #else  #else
304  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
305    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
306  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
307  #endif  #endif
308    
# Line 281  argument of match(), which never changes Line 315  argument of match(), which never changes
315    
316  #define REGISTER  #define REGISTER
317    
318  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
319    {\    {\
320    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
321    frame->Xwhere = rw; \    if (newframe == NULL)\
322        {\
323        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
324        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
325        newframe->Xnextframe = NULL;\
326        frame->Xnextframe = newframe;\
327        }\
328      frame->Xwhere = rw;\
329    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
330    newframe->Xecode = rb;\    newframe->Xecode = rb;\
331    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
332    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
333    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
334    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
335    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
336    frame = newframe;\    frame = newframe;\
# Line 303  argument of match(), which never changes Line 342  argument of match(), which never changes
342    
343  #define RRETURN(ra)\  #define RRETURN(ra)\
344    {\    {\
345    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
346    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
347    if (frame != NULL)\    if (frame != NULL)\
348      {\      {\
349      rrc = ra;\      rrc = ra;\
# Line 319  argument of match(), which never changes Line 357  argument of match(), which never changes
357    
358  typedef struct heapframe {  typedef struct heapframe {
359    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
360      struct heapframe *Xnextframe;
361    
362    /* Function arguments that may change */    /* Function arguments that may change */
363    
364    USPTR Xeptr;    PCRE_PUCHAR Xeptr;
365    const uschar *Xecode;    const pcre_uchar *Xecode;
366    USPTR Xmstart;    PCRE_PUCHAR Xmstart;
367    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
368    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
369    unsigned int Xrdepth;    unsigned int Xrdepth;
370    
371    /* Function local variables */    /* Function local variables */
372    
373    USPTR Xcallpat;    PCRE_PUCHAR Xcallpat;
374  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
375    USPTR Xcharptr;    PCRE_PUCHAR Xcharptr;
376  #endif  #endif
377    USPTR Xdata;    PCRE_PUCHAR Xdata;
378    USPTR Xnext;    PCRE_PUCHAR Xnext;
379    USPTR Xpp;    PCRE_PUCHAR Xpp;
380    USPTR Xprev;    PCRE_PUCHAR Xprev;
381    USPTR Xsaved_eptr;    PCRE_PUCHAR Xsaved_eptr;
382    
383    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
384    
# Line 349  typedef struct heapframe { Line 386  typedef struct heapframe {
386    BOOL Xcondition;    BOOL Xcondition;
387    BOOL Xprev_is_word;    BOOL Xprev_is_word;
388    
   unsigned long int Xoriginal_ims;  
   
389  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
390    int Xprop_type;    int Xprop_type;
391    int Xprop_value;    int Xprop_value;
392    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
393    int Xoclength;    int Xoclength;
394    uschar Xocchars[8];    pcre_uchar Xocchars[6];
395  #endif  #endif
396    
397    int Xcodelink;    int Xcodelink;
# Line 398  typedef struct heapframe { Line 430  typedef struct heapframe {
430    
431  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
432  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
433  same response.  same response. */
434    
435  Performance note: It might be tempting to extract commonly used fields from the  /* These macros pack up tests that are used for partial matching, and which
436  md structure (e.g. utf8, end_subject) into individual variables to improve  appear several times in the code. We set the "hit end" flag if the pointer is
437    at the end of the subject and also past the start of the subject (i.e.
438    something has been matched). For hard partial matching, we then return
439    immediately. The second one is used when we already know we are past the end of
440    the subject. */
441    
442    #define CHECK_PARTIAL()\
443      if (md->partial != 0 && eptr >= md->end_subject && \
444          eptr > md->start_used_ptr) \
445        { \
446        md->hitend = TRUE; \
447        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
448        }
449    
450    #define SCHECK_PARTIAL()\
451      if (md->partial != 0 && eptr > md->start_used_ptr) \
452        { \
453        md->hitend = TRUE; \
454        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
455        }
456    
457    
458    /* Performance note: It might be tempting to extract commonly used fields from
459    the md structure (e.g. utf, end_subject) into individual variables to improve
460  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
461  made performance worse.  made performance worse.
462    
# Line 412  Arguments: Line 467  Arguments:
467                   by encountering \K)                   by encountering \K)
468     offset_top  current top pointer     offset_top  current top pointer
469     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
470     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
471                   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  
472     rdepth      the recursion depth     rdepth      the recursion depth
473    
474  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
475                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
476                   a negative MATCH_xxx value for PRUNE, SKIP, etc
477                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
478                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
479  */  */
480    
481  static int  static int
482  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
483    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
484    int flags, unsigned int rdepth)    unsigned int rdepth)
485  {  {
486  /* 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,
487  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 490  so they can be ordinary variables in all
490  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
491  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
492  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
493  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
494    
495  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
496    BOOL caseless;
497  int condcode;  int condcode;
498    
499  /* 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
500  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
501  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
502  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
503    the top-level on the stack rather than malloc-ing them all gives a performance
504    boost in many cases where there is not much "recursion". */
505    
506  #ifdef NO_RECURSE  #ifdef NO_RECURSE
507  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
508    
509  /* Copy in the original argument variables */  /* Copy in the original argument variables */
510    
# Line 459  frame->Xeptr = eptr; Line 512  frame->Xeptr = eptr;
512  frame->Xecode = ecode;  frame->Xecode = ecode;
513  frame->Xmstart = mstart;  frame->Xmstart = mstart;
514  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
515  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
516  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
517    
518  /* 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 525  HEAP_RECURSE:
525  #define ecode              frame->Xecode  #define ecode              frame->Xecode
526  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
527  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
528  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
529  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
530    
531  /* Ditto for the local variables */  /* Ditto for the local variables */
532    
533  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
534  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
535  #endif  #endif
536  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
# Line 498  HEAP_RECURSE: Line 547  HEAP_RECURSE:
547  #define condition          frame->Xcondition  #define condition          frame->Xcondition
548  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
549    
 #define original_ims       frame->Xoriginal_ims  
   
550  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
551  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
552  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
553  #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  
554  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
555  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
556  #endif  #endif
# Line 536  i, and fc and c, can be the same variabl Line 580  i, and fc and c, can be the same variabl
580  #define fi i  #define fi i
581  #define fc c  #define fc c
582    
583    /* Many of the following variables are used only in small blocks of the code.
584    My normal style of coding would have declared them within each of those blocks.
585    However, in order to accommodate the version of this code that uses an external
586    "stack" implemented on the heap, it is easier to declare them all here, so the
587    declarations can be cut out in a block. The only declarations within blocks
588    below are for variables that do not have to be preserved over a recursive call
589    to RMATCH(). */
590    
591    #ifdef SUPPORT_UTF
592    const pcre_uchar *charptr;
593    #endif
594    const pcre_uchar *callpat;
595    const pcre_uchar *data;
596    const pcre_uchar *next;
597    PCRE_PUCHAR       pp;
598    const pcre_uchar *prev;
599    PCRE_PUCHAR       saved_eptr;
600    
601  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
602  const uschar *charptr;             /* in small blocks of the code. My normal */  
603  #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().          */  
604  BOOL condition;  BOOL condition;
605  BOOL prev_is_word;  BOOL prev_is_word;
606    
 unsigned long int original_ims;  
   
607  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
608  int prop_type;  int prop_type;
609  int prop_value;  int prop_value;
610  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
611  int oclength;  int oclength;
612  uschar occhars[8];  pcre_uchar occhars[6];
613  #endif  #endif
614    
615  int codelink;  int codelink;
# Line 579  int save_offset1, save_offset2, save_off Line 625  int save_offset1, save_offset2, save_off
625  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
626    
627  eptrblock newptrb;  eptrblock newptrb;
628    
629    /* There is a special fudge for calling match() in a way that causes it to
630    measure the size of its basic stack frame when the stack is being used for
631    recursion. The second argument (ecode) being NULL triggers this behaviour. It
632    cannot normally ever be NULL. The return is the negated value of the frame
633    size. */
634    
635    if (ecode == NULL)
636      {
637      if (rdepth == 0)
638        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
639      else
640        {
641        int len = (char *)&rdepth - (char *)eptr;
642        return (len > 0)? -len : len;
643        }
644      }
645  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
646    
647    /* To save space on the stack and in the heap frame, I have doubled up on some
648    of the local variables that are used only in localised parts of the code, but
649    still need to be preserved over recursive calls of match(). These macros define
650    the alternative names that are used. */
651    
652    #define allow_zero    cur_is_word
653    #define cbegroup      condition
654    #define code_offset   codelink
655    #define condassert    condition
656    #define matched_once  prev_is_word
657    #define foc           number
658    #define save_mark     data
659    
660  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
661  variables. */  variables. */
662    
# Line 600  TAIL_RECURSE: Line 676  TAIL_RECURSE:
676  /* 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
677  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
678  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()
679  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
680  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
681  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,
682  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
683    
684  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
685  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
686  #else  #else
687  utf8 = FALSE;  utf = FALSE;
688  #endif  #endif
689    
690  /* 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 693  haven't exceeded the recursive call limi
693  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
694  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
695    
 original_ims = ims;    /* Save for resetting on ')' */  
   
696  /* 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
697  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
698  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
699  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
700  When match() is called in other circumstances, don't add to the chain. The  
701  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
702  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
703  match(). */  to break infinite loops that match no characters. When match() is called in
704    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
705    NOT be used with tail recursion, because the memory block that is used is on
706    the stack, so a new one may be required for each match(). */
707    
708  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
709    {    {
710    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
711    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
712    eptrb = &newptrb;    eptrb = &newptrb;
713      md->match_function_type = 0;
714    }    }
715    
716  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 642  for (;;) Line 720  for (;;)
720    minimize = possessive = FALSE;    minimize = possessive = FALSE;
721    op = *ecode;    op = *ecode;
722    
   /* 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;  
   
723    switch(op)    switch(op)
724      {      {
725        case OP_MARK:
726        md->nomatch_mark = ecode + 2;
727        md->mark = NULL;    /* In case previously set by assertion */
728        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
729          eptrb, RM55);
730        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
731             md->mark == NULL) md->mark = ecode + 2;
732    
733        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
734        argument, and we must check whether that argument matches this MARK's
735        argument. It is passed back in md->start_match_ptr (an overloading of that
736        variable). If it does match, we reset that variable to the current subject
737        position and return MATCH_SKIP. Otherwise, pass back the return code
738        unaltered. */
739    
740        else if (rrc == MATCH_SKIP_ARG &&
741            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
742          {
743          md->start_match_ptr = eptr;
744          RRETURN(MATCH_SKIP);
745          }
746        RRETURN(rrc);
747    
748      case OP_FAIL:      case OP_FAIL:
749      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
750    
751      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);  
752    
753      case OP_COMMIT:      case OP_COMMIT:
754      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
755        ims, eptrb, flags, RM52);        eptrb, RM52);
756      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
757            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
758            rrc != MATCH_THEN)
759          RRETURN(rrc);
760      RRETURN(MATCH_COMMIT);      RRETURN(MATCH_COMMIT);
761    
762        /* PRUNE overrides THEN */
763    
764        case OP_PRUNE:
765        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
766          eptrb, RM51);
767        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
768        RRETURN(MATCH_PRUNE);
769    
770        case OP_PRUNE_ARG:
771        md->nomatch_mark = ecode + 2;
772        md->mark = NULL;    /* In case previously set by assertion */
773        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
774          eptrb, RM56);
775        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
776             md->mark == NULL) md->mark = ecode + 2;
777        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
778        RRETURN(MATCH_PRUNE);
779    
780        /* SKIP overrides PRUNE and THEN */
781    
782      case OP_SKIP:      case OP_SKIP:
783      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
784        ims, eptrb, flags, RM53);        eptrb, RM53);
785      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
786          RRETURN(rrc);
787      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
788      RRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
789    
790        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
791        nomatch_mark. There is a flag that disables this opcode when re-matching a
792        pattern that ended with a SKIP for which there was not a matching MARK. */
793    
794        case OP_SKIP_ARG:
795        if (md->ignore_skip_arg)
796          {
797          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
798          break;
799          }
800        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
801          eptrb, RM57);
802        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
803          RRETURN(rrc);
804    
805        /* Pass back the current skip name by overloading md->start_match_ptr and
806        returning the special MATCH_SKIP_ARG return code. This will either be
807        caught by a matching MARK, or get to the top, where it causes a rematch
808        with the md->ignore_skip_arg flag set. */
809    
810        md->start_match_ptr = ecode + 2;
811        RRETURN(MATCH_SKIP_ARG);
812    
813        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
814        the branch in which it occurs can be determined. Overload the start of
815        match pointer to do this. */
816    
817      case OP_THEN:      case OP_THEN:
818      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
819        ims, eptrb, flags, RM54);        eptrb, RM54);
820      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
821        md->start_match_ptr = ecode;
822      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
823    
824      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_THEN_ARG:
825      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
826      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
827      set from a previous iteration of this group, and be referred to by a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
828      reference inside the group.        md, eptrb, RM58);
829        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
830      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
831      values of the final offsets, in case they were set by a previous iteration      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
832      of the same bracket.      md->start_match_ptr = ecode;
833        RRETURN(MATCH_THEN);
834    
835        /* Handle an atomic group that does not contain any capturing parentheses.
836        This can be handled like an assertion. Prior to 8.13, all atomic groups
837        were handled this way. In 8.13, the code was changed as below for ONCE, so
838        that backups pass through the group and thereby reset captured values.
839        However, this uses a lot more stack, so in 8.20, atomic groups that do not
840        contain any captures generate OP_ONCE_NC, which can be handled in the old,
841        less stack intensive way.
842    
843        Check the alternative branches in turn - the matching won't pass the KET
844        for this kind of subpattern. If any one branch matches, we carry on as at
845        the end of a normal bracket, leaving the subject pointer, but resetting
846        the start-of-match value in case it was changed by \K. */
847    
848        case OP_ONCE_NC:
849        prev = ecode;
850        saved_eptr = eptr;
851        save_mark = md->mark;
852        do
853          {
854          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
855          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
856            {
857            mstart = md->start_match_ptr;
858            break;
859            }
860          if (rrc == MATCH_THEN)
861            {
862            next = ecode + GET(ecode,1);
863            if (md->start_match_ptr < next &&
864                (*ecode == OP_ALT || *next == OP_ALT))
865              rrc = MATCH_NOMATCH;
866            }
867    
868          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
869          ecode += GET(ecode,1);
870          md->mark = save_mark;
871          }
872        while (*ecode == OP_ALT);
873    
874        /* If hit the end of the group (which could be repeated), fail */
875    
876        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
877    
878        /* Continue as from after the group, updating the offsets high water
879        mark, since extracts may have been taken. */
880    
881        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
882    
883        offset_top = md->end_offset_top;
884        eptr = md->end_match_ptr;
885    
886        /* For a non-repeating ket, just continue at this level. This also
887        happens for a repeating ket if no characters were matched in the group.
888        This is the forcible breaking of infinite loops as implemented in Perl
889        5.005. */
890    
891        if (*ecode == OP_KET || eptr == saved_eptr)
892          {
893          ecode += 1+LINK_SIZE;
894          break;
895          }
896    
897        /* The repeating kets try the rest of the pattern or restart from the
898        preceding bracket, in the appropriate order. The second "call" of match()
899        uses tail recursion, to avoid using another stack frame. */
900    
901        if (*ecode == OP_KETRMIN)
902          {
903          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
904          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
905          ecode = prev;
906          goto TAIL_RECURSE;
907          }
908        else  /* OP_KETRMAX */
909          {
910          md->match_function_type = MATCH_CBEGROUP;
911          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
912          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
913          ecode += 1 + LINK_SIZE;
914          goto TAIL_RECURSE;
915          }
916        /* Control never gets here */
917    
918        /* Handle a capturing bracket, other than those that are possessive with an
919        unlimited repeat. If there is space in the offset vector, save the current
920        subject position in the working slot at the top of the vector. We mustn't
921        change the current values of the data slot, because they may be set from a
922        previous iteration of this group, and be referred to by a reference inside
923        the group. A failure to match might occur after the group has succeeded,
924        if something later on doesn't match. For this reason, we need to restore
925        the working value and also the values of the final offsets, in case they
926        were set by a previous iteration of the same bracket.
927    
928      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
929      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 934  for (;;)
934      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
935      offset = number << 1;      offset = number << 1;
936    
937  #ifdef DEBUG  #ifdef PCRE_DEBUG
938      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
939      printf("subject=");      printf("subject=");
940      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 712  for (;;) Line 947  for (;;)
947        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
948        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
949        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
950          save_mark = md->mark;
951    
952        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
953        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
954            (int)(eptr - md->start_subject);
955    
956        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
957          {          {
958          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
959            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
960          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
961            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
962    
963            /* If we backed up to a THEN, check whether it is within the current
964            branch by comparing the address of the THEN that is passed back with
965            the end of the branch. If it is within the current branch, and the
966            branch is one of two or more alternatives (it either starts or ends
967            with OP_ALT), we have reached the limit of THEN's action, so convert
968            the return code to NOMATCH, which will cause normal backtracking to
969            happen from now on. Otherwise, THEN is passed back to an outer
970            alternative. This implements Perl's treatment of parenthesized groups,
971            where a group not containing | does not affect the current alternative,
972            that is, (X) is NOT the same as (X|(*F)). */
973    
974            if (rrc == MATCH_THEN)
975              {
976              next = ecode + GET(ecode,1);
977              if (md->start_match_ptr < next &&
978                  (*ecode == OP_ALT || *next == OP_ALT))
979                rrc = MATCH_NOMATCH;
980              }
981    
982            /* Anything other than NOMATCH is passed back. */
983    
984            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
985          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
986          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
987            md->mark = save_mark;
988            if (*ecode != OP_ALT) break;
989          }          }
       while (*ecode == OP_ALT);  
990    
991        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
992        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
993        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
994        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
995    
996        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
997    
998          RRETURN(rrc);
999        }        }
1000    
1001      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 747  for (;;) Line 1009  for (;;)
1009      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1010      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1011    
1012      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1013      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion, except in the case when match_cbegroup  
     is set.*/  
1014    
1015        When we get to the final alternative within the brackets, we used to return
1016        the result of a recursive call to match() whatever happened so it was
1017        possible to reduce stack usage by turning this into a tail recursion,
1018        except in the case of a possibly empty group. However, now that there is
1019        the possiblity of (*THEN) occurring in the final alternative, this
1020        optimization is no longer always possible.
1021    
1022        We can optimize if we know there are no (*THEN)s in the pattern; at present
1023        this is the best that can be done.
1024    
1025        MATCH_ONCE is returned when the end of an atomic group is successfully
1026        reached, but subsequent matching fails. It passes back up the tree (causing
1027        captured values to be reset) until the original atomic group level is
1028        reached. This is tested by comparing md->once_target with the start of the
1029        group. At this point, the return is converted into MATCH_NOMATCH so that
1030        previous backup points can be taken. */
1031    
1032        case OP_ONCE:
1033      case OP_BRA:      case OP_BRA:
1034      case OP_SBRA:      case OP_SBRA:
1035      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1036      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1037      for (;;)      for (;;)
1038        {        {
1039        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1040    
1041          /* If this is not a possibly empty group, and there are no (*THEN)s in
1042          the pattern, and this is the final alternative, optimize as described
1043          above. */
1044    
1045          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1046            {
1047            ecode += PRIV(OP_lengths)[*ecode];
1048            goto TAIL_RECURSE;
1049            }
1050    
1051          /* In all other cases, we have to make another call to match(). */
1052    
1053          save_mark = md->mark;
1054          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1055            RM2);
1056    
1057          /* See comment in the code for capturing groups above about handling
1058          THEN. */
1059    
1060          if (rrc == MATCH_THEN)
1061            {
1062            next = ecode + GET(ecode,1);
1063            if (md->start_match_ptr < next &&
1064                (*ecode == OP_ALT || *next == OP_ALT))
1065              rrc = MATCH_NOMATCH;
1066            }
1067    
1068          if (rrc != MATCH_NOMATCH)
1069          {          {
1070          if (flags == 0)    /* Not a possibly empty group */          if (rrc == MATCH_ONCE)
1071            {            {
1072            ecode += _pcre_OP_lengths[*ecode];            const pcre_uchar *scode = ecode;
1073            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
1074            goto TAIL_RECURSE;              {
1075                while (*scode == OP_ALT) scode += GET(scode, 1);
1076                scode -= GET(scode, 1);
1077                }
1078              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1079            }            }
1080            RRETURN(rrc);
1081            }
1082          ecode += GET(ecode, 1);
1083          md->mark = save_mark;
1084          if (*ecode != OP_ALT) break;
1085          }
1086    
1087        RRETURN(MATCH_NOMATCH);
1088    
1089          /* Possibly empty group; can't use tail recursion. */      /* Handle possessive capturing brackets with an unlimited repeat. We come
1090        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1091        handled similarly to the normal case above. However, the matching is
1092        different. The end of these brackets will always be OP_KETRPOS, which
1093        returns MATCH_KETRPOS without going further in the pattern. By this means
1094        we can handle the group by iteration rather than recursion, thereby
1095        reducing the amount of stack needed. */
1096    
1097        case OP_CBRAPOS:
1098        case OP_SCBRAPOS:
1099        allow_zero = FALSE;
1100    
1101          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      POSSESSIVE_CAPTURE:
1102            eptrb, flags, RM48);      number = GET2(ecode, 1+LINK_SIZE);
1103          RRETURN(rrc);      offset = number << 1;
1104    
1105    #ifdef PCRE_DEBUG
1106        printf("start possessive bracket %d\n", number);
1107        printf("subject=");
1108        pchars(eptr, 16, TRUE, md);
1109        printf("\n");
1110    #endif
1111    
1112        if (offset < md->offset_max)
1113          {
1114          matched_once = FALSE;
1115          code_offset = (int)(ecode - md->start_code);
1116    
1117          save_offset1 = md->offset_vector[offset];
1118          save_offset2 = md->offset_vector[offset+1];
1119          save_offset3 = md->offset_vector[md->offset_end - number];
1120          save_capture_last = md->capture_last;
1121    
1122          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1123    
1124          /* Each time round the loop, save the current subject position for use
1125          when the group matches. For MATCH_MATCH, the group has matched, so we
1126          restart it with a new subject starting position, remembering that we had
1127          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1128          usual. If we haven't matched any alternatives in any iteration, check to
1129          see if a previous iteration matched. If so, the group has matched;
1130          continue from afterwards. Otherwise it has failed; restore the previous
1131          capture values before returning NOMATCH. */
1132    
1133          for (;;)
1134            {
1135            md->offset_vector[md->offset_end - number] =
1136              (int)(eptr - md->start_subject);
1137            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1138            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1139              eptrb, RM63);
1140            if (rrc == MATCH_KETRPOS)
1141              {
1142              offset_top = md->end_offset_top;
1143              eptr = md->end_match_ptr;
1144              ecode = md->start_code + code_offset;
1145              save_capture_last = md->capture_last;
1146              matched_once = TRUE;
1147              continue;
1148              }
1149    
1150            /* See comment in the code for capturing groups above about handling
1151            THEN. */
1152    
1153            if (rrc == MATCH_THEN)
1154              {
1155              next = ecode + GET(ecode,1);
1156              if (md->start_match_ptr < next &&
1157                  (*ecode == OP_ALT || *next == OP_ALT))
1158                rrc = MATCH_NOMATCH;
1159              }
1160    
1161            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1162            md->capture_last = save_capture_last;
1163            ecode += GET(ecode, 1);
1164            if (*ecode != OP_ALT) break;
1165          }          }
1166    
1167        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (!matched_once)
1168        otherwise return. */          {
1169            md->offset_vector[offset] = save_offset1;
1170            md->offset_vector[offset+1] = save_offset2;
1171            md->offset_vector[md->offset_end - number] = save_offset3;
1172            }
1173    
1174        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        if (allow_zero || matched_once)
1175          eptrb, flags, RM2);          {
1176        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);          ecode += 1 + LINK_SIZE;
1177            break;
1178            }
1179    
1180          RRETURN(MATCH_NOMATCH);
1181          }
1182    
1183        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1184        as a non-capturing bracket. */
1185    
1186        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1187        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1188    
1189        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1190    
1191        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1192        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1193    
1194        /* Non-capturing possessive bracket with unlimited repeat. We come here
1195        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1196        without the capturing complication. It is written out separately for speed
1197        and cleanliness. */
1198    
1199        case OP_BRAPOS:
1200        case OP_SBRAPOS:
1201        allow_zero = FALSE;
1202    
1203        POSSESSIVE_NON_CAPTURE:
1204        matched_once = FALSE;
1205        code_offset = (int)(ecode - md->start_code);
1206    
1207        for (;;)
1208          {
1209          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1210          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1211            eptrb, RM48);
1212          if (rrc == MATCH_KETRPOS)
1213            {
1214            offset_top = md->end_offset_top;
1215            eptr = md->end_match_ptr;
1216            ecode = md->start_code + code_offset;
1217            matched_once = TRUE;
1218            continue;
1219            }
1220    
1221          /* See comment in the code for capturing groups above about handling
1222          THEN. */
1223    
1224          if (rrc == MATCH_THEN)
1225            {
1226            next = ecode + GET(ecode,1);
1227            if (md->start_match_ptr < next &&
1228                (*ecode == OP_ALT || *next == OP_ALT))
1229              rrc = MATCH_NOMATCH;
1230            }
1231    
1232          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1233        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1234          if (*ecode != OP_ALT) break;
1235          }
1236    
1237        if (matched_once || allow_zero)
1238          {
1239          ecode += 1 + LINK_SIZE;
1240          break;
1241        }        }
1242        RRETURN(MATCH_NOMATCH);
1243    
1244      /* Control never reaches here. */      /* Control never reaches here. */
1245    
1246      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1247      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
1248      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
1249      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. */  
1250    
1251      case OP_COND:      case OP_COND:
1252      case OP_SCOND:      case OP_SCOND:
1253      codelink= GET(ecode, 1);      codelink = GET(ecode, 1);
1254    
1255      /* 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
1256      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
1257    
1258      if (ecode[LINK_SIZE+1] == OP_CALLOUT)      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1259        {        {
1260        if (pcre_callout != NULL)        if (PUBL(callout) != NULL)
1261          {          {
1262          pcre_callout_block cb;          PUBL(callout_block) cb;
1263          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1264          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1265          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1266    #ifdef COMPILE_PCRE8
1267          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
1268          cb.subject_length   = md->end_subject - md->start_subject;  #else
1269          cb.start_match      = mstart - md->start_subject;          cb.subject          = (PCRE_SPTR16)md->start_subject;
1270          cb.current_position = eptr - md->start_subject;  #endif
1271            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1272            cb.start_match      = (int)(mstart - md->start_subject);
1273            cb.current_position = (int)(eptr - md->start_subject);
1274          cb.pattern_position = GET(ecode, LINK_SIZE + 3);          cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1275          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1276          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1277          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1278          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1279          if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);          cb.mark             = md->nomatch_mark;
1280            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1281          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1282          }          }
1283        ecode += _pcre_OP_lengths[OP_CALLOUT];        ecode += PRIV(OP_lengths)[OP_CALLOUT];
1284        }        }
1285    
1286      condcode = ecode[LINK_SIZE+1];      condcode = ecode[LINK_SIZE+1];
1287    
1288      /* Now see what the actual condition is */      /* Now see what the actual condition is */
1289    
1290      if (condcode == OP_RREF)         /* Recursion test */      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1291        {        {
1292        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/        if (md->recursive == NULL)                /* Not recursing => FALSE */
1293        condition = md->recursive != NULL &&          {
1294          (offset == RREF_ANY || offset == md->recursive->group_num);          condition = FALSE;
1295        ecode += condition? 3 : GET(ecode, 1);          ecode += GET(ecode, 1);
1296            }
1297          else
1298            {
1299            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1300            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1301    
1302            /* If the test is for recursion into a specific subpattern, and it is
1303            false, but the test was set up by name, scan the table to see if the
1304            name refers to any other numbers, and test them. The condition is true
1305            if any one is set. */
1306    
1307            if (!condition && condcode == OP_NRREF)
1308              {
1309              pcre_uchar *slotA = md->name_table;
1310              for (i = 0; i < md->name_count; i++)
1311                {
1312                if (GET2(slotA, 0) == recno) break;
1313                slotA += md->name_entry_size;
1314                }
1315    
1316              /* Found a name for the number - there can be only one; duplicate
1317              names for different numbers are allowed, but not vice versa. First
1318              scan down for duplicates. */
1319    
1320              if (i < md->name_count)
1321                {
1322                pcre_uchar *slotB = slotA;
1323                while (slotB > md->name_table)
1324                  {
1325                  slotB -= md->name_entry_size;
1326                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1327                    {
1328                    condition = GET2(slotB, 0) == md->recursive->group_num;
1329                    if (condition) break;
1330                    }
1331                  else break;
1332                  }
1333    
1334                /* Scan up for duplicates */
1335    
1336                if (!condition)
1337                  {
1338                  slotB = slotA;
1339                  for (i++; i < md->name_count; i++)
1340                    {
1341                    slotB += md->name_entry_size;
1342                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1343                      {
1344                      condition = GET2(slotB, 0) == md->recursive->group_num;
1345                      if (condition) break;
1346                      }
1347                    else break;
1348                    }
1349                  }
1350                }
1351              }
1352    
1353            /* Chose branch according to the condition */
1354    
1355            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1356            }
1357        }        }
1358    
1359      else if (condcode == OP_CREF)    /* Group used test */      else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1360        {        {
1361        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1362        condition = offset < offset_top && md->offset_vector[offset] >= 0;        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1363        ecode += condition? 3 : GET(ecode, 1);  
1364          /* If the numbered capture is unset, but the reference was by name,
1365          scan the table to see if the name refers to any other numbers, and test
1366          them. The condition is true if any one is set. This is tediously similar
1367          to the code above, but not close enough to try to amalgamate. */
1368    
1369          if (!condition && condcode == OP_NCREF)
1370            {
1371            int refno = offset >> 1;
1372            pcre_uchar *slotA = md->name_table;
1373    
1374            for (i = 0; i < md->name_count; i++)
1375              {
1376              if (GET2(slotA, 0) == refno) break;
1377              slotA += md->name_entry_size;
1378              }
1379    
1380            /* Found a name for the number - there can be only one; duplicate names
1381            for different numbers are allowed, but not vice versa. First scan down
1382            for duplicates. */
1383    
1384            if (i < md->name_count)
1385              {
1386              pcre_uchar *slotB = slotA;
1387              while (slotB > md->name_table)
1388                {
1389                slotB -= md->name_entry_size;
1390                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1391                  {
1392                  offset = GET2(slotB, 0) << 1;
1393                  condition = offset < offset_top &&
1394                    md->offset_vector[offset] >= 0;
1395                  if (condition) break;
1396                  }
1397                else break;
1398                }
1399    
1400              /* Scan up for duplicates */
1401    
1402              if (!condition)
1403                {
1404                slotB = slotA;
1405                for (i++; i < md->name_count; i++)
1406                  {
1407                  slotB += md->name_entry_size;
1408                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1409                    {
1410                    offset = GET2(slotB, 0) << 1;
1411                    condition = offset < offset_top &&
1412                      md->offset_vector[offset] >= 0;
1413                    if (condition) break;
1414                    }
1415                  else break;
1416                  }
1417                }
1418              }
1419            }
1420    
1421          /* Chose branch according to the condition */
1422    
1423          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1424        }        }
1425    
1426      else if (condcode == OP_DEF)     /* DEFINE - always false */      else if (condcode == OP_DEF)     /* DEFINE - always false */
# Line 847  for (;;) Line 1430  for (;;)
1430        }        }
1431    
1432      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1433      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
1434      assertion. */      an assertion. */
1435    
1436      else      else
1437        {        {
1438        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1439            match_condassert, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1440        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1441          {          {
1442            if (md->end_offset_top > offset_top)
1443              offset_top = md->end_offset_top;  /* Captures may have happened */
1444          condition = TRUE;          condition = TRUE;
1445          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1446          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1447          }          }
1448    
1449          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1450          assertion; it is therefore treated as NOMATCH. */
1451    
1452        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1453          {          {
1454          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
# Line 871  for (;;) Line 1460  for (;;)
1460          }          }
1461        }        }
1462    
1463      /* 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
1464      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
1465      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
1466      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
1467        exist, in which case we can just plough on. Note that, for compatibility
1468        with Perl, the | in a conditional group is NOT treated as creating two
1469        alternatives. If a THEN is encountered in the branch, it propagates out to
1470        the enclosing alternative (unless nested in a deeper set of alternatives,
1471        of course). */
1472    
1473      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1474        {        {
1475        ecode += 1 + LINK_SIZE;        if (op != OP_SCOND)
       if (op == OP_SCOND)        /* Possibly empty group */  
1476          {          {
1477          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);          ecode += 1 + LINK_SIZE;
         RRETURN(rrc);  
         }  
       else                       /* Group must match something */  
         {  
         flags = 0;  
1478          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1479          }          }
1480    
1481          md->match_function_type = MATCH_CBEGROUP;
1482          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1483          RRETURN(rrc);
1484        }        }
1485      else                         /* Condition false & no alternative */  
1486         /* Condition false & no alternative; continue after the group. */
1487    
1488        else
1489        {        {
1490        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1491        }        }
1492      break;      break;
1493    
1494    
1495      /* 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,
1496      recursion, we should restore the offsets appropriately and continue from      to close any currently open capturing brackets. */
     after the call. */  
1497    
1498      case OP_ACCEPT:      case OP_CLOSE:
1499      case OP_END:      number = GET2(ecode, 1);
1500      if (md->recursive != NULL && md->recursive->group_num == 0)      offset = number << 1;
1501    
1502    #ifdef PCRE_DEBUG
1503          printf("end bracket %d at *ACCEPT", number);
1504          printf("\n");
1505    #endif
1506    
1507        md->capture_last = number;
1508        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1509        {        {
1510        recursion_info *rec = md->recursive;        md->offset_vector[offset] =
1511        DPRINTF(("End of pattern in a (?0) recursion\n"));          md->offset_vector[md->offset_end - number];
1512        md->recursive = rec->prevrec;        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1513        memmove(md->offset_vector, rec->offset_save,        if (offset_top <= offset) offset_top = offset + 2;
         rec->saved_max * sizeof(int));  
       mstart = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
1514        }        }
1515        ecode += 1 + IMM2_SIZE;
1516        break;
1517    
1518    
1519        /* End of the pattern, either real or forced. */
1520    
1521        case OP_END:
1522        case OP_ACCEPT:
1523        case OP_ASSERT_ACCEPT:
1524    
1525        /* If we have matched an empty string, fail if not in an assertion and not
1526        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1527        is set and we have matched at the start of the subject. In both cases,
1528        backtracking will then try other alternatives, if any. */
1529    
1530        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1531             md->recursive == NULL &&
1532             (md->notempty ||
1533               (md->notempty_atstart &&
1534                 mstart == md->start_subject + md->start_offset)))
1535          RRETURN(MATCH_NOMATCH);
1536    
1537      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Otherwise, we have a match. */
     string - backtracking will then try other alternatives, if any. */  
1538    
     if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);  
1539      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1540      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1541      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);  
1542    
1543      /* Change option settings */      /* For some reason, the macros don't work properly if an expression is
1544        given as the argument to RRETURN when the heap is in use. */
1545    
1546      case OP_OPT:      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1547      ims = ecode[1];      RRETURN(rrc);
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1548    
1549      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1550      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,
1551      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
1552      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
1553      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1554        of a condition, we want to return immediately afterwards. The caller of
1555        this incarnation of the match() function will have set MATCH_CONDASSERT in
1556        md->match_function type, and one of these opcodes will be the first opcode
1557        that is processed. We use a local variable that is preserved over calls to
1558        match() to remember this case. */
1559    
1560      case OP_ASSERT:      case OP_ASSERT:
1561      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1562        save_mark = md->mark;
1563        if (md->match_function_type == MATCH_CONDASSERT)
1564          {
1565          condassert = TRUE;
1566          md->match_function_type = 0;
1567          }
1568        else condassert = FALSE;
1569    
1570      do      do
1571        {        {
1572        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1573          RM4);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1574        if (rrc == MATCH_MATCH) break;          {
1575        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);          mstart = md->start_match_ptr;   /* In case \K reset it */
1576            break;
1577            }
1578    
1579          /* PCRE does not allow THEN or COMMIT to escape beyond an assertion; it
1580          is treated as NOMATCH. */
1581    
1582          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1583              rrc != MATCH_COMMIT) RRETURN(rrc);
1584    
1585        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1586          md->mark = save_mark;
1587        }        }
1588      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1589    
1590      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1591    
1592      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1593    
1594      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1595    
1596      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1597      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 964  for (;;) Line 1601  for (;;)
1601      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1602      continue;      continue;
1603    
1604      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1605        PRUNE, or COMMIT means we must assume failure without checking subsequent
1606        branches. */
1607    
1608      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1609      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1610        save_mark = md->mark;
1611        if (md->match_function_type == MATCH_CONDASSERT)
1612          {
1613          condassert = TRUE;
1614          md->match_function_type = 0;
1615          }
1616        else condassert = FALSE;
1617    
1618      do      do
1619        {        {
1620        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1621          RM5);        md->mark = save_mark;
1622        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1623          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1624            {
1625            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1626            break;
1627            }
1628    
1629          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1630          as NOMATCH. */
1631    
1632        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1633        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1634        }        }
1635      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1636    
1637      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1638    
1639      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1640      continue;      continue;
# Line 989  for (;;) Line 1645  for (;;)
1645      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1646    
1647      case OP_REVERSE:      case OP_REVERSE:
1648  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1649      if (utf8)      if (utf)
1650        {        {
1651        i = GET(ecode, 1);        i = GET(ecode, 1);
1652        while (i-- > 0)        while (i-- > 0)
# Line 1010  for (;;) Line 1666  for (;;)
1666        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1667        }        }
1668    
1669      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1670    
1671        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1672      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1673      break;      break;
1674    
# Line 1020  for (;;) Line 1677  for (;;)
1677      function is able to force a failure. */      function is able to force a failure. */
1678    
1679      case OP_CALLOUT:      case OP_CALLOUT:
1680      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1681        {        {
1682        pcre_callout_block cb;        PUBL(callout_block) cb;
1683        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1684        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1685        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1686    #ifdef COMPILE_PCRE8
1687        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1688        cb.subject_length   = md->end_subject - md->start_subject;  #else
1689        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1690        cb.current_position = eptr - md->start_subject;  #endif
1691          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1692          cb.start_match      = (int)(mstart - md->start_subject);
1693          cb.current_position = (int)(eptr - md->start_subject);
1694        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1695        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1696        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1697        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1698        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1699        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1700          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1701        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1702        }        }
1703      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 1045  for (;;) Line 1707  for (;;)
1707      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
1708      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1709    
1710      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1711      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
1712      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
1713      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
1714      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
1715      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
1716      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.  
1717    
1718      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
1719      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
1720      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1721        a lot, so he is not to blame for the current way it works. */
1722    
1723      case OP_RECURSE:      case OP_RECURSE:
1724        {        {
1725          recursion_info *ri;
1726          int recno;
1727    
1728        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1729        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1730          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1731    
1732          /* Check for repeating a recursion without advancing the subject pointer.
1733          This should catch convoluted mutual recursions. (Some simple cases are
1734          caught at compile time.) */
1735    
1736          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1737            if (recno == ri->group_num && eptr == ri->subject_position)
1738              RRETURN(PCRE_ERROR_RECURSELOOP);
1739    
1740        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1741    
1742          new_recursive.group_num = recno;
1743          new_recursive.subject_position = eptr;
1744        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1745        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1746    
1747        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1748    
1749        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1750    
1751        /* Now save the offset data. */        /* Now save the offset data */
1752    
1753        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1754        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1084  for (;;) Line 1756  for (;;)
1756        else        else
1757          {          {
1758          new_recursive.offset_save =          new_recursive.offset_save =
1759            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1760          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1761          }          }
   
1762        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1763              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1764    
1765        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1766        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1767          might be changed, so reset it before looping. */
1768    
1769        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1770        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1771        do        do
1772          {          {
1773          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1774            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1775          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1776            memcpy(md->offset_vector, new_recursive.offset_save,
1777                new_recursive.saved_max * sizeof(int));
1778            md->recursive = new_recursive.prevrec;
1779            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1780            {            {
1781            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1782            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1783              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1784            RRETURN(MATCH_MATCH);  
1785              /* Set where we got to in the subject, and reset the start in case
1786              it was changed by \K. This *is* propagated back out of a recursion,
1787              for Perl compatibility. */
1788    
1789              eptr = md->end_match_ptr;
1790              mstart = md->start_match_ptr;
1791              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1792            }            }
1793          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1794            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1795            is treated as NOMATCH. */
1796    
1797            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1798                     rrc != MATCH_COMMIT)
1799            {            {
1800            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1801            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1802              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1803            RRETURN(rrc);            RRETURN(rrc);
1804            }            }
1805    
1806          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1807          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1808          }          }
1809        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1128  for (;;) Line 1811  for (;;)
1811        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1812        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1813        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1814          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1815        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1816        }        }
     /* Control never reaches here */  
1817    
1818      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1819      the point in the subject string is not moved back. Thus there can never be      break;
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1820    
1821      /* 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
1822      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1221  for (;;) Line 1832  for (;;)
1832      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1833    
1834      case OP_BRAZERO:      case OP_BRAZERO:
1835        {      next = ecode + 1;
1836        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1837        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1838        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1839        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1840      break;      break;
1841    
1842      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1843        {      next = ecode + 1;
1844        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1845        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1846        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1847        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1848      break;      break;
1849    
1850      case OP_SKIPZERO:      case OP_SKIPZERO:
1851        {      next = ecode+1;
1852        next = ecode+1;      do next += GET(next,1); while (*next == OP_ALT);
1853        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1854      break;      break;
1855    
1856        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1857        here; just jump to the group, with allow_zero set TRUE. */
1858    
1859        case OP_BRAPOSZERO:
1860        op = *(++ecode);
1861        allow_zero = TRUE;
1862        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1863          goto POSSESSIVE_NON_CAPTURE;
1864    
1865      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1866    
1867      case OP_KET:      case OP_KET:
1868      case OP_KETRMIN:      case OP_KETRMIN:
1869      case OP_KETRMAX:      case OP_KETRMAX:
1870        case OP_KETRPOS:
1871      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1872    
1873      /* 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
1874      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1875      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1876    
1877      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1878        {        {
1879        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1880        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1881        }        }
1882      else saved_eptr = NULL;      else saved_eptr = NULL;
1883    
1884      /* 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
1885      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
1886      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1887        start in case it was changed by \K. */
1888      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1889          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1890          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1891        {        {
1892        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1893        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1894        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1895          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1896        }        }
1897    
1898      /* 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
1899      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1900      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
1901      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
1902      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
1903        the current subject position and start match pointer and give a MATCH
1904        return. */
1905    
1906      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1907            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1908        {        {
1909        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1910        offset = number << 1;        offset = number << 1;
1911    
1912  #ifdef DEBUG  #ifdef PCRE_DEBUG
1913        printf("end bracket %d", number);        printf("end bracket %d", number);
1914        printf("\n");        printf("\n");
1915  #endif  #endif
1916    
1917          /* Handle a recursively called group. */
1918    
1919          if (md->recursive != NULL && md->recursive->group_num == number)
1920            {
1921            md->end_match_ptr = eptr;
1922            md->start_match_ptr = mstart;
1923            RRETURN(MATCH_MATCH);
1924            }
1925    
1926          /* Deal with capturing */
1927    
1928        md->capture_last = number;        md->capture_last = number;
1929        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1930          {          {
1931            /* If offset is greater than offset_top, it means that we are
1932            "skipping" a capturing group, and that group's offsets must be marked
1933            unset. In earlier versions of PCRE, all the offsets were unset at the
1934            start of matching, but this doesn't work because atomic groups and
1935            assertions can cause a value to be set that should later be unset.
1936            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1937            part of the atomic group, but this is not on the final matching path,
1938            so must be unset when 2 is set. (If there is no group 2, there is no
1939            problem, because offset_top will then be 2, indicating no capture.) */
1940    
1941            if (offset > offset_top)
1942              {
1943              register int *iptr = md->offset_vector + offset_top;
1944              register int *iend = md->offset_vector + offset;
1945              while (iptr < iend) *iptr++ = -1;
1946              }
1947    
1948            /* Now make the extraction */
1949    
1950          md->offset_vector[offset] =          md->offset_vector[offset] =
1951            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1952          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1953          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1954          }          }
1955          }
1956    
1957        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1958        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1959        group. This is the forcible breaking of infinite loops as implemented in
1960        Perl 5.005. For a non-repeating atomic group that includes captures,
1961        establish a backup point by processing the rest of the pattern at a lower
1962        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1963        original OP_ONCE level, thereby bypassing intermediate backup points, but
1964        resetting any captures that happened along the way. */
1965    
1966        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1967          {
1968          if (*prev == OP_ONCE)
1969          {          {
1970          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1971          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1972          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1973          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;  
1974          }          }
1975          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1976          break;
1977        }        }
1978    
1979      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1980      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1981        at a time from the outer level, thus saving stack. */
1982    
1983      ims = original_ims;      if (*ecode == OP_KETRPOS)
     DPRINTF(("ims reset to %02lx\n", ims));  
   
     /* 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)  
1984        {        {
1985        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1986        break;        md->end_offset_top = offset_top;
1987          RRETURN(MATCH_KETRPOS);
1988        }        }
1989    
1990      /* 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
1991      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
1992      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
1993      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
1994        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
1995    
1996      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1997        {        {
1998        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1999        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2000        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2001          {          {
2002          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2003            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2004            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2005            RRETURN(MATCH_ONCE);
2006            }
2007          if (*prev >= OP_SBRA)    /* Could match an empty string */
2008            {
2009            md->match_function_type = MATCH_CBEGROUP;
2010            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2011          RRETURN(rrc);          RRETURN(rrc);
2012          }          }
2013        ecode = prev;        ecode = prev;
# Line 1360  for (;;) Line 2015  for (;;)
2015        }        }
2016      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2017        {        {
2018        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
2019          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2020          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2021        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2022          if (*prev == OP_ONCE)
2023            {
2024            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2025            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2026            md->once_target = prev;
2027            RRETURN(MATCH_ONCE);
2028            }
2029        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2030        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2031        }        }
2032      /* Control never gets here */      /* Control never gets here */
2033    
2034      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2035    
2036      case OP_CIRC:      case OP_CIRC:
2037      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 */  
2038    
2039      /* Start of subject assertion */      /* Start of subject assertion */
2040    
# Line 1389  for (;;) Line 2043  for (;;)
2043      ecode++;      ecode++;
2044      break;      break;
2045    
2046        /* Multiline mode: start of subject unless notbol, or after any newline. */
2047    
2048        case OP_CIRCM:
2049        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2050        if (eptr != md->start_subject &&
2051            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2052          RRETURN(MATCH_NOMATCH);
2053        ecode++;
2054        break;
2055    
2056      /* Start of match assertion */      /* Start of match assertion */
2057    
2058      case OP_SOM:      case OP_SOM:
# Line 1403  for (;;) Line 2067  for (;;)
2067      ecode++;      ecode++;
2068      break;      break;
2069    
2070      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2071      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2072    
2073      case OP_DOLL:      case OP_DOLLM:
2074      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2075        {        {
2076        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2077          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2078        else          if (md->partial != 0 &&
2079          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2080        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2081        break;              NLBLOCK->nllen == 2 &&
2082                *eptr == NLBLOCK->nl[0])
2083              {
2084              md->hitend = TRUE;
2085              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2086              }
2087            RRETURN(MATCH_NOMATCH);
2088            }
2089        }        }
2090      else      else
2091        {        {
2092        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2093        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2094        }        }
2095        ecode++;
2096        break;
2097    
2098        /* Not multiline mode: assert before a terminating newline or before end of
2099        subject unless noteol is set. */
2100    
2101        case OP_DOLL:
2102        if (md->noteol) RRETURN(MATCH_NOMATCH);
2103        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2104    
2105      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2106    
2107      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2108    
2109      case OP_EOD:      case OP_EOD:
2110      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2111        SCHECK_PARTIAL();
2112      ecode++;      ecode++;
2113      break;      break;
2114    
2115      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2116    
2117      case OP_EODN:      case OP_EODN:
2118      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2119        if (eptr < md->end_subject &&
2120          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2121          {
2122          if (md->partial != 0 &&
2123              eptr + 1 >= md->end_subject &&
2124              NLBLOCK->nltype == NLTYPE_FIXED &&
2125              NLBLOCK->nllen == 2 &&
2126              *eptr == NLBLOCK->nl[0])
2127            {
2128            md->hitend = TRUE;
2129            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2130            }
2131        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2132          }
2133    
2134        /* Either at end of string or \n before end. */
2135    
2136        SCHECK_PARTIAL();
2137      ecode++;      ecode++;
2138      break;      break;
2139    
# Line 1454  for (;;) Line 2145  for (;;)
2145    
2146        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2147        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
2148        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2149          partial matching. */
2150    
2151  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2152        if (utf8)        if (utf)
2153          {          {
2154            /* Get status of previous character */
2155    
2156          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2157            {            {
2158            USPTR lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2159            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2160              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2161            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2162    #ifdef SUPPORT_UCP
2163              if (md->use_ucp)
2164                {
2165                if (c == '_') prev_is_word = TRUE; else
2166                  {
2167                  int cat = UCD_CATEGORY(c);
2168                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2169                  }
2170                }
2171              else
2172    #endif
2173            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2174            }            }
2175          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2176            /* Get status of next character */
2177    
2178            if (eptr >= md->end_subject)
2179              {
2180              SCHECK_PARTIAL();
2181              cur_is_word = FALSE;
2182              }
2183            else
2184            {            {
2185            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2186    #ifdef SUPPORT_UCP
2187              if (md->use_ucp)
2188                {
2189                if (c == '_') cur_is_word = TRUE; else
2190                  {
2191                  int cat = UCD_CATEGORY(c);
2192                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2193                  }
2194                }
2195              else
2196    #endif
2197            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2198            }            }
2199          }          }
2200        else        else
2201  #endif  #endif
2202    
2203        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2204          consistency with the behaviour of \w we do use it in this case. */
2205    
2206          {          {
2207          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2208            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2209          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2210            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2211              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2212    #ifdef SUPPORT_UCP
2213              if (md->use_ucp)
2214                {
2215                c = eptr[-1];
2216                if (c == '_') prev_is_word = TRUE; else
2217                  {
2218                  int cat = UCD_CATEGORY(c);
2219                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2220                  }
2221                }
2222              else
2223    #endif
2224              prev_is_word = MAX_255(eptr[-1])
2225                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2226              }
2227    
2228            /* Get status of next character */
2229    
2230            if (eptr >= md->end_subject)
2231              {
2232              SCHECK_PARTIAL();
2233              cur_is_word = FALSE;
2234              }
2235            else
2236    #ifdef SUPPORT_UCP
2237            if (md->use_ucp)
2238              {
2239              c = *eptr;
2240              if (c == '_') cur_is_word = TRUE; else
2241                {
2242                int cat = UCD_CATEGORY(c);
2243                cur_is_word = (cat == ucp_L || cat == ucp_N);
2244                }
2245              }
2246            else
2247    #endif
2248            cur_is_word = MAX_255(*eptr)
2249              && ((md->ctypes[*eptr] & ctype_word) != 0);
2250          }          }
2251    
2252        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1492  for (;;) Line 2257  for (;;)
2257        }        }
2258      break;      break;
2259    
2260      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2261        CRLF newlines and partial matching. */
2262    
2263      case OP_ANY:      case OP_ANY:
2264      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2265        if (md->partial != 0 &&
2266            eptr + 1 >= md->end_subject &&
2267            NLBLOCK->nltype == NLTYPE_FIXED &&
2268            NLBLOCK->nllen == 2 &&
2269            *eptr == NLBLOCK->nl[0])
2270          {
2271          md->hitend = TRUE;
2272          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2273          }
2274    
2275      /* Fall through */      /* Fall through */
2276    
2277        /* Match any single character whatsoever. */
2278    
2279      case OP_ALLANY:      case OP_ALLANY:
2280      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2281      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;        {                            /* not be updated before SCHECK_PARTIAL. */
2282          SCHECK_PARTIAL();
2283          RRETURN(MATCH_NOMATCH);
2284          }
2285        eptr++;
2286    #ifdef SUPPORT_UTF
2287        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2288    #endif
2289      ecode++;      ecode++;
2290      break;      break;
2291    
# Line 1508  for (;;) Line 2293  for (;;)
2293      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2294    
2295      case OP_ANYBYTE:      case OP_ANYBYTE:
2296      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2297          {                            /* not be updated before SCHECK_PARTIAL. */
2298          SCHECK_PARTIAL();
2299          RRETURN(MATCH_NOMATCH);
2300          }
2301        eptr++;
2302      ecode++;      ecode++;
2303      break;      break;
2304    
2305      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2306      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2307          {
2308          SCHECK_PARTIAL();
2309          RRETURN(MATCH_NOMATCH);
2310          }
2311      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2312      if (      if (
2313  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2314         c < 256 &&         c < 256 &&
2315  #endif  #endif
2316         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1526  for (;;) Line 2320  for (;;)
2320      break;      break;
2321    
2322      case OP_DIGIT:      case OP_DIGIT:
2323      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2324          {
2325          SCHECK_PARTIAL();
2326          RRETURN(MATCH_NOMATCH);
2327          }
2328      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2329      if (      if (
2330  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2331         c >= 256 ||         c > 255 ||
2332  #endif  #endif
2333         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2334         )         )
# Line 1539  for (;;) Line 2337  for (;;)
2337      break;      break;
2338    
2339      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2340      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2341          {
2342          SCHECK_PARTIAL();
2343          RRETURN(MATCH_NOMATCH);
2344          }
2345      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2346      if (      if (
2347  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2348         c < 256 &&         c < 256 &&
2349  #endif  #endif
2350         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1552  for (;;) Line 2354  for (;;)
2354      break;      break;
2355    
2356      case OP_WHITESPACE:      case OP_WHITESPACE:
2357      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2358          {
2359          SCHECK_PARTIAL();
2360          RRETURN(MATCH_NOMATCH);
2361          }
2362      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2363      if (      if (
2364  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2365         c >= 256 ||         c > 255 ||
2366  #endif  #endif
2367         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2368         )         )
# Line 1565  for (;;) Line 2371  for (;;)
2371      break;      break;
2372    
2373      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2374      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2375          {
2376          SCHECK_PARTIAL();
2377          RRETURN(MATCH_NOMATCH);
2378          }
2379      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2380      if (      if (
2381  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2382         c < 256 &&         c < 256 &&
2383  #endif  #endif
2384         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1578  for (;;) Line 2388  for (;;)
2388      break;      break;
2389    
2390      case OP_WORDCHAR:      case OP_WORDCHAR:
2391      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2392          {
2393          SCHECK_PARTIAL();
2394          RRETURN(MATCH_NOMATCH);
2395          }
2396      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2397      if (      if (
2398  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2399         c >= 256 ||         c > 255 ||
2400  #endif  #endif
2401         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2402         )         )
# Line 1591  for (;;) Line 2405  for (;;)
2405      break;      break;
2406    
2407      case OP_ANYNL:      case OP_ANYNL:
2408      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2409          {
2410          SCHECK_PARTIAL();
2411          RRETURN(MATCH_NOMATCH);
2412          }
2413      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2414      switch(c)      switch(c)
2415        {        {
2416        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2417    
2418        case 0x000d:        case 0x000d:
2419        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr >= md->end_subject)
2420            {
2421            SCHECK_PARTIAL();
2422            }
2423          else if (*eptr == 0x0a) eptr++;
2424        break;        break;
2425    
2426        case 0x000a:        case 0x000a:
# Line 1615  for (;;) Line 2438  for (;;)
2438      break;      break;
2439    
2440      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2441      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2442          {
2443          SCHECK_PARTIAL();
2444          RRETURN(MATCH_NOMATCH);
2445          }
2446      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2447      switch(c)      switch(c)
2448        {        {
# Line 1645  for (;;) Line 2472  for (;;)
2472      break;      break;
2473    
2474      case OP_HSPACE:      case OP_HSPACE:
2475      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2476          {
2477          SCHECK_PARTIAL();
2478          RRETURN(MATCH_NOMATCH);
2479          }
2480      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2481      switch(c)      switch(c)
2482        {        {
# Line 1675  for (;;) Line 2506  for (;;)
2506      break;      break;
2507    
2508      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2509      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2510          {
2511          SCHECK_PARTIAL();
2512          RRETURN(MATCH_NOMATCH);
2513          }
2514      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2515      switch(c)      switch(c)
2516        {        {
# Line 1693  for (;;) Line 2528  for (;;)
2528      break;      break;
2529    
2530      case OP_VSPACE:      case OP_VSPACE:
2531      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2532          {
2533          SCHECK_PARTIAL();
2534          RRETURN(MATCH_NOMATCH);
2535          }
2536      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2537      switch(c)      switch(c)
2538        {        {
# Line 1716  for (;;) Line 2555  for (;;)
2555    
2556      case OP_PROP:      case OP_PROP:
2557      case OP_NOTPROP:      case OP_NOTPROP:
2558      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2559          {
2560          SCHECK_PARTIAL();
2561          RRETURN(MATCH_NOMATCH);
2562          }
2563      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2564        {        {
2565        const ucd_record *prop = GET_UCD(c);        const ucd_record *prop = GET_UCD(c);
# Line 1732  for (;;) Line 2575  for (;;)
2575               prop->chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2576               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2577            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2578           break;          break;
2579    
2580          case PT_GC:          case PT_GC:
2581          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2582            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2583          break;          break;
2584    
# Line 1749  for (;;) Line 2592  for (;;)
2592            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2593          break;          break;
2594    
2595            /* These are specials */
2596    
2597            case PT_ALNUM:
2598            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2599                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2600              RRETURN(MATCH_NOMATCH);
2601            break;
2602    
2603            case PT_SPACE:    /* Perl space */
2604            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2605                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2606                   == (op == OP_NOTPROP))
2607              RRETURN(MATCH_NOMATCH);
2608            break;
2609    
2610            case PT_PXSPACE:  /* POSIX space */
2611            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2612                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2613                 c == CHAR_FF || c == CHAR_CR)
2614                   == (op == OP_NOTPROP))
2615              RRETURN(MATCH_NOMATCH);
2616            break;
2617    
2618            case PT_WORD:
2619            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2620                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2621                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2622              RRETURN(MATCH_NOMATCH);
2623            break;
2624    
2625            /* This should never occur */
2626    
2627          default:          default:
2628          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2629          }          }
# Line 1761  for (;;) Line 2636  for (;;)
2636      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2637    
2638      case OP_EXTUNI:      case OP_EXTUNI:
2639      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2640          {
2641          SCHECK_PARTIAL();
2642          RRETURN(MATCH_NOMATCH);
2643          }
2644      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2645        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2646        while (eptr < md->end_subject)
2647        {        {
2648        int category = UCD_CATEGORY(c);        int len = 1;
2649        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2650        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2651          {        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;  
         }  
2652        }        }
2653        CHECK_PARTIAL();
2654      ecode++;      ecode++;
2655      break;      break;
2656  #endif  #endif
# Line 1792  for (;;) Line 2665  for (;;)
2665      loops). */      loops). */
2666    
2667      case OP_REF:      case OP_REF:
2668        {      case OP_REFI:
2669        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2670        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2671        ecode += 1 + IMM2_SIZE;
2672    
2673        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2674    
2675        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2676        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
2677        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.  
2678    
2679        (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
2680        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2681    
2682        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
2683        referenced subpattern. */      referenced subpattern. */
2684    
2685        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2686          length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;        length = (md->jscript_compat)? 0 : -1;
2687        else      else
2688          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2689    
2690        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2691    
2692        switch (*ecode)      switch (*ecode)
2693          {        {
2694          case OP_CRSTAR:        case OP_CRSTAR:
2695          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2696          case OP_CRPLUS:        case OP_CRPLUS:
2697          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2698          case OP_CRQUERY:        case OP_CRQUERY:
2699          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2700          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2701          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2702          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2703          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2704          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2705          break;        break;
2706    
2707          case OP_CRRANGE:        case OP_CRRANGE:
2708          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2709          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2710          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2711          max = GET2(ecode, 3);        max = GET2(ecode, 1 + IMM2_SIZE);
2712          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2713          ecode += 5;        ecode += 1 + 2 * IMM2_SIZE;
2714          break;        break;
2715    
2716          default:               /* No repeat follows */        default:               /* No repeat follows */
2717          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2718          eptr += length;          {
2719          continue;              /* With the main loop */          if (length == -2) eptr = md->end_subject;   /* Partial match */
2720            CHECK_PARTIAL();
2721            RRETURN(MATCH_NOMATCH);
2722          }          }
2723          eptr += length;
2724          continue;              /* With the main loop */
2725          }
2726    
2727        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2728        main loop. */      zero, just continue with the main loop. If the length is negative, it
2729        means the reference is unset in non-Java-compatible mode. If the minimum is
2730        zero, we can continue at the same level without recursion. For any other
2731        minimum, carrying on will result in NOMATCH. */
2732    
2733        if (length == 0) continue;      if (length == 0) continue;
2734        if (length < 0 && min == 0) continue;
2735    
2736        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2737        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2738        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2739    
2740        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2741          {
2742          int slength;
2743          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2744          {          {
2745          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2746          eptr += length;          CHECK_PARTIAL();
2747            RRETURN(MATCH_NOMATCH);
2748          }          }
2749          eptr += slength;
2750          }
2751    
2752        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2753        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2754    
2755        if (min == max) continue;      if (min == max) continue;
2756    
2757        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2758    
2759        if (minimize)      if (minimize)
2760          {
2761          for (fi = min;; fi++)
2762          {          {
2763          for (fi = min;; fi++)          int slength;
2764            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2765            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2766            if (fi >= max) RRETURN(MATCH_NOMATCH);
2767            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2768            {            {
2769            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2770            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2771            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2772            }            }
2773          /* Control never gets here */          eptr += slength;
2774          }          }
2775          /* Control never gets here */
2776          }
2777    
2778        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2779    
2780        else      else
2781          {
2782          pp = eptr;
2783          for (i = min; i < max; i++)
2784          {          {
2785          pp = eptr;          int slength;
2786          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2787            {            {
2788            if (!match_ref(offset, eptr, length, md, ims)) break;            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2789            eptr += length;            the soft partial matching case. */
2790            }  
2791          while (eptr >= pp)            if (slength == -2 && md->partial != 0 &&
2792            {                md->end_subject > md->start_used_ptr)
2793            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);              {
2794            if (rrc != MATCH_NOMATCH) RRETURN(rrc);              md->hitend = TRUE;
2795            eptr -= length;              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2796                }
2797              break;
2798            }            }
2799          RRETURN(MATCH_NOMATCH);          eptr += slength;
2800          }          }
2801    
2802          while (eptr >= pp)
2803            {
2804            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2805            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2806            eptr -= length;
2807            }
2808          RRETURN(MATCH_NOMATCH);
2809        }        }
2810      /* Control never gets here */      /* Control never gets here */
2811    
   
   
2812      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2813      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,
2814      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 2823  for (;;)
2823      case OP_NCLASS:      case OP_NCLASS:
2824      case OP_CLASS:      case OP_CLASS:
2825        {        {
2826          /* The data variable is saved across frames, so the byte map needs to
2827          be stored there. */
2828    #define BYTE_MAP ((pcre_uint8 *)data)
2829        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2830        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2831    
2832        switch (*ecode)        switch (*ecode)
2833          {          {
# Line 1940  for (;;) Line 2848  for (;;)
2848          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2849          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2850          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2851          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2852          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2853          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2854          break;          break;
2855    
2856          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1952  for (;;) Line 2860  for (;;)
2860    
2861        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2862    
2863  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2864        /* UTF-8 mode */        if (utf)
       if (utf8)  
2865          {          {
2866          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2867            {            {
2868            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2869                {
2870                SCHECK_PARTIAL();
2871                RRETURN(MATCH_NOMATCH);
2872                }
2873            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2874            if (c > 255)            if (c > 255)
2875              {              {
2876              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2877              }              }
2878            else            else
2879              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2880            }            }
2881          }          }
2882        else        else
2883  #endif  #endif
2884        /* Not UTF-8 mode */        /* Not UTF mode */
2885          {          {
2886          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2887            {            {
2888            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2889                {
2890                SCHECK_PARTIAL();
2891                RRETURN(MATCH_NOMATCH);
2892                }
2893            c = *eptr++;            c = *eptr++;
2894            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2895              if (c > 255)
2896                {
2897                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2898                }
2899              else
2900    #endif
2901                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2902            }            }
2903          }          }
2904    
# Line 1992  for (;;) Line 2912  for (;;)
2912    
2913        if (minimize)        if (minimize)
2914          {          {
2915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2916          /* UTF-8 mode */          if (utf)
         if (utf8)  
2917            {            {
2918            for (fi = min;; fi++)            for (fi = min;; fi++)
2919              {              {
2920              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2921              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2922              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2923                if (eptr >= md->end_subject)
2924                  {
2925                  SCHECK_PARTIAL();
2926                  RRETURN(MATCH_NOMATCH);
2927                  }
2928              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2929              if (c > 255)              if (c > 255)
2930                {                {
2931                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2932                }                }
2933              else              else
2934                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2935              }              }
2936            }            }
2937          else          else
2938  #endif  #endif
2939          /* Not UTF-8 mode */          /* Not UTF mode */
2940            {            {
2941            for (fi = min;; fi++)            for (fi = min;; fi++)
2942              {              {
2943              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2944              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2945              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2946                if (eptr >= md->end_subject)
2947                  {
2948                  SCHECK_PARTIAL();
2949                  RRETURN(MATCH_NOMATCH);
2950                  }
2951              c = *eptr++;              c = *eptr++;
2952              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2953                if (c > 255)
2954                  {
2955                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2956                  }
2957                else
2958    #endif
2959                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2960              }              }
2961            }            }
2962          /* Control never gets here */          /* Control never gets here */
# Line 2034  for (;;) Line 2968  for (;;)
2968          {          {
2969          pp = eptr;          pp = eptr;
2970    
2971  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2972          /* UTF-8 mode */          if (utf)
         if (utf8)  
2973            {            {
2974            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2975              {              {
2976              int len = 1;              int len = 1;
2977              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2978                  {
2979                  SCHECK_PARTIAL();
2980                  break;
2981                  }
2982              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2983              if (c > 255)              if (c > 255)
2984                {                {
2985                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2986                }                }
2987              else              else
2988                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2989              eptr += len;              eptr += len;
2990              }              }
2991            for (;;)            for (;;)
2992              {              {
2993              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2994              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2995              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2996              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2063  for (;;) Line 2998  for (;;)
2998            }            }
2999          else          else
3000  #endif  #endif
3001            /* Not UTF-8 mode */            /* Not UTF mode */
3002            {            {
3003            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3004              {              {
3005              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3006                  {
3007                  SCHECK_PARTIAL();
3008                  break;
3009                  }
3010              c = *eptr;              c = *eptr;
3011              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3012                if (c > 255)
3013                  {
3014                  if (op == OP_CLASS) break;
3015                  }
3016                else
3017    #endif
3018                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3019              eptr++;              eptr++;
3020              }              }
3021            while (eptr >= pp)            while (eptr >= pp)
3022              {              {
3023              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3024              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3025              eptr--;              eptr--;
3026              }              }
# Line 2082  for (;;) Line 3028  for (;;)
3028    
3029          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3030          }          }
3031    #undef BYTE_MAP
3032        }        }
3033      /* Control never gets here */      /* Control never gets here */
3034    
# Line 2090  for (;;) Line 3037  for (;;)
3037      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
3038      mode, because Unicode properties are supported in non-UTF-8 mode. */      mode, because Unicode properties are supported in non-UTF-8 mode. */
3039    
3040  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3041      case OP_XCLASS:      case OP_XCLASS:
3042        {        {
3043        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2115  for (;;) Line 3062  for (;;)
3062          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3063          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3064          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3065          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3066          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3067          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3068          break;          break;
3069    
3070          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2129  for (;;) Line 3076  for (;;)
3076    
3077        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3078          {          {
3079          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3080              {
3081              SCHECK_PARTIAL();
3082              RRETURN(MATCH_NOMATCH);
3083              }
3084          GETCHARINCTEST(c, eptr);          GETCHARINCTEST(c, eptr);
3085          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);          if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3086          }          }
3087    
3088        /* 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 3097  for (;;)
3097          {          {
3098          for (fi = min;; fi++)          for (fi = min;; fi++)
3099            {            {
3100            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3101            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3102            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3103              if (eptr >= md->end_subject)
3104                {
3105                SCHECK_PARTIAL();
3106                RRETURN(MATCH_NOMATCH);
3107                }
3108            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3109            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3110            }            }
3111          /* Control never gets here */          /* Control never gets here */
3112          }          }
# Line 2163  for (;;) Line 3119  for (;;)
3119          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3120            {            {
3121            int len = 1;            int len = 1;
3122            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3123                {
3124                SCHECK_PARTIAL();
3125                break;
3126                }
3127    #ifdef SUPPORT_UTF
3128            GETCHARLENTEST(c, eptr, len);            GETCHARLENTEST(c, eptr, len);
3129            if (!_pcre_xclass(c, data)) break;  #else
3130              c = *eptr;
3131    #endif
3132              if (!PRIV(xclass)(c, data, utf)) break;
3133            eptr += len;            eptr += len;
3134            }            }
3135          for(;;)          for(;;)
3136            {            {
3137            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3138            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3139            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3140            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3141              if (utf) BACKCHAR(eptr);
3142    #endif
3143            }            }
3144          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3145          }          }
# Line 2185  for (;;) Line 3151  for (;;)
3151      /* Match a single character, casefully */      /* Match a single character, casefully */
3152    
3153      case OP_CHAR:      case OP_CHAR:
3154  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3155      if (utf8)      if (utf)
3156        {        {
3157        length = 1;        length = 1;
3158        ecode++;        ecode++;
3159        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3160        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3161            {
3162            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3163            RRETURN(MATCH_NOMATCH);
3164            }
3165        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3166        }        }
3167      else      else
3168  #endif  #endif
3169        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3170        {        {
3171        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3172            {
3173            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3174            RRETURN(MATCH_NOMATCH);
3175            }
3176        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3177        ecode += 2;        ecode += 2;
3178        }        }
3179      break;      break;
3180    
3181      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3182        subject, give up immediately. */
3183    
3184      case OP_CHARNC:      case OP_CHARI:
3185  #ifdef SUPPORT_UTF8      if (eptr >= md->end_subject)
3186      if (utf8)        {
3187          SCHECK_PARTIAL();
3188          RRETURN(MATCH_NOMATCH);
3189          }
3190    
3191    #ifdef SUPPORT_UTF
3192        if (utf)
3193        {        {
3194        length = 1;        length = 1;
3195        ecode++;        ecode++;
3196        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3197    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3198        /* 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
3199        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3200          fast lookup table. We know that there is at least one byte left in the
3201          subject. */
3202    
3203        if (fc < 128)        if (fc < 128)
3204          {          {
3205          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3206                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3207            ecode++;
3208            eptr++;
3209          }          }
3210    
3211        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3212          use the value of "length" to check for sufficient bytes left, because the
3213          other case of the character may have more or fewer bytes.  */
3214    
3215        else        else
3216          {          {
# Line 2246  for (;;) Line 3231  for (;;)
3231          }          }
3232        }        }
3233      else      else
3234  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3235    
3236      /* Non-UTF-8 mode */      /* Not UTF mode */
3237        {        {
3238        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3239        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3240          eptr++;
3241        ecode += 2;        ecode += 2;
3242        }        }
3243      break;      break;
# Line 2259  for (;;) Line 3245  for (;;)
3245      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3246    
3247      case OP_EXACT:      case OP_EXACT:
3248        case OP_EXACTI:
3249      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3250      ecode += 3;      ecode += 1 + IMM2_SIZE;
3251      goto REPEATCHAR;      goto REPEATCHAR;
3252    
3253      case OP_POSUPTO:      case OP_POSUPTO:
3254        case OP_POSUPTOI:
3255      possessive = TRUE;      possessive = TRUE;
3256      /* Fall through */      /* Fall through */
3257    
3258      case OP_UPTO:      case OP_UPTO:
3259        case OP_UPTOI:
3260      case OP_MINUPTO:      case OP_MINUPTO:
3261        case OP_MINUPTOI:
3262      min = 0;      min = 0;
3263      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3264      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3265      ecode += 3;      ecode += 1 + IMM2_SIZE;
3266      goto REPEATCHAR;      goto REPEATCHAR;
3267    
3268      case OP_POSSTAR:      case OP_POSSTAR:
3269        case OP_POSSTARI:
3270      possessive = TRUE;      possessive = TRUE;
3271      min = 0;      min = 0;
3272      max = INT_MAX;      max = INT_MAX;
# Line 2283  for (;;) Line 3274  for (;;)
3274      goto REPEATCHAR;      goto REPEATCHAR;
3275    
3276      case OP_POSPLUS:      case OP_POSPLUS:
3277        case OP_POSPLUSI:
3278      possessive = TRUE;      possessive = TRUE;
3279      min = 1;      min = 1;
3280      max = INT_MAX;      max = INT_MAX;
# Line 2290  for (;;) Line 3282  for (;;)
3282      goto REPEATCHAR;      goto REPEATCHAR;
3283    
3284      case OP_POSQUERY:      case OP_POSQUERY:
3285        case OP_POSQUERYI:
3286      possessive = TRUE;      possessive = TRUE;
3287      min = 0;      min = 0;
3288      max = 1;      max = 1;
# Line 2297  for (;;) Line 3290  for (;;)
3290      goto REPEATCHAR;      goto REPEATCHAR;
3291    
3292      case OP_STAR:      case OP_STAR:
3293        case OP_STARI:
3294      case OP_MINSTAR:      case OP_MINSTAR:
3295        case OP_MINSTARI:
3296      case OP_PLUS:      case OP_PLUS:
3297        case OP_PLUSI:
3298      case OP_MINPLUS:      case OP_MINPLUS:
3299        case OP_MINPLUSI:
3300      case OP_QUERY:      case OP_QUERY:
3301        case OP_QUERYI:
3302      case OP_MINQUERY:      case OP_MINQUERY:
3303      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3304        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3305      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3306      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3307      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3308      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3309    
3310      /* 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. */  
3311    
3312      REPEATCHAR:      REPEATCHAR:
3313  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3314      if (utf8)      if (utf)
3315        {        {
3316        length = 1;        length = 1;
3317        charptr = ecode;        charptr = ecode;
3318        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3319        ecode += length;        ecode += length;
3320    
3321        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2329  for (;;) Line 3325  for (;;)
3325          {          {
3326  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3327          unsigned int othercase;          unsigned int othercase;
3328          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3329              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3330            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3331          else oclength = 0;          else oclength = 0;
3332  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3333    
3334          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3335            {            {
3336            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3337                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3338  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3339            /* Need braces because of following else */            else if (oclength > 0 &&
3340            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3341                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3342    #endif  /* SUPPORT_UCP */
3343            else            else
3344              {              {
3345              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3346              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3347              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3348            }            }
3349    
3350          if (min == max) continue;          if (min == max) continue;
# Line 2357  for (;;) Line 3353  for (;;)
3353            {            {
3354            for (fi = min;; fi++)            for (fi = min;; fi++)
3355              {              {
3356              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3357              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3358              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3359              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3360                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3361  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3362              /* Need braces because of following else */              else if (oclength > 0 &&
3363              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3364                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3365    #endif  /* SUPPORT_UCP */
3366              else              else
3367                {                {
3368                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3369                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3370                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3371              }              }
3372            /* Control never gets here */            /* Control never gets here */
3373            }            }
# Line 2381  for (;;) Line 3377  for (;;)
3377            pp = eptr;            pp = eptr;
3378            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3379              {              {
3380              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3381              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3382  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3383              else if (oclength == 0) break;              else if (oclength > 0 &&
3384                         eptr <= md->end_subject - oclength &&
3385                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3386    #endif  /* SUPPORT_UCP */
3387              else              else
3388                {                {
3389                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3390                eptr += oclength;                break;
3391                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3392              }              }
3393    
3394            if (possessive) continue;            if (possessive) continue;
3395    
3396            for(;;)            for(;;)
3397             {              {
3398             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3399             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3400             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3401  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3402             eptr--;              eptr--;
3403             BACKCHAR(eptr);              BACKCHAR(eptr);
3404  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3405             eptr -= length;              eptr -= length;
3406  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3407             }              }
3408            }            }
3409          /* Control never gets here */          /* Control never gets here */
3410          }          }
# Line 2417  for (;;) Line 3414  for (;;)
3414        value of fc will always be < 128. */        value of fc will always be < 128. */
3415        }        }
3416      else      else
3417  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3418          /* 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);  
3419        fc = *ecode++;        fc = *ecode++;
       }  
3420    
3421      /* 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
3422      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
3423      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3424      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3425      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2437  for (;;) Line 3430  for (;;)
3430      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3431        max, eptr));        max, eptr));
3432    
3433      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3434        {        {
3435        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3436          /* fc must be < 128 if UTF is enabled. */
3437          foc = md->fcc[fc];
3438    #else
3439    #ifdef SUPPORT_UTF
3440    #ifdef SUPPORT_UCP
3441          if (utf && fc > 127)
3442            foc = UCD_OTHERCASE(fc);
3443    #else
3444          if (utf && fc > 127)
3445            foc = fc;
3446    #endif /* SUPPORT_UCP */
3447          else
3448    #endif /* SUPPORT_UTF */
3449            foc = TABLE_GET(fc, md->fcc, fc);
3450    #endif /* COMPILE_PCRE8 */
3451    
3452        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3453          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3454            if (eptr >= md->end_subject)
3455              {
3456              SCHECK_PARTIAL();
3457              RRETURN(MATCH_NOMATCH);
3458              }
3459            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3460            eptr++;
3461            }
3462        if (min == max) continue;        if (min == max) continue;
3463        if (minimize)        if (minimize)
3464          {          {
3465          for (fi = min;; fi++)          for (fi = min;; fi++)
3466            {            {
3467            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3468            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3469            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3470                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3471                {
3472                SCHECK_PARTIAL();
3473              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3474                }
3475              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3476              eptr++;
3477            }            }
3478          /* Control never gets here */          /* Control never gets here */
3479          }          }
# Line 2460  for (;;) Line 3482  for (;;)
3482          pp = eptr;          pp = eptr;
3483          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3484            {            {
3485            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3486                {
3487                SCHECK_PARTIAL();
3488                break;
3489                }
3490              if (fc != *eptr && foc != *eptr) break;
3491            eptr++;            eptr++;
3492            }            }
3493    
3494          if (possessive) continue;          if (possessive) continue;
3495    
3496          while (eptr >= pp)          while (eptr >= pp)
3497            {            {
3498            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3499            eptr--;            eptr--;
3500            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3501            }            }
# Line 2479  for (;;) Line 3508  for (;;)
3508    
3509      else      else
3510        {        {
3511        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3512            {
3513            if (eptr >= md->end_subject)
3514              {
3515              SCHECK_PARTIAL();
3516              RRETURN(MATCH_NOMATCH);
3517              }
3518            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3519            }
3520    
3521        if (min == max) continue;        if (min == max) continue;
3522    
3523        if (minimize)        if (minimize)
3524          {          {
3525          for (fi = min;; fi++)          for (fi = min;; fi++)
3526            {            {
3527            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3528            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3529            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3530              if (eptr >= md->end_subject)
3531                {
3532                SCHECK_PARTIAL();
3533              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3534                }
3535              if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3536            }            }
3537          /* Control never gets here */          /* Control never gets here */
3538          }          }
# Line 2497  for (;;) Line 3541  for (;;)
3541          pp = eptr;          pp = eptr;
3542          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3543            {            {
3544            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3545                {
3546                SCHECK_PARTIAL();
3547                break;
3548                }
3549              if (fc != *eptr) break;
3550            eptr++;            eptr++;
3551            }            }
3552          if (possessive) continue;          if (possessive) continue;
3553    
3554          while (eptr >= pp)          while (eptr >= pp)
3555            {            {
3556            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3557            eptr--;            eptr--;
3558            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3559            }            }
# Line 2516  for (;;) Line 3566  for (;;)
3566      checking can be multibyte. */      checking can be multibyte. */
3567    
3568      case OP_NOT:      case OP_NOT:
3569      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3570      ecode++;      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
     if ((ims & PCRE_CASELESS) != 0)  
3571        {        {
3572  #ifdef SUPPORT_UTF8        SCHECK_PARTIAL();
3573        if (c < 256)        RRETURN(MATCH_NOMATCH);
3574  #endif        }
3575        c = md->lcc[c];  #ifdef SUPPORT_UTF
3576        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);      if (utf)
3577          {
3578          register unsigned int ch, och;
3579    
3580          ecode++;
3581          GETCHARINC(ch, ecode);
3582          GETCHARINC(c, eptr);
3583    
3584          if (op == OP_NOT)
3585            {
3586            if (ch == c) RRETURN(MATCH_NOMATCH);
3587            }
3588          else
3589            {
3590    #ifdef SUPPORT_UCP
3591            if (ch > 127)
3592              och = UCD_OTHERCASE(ch);
3593    #else
3594            if (ch > 127)
3595              och = ch;
3596    #endif /* SUPPORT_UCP */
3597            else
3598              och = TABLE_GET(ch, md->fcc, ch);
3599            if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3600            }
3601        }        }
3602      else      else
3603    #endif
3604        {        {
3605        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        register unsigned int ch = ecode[1];
3606          c = *eptr++;
3607          if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3608            RRETURN(MATCH_NOMATCH);
3609          ecode += 2;
3610        }        }
3611      break;      break;
3612    
# Line 2541  for (;;) Line 3618  for (;;)
3618      about... */      about... */
3619