/[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 733 by ph10, Tue Oct 11 10:29:36 2011 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) 1997-2011 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    /* This is a convenience macro for code that occurs many times. */
86    
87    #define MRRETURN(ra) \
88      { \
89      md->mark = markptr; \
90      RRETURN(ra); \
91      }
92    
93  /* 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.
94  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,
95  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 103  static const char rep_max[] = { 0, 0, 0,
103    
104    
105    
106  #ifdef DEBUG  #ifdef PCRE_DEBUG
107  /*************************************************  /*************************************************
108  *        Debugging function to print chars       *  *        Debugging function to print chars       *
109  *************************************************/  *************************************************/
# Line 101  Returns: nothing Line 123  Returns: nothing
123  static void  static void
124  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)
125  {  {
126  int c;  unsigned int c;
127  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;
128  while (length-- > 0)  while (length-- > 0)
129    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 136  while (length-- > 0)
136  *          Match a back-reference                *  *          Match a back-reference                *
137  *************************************************/  *************************************************/
138    
139  /* 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
140  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
141    the length passed is zero. Note that in caseless UTF-8 mode, the number of
142    subject bytes matched may be different to the number of reference bytes.
143    
144  Arguments:  Arguments:
145    offset      index into the offset vector    offset      index into the offset vector
146    eptr        points into the subject    eptr        pointer into the subject
147    length      length to be matched    length      length of reference to be matched (number of bytes)
148    md          points to match data block    md          points to match data block
149    ims         the ims flags    caseless    TRUE if caseless
150    
151  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
152  */  */
153    
154  static BOOL  static int
155  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register USPTR eptr, int length, match_data *md,
156    unsigned long int ims)    BOOL caseless)
157  {  {
158  USPTR p = md->start_subject + md->offset_vector[offset];  USPTR eptr_start = eptr;
159    register USPTR p = md->start_subject + md->offset_vector[offset];
160    
161  #ifdef DEBUG  #ifdef PCRE_DEBUG
162  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
163    printf("matching subject <null>");    printf("matching subject <null>");
164  else  else
# Line 146  pchars(p, length, FALSE, md); Line 171  pchars(p, length, FALSE, md);
171  printf("\n");  printf("\n");
172  #endif  #endif
173    
174  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
175    
176  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
177    
178  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
179    properly if Unicode properties are supported. Otherwise, we can check only
180    ASCII characters. */
181    
182  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
183    {    {
184    while (length-- > 0)  #ifdef SUPPORT_UTF8
185      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
186      if (md->utf8)
187        {
188        /* Match characters up to the end of the reference. NOTE: the number of
189        bytes matched may differ, because there are some characters whose upper and
190        lower case versions code as different numbers of bytes. For example, U+023A
191        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
192        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
193        the latter. It is important, therefore, to check the length along the
194        reference, not along the subject (earlier code did this wrong). */
195    
196        USPTR endptr = p + length;
197        while (p < endptr)
198          {
199          int c, d;
200          if (eptr >= md->end_subject) return -1;
201          GETCHARINC(c, eptr);
202          GETCHARINC(d, p);
203          if (c != d && c != UCD_OTHERCASE(d)) return -1;
204          }
205        }
206      else
207    #endif
208    #endif
209    
210      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
211      is no UCP support. */
212        {
213        if (eptr + length > md->end_subject) return -1;
214        while (length-- > 0)
215          { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }
216        }
217    }    }
218    
219    /* In the caseful case, we can just compare the bytes, whether or not we
220    are in UTF-8 mode. */
221    
222  else  else
223    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
224      if (eptr + length > md->end_subject) return -1;
225      while (length-- > 0) if (*p++ != *eptr++) return -1;
226      }
227    
228  return TRUE;  return eptr - eptr_start;
229  }  }
230    
231    
# Line 186  calls by keeping local variables that ne Line 251  calls by keeping local variables that ne
251  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
252  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
253  always used to.  always used to.
254    
255    The original heap-recursive code used longjmp(). However, it seems that this
256    can be very slow on some operating systems. Following a suggestion from Stan
257    Switzer, the use of longjmp() has been abolished, at the cost of having to
258    provide a unique number for each call to RMATCH. There is no way of generating
259    a sequence of numbers at compile time in C. I have given them names, to make
260    them stand out more clearly.
261    
262    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
263    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
264    tests. Furthermore, not using longjmp() means that local dynamic variables
265    don't have indeterminate values; this has meant that the frame size can be
266    reduced because the result can be "passed back" by straight setting of the
267    variable instead of being passed in the frame.
268  ****************************************************************************  ****************************************************************************
269  ***************************************************************************/  ***************************************************************************/
270    
271    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
272    below must be updated in sync.  */
273    
274    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
275           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
276           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
277           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
278           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
279           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
280           RM61,  RM62, RM63, RM64, RM65, RM66 };
281    
282  /* 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
283  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
284    actually used in this definition. */
285    
286  #ifndef NO_RECURSE  #ifndef NO_RECURSE
287  #define REGISTER register  #define REGISTER register
288  #ifdef DEBUG  
289  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
290    #define RMATCH(ra,rb,rc,rd,re,rw) \
291    { \    { \
292    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
293    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1); \
294    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
295    }    }
296  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 208  versions and production versions. */ Line 299  versions and production versions. */
299    return ra; \    return ra; \
300    }    }
301  #else  #else
302  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
303    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1)
304  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
305  #endif  #endif
306    
307  #else  #else
308    
309    
310  /* 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
311  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
312  match(), which never changes. */  argument of match(), which never changes. */
313    
314  #define REGISTER  #define REGISTER
315    
316  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
317    {\    {\
318    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\
319    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
320      {\    frame->Xwhere = rw; \
321      newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
322      newframe->Xecode = rb;\    newframe->Xecode = rb;\
323      newframe->Xoffset_top = rc;\    newframe->Xmstart = mstart;\
324      newframe->Xims = re;\    newframe->Xmarkptr = markptr;\
325      newframe->Xeptrb = rf;\    newframe->Xoffset_top = rc;\
326      newframe->Xflags = rg;\    newframe->Xeptrb = re;\
327      newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
328      newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
329      frame = newframe;\    frame = newframe;\
330      DPRINTF(("restarting from line %d\n", __LINE__));\    DPRINTF(("restarting from line %d\n", __LINE__));\
331      goto HEAP_RECURSE;\    goto HEAP_RECURSE;\
332      }\    L_##rw:\
333    else\    DPRINTF(("jumped back to line %d\n", __LINE__));\
     {\  
     DPRINTF(("longjumped back to line %d\n", __LINE__));\  
     frame = md->thisframe;\  
     rx = frame->Xresult;\  
     }\  
334    }    }
335    
336  #define RRETURN(ra)\  #define RRETURN(ra)\
337    {\    {\
338    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
339    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
340    (pcre_stack_free)(newframe);\    (pcre_stack_free)(oldframe);\
341    if (frame != NULL)\    if (frame != NULL)\
342      {\      {\
343      frame->Xresult = ra;\      rrc = ra;\
344      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
345      }\      }\
346    return ra;\    return ra;\
347    }    }
# Line 269  typedef struct heapframe { Line 354  typedef struct heapframe {
354    
355    /* Function arguments that may change */    /* Function arguments that may change */
356    
357    const uschar *Xeptr;    USPTR Xeptr;
358    const uschar *Xecode;    const uschar *Xecode;
359      USPTR Xmstart;
360      USPTR Xmarkptr;
361    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
362    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
363    unsigned int Xrdepth;    unsigned int Xrdepth;
364    
365    /* Function local variables */    /* Function local variables */
366    
367    const uschar *Xcallpat;    USPTR Xcallpat;
368    const uschar *Xcharptr;  #ifdef SUPPORT_UTF8
369    const uschar *Xdata;    USPTR Xcharptr;
370    const uschar *Xnext;  #endif
371    const uschar *Xpp;    USPTR Xdata;
372    const uschar *Xprev;    USPTR Xnext;
373    const uschar *Xsaved_eptr;    USPTR Xpp;
374      USPTR Xprev;
375      USPTR Xsaved_eptr;
376    
377    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
378    
379    BOOL Xcur_is_word;    BOOL Xcur_is_word;
380    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
381    BOOL Xprev_is_word;    BOOL Xprev_is_word;
382    
   unsigned long int Xoriginal_ims;  
   
383  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
384    int Xprop_type;    int Xprop_type;
385    int Xprop_value;    int Xprop_value;
386    int Xprop_fail_result;    int Xprop_fail_result;
387    int Xprop_category;    int Xoclength;
388    int Xprop_chartype;    uschar Xocchars[8];
   int Xprop_script;  
   int *Xprop_test_variable;  
389  #endif  #endif
390    
391      int Xcodelink;
392    int Xctype;    int Xctype;
393    int Xfc;    unsigned int Xfc;
394    int Xfi;    int Xfi;
395    int Xlength;    int Xlength;
396    int Xmax;    int Xmax;
# Line 321  typedef struct heapframe { Line 404  typedef struct heapframe {
404    
405    eptrblock Xnewptrb;    eptrblock Xnewptrb;
406    
407    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
408    
409    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
410    
411  } heapframe;  } heapframe;
412    
# Line 340  typedef struct heapframe { Line 422  typedef struct heapframe {
422  *         Match from current position            *  *         Match from current position            *
423  *************************************************/  *************************************************/
424    
425  /* 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  
426  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
427  same response.  same response. */
428    
429    /* These macros pack up tests that are used for partial matching, and which
430    appears several times in the code. We set the "hit end" flag if the pointer is
431    at the end of the subject and also past the start of the subject (i.e.
432    something has been matched). For hard partial matching, we then return
433    immediately. The second one is used when we already know we are past the end of
434    the subject. */
435    
436    #define CHECK_PARTIAL()\
437      if (md->partial != 0 && eptr >= md->end_subject && \
438          eptr > md->start_used_ptr) \
439        { \
440        md->hitend = TRUE; \
441        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
442        }
443    
444    #define SCHECK_PARTIAL()\
445      if (md->partial != 0 && eptr > md->start_used_ptr) \
446        { \
447        md->hitend = TRUE; \
448        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
449        }
450    
451    
452  Performance note: It might be tempting to extract commonly used fields from the  /* Performance note: It might be tempting to extract commonly used fields from
453  md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf8, end_subject) into individual variables to improve
454  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
455  made performance worse.  made performance worse.
456    
457  Arguments:  Arguments:
458     eptr        pointer in subject     eptr        pointer to current character in subject
459     ecode       position in code     ecode       pointer to current position in compiled code
460       mstart      pointer to the current match start position (can be modified
461                     by encountering \K)
462       markptr     pointer to the most recent MARK name, or NULL
463     offset_top  current top pointer     offset_top  current top pointer
464     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
465     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
466                   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  
467     rdepth      the recursion depth     rdepth      the recursion depth
468    
469  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
470                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
471                   a negative MATCH_xxx value for PRUNE, SKIP, etc
472                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
473                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
474  */  */
475    
476  static int  static int
477  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
478    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
479    int flags, unsigned int rdepth)    unsigned int rdepth)
480  {  {
481  /* 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,
482  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
483  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
484    
485  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
486  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
487  register unsigned int  c;  /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
488  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
489    
490    BOOL minimize, possessive; /* Quantifier options */
491    BOOL caseless;
492    int condcode;
493    
494  /* 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
495  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame" which is obtained from
496  heap storage. Set up the top-level frame here; others are obtained from the  heap storage. Set up the top-level frame here; others are obtained from the
497  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
498    
499  #ifdef NO_RECURSE  #ifdef NO_RECURSE
500  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));
501    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
502  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
503    
504  /* Copy in the original argument variables */  /* Copy in the original argument variables */
505    
506  frame->Xeptr = eptr;  frame->Xeptr = eptr;
507  frame->Xecode = ecode;  frame->Xecode = ecode;
508    frame->Xmstart = mstart;
509    frame->Xmarkptr = markptr;
510  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
511  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
512  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
513    
514  /* 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 519  HEAP_RECURSE:
519    
520  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
521  #define ecode              frame->Xecode  #define ecode              frame->Xecode
522    #define mstart             frame->Xmstart
523    #define markptr            frame->Xmarkptr
524  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
525  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
526  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
527    
528  /* Ditto for the local variables */  /* Ditto for the local variables */
# Line 424  HEAP_RECURSE: Line 531  HEAP_RECURSE:
531  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
532  #endif  #endif
533  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
534    #define codelink           frame->Xcodelink
535  #define data               frame->Xdata  #define data               frame->Xdata
536  #define next               frame->Xnext  #define next               frame->Xnext
537  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 434  HEAP_RECURSE: Line 542  HEAP_RECURSE:
542    
543  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
544  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
545  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
546    
 #define original_ims       frame->Xoriginal_ims  
   
547  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
548  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
549  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
550  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
551  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
552  #define prop_chartype      frame->Xprop_chartype  #define occhars            frame->Xocchars
 #define prop_script        frame->Xprop_script  
 #define prop_test_variable frame->Xprop_test_variable  
553  #endif  #endif
554    
555  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 470  HEAP_RECURSE: Line 573  HEAP_RECURSE:
573  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
574  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
575    
576  #else  #else         /* NO_RECURSE not defined */
577  #define fi i  #define fi i
578  #define fc c  #define fc c
579    
580    /* Many of the following variables are used only in small blocks of the code.
581    My normal style of coding would have declared them within each of those blocks.
582    However, in order to accommodate the version of this code that uses an external
583    "stack" implemented on the heap, it is easier to declare them all here, so the
584    declarations can be cut out in a block. The only declarations within blocks
585    below are for variables that do not have to be preserved over a recursive call
586    to RMATCH(). */
587    
588    #ifdef SUPPORT_UTF8
589    const uschar *charptr;
590    #endif
591    const uschar *callpat;
592    const uschar *data;
593    const uschar *next;
594    USPTR         pp;
595    const uschar *prev;
596    USPTR         saved_eptr;
597    
598    recursion_info new_recursive;
599    
600  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  BOOL cur_is_word;
 const uschar *charptr;             /* in small blocks of the code. My normal */  
 #endif                             /* style of coding would have declared    */  
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
601  BOOL condition;  BOOL condition;
 BOOL minimize;  
602  BOOL prev_is_word;  BOOL prev_is_word;
603    
 unsigned long int original_ims;  
   
604  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
605  int prop_type;  int prop_type;
606  int prop_value;  int prop_value;
607  int prop_fail_result;  int prop_fail_result;
608  int prop_category;  int oclength;
609  int prop_chartype;  uschar occhars[8];
 int prop_script;  
 int *prop_test_variable;  
610  #endif  #endif
611    
612    int codelink;
613  int ctype;  int ctype;
614  int length;  int length;
615  int max;  int max;
# Line 516  int save_offset1, save_offset2, save_off Line 622  int save_offset1, save_offset2, save_off
622  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
623    
624  eptrblock newptrb;  eptrblock newptrb;
625  #endif  #endif     /* NO_RECURSE */
626    
627    /* To save space on the stack and in the heap frame, I have doubled up on some
628    of the local variables that are used only in localised parts of the code, but
629    still need to be preserved over recursive calls of match(). These macros define
630    the alternative names that are used. */
631    
632    #define allow_zero    cur_is_word
633    #define cbegroup      condition
634    #define code_offset   codelink
635    #define condassert    condition
636    #define matched_once  prev_is_word
637    
638  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
639  variables. */  variables. */
# Line 524  variables. */ Line 641  variables. */
641  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
642  prop_value = 0;  prop_value = 0;
643  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_variable = NULL;  
644  #endif  #endif
645    
646    
647  /* 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
648  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
649  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 654  TAIL_RECURSE:
654  /* 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
655  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
656  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()
657  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
658  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
659  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,
660  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
661    
 /* First check that we haven't called match() too many times, or that we  
 haven't exceeded the recursive call limit. */  
   
 if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  
 if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  
   
 original_ims = ims;    /* Save for resetting on ')' */  
   
662  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
663  utf8 = md->utf8;       /* Local copy of the flag */  utf8 = md->utf8;       /* Local copy of the flag */
664  #else  #else
665  utf8 = FALSE;  utf8 = FALSE;
666  #endif  #endif
667    
668  /* At the start of a bracketed group, add the current subject pointer to the  /* First check that we haven't called match() too many times, or that we
669  stack of such pointers, to be re-instated at the end of the group when we hit  haven't exceeded the recursive call limit. */
670  the closing ket. When match() is called in other circumstances, we don't add to  
671  this stack. */  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
672    if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
673    
674    /* At the start of a group with an unlimited repeat that may match an empty
675    string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
676    done this way to save having to use another function argument, which would take
677    up space on the stack. See also MATCH_CONDASSERT below.
678    
679    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
680    such remembered pointers, to be checked when we hit the closing ket, in order
681    to break infinite loops that match no characters. When match() is called in
682    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
683    NOT be used with tail recursion, because the memory block that is used is on
684    the stack, so a new one may be required for each match(). */
685    
686  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
687    {    {
   newptrb.epb_prev = eptrb;  
688    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
689      newptrb.epb_prev = eptrb;
690    eptrb = &newptrb;    eptrb = &newptrb;
691      md->match_function_type = 0;
692    }    }
693    
694  /* Now start processing the operations. */  /* Now start processing the opcodes. */
695    
696  for (;;)  for (;;)
697    {    {
698      minimize = possessive = FALSE;
699    op = *ecode;    op = *ecode;
   minimize = FALSE;  
700    
701    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
702    matching at least one subject character. */      {
703        case OP_MARK:
704        markptr = ecode + 2;
705        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
706          eptrb, RM55);
707    
708        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
709        argument, and we must check whether that argument matches this MARK's
710        argument. It is passed back in md->start_match_ptr (an overloading of that
711        variable). If it does match, we reset that variable to the current subject
712        position and return MATCH_SKIP. Otherwise, pass back the return code
713        unaltered. */
714    
715        if (rrc == MATCH_SKIP_ARG &&
716            strcmp((char *)markptr, (char *)(md->start_match_ptr)) == 0)
717          {
718          md->start_match_ptr = eptr;
719          RRETURN(MATCH_SKIP);
720          }
721    
722        if (md->mark == NULL) md->mark = markptr;
723        RRETURN(rrc);
724    
725        case OP_FAIL:
726        MRRETURN(MATCH_NOMATCH);
727    
728        /* COMMIT overrides PRUNE, SKIP, and THEN */
729    
730        case OP_COMMIT:
731        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
732          eptrb, RM52);
733        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
734            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
735            rrc != MATCH_THEN)
736          RRETURN(rrc);
737        MRRETURN(MATCH_COMMIT);
738    
739        /* PRUNE overrides THEN */
740    
741        case OP_PRUNE:
742        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
743          eptrb, RM51);
744        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
745        MRRETURN(MATCH_PRUNE);
746    
747        case OP_PRUNE_ARG:
748        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
749          eptrb, RM56);
750        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
751        md->mark = ecode + 2;
752        RRETURN(MATCH_PRUNE);
753    
754        /* SKIP overrides PRUNE and THEN */
755    
756        case OP_SKIP:
757        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
758          eptrb, RM53);
759        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
760          RRETURN(rrc);
761        md->start_match_ptr = eptr;   /* Pass back current position */
762        MRRETURN(MATCH_SKIP);
763    
764        case OP_SKIP_ARG:
765        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
766          eptrb, RM57);
767        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
768          RRETURN(rrc);
769    
770        /* Pass back the current skip name by overloading md->start_match_ptr and
771        returning the special MATCH_SKIP_ARG return code. This will either be
772        caught by a matching MARK, or get to the top, where it is treated the same
773        as PRUNE. */
774    
775        md->start_match_ptr = ecode + 2;
776        RRETURN(MATCH_SKIP_ARG);
777    
778        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
779        the branch in which it occurs can be determined. Overload the start of
780        match pointer to do this. */
781    
782        case OP_THEN:
783        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
784          eptrb, RM54);
785        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
786        md->start_match_ptr = ecode;
787        MRRETURN(MATCH_THEN);
788    
789        case OP_THEN_ARG:
790        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top,
791          md, eptrb, RM58);
792        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
793        md->start_match_ptr = ecode;
794        md->mark = ecode + 2;
795        RRETURN(MATCH_THEN);
796    
797        /* Handle an atomic group that does not contain any capturing parentheses.
798        This can be handled like an assertion. Prior to 8.13, all atomic groups
799        were handled this way. In 8.13, the code was changed as below for ONCE, so
800        that backups pass through the group and thereby reset captured values.
801        However, this uses a lot more stack, so in 8.20, atomic groups that do not
802        contain any captures generate OP_ONCE_NC, which can be handled in the old,
803        less stack intensive way.
804    
805    if (md->partial &&      Check the alternative branches in turn - the matching won't pass the KET
806        eptr >= md->end_subject &&      for this kind of subpattern. If any one branch matches, we carry on as at
807        eptr > md->start_match)      the end of a normal bracket, leaving the subject pointer, but resetting
808      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. */  
809    
810    if (op > OP_BRA)      case OP_ONCE_NC:
811      {      prev = ecode;
812      number = op - OP_BRA;      saved_eptr = eptr;
813        do
814          {
815          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
816          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
817            {
818            mstart = md->start_match_ptr;
819            break;
820            }
821          if (rrc == MATCH_THEN)
822            {
823            next = ecode + GET(ecode,1);
824            if (md->start_match_ptr < next &&
825                (*ecode == OP_ALT || *next == OP_ALT))
826              rrc = MATCH_NOMATCH;
827            }
828    
829          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
830          ecode += GET(ecode,1);
831          }
832        while (*ecode == OP_ALT);
833    
834        /* If hit the end of the group (which could be repeated), fail */
835    
836        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
837    
838        /* Continue as from after the group, updating the offsets high water
839        mark, since extracts may have been taken. */
840    
841        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
842    
843        offset_top = md->end_offset_top;
844        eptr = md->end_match_ptr;
845    
846      /* For extended extraction brackets (large number), we have to fish out the      /* For a non-repeating ket, just continue at this level. This also
847      number from a dummy opcode at the start. */      happens for a repeating ket if no characters were matched in the group.
848        This is the forcible breaking of infinite loops as implemented in Perl
849        5.005. */
850    
851        if (*ecode == OP_KET || eptr == saved_eptr)
852          {
853          ecode += 1+LINK_SIZE;
854          break;
855          }
856    
857        /* The repeating kets try the rest of the pattern or restart from the
858        preceding bracket, in the appropriate order. The second "call" of match()
859        uses tail recursion, to avoid using another stack frame. */
860    
861        if (*ecode == OP_KETRMIN)
862          {
863          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
864          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
865          ecode = prev;
866          goto TAIL_RECURSE;
867          }
868        else  /* OP_KETRMAX */
869          {
870          md->match_function_type = MATCH_CBEGROUP;
871          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
872          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
873          ecode += 1 + LINK_SIZE;
874          goto TAIL_RECURSE;
875          }
876        /* Control never gets here */
877    
878      if (number > EXTRACT_BASIC_MAX)      /* Handle a capturing bracket, other than those that are possessive with an
879        number = GET2(ecode, 2+LINK_SIZE);      unlimited repeat. If there is space in the offset vector, save the current
880        subject position in the working slot at the top of the vector. We mustn't
881        change the current values of the data slot, because they may be set from a
882        previous iteration of this group, and be referred to by a reference inside
883        the group. A failure to match might occur after the group has succeeded,
884        if something later on doesn't match. For this reason, we need to restore
885        the working value and also the values of the final offsets, in case they
886        were set by a previous iteration of the same bracket.
887    
888        If there isn't enough space in the offset vector, treat this as if it were
889        a non-capturing bracket. Don't worry about setting the flag for the error
890        case here; that is handled in the code for KET. */
891    
892        case OP_CBRA:
893        case OP_SCBRA:
894        number = GET2(ecode, 1+LINK_SIZE);
895      offset = number << 1;      offset = number << 1;
896    
897  #ifdef DEBUG  #ifdef PCRE_DEBUG
898      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
899        printf("subject=");
900      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
901      printf("\n");      printf("\n");
902  #endif  #endif
# Line 622  for (;;) Line 909  for (;;)
909        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
910    
911        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
912        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
913            (int)(eptr - md->start_subject);
914    
915        do        for (;;)
916          {          {
917          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
918            match_isgroup);          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
919              eptrb, RM1);
920            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
921    
922            /* If we backed up to a THEN, check whether it is within the current
923            branch by comparing the address of the THEN that is passed back with
924            the end of the branch. If it is within the current branch, and the
925            branch is one of two or more alternatives (it either starts or ends
926            with OP_ALT), we have reached the limit of THEN's action, so convert
927            the return code to NOMATCH, which will cause normal backtracking to
928            happen from now on. Otherwise, THEN is passed back to an outer
929            alternative. This implements Perl's treatment of parenthesized groups,
930            where a group not containing | does not affect the current alternative,
931            that is, (X) is NOT the same as (X|(*F)). */
932    
933            if (rrc == MATCH_THEN)
934              {
935              next = ecode + GET(ecode,1);
936              if (md->start_match_ptr < next &&
937                  (*ecode == OP_ALT || *next == OP_ALT))
938                rrc = MATCH_NOMATCH;
939              }
940    
941            /* Anything other than NOMATCH is passed back. */
942    
943          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
944          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
945          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
946            if (*ecode != OP_ALT) break;
947          }          }
       while (*ecode == OP_ALT);  
948    
949        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
950        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
951        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
952        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
953    
954        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
955    
956          if (md->mark == NULL) md->mark = markptr;
957          RRETURN(rrc);
958        }        }
959    
960      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
961        as a non-capturing bracket. */
962    
963      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
964      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
965    
966    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
967    
968    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
969      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
970      case OP_BRA:     /* Non-capturing bracket: optimized */  
971      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
972        repeat and ONCE group with no captures. Loop for all the alternatives.
973    
974        When we get to the final alternative within the brackets, we used to return
975        the result of a recursive call to match() whatever happened so it was
976        possible to reduce stack usage by turning this into a tail recursion,
977        except in the case of a possibly empty group. However, now that there is
978        the possiblity of (*THEN) occurring in the final alternative, this
979        optimization is no longer always possible.
980    
981        We can optimize if we know there are no (*THEN)s in the pattern; at present
982        this is the best that can be done.
983    
984        MATCH_ONCE is returned when the end of an atomic group is successfully
985        reached, but subsequent matching fails. It passes back up the tree (causing
986        captured values to be reset) until the original atomic group level is
987        reached. This is tested by comparing md->once_target with the start of the
988        group. At this point, the return is converted into MATCH_NOMATCH so that
989        previous backup points can be taken. */
990    
991      /* Loop for all the alternatives */      case OP_ONCE:
992        case OP_BRA:
993        case OP_SBRA:
994        DPRINTF(("start non-capturing bracket\n"));
995    
996      for (;;)      for (;;)
997        {        {
998        /* 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;  
        }  
999    
1000        /* 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
1001        otherwise return. */        the pattern, and this is the final alternative, optimize as described
1002          above. */
1003    
1004        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1005          match_isgroup);          {
1006        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          ecode += _pcre_OP_lengths[*ecode];
1007            goto TAIL_RECURSE;
1008            }
1009    
1010          /* In all other cases, we have to make another call to match(). */
1011    
1012          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
1013            RM2);
1014    
1015          /* See comment in the code for capturing groups above about handling
1016          THEN. */
1017    
1018          if (rrc == MATCH_THEN)
1019            {
1020            next = ecode + GET(ecode,1);
1021            if (md->start_match_ptr < next &&
1022                (*ecode == OP_ALT || *next == OP_ALT))
1023              rrc = MATCH_NOMATCH;
1024            }
1025    
1026          if (rrc != MATCH_NOMATCH)
1027            {
1028            if (rrc == MATCH_ONCE)
1029              {
1030              const uschar *scode = ecode;
1031              if (*scode != OP_ONCE)           /* If not at start, find it */
1032                {
1033                while (*scode == OP_ALT) scode += GET(scode, 1);
1034                scode -= GET(scode, 1);
1035                }
1036              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1037              }
1038            RRETURN(rrc);
1039            }
1040        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1041          if (*ecode != OP_ALT) break;
1042        }        }
     /* Control never reaches here. */  
1043    
1044      /* Conditional group: compilation checked that there are no more than      if (md->mark == NULL) md->mark = markptr;
1045      two branches. If the condition is false, skipping the first branch takes us      RRETURN(MATCH_NOMATCH);
     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. */  
1046    
1047      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1048      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
1049        {      handled similarly to the normal case above. However, the matching is
1050        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      different. The end of these brackets will always be OP_KETRPOS, which
1051        condition = (offset == CREF_RECURSE * 2)?      returns MATCH_KETRPOS without going further in the pattern. By this means
1052          (md->recursive != NULL) :      we can handle the group by iteration rather than recursion, thereby
1053          (offset < offset_top && md->offset_vector[offset] >= 0);      reducing the amount of stack needed. */
1054        ecode += condition? (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1));  
1055        flags = match_isgroup;      case OP_CBRAPOS:
1056        goto TAIL_RECURSE;      case OP_SCBRAPOS:
1057        }      allow_zero = FALSE;
1058    
1059      /* The condition is an assertion. Call match() to evaluate it - setting      POSSESSIVE_CAPTURE:
1060      the final argument TRUE causes it to stop at the end of an assertion. */      number = GET2(ecode, 1+LINK_SIZE);
1061        offset = number << 1;
1062    
1063      else  #ifdef PCRE_DEBUG
1064        printf("start possessive bracket %d\n", number);
1065        printf("subject=");
1066        pchars(eptr, 16, TRUE, md);
1067        printf("\n");
1068    #endif
1069    
1070        if (offset < md->offset_max)
1071        {        {
1072        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1073            match_condassert | match_isgroup);        code_offset = ecode - md->start_code;
1074        if (rrc == MATCH_MATCH)  
1075          {        save_offset1 = md->offset_vector[offset];
1076          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);        save_offset2 = md->offset_vector[offset+1];
1077          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_offset3 = md->offset_vector[md->offset_end - number];
1078          save_capture_last = md->capture_last;
1079    
1080          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1081    
1082          /* Each time round the loop, save the current subject position for use
1083          when the group matches. For MATCH_MATCH, the group has matched, so we
1084          restart it with a new subject starting position, remembering that we had
1085          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1086          usual. If we haven't matched any alternatives in any iteration, check to
1087          see if a previous iteration matched. If so, the group has matched;
1088          continue from afterwards. Otherwise it has failed; restore the previous
1089          capture values before returning NOMATCH. */
1090    
1091          for (;;)
1092            {
1093            md->offset_vector[md->offset_end - number] =
1094              (int)(eptr - md->start_subject);
1095            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1096            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1097              eptrb, RM63);
1098            if (rrc == MATCH_KETRPOS)
1099              {
1100              offset_top = md->end_offset_top;
1101              eptr = md->end_match_ptr;
1102              ecode = md->start_code + code_offset;
1103              save_capture_last = md->capture_last;
1104              matched_once = TRUE;
1105              continue;
1106              }
1107    
1108            /* See comment in the code for capturing groups above about handling
1109            THEN. */
1110    
1111            if (rrc == MATCH_THEN)
1112              {
1113              next = ecode + GET(ecode,1);
1114              if (md->start_match_ptr < next &&
1115                  (*ecode == OP_ALT || *next == OP_ALT))
1116                rrc = MATCH_NOMATCH;
1117              }
1118    
1119            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1120            md->capture_last = save_capture_last;
1121            ecode += GET(ecode, 1);
1122            if (*ecode != OP_ALT) break;
1123          }          }
1124        else if (rrc != MATCH_NOMATCH)  
1125          if (!matched_once)
1126          {          {
1127          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1128            md->offset_vector[offset+1] = save_offset2;
1129            md->offset_vector[md->offset_end - number] = save_offset3;
1130          }          }
       else ecode += GET(ecode, 1);  
1131    
1132        /* We are now at the branch that is to be obeyed. As there is only one,        if (md->mark == NULL) md->mark = markptr;
1133        we can use tail recursion to avoid using another stack frame. */        if (allow_zero || matched_once)
1134            {
1135            ecode += 1 + LINK_SIZE;
1136            break;
1137            }
1138    
1139        ecode += 1 + LINK_SIZE;        RRETURN(MATCH_NOMATCH);
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
1140        }        }
     /* Control never reaches here */  
1141    
1142      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1143      encountered. */      as a non-capturing bracket. */
1144    
1145      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1146      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1147    
1148      /* 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. */  
1149    
1150      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1151      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;  
       }  
1152    
1153      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1154      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1155        without the capturing complication. It is written out separately for speed
1156        and cleanliness. */
1157    
1158      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1159      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1160      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;  
1161    
1162      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1163      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1164      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      code_offset = 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. */  
1165    
1166      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1167        {        {
1168        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1169          match_isgroup);        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1170        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1171          if (rrc == MATCH_KETRPOS)
1172            {
1173            offset_top = md->end_offset_top;
1174            eptr = md->end_match_ptr;
1175            ecode = md->start_code + code_offset;
1176            matched_once = TRUE;
1177            continue;
1178            }
1179    
1180          /* See comment in the code for capturing groups above about handling
1181          THEN. */
1182    
1183          if (rrc == MATCH_THEN)
1184            {
1185            next = ecode + GET(ecode,1);
1186            if (md->start_match_ptr < next &&
1187                (*ecode == OP_ALT || *next == OP_ALT))
1188              rrc = MATCH_NOMATCH;
1189            }
1190    
1191        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1192        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1193          if (*ecode != OP_ALT) break;
1194        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1195    
1196      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1197          {
1198          ecode += 1 + LINK_SIZE;
1199          break;
1200          }
1201        RRETURN(MATCH_NOMATCH);
1202    
1203      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1204    
1205      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1206      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1207        past the end if there is only one branch, but that's OK because that is
1208        exactly what going to the ket would do. */
1209    
1210      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1211      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1212      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1213    
1214      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1215        inserted between OP_COND and an assertion condition. */
1216    
1217      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1218        {        {
1219        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (pcre_callout != NULL)
1220          match_isgroup);          {
1221        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          pcre_callout_block cb;
1222        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1223        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1224            cb.offset_vector    = md->offset_vector;
1225            cb.subject          = (PCRE_SPTR)md->start_subject;
1226            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1227            cb.start_match      = (int)(mstart - md->start_subject);
1228            cb.current_position = (int)(eptr - md->start_subject);
1229            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1230            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1231            cb.capture_top      = offset_top/2;
1232            cb.capture_last     = md->capture_last;
1233            cb.callout_data     = md->callout_data;
1234            cb.mark             = markptr;
1235            if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1236            if (rrc < 0) RRETURN(rrc);
1237            }
1238          ecode += _pcre_OP_lengths[OP_CALLOUT];
1239        }        }
     while (*ecode == OP_ALT);  
1240    
1241      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      condcode = ecode[LINK_SIZE+1];
1242    
1243      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. */  
1244    
1245      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1246        {        {
1247        c = GET(ecode,1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
       for (i = 0; i < c; i++)  
1248          {          {
1249          eptr--;          condition = FALSE;
1250          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          ecode += GET(ecode, 1);
         BACKCHAR(eptr)  
1251          }          }
1252        }        else
1253      else          {
1254  #endif          int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1255            condition =  (recno == RREF_ANY || recno == md->recursive->group_num);
1256    
1257      /* 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
1258            false, but the test was set up by name, scan the table to see if the
1259            name refers to any other numbers, and test them. The condition is true
1260            if any one is set. */
1261    
1262        {          if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
1263        eptr -= GET(ecode,1);            {
1264        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);            uschar *slotA = md->name_table;
1265        }            for (i = 0; i < md->name_count; i++)
1266                {
1267                if (GET2(slotA, 0) == recno) break;
1268                slotA += md->name_entry_size;
1269                }
1270    
1271              /* Found a name for the number - there can be only one; duplicate
1272              names for different numbers are allowed, but not vice versa. First
1273              scan down for duplicates. */
1274    
1275              if (i < md->name_count)
1276                {
1277                uschar *slotB = slotA;
1278                while (slotB > md->name_table)
1279                  {
1280                  slotB -= md->name_entry_size;
1281                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1282                    {
1283                    condition = GET2(slotB, 0) == md->recursive->group_num;
1284                    if (condition) break;
1285                    }
1286                  else break;
1287                  }
1288    
1289                /* Scan up for duplicates */
1290    
1291                if (!condition)
1292                  {
1293                  slotB = slotA;
1294                  for (i++; i < md->name_count; i++)
1295                    {
1296                    slotB += md->name_entry_size;
1297                    if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1298                      {
1299                      condition = GET2(slotB, 0) == md->recursive->group_num;
1300                      if (condition) break;
1301                      }
1302                    else break;
1303                    }
1304                  }
1305                }
1306              }
1307    
1308            /* Chose branch according to the condition */
1309    
1310            ecode += condition? 3 : GET(ecode, 1);
1311            }
1312          }
1313    
1314        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1315          {
1316          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1317          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1318    
1319          /* If the numbered capture is unset, but the reference was by name,
1320          scan the table to see if the name refers to any other numbers, and test
1321          them. The condition is true if any one is set. This is tediously similar
1322          to the code above, but not close enough to try to amalgamate. */
1323    
1324          if (!condition && condcode == OP_NCREF)
1325            {
1326            int refno = offset >> 1;
1327            uschar *slotA = md->name_table;
1328    
1329            for (i = 0; i < md->name_count; i++)
1330              {
1331              if (GET2(slotA, 0) == refno) break;
1332              slotA += md->name_entry_size;
1333              }
1334    
1335            /* Found a name for the number - there can be only one; duplicate names
1336            for different numbers are allowed, but not vice versa. First scan down
1337            for duplicates. */
1338    
1339            if (i < md->name_count)
1340              {
1341              uschar *slotB = slotA;
1342              while (slotB > md->name_table)
1343                {
1344                slotB -= md->name_entry_size;
1345                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1346                  {
1347                  offset = GET2(slotB, 0) << 1;
1348                  condition = offset < offset_top &&
1349                    md->offset_vector[offset] >= 0;
1350                  if (condition) break;
1351                  }
1352                else break;
1353                }
1354    
1355              /* Scan up for duplicates */
1356    
1357              if (!condition)
1358                {
1359                slotB = slotA;
1360                for (i++; i < md->name_count; i++)
1361                  {
1362                  slotB += md->name_entry_size;
1363                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1364                    {
1365                    offset = GET2(slotB, 0) << 1;
1366                    condition = offset < offset_top &&
1367                      md->offset_vector[offset] >= 0;
1368                    if (condition) break;
1369                    }
1370                  else break;
1371                  }
1372                }
1373              }
1374            }
1375    
1376          /* Chose branch according to the condition */
1377    
1378          ecode += condition? 3 : GET(ecode, 1);
1379          }
1380    
1381        else if (condcode == OP_DEF)     /* DEFINE - always false */
1382          {
1383          condition = FALSE;
1384          ecode += GET(ecode, 1);
1385          }
1386    
1387        /* The condition is an assertion. Call match() to evaluate it - setting
1388        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1389        an assertion. */
1390    
1391        else
1392          {
1393          md->match_function_type = MATCH_CONDASSERT;
1394          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1395          if (rrc == MATCH_MATCH)
1396            {
1397            if (md->end_offset_top > offset_top)
1398              offset_top = md->end_offset_top;  /* Captures may have happened */
1399            condition = TRUE;
1400            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1401            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1402            }
1403    
1404          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1405          assertion; it is therefore treated as NOMATCH. */
1406    
1407          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1408            {
1409            RRETURN(rrc);         /* Need braces because of following else */
1410            }
1411          else
1412            {
1413            condition = FALSE;
1414            ecode += codelink;
1415            }
1416          }
1417    
1418        /* We are now at the branch that is to be obeyed. As there is only one, can
1419        use tail recursion to avoid using another stack frame, except when there is
1420        unlimited repeat of a possibly empty group. In the latter case, a recursive
1421        call to match() is always required, unless the second alternative doesn't
1422        exist, in which case we can just plough on. Note that, for compatibility
1423        with Perl, the | in a conditional group is NOT treated as creating two
1424        alternatives. If a THEN is encountered in the branch, it propagates out to
1425        the enclosing alternative (unless nested in a deeper set of alternatives,
1426        of course). */
1427    
1428        if (condition || *ecode == OP_ALT)
1429          {
1430          if (op != OP_SCOND)
1431            {
1432            ecode += 1 + LINK_SIZE;
1433            goto TAIL_RECURSE;
1434            }
1435    
1436          md->match_function_type = MATCH_CBEGROUP;
1437          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1438          RRETURN(rrc);
1439          }
1440    
1441         /* Condition false & no alternative; continue after the group. */
1442    
1443        else
1444          {
1445          ecode += 1 + LINK_SIZE;
1446          }
1447        break;
1448    
1449    
1450        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1451        to close any currently open capturing brackets. */
1452    
1453        case OP_CLOSE:
1454        number = GET2(ecode, 1);
1455        offset = number << 1;
1456    
1457    #ifdef PCRE_DEBUG
1458          printf("end bracket %d at *ACCEPT", number);
1459          printf("\n");
1460    #endif
1461    
1462        md->capture_last = number;
1463        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1464          {
1465          md->offset_vector[offset] =
1466            md->offset_vector[md->offset_end - number];
1467          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1468          if (offset_top <= offset) offset_top = offset + 2;
1469          }
1470        ecode += 3;
1471        break;
1472    
1473    
1474        /* End of the pattern, either real or forced. */
1475    
1476        case OP_END:
1477        case OP_ACCEPT:
1478        case OP_ASSERT_ACCEPT:
1479    
1480        /* If we have matched an empty string, fail if not in an assertion and not
1481        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1482        is set and we have matched at the start of the subject. In both cases,
1483        backtracking will then try other alternatives, if any. */
1484    
1485        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1486             md->recursive == NULL &&
1487             (md->notempty ||
1488               (md->notempty_atstart &&
1489                 mstart == md->start_subject + md->start_offset)))
1490          MRRETURN(MATCH_NOMATCH);
1491    
1492        /* Otherwise, we have a match. */
1493    
1494        md->end_match_ptr = eptr;           /* Record where we ended */
1495        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1496        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1497    
1498        /* For some reason, the macros don't work properly if an expression is
1499        given as the argument to MRRETURN when the heap is in use. */
1500    
1501        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1502        MRRETURN(rrc);
1503    
1504        /* Assertion brackets. Check the alternative branches in turn - the
1505        matching won't pass the KET for an assertion. If any one branch matches,
1506        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1507        start of each branch to move the current point backwards, so the code at
1508        this level is identical to the lookahead case. When the assertion is part
1509        of a condition, we want to return immediately afterwards. The caller of
1510        this incarnation of the match() function will have set MATCH_CONDASSERT in
1511        md->match_function type, and one of these opcodes will be the first opcode
1512        that is processed. We use a local variable that is preserved over calls to
1513        match() to remember this case. */
1514    
1515        case OP_ASSERT:
1516        case OP_ASSERTBACK:
1517        if (md->match_function_type == MATCH_CONDASSERT)
1518          {
1519          condassert = TRUE;
1520          md->match_function_type = 0;
1521          }
1522        else condassert = FALSE;
1523    
1524        do
1525          {
1526          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1527          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1528            {
1529            mstart = md->start_match_ptr;   /* In case \K reset it */
1530            markptr = md->mark;
1531            break;
1532            }
1533    
1534          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1535          as NOMATCH. */
1536    
1537          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1538          ecode += GET(ecode, 1);
1539          }
1540        while (*ecode == OP_ALT);
1541    
1542        if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1543    
1544        /* If checking an assertion for a condition, return MATCH_MATCH. */
1545    
1546        if (condassert) RRETURN(MATCH_MATCH);
1547    
1548        /* Continue from after the assertion, updating the offsets high water
1549        mark, since extracts may have been taken during the assertion. */
1550    
1551        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1552        ecode += 1 + LINK_SIZE;
1553        offset_top = md->end_offset_top;
1554        continue;
1555    
1556        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1557        PRUNE, or COMMIT means we must assume failure without checking subsequent
1558        branches. */
1559    
1560        case OP_ASSERT_NOT:
1561        case OP_ASSERTBACK_NOT:
1562        if (md->match_function_type == MATCH_CONDASSERT)
1563          {
1564          condassert = TRUE;
1565          md->match_function_type = 0;
1566          }
1567        else condassert = FALSE;
1568    
1569        do
1570          {
1571          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1572          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1573          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1574            {
1575            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1576            break;
1577            }
1578    
1579          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1580          as NOMATCH. */
1581    
1582          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1583          ecode += GET(ecode,1);
1584          }
1585        while (*ecode == OP_ALT);
1586    
1587        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1588    
1589        ecode += 1 + LINK_SIZE;
1590        continue;
1591    
1592        /* Move the subject pointer back. This occurs only at the start of
1593        each branch of a lookbehind assertion. If we are too close to the start to
1594        move back, this match function fails. When working with UTF-8 we move
1595        back a number of characters, not bytes. */
1596    
1597        case OP_REVERSE:
1598    #ifdef SUPPORT_UTF8
1599        if (utf8)
1600          {
1601          i = GET(ecode, 1);
1602          while (i-- > 0)
1603            {
1604            eptr--;
1605            if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1606            BACKCHAR(eptr);
1607            }
1608          }
1609        else
1610    #endif
1611    
1612        /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1613    
1614          {
1615          eptr -= GET(ecode, 1);
1616          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1617          }
1618    
1619      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1620    
1621        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1622      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1623      break;      break;
1624    
# Line 857  for (;;) Line 1630  for (;;)
1630      if (pcre_callout != NULL)      if (pcre_callout != NULL)
1631        {        {
1632        pcre_callout_block cb;        pcre_callout_block cb;
1633        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1634        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1635        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1636        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1637        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1638        cb.start_match      = md->start_match - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1639        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1640        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1641        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1642        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1643        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1644        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1645        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = markptr;
1646          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1647        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1648        }        }
1649      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 879  for (;;) Line 1653  for (;;)
1653      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
1654      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1655    
1656      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1657      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
1658      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
1659      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
1660      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
1661      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
1662      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.  
1663    
1664      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
1665      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
1666      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1667        a lot, so he is not to blame for the current way it works. */
1668    
1669      case OP_RECURSE:      case OP_RECURSE:
1670        {        {
1671        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1672        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1673    
1674        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1675        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1676            GET2(callpat, 1 + LINK_SIZE);
1677    
1678        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1679          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1680          caught at compile time.) */
1681    
1682          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1683            if (recno == ri->group_num && eptr == ri->subject_position)
1684              RRETURN(PCRE_ERROR_RECURSELOOP);
1685    
1686        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1687    
1688          new_recursive.group_num = recno;
1689          new_recursive.subject_position = eptr;
1690        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1691        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1692    
1693        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1694    
1695        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1696    
1697        /* Now save the offset data. */        /* Now save the offset data */
1698    
1699        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1700        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 926  for (;;) Line 1705  for (;;)
1705            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1706          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1707          }          }
   
1708        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1709              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1710    
1711        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1712        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1713          might be changed, so reset it before looping. */
1714    
1715        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1716          cbegroup = (*callpat >= OP_SBRA);
1717        do        do
1718          {          {
1719          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1720              eptrb, match_isgroup);          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1721          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1722            memcpy(md->offset_vector, new_recursive.offset_save,
1723                new_recursive.saved_max * sizeof(int));
1724            md->recursive = new_recursive.prevrec;
1725            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1726            {            {
1727            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1728            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1729              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1730            RRETURN(MATCH_MATCH);  
1731              /* Set where we got to in the subject, and reset the start in case
1732              it was changed by \K. This *is* propagated back out of a recursion,
1733              for Perl compatibility. */
1734    
1735              eptr = md->end_match_ptr;
1736              mstart = md->start_match_ptr;
1737              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1738            }            }
1739          else if (rrc != MATCH_NOMATCH)  
1740            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1741            as NOMATCH. */
1742    
1743            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1744            {            {
1745            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1746              if (new_recursive.offset_save != stacksave)
1747                (pcre_free)(new_recursive.offset_save);
1748            RRETURN(rrc);            RRETURN(rrc);
1749            }            }
1750    
1751          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1752          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1753          }          }
1754        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 965  for (;;) Line 1757  for (;;)
1757        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1758        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1759          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1760        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
       }  
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
         eptrb, match_isgroup);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
1761        }        }
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
1762    
1763      if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);      RECURSION_MATCHED:
1764        break;
     /* 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 */  
1765    
1766      /* 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
1767      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1051  for (;;) Line 1770  for (;;)
1770      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1771      break;      break;
1772    
1773      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1774      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
1775      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
1776      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
1777      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1778    
1779      case OP_BRAZERO:      case OP_BRAZERO:
1780        {      next = ecode + 1;
1781        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1782        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1783        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1784        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1785      break;      break;
1786    
1787      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1788        {      next = ecode + 1;
1789        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1790        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1791        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1792          match_isgroup);      ecode++;
1793        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1794        ecode++;  
1795        }      case OP_SKIPZERO:
1796        next = ecode+1;
1797        do next += GET(next,1); while (*next == OP_ALT);
1798        ecode = next + 1 + LINK_SIZE;
1799      break;      break;
1800    
1801      /* 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
1802      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1803      current high water mark for use by positive assertions. Do this also  
1804      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1805        op = *(++ecode);
1806        allow_zero = TRUE;
1807        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1808          goto POSSESSIVE_NON_CAPTURE;
1809    
1810        /* End of a group, repeated or non-repeating. */
1811    
1812      case OP_KET:      case OP_KET:
1813      case OP_KETRMIN:      case OP_KETRMIN:
1814      case OP_KETRMAX:      case OP_KETRMAX:
1815        case OP_KETRPOS:
1816      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
     saved_eptr = eptrb->epb_saved_eptr;  
   
     /* Back up the stack of bracket start pointers. */  
1817    
1818      eptrb = eptrb->epb_prev;      /* If this was a group that remembered the subject start, in order to break
1819        infinite repeats of empty string matches, retrieve the subject start from
1820        the chain. Otherwise, set it NULL. */
1821    
1822      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev >= OP_SBRA || *prev == OP_ONCE)
         *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||  
         *prev == OP_ONCE)  
1823        {        {
1824        md->end_match_ptr = eptr;      /* For ONCE */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1825        md->end_offset_top = offset_top;        eptrb = eptrb->epb_prev;              /* Backup to previous group */
       RRETURN(MATCH_MATCH);  
1826        }        }
1827        else saved_eptr = NULL;
1828    
1829      /* 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
1830      group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1831      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
1832        start in case it was changed by \K. */
1833    
1834      if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1835             *prev == OP_ONCE_NC)
1836        {        {
1837        number = *prev - OP_BRA;        md->end_match_ptr = eptr;      /* For ONCE_NC */
1838          md->end_offset_top = offset_top;
1839          md->start_match_ptr = mstart;
1840          MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1841          }
1842    
1843        /* 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
1844        the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1845        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1846        into group 0, so it won't be picked up here. Instead, we catch it when the
1847        OP_END is reached. Other recursion is handled here. We just have to record
1848        the current subject position and start match pointer and give a MATCH
1849        return. */
1850    
1851        if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1852            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1853          {
1854          number = GET2(prev, 1+LINK_SIZE);
1855        offset = number << 1;        offset = number << 1;
1856    
1857  #ifdef DEBUG  #ifdef PCRE_DEBUG
1858        printf("end bracket %d", number);        printf("end bracket %d", number);
1859        printf("\n");        printf("\n");
1860  #endif  #endif
1861    
1862        /* Test for a numbered group. This includes groups called as a result        /* Handle a recursively called group. */
       of recursion. Note that whole-pattern recursion is coded as a recurse  
       into group 0, so it won't be picked up here. Instead, we catch it when  
       the OP_END is reached. */  
1863    
1864        if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1865          {          {
1866          md->capture_last = number;          md->end_match_ptr = eptr;
1867          if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1868            {          RRETURN(MATCH_MATCH);
1869            md->offset_vector[offset] =          }
             md->offset_vector[md->offset_end - number];  
           md->offset_vector[offset+1] = eptr - md->start_subject;  
           if (offset_top <= offset) offset_top = offset + 2;  
           }  
1870    
1871          /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
         appropriately and continue from after the call. */  
1872    
1873          if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = number;
1874          if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1875            {
1876            /* If offset is greater than offset_top, it means that we are
1877            "skipping" a capturing group, and that group's offsets must be marked
1878            unset. In earlier versions of PCRE, all the offsets were unset at the
1879            start of matching, but this doesn't work because atomic groups and
1880            assertions can cause a value to be set that should later be unset.
1881            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1882            part of the atomic group, but this is not on the final matching path,
1883            so must be unset when 2 is set. (If there is no group 2, there is no
1884            problem, because offset_top will then be 2, indicating no capture.) */
1885    
1886            if (offset > offset_top)
1887            {            {
1888            recursion_info *rec = md->recursive;            register int *iptr = md->offset_vector + offset_top;
1889            DPRINTF(("Recursion (%d) succeeded - continuing\n", number));            register int *iend = md->offset_vector + offset;
1890            md->recursive = rec->prevrec;            while (iptr < iend) *iptr++ = -1;
           md->start_match = rec->save_start;  
           memcpy(md->offset_vector, rec->offset_save,  
             rec->saved_max * sizeof(int));  
           ecode = rec->after_call;  
           ims = original_ims;  
           break;  
1891            }            }
1892    
1893            /* Now make the extraction */
1894    
1895            md->offset_vector[offset] =
1896              md->offset_vector[md->offset_end - number];
1897            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1898            if (offset_top <= offset) offset_top = offset + 2;
1899          }          }
1900        }        }
1901    
1902      /* 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
1903      the group. */      also happens for a repeating ket if no characters were matched in the
1904        group. This is the forcible breaking of infinite loops as implemented in
1905      ims = original_ims;      Perl 5.005. For a non-repeating atomic group that includes captures,
1906      DPRINTF(("ims reset to %02lx\n", ims));      establish a backup point by processing the rest of the pattern at a lower
1907        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1908      /* For a non-repeating ket, just continue at this level. This also      original OP_ONCE level, thereby bypassing intermediate backup points, but
1909      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. */  
1910    
1911      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1912        {        {
1913        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1914            {
1915            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1916            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1917            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1918            RRETURN(MATCH_ONCE);
1919            }
1920          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1921        break;        break;
1922        }        }
1923    
1924      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1925      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
1926      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
1927    
1928        if (*ecode == OP_KETRPOS)
1929          {
1930          md->end_match_ptr = eptr;
1931          md->end_offset_top = offset_top;
1932          RRETURN(MATCH_KETRPOS);
1933          }
1934    
1935        /* The normal repeating kets try the rest of the pattern or restart from
1936        the preceding bracket, in the appropriate order. In the second case, we can
1937        use tail recursion to avoid using another stack frame, unless we have an
1938        an atomic group or an unlimited repeat of a group that can match an empty
1939        string. */
1940    
1941      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1942        {        {
1943        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1944        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1945          if (*prev == OP_ONCE)
1946            {
1947            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1948            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1949            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1950            RRETURN(MATCH_ONCE);
1951            }
1952          if (*prev >= OP_SBRA)    /* Could match an empty string */
1953            {
1954            md->match_function_type = MATCH_CBEGROUP;
1955            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1956            RRETURN(rrc);
1957            }
1958        ecode = prev;        ecode = prev;
       flags = match_isgroup;  
1959        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1960        }        }
1961      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1962        {        {
1963        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1964          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1965          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1966        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1967          if (*prev == OP_ONCE)
1968            {
1969            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1970            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1971            md->once_target = prev;
1972            RRETURN(MATCH_ONCE);
1973            }
1974        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1975        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1976        }        }
1977      /* Control never gets here */      /* Control never gets here */
1978    
1979      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1980    
1981      case OP_CIRC:      case OP_CIRC:
1982      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(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 */  
1983    
1984      /* Start of subject assertion */      /* Start of subject assertion */
1985    
1986      case OP_SOD:      case OP_SOD:
1987      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1988        ecode++;
1989        break;
1990    
1991        /* Multiline mode: start of subject unless notbol, or after any newline. */
1992    
1993        case OP_CIRCM:
1994        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1995        if (eptr != md->start_subject &&
1996            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1997          MRRETURN(MATCH_NOMATCH);
1998      ecode++;      ecode++;
1999      break;      break;
2000    
2001      /* Start of match assertion */      /* Start of match assertion */
2002    
2003      case OP_SOM:      case OP_SOM:
2004      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
2005      ecode++;      ecode++;
2006      break;      break;
2007    
2008      /* 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. */  
2009    
2010      case OP_DOLL:      case OP_SET_SOM:
2011      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2012        {      ecode++;
2013        if (eptr < md->end_subject)      break;
2014          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
2015        else      /* Multiline mode: assert before any newline, or before end of subject
2016          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
2017        ecode++;  
2018        break;      case OP_DOLLM:
2019        }      if (eptr < md->end_subject)
2020          { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
2021      else      else
2022        {        {
2023        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
2024        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2025        }        }
2026        ecode++;
2027        break;
2028    
2029        /* Not multiline mode: assert before a terminating newline or before end of
2030        subject unless noteol is set. */
2031    
2032        case OP_DOLL:
2033        if (md->noteol) MRRETURN(MATCH_NOMATCH);
2034        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2035    
2036      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2037    
2038      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2039    
2040      case OP_EOD:      case OP_EOD:
2041      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
2042        SCHECK_PARTIAL();
2043      ecode++;      ecode++;
2044      break;      break;
2045    
2046      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2047    
2048      case OP_EODN:      case OP_EODN:
2049      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2050          (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))      if (eptr < md->end_subject &&
2051        RRETURN(MATCH_NOMATCH);          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2052          MRRETURN(MATCH_NOMATCH);
2053    
2054        /* Either at end of string or \n before end. */
2055    
2056        SCHECK_PARTIAL();
2057      ecode++;      ecode++;
2058      break;      break;
2059    
# Line 1276  for (;;) Line 2065  for (;;)
2065    
2066        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2067        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
2068        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2069          partial matching. */
2070    
2071  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2072        if (utf8)        if (utf8)
2073          {          {
2074            /* Get status of previous character */
2075    
2076          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2077            {            {
2078            const uschar *lastptr = eptr - 1;            USPTR lastptr = eptr - 1;
2079            while((*lastptr & 0xc0) == 0x80) lastptr--;            while((*lastptr & 0xc0) == 0x80) lastptr--;
2080              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2081            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2082    #ifdef SUPPORT_UCP
2083              if (md->use_ucp)
2084                {
2085                if (c == '_') prev_is_word = TRUE; else
2086                  {
2087                  int cat = UCD_CATEGORY(c);
2088                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2089                  }
2090                }
2091              else
2092    #endif
2093            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2094            }            }
2095          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2096            /* Get status of next character */
2097    
2098            if (eptr >= md->end_subject)
2099              {
2100              SCHECK_PARTIAL();
2101              cur_is_word = FALSE;
2102              }
2103            else
2104            {            {
2105            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2106    #ifdef SUPPORT_UCP
2107              if (md->use_ucp)
2108                {
2109                if (c == '_') cur_is_word = TRUE; else
2110                  {
2111                  int cat = UCD_CATEGORY(c);
2112                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2113                  }
2114                }
2115              else
2116    #endif
2117            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2118            }            }
2119          }          }
2120        else        else
2121  #endif  #endif
2122    
2123        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2124          consistency with the behaviour of \w we do use it in this case. */
2125    
2126          {          {
2127          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2128            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2129          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2130            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2131              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2132    #ifdef SUPPORT_UCP
2133              if (md->use_ucp)
2134                {
2135                c = eptr[-1];
2136                if (c == '_') prev_is_word = TRUE; else
2137                  {
2138                  int cat = UCD_CATEGORY(c);
2139                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2140                  }
2141                }
2142              else
2143    #endif
2144              prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2145              }
2146    
2147            /* Get status of next character */
2148    
2149            if (eptr >= md->end_subject)
2150              {
2151              SCHECK_PARTIAL();
2152              cur_is_word = FALSE;
2153              }
2154            else
2155    #ifdef SUPPORT_UCP
2156            if (md->use_ucp)
2157              {
2158              c = *eptr;
2159              if (c == '_') cur_is_word = TRUE; else
2160                {
2161                int cat = UCD_CATEGORY(c);
2162                cur_is_word = (cat == ucp_L || cat == ucp_N);
2163                }
2164              }
2165            else
2166    #endif
2167            cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
2168          }          }
2169    
2170        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
2171    
2172        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
2173             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2174          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2175        }        }
2176      break;      break;
2177    
2178      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2179    
2180      case OP_ANY:      case OP_ANY:
2181      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
2182        {      /* Fall through */
2183        if (eptr <= md->end_subject - md->nllen && IS_NEWLINE(eptr))  
2184          RRETURN(MATCH_NOMATCH);      case OP_ALLANY:
2185        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2186          {                            /* not be updated before SCHECK_PARTIAL. */
2187          SCHECK_PARTIAL();
2188          MRRETURN(MATCH_NOMATCH);
2189        }        }
2190      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2191      if (utf8)      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
       while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;  
2192      ecode++;      ecode++;
2193      break;      break;
2194    
# Line 1332  for (;;) Line 2196  for (;;)
2196      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2197    
2198      case OP_ANYBYTE:      case OP_ANYBYTE:
2199      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2200          {                            /* not be updated before SCHECK_PARTIAL. */
2201          SCHECK_PARTIAL();
2202          MRRETURN(MATCH_NOMATCH);
2203          }
2204        eptr++;
2205      ecode++;      ecode++;
2206      break;      break;
2207    
2208      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2209      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2210          {
2211          SCHECK_PARTIAL();
2212          MRRETURN(MATCH_NOMATCH);
2213          }
2214      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2215      if (      if (
2216  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1345  for (;;) Line 2218  for (;;)
2218  #endif  #endif
2219         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2220         )         )
2221        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2222      ecode++;      ecode++;
2223      break;      break;
2224    
2225      case OP_DIGIT:      case OP_DIGIT:
2226      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2227          {
2228          SCHECK_PARTIAL();
2229          MRRETURN(MATCH_NOMATCH);
2230          }
2231      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2232      if (      if (
2233  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1358  for (;;) Line 2235  for (;;)
2235  #endif  #endif
2236         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2237         )         )
2238        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2239      ecode++;      ecode++;
2240      break;      break;
2241    
2242      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2243      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2244          {
2245          SCHECK_PARTIAL();
2246          MRRETURN(MATCH_NOMATCH);
2247          }
2248      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2249      if (      if (
2250  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1371  for (;;) Line 2252  for (;;)
2252  #endif  #endif
2253         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2254         )         )
2255        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2256      ecode++;      ecode++;
2257      break;      break;
2258    
2259      case OP_WHITESPACE:      case OP_WHITESPACE:
2260      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2261          {
2262          SCHECK_PARTIAL();
2263          MRRETURN(MATCH_NOMATCH);
2264          }
2265      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2266      if (      if (
2267  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1384  for (;;) Line 2269  for (;;)
2269  #endif  #endif
2270         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2271         )         )
2272        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2273      ecode++;      ecode++;
2274      break;      break;
2275    
2276      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2277      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2278          {
2279          SCHECK_PARTIAL();
2280          MRRETURN(MATCH_NOMATCH);
2281          }
2282      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2283      if (      if (
2284  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1397  for (;;) Line 2286  for (;;)
2286  #endif  #endif
2287         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2288         )         )
2289        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2290      ecode++;      ecode++;
2291      break;      break;
2292    
2293      case OP_WORDCHAR:      case OP_WORDCHAR:
2294      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2295          {
2296          SCHECK_PARTIAL();
2297          MRRETURN(MATCH_NOMATCH);
2298          }
2299      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2300      if (      if (
2301  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1410  for (;;) Line 2303  for (;;)
2303  #endif  #endif
2304         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2305         )         )
2306        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2307        ecode++;
2308        break;
2309    
2310        case OP_ANYNL:
2311        if (eptr >= md->end_subject)
2312          {
2313          SCHECK_PARTIAL();
2314          MRRETURN(MATCH_NOMATCH);
2315          }
2316        GETCHARINCTEST(c, eptr);
2317        switch(c)
2318          {
2319          default: MRRETURN(MATCH_NOMATCH);
2320    
2321          case 0x000d:
2322          if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2323          break;
2324    
2325          case 0x000a:
2326          break;
2327    
2328          case 0x000b:
2329          case 0x000c:
2330          case 0x0085:
2331          case 0x2028:
2332          case 0x2029:
2333          if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
2334          break;
2335          }
2336        ecode++;
2337        break;
2338    
2339        case OP_NOT_HSPACE:
2340        if (eptr >= md->end_subject)
2341          {
2342          SCHECK_PARTIAL();
2343          MRRETURN(MATCH_NOMATCH);
2344          }
2345        GETCHARINCTEST(c, eptr);
2346        switch(c)
2347          {
2348          default: break;
2349          case 0x09:      /* HT */
2350          case 0x20:      /* SPACE */
2351          case 0xa0:      /* NBSP */
2352          case 0x1680:    /* OGHAM SPACE MARK */
2353          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2354          case 0x2000:    /* EN QUAD */
2355          case 0x2001:    /* EM QUAD */
2356          case 0x2002:    /* EN SPACE */
2357          case 0x2003:    /* EM SPACE */
2358          case 0x2004:    /* THREE-PER-EM SPACE */
2359          case 0x2005:    /* FOUR-PER-EM SPACE */
2360          case 0x2006:    /* SIX-PER-EM SPACE */
2361          case 0x2007:    /* FIGURE SPACE */
2362          case 0x2008:    /* PUNCTUATION SPACE */
2363          case 0x2009:    /* THIN SPACE */
2364          case 0x200A:    /* HAIR SPACE */
2365          case 0x202f:    /* NARROW NO-BREAK SPACE */
2366          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2367          case 0x3000:    /* IDEOGRAPHIC SPACE */
2368          MRRETURN(MATCH_NOMATCH);
2369          }
2370        ecode++;
2371        break;
2372    
2373        case OP_HSPACE:
2374        if (eptr >= md->end_subject)
2375          {
2376          SCHECK_PARTIAL();
2377          MRRETURN(MATCH_NOMATCH);
2378          }
2379        GETCHARINCTEST(c, eptr);
2380        switch(c)
2381          {
2382          default: MRRETURN(MATCH_NOMATCH);
2383          case 0x09:      /* HT */
2384          case 0x20:      /* SPACE */
2385          case 0xa0:      /* NBSP */
2386          case 0x1680:    /* OGHAM SPACE MARK */
2387          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2388          case 0x2000:    /* EN QUAD */
2389          case 0x2001:    /* EM QUAD */
2390          case 0x2002:    /* EN SPACE */
2391          case 0x2003:    /* EM SPACE */
2392          case 0x2004:    /* THREE-PER-EM SPACE */
2393          case 0x2005:    /* FOUR-PER-EM SPACE */
2394          case 0x2006:    /* SIX-PER-EM SPACE */
2395          case 0x2007:    /* FIGURE SPACE */
2396          case 0x2008:    /* PUNCTUATION SPACE */
2397          case 0x2009:    /* THIN SPACE */
2398          case 0x200A:    /* HAIR SPACE */
2399          case 0x202f:    /* NARROW NO-BREAK SPACE */
2400          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2401          case 0x3000:    /* IDEOGRAPHIC SPACE */
2402          break;
2403          }
2404        ecode++;
2405        break;
2406    
2407        case OP_NOT_VSPACE:
2408        if (eptr >= md->end_subject)
2409          {
2410          SCHECK_PARTIAL();
2411          MRRETURN(MATCH_NOMATCH);
2412          }
2413        GETCHARINCTEST(c, eptr);
2414        switch(c)
2415          {
2416          default: break;
2417          case 0x0a:      /* LF */
2418          case 0x0b:      /* VT */
2419          case 0x0c:      /* FF */
2420          case 0x0d:      /* CR */
2421          case 0x85:      /* NEL */
2422          case 0x2028:    /* LINE SEPARATOR */
2423          case 0x2029:    /* PARAGRAPH SEPARATOR */
2424          MRRETURN(MATCH_NOMATCH);
2425          }
2426        ecode++;
2427        break;
2428    
2429        case OP_VSPACE:
2430        if (eptr >= md->end_subject)
2431          {
2432          SCHECK_PARTIAL();
2433          MRRETURN(MATCH_NOMATCH);
2434          }
2435        GETCHARINCTEST(c, eptr);
2436        switch(c)
2437          {
2438          default: MRRETURN(MATCH_NOMATCH);
2439          case 0x0a:      /* LF */
2440          case 0x0b:      /* VT */
2441          case 0x0c:      /* FF */
2442          case 0x0d:      /* CR */
2443          case 0x85:      /* NEL */
2444          case 0x2028:    /* LINE SEPARATOR */
2445          case 0x2029:    /* PARAGRAPH SEPARATOR */
2446          break;
2447          }
2448      ecode++;      ecode++;
2449      break;      break;
2450    
# Line 1420  for (;;) Line 2454  for (;;)
2454    
2455      case OP_PROP:      case OP_PROP:
2456      case OP_NOTPROP:      case OP_NOTPROP:
2457      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2458          {
2459          SCHECK_PARTIAL();
2460          MRRETURN(MATCH_NOMATCH);
2461          }
2462      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2463        {        {
2464        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2465    
2466        switch(ecode[1])        switch(ecode[1])
2467          {          {
2468          case PT_ANY:          case PT_ANY:
2469          if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
2470          break;          break;
2471    
2472          case PT_LAMP:          case PT_LAMP:
2473          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2474               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2475               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2476            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2477           break;          break;
2478    
2479          case PT_GC:          case PT_GC:
2480          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
2481            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2482          break;          break;
2483    
2484          case PT_PC:          case PT_PC:
2485          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2486            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2487          break;          break;
2488    
2489          case PT_SC:          case PT_SC:
2490          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2491            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2492            break;
2493    
2494            /* These are specials */
2495    
2496            case PT_ALNUM:
2497            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2498                 _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2499              MRRETURN(MATCH_NOMATCH);
2500          break;          break;
2501    
2502            case PT_SPACE:    /* Perl space */
2503            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2504                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2505                   == (op == OP_NOTPROP))
2506              MRRETURN(MATCH_NOMATCH);
2507            break;
2508    
2509            case PT_PXSPACE:  /* POSIX space */
2510            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2511                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2512                 c == CHAR_FF || c == CHAR_CR)
2513                   == (op == OP_NOTPROP))
2514              MRRETURN(MATCH_NOMATCH);
2515            break;
2516    
2517            case PT_WORD:
2518            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2519                 _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2520                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2521              MRRETURN(MATCH_NOMATCH);
2522            break;
2523    
2524            /* This should never occur */
2525    
2526          default:          default:
2527          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
         break;  
2528          }          }
2529    
2530        ecode += 3;        ecode += 3;
# Line 1467  for (;;) Line 2535  for (;;)
2535      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2536    
2537      case OP_EXTUNI:      case OP_EXTUNI:
2538      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2539          {
2540          SCHECK_PARTIAL();
2541          MRRETURN(MATCH_NOMATCH);
2542          }
2543      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2544        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2545        while (eptr < md->end_subject)
2546        {        {
2547        int chartype, script;        int len = 1;
2548        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2549        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2550        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;  
         }  
2551        }        }
2552      ecode++;      ecode++;
2553      break;      break;
# Line 1499  for (;;) Line 2563  for (;;)
2563      loops). */      loops). */
2564    
2565      case OP_REF:      case OP_REF:
2566        {      case OP_REFI:
2567        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2568        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2569        ecode += 3;
2570    
2571        /* 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];  
2572    
2573        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2574        this ensures that every attempt at a match fails. We can't just fail
2575        here, because of the possibility of quantifiers with zero minima.
2576    
2577        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2578          {      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;  
2579    
2580          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2581          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;  
2582    
2583          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2584          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2585          eptr += length;      else
2586          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2587    
2588        /* Set up for repetition, or handle the non-repeated case */
2589    
2590        switch (*ecode)
2591          {
2592          case OP_CRSTAR:
2593          case OP_CRMINSTAR:
2594          case OP_CRPLUS:
2595          case OP_CRMINPLUS:
2596          case OP_CRQUERY:
2597          case OP_CRMINQUERY:
2598          c = *ecode++ - OP_CRSTAR;
2599          minimize = (c & 1) != 0;
2600          min = rep_min[c];                 /* Pick up values from tables; */
2601          max = rep_max[c];                 /* zero for max => infinity */
2602          if (max == 0) max = INT_MAX;
2603          break;
2604    
2605          case OP_CRRANGE:
2606          case OP_CRMINRANGE:
2607          minimize = (*ecode == OP_CRMINRANGE);
2608          min = GET2(ecode, 1);
2609          max = GET2(ecode, 3);
2610          if (max == 0) max = INT_MAX;
2611          ecode += 5;
2612          break;
2613    
2614          default:               /* No repeat follows */
2615          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2616            {
2617            CHECK_PARTIAL();
2618            MRRETURN(MATCH_NOMATCH);
2619          }          }
2620          eptr += length;
2621          continue;              /* With the main loop */
2622          }
2623    
2624        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2625        main loop. */      zero, just continue with the main loop. */
2626    
2627        if (length == 0) continue;      if (length == 0) continue;
2628    
2629        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2630        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2631        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2632    
2633        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2634          {
2635          int slength;
2636          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2637          {          {
2638          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2639          eptr += length;          MRRETURN(MATCH_NOMATCH);
2640          }          }
2641          eptr += slength;
2642          }
2643    
2644        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2645        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2646    
2647        if (min == max) continue;      if (min == max) continue;
2648    
2649        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2650    
2651        if (minimize)      if (minimize)
2652          {
2653          for (fi = min;; fi++)
2654          {          {
2655          for (fi = min;; fi++)          int slength;
2656            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2657            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2658            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2659            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2660            {            {
2661            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2662            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2663            }            }
2664          /* Control never gets here */          eptr += slength;
2665          }          }
2666          /* Control never gets here */
2667          }
2668    
2669        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2670    
2671        else      else
2672          {
2673          pp = eptr;
2674          for (i = min; i < max; i++)
2675          {          {
2676          pp = eptr;          int slength;
2677          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)  
2678            {            {
2679            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2680            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2681            }            }
2682          RRETURN(MATCH_NOMATCH);          eptr += slength;
2683            }
2684          while (eptr >= pp)
2685            {
2686            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2687            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2688            eptr -= length;
2689          }          }
2690          MRRETURN(MATCH_NOMATCH);
2691        }        }
2692      /* Control never gets here */      /* Control never gets here */
2693    
   
   
2694      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2695      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,
2696      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 1656  for (;;) Line 2745  for (;;)
2745          {          {
2746          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2747            {            {
2748            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2749                {
2750                SCHECK_PARTIAL();
2751                MRRETURN(MATCH_NOMATCH);
2752                }
2753            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2754            if (c > 255)            if (c > 255)
2755              {              {
2756              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2757              }              }
2758            else            else
2759              {              {
2760              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2761              }              }
2762            }            }
2763          }          }
# Line 1674  for (;;) Line 2767  for (;;)
2767          {          {
2768          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2769            {            {
2770            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2771                {
2772                SCHECK_PARTIAL();
2773                MRRETURN(MATCH_NOMATCH);
2774                }
2775            c = *eptr++;            c = *eptr++;
2776            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2777            }            }
2778          }          }
2779    
# Line 1696  for (;;) Line 2793  for (;;)
2793            {            {
2794            for (fi = min;; fi++)            for (fi = min;; fi++)
2795              {              {
2796              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2797              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2798              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2799                if (eptr >= md->end_subject)
2800                  {
2801                  SCHECK_PARTIAL();
2802                  MRRETURN(MATCH_NOMATCH);
2803                  }
2804              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2805              if (c > 255)              if (c > 255)
2806                {                {
2807                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2808                }                }
2809              else              else
2810                {                {
2811                if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);                if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2812                }                }
2813              }              }
2814            }            }
# Line 1716  for (;;) Line 2818  for (;;)
2818            {            {
2819            for (fi = min;; fi++)            for (fi = min;; fi++)
2820              {              {
2821              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2822              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2823              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2824                if (eptr >= md->end_subject)
2825                  {
2826                  SCHECK_PARTIAL();
2827                  MRRETURN(MATCH_NOMATCH);
2828                  }
2829              c = *eptr++;              c = *eptr++;
2830              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2831              }              }
2832            }            }
2833          /* Control never gets here */          /* Control never gets here */
# Line 1739  for (;;) Line 2846  for (;;)
2846            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2847              {              {
2848              int len = 1;              int len = 1;
2849              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2850                  {
2851                  SCHECK_PARTIAL();
2852                  break;
2853                  }
2854              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2855              if (c > 255)              if (c > 255)
2856                {                {
# Line 1753  for (;;) Line 2864  for (;;)
2864              }              }
2865            for (;;)            for (;;)
2866              {              {
2867              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2868              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2869              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2870              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1765  for (;;) Line 2876  for (;;)
2876            {            {
2877            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2878              {              {
2879              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2880                  {
2881                  SCHECK_PARTIAL();
2882                  break;
2883                  }
2884              c = *eptr;              c = *eptr;
2885              if ((data[c/8] & (1 << (c&7))) == 0) break;              if ((data[c/8] & (1 << (c&7))) == 0) break;
2886              eptr++;              eptr++;
2887              }              }
2888            while (eptr >= pp)            while (eptr >= pp)
2889              {              {
2890              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2891              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2892              eptr--;              eptr--;
2893              }              }
2894            }            }
2895    
2896          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2897          }          }
2898        }        }
2899      /* Control never gets here */      /* Control never gets here */
2900    
2901    
2902      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2903      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
2904        mode, because Unicode properties are supported in non-UTF-8 mode. */
2905    
2906  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2907      case OP_XCLASS:      case OP_XCLASS:
# Line 1826  for (;;) Line 2942  for (;;)
2942    
2943        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2944          {          {
2945          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2946          GETCHARINC(c, eptr);            {
2947          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2948              MRRETURN(MATCH_NOMATCH);
2949              }
2950            GETCHARINCTEST(c, eptr);
2951            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2952          }          }
2953    
2954        /* 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 2963  for (;;)
2963          {          {
2964          for (fi = min;; fi++)          for (fi = min;; fi++)
2965            {            {
2966            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2967            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2968            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2969            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
2970            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
2971                SCHECK_PARTIAL();
2972                MRRETURN(MATCH_NOMATCH);
2973                }
2974              GETCHARINCTEST(c, eptr);
2975              if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2976            }            }
2977          /* Control never gets here */          /* Control never gets here */
2978          }          }
# Line 1860  for (;;) Line 2985  for (;;)
2985          for (i = min; i < max; i++)          for (i = min; i < max; i++)
2986            {            {
2987            int len = 1;            int len = 1;
2988            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
2989            GETCHARLEN(c, eptr, len);              {
2990                SCHECK_PARTIAL();
2991                break;
2992                }
2993              GETCHARLENTEST(c, eptr, len);
2994            if (!_pcre_xclass(c, data)) break;            if (!_pcre_xclass(c, data)) break;
2995            eptr += len;            eptr += len;
2996            }            }
2997          for(;;)          for(;;)
2998            {            {
2999            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3000            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3001            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3002            BACKCHAR(eptr)            if (utf8) BACKCHAR(eptr);
3003            }            }
3004          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3005          }          }
3006    
3007        /* Control never gets here */        /* Control never gets here */
# Line 1888  for (;;) Line 3017  for (;;)
3017        length = 1;        length = 1;
3018        ecode++;        ecode++;
3019        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3020        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3021        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3022            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3023            MRRETURN(MATCH_NOMATCH);
3024            }
3025          while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
3026        }        }
3027      else      else
3028  #endif  #endif
3029    
3030      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
3031        {        {
3032        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3033        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);          {
3034            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3035            MRRETURN(MATCH_NOMATCH);
3036            }
3037          if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
3038        ecode += 2;        ecode += 2;
3039        }        }
3040      break;      break;
3041    
3042      /* Match a single character, caselessly */      /* Match a single character, caselessly */
3043    
3044      case OP_CHARNC:      case OP_CHARI:
3045  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3046      if (utf8)      if (utf8)
3047        {        {
# Line 1912  for (;;) Line 3049  for (;;)
3049        ecode++;        ecode++;
3050        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3051    
3052        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3053            {
3054            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3055            MRRETURN(MATCH_NOMATCH);
3056            }
3057    
3058        /* 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
3059        can use the fast lookup table. */        can use the fast lookup table. */
3060    
3061        if (fc < 128)        if (fc < 128)
3062          {          {
3063          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3064          }          }
3065    
3066        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character */
3067    
3068        else        else
3069          {          {
3070          int dc;          unsigned int dc;
3071          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3072          ecode += length;          ecode += length;
3073    
# Line 1936  for (;;) Line 3077  for (;;)
3077          if (fc != dc)          if (fc != dc)
3078            {            {
3079  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3080            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3081  #endif  #endif
3082              RRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3083            }            }
3084          }          }
3085        }        }
# Line 1947  for (;;) Line 3088  for (;;)
3088    
3089      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
3090        {        {
3091        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3092        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3093            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3094            MRRETURN(MATCH_NOMATCH);
3095            }
3096          if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3097        ecode += 2;        ecode += 2;
3098        }        }
3099      break;      break;
3100    
3101      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3102    
3103      case OP_EXACT:      case OP_EXACT:
3104        case OP_EXACTI:
3105      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3106      ecode += 3;      ecode += 3;
3107      goto REPEATCHAR;      goto REPEATCHAR;
3108    
3109        case OP_POSUPTO:
3110        case OP_POSUPTOI:
3111        possessive = TRUE;
3112        /* Fall through */
3113    
3114      case OP_UPTO:      case OP_UPTO:
3115        case OP_UPTOI:
3116      case OP_MINUPTO:      case OP_MINUPTO:
3117        case OP_MINUPTOI:
3118      min = 0;      min = 0;
3119      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3120      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3121      ecode += 3;      ecode += 3;
3122      goto REPEATCHAR;      goto REPEATCHAR;
3123    
3124        case OP_POSSTAR:
3125        case OP_POSSTARI:
3126        possessive = TRUE;
3127        min = 0;
3128        max = INT_MAX;
3129        ecode++;
3130        goto REPEATCHAR;
3131    
3132        case OP_POSPLUS:
3133        case OP_POSPLUSI:
3134        possessive = TRUE;
3135        min = 1;
3136        max = INT_MAX;
3137        ecode++;
3138        goto REPEATCHAR;
3139    
3140        case OP_POSQUERY:
3141        case OP_POSQUERYI:
3142        possessive = TRUE;
3143        min = 0;
3144        max = 1;
3145        ecode++;
3146        goto REPEATCHAR;
3147    
3148      case OP_STAR:      case OP_STAR:
3149        case OP_STARI:
3150      case OP_MINSTAR:      case OP_MINSTAR:
3151        case OP_MINSTARI:
3152      case OP_PLUS:      case OP_PLUS:
3153        case OP_PLUSI:
3154      case OP_MINPLUS:      case OP_MINPLUS:
3155        case OP_MINPLUSI:
3156      case OP_QUERY:      case OP_QUERY:
3157        case OP_QUERYI:
3158      case OP_MINQUERY:      case OP_MINQUERY:
3159      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3160        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3161      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3162      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3163      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3164      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3165    
3166      /* 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. */  
3167    
3168      REPEATCHAR:      REPEATCHAR:
3169  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1991  for (;;) Line 3172  for (;;)
3172        length = 1;        length = 1;
3173        charptr = ecode;        charptr = ecode;
3174        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3175        ecode += length;        ecode += length;
3176    
3177        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1999  for (;;) Line 3179  for (;;)
3179    
3180        if (length > 1)        if (length > 1)
3181          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3182  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3183          int othercase;          unsigned int othercase;
3184          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3185              (othercase = _pcre_ucp_othercase(fc)) >= 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
              othercase >= 0)  
3186            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3187            else oclength = 0;
3188  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3189    
3190          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3191            {            {
3192            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3193            /* Need braces because of following else */              memcmp(eptr, charptr, length) == 0) eptr += length;
3194            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3195              else if (oclength > 0 &&
3196                       eptr <= md->end_subject - oclength &&
3197                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3198    #endif  /* SUPPORT_UCP */
3199            else            else
3200              {              {
3201              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3202              eptr += oclength;              MRRETURN(MATCH_NOMATCH);
3203              }              }
3204            }            }
3205    
# Line 2028  for (;;) Line 3209  for (;;)
3209            {            {
3210            for (fi = min;; fi++)            for (fi = min;; fi++)
3211              {              {
3212              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3213              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3214              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3215              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3216              /* Need braces because of following else */                memcmp(eptr, charptr, length) == 0) eptr += length;
3217              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3218                else if (oclength > 0 &&
3219                         eptr <= md->end_subject - oclength &&
3220                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3221    #endif  /* SUPPORT_UCP */
3222              else              else
3223                {                {
3224                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3225                eptr += oclength;                MRRETURN(MATCH_NOMATCH);
3226                }                }
3227              }              }
3228            /* Control never gets here */            /* Control never gets here */
3229            }            }
3230          else  
3231            else  /* Maximize */
3232            {            {
3233            pp = eptr;            pp = eptr;
3234            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3235              {              {
3236              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3237              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, length) == 0) eptr += length;
3238              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3239                else if (oclength > 0 &&
3240                         eptr <= md->end_subject - oclength &&
3241                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3242    #endif  /* SUPPORT_UCP */
3243              else              else
3244                {                {
3245                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3246                eptr += oclength;                break;
3247                }                }
3248              }              }
3249            while (eptr >= pp)  
3250             {            if (possessive) continue;
3251             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3252             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3253             eptr -= length;              {
3254             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3255            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3256                if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3257    #ifdef SUPPORT_UCP
3258                eptr--;
3259                BACKCHAR(eptr);
3260    #else   /* without SUPPORT_UCP */
3261                eptr -= length;
3262    #endif  /* SUPPORT_UCP */
3263                }
3264            }            }
3265          /* Control never gets here */          /* Control never gets here */
3266          }          }
# Line 2075  for (;;) Line 3273  for (;;)
3273  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3274    
3275      /* When not in UTF-8 mode, load a single-byte character. */      /* When not in UTF-8 mode, load a single-byte character. */
3276        {  
3277        if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);      fc = *ecode++;
       fc = *ecode++;  
       }  
3278    
3279      /* The value of fc at this point is always less than 256, though we may or      /* The value of fc at this point is always less than 256, though we may or
3280      may not be in UTF-8 mode. The code is duplicated for the caseless and      may not be in UTF-8 mode. The code is duplicated for the caseless and
# Line 2092  for (;;) Line 3288  for (;;)
3288      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3289        max, eptr));        max, eptr));
3290    
3291      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3292        {        {
3293        fc = md->lcc[fc];        fc = md->lcc[fc];
3294        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3295          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3296            if (eptr >= md->end_subject)
3297              {
3298              SCHECK_PARTIAL();
3299              MRRETURN(MATCH_NOMATCH);
3300              }
3301            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3302            }
3303        if (min == max) continue;        if (min == max) continue;
3304        if (minimize)        if (minimize)
3305          {          {
3306          for (fi = min;; fi++)          for (fi = min;; fi++)
3307            {            {
3308            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3309            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3310            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3311                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3312              RRETURN(MATCH_NOMATCH);              {
3313                SCHECK_PARTIAL();
3314                MRRETURN(MATCH_NOMATCH);
3315                }
3316              if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3317            }            }
3318          /* Control never gets here */          /* Control never gets here */
3319          }          }
3320        else        else  /* Maximize */
3321          {          {
3322          pp = eptr;          pp = eptr;
3323          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3324            {            {
3325            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3326                {
3327                SCHECK_PARTIAL();
3328                break;
3329                }
3330              if (fc != md->lcc[*eptr]) break;
3331            eptr++;            eptr++;
3332            }            }
3333    
3334            if (possessive) continue;
3335    
3336          while (eptr >= pp)          while (eptr >= pp)
3337            {            {
3338            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3339            eptr--;            eptr--;
3340            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3341            }            }
3342          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3343          }          }
3344        /* Control never gets here */        /* Control never gets here */
3345        }        }
# Line 2133  for (;;) Line 3348  for (;;)
3348    
3349      else      else
3350        {        {
3351        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3352            {
3353            if (eptr >= md->end_subject)
3354              {
3355              SCHECK_PARTIAL();
3356              MRRETURN(MATCH_NOMATCH);
3357              }
3358            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3359            }
3360    
3361        if (min == max) continue;        if (min == max) continue;
3362    
3363        if (minimize)        if (minimize)
3364          {          {
3365          for (fi = min;; fi++)          for (fi = min;; fi++)
3366            {            {
3367            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3368            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3369            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3370              RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
3371                {
3372                SCHECK_PARTIAL();
3373                MRRETURN(MATCH_NOMATCH);
3374                }
3375              if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3376            }            }
3377          /* Control never gets here */          /* Control never gets here */
3378          }          }
3379        else        else  /* Maximize */
3380          {          {
3381          pp = eptr;          pp = eptr;
3382          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3383            {            {
3384            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3385                {
3386                SCHECK_PARTIAL();
3387                break;
3388                }
3389              if (fc != *eptr) break;
3390            eptr++;            eptr++;
3391            }            }
3392            if (possessive) continue;
3393    
3394          while (eptr >= pp)          while (eptr >= pp)
3395            {            {
3396            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3397            eptr--;            eptr--;
3398            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3399            }            }
3400          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3401          }          }
3402        }        }
3403      /* Control never gets here */      /* Control never gets here */
# Line 2169  for (;;) Line 3406  for (;;)
3406      checking can be multibyte. */      checking can be multibyte. */
3407    
3408      case OP_NOT:      case OP_NOT:
3409      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3410        if (eptr >= md->end_subject)
3411          {
3412          SCHECK_PARTIAL();
3413          MRRETURN(MATCH_NOMATCH);
3414          }
3415      ecode++;      ecode++;
3416      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3417      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3418        {        {
3419  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3420        if (c < 256)        if (c < 256)
3421  #endif  #endif
3422        c = md->lcc[c];        c = md->lcc[c];
3423        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3424        }        }
3425      else      else    /* Caseful */
3426        {        {
3427        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3428        }        }
3429      break;      break;
3430    
# Line 2194  for (;;) Line 3436  for (;;)
3436      about... */      about... */
3437    
3438      case OP_NOTEXACT:      case OP_NOTEXACT:
3439        case OP_NOTEXACTI:
3440      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3441      ecode += 3;      ecode += 3;
3442      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3443    
3444      case OP_NOTUPTO:      case OP_NOTUPTO:
3445        case OP_NOTUPTOI:
3446      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3447        case OP_NOTMINUPTOI:
3448        min = 0;
3449        max = GET2(ecode, 1);
3450        minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3451        ecode += 3;
3452        goto REPEATNOTCHAR;
3453    
3454        case OP_NOTPOSSTAR:
3455        case OP_NOTPOSSTARI:
3456        possessive = TRUE;
3457        min = 0;
3458        max = INT_MAX;
3459        ecode++;
3460        goto REPEATNOTCHAR;
3461    
3462        case OP_NOTPOSPLUS:
3463        case OP_NOTPOSPLUSI:
3464        possessive = TRUE;
3465        min = 1;
3466        max = INT_MAX;
3467        ecode++;
3468        goto REPEATNOTCHAR;
3469    
3470        case OP_NOTPOSQUERY:
3471        case OP_NOTPOSQUERYI:
3472        possessive = TRUE;
3473        min = 0;
3474        max = 1;
3475        ecode++;
3476        goto REPEATNOTCHAR;
3477    
3478        case OP_NOTPOSUPTO:
3479        case OP_NOTPOSUPTOI:
3480        possessive = TRUE;
3481      min = 0;      min = 0;
3482      max = GET2(ecode, 1);      max = GET2(ecode, 1);
     minimize = *ecode == OP_NOTMINUPTO;  
3483      ecode += 3;      ecode += 3;
3484      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3485    
3486      case OP_NOTSTAR:      case OP_NOTSTAR:
3487        case OP_NOTSTARI:
3488      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3489        case OP_NOTMINSTARI:
3490      case OP_NOTPLUS:      case OP_NOTPLUS:
3491        case OP_NOTPLUSI:
3492      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3493        case OP_NOTMINPLUSI:
3494      case OP_NOTQUERY:      case OP_NOTQUERY:
3495        case OP_NOTQUERYI:
3496      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3497      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3498        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3499      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3500      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3501      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3502      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3503    
3504      /* Common code for all repeated single-byte matches. We can give up quickly      /* Common code for all repeated single-byte matches. */
     if there are fewer than the minimum number of bytes left in the  
     subject. */  
3505    
3506      REPEATNOTCHAR:      REPEATNOTCHAR:
     if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3507      fc = *ecode++;      fc = *ecode++;
3508    
3509      /* The code is duplicated for the caseless and caseful cases, for speed,      /* The code is duplicated for the caseless and caseful cases, for speed,
# Line 2237  for (;;) Line 3517  for (;;)
3517      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3518        max, eptr));        max, eptr));
3519    
3520      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3521        {        {
3522        fc = md->lcc[fc];        fc = md->lcc[fc];
3523