/[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 91 by nigel, Sat Feb 24 21:41:34 2007 UTC revision 892 by ph10, Wed Jan 18 17:23:20 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  #define NLBLOCK md           /* The block containing newline information */  #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK md             /* Block containing newline information */
50    #define PSSTART start_subject  /* Field containing processed string start */
51    #define PSEND   end_subject    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    /* Undefine some potentially clashing cpp symbols */
56    
57  /* Structure for building a chain of data that actually lives on the  #undef min
58  stack, for holding the values of the subject pointer at the start of each  #undef max
 subpattern, so as to detect when an empty string has been matched by a  
 subpattern - to break infinite loops. When NO_RECURSE is set, these blocks  
 are on the heap, not on the stack. */  
   
 typedef struct eptrblock {  
   struct eptrblock *epb_prev;  
   USPTR epb_saved_eptr;  
 } eptrblock;  
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_isgroup      0x02    /* Set if start of bracketed group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 68  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 81  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 99  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  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;
120  while (length-- > 0)  while (length-- > 0)
121    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);
# Line 114  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 146  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 186  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 208  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 269  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    
373    BOOL Xcur_is_word;    BOOL Xcur_is_word;
374    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
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;  
   int *Xprop_test_variable;  
383  #endif  #endif
384    
385      int Xcodelink;
386    int Xctype;    int Xctype;
387    int Xfc;    unsigned int Xfc;
388    int Xfi;    int Xfi;
389    int Xlength;    int Xlength;
390    int Xmax;    int Xmax;
# Line 321  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 340  typedef struct heapframe { Line 416  typedef struct heapframe {
416  *         Match from current position            *  *         Match from current position            *
417  *************************************************/  *************************************************/
418    
419  /* On entry ecode points to the first opcode, and eptr to the first character  /* This function is called recursively in many circumstances. Whenever it
 in the subject string, while eptrb holds the value of eptr at the start of the  
 last bracketed group - used for breaking infinite loops matching zero-length  
 strings. 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 in subject     eptr        pointer to current character in subject
453     ecode       position in 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_isgroup - this is the start of a bracketed 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 them with "register"  so they can be ordinary variables in all cases. Mark some of them with
476  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
477    
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 */
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 412  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 434  HEAP_RECURSE: Line 535  HEAP_RECURSE:
535    
536  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
537  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
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  
 #define prop_test_variable frame->Xprop_test_variable  
546  #endif  #endif
547    
548  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 470  HEAP_RECURSE: Line 566  HEAP_RECURSE:
566  get preserved during recursion in the normal way. In this environment, fi and  get preserved during recursion in the normal way. In this environment, fi and
567  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
568    
569  #else  #else         /* NO_RECURSE not defined */
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  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
592  const uschar *charptr;             /* in small blocks of the code. My normal */  
593  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
594  BOOL condition;  BOOL condition;
 BOOL minimize;  
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;  
 int *prop_test_variable;  
603  #endif  #endif
604    
605    int codelink;
606  int ctype;  int ctype;
607  int length;  int length;
608  int max;  int max;
# Line 516  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  #endif  #endif     /* NO_RECURSE */
619    
620    /* To save space on the stack and in the heap frame, I have doubled up on some
621    of the local variables that are used only in localised parts of the code, but
622    still need to be preserved over recursive calls of match(). These macros define
623    the alternative names that are used. */
624    
625    #define allow_zero    cur_is_word
626    #define cbegroup      condition
627    #define code_offset   codelink
628    #define condassert    condition
629    #define matched_once  prev_is_word
630    #define foc           number
631    #define save_mark     data
632    
633  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
634  variables. */  variables. */
# Line 524  variables. */ Line 636  variables. */
636  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
637  prop_value = 0;  prop_value = 0;
638  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_variable = NULL;  
639  #endif  #endif
640    
641    
642  /* This label is used for tail recursion, which is used in a few cases even  /* This label is used for tail recursion, which is used in a few cases even
643  when NO_RECURSE is not defined, in order to reduce the amount of stack that is  when NO_RECURSE is not defined, in order to reduce the amount of stack that is
644  used. Thanks to Ian Taylor for noticing this possibility and sending the  used. Thanks to Ian Taylor for noticing this possibility and sending the
# Line 537  TAIL_RECURSE: Line 649  TAIL_RECURSE:
649  /* 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
650  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
651  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()
652  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
653  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
654  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,
655  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
656    
657    #ifdef SUPPORT_UTF
658    utf = md->utf;       /* Local copy of the flag */
659    #else
660    utf = FALSE;
661    #endif
662    
663  /* 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
664  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
665    
666  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
667  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
668    
669  original_ims = ims;    /* Save for resetting on ')' */  /* At the start of a group with an unlimited repeat that may match an empty
670    string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
671    done this way to save having to use another function argument, which would take
672    up space on the stack. See also MATCH_CONDASSERT below.
673    
674    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
675    such remembered pointers, to be checked when we hit the closing ket, in order
676    to break infinite loops that match no characters. When match() is called in
677    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
678    NOT be used with tail recursion, because the memory block that is used is on
679    the stack, so a new one may be required for each match(). */
680    
681  #ifdef SUPPORT_UTF8  if (md->match_function_type == MATCH_CBEGROUP)
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
 /* At the start of a bracketed group, add the current subject pointer to the  
 stack of such pointers, to be re-instated at the end of the group when we hit  
 the closing ket. When match() is called in other circumstances, we don't add to  
 this stack. */  
   
 if ((flags & match_isgroup) != 0)  
682    {    {
   newptrb.epb_prev = eptrb;  
683    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
684      newptrb.epb_prev = eptrb;
685    eptrb = &newptrb;    eptrb = &newptrb;
686      md->match_function_type = 0;
687    }    }
688    
689  /* Now start processing the operations. */  /* Now start processing the opcodes. */
690    
691  for (;;)  for (;;)
692    {    {
693      minimize = possessive = FALSE;
694    op = *ecode;    op = *ecode;
   minimize = FALSE;  
695    
696    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
697    matching at least one subject character. */      {
698        case OP_MARK:
699        md->nomatch_mark = ecode + 2;
700        md->mark = NULL;    /* In case previously set by assertion */
701        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
702          eptrb, RM55);
703        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
704             md->mark == NULL) md->mark = ecode + 2;
705    
706        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
707        argument, and we must check whether that argument matches this MARK's
708        argument. It is passed back in md->start_match_ptr (an overloading of that
709        variable). If it does match, we reset that variable to the current subject
710        position and return MATCH_SKIP. Otherwise, pass back the return code
711        unaltered. */
712    
713        else if (rrc == MATCH_SKIP_ARG &&
714            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
715          {
716          md->start_match_ptr = eptr;
717          RRETURN(MATCH_SKIP);
718          }
719        RRETURN(rrc);
720    
721        case OP_FAIL:
722        RRETURN(MATCH_NOMATCH);
723    
724        /* COMMIT overrides PRUNE, SKIP, and THEN */
725    
726        case OP_COMMIT:
727        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
728          eptrb, RM52);
729        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
730            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
731            rrc != MATCH_THEN)
732          RRETURN(rrc);
733        RRETURN(MATCH_COMMIT);
734    
735        /* PRUNE overrides THEN */
736    
737        case OP_PRUNE:
738        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
739          eptrb, RM51);
740        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
741        RRETURN(MATCH_PRUNE);
742    
743        case OP_PRUNE_ARG:
744        md->nomatch_mark = ecode + 2;
745        md->mark = NULL;    /* In case previously set by assertion */
746        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
747          eptrb, RM56);
748        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
749             md->mark == NULL) md->mark = ecode + 2;
750        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
751        RRETURN(MATCH_PRUNE);
752    
753        /* SKIP overrides PRUNE and THEN */
754    
755        case OP_SKIP:
756        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
757          eptrb, RM53);
758        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
759          RRETURN(rrc);
760        md->start_match_ptr = eptr;   /* Pass back current position */
761        RRETURN(MATCH_SKIP);
762    
763        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
764        nomatch_mark. There is a flag that disables this opcode when re-matching a
765        pattern that ended with a SKIP for which there was not a matching MARK. */
766    
767        case OP_SKIP_ARG:
768        if (md->ignore_skip_arg)
769          {
770          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
771          break;
772          }
773        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
774          eptrb, RM57);
775        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
776          RRETURN(rrc);
777    
778        /* Pass back the current skip name by overloading md->start_match_ptr and
779        returning the special MATCH_SKIP_ARG return code. This will either be
780        caught by a matching MARK, or get to the top, where it causes a rematch
781        with the md->ignore_skip_arg flag set. */
782    
783        md->start_match_ptr = ecode + 2;
784        RRETURN(MATCH_SKIP_ARG);
785    
786        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
787        the branch in which it occurs can be determined. Overload the start of
788        match pointer to do this. */
789    
790        case OP_THEN:
791        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792          eptrb, RM54);
793        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794        md->start_match_ptr = ecode;
795        RRETURN(MATCH_THEN);
796    
797        case OP_THEN_ARG:
798        md->nomatch_mark = ecode + 2;
799        md->mark = NULL;    /* In case previously set by assertion */
800        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
801          md, eptrb, RM58);
802        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
803             md->mark == NULL) md->mark = ecode + 2;
804        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
805        md->start_match_ptr = ecode;
806        RRETURN(MATCH_THEN);
807    
808        /* Handle an atomic group that does not contain any capturing parentheses.
809        This can be handled like an assertion. Prior to 8.13, all atomic groups
810        were handled this way. In 8.13, the code was changed as below for ONCE, so
811        that backups pass through the group and thereby reset captured values.
812        However, this uses a lot more stack, so in 8.20, atomic groups that do not
813        contain any captures generate OP_ONCE_NC, which can be handled in the old,
814        less stack intensive way.
815    
816    if (md->partial &&      Check the alternative branches in turn - the matching won't pass the KET
817        eptr >= md->end_subject &&      for this kind of subpattern. If any one branch matches, we carry on as at
818        eptr > md->start_match)      the end of a normal bracket, leaving the subject pointer, but resetting
819      md->hitend = TRUE;      the start-of-match value in case it was changed by \K. */
   
   /* Opening capturing bracket. If there is space in the offset vector, save  
   the current subject position in the working slot at the top of the vector. We  
   mustn't change the current values of the data slot, because they may be set  
   from a previous iteration of this group, and be referred to by a reference  
   inside the group.  
   
   If the bracket fails to match, we need to restore this value and also the  
   values of the final offsets, in case they were set by a previous iteration of  
   the same bracket.  
   
   If there isn't enough space in the offset vector, treat this as if it were a  
   non-capturing bracket. Don't worry about setting the flag for the error case  
   here; that is handled in the code for KET. */  
820    
821    if (op > OP_BRA)      case OP_ONCE_NC:
822      {      prev = ecode;
823      number = op - OP_BRA;      saved_eptr = eptr;
824        save_mark = md->mark;
825        do
826          {
827          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
828          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
829            {
830            mstart = md->start_match_ptr;
831            break;
832            }
833          if (rrc == MATCH_THEN)
834            {
835            next = ecode + GET(ecode,1);
836            if (md->start_match_ptr < next &&
837                (*ecode == OP_ALT || *next == OP_ALT))
838              rrc = MATCH_NOMATCH;
839            }
840    
841          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
842          ecode += GET(ecode,1);
843          md->mark = save_mark;
844          }
845        while (*ecode == OP_ALT);
846    
847        /* If hit the end of the group (which could be repeated), fail */
848    
849        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
850    
851        /* Continue as from after the group, updating the offsets high water
852        mark, since extracts may have been taken. */
853    
854        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
855    
856        offset_top = md->end_offset_top;
857        eptr = md->end_match_ptr;
858    
859        /* For a non-repeating ket, just continue at this level. This also
860        happens for a repeating ket if no characters were matched in the group.
861        This is the forcible breaking of infinite loops as implemented in Perl
862        5.005. */
863    
864        if (*ecode == OP_KET || eptr == saved_eptr)
865          {
866          ecode += 1+LINK_SIZE;
867          break;
868          }
869    
870      /* For extended extraction brackets (large number), we have to fish out the      /* The repeating kets try the rest of the pattern or restart from the
871      number from a dummy opcode at the start. */      preceding bracket, in the appropriate order. The second "call" of match()
872        uses tail recursion, to avoid using another stack frame. */
873    
874        if (*ecode == OP_KETRMIN)
875          {
876          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
877          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
878          ecode = prev;
879          goto TAIL_RECURSE;
880          }
881        else  /* OP_KETRMAX */
882          {
883          md->match_function_type = MATCH_CBEGROUP;
884          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
885          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
886          ecode += 1 + LINK_SIZE;
887          goto TAIL_RECURSE;
888          }
889        /* Control never gets here */
890    
891      if (number > EXTRACT_BASIC_MAX)      /* Handle a capturing bracket, other than those that are possessive with an
892        number = GET2(ecode, 2+LINK_SIZE);      unlimited repeat. If there is space in the offset vector, save the current
893        subject position in the working slot at the top of the vector. We mustn't
894        change the current values of the data slot, because they may be set from a
895        previous iteration of this group, and be referred to by a reference inside
896        the group. A failure to match might occur after the group has succeeded,
897        if something later on doesn't match. For this reason, we need to restore
898        the working value and also the values of the final offsets, in case they
899        were set by a previous iteration of the same bracket.
900    
901        If there isn't enough space in the offset vector, treat this as if it were
902        a non-capturing bracket. Don't worry about setting the flag for the error
903        case here; that is handled in the code for KET. */
904    
905        case OP_CBRA:
906        case OP_SCBRA:
907        number = GET2(ecode, 1+LINK_SIZE);
908      offset = number << 1;      offset = number << 1;
909    
910  #ifdef DEBUG  #ifdef PCRE_DEBUG
911      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
912        printf("subject=");
913      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
914      printf("\n");      printf("\n");
915  #endif  #endif
# Line 620  for (;;) Line 920  for (;;)
920        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
921        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
922        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
923          save_mark = md->mark;
924    
925        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
926        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
927            (int)(eptr - md->start_subject);
928    
929        do        for (;;)
930          {          {
931          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
932            match_isgroup);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
933              eptrb, RM1);
934            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
935    
936            /* If we backed up to a THEN, check whether it is within the current
937            branch by comparing the address of the THEN that is passed back with
938            the end of the branch. If it is within the current branch, and the
939            branch is one of two or more alternatives (it either starts or ends
940            with OP_ALT), we have reached the limit of THEN's action, so convert
941            the return code to NOMATCH, which will cause normal backtracking to
942            happen from now on. Otherwise, THEN is passed back to an outer
943            alternative. This implements Perl's treatment of parenthesized groups,
944            where a group not containing | does not affect the current alternative,
945            that is, (X) is NOT the same as (X|(*F)). */
946    
947            if (rrc == MATCH_THEN)
948              {
949              next = ecode + GET(ecode,1);
950              if (md->start_match_ptr < next &&
951                  (*ecode == OP_ALT || *next == OP_ALT))
952                rrc = MATCH_NOMATCH;
953              }
954    
955            /* Anything other than NOMATCH is passed back. */
956    
957          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
958          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
959          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
960            md->mark = save_mark;
961            if (*ecode != OP_ALT) break;
962          }          }
       while (*ecode == OP_ALT);  
963    
964        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
965        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
966        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
967        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
968    
969        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
970    
971          RRETURN(rrc);
972        }        }
973    
974      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
975        as a non-capturing bracket. */
976    
977      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
978      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
979    
980    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
981    
982    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
983      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
984      case OP_BRA:     /* Non-capturing bracket: optimized */  
985      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
986        repeat and ONCE group with no captures. Loop for all the alternatives.
987    
988        When we get to the final alternative within the brackets, we used to return
989        the result of a recursive call to match() whatever happened so it was
990        possible to reduce stack usage by turning this into a tail recursion,
991        except in the case of a possibly empty group. However, now that there is
992        the possiblity of (*THEN) occurring in the final alternative, this
993        optimization is no longer always possible.
994    
995      /* Loop for all the alternatives */      We can optimize if we know there are no (*THEN)s in the pattern; at present
996        this is the best that can be done.
997    
998        MATCH_ONCE is returned when the end of an atomic group is successfully
999        reached, but subsequent matching fails. It passes back up the tree (causing
1000        captured values to be reset) until the original atomic group level is
1001        reached. This is tested by comparing md->once_target with the start of the
1002        group. At this point, the return is converted into MATCH_NOMATCH so that
1003        previous backup points can be taken. */
1004    
1005        case OP_ONCE:
1006        case OP_BRA:
1007        case OP_SBRA:
1008        DPRINTF(("start non-capturing bracket\n"));
1009    
1010      for (;;)      for (;;)
1011        {        {
1012        /* When we get to the final alternative within the brackets, we would        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
       return the result of a recursive call to match() whatever happened. We  
       can reduce stack usage by turning this into a tail recursion. */  
   
       if (ecode[GET(ecode, 1)] != OP_ALT)  
        {  
        ecode += 1 + LINK_SIZE;  
        flags = match_isgroup;  
        DPRINTF(("bracket 0 tail recursion\n"));  
        goto TAIL_RECURSE;  
        }  
1013    
1014        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* If this is not a possibly empty group, and there are no (*THEN)s in
1015        otherwise return. */        the pattern, and this is the final alternative, optimize as described
1016          above. */
1017    
1018        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1019          match_isgroup);          {
1020        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1021            goto TAIL_RECURSE;
1022            }
1023    
1024          /* In all other cases, we have to make another call to match(). */
1025    
1026          save_mark = md->mark;
1027          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1028            RM2);
1029    
1030          /* See comment in the code for capturing groups above about handling
1031          THEN. */
1032    
1033          if (rrc == MATCH_THEN)
1034            {
1035            next = ecode + GET(ecode,1);
1036            if (md->start_match_ptr < next &&
1037                (*ecode == OP_ALT || *next == OP_ALT))
1038              rrc = MATCH_NOMATCH;
1039            }
1040    
1041          if (rrc != MATCH_NOMATCH)
1042            {
1043            if (rrc == MATCH_ONCE)
1044              {
1045              const pcre_uchar *scode = ecode;
1046              if (*scode != OP_ONCE)           /* If not at start, find it */
1047                {
1048                while (*scode == OP_ALT) scode += GET(scode, 1);
1049                scode -= GET(scode, 1);
1050                }
1051              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1052              }
1053            RRETURN(rrc);
1054            }
1055        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1056          md->mark = save_mark;
1057          if (*ecode != OP_ALT) break;
1058        }        }
     /* Control never reaches here. */  
1059    
1060      /* 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. */  
1061    
1062      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1063      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1064        {      handled similarly to the normal case above. However, the matching is
1065        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      different. The end of these brackets will always be OP_KETRPOS, which
1066        condition = (offset == CREF_RECURSE * 2)?      returns MATCH_KETRPOS without going further in the pattern. By this means
1067          (md->recursive != NULL) :      we can handle the group by iteration rather than recursion, thereby
1068          (offset < offset_top && md->offset_vector[offset] >= 0);      reducing the amount of stack needed. */
1069        ecode += condition? (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1));  
1070        flags = match_isgroup;      case OP_CBRAPOS:
1071        goto TAIL_RECURSE;      case OP_SCBRAPOS:
1072        }      allow_zero = FALSE;
1073    
1074      /* The condition is an assertion. Call match() to evaluate it - setting      POSSESSIVE_CAPTURE:
1075      the final argument TRUE causes it to stop at the end of an assertion. */      number = GET2(ecode, 1+LINK_SIZE);
1076        offset = number << 1;
1077    
1078      else  #ifdef PCRE_DEBUG
1079        printf("start possessive bracket %d\n", number);
1080        printf("subject=");
1081        pchars(eptr, 16, TRUE, md);
1082        printf("\n");
1083    #endif
1084    
1085        if (offset < md->offset_max)
1086        {        {
1087        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1088            match_condassert | match_isgroup);        code_offset = (int)(ecode - md->start_code);
1089        if (rrc == MATCH_MATCH)  
1090          {        save_offset1 = md->offset_vector[offset];
1091          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);        save_offset2 = md->offset_vector[offset+1];
1092          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_offset3 = md->offset_vector[md->offset_end - number];
1093          save_capture_last = md->capture_last;
1094    
1095          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1096    
1097          /* Each time round the loop, save the current subject position for use
1098          when the group matches. For MATCH_MATCH, the group has matched, so we
1099          restart it with a new subject starting position, remembering that we had
1100          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1101          usual. If we haven't matched any alternatives in any iteration, check to
1102          see if a previous iteration matched. If so, the group has matched;
1103          continue from afterwards. Otherwise it has failed; restore the previous
1104          capture values before returning NOMATCH. */
1105    
1106          for (;;)
1107            {
1108            md->offset_vector[md->offset_end - number] =
1109              (int)(eptr - md->start_subject);
1110            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1111            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1112              eptrb, RM63);
1113            if (rrc == MATCH_KETRPOS)
1114              {
1115              offset_top = md->end_offset_top;
1116              eptr = md->end_match_ptr;
1117              ecode = md->start_code + code_offset;
1118              save_capture_last = md->capture_last;
1119              matched_once = TRUE;
1120              continue;
1121              }
1122    
1123            /* See comment in the code for capturing groups above about handling
1124            THEN. */
1125    
1126            if (rrc == MATCH_THEN)
1127              {
1128              next = ecode + GET(ecode,1);
1129              if (md->start_match_ptr < next &&
1130                  (*ecode == OP_ALT || *next == OP_ALT))
1131                rrc = MATCH_NOMATCH;
1132              }
1133    
1134            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1135            md->capture_last = save_capture_last;
1136            ecode += GET(ecode, 1);
1137            if (*ecode != OP_ALT) break;
1138          }          }
1139        else if (rrc != MATCH_NOMATCH)  
1140          if (!matched_once)
1141          {          {
1142          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1143            md->offset_vector[offset+1] = save_offset2;
1144            md->offset_vector[md->offset_end - number] = save_offset3;
1145          }          }
       else ecode += GET(ecode, 1);  
1146    
1147        /* We are now at the branch that is to be obeyed. As there is only one,        if (allow_zero || matched_once)
1148        we can use tail recursion to avoid using another stack frame. */          {
1149            ecode += 1 + LINK_SIZE;
1150            break;
1151            }
1152    
1153        ecode += 1 + LINK_SIZE;        RRETURN(MATCH_NOMATCH);
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
1154        }        }
     /* Control never reaches here */  
1155    
1156      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1157      encountered. */      as a non-capturing bracket. */
1158    
1159      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1160      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1161    
1162      /* End of the pattern. If we are in a recursion, we should restore the      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     offsets appropriately and continue from after the call. */  
1163    
1164      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1165      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;  
       }  
1166    
1167      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1168      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1169        without the capturing complication. It is written out separately for speed
1170        and cleanliness. */
1171    
1172      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1173      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1174      md->end_offset_top = offset_top;   /* and how many extracts were taken */      allow_zero = FALSE;
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1175    
1176      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1177      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1178      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      code_offset = (int)(ecode - md->start_code);
     start of each branch to move the current point backwards, so the code at  
     this level is identical to the lookahead case. */  
1179    
1180      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1181        {        {
1182        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1183          match_isgroup);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1184        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1185          if (rrc == MATCH_KETRPOS)
1186            {
1187            offset_top = md->end_offset_top;
1188            eptr = md->end_match_ptr;
1189            ecode = md->start_code + code_offset;
1190            matched_once = TRUE;
1191            continue;
1192            }
1193    
1194          /* See comment in the code for capturing groups above about handling
1195          THEN. */
1196    
1197          if (rrc == MATCH_THEN)
1198            {
1199            next = ecode + GET(ecode,1);
1200            if (md->start_match_ptr < next &&
1201                (*ecode == OP_ALT || *next == OP_ALT))
1202              rrc = MATCH_NOMATCH;
1203            }
1204    
1205        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1206        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1207          if (*ecode != OP_ALT) break;
1208        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1209    
1210      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1211          {
1212          ecode += 1 + LINK_SIZE;
1213          break;
1214          }
1215        RRETURN(MATCH_NOMATCH);
1216    
1217      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1218    
1219      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1220      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1221        past the end if there is only one branch, but that's OK because that is
1222        exactly what going to the ket would do. */
1223    
1224      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1225      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1226      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1227    
1228      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1229        inserted between OP_COND and an assertion condition. */
1230    
1231      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1232        {        {
1233        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (PUBL(callout) != NULL)
1234          match_isgroup);          {
1235        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1236        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1237        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1238            cb.offset_vector    = md->offset_vector;
1239    #ifdef COMPILE_PCRE8
1240            cb.subject          = (PCRE_SPTR)md->start_subject;
1241    #else
1242            cb.subject          = (PCRE_SPTR16)md->start_subject;
1243    #endif
1244            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1245            cb.start_match      = (int)(mstart - md->start_subject);
1246            cb.current_position = (int)(eptr - md->start_subject);
1247            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1248            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1249            cb.capture_top      = offset_top/2;
1250            cb.capture_last     = md->capture_last;
1251            cb.callout_data     = md->callout_data;
1252            cb.mark             = md->nomatch_mark;
1253            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1254            if (rrc < 0) RRETURN(rrc);
1255            }
1256          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1257        }        }
     while (*ecode == OP_ALT);  
1258    
1259      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      condcode = ecode[LINK_SIZE+1];
   
     ecode += 1 + LINK_SIZE;  
     continue;  
1260    
1261      /* Move the subject pointer back. This occurs only at the start of      /* Now see what the actual condition is */
     each branch of a lookbehind assertion. If we are too close to the start to  
     move back, this match function fails. When working with UTF-8 we move  
     back a number of characters, not bytes. */  
1262    
1263      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1264        {        {
1265        c = GET(ecode,1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
       for (i = 0; i < c; i++)  
1266          {          {
1267          eptr--;          condition = FALSE;
1268          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          ecode += GET(ecode, 1);
         BACKCHAR(eptr)  
1269          }          }
1270        }        else
1271      else          {
1272  #endif          int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1273            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1274    
1275      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */          /* If the test is for recursion into a specific subpattern, and it is
1276            false, but the test was set up by name, scan the table to see if the
1277            name refers to any other numbers, and test them. The condition is true
1278            if any one is set. */
1279    
1280        {          if (!condition && condcode == OP_NRREF)
1281        eptr -= GET(ecode,1);            {
1282        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);            pcre_uchar *slotA = md->name_table;
1283        }            for (i = 0; i < md->name_count; i++)
1284                {
1285                if (GET2(slotA, 0) == recno) break;
1286                slotA += md->name_entry_size;
1287                }
1288    
1289      /* Skip to next op code */            /* Found a name for the number - there can be only one; duplicate
1290              names for different numbers are allowed, but not vice versa. First
1291              scan down for duplicates. */
1292    
1293      ecode += 1 + LINK_SIZE;            if (i < md->name_count)
1294      break;              {
1295                pcre_uchar *slotB = slotA;
1296                while (slotB > md->name_table)
1297                  {
1298                  slotB -= md->name_entry_size;
1299                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1300                    {
1301                    condition = GET2(slotB, 0) == md->recursive->group_num;
1302                    if (condition) break;
1303                    }
1304                  else break;
1305                  }
1306    
1307                /* Scan up for duplicates */
1308    
1309                if (!condition)
1310                  {
1311                  slotB = slotA;
1312                  for (i++; i < md->name_count; i++)
1313                    {
1314                    slotB += md->name_entry_size;
1315                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1316                      {
1317                      condition = GET2(slotB, 0) == md->recursive->group_num;
1318                      if (condition) break;
1319                      }
1320                    else break;
1321                    }
1322                  }
1323                }
1324              }
1325    
1326            /* Chose branch according to the condition */
1327    
1328            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1329            }
1330          }
1331    
1332        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1333          {
1334          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1335          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1336    
1337          /* If the numbered capture is unset, but the reference was by name,
1338          scan the table to see if the name refers to any other numbers, and test
1339          them. The condition is true if any one is set. This is tediously similar
1340          to the code above, but not close enough to try to amalgamate. */
1341    
1342          if (!condition && condcode == OP_NCREF)
1343            {
1344            int refno = offset >> 1;
1345            pcre_uchar *slotA = md->name_table;
1346    
1347            for (i = 0; i < md->name_count; i++)
1348              {
1349              if (GET2(slotA, 0) == refno) break;
1350              slotA += md->name_entry_size;
1351              }
1352    
1353            /* Found a name for the number - there can be only one; duplicate names
1354            for different numbers are allowed, but not vice versa. First scan down
1355            for duplicates. */
1356    
1357            if (i < md->name_count)
1358              {
1359              pcre_uchar *slotB = slotA;
1360              while (slotB > md->name_table)
1361                {
1362                slotB -= md->name_entry_size;
1363                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1364                  {
1365                  offset = GET2(slotB, 0) << 1;
1366                  condition = offset < offset_top &&
1367                    md->offset_vector[offset] >= 0;
1368                  if (condition) break;
1369                  }
1370                else break;
1371                }
1372    
1373              /* Scan up for duplicates */
1374    
1375              if (!condition)
1376                {
1377                slotB = slotA;
1378                for (i++; i < md->name_count; i++)
1379                  {
1380                  slotB += md->name_entry_size;
1381                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1382                    {
1383                    offset = GET2(slotB, 0) << 1;
1384                    condition = offset < offset_top &&
1385                      md->offset_vector[offset] >= 0;
1386                    if (condition) break;
1387                    }
1388                  else break;
1389                  }
1390                }
1391              }
1392            }
1393    
1394          /* Chose branch according to the condition */
1395    
1396          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1397          }
1398    
1399        else if (condcode == OP_DEF)     /* DEFINE - always false */
1400          {
1401          condition = FALSE;
1402          ecode += GET(ecode, 1);
1403          }
1404    
1405        /* The condition is an assertion. Call match() to evaluate it - setting
1406        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1407        an assertion. */
1408    
1409        else
1410          {
1411          md->match_function_type = MATCH_CONDASSERT;
1412          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1413          if (rrc == MATCH_MATCH)
1414            {
1415            if (md->end_offset_top > offset_top)
1416              offset_top = md->end_offset_top;  /* Captures may have happened */
1417            condition = TRUE;
1418            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1419            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1420            }
1421    
1422          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1423          assertion; it is therefore treated as NOMATCH. */
1424    
1425          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1426            {
1427            RRETURN(rrc);         /* Need braces because of following else */
1428            }
1429          else
1430            {
1431            condition = FALSE;
1432            ecode += codelink;
1433            }
1434          }
1435    
1436        /* We are now at the branch that is to be obeyed. As there is only one, can
1437        use tail recursion to avoid using another stack frame, except when there is
1438        unlimited repeat of a possibly empty group. In the latter case, a recursive
1439        call to match() is always required, unless the second alternative doesn't
1440        exist, in which case we can just plough on. Note that, for compatibility
1441        with Perl, the | in a conditional group is NOT treated as creating two
1442        alternatives. If a THEN is encountered in the branch, it propagates out to
1443        the enclosing alternative (unless nested in a deeper set of alternatives,
1444        of course). */
1445    
1446        if (condition || *ecode == OP_ALT)
1447          {
1448          if (op != OP_SCOND)
1449            {
1450            ecode += 1 + LINK_SIZE;
1451            goto TAIL_RECURSE;
1452            }
1453    
1454          md->match_function_type = MATCH_CBEGROUP;
1455          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1456          RRETURN(rrc);
1457          }
1458    
1459         /* Condition false & no alternative; continue after the group. */
1460    
1461        else
1462          {
1463          ecode += 1 + LINK_SIZE;
1464          }
1465        break;
1466    
1467    
1468        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1469        to close any currently open capturing brackets. */
1470    
1471        case OP_CLOSE:
1472        number = GET2(ecode, 1);
1473        offset = number << 1;
1474    
1475    #ifdef PCRE_DEBUG
1476          printf("end bracket %d at *ACCEPT", number);
1477          printf("\n");
1478    #endif
1479    
1480        md->capture_last = number;
1481        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1482          {
1483          md->offset_vector[offset] =
1484            md->offset_vector[md->offset_end - number];
1485          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1486          if (offset_top <= offset) offset_top = offset + 2;
1487          }
1488        ecode += 1 + IMM2_SIZE;
1489        break;
1490    
1491    
1492        /* End of the pattern, either real or forced. */
1493    
1494        case OP_END:
1495        case OP_ACCEPT:
1496        case OP_ASSERT_ACCEPT:
1497    
1498        /* If we have matched an empty string, fail if not in an assertion and not
1499        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1500        is set and we have matched at the start of the subject. In both cases,
1501        backtracking will then try other alternatives, if any. */
1502    
1503        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1504             md->recursive == NULL &&
1505             (md->notempty ||
1506               (md->notempty_atstart &&
1507                 mstart == md->start_subject + md->start_offset)))
1508          RRETURN(MATCH_NOMATCH);
1509    
1510        /* Otherwise, we have a match. */
1511    
1512        md->end_match_ptr = eptr;           /* Record where we ended */
1513        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1514        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1515    
1516        /* For some reason, the macros don't work properly if an expression is
1517        given as the argument to RRETURN when the heap is in use. */
1518    
1519        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1520        RRETURN(rrc);
1521    
1522        /* Assertion brackets. Check the alternative branches in turn - the
1523        matching won't pass the KET for an assertion. If any one branch matches,
1524        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1525        start of each branch to move the current point backwards, so the code at
1526        this level is identical to the lookahead case. When the assertion is part
1527        of a condition, we want to return immediately afterwards. The caller of
1528        this incarnation of the match() function will have set MATCH_CONDASSERT in
1529        md->match_function type, and one of these opcodes will be the first opcode
1530        that is processed. We use a local variable that is preserved over calls to
1531        match() to remember this case. */
1532    
1533        case OP_ASSERT:
1534        case OP_ASSERTBACK:
1535        save_mark = md->mark;
1536        if (md->match_function_type == MATCH_CONDASSERT)
1537          {
1538          condassert = TRUE;
1539          md->match_function_type = 0;
1540          }
1541        else condassert = FALSE;
1542    
1543        do
1544          {
1545          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1546          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1547            {
1548            mstart = md->start_match_ptr;   /* In case \K reset it */
1549            break;
1550            }
1551    
1552          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1553          as NOMATCH. */
1554    
1555          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1556          ecode += GET(ecode, 1);
1557          md->mark = save_mark;
1558          }
1559        while (*ecode == OP_ALT);
1560    
1561        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1562    
1563        /* If checking an assertion for a condition, return MATCH_MATCH. */
1564    
1565        if (condassert) RRETURN(MATCH_MATCH);
1566    
1567        /* Continue from after the assertion, updating the offsets high water
1568        mark, since extracts may have been taken during the assertion. */
1569    
1570        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1571        ecode += 1 + LINK_SIZE;
1572        offset_top = md->end_offset_top;
1573        continue;
1574    
1575        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1576        PRUNE, or COMMIT means we must assume failure without checking subsequent
1577        branches. */
1578    
1579        case OP_ASSERT_NOT:
1580        case OP_ASSERTBACK_NOT:
1581        save_mark = md->mark;
1582        if (md->match_function_type == MATCH_CONDASSERT)
1583          {
1584          condassert = TRUE;
1585          md->match_function_type = 0;
1586          }
1587        else condassert = FALSE;
1588    
1589        do
1590          {
1591          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1592          md->mark = save_mark;
1593          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1594          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1595            {
1596            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1597            break;
1598            }
1599    
1600          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1601          as NOMATCH. */
1602    
1603          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1604          ecode += GET(ecode,1);
1605          }
1606        while (*ecode == OP_ALT);
1607    
1608        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1609    
1610        ecode += 1 + LINK_SIZE;
1611        continue;
1612    
1613        /* Move the subject pointer back. This occurs only at the start of
1614        each branch of a lookbehind assertion. If we are too close to the start to
1615        move back, this match function fails. When working with UTF-8 we move
1616        back a number of characters, not bytes. */
1617    
1618        case OP_REVERSE:
1619    #ifdef SUPPORT_UTF
1620        if (utf)
1621          {
1622          i = GET(ecode, 1);
1623          while (i-- > 0)
1624            {
1625            eptr--;
1626            if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1627            BACKCHAR(eptr);
1628            }
1629          }
1630        else
1631    #endif
1632    
1633        /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1634    
1635          {
1636          eptr -= GET(ecode, 1);
1637          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1638          }
1639    
1640        /* Save the earliest consulted character, then skip to next op code */
1641    
1642        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1643        ecode += 1 + LINK_SIZE;
1644        break;
1645    
1646      /* The callout item calls an external function, if one is provided, passing      /* The callout item calls an external function, if one is provided, passing
1647      details of the match so far. This is mainly for debugging, though the      details of the match so far. This is mainly for debugging, though the
1648      function is able to force a failure. */      function is able to force a failure. */
1649    
1650      case OP_CALLOUT:      case OP_CALLOUT:
1651      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1652        {        {
1653        pcre_callout_block cb;        PUBL(callout_block) cb;
1654        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1655        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1656        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1657    #ifdef COMPILE_PCRE8
1658        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1659        cb.subject_length   = md->end_subject - md->start_subject;  #else
1660        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1661        cb.current_position = eptr - md->start_subject;  #endif
1662          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1663          cb.start_match      = (int)(mstart - md->start_subject);
1664          cb.current_position = (int)(eptr - md->start_subject);
1665        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1666        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1667        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1668        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1669        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1670        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1671          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1672        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1673        }        }
1674      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 879  for (;;) Line 1678  for (;;)
1678      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
1679      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1680    
1681      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1682      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
1683      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
1684      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
1685      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
1686      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
1687      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.  
1688    
1689      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
1690      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
1691      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1692        a lot, so he is not to blame for the current way it works. */
1693    
1694      case OP_RECURSE:      case OP_RECURSE:
1695        {        {
1696        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1697        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1698    
1699        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1700        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1701            GET2(callpat, 1 + LINK_SIZE);
1702    
1703        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1704          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1705          caught at compile time.) */
1706    
1707          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1708            if (recno == ri->group_num && eptr == ri->subject_position)
1709              RRETURN(PCRE_ERROR_RECURSELOOP);
1710    
1711        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1712    
1713          new_recursive.group_num = recno;
1714          new_recursive.subject_position = eptr;
1715        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1716        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1717    
1718        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1719    
1720        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1721    
1722        /* Now save the offset data. */        /* Now save the offset data */
1723    
1724        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1725        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 923  for (;;) Line 1727  for (;;)
1727        else        else
1728          {          {
1729          new_recursive.offset_save =          new_recursive.offset_save =
1730            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1731          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1732          }          }
   
1733        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1734              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1735    
1736        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1737        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1738          might be changed, so reset it before looping. */
1739    
1740        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1741          cbegroup = (*callpat >= OP_SBRA);
1742        do        do
1743          {          {
1744          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1745              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1746          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1747            memcpy(md->offset_vector, new_recursive.offset_save,
1748                new_recursive.saved_max * sizeof(int));
1749            md->recursive = new_recursive.prevrec;
1750            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1751            {            {
1752            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1753            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1754              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1755            RRETURN(MATCH_MATCH);  
1756              /* Set where we got to in the subject, and reset the start in case
1757              it was changed by \K. This *is* propagated back out of a recursion,
1758              for Perl compatibility. */
1759    
1760              eptr = md->end_match_ptr;
1761              mstart = md->start_match_ptr;
1762              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1763            }            }
1764          else if (rrc != MATCH_NOMATCH)  
1765            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1766            as NOMATCH. */
1767    
1768            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1769            {            {
1770            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1771              if (new_recursive.offset_save != stacksave)
1772                (PUBL(free))(new_recursive.offset_save);
1773            RRETURN(rrc);            RRETURN(rrc);
1774            }            }
1775    
1776          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1777          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1778          }          }
1779        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 964  for (;;) Line 1781  for (;;)
1781        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1782        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1783        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1784          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1785        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1786        }        }
     /* Control never reaches here */  
1787    
1788      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1789      the point in the subject string is not moved back. Thus there can never be      break;
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
         eptrb, match_isgroup);  
       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;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1790    
1791      /* 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
1792      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1051  for (;;) Line 1795  for (;;)
1795      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1796      break;      break;
1797    
1798      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1799      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
1800      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
1801      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
1802      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1803    
1804      case OP_BRAZERO:      case OP_BRAZERO:
1805        {      next = ecode + 1;
1806        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1807        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1808        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1809        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1810      break;      break;
1811    
1812      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1813        {      next = ecode + 1;
1814        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1815        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1816        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1817          match_isgroup);      ecode++;
1818        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1819        ecode++;  
1820        }      case OP_SKIPZERO:
1821        next = ecode+1;
1822        do next += GET(next,1); while (*next == OP_ALT);
1823        ecode = next + 1 + LINK_SIZE;
1824      break;      break;
1825    
1826      /* End of a group, repeated or non-repeating. If we are at the end of      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1827      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1828      current high water mark for use by positive assertions. Do this also  
1829      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1830        op = *(++ecode);
1831        allow_zero = TRUE;
1832        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1833          goto POSSESSIVE_NON_CAPTURE;
1834    
1835        /* End of a group, repeated or non-repeating. */
1836    
1837      case OP_KET:      case OP_KET:
1838      case OP_KETRMIN:      case OP_KETRMIN:
1839      case OP_KETRMAX:      case OP_KETRMAX:
1840        case OP_KETRPOS:
1841      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
     saved_eptr = eptrb->epb_saved_eptr;  
   
     /* Back up the stack of bracket start pointers. */  
1842    
1843      eptrb = eptrb->epb_prev;      /* If this was a group that remembered the subject start, in order to break
1844        infinite repeats of empty string matches, retrieve the subject start from
1845        the chain. Otherwise, set it NULL. */
1846    
1847      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev >= OP_SBRA || *prev == OP_ONCE)
         *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||  
         *prev == OP_ONCE)  
1848        {        {
1849        md->end_match_ptr = eptr;      /* For ONCE */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1850        md->end_offset_top = offset_top;        eptrb = eptrb->epb_prev;              /* Backup to previous group */
       RRETURN(MATCH_MATCH);  
1851        }        }
1852        else saved_eptr = NULL;
1853    
1854      /* In all other cases except a conditional group we have to check the      /* If we are at the end of an assertion group or a non-capturing atomic
1855      group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1856      extraction by setting the offsets and bumping the high water mark. */      water mark for use by positive assertions. We also need to record the match
1857        start in case it was changed by \K. */
1858    
1859      if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1860             *prev == OP_ONCE_NC)
1861        {        {
1862        number = *prev - OP_BRA;        md->end_match_ptr = eptr;      /* For ONCE_NC */
1863          md->end_offset_top = offset_top;
1864          md->start_match_ptr = mstart;
1865          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1866          }
1867    
1868        /* For extended extraction brackets (large number), we have to fish out      /* For capturing groups we have to check the group number back at the start
1869        the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1870        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1871        into group 0, so it won't be picked up here. Instead, we catch it when the
1872        OP_END is reached. Other recursion is handled here. We just have to record
1873        the current subject position and start match pointer and give a MATCH
1874        return. */
1875    
1876        if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1877            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1878          {
1879          number = GET2(prev, 1+LINK_SIZE);
1880        offset = number << 1;        offset = number << 1;
1881    
1882  #ifdef DEBUG  #ifdef PCRE_DEBUG
1883        printf("end bracket %d", number);        printf("end bracket %d", number);
1884        printf("\n");        printf("\n");
1885  #endif  #endif
1886    
1887        /* Test for a numbered group. This includes groups called as a result        /* Handle a recursively called group. */
       of recursion. Note that whole-pattern recursion is coded as a recurse  
       into group 0, so it won't be picked up here. Instead, we catch it when  
       the OP_END is reached. */  
1888    
1889        if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1890          {          {
1891          md->capture_last = number;          md->end_match_ptr = eptr;
1892          if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1893            {          RRETURN(MATCH_MATCH);
1894            md->offset_vector[offset] =          }
1895              md->offset_vector[md->offset_end - number];  
1896            md->offset_vector[offset+1] = eptr - md->start_subject;        /* Deal with capturing */
           if (offset_top <= offset) offset_top = offset + 2;  
           }  
1897    
1898          /* Handle a recursively called group. Restore the offsets        md->capture_last = number;
1899          appropriately and continue from after the call. */        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1900            {
1901            /* If offset is greater than offset_top, it means that we are
1902            "skipping" a capturing group, and that group's offsets must be marked
1903            unset. In earlier versions of PCRE, all the offsets were unset at the
1904            start of matching, but this doesn't work because atomic groups and
1905            assertions can cause a value to be set that should later be unset.
1906            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1907            part of the atomic group, but this is not on the final matching path,
1908            so must be unset when 2 is set. (If there is no group 2, there is no
1909            problem, because offset_top will then be 2, indicating no capture.) */
1910    
1911          if (md->recursive != NULL && md->recursive->group_num == number)          if (offset > offset_top)
1912            {            {
1913            recursion_info *rec = md->recursive;            register int *iptr = md->offset_vector + offset_top;
1914            DPRINTF(("Recursion (%d) succeeded - continuing\n", number));            register int *iend = md->offset_vector + offset;
1915            md->recursive = rec->prevrec;            while (iptr < iend) *iptr++ = -1;
           md->start_match = rec->save_start;  
           memcpy(md->offset_vector, rec->offset_save,  
             rec->saved_max * sizeof(int));  
           ecode = rec->after_call;  
           ims = original_ims;  
           break;  
1916            }            }
         }  
       }  
1917    
1918      /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
     the group. */  
1919    
1920      ims = original_ims;          md->offset_vector[offset] =
1921      DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1922            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1923            if (offset_top <= offset) offset_top = offset + 2;
1924            }
1925          }
1926    
1927      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1928      happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1929      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1930      5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group that includes captures,
1931      course of events. */      establish a backup point by processing the rest of the pattern at a lower
1932        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1933        original OP_ONCE level, thereby bypassing intermediate backup points, but
1934        resetting any captures that happened along the way. */
1935    
1936      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1937        {        {
1938        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1939            {
1940            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1941            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1942            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1943            RRETURN(MATCH_ONCE);
1944            }
1945          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1946        break;        break;
1947        }        }
1948    
1949      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1950      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1951      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
1952    
1953        if (*ecode == OP_KETRPOS)
1954          {
1955          md->end_match_ptr = eptr;
1956          md->end_offset_top = offset_top;
1957          RRETURN(MATCH_KETRPOS);
1958          }
1959    
1960        /* The normal repeating kets try the rest of the pattern or restart from
1961        the preceding bracket, in the appropriate order. In the second case, we can
1962        use tail recursion to avoid using another stack frame, unless we have an
1963        an atomic group or an unlimited repeat of a group that can match an empty
1964        string. */
1965    
1966      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1967        {        {
1968        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1969        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1970          if (*prev == OP_ONCE)
1971            {
1972            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1973            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1974            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1975            RRETURN(MATCH_ONCE);
1976            }
1977          if (*prev >= OP_SBRA)    /* Could match an empty string */
1978            {
1979            md->match_function_type = MATCH_CBEGROUP;
1980            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1981            RRETURN(rrc);
1982            }
1983        ecode = prev;        ecode = prev;
       flags = match_isgroup;  
1984        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1985        }        }
1986      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1987        {        {
1988        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1989          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1990          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1991        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1992          if (*prev == OP_ONCE)
1993            {
1994            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1995            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1996            md->once_target = prev;
1997            RRETURN(MATCH_ONCE);
1998            }
1999        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2000        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2001        }        }
2002      /* Control never gets here */      /* Control never gets here */
2003    
2004      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2005    
2006      case OP_CIRC:      case OP_CIRC:
2007      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 ||  
            eptr < md->start_subject + md->nllen ||  
            !IS_NEWLINE(eptr - md->nllen)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2008    
2009      /* Start of subject assertion */      /* Start of subject assertion */
2010    
# Line 1218  for (;;) Line 2013  for (;;)
2013      ecode++;      ecode++;
2014      break;      break;
2015    
2016        /* Multiline mode: start of subject unless notbol, or after any newline. */
2017    
2018        case OP_CIRCM:
2019        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2020        if (eptr != md->start_subject &&
2021            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2022          RRETURN(MATCH_NOMATCH);
2023        ecode++;
2024        break;
2025    
2026      /* Start of match assertion */      /* Start of match assertion */
2027    
2028      case OP_SOM:      case OP_SOM:
# Line 1225  for (;;) Line 2030  for (;;)
2030      ecode++;      ecode++;
2031      break;      break;
2032    
2033      /* 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. */  
2034    
2035      case OP_DOLL:      case OP_SET_SOM:
2036      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2037        {      ecode++;
2038        if (eptr < md->end_subject)      break;
2039          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
2040        else      /* Multiline mode: assert before any newline, or before end of subject
2041          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
2042        ecode++;  
2043        break;      case OP_DOLLM:
2044        }      if (eptr < md->end_subject)
2045          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
2046      else      else
2047        {        {
2048        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2049        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2050        }        }
2051        ecode++;
2052        break;
2053    
2054        /* Not multiline mode: assert before a terminating newline or before end of
2055        subject unless noteol is set. */
2056    
2057        case OP_DOLL:
2058        if (md->noteol) RRETURN(MATCH_NOMATCH);
2059        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2060    
2061      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2062    
2063      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2064    
2065      case OP_EOD:      case OP_EOD:
2066      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2067        SCHECK_PARTIAL();
2068      ecode++;      ecode++;
2069      break;      break;
2070    
2071      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2072    
2073      case OP_EODN:      case OP_EODN:
2074      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2075          (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))      if (eptr < md->end_subject &&
2076            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2077        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2078    
2079        /* Either at end of string or \n before end. */
2080    
2081        SCHECK_PARTIAL();
2082      ecode++;      ecode++;
2083      break;      break;
2084    
# Line 1276  for (;;) Line 2090  for (;;)
2090    
2091        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2092        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
2093        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2094          partial matching. */
2095    
2096  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2097        if (utf8)        if (utf)
2098          {          {
2099            /* Get status of previous character */
2100    
2101          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2102            {            {
2103            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2104            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2105              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2106            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2107    #ifdef SUPPORT_UCP
2108              if (md->use_ucp)
2109                {
2110                if (c == '_') prev_is_word = TRUE; else
2111                  {
2112                  int cat = UCD_CATEGORY(c);
2113                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2114                  }
2115                }
2116              else
2117    #endif
2118            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2119            }            }
2120          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2121            /* Get status of next character */
2122    
2123            if (eptr >= md->end_subject)
2124              {
2125              SCHECK_PARTIAL();
2126              cur_is_word = FALSE;
2127              }
2128            else
2129            {            {
2130            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2131    #ifdef SUPPORT_UCP
2132              if (md->use_ucp)
2133                {
2134                if (c == '_') cur_is_word = TRUE; else
2135                  {
2136                  int cat = UCD_CATEGORY(c);
2137                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2138                  }
2139                }
2140              else
2141    #endif
2142            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2143            }            }
2144          }          }
2145        else        else
2146  #endif  #endif
2147    
2148        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2149          consistency with the behaviour of \w we do use it in this case. */
2150    
2151          {          {
2152          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2153            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2154          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2155            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2156              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2157    #ifdef SUPPORT_UCP
2158              if (md->use_ucp)
2159                {
2160                c = eptr[-1];
2161                if (c == '_') prev_is_word = TRUE; else
2162                  {
2163                  int cat = UCD_CATEGORY(c);
2164                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2165                  }
2166                }
2167              else
2168    #endif
2169              prev_is_word = MAX_255(eptr[-1])
2170                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2171              }
2172    
2173            /* Get status of next character */
2174    
2175            if (eptr >= md->end_subject)
2176              {
2177              SCHECK_PARTIAL();
2178              cur_is_word = FALSE;
2179              }
2180            else
2181    #ifdef SUPPORT_UCP
2182            if (md->use_ucp)
2183              {
2184              c = *eptr;
2185              if (c == '_') cur_is_word = TRUE; else
2186                {
2187                int cat = UCD_CATEGORY(c);
2188                cur_is_word = (cat == ucp_L || cat == ucp_N);
2189                }
2190              }
2191            else
2192    #endif
2193            cur_is_word = MAX_255(*eptr)
2194              && ((md->ctypes[*eptr] & ctype_word) != 0);
2195          }          }
2196    
2197        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1317  for (;;) Line 2205  for (;;)
2205      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2206    
2207      case OP_ANY:      case OP_ANY:
2208      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2209        {      /* Fall through */
2210        if (eptr <= md->end_subject - md->nllen && IS_NEWLINE(eptr))  
2211          RRETURN(MATCH_NOMATCH);      case OP_ALLANY:
2212        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2213          {                            /* not be updated before SCHECK_PARTIAL. */
2214          SCHECK_PARTIAL();
2215          RRETURN(MATCH_NOMATCH);
2216        }        }
2217      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2218      if (utf8)  #ifdef SUPPORT_UTF
2219        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2220    #endif
2221      ecode++;      ecode++;
2222      break;      break;
2223    
# Line 1332  for (;;) Line 2225  for (;;)
2225      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2226    
2227      case OP_ANYBYTE:      case OP_ANYBYTE:
2228      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2229          {                            /* not be updated before SCHECK_PARTIAL. */
2230          SCHECK_PARTIAL();
2231          RRETURN(MATCH_NOMATCH);
2232          }
2233        eptr++;
2234      ecode++;      ecode++;
2235      break;      break;
2236    
2237      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2238      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2239          {
2240          SCHECK_PARTIAL();
2241          RRETURN(MATCH_NOMATCH);
2242          }
2243      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2244      if (      if (
2245  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2246         c < 256 &&         c < 256 &&
2247  #endif  #endif
2248         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1350  for (;;) Line 2252  for (;;)
2252      break;      break;
2253    
2254      case OP_DIGIT:      case OP_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 > 255 ||
2264  #endif  #endif
2265         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2266         )         )
# Line 1363  for (;;) Line 2269  for (;;)
2269      break;      break;
2270    
2271      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
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 < 256 &&
2281  #endif  #endif
2282         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1376  for (;;) Line 2286  for (;;)
2286      break;      break;
2287    
2288      case OP_WHITESPACE:      case OP_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 > 255 ||
2298  #endif  #endif
2299         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2300         )         )
# Line 1389  for (;;) Line 2303  for (;;)
2303      break;      break;
2304    
2305      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2306      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2307          {
2308          SCHECK_PARTIAL();
2309          RRETURN(MATCH_NOMATCH);
2310          }
2311      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2312      if (      if (
2313  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2314         c < 256 &&         c < 256 &&
2315  #endif  #endif
2316         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1402  for (;;) Line 2320  for (;;)
2320      break;      break;
2321    
2322      case OP_WORDCHAR:      case OP_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 > 255 ||
2332  #endif  #endif
2333         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2334         )         )
# Line 1414  for (;;) Line 2336  for (;;)
2336      ecode++;      ecode++;
2337      break;      break;
2338    
2339        case OP_ANYNL:
2340        if (eptr >= md->end_subject)
2341          {
2342          SCHECK_PARTIAL();
2343          RRETURN(MATCH_NOMATCH);
2344          }
2345        GETCHARINCTEST(c, eptr);
2346        switch(c)
2347          {
2348          default: RRETURN(MATCH_NOMATCH);
2349    
2350          case 0x000d:
2351          if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2352          break;
2353    
2354          case 0x000a:
2355          break;
2356    
2357          case 0x000b:
2358          case 0x000c:
2359          case 0x0085:
2360          case 0x2028:
2361          case 0x2029:
2362          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2363          break;
2364          }
2365        ecode++;
2366        break;
2367    
2368        case OP_NOT_HSPACE:
2369        if (eptr >= md->end_subject)
2370          {
2371          SCHECK_PARTIAL();
2372          RRETURN(MATCH_NOMATCH);
2373          }
2374        GETCHARINCTEST(c, eptr);
2375        switch(c)
2376          {
2377          default: break;
2378          case 0x09:      /* HT */
2379          case 0x20:      /* SPACE */
2380          case 0xa0:      /* NBSP */
2381          case 0x1680:    /* OGHAM SPACE MARK */
2382          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2383          case 0x2000:    /* EN QUAD */
2384          case 0x2001:    /* EM QUAD */
2385          case 0x2002:    /* EN SPACE */
2386          case 0x2003:    /* EM SPACE */
2387          case 0x2004:    /* THREE-PER-EM SPACE */
2388          case 0x2005:    /* FOUR-PER-EM SPACE */
2389          case 0x2006:    /* SIX-PER-EM SPACE */
2390          case 0x2007:    /* FIGURE SPACE */
2391          case 0x2008:    /* PUNCTUATION SPACE */
2392          case 0x2009:    /* THIN SPACE */
2393          case 0x200A:    /* HAIR SPACE */
2394          case 0x202f:    /* NARROW NO-BREAK SPACE */
2395          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2396          case 0x3000:    /* IDEOGRAPHIC SPACE */
2397          RRETURN(MATCH_NOMATCH);
2398          }
2399        ecode++;
2400        break;
2401    
2402        case OP_HSPACE:
2403        if (eptr >= md->end_subject)
2404          {
2405          SCHECK_PARTIAL();
2406          RRETURN(MATCH_NOMATCH);
2407          }
2408        GETCHARINCTEST(c, eptr);
2409        switch(c)
2410          {
2411          default: RRETURN(MATCH_NOMATCH);
2412          case 0x09:      /* HT */
2413          case 0x20:      /* SPACE */
2414          case 0xa0:      /* NBSP */
2415          case 0x1680:    /* OGHAM SPACE MARK */
2416          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2417          case 0x2000:    /* EN QUAD */
2418          case 0x2001:    /* EM QUAD */
2419          case 0x2002:    /* EN SPACE */
2420          case 0x2003:    /* EM SPACE */
2421          case 0x2004:    /* THREE-PER-EM SPACE */
2422          case 0x2005:    /* FOUR-PER-EM SPACE */
2423          case 0x2006:    /* SIX-PER-EM SPACE */
2424          case 0x2007:    /* FIGURE SPACE */
2425          case 0x2008:    /* PUNCTUATION SPACE */
2426          case 0x2009:    /* THIN SPACE */
2427          case 0x200A:    /* HAIR SPACE */
2428          case 0x202f:    /* NARROW NO-BREAK SPACE */
2429          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2430          case 0x3000:    /* IDEOGRAPHIC SPACE */
2431          break;
2432          }
2433        ecode++;
2434        break;
2435    
2436        case OP_NOT_VSPACE:
2437        if (eptr >= md->end_subject)
2438          {
2439          SCHECK_PARTIAL();
2440          RRETURN(MATCH_NOMATCH);
2441          }
2442        GETCHARINCTEST(c, eptr);
2443        switch(c)
2444          {
2445          default: break;
2446          case 0x0a:      /* LF */
2447          case 0x0b:      /* VT */
2448          case 0x0c:      /* FF */
2449          case 0x0d:      /* CR */
2450          case 0x85:      /* NEL */
2451          case 0x2028:    /* LINE SEPARATOR */
2452          case 0x2029:    /* PARAGRAPH SEPARATOR */
2453          RRETURN(MATCH_NOMATCH);
2454          }
2455        ecode++;
2456        break;
2457    
2458        case OP_VSPACE:
2459        if (eptr >= md->end_subject)
2460          {
2461          SCHECK_PARTIAL();
2462          RRETURN(MATCH_NOMATCH);
2463          }
2464        GETCHARINCTEST(c, eptr);
2465        switch(c)
2466          {
2467          default: RRETURN(MATCH_NOMATCH);
2468          case 0x0a:      /* LF */
2469          case 0x0b:      /* VT */
2470          case 0x0c:      /* FF */
2471          case 0x0d:      /* CR */
2472          case 0x85:      /* NEL */
2473          case 0x2028:    /* LINE SEPARATOR */
2474          case 0x2029:    /* PARAGRAPH SEPARATOR */
2475          break;
2476          }
2477        ecode++;
2478        break;
2479    
2480  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2481      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2482      if the support is in the binary; otherwise a compile-time error occurs. */      if the support is in the binary; otherwise a compile-time error occurs. */
2483    
2484      case OP_PROP:      case OP_PROP:
2485      case OP_NOTPROP:      case OP_NOTPROP:
2486      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2487          {
2488          SCHECK_PARTIAL();
2489          RRETURN(MATCH_NOMATCH);
2490          }
2491      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2492        {        {
2493        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2494    
2495        switch(ecode[1])        switch(ecode[1])
2496          {          {
# Line 1433  for (;;) Line 2499  for (;;)
2499          break;          break;
2500    
2501          case PT_LAMP:          case PT_LAMP:
2502          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2503               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2504               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2505            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2506           break;          break;
2507    
2508          case PT_GC:          case PT_GC:
2509          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2510            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2511          break;          break;
2512    
2513          case PT_PC:          case PT_PC:
2514          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2515            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2516          break;          break;
2517    
2518          case PT_SC:          case PT_SC:
2519          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2520            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2521          break;          break;
2522    
2523            /* These are specials */
2524    
2525            case PT_ALNUM:
2526            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2527                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2528              RRETURN(MATCH_NOMATCH);
2529            break;
2530    
2531            case PT_SPACE:    /* Perl space */
2532            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2533                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2534                   == (op == OP_NOTPROP))
2535              RRETURN(MATCH_NOMATCH);
2536            break;
2537    
2538            case PT_PXSPACE:  /* POSIX space */
2539            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2540                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2541                 c == CHAR_FF || c == CHAR_CR)
2542                   == (op == OP_NOTPROP))
2543              RRETURN(MATCH_NOMATCH);
2544            break;
2545    
2546            case PT_WORD:
2547            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2548                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2549                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2550              RRETURN(MATCH_NOMATCH);
2551            break;
2552    
2553            /* This should never occur */
2554    
2555          default:          default:
2556          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
         break;  
2557          }          }
2558    
2559        ecode += 3;        ecode += 3;
# Line 1467  for (;;) Line 2564  for (;;)
2564      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2565    
2566      case OP_EXTUNI:      case OP_EXTUNI:
2567      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2568          {
2569          SCHECK_PARTIAL();
2570          RRETURN(MATCH_NOMATCH);
2571          }
2572      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2573        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2574        while (eptr < md->end_subject)
2575        {        {
2576        int chartype, script;        int len = 1;
2577        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2578        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2579        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;  
         }  
2580        }        }
2581      ecode++;      ecode++;
2582      break;      break;
# Line 1499  for (;;) Line 2592  for (;;)
2592      loops). */      loops). */
2593    
2594      case OP_REF:      case OP_REF:
2595        {      case OP_REFI:
2596        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2597        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2598        ecode += 1 + IMM2_SIZE;
2599    
2600        /* 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];  
2601    
2602        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2603        this ensures that every attempt at a match fails. We can't just fail
2604        here, because of the possibility of quantifiers with zero minima.
2605    
2606        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2607          {      so that the back reference matches an empty string.
2608          case OP_CRSTAR:  
2609          case OP_CRMINSTAR:      Otherwise, set the length to the length of what was matched by the
2610          case OP_CRPLUS:      referenced subpattern. */
         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;  
2611    
2612          case OP_CRRANGE:      if (offset >= offset_top || md->offset_vector[offset] < 0)
2613          case OP_CRMINRANGE:        length = (md->jscript_compat)? 0 : -1;
2614          minimize = (*ecode == OP_CRMINRANGE);      else
2615          min = GET2(ecode, 1);        length = md->offset_vector[offset+1] - md->offset_vector[offset];
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2616    
2617          default:               /* No repeat follows */      /* Set up for repetition, or handle the non-repeated case */
2618          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);  
2619          eptr += length;      switch (*ecode)
2620          continue;              /* With the main loop */        {
2621          case OP_CRSTAR:
2622          case OP_CRMINSTAR:
2623          case OP_CRPLUS:
2624          case OP_CRMINPLUS:
2625          case OP_CRQUERY:
2626          case OP_CRMINQUERY:
2627          c = *ecode++ - OP_CRSTAR;
2628          minimize = (c & 1) != 0;
2629          min = rep_min[c];                 /* Pick up values from tables; */
2630          max = rep_max[c];                 /* zero for max => infinity */
2631          if (max == 0) max = INT_MAX;
2632          break;
2633    
2634          case OP_CRRANGE:
2635          case OP_CRMINRANGE:
2636          minimize = (*ecode == OP_CRMINRANGE);
2637          min = GET2(ecode, 1);
2638          max = GET2(ecode, 1 + IMM2_SIZE);
2639          if (max == 0) max = INT_MAX;
2640          ecode += 1 + 2 * IMM2_SIZE;
2641          break;
2642    
2643          default:               /* No repeat follows */
2644          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2645            {
2646            CHECK_PARTIAL();
2647            RRETURN(MATCH_NOMATCH);
2648          }          }
2649          eptr += length;
2650          continue;              /* With the main loop */
2651          }
2652    
2653        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2654        main loop. */      zero, just continue with the main loop. If the length is negative, it
2655        means the reference is unset in non-Java-compatible mode. If the minimum is
2656        zero, we can continue at the same level without recursion. For any other
2657        minimum, carrying on will result in NOMATCH. */
2658    
2659        if (length == 0) continue;      if (length == 0) continue;
2660        if (length < 0 && min == 0) continue;
2661    
2662        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2663        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2664        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2665    
2666        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2667          {
2668          int slength;
2669          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2670          {          {
2671          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2672          eptr += length;          RRETURN(MATCH_NOMATCH);
2673          }          }
2674          eptr += slength;
2675          }
2676    
2677        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2678        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2679    
2680        if (min == max) continue;      if (min == max) continue;
2681    
2682        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2683    
2684        if (minimize)      if (minimize)
2685          {
2686          for (fi = min;; fi++)
2687          {          {
2688          for (fi = min;; fi++)          int slength;
2689            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2690            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2691            if (fi >= max) RRETURN(MATCH_NOMATCH);
2692            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2693            {            {
2694            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2695            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            RRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2696            }            }
2697          /* Control never gets here */          eptr += slength;
2698          }          }
2699          /* Control never gets here */
2700          }
2701    
2702        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2703    
2704        else      else
2705          {
2706          pp = eptr;
2707          for (i = min; i < max; i++)
2708          {          {
2709          pp = eptr;          int slength;
2710          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)  
2711            {            {
2712            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2713            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2714            }            }
2715          RRETURN(MATCH_NOMATCH);          eptr += slength;
2716            }
2717          while (eptr >= pp)
2718            {
2719            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2720            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2721            eptr -= length;
2722          }          }
2723          RRETURN(MATCH_NOMATCH);
2724        }        }
2725      /* Control never gets here */      /* Control never gets here */
2726    
   
   
2727      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2728      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,
2729      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 1616  for (;;) Line 2738  for (;;)
2738      case OP_NCLASS:      case OP_NCLASS:
2739      case OP_CLASS:      case OP_CLASS:
2740        {        {
2741          /* The data variable is saved across frames, so the byte map needs to
2742          be stored there. */
2743    #define BYTE_MAP ((pcre_uint8 *)data)
2744        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2745        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2746    
2747        switch (*ecode)        switch (*ecode)
2748          {          {
# Line 1638  for (;;) Line 2763  for (;;)
2763          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2764          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2765          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2766          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2767          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2768          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2769          break;          break;
2770    
2771          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1650  for (;;) Line 2775  for (;;)
2775    
2776        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2777    
2778  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2779        /* UTF-8 mode */        if (utf)
       if (utf8)  
2780          {          {
2781          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2782            {            {
2783            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2784                {
2785                SCHECK_PARTIAL();
2786                RRETURN(MATCH_NOMATCH);
2787                }
2788            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2789            if (c > 255)            if (c > 255)
2790              {              {
2791              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2792              }              }
2793            else            else
2794              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2795            }            }
2796          }          }
2797        else        else
2798  #endif  #endif
2799        /* Not UTF-8 mode */        /* Not UTF mode */
2800          {          {
2801          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2802            {            {
2803            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2804                {
2805                SCHECK_PARTIAL();
2806                RRETURN(MATCH_NOMATCH);
2807                }
2808            c = *eptr++;            c = *eptr++;
2809            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2810              if (c > 255)
2811                {
2812                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2813                }
2814              else
2815    #endif
2816                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2817            }            }
2818          }          }
2819    
# Line 1690  for (;;) Line 2827  for (;;)
2827    
2828        if (minimize)        if (minimize)
2829          {          {
2830  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2831          /* UTF-8 mode */          if (utf)
         if (utf8)  
2832            {            {
2833            for (fi = min;; fi++)            for (fi = min;; fi++)
2834              {              {
2835              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2836              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2837              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2838                if (eptr >= md->end_subject)
2839                  {
2840                  SCHECK_PARTIAL();
2841                  RRETURN(MATCH_NOMATCH);
2842                  }
2843              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2844              if (c > 255)              if (c > 255)
2845                {                {
2846                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2847                }                }
2848              else              else
2849                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2850              }              }
2851            }            }
2852          else          else
2853  #endif  #endif
2854          /* Not UTF-8 mode */          /* Not UTF mode */
2855            {            {
2856            for (fi = min;; fi++)            for (fi = min;; fi++)
2857              {              {
2858              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2859              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2860              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2861                if (eptr >= md->end_subject)
2862                  {
2863                  SCHECK_PARTIAL();
2864                  RRETURN(MATCH_NOMATCH);
2865                  }
2866              c = *eptr++;              c = *eptr++;
2867              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2868                if (c > 255)
2869                  {
2870                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2871                  }
2872                else
2873    #endif
2874                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2875              }              }
2876            }            }
2877          /* Control never gets here */          /* Control never gets here */
# Line 1732  for (;;) Line 2883  for (;;)
2883          {          {
2884          pp = eptr;          pp = eptr;
2885    
2886  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2887          /* UTF-8 mode */          if (utf)
         if (utf8)  
2888            {            {
2889            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2890              {              {
2891              int len = 1;              int len = 1;
2892              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2893                  {
2894                  SCHECK_PARTIAL();
2895                  break;
2896                  }
2897              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2898              if (c > 255)              if (c > 255)
2899                {                {
2900                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2901                }                }
2902              else              else
2903                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2904              eptr += len;              eptr += len;
2905              }              }
2906            for (;;)            for (;;)
2907              {              {
2908              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2909              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2910              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2911              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1761  for (;;) Line 2913  for (;;)
2913            }            }
2914          else          else
2915  #endif  #endif
2916            /* Not UTF-8 mode */            /* Not UTF mode */
2917            {            {
2918            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2919              {              {
2920              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2921                  {
2922                  SCHECK_PARTIAL();
2923                  break;
2924                  }
2925              c = *eptr;              c = *eptr;
2926              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2927                if (c > 255)
2928                  {
2929                  if (op == OP_CLASS) break;
2930                  }
2931                else
2932    #endif
2933                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2934              eptr++;              eptr++;
2935              }              }
2936            while (eptr >= pp)            while (eptr >= pp)
2937              {              {
2938              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2939              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2940              eptr--;              eptr--;
2941              }              }
# Line 1780  for (;;) Line 2943  for (;;)
2943    
2944          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2945          }          }
2946    #undef BYTE_MAP
2947        }        }
2948      /* Control never gets here */      /* Control never gets here */
2949    
2950    
2951      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2952      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
2953        mode, because Unicode properties are supported in non-UTF-8 mode. */
2954    
2955  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2956      case OP_XCLASS:      case OP_XCLASS:
2957        {        {
2958        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1812  for (;;) Line 2977  for (;;)
2977          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2978          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2979          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2980          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2981          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2982          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2983          break;          break;
2984    
2985          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1826  for (;;) Line 2991  for (;;)
2991    
2992        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2993          {          {
2994          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2995          GETCHARINC(c, eptr);            {
2996          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2997              RRETURN(MATCH_NOMATCH);
2998              }
2999            GETCHARINCTEST(c, eptr);
3000            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3001          }          }
3002    
3003        /* 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 1843  for (;;) Line 3012  for (;;)
3012          {          {
3013          for (fi = min;; fi++)          for (fi = min;; fi++)
3014            {            {
3015            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3016            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3017            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3018            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3019            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3020                SCHECK_PARTIAL();
3021                RRETURN(MATCH_NOMATCH);
3022                }
3023              GETCHARINCTEST(c, eptr);
3024              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3025            }            }
3026          /* Control never gets here */          /* Control never gets here */
3027          }          }
# Line 1860  for (;;) Line 3034  for (;;)
3034          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3035            {            {
3036            int len = 1;            int len = 1;
3037            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3038            GETCHARLEN(c, eptr, len);              {
3039            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3040                break;
3041                }
3042    #ifdef SUPPORT_UTF
3043              GETCHARLENTEST(c, eptr, len);
3044    #else
3045              c = *eptr;
3046    #endif
3047              if (!PRIV(xclass)(c, data, utf)) break;
3048            eptr += len;            eptr += len;
3049            }            }
3050          for(;;)          for(;;)
3051            {            {
3052            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3053            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3054            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3055            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3056              if (utf) BACKCHAR(eptr);
3057    #endif
3058            }            }
3059          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3060          }          }
# Line 1882  for (;;) Line 3066  for (;;)
3066      /* Match a single character, casefully */      /* Match a single character, casefully */
3067    
3068      case OP_CHAR:      case OP_CHAR:
3069  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3070      if (utf8)      if (utf)
3071        {        {
3072        length = 1;        length = 1;
3073        ecode++;        ecode++;
3074        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3075        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3076            {
3077            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3078            RRETURN(MATCH_NOMATCH);
3079            }
3080        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3081        }        }
3082      else      else
3083  #endif  #endif
3084        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3085        {        {
3086        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3087            {
3088            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3089            RRETURN(MATCH_NOMATCH);
3090            }
3091        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3092        ecode += 2;        ecode += 2;
3093        }        }
3094      break;      break;
3095    
3096      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3097        subject, give up immediately. */
3098    
3099        case OP_CHARI:
3100        if (eptr >= md->end_subject)
3101          {
3102          SCHECK_PARTIAL();
3103          RRETURN(MATCH_NOMATCH);
3104          }
3105    
3106      case OP_CHARNC:  #ifdef SUPPORT_UTF
3107  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3108        {        {
3109        length = 1;        length = 1;
3110        ecode++;        ecode++;
3111        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3112    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3113        /* 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
3114        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3115          fast lookup table. We know that there is at least one byte left in the
3116          subject. */
3117    
3118        if (fc < 128)        if (fc < 128)
3119          {          {
3120          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3121                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3122            ecode++;
3123            eptr++;
3124          }          }
3125    
3126        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3127          use the value of "length" to check for sufficient bytes left, because the
3128          other case of the character may have more or fewer bytes.  */
3129    
3130        else        else
3131          {          {
3132          int dc;          unsigned int dc;
3133          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3134          ecode += length;          ecode += length;
3135    
# Line 1936  for (;;) Line 3139  for (;;)
3139          if (fc != dc)          if (fc != dc)
3140            {            {
3141  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3142            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3143  #endif  #endif
3144              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3145            }            }
3146          }          }
3147        }        }
3148      else      else
3149  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3150    
3151      /* Non-UTF-8 mode */      /* Not UTF mode */
3152        {        {
3153        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3154        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3155          eptr++;
3156        ecode += 2;        ecode += 2;
3157        }        }
3158      break;      break;
3159    
3160      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3161    
3162      case OP_EXACT:      case OP_EXACT:
3163        case OP_EXACTI:
3164      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3165      ecode += 3;      ecode += 1 + IMM2_SIZE;
3166      goto REPEATCHAR;      goto REPEATCHAR;
3167    
3168        case OP_POSUPTO:
3169        case OP_POSUPTOI:
3170        possessive = TRUE;
3171        /* Fall through */
3172    
3173      case OP_UPTO:      case OP_UPTO:
3174        case OP_UPTOI:
3175      case OP_MINUPTO:      case OP_MINUPTO:
3176        case OP_MINUPTOI:
3177      min = 0;      min = 0;
3178      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3179      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3180      ecode += 3;      ecode += 1 + IMM2_SIZE;
3181        goto REPEATCHAR;
3182    
3183        case OP_POSSTAR:
3184        case OP_POSSTARI:
3185        possessive = TRUE;
3186        min = 0;
3187        max = INT_MAX;
3188        ecode++;
3189        goto REPEATCHAR;
3190    
3191        case OP_POSPLUS:
3192        case OP_POSPLUSI:
3193        possessive = TRUE;
3194        min = 1;
3195        max = INT_MAX;
3196        ecode++;
3197        goto REPEATCHAR;
3198    
3199        case OP_POSQUERY:
3200        case OP_POSQUERYI:
3201        possessive = TRUE;
3202        min = 0;
3203        max = 1;
3204        ecode++;
3205      goto REPEATCHAR;      goto REPEATCHAR;
3206    
3207      case OP_STAR:      case OP_STAR:
3208        case OP_STARI:
3209      case OP_MINSTAR:      case OP_MINSTAR:
3210        case OP_MINSTARI:
3211      case OP_PLUS:      case OP_PLUS:
3212        case OP_PLUSI:
3213      case OP_MINPLUS:      case OP_MINPLUS:
3214        case OP_MINPLUSI:
3215      case OP_QUERY:      case OP_QUERY:
3216        case OP_QUERYI:
3217      case OP_MINQUERY:      case OP_MINQUERY:
3218      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3219        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3220      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3221      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3222      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3223      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3224    
3225      /* 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. */  
3226    
3227      REPEATCHAR:      REPEATCHAR:
3228  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3229      if (utf8)      if (utf)
3230        {        {
3231        length = 1;        length = 1;
3232        charptr = ecode;        charptr = ecode;
3233        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3234        ecode += length;        ecode += length;
3235    
3236        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1999  for (;;) Line 3238  for (;;)
3238    
3239        if (length > 1)        if (length > 1)
3240          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3241  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3242          int othercase;          unsigned int othercase;
3243          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3244              (othercase = _pcre_ucp_othercase(fc)) >= 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
3245               othercase >= 0)            oclength = PRIV(ord2utf)(othercase, occhars);
3246            oclength = _pcre_ord2utf8(othercase, occhars);          else oclength = 0;
3247  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3248    
3249          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3250            {            {
3251            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3252            /* Need braces because of following else */              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3253            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3254              else if (oclength > 0 &&
3255                       eptr <= md->end_subject - oclength &&
3256                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3257    #endif  /* SUPPORT_UCP */
3258            else            else
3259              {              {
3260              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3261              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3262              }              }
3263            }            }
3264    
# Line 2028  for (;;) Line 3268  for (;;)
3268            {            {
3269            for (fi = min;; fi++)            for (fi = min;; fi++)
3270              {              {
3271              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3272              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3273              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3274              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3275              /* Need braces because of following else */                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3276              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3277                else if (oclength > 0 &&
3278                         eptr <= md->end_subject - oclength &&
3279                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3280    #endif  /* SUPPORT_UCP */
3281              else              else
3282                {                {
3283                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3284                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3285                }                }
3286              }              }
3287            /* Control never gets here */            /* Control never gets here */
3288            }            }
3289          else  
3290            else  /* Maximize */
3291            {            {
3292            pp = eptr;            pp = eptr;
3293            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3294              {              {
3295              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3296              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3297              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3298                else if (oclength > 0 &&
3299                         eptr <= md->end_subject - oclength &&
3300                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3301    #endif  /* SUPPORT_UCP */
3302              else              else
3303                {                {
3304                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3305                eptr += oclength;                break;
3306                }                }
3307              }              }
3308            while (eptr >= pp)  
3309             {            if (possessive) continue;
3310             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3311             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3312             eptr -= length;              {
3313             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3314            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3315                if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3316    #ifdef SUPPORT_UCP
3317                eptr--;
3318                BACKCHAR(eptr);
3319    #else   /* without SUPPORT_UCP */
3320                eptr -= length;
3321    #endif  /* SUPPORT_UCP */
3322                }
3323            }            }
3324          /* Control never gets here */          /* Control never gets here */
3325          }          }
# Line 2072  for (;;) Line 3329