/[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 97 by ph10, Mon Mar 5 12:36:47 2007 UTC revision 914 by zherczeg, Mon Feb 13 06:04:50 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) 1997-2006 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
43  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK md             /* Block containing newline information */  #define NLBLOCK md             /* Block containing newline information */
50  #define PSSTART start_subject  /* Field containing processed string start */  #define PSSTART start_subject  /* Field containing processed string start */
51  #define PSEND   end_subject    /* Field containing processed string end */  #define PSEND   end_subject    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55  /* The chain of eptrblocks for tail recursions uses memory in stack workspace,  /* Undefine some potentially clashing cpp symbols */
 obtained at top level, the size of which is defined by EPTR_WORK_SIZE. */  
56    
57  #define EPTR_WORK_SIZE (1000)  #undef min
58    #undef max
59    
60  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
61    of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
63    
64  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
 #define match_tail_recursed  0x04  /* Tail recursive call */  
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 65  defined PCRE_ERROR_xxx codes, which are Line 70  defined PCRE_ERROR_xxx codes, which are
70  #define MATCH_MATCH        1  #define MATCH_MATCH        1
71  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
72    
73    /* Special internal returns from the match() function. Make them sufficiently
74    negative to avoid the external error codes. */
75    
76    #define MATCH_ACCEPT       (-999)
77    #define MATCH_COMMIT       (-998)
78    #define MATCH_KETRPOS      (-997)
79    #define MATCH_ONCE         (-996)
80    #define MATCH_PRUNE        (-995)
81    #define MATCH_SKIP         (-994)
82    #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
86  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
87  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 78  static const char rep_max[] = { 0, 0, 0, Line 95  static const char rep_max[] = { 0, 0, 0,
95    
96    
97    
98  #ifdef DEBUG  #ifdef PCRE_DEBUG
99  /*************************************************  /*************************************************
100  *        Debugging function to print chars       *  *        Debugging function to print chars       *
101  *************************************************/  *************************************************/
# Line 96  Returns: nothing Line 113  Returns: nothing
113  */  */
114    
115  static void  static void
116  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
117  {  {
118  unsigned int c;  unsigned int c;
119  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
# Line 111  while (length-- > 0) Line 128  while (length-- > 0)
128  *          Match a back-reference                *  *          Match a back-reference                *
129  *************************************************/  *************************************************/
130    
131  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
132  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
133    the length passed is zero. Note that in caseless UTF-8 mode, the number of
134    subject bytes matched may be different to the number of reference bytes.
135    
136  Arguments:  Arguments:
137    offset      index into the offset vector    offset      index into the offset vector
138    eptr        points into the subject    eptr        pointer into the subject
139    length      length to be matched    length      length of reference to be matched (number of bytes)
140    md          points to match data block    md          points to match data block
141    ims         the ims flags    caseless    TRUE if caseless
142    
143  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
144  */  */
145    
146  static BOOL  static int
147  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
148    unsigned long int ims)    BOOL caseless)
149  {  {
150  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
151    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
152    
153  #ifdef DEBUG  #ifdef PCRE_DEBUG
154  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
155    printf("matching subject <null>");    printf("matching subject <null>");
156  else  else
# Line 143  pchars(p, length, FALSE, md); Line 163  pchars(p, length, FALSE, md);
163  printf("\n");  printf("\n");
164  #endif  #endif
165    
166  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
167    
168  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
169    
170  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
171    properly if Unicode properties are supported. Otherwise, we can check only
172    ASCII characters. */
173    
174  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
175    {    {
176    while (length-- > 0)  #ifdef SUPPORT_UTF
177      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
178      if (md->utf)
179        {
180        /* Match characters up to the end of the reference. NOTE: the number of
181        bytes matched may differ, because there are some characters whose upper and
182        lower case versions code as different numbers of bytes. For example, U+023A
183        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
184        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
185        the latter. It is important, therefore, to check the length along the
186        reference, not along the subject (earlier code did this wrong). */
187    
188        PCRE_PUCHAR endptr = p + length;
189        while (p < endptr)
190          {
191          int c, d;
192          if (eptr >= md->end_subject) return -1;
193          GETCHARINC(c, eptr);
194          GETCHARINC(d, p);
195          if (c != d && c != UCD_OTHERCASE(d)) return -1;
196          }
197        }
198      else
199    #endif
200    #endif
201    
202      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
203      is no UCP support. */
204        {
205        if (eptr + length > md->end_subject) return -1;
206        while (length-- > 0)
207          {
208          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
209          p++;
210          eptr++;
211          }
212        }
213    }    }
214    
215    /* In the caseful case, we can just compare the bytes, whether or not we
216    are in UTF-8 mode. */
217    
218  else  else
219    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
220      if (eptr + length > md->end_subject) return -1;
221      while (length-- > 0) if (*p++ != *eptr++) return -1;
222      }
223    
224  return TRUE;  return (int)(eptr - eptr_start);
225  }  }
226    
227    
# Line 183  calls by keeping local variables that ne Line 247  calls by keeping local variables that ne
247  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
248  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
249  always used to.  always used to.
250    
251    The original heap-recursive code used longjmp(). However, it seems that this
252    can be very slow on some operating systems. Following a suggestion from Stan
253    Switzer, the use of longjmp() has been abolished, at the cost of having to
254    provide a unique number for each call to RMATCH. There is no way of generating
255    a sequence of numbers at compile time in C. I have given them names, to make
256    them stand out more clearly.
257    
258    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
259    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
260    tests. Furthermore, not using longjmp() means that local dynamic variables
261    don't have indeterminate values; this has meant that the frame size can be
262    reduced because the result can be "passed back" by straight setting of the
263    variable instead of being passed in the frame.
264  ****************************************************************************  ****************************************************************************
265  ***************************************************************************/  ***************************************************************************/
266    
267    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
268    below must be updated in sync.  */
269    
270    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
271           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
272           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
273           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
274           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
275           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
276           RM61,  RM62, RM63, RM64, RM65, RM66 };
277    
278  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
279  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
280    actually used in this definition. */
281    
282  #ifndef NO_RECURSE  #ifndef NO_RECURSE
283  #define REGISTER register  #define REGISTER register
284  #ifdef DEBUG  
285  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
286    #define RMATCH(ra,rb,rc,rd,re,rw) \
287    { \    { \
288    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
289    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
290    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
291    }    }
292  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 205  versions and production versions. */ Line 295  versions and production versions. */
295    return ra; \    return ra; \
296    }    }
297  #else  #else
298  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
299    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
300  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
301  #endif  #endif
302    
303  #else  #else
304    
305    
306  /* 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
307  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
308  match(), which never changes. */  argument of match(), which never changes. */
309    
310  #define REGISTER  #define REGISTER
311    
312  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
313    {\    {\
314    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
315    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
316      {\    frame->Xwhere = rw; \
317      newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
318      newframe->Xecode = rb;\    newframe->Xecode = rb;\
319      newframe->Xoffset_top = rc;\    newframe->Xmstart = mstart;\
320      newframe->Xims = re;\    newframe->Xoffset_top = rc;\
321      newframe->Xeptrb = rf;\    newframe->Xeptrb = re;\
322      newframe->Xflags = rg;\    newframe->Xrdepth = frame->Xrdepth + 1;\
323      newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xprevframe = frame;\
324      newframe->Xprevframe = frame;\    frame = newframe;\
325      frame = newframe;\    DPRINTF(("restarting from line %d\n", __LINE__));\
326      DPRINTF(("restarting from line %d\n", __LINE__));\    goto HEAP_RECURSE;\
327      goto HEAP_RECURSE;\    L_##rw:\
328      }\    DPRINTF(("jumped back to line %d\n", __LINE__));\
   else\  
     {\  
     DPRINTF(("longjumped back to line %d\n", __LINE__));\  
     frame = md->thisframe;\  
     rx = frame->Xresult;\  
     }\  
329    }    }
330    
331  #define RRETURN(ra)\  #define RRETURN(ra)\
332    {\    {\
333    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
334    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
335    (pcre_stack_free)(newframe);\    if (oldframe != &frame_zero) (PUBL(stack_free))(oldframe);\
336    if (frame != NULL)\    if (frame != NULL)\
337      {\      {\
338      frame->Xresult = ra;\      rrc = ra;\
339      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
340      }\      }\
341    return ra;\    return ra;\
342    }    }
# Line 266  typedef struct heapframe { Line 349  typedef struct heapframe {
349    
350    /* Function arguments that may change */    /* Function arguments that may change */
351    
352    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
353    const uschar *Xecode;    const pcre_uchar *Xecode;
354      PCRE_PUCHAR Xmstart;
355    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
356    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
357    unsigned int Xrdepth;    unsigned int Xrdepth;
358    
359    /* Function local variables */    /* Function local variables */
360    
361    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
362    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
363    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
364    const uschar *Xnext;  #endif
365    const uschar *Xpp;    PCRE_PUCHAR Xdata;
366    const uschar *Xprev;    PCRE_PUCHAR Xnext;
367    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
368      PCRE_PUCHAR Xprev;
369      PCRE_PUCHAR Xsaved_eptr;
370    
371    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
372    
# Line 290  typedef struct heapframe { Line 374  typedef struct heapframe {
374    BOOL Xcondition;    BOOL Xcondition;
375    BOOL Xprev_is_word;    BOOL Xprev_is_word;
376    
   unsigned long int Xoriginal_ims;  
   
377  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
378    int Xprop_type;    int Xprop_type;
379    int Xprop_value;    int Xprop_value;
380    int Xprop_fail_result;    int Xprop_fail_result;
381    int Xprop_category;    int Xoclength;
382    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_script;  
383  #endif  #endif
384    
385      int Xcodelink;
386    int Xctype;    int Xctype;
387    unsigned int Xfc;    unsigned int Xfc;
388    int Xfi;    int Xfi;
# Line 316  typedef struct heapframe { Line 398  typedef struct heapframe {
398    
399    eptrblock Xnewptrb;    eptrblock Xnewptrb;
400    
401    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
402    
403    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
404    
405  } heapframe;  } heapframe;
406    
# Line 337  typedef struct heapframe { Line 418  typedef struct heapframe {
418    
419  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
420  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
421  same response.  same response. */
422    
423  Performance note: It might be tempting to extract commonly used fields from the  /* These macros pack up tests that are used for partial matching, and which
424  md structure (e.g. utf8, end_subject) into individual variables to improve  appear several times in the code. We set the "hit end" flag if the pointer is
425    at the end of the subject and also past the start of the subject (i.e.
426    something has been matched). For hard partial matching, we then return
427    immediately. The second one is used when we already know we are past the end of
428    the subject. */
429    
430    #define CHECK_PARTIAL()\
431      if (md->partial != 0 && eptr >= md->end_subject && \
432          eptr > md->start_used_ptr) \
433        { \
434        md->hitend = TRUE; \
435        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
436        }
437    
438    #define SCHECK_PARTIAL()\
439      if (md->partial != 0 && eptr > md->start_used_ptr) \
440        { \
441        md->hitend = TRUE; \
442        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
443        }
444    
445    
446    /* Performance note: It might be tempting to extract commonly used fields from
447    the md structure (e.g. utf, end_subject) into individual variables to improve
448  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
449  made performance worse.  made performance worse.
450    
451  Arguments:  Arguments:
452     eptr        pointer to current character in subject     eptr        pointer to current character in subject
453     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
454       mstart      pointer to the current match start position (can be modified
455                     by encountering \K)
456     offset_top  current top pointer     offset_top  current top pointer
457     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
458     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
459                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
                  match_tail_recursed - this is a tail_recursed group  
460     rdepth      the recursion depth     rdepth      the recursion depth
461    
462  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
463                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
464                   a negative MATCH_xxx value for PRUNE, SKIP, etc
465                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
466                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
467  */  */
468    
469  static int  static int
470  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
471    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
472    int flags, unsigned int rdepth)    unsigned int rdepth)
473  {  {
474  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
475  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 377  so they can be ordinary variables in all Line 478  so they can be ordinary variables in all
478  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
479  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
480  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
481  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
482    
483  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
484    BOOL caseless;
485    int condcode;
486    
487  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
488  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
489  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
490  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
491    the top-level on the stack rather than malloc-ing them all gives a performance
492    boost in many cases where there is not much "recursion". */
493    
494  #ifdef NO_RECURSE  #ifdef NO_RECURSE
495  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe frame_zero;
496    heapframe *frame = &frame_zero;
497  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
498    
499  /* Copy in the original argument variables */  /* Copy in the original argument variables */
500    
501  frame->Xeptr = eptr;  frame->Xeptr = eptr;
502  frame->Xecode = ecode;  frame->Xecode = ecode;
503    frame->Xmstart = mstart;
504  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
505  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
506  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
507    
508  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 408  HEAP_RECURSE: Line 513  HEAP_RECURSE:
513    
514  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
515  #define ecode              frame->Xecode  #define ecode              frame->Xecode
516    #define mstart             frame->Xmstart
517  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
518  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
519  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
520    
521  /* Ditto for the local variables */  /* Ditto for the local variables */
522    
523  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
524  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
525  #endif  #endif
526  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
527    #define codelink           frame->Xcodelink
528  #define data               frame->Xdata  #define data               frame->Xdata
529  #define next               frame->Xnext  #define next               frame->Xnext
530  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 432  HEAP_RECURSE: Line 537  HEAP_RECURSE:
537  #define condition          frame->Xcondition  #define condition          frame->Xcondition
538  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
539    
 #define original_ims       frame->Xoriginal_ims  
   
540  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
541  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
542  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
543  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
544  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
545  #define prop_chartype      frame->Xprop_chartype  #define occhars            frame->Xocchars
 #define prop_script        frame->Xprop_script  
546  #endif  #endif
547    
548  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 468  i, and fc and c, can be the same variabl Line 570  i, and fc and c, can be the same variabl
570  #define fi i  #define fi i
571  #define fc c  #define fc c
572    
573    /* Many of the following variables are used only in small blocks of the code.
574    My normal style of coding would have declared them within each of those blocks.
575    However, in order to accommodate the version of this code that uses an external
576    "stack" implemented on the heap, it is easier to declare them all here, so the
577    declarations can be cut out in a block. The only declarations within blocks
578    below are for variables that do not have to be preserved over a recursive call
579    to RMATCH(). */
580    
581    #ifdef SUPPORT_UTF
582    const pcre_uchar *charptr;
583    #endif
584    const pcre_uchar *callpat;
585    const pcre_uchar *data;
586    const pcre_uchar *next;
587    PCRE_PUCHAR       pp;
588    const pcre_uchar *prev;
589    PCRE_PUCHAR       saved_eptr;
590    
591    recursion_info new_recursive;
592    
593  #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().          */  
594  BOOL condition;  BOOL condition;
595  BOOL prev_is_word;  BOOL prev_is_word;
596    
 unsigned long int original_ims;  
   
597  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
598  int prop_type;  int prop_type;
599  int prop_value;  int prop_value;
600  int prop_fail_result;  int prop_fail_result;
601  int prop_category;  int oclength;
602  int prop_chartype;  pcre_uchar occhars[6];
 int prop_script;  
603  #endif  #endif
604    
605    int codelink;
606  int ctype;  int ctype;
607  int length;  int length;
608  int max;  int max;
# Line 508  int save_offset1, save_offset2, save_off Line 615  int save_offset1, save_offset2, save_off
615  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
616    
617  eptrblock newptrb;  eptrblock newptrb;
618    
619    /* There is a special fudge for calling match() in a way that causes it to
620    measure the size of its basic stack frame when the stack is being used for
621    recursion. The second argument (ecode) being NULL triggers this behaviour. It
622    cannot normally ever be NULL. The return is the negated value of the frame
623    size. */
624    
625    if (ecode == NULL)
626      {
627      if (rdepth == 0)
628        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
629      else
630        {
631        int len = (char *)&rdepth - (char *)eptr;
632        return (len > 0)? -len : len;
633        }
634      }
635  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
636    
637    /* To save space on the stack and in the heap frame, I have doubled up on some
638    of the local variables that are used only in localised parts of the code, but
639    still need to be preserved over recursive calls of match(). These macros define
640    the alternative names that are used. */
641    
642    #define allow_zero    cur_is_word
643    #define cbegroup      condition
644    #define code_offset   codelink
645    #define condassert    condition
646    #define matched_once  prev_is_word
647    #define foc           number
648    #define save_mark     data
649    
650  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
651  variables. */  variables. */
652    
# Line 529  TAIL_RECURSE: Line 666  TAIL_RECURSE:
666  /* 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
667  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
668  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()
669  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
670  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
671  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,
672  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
673    
674    #ifdef SUPPORT_UTF
675    utf = md->utf;       /* Local copy of the flag */
676    #else
677    utf = FALSE;
678    #endif
679    
680  /* 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
681  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
682    
683  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
684  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
685    
 original_ims = ims;    /* Save for resetting on ')' */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
686  /* 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
687  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
688  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
689  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
690  When match() is called in other circumstances, don't add to the chain. If this  
691  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
692  already used. */  such remembered pointers, to be checked when we hit the closing ket, in order
693    to break infinite loops that match no characters. When match() is called in
694    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
695    NOT be used with tail recursion, because the memory block that is used is on
696    the stack, so a new one may be required for each match(). */
697    
698  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
699    {    {
700    eptrblock *p;    newptrb.epb_saved_eptr = eptr;
701    if ((flags & match_tail_recursed) != 0)    newptrb.epb_prev = eptrb;
702      {    eptrb = &newptrb;
703      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;  
704    }    }
705    
706  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 577  for (;;) Line 710  for (;;)
710    minimize = possessive = FALSE;    minimize = possessive = FALSE;
711    op = *ecode;    op = *ecode;
712    
   /* 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;  
   
713    switch(op)    switch(op)
714      {      {
715      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
716      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
717      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
718      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,
719      reference inside the group.        eptrb, RM55);
720        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
721      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
722      values of the final offsets, in case they were set by a previous iteration  
723      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
724        argument, and we must check whether that argument matches this MARK's
725        argument. It is passed back in md->start_match_ptr (an overloading of that
726        variable). If it does match, we reset that variable to the current subject
727        position and return MATCH_SKIP. Otherwise, pass back the return code
728        unaltered. */
729    
730        else if (rrc == MATCH_SKIP_ARG &&
731            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
732          {
733          md->start_match_ptr = eptr;
734          RRETURN(MATCH_SKIP);
735          }
736        RRETURN(rrc);
737    
738        case OP_FAIL:
739        RRETURN(MATCH_NOMATCH);
740    
741        /* COMMIT overrides PRUNE, SKIP, and THEN */
742    
743        case OP_COMMIT:
744        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
745          eptrb, RM52);
746        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
747            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
748            rrc != MATCH_THEN)
749          RRETURN(rrc);
750        RRETURN(MATCH_COMMIT);
751    
752        /* PRUNE overrides THEN */
753    
754        case OP_PRUNE:
755        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
756          eptrb, RM51);
757        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
758        RRETURN(MATCH_PRUNE);
759    
760        case OP_PRUNE_ARG:
761        md->nomatch_mark = ecode + 2;
762        md->mark = NULL;    /* In case previously set by assertion */
763        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
764          eptrb, RM56);
765        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
766             md->mark == NULL) md->mark = ecode + 2;
767        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
768        RRETURN(MATCH_PRUNE);
769    
770        /* SKIP overrides PRUNE and THEN */
771    
772        case OP_SKIP:
773        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
774          eptrb, RM53);
775        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
776          RRETURN(rrc);
777        md->start_match_ptr = eptr;   /* Pass back current position */
778        RRETURN(MATCH_SKIP);
779    
780        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
781        nomatch_mark. There is a flag that disables this opcode when re-matching a
782        pattern that ended with a SKIP for which there was not a matching MARK. */
783    
784        case OP_SKIP_ARG:
785        if (md->ignore_skip_arg)
786          {
787          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
788          break;
789          }
790        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
791          eptrb, RM57);
792        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
793          RRETURN(rrc);
794    
795        /* Pass back the current skip name by overloading md->start_match_ptr and
796        returning the special MATCH_SKIP_ARG return code. This will either be
797        caught by a matching MARK, or get to the top, where it causes a rematch
798        with the md->ignore_skip_arg flag set. */
799    
800        md->start_match_ptr = ecode + 2;
801        RRETURN(MATCH_SKIP_ARG);
802    
803        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
804        the branch in which it occurs can be determined. Overload the start of
805        match pointer to do this. */
806    
807        case OP_THEN:
808        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
809          eptrb, RM54);
810        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
811        md->start_match_ptr = ecode;
812        RRETURN(MATCH_THEN);
813    
814        case OP_THEN_ARG:
815        md->nomatch_mark = ecode + 2;
816        md->mark = NULL;    /* In case previously set by assertion */
817        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
818          md, eptrb, RM58);
819        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
820             md->mark == NULL) md->mark = ecode + 2;
821        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
822        md->start_match_ptr = ecode;
823        RRETURN(MATCH_THEN);
824    
825        /* Handle an atomic group that does not contain any capturing parentheses.
826        This can be handled like an assertion. Prior to 8.13, all atomic groups
827        were handled this way. In 8.13, the code was changed as below for ONCE, so
828        that backups pass through the group and thereby reset captured values.
829        However, this uses a lot more stack, so in 8.20, atomic groups that do not
830        contain any captures generate OP_ONCE_NC, which can be handled in the old,
831        less stack intensive way.
832    
833        Check the alternative branches in turn - the matching won't pass the KET
834        for this kind of subpattern. If any one branch matches, we carry on as at
835        the end of a normal bracket, leaving the subject pointer, but resetting
836        the start-of-match value in case it was changed by \K. */
837    
838        case OP_ONCE_NC:
839        prev = ecode;
840        saved_eptr = eptr;
841        save_mark = md->mark;
842        do
843          {
844          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
845          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
846            {
847            mstart = md->start_match_ptr;
848            break;
849            }
850          if (rrc == MATCH_THEN)
851            {
852            next = ecode + GET(ecode,1);
853            if (md->start_match_ptr < next &&
854                (*ecode == OP_ALT || *next == OP_ALT))
855              rrc = MATCH_NOMATCH;
856            }
857    
858          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859          ecode += GET(ecode,1);
860          md->mark = save_mark;
861          }
862        while (*ecode == OP_ALT);
863    
864        /* If hit the end of the group (which could be repeated), fail */
865    
866        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
867    
868        /* Continue as from after the group, updating the offsets high water
869        mark, since extracts may have been taken. */
870    
871        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
872    
873        offset_top = md->end_offset_top;
874        eptr = md->end_match_ptr;
875    
876        /* For a non-repeating ket, just continue at this level. This also
877        happens for a repeating ket if no characters were matched in the group.
878        This is the forcible breaking of infinite loops as implemented in Perl
879        5.005. */
880    
881        if (*ecode == OP_KET || eptr == saved_eptr)
882          {
883          ecode += 1+LINK_SIZE;
884          break;
885          }
886    
887        /* The repeating kets try the rest of the pattern or restart from the
888        preceding bracket, in the appropriate order. The second "call" of match()
889        uses tail recursion, to avoid using another stack frame. */
890    
891        if (*ecode == OP_KETRMIN)
892          {
893          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
894          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
895          ecode = prev;
896          goto TAIL_RECURSE;
897          }
898        else  /* OP_KETRMAX */
899          {
900          md->match_function_type = MATCH_CBEGROUP;
901          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
902          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
903          ecode += 1 + LINK_SIZE;
904          goto TAIL_RECURSE;
905          }
906        /* Control never gets here */
907    
908        /* Handle a capturing bracket, other than those that are possessive with an
909        unlimited repeat. If there is space in the offset vector, save the current
910        subject position in the working slot at the top of the vector. We mustn't
911        change the current values of the data slot, because they may be set from a
912        previous iteration of this group, and be referred to by a reference inside
913        the group. A failure to match might occur after the group has succeeded,
914        if something later on doesn't match. For this reason, we need to restore
915        the working value and also the values of the final offsets, in case they
916        were set by a previous iteration of the same bracket.
917    
918      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
919      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 606  for (;;) Line 924  for (;;)
924      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
925      offset = number << 1;      offset = number << 1;
926    
927  #ifdef DEBUG  #ifdef PCRE_DEBUG
928      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
929      printf("subject=");      printf("subject=");
930      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 619  for (;;) Line 937  for (;;)
937        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
938        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
939        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
940          save_mark = md->mark;
941    
942        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
943        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
944            (int)(eptr - md->start_subject);
945    
946        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
947          {          {
948          RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
949            ims, eptrb, flags);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
950              eptrb, RM1);
951            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
952    
953            /* If we backed up to a THEN, check whether it is within the current
954            branch by comparing the address of the THEN that is passed back with
955            the end of the branch. If it is within the current branch, and the
956            branch is one of two or more alternatives (it either starts or ends
957            with OP_ALT), we have reached the limit of THEN's action, so convert
958            the return code to NOMATCH, which will cause normal backtracking to
959            happen from now on. Otherwise, THEN is passed back to an outer
960            alternative. This implements Perl's treatment of parenthesized groups,
961            where a group not containing | does not affect the current alternative,
962            that is, (X) is NOT the same as (X|(*F)). */
963    
964            if (rrc == MATCH_THEN)
965              {
966              next = ecode + GET(ecode,1);
967              if (md->start_match_ptr < next &&
968                  (*ecode == OP_ALT || *next == OP_ALT))
969                rrc = MATCH_NOMATCH;
970              }
971    
972            /* Anything other than NOMATCH is passed back. */
973    
974          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
975          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
976          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
977            md->mark = save_mark;
978            if (*ecode != OP_ALT) break;
979          }          }
       while (*ecode == OP_ALT);  
980    
981        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
982        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
983        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
984        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
985    
986        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
987    
988          RRETURN(rrc);
989        }        }
990    
991      /* Insufficient room for saving captured contents. Treat as a non-capturing      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
992      bracket. */      as a non-capturing bracket. */
993    
994        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
995        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
996    
997      DPRINTF(("insufficient capture room: treat as non-capturing\n"));      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
998    
999      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1000      final alternative within the brackets, we would return the result of a      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1001      recursive call to match() whatever happened. We can reduce stack usage by  
1002      turning this into a tail recursion. */      /* Non-capturing or atomic group, except for possessive with unlimited
1003        repeat and ONCE group with no captures. Loop for all the alternatives.
1004    
1005        When we get to the final alternative within the brackets, we used to return
1006        the result of a recursive call to match() whatever happened so it was
1007        possible to reduce stack usage by turning this into a tail recursion,
1008        except in the case of a possibly empty group. However, now that there is
1009        the possiblity of (*THEN) occurring in the final alternative, this
1010        optimization is no longer always possible.
1011    
1012        We can optimize if we know there are no (*THEN)s in the pattern; at present
1013        this is the best that can be done.
1014    
1015        MATCH_ONCE is returned when the end of an atomic group is successfully
1016        reached, but subsequent matching fails. It passes back up the tree (causing
1017        captured values to be reset) until the original atomic group level is
1018        reached. This is tested by comparing md->once_target with the start of the
1019        group. At this point, the return is converted into MATCH_NOMATCH so that
1020        previous backup points can be taken. */
1021    
1022        case OP_ONCE:
1023      case OP_BRA:      case OP_BRA:
1024      case OP_SBRA:      case OP_SBRA:
1025      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1026      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1027      for (;;)      for (;;)
1028        {        {
1029        if (ecode[GET(ecode, 1)] != OP_ALT)        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1030    
1031          /* If this is not a possibly empty group, and there are no (*THEN)s in
1032          the pattern, and this is the final alternative, optimize as described
1033          above. */
1034    
1035          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1036          {          {
1037          ecode += _pcre_OP_lengths[*ecode];          ecode += PRIV(OP_lengths)[*ecode];
         flags |= match_tail_recursed;  
         DPRINTF(("bracket 0 tail recursion\n"));  
1038          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1039          }          }
1040    
1041        /* 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. */  
1042    
1043        RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1044          eptrb, flags);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1045        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RM2);
1046    
1047          /* See comment in the code for capturing groups above about handling
1048          THEN. */
1049    
1050          if (rrc == MATCH_THEN)
1051            {
1052            next = ecode + GET(ecode,1);
1053            if (md->start_match_ptr < next &&
1054                (*ecode == OP_ALT || *next == OP_ALT))
1055              rrc = MATCH_NOMATCH;
1056            }
1057    
1058          if (rrc != MATCH_NOMATCH)
1059            {
1060            if (rrc == MATCH_ONCE)
1061              {
1062              const pcre_uchar *scode = ecode;
1063              if (*scode != OP_ONCE)           /* If not at start, find it */
1064                {
1065                while (*scode == OP_ALT) scode += GET(scode, 1);
1066                scode -= GET(scode, 1);
1067                }
1068              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1069              }
1070            RRETURN(rrc);
1071            }
1072        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1073          md->mark = save_mark;
1074          if (*ecode != OP_ALT) break;
1075        }        }
     /* Control never reaches here. */  
1076    
1077      /* 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. */  
1078    
1079      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1080      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1081      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1082        {      different. The end of these brackets will always be OP_KETRPOS, which
1083        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1084        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1085          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1086    
1087      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      case OP_CBRAPOS:
1088        {      case OP_SCBRAPOS:
1089        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);  
       }  
1090    
1091      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      POSSESSIVE_CAPTURE:
1092        {      number = GET2(ecode, 1+LINK_SIZE);
1093        condition = FALSE;      offset = number << 1;
       ecode += GET(ecode, 1);  
       }  
1094    
1095      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1096      the final argument match_condassert causes it to stop at the end of an      printf("start possessive bracket %d\n", number);
1097      assertion. */      printf("subject=");
1098        pchars(eptr, 16, TRUE, md);
1099        printf("\n");
1100    #endif
1101    
1102      else      if (offset < md->offset_max)
1103        {        {
1104        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1105            match_condassert);        code_offset = (int)(ecode - md->start_code);
1106        if (rrc == MATCH_MATCH)  
1107          {        save_offset1 = md->offset_vector[offset];
1108          condition = TRUE;        save_offset2 = md->offset_vector[offset+1];
1109          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        save_offset3 = md->offset_vector[md->offset_end - number];
1110          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
1111    
1112          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1113    
1114          /* Each time round the loop, save the current subject position for use
1115          when the group matches. For MATCH_MATCH, the group has matched, so we
1116          restart it with a new subject starting position, remembering that we had
1117          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1118          usual. If we haven't matched any alternatives in any iteration, check to
1119          see if a previous iteration matched. If so, the group has matched;
1120          continue from afterwards. Otherwise it has failed; restore the previous
1121          capture values before returning NOMATCH. */
1122    
1123          for (;;)
1124            {
1125            md->offset_vector[md->offset_end - number] =
1126              (int)(eptr - md->start_subject);
1127            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1128            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1129              eptrb, RM63);
1130            if (rrc == MATCH_KETRPOS)
1131              {
1132              offset_top = md->end_offset_top;
1133              eptr = md->end_match_ptr;
1134              ecode = md->start_code + code_offset;
1135              save_capture_last = md->capture_last;
1136              matched_once = TRUE;
1137              continue;
1138              }
1139    
1140            /* See comment in the code for capturing groups above about handling
1141            THEN. */
1142    
1143            if (rrc == MATCH_THEN)
1144              {
1145              next = ecode + GET(ecode,1);
1146              if (md->start_match_ptr < next &&
1147                  (*ecode == OP_ALT || *next == OP_ALT))
1148                rrc = MATCH_NOMATCH;
1149              }
1150    
1151            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1152            md->capture_last = save_capture_last;
1153            ecode += GET(ecode, 1);
1154            if (*ecode != OP_ALT) break;
1155          }          }
1156        else if (rrc != MATCH_NOMATCH)  
1157          if (!matched_once)
1158          {          {
1159          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1160            md->offset_vector[offset+1] = save_offset2;
1161            md->offset_vector[md->offset_end - number] = save_offset3;
1162          }          }
1163        else  
1164          if (allow_zero || matched_once)
1165          {          {
1166          condition = FALSE;          ecode += 1 + LINK_SIZE;
1167          ecode += GET(ecode, 1);          break;
1168          }          }
       }  
1169    
1170      /* We are now at the branch that is to be obeyed. As there is only one,        RRETURN(MATCH_NOMATCH);
     we can use tail recursion to avoid using another stack frame. If the second  
     alternative doesn't exist, we can just plough on. */  
   
     if (condition || *ecode == OP_ALT)  
       {  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed | ((op == OP_SCOND)? match_cbegroup : 0);  
       goto TAIL_RECURSE;  
       }  
     else  
       {  
       ecode += 1 + LINK_SIZE;  
1171        }        }
     break;  
   
1172    
1173      /* End of the pattern. If we are in a top-level recursion, we should      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1174      restore the offsets appropriately and continue from after the call. */      as a non-capturing bracket. */
1175    
1176      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1177      if (md->recursive != NULL && md->recursive->group_num == 0)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       md->start_match = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1178    
1179      /* 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. */  
1180    
1181      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1182      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;  
1183    
1184      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1185      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,
1186      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1187      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1188      this level is identical to the lookahead case. */  
1189        case OP_BRAPOS:
1190        case OP_SBRAPOS:
1191        allow_zero = FALSE;
1192    
1193        POSSESSIVE_NON_CAPTURE:
1194        matched_once = FALSE;
1195        code_offset = (int)(ecode - md->start_code);
1196    
1197      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1198        {        {
1199        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1200        if (rrc == MATCH_MATCH) break;        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1201            eptrb, RM48);
1202          if (rrc == MATCH_KETRPOS)
1203            {
1204            offset_top = md->end_offset_top;
1205            eptr = md->end_match_ptr;
1206            ecode = md->start_code + code_offset;
1207            matched_once = TRUE;
1208            continue;
1209            }
1210    
1211          /* See comment in the code for capturing groups above about handling
1212          THEN. */
1213    
1214          if (rrc == MATCH_THEN)
1215            {
1216            next = ecode + GET(ecode,1);
1217            if (md->start_match_ptr < next &&
1218                (*ecode == OP_ALT || *next == OP_ALT))
1219              rrc = MATCH_NOMATCH;
1220            }
1221    
1222        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1223        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1224          if (*ecode != OP_ALT) break;
1225        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1226    
1227      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1228          {
1229          ecode += 1 + LINK_SIZE;
1230          break;
1231          }
1232        RRETURN(MATCH_NOMATCH);
1233    
1234      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1235    
1236      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1237      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1238        past the end if there is only one branch, but that's OK because that is
1239        exactly what going to the ket would do. */
1240    
1241      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1242      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1243      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1244    
1245      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1246        inserted between OP_COND and an assertion condition. */
1247    
1248      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1249        {        {
1250        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (PUBL(callout) != NULL)
1251        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          {
1252        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          PUBL(callout_block) cb;
1253        ecode += GET(ecode,1);          cb.version          = 2;   /* Version 1 of the callout block */
1254            cb.callout_number   = ecode[LINK_SIZE+2];
1255            cb.offset_vector    = md->offset_vector;
1256    #ifdef COMPILE_PCRE8
1257            cb.subject          = (PCRE_SPTR)md->start_subject;
1258    #else
1259            cb.subject          = (PCRE_SPTR16)md->start_subject;
1260    #endif
1261            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1262            cb.start_match      = (int)(mstart - md->start_subject);
1263            cb.current_position = (int)(eptr - md->start_subject);
1264            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1265            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1266            cb.capture_top      = offset_top/2;
1267            cb.capture_last     = md->capture_last;
1268            cb.callout_data     = md->callout_data;
1269            cb.mark             = md->nomatch_mark;
1270            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1271            if (rrc < 0) RRETURN(rrc);
1272            }
1273          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1274        }        }
     while (*ecode == OP_ALT);  
1275    
1276      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      condcode = ecode[LINK_SIZE+1];
1277    
1278      ecode += 1 + LINK_SIZE;      /* Now see what the actual condition is */
1279    
1280        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1281          {
1282          if (md->recursive == NULL)                /* Not recursing => FALSE */
1283            {
1284            condition = FALSE;
1285            ecode += GET(ecode, 1);
1286            }
1287          else
1288            {
1289            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1290            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1291    
1292            /* If the test is for recursion into a specific subpattern, and it is
1293            false, but the test was set up by name, scan the table to see if the
1294            name refers to any other numbers, and test them. The condition is true
1295            if any one is set. */
1296    
1297            if (!condition && condcode == OP_NRREF)
1298              {
1299              pcre_uchar *slotA = md->name_table;
1300              for (i = 0; i < md->name_count; i++)
1301                {
1302                if (GET2(slotA, 0) == recno) break;
1303                slotA += md->name_entry_size;
1304                }
1305    
1306              /* Found a name for the number - there can be only one; duplicate
1307              names for different numbers are allowed, but not vice versa. First
1308              scan down for duplicates. */
1309    
1310              if (i < md->name_count)
1311                {
1312                pcre_uchar *slotB = slotA;
1313                while (slotB > md->name_table)
1314                  {
1315                  slotB -= md->name_entry_size;
1316                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1317                    {
1318                    condition = GET2(slotB, 0) == md->recursive->group_num;
1319                    if (condition) break;
1320                    }
1321                  else break;
1322                  }
1323    
1324                /* Scan up for duplicates */
1325    
1326                if (!condition)
1327                  {
1328                  slotB = slotA;
1329                  for (i++; i < md->name_count; i++)
1330                    {
1331                    slotB += md->name_entry_size;
1332                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1333                      {
1334                      condition = GET2(slotB, 0) == md->recursive->group_num;
1335                      if (condition) break;
1336                      }
1337                    else break;
1338                    }
1339                  }
1340                }
1341              }
1342    
1343            /* Chose branch according to the condition */
1344    
1345            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1346            }
1347          }
1348    
1349        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1350          {
1351          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1352          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1353    
1354          /* If the numbered capture is unset, but the reference was by name,
1355          scan the table to see if the name refers to any other numbers, and test
1356          them. The condition is true if any one is set. This is tediously similar
1357          to the code above, but not close enough to try to amalgamate. */
1358    
1359          if (!condition && condcode == OP_NCREF)
1360            {
1361            int refno = offset >> 1;
1362            pcre_uchar *slotA = md->name_table;
1363    
1364            for (i = 0; i < md->name_count; i++)
1365              {
1366              if (GET2(slotA, 0) == refno) break;
1367              slotA += md->name_entry_size;
1368              }
1369    
1370            /* Found a name for the number - there can be only one; duplicate names
1371            for different numbers are allowed, but not vice versa. First scan down
1372            for duplicates. */
1373    
1374            if (i < md->name_count)
1375              {
1376              pcre_uchar *slotB = slotA;
1377              while (slotB > md->name_table)
1378                {
1379                slotB -= md->name_entry_size;
1380                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1381                  {
1382                  offset = GET2(slotB, 0) << 1;
1383                  condition = offset < offset_top &&
1384                    md->offset_vector[offset] >= 0;
1385                  if (condition) break;
1386                  }
1387                else break;
1388                }
1389    
1390              /* Scan up for duplicates */
1391    
1392              if (!condition)
1393                {
1394                slotB = slotA;
1395                for (i++; i < md->name_count; i++)
1396                  {
1397                  slotB += md->name_entry_size;
1398                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1399                    {
1400                    offset = GET2(slotB, 0) << 1;
1401                    condition = offset < offset_top &&
1402                      md->offset_vector[offset] >= 0;
1403                    if (condition) break;
1404                    }
1405                  else break;
1406                  }
1407                }
1408              }
1409            }
1410    
1411          /* Chose branch according to the condition */
1412    
1413          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1414          }
1415    
1416        else if (condcode == OP_DEF)     /* DEFINE - always false */
1417          {
1418          condition = FALSE;
1419          ecode += GET(ecode, 1);
1420          }
1421    
1422        /* The condition is an assertion. Call match() to evaluate it - setting
1423        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1424        an assertion. */
1425    
1426        else
1427          {
1428          md->match_function_type = MATCH_CONDASSERT;
1429          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1430          if (rrc == MATCH_MATCH)
1431            {
1432            if (md->end_offset_top > offset_top)
1433              offset_top = md->end_offset_top;  /* Captures may have happened */
1434            condition = TRUE;
1435            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1436            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1437            }
1438    
1439          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1440          assertion; it is therefore treated as NOMATCH. */
1441    
1442          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1443            {
1444            RRETURN(rrc);         /* Need braces because of following else */
1445            }
1446          else
1447            {
1448            condition = FALSE;
1449            ecode += codelink;
1450            }
1451          }
1452    
1453        /* We are now at the branch that is to be obeyed. As there is only one, can
1454        use tail recursion to avoid using another stack frame, except when there is
1455        unlimited repeat of a possibly empty group. In the latter case, a recursive
1456        call to match() is always required, unless the second alternative doesn't
1457        exist, in which case we can just plough on. Note that, for compatibility
1458        with Perl, the | in a conditional group is NOT treated as creating two
1459        alternatives. If a THEN is encountered in the branch, it propagates out to
1460        the enclosing alternative (unless nested in a deeper set of alternatives,
1461        of course). */
1462    
1463        if (condition || *ecode == OP_ALT)
1464          {
1465          if (op != OP_SCOND)
1466            {
1467            ecode += 1 + LINK_SIZE;
1468            goto TAIL_RECURSE;
1469            }
1470    
1471          md->match_function_type = MATCH_CBEGROUP;
1472          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1473          RRETURN(rrc);
1474          }
1475    
1476         /* Condition false & no alternative; continue after the group. */
1477    
1478        else
1479          {
1480          ecode += 1 + LINK_SIZE;
1481          }
1482        break;
1483    
1484    
1485        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1486        to close any currently open capturing brackets. */
1487    
1488        case OP_CLOSE:
1489        number = GET2(ecode, 1);
1490        offset = number << 1;
1491    
1492    #ifdef PCRE_DEBUG
1493          printf("end bracket %d at *ACCEPT", number);
1494          printf("\n");
1495    #endif
1496    
1497        md->capture_last = number;
1498        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1499          {
1500          md->offset_vector[offset] =
1501            md->offset_vector[md->offset_end - number];
1502          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1503          if (offset_top <= offset) offset_top = offset + 2;
1504          }
1505        ecode += 1 + IMM2_SIZE;
1506        break;
1507    
1508    
1509        /* End of the pattern, either real or forced. */
1510    
1511        case OP_END:
1512        case OP_ACCEPT:
1513        case OP_ASSERT_ACCEPT:
1514    
1515        /* If we have matched an empty string, fail if not in an assertion and not
1516        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1517        is set and we have matched at the start of the subject. In both cases,
1518        backtracking will then try other alternatives, if any. */
1519    
1520        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1521             md->recursive == NULL &&
1522             (md->notempty ||
1523               (md->notempty_atstart &&
1524                 mstart == md->start_subject + md->start_offset)))
1525          RRETURN(MATCH_NOMATCH);
1526    
1527        /* Otherwise, we have a match. */
1528    
1529        md->end_match_ptr = eptr;           /* Record where we ended */
1530        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1531        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1532    
1533        /* For some reason, the macros don't work properly if an expression is
1534        given as the argument to RRETURN when the heap is in use. */
1535    
1536        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1537        RRETURN(rrc);
1538    
1539        /* Assertion brackets. Check the alternative branches in turn - the
1540        matching won't pass the KET for an assertion. If any one branch matches,
1541        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1542        start of each branch to move the current point backwards, so the code at
1543        this level is identical to the lookahead case. When the assertion is part
1544        of a condition, we want to return immediately afterwards. The caller of
1545        this incarnation of the match() function will have set MATCH_CONDASSERT in
1546        md->match_function type, and one of these opcodes will be the first opcode
1547        that is processed. We use a local variable that is preserved over calls to
1548        match() to remember this case. */
1549    
1550        case OP_ASSERT:
1551        case OP_ASSERTBACK:
1552        save_mark = md->mark;
1553        if (md->match_function_type == MATCH_CONDASSERT)
1554          {
1555          condassert = TRUE;
1556          md->match_function_type = 0;
1557          }
1558        else condassert = FALSE;
1559    
1560        do
1561          {
1562          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1563          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1564            {
1565            mstart = md->start_match_ptr;   /* In case \K reset it */
1566            break;
1567            }
1568    
1569          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1570          as NOMATCH. */
1571    
1572          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1573          ecode += GET(ecode, 1);
1574          md->mark = save_mark;
1575          }
1576        while (*ecode == OP_ALT);
1577    
1578        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1579    
1580        /* If checking an assertion for a condition, return MATCH_MATCH. */
1581    
1582        if (condassert) RRETURN(MATCH_MATCH);
1583    
1584        /* Continue from after the assertion, updating the offsets high water
1585        mark, since extracts may have been taken during the assertion. */
1586    
1587        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1588        ecode += 1 + LINK_SIZE;
1589        offset_top = md->end_offset_top;
1590        continue;
1591    
1592        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1593        PRUNE, or COMMIT means we must assume failure without checking subsequent
1594        branches. */
1595    
1596        case OP_ASSERT_NOT:
1597        case OP_ASSERTBACK_NOT:
1598        save_mark = md->mark;
1599        if (md->match_function_type == MATCH_CONDASSERT)
1600          {
1601          condassert = TRUE;
1602          md->match_function_type = 0;
1603          }
1604        else condassert = FALSE;
1605    
1606        do
1607          {
1608          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1609          md->mark = save_mark;
1610          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1611          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1612            {
1613            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1614            break;
1615            }
1616    
1617          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1618          as NOMATCH. */
1619    
1620          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1621          ecode += GET(ecode,1);
1622          }
1623        while (*ecode == OP_ALT);
1624    
1625        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1626    
1627        ecode += 1 + LINK_SIZE;
1628      continue;      continue;
1629    
1630      /* Move the subject pointer back. This occurs only at the start of      /* Move the subject pointer back. This occurs only at the start of
# Line 835  for (;;) Line 1633  for (;;)
1633      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1634    
1635      case OP_REVERSE:      case OP_REVERSE:
1636  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1637      if (utf8)      if (utf)
1638        {        {
1639        i = GET(ecode, 1);        i = GET(ecode, 1);
1640        while (i-- > 0)        while (i-- > 0)
1641          {          {
1642          eptr--;          eptr--;
1643          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1644          BACKCHAR(eptr)          BACKCHAR(eptr);
1645          }          }
1646        }        }
1647      else      else
# Line 856  for (;;) Line 1654  for (;;)
1654        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1655        }        }
1656    
1657      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1658    
1659        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1660      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1661      break;      break;
1662    
# Line 866  for (;;) Line 1665  for (;;)
1665      function is able to force a failure. */      function is able to force a failure. */
1666    
1667      case OP_CALLOUT:      case OP_CALLOUT:
1668      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1669        {        {
1670        pcre_callout_block cb;        PUBL(callout_block) cb;
1671        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1672        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1673        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1674    #ifdef COMPILE_PCRE8
1675        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1676        cb.subject_length   = md->end_subject - md->start_subject;  #else
1677        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1678        cb.current_position = eptr - md->start_subject;  #endif
1679          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1680          cb.start_match      = (int)(mstart - md->start_subject);
1681          cb.current_position = (int)(eptr - md->start_subject);
1682        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1683        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1684        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1685        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1686        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1687        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1688          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1689        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1690        }        }
1691      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 891  for (;;) Line 1695  for (;;)
1695      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
1696      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1697    
1698      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1699      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
1700      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
1701      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
1702      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
1703      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
1704      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.  
1705    
1706      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
1707      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
1708      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1709        a lot, so he is not to blame for the current way it works. */
1710    
1711      case OP_RECURSE:      case OP_RECURSE:
1712        {        {
1713          recursion_info *ri;
1714          int recno;
1715    
1716        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1717        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1718          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1719    
1720          /* Check for repeating a recursion without advancing the subject pointer.
1721          This should catch convoluted mutual recursions. (Some simple cases are
1722          caught at compile time.) */
1723    
1724          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1725            if (recno == ri->group_num && eptr == ri->subject_position)
1726              RRETURN(PCRE_ERROR_RECURSELOOP);
1727    
1728        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1729    
1730          new_recursive.group_num = recno;
1731          new_recursive.subject_position = eptr;
1732        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1733        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1734    
1735        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1736    
1737        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1738    
1739        /* Now save the offset data. */        /* Now save the offset data */
1740    
1741        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1742        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 930  for (;;) Line 1744  for (;;)
1744        else        else
1745          {          {
1746          new_recursive.offset_save =          new_recursive.offset_save =
1747            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1748          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1749          }          }
   
1750        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1751              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1752    
1753        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1754        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1755          might be changed, so reset it before looping. */
1756    
1757        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1758        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1759        do        do
1760          {          {
1761          RMATCH(rrc, eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1762            md, ims, eptrb, flags);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1763          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1764            memcpy(md->offset_vector, new_recursive.offset_save,
1765                new_recursive.saved_max * sizeof(int));
1766            md->recursive = new_recursive.prevrec;
1767            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1768            {            {
1769            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1770            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1771              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1772            RRETURN(MATCH_MATCH);  
1773              /* Set where we got to in the subject, and reset the start in case
1774              it was changed by \K. This *is* propagated back out of a recursion,
1775              for Perl compatibility. */
1776    
1777              eptr = md->end_match_ptr;
1778              mstart = md->start_match_ptr;
1779              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1780            }            }
1781          else if (rrc != MATCH_NOMATCH)  
1782            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1783            as NOMATCH. */
1784    
1785            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1786            {            {
1787            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1788              if (new_recursive.offset_save != stacksave)
1789                (PUBL(free))(new_recursive.offset_save);
1790            RRETURN(rrc);            RRETURN(rrc);
1791            }            }
1792    
1793          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1794          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1795          }          }
1796        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 972  for (;;) Line 1798  for (;;)
1798        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1799        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1800        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1801          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1802        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1803        }        }
     /* 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. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
1804    
1805      /* The repeating kets try the rest of the pattern or restart from the      RECURSION_MATCHED:
1806      preceding bracket, in the appropriate order. The second "call" of match()      break;
     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 */  
1807    
1808      /* 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
1809      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1059  for (;;) Line 1812  for (;;)
1812      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1813      break;      break;
1814    
1815      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1816      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
1817      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
1818      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
1819      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1820    
1821      case OP_BRAZERO:      case OP_BRAZERO:
1822        {      next = ecode + 1;
1823        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1824        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1825        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1826        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1827      break;      break;
1828    
1829      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1830        {      next = ecode + 1;
1831        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1832        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1833        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1834        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1835      break;      break;
1836    
1837        case OP_SKIPZERO:
1838        next = ecode+1;
1839        do next += GET(next,1); while (*next == OP_ALT);
1840        ecode = next + 1 + LINK_SIZE;
1841        break;
1842    
1843        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1844        here; just jump to the group, with allow_zero set TRUE. */
1845    
1846        case OP_BRAPOSZERO:
1847        op = *(++ecode);
1848        allow_zero = TRUE;
1849        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1850          goto POSSESSIVE_NON_CAPTURE;
1851    
1852      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1853    
1854      case OP_KET:      case OP_KET:
1855      case OP_KETRMIN:      case OP_KETRMIN:
1856      case OP_KETRMAX:      case OP_KETRMAX:
1857        case OP_KETRPOS:
1858      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1859    
1860      /* 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
1861      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1862      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1863    
1864      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1865        {        {
1866        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1867        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1868        }        }
1869      else saved_eptr = NULL;      else saved_eptr = NULL;
1870    
1871      /* 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
1872      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
1873      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1874        start in case it was changed by \K. */
1875      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1876          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1877          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1878        {        {
1879        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1880        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1881        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1882          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1883        }        }
1884    
1885      /* 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
1886      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1887      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
1888      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
1889      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
1890        the current subject position and start match pointer and give a MATCH
1891        return. */
1892    
1893      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1894            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1895        {        {
1896        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1897        offset = number << 1;        offset = number << 1;
1898    
1899  #ifdef DEBUG  #ifdef PCRE_DEBUG
1900        printf("end bracket %d", number);        printf("end bracket %d", number);
1901        printf("\n");        printf("\n");
1902  #endif  #endif
1903    
1904          /* Handle a recursively called group. */
1905    
1906          if (md->recursive != NULL && md->recursive->group_num == number)
1907            {
1908            md->end_match_ptr = eptr;
1909            md->start_match_ptr = mstart;
1910            RRETURN(MATCH_MATCH);
1911            }
1912    
1913          /* Deal with capturing */
1914    
1915        md->capture_last = number;        md->capture_last = number;
1916        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1917          {          {
1918            /* If offset is greater than offset_top, it means that we are
1919            "skipping" a capturing group, and that group's offsets must be marked
1920            unset. In earlier versions of PCRE, all the offsets were unset at the
1921            start of matching, but this doesn't work because atomic groups and
1922            assertions can cause a value to be set that should later be unset.
1923            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1924            part of the atomic group, but this is not on the final matching path,
1925            so must be unset when 2 is set. (If there is no group 2, there is no
1926            problem, because offset_top will then be 2, indicating no capture.) */
1927    
1928            if (offset > offset_top)
1929              {
1930              register int *iptr = md->offset_vector + offset_top;
1931              register int *iend = md->offset_vector + offset;
1932              while (iptr < iend) *iptr++ = -1;
1933              }
1934    
1935            /* Now make the extraction */
1936    
1937          md->offset_vector[offset] =          md->offset_vector[offset] =
1938            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1939          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1940          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1941          }          }
1942          }
1943    
1944        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1945        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1946        group. This is the forcible breaking of infinite loops as implemented in
1947        Perl 5.005. For a non-repeating atomic group that includes captures,
1948        establish a backup point by processing the rest of the pattern at a lower
1949        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1950        original OP_ONCE level, thereby bypassing intermediate backup points, but
1951        resetting any captures that happened along the way. */
1952    
1953        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1954          {
1955          if (*prev == OP_ONCE)
1956          {          {
1957          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1958          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1959          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1960          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;  
1961          }          }
1962          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1963          break;
1964        }        }
1965    
1966      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1967      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1968        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. */  
1969    
1970      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KETRPOS)
1971        {        {
1972        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1973        break;        md->end_offset_top = offset_top;
1974          RRETURN(MATCH_KETRPOS);
1975        }        }
1976    
1977      /* 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
1978      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
1979      tail recursion to avoid using another stack frame. */      use tail recursion to avoid using another stack frame, unless we have an
1980        an atomic group or an unlimited repeat of a group that can match an empty
1981      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      string. */
1982    
1983      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1984        {        {
1985        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1986        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1987          if (*prev == OP_ONCE)
1988            {
1989            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1990            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1991            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1992            RRETURN(MATCH_ONCE);
1993            }
1994          if (*prev >= OP_SBRA)    /* Could match an empty string */
1995            {
1996            md->match_function_type = MATCH_CBEGROUP;
1997            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1998            RRETURN(rrc);
1999            }
2000        ecode = prev;        ecode = prev;
       flags |= match_tail_recursed;  
2001        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2002        }        }
2003      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2004        {        {
2005        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, flags);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
2006          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2007          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2008        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2009          if (*prev == OP_ONCE)
2010            {
2011            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2012            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2013            md->once_target = prev;
2014            RRETURN(MATCH_ONCE);
2015            }
2016        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_tail_recursed;  
2017        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2018        }        }
2019      /* Control never gets here */      /* Control never gets here */
2020    
2021      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2022    
2023      case OP_CIRC:      case OP_CIRC:
2024      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 */  
2025    
2026      /* Start of subject assertion */      /* Start of subject assertion */
2027    
# Line 1221  for (;;) Line 2030  for (;;)
2030      ecode++;      ecode++;
2031      break;      break;
2032    
2033        /* Multiline mode: start of subject unless notbol, or after any newline. */
2034    
2035        case OP_CIRCM:
2036        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2037        if (eptr != md->start_subject &&
2038            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2039          RRETURN(MATCH_NOMATCH);
2040        ecode++;
2041        break;
2042    
2043      /* Start of match assertion */      /* Start of match assertion */
2044    
2045      case OP_SOM:      case OP_SOM:
# Line 1228  for (;;) Line 2047  for (;;)
2047      ecode++;      ecode++;
2048      break;      break;
2049    
2050      /* 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. */  
2051    
2052      case OP_DOLL:      case OP_SET_SOM:
2053      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2054        {      ecode++;
2055        if (eptr < md->end_subject)      break;
2056          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
2057        else      /* Multiline mode: assert before any newline, or before end of subject
2058          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
2059        ecode++;  
2060        break;      case OP_DOLLM:
2061        }      if (eptr < md->end_subject)
2062          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
2063      else      else
2064        {        {
2065        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2066        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2067        }        }
2068        ecode++;
2069        break;
2070    
2071        /* Not multiline mode: assert before a terminating newline or before end of
2072        subject unless noteol is set. */
2073    
2074        case OP_DOLL:
2075        if (md->noteol) RRETURN(MATCH_NOMATCH);
2076        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2077    
2078      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2079    
2080      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2081    
2082      case OP_EOD:      case OP_EOD:
2083      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2084        SCHECK_PARTIAL();
2085      ecode++;      ecode++;
2086      break;      break;
2087    
2088      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2089    
2090      case OP_EODN:      case OP_EODN:
2091      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2092        if (eptr < md->end_subject &&
2093          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2094        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2095    
2096        /* Either at end of string or \n before end. */
2097    
2098        SCHECK_PARTIAL();
2099      ecode++;      ecode++;
2100      break;      break;
2101    
# Line 1279  for (;;) Line 2107  for (;;)
2107    
2108        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2109        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
2110        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2111          partial matching. */
2112    
2113  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2114        if (utf8)        if (utf)
2115          {          {
2116            /* Get status of previous character */
2117    
2118          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2119            {            {
2120            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2121            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2122              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2123            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2124    #ifdef SUPPORT_UCP
2125              if (md->use_ucp)
2126                {
2127                if (c == '_') prev_is_word = TRUE; else
2128                  {
2129                  int cat = UCD_CATEGORY(c);
2130                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2131                  }
2132                }
2133              else
2134    #endif
2135            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2136            }            }
2137          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2138            /* Get status of next character */
2139    
2140            if (eptr >= md->end_subject)
2141              {
2142              SCHECK_PARTIAL();
2143              cur_is_word = FALSE;
2144              }
2145            else
2146            {            {
2147            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2148    #ifdef SUPPORT_UCP
2149              if (md->use_ucp)
2150                {
2151                if (c == '_') cur_is_word = TRUE; else
2152                  {
2153                  int cat = UCD_CATEGORY(c);
2154                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2155                  }
2156                }
2157              else
2158    #endif
2159            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2160            }            }
2161          }          }
2162        else        else
2163  #endif  #endif
2164    
2165        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2166          consistency with the behaviour of \w we do use it in this case. */
2167    
2168          {          {
2169          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2170            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2171          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2172            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2173              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2174    #ifdef SUPPORT_UCP
2175              if (md->use_ucp)
2176                {
2177                c = eptr[-1];
2178                if (c == '_') prev_is_word = TRUE; else
2179                  {
2180                  int cat = UCD_CATEGORY(c);
2181                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2182                  }
2183                }
2184              else
2185    #endif
2186              prev_is_word = MAX_255(eptr[-1])
2187                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2188              }
2189    
2190            /* Get status of next character */
2191    
2192            if (eptr >= md->end_subject)
2193              {
2194              SCHECK_PARTIAL();
2195              cur_is_word = FALSE;
2196              }
2197            else
2198    #ifdef SUPPORT_UCP
2199            if (md->use_ucp)
2200              {
2201              c = *eptr;
2202              if (c == '_') cur_is_word = TRUE; else
2203                {
2204                int cat = UCD_CATEGORY(c);
2205                cur_is_word = (cat == ucp_L || cat == ucp_N);
2206                }
2207              }
2208            else
2209    #endif
2210            cur_is_word = MAX_255(*eptr)
2211              && ((md->ctypes[*eptr] & ctype_word) != 0);
2212          }          }
2213    
2214        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1320  for (;;) Line 2222  for (;;)
2222      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2223    
2224      case OP_ANY:      case OP_ANY:
2225      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2226        {      /* Fall through */
2227        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);  
2228        case OP_ALLANY:
2229        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2230          {                            /* not be updated before SCHECK_PARTIAL. */
2231          SCHECK_PARTIAL();
2232          RRETURN(MATCH_NOMATCH);
2233        }        }
2234      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2235      if (utf8)  #ifdef SUPPORT_UTF
2236        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2237    #endif
2238      ecode++;      ecode++;
2239      break;      break;
2240    
# Line 1334  for (;;) Line 2242  for (;;)
2242      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2243    
2244      case OP_ANYBYTE:      case OP_ANYBYTE:
2245      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2246          {                            /* not be updated before SCHECK_PARTIAL. */
2247          SCHECK_PARTIAL();
2248          RRETURN(MATCH_NOMATCH);
2249          }
2250        eptr++;
2251      ecode++;      ecode++;
2252      break;      break;
2253    
2254      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2255      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2256          {
2257          SCHECK_PARTIAL();
2258          RRETURN(MATCH_NOMATCH);
2259          }
2260      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2261      if (      if (
2262  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2263         c < 256 &&         c < 256 &&
2264  #endif  #endif
2265         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1352  for (;;) Line 2269  for (;;)
2269      break;      break;
2270    
2271      case OP_DIGIT:      case OP_DIGIT:
2272      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2273          {
2274          SCHECK_PARTIAL();
2275          RRETURN(MATCH_NOMATCH);
2276          }
2277      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2278      if (      if (
2279  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2280         c >= 256 ||         c > 255 ||
2281  #endif  #endif
2282         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2283         )         )
# Line 1365  for (;;) Line 2286  for (;;)
2286      break;      break;
2287    
2288      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2289      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2290          {
2291          SCHECK_PARTIAL();
2292          RRETURN(MATCH_NOMATCH);
2293          }
2294      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2295      if (      if (
2296  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2297         c < 256 &&         c < 256 &&
2298  #endif  #endif
2299         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1378  for (;;) Line 2303  for (;;)
2303      break;      break;
2304    
2305      case OP_WHITESPACE:      case OP_WHITESPACE:
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 > 255 ||
2315  #endif  #endif
2316         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2317         )         )
# Line 1391  for (;;) Line 2320  for (;;)
2320      break;      break;
2321    
2322      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
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 < 256 &&
2332  #endif  #endif
2333         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1404  for (;;) Line 2337  for (;;)
2337      break;      break;
2338    
2339      case OP_WORDCHAR:      case OP_WORDCHAR:
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 > 255 ||
2349  #endif  #endif
2350         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2351         )         )
# Line 1417  for (;;) Line 2354  for (;;)
2354      break;      break;
2355    
2356      case OP_ANYNL:      case OP_ANYNL:
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      switch(c)      switch(c)
2364        {        {
2365        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2366    
2367        case 0x000d:        case 0x000d:
2368        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2369        break;        break;
2370    
2371        case 0x000a:        case 0x000a:
2372          break;
2373    
2374        case 0x000b:        case 0x000b:
2375        case 0x000c:        case 0x000c:
2376        case 0x0085:        case 0x0085:
2377        case 0x2028:        case 0x2028:
2378        case 0x2029:        case 0x2029:
2379          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2380          break;
2381          }
2382        ecode++;
2383        break;
2384    
2385        case OP_NOT_HSPACE:
2386        if (eptr >= md->end_subject)
2387          {
2388          SCHECK_PARTIAL();
2389          RRETURN(MATCH_NOMATCH);
2390          }
2391        GETCHARINCTEST(c, eptr);
2392        switch(c)
2393          {
2394          default: break;
2395          case 0x09:      /* HT */
2396          case 0x20:      /* SPACE */
2397          case 0xa0:      /* NBSP */
2398          case 0x1680:    /* OGHAM SPACE MARK */
2399          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2400          case 0x2000:    /* EN QUAD */
2401          case 0x2001:    /* EM QUAD */
2402          case 0x2002:    /* EN SPACE */
2403          case 0x2003:    /* EM SPACE */
2404          case 0x2004:    /* THREE-PER-EM SPACE */
2405          case 0x2005:    /* FOUR-PER-EM SPACE */
2406          case 0x2006:    /* SIX-PER-EM SPACE */
2407          case 0x2007:    /* FIGURE SPACE */
2408          case 0x2008:    /* PUNCTUATION SPACE */
2409          case 0x2009:    /* THIN SPACE */
2410          case 0x200A:    /* HAIR SPACE */
2411          case 0x202f:    /* NARROW NO-BREAK SPACE */
2412          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2413          case 0x3000:    /* IDEOGRAPHIC SPACE */
2414          RRETURN(MATCH_NOMATCH);
2415          }
2416        ecode++;
2417        break;
2418    
2419        case OP_HSPACE:
2420        if (eptr >= md->end_subject)
2421          {
2422          SCHECK_PARTIAL();
2423          RRETURN(MATCH_NOMATCH);
2424          }
2425        GETCHARINCTEST(c, eptr);
2426        switch(c)
2427          {
2428          default: RRETURN(MATCH_NOMATCH);
2429          case 0x09:      /* HT */
2430          case 0x20:      /* SPACE */
2431          case 0xa0:      /* NBSP */
2432          case 0x1680:    /* OGHAM SPACE MARK */
2433          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2434          case 0x2000:    /* EN QUAD */
2435          case 0x2001:    /* EM QUAD */
2436          case 0x2002:    /* EN SPACE */
2437          case 0x2003:    /* EM SPACE */
2438          case 0x2004:    /* THREE-PER-EM SPACE */
2439          case 0x2005:    /* FOUR-PER-EM SPACE */
2440          case 0x2006:    /* SIX-PER-EM SPACE */
2441          case 0x2007:    /* FIGURE SPACE */
2442          case 0x2008:    /* PUNCTUATION SPACE */
2443          case 0x2009:    /* THIN SPACE */
2444          case 0x200A:    /* HAIR SPACE */
2445          case 0x202f:    /* NARROW NO-BREAK SPACE */
2446          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2447          case 0x3000:    /* IDEOGRAPHIC SPACE */
2448          break;
2449          }
2450        ecode++;
2451        break;
2452    
2453        case OP_NOT_VSPACE:
2454        if (eptr >= md->end_subject)
2455          {
2456          SCHECK_PARTIAL();
2457          RRETURN(MATCH_NOMATCH);
2458          }
2459        GETCHARINCTEST(c, eptr);
2460        switch(c)
2461          {
2462          default: break;
2463          case 0x0a:      /* LF */
2464          case 0x0b:      /* VT */
2465          case 0x0c:      /* FF */
2466          case 0x0d:      /* CR */
2467          case 0x85:      /* NEL */
2468          case 0x2028:    /* LINE SEPARATOR */
2469          case 0x2029:    /* PARAGRAPH SEPARATOR */
2470          RRETURN(MATCH_NOMATCH);
2471          }
2472        ecode++;
2473        break;
2474    
2475        case OP_VSPACE:
2476        if (eptr >= md->end_subject)
2477          {
2478          SCHECK_PARTIAL();
2479          RRETURN(MATCH_NOMATCH);
2480          }
2481        GETCHARINCTEST(c, eptr);
2482        switch(c)
2483          {
2484          default: RRETURN(MATCH_NOMATCH);
2485          case 0x0a:      /* LF */
2486          case 0x0b:      /* VT */
2487          case 0x0c:      /* FF */
2488          case 0x0d:      /* CR */
2489          case 0x85:      /* NEL */
2490          case 0x2028:    /* LINE SEPARATOR */
2491          case 0x2029:    /* PARAGRAPH SEPARATOR */
2492        break;        break;
2493        }        }
2494      ecode++;      ecode++;
# Line 1442  for (;;) Line 2500  for (;;)
2500    
2501      case OP_PROP:      case OP_PROP:
2502      case OP_NOTPROP:      case OP_NOTPROP:
2503      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2504          {
2505          SCHECK_PARTIAL();
2506          RRETURN(MATCH_NOMATCH);
2507          }
2508      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2509        {        {
2510        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2511    
2512        switch(ecode[1])        switch(ecode[1])
2513          {          {
# Line 1455  for (;;) Line 2516  for (;;)
2516          break;          break;
2517    
2518          case PT_LAMP:          case PT_LAMP:
2519          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2520               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2521               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2522            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2523           break;          break;
2524    
2525          case PT_GC:          case PT_GC:
2526          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2527            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2528          break;          break;
2529    
2530          case PT_PC:          case PT_PC:
2531          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2532            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2533          break;          break;
2534    
2535          case PT_SC:          case PT_SC:
2536          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2537              RRETURN(MATCH_NOMATCH);
2538            break;
2539    
2540            /* These are specials */
2541    
2542            case PT_ALNUM:
2543            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2544                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2545              RRETURN(MATCH_NOMATCH);
2546            break;
2547    
2548            case PT_SPACE:    /* Perl space */
2549            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2550                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2551                   == (op == OP_NOTPROP))
2552              RRETURN(MATCH_NOMATCH);
2553            break;
2554    
2555            case PT_PXSPACE:  /* POSIX space */
2556            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2557                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2558                 c == CHAR_FF || c == CHAR_CR)
2559                   == (op == OP_NOTPROP))
2560              RRETURN(MATCH_NOMATCH);
2561            break;
2562    
2563            case PT_WORD:
2564            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2565                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2566                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2567            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2568          break;          break;
2569    
2570            /* This should never occur */
2571    
2572          default:          default:
2573          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2574          }          }
# Line 1488  for (;;) Line 2581  for (;;)
2581      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2582    
2583      case OP_EXTUNI:      case OP_EXTUNI:
2584      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2585          {
2586          SCHECK_PARTIAL();
2587          RRETURN(MATCH_NOMATCH);
2588          }
2589      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2590        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2591        while (eptr < md->end_subject)
2592        {        {
2593        int chartype, script;        int len = 1;
2594        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2595        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2596        while (eptr < md->end_subject)        eptr += len;
         {  
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2597        }        }
2598      ecode++;      ecode++;
2599      break;      break;
# Line 1520  for (;;) Line 2609  for (;;)
2609      loops). */      loops). */
2610    
2611      case OP_REF:      case OP_REF:
2612        {      case OP_REFI:
2613        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2614        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2615        ecode += 1 + IMM2_SIZE;
2616    
2617        /* 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];  
2618    
2619        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2620        this ensures that every attempt at a match fails. We can't just fail
2621        here, because of the possibility of quantifiers with zero minima.
2622    
2623        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2624          {      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;  
2625    
2626          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2627          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;  
2628    
2629          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2630          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2631          eptr += length;      else
2632          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2633    
2634        /* Set up for repetition, or handle the non-repeated case */
2635    
2636        switch (*ecode)
2637          {
2638          case OP_CRSTAR:
2639          case OP_CRMINSTAR:
2640          case OP_CRPLUS:
2641          case OP_CRMINPLUS:
2642          case OP_CRQUERY:
2643          case OP_CRMINQUERY:
2644          c = *ecode++ - OP_CRSTAR;
2645          minimize = (c & 1) != 0;
2646          min = rep_min[c];                 /* Pick up values from tables; */
2647          max = rep_max[c];                 /* zero for max => infinity */
2648          if (max == 0) max = INT_MAX;
2649          break;
2650    
2651          case OP_CRRANGE:
2652          case OP_CRMINRANGE:
2653          minimize = (*ecode == OP_CRMINRANGE);
2654          min = GET2(ecode, 1);
2655          max = GET2(ecode, 1 + IMM2_SIZE);
2656          if (max == 0) max = INT_MAX;
2657          ecode += 1 + 2 * IMM2_SIZE;
2658          break;
2659    
2660          default:               /* No repeat follows */
2661          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2662            {
2663            CHECK_PARTIAL();
2664            RRETURN(MATCH_NOMATCH);
2665          }          }
2666          eptr += length;
2667          continue;              /* With the main loop */
2668          }
2669    
2670        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2671        main loop. */      zero, just continue with the main loop. If the length is negative, it
2672        means the reference is unset in non-Java-compatible mode. If the minimum is
2673        zero, we can continue at the same level without recursion. For any other
2674        minimum, carrying on will result in NOMATCH. */
2675    
2676        if (length == 0) continue;      if (length == 0) continue;
2677        if (length < 0 && min == 0) continue;
2678    
2679        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2680        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2681        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2682    
2683        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2684          {
2685          int slength;
2686          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2687          {          {
2688          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2689          eptr += length;          RRETURN(MATCH_NOMATCH);
2690          }          }
2691          eptr += slength;
2692          }
2693    
2694        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2695        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2696    
2697        if (min == max) continue;      if (min == max) continue;
2698    
2699        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2700    
2701        if (minimize)      if (minimize)
2702          {
2703          for (fi = min;; fi++)
2704          {          {
2705          for (fi = min;; fi++)          int slength;
2706            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2707            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2708            if (fi >= max) RRETURN(MATCH_NOMATCH);
2709            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2710            {            {
2711            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2712            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            RRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2713            }            }
2714          /* Control never gets here */          eptr += slength;
2715          }          }
2716          /* Control never gets here */
2717          }
2718    
2719        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2720    
2721        else      else
2722          {
2723          pp = eptr;
2724          for (i = min; i < max; i++)
2725          {          {
2726          pp = eptr;          int slength;
2727          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)  
2728            {            {
2729            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2730            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2731            }            }
2732          RRETURN(MATCH_NOMATCH);          eptr += slength;
2733          }          }
2734          while (eptr >= pp)
2735            {
2736            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2737            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2738            eptr -= length;
2739            }
2740          RRETURN(MATCH_NOMATCH);
2741        }        }
2742      /* Control never gets here */      /* Control never gets here */
2743    
   
   
2744      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2745      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,
2746      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 1637  for (;;) Line 2755  for (;;)
2755      case OP_NCLASS:      case OP_NCLASS:
2756      case OP_CLASS:      case OP_CLASS:
2757        {        {
2758          /* The data variable is saved across frames, so the byte map needs to
2759          be stored there. */
2760    #define BYTE_MAP ((pcre_uint8 *)data)
2761        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2762        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2763    
2764        switch (*ecode)        switch (*ecode)
2765          {          {
# Line 1659  for (;;) Line 2780  for (;;)
2780          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2781          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2782          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2783          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2784          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2785          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2786          break;          break;
2787    
2788          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1671  for (;;) Line 2792  for (;;)
2792    
2793        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2794    
2795  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2796        /* UTF-8 mode */        if (utf)
       if (utf8)  
2797          {          {
2798          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2799            {            {
2800            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2801                {
2802                SCHECK_PARTIAL();
2803                RRETURN(MATCH_NOMATCH);
2804                }
2805            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2806            if (c > 255)            if (c > 255)
2807              {              {
2808              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2809              }              }
2810            else            else
2811              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2812            }            }
2813          }          }
2814        else        else
2815  #endif  #endif
2816        /* Not UTF-8 mode */        /* Not UTF mode */
2817          {          {
2818          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2819            {            {
2820            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2821                {
2822                SCHECK_PARTIAL();
2823                RRETURN(MATCH_NOMATCH);
2824                }
2825            c = *eptr++;            c = *eptr++;
2826            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2827              if (c > 255)
2828                {
2829                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2830                }
2831              else
2832    #endif
2833                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2834            }            }
2835          }          }
2836    
# Line 1711  for (;;) Line 2844  for (;;)
2844    
2845        if (minimize)        if (minimize)
2846          {          {
2847  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2848          /* UTF-8 mode */          if (utf)
         if (utf8)  
2849            {            {
2850            for (fi = min;; fi++)            for (fi = min;; fi++)
2851              {              {
2852              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2853              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2854              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2855                if (eptr >= md->end_subject)
2856                  {
2857                  SCHECK_PARTIAL();
2858                  RRETURN(MATCH_NOMATCH);
2859                  }
2860              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2861              if (c > 255)              if (c > 255)
2862                {                {
2863                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2864                }                }
2865              else              else
2866                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2867              }              }
2868            }            }
2869          else          else
2870  #endif  #endif
2871          /* Not UTF-8 mode */          /* Not UTF mode */
2872            {            {
2873            for (fi = min;; fi++)            for (fi = min;; fi++)
2874              {              {
2875              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2876              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2877              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2878                if (eptr >= md->end_subject)
2879                  {
2880                  SCHECK_PARTIAL();
2881                  RRETURN(MATCH_NOMATCH);
2882                  }
2883              c = *eptr++;              c = *eptr++;
2884              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2885                if (c > 255)
2886                  {
2887                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2888                  }
2889                else
2890    #endif
2891                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2892              }              }
2893            }            }
2894          /* Control never gets here */          /* Control never gets here */
# Line 1753  for (;;) Line 2900  for (;;)
2900          {          {
2901          pp = eptr;          pp = eptr;
2902    
2903  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2904          /* UTF-8 mode */          if (utf)
         if (utf8)  
2905            {            {
2906            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2907              {              {
2908              int len = 1;              int len = 1;
2909              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2910                  {
2911                  SCHECK_PARTIAL();
2912                  break;
2913                  }
2914              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2915              if (c > 255)              if (c > 255)
2916                {                {
2917                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2918                }                }
2919              else              else
2920                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2921              eptr += len;              eptr += len;
2922              }              }
2923            for (;;)            for (;;)
2924              {              {
2925              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2926              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2927              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2928              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1782  for (;;) Line 2930  for (;;)
2930            }            }
2931          else          else
2932  #endif  #endif
2933            /* Not UTF-8 mode */            /* Not UTF mode */
2934            {            {
2935            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2936              {              {
2937              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2938                  {
2939                  SCHECK_PARTIAL();
2940                  break;
2941                  }
2942              c = *eptr;              c = *eptr;
2943              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2944                if (c > 255)
2945                  {
2946                  if (op == OP_CLASS) break;
2947                  }
2948                else
2949    #endif
2950                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2951              eptr++;              eptr++;
2952              }              }
2953            while (eptr >= pp)            while (eptr >= pp)
2954              {              {
2955              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2956              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2957              eptr--;              eptr--;
2958              }              }
# Line 1801  for (;;) Line 2960  for (;;)
2960    
2961          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2962          }          }
2963    #undef BYTE_MAP
2964        }        }
2965      /* Control never gets here */      /* Control never gets here */
2966    
2967    
2968      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2969      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
2970        mode, because Unicode properties are supported in non-UTF-8 mode. */
2971    
2972  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2973      case OP_XCLASS:      case OP_XCLASS:
2974        {        {
2975        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1833  for (;;) Line 2994  for (;;)
2994          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2995          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2996          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2997          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2998          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2999          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3000          break;          break;
3001    
3002          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1847  for (;;) Line 3008  for (;;)
3008    
3009        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3010          {          {
3011          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3012          GETCHARINC(c, eptr);            {
3013          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3014              RRETURN(MATCH_NOMATCH);
3015              }
3016            GETCHARINCTEST(c, eptr);
3017            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3018          }          }
3019    
3020        /* 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 1864  for (;;) Line 3029  for (;;)
3029          {          {
3030          for (fi = min;; fi++)          for (fi = min;; fi++)
3031            {            {
3032            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3033            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3034            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3035            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3036            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3037                SCHECK_PARTIAL();
3038                RRETURN(MATCH_NOMATCH);
3039                }
3040              GETCHARINCTEST(c, eptr);
3041              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3042            }            }
3043          /* Control never gets here */          /* Control never gets here */
3044          }          }
# Line 1881  for (;;) Line 3051  for (;;)
3051          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3052            {            {
3053            int len = 1;            int len = 1;
3054            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3055            GETCHARLEN(c, eptr, len);              {
3056            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3057                break;
3058                }
3059    #ifdef SUPPORT_UTF
3060              GETCHARLENTEST(c, eptr, len);
3061    #else
3062              c = *eptr;
3063    #endif
3064              if (!PRIV(xclass)(c, data, utf)) break;
3065            eptr += len;            eptr += len;
3066            }            }
3067          for(;;)          for(;;)
3068            {            {
3069            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3070            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3071            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3072            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3073              if (utf) BACKCHAR(eptr);
3074    #endif
3075            }            }
3076          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3077          }          }
# Line 1903  for (;;) Line 3083  for (;;)
3083      /* Match a single character, casefully */      /* Match a single character, casefully */
3084    
3085      case OP_CHAR:      case OP_CHAR:
3086  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3087      if (utf8)      if (utf)
3088        {        {
3089        length = 1;        length = 1;
3090        ecode++;        ecode++;
3091        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3092        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3093            {
3094            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3095            RRETURN(MATCH_NOMATCH);
3096            }
3097        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3098        }        }
3099      else      else
3100  #endif  #endif
3101        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3102        {        {
3103        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3104            {
3105            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3106            RRETURN(MATCH_NOMATCH);
3107            }
3108        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3109        ecode += 2;        ecode += 2;
3110        }        }
3111      break;      break;
3112    
3113      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3114        subject, give up immediately. */
3115    
3116        case OP_CHARI:
3117        if (eptr >= md->end_subject)
3118          {
3119          SCHECK_PARTIAL();
3120          RRETURN(MATCH_NOMATCH);
3121          }
3122    
3123      case OP_CHARNC:  #ifdef SUPPORT_UTF
3124  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3125        {        {
3126        length = 1;        length = 1;
3127        ecode++;        ecode++;
3128        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3129    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3130        /* 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
3131        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3132          fast lookup table. We know that there is at least one byte left in the
3133          subject. */
3134    
3135        if (fc < 128)        if (fc < 128)
3136          {          {
3137          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3138                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3139            ecode++;
3140            eptr++;
3141          }          }
3142    
3143        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3144          use the value of "length" to check for sufficient bytes left, because the
3145          other case of the character may have more or fewer bytes.  */
3146    
3147        else        else
3148          {          {
# Line 1957  for (;;) Line 3156  for (;;)
3156          if (fc != dc)          if (fc != dc)
3157            {            {
3158  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3159            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3160  #endif  #endif
3161              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3162            }            }
3163          }          }
3164        }        }
3165      else      else
3166  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3167    
3168      /* Non-UTF-8 mode */      /* Not UTF mode */
3169        {        {
3170        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3171        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3172          eptr++;
3173        ecode += 2;        ecode += 2;
3174        }        }
3175      break;      break;
# Line 1977  for (;;) Line 3177  for (;;)
3177      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3178    
3179      case OP_EXACT:      case OP_EXACT:
3180        case OP_EXACTI:
3181      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3182      ecode += 3;      ecode += 1 + IMM2_SIZE;
3183      goto REPEATCHAR;      goto REPEATCHAR;
3184    
3185      case OP_POSUPTO:      case OP_POSUPTO:
3186        case OP_POSUPTOI:
3187      possessive = TRUE;      possessive = TRUE;
3188      /* Fall through */      /* Fall through */
3189    
3190      case OP_UPTO:      case OP_UPTO:
3191        case OP_UPTOI:
3192      case OP_MINUPTO:      case OP_MINUPTO:
3193        case OP_MINUPTOI:
3194      min = 0;      min = 0;
3195      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3196      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3197      ecode += 3;      ecode += 1 + IMM2_SIZE;
3198      goto REPEATCHAR;      goto REPEATCHAR;
3199    
3200      case OP_POSSTAR:      case OP_POSSTAR:
3201        case OP_POSSTARI:
3202      possessive = TRUE;      possessive = TRUE;
3203      min = 0;      min = 0;
3204      max = INT_MAX;      max = INT_MAX;
# Line 2001  for (;;) Line 3206  for (;;)
3206      goto REPEATCHAR;      goto REPEATCHAR;
3207    
3208      case OP_POSPLUS:      case OP_POSPLUS:
3209        case OP_POSPLUSI:
3210      possessive = TRUE;      possessive = TRUE;
3211      min = 1;      min = 1;
3212      max = INT_MAX;      max = INT_MAX;
# Line 2008  for (;;) Line 3214  for (;;)
3214      goto REPEATCHAR;      goto REPEATCHAR;
3215    
3216      case OP_POSQUERY:      case OP_POSQUERY:
3217        case OP_POSQUERYI:
3218      possessive = TRUE;      possessive = TRUE;
3219      min = 0;      min = 0;
3220      max = 1;      max = 1;
# Line 2015  for (;;) Line 3222  for (;;)
3222      goto REPEATCHAR;      goto REPEATCHAR;
3223    
3224      case OP_STAR:      case OP_STAR:
3225        case OP_STARI:
3226      case OP_MINSTAR:      case OP_MINSTAR:
3227        case OP_MINSTARI:
3228      case OP_PLUS:      case OP_PLUS:
3229        case OP_PLUSI:
3230      case OP_MINPLUS:      case OP_MINPLUS:
3231        case OP_MINPLUSI:
3232      case OP_QUERY:      case OP_QUERY:
3233        case OP_QUERYI:
3234      case OP_MINQUERY:      case OP_MINQUERY:
3235      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3236        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3237      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3238      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3239      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3240      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3241    
3242      /* 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. */  
3243    
3244      REPEATCHAR:      REPEATCHAR:
3245  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3246      if (utf8)      if (utf)
3247        {        {
3248        length = 1;        length = 1;
3249        charptr = ecode;        charptr = ecode;
3250        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3251        ecode += length;        ecode += length;
3252    
3253        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2045  for (;;) Line 3255  for (;;)
3255    
3256        if (length > 1)        if (length > 1)
3257          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3258  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3259          unsigned int othercase;          unsigned int othercase;
3260          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3261              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3262            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3263            else oclength = 0;
3264  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3265    
3266          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3267            {            {
3268            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3269            /* Need braces because of following else */              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3270            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3271              else if (oclength > 0 &&
3272                       eptr <= md->end_subject - oclength &&
3273                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3274    #endif  /* SUPPORT_UCP */
3275            else            else
3276              {              {
3277              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3278              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3279              }              }
3280            }            }
3281    
# Line 2073  for (;;) Line 3285  for (;;)
3285            {            {
3286            for (fi = min;; fi++)            for (fi = min;; fi++)
3287              {              {
3288              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3289              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3290              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3291              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3292              /* Need braces because of following else */                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3293              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3294                else if (oclength > 0 &&
3295                         eptr <= md->end_subject - oclength &&
3296                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3297    #endif  /* SUPPORT_UCP */
3298              else              else
3299                {                {
3300                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3301                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3302                }                }
3303              }              }
3304            /* Control never gets here */            /* Control never gets here */
# Line 2093  for (;;) Line 3309  for (;;)
3309            pp = eptr;            pp = eptr;
3310            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3311              {              {
3312              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3313              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3314              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3315                else if (oclength > 0 &&
3316                         eptr <= md->end_subject - oclength &&
3317                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3318    #endif  /* SUPPORT_UCP */
3319              else              else
3320                {                {
3321                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3322                eptr += oclength;                break;
3323                }                }
3324              }              }
3325    
3326            if (possessive) continue;            if (possessive) continue;
3327            while (eptr >= pp)  
3328             {            for(;;)
3329             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              {
3330             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3331             eptr -= length;              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3332             }              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3333            RRETURN(MATCH_NOMATCH);  #ifdef SUPPORT_UCP
3334                eptr--;
3335                BACKCHAR(eptr);
3336    #else   /* without SUPPORT_UCP */
3337                eptr -= length;
3338    #endif  /* SUPPORT_UCP */
3339                }
3340            }            }
3341          /* Control never gets here */          /* Control never gets here */
3342          }          }
# Line 2120  for (;;) Line 3346  for (;;)
3346        value of fc will always be < 128. */        value of fc will always be < 128. */
3347        }        }
3348      else      else
3349  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3350          /* 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);  
3351        fc = *ecode++;        fc = *ecode++;
       }  
3352    
3353      /* 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
3354      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
3355      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3356      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3357      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2140  for (;;) Line 3362  for (;;)
3362      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3363        max, eptr));        max, eptr));
3364    
3365      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3366        {        {
3367        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3368          /* fc must be < 128 if UTF is enabled. */
3369          foc = md->fcc[fc];
3370    #else
3371    #ifdef SUPPORT_UTF
3372    #ifdef SUPPORT_UCP
3373          if (utf && fc > 127)
3374            foc = UCD_OTHERCASE(fc);
3375    #else
3376          if (utf && fc > 127)
3377            foc = fc;
3378    #endif /* SUPPORT_UCP */
3379          else
3380    #endif /* SUPPORT_UTF */
3381            foc = TABLE_GET(fc, md->fcc, fc);
3382    #endif /* COMPILE_PCRE8 */
3383    
3384        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3385          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3386            if (eptr >= md->end_subject)
3387              {
3388              SCHECK_PARTIAL();
3389              RRETURN(MATCH_NOMATCH);
3390              }
3391            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3392            eptr++;
3393            }
3394        if (min == max) continue;        if (min == max) continue;
3395        if (minimize)        if (minimize)
3396          {          {
3397          for (fi = min;; fi++)          for (fi = min;; fi++)
3398            {            {
3399            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3400            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3401            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3402                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3403                {
3404                SCHECK_PARTIAL();
3405              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3406                }
3407              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3408              eptr++;
3409            }            }
3410          /* Control never gets here */          /* Control never gets here */
3411          }          }
# Line 2163  for (;;) Line 3414  for (;;)
3414          pp = eptr;          pp = eptr;
3415          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3416            {            {
3417            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3418                {
3419                SCHECK_PARTIAL();
3420                break;
3421                }
3422              if (fc != *eptr && foc != *eptr) break;
3423            eptr++;            eptr++;
3424            }            }
3425    
3426          if (possessive) continue;          if (possessive) continue;
3427    
3428          while (eptr >= pp)          while (eptr >= pp)
3429            {            {
3430            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3431            eptr--;            eptr--;
3432            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3433            }            }
# Line 2182  for (;;) Line 3440  for (;;)
3440    
3441      else      else
3442        {        {
3443        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3444            {
3445            if (eptr >= md->end_subject)
3446              {
3447              SCHECK_PARTIAL();
3448              RRETURN(MATCH_NOMATCH);
3449              }
3450            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3451            }
3452    
3453        if (min == max) continue;        if (min == max) continue;
3454    
3455        if (minimize)        if (minimize)
3456          {          {
3457          for (fi = min;; fi++)          for (fi = min;; fi++)
3458            {            {
3459            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3460            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3461            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3462              if (eptr >= md->end_subject)
3463                {
3464                SCHECK_PARTIAL();
3465              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3466                }
3467              if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3468            }            }
3469          /* Control never gets here */          /* Control never gets here */
3470          }          }
# Line 2200  for (;;) Line 3473  for (;;)
3473          pp = eptr;          pp = eptr;
3474          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3475            {            {
3476            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3477                {
3478                SCHECK_PARTIAL();
3479                break;
3480                }
3481              if (fc != *eptr) break;
3482            eptr++;            eptr++;
3483            }            }
3484          if (possessive) continue;          if (possessive) continue;
3485    
3486          while (eptr >= pp)          while (eptr >= pp)
3487            {            {
3488            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);