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

Diff of /code/trunk/pcre_exec.c

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

revision 150 by ph10, Tue Apr 17 08:22:40 2007 UTC revision 915 by zherczeg, Tue Feb 14 13:05:39 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
43  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK md             /* Block containing newline information */  #define NLBLOCK md             /* Block containing newline information */
50  #define PSSTART start_subject  /* Field containing processed string start */  #define PSSTART start_subject  /* Field containing processed string start */
51  #define PSEND   end_subject    /* Field containing processed string end */  #define PSEND   end_subject    /* Field containing processed string end */
# Line 53  possible. There are also some static sup Line 57  possible. There are also some static sup
57  #undef min  #undef min
58  #undef max  #undef max
59    
60  /* The chain of eptrblocks for tail recursions uses memory in stack workspace,  /* Values for setting in md->match_function_type to indicate two special types
61  obtained at top level, the size of which is defined by EPTR_WORK_SIZE. */  of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
 #define EPTR_WORK_SIZE (1000)  
   
 /* Flag bits for the match() function */  
63    
64  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
 #define match_tail_recursed  0x04  /* Tail recursive call */  
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 70  defined PCRE_ERROR_xxx codes, which are Line 70  defined PCRE_ERROR_xxx codes, which are
70  #define MATCH_MATCH        1  #define MATCH_MATCH        1
71  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
72    
73    /* Special internal returns from the match() function. Make them sufficiently
74    negative to avoid the external error codes. */
75    
76    #define MATCH_ACCEPT       (-999)
77    #define MATCH_COMMIT       (-998)
78    #define MATCH_KETRPOS      (-997)
79    #define MATCH_ONCE         (-996)
80    #define MATCH_PRUNE        (-995)
81    #define MATCH_SKIP         (-994)
82    #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
86  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
87  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 83  static const char rep_max[] = { 0, 0, 0, Line 95  static const char rep_max[] = { 0, 0, 0,
95    
96    
97    
98  #ifdef DEBUG  #ifdef PCRE_DEBUG
99  /*************************************************  /*************************************************
100  *        Debugging function to print chars       *  *        Debugging function to print chars       *
101  *************************************************/  *************************************************/
# Line 101  Returns:     nothing Line 113  Returns:     nothing
113  */  */
114    
115  static void  static void
116  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
117  {  {
118  unsigned int c;  unsigned int c;
119  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
# Line 116  while (length-- > 0) Line 128  while (length-- > 0)
128  *          Match a back-reference                *  *          Match a back-reference                *
129  *************************************************/  *************************************************/
130    
131  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
132  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
133    the length passed is zero. Note that in caseless UTF-8 mode, the number of
134    subject bytes matched may be different to the number of reference bytes.
135    
136  Arguments:  Arguments:
137    offset      index into the offset vector    offset      index into the offset vector
138    eptr        points into the subject    eptr        pointer into the subject
139    length      length to be matched    length      length of reference to be matched (number of bytes)
140    md          points to match data block    md          points to match data block
141    ims         the ims flags    caseless    TRUE if caseless
142    
143  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
144  */  */
145    
146  static BOOL  static int
147  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
148    unsigned long int ims)    BOOL caseless)
149  {  {
150  USPTR p = md->start_subject + md->offset_vector[offset];  int matched_length = length;
151    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
152    
153  #ifdef DEBUG  #ifdef PCRE_DEBUG
154  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
155    printf("matching subject <null>");    printf("matching subject <null>");
156  else  else
# Line 148  pchars(p, length, FALSE, md); Line 163  pchars(p, length, FALSE, md);
163  printf("\n");  printf("\n");
164  #endif  #endif
165    
166  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
167    
168  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
169    
170  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
171    properly if Unicode properties are supported. Otherwise, we can check only
172    ASCII characters. */
173    
174  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
175    {    {
176    while (length-- > 0)  #ifdef SUPPORT_UTF
177      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
178      if (md->utf)
179        {
180        /* Match characters up to the end of the reference. NOTE: the number of
181        bytes matched may differ, because there are some characters whose upper and
182        lower case versions code as different numbers of bytes. For example, U+023A
183        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
184        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
185        the latter. It is important, therefore, to check the length along the
186        reference, not along the subject (earlier code did this wrong). */
187    
188        PCRE_PUCHAR endptr = p + length;
189        PCRE_PUCHAR eptr_start = eptr;
190        while (p < endptr)
191          {
192          int c, d;
193          if (eptr >= md->end_subject) return -((int)(eptr - eptr_start) + 1);
194          GETCHARINC(c, eptr);
195          GETCHARINC(d, p);
196          if (c != d && c != UCD_OTHERCASE(d)) return -1;
197          }
198        matched_length = (int)(eptr - eptr_start);
199        }
200      else
201    #endif
202    #endif
203    
204      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
205      is no UCP support. */
206        {
207        if (eptr + length > md->end_subject)
208          {
209          if (md->partial == 0)
210            return -1;
211          length = (int)(md->end_subject - eptr);
212          matched_length = -(length + 1);
213          }
214        while (length-- > 0)
215          {
216          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
217          p++;
218          eptr++;
219          }
220        }
221    }    }
222    
223    /* In the caseful case, we can just compare the bytes, whether or not we
224    are in UTF-8 mode. */
225    
226  else  else
227    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
228      if (eptr + length > md->end_subject)
229        {
230        if (md->partial == 0)
231          return -1;
232        length = (int)(md->end_subject - eptr);
233        matched_length = -(length + 1);
234        }
235      while (length-- > 0) if (*p++ != *eptr++) return -1;
236      }
237    
238  return TRUE;  return matched_length;
239  }  }
240    
241    
# Line 188  calls by keeping local variables that ne Line 261  calls by keeping local variables that ne
261  obtained from malloc() instead instead of on the stack. Macros are used to  obtained from malloc() instead instead of on the stack. Macros are used to
262  achieve this so that the actual code doesn't look very different to what it  achieve this so that the actual code doesn't look very different to what it
263  always used to.  always used to.
264    
265    The original heap-recursive code used longjmp(). However, it seems that this
266    can be very slow on some operating systems. Following a suggestion from Stan
267    Switzer, the use of longjmp() has been abolished, at the cost of having to
268    provide a unique number for each call to RMATCH. There is no way of generating
269    a sequence of numbers at compile time in C. I have given them names, to make
270    them stand out more clearly.
271    
272    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
273    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
274    tests. Furthermore, not using longjmp() means that local dynamic variables
275    don't have indeterminate values; this has meant that the frame size can be
276    reduced because the result can be "passed back" by straight setting of the
277    variable instead of being passed in the frame.
278  ****************************************************************************  ****************************************************************************
279  ***************************************************************************/  ***************************************************************************/
280    
281    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
282    below must be updated in sync.  */
283    
284    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
285           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
286           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
287           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
288           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
289           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
290           RM61,  RM62, RM63, RM64, RM65, RM66 };
291    
292  /* 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
293  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
294    actually used in this definition. */
295    
296  #ifndef NO_RECURSE  #ifndef NO_RECURSE
297  #define REGISTER register  #define REGISTER register
298  #ifdef DEBUG  
299  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
300    #define RMATCH(ra,rb,rc,rd,re,rw) \
301    { \    { \
302    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
303    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
304    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
305    }    }
306  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 210  versions and production versions. */ Line 309  versions and production versions. */
309    return ra; \    return ra; \
310    }    }
311  #else  #else
312  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
313    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
314  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
315  #endif  #endif
316    
317  #else  #else
318    
319    
320  /* These versions of the macros manage a private stack on the heap. Note  /* These versions of the macros manage a private stack on the heap. Note that
321  that the rd argument of RMATCH isn't actually used. It's the md argument of  the "rd" argument of RMATCH isn't actually used in this definition. It's the md
322  match(), which never changes. */  argument of match(), which never changes. */
323    
324  #define REGISTER  #define REGISTER
325    
326  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
327    {\    {\
328    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
329    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
330      {\    frame->Xwhere = rw; \
331      newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
332      newframe->Xecode = rb;\    newframe->Xecode = rb;\
333      newframe->Xoffset_top = rc;\    newframe->Xmstart = mstart;\
334      newframe->Xims = re;\    newframe->Xoffset_top = rc;\
335      newframe->Xeptrb = rf;\    newframe->Xeptrb = re;\
336      newframe->Xflags = rg;\    newframe->Xrdepth = frame->Xrdepth + 1;\
337      newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xprevframe = frame;\
338      newframe->Xprevframe = frame;\    frame = newframe;\
339      frame = newframe;\    DPRINTF(("restarting from line %d\n", __LINE__));\
340      DPRINTF(("restarting from line %d\n", __LINE__));\    goto HEAP_RECURSE;\
341      goto HEAP_RECURSE;\    L_##rw:\
342      }\    DPRINTF(("jumped back to line %d\n", __LINE__));\
   else\  
     {\  
     DPRINTF(("longjumped back to line %d\n", __LINE__));\  
     frame = md->thisframe;\  
     rx = frame->Xresult;\  
     }\  
343    }    }
344    
345  #define RRETURN(ra)\  #define RRETURN(ra)\
346    {\    {\
347    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
348    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
349    (pcre_stack_free)(newframe);\    if (oldframe != &frame_zero) (PUBL(stack_free))(oldframe);\
350    if (frame != NULL)\    if (frame != NULL)\
351      {\      {\
352      frame->Xresult = ra;\      rrc = ra;\
353      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
354      }\      }\
355    return ra;\    return ra;\
356    }    }
# Line 271  typedef struct heapframe { Line 363  typedef struct heapframe {
363    
364    /* Function arguments that may change */    /* Function arguments that may change */
365    
366    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
367    const uschar *Xecode;    const pcre_uchar *Xecode;
368      PCRE_PUCHAR Xmstart;
369    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
370    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
371    unsigned int Xrdepth;    unsigned int Xrdepth;
372    
373    /* Function local variables */    /* Function local variables */
374    
375    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
376    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
377    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
378    const uschar *Xnext;  #endif
379    const uschar *Xpp;    PCRE_PUCHAR Xdata;
380    const uschar *Xprev;    PCRE_PUCHAR Xnext;
381    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
382      PCRE_PUCHAR Xprev;
383      PCRE_PUCHAR Xsaved_eptr;
384    
385    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
386    
# Line 295  typedef struct heapframe { Line 388  typedef struct heapframe {
388    BOOL Xcondition;    BOOL Xcondition;
389    BOOL Xprev_is_word;    BOOL Xprev_is_word;
390    
   unsigned long int Xoriginal_ims;  
   
391  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
392    int Xprop_type;    int Xprop_type;
393    int Xprop_value;    int Xprop_value;
394    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
395    int Xoclength;    int Xoclength;
396    uschar Xocchars[8];    pcre_uchar Xocchars[6];
397  #endif  #endif
398    
399      int Xcodelink;
400    int Xctype;    int Xctype;
401    unsigned int Xfc;    unsigned int Xfc;
402    int Xfi;    int Xfi;
# Line 323  typedef struct heapframe { Line 412  typedef struct heapframe {
412    
413    eptrblock Xnewptrb;    eptrblock Xnewptrb;
414    
415    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
416    
417    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
418    
419  } heapframe;  } heapframe;
420    
# Line 344  typedef struct heapframe { Line 432  typedef struct heapframe {
432    
433  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
434  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
435  same response.  same response. */
436    
437    /* These macros pack up tests that are used for partial matching, and which
438    appear several times in the code. We set the "hit end" flag if the pointer is
439    at the end of the subject and also past the start of the subject (i.e.
440    something has been matched). For hard partial matching, we then return
441    immediately. The second one is used when we already know we are past the end of
442    the subject. */
443    
444    #define CHECK_PARTIAL()\
445      if (md->partial != 0 && eptr >= md->end_subject && \
446          eptr > md->start_used_ptr) \
447        { \
448        md->hitend = TRUE; \
449        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
450        }
451    
452  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
453  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
454        { \
455        md->hitend = TRUE; \
456        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
457        }
458    
459    
460    /* Performance note: It might be tempting to extract commonly used fields from
461    the md structure (e.g. utf, end_subject) into individual variables to improve
462  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
463  made performance worse.  made performance worse.
464    
465  Arguments:  Arguments:
466     eptr        pointer to current character in subject     eptr        pointer to current character in subject
467     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
468       mstart      pointer to the current match start position (can be modified
469                     by encountering \K)
470     offset_top  current top pointer     offset_top  current top pointer
471     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
472     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
473                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
                  match_tail_recursed - this is a tail_recursed group  
474     rdepth      the recursion depth     rdepth      the recursion depth
475    
476  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
477                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
478                   a negative MATCH_xxx value for PRUNE, SKIP, etc
479                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
480                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
481  */  */
482    
483  static int  static int
484  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
485    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
486    int flags, unsigned int rdepth)    unsigned int rdepth)
487  {  {
488  /* 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,
489  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 384  so they can be ordinary variables in all Line 492  so they can be ordinary variables in all
492  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
493  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
494  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
495  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
496    
497  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
498    BOOL caseless;
499    int condcode;
500    
501  /* 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
502  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
503  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
504  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
505    the top-level on the stack rather than malloc-ing them all gives a performance
506    boost in many cases where there is not much "recursion". */
507    
508  #ifdef NO_RECURSE  #ifdef NO_RECURSE
509  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe frame_zero;
510    heapframe *frame = &frame_zero;
511  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
512    
513  /* Copy in the original argument variables */  /* Copy in the original argument variables */
514    
515  frame->Xeptr = eptr;  frame->Xeptr = eptr;
516  frame->Xecode = ecode;  frame->Xecode = ecode;
517    frame->Xmstart = mstart;
518  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
519  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
520  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
521    
522  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 415  HEAP_RECURSE: Line 527  HEAP_RECURSE:
527    
528  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
529  #define ecode              frame->Xecode  #define ecode              frame->Xecode
530    #define mstart             frame->Xmstart
531  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
532  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
533  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
534    
535  /* Ditto for the local variables */  /* Ditto for the local variables */
536    
537  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
538  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
539  #endif  #endif
540  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
541    #define codelink           frame->Xcodelink
542  #define data               frame->Xdata  #define data               frame->Xdata
543  #define next               frame->Xnext  #define next               frame->Xnext
544  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 439  HEAP_RECURSE: Line 551  HEAP_RECURSE:
551  #define condition          frame->Xcondition  #define condition          frame->Xcondition
552  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
553    
 #define original_ims       frame->Xoriginal_ims  
   
554  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
555  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
556  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
557  #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  
558  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
559  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
560  #endif  #endif
# Line 477  i, and fc and c, can be the same variabl Line 584  i, and fc and c, can be the same variabl
584  #define fi i  #define fi i
585  #define fc c  #define fc c
586    
587    /* Many of the following variables are used only in small blocks of the code.
588    My normal style of coding would have declared them within each of those blocks.
589    However, in order to accommodate the version of this code that uses an external
590    "stack" implemented on the heap, it is easier to declare them all here, so the
591    declarations can be cut out in a block. The only declarations within blocks
592    below are for variables that do not have to be preserved over a recursive call
593    to RMATCH(). */
594    
595  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  #ifdef SUPPORT_UTF
596  const uschar *charptr;             /* in small blocks of the code. My normal */  const pcre_uchar *charptr;
597  #endif                             /* style of coding would have declared    */  #endif
598  const uschar *callpat;             /* them within each of those blocks.      */  const pcre_uchar *callpat;
599  const uschar *data;                /* However, in order to accommodate the   */  const pcre_uchar *data;
600  const uschar *next;                /* version of this code that uses an      */  const pcre_uchar *next;
601  USPTR         pp;                  /* external "stack" implemented on the    */  PCRE_PUCHAR       pp;
602  const uschar *prev;                /* heap, it is easier to declare them all */  const pcre_uchar *prev;
603  USPTR         saved_eptr;          /* here, so the declarations can be cut   */  PCRE_PUCHAR       saved_eptr;
604                                     /* out in a block. The only declarations  */  
605  recursion_info new_recursive;      /* within blocks below are for variables  */  recursion_info new_recursive;
606                                     /* that do not have to be preserved over  */  
607  BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  BOOL cur_is_word;
608  BOOL condition;  BOOL condition;
609  BOOL prev_is_word;  BOOL prev_is_word;
610    
 unsigned long int original_ims;  
   
611  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
612  int prop_type;  int prop_type;
613  int prop_value;  int prop_value;
614  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
615  int oclength;  int oclength;
616  uschar occhars[8];  pcre_uchar occhars[6];
617  #endif  #endif
618    
619    int codelink;
620  int ctype;  int ctype;
621  int length;  int length;
622  int max;  int max;
# Line 519  int save_offset1, save_offset2, save_off Line 629  int save_offset1, save_offset2, save_off
629  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
630    
631  eptrblock newptrb;  eptrblock newptrb;
632    
633    /* There is a special fudge for calling match() in a way that causes it to
634    measure the size of its basic stack frame when the stack is being used for
635    recursion. The second argument (ecode) being NULL triggers this behaviour. It
636    cannot normally ever be NULL. The return is the negated value of the frame
637    size. */
638    
639    if (ecode == NULL)
640      {
641      if (rdepth == 0)
642        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
643      else
644        {
645        int len = (char *)&rdepth - (char *)eptr;
646        return (len > 0)? -len : len;
647        }
648      }
649  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
650    
651    /* To save space on the stack and in the heap frame, I have doubled up on some
652    of the local variables that are used only in localised parts of the code, but
653    still need to be preserved over recursive calls of match(). These macros define
654    the alternative names that are used. */
655    
656    #define allow_zero    cur_is_word
657    #define cbegroup      condition
658    #define code_offset   codelink
659    #define condassert    condition
660    #define matched_once  prev_is_word
661    #define foc           number
662    #define save_mark     data
663    
664  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
665  variables. */  variables. */
666    
# Line 540  TAIL_RECURSE: Line 680  TAIL_RECURSE:
680  /* 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
681  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
682  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()
683  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
684  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
685  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,
686  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
687    
688    #ifdef SUPPORT_UTF
689    utf = md->utf;       /* Local copy of the flag */
690    #else
691    utf = FALSE;
692    #endif
693    
694  /* 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
695  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
696    
697  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
698  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
699    
 original_ims = ims;    /* Save for resetting on ')' */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
700  /* 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
701  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
702  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
703  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
704  When match() is called in other circumstances, don't add to the chain. If this  
705  is a tail recursion, use a block from the workspace, as the one on the stack is  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
706  already used. */  such remembered pointers, to be checked when we hit the closing ket, in order
707    to break infinite loops that match no characters. When match() is called in
708    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
709    NOT be used with tail recursion, because the memory block that is used is on
710    the stack, so a new one may be required for each match(). */
711    
712  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
713    {    {
714    eptrblock *p;    newptrb.epb_saved_eptr = eptr;
715    if ((flags & match_tail_recursed) != 0)    newptrb.epb_prev = eptrb;
716      {    eptrb = &newptrb;
717      if (md->eptrn >= EPTR_WORK_SIZE) RRETURN(PCRE_ERROR_NULLWSLIMIT);    md->match_function_type = 0;
     p = md->eptrchain + md->eptrn++;  
     }  
   else p = &newptrb;  
   p->epb_saved_eptr = eptr;  
   p->epb_prev = eptrb;  
   eptrb = p;  
718    }    }
719    
720  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 588  for (;;) Line 724  for (;;)
724    minimize = possessive = FALSE;    minimize = possessive = FALSE;
725    op = *ecode;    op = *ecode;
726    
   /* For partial matching, remember if we ever hit the end of the subject after  
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > md->start_match)  
     md->hitend = TRUE;  
   
727    switch(op)    switch(op)
728      {      {
729      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
730      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
731      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
732      set from a previous iteration of this group, and be referred to by a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
733      reference inside the group.        eptrb, RM55);
734        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
735      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
736      values of the final offsets, in case they were set by a previous iteration  
737      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
738        argument, and we must check whether that argument matches this MARK's
739        argument. It is passed back in md->start_match_ptr (an overloading of that
740        variable). If it does match, we reset that variable to the current subject
741        position and return MATCH_SKIP. Otherwise, pass back the return code
742        unaltered. */
743    
744        else if (rrc == MATCH_SKIP_ARG &&
745            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
746          {
747          md->start_match_ptr = eptr;
748          RRETURN(MATCH_SKIP);
749          }
750        RRETURN(rrc);
751    
752        case OP_FAIL:
753        RRETURN(MATCH_NOMATCH);
754    
755        /* COMMIT overrides PRUNE, SKIP, and THEN */
756    
757        case OP_COMMIT:
758        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
759          eptrb, RM52);
760        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
761            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
762            rrc != MATCH_THEN)
763          RRETURN(rrc);
764        RRETURN(MATCH_COMMIT);
765    
766        /* PRUNE overrides THEN */
767    
768        case OP_PRUNE:
769        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
770          eptrb, RM51);
771        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
772        RRETURN(MATCH_PRUNE);
773    
774        case OP_PRUNE_ARG:
775        md->nomatch_mark = ecode + 2;
776        md->mark = NULL;    /* In case previously set by assertion */
777        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
778          eptrb, RM56);
779        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
780             md->mark == NULL) md->mark = ecode + 2;
781        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
782        RRETURN(MATCH_PRUNE);
783    
784        /* SKIP overrides PRUNE and THEN */
785    
786        case OP_SKIP:
787        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
788          eptrb, RM53);
789        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
790          RRETURN(rrc);
791        md->start_match_ptr = eptr;   /* Pass back current position */
792        RRETURN(MATCH_SKIP);
793    
794        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
795        nomatch_mark. There is a flag that disables this opcode when re-matching a
796        pattern that ended with a SKIP for which there was not a matching MARK. */
797    
798        case OP_SKIP_ARG:
799        if (md->ignore_skip_arg)
800          {
801          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
802          break;
803          }
804        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
805          eptrb, RM57);
806        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
807          RRETURN(rrc);
808    
809        /* Pass back the current skip name by overloading md->start_match_ptr and
810        returning the special MATCH_SKIP_ARG return code. This will either be
811        caught by a matching MARK, or get to the top, where it causes a rematch
812        with the md->ignore_skip_arg flag set. */
813    
814        md->start_match_ptr = ecode + 2;
815        RRETURN(MATCH_SKIP_ARG);
816    
817        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
818        the branch in which it occurs can be determined. Overload the start of
819        match pointer to do this. */
820    
821        case OP_THEN:
822        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
823          eptrb, RM54);
824        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
825        md->start_match_ptr = ecode;
826        RRETURN(MATCH_THEN);
827    
828        case OP_THEN_ARG:
829        md->nomatch_mark = ecode + 2;
830        md->mark = NULL;    /* In case previously set by assertion */
831        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
832          md, eptrb, RM58);
833        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
834             md->mark == NULL) md->mark = ecode + 2;
835        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
836        md->start_match_ptr = ecode;
837        RRETURN(MATCH_THEN);
838    
839        /* Handle an atomic group that does not contain any capturing parentheses.
840        This can be handled like an assertion. Prior to 8.13, all atomic groups
841        were handled this way. In 8.13, the code was changed as below for ONCE, so
842        that backups pass through the group and thereby reset captured values.
843        However, this uses a lot more stack, so in 8.20, atomic groups that do not
844        contain any captures generate OP_ONCE_NC, which can be handled in the old,
845        less stack intensive way.
846    
847        Check the alternative branches in turn - the matching won't pass the KET
848        for this kind of subpattern. If any one branch matches, we carry on as at
849        the end of a normal bracket, leaving the subject pointer, but resetting
850        the start-of-match value in case it was changed by \K. */
851    
852        case OP_ONCE_NC:
853        prev = ecode;
854        saved_eptr = eptr;
855        save_mark = md->mark;
856        do
857          {
858          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
859          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
860            {
861            mstart = md->start_match_ptr;
862            break;
863            }
864          if (rrc == MATCH_THEN)
865            {
866            next = ecode + GET(ecode,1);
867            if (md->start_match_ptr < next &&
868                (*ecode == OP_ALT || *next == OP_ALT))
869              rrc = MATCH_NOMATCH;
870            }
871    
872          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
873          ecode += GET(ecode,1);
874          md->mark = save_mark;
875          }
876        while (*ecode == OP_ALT);
877    
878        /* If hit the end of the group (which could be repeated), fail */
879    
880        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
881    
882        /* Continue as from after the group, updating the offsets high water
883        mark, since extracts may have been taken. */
884    
885        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
886    
887        offset_top = md->end_offset_top;
888        eptr = md->end_match_ptr;
889    
890        /* For a non-repeating ket, just continue at this level. This also
891        happens for a repeating ket if no characters were matched in the group.
892        This is the forcible breaking of infinite loops as implemented in Perl
893        5.005. */
894    
895        if (*ecode == OP_KET || eptr == saved_eptr)
896          {
897          ecode += 1+LINK_SIZE;
898          break;
899          }
900    
901        /* The repeating kets try the rest of the pattern or restart from the
902        preceding bracket, in the appropriate order. The second "call" of match()
903        uses tail recursion, to avoid using another stack frame. */
904    
905        if (*ecode == OP_KETRMIN)
906          {
907          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
908          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
909          ecode = prev;
910          goto TAIL_RECURSE;
911          }
912        else  /* OP_KETRMAX */
913          {
914          md->match_function_type = MATCH_CBEGROUP;
915          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
916          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
917          ecode += 1 + LINK_SIZE;
918          goto TAIL_RECURSE;
919          }
920        /* Control never gets here */
921    
922        /* Handle a capturing bracket, other than those that are possessive with an
923        unlimited repeat. If there is space in the offset vector, save the current
924        subject position in the working slot at the top of the vector. We mustn't
925        change the current values of the data slot, because they may be set from a
926        previous iteration of this group, and be referred to by a reference inside
927        the group. A failure to match might occur after the group has succeeded,
928        if something later on doesn't match. For this reason, we need to restore
929        the working value and also the values of the final offsets, in case they
930        were set by a previous iteration of the same bracket.
931    
932      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
933      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 617  for (;;) Line 938  for (;;)
938      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
939      offset = number << 1;      offset = number << 1;
940    
941  #ifdef DEBUG  #ifdef PCRE_DEBUG
942      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
943      printf("subject=");      printf("subject=");
944      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 630  for (;;) Line 951  for (;;)
951        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
952        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
953        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
954          save_mark = md->mark;
955    
956        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
957        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
958            (int)(eptr - md->start_subject);
959    
960        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
961          {          {
962          RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
963            ims, eptrb, flags);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
964              eptrb, RM1);
965            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
966    
967            /* If we backed up to a THEN, check whether it is within the current
968            branch by comparing the address of the THEN that is passed back with
969            the end of the branch. If it is within the current branch, and the
970            branch is one of two or more alternatives (it either starts or ends
971            with OP_ALT), we have reached the limit of THEN's action, so convert
972            the return code to NOMATCH, which will cause normal backtracking to
973            happen from now on. Otherwise, THEN is passed back to an outer
974            alternative. This implements Perl's treatment of parenthesized groups,
975            where a group not containing | does not affect the current alternative,
976            that is, (X) is NOT the same as (X|(*F)). */
977    
978            if (rrc == MATCH_THEN)
979              {
980              next = ecode + GET(ecode,1);
981              if (md->start_match_ptr < next &&
982                  (*ecode == OP_ALT || *next == OP_ALT))
983                rrc = MATCH_NOMATCH;
984              }
985    
986            /* Anything other than NOMATCH is passed back. */
987    
988          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
989          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
990          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
991            md->mark = save_mark;
992            if (*ecode != OP_ALT) break;
993          }          }
       while (*ecode == OP_ALT);  
994    
995        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
996        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
997        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
998        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
999    
1000        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1001    
1002          RRETURN(rrc);
1003        }        }
1004    
1005      /* Insufficient room for saving captured contents. Treat as a non-capturing      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1006      bracket. */      as a non-capturing bracket. */
1007    
1008        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1009        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1010    
1011      DPRINTF(("insufficient capture room: treat as non-capturing\n"));      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1012    
1013      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1014      final alternative within the brackets, we would return the result of a      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1015      recursive call to match() whatever happened. We can reduce stack usage by  
1016      turning this into a tail recursion. */      /* Non-capturing or atomic group, except for possessive with unlimited
1017        repeat and ONCE group with no captures. Loop for all the alternatives.
1018    
1019        When we get to the final alternative within the brackets, we used to return
1020        the result of a recursive call to match() whatever happened so it was
1021        possible to reduce stack usage by turning this into a tail recursion,
1022        except in the case of a possibly empty group. However, now that there is
1023        the possiblity of (*THEN) occurring in the final alternative, this
1024        optimization is no longer always possible.
1025    
1026        We can optimize if we know there are no (*THEN)s in the pattern; at present
1027        this is the best that can be done.
1028    
1029        MATCH_ONCE is returned when the end of an atomic group is successfully
1030        reached, but subsequent matching fails. It passes back up the tree (causing
1031        captured values to be reset) until the original atomic group level is
1032        reached. This is tested by comparing md->once_target with the start of the
1033        group. At this point, the return is converted into MATCH_NOMATCH so that
1034        previous backup points can be taken. */
1035    
1036        case OP_ONCE:
1037      case OP_BRA:      case OP_BRA:
1038      case OP_SBRA:      case OP_SBRA:
1039      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1040      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1041      for (;;)      for (;;)
1042        {        {
1043        if (ecode[GET(ecode, 1)] != OP_ALT)        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1044    
1045          /* If this is not a possibly empty group, and there are no (*THEN)s in
1046          the pattern, and this is the final alternative, optimize as described
1047          above. */
1048    
1049          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1050          {          {
1051          ecode += _pcre_OP_lengths[*ecode];          ecode += PRIV(OP_lengths)[*ecode];
         flags |= match_tail_recursed;  
         DPRINTF(("bracket 0 tail recursion\n"));  
1052          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1053          }          }
1054    
1055        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* In all other cases, we have to make another call to match(). */
       otherwise return. */  
1056    
1057        RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1058          eptrb, flags);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1059        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RM2);
1060    
1061          /* See comment in the code for capturing groups above about handling
1062          THEN. */
1063    
1064          if (rrc == MATCH_THEN)
1065            {
1066            next = ecode + GET(ecode,1);
1067            if (md->start_match_ptr < next &&
1068                (*ecode == OP_ALT || *next == OP_ALT))
1069              rrc = MATCH_NOMATCH;
1070            }
1071    
1072          if (rrc != MATCH_NOMATCH)
1073            {
1074            if (rrc == MATCH_ONCE)
1075              {
1076              const pcre_uchar *scode = ecode;
1077              if (*scode != OP_ONCE)           /* If not at start, find it */
1078                {
1079                while (*scode == OP_ALT) scode += GET(scode, 1);
1080                scode -= GET(scode, 1);
1081                }
1082              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1083              }
1084            RRETURN(rrc);
1085            }
1086        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1087          md->mark = save_mark;
1088          if (*ecode != OP_ALT) break;
1089        }        }
     /* Control never reaches here. */  
1090    
1091      /* Conditional group: compilation checked that there are no more than      RRETURN(MATCH_NOMATCH);
     two branches. If the condition is false, skipping the first branch takes us  
     past the end if there is only one branch, but that's OK because that is  
     exactly what going to the ket would do. As there is only one branch to be  
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1092    
1093      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1094      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1095      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1096        {      different. The end of these brackets will always be OP_KETRPOS, which
1097        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1098        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1099          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1100    
1101      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      case OP_CBRAPOS:
1102        {      case OP_SCBRAPOS:
1103        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      allow_zero = FALSE;
       condition = offset < offset_top && md->offset_vector[offset] >= 0;  
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1104    
1105      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      POSSESSIVE_CAPTURE:
1106        {      number = GET2(ecode, 1+LINK_SIZE);
1107        condition = FALSE;      offset = number << 1;
       ecode += GET(ecode, 1);  
       }  
1108    
1109      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1110      the final argument match_condassert causes it to stop at the end of an      printf("start possessive bracket %d\n", number);
1111      assertion. */      printf("subject=");
1112        pchars(eptr, 16, TRUE, md);
1113        printf("\n");
1114    #endif
1115    
1116      else      if (offset < md->offset_max)
1117        {        {
1118        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1119            match_condassert);        code_offset = (int)(ecode - md->start_code);
1120        if (rrc == MATCH_MATCH)  
1121          {        save_offset1 = md->offset_vector[offset];
1122          condition = TRUE;        save_offset2 = md->offset_vector[offset+1];
1123          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        save_offset3 = md->offset_vector[md->offset_end - number];
1124          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
1125    
1126          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1127    
1128          /* Each time round the loop, save the current subject position for use
1129          when the group matches. For MATCH_MATCH, the group has matched, so we
1130          restart it with a new subject starting position, remembering that we had
1131          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1132          usual. If we haven't matched any alternatives in any iteration, check to
1133          see if a previous iteration matched. If so, the group has matched;
1134          continue from afterwards. Otherwise it has failed; restore the previous
1135          capture values before returning NOMATCH. */
1136    
1137          for (;;)
1138            {
1139            md->offset_vector[md->offset_end - number] =
1140              (int)(eptr - md->start_subject);
1141            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1142            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1143              eptrb, RM63);
1144            if (rrc == MATCH_KETRPOS)
1145              {
1146              offset_top = md->end_offset_top;
1147              eptr = md->end_match_ptr;
1148              ecode = md->start_code + code_offset;
1149              save_capture_last = md->capture_last;
1150              matched_once = TRUE;
1151              continue;
1152              }
1153    
1154            /* See comment in the code for capturing groups above about handling
1155            THEN. */
1156    
1157            if (rrc == MATCH_THEN)
1158              {
1159              next = ecode + GET(ecode,1);
1160              if (md->start_match_ptr < next &&
1161                  (*ecode == OP_ALT || *next == OP_ALT))
1162                rrc = MATCH_NOMATCH;
1163              }
1164    
1165            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1166            md->capture_last = save_capture_last;
1167            ecode += GET(ecode, 1);
1168            if (*ecode != OP_ALT) break;
1169          }          }
1170        else if (rrc != MATCH_NOMATCH)  
1171          if (!matched_once)
1172          {          {
1173          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1174            md->offset_vector[offset+1] = save_offset2;
1175            md->offset_vector[md->offset_end - number] = save_offset3;
1176          }          }
1177        else  
1178          if (allow_zero || matched_once)
1179          {          {
1180          condition = FALSE;          ecode += 1 + LINK_SIZE;
1181          ecode += GET(ecode, 1);          break;
1182          }          }
       }  
1183    
1184      /* We are now at the branch that is to be obeyed. As there is only one,        RRETURN(MATCH_NOMATCH);
     we can use tail recursion to avoid using another stack frame. If the second  
     alternative doesn't exist, we can just plough on. */  
   
     if (condition || *ecode == OP_ALT)  
       {  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed | ((op == OP_SCOND)? match_cbegroup : 0);  
       goto TAIL_RECURSE;  
       }  
     else  
       {  
       ecode += 1 + LINK_SIZE;  
1185        }        }
     break;  
   
1186    
1187      /* End of the pattern. If we are in a top-level recursion, we should      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1188      restore the offsets appropriately and continue from after the call. */      as a non-capturing bracket. */
1189    
1190      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1191      if (md->recursive != NULL && md->recursive->group_num == 0)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       md->start_match = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1192    
1193      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     string - backtracking will then try other alternatives, if any. */  
1194    
1195      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1196      md->end_match_ptr = eptr;          /* Record where we ended */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     md->end_offset_top = offset_top;   /* and how many extracts were taken */  
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1197    
1198      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1199      matching won't pass the KET for an assertion. If any one branch matches,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1200      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1201      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1202      this level is identical to the lookahead case. */  
1203        case OP_BRAPOS:
1204        case OP_SBRAPOS:
1205        allow_zero = FALSE;
1206    
1207        POSSESSIVE_NON_CAPTURE:
1208        matched_once = FALSE;
1209        code_offset = (int)(ecode - md->start_code);
1210    
1211      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1212        {        {
1213        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1214        if (rrc == MATCH_MATCH) break;        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1215            eptrb, RM48);
1216          if (rrc == MATCH_KETRPOS)
1217            {
1218            offset_top = md->end_offset_top;
1219            eptr = md->end_match_ptr;
1220            ecode = md->start_code + code_offset;
1221            matched_once = TRUE;
1222            continue;
1223            }
1224    
1225          /* See comment in the code for capturing groups above about handling
1226          THEN. */
1227    
1228          if (rrc == MATCH_THEN)
1229            {
1230            next = ecode + GET(ecode,1);
1231            if (md->start_match_ptr < next &&
1232                (*ecode == OP_ALT || *next == OP_ALT))
1233              rrc = MATCH_NOMATCH;
1234            }
1235    
1236        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1237        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1238          if (*ecode != OP_ALT) break;
1239        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1240    
1241      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1242          {
1243          ecode += 1 + LINK_SIZE;
1244          break;
1245          }
1246        RRETURN(MATCH_NOMATCH);
1247    
1248      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1249    
1250      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1251      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1252        past the end if there is only one branch, but that's OK because that is
1253        exactly what going to the ket would do. */
1254    
1255      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1256      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1257      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1258    
1259      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1260        inserted between OP_COND and an assertion condition. */
1261    
1262      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1263        {        {
1264        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (PUBL(callout) != NULL)
1265        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          {
1266        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          PUBL(callout_block) cb;
1267        ecode += GET(ecode,1);          cb.version          = 2;   /* Version 1 of the callout block */
1268            cb.callout_number   = ecode[LINK_SIZE+2];
1269            cb.offset_vector    = md->offset_vector;
1270    #ifdef COMPILE_PCRE8
1271            cb.subject          = (PCRE_SPTR)md->start_subject;
1272    #else
1273            cb.subject          = (PCRE_SPTR16)md->start_subject;
1274    #endif
1275            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1276            cb.start_match      = (int)(mstart - md->start_subject);
1277            cb.current_position = (int)(eptr - md->start_subject);
1278            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1279            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1280            cb.capture_top      = offset_top/2;
1281            cb.capture_last     = md->capture_last;
1282            cb.callout_data     = md->callout_data;
1283            cb.mark             = md->nomatch_mark;
1284            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1285            if (rrc < 0) RRETURN(rrc);
1286            }
1287          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1288          }
1289    
1290        condcode = ecode[LINK_SIZE+1];
1291    
1292        /* Now see what the actual condition is */
1293    
1294        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1295          {
1296          if (md->recursive == NULL)                /* Not recursing => FALSE */
1297            {
1298            condition = FALSE;
1299            ecode += GET(ecode, 1);
1300            }
1301          else
1302            {
1303            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1304            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1305    
1306            /* If the test is for recursion into a specific subpattern, and it is
1307            false, but the test was set up by name, scan the table to see if the
1308            name refers to any other numbers, and test them. The condition is true
1309            if any one is set. */
1310    
1311            if (!condition && condcode == OP_NRREF)
1312              {
1313              pcre_uchar *slotA = md->name_table;
1314              for (i = 0; i < md->name_count; i++)
1315                {
1316                if (GET2(slotA, 0) == recno) break;
1317                slotA += md->name_entry_size;
1318                }
1319    
1320              /* Found a name for the number - there can be only one; duplicate
1321              names for different numbers are allowed, but not vice versa. First
1322              scan down for duplicates. */
1323    
1324              if (i < md->name_count)
1325                {
1326                pcre_uchar *slotB = slotA;
1327                while (slotB > md->name_table)
1328                  {
1329                  slotB -= md->name_entry_size;
1330                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1331                    {
1332                    condition = GET2(slotB, 0) == md->recursive->group_num;
1333                    if (condition) break;
1334                    }
1335                  else break;
1336                  }
1337    
1338                /* Scan up for duplicates */
1339    
1340                if (!condition)
1341                  {
1342                  slotB = slotA;
1343                  for (i++; i < md->name_count; i++)
1344                    {
1345                    slotB += md->name_entry_size;
1346                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1347                      {
1348                      condition = GET2(slotB, 0) == md->recursive->group_num;
1349                      if (condition) break;
1350                      }
1351                    else break;
1352                    }
1353                  }
1354                }
1355              }
1356    
1357            /* Chose branch according to the condition */
1358    
1359            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1360            }
1361          }
1362    
1363        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1364          {
1365          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1366          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1367    
1368          /* If the numbered capture is unset, but the reference was by name,
1369          scan the table to see if the name refers to any other numbers, and test
1370          them. The condition is true if any one is set. This is tediously similar
1371          to the code above, but not close enough to try to amalgamate. */
1372    
1373          if (!condition && condcode == OP_NCREF)
1374            {
1375            int refno = offset >> 1;
1376            pcre_uchar *slotA = md->name_table;
1377    
1378            for (i = 0; i < md->name_count; i++)
1379              {
1380              if (GET2(slotA, 0) == refno) break;
1381              slotA += md->name_entry_size;
1382              }
1383    
1384            /* Found a name for the number - there can be only one; duplicate names
1385            for different numbers are allowed, but not vice versa. First scan down
1386            for duplicates. */
1387    
1388            if (i < md->name_count)
1389              {
1390              pcre_uchar *slotB = slotA;
1391              while (slotB > md->name_table)
1392                {
1393                slotB -= md->name_entry_size;
1394                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1395                  {
1396                  offset = GET2(slotB, 0) << 1;
1397                  condition = offset < offset_top &&
1398                    md->offset_vector[offset] >= 0;
1399                  if (condition) break;
1400                  }
1401                else break;
1402                }
1403    
1404              /* Scan up for duplicates */
1405    
1406              if (!condition)
1407                {
1408                slotB = slotA;
1409                for (i++; i < md->name_count; i++)
1410                  {
1411                  slotB += md->name_entry_size;
1412                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1413                    {
1414                    offset = GET2(slotB, 0) << 1;
1415                    condition = offset < offset_top &&
1416                      md->offset_vector[offset] >= 0;
1417                    if (condition) break;
1418                    }
1419                  else break;
1420                  }
1421                }
1422              }
1423            }
1424    
1425          /* Chose branch according to the condition */
1426    
1427          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1428          }
1429    
1430        else if (condcode == OP_DEF)     /* DEFINE - always false */
1431          {
1432          condition = FALSE;
1433          ecode += GET(ecode, 1);
1434          }
1435    
1436        /* The condition is an assertion. Call match() to evaluate it - setting
1437        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1438        an assertion. */
1439    
1440        else
1441          {
1442          md->match_function_type = MATCH_CONDASSERT;
1443          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1444          if (rrc == MATCH_MATCH)
1445            {
1446            if (md->end_offset_top > offset_top)
1447              offset_top = md->end_offset_top;  /* Captures may have happened */
1448            condition = TRUE;
1449            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1450            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1451            }
1452    
1453          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1454          assertion; it is therefore treated as NOMATCH. */
1455    
1456          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1457            {
1458            RRETURN(rrc);         /* Need braces because of following else */
1459            }
1460          else
1461            {
1462            condition = FALSE;
1463            ecode += codelink;
1464            }
1465          }
1466    
1467        /* We are now at the branch that is to be obeyed. As there is only one, can
1468        use tail recursion to avoid using another stack frame, except when there is
1469        unlimited repeat of a possibly empty group. In the latter case, a recursive
1470        call to match() is always required, unless the second alternative doesn't
1471        exist, in which case we can just plough on. Note that, for compatibility
1472        with Perl, the | in a conditional group is NOT treated as creating two
1473        alternatives. If a THEN is encountered in the branch, it propagates out to
1474        the enclosing alternative (unless nested in a deeper set of alternatives,
1475        of course). */
1476    
1477        if (condition || *ecode == OP_ALT)
1478          {
1479          if (op != OP_SCOND)
1480            {
1481            ecode += 1 + LINK_SIZE;
1482            goto TAIL_RECURSE;
1483            }
1484    
1485          md->match_function_type = MATCH_CBEGROUP;
1486          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1487          RRETURN(rrc);
1488          }
1489    
1490         /* Condition false & no alternative; continue after the group. */
1491    
1492        else
1493          {
1494          ecode += 1 + LINK_SIZE;
1495          }
1496        break;
1497    
1498    
1499        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1500        to close any currently open capturing brackets. */
1501    
1502        case OP_CLOSE:
1503        number = GET2(ecode, 1);
1504        offset = number << 1;
1505    
1506    #ifdef PCRE_DEBUG
1507          printf("end bracket %d at *ACCEPT", number);
1508          printf("\n");
1509    #endif
1510    
1511        md->capture_last = number;
1512        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1513          {
1514          md->offset_vector[offset] =
1515            md->offset_vector[md->offset_end - number];
1516          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1517          if (offset_top <= offset) offset_top = offset + 2;
1518          }
1519        ecode += 1 + IMM2_SIZE;
1520        break;
1521    
1522    
1523        /* End of the pattern, either real or forced. */
1524    
1525        case OP_END:
1526        case OP_ACCEPT:
1527        case OP_ASSERT_ACCEPT:
1528    
1529        /* If we have matched an empty string, fail if not in an assertion and not
1530        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1531        is set and we have matched at the start of the subject. In both cases,
1532        backtracking will then try other alternatives, if any. */
1533    
1534        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1535             md->recursive == NULL &&
1536             (md->notempty ||
1537               (md->notempty_atstart &&
1538                 mstart == md->start_subject + md->start_offset)))
1539          RRETURN(MATCH_NOMATCH);
1540    
1541        /* Otherwise, we have a match. */
1542    
1543        md->end_match_ptr = eptr;           /* Record where we ended */
1544        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1545        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1546    
1547        /* For some reason, the macros don't work properly if an expression is
1548        given as the argument to RRETURN when the heap is in use. */
1549    
1550        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1551        RRETURN(rrc);
1552    
1553        /* Assertion brackets. Check the alternative branches in turn - the
1554        matching won't pass the KET for an assertion. If any one branch matches,
1555        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1556        start of each branch to move the current point backwards, so the code at
1557        this level is identical to the lookahead case. When the assertion is part
1558        of a condition, we want to return immediately afterwards. The caller of
1559        this incarnation of the match() function will have set MATCH_CONDASSERT in
1560        md->match_function type, and one of these opcodes will be the first opcode
1561        that is processed. We use a local variable that is preserved over calls to
1562        match() to remember this case. */
1563    
1564        case OP_ASSERT:
1565        case OP_ASSERTBACK:
1566        save_mark = md->mark;
1567        if (md->match_function_type == MATCH_CONDASSERT)
1568          {
1569          condassert = TRUE;
1570          md->match_function_type = 0;
1571          }
1572        else condassert = FALSE;
1573    
1574        do
1575          {
1576          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1577          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1578            {
1579            mstart = md->start_match_ptr;   /* In case \K reset it */
1580            break;
1581            }
1582    
1583          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1584          as NOMATCH. */
1585    
1586          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1587          ecode += GET(ecode, 1);
1588          md->mark = save_mark;
1589        }        }
1590      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1591    
1592      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1593    
1594        /* If checking an assertion for a condition, return MATCH_MATCH. */
1595    
1596        if (condassert) RRETURN(MATCH_MATCH);
1597    
1598        /* Continue from after the assertion, updating the offsets high water
1599        mark, since extracts may have been taken during the assertion. */
1600    
1601        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1602        ecode += 1 + LINK_SIZE;
1603        offset_top = md->end_offset_top;
1604        continue;
1605    
1606        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1607        PRUNE, or COMMIT means we must assume failure without checking subsequent
1608        branches. */
1609    
1610        case OP_ASSERT_NOT:
1611        case OP_ASSERTBACK_NOT:
1612        save_mark = md->mark;
1613        if (md->match_function_type == MATCH_CONDASSERT)
1614          {
1615          condassert = TRUE;
1616          md->match_function_type = 0;
1617          }
1618        else condassert = FALSE;
1619    
1620        do
1621          {
1622          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1623          md->mark = save_mark;
1624          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1625          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1626            {
1627            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1628            break;
1629            }
1630    
1631          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1632          as NOMATCH. */
1633    
1634          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1635          ecode += GET(ecode,1);
1636          }
1637        while (*ecode == OP_ALT);
1638    
1639        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1640    
1641      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1642      continue;      continue;
# Line 846  for (;;) Line 1647  for (;;)
1647      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1648    
1649      case OP_REVERSE:      case OP_REVERSE:
1650  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1651      if (utf8)      if (utf)
1652        {        {
1653        i = GET(ecode, 1);        i = GET(ecode, 1);
1654        while (i-- > 0)        while (i-- > 0)
1655          {          {
1656          eptr--;          eptr--;
1657          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1658          BACKCHAR(eptr)          BACKCHAR(eptr);
1659          }          }
1660        }        }
1661      else      else
# Line 867  for (;;) Line 1668  for (;;)
1668        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1669        }        }
1670    
1671      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1672    
1673        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1674      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1675      break;      break;
1676    
# Line 877  for (;;) Line 1679  for (;;)
1679      function is able to force a failure. */      function is able to force a failure. */
1680    
1681      case OP_CALLOUT:      case OP_CALLOUT:
1682      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1683        {        {
1684        pcre_callout_block cb;        PUBL(callout_block) cb;
1685        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1686        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1687        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1688    #ifdef COMPILE_PCRE8
1689        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1690        cb.subject_length   = md->end_subject - md->start_subject;  #else
1691        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1692        cb.current_position = eptr - md->start_subject;  #endif
1693          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1694          cb.start_match      = (int)(mstart - md->start_subject);
1695          cb.current_position = (int)(eptr - md->start_subject);
1696        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1697        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1698        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1699        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1700        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1701        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1702          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1703        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1704        }        }
1705      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 902  for (;;) Line 1709  for (;;)
1709      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
1710      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1711    
1712      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1713      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
1714      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
1715      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
1716      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
1717      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
1718      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.  
1719    
1720      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
1721      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
1722      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1723        a lot, so he is not to blame for the current way it works. */
1724    
1725      case OP_RECURSE:      case OP_RECURSE:
1726        {        {
1727          recursion_info *ri;
1728          int recno;
1729    
1730        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1731        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1732          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1733    
1734          /* Check for repeating a recursion without advancing the subject pointer.
1735          This should catch convoluted mutual recursions. (Some simple cases are
1736          caught at compile time.) */
1737    
1738          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1739            if (recno == ri->group_num && eptr == ri->subject_position)
1740              RRETURN(PCRE_ERROR_RECURSELOOP);
1741    
1742        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1743    
1744          new_recursive.group_num = recno;
1745          new_recursive.subject_position = eptr;
1746        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1747        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1748    
1749        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1750    
1751        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1752    
1753        /* Now save the offset data. */        /* Now save the offset data */
1754    
1755        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1756        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 941  for (;;) Line 1758  for (;;)
1758        else        else
1759          {          {
1760          new_recursive.offset_save =          new_recursive.offset_save =
1761            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1762          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1763          }          }
   
1764        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1765              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1766    
1767        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1768        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1769          might be changed, so reset it before looping. */
1770    
1771        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1772        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1773        do        do
1774          {          {
1775          RMATCH(rrc, eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1776            md, ims, eptrb, flags);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1777          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1778            memcpy(md->offset_vector, new_recursive.offset_save,
1779                new_recursive.saved_max * sizeof(int));
1780            md->recursive = new_recursive.prevrec;
1781            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1782            {            {
1783            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1784            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1785              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1786            RRETURN(MATCH_MATCH);  
1787              /* Set where we got to in the subject, and reset the start in case
1788              it was changed by \K. This *is* propagated back out of a recursion,
1789              for Perl compatibility. */
1790    
1791              eptr = md->end_match_ptr;
1792              mstart = md->start_match_ptr;
1793              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1794            }            }
1795          else if (rrc != MATCH_NOMATCH)  
1796            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1797            as NOMATCH. */
1798    
1799            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1800            {            {
1801            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1802              if (new_recursive.offset_save != stacksave)
1803                (PUBL(free))(new_recursive.offset_save);
1804            RRETURN(rrc);            RRETURN(rrc);
1805            }            }
1806    
1807          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1808          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1809          }          }
1810        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 983  for (;;) Line 1812  for (;;)
1812        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1813        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1814        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1815          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1816        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1817        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
         eptrb, 0);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
1818    
1819      /* Continue as from after the assertion, updating the offsets high water      RECURSION_MATCHED:
1820      mark, since extracts may have been taken. */      break;
   
     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(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_cbegroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1821    
1822      /* 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
1823      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1070  for (;;) Line 1826  for (;;)
1826      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1827      break;      break;
1828    
1829      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1830      that it may occur zero times. It may repeat infinitely, or not at all -      indicating that it may occur zero times. It may repeat infinitely, or not
1831      i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1832      repeat limits are compiled as a number of copies, with the optional ones      with fixed upper repeat limits are compiled as a number of copies, with the
1833      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1834    
1835      case OP_BRAZERO:      case OP_BRAZERO:
1836        {      next = ecode + 1;
1837        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1838        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1839        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1840        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1841      break;      break;
1842    
1843      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1844        {      next = ecode + 1;
1845        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1846        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1847        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1848        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1849      break;      break;
1850    
1851        case OP_SKIPZERO:
1852        next = ecode+1;
1853        do next += GET(next,1); while (*next == OP_ALT);
1854        ecode = next + 1 + LINK_SIZE;
1855        break;
1856    
1857        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1858        here; just jump to the group, with allow_zero set TRUE. */
1859    
1860        case OP_BRAPOSZERO:
1861        op = *(++ecode);
1862        allow_zero = TRUE;
1863        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1864          goto POSSESSIVE_NON_CAPTURE;
1865    
1866      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1867    
1868      case OP_KET:      case OP_KET:
1869      case OP_KETRMIN:      case OP_KETRMIN:
1870      case OP_KETRMAX:      case OP_KETRMAX:
1871        case OP_KETRPOS:
1872      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1873    
1874      /* 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
1875      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1876      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1877    
1878      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1879        {        {
1880        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1881        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1882        }        }
1883      else saved_eptr = NULL;      else saved_eptr = NULL;
1884    
1885      /* 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
1886      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
1887      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1888        start in case it was changed by \K. */
1889      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1890          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1891          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1892        {        {
1893        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1894        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1895        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1896          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1897        }        }
1898    
1899      /* 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
1900      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1901      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
1902      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
1903      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
1904        the current subject position and start match pointer and give a MATCH
1905        return. */
1906    
1907      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1908            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1909        {        {
1910        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1911        offset = number << 1;        offset = number << 1;
1912    
1913  #ifdef DEBUG  #ifdef PCRE_DEBUG
1914        printf("end bracket %d", number);        printf("end bracket %d", number);
1915        printf("\n");        printf("\n");
1916  #endif  #endif
1917    
1918          /* Handle a recursively called group. */
1919    
1920          if (md->recursive != NULL && md->recursive->group_num == number)
1921            {
1922            md->end_match_ptr = eptr;
1923            md->start_match_ptr = mstart;
1924            RRETURN(MATCH_MATCH);
1925            }
1926    
1927          /* Deal with capturing */
1928    
1929        md->capture_last = number;        md->capture_last = number;
1930        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1931          {          {
1932            /* If offset is greater than offset_top, it means that we are
1933            "skipping" a capturing group, and that group's offsets must be marked
1934            unset. In earlier versions of PCRE, all the offsets were unset at the
1935            start of matching, but this doesn't work because atomic groups and
1936            assertions can cause a value to be set that should later be unset.
1937            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1938            part of the atomic group, but this is not on the final matching path,
1939            so must be unset when 2 is set. (If there is no group 2, there is no
1940            problem, because offset_top will then be 2, indicating no capture.) */
1941    
1942            if (offset > offset_top)
1943              {
1944              register int *iptr = md->offset_vector + offset_top;
1945              register int *iend = md->offset_vector + offset;
1946              while (iptr < iend) *iptr++ = -1;
1947              }
1948    
1949            /* Now make the extraction */
1950    
1951          md->offset_vector[offset] =          md->offset_vector[offset] =
1952            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1953          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1954          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1955          }          }
1956          }
1957    
1958        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1959        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1960        group. This is the forcible breaking of infinite loops as implemented in
1961        Perl 5.005. For a non-repeating atomic group that includes captures,
1962        establish a backup point by processing the rest of the pattern at a lower
1963        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1964        original OP_ONCE level, thereby bypassing intermediate backup points, but
1965        resetting any captures that happened along the way. */
1966    
1967        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1968          {
1969          if (*prev == OP_ONCE)
1970          {          {
1971          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1972          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1973          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1974          md->start_match = rec->save_start;          RRETURN(MATCH_ONCE);
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
1975          }          }
1976          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1977          break;
1978        }        }
1979    
1980      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1981      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1982        at a time from the outer level, thus saving stack. */
1983    
1984      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)  
1985        {        {
1986        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1987        break;        md->end_offset_top = offset_top;
1988          RRETURN(MATCH_KETRPOS);
1989        }        }
1990    
1991      /* 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
1992      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
1993      tail recursion to avoid using another stack frame. */      use tail recursion to avoid using another stack frame, unless we have an
1994        an atomic group or an unlimited repeat of a group that can match an empty
1995      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      string. */
1996    
1997      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1998        {        {
1999        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2000        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2001          if (*prev == OP_ONCE)
2002            {
2003            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2004            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2005            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2006            RRETURN(MATCH_ONCE);
2007            }
2008          if (*prev >= OP_SBRA)    /* Could match an empty string */
2009            {
2010            md->match_function_type = MATCH_CBEGROUP;
2011            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2012            RRETURN(rrc);
2013            }
2014        ecode = prev;        ecode = prev;
       flags |= match_tail_recursed;  
2015        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2016        }        }
2017      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2018        {        {
2019        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, flags);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
2020          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2021          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2022        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2023          if (*prev == OP_ONCE)
2024            {
2025            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2026            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2027            md->once_target = prev;
2028            RRETURN(MATCH_ONCE);
2029            }
2030        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_tail_recursed;  
2031        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2032        }        }
2033      /* Control never gets here */      /* Control never gets here */
2034    
2035      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2036    
2037      case OP_CIRC:      case OP_CIRC:
2038      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 */  
2039    
2040      /* Start of subject assertion */      /* Start of subject assertion */
2041    
# Line 1232  for (;;) Line 2044  for (;;)
2044      ecode++;      ecode++;
2045      break;      break;
2046    
2047        /* Multiline mode: start of subject unless notbol, or after any newline. */
2048    
2049        case OP_CIRCM:
2050        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2051        if (eptr != md->start_subject &&
2052            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2053          RRETURN(MATCH_NOMATCH);
2054        ecode++;
2055        break;
2056    
2057      /* Start of match assertion */      /* Start of match assertion */
2058    
2059      case OP_SOM:      case OP_SOM:
# Line 1239  for (;;) Line 2061  for (;;)
2061      ecode++;      ecode++;
2062      break;      break;
2063    
2064      /* Assert before internal newline if multiline, or before a terminating      /* Reset the start of match point */
     newline unless endonly is set, else end of subject unless noteol is set. */  
2065    
2066      case OP_DOLL:      case OP_SET_SOM:
2067      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2068        {      ecode++;
2069        if (eptr < md->end_subject)      break;
2070          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
2071        else      /* Multiline mode: assert before any newline, or before end of subject
2072          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
2073        ecode++;  
2074        break;      case OP_DOLLM:
2075        }      if (eptr < md->end_subject)
2076          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
2077      else      else
2078        {        {
2079        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2080        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2081        }        }
2082        ecode++;
2083        break;
2084    
2085        /* Not multiline mode: assert before a terminating newline or before end of
2086        subject unless noteol is set. */
2087    
2088        case OP_DOLL:
2089        if (md->noteol) RRETURN(MATCH_NOMATCH);
2090        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2091    
2092      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2093    
2094      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2095    
2096      case OP_EOD:      case OP_EOD:
2097      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2098        SCHECK_PARTIAL();
2099      ecode++;      ecode++;
2100      break;      break;
2101    
2102      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2103    
2104      case OP_EODN:      case OP_EODN:
2105      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2106        if (eptr < md->end_subject &&
2107          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2108        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2109    
2110        /* Either at end of string or \n before end. */
2111    
2112        SCHECK_PARTIAL();
2113      ecode++;      ecode++;
2114      break;      break;
2115    
# Line 1290  for (;;) Line 2121  for (;;)
2121    
2122        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2123        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
2124        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2125          partial matching. */
2126    
2127  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2128        if (utf8)        if (utf)
2129          {          {
2130            /* Get status of previous character */
2131    
2132          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2133            {            {
2134            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2135            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2136              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2137            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2138    #ifdef SUPPORT_UCP
2139              if (md->use_ucp)
2140                {
2141                if (c == '_') prev_is_word = TRUE; else
2142                  {
2143                  int cat = UCD_CATEGORY(c);
2144                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2145                  }
2146                }
2147              else
2148    #endif
2149            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2150            }            }
2151          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2152            /* Get status of next character */
2153    
2154            if (eptr >= md->end_subject)
2155              {
2156              SCHECK_PARTIAL();
2157              cur_is_word = FALSE;
2158              }
2159            else
2160            {            {
2161            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2162    #ifdef SUPPORT_UCP
2163              if (md->use_ucp)
2164                {
2165                if (c == '_') cur_is_word = TRUE; else
2166                  {
2167                  int cat = UCD_CATEGORY(c);
2168                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2169                  }
2170                }
2171              else
2172    #endif
2173            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2174            }            }
2175          }          }
2176        else        else
2177  #endif  #endif
2178    
2179        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2180          consistency with the behaviour of \w we do use it in this case. */
2181    
2182          {          {
2183          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2184            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2185          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2186            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2187              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2188    #ifdef SUPPORT_UCP
2189              if (md->use_ucp)
2190                {
2191                c = eptr[-1];
2192                if (c == '_') prev_is_word = TRUE; else
2193                  {
2194                  int cat = UCD_CATEGORY(c);
2195                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2196                  }
2197                }
2198              else
2199    #endif
2200              prev_is_word = MAX_255(eptr[-1])
2201                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2202              }
2203    
2204            /* Get status of next character */
2205    
2206            if (eptr >= md->end_subject)
2207              {
2208              SCHECK_PARTIAL();
2209              cur_is_word = FALSE;
2210              }
2211            else
2212    #ifdef SUPPORT_UCP
2213            if (md->use_ucp)
2214              {
2215              c = *eptr;
2216              if (c == '_') cur_is_word = TRUE; else
2217                {
2218                int cat = UCD_CATEGORY(c);
2219                cur_is_word = (cat == ucp_L || cat == ucp_N);
2220                }
2221              }
2222            else
2223    #endif
2224            cur_is_word = MAX_255(*eptr)
2225              && ((md->ctypes[*eptr] & ctype_word) != 0);
2226          }          }
2227    
2228        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1331  for (;;) Line 2236  for (;;)
2236      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2237    
2238      case OP_ANY:      case OP_ANY:
2239      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2240        {      /* Fall through */
2241        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);  
2242        case OP_ALLANY:
2243        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2244          {                            /* not be updated before SCHECK_PARTIAL. */
2245          SCHECK_PARTIAL();
2246          RRETURN(MATCH_NOMATCH);
2247        }        }
2248      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2249      if (utf8)  #ifdef SUPPORT_UTF
2250        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2251    #endif
2252      ecode++;      ecode++;
2253      break;      break;
2254    
# Line 1345  for (;;) Line 2256  for (;;)
2256      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2257    
2258      case OP_ANYBYTE:      case OP_ANYBYTE:
2259      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2260          {                            /* not be updated before SCHECK_PARTIAL. */
2261          SCHECK_PARTIAL();
2262          RRETURN(MATCH_NOMATCH);
2263          }
2264        eptr++;
2265      ecode++;      ecode++;
2266      break;      break;
2267    
2268      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2269      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2270          {
2271          SCHECK_PARTIAL();
2272          RRETURN(MATCH_NOMATCH);
2273          }
2274      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2275      if (      if (
2276  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2277         c < 256 &&         c < 256 &&
2278  #endif  #endif
2279         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1363  for (;;) Line 2283  for (;;)
2283      break;      break;
2284    
2285      case OP_DIGIT:      case OP_DIGIT:
2286      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2287          {
2288          SCHECK_PARTIAL();
2289          RRETURN(MATCH_NOMATCH);
2290          }
2291      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2292      if (      if (
2293  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2294         c >= 256 ||         c > 255 ||
2295  #endif  #endif
2296         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2297         )         )
# Line 1376  for (;;) Line 2300  for (;;)
2300      break;      break;
2301    
2302      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2303      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2304          {
2305          SCHECK_PARTIAL();
2306          RRETURN(MATCH_NOMATCH);
2307          }
2308      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2309      if (      if (
2310  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2311         c < 256 &&         c < 256 &&
2312  #endif  #endif
2313         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1389  for (;;) Line 2317  for (;;)
2317      break;      break;
2318    
2319      case OP_WHITESPACE:      case OP_WHITESPACE:
2320      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2321          {
2322          SCHECK_PARTIAL();
2323          RRETURN(MATCH_NOMATCH);
2324          }
2325      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2326      if (      if (
2327  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2328         c >= 256 ||         c > 255 ||
2329  #endif  #endif
2330         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2331         )         )
# Line 1402  for (;;) Line 2334  for (;;)
2334      break;      break;
2335    
2336      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2337      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2338          {
2339          SCHECK_PARTIAL();
2340          RRETURN(MATCH_NOMATCH);
2341          }
2342      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2343      if (      if (
2344  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2345         c < 256 &&         c < 256 &&
2346  #endif  #endif
2347         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1415  for (;;) Line 2351  for (;;)
2351      break;      break;
2352    
2353      case OP_WORDCHAR:      case OP_WORDCHAR:
2354      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2355          {
2356          SCHECK_PARTIAL();
2357          RRETURN(MATCH_NOMATCH);
2358          }
2359      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2360      if (      if (
2361  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2362         c >= 256 ||         c > 255 ||
2363  #endif  #endif
2364         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2365         )         )
# Line 1428  for (;;) Line 2368  for (;;)
2368      break;      break;
2369    
2370      case OP_ANYNL:      case OP_ANYNL:
2371      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2372          {
2373          SCHECK_PARTIAL();
2374          RRETURN(MATCH_NOMATCH);
2375          }
2376      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2377      switch(c)      switch(c)
2378        {        {
2379        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2380    
2381        case 0x000d:        case 0x000d:
2382        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2383        break;        break;
2384    
2385        case 0x000a:        case 0x000a:
2386          break;
2387    
2388        case 0x000b:        case 0x000b:
2389        case 0x000c:        case 0x000c:
2390        case 0x0085:        case 0x0085:
2391        case 0x2028:        case 0x2028:
2392        case 0x2029:        case 0x2029:
2393          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2394          break;
2395          }
2396        ecode++;
2397        break;
2398    
2399        case OP_NOT_HSPACE:
2400        if (eptr >= md->end_subject)
2401          {
2402          SCHECK_PARTIAL();
2403          RRETURN(MATCH_NOMATCH);
2404          }
2405        GETCHARINCTEST(c, eptr);
2406        switch(c)
2407          {
2408          default: break;
2409          case 0x09:      /* HT */
2410          case 0x20:      /* SPACE */
2411          case 0xa0:      /* NBSP */
2412          case 0x1680:    /* OGHAM SPACE MARK */
2413          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2414          case 0x2000:    /* EN QUAD */
2415          case 0x2001:    /* EM QUAD */
2416          case 0x2002:    /* EN SPACE */
2417          case 0x2003:    /* EM SPACE */
2418          case 0x2004:    /* THREE-PER-EM SPACE */
2419          case 0x2005:    /* FOUR-PER-EM SPACE */
2420          case 0x2006:    /* SIX-PER-EM SPACE */
2421          case 0x2007:    /* FIGURE SPACE */
2422          case 0x2008:    /* PUNCTUATION SPACE */
2423          case 0x2009:    /* THIN SPACE */
2424          case 0x200A:    /* HAIR SPACE */
2425          case 0x202f:    /* NARROW NO-BREAK SPACE */
2426          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2427          case 0x3000:    /* IDEOGRAPHIC SPACE */
2428          RRETURN(MATCH_NOMATCH);
2429          }
2430        ecode++;
2431        break;
2432    
2433        case OP_HSPACE:
2434        if (eptr >= md->end_subject)
2435          {
2436          SCHECK_PARTIAL();
2437          RRETURN(MATCH_NOMATCH);
2438          }
2439        GETCHARINCTEST(c, eptr);
2440        switch(c)
2441          {
2442          default: RRETURN(MATCH_NOMATCH);
2443          case 0x09:      /* HT */
2444          case 0x20:      /* SPACE */
2445          case 0xa0:      /* NBSP */
2446          case 0x1680:    /* OGHAM SPACE MARK */
2447          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2448          case 0x2000:    /* EN QUAD */
2449          case 0x2001:    /* EM QUAD */
2450          case 0x2002:    /* EN SPACE */
2451          case 0x2003:    /* EM SPACE */
2452          case 0x2004:    /* THREE-PER-EM SPACE */
2453          case 0x2005:    /* FOUR-PER-EM SPACE */
2454          case 0x2006:    /* SIX-PER-EM SPACE */
2455          case 0x2007:    /* FIGURE SPACE */
2456          case 0x2008:    /* PUNCTUATION SPACE */
2457          case 0x2009:    /* THIN SPACE */
2458          case 0x200A:    /* HAIR SPACE */
2459          case 0x202f:    /* NARROW NO-BREAK SPACE */
2460          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2461          case 0x3000:    /* IDEOGRAPHIC SPACE */
2462          break;
2463          }
2464        ecode++;
2465        break;
2466    
2467        case OP_NOT_VSPACE:
2468        if (eptr >= md->end_subject)
2469          {
2470          SCHECK_PARTIAL();
2471          RRETURN(MATCH_NOMATCH);
2472          }
2473        GETCHARINCTEST(c, eptr);
2474        switch(c)
2475          {
2476          default: break;
2477          case 0x0a:      /* LF */
2478          case 0x0b:      /* VT */
2479          case 0x0c:      /* FF */
2480          case 0x0d:      /* CR */
2481          case 0x85:      /* NEL */
2482          case 0x2028:    /* LINE SEPARATOR */
2483          case 0x2029:    /* PARAGRAPH SEPARATOR */
2484          RRETURN(MATCH_NOMATCH);
2485          }
2486        ecode++;
2487        break;
2488    
2489        case OP_VSPACE:
2490        if (eptr >= md->end_subject)
2491          {
2492          SCHECK_PARTIAL();
2493          RRETURN(MATCH_NOMATCH);
2494          }
2495        GETCHARINCTEST(c, eptr);
2496        switch(c)
2497          {
2498          default: RRETURN(MATCH_NOMATCH);
2499          case 0x0a:      /* LF */
2500          case 0x0b:      /* VT */
2501          case 0x0c:      /* FF */
2502          case 0x0d:      /* CR */
2503          case 0x85:      /* NEL */
2504          case 0x2028:    /* LINE SEPARATOR */
2505          case 0x2029:    /* PARAGRAPH SEPARATOR */
2506        break;        break;
2507        }        }
2508      ecode++;      ecode++;
# Line 1453  for (;;) Line 2514  for (;;)
2514    
2515      case OP_PROP:      case OP_PROP:
2516      case OP_NOTPROP:      case OP_NOTPROP:
2517      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2518          {
2519          SCHECK_PARTIAL();
2520          RRETURN(MATCH_NOMATCH);
2521          }
2522      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2523        {        {
2524        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2525    
2526        switch(ecode[1])        switch(ecode[1])
2527          {          {
# Line 1466  for (;;) Line 2530  for (;;)
2530          break;          break;
2531    
2532          case PT_LAMP:          case PT_LAMP:
2533          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2534               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2535               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2536            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2537           break;          break;
2538    
2539          case PT_GC:          case PT_GC:
2540          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2541            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2542          break;          break;
2543    
2544          case PT_PC:          case PT_PC:
2545          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2546            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2547          break;          break;
2548    
2549          case PT_SC:          case PT_SC:
2550          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2551              RRETURN(MATCH_NOMATCH);
2552            break;
2553    
2554            /* These are specials */
2555    
2556            case PT_ALNUM:
2557            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2558                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2559              RRETURN(MATCH_NOMATCH);
2560            break;
2561    
2562            case PT_SPACE:    /* Perl space */
2563            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2564                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2565                   == (op == OP_NOTPROP))
2566              RRETURN(MATCH_NOMATCH);
2567            break;
2568    
2569            case PT_PXSPACE:  /* POSIX space */
2570            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2571                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2572                 c == CHAR_FF || c == CHAR_CR)
2573                   == (op == OP_NOTPROP))
2574              RRETURN(MATCH_NOMATCH);
2575            break;
2576    
2577            case PT_WORD:
2578            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2579                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2580                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2581            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2582          break;          break;
2583    
2584            /* This should never occur */
2585    
2586          default:          default:
2587          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2588          }          }
# Line 1499  for (;;) Line 2595  for (;;)
2595      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2596    
2597      case OP_EXTUNI:      case OP_EXTUNI:
2598      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2599          {
2600          SCHECK_PARTIAL();
2601          RRETURN(MATCH_NOMATCH);
2602          }
2603      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2604        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2605        while (eptr < md->end_subject)
2606        {        {
2607        int chartype, script;        int len = 1;
2608        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2609        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2610        while (eptr < md->end_subject)        eptr += len;
2611          {        }
2612          int len = 1;      if (md->partial != 0 && eptr >= md->end_subject)
2613          if (!utf8) c = *eptr; else        {
2614            {        SCHECK_PARTIAL();
           GETCHARLEN(c, eptr, len);  
           }  
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2615        }        }
2616      ecode++;      ecode++;
2617      break;      break;
# Line 1531  for (;;) Line 2627  for (;;)
2627      loops). */      loops). */
2628    
2629      case OP_REF:      case OP_REF:
2630        {      case OP_REFI:
2631        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2632        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2633        ecode += 1 + IMM2_SIZE;
2634    
2635        /* If the reference is unset, set the length to be longer than the amount      /* If the reference is unset, there are two possibilities:
       of subject left; this ensures that every attempt at a match fails. We  
       can't just fail here, because of the possibility of quantifiers with zero  
       minima. */  
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2636    
2637        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2638        this ensures that every attempt at a match fails. We can't just fail
2639        here, because of the possibility of quantifiers with zero minima.
2640    
2641        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2642          {      so that the back reference matches an empty string.
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
2643    
2644          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2645          case OP_CRMINRANGE:      referenced subpattern. */
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2646    
2647          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2648          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2649          eptr += length;      else
2650          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2651    
2652        /* Set up for repetition, or handle the non-repeated case */
2653    
2654        switch (*ecode)
2655          {
2656          case OP_CRSTAR:
2657          case OP_CRMINSTAR:
2658          case OP_CRPLUS:
2659          case OP_CRMINPLUS:
2660          case OP_CRQUERY:
2661          case OP_CRMINQUERY:
2662          c = *ecode++ - OP_CRSTAR;
2663          minimize = (c & 1) != 0;
2664          min = rep_min[c];                 /* Pick up values from tables; */
2665          max = rep_max[c];                 /* zero for max => infinity */
2666          if (max == 0) max = INT_MAX;
2667          break;
2668    
2669          case OP_CRRANGE:
2670          case OP_CRMINRANGE:
2671          minimize = (*ecode == OP_CRMINRANGE);
2672          min = GET2(ecode, 1);
2673          max = GET2(ecode, 1 + IMM2_SIZE);
2674          if (max == 0) max = INT_MAX;
2675          ecode += 1 + 2 * IMM2_SIZE;
2676          break;
2677    
2678          default:               /* No repeat follows */
2679          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2680            {
2681            eptr += -(length + 1);
2682            CHECK_PARTIAL();
2683            RRETURN(MATCH_NOMATCH);
2684          }          }
2685          eptr += length;
2686          continue;              /* With the main loop */
2687          }
2688    
2689        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2690        main loop. */      zero, just continue with the main loop. If the length is negative, it
2691        means the reference is unset in non-Java-compatible mode. If the minimum is
2692        zero, we can continue at the same level without recursion. For any other
2693        minimum, carrying on will result in NOMATCH. */
2694    
2695        if (length == 0) continue;      if (length == 0) continue;
2696        if (length < 0 && min == 0) continue;
2697    
2698        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2699        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2700        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2701    
2702        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2703          {
2704          int slength;
2705          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2706          {          {
2707          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          eptr += -(slength + 1);
2708          eptr += length;          CHECK_PARTIAL();
2709            RRETURN(MATCH_NOMATCH);
2710          }          }
2711          eptr += slength;
2712          }
2713    
2714        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2715        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2716    
2717        if (min == max) continue;      if (min == max) continue;
2718    
2719        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2720    
2721        if (minimize)      if (minimize)
2722          {
2723          for (fi = min;; fi++)
2724          {          {
2725          for (fi = min;; fi++)          int slength;
2726            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2727            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2728            if (fi >= max) RRETURN(MATCH_NOMATCH);
2729            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2730            {            {
2731            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            eptr += -(slength + 1);
2732            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2733            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2734            }            }
2735          /* Control never gets here */          eptr += slength;
2736          }          }
2737          /* Control never gets here */
2738          }
2739    
2740        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2741    
2742        else      else
2743          {
2744          pp = eptr;
2745          for (i = min; i < max; i++)
2746          {          {
2747          pp = eptr;          int slength;
2748          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
2749            {            {
2750            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Restore the eptr after the check. */
2751            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            eptr += -(slength + 1);
2752            eptr -= length;            CHECK_PARTIAL();
2753              eptr -= -(slength + 1);
2754              break;
2755            }            }
2756          RRETURN(MATCH_NOMATCH);          eptr += slength;
2757          }          }
2758          while (eptr >= pp)
2759            {
2760            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2761            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2762            eptr -= length;
2763            }
2764          RRETURN(MATCH_NOMATCH);
2765        }        }
2766      /* Control never gets here */      /* Control never gets here */
2767    
   
   
2768      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2769      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,
2770      and either the matching is caseful, or the characters are in the range      and either the matching is caseful, or the characters are in the range
# Line 1648  for (;;) Line 2779  for (;;)
2779      case OP_NCLASS:      case OP_NCLASS:
2780      case OP_CLASS:      case OP_CLASS:
2781        {        {
2782          /* The data variable is saved across frames, so the byte map needs to
2783          be stored there. */
2784    #define BYTE_MAP ((pcre_uint8 *)data)
2785        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2786        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2787    
2788        switch (*ecode)        switch (*ecode)
2789          {          {
# Line 1670  for (;;) Line 2804  for (;;)
2804          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2805          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2806          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2807          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2808          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2809          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2810          break;          break;
2811    
2812          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1682  for (;;) Line 2816  for (;;)
2816    
2817        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2818    
2819  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2820        /* UTF-8 mode */        if (utf)
       if (utf8)  
2821          {          {
2822          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2823            {            {
2824            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2825                {
2826                SCHECK_PARTIAL();
2827                RRETURN(MATCH_NOMATCH);
2828                }
2829            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2830            if (c > 255)            if (c > 255)
2831              {              {
2832              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2833              }              }
2834            else            else
2835              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2836            }            }
2837          }          }
2838        else        else
2839  #endif  #endif
2840        /* Not UTF-8 mode */        /* Not UTF mode */
2841          {          {
2842          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2843            {            {
2844            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2845                {
2846                SCHECK_PARTIAL();
2847                RRETURN(MATCH_NOMATCH);
2848                }
2849            c = *eptr++;            c = *eptr++;
2850            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2851              if (c > 255)
2852                {
2853                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2854                }
2855              else
2856    #endif
2857                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2858            }            }
2859          }          }
2860    
# Line 1722  for (;;) Line 2868  for (;;)
2868    
2869        if (minimize)        if (minimize)
2870          {          {
2871  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2872          /* UTF-8 mode */          if (utf)
         if (utf8)  
2873            {            {
2874            for (fi = min;; fi++)            for (fi = min;; fi++)
2875              {              {
2876              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2877              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2878              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2879                if (eptr >= md->end_subject)
2880                  {
2881                  SCHECK_PARTIAL();
2882                  RRETURN(MATCH_NOMATCH);
2883                  }
2884              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2885              if (c > 255)              if (c > 255)
2886                {                {
2887                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2888                }                }
2889              else              else
2890                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2891              }              }
2892            }            }
2893          else          else
2894  #endif  #endif
2895          /* Not UTF-8 mode */          /* Not UTF mode */
2896            {            {
2897            for (fi = min;; fi++)            for (fi = min;; fi++)
2898              {              {
2899              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2900              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2901              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2902                if (eptr >= md->end_subject)
2903                  {
2904                  SCHECK_PARTIAL();
2905                  RRETURN(MATCH_NOMATCH);
2906                  }
2907              c = *eptr++;              c = *eptr++;
2908              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2909                if (c > 255)
2910                  {
2911                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2912                  }
2913                else
2914    #endif
2915                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2916              }              }
2917            }            }
2918          /* Control never gets here */          /* Control never gets here */
# Line 1764  for (;;) Line 2924  for (;;)
2924          {          {
2925          pp = eptr;          pp = eptr;
2926    
2927  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2928          /* UTF-8 mode */          if (utf)
         if (utf8)  
2929            {            {
2930            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2931              {              {
2932              int len = 1;              int len = 1;
2933              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2934                  {
2935                  SCHECK_PARTIAL();
2936                  break;
2937                  }
2938              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2939              if (c > 255)              if (c > 255)
2940                {                {
2941                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2942                }                }
2943              else              else
2944                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2945              eptr += len;              eptr += len;
2946              }              }
2947            for (;;)            for (;;)
2948              {              {
2949              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2950              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2951              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2952              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1793  for (;;) Line 2954  for (;;)
2954            }            }
2955          else          else
2956  #endif  #endif
2957            /* Not UTF-8 mode */            /* Not UTF mode */
2958            {            {
2959            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2960              {              {
2961              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2962                  {
2963                  SCHECK_PARTIAL();
2964                  break;
2965                  }
2966              c = *eptr;              c = *eptr;
2967              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2968                if (c > 255)
2969                  {
2970                  if (op == OP_CLASS) break;
2971                  }
2972                else
2973    #endif
2974                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2975              eptr++;              eptr++;
2976              }              }
2977            while (eptr >= pp)            while (eptr >= pp)
2978              {              {
2979              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2980              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2981              eptr--;              eptr--;
2982              }              }
# Line 1812  for (;;) Line 2984  for (;;)
2984    
2985          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2986          }          }
2987    #undef BYTE_MAP
2988        }        }
2989      /* Control never gets here */      /* Control never gets here */
2990    
2991    
2992      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2993      in UTF-8 mode, because that's the only time it is compiled. */      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
2994        mode, because Unicode properties are supported in non-UTF-8 mode. */
2995    
2996  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2997      case OP_XCLASS:      case OP_XCLASS:
2998        {        {
2999        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1844  for (;;) Line 3018  for (;;)
3018          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3019          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3020          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3021          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3022          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3023          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3024          break;          break;
3025    
3026          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1858  for (;;) Line 3032  for (;;)
3032    
3033        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3034          {          {
3035          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3036          GETCHARINC(c, eptr);            {
3037          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3038              RRETURN(MATCH_NOMATCH);
3039              }
3040            GETCHARINCTEST(c, eptr);
3041            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3042          }          }
3043    
3044        /* If max == min we can continue with the main loop without the        /* If max == min we can continue with the main loop without the
# Line 1875  for (;;) Line 3053  for (;;)
3053          {          {
3054          for (fi = min;; fi++)          for (fi = min;; fi++)
3055            {            {
3056            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3057            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3058            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3059            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3060            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3061                SCHECK_PARTIAL();
3062                RRETURN(MATCH_NOMATCH);
3063                }
3064              GETCHARINCTEST(c, eptr);
3065              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3066            }            }
3067          /* Control never gets here */          /* Control never gets here */
3068          }          }
# Line 1892  for (;;) Line 3075  for (;;)
3075          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3076            {            {
3077            int len = 1;            int len = 1;
3078            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3079            GETCHARLEN(c, eptr, len);              {
3080            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3081                break;
3082                }
3083    #ifdef SUPPORT_UTF
3084              GETCHARLENTEST(c, eptr, len);
3085    #else
3086              c = *eptr;
3087    #endif
3088              if (!PRIV(xclass)(c, data, utf)) break;
3089            eptr += len;            eptr += len;
3090            }            }
3091          for(;;)          for(;;)
3092            {            {
3093            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3094            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3095            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3096            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3097              if (utf) BACKCHAR(eptr);
3098    #endif
3099            }            }
3100          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3101          }          }
# Line 1914  for (;;) Line 3107  for (;;)
3107      /* Match a single character, casefully */      /* Match a single character, casefully */
3108    
3109      case OP_CHAR:      case OP_CHAR:
3110  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3111      if (utf8)      if (utf)
3112        {        {
3113        length = 1;        length = 1;
3114        ecode++;        ecode++;
3115        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3116        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3117            {
3118            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3119            RRETURN(MATCH_NOMATCH);
3120            }
3121        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3122        }        }
3123      else      else
3124  #endif  #endif
3125        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3126        {        {
3127        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3128            {
3129            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3130            RRETURN(MATCH_NOMATCH);
3131            }
3132        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3133        ecode += 2;        ecode += 2;
3134        }        }
3135      break;      break;
3136    
3137      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3138        subject, give up immediately. */
3139    
3140        case OP_CHARI:
3141        if (eptr >= md->end_subject)
3142          {
3143          SCHECK_PARTIAL();
3144          RRETURN(MATCH_NOMATCH);
3145          }
3146    
3147      case OP_CHARNC:  #ifdef SUPPORT_UTF
3148  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3149        {        {
3150        length = 1;        length = 1;
3151        ecode++;        ecode++;
3152        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3153    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3154        /* 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
3155        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3156          fast lookup table. We know that there is at least one byte left in the
3157          subject. */
3158    
3159        if (fc < 128)        if (fc < 128)
3160          {          {
3161          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3162                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3163            ecode++;
3164            eptr++;
3165          }          }
3166    
3167        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3168          use the value of "length" to check for sufficient bytes left, because the
3169          other case of the character may have more or fewer bytes.  */
3170    
3171        else        else
3172          {          {
# Line 1968  for (;;) Line 3180  for (;;)
3180          if (fc != dc)          if (fc != dc)
3181            {            {
3182  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3183            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3184  #endif  #endif
3185              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3186            }            }
3187          }          }
3188        }        }
3189      else      else
3190  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3191    
3192      /* Non-UTF-8 mode */      /* Not UTF mode */
3193        {        {
3194        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3195        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3196          eptr++;
3197        ecode += 2;        ecode += 2;
3198        }        }
3199      break;      break;
# Line 1988  for (;;) Line 3201  for (;;)
3201      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3202    
3203      case OP_EXACT:      case OP_EXACT:
3204        case OP_EXACTI:
3205      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3206      ecode += 3;      ecode += 1 + IMM2_SIZE;
3207      goto REPEATCHAR;      goto REPEATCHAR;
3208    
3209      case OP_POSUPTO:      case OP_POSUPTO:
3210        case OP_POSUPTOI:
3211      possessive = TRUE;      possessive = TRUE;
3212      /* Fall through */      /* Fall through */
3213    
3214      case OP_UPTO:      case OP_UPTO:
3215        case OP_UPTOI:
3216      case OP_MINUPTO:      case OP_MINUPTO:
3217        case OP_MINUPTOI:
3218      min = 0;      min = 0;
3219      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3220      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3221      ecode += 3;      ecode += 1 + IMM2_SIZE;
3222      goto REPEATCHAR;      goto REPEATCHAR;
3223    
3224      case OP_POSSTAR:      case OP_POSSTAR:
3225        case OP_POSSTARI:
3226      possessive = TRUE;      possessive = TRUE;
3227      min = 0;      min = 0;
3228      max = INT_MAX;      max = INT_MAX;
# Line 2012  for (;;) Line 3230  for (;;)
3230      goto REPEATCHAR;      goto REPEATCHAR;
3231    
3232      case OP_POSPLUS:      case OP_POSPLUS:
3233        case OP_POSPLUSI:
3234      possessive = TRUE;      possessive = TRUE;
3235      min = 1;      min = 1;
3236      max = INT_MAX;      max = INT_MAX;
# Line 2019  for (;;) Line 3238  for (;;)
3238      goto REPEATCHAR;      goto REPEATCHAR;
3239    
3240      case OP_POSQUERY:      case OP_POSQUERY:
3241        case OP_POSQUERYI:
3242      possessive = TRUE;      possessive = TRUE;
3243      min = 0;      min = 0;
3244      max = 1;      max = 1;
# Line 2026  for (;;) Line 3246  for (;;)
3246      goto REPEATCHAR;      goto REPEATCHAR;
3247    
3248      case OP_STAR:      case OP_STAR:
3249        case OP_STARI:
3250      case OP_MINSTAR:      case OP_MINSTAR:
3251        case OP_MINSTARI:
3252      case OP_PLUS:      case OP_PLUS:
3253        case OP_PLUSI:
3254      case OP_MINPLUS:      case OP_MINPLUS:
3255        case OP_MINPLUSI:
3256      case OP_QUERY:      case OP_QUERY:
3257        case OP_QUERYI:
3258      case OP_MINQUERY:      case OP_MINQUERY:
3259      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3260        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3261      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3262      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3263      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3264      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3265    
3266      /* 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. */  
3267    
3268      REPEATCHAR:      REPEATCHAR:
3269  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3270      if (utf8)      if (utf)
3271        {        {
3272        length = 1;        length = 1;
3273        charptr = ecode;        charptr = ecode;
3274        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3275        ecode += length;        ecode += length;
3276    
3277        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2058  for (;;) Line 3281  for (;;)
3281          {          {
3282  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3283          unsigned int othercase;          unsigned int othercase;
3284          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3285              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3286            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3287          else oclength = 0;          else oclength = 0;
3288  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3289    
3290          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3291            {            {
3292            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3293                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3294  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3295            /* Need braces because of following else */            else if (oclength > 0 &&
3296            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3297                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3298    #endif  /* SUPPORT_UCP */
3299            else            else
3300              {              {
3301              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3302              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3303              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3304            }            }
3305    
3306          if (min == max) continue;          if (min == max) continue;
# Line 2086  for (;;) Line 3309  for (;;)
3309            {            {
3310            for (fi = min;; fi++)            for (fi = min;; fi++)
3311              {              {
3312              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3313              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3314              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3315              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3316                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3317  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3318              /* Need braces because of following else */              else if (oclength > 0 &&
3319              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3320                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3321    #endif  /* SUPPORT_UCP */
3322              else              else
3323                {                {
3324                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3325                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3326                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3327              }              }
3328            /* Control never gets here */            /* Control never gets here */
3329            }            }
# Line 2110  for (;;) Line 3333  for (;;)
3333            pp = eptr;            pp = eptr;
3334            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3335              {              {
3336              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3337              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3338  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3339              else if (oclength == 0) break;              else if (oclength > 0 &&
3340                         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) break;                CHECK_PARTIAL();
3346                eptr += oclength;                break;
3347                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3348              }              }
3349    
3350            if (possessive) continue;            if (possessive) continue;
3351    
3352            for(;;)            for(;;)
3353             {              {
3354             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3355             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3356             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3357  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3358             eptr--;              eptr--;
3359             BACKCHAR(eptr);              BACKCHAR(eptr);
3360  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3361             eptr -= length;              eptr -= length;
3362  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3363             }              }
3364            }            }
3365          /* Control never gets here */          /* Control never gets here */
3366          }          }
# Line 2146  for (;;) Line 3370  for (;;)
3370        value of fc will always be < 128. */        value of fc will always be < 128. */
3371        }        }
3372      else      else
3373  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3374          /* 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);  
3375        fc = *ecode++;        fc = *ecode++;
       }  
3376    
3377      /* 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
3378      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
3379      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3380      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3381      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2166  for (;;) Line 3386  for (;;)
3386      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3387        max, eptr));        max, eptr));
3388    
3389      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3390        {        {
3391        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3392          /* fc must be < 128 if UTF is enabled. */
3393          foc = md->fcc[fc];