/[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 137 by ph10, Thu Mar 29 13:56:00 2007 UTC revision 1041 by ph10, Sun Sep 16 10:16:27 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 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
43    
44    #ifdef HAVE_CONFIG_H
45    #include "config.h"
46    #endif
47    
48  #define NLBLOCK md             /* Block containing newline information */  #define NLBLOCK md             /* Block containing newline information */
49  #define PSSTART start_subject  /* Field containing processed string start */  #define PSSTART start_subject  /* Field containing processed string start */
50  #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 56  possible. There are also some static sup
56  #undef min  #undef min
57  #undef max  #undef max
58    
59  /* 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
60  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,
61    as stack usage is to be discouraged. */
 #define EPTR_WORK_SIZE (1000)  
62    
63  /* Flag bits for the match() function */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
64    #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
 #define match_condassert     0x01  /* Called to check a condition assertion */  
 #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  
 #define match_tail_recursed  0x04  /* Tail recursive call */  
65    
66  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
67  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 69  defined PCRE_ERROR_xxx codes, which are
69  #define MATCH_MATCH        1  #define MATCH_MATCH        1
70  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
71    
72    /* Special internal returns from the match() function. Make them sufficiently
73    negative to avoid the external error codes. */
74    
75    #define MATCH_ACCEPT       (-999)
76    #define MATCH_COMMIT       (-998)
77    #define MATCH_KETRPOS      (-997)
78    #define MATCH_ONCE         (-996)
79    #define MATCH_PRUNE        (-995)
80    #define MATCH_SKIP         (-994)
81    #define MATCH_SKIP_ARG     (-993)
82    #define MATCH_THEN         (-992)
83    
84  /* 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.
85  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,
86  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 94  static const char rep_max[] = { 0, 0, 0,
94    
95    
96    
97  #ifdef DEBUG  #ifdef PCRE_DEBUG
98  /*************************************************  /*************************************************
99  *        Debugging function to print chars       *  *        Debugging function to print chars       *
100  *************************************************/  *************************************************/
# Line 101  Returns:     nothing Line 112  Returns:     nothing
112  */  */
113    
114  static void  static void
115  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)
116  {  {
117  unsigned int c;  unsigned int c;
118  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 127  while (length-- > 0)
127  *          Match a back-reference                *  *          Match a back-reference                *
128  *************************************************/  *************************************************/
129    
130  /* 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
131  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
132    the length passed is zero. Note that in caseless UTF-8 mode, the number of
133    subject bytes matched may be different to the number of reference bytes.
134    
135  Arguments:  Arguments:
136    offset      index into the offset vector    offset      index into the offset vector
137    eptr        points into the subject    eptr        pointer into the subject
138    length      length to be matched    length      length of reference to be matched (number of bytes)
139    md          points to match data block    md          points to match data block
140    ims         the ims flags    caseless    TRUE if caseless
141    
142  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
143                  -1 no match
144                  -2 partial match; always given if at end subject
145  */  */
146    
147  static BOOL  static int
148  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,
149    unsigned long int ims)    BOOL caseless)
150  {  {
151  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
152    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
153    
154  #ifdef DEBUG  #ifdef PCRE_DEBUG
155  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
156    printf("matching subject <null>");    printf("matching subject <null>");
157  else  else
# Line 148  pchars(p, length, FALSE, md); Line 164  pchars(p, length, FALSE, md);
164  printf("\n");  printf("\n");
165  #endif  #endif
166    
167  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
168    case the length is passed as zero). */
169    
170  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
171    
172  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
173    properly if Unicode properties are supported. Otherwise, we can check only
174    ASCII characters. */
175    
176  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
177    {    {
178    while (length-- > 0)  #ifdef SUPPORT_UTF
179      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
180      if (md->utf)
181        {
182        /* Match characters up to the end of the reference. NOTE: the number of
183        bytes matched may differ, because there are some characters whose upper and
184        lower case versions code as different numbers of bytes. For example, U+023A
185        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
186        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
187        the latter. It is important, therefore, to check the length along the
188        reference, not along the subject (earlier code did this wrong). */
189    
190        PCRE_PUCHAR endptr = p + length;
191        while (p < endptr)
192          {
193          int c, d;
194          if (eptr >= md->end_subject) return -2;   /* Partial match */
195          GETCHARINC(c, eptr);
196          GETCHARINC(d, p);
197          if (c != d && c != UCD_OTHERCASE(d)) return -1;
198          }
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        while (length-- > 0)
208          {
209          if (eptr >= md->end_subject) return -2;   /* Partial match */
210          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
211          p++;
212          eptr++;
213          }
214        }
215    }    }
216    
217    /* In the caseful case, we can just compare the bytes, whether or not we
218    are in UTF-8 mode. */
219    
220  else  else
221    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
222      while (length-- > 0)
223        {
224        if (eptr >= md->end_subject) return -2;   /* Partial match */
225        if (*p++ != *eptr++) return -1;
226        }
227      }
228    
229  return TRUE;  return (int)(eptr - eptr_start);
230  }  }
231    
232    
# Line 188  calls by keeping local variables that ne Line 252  calls by keeping local variables that ne
252  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
253  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
254  always used to.  always used to.
255    
256    The original heap-recursive code used longjmp(). However, it seems that this
257    can be very slow on some operating systems. Following a suggestion from Stan
258    Switzer, the use of longjmp() has been abolished, at the cost of having to
259    provide a unique number for each call to RMATCH. There is no way of generating
260    a sequence of numbers at compile time in C. I have given them names, to make
261    them stand out more clearly.
262    
263    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
264    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
265    tests. Furthermore, not using longjmp() means that local dynamic variables
266    don't have indeterminate values; this has meant that the frame size can be
267    reduced because the result can be "passed back" by straight setting of the
268    variable instead of being passed in the frame.
269  ****************************************************************************  ****************************************************************************
270  ***************************************************************************/  ***************************************************************************/
271    
272    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
273    below must be updated in sync.  */
274    
275    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
276           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
277           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
278           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
279           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
280           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
281           RM61,  RM62, RM63, RM64, RM65, RM66 };
282    
283  /* 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
284  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
285    actually used in this definition. */
286    
287  #ifndef NO_RECURSE  #ifndef NO_RECURSE
288  #define REGISTER register  #define REGISTER register
289  #ifdef DEBUG  
290  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
291    #define RMATCH(ra,rb,rc,rd,re,rw) \
292    { \    { \
293    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
294    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
295    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
296    }    }
297  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 210  versions and production versions. */ Line 300  versions and production versions. */
300    return ra; \    return ra; \
301    }    }
302  #else  #else
303  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
304    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
305  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
306  #endif  #endif
307    
308  #else  #else
309    
310    
311  /* 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
312  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
313  match(), which never changes. */  argument of match(), which never changes. */
314    
315  #define REGISTER  #define REGISTER
316    
317  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
318    {\    {\
319    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
320    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL)\
     {\  
     newframe->Xeptr = ra;\  
     newframe->Xecode = rb;\  
     newframe->Xoffset_top = rc;\  
     newframe->Xims = re;\  
     newframe->Xeptrb = rf;\  
     newframe->Xflags = rg;\  
     newframe->Xrdepth = frame->Xrdepth + 1;\  
     newframe->Xprevframe = frame;\  
     frame = newframe;\  
     DPRINTF(("restarting from line %d\n", __LINE__));\  
     goto HEAP_RECURSE;\  
     }\  
   else\  
321      {\      {\
322      DPRINTF(("longjumped back to line %d\n", __LINE__));\      newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
323      frame = md->thisframe;\      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
324      rx = frame->Xresult;\      newframe->Xnextframe = NULL;\
325        frame->Xnextframe = newframe;\
326      }\      }\
327      frame->Xwhere = rw;\
328      newframe->Xeptr = ra;\
329      newframe->Xecode = rb;\
330      newframe->Xmstart = mstart;\
331      newframe->Xoffset_top = rc;\
332      newframe->Xeptrb = re;\
333      newframe->Xrdepth = frame->Xrdepth + 1;\
334      newframe->Xprevframe = frame;\
335      frame = newframe;\
336      DPRINTF(("restarting from line %d\n", __LINE__));\
337      goto HEAP_RECURSE;\
338      L_##rw:\
339      DPRINTF(("jumped back to line %d\n", __LINE__));\
340    }    }
341    
342  #define RRETURN(ra)\  #define RRETURN(ra)\
343    {\    {\
344    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
345    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
346    if (frame != NULL)\    if (frame != NULL)\
347      {\      {\
348      frame->Xresult = ra;\      rrc = ra;\
349      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
350      }\      }\
351    return ra;\    return ra;\
352    }    }
# Line 268  match(), which never changes. */ Line 356  match(), which never changes. */
356    
357  typedef struct heapframe {  typedef struct heapframe {
358    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
359      struct heapframe *Xnextframe;
360    
361    /* Function arguments that may change */    /* Function arguments that may change */
362    
363    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
364    const uschar *Xecode;    const pcre_uchar *Xecode;
365      PCRE_PUCHAR Xmstart;
366    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
367    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
368    unsigned int Xrdepth;    unsigned int Xrdepth;
369    
370    /* Function local variables */    /* Function local variables */
371    
372    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
373    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
374    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
375    const uschar *Xnext;  #endif
376    const uschar *Xpp;    PCRE_PUCHAR Xdata;
377    const uschar *Xprev;    PCRE_PUCHAR Xnext;
378    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
379      PCRE_PUCHAR Xprev;
380      PCRE_PUCHAR Xsaved_eptr;
381    
382    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
383    
# Line 295  typedef struct heapframe { Line 385  typedef struct heapframe {
385    BOOL Xcondition;    BOOL Xcondition;
386    BOOL Xprev_is_word;    BOOL Xprev_is_word;
387    
   unsigned long int Xoriginal_ims;  
   
388  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
389    int Xprop_type;    int Xprop_type;
390    int Xprop_value;    int Xprop_value;
391    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
392    int Xoclength;    int Xoclength;
393    uschar Xocchars[8];    pcre_uchar Xocchars[6];
394  #endif  #endif
395    
396      int Xcodelink;
397    int Xctype;    int Xctype;
398    unsigned int Xfc;    unsigned int Xfc;
399    int Xfi;    int Xfi;
# Line 323  typedef struct heapframe { Line 409  typedef struct heapframe {
409    
410    eptrblock Xnewptrb;    eptrblock Xnewptrb;
411    
412    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
413    
414    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
415    
416  } heapframe;  } heapframe;
417    
# Line 344  typedef struct heapframe { Line 429  typedef struct heapframe {
429    
430  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
431  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
432  same response.  same response. */
433    
434    /* These macros pack up tests that are used for partial matching, and which
435    appear several times in the code. We set the "hit end" flag if the pointer is
436    at the end of the subject and also past the start of the subject (i.e.
437    something has been matched). For hard partial matching, we then return
438    immediately. The second one is used when we already know we are past the end of
439    the subject. */
440    
441    #define CHECK_PARTIAL()\
442      if (md->partial != 0 && eptr >= md->end_subject && \
443          eptr > md->start_used_ptr) \
444        { \
445        md->hitend = TRUE; \
446        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
447        }
448    
449    #define SCHECK_PARTIAL()\
450      if (md->partial != 0 && eptr > md->start_used_ptr) \
451        { \
452        md->hitend = TRUE; \
453        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
454        }
455    
456    
457  Performance note: It might be tempting to extract commonly used fields from the  /* Performance note: It might be tempting to extract commonly used fields from
458  md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf, end_subject) into individual variables to improve
459  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
460  made performance worse.  made performance worse.
461    
462  Arguments:  Arguments:
463     eptr        pointer to current character in subject     eptr        pointer to current character in subject
464     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
465       mstart      pointer to the current match start position (can be modified
466                     by encountering \K)
467     offset_top  current top pointer     offset_top  current top pointer
468     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
469     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
470                   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  
471     rdepth      the recursion depth     rdepth      the recursion depth
472    
473  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
474                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
475                   a negative MATCH_xxx value for PRUNE, SKIP, etc
476                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
477                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
478  */  */
479    
480  static int  static int
481  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
482    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
483    int flags, unsigned int rdepth)    unsigned int rdepth)
484  {  {
485  /* 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,
486  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 489  so they can be ordinary variables in all
489  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
490  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
491  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
492  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
493    
494  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
495    BOOL caseless;
496    int condcode;
497    
498  /* 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
499  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
500  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
501  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
502    the top-level on the stack rather than malloc-ing them all gives a performance
503    boost in many cases where there is not much "recursion". */
504    
505  #ifdef NO_RECURSE  #ifdef NO_RECURSE
506  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
507    
508  /* Copy in the original argument variables */  /* Copy in the original argument variables */
509    
510  frame->Xeptr = eptr;  frame->Xeptr = eptr;
511  frame->Xecode = ecode;  frame->Xecode = ecode;
512    frame->Xmstart = mstart;
513  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
514  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
515  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
516    
517  /* 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 522  HEAP_RECURSE:
522    
523  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
524  #define ecode              frame->Xecode  #define ecode              frame->Xecode
525    #define mstart             frame->Xmstart
526  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
527  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
528  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
529    
530  /* Ditto for the local variables */  /* Ditto for the local variables */
531    
532  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
533  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
534  #endif  #endif
535  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
536    #define codelink           frame->Xcodelink
537  #define data               frame->Xdata  #define data               frame->Xdata
538  #define next               frame->Xnext  #define next               frame->Xnext
539  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 439  HEAP_RECURSE: Line 546  HEAP_RECURSE:
546  #define condition          frame->Xcondition  #define condition          frame->Xcondition
547  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
548    
 #define original_ims       frame->Xoriginal_ims  
   
549  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
550  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
551  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
552  #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  
553  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
554  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
555  #endif  #endif
# Line 477  i, and fc and c, can be the same variabl Line 579  i, and fc and c, can be the same variabl
579  #define fi i  #define fi i
580  #define fc c  #define fc c
581    
582    /* Many of the following variables are used only in small blocks of the code.
583    My normal style of coding would have declared them within each of those blocks.
584    However, in order to accommodate the version of this code that uses an external
585    "stack" implemented on the heap, it is easier to declare them all here, so the
586    declarations can be cut out in a block. The only declarations within blocks
587    below are for variables that do not have to be preserved over a recursive call
588    to RMATCH(). */
589    
590    #ifdef SUPPORT_UTF
591    const pcre_uchar *charptr;
592    #endif
593    const pcre_uchar *callpat;
594    const pcre_uchar *data;
595    const pcre_uchar *next;
596    PCRE_PUCHAR       pp;
597    const pcre_uchar *prev;
598    PCRE_PUCHAR       saved_eptr;
599    
600    recursion_info new_recursive;
601    
602  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  BOOL cur_is_word;
 const uschar *charptr;             /* in small blocks of the code. My normal */  
 #endif                             /* style of coding would have declared    */  
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
603  BOOL condition;  BOOL condition;
604  BOOL prev_is_word;  BOOL prev_is_word;
605    
 unsigned long int original_ims;  
   
606  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
607  int prop_type;  int prop_type;
608  int prop_value;  int prop_value;
609  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
610  int oclength;  int oclength;
611  uschar occhars[8];  pcre_uchar occhars[6];
612  #endif  #endif
613    
614    int codelink;
615  int ctype;  int ctype;
616  int length;  int length;
617  int max;  int max;
# Line 519  int save_offset1, save_offset2, save_off Line 624  int save_offset1, save_offset2, save_off
624  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
625    
626  eptrblock newptrb;  eptrblock newptrb;
627    
628    /* There is a special fudge for calling match() in a way that causes it to
629    measure the size of its basic stack frame when the stack is being used for
630    recursion. The second argument (ecode) being NULL triggers this behaviour. It
631    cannot normally ever be NULL. The return is the negated value of the frame
632    size. */
633    
634    if (ecode == NULL)
635      {
636      if (rdepth == 0)
637        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
638      else
639        {
640        int len = (char *)&rdepth - (char *)eptr;
641        return (len > 0)? -len : len;
642        }
643      }
644  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
645    
646    /* To save space on the stack and in the heap frame, I have doubled up on some
647    of the local variables that are used only in localised parts of the code, but
648    still need to be preserved over recursive calls of match(). These macros define
649    the alternative names that are used. */
650    
651    #define allow_zero    cur_is_word
652    #define cbegroup      condition
653    #define code_offset   codelink
654    #define condassert    condition
655    #define matched_once  prev_is_word
656    #define foc           number
657    #define save_mark     data
658    
659  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
660  variables. */  variables. */
661    
# Line 540  TAIL_RECURSE: Line 675  TAIL_RECURSE:
675  /* 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
676  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
677  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()
678  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
679  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
680  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,
681  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
682    
683    #ifdef SUPPORT_UTF
684    utf = md->utf;       /* Local copy of the flag */
685    #else
686    utf = FALSE;
687    #endif
688    
689  /* 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
690  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
691    
692  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
693  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
694    
 original_ims = ims;    /* Save for resetting on ')' */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
695  /* 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
696  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
697  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
698  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
699  When match() is called in other circumstances, don't add to the chain. If this  
700  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
701  already used. */  such remembered pointers, to be checked when we hit the closing ket, in order
702    to break infinite loops that match no characters. When match() is called in
703    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
704    NOT be used with tail recursion, because the memory block that is used is on
705    the stack, so a new one may be required for each match(). */
706    
707  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
708    {    {
709    eptrblock *p;    newptrb.epb_saved_eptr = eptr;
710    if ((flags & match_tail_recursed) != 0)    newptrb.epb_prev = eptrb;
711      {    eptrb = &newptrb;
712      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;  
713    }    }
714    
715  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 588  for (;;) Line 719  for (;;)
719    minimize = possessive = FALSE;    minimize = possessive = FALSE;
720    op = *ecode;    op = *ecode;
721    
   /* 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;  
   
722    switch(op)    switch(op)
723      {      {
724      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
725      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
726      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
727      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,
728      reference inside the group.        eptrb, RM55);
729        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
730      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
731      values of the final offsets, in case they were set by a previous iteration  
732      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
733        argument, and we must check whether that argument matches this MARK's
734        argument. It is passed back in md->start_match_ptr (an overloading of that
735        variable). If it does match, we reset that variable to the current subject
736        position and return MATCH_SKIP. Otherwise, pass back the return code
737        unaltered. */
738    
739        else if (rrc == MATCH_SKIP_ARG &&
740            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
741          {
742          md->start_match_ptr = eptr;
743          RRETURN(MATCH_SKIP);
744          }
745        RRETURN(rrc);
746    
747        case OP_FAIL:
748        RRETURN(MATCH_NOMATCH);
749    
750        /* COMMIT overrides PRUNE, SKIP, and THEN */
751    
752        case OP_COMMIT:
753        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
754          eptrb, RM52);
755        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
756            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
757            rrc != MATCH_THEN)
758          RRETURN(rrc);
759        RRETURN(MATCH_COMMIT);
760    
761        /* PRUNE overrides THEN */
762    
763        case OP_PRUNE:
764        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
765          eptrb, RM51);
766        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
767        RRETURN(MATCH_PRUNE);
768    
769        case OP_PRUNE_ARG:
770        md->nomatch_mark = ecode + 2;
771        md->mark = NULL;    /* In case previously set by assertion */
772        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
773          eptrb, RM56);
774        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
775             md->mark == NULL) md->mark = ecode + 2;
776        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
777        RRETURN(MATCH_PRUNE);
778    
779        /* SKIP overrides PRUNE and THEN */
780    
781        case OP_SKIP:
782        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
783          eptrb, RM53);
784        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
785          RRETURN(rrc);
786        md->start_match_ptr = eptr;   /* Pass back current position */
787        RRETURN(MATCH_SKIP);
788    
789        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
790        nomatch_mark. There is a flag that disables this opcode when re-matching a
791        pattern that ended with a SKIP for which there was not a matching MARK. */
792    
793        case OP_SKIP_ARG:
794        if (md->ignore_skip_arg)
795          {
796          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
797          break;
798          }
799        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800          eptrb, RM57);
801        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
802          RRETURN(rrc);
803    
804        /* Pass back the current skip name by overloading md->start_match_ptr and
805        returning the special MATCH_SKIP_ARG return code. This will either be
806        caught by a matching MARK, or get to the top, where it causes a rematch
807        with the md->ignore_skip_arg flag set. */
808    
809        md->start_match_ptr = ecode + 2;
810        RRETURN(MATCH_SKIP_ARG);
811    
812        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
813        the branch in which it occurs can be determined. Overload the start of
814        match pointer to do this. */
815    
816        case OP_THEN:
817        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
818          eptrb, RM54);
819        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
820        md->start_match_ptr = ecode;
821        RRETURN(MATCH_THEN);
822    
823        case OP_THEN_ARG:
824        md->nomatch_mark = ecode + 2;
825        md->mark = NULL;    /* In case previously set by assertion */
826        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
827          md, eptrb, RM58);
828        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
829             md->mark == NULL) md->mark = ecode + 2;
830        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831        md->start_match_ptr = ecode;
832        RRETURN(MATCH_THEN);
833    
834        /* Handle an atomic group that does not contain any capturing parentheses.
835        This can be handled like an assertion. Prior to 8.13, all atomic groups
836        were handled this way. In 8.13, the code was changed as below for ONCE, so
837        that backups pass through the group and thereby reset captured values.
838        However, this uses a lot more stack, so in 8.20, atomic groups that do not
839        contain any captures generate OP_ONCE_NC, which can be handled in the old,
840        less stack intensive way.
841    
842        Check the alternative branches in turn - the matching won't pass the KET
843        for this kind of subpattern. If any one branch matches, we carry on as at
844        the end of a normal bracket, leaving the subject pointer, but resetting
845        the start-of-match value in case it was changed by \K. */
846    
847        case OP_ONCE_NC:
848        prev = ecode;
849        saved_eptr = eptr;
850        save_mark = md->mark;
851        do
852          {
853          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
854          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
855            {
856            mstart = md->start_match_ptr;
857            break;
858            }
859          if (rrc == MATCH_THEN)
860            {
861            next = ecode + GET(ecode,1);
862            if (md->start_match_ptr < next &&
863                (*ecode == OP_ALT || *next == OP_ALT))
864              rrc = MATCH_NOMATCH;
865            }
866    
867          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
868          ecode += GET(ecode,1);
869          md->mark = save_mark;
870          }
871        while (*ecode == OP_ALT);
872    
873        /* If hit the end of the group (which could be repeated), fail */
874    
875        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
876    
877        /* Continue as from after the group, updating the offsets high water
878        mark, since extracts may have been taken. */
879    
880        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
881    
882        offset_top = md->end_offset_top;
883        eptr = md->end_match_ptr;
884    
885        /* For a non-repeating ket, just continue at this level. This also
886        happens for a repeating ket if no characters were matched in the group.
887        This is the forcible breaking of infinite loops as implemented in Perl
888        5.005. */
889    
890        if (*ecode == OP_KET || eptr == saved_eptr)
891          {
892          ecode += 1+LINK_SIZE;
893          break;
894          }
895    
896        /* The repeating kets try the rest of the pattern or restart from the
897        preceding bracket, in the appropriate order. The second "call" of match()
898        uses tail recursion, to avoid using another stack frame. */
899    
900        if (*ecode == OP_KETRMIN)
901          {
902          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
903          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
904          ecode = prev;
905          goto TAIL_RECURSE;
906          }
907        else  /* OP_KETRMAX */
908          {
909          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
910          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
911          ecode += 1 + LINK_SIZE;
912          goto TAIL_RECURSE;
913          }
914        /* Control never gets here */
915    
916        /* Handle a capturing bracket, other than those that are possessive with an
917        unlimited repeat. If there is space in the offset vector, save the current
918        subject position in the working slot at the top of the vector. We mustn't
919        change the current values of the data slot, because they may be set from a
920        previous iteration of this group, and be referred to by a reference inside
921        the group. A failure to match might occur after the group has succeeded,
922        if something later on doesn't match. For this reason, we need to restore
923        the working value and also the values of the final offsets, in case they
924        were set by a previous iteration of the same bracket.
925    
926      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
927      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 932  for (;;)
932      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
933      offset = number << 1;      offset = number << 1;
934    
935  #ifdef DEBUG  #ifdef PCRE_DEBUG
936      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
937      printf("subject=");      printf("subject=");
938      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 630  for (;;) Line 945  for (;;)
945        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
946        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
947        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
948          save_mark = md->mark;
949    
950        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
951        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
952            (int)(eptr - md->start_subject);
953    
954        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
955          {          {
956          RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
957            ims, eptrb, flags);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
958              eptrb, RM1);
959            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
960    
961            /* If we backed up to a THEN, check whether it is within the current
962            branch by comparing the address of the THEN that is passed back with
963            the end of the branch. If it is within the current branch, and the
964            branch is one of two or more alternatives (it either starts or ends
965            with OP_ALT), we have reached the limit of THEN's action, so convert
966            the return code to NOMATCH, which will cause normal backtracking to
967            happen from now on. Otherwise, THEN is passed back to an outer
968            alternative. This implements Perl's treatment of parenthesized groups,
969            where a group not containing | does not affect the current alternative,
970            that is, (X) is NOT the same as (X|(*F)). */
971    
972            if (rrc == MATCH_THEN)
973              {
974              next = ecode + GET(ecode,1);
975              if (md->start_match_ptr < next &&
976                  (*ecode == OP_ALT || *next == OP_ALT))
977                rrc = MATCH_NOMATCH;
978              }
979    
980            /* Anything other than NOMATCH is passed back. */
981    
982          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
983          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
984          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
985            md->mark = save_mark;
986            if (*ecode != OP_ALT) break;
987          }          }
       while (*ecode == OP_ALT);  
988    
989        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
990        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
991        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
992        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
993    
994        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
995    
996          RRETURN(rrc);
997        }        }
998    
999      /* Insufficient room for saving captured contents. Treat as a non-capturing      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1000      bracket. */      as a non-capturing bracket. */
1001    
1002        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1003        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1004    
1005      DPRINTF(("insufficient capture room: treat as non-capturing\n"));      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1006    
1007      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1008      final alternative within the brackets, we would return the result of a      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion. */  
1009    
1010        /* Non-capturing or atomic group, except for possessive with unlimited
1011        repeat and ONCE group with no captures. Loop for all the alternatives.
1012    
1013        When we get to the final alternative within the brackets, we used to return
1014        the result of a recursive call to match() whatever happened so it was
1015        possible to reduce stack usage by turning this into a tail recursion,
1016        except in the case of a possibly empty group. However, now that there is
1017        the possiblity of (*THEN) occurring in the final alternative, this
1018        optimization is no longer always possible.
1019    
1020        We can optimize if we know there are no (*THEN)s in the pattern; at present
1021        this is the best that can be done.
1022    
1023        MATCH_ONCE is returned when the end of an atomic group is successfully
1024        reached, but subsequent matching fails. It passes back up the tree (causing
1025        captured values to be reset) until the original atomic group level is
1026        reached. This is tested by comparing md->once_target with the start of the
1027        group. At this point, the return is converted into MATCH_NOMATCH so that
1028        previous backup points can be taken. */
1029    
1030        case OP_ONCE:
1031      case OP_BRA:      case OP_BRA:
1032      case OP_SBRA:      case OP_SBRA:
1033      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1034      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1035      for (;;)      for (;;)
1036        {        {
1037        if (ecode[GET(ecode, 1)] != OP_ALT)        if (op >= OP_SBRA || op == OP_ONCE)
1038            md->match_function_type = MATCH_CBEGROUP;
1039    
1040          /* If this is not a possibly empty group, and there are no (*THEN)s in
1041          the pattern, and this is the final alternative, optimize as described
1042          above. */
1043    
1044          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1045          {          {
1046          ecode += _pcre_OP_lengths[*ecode];          ecode += PRIV(OP_lengths)[*ecode];
         flags |= match_tail_recursed;  
         DPRINTF(("bracket 0 tail recursion\n"));  
1047          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1048          }          }
1049    
1050        /* 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. */  
1051    
1052        RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1053          eptrb, flags);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1054        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RM2);
1055    
1056          /* See comment in the code for capturing groups above about handling
1057          THEN. */
1058    
1059          if (rrc == MATCH_THEN)
1060            {
1061            next = ecode + GET(ecode,1);
1062            if (md->start_match_ptr < next &&
1063                (*ecode == OP_ALT || *next == OP_ALT))
1064              rrc = MATCH_NOMATCH;
1065            }
1066    
1067          if (rrc != MATCH_NOMATCH)
1068            {
1069            if (rrc == MATCH_ONCE)
1070              {
1071              const pcre_uchar *scode = ecode;
1072              if (*scode != OP_ONCE)           /* If not at start, find it */
1073                {
1074                while (*scode == OP_ALT) scode += GET(scode, 1);
1075                scode -= GET(scode, 1);
1076                }
1077              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1078              }
1079            RRETURN(rrc);
1080            }
1081        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1082          md->mark = save_mark;
1083          if (*ecode != OP_ALT) break;
1084        }        }
     /* Control never reaches here. */  
1085    
1086      /* 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. */  
1087    
1088      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1089      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1090      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1091        {      different. The end of these brackets will always be OP_KETRPOS, which
1092        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1093        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1094          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1095    
1096      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      case OP_CBRAPOS:
1097        {      case OP_SCBRAPOS:
1098        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);  
       }  
1099    
1100      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      POSSESSIVE_CAPTURE:
1101        {      number = GET2(ecode, 1+LINK_SIZE);
1102        condition = FALSE;      offset = number << 1;
       ecode += GET(ecode, 1);  
       }  
1103    
1104      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1105      the final argument match_condassert causes it to stop at the end of an      printf("start possessive bracket %d\n", number);
1106      assertion. */      printf("subject=");
1107        pchars(eptr, 16, TRUE, md);
1108        printf("\n");
1109    #endif
1110    
1111      else      if (offset < md->offset_max)
1112        {        {
1113        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1114            match_condassert);        code_offset = (int)(ecode - md->start_code);
1115        if (rrc == MATCH_MATCH)  
1116          {        save_offset1 = md->offset_vector[offset];
1117          condition = TRUE;        save_offset2 = md->offset_vector[offset+1];
1118          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        save_offset3 = md->offset_vector[md->offset_end - number];
1119          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
1120    
1121          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1122    
1123          /* Each time round the loop, save the current subject position for use
1124          when the group matches. For MATCH_MATCH, the group has matched, so we
1125          restart it with a new subject starting position, remembering that we had
1126          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1127          usual. If we haven't matched any alternatives in any iteration, check to
1128          see if a previous iteration matched. If so, the group has matched;
1129          continue from afterwards. Otherwise it has failed; restore the previous
1130          capture values before returning NOMATCH. */
1131    
1132          for (;;)
1133            {
1134            md->offset_vector[md->offset_end - number] =
1135              (int)(eptr - md->start_subject);
1136            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1137            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1138              eptrb, RM63);
1139            if (rrc == MATCH_KETRPOS)
1140              {
1141              offset_top = md->end_offset_top;
1142              eptr = md->end_match_ptr;
1143              ecode = md->start_code + code_offset;
1144              save_capture_last = md->capture_last;
1145              matched_once = TRUE;
1146              continue;
1147              }
1148    
1149            /* See comment in the code for capturing groups above about handling
1150            THEN. */
1151    
1152            if (rrc == MATCH_THEN)
1153              {
1154              next = ecode + GET(ecode,1);
1155              if (md->start_match_ptr < next &&
1156                  (*ecode == OP_ALT || *next == OP_ALT))
1157                rrc = MATCH_NOMATCH;
1158              }
1159    
1160            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1161            md->capture_last = save_capture_last;
1162            ecode += GET(ecode, 1);
1163            if (*ecode != OP_ALT) break;
1164          }          }
1165        else if (rrc != MATCH_NOMATCH)  
1166          if (!matched_once)
1167          {          {
1168          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1169            md->offset_vector[offset+1] = save_offset2;
1170            md->offset_vector[md->offset_end - number] = save_offset3;
1171          }          }
1172        else  
1173          if (allow_zero || matched_once)
1174          {          {
1175          condition = FALSE;          ecode += 1 + LINK_SIZE;
1176          ecode += GET(ecode, 1);          break;
1177          }          }
       }  
   
     /* We are now at the branch that is to be obeyed. As there is only one,  
     we can use tail recursion to avoid using another stack frame. If the second  
     alternative doesn't exist, we can just plough on. */  
1178    
1179      if (condition || *ecode == OP_ALT)        RRETURN(MATCH_NOMATCH);
       {  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed | ((op == OP_SCOND)? match_cbegroup : 0);  
       goto TAIL_RECURSE;  
       }  
     else  
       {  
       ecode += 1 + LINK_SIZE;  
1180        }        }
     break;  
1181    
1182        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1183        as a non-capturing bracket. */
1184    
1185      /* End of the pattern. If we are in a top-level recursion, we should      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1186      restore the offsets appropriately and continue from after the call. */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
   
     case OP_END:  
     if (md->recursive != NULL && md->recursive->group_num == 0)  
       {  
       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;  
       }  
1187    
1188      /* 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. */  
1189    
1190      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1191      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;  
1192    
1193      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1194      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,
1195      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1196      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1197      this level is identical to the lookahead case. */  
1198        case OP_BRAPOS:
1199        case OP_SBRAPOS:
1200        allow_zero = FALSE;
1201    
1202        POSSESSIVE_NON_CAPTURE:
1203        matched_once = FALSE;
1204        code_offset = (int)(ecode - md->start_code);
1205    
1206      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1207        {        {
1208        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1209        if (rrc == MATCH_MATCH) break;        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1210            eptrb, RM48);
1211          if (rrc == MATCH_KETRPOS)
1212            {
1213            offset_top = md->end_offset_top;
1214            eptr = md->end_match_ptr;
1215            ecode = md->start_code + code_offset;
1216            matched_once = TRUE;
1217            continue;
1218            }
1219    
1220          /* See comment in the code for capturing groups above about handling
1221          THEN. */
1222    
1223          if (rrc == MATCH_THEN)
1224            {
1225            next = ecode + GET(ecode,1);
1226            if (md->start_match_ptr < next &&
1227                (*ecode == OP_ALT || *next == OP_ALT))
1228              rrc = MATCH_NOMATCH;
1229            }
1230    
1231        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1232        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1233          if (*ecode != OP_ALT) break;
1234        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1235    
1236      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1237          {
1238          ecode += 1 + LINK_SIZE;
1239          break;
1240          }
1241        RRETURN(MATCH_NOMATCH);
1242    
1243      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1244    
1245      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1246      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1247        past the end if there is only one branch, but that's OK because that is
1248        exactly what going to the ket would do. */
1249    
1250      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1251      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1252      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1253    
1254      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1255        inserted between OP_COND and an assertion condition. */
1256    
1257      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1258        {        {
1259        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (PUBL(callout) != NULL)
1260        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          {
1261        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          PUBL(callout_block) cb;
1262        ecode += GET(ecode,1);          cb.version          = 2;   /* Version 1 of the callout block */
1263        }          cb.callout_number   = ecode[LINK_SIZE+2];
1264      while (*ecode == OP_ALT);          cb.offset_vector    = md->offset_vector;
1265    #ifdef COMPILE_PCRE8
1266      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);          cb.subject          = (PCRE_SPTR)md->start_subject;
1267    #else
1268            cb.subject          = (PCRE_SPTR16)md->start_subject;
1269    #endif
1270            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1271            cb.start_match      = (int)(mstart - md->start_subject);
1272            cb.current_position = (int)(eptr - md->start_subject);
1273            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1274            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1275            cb.capture_top      = offset_top/2;
1276            cb.capture_last     = md->capture_last;
1277            cb.callout_data     = md->callout_data;
1278            cb.mark             = md->nomatch_mark;
1279            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1280            if (rrc < 0) RRETURN(rrc);
1281            }
1282          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1283          }
1284    
1285        condcode = ecode[LINK_SIZE+1];
1286    
1287        /* Now see what the actual condition is */
1288    
1289        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1290          {
1291          if (md->recursive == NULL)                /* Not recursing => FALSE */
1292            {
1293            condition = FALSE;
1294            ecode += GET(ecode, 1);
1295            }
1296          else
1297            {
1298            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1299            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1300    
1301            /* If the test is for recursion into a specific subpattern, and it is
1302            false, but the test was set up by name, scan the table to see if the
1303            name refers to any other numbers, and test them. The condition is true
1304            if any one is set. */
1305    
1306            if (!condition && condcode == OP_NRREF)
1307              {
1308              pcre_uchar *slotA = md->name_table;
1309              for (i = 0; i < md->name_count; i++)
1310                {
1311                if (GET2(slotA, 0) == recno) break;
1312                slotA += md->name_entry_size;
1313                }
1314    
1315              /* Found a name for the number - there can be only one; duplicate
1316              names for different numbers are allowed, but not vice versa. First
1317              scan down for duplicates. */
1318    
1319              if (i < md->name_count)
1320                {
1321                pcre_uchar *slotB = slotA;
1322                while (slotB > md->name_table)
1323                  {
1324                  slotB -= md->name_entry_size;
1325                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1326                    {
1327                    condition = GET2(slotB, 0) == md->recursive->group_num;
1328                    if (condition) break;
1329                    }
1330                  else break;
1331                  }
1332    
1333                /* Scan up for duplicates */
1334    
1335                if (!condition)
1336                  {
1337                  slotB = slotA;
1338                  for (i++; i < md->name_count; i++)
1339                    {
1340                    slotB += md->name_entry_size;
1341                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1342                      {
1343                      condition = GET2(slotB, 0) == md->recursive->group_num;
1344                      if (condition) break;
1345                      }
1346                    else break;
1347                    }
1348                  }
1349                }
1350              }
1351    
1352            /* Chose branch according to the condition */
1353    
1354            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1355            }
1356          }
1357    
1358        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1359          {
1360          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1361          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1362    
1363          /* If the numbered capture is unset, but the reference was by name,
1364          scan the table to see if the name refers to any other numbers, and test
1365          them. The condition is true if any one is set. This is tediously similar
1366          to the code above, but not close enough to try to amalgamate. */
1367    
1368          if (!condition && condcode == OP_NCREF)
1369            {
1370            int refno = offset >> 1;
1371            pcre_uchar *slotA = md->name_table;
1372    
1373            for (i = 0; i < md->name_count; i++)
1374              {
1375              if (GET2(slotA, 0) == refno) break;
1376              slotA += md->name_entry_size;
1377              }
1378    
1379            /* Found a name for the number - there can be only one; duplicate names
1380            for different numbers are allowed, but not vice versa. First scan down
1381            for duplicates. */
1382    
1383            if (i < md->name_count)
1384              {
1385              pcre_uchar *slotB = slotA;
1386              while (slotB > md->name_table)
1387                {
1388                slotB -= md->name_entry_size;
1389                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1390                  {
1391                  offset = GET2(slotB, 0) << 1;
1392                  condition = offset < offset_top &&
1393                    md->offset_vector[offset] >= 0;
1394                  if (condition) break;
1395                  }
1396                else break;
1397                }
1398    
1399              /* Scan up for duplicates */
1400    
1401              if (!condition)
1402                {
1403                slotB = slotA;
1404                for (i++; i < md->name_count; i++)
1405                  {
1406                  slotB += md->name_entry_size;
1407                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1408                    {
1409                    offset = GET2(slotB, 0) << 1;
1410                    condition = offset < offset_top &&
1411                      md->offset_vector[offset] >= 0;
1412                    if (condition) break;
1413                    }
1414                  else break;
1415                  }
1416                }
1417              }
1418            }
1419    
1420          /* Chose branch according to the condition */
1421    
1422          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1423          }
1424    
1425        else if (condcode == OP_DEF)     /* DEFINE - always false */
1426          {
1427          condition = FALSE;
1428          ecode += GET(ecode, 1);
1429          }
1430    
1431        /* The condition is an assertion. Call match() to evaluate it - setting
1432        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1433        an assertion. */
1434    
1435        else
1436          {
1437          md->match_function_type = MATCH_CONDASSERT;
1438          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1439          if (rrc == MATCH_MATCH)
1440            {
1441            if (md->end_offset_top > offset_top)
1442              offset_top = md->end_offset_top;  /* Captures may have happened */
1443            condition = TRUE;
1444            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1445            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1446            }
1447    
1448          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1449          assertion; it is therefore treated as NOMATCH. */
1450    
1451          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1452            {
1453            RRETURN(rrc);         /* Need braces because of following else */
1454            }
1455          else
1456            {
1457            condition = FALSE;
1458            ecode += codelink;
1459            }
1460          }
1461    
1462        /* We are now at the branch that is to be obeyed. As there is only one, can
1463        use tail recursion to avoid using another stack frame, except when there is
1464        unlimited repeat of a possibly empty group. In the latter case, a recursive
1465        call to match() is always required, unless the second alternative doesn't
1466        exist, in which case we can just plough on. Note that, for compatibility
1467        with Perl, the | in a conditional group is NOT treated as creating two
1468        alternatives. If a THEN is encountered in the branch, it propagates out to
1469        the enclosing alternative (unless nested in a deeper set of alternatives,
1470        of course). */
1471    
1472        if (condition || *ecode == OP_ALT)
1473          {
1474          if (op != OP_SCOND)
1475            {
1476            ecode += 1 + LINK_SIZE;
1477            goto TAIL_RECURSE;
1478            }
1479    
1480          md->match_function_type = MATCH_CBEGROUP;
1481          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1482          RRETURN(rrc);
1483          }
1484    
1485         /* Condition false & no alternative; continue after the group. */
1486    
1487        else
1488          {
1489          ecode += 1 + LINK_SIZE;
1490          }
1491        break;
1492    
1493    
1494        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1495        to close any currently open capturing brackets. */
1496    
1497        case OP_CLOSE:
1498        number = GET2(ecode, 1);
1499        offset = number << 1;
1500    
1501    #ifdef PCRE_DEBUG
1502          printf("end bracket %d at *ACCEPT", number);
1503          printf("\n");
1504    #endif
1505    
1506        md->capture_last = number;
1507        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1508          {
1509          md->offset_vector[offset] =
1510            md->offset_vector[md->offset_end - number];
1511          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1512          if (offset_top <= offset) offset_top = offset + 2;
1513          }
1514        ecode += 1 + IMM2_SIZE;
1515        break;
1516    
1517    
1518        /* End of the pattern, either real or forced. */
1519    
1520        case OP_END:
1521        case OP_ACCEPT:
1522        case OP_ASSERT_ACCEPT:
1523    
1524        /* If we have matched an empty string, fail if not in an assertion and not
1525        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1526        is set and we have matched at the start of the subject. In both cases,
1527        backtracking will then try other alternatives, if any. */
1528    
1529        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1530             md->recursive == NULL &&
1531             (md->notempty ||
1532               (md->notempty_atstart &&
1533                 mstart == md->start_subject + md->start_offset)))
1534          RRETURN(MATCH_NOMATCH);
1535    
1536        /* Otherwise, we have a match. */
1537    
1538        md->end_match_ptr = eptr;           /* Record where we ended */
1539        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1540        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1541    
1542        /* For some reason, the macros don't work properly if an expression is
1543        given as the argument to RRETURN when the heap is in use. */
1544    
1545        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1546        RRETURN(rrc);
1547    
1548        /* Assertion brackets. Check the alternative branches in turn - the
1549        matching won't pass the KET for an assertion. If any one branch matches,
1550        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1551        start of each branch to move the current point backwards, so the code at
1552        this level is identical to the lookahead case. When the assertion is part
1553        of a condition, we want to return immediately afterwards. The caller of
1554        this incarnation of the match() function will have set MATCH_CONDASSERT in
1555        md->match_function type, and one of these opcodes will be the first opcode
1556        that is processed. We use a local variable that is preserved over calls to
1557        match() to remember this case. */
1558    
1559        case OP_ASSERT:
1560        case OP_ASSERTBACK:
1561        save_mark = md->mark;
1562        if (md->match_function_type == MATCH_CONDASSERT)
1563          {
1564          condassert = TRUE;
1565          md->match_function_type = 0;
1566          }
1567        else condassert = FALSE;
1568    
1569        do
1570          {
1571          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1572          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1573            {
1574            mstart = md->start_match_ptr;   /* In case \K reset it */
1575            break;
1576            }
1577          md->mark = save_mark;
1578    
1579          /* A COMMIT failure must fail the entire assertion, without trying any
1580          subsequent branches. */
1581    
1582          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1583    
1584          /* PCRE does not allow THEN to escape beyond an assertion; it
1585          is treated as NOMATCH. */
1586    
1587          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1588          ecode += GET(ecode, 1);
1589          }
1590        while (*ecode == OP_ALT);
1591    
1592        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 or COMMIT to escape beyond a recursion; it
1797            is treated as NOMATCH. */
1798    
1799            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1800                     rrc != MATCH_COMMIT)
1801            {            {
1802            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1803              if (new_recursive.offset_save != stacksave)
1804                (PUBL(free))(new_recursive.offset_save);
1805            RRETURN(rrc);            RRETURN(rrc);
1806            }            }
1807    
1808          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1809          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1810          }          }
1811        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 983  for (;;) Line 1813  for (;;)
1813        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1814        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1815        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1816          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1817        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1818        }        }
     /* 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);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
1819    
1820      do ecode += GET(ecode, 1); while (*ecode == OP_ALT);      RECURSION_MATCHED:
1821        break;
     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 */  
1822    
1823      /* 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
1824      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1070  for (;;) Line 1827  for (;;)
1827      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1828      break;      break;
1829    
1830      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1831      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
1832      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
1833      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
1834      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1835    
1836      case OP_BRAZERO:      case OP_BRAZERO:
1837        {      next = ecode + 1;
1838        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1839        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1840        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1841        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1842      break;      break;
1843    
1844      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1845        {      next = ecode + 1;
1846        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1847        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1848        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1849        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1850      break;      break;
1851    
1852        case OP_SKIPZERO:
1853        next = ecode+1;
1854        do next += GET(next,1); while (*next == OP_ALT);
1855        ecode = next + 1 + LINK_SIZE;
1856        break;
1857    
1858        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1859        here; just jump to the group, with allow_zero set TRUE. */
1860    
1861        case OP_BRAPOSZERO:
1862        op = *(++ecode);
1863        allow_zero = TRUE;
1864        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1865          goto POSSESSIVE_NON_CAPTURE;
1866    
1867      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1868    
1869      case OP_KET:      case OP_KET:
1870      case OP_KETRMIN:      case OP_KETRMIN:
1871      case OP_KETRMAX:      case OP_KETRMAX:
1872        case OP_KETRPOS:
1873      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1874    
1875      /* 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
1876      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1877      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1878    
1879      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1880        {        {
1881        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1882        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1883        }        }
1884      else saved_eptr = NULL;      else saved_eptr = NULL;
1885    
1886      /* 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
1887      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
1888      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1889        start in case it was changed by \K. */
1890      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1891          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1892          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1893        {        {
1894        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1895        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1896        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1897          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1898        }        }
1899    
1900      /* 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
1901      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1902      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
1903      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
1904      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
1905        the current subject position and start match pointer and give a MATCH
1906        return. */
1907    
1908      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1909            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1910        {        {
1911        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1912        offset = number << 1;        offset = number << 1;
1913    
1914  #ifdef DEBUG  #ifdef PCRE_DEBUG
1915        printf("end bracket %d", number);        printf("end bracket %d", number);
1916        printf("\n");        printf("\n");
1917  #endif  #endif
1918    
1919          /* Handle a recursively called group. */
1920    
1921          if (md->recursive != NULL && md->recursive->group_num == number)
1922            {
1923            md->end_match_ptr = eptr;
1924            md->start_match_ptr = mstart;
1925            RRETURN(MATCH_MATCH);
1926            }
1927    
1928          /* Deal with capturing */
1929    
1930        md->capture_last = number;        md->capture_last = number;
1931        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1932          {          {
1933            /* If offset is greater than offset_top, it means that we are
1934            "skipping" a capturing group, and that group's offsets must be marked
1935            unset. In earlier versions of PCRE, all the offsets were unset at the
1936            start of matching, but this doesn't work because atomic groups and
1937            assertions can cause a value to be set that should later be unset.
1938            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1939            part of the atomic group, but this is not on the final matching path,
1940            so must be unset when 2 is set. (If there is no group 2, there is no
1941            problem, because offset_top will then be 2, indicating no capture.) */
1942    
1943            if (offset > offset_top)
1944              {
1945              register int *iptr = md->offset_vector + offset_top;
1946              register int *iend = md->offset_vector + offset;
1947              while (iptr < iend) *iptr++ = -1;
1948              }
1949    
1950            /* Now make the extraction */
1951    
1952          md->offset_vector[offset] =          md->offset_vector[offset] =
1953            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1954          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1955          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1956          }          }
1957          }
1958    
1959        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1960        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1961        group. This is the forcible breaking of infinite loops as implemented in
1962        Perl 5.005. For a non-repeating atomic group that includes captures,
1963        establish a backup point by processing the rest of the pattern at a lower
1964        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1965        original OP_ONCE level, thereby bypassing intermediate backup points, but
1966        resetting any captures that happened along the way. */
1967    
1968        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1969          {
1970          if (*prev == OP_ONCE)
1971          {          {
1972          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1973          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1974          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1975          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;  
1976          }          }
1977          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1978          break;
1979        }        }
1980    
1981      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1982      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1983        at a time from the outer level, thus saving stack. */
     ims = original_ims;  
     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. */  
1984    
1985      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KETRPOS)
1986        {        {
1987        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1988        break;        md->end_offset_top = offset_top;
1989          RRETURN(MATCH_KETRPOS);
1990        }        }
1991    
1992      /* 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
1993      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
1994      tail recursion to avoid using another stack frame. */      use tail recursion to avoid using another stack frame, unless we have an
1995        an atomic group or an unlimited repeat of a group that can match an empty
1996      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      string. */
1997    
1998      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1999        {        {
2000        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2001        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2002          if (*prev == OP_ONCE)
2003            {
2004            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2005            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2006            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2007            RRETURN(MATCH_ONCE);
2008            }
2009          if (*prev >= OP_SBRA)    /* Could match an empty string */
2010            {
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);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2020          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2021        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2022          if (*prev == OP_ONCE)
2023            {
2024            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2025            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2026            md->once_target = prev;
2027            RRETURN(MATCH_ONCE);
2028            }
2029        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_tail_recursed;  
2030        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2031        }        }
2032      /* Control never gets here */      /* Control never gets here */
2033    
2034      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2035    
2036      case OP_CIRC:      case OP_CIRC:
2037      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2038    
2039      /* Start of subject assertion */      /* Start of subject assertion */
2040    
# Line 1232  for (;;) Line 2043  for (;;)
2043      ecode++;      ecode++;
2044      break;      break;
2045    
2046        /* Multiline mode: start of subject unless notbol, or after any newline. */
2047    
2048        case OP_CIRCM:
2049        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2050        if (eptr != md->start_subject &&
2051            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2052          RRETURN(MATCH_NOMATCH);
2053        ecode++;
2054        break;
2055    
2056      /* Start of match assertion */      /* Start of match assertion */
2057    
2058      case OP_SOM:      case OP_SOM:
# Line 1239  for (;;) Line 2060  for (;;)
2060      ecode++;      ecode++;
2061      break;      break;
2062    
2063      /* 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. */  
2064    
2065      case OP_DOLL:      case OP_SET_SOM:
2066      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2067        ecode++;
2068        break;
2069    
2070        /* Multiline mode: assert before any newline, or before end of subject
2071        unless noteol is set. */
2072    
2073        case OP_DOLLM:
2074        if (eptr < md->end_subject)
2075        {        {
2076        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2077          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2078        else          if (md->partial != 0 &&
2079          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2080        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2081        break;              NLBLOCK->nllen == 2 &&
2082                *eptr == NLBLOCK->nl[0])
2083              {
2084              md->hitend = TRUE;
2085              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2086              }
2087            RRETURN(MATCH_NOMATCH);
2088            }
2089        }        }
2090      else      else
2091        {        {
2092        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2093        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2094        }        }
2095        ecode++;
2096        break;
2097    
2098        /* Not multiline mode: assert before a terminating newline or before end of
2099        subject unless noteol is set. */
2100    
2101        case OP_DOLL:
2102        if (md->noteol) RRETURN(MATCH_NOMATCH);
2103        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2104    
2105      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2106    
2107      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2108    
2109      case OP_EOD:      case OP_EOD:
2110      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2111        SCHECK_PARTIAL();
2112      ecode++;      ecode++;
2113      break;      break;
2114    
2115      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2116    
2117      case OP_EODN:      case OP_EODN:
2118      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2119        if (eptr < md->end_subject &&
2120          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2121          {
2122          if (md->partial != 0 &&
2123              eptr + 1 >= md->end_subject &&
2124              NLBLOCK->nltype == NLTYPE_FIXED &&
2125              NLBLOCK->nllen == 2 &&
2126              *eptr == NLBLOCK->nl[0])
2127            {
2128            md->hitend = TRUE;
2129            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2130            }
2131        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2132          }
2133    
2134        /* Either at end of string or \n before end. */
2135    
2136        SCHECK_PARTIAL();
2137      ecode++;      ecode++;
2138      break;      break;
2139    
# Line 1290  for (;;) Line 2145  for (;;)
2145    
2146        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2147        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2148        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2149          partial matching. */
2150    
2151  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2152        if (utf8)        if (utf)
2153          {          {
2154            /* Get status of previous character */
2155    
2156          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2157            {            {
2158            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2159            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2160              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2161            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2162    #ifdef SUPPORT_UCP
2163              if (md->use_ucp)
2164                {
2165                if (c == '_') prev_is_word = TRUE; else
2166                  {
2167                  int cat = UCD_CATEGORY(c);
2168                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2169                  }
2170                }
2171              else
2172    #endif
2173            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2174            }            }
2175          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2176            /* Get status of next character */
2177    
2178            if (eptr >= md->end_subject)
2179              {
2180              SCHECK_PARTIAL();
2181              cur_is_word = FALSE;
2182              }
2183            else
2184            {            {
2185            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2186    #ifdef SUPPORT_UCP
2187              if (md->use_ucp)
2188                {
2189                if (c == '_') cur_is_word = TRUE; else
2190                  {
2191                  int cat = UCD_CATEGORY(c);
2192                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2193                  }
2194                }
2195              else
2196    #endif
2197            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2198            }            }
2199          }          }
2200        else        else
2201    #endif
2202    
2203          /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2204          consistency with the behaviour of \w we do use it in this case. */
2205    
2206            {
2207            /* Get status of previous character */
2208    
2209            if (eptr == md->start_subject) prev_is_word = FALSE; else
2210              {
2211              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2212    #ifdef SUPPORT_UCP
2213              if (md->use_ucp)
2214                {
2215                c = eptr[-1];
2216                if (c == '_') prev_is_word = TRUE; else
2217                  {
2218                  int cat = UCD_CATEGORY(c);
2219                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2220                  }
2221                }
2222              else
2223    #endif
2224              prev_is_word = MAX_255(eptr[-1])
2225                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2226              }
2227    
2228            /* Get status of next character */
2229    
2230            if (eptr >= md->end_subject)
2231              {
2232              SCHECK_PARTIAL();
2233              cur_is_word = FALSE;
2234              }
2235            else
2236    #ifdef SUPPORT_UCP
2237            if (md->use_ucp)
2238              {
2239              c = *eptr;
2240              if (c == '_') cur_is_word = TRUE; else
2241                {
2242                int cat = UCD_CATEGORY(c);
2243                cur_is_word = (cat == ucp_L || cat == ucp_N);
2244                }
2245              }
2246            else
2247  #endif  #endif
2248            cur_is_word = MAX_255(*eptr)
2249        /* More streamlined when not in UTF-8 mode */            && ((md->ctypes[*eptr] & ctype_word) != 0);
   
         {  
         prev_is_word = (eptr != md->start_subject) &&  
           ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
         cur_is_word = (eptr < md->end_subject) &&  
           ((md->ctypes[*eptr] & ctype_word) != 0);  
2250          }          }
2251    
2252        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1328  for (;;) Line 2257  for (;;)
2257        }        }
2258      break;      break;
2259    
2260      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2261        CRLF newlines and partial matching. */
2262    
2263      case OP_ANY:      case OP_ANY:
2264      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2265        if (md->partial != 0 &&
2266            eptr + 1 >= md->end_subject &&
2267            NLBLOCK->nltype == NLTYPE_FIXED &&
2268            NLBLOCK->nllen == 2 &&
2269            *eptr == NLBLOCK->nl[0])
2270        {        {
2271        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);        md->hitend = TRUE;
2272          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2273          }
2274    
2275        /* Fall through */
2276    
2277        /* Match any single character whatsoever. */
2278    
2279        case OP_ALLANY:
2280        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2281          {                            /* not be updated before SCHECK_PARTIAL. */
2282          SCHECK_PARTIAL();
2283          RRETURN(MATCH_NOMATCH);
2284        }        }
2285      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2286      if (utf8)  #ifdef SUPPORT_UTF
2287        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2288    #endif
2289      ecode++;      ecode++;
2290      break;      break;
2291    
# Line 1345  for (;;) Line 2293  for (;;)
2293      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2294    
2295      case OP_ANYBYTE:      case OP_ANYBYTE:
2296      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2297          {                            /* not be updated before SCHECK_PARTIAL. */
2298          SCHECK_PARTIAL();
2299          RRETURN(MATCH_NOMATCH);
2300          }
2301        eptr++;
2302      ecode++;      ecode++;
2303      break;      break;
2304    
2305      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2306      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2307          {
2308          SCHECK_PARTIAL();
2309          RRETURN(MATCH_NOMATCH);
2310          }
2311      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2312      if (      if (
2313  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2314         c < 256 &&         c < 256 &&
2315  #endif  #endif
2316         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1363  for (;;) Line 2320  for (;;)
2320      break;      break;
2321    
2322      case OP_DIGIT:      case OP_DIGIT:
2323      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2324          {
2325          SCHECK_PARTIAL();
2326          RRETURN(MATCH_NOMATCH);
2327          }
2328      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2329      if (      if (
2330  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2331         c >= 256 ||         c > 255 ||
2332  #endif  #endif
2333         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2334         )         )
# Line 1376  for (;;) Line 2337  for (;;)
2337      break;      break;
2338    
2339      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2340      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2341          {
2342          SCHECK_PARTIAL();
2343          RRETURN(MATCH_NOMATCH);
2344          }
2345      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2346      if (      if (
2347  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2348         c < 256 &&         c < 256 &&
2349  #endif  #endif
2350         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1389  for (;;) Line 2354  for (;;)
2354      break;      break;
2355    
2356      case OP_WHITESPACE:      case OP_WHITESPACE:
2357      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2358          {
2359          SCHECK_PARTIAL();
2360          RRETURN(MATCH_NOMATCH);
2361          }
2362      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2363      if (      if (
2364  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2365         c >= 256 ||         c > 255 ||
2366  #endif  #endif
2367         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2368         )         )
# Line 1402  for (;;) Line 2371  for (;;)
2371      break;      break;
2372    
2373      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2374      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2375          {
2376          SCHECK_PARTIAL();
2377          RRETURN(MATCH_NOMATCH);
2378          }
2379      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2380      if (      if (
2381  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2382         c < 256 &&         c < 256 &&
2383  #endif  #endif
2384         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1415  for (;;) Line 2388  for (;;)
2388      break;      break;
2389    
2390      case OP_WORDCHAR:      case OP_WORDCHAR:
2391      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2392          {
2393          SCHECK_PARTIAL();
2394          RRETURN(MATCH_NOMATCH);
2395          }
2396      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2397      if (      if (
2398  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2399         c >= 256 ||         c > 255 ||
2400  #endif  #endif
2401         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2402         )         )
# Line 1428  for (;;) Line 2405  for (;;)
2405      break;      break;
2406    
2407      case OP_ANYNL:      case OP_ANYNL:
2408      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2409          {
2410          SCHECK_PARTIAL();
2411          RRETURN(MATCH_NOMATCH);
2412          }
2413      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2414      switch(c)      switch(c)
2415        {        {
2416        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2417        case 0x000d:  
2418        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2419          if (eptr >= md->end_subject)
2420            {
2421            SCHECK_PARTIAL();
2422            }
2423          else if (*eptr == CHAR_LF) eptr++;
2424        break;        break;
2425        case 0x000a:  
2426        case 0x000b:        case CHAR_LF:
2427        case 0x000c:        break;
2428        case 0x0085:  
2429          case CHAR_VT:
2430          case CHAR_FF:
2431          case CHAR_NEL:
2432    #ifndef EBCDIC
2433        case 0x2028:        case 0x2028:
2434        case 0x2029:        case 0x2029:
2435    #endif  /* Not EBCDIC */
2436          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2437        break;        break;
2438        }        }
2439      ecode++;      ecode++;
2440      break;      break;
2441    
2442        case OP_NOT_HSPACE:
2443        if (eptr >= md->end_subject)
2444          {
2445          SCHECK_PARTIAL();
2446          RRETURN(MATCH_NOMATCH);
2447          }
2448        GETCHARINCTEST(c, eptr);
2449        switch(c)
2450          {
2451          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2452          default: break;
2453          }
2454        ecode++;
2455        break;
2456    
2457        case OP_HSPACE:
2458        if (eptr >= md->end_subject)
2459          {
2460          SCHECK_PARTIAL();
2461          RRETURN(MATCH_NOMATCH);
2462          }
2463        GETCHARINCTEST(c, eptr);
2464        switch(c)
2465          {
2466          HSPACE_CASES: break;  /* Byte and multibyte cases */
2467          default: RRETURN(MATCH_NOMATCH);
2468          }
2469        ecode++;
2470        break;
2471    
2472        case OP_NOT_VSPACE:
2473        if (eptr >= md->end_subject)
2474          {
2475          SCHECK_PARTIAL();
2476          RRETURN(MATCH_NOMATCH);
2477          }
2478        GETCHARINCTEST(c, eptr);
2479        switch(c)
2480          {
2481          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2482          default: break;
2483          }
2484        ecode++;
2485        break;
2486    
2487        case OP_VSPACE:
2488        if (eptr >= md->end_subject)
2489          {
2490          SCHECK_PARTIAL();
2491          RRETURN(MATCH_NOMATCH);
2492          }
2493        GETCHARINCTEST(c, eptr);
2494        switch(c)
2495          {
2496          VSPACE_CASES: break;
2497          default: RRETURN(MATCH_NOMATCH);
2498          }
2499        ecode++;
2500        break;
2501    
2502  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2503      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2504      if the support is in the binary; otherwise a compile-time error occurs. */      if the support is in the binary; otherwise a compile-time error occurs. */
2505    
2506      case OP_PROP:      case OP_PROP:
2507      case OP_NOTPROP:      case OP_NOTPROP:
2508      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2509          {
2510          SCHECK_PARTIAL();
2511          RRETURN(MATCH_NOMATCH);
2512          }
2513      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2514        {        {
2515        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2516    
2517        switch(ecode[1])        switch(ecode[1])
2518          {          {
# Line 1466  for (;;) Line 2521  for (;;)
2521          break;          break;
2522    
2523          case PT_LAMP:          case PT_LAMP:
2524          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2525               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2526               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2527            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2528           break;          break;
2529    
2530          case PT_GC:          case PT_GC:
2531          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2532            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2533          break;          break;
2534    
2535          case PT_PC:          case PT_PC:
2536          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2537            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2538          break;          break;
2539    
2540          case PT_SC:          case PT_SC:
2541          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2542              RRETURN(MATCH_NOMATCH);
2543            break;
2544    
2545            /* These are specials */
2546    
2547            case PT_ALNUM:
2548            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2549                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2550              RRETURN(MATCH_NOMATCH);
2551            break;
2552    
2553            case PT_SPACE:    /* Perl space */
2554            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2555                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2556                   == (op == OP_NOTPROP))
2557            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2558          break;          break;
2559    
2560            case PT_PXSPACE:  /* POSIX space */
2561            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2562                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2563                 c == CHAR_FF || c == CHAR_CR)
2564                   == (op == OP_NOTPROP))
2565              RRETURN(MATCH_NOMATCH);
2566            break;
2567    
2568            case PT_WORD:
2569            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2570                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2571                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2572              RRETURN(MATCH_NOMATCH);
2573            break;
2574    
2575            /* This should never occur */
2576    
2577          default:          default:
2578          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2579          }          }
# Line 1499  for (;;) Line 2586  for (;;)
2586      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2587    
2588      case OP_EXTUNI:      case OP_EXTUNI:
2589      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
2590        {        {
2591        int chartype, script;        SCHECK_PARTIAL();
2592        int category = _pcre_ucp_findprop(c, &chartype, &script);        RRETURN(MATCH_NOMATCH);
2593        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        }
2594        else
2595          {
2596          int lgb, rgb;
2597          GETCHARINCTEST(c, eptr);
2598          lgb = UCD_GRAPHBREAK(c);
2599        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2600          {          {
2601          int len = 1;          int len = 1;
2602          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2603            {          rgb = UCD_GRAPHBREAK(c);
2604            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2605            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2606          eptr += len;          eptr += len;
2607          }          }
2608        }        }
2609        CHECK_PARTIAL();
2610      ecode++;      ecode++;
2611      break;      break;
2612  #endif  #endif
# Line 1531  for (;;) Line 2621  for (;;)
2621      loops). */      loops). */
2622    
2623      case OP_REF:      case OP_REF:
2624        {      case OP_REFI:
2625        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2626        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2627        ecode += 1 + IMM2_SIZE;
2628    
2629        /* 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];  
2630    
2631        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2632        this ensures that every attempt at a match fails. We can't just fail
2633        here, because of the possibility of quantifiers with zero minima.
2634    
2635        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2636          {      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;  
2637    
2638          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2639          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;  
2640    
2641          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2642          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2643          eptr += length;      else
2644          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2645    
2646        /* Set up for repetition, or handle the non-repeated case */
2647    
2648        switch (*ecode)
2649          {
2650          case OP_CRSTAR:
2651          case OP_CRMINSTAR:
2652          case OP_CRPLUS:
2653          case OP_CRMINPLUS:
2654          case OP_CRQUERY:
2655          case OP_CRMINQUERY:
2656          c = *ecode++ - OP_CRSTAR;
2657          minimize = (c & 1) != 0;
2658          min = rep_min[c];                 /* Pick up values from tables; */
2659          max = rep_max[c];                 /* zero for max => infinity */
2660          if (max == 0) max = INT_MAX;
2661          break;
2662    
2663          case OP_CRRANGE:
2664          case OP_CRMINRANGE:
2665          minimize = (*ecode == OP_CRMINRANGE);
2666          min = GET2(ecode, 1);
2667          max = GET2(ecode, 1 + IMM2_SIZE);
2668          if (max == 0) max = INT_MAX;
2669          ecode += 1 + 2 * IMM2_SIZE;
2670          break;
2671    
2672          default:               /* No repeat follows */
2673          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2674            {
2675            if (length == -2) eptr = md->end_subject;   /* Partial match */
2676            CHECK_PARTIAL();
2677            RRETURN(MATCH_NOMATCH);
2678          }          }
2679          eptr += length;
2680          continue;              /* With the main loop */
2681          }
2682    
2683        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2684        main loop. */      zero, just continue with the main loop. If the length is negative, it
2685        means the reference is unset in non-Java-compatible mode. If the minimum is
2686        zero, we can continue at the same level without recursion. For any other
2687        minimum, carrying on will result in NOMATCH. */
2688    
2689        if (length == 0) continue;      if (length == 0) continue;
2690        if (length < 0 && min == 0) continue;
2691    
2692        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2693        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2694        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2695    
2696        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2697          {
2698          int slength;
2699          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2700          {          {
2701          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2702          eptr += length;          CHECK_PARTIAL();
2703            RRETURN(MATCH_NOMATCH);
2704          }          }
2705          eptr += slength;
2706          }
2707    
2708        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2709        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2710    
2711        if (min == max) continue;      if (min == max) continue;
2712    
2713        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2714    
2715        if (minimize)      if (minimize)
2716          {
2717          for (fi = min;; fi++)
2718          {          {
2719          for (fi = min;; fi++)          int slength;
2720            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2721            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2722            if (fi >= max) RRETURN(MATCH_NOMATCH);
2723            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2724            {            {
2725            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2726            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2727            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2728            }            }
2729          /* Control never gets here */          eptr += slength;
2730          }          }
2731          /* Control never gets here */
2732          }
2733    
2734        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2735    
2736        else      else
2737          {
2738          pp = eptr;
2739          for (i = min; i < max; i++)
2740          {          {
2741          pp = eptr;          int slength;
2742          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)  
2743            {            {
2744            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2745            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2746            eptr -= length;  
2747              if (slength == -2 && md->partial != 0 &&
2748                  md->end_subject > md->start_used_ptr)
2749                {
2750                md->hitend = TRUE;
2751                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2752                }
2753              break;
2754            }            }
2755          RRETURN(MATCH_NOMATCH);          eptr += slength;
2756            }
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 2164  for (;;) Line 3384  for (;;)
3384      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3385    
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, (char *)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];
3394    #else
3395    #ifdef SUPPORT_UTF
3396    #ifdef SUPPORT_UCP
3397          if (utf && fc > 127)
3398            foc = UCD_OTHERCASE(fc);
3399    #else
3400          if (utf && fc > 127)
3401            foc = fc;
3402    #endif /* SUPPORT_UCP */
3403          else
3404    #endif /* SUPPORT_UTF */
3405            foc = TABLE_GET(fc, md->fcc, fc);
3406    #endif /* COMPILE_PCRE8 */
3407    
3408        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3409          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3410            if (eptr >= md->end_subject)
3411              {
3412              SCHECK_PARTIAL();
3413              RRETURN(MATCH_NOMATCH);
3414              }
3415            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3416            eptr++;
3417            }
3418        if (min == max) continue;        if (min == max) continue;
3419        if (minimize)        if (minimize)
3420          {          {
3421          for (fi = min;; fi++)          for (fi = min;; fi++)
3422            {            {
3423            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3424            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3425            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3426                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3427                {
3428                SCHECK_PARTIAL();
3429              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3430                }
3431              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3432              eptr++;
3433            }            }
3434          /* Control never gets here */          /* Control never gets here */
3435          }          }
# Line 2189  for (;;) Line 3438  for (;;)
3438          pp = eptr;          pp = eptr;
3439          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3440            {            {
3441            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3442                {
3443                SCHECK_PARTIAL();
3444                break;
3445                }
3446              if (fc != *eptr && foc != *eptr) break;
3447            eptr++;            eptr++;
3448            }            }
3449    
3450          if (possessive) continue;          if (possessive) continue;
3451    
3452          while (eptr >= pp)          while (eptr >= pp)
3453            {            {
3454            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3455            eptr--;            eptr--;
3456            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3457            }            }
# Line 2208  for (;;) Line 3464  for (;;)
3464    
3465      else      else
3466        {        {
3467        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3468            {
3469            if (eptr >= md->end_subject)
3470              {
3471              SCHECK_PARTIAL();
3472              RRETURN(MATCH_NOMATCH);
3473              }
3474            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3475            }
3476    
3477        if (min == max) continue;        if (min == max) continue;
3478    
3479        if (minimize)        if (minimize)
3480          {          {
3481          for (fi = min;; fi++)          for (fi = min;; fi++)
3482            {            {
3483            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3484            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3485            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3486              if (eptr >= md->end_subject)
3487                {
3488                SCHECK_PARTIAL();
3489              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3490                }
3491              if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3492            }            }
3493          /* Control never gets here */          /* Control never gets here */
3494          }          }
# Line 2226  for (;;) Line 3497  for (;;)
3497          pp = eptr;          pp = eptr;
3498          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3499            {            {
3500            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3501                {
3502                SCHECK_PARTIAL();
3503                break;
3504                }
3505              if (fc != *eptr) break;
3506            eptr++;            eptr++;
3507            }            }
3508          if (possessive) continue;          if (possessive) continue;
3509    
3510          while (eptr >= pp)          while (eptr >= pp)
3511            {            {
3512            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3513            eptr--;            eptr--;
3514            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3515            }            }
# Line 2245  for (;;) Line 3522  for (;;)
3522      checking can be multibyte. */      checking can be multibyte. */
3523    
3524      case OP_NOT:      case OP_NOT:
3525      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3526      ecode++;      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
     if ((ims & PCRE_CASELESS) != 0)  
3527        {        {
3528  #ifdef SUPPORT_UTF8        SCHECK_PARTIAL();
3529        if (c < 256)        RRETURN(MATCH_NOMATCH);
3530  #endif        }
3531        c = md->lcc[c];  #ifdef SUPPORT_UTF
3532        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);      if (utf)
3533          {
3534          register unsigned int ch, och;
3535    
3536          ecode++;
3537          GETCHARINC(ch, ecode);
3538          GETCHARINC(c, eptr);
3539    
3540          if (op == OP_NOT)
3541            {
3542            if (ch == c) RRETURN(MATCH_NOMATCH);
3543            }
3544          else
3545            {