/[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 963 by ph10, Sat Apr 21 18:06:31 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 the number of subject bytes matched
144                  -1 no match
145                  -2 partial match; always given if at end subject
146  */  */
147    
148  static BOOL  static int
149  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
150    unsigned long int ims)    BOOL caseless)
151  {  {
152  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
153    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
154    
155  #ifdef DEBUG  #ifdef PCRE_DEBUG
156  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
157    printf("matching subject <null>");    printf("matching subject <null>");
158  else  else
# Line 146  pchars(p, length, FALSE, md); Line 165  pchars(p, length, FALSE, md);
165  printf("\n");  printf("\n");
166  #endif  #endif
167    
168  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
169    case the length is passed as zero). */
170    
171  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
172    
173  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
174    properly if Unicode properties are supported. Otherwise, we can check only
175    ASCII characters. */
176    
177  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
178    {    {
179    while (length-- > 0)  #ifdef SUPPORT_UTF
180      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
181      if (md->utf)
182        {
183        /* Match characters up to the end of the reference. NOTE: the number of
184        bytes matched may differ, because there are some characters whose upper and
185        lower case versions code as different numbers of bytes. For example, U+023A
186        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
187        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
188        the latter. It is important, therefore, to check the length along the
189        reference, not along the subject (earlier code did this wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193          {
194          int c, d;
195          if (eptr >= md->end_subject) return -2;   /* Partial match */
196          GETCHARINC(c, eptr);
197          GETCHARINC(d, p);
198          if (c != d && c != UCD_OTHERCASE(d)) return -1;
199          }
200        }
201      else
202    #endif
203    #endif
204    
205      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
206      is no UCP support. */
207        {
208        while (length-- > 0)
209          {
210          if (eptr >= md->end_subject) return -2;   /* Partial match */
211          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
212          p++;
213          eptr++;
214          }
215        }
216    }    }
217    
218    /* In the caseful case, we can just compare the bytes, whether or not we
219    are in UTF-8 mode. */
220    
221  else  else
222    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
223      while (length-- > 0)
224        {
225        if (eptr >= md->end_subject) return -2;   /* Partial match */
226        if (*p++ != *eptr++) return -1;
227        }
228      }
229    
230  return TRUE;  return (int)(eptr - eptr_start);
231  }  }
232    
233    
# Line 186  calls by keeping local variables that ne Line 253  calls by keeping local variables that ne
253  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
254  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
255  always used to.  always used to.
256    
257    The original heap-recursive code used longjmp(). However, it seems that this
258    can be very slow on some operating systems. Following a suggestion from Stan
259    Switzer, the use of longjmp() has been abolished, at the cost of having to
260    provide a unique number for each call to RMATCH. There is no way of generating
261    a sequence of numbers at compile time in C. I have given them names, to make
262    them stand out more clearly.
263    
264    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
265    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
266    tests. Furthermore, not using longjmp() means that local dynamic variables
267    don't have indeterminate values; this has meant that the frame size can be
268    reduced because the result can be "passed back" by straight setting of the
269    variable instead of being passed in the frame.
270  ****************************************************************************  ****************************************************************************
271  ***************************************************************************/  ***************************************************************************/
272    
273    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
274    below must be updated in sync.  */
275    
276    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
277           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
278           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
279           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
280           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
281           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
282           RM61,  RM62, RM63, RM64, RM65, RM66 };
283    
284  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
285  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
286    actually used in this definition. */
287    
288  #ifndef NO_RECURSE  #ifndef NO_RECURSE
289  #define REGISTER register  #define REGISTER register
290  #ifdef DEBUG  
291  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
292    #define RMATCH(ra,rb,rc,rd,re,rw) \
293    { \    { \
294    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
295    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
296    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
297    }    }
298  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 208  versions and production versions. */ Line 301  versions and production versions. */
301    return ra; \    return ra; \
302    }    }
303  #else  #else
304  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
305    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
306  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
307  #endif  #endif
308    
309  #else  #else
310    
311    
312  /* 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
313  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
314  match(), which never changes. */  argument of match(), which never changes. */
315    
316  #define REGISTER  #define REGISTER
317    
318  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
319    {\    {\
320    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
321    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL)\
     {\  
     newframe->Xeptr = ra;\  
     newframe->Xecode = rb;\  
     newframe->Xoffset_top = rc;\  
     newframe->Xims = re;\  
     newframe->Xeptrb = rf;\  
     newframe->Xflags = rg;\  
     newframe->Xrdepth = frame->Xrdepth + 1;\  
     newframe->Xprevframe = frame;\  
     frame = newframe;\  
     DPRINTF(("restarting from line %d\n", __LINE__));\  
     goto HEAP_RECURSE;\  
     }\  
   else\  
322      {\      {\
323      DPRINTF(("longjumped back to line %d\n", __LINE__));\      newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
324      frame = md->thisframe;\      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
325      rx = frame->Xresult;\      newframe->Xnextframe = NULL;\
326        frame->Xnextframe = newframe;\
327      }\      }\
328      frame->Xwhere = rw;\
329      newframe->Xeptr = ra;\
330      newframe->Xecode = rb;\
331      newframe->Xmstart = mstart;\
332      newframe->Xoffset_top = rc;\
333      newframe->Xeptrb = re;\
334      newframe->Xrdepth = frame->Xrdepth + 1;\
335      newframe->Xprevframe = frame;\
336      frame = newframe;\
337      DPRINTF(("restarting from line %d\n", __LINE__));\
338      goto HEAP_RECURSE;\
339      L_##rw:\
340      DPRINTF(("jumped back to line %d\n", __LINE__));\
341    }    }
342    
343  #define RRETURN(ra)\  #define RRETURN(ra)\
344    {\    {\
345    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
346    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
347    if (frame != NULL)\    if (frame != NULL)\
348      {\      {\
349      frame->Xresult = ra;\      rrc = ra;\
350      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
351      }\      }\
352    return ra;\    return ra;\
353    }    }
# Line 266  match(), which never changes. */ Line 357  match(), which never changes. */
357    
358  typedef struct heapframe {  typedef struct heapframe {
359    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
360      struct heapframe *Xnextframe;
361    
362    /* Function arguments that may change */    /* Function arguments that may change */
363    
364    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
365    const uschar *Xecode;    const pcre_uchar *Xecode;
366      PCRE_PUCHAR Xmstart;
367    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
368    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
369    unsigned int Xrdepth;    unsigned int Xrdepth;
370    
371    /* Function local variables */    /* Function local variables */
372    
373    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
374    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
375    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
376    const uschar *Xnext;  #endif
377    const uschar *Xpp;    PCRE_PUCHAR Xdata;
378    const uschar *Xprev;    PCRE_PUCHAR Xnext;
379    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
380      PCRE_PUCHAR Xprev;
381      PCRE_PUCHAR Xsaved_eptr;
382    
383    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
384    
385    BOOL Xcur_is_word;    BOOL Xcur_is_word;
386    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
387    BOOL Xprev_is_word;    BOOL Xprev_is_word;
388    
   unsigned long int Xoriginal_ims;  
   
389  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
390    int Xprop_type;    int Xprop_type;
391    int Xprop_value;    int Xprop_value;
392    int Xprop_fail_result;    int Xprop_fail_result;
393    int Xprop_category;    int Xoclength;
394    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_script;  
   int *Xprop_test_variable;  
395  #endif  #endif
396    
397      int Xcodelink;
398    int Xctype;    int Xctype;
399    int Xfc;    unsigned int Xfc;
400    int Xfi;    int Xfi;
401    int Xlength;    int Xlength;
402    int Xmax;    int Xmax;
# Line 321  typedef struct heapframe { Line 410  typedef struct heapframe {
410    
411    eptrblock Xnewptrb;    eptrblock Xnewptrb;
412    
413    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
414    
415    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
416    
417  } heapframe;  } heapframe;
418    
# Line 340  typedef struct heapframe { Line 428  typedef struct heapframe {
428  *         Match from current position            *  *         Match from current position            *
429  *************************************************/  *************************************************/
430    
431  /* 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  
432  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
433  same response.  same response. */
434    
435    /* These macros pack up tests that are used for partial matching, and which
436    appear several times in the code. We set the "hit end" flag if the pointer is
437    at the end of the subject and also past the start of the subject (i.e.
438    something has been matched). For hard partial matching, we then return
439    immediately. The second one is used when we already know we are past the end of
440    the subject. */
441    
442    #define CHECK_PARTIAL()\
443      if (md->partial != 0 && eptr >= md->end_subject && \
444          eptr > md->start_used_ptr) \
445        { \
446        md->hitend = TRUE; \
447        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
448        }
449    
450  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
451  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
452        { \
453        md->hitend = TRUE; \
454        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
455        }
456    
457    
458    /* Performance note: It might be tempting to extract commonly used fields from
459    the md structure (e.g. utf, end_subject) into individual variables to improve
460  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
461  made performance worse.  made performance worse.
462    
463  Arguments:  Arguments:
464     eptr        pointer in subject     eptr        pointer to current character in subject
465     ecode       position in code     ecode       pointer to current position in compiled code
466       mstart      pointer to the current match start position (can be modified
467                     by encountering \K)
468     offset_top  current top pointer     offset_top  current top pointer
469     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
470     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
471                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_isgroup - this is the start of a bracketed group  
472     rdepth      the recursion depth     rdepth      the recursion depth
473    
474  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
475                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
476                   a negative MATCH_xxx value for PRUNE, SKIP, etc
477                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
478                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
479  */  */
480    
481  static int  static int
482  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
483    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
484    int flags, unsigned int rdepth)    unsigned int rdepth)
485  {  {
486  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
487  so they can be ordinary variables in all cases. Mark them with "register"  so they can be ordinary variables in all cases. Mark some of them with
488  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
489    
490  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
491  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
492  register unsigned int  c;  /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
493  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
494    
495    BOOL minimize, possessive; /* Quantifier options */
496    BOOL caseless;
497    int condcode;
498    
499  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
500  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
501  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
502  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
503    the top-level on the stack rather than malloc-ing them all gives a performance
504    boost in many cases where there is not much "recursion". */
505    
506  #ifdef NO_RECURSE  #ifdef NO_RECURSE
507  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
508    
509  /* Copy in the original argument variables */  /* Copy in the original argument variables */
510    
511  frame->Xeptr = eptr;  frame->Xeptr = eptr;
512  frame->Xecode = ecode;  frame->Xecode = ecode;
513    frame->Xmstart = mstart;
514  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
515  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
516  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
517    
518  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 412  HEAP_RECURSE: Line 523  HEAP_RECURSE:
523    
524  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
525  #define ecode              frame->Xecode  #define ecode              frame->Xecode
526    #define mstart             frame->Xmstart
527  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
528  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
529  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
530    
531  /* Ditto for the local variables */  /* Ditto for the local variables */
532    
533  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
534  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
535  #endif  #endif
536  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
537    #define codelink           frame->Xcodelink
538  #define data               frame->Xdata  #define data               frame->Xdata
539  #define next               frame->Xnext  #define next               frame->Xnext
540  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 434  HEAP_RECURSE: Line 545  HEAP_RECURSE:
545    
546  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
547  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
548  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
549    
 #define original_ims       frame->Xoriginal_ims  
   
550  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
551  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
552  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
553  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
554  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
555  #define prop_chartype      frame->Xprop_chartype  #define occhars            frame->Xocchars
 #define prop_script        frame->Xprop_script  
 #define prop_test_variable frame->Xprop_test_variable  
556  #endif  #endif
557    
558  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 470  HEAP_RECURSE: Line 576  HEAP_RECURSE:
576  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
577  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
578    
579  #else  #else         /* NO_RECURSE not defined */
580  #define fi i  #define fi i
581  #define fc c  #define fc c
582    
583    /* Many of the following variables are used only in small blocks of the code.
584    My normal style of coding would have declared them within each of those blocks.
585    However, in order to accommodate the version of this code that uses an external
586    "stack" implemented on the heap, it is easier to declare them all here, so the
587    declarations can be cut out in a block. The only declarations within blocks
588    below are for variables that do not have to be preserved over a recursive call
589    to RMATCH(). */
590    
591    #ifdef SUPPORT_UTF
592    const pcre_uchar *charptr;
593    #endif
594    const pcre_uchar *callpat;
595    const pcre_uchar *data;
596    const pcre_uchar *next;
597    PCRE_PUCHAR       pp;
598    const pcre_uchar *prev;
599    PCRE_PUCHAR       saved_eptr;
600    
601  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
602  const uschar *charptr;             /* in small blocks of the code. My normal */  
603  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
604  BOOL condition;  BOOL condition;
 BOOL minimize;  
605  BOOL prev_is_word;  BOOL prev_is_word;
606    
 unsigned long int original_ims;  
   
607  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
608  int prop_type;  int prop_type;
609  int prop_value;  int prop_value;
610  int prop_fail_result;  int prop_fail_result;
611  int prop_category;  int oclength;
612  int prop_chartype;  pcre_uchar occhars[6];
 int prop_script;  
 int *prop_test_variable;  
613  #endif  #endif
614    
615    int codelink;
616  int ctype;  int ctype;
617  int length;  int length;
618  int max;  int max;
# Line 516  int save_offset1, save_offset2, save_off Line 625  int save_offset1, save_offset2, save_off
625  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
626    
627  eptrblock newptrb;  eptrblock newptrb;
628  #endif  
629    /* There is a special fudge for calling match() in a way that causes it to
630    measure the size of its basic stack frame when the stack is being used for
631    recursion. The second argument (ecode) being NULL triggers this behaviour. It
632    cannot normally ever be NULL. The return is the negated value of the frame
633    size. */
634    
635    if (ecode == NULL)
636      {
637      if (rdepth == 0)
638        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
639      else
640        {
641        int len = (char *)&rdepth - (char *)eptr;
642        return (len > 0)? -len : len;
643        }
644      }
645    #endif     /* NO_RECURSE */
646    
647    /* To save space on the stack and in the heap frame, I have doubled up on some
648    of the local variables that are used only in localised parts of the code, but
649    still need to be preserved over recursive calls of match(). These macros define
650    the alternative names that are used. */
651    
652    #define allow_zero    cur_is_word
653    #define cbegroup      condition
654    #define code_offset   codelink
655    #define condassert    condition
656    #define matched_once  prev_is_word
657    #define foc           number
658    #define save_mark     data
659    
660  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
661  variables. */  variables. */
# Line 524  variables. */ Line 663  variables. */
663  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
664  prop_value = 0;  prop_value = 0;
665  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_variable = NULL;  
666  #endif  #endif
667    
668    
669  /* 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
670  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
671  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 676  TAIL_RECURSE:
676  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
677  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
678  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
679  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
680  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
681  complicated macro. It has to be used in one particular way. This shouldn't,  complicated macro. It has to be used in one particular way. This shouldn't,
682  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
683    
684    #ifdef SUPPORT_UTF
685    utf = md->utf;       /* Local copy of the flag */
686    #else
687    utf = FALSE;
688    #endif
689    
690  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
691  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
692    
693  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
694  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
695    
696  original_ims = ims;    /* Save for resetting on ')' */  /* At the start of a group with an unlimited repeat that may match an empty
697    string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
698  #ifdef SUPPORT_UTF8  done this way to save having to use another function argument, which would take
699  utf8 = md->utf8;       /* Local copy of the flag */  up space on the stack. See also MATCH_CONDASSERT below.
700  #else  
701  utf8 = FALSE;  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
702  #endif  such remembered pointers, to be checked when we hit the closing ket, in order
703    to break infinite loops that match no characters. When match() is called in
704  /* At the start of a bracketed group, add the current subject pointer to the  other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
705  stack of such pointers, to be re-instated at the end of the group when we hit  NOT be used with tail recursion, because the memory block that is used is on
706  the closing ket. When match() is called in other circumstances, we don't add to  the stack, so a new one may be required for each match(). */
 this stack. */  
707    
708  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
709    {    {
   newptrb.epb_prev = eptrb;  
710    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
711      newptrb.epb_prev = eptrb;
712    eptrb = &newptrb;    eptrb = &newptrb;
713      md->match_function_type = 0;
714    }    }
715    
716  /* Now start processing the operations. */  /* Now start processing the opcodes. */
717    
718  for (;;)  for (;;)
719    {    {
720      minimize = possessive = FALSE;
721    op = *ecode;    op = *ecode;
   minimize = FALSE;  
722    
723    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
724    matching at least one subject character. */      {
725        case OP_MARK:
726        md->nomatch_mark = ecode + 2;
727        md->mark = NULL;    /* In case previously set by assertion */
728        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
729          eptrb, RM55);
730        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
731             md->mark == NULL) md->mark = ecode + 2;
732    
733        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
734        argument, and we must check whether that argument matches this MARK's
735        argument. It is passed back in md->start_match_ptr (an overloading of that
736        variable). If it does match, we reset that variable to the current subject
737        position and return MATCH_SKIP. Otherwise, pass back the return code
738        unaltered. */
739    
740        else if (rrc == MATCH_SKIP_ARG &&
741            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
742          {
743          md->start_match_ptr = eptr;
744          RRETURN(MATCH_SKIP);
745          }
746        RRETURN(rrc);
747    
748        case OP_FAIL:
749        RRETURN(MATCH_NOMATCH);
750    
751        /* COMMIT overrides PRUNE, SKIP, and THEN */
752    
753        case OP_COMMIT:
754        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
755          eptrb, RM52);
756        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
757            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
758            rrc != MATCH_THEN)
759          RRETURN(rrc);
760        RRETURN(MATCH_COMMIT);
761    
762        /* PRUNE overrides THEN */
763    
764        case OP_PRUNE:
765        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
766          eptrb, RM51);
767        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
768        RRETURN(MATCH_PRUNE);
769    
770        case OP_PRUNE_ARG:
771        md->nomatch_mark = ecode + 2;
772        md->mark = NULL;    /* In case previously set by assertion */
773        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
774          eptrb, RM56);
775        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
776             md->mark == NULL) md->mark = ecode + 2;
777        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
778        RRETURN(MATCH_PRUNE);
779    
780        /* SKIP overrides PRUNE and THEN */
781    
782        case OP_SKIP:
783        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
784          eptrb, RM53);
785        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
786          RRETURN(rrc);
787        md->start_match_ptr = eptr;   /* Pass back current position */
788        RRETURN(MATCH_SKIP);
789    
790        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
791        nomatch_mark. There is a flag that disables this opcode when re-matching a
792        pattern that ended with a SKIP for which there was not a matching MARK. */
793    
794        case OP_SKIP_ARG:
795        if (md->ignore_skip_arg)
796          {
797          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
798          break;
799          }
800        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
801          eptrb, RM57);
802        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
803          RRETURN(rrc);
804    
805        /* Pass back the current skip name by overloading md->start_match_ptr and
806        returning the special MATCH_SKIP_ARG return code. This will either be
807        caught by a matching MARK, or get to the top, where it causes a rematch
808        with the md->ignore_skip_arg flag set. */
809    
810        md->start_match_ptr = ecode + 2;
811        RRETURN(MATCH_SKIP_ARG);
812    
813        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
814        the branch in which it occurs can be determined. Overload the start of
815        match pointer to do this. */
816    
817        case OP_THEN:
818        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
819          eptrb, RM54);
820        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
821        md->start_match_ptr = ecode;
822        RRETURN(MATCH_THEN);
823    
824        case OP_THEN_ARG:
825        md->nomatch_mark = ecode + 2;
826        md->mark = NULL;    /* In case previously set by assertion */
827        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
828          md, eptrb, RM58);
829        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
830             md->mark == NULL) md->mark = ecode + 2;
831        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
832        md->start_match_ptr = ecode;
833        RRETURN(MATCH_THEN);
834    
835        /* Handle an atomic group that does not contain any capturing parentheses.
836        This can be handled like an assertion. Prior to 8.13, all atomic groups
837        were handled this way. In 8.13, the code was changed as below for ONCE, so
838        that backups pass through the group and thereby reset captured values.
839        However, this uses a lot more stack, so in 8.20, atomic groups that do not
840        contain any captures generate OP_ONCE_NC, which can be handled in the old,
841        less stack intensive way.
842    
843    if (md->partial &&      Check the alternative branches in turn - the matching won't pass the KET
844        eptr >= md->end_subject &&      for this kind of subpattern. If any one branch matches, we carry on as at
845        eptr > md->start_match)      the end of a normal bracket, leaving the subject pointer, but resetting
846      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. */  
847    
848    if (op > OP_BRA)      case OP_ONCE_NC:
849      {      prev = ecode;
850      number = op - OP_BRA;      saved_eptr = eptr;
851        save_mark = md->mark;
852        do
853          {
854          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
855          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
856            {
857            mstart = md->start_match_ptr;
858            break;
859            }
860          if (rrc == MATCH_THEN)
861            {
862            next = ecode + GET(ecode,1);
863            if (md->start_match_ptr < next &&
864                (*ecode == OP_ALT || *next == OP_ALT))
865              rrc = MATCH_NOMATCH;
866            }
867    
868          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
869          ecode += GET(ecode,1);
870          md->mark = save_mark;
871          }
872        while (*ecode == OP_ALT);
873    
874        /* If hit the end of the group (which could be repeated), fail */
875    
876        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
877    
878        /* Continue as from after the group, updating the offsets high water
879        mark, since extracts may have been taken. */
880    
881        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
882    
883        offset_top = md->end_offset_top;
884        eptr = md->end_match_ptr;
885    
886        /* For a non-repeating ket, just continue at this level. This also
887        happens for a repeating ket if no characters were matched in the group.
888        This is the forcible breaking of infinite loops as implemented in Perl
889        5.005. */
890    
891        if (*ecode == OP_KET || eptr == saved_eptr)
892          {
893          ecode += 1+LINK_SIZE;
894          break;
895          }
896    
897        /* The repeating kets try the rest of the pattern or restart from the
898        preceding bracket, in the appropriate order. The second "call" of match()
899        uses tail recursion, to avoid using another stack frame. */
900    
901      /* For extended extraction brackets (large number), we have to fish out the      if (*ecode == OP_KETRMIN)
902      number from a dummy opcode at the start. */        {
903          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
904          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
905          ecode = prev;
906          goto TAIL_RECURSE;
907          }
908        else  /* OP_KETRMAX */
909          {
910          md->match_function_type = MATCH_CBEGROUP;
911          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
912          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
913          ecode += 1 + LINK_SIZE;
914          goto TAIL_RECURSE;
915          }
916        /* Control never gets here */
917    
918      if (number > EXTRACT_BASIC_MAX)      /* Handle a capturing bracket, other than those that are possessive with an
919        number = GET2(ecode, 2+LINK_SIZE);      unlimited repeat. If there is space in the offset vector, save the current
920        subject position in the working slot at the top of the vector. We mustn't
921        change the current values of the data slot, because they may be set from a
922        previous iteration of this group, and be referred to by a reference inside
923        the group. A failure to match might occur after the group has succeeded,
924        if something later on doesn't match. For this reason, we need to restore
925        the working value and also the values of the final offsets, in case they
926        were set by a previous iteration of the same bracket.
927    
928        If there isn't enough space in the offset vector, treat this as if it were
929        a non-capturing bracket. Don't worry about setting the flag for the error
930        case here; that is handled in the code for KET. */
931    
932        case OP_CBRA:
933        case OP_SCBRA:
934        number = GET2(ecode, 1+LINK_SIZE);
935      offset = number << 1;      offset = number << 1;
936    
937  #ifdef DEBUG  #ifdef PCRE_DEBUG
938      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
939        printf("subject=");
940      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
941      printf("\n");      printf("\n");
942  #endif  #endif
# Line 620  for (;;) Line 947  for (;;)
947        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
948        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
949        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
950          save_mark = md->mark;
951    
952        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
953        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
954            (int)(eptr - md->start_subject);
955    
956        do        for (;;)
957          {          {
958          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
959            match_isgroup);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
960              eptrb, RM1);
961            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
962    
963            /* If we backed up to a THEN, check whether it is within the current
964            branch by comparing the address of the THEN that is passed back with
965            the end of the branch. If it is within the current branch, and the
966            branch is one of two or more alternatives (it either starts or ends
967            with OP_ALT), we have reached the limit of THEN's action, so convert
968            the return code to NOMATCH, which will cause normal backtracking to
969            happen from now on. Otherwise, THEN is passed back to an outer
970            alternative. This implements Perl's treatment of parenthesized groups,
971            where a group not containing | does not affect the current alternative,
972            that is, (X) is NOT the same as (X|(*F)). */
973    
974            if (rrc == MATCH_THEN)
975              {
976              next = ecode + GET(ecode,1);
977              if (md->start_match_ptr < next &&
978                  (*ecode == OP_ALT || *next == OP_ALT))
979                rrc = MATCH_NOMATCH;
980              }
981    
982            /* Anything other than NOMATCH is passed back. */
983    
984          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
985          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
986          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
987            md->mark = save_mark;
988            if (*ecode != OP_ALT) break;
989          }          }
       while (*ecode == OP_ALT);  
990    
991        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
992        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
993        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
994        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
995    
996        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
997    
998          RRETURN(rrc);
999        }        }
1000    
1001      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1002        as a non-capturing bracket. */
1003    
1004      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1005      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1006    
1007    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1008    
1009    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1010      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1011      case OP_BRA:     /* Non-capturing bracket: optimized */  
1012      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
1013        repeat and ONCE group with no captures. Loop for all the alternatives.
1014    
1015      /* Loop for all the alternatives */      When we get to the final alternative within the brackets, we used to return
1016        the result of a recursive call to match() whatever happened so it was
1017        possible to reduce stack usage by turning this into a tail recursion,
1018        except in the case of a possibly empty group. However, now that there is
1019        the possiblity of (*THEN) occurring in the final alternative, this
1020        optimization is no longer always possible.
1021    
1022        We can optimize if we know there are no (*THEN)s in the pattern; at present
1023        this is the best that can be done.
1024    
1025        MATCH_ONCE is returned when the end of an atomic group is successfully
1026        reached, but subsequent matching fails. It passes back up the tree (causing
1027        captured values to be reset) until the original atomic group level is
1028        reached. This is tested by comparing md->once_target with the start of the
1029        group. At this point, the return is converted into MATCH_NOMATCH so that
1030        previous backup points can be taken. */
1031    
1032        case OP_ONCE:
1033        case OP_BRA:
1034        case OP_SBRA:
1035        DPRINTF(("start non-capturing bracket\n"));
1036    
1037      for (;;)      for (;;)
1038        {        {
1039        /* 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;  
        }  
1040    
1041        /* 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
1042        otherwise return. */        the pattern, and this is the final alternative, optimize as described
1043          above. */
1044    
1045        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1046          match_isgroup);          {
1047        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1048            goto TAIL_RECURSE;
1049            }
1050    
1051          /* In all other cases, we have to make another call to match(). */
1052    
1053          save_mark = md->mark;
1054          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1055            RM2);
1056    
1057          /* See comment in the code for capturing groups above about handling
1058          THEN. */
1059    
1060          if (rrc == MATCH_THEN)
1061            {
1062            next = ecode + GET(ecode,1);
1063            if (md->start_match_ptr < next &&
1064                (*ecode == OP_ALT || *next == OP_ALT))
1065              rrc = MATCH_NOMATCH;
1066            }
1067    
1068          if (rrc != MATCH_NOMATCH)
1069            {
1070            if (rrc == MATCH_ONCE)
1071              {
1072              const pcre_uchar *scode = ecode;
1073              if (*scode != OP_ONCE)           /* If not at start, find it */
1074                {
1075                while (*scode == OP_ALT) scode += GET(scode, 1);
1076                scode -= GET(scode, 1);
1077                }
1078              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1079              }
1080            RRETURN(rrc);
1081            }
1082        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1083          md->mark = save_mark;
1084          if (*ecode != OP_ALT) break;
1085        }        }
     /* Control never reaches here. */  
1086    
1087      /* 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. */  
1088    
1089      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1090      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
1091        {      handled similarly to the normal case above. However, the matching is
1092        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      different. The end of these brackets will always be OP_KETRPOS, which
1093        condition = (offset == CREF_RECURSE * 2)?      returns MATCH_KETRPOS without going further in the pattern. By this means
1094          (md->recursive != NULL) :      we can handle the group by iteration rather than recursion, thereby
1095          (offset < offset_top && md->offset_vector[offset] >= 0);      reducing the amount of stack needed. */
1096        ecode += condition? (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1));  
1097        flags = match_isgroup;      case OP_CBRAPOS:
1098        goto TAIL_RECURSE;      case OP_SCBRAPOS:
1099        }      allow_zero = FALSE;
1100    
1101      /* The condition is an assertion. Call match() to evaluate it - setting      POSSESSIVE_CAPTURE:
1102      the final argument TRUE causes it to stop at the end of an assertion. */      number = GET2(ecode, 1+LINK_SIZE);
1103        offset = number << 1;
1104    
1105      else  #ifdef PCRE_DEBUG
1106        printf("start possessive bracket %d\n", number);
1107        printf("subject=");
1108        pchars(eptr, 16, TRUE, md);
1109        printf("\n");
1110    #endif
1111    
1112        if (offset < md->offset_max)
1113        {        {
1114        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1115            match_condassert | match_isgroup);        code_offset = (int)(ecode - md->start_code);
1116        if (rrc == MATCH_MATCH)  
1117          {        save_offset1 = md->offset_vector[offset];
1118          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);        save_offset2 = md->offset_vector[offset+1];
1119          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_offset3 = md->offset_vector[md->offset_end - number];
1120          save_capture_last = md->capture_last;
1121    
1122          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1123    
1124          /* Each time round the loop, save the current subject position for use
1125          when the group matches. For MATCH_MATCH, the group has matched, so we
1126          restart it with a new subject starting position, remembering that we had
1127          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1128          usual. If we haven't matched any alternatives in any iteration, check to
1129          see if a previous iteration matched. If so, the group has matched;
1130          continue from afterwards. Otherwise it has failed; restore the previous
1131          capture values before returning NOMATCH. */
1132    
1133          for (;;)
1134            {
1135            md->offset_vector[md->offset_end - number] =
1136              (int)(eptr - md->start_subject);
1137            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1138            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1139              eptrb, RM63);
1140            if (rrc == MATCH_KETRPOS)
1141              {
1142              offset_top = md->end_offset_top;
1143              eptr = md->end_match_ptr;
1144              ecode = md->start_code + code_offset;
1145              save_capture_last = md->capture_last;
1146              matched_once = TRUE;
1147              continue;
1148              }
1149    
1150            /* See comment in the code for capturing groups above about handling
1151            THEN. */
1152    
1153            if (rrc == MATCH_THEN)
1154              {
1155              next = ecode + GET(ecode,1);
1156              if (md->start_match_ptr < next &&
1157                  (*ecode == OP_ALT || *next == OP_ALT))
1158                rrc = MATCH_NOMATCH;
1159              }
1160    
1161            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1162            md->capture_last = save_capture_last;
1163            ecode += GET(ecode, 1);
1164            if (*ecode != OP_ALT) break;
1165          }          }
1166        else if (rrc != MATCH_NOMATCH)  
1167          if (!matched_once)
1168          {          {
1169          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1170            md->offset_vector[offset+1] = save_offset2;
1171            md->offset_vector[md->offset_end - number] = save_offset3;
1172          }          }
       else ecode += GET(ecode, 1);  
1173    
1174        /* We are now at the branch that is to be obeyed. As there is only one,        if (allow_zero || matched_once)
1175        we can use tail recursion to avoid using another stack frame. */          {
1176            ecode += 1 + LINK_SIZE;
1177            break;
1178            }
1179    
1180        ecode += 1 + LINK_SIZE;        RRETURN(MATCH_NOMATCH);
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
1181        }        }
     /* Control never reaches here */  
1182    
1183      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1184      encountered. */      as a non-capturing bracket. */
1185    
1186      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1187      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1188    
1189      /* 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. */  
1190    
1191      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1192      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;  
       }  
1193    
1194      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1195      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1196        without the capturing complication. It is written out separately for speed
1197        and cleanliness. */
1198    
1199      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1200      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1201      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;  
1202    
1203      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1204      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1205      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. */  
1206    
1207      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1208        {        {
1209        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1210          match_isgroup);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1211        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1212          if (rrc == MATCH_KETRPOS)
1213            {
1214            offset_top = md->end_offset_top;
1215            eptr = md->end_match_ptr;
1216            ecode = md->start_code + code_offset;
1217            matched_once = TRUE;
1218            continue;
1219            }
1220    
1221          /* See comment in the code for capturing groups above about handling
1222          THEN. */
1223    
1224          if (rrc == MATCH_THEN)
1225            {
1226            next = ecode + GET(ecode,1);
1227            if (md->start_match_ptr < next &&
1228                (*ecode == OP_ALT || *next == OP_ALT))
1229              rrc = MATCH_NOMATCH;
1230            }
1231    
1232        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1233        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1234          if (*ecode != OP_ALT) break;
1235        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1236    
1237      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1238          {
1239          ecode += 1 + LINK_SIZE;
1240          break;
1241          }
1242        RRETURN(MATCH_NOMATCH);
1243    
1244      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1245    
1246      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1247      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1248        past the end if there is only one branch, but that's OK because that is
1249        exactly what going to the ket would do. */
1250    
1251      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1252      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1253      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1254    
1255      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1256        inserted between OP_COND and an assertion condition. */
1257    
1258      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1259        {        {
1260        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (PUBL(callout) != NULL)
1261          match_isgroup);          {
1262        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1263        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1264        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1265            cb.offset_vector    = md->offset_vector;
1266    #ifdef COMPILE_PCRE8
1267            cb.subject          = (PCRE_SPTR)md->start_subject;
1268    #else
1269            cb.subject          = (PCRE_SPTR16)md->start_subject;
1270    #endif
1271            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1272            cb.start_match      = (int)(mstart - md->start_subject);
1273            cb.current_position = (int)(eptr - md->start_subject);
1274            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1275            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1276            cb.capture_top      = offset_top/2;
1277            cb.capture_last     = md->capture_last;
1278            cb.callout_data     = md->callout_data;
1279            cb.mark             = md->nomatch_mark;
1280            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1281            if (rrc < 0) RRETURN(rrc);
1282            }
1283          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1284        }        }
     while (*ecode == OP_ALT);  
1285    
1286      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      condcode = ecode[LINK_SIZE+1];
1287    
1288      ecode += 1 + LINK_SIZE;      /* Now see what the actual condition is */
     continue;  
   
     /* Move the subject pointer back. This occurs only at the start of  
     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. */  
1289    
1290      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1291        {        {
1292        c = GET(ecode,1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
       for (i = 0; i < c; i++)  
1293          {          {
1294          eptr--;          condition = FALSE;
1295          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          ecode += GET(ecode, 1);
         BACKCHAR(eptr)  
1296          }          }
1297        }        else
1298      else          {
1299  #endif          int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1300            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1301    
1302      /* 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
1303            false, but the test was set up by name, scan the table to see if the
1304            name refers to any other numbers, and test them. The condition is true
1305            if any one is set. */
1306    
1307        {          if (!condition && condcode == OP_NRREF)
1308        eptr -= GET(ecode,1);            {
1309        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);            pcre_uchar *slotA = md->name_table;
1310        }            for (i = 0; i < md->name_count; i++)
1311                {
1312                if (GET2(slotA, 0) == recno) break;
1313                slotA += md->name_entry_size;
1314                }
1315    
1316              /* Found a name for the number - there can be only one; duplicate
1317              names for different numbers are allowed, but not vice versa. First
1318              scan down for duplicates. */
1319    
1320              if (i < md->name_count)
1321                {
1322                pcre_uchar *slotB = slotA;
1323                while (slotB > md->name_table)
1324                  {
1325                  slotB -= md->name_entry_size;
1326                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1327                    {
1328                    condition = GET2(slotB, 0) == md->recursive->group_num;
1329                    if (condition) break;
1330                    }
1331                  else break;
1332                  }
1333    
1334                /* Scan up for duplicates */
1335    
1336                if (!condition)
1337                  {
1338                  slotB = slotA;
1339                  for (i++; i < md->name_count; i++)
1340                    {
1341                    slotB += md->name_entry_size;
1342                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1343                      {
1344                      condition = GET2(slotB, 0) == md->recursive->group_num;
1345                      if (condition) break;
1346                      }
1347                    else break;
1348                    }
1349                  }
1350                }
1351              }
1352    
1353            /* Chose branch according to the condition */
1354    
1355            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1356            }
1357          }
1358    
1359        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1360          {
1361          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1362          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1363    
1364          /* If the numbered capture is unset, but the reference was by name,
1365          scan the table to see if the name refers to any other numbers, and test
1366          them. The condition is true if any one is set. This is tediously similar
1367          to the code above, but not close enough to try to amalgamate. */
1368    
1369          if (!condition && condcode == OP_NCREF)
1370            {
1371            int refno = offset >> 1;
1372            pcre_uchar *slotA = md->name_table;
1373    
1374            for (i = 0; i < md->name_count; i++)
1375              {
1376              if (GET2(slotA, 0) == refno) break;
1377              slotA += md->name_entry_size;
1378              }
1379    
1380            /* Found a name for the number - there can be only one; duplicate names
1381            for different numbers are allowed, but not vice versa. First scan down
1382            for duplicates. */
1383    
1384            if (i < md->name_count)
1385              {
1386              pcre_uchar *slotB = slotA;
1387              while (slotB > md->name_table)
1388                {
1389                slotB -= md->name_entry_size;
1390                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1391                  {
1392                  offset = GET2(slotB, 0) << 1;
1393                  condition = offset < offset_top &&
1394                    md->offset_vector[offset] >= 0;
1395                  if (condition) break;
1396                  }
1397                else break;
1398                }
1399    
1400              /* Scan up for duplicates */
1401    
1402              if (!condition)
1403                {
1404                slotB = slotA;
1405                for (i++; i < md->name_count; i++)
1406                  {
1407                  slotB += md->name_entry_size;
1408                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1409                    {
1410                    offset = GET2(slotB, 0) << 1;
1411                    condition = offset < offset_top &&
1412                      md->offset_vector[offset] >= 0;
1413                    if (condition) break;
1414                    }
1415                  else break;
1416                  }
1417                }
1418              }
1419            }
1420    
1421          /* Chose branch according to the condition */
1422    
1423          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1424          }
1425    
1426        else if (condcode == OP_DEF)     /* DEFINE - always false */
1427          {
1428          condition = FALSE;
1429          ecode += GET(ecode, 1);
1430          }
1431    
1432        /* The condition is an assertion. Call match() to evaluate it - setting
1433        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1434        an assertion. */
1435    
1436        else
1437          {
1438          md->match_function_type = MATCH_CONDASSERT;
1439          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1440          if (rrc == MATCH_MATCH)
1441            {
1442            if (md->end_offset_top > offset_top)
1443              offset_top = md->end_offset_top;  /* Captures may have happened */
1444            condition = TRUE;
1445            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1446            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1447            }
1448    
1449          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1450          assertion; it is therefore treated as NOMATCH. */
1451    
1452          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1453            {
1454            RRETURN(rrc);         /* Need braces because of following else */
1455            }
1456          else
1457            {
1458            condition = FALSE;
1459            ecode += codelink;
1460            }
1461          }
1462    
1463        /* We are now at the branch that is to be obeyed. As there is only one, can
1464        use tail recursion to avoid using another stack frame, except when there is
1465        unlimited repeat of a possibly empty group. In the latter case, a recursive
1466        call to match() is always required, unless the second alternative doesn't
1467        exist, in which case we can just plough on. Note that, for compatibility
1468        with Perl, the | in a conditional group is NOT treated as creating two
1469        alternatives. If a THEN is encountered in the branch, it propagates out to
1470        the enclosing alternative (unless nested in a deeper set of alternatives,
1471        of course). */
1472    
1473        if (condition || *ecode == OP_ALT)
1474          {
1475          if (op != OP_SCOND)
1476            {
1477            ecode += 1 + LINK_SIZE;
1478            goto TAIL_RECURSE;
1479            }
1480    
1481          md->match_function_type = MATCH_CBEGROUP;
1482          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1483          RRETURN(rrc);
1484          }
1485    
1486         /* Condition false & no alternative; continue after the group. */
1487    
1488        else
1489          {
1490          ecode += 1 + LINK_SIZE;
1491          }
1492        break;
1493    
1494    
1495        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1496        to close any currently open capturing brackets. */
1497    
1498        case OP_CLOSE:
1499        number = GET2(ecode, 1);
1500        offset = number << 1;
1501    
1502    #ifdef PCRE_DEBUG
1503          printf("end bracket %d at *ACCEPT", number);
1504          printf("\n");
1505    #endif
1506    
1507        md->capture_last = number;
1508        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1509          {
1510          md->offset_vector[offset] =
1511            md->offset_vector[md->offset_end - number];
1512          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1513          if (offset_top <= offset) offset_top = offset + 2;
1514          }
1515        ecode += 1 + IMM2_SIZE;
1516        break;
1517    
1518    
1519        /* End of the pattern, either real or forced. */
1520    
1521        case OP_END:
1522        case OP_ACCEPT:
1523        case OP_ASSERT_ACCEPT:
1524    
1525        /* If we have matched an empty string, fail if not in an assertion and not
1526        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1527        is set and we have matched at the start of the subject. In both cases,
1528        backtracking will then try other alternatives, if any. */
1529    
1530        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1531             md->recursive == NULL &&
1532             (md->notempty ||
1533               (md->notempty_atstart &&
1534                 mstart == md->start_subject + md->start_offset)))
1535          RRETURN(MATCH_NOMATCH);
1536    
1537      /* Skip to next op code */      /* Otherwise, we have a match. */
1538    
1539        md->end_match_ptr = eptr;           /* Record where we ended */
1540        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1541        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1542    
1543        /* For some reason, the macros don't work properly if an expression is
1544        given as the argument to RRETURN when the heap is in use. */
1545    
1546        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1547        RRETURN(rrc);
1548    
1549        /* Assertion brackets. Check the alternative branches in turn - the
1550        matching won't pass the KET for an assertion. If any one branch matches,
1551        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1552        start of each branch to move the current point backwards, so the code at
1553        this level is identical to the lookahead case. When the assertion is part
1554        of a condition, we want to return immediately afterwards. The caller of
1555        this incarnation of the match() function will have set MATCH_CONDASSERT in
1556        md->match_function type, and one of these opcodes will be the first opcode
1557        that is processed. We use a local variable that is preserved over calls to
1558        match() to remember this case. */
1559    
1560        case OP_ASSERT:
1561        case OP_ASSERTBACK:
1562        save_mark = md->mark;
1563        if (md->match_function_type == MATCH_CONDASSERT)
1564          {
1565          condassert = TRUE;
1566          md->match_function_type = 0;
1567          }
1568        else condassert = FALSE;
1569    
1570        do
1571          {
1572          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1573          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1574            {
1575            mstart = md->start_match_ptr;   /* In case \K reset it */
1576            break;
1577            }
1578          md->mark = save_mark;
1579    
1580          /* A COMMIT failure must fail the entire assertion, without trying any
1581          subsequent branches. */
1582    
1583          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1584    
1585          /* PCRE does not allow THEN to escape beyond an assertion; it
1586          is treated as NOMATCH. */
1587    
1588          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1589          ecode += GET(ecode, 1);
1590          }
1591        while (*ecode == OP_ALT);
1592    
1593        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1594    
1595        /* If checking an assertion for a condition, return MATCH_MATCH. */
1596    
1597        if (condassert) RRETURN(MATCH_MATCH);
1598    
1599        /* Continue from after the assertion, updating the offsets high water
1600        mark, since extracts may have been taken during the assertion. */
1601    
1602        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1603        ecode += 1 + LINK_SIZE;
1604        offset_top = md->end_offset_top;
1605        continue;
1606    
1607        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1608        PRUNE, or COMMIT means we must assume failure without checking subsequent
1609        branches. */
1610    
1611        case OP_ASSERT_NOT:
1612        case OP_ASSERTBACK_NOT:
1613        save_mark = md->mark;
1614        if (md->match_function_type == MATCH_CONDASSERT)
1615          {
1616          condassert = TRUE;
1617          md->match_function_type = 0;
1618          }
1619        else condassert = FALSE;
1620    
1621        do
1622          {
1623          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1624          md->mark = save_mark;
1625          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1626          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1627            {
1628            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1629            break;
1630            }
1631    
1632          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1633          as NOMATCH. */
1634    
1635          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1636          ecode += GET(ecode,1);
1637          }
1638        while (*ecode == OP_ALT);
1639    
1640        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1641    
1642        ecode += 1 + LINK_SIZE;
1643        continue;
1644    
1645        /* Move the subject pointer back. This occurs only at the start of
1646        each branch of a lookbehind assertion. If we are too close to the start to
1647        move back, this match function fails. When working with UTF-8 we move
1648        back a number of characters, not bytes. */
1649    
1650        case OP_REVERSE:
1651    #ifdef SUPPORT_UTF
1652        if (utf)
1653          {
1654          i = GET(ecode, 1);
1655          while (i-- > 0)
1656            {
1657            eptr--;
1658            if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1659            BACKCHAR(eptr);
1660            }
1661          }
1662        else
1663    #endif
1664    
1665        /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1666    
1667          {
1668          eptr -= GET(ecode, 1);
1669          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1670          }
1671    
1672        /* Save the earliest consulted character, then skip to next op code */
1673    
1674        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1675      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1676      break;      break;
1677    
# Line 854  for (;;) Line 1680  for (;;)
1680      function is able to force a failure. */      function is able to force a failure. */
1681    
1682      case OP_CALLOUT:      case OP_CALLOUT:
1683      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1684        {        {
1685        pcre_callout_block cb;        PUBL(callout_block) cb;
1686        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1687        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1688        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1689    #ifdef COMPILE_PCRE8
1690        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1691        cb.subject_length   = md->end_subject - md->start_subject;  #else
1692        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1693        cb.current_position = eptr - md->start_subject;  #endif
1694          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1695          cb.start_match      = (int)(mstart - md->start_subject);
1696          cb.current_position = (int)(eptr - md->start_subject);
1697        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1698        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1699        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1700        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1701        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1702        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1703          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1704        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1705        }        }
1706      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 879  for (;;) Line 1710  for (;;)
1710      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
1711      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1712    
1713      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1714      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
1715      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
1716      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
1717      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
1718      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
1719      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.  
1720    
1721      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
1722      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
1723      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1724        a lot, so he is not to blame for the current way it works. */
1725    
1726      case OP_RECURSE:      case OP_RECURSE:
1727        {        {
1728        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1729        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1730    
1731        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1732        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1733            GET2(callpat, 1 + LINK_SIZE);
1734    
1735        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1736          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1737          caught at compile time.) */
1738    
1739          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1740            if (recno == ri->group_num && eptr == ri->subject_position)
1741              RRETURN(PCRE_ERROR_RECURSELOOP);
1742    
1743        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1744    
1745          new_recursive.group_num = recno;
1746          new_recursive.subject_position = eptr;
1747        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1748        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1749    
1750        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1751    
1752        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1753    
1754        /* Now save the offset data. */        /* Now save the offset data */
1755    
1756        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1757        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 923  for (;;) Line 1759  for (;;)
1759        else        else
1760          {          {
1761          new_recursive.offset_save =          new_recursive.offset_save =
1762            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1763          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1764          }          }
   
1765        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1766              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1767    
1768        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1769        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1770          might be changed, so reset it before looping. */
1771    
1772        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1773          cbegroup = (*callpat >= OP_SBRA);
1774        do        do
1775          {          {
1776          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1777              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1778          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1779            memcpy(md->offset_vector, new_recursive.offset_save,
1780                new_recursive.saved_max * sizeof(int));
1781            md->recursive = new_recursive.prevrec;
1782            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1783            {            {
1784            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1785            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1786              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1787            RRETURN(MATCH_MATCH);  
1788              /* Set where we got to in the subject, and reset the start in case
1789              it was changed by \K. This *is* propagated back out of a recursion,
1790              for Perl compatibility. */
1791    
1792              eptr = md->end_match_ptr;
1793              mstart = md->start_match_ptr;
1794              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1795            }            }
1796          else if (rrc != MATCH_NOMATCH)  
1797            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1798            is treated as NOMATCH. */
1799    
1800            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1801                     rrc != MATCH_COMMIT)
1802            {            {
1803            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1804              if (new_recursive.offset_save != stacksave)
1805                (PUBL(free))(new_recursive.offset_save);
1806            RRETURN(rrc);            RRETURN(rrc);
1807            }            }
1808    
1809          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1810          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1811          }          }
1812        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 964  for (;;) Line 1814  for (;;)
1814        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1815        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1816        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1817          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1818        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1819        }        }
     /* Control never reaches here */  
1820    
1821      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1822      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 */  
1823    
1824      /* 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
1825      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1051  for (;;) Line 1828  for (;;)
1828      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1829      break;      break;
1830    
1831      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1832      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
1833      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
1834      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
1835      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1836    
1837      case OP_BRAZERO:      case OP_BRAZERO:
1838        {      next = ecode + 1;
1839        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1840        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1841        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1842        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1843      break;      break;
1844    
1845      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1846        {      next = ecode + 1;
1847        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1848        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1849        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1850          match_isgroup);      ecode++;
1851        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1852        ecode++;  
1853        }      case OP_SKIPZERO:
1854        next = ecode+1;
1855        do next += GET(next,1); while (*next == OP_ALT);
1856        ecode = next + 1 + LINK_SIZE;
1857      break;      break;
1858    
1859      /* 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
1860      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1861      current high water mark for use by positive assertions. Do this also  
1862      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1863        op = *(++ecode);
1864        allow_zero = TRUE;
1865        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1866          goto POSSESSIVE_NON_CAPTURE;
1867    
1868        /* End of a group, repeated or non-repeating. */
1869    
1870      case OP_KET:      case OP_KET:
1871      case OP_KETRMIN:      case OP_KETRMIN:
1872      case OP_KETRMAX:      case OP_KETRMAX:
1873        case OP_KETRPOS:
1874      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
     saved_eptr = eptrb->epb_saved_eptr;  
1875    
1876      /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1877        infinite repeats of empty string matches, retrieve the subject start from
1878        the chain. Otherwise, set it NULL. */
1879    
1880      eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
   
     if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
         *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||  
         *prev == OP_ONCE)  
1881        {        {
1882        md->end_match_ptr = eptr;      /* For ONCE */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1883        md->end_offset_top = offset_top;        eptrb = eptrb->epb_prev;              /* Backup to previous group */
       RRETURN(MATCH_MATCH);  
1884        }        }
1885        else saved_eptr = NULL;
1886    
1887      /* 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
1888      group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1889      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
1890        start in case it was changed by \K. */
1891    
1892      if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1893             *prev == OP_ONCE_NC)
1894        {        {
1895        number = *prev - OP_BRA;        md->end_match_ptr = eptr;      /* For ONCE_NC */
1896          md->end_offset_top = offset_top;
1897          md->start_match_ptr = mstart;
1898          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1899          }
1900    
1901        /* 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
1902        the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1903        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1904        into group 0, so it won't be picked up here. Instead, we catch it when the
1905        OP_END is reached. Other recursion is handled here. We just have to record
1906        the current subject position and start match pointer and give a MATCH
1907        return. */
1908    
1909        if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1910            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1911          {
1912          number = GET2(prev, 1+LINK_SIZE);
1913        offset = number << 1;        offset = number << 1;
1914    
1915  #ifdef DEBUG  #ifdef PCRE_DEBUG
1916        printf("end bracket %d", number);        printf("end bracket %d", number);
1917        printf("\n");        printf("\n");
1918  #endif  #endif
1919    
1920        /* Test for a numbered group. This includes groups called as a result        /* Handle a recursively called group. */
1921        of recursion. Note that whole-pattern recursion is coded as a recurse  
1922        into group 0, so it won't be picked up here. Instead, we catch it when        if (md->recursive != NULL && md->recursive->group_num == number)
1923        the OP_END is reached. */          {
1924            md->end_match_ptr = eptr;
1925            md->start_match_ptr = mstart;
1926            RRETURN(MATCH_MATCH);
1927            }
1928    
1929        if (number > 0)        /* Deal with capturing */
1930    
1931          md->capture_last = number;
1932          if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1933          {          {
1934          md->capture_last = number;          /* If offset is greater than offset_top, it means that we are
1935          if (offset >= md->offset_max) md->offset_overflow = TRUE; else          "skipping" a capturing group, and that group's offsets must be marked
1936            unset. In earlier versions of PCRE, all the offsets were unset at the
1937            start of matching, but this doesn't work because atomic groups and
1938            assertions can cause a value to be set that should later be unset.
1939            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1940            part of the atomic group, but this is not on the final matching path,
1941            so must be unset when 2 is set. (If there is no group 2, there is no
1942            problem, because offset_top will then be 2, indicating no capture.) */
1943    
1944            if (offset > offset_top)
1945            {            {
1946            md->offset_vector[offset] =            register int *iptr = md->offset_vector + offset_top;
1947              md->offset_vector[md->offset_end - number];            register int *iend = md->offset_vector + offset;
1948            md->offset_vector[offset+1] = eptr - md->start_subject;            while (iptr < iend) *iptr++ = -1;
           if (offset_top <= offset) offset_top = offset + 2;  
1949            }            }
1950    
1951          /* Handle a recursively called group. Restore the offsets          /* Now make the extraction */
         appropriately and continue from after the call. */  
1952    
1953          if (md->recursive != NULL && md->recursive->group_num == number)          md->offset_vector[offset] =
1954            {            md->offset_vector[md->offset_end - number];
1955            recursion_info *rec = md->recursive;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1956            DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (offset_top <= offset) offset_top = offset + 2;
           md->recursive = rec->prevrec;  
           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;  
           }  
1957          }          }
1958        }        }
1959    
1960      /* Reset the value of the ims flags, in case they got changed during      /* For an ordinary non-repeating ket, just continue at this level. This
1961      the group. */      also happens for a repeating ket if no characters were matched in the
1962        group. This is the forcible breaking of infinite loops as implemented in
1963      ims = original_ims;      Perl 5.005. For a non-repeating atomic group that includes captures,
1964      DPRINTF(("ims reset to %02lx\n", ims));      establish a backup point by processing the rest of the pattern at a lower
1965        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1966      /* For a non-repeating ket, just continue at this level. This also      original OP_ONCE level, thereby bypassing intermediate backup points, but
1967      happens for a repeating ket if no characters were matched in the group.      resetting any captures that happened along the way. */
     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. */  
1968    
1969      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1970        {        {
1971        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1972            {
1973            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1974            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1975            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1976            RRETURN(MATCH_ONCE);
1977            }
1978          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1979        break;        break;
1980        }        }
1981    
1982      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1983      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
1984      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
1985    
1986        if (*ecode == OP_KETRPOS)
1987          {
1988          md->end_match_ptr = eptr;
1989          md->end_offset_top = offset_top;
1990          RRETURN(MATCH_KETRPOS);
1991          }
1992    
1993        /* The normal repeating kets try the rest of the pattern or restart from
1994        the preceding bracket, in the appropriate order. In the second case, we can
1995        use tail recursion to avoid using another stack frame, unless we have an
1996        an atomic group or an unlimited repeat of a group that can match an empty
1997        string. */
1998    
1999      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2000        {        {
2001        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2002        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2003          if (*prev == OP_ONCE)
2004            {
2005            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2006            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2007            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2008            RRETURN(MATCH_ONCE);
2009            }
2010          if (*prev >= OP_SBRA)    /* Could match an empty string */
2011            {
2012            md->match_function_type = MATCH_CBEGROUP;
2013            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2014            RRETURN(rrc);
2015            }
2016        ecode = prev;        ecode = prev;
       flags = match_isgroup;  
2017        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2018        }        }
2019      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2020        {        {
2021        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
2022          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2023          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2024        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2025          if (*prev == OP_ONCE)
2026            {
2027            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2028            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2029            md->once_target = prev;
2030            RRETURN(MATCH_ONCE);
2031            }
2032        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2033        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2034        }        }
2035      /* Control never gets here */      /* Control never gets here */
2036    
2037      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2038    
2039      case OP_CIRC:      case OP_CIRC:
2040      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 */  
2041    
2042      /* Start of subject assertion */      /* Start of subject assertion */
2043    
# Line 1218  for (;;) Line 2046  for (;;)
2046      ecode++;      ecode++;
2047      break;      break;
2048    
2049        /* Multiline mode: start of subject unless notbol, or after any newline. */
2050    
2051        case OP_CIRCM:
2052        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2053        if (eptr != md->start_subject &&
2054            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2055          RRETURN(MATCH_NOMATCH);
2056        ecode++;
2057        break;
2058    
2059      /* Start of match assertion */      /* Start of match assertion */
2060    
2061      case OP_SOM:      case OP_SOM:
# Line 1225  for (;;) Line 2063  for (;;)
2063      ecode++;      ecode++;
2064      break;      break;
2065    
2066      /* 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. */  
2067    
2068      case OP_DOLL:      case OP_SET_SOM:
2069      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2070        ecode++;
2071        break;
2072    
2073        /* Multiline mode: assert before any newline, or before end of subject
2074        unless noteol is set. */
2075    
2076        case OP_DOLLM:
2077        if (eptr < md->end_subject)
2078        {        {
2079        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2080          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2081        else          if (md->partial != 0 &&
2082          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2083        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2084        break;              NLBLOCK->nllen == 2 &&
2085                *eptr == NLBLOCK->nl[0])
2086              {
2087              md->hitend = TRUE;
2088              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2089              }
2090            RRETURN(MATCH_NOMATCH);
2091            }
2092        }        }
2093      else      else
2094        {        {
2095        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2096        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2097        }        }
2098        ecode++;
2099        break;
2100    
2101        /* Not multiline mode: assert before a terminating newline or before end of
2102        subject unless noteol is set. */
2103    
2104        case OP_DOLL:
2105        if (md->noteol) RRETURN(MATCH_NOMATCH);
2106        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2107    
2108      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2109    
2110      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2111    
2112      case OP_EOD:      case OP_EOD:
2113      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2114        SCHECK_PARTIAL();
2115      ecode++;      ecode++;
2116      break;      break;
2117    
2118      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2119    
2120      case OP_EODN:      case OP_EODN:
2121      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2122          (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))      if (eptr < md->end_subject &&
2123            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2124          {
2125          if (md->partial != 0 &&
2126              eptr + 1 >= md->end_subject &&
2127              NLBLOCK->nltype == NLTYPE_FIXED &&
2128              NLBLOCK->nllen == 2 &&
2129              *eptr == NLBLOCK->nl[0])
2130            {
2131            md->hitend = TRUE;
2132            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2133            }
2134        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2135          }
2136    
2137        /* Either at end of string or \n before end. */
2138    
2139        SCHECK_PARTIAL();
2140      ecode++;      ecode++;
2141      break;      break;
2142    
# Line 1276  for (;;) Line 2148  for (;;)
2148    
2149        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2150        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
2151        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2152          partial matching. */
2153    
2154  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2155        if (utf8)        if (utf)
2156          {          {
2157            /* Get status of previous character */
2158    
2159          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2160            {            {
2161            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2162            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2163              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2164            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2165    #ifdef SUPPORT_UCP
2166              if (md->use_ucp)
2167                {
2168                if (c == '_') prev_is_word = TRUE; else
2169                  {
2170                  int cat = UCD_CATEGORY(c);
2171                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2172                  }
2173                }
2174              else
2175    #endif
2176            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2177            }            }
2178          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2179            /* Get status of next character */
2180    
2181            if (eptr >= md->end_subject)
2182              {
2183              SCHECK_PARTIAL();
2184              cur_is_word = FALSE;
2185              }
2186            else
2187            {            {
2188            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2189    #ifdef SUPPORT_UCP
2190              if (md->use_ucp)
2191                {
2192                if (c == '_') cur_is_word = TRUE; else
2193                  {
2194                  int cat = UCD_CATEGORY(c);
2195                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2196                  }
2197                }
2198              else
2199    #endif
2200            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2201            }            }
2202          }          }
2203        else        else
2204  #endif  #endif
2205    
2206        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2207          consistency with the behaviour of \w we do use it in this case. */
2208    
2209          {          {
2210          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2211            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2212          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2213            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2214              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2215    #ifdef SUPPORT_UCP
2216              if (md->use_ucp)
2217                {
2218                c = eptr[-1];
2219                if (c == '_') prev_is_word = TRUE; else
2220                  {
2221                  int cat = UCD_CATEGORY(c);
2222                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2223                  }
2224                }
2225              else
2226    #endif
2227              prev_is_word = MAX_255(eptr[-1])
2228                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2229              }
2230    
2231            /* Get status of next character */
2232    
2233            if (eptr >= md->end_subject)
2234              {
2235              SCHECK_PARTIAL();
2236              cur_is_word = FALSE;
2237              }
2238            else
2239    #ifdef SUPPORT_UCP
2240            if (md->use_ucp)
2241              {
2242              c = *eptr;
2243              if (c == '_') cur_is_word = TRUE; else
2244                {
2245                int cat = UCD_CATEGORY(c);
2246                cur_is_word = (cat == ucp_L || cat == ucp_N);
2247                }
2248              }
2249            else
2250    #endif
2251            cur_is_word = MAX_255(*eptr)
2252              && ((md->ctypes[*eptr] & ctype_word) != 0);
2253          }          }
2254    
2255        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1314  for (;;) Line 2260  for (;;)
2260        }        }
2261      break;      break;
2262    
2263      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2264        CRLF newlines and partial matching. */
2265    
2266      case OP_ANY:      case OP_ANY:
2267      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2268        if (md->partial != 0 &&
2269            eptr + 1 >= md->end_subject &&
2270            NLBLOCK->nltype == NLTYPE_FIXED &&
2271            NLBLOCK->nllen == 2 &&
2272            *eptr == NLBLOCK->nl[0])
2273        {        {
2274        if (eptr <= md->end_subject - md->nllen && IS_NEWLINE(eptr))        md->hitend = TRUE;
2275          RRETURN(MATCH_NOMATCH);        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2276          }
2277    
2278        /* Fall through */
2279    
2280        /* Match any single character whatsoever. */
2281    
2282        case OP_ALLANY:
2283        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2284          {                            /* not be updated before SCHECK_PARTIAL. */
2285          SCHECK_PARTIAL();
2286          RRETURN(MATCH_NOMATCH);
2287        }        }
2288      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2289      if (utf8)  #ifdef SUPPORT_UTF
2290        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2291    #endif
2292      ecode++;      ecode++;
2293      break;      break;
2294    
# Line 1332  for (;;) Line 2296  for (;;)
2296      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2297    
2298      case OP_ANYBYTE:      case OP_ANYBYTE:
2299      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2300          {                            /* not be updated before SCHECK_PARTIAL. */
2301          SCHECK_PARTIAL();
2302          RRETURN(MATCH_NOMATCH);
2303          }
2304        eptr++;
2305      ecode++;      ecode++;
2306      break;      break;
2307    
2308      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2309      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2310          {
2311          SCHECK_PARTIAL();
2312          RRETURN(MATCH_NOMATCH);
2313          }
2314      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2315      if (      if (
2316  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2317         c < 256 &&         c < 256 &&
2318  #endif  #endif
2319         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1350  for (;;) Line 2323  for (;;)
2323      break;      break;
2324    
2325      case OP_DIGIT:      case OP_DIGIT:
2326      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2327          {
2328          SCHECK_PARTIAL();
2329          RRETURN(MATCH_NOMATCH);
2330          }
2331      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2332      if (      if (
2333  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2334         c >= 256 ||         c > 255 ||
2335  #endif  #endif
2336         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2337         )         )
# Line 1363  for (;;) Line 2340  for (;;)
2340      break;      break;
2341    
2342      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2343      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2344          {
2345          SCHECK_PARTIAL();
2346          RRETURN(MATCH_NOMATCH);
2347          }
2348      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2349      if (      if (
2350  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2351         c < 256 &&         c < 256 &&
2352  #endif  #endif
2353         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1376  for (;;) Line 2357  for (;;)
2357      break;      break;
2358    
2359      case OP_WHITESPACE:      case OP_WHITESPACE:
2360      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2361          {
2362          SCHECK_PARTIAL();
2363          RRETURN(MATCH_NOMATCH);
2364          }
2365      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2366      if (      if (
2367  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2368         c >= 256 ||         c > 255 ||
2369  #endif  #endif
2370         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2371         )         )
# Line 1389  for (;;) Line 2374  for (;;)
2374      break;      break;
2375    
2376      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2377      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2378          {
2379          SCHECK_PARTIAL();
2380          RRETURN(MATCH_NOMATCH);
2381          }
2382      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2383      if (      if (
2384  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2385         c < 256 &&         c < 256 &&
2386  #endif  #endif
2387         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1402  for (;;) Line 2391  for (;;)
2391      break;      break;
2392    
2393      case OP_WORDCHAR:      case OP_WORDCHAR:
2394      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2395          {
2396          SCHECK_PARTIAL();
2397          RRETURN(MATCH_NOMATCH);
2398          }
2399      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2400      if (      if (
2401  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2402         c >= 256 ||         c > 255 ||
2403  #endif  #endif
2404         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2405         )         )
# Line 1414  for (;;) Line 2407  for (;;)
2407      ecode++;      ecode++;
2408      break;      break;
2409    
2410        case OP_ANYNL:
2411        if (eptr >= md->end_subject)
2412          {
2413          SCHECK_PARTIAL();
2414          RRETURN(MATCH_NOMATCH);
2415          }
2416        GETCHARINCTEST(c, eptr);
2417        switch(c)
2418          {
2419          default: RRETURN(MATCH_NOMATCH);
2420    
2421          case 0x000d:
2422          if (eptr >= md->end_subject)
2423            {
2424            SCHECK_PARTIAL();
2425            }
2426          else if (*eptr == 0x0a) eptr++;
2427          break;
2428    
2429          case 0x000a:
2430          break;
2431    
2432          case 0x000b:
2433          case 0x000c:
2434          case 0x0085:
2435          case 0x2028:
2436          case 0x2029:
2437          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2438          break;
2439          }
2440        ecode++;
2441        break;
2442    
2443        case OP_NOT_HSPACE:
2444        if (eptr >= md->end_subject)
2445          {
2446          SCHECK_PARTIAL();
2447          RRETURN(MATCH_NOMATCH);
2448          }
2449        GETCHARINCTEST(c, eptr);
2450        switch(c)
2451          {
2452          default: break;
2453          case 0x09:      /* HT */
2454          case 0x20:      /* SPACE */
2455          case 0xa0:      /* NBSP */
2456          case 0x1680:    /* OGHAM SPACE MARK */
2457          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2458          case 0x2000:    /* EN QUAD */
2459          case 0x2001:    /* EM QUAD */
2460          case 0x2002:    /* EN SPACE */
2461          case 0x2003:    /* EM SPACE */
2462          case 0x2004:    /* THREE-PER-EM SPACE */
2463          case 0x2005:    /* FOUR-PER-EM SPACE */
2464          case 0x2006:    /* SIX-PER-EM SPACE */
2465          case 0x2007:    /* FIGURE SPACE */
2466          case 0x2008:    /* PUNCTUATION SPACE */
2467          case 0x2009:    /* THIN SPACE */
2468          case 0x200A:    /* HAIR SPACE */
2469          case 0x202f:    /* NARROW NO-BREAK SPACE */
2470          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2471          case 0x3000:    /* IDEOGRAPHIC SPACE */
2472          RRETURN(MATCH_NOMATCH);
2473          }
2474        ecode++;
2475        break;
2476    
2477        case OP_HSPACE:
2478        if (eptr >= md->end_subject)
2479          {
2480          SCHECK_PARTIAL();
2481          RRETURN(MATCH_NOMATCH);
2482          }
2483        GETCHARINCTEST(c, eptr);
2484        switch(c)
2485          {
2486          default: RRETURN(MATCH_NOMATCH);
2487          case 0x09:      /* HT */
2488          case 0x20:      /* SPACE */
2489          case 0xa0:      /* NBSP */
2490          case 0x1680:    /* OGHAM SPACE MARK */
2491          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2492          case 0x2000:    /* EN QUAD */
2493          case 0x2001:    /* EM QUAD */
2494          case 0x2002:    /* EN SPACE */
2495          case 0x2003:    /* EM SPACE */
2496          case 0x2004:    /* THREE-PER-EM SPACE */
2497          case 0x2005:    /* FOUR-PER-EM SPACE */
2498          case 0x2006:    /* SIX-PER-EM SPACE */
2499          case 0x2007:    /* FIGURE SPACE */
2500          case 0x2008:    /* PUNCTUATION SPACE */
2501          case 0x2009:    /* THIN SPACE */
2502          case 0x200A:    /* HAIR SPACE */
2503          case 0x202f:    /* NARROW NO-BREAK SPACE */
2504          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2505          case 0x3000:    /* IDEOGRAPHIC SPACE */
2506          break;
2507          }
2508        ecode++;
2509        break;
2510    
2511        case OP_NOT_VSPACE:
2512        if (eptr >= md->end_subject)
2513          {
2514          SCHECK_PARTIAL();
2515          RRETURN(MATCH_NOMATCH);
2516          }
2517        GETCHARINCTEST(c, eptr);
2518        switch(c)
2519          {
2520          default: break;
2521          case 0x0a:      /* LF */
2522          case 0x0b:      /* VT */
2523          case 0x0c:      /* FF */
2524          case 0x0d:      /* CR */
2525          case 0x85:      /* NEL */
2526          case 0x2028:    /* LINE SEPARATOR */
2527          case 0x2029:    /* PARAGRAPH SEPARATOR */
2528          RRETURN(MATCH_NOMATCH);
2529          }
2530        ecode++;
2531        break;
2532    
2533        case OP_VSPACE:
2534        if (eptr >= md->end_subject)
2535          {
2536          SCHECK_PARTIAL();
2537          RRETURN(MATCH_NOMATCH);
2538          }
2539        GETCHARINCTEST(c, eptr);
2540        switch(c)
2541          {
2542          default: RRETURN(MATCH_NOMATCH);
2543          case 0x0a:      /* LF */
2544          case 0x0b:      /* VT */
2545          case 0x0c:      /* FF */
2546          case 0x0d:      /* CR */
2547          case 0x85:      /* NEL */
2548          case 0x2028:    /* LINE SEPARATOR */
2549          case 0x2029:    /* PARAGRAPH SEPARATOR */
2550          break;
2551          }
2552        ecode++;
2553        break;
2554    
2555  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2556      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2557      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. */
2558    
2559      case OP_PROP:      case OP_PROP:
2560      case OP_NOTPROP:      case OP_NOTPROP:
2561      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2562          {
2563          SCHECK_PARTIAL();
2564          RRETURN(MATCH_NOMATCH);
2565          }
2566      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2567        {        {
2568        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2569    
2570        switch(ecode[1])        switch(ecode[1])
2571          {          {
# Line 1433  for (;;) Line 2574  for (;;)
2574          break;          break;
2575    
2576          case PT_LAMP:          case PT_LAMP:
2577          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2578               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2579               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2580            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2581           break;          break;
2582    
2583          case PT_GC:          case PT_GC:
2584          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2585            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2586          break;          break;
2587    
2588          case PT_PC:          case PT_PC:
2589          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2590            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2591          break;          break;
2592    
2593          case PT_SC:          case PT_SC:
2594          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2595              RRETURN(MATCH_NOMATCH);
2596            break;
2597    
2598            /* These are specials */
2599    
2600            case PT_ALNUM:
2601            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2602                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2603              RRETURN(MATCH_NOMATCH);
2604            break;
2605    
2606            case PT_SPACE:    /* Perl space */
2607            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2608                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2609                   == (op == OP_NOTPROP))
2610              RRETURN(MATCH_NOMATCH);
2611            break;
2612    
2613            case PT_PXSPACE:  /* POSIX space */
2614            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2615                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2616                 c == CHAR_FF || c == CHAR_CR)
2617                   == (op == OP_NOTPROP))
2618              RRETURN(MATCH_NOMATCH);
2619            break;
2620    
2621            case PT_WORD:
2622            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2623                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2624                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2625            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2626          break;          break;
2627    
2628            /* This should never occur */
2629    
2630          default:          default:
2631          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
         break;  
2632          }          }
2633    
2634        ecode += 3;        ecode += 3;
# Line 1467  for (;;) Line 2639  for (;;)
2639      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2640    
2641      case OP_EXTUNI:      case OP_EXTUNI:
2642      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2643          {
2644          SCHECK_PARTIAL();
2645          RRETURN(MATCH_NOMATCH);
2646          }
2647      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2648        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2649        while (eptr < md->end_subject)
2650        {        {
2651        int chartype, script;        int len = 1;
2652        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2653        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2654        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;  
         }  
2655        }        }
2656        CHECK_PARTIAL();
2657      ecode++;      ecode++;
2658      break;      break;
2659  #endif  #endif
# Line 1499  for (;;) Line 2668  for (;;)
2668      loops). */      loops). */
2669    
2670      case OP_REF:      case OP_REF:
2671        {      case OP_REFI:
2672        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2673        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2674        ecode += 1 + IMM2_SIZE;
2675    
2676        /* 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];  
2677    
2678        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2679        this ensures that every attempt at a match fails. We can't just fail
2680        here, because of the possibility of quantifiers with zero minima.
2681    
2682        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2683          {      so that the back reference matches an empty string.
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
2684    
2685          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2686          case OP_CRMINRANGE:      referenced subpattern. */
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2687    
2688          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2689          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2690          eptr += length;      else
2691          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2692    
2693        /* Set up for repetition, or handle the non-repeated case */
2694    
2695        switch (*ecode)
2696          {
2697          case OP_CRSTAR:
2698          case OP_CRMINSTAR:
2699          case OP_CRPLUS:
2700          case OP_CRMINPLUS:
2701          case OP_CRQUERY:
2702          case OP_CRMINQUERY:
2703          c = *ecode++ - OP_CRSTAR;
2704          minimize = (c & 1) != 0;
2705          min = rep_min[c];                 /* Pick up values from tables; */
2706          max = rep_max[c];                 /* zero for max => infinity */
2707          if (max == 0) max = INT_MAX;
2708          break;
2709    
2710          case OP_CRRANGE:
2711          case OP_CRMINRANGE:
2712          minimize = (*ecode == OP_CRMINRANGE);
2713          min = GET2(ecode, 1);
2714          max = GET2(ecode, 1 + IMM2_SIZE);
2715          if (max == 0) max = INT_MAX;
2716          ecode += 1 + 2 * IMM2_SIZE;
2717          break;
2718    
2719          default:               /* No repeat follows */
2720          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2721            {
2722            if (length == -2) eptr = md->end_subject;   /* Partial match */
2723            CHECK_PARTIAL();
2724            RRETURN(MATCH_NOMATCH);
2725          }          }
2726          eptr += length;
2727          continue;              /* With the main loop */
2728          }
2729    
2730        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2731        main loop. */      zero, just continue with the main loop. If the length is negative, it
2732        means the reference is unset in non-Java-compatible mode. If the minimum is
2733        zero, we can continue at the same level without recursion. For any other
2734        minimum, carrying on will result in NOMATCH. */
2735    
2736        if (length == 0) continue;      if (length == 0) continue;
2737        if (length < 0 && min == 0) continue;
2738    
2739        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2740        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2741        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2742    
2743        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2744          {
2745          int slength;
2746          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2747          {          {
2748          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2749          eptr += length;          CHECK_PARTIAL();
2750            RRETURN(MATCH_NOMATCH);
2751          }          }
2752          eptr += slength;
2753          }
2754    
2755        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2756        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2757    
2758        if (min == max) continue;      if (min == max) continue;
2759    
2760        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2761    
2762        if (minimize)      if (minimize)
2763          {
2764          for (fi = min;; fi++)
2765          {          {
2766          for (fi = min;; fi++)          int slength;
2767            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2768            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2769            if (fi >= max) RRETURN(MATCH_NOMATCH);
2770            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2771            {            {
2772            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2773            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2774            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2775            }            }
2776          /* Control never gets here */          eptr += slength;
2777          }          }
2778          /* Control never gets here */
2779          }
2780    
2781        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2782    
2783        else      else
2784          {
2785          pp = eptr;
2786          for (i = min; i < max; i++)
2787          {          {
2788          pp = eptr;          int slength;
2789          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)  
2790            {            {
2791            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2792            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2793            eptr -= length;  
2794              if (slength == -2 && md->partial != 0 &&
2795                  md->end_subject > md->start_used_ptr)
2796                {
2797                md->hitend = TRUE;
2798                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2799                }
2800              break;
2801            }            }
2802          RRETURN(MATCH_NOMATCH);          eptr += slength;
2803          }          }
2804    
2805          while (eptr >= pp)
2806            {
2807            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2808            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2809            eptr -= length;
2810            }
2811          RRETURN(MATCH_NOMATCH);
2812        }        }
2813      /* Control never gets here */      /* Control never gets here */
2814    
   
   
2815      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2816      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,
2817      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 2826  for (;;)
2826      case OP_NCLASS:      case OP_NCLASS:
2827      case OP_CLASS:      case OP_CLASS:
2828        {        {
2829          /* The data variable is saved across frames, so the byte map needs to
2830          be stored there. */
2831    #define BYTE_MAP ((pcre_uint8 *)data)
2832        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2833        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2834    
2835        switch (*ecode)        switch (*ecode)
2836          {          {
# Line 1638  for (;;) Line 2851  for (;;)
2851          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2852          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2853          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2854          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2855          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2856          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2857          break;          break;
2858    
2859          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1650  for (;;) Line 2863  for (;;)
2863    
2864        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2865    
2866  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2867        /* UTF-8 mode */        if (utf)
       if (utf8)  
2868          {          {
2869          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2870            {            {
2871            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2872                {
2873                SCHECK_PARTIAL();
2874                RRETURN(MATCH_NOMATCH);
2875                }
2876            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2877            if (c > 255)            if (c > 255)
2878              {              {
2879              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2880              }              }
2881            else            else
2882              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2883            }            }
2884          }          }
2885        else        else
2886  #endif  #endif
2887        /* Not UTF-8 mode */        /* Not UTF mode */
2888          {          {
2889          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2890            {            {
2891            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2892                {
2893                SCHECK_PARTIAL();
2894                RRETURN(MATCH_NOMATCH);
2895                }
2896            c = *eptr++;            c = *eptr++;
2897            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2898              if (c > 255)
2899                {
2900                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2901                }
2902              else
2903    #endif
2904                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2905            }            }
2906          }          }
2907    
# Line 1690  for (;;) Line 2915  for (;;)
2915    
2916        if (minimize)        if (minimize)
2917          {          {
2918  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2919          /* UTF-8 mode */          if (utf)
         if (utf8)  
2920            {            {
2921            for (fi = min;; fi++)            for (fi = min;; fi++)
2922              {              {
2923              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2924              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2925              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2926                if (eptr >= md->end_subject)
2927                  {
2928                  SCHECK_PARTIAL();
2929                  RRETURN(MATCH_NOMATCH);
2930                  }
2931              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2932              if (c > 255)              if (c > 255)
2933                {                {
2934                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2935                }                }
2936              else              else
2937                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2938              }              }
2939            }            }
2940          else          else
2941  #endif  #endif
2942          /* Not UTF-8 mode */          /* Not UTF mode */
2943            {            {
2944            for (fi = min;; fi++)            for (fi = min;; fi++)
2945              {              {
2946              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2947              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2948              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2949                if (eptr >= md->end_subject)
2950                  {
2951                  SCHECK_PARTIAL();
2952                  RRETURN(MATCH_NOMATCH);
2953                  }
2954              c = *eptr++;              c = *eptr++;
2955              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2956                if (c > 255)
2957                  {
2958                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2959                  }
2960                else
2961    #endif
2962                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2963              }              }
2964            }            }
2965          /* Control never gets here */          /* Control never gets here */
# Line 1732  for (;;) Line 2971  for (;;)
2971          {          {
2972          pp = eptr;          pp = eptr;
2973    
2974  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2975          /* UTF-8 mode */          if (utf)
         if (utf8)  
2976            {            {
2977            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2978              {              {
2979              int len = 1;              int len = 1;
2980              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2981                  {
2982                  SCHECK_PARTIAL();
2983                  break;
2984                  }
2985              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2986              if (c > 255)              if (c > 255)
2987                {                {
2988                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2989                }                }
2990              else              else
2991                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2992              eptr += len;              eptr += len;
2993              }              }
2994            for (;;)            for (;;)
2995              {              {
2996              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2997              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2998              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2999              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1761  for (;;) Line 3001  for (;;)
3001            }            }
3002          else          else
3003  #endif  #endif
3004            /* Not UTF-8 mode */            /* Not UTF mode */
3005            {            {
3006            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3007              {              {
3008              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3009                  {
3010                  SCHECK_PARTIAL();
3011                  break;
3012                  }
3013              c = *eptr;              c = *eptr;
3014              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3015                if (c > 255)
3016                  {
3017                  if (op == OP_CLASS) break;
3018                  }
3019                else
3020    #endif
3021                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3022              eptr++;              eptr++;
3023              }              }
3024            while (eptr >= pp)            while (eptr >= pp)
3025              {              {
3026              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3027              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3028              eptr--;              eptr--;
3029              }              }
# Line 1780  for (;;) Line 3031  for (;;)
3031    
3032          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3033          }          }
3034    #undef BYTE_MAP
3035        }        }
3036      /* Control never gets here */      /* Control never gets here */
3037    
3038    
3039      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3040      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
3041        mode, because Unicode properties are supported in non-UTF-8 mode. */
3042    
3043  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3044      case OP_XCLASS:      case OP_XCLASS:
3045        {        {
3046        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1812  for (;;) Line 3065  for (;;)
3065          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3066          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3067          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3068          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3069          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3070          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3071          break;          break;
3072    
3073          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1826  for (;;) Line 3079  for (;;)
3079    
3080        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3081          {          {
3082          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3083          GETCHARINC(c, eptr);            {
3084          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3085              RRETURN(MATCH_NOMATCH);
3086              }
3087            GETCHARINCTEST(c, eptr);
3088            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3089          }          }
3090    
3091        /* 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 3100  for (;;)
3100          {          {
3101          for (fi = min;; fi++)          for (fi = min;; fi++)
3102            {            {
3103            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3104            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3105            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3106            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3107            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3108                SCHECK_PARTIAL();
3109                RRETURN(MATCH_NOMATCH);
3110                }
3111              GETCHARINCTEST(c, eptr);
3112              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3113            }            }
3114          /* Control never gets here */          /* Control never gets here */
3115          }          }
# Line 1860  for (;;) Line 3122  for (;;)
3122          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3123            {            {
3124            int len = 1;            int len = 1;
3125            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3126            GETCHARLEN(c, eptr, len);              {
3127            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3128                break;
3129                }
3130    #ifdef SUPPORT_UTF
3131              GETCHARLENTEST(c, eptr, len);
3132    #else
3133              c = *eptr;
3134    #endif
3135              if (!PRIV(xclass)(c, data, utf)) break;
3136            eptr += len;            eptr += len;
3137            }            }
3138          for(;;)          for(;;)
3139            {            {
3140            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3141            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3142            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3143            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3144              if (utf) BACKCHAR(eptr);
3145    #endif
3146            }            }
3147          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3148          }          }
# Line 1882  for (;;) Line 3154  for (;;)
3154      /* Match a single character, casefully */      /* Match a single character, casefully */
3155    
3156      case OP_CHAR:      case OP_CHAR:
3157  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3158      if (utf8)      if (utf)
3159        {        {
3160        length = 1;        length = 1;
3161        ecode++;        ecode++;
3162        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3163        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3164            {
3165            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3166            RRETURN(MATCH_NOMATCH);
3167            }
3168        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3169        }        }
3170      else      else
3171  #endif  #endif
3172        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3173        {        {
3174        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3175            {
3176            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3177            RRETURN(MATCH_NOMATCH);
3178            }
3179        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3180        ecode += 2;        ecode += 2;
3181        }        }
3182      break;      break;
3183    
3184      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3185        subject, give up immediately. */
3186    
3187        case OP_CHARI:
3188        if (eptr >= md->end_subject)
3189          {
3190          SCHECK_PARTIAL();
3191          RRETURN(MATCH_NOMATCH);
3192          }
3193    
3194      case OP_CHARNC:  #ifdef SUPPORT_UTF
3195  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3196        {        {
3197        length = 1;        length = 1;
3198        ecode++;        ecode++;
3199        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3200    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3201        /* 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
3202        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3203          fast lookup table. We know that there is at least one byte left in the
3204          subject. */
3205    
3206        if (fc < 128)        if (fc < 128)
3207          {          {
3208          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3209                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3210            ecode++;
3211            eptr++;
3212          }          }
3213    
3214        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3215          use the value of "length" to check for sufficient bytes left, because the
3216          other case of the character may have more or fewer bytes.  */
3217    
3218        else        else
3219          {          {
3220          int dc;          unsigned int dc;
3221          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3222          ecode += length;          ecode += length;
3223    
# Line 1936  for (;;) Line 3227  for (;;)
3227          if (fc != dc)          if (fc != dc)
3228            {            {
3229  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3230            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3231  #endif  #endif
3232              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3233            }            }
3234          }          }
3235        }        }
3236      else      else
3237  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3238    
3239      /* Non-UTF-8 mode */      /* Not UTF mode */
3240        {        {
3241        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3242        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3243          eptr++;
3244        ecode += 2;        ecode += 2;
3245        }        }
3246      break;      break;
3247    
3248      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3249    
3250      case OP_EXACT:      case OP_EXACT:
3251        case OP_EXACTI:
3252      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3253      ecode += 3;      ecode += 1 + IMM2_SIZE;
3254      goto REPEATCHAR;      goto REPEATCHAR;
3255    
3256        case OP_POSUPTO:
3257        case OP_POSUPTOI:
3258        possessive = TRUE;
3259        /* Fall through */
3260    
3261      case OP_UPTO:      case OP_UPTO:
3262        case OP_UPTOI:
3263      case OP_MINUPTO:      case OP_MINUPTO:
3264        case OP_MINUPTOI:
3265      min = 0;      min = 0;
3266      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3267      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3268      ecode += 3;      ecode += 1 + IMM2_SIZE;
3269        goto REPEATCHAR;
3270    
3271        case OP_POSSTAR:
3272        case OP_POSSTARI:
3273        possessive = TRUE;
3274        min = 0;
3275        max = INT_MAX;
3276        ecode++;
3277        goto REPEATCHAR;
3278    
3279        case OP_POSPLUS:
3280        case OP_POSPLUSI:
3281        possessive = TRUE;
3282        min = 1;
3283        max = INT_MAX;
3284        ecode++;
3285        goto REPEATCHAR;
3286    
3287        case OP_POSQUERY:
3288        case OP_POSQUERYI:
3289        possessive = TRUE;
3290        min = 0;
3291        max = 1;
3292        ecode++;
3293      goto REPEATCHAR;      goto REPEATCHAR;
3294    
3295      case OP_STAR:      case OP_STAR:
3296        case OP_STARI:
3297      case OP_MINSTAR:      case OP_MINSTAR:
3298        case OP_MINSTARI:
3299      case OP_PLUS:      case OP_PLUS:
3300        case OP_PLUSI:
3301      case OP_MINPLUS:      case OP_MINPLUS:
3302        case OP_MINPLUSI:
3303      case OP_QUERY:      case OP_QUERY:
3304        case OP_QUERYI:
3305      case OP_MINQUERY:      case OP_MINQUERY:
3306      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3307        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3308      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3309      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3310      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3311      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3312    
3313      /* 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. */  
3314    
3315      REPEATCHAR:      REPEATCHAR:
3316  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3317      if (utf8)      if (utf)
3318        {        {
3319        length = 1;        length = 1;
3320        charptr = ecode;        charptr = ecode;
3321        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3322        ecode += length;        ecode += length;
3323    
3324        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1999  for (;;) Line 3326  for (;;)
3326    
3327        if (length > 1)        if (length > 1)
3328          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3329  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3330          int othercase;          unsigned int othercase;
3331          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3332              (othercase = _pcre_ucp_othercase(fc)) >= 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
3333               othercase >= 0)            oclength = PRIV(ord2utf)(othercase, occhars);
3334            oclength = _pcre_ord2utf8(othercase, occhars);          else oclength = 0;
3335  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3336    
3337          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3338            {            {
3339           &nb