/[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 716 by ph10, Tue Oct 4 16:38:05 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 };
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)
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > md->start_match)  
     md->hitend = TRUE;  
   
   /* 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. */  
   
   if (op > OP_BRA)  
702      {      {
703      number = op - OP_BRA;      case OP_MARK:
704        markptr = ecode + 2;
705      /* For extended extraction brackets (large number), we have to fish out the      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
706      number from a dummy opcode at the start. */        eptrb, RM55);
707    
708      if (number > EXTRACT_BASIC_MAX)      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
709        number = GET2(ecode, 2+LINK_SIZE);      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 a capturing bracket, other than those that are possessive with an
798        unlimited repeat. If there is space in the offset vector, save the current
799        subject position in the working slot at the top of the vector. We mustn't
800        change the current values of the data slot, because they may be set from a
801        previous iteration of this group, and be referred to by a reference inside
802        the group. A failure to match might occur after the group has succeeded,
803        if something later on doesn't match. For this reason, we need to restore
804        the working value and also the values of the final offsets, in case they
805        were set by a previous iteration of the same bracket.
806    
807        If there isn't enough space in the offset vector, treat this as if it were
808        a non-capturing bracket. Don't worry about setting the flag for the error
809        case here; that is handled in the code for KET. */
810    
811        case OP_CBRA:
812        case OP_SCBRA:
813        number = GET2(ecode, 1+LINK_SIZE);
814      offset = number << 1;      offset = number << 1;
815    
816  #ifdef DEBUG  #ifdef PCRE_DEBUG
817      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
818        printf("subject=");
819      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
820      printf("\n");      printf("\n");
821  #endif  #endif
# Line 622  for (;;) Line 828  for (;;)
828        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
829    
830        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
831        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
832            (int)(eptr - md->start_subject);
833    
834        do        for (;;)
835          {          {
836          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
837            match_isgroup);          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
838              eptrb, RM1);
839            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
840    
841            /* If we backed up to a THEN, check whether it is within the current
842            branch by comparing the address of the THEN that is passed back with
843            the end of the branch. If it is within the current branch, and the
844            branch is one of two or more alternatives (it either starts or ends
845            with OP_ALT), we have reached the limit of THEN's action, so convert
846            the return code to NOMATCH, which will cause normal backtracking to
847            happen from now on. Otherwise, THEN is passed back to an outer
848            alternative. This implements Perl's treatment of parenthesized groups,
849            where a group not containing | does not affect the current alternative,
850            that is, (X) is NOT the same as (X|(*F)). */
851    
852            if (rrc == MATCH_THEN)
853              {
854              next = ecode + GET(ecode,1);
855              if (md->start_match_ptr < next &&
856                  (*ecode == OP_ALT || *next == OP_ALT))
857                rrc = MATCH_NOMATCH;
858              }
859    
860            /* Anything other than NOMATCH is passed back. */
861    
862          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
863          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
864          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
865            if (*ecode != OP_ALT) break;
866          }          }
       while (*ecode == OP_ALT);  
867    
868        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
869        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
870        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
871        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
872    
873        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
874    
875          if (md->mark == NULL) md->mark = markptr;
876          RRETURN(rrc);
877        }        }
878    
879      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
880        as a non-capturing bracket. */
881    
882      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
883      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
884    
885    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
886    
887    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
888      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
889      case OP_BRA:     /* Non-capturing bracket: optimized */  
890      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
891        repeat. Loop for all the alternatives.
892    
893        When we get to the final alternative within the brackets, we used to return
894        the result of a recursive call to match() whatever happened so it was
895        possible to reduce stack usage by turning this into a tail recursion,
896        except in the case of a possibly empty group. However, now that there is
897        the possiblity of (*THEN) occurring in the final alternative, this
898        optimization is no longer always possible.
899    
900        We can optimize if we know there are no (*THEN)s in the pattern; at present
901        this is the best that can be done.
902    
903      /* Loop for all the alternatives */      MATCH_ONCE is returned when the end of an atomic group is successfully
904        reached, but subsequent matching fails. It passes back up the tree (causing
905        captured values to be reset) until the original atomic group level is
906        reached. This is tested by comparing md->once_target with the start of the
907        group. At this point, the return is converted into MATCH_NOMATCH so that
908        previous backup points can be taken. */
909    
910        case OP_ONCE:
911        case OP_BRA:
912        case OP_SBRA:
913        DPRINTF(("start non-capturing bracket\n"));
914    
915      for (;;)      for (;;)
916        {        {
917        /* 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;
918        return the result of a recursive call to match() whatever happened. We  
919        can reduce stack usage by turning this into a tail recursion. */        /* If this is not a possibly empty group, and there are no (*THEN)s in
920          the pattern, and this is the final alternative, optimize as described
921        if (ecode[GET(ecode, 1)] != OP_ALT)        above. */
922         {  
923         ecode += 1 + LINK_SIZE;        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
924         flags = match_isgroup;          {
925         DPRINTF(("bracket 0 tail recursion\n"));          ecode += _pcre_OP_lengths[*ecode];
926         goto TAIL_RECURSE;          goto TAIL_RECURSE;
927         }          }
928    
929          /* In all other cases, we have to make another call to match(). */
930    
931          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
932            RM2);
933    
934          /* See comment in the code for capturing groups above about handling
935          THEN. */
936    
937          if (rrc == MATCH_THEN)
938            {
939            next = ecode + GET(ecode,1);
940            if (md->start_match_ptr < next &&
941                (*ecode == OP_ALT || *next == OP_ALT))
942              rrc = MATCH_NOMATCH;
943            }
944    
945          if (rrc != MATCH_NOMATCH)
946            {
947            if (rrc == MATCH_ONCE)
948              {
949              const uschar *scode = ecode;
950              if (*scode != OP_ONCE)           /* If not at start, find it */
951                {
952                while (*scode == OP_ALT) scode += GET(scode, 1);
953                scode -= GET(scode, 1);
954                }
955              if (md->once_target == scode) rrc = MATCH_NOMATCH;
956              }
957            RRETURN(rrc);
958            }
959          ecode += GET(ecode, 1);
960          if (*ecode != OP_ALT) break;
961          }
962    
963        if (md->mark == NULL) md->mark = markptr;
964        RRETURN(MATCH_NOMATCH);
965    
966        /* Handle possessive capturing brackets with an unlimited repeat. We come
967        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
968        handled similarly to the normal case above. However, the matching is
969        different. The end of these brackets will always be OP_KETRPOS, which
970        returns MATCH_KETRPOS without going further in the pattern. By this means
971        we can handle the group by iteration rather than recursion, thereby
972        reducing the amount of stack needed. */
973    
974        case OP_CBRAPOS:
975        case OP_SCBRAPOS:
976        allow_zero = FALSE;
977    
978        POSSESSIVE_CAPTURE:
979        number = GET2(ecode, 1+LINK_SIZE);
980        offset = number << 1;
981    
982    #ifdef PCRE_DEBUG
983        printf("start possessive bracket %d\n", number);
984        printf("subject=");
985        pchars(eptr, 16, TRUE, md);
986        printf("\n");
987    #endif
988    
989        if (offset < md->offset_max)
990          {
991          matched_once = FALSE;
992          code_offset = ecode - md->start_code;
993    
994          save_offset1 = md->offset_vector[offset];
995          save_offset2 = md->offset_vector[offset+1];
996          save_offset3 = md->offset_vector[md->offset_end - number];
997          save_capture_last = md->capture_last;
998    
999          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1000    
1001          /* Each time round the loop, save the current subject position for use
1002          when the group matches. For MATCH_MATCH, the group has matched, so we
1003          restart it with a new subject starting position, remembering that we had
1004          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1005          usual. If we haven't matched any alternatives in any iteration, check to
1006          see if a previous iteration matched. If so, the group has matched;
1007          continue from afterwards. Otherwise it has failed; restore the previous
1008          capture values before returning NOMATCH. */
1009    
1010          for (;;)
1011            {
1012            md->offset_vector[md->offset_end - number] =
1013              (int)(eptr - md->start_subject);
1014            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1015            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1016              eptrb, RM63);
1017            if (rrc == MATCH_KETRPOS)
1018              {
1019              offset_top = md->end_offset_top;
1020              eptr = md->end_match_ptr;
1021              ecode = md->start_code + code_offset;
1022              save_capture_last = md->capture_last;
1023              matched_once = TRUE;
1024              continue;
1025              }
1026    
1027            /* See comment in the code for capturing groups above about handling
1028            THEN. */
1029    
1030            if (rrc == MATCH_THEN)
1031              {
1032              next = ecode + GET(ecode,1);
1033              if (md->start_match_ptr < next &&
1034                  (*ecode == OP_ALT || *next == OP_ALT))
1035                rrc = MATCH_NOMATCH;
1036              }
1037    
1038            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1039            md->capture_last = save_capture_last;
1040            ecode += GET(ecode, 1);
1041            if (*ecode != OP_ALT) break;
1042            }
1043    
1044          if (!matched_once)
1045            {
1046            md->offset_vector[offset] = save_offset1;
1047            md->offset_vector[offset+1] = save_offset2;
1048            md->offset_vector[md->offset_end - number] = save_offset3;
1049            }
1050    
1051          if (md->mark == NULL) md->mark = markptr;
1052          if (allow_zero || matched_once)
1053            {
1054            ecode += 1 + LINK_SIZE;
1055            break;
1056            }
1057    
1058          RRETURN(MATCH_NOMATCH);
1059          }
1060    
1061        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1062        as a non-capturing bracket. */
1063    
1064        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1065        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1066    
1067        /* For non-final alternatives, continue the loop for a NOMATCH result;      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1068        otherwise return. */  
1069        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1070        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1071    
1072        /* Non-capturing possessive bracket with unlimited repeat. We come here
1073        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1074        without the capturing complication. It is written out separately for speed
1075        and cleanliness. */
1076    
1077        case OP_BRAPOS:
1078        case OP_SBRAPOS:
1079        allow_zero = FALSE;
1080    
1081        POSSESSIVE_NON_CAPTURE:
1082        matched_once = FALSE;
1083        code_offset = ecode - md->start_code;
1084    
1085        for (;;)
1086          {
1087          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1088          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1089            eptrb, RM48);
1090          if (rrc == MATCH_KETRPOS)
1091            {
1092            offset_top = md->end_offset_top;
1093            eptr = md->end_match_ptr;
1094            ecode = md->start_code + code_offset;
1095            matched_once = TRUE;
1096            continue;
1097            }
1098    
1099          /* See comment in the code for capturing groups above about handling
1100          THEN. */
1101    
1102          if (rrc == MATCH_THEN)
1103            {
1104            next = ecode + GET(ecode,1);
1105            if (md->start_match_ptr < next &&
1106                (*ecode == OP_ALT || *next == OP_ALT))
1107              rrc = MATCH_NOMATCH;
1108            }
1109    
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,  
         match_isgroup);  
1110        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1111        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1112          if (*ecode != OP_ALT) break;
1113          }
1114    
1115        if (matched_once || allow_zero)
1116          {
1117          ecode += 1 + LINK_SIZE;
1118          break;
1119        }        }
1120        RRETURN(MATCH_NOMATCH);
1121    
1122      /* Control never reaches here. */      /* Control never reaches here. */
1123    
1124      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1125      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1126      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1127      exactly what going to the ket would do. As there is only one branch to be      exactly what going to the ket would do. */
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1128    
1129      case OP_COND:      case OP_COND:
1130      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      case OP_SCOND:
1131        codelink = GET(ecode, 1);
1132    
1133        /* Because of the way auto-callout works during compile, a callout item is
1134        inserted between OP_COND and an assertion condition. */
1135    
1136        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1137          {
1138          if (pcre_callout != NULL)
1139            {
1140            pcre_callout_block cb;
1141            cb.version          = 2;   /* Version 1 of the callout block */
1142            cb.callout_number   = ecode[LINK_SIZE+2];
1143            cb.offset_vector    = md->offset_vector;
1144            cb.subject          = (PCRE_SPTR)md->start_subject;
1145            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1146            cb.start_match      = (int)(mstart - md->start_subject);
1147            cb.current_position = (int)(eptr - md->start_subject);
1148            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1149            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1150            cb.capture_top      = offset_top/2;
1151            cb.capture_last     = md->capture_last;
1152            cb.callout_data     = md->callout_data;
1153            cb.mark             = markptr;
1154            if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1155            if (rrc < 0) RRETURN(rrc);
1156            }
1157          ecode += _pcre_OP_lengths[OP_CALLOUT];
1158          }
1159    
1160        condcode = ecode[LINK_SIZE+1];
1161    
1162        /* Now see what the actual condition is */
1163    
1164        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1165          {
1166          if (md->recursive == NULL)                /* Not recursing => FALSE */
1167            {
1168            condition = FALSE;
1169            ecode += GET(ecode, 1);
1170            }
1171          else
1172            {
1173            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1174            condition =  (recno == RREF_ANY || recno == md->recursive->group_num);
1175    
1176            /* If the test is for recursion into a specific subpattern, and it is
1177            false, but the test was set up by name, scan the table to see if the
1178            name refers to any other numbers, and test them. The condition is true
1179            if any one is set. */
1180    
1181            if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
1182              {
1183              uschar *slotA = md->name_table;
1184              for (i = 0; i < md->name_count; i++)
1185                {
1186                if (GET2(slotA, 0) == recno) break;
1187                slotA += md->name_entry_size;
1188                }
1189    
1190              /* Found a name for the number - there can be only one; duplicate
1191              names for different numbers are allowed, but not vice versa. First
1192              scan down for duplicates. */
1193    
1194              if (i < md->name_count)
1195                {
1196                uschar *slotB = slotA;
1197                while (slotB > md->name_table)
1198                  {
1199                  slotB -= md->name_entry_size;
1200                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1201                    {
1202                    condition = GET2(slotB, 0) == md->recursive->group_num;
1203                    if (condition) break;
1204                    }
1205                  else break;
1206                  }
1207    
1208                /* Scan up for duplicates */
1209    
1210                if (!condition)
1211                  {
1212                  slotB = slotA;
1213                  for (i++; i < md->name_count; i++)
1214                    {
1215                    slotB += md->name_entry_size;
1216                    if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1217                      {
1218                      condition = GET2(slotB, 0) == md->recursive->group_num;
1219                      if (condition) break;
1220                      }
1221                    else break;
1222                    }
1223                  }
1224                }
1225              }
1226    
1227            /* Chose branch according to the condition */
1228    
1229            ecode += condition? 3 : GET(ecode, 1);
1230            }
1231          }
1232    
1233        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1234        {        {
1235        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1236        condition = (offset == CREF_RECURSE * 2)?        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1237          (md->recursive != NULL) :  
1238          (offset < offset_top && md->offset_vector[offset] >= 0);        /* If the numbered capture is unset, but the reference was by name,
1239        ecode += condition? (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1));        scan the table to see if the name refers to any other numbers, and test
1240        flags = match_isgroup;        them. The condition is true if any one is set. This is tediously similar
1241        goto TAIL_RECURSE;        to the code above, but not close enough to try to amalgamate. */
1242    
1243          if (!condition && condcode == OP_NCREF)
1244            {
1245            int refno = offset >> 1;
1246            uschar *slotA = md->name_table;
1247    
1248            for (i = 0; i < md->name_count; i++)
1249              {
1250              if (GET2(slotA, 0) == refno) break;
1251              slotA += md->name_entry_size;
1252              }
1253    
1254            /* Found a name for the number - there can be only one; duplicate names
1255            for different numbers are allowed, but not vice versa. First scan down
1256            for duplicates. */
1257    
1258            if (i < md->name_count)
1259              {
1260              uschar *slotB = slotA;
1261              while (slotB > md->name_table)
1262                {
1263                slotB -= md->name_entry_size;
1264                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1265                  {
1266                  offset = GET2(slotB, 0) << 1;
1267                  condition = offset < offset_top &&
1268                    md->offset_vector[offset] >= 0;
1269                  if (condition) break;
1270                  }
1271                else break;
1272                }
1273    
1274              /* Scan up for duplicates */
1275    
1276              if (!condition)
1277                {
1278                slotB = slotA;
1279                for (i++; i < md->name_count; i++)
1280                  {
1281                  slotB += md->name_entry_size;
1282                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1283                    {
1284                    offset = GET2(slotB, 0) << 1;
1285                    condition = offset < offset_top &&
1286                      md->offset_vector[offset] >= 0;
1287                    if (condition) break;
1288                    }
1289                  else break;
1290                  }
1291                }
1292              }
1293            }
1294    
1295          /* Chose branch according to the condition */
1296    
1297          ecode += condition? 3 : GET(ecode, 1);
1298          }
1299    
1300        else if (condcode == OP_DEF)     /* DEFINE - always false */
1301          {
1302          condition = FALSE;
1303          ecode += GET(ecode, 1);
1304        }        }
1305    
1306      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1307      the final argument TRUE causes it to stop at the end of an assertion. */      md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1308        an assertion. */
1309    
1310      else      else
1311        {        {
1312        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1313            match_condassert | match_isgroup);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1314        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1315          {          {
1316          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          if (md->end_offset_top > offset_top)
1317              offset_top = md->end_offset_top;  /* Captures may have happened */
1318            condition = TRUE;
1319            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1320          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1321          }          }
1322        else if (rrc != MATCH_NOMATCH)  
1323          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1324          assertion; it is therefore treated as NOMATCH. */
1325    
1326          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1327          {          {
1328          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1329          }          }
1330        else ecode += GET(ecode, 1);        else
1331            {
1332            condition = FALSE;
1333            ecode += codelink;
1334            }
1335          }
1336    
1337        /* We are now at the branch that is to be obeyed. As there is only one, can
1338        use tail recursion to avoid using another stack frame, except when there is
1339        unlimited repeat of a possibly empty group. In the latter case, a recursive
1340        call to match() is always required, unless the second alternative doesn't
1341        exist, in which case we can just plough on. Note that, for compatibility
1342        with Perl, the | in a conditional group is NOT treated as creating two
1343        alternatives. If a THEN is encountered in the branch, it propagates out to
1344        the enclosing alternative (unless nested in a deeper set of alternatives,
1345        of course). */
1346    
1347        if (condition || *ecode == OP_ALT)
1348          {
1349          if (op != OP_SCOND)
1350            {
1351            ecode += 1 + LINK_SIZE;
1352            goto TAIL_RECURSE;
1353            }
1354    
1355          md->match_function_type = MATCH_CBEGROUP;
1356          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1357          RRETURN(rrc);
1358          }
1359    
1360        /* We are now at the branch that is to be obeyed. As there is only one,       /* Condition false & no alternative; continue after the group. */
       we can use tail recursion to avoid using another stack frame. */  
1361    
1362        else
1363          {
1364        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
1365        }        }
1366      /* Control never reaches here */      break;
1367    
1368    
1369        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1370        to close any currently open capturing brackets. */
1371    
1372      /* Skip over conditional reference or large extraction number data if      case OP_CLOSE:
1373      encountered. */      number = GET2(ecode, 1);
1374        offset = number << 1;
1375    
1376    #ifdef PCRE_DEBUG
1377          printf("end bracket %d at *ACCEPT", number);
1378          printf("\n");
1379    #endif
1380    
1381      case OP_CREF:      md->capture_last = number;
1382      case OP_BRANUMBER:      if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1383          {
1384          md->offset_vector[offset] =
1385            md->offset_vector[md->offset_end - number];
1386          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1387          if (offset_top <= offset) offset_top = offset + 2;
1388          }
1389      ecode += 3;      ecode += 3;
1390      break;      break;
1391    
1392      /* End of the pattern. If we are in a recursion, we should restore the  
1393      offsets appropriately and continue from after the call. */      /* End of the pattern, either real or forced. */
1394    
1395      case OP_END:      case OP_END:
1396      if (md->recursive != NULL && md->recursive->group_num == 0)      case OP_ACCEPT:
1397        {      case OP_ASSERT_ACCEPT:
       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;  
       }  
1398    
1399      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* If we have matched an empty string, fail if not in an assertion and not
1400      string - backtracking will then try other alternatives, if any. */      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1401        is set and we have matched at the start of the subject. In both cases,
1402        backtracking will then try other alternatives, if any. */
1403    
1404        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1405             md->recursive == NULL &&
1406             (md->notempty ||
1407               (md->notempty_atstart &&
1408                 mstart == md->start_subject + md->start_offset)))
1409          MRRETURN(MATCH_NOMATCH);
1410    
1411        /* Otherwise, we have a match. */
1412    
1413        md->end_match_ptr = eptr;           /* Record where we ended */
1414        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1415        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1416    
1417      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* For some reason, the macros don't work properly if an expression is
1418      md->end_match_ptr = eptr;          /* Record where we ended */      given as the argument to MRRETURN when the heap is in use. */
1419      md->end_offset_top = offset_top;   /* and how many extracts were taken */  
1420      RRETURN(MATCH_MATCH);      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1421        MRRETURN(rrc);
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1422    
1423      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1424      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1425      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1426      start of each branch to move the current point backwards, so the code at      start of each branch to move the current point backwards, so the code at
1427      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1428        of a condition, we want to return immediately afterwards. The caller of
1429        this incarnation of the match() function will have set MATCH_CONDASSERT in
1430        md->match_function type, and one of these opcodes will be the first opcode
1431        that is processed. We use a local variable that is preserved over calls to
1432        match() to remember this case. */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
1435      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1436        if (md->match_function_type == MATCH_CONDASSERT)
1437          {
1438          condassert = TRUE;
1439          md->match_function_type = 0;
1440          }
1441        else condassert = FALSE;
1442    
1443      do      do
1444        {        {
1445        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1446          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1447        if (rrc == MATCH_MATCH) break;          {
1448        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          mstart = md->start_match_ptr;   /* In case \K reset it */
1449            markptr = md->mark;
1450            break;
1451            }
1452    
1453          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1454          as NOMATCH. */
1455    
1456          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1457        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1458        }        }
1459      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1460      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1461        if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1462    
1463      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1464    
1465      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1466    
1467      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1468      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 798  for (;;) Line 1472  for (;;)
1472      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1473      continue;      continue;
1474    
1475      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1476        PRUNE, or COMMIT means we must assume failure without checking subsequent
1477        branches. */
1478    
1479      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1480      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1481        if (md->match_function_type == MATCH_CONDASSERT)
1482          {
1483          condassert = TRUE;
1484          md->match_function_type = 0;
1485          }
1486        else condassert = FALSE;
1487    
1488      do      do
1489        {        {
1490        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1491          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1492        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1493        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          {
1494            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1495            break;
1496            }
1497    
1498          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1499          as NOMATCH. */
1500    
1501          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1502        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1503        }        }
1504      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1505    
1506      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1507    
1508      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1509      continue;      continue;
# Line 826  for (;;) Line 1517  for (;;)
1517  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1518      if (utf8)      if (utf8)
1519        {        {
1520        c = GET(ecode,1);        i = GET(ecode, 1);
1521        for (i = 0; i < c; i++)        while (i-- > 0)
1522          {          {
1523          eptr--;          eptr--;
1524          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1525          BACKCHAR(eptr)          BACKCHAR(eptr);
1526          }          }
1527        }        }
1528      else      else
# Line 840  for (;;) Line 1531  for (;;)
1531      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1532    
1533        {        {
1534        eptr -= GET(ecode,1);        eptr -= GET(ecode, 1);
1535        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1536        }        }
1537    
1538      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1539    
1540        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1541      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1542      break;      break;
1543    
# Line 857  for (;;) Line 1549  for (;;)
1549      if (pcre_callout != NULL)      if (pcre_callout != NULL)
1550        {        {
1551        pcre_callout_block cb;        pcre_callout_block cb;
1552        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1553        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1554        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1555        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1556        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1557        cb.start_match      = md->start_match - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1558        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1559        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1560        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1561        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1562        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1563        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1564        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = markptr;
1565          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1566        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1567        }        }
1568      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 879  for (;;) Line 1572  for (;;)
1572      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
1573      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1574    
1575      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1576      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
1577      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
1578      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
1579      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
1580      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
1581      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.  
1582    
1583      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
1584      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
1585      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1586        a lot, so he is not to blame for the current way it works. */
1587    
1588      case OP_RECURSE:      case OP_RECURSE:
1589        {        {
1590        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1591        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1592    
1593        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1594        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1595            GET2(callpat, 1 + LINK_SIZE);
1596    
1597        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1598          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1599          caught at compile time.) */
1600    
1601          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1602            if (recno == ri->group_num && eptr == ri->subject_position)
1603              RRETURN(PCRE_ERROR_RECURSELOOP);
1604    
1605        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1606    
1607          new_recursive.group_num = recno;
1608          new_recursive.subject_position = eptr;
1609        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1610        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1611    
1612        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1613    
1614        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1615    
1616        /* Now save the offset data. */        /* Now save the offset data */
1617    
1618        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1619        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 926  for (;;) Line 1624  for (;;)
1624            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1625          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1626          }          }
   
1627        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1628              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1629    
1630        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1631        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1632          might be changed, so reset it before looping. */
1633    
1634        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1635          cbegroup = (*callpat >= OP_SBRA);
1636        do        do
1637          {          {
1638          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1639              eptrb, match_isgroup);          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1640          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1641            memcpy(md->offset_vector, new_recursive.offset_save,
1642                new_recursive.saved_max * sizeof(int));
1643            md->recursive = new_recursive.prevrec;
1644            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1645            {            {
1646            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1647            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1648              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1649            RRETURN(MATCH_MATCH);  
1650              /* Set where we got to in the subject, and reset the start in case
1651              it was changed by \K. This *is* propagated back out of a recursion,
1652              for Perl compatibility. */
1653    
1654              eptr = md->end_match_ptr;
1655              mstart = md->start_match_ptr;
1656              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1657            }            }
1658          else if (rrc != MATCH_NOMATCH)  
1659            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1660            as NOMATCH. */
1661    
1662            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1663            {            {
1664            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1665              if (new_recursive.offset_save != stacksave)
1666                (pcre_free)(new_recursive.offset_save);
1667            RRETURN(rrc);            RRETURN(rrc);
1668            }            }
1669    
1670          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1671          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1672          }          }
1673        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 965  for (;;) Line 1676  for (;;)
1676        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1677        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1678          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1679        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);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode,1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
1680        }        }
1681    
1682      if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1683        {      break;
       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 */  
1684    
1685      /* 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
1686      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1051  for (;;) Line 1689  for (;;)
1689      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1690      break;      break;
1691    
1692      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1693      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
1694      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
1695      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
1696      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1697    
1698      case OP_BRAZERO:      case OP_BRAZERO:
1699        {      next = ecode + 1;
1700        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1701        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1702        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1703        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1704      break;      break;
1705    
1706      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1707        {      next = ecode + 1;
1708        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1709        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1710        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1711          match_isgroup);      ecode++;
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode++;  
       }  
1712      break;      break;
1713    
1714      /* End of a group, repeated or non-repeating. If we are at the end of      case OP_SKIPZERO:
1715      an assertion "group", stop matching and return MATCH_MATCH, but record the      next = ecode+1;
1716      current high water mark for use by positive assertions. Do this also      do next += GET(next,1); while (*next == OP_ALT);
1717      for the "once" (not-backup up) groups. */      ecode = next + 1 + LINK_SIZE;
1718        break;
1719    
1720        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1721        here; just jump to the group, with allow_zero set TRUE. */
1722    
1723        case OP_BRAPOSZERO:
1724        op = *(++ecode);
1725        allow_zero = TRUE;
1726        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1727          goto POSSESSIVE_NON_CAPTURE;
1728    
1729        /* End of a group, repeated or non-repeating. */
1730    
1731      case OP_KET:      case OP_KET:
1732      case OP_KETRMIN:      case OP_KETRMIN:
1733      case OP_KETRMAX:      case OP_KETRMAX:
1734        case OP_KETRPOS:
1735      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
     saved_eptr = eptrb->epb_saved_eptr;  
1736    
1737      /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1738        infinite repeats of empty string matches, retrieve the subject start from
1739        the chain. Otherwise, set it NULL. */
1740    
1741      eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1742          {
1743          saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1744          eptrb = eptrb->epb_prev;              /* Backup to previous group */
1745          }
1746        else saved_eptr = NULL;
1747    
1748        /* If we are at the end of an assertion group, stop matching and return
1749        MATCH_MATCH, but record the current high water mark for use by positive
1750        assertions. We also need to record the match start in case it was changed
1751        by \K. */
1752    
1753      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1754          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1755        {        {
1756        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1757        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1758        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1759          MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1760        }        }
1761    
1762      /* In all other cases except a conditional group we have to check the      /* For capturing groups we have to check the group number back at the start
1763      group number back at the start and if necessary complete handling an      and if necessary complete handling an extraction by setting the offsets and
1764      extraction by setting the offsets and bumping the high water mark. */      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1765        into group 0, so it won't be picked up here. Instead, we catch it when the
1766        OP_END is reached. Other recursion is handled here. We just have to record
1767        the current subject position and start match pointer and give a MATCH
1768        return. */
1769    
1770      if (*prev != OP_COND)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1771            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1772        {        {
1773        number = *prev - OP_BRA;        number = GET2(prev, 1+LINK_SIZE);
   
       /* For extended extraction brackets (large number), we have to fish out  
       the number from a dummy opcode at the start. */  
   
       if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);  
1774        offset = number << 1;        offset = number << 1;
1775    
1776  #ifdef DEBUG  #ifdef PCRE_DEBUG
1777        printf("end bracket %d", number);        printf("end bracket %d", number);
1778        printf("\n");        printf("\n");
1779  #endif  #endif
1780    
1781        /* 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. */  
1782    
1783        if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1784          {          {
1785          md->capture_last = number;          md->end_match_ptr = eptr;
1786          if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1787            {          RRETURN(MATCH_MATCH);
1788            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;  
           }  
1789    
1790          /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
         appropriately and continue from after the call. */  
1791    
1792          if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = number;
1793          if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1794            {
1795            /* If offset is greater than offset_top, it means that we are
1796            "skipping" a capturing group, and that group's offsets must be marked
1797            unset. In earlier versions of PCRE, all the offsets were unset at the
1798            start of matching, but this doesn't work because atomic groups and
1799            assertions can cause a value to be set that should later be unset.
1800            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1801            part of the atomic group, but this is not on the final matching path,
1802            so must be unset when 2 is set. (If there is no group 2, there is no
1803            problem, because offset_top will then be 2, indicating no capture.) */
1804    
1805            if (offset > offset_top)
1806            {            {
1807            recursion_info *rec = md->recursive;            register int *iptr = md->offset_vector + offset_top;
1808            DPRINTF(("Recursion (%d) succeeded - continuing\n", number));            register int *iend = md->offset_vector + offset;
1809            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;  
1810            }            }
         }  
       }  
1811    
1812      /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
     the group. */  
1813    
1814      ims = original_ims;          md->offset_vector[offset] =
1815      DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1816            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1817            if (offset_top <= offset) offset_top = offset + 2;
1818            }
1819          }
1820    
1821      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1822      happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1823      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1824      5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group, establish a backup point by
1825      course of events. */      processing the rest of the pattern at a lower level. If this results in a
1826        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1827        bypassing intermediate backup points, but resetting any captures that
1828        happened along the way. */
1829    
1830      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1831        {        {
1832        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1833            {
1834            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1835            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1836            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1837            RRETURN(MATCH_ONCE);
1838            }
1839          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1840        break;        break;
1841        }        }
1842    
1843      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1844      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
1845      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
1846    
1847        if (*ecode == OP_KETRPOS)
1848          {
1849          md->end_match_ptr = eptr;
1850          md->end_offset_top = offset_top;
1851          RRETURN(MATCH_KETRPOS);
1852          }
1853    
1854        /* The normal repeating kets try the rest of the pattern or restart from
1855        the preceding bracket, in the appropriate order. In the second case, we can
1856        use tail recursion to avoid using another stack frame, unless we have an
1857        an atomic group or an unlimited repeat of a group that can match an empty
1858        string. */
1859    
1860      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1861        {        {
1862        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1863        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1864          if (*prev == OP_ONCE)
1865            {
1866            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1867            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1868            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1869            RRETURN(MATCH_ONCE);
1870            }
1871          if (*prev >= OP_SBRA)    /* Could match an empty string */
1872            {
1873            md->match_function_type = MATCH_CBEGROUP;
1874            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1875            RRETURN(rrc);
1876            }
1877        ecode = prev;        ecode = prev;
       flags = match_isgroup;  
1878        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1879        }        }
1880      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1881        {        {
1882        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1883          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1884          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1885        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1886          if (*prev == OP_ONCE)
1887            {
1888            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1889            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1890            md->once_target = prev;
1891            RRETURN(MATCH_ONCE);
1892            }
1893        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1894        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1895        }        }
1896      /* Control never gets here */      /* Control never gets here */
1897    
1898      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1899    
1900      case OP_CIRC:      case OP_CIRC:
1901      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 */  
1902    
1903      /* Start of subject assertion */      /* Start of subject assertion */
1904    
1905      case OP_SOD:      case OP_SOD:
1906      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1907        ecode++;
1908        break;
1909    
1910        /* Multiline mode: start of subject unless notbol, or after any newline. */
1911    
1912        case OP_CIRCM:
1913        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1914        if (eptr != md->start_subject &&
1915            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1916          MRRETURN(MATCH_NOMATCH);
1917      ecode++;      ecode++;
1918      break;      break;
1919    
1920      /* Start of match assertion */      /* Start of match assertion */
1921    
1922      case OP_SOM:      case OP_SOM:
1923      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
1924      ecode++;      ecode++;
1925      break;      break;
1926    
1927      /* 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. */  
1928    
1929      case OP_DOLL:      case OP_SET_SOM:
1930      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
1931        {      ecode++;
1932        if (eptr < md->end_subject)      break;
1933          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
1934        else      /* Multiline mode: assert before any newline, or before end of subject
1935          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
1936        ecode++;  
1937        break;      case OP_DOLLM:
1938        }      if (eptr < md->end_subject)
1939          { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
1940      else      else
1941        {        {
1942        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1943        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
1944        }        }
1945        ecode++;
1946        break;
1947    
1948        /* Not multiline mode: assert before a terminating newline or before end of
1949        subject unless noteol is set. */
1950    
1951        case OP_DOLL:
1952        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1953        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1954    
1955      /* ... else fall through for endonly */      /* ... else fall through for endonly */
1956    
1957      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
1958    
1959      case OP_EOD:      case OP_EOD:
1960      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
1961        SCHECK_PARTIAL();
1962      ecode++;      ecode++;
1963      break;      break;
1964    
1965      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
1966    
1967      case OP_EODN:      case OP_EODN:
1968      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
1969          (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))      if (eptr < md->end_subject &&
1970        RRETURN(MATCH_NOMATCH);          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1971          MRRETURN(MATCH_NOMATCH);
1972    
1973        /* Either at end of string or \n before end. */
1974    
1975        SCHECK_PARTIAL();
1976      ecode++;      ecode++;
1977      break;      break;
1978    
# Line 1276  for (;;) Line 1984  for (;;)
1984    
1985        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
1986        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
1987        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
1988          partial matching. */
1989    
1990  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1991        if (utf8)        if (utf8)
1992          {          {
1993            /* Get status of previous character */
1994    
1995          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
1996            {            {
1997            const uschar *lastptr = eptr - 1;            USPTR lastptr = eptr - 1;
1998            while((*lastptr & 0xc0) == 0x80) lastptr--;            while((*lastptr & 0xc0) == 0x80) lastptr--;
1999              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2000            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2001    #ifdef SUPPORT_UCP
2002              if (md->use_ucp)
2003                {
2004                if (c == '_') prev_is_word = TRUE; else
2005                  {
2006                  int cat = UCD_CATEGORY(c);
2007                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2008                  }
2009                }
2010              else
2011    #endif
2012            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2013            }            }
2014          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2015            /* Get status of next character */
2016    
2017            if (eptr >= md->end_subject)
2018              {
2019              SCHECK_PARTIAL();
2020              cur_is_word = FALSE;
2021              }
2022            else
2023            {            {
2024            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2025    #ifdef SUPPORT_UCP
2026              if (md->use_ucp)
2027                {
2028                if (c == '_') cur_is_word = TRUE; else
2029                  {
2030                  int cat = UCD_CATEGORY(c);
2031                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2032                  }
2033                }
2034              else
2035    #endif
2036            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2037            }            }
2038          }          }
2039        else        else
2040  #endif  #endif
2041    
2042        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2043          consistency with the behaviour of \w we do use it in this case. */
2044    
2045          {          {
2046          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2047            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2048          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2049            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2050              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2051    #ifdef SUPPORT_UCP
2052              if (md->use_ucp)
2053                {
2054                c = eptr[-1];
2055                if (c == '_') prev_is_word = TRUE; else
2056                  {
2057                  int cat = UCD_CATEGORY(c);
2058                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2059                  }
2060                }
2061              else
2062    #endif
2063              prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2064              }
2065    
2066            /* Get status of next character */
2067    
2068            if (eptr >= md->end_subject)
2069              {
2070              SCHECK_PARTIAL();
2071              cur_is_word = FALSE;
2072              }
2073            else
2074    #ifdef SUPPORT_UCP
2075            if (md->use_ucp)
2076              {
2077              c = *eptr;
2078              if (c == '_') cur_is_word = TRUE; else
2079                {
2080                int cat = UCD_CATEGORY(c);
2081                cur_is_word = (cat == ucp_L || cat == ucp_N);
2082                }
2083              }
2084            else
2085    #endif
2086            cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
2087          }          }
2088    
2089        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
2090    
2091        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
2092             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2093          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2094        }        }
2095      break;      break;
2096    
2097      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2098    
2099      case OP_ANY:      case OP_ANY:
2100      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
2101        {      /* Fall through */
2102        if (eptr <= md->end_subject - md->nllen && IS_NEWLINE(eptr))  
2103          RRETURN(MATCH_NOMATCH);      case OP_ALLANY:
2104        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2105          {                            /* not be updated before SCHECK_PARTIAL. */
2106          SCHECK_PARTIAL();
2107          MRRETURN(MATCH_NOMATCH);
2108        }        }
2109      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2110      if (utf8)      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
       while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;  
2111      ecode++;      ecode++;
2112      break;      break;
2113    
# Line 1332  for (;;) Line 2115  for (;;)
2115      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2116    
2117      case OP_ANYBYTE:      case OP_ANYBYTE:
2118      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2119          {                            /* not be updated before SCHECK_PARTIAL. */
2120          SCHECK_PARTIAL();
2121          MRRETURN(MATCH_NOMATCH);
2122          }
2123        eptr++;
2124      ecode++;      ecode++;
2125      break;      break;
2126    
2127      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2128      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2129          {
2130          SCHECK_PARTIAL();
2131          MRRETURN(MATCH_NOMATCH);
2132          }
2133      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2134      if (      if (
2135  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1345  for (;;) Line 2137  for (;;)
2137  #endif  #endif
2138         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2139         )         )
2140        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2141      ecode++;      ecode++;
2142      break;      break;
2143    
2144      case OP_DIGIT:      case OP_DIGIT:
2145      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2146          {
2147          SCHECK_PARTIAL();
2148          MRRETURN(MATCH_NOMATCH);
2149          }
2150      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2151      if (      if (
2152  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1358  for (;;) Line 2154  for (;;)
2154  #endif  #endif
2155         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2156         )         )
2157        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2158      ecode++;      ecode++;
2159      break;      break;
2160    
2161      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2162      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2163          {
2164          SCHECK_PARTIAL();
2165          MRRETURN(MATCH_NOMATCH);
2166          }
2167      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2168      if (      if (
2169  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1371  for (;;) Line 2171  for (;;)
2171  #endif  #endif
2172         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2173         )         )
2174        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2175      ecode++;      ecode++;
2176      break;      break;
2177    
2178      case OP_WHITESPACE:      case OP_WHITESPACE:
2179      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2180          {
2181          SCHECK_PARTIAL();
2182          MRRETURN(MATCH_NOMATCH);
2183          }
2184      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2185      if (      if (
2186  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1384  for (;;) Line 2188  for (;;)
2188  #endif  #endif
2189         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2190         )         )
2191        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2192      ecode++;      ecode++;
2193      break;      break;
2194    
2195      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2196      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2197          {
2198          SCHECK_PARTIAL();
2199          MRRETURN(MATCH_NOMATCH);
2200          }
2201      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2202      if (      if (
2203  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1397  for (;;) Line 2205  for (;;)
2205  #endif  #endif
2206         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2207         )         )
2208        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2209      ecode++;      ecode++;
2210      break;      break;
2211    
2212      case OP_WORDCHAR:      case OP_WORDCHAR:
2213      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2214          {
2215          SCHECK_PARTIAL();
2216          MRRETURN(MATCH_NOMATCH);
2217          }
2218      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2219      if (      if (
2220  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1410  for (;;) Line 2222  for (;;)
2222  #endif  #endif
2223         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2224         )         )
2225        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2226        ecode++;
2227        break;
2228    
2229        case OP_ANYNL:
2230        if (eptr >= md->end_subject)
2231          {
2232          SCHECK_PARTIAL();
2233          MRRETURN(MATCH_NOMATCH);
2234          }
2235        GETCHARINCTEST(c, eptr);
2236        switch(c)
2237          {
2238          default: MRRETURN(MATCH_NOMATCH);
2239    
2240          case 0x000d:
2241          if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2242          break;
2243    
2244          case 0x000a:
2245          break;
2246    
2247          case 0x000b:
2248          case 0x000c:
2249          case 0x0085:
2250          case 0x2028:
2251          case 0x2029:
2252          if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
2253          break;
2254          }
2255        ecode++;
2256        break;
2257    
2258        case OP_NOT_HSPACE:
2259        if (eptr >= md->end_subject)
2260          {
2261          SCHECK_PARTIAL();
2262          MRRETURN(MATCH_NOMATCH);
2263          }
2264        GETCHARINCTEST(c, eptr);
2265        switch(c)
2266          {
2267          default: break;
2268          case 0x09:      /* HT */
2269          case 0x20:      /* SPACE */
2270          case 0xa0:      /* NBSP */
2271          case 0x1680:    /* OGHAM SPACE MARK */
2272          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2273          case 0x2000:    /* EN QUAD */
2274          case 0x2001:    /* EM QUAD */
2275          case 0x2002:    /* EN SPACE */
2276          case 0x2003:    /* EM SPACE */
2277          case 0x2004:    /* THREE-PER-EM SPACE */
2278          case 0x2005:    /* FOUR-PER-EM SPACE */
2279          case 0x2006:    /* SIX-PER-EM SPACE */
2280          case 0x2007:    /* FIGURE SPACE */
2281          case 0x2008:    /* PUNCTUATION SPACE */
2282          case 0x2009:    /* THIN SPACE */
2283          case 0x200A:    /* HAIR SPACE */
2284          case 0x202f:    /* NARROW NO-BREAK SPACE */
2285          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2286          case 0x3000:    /* IDEOGRAPHIC SPACE */
2287          MRRETURN(MATCH_NOMATCH);
2288          }
2289        ecode++;
2290        break;
2291    
2292        case OP_HSPACE:
2293        if (eptr >= md->end_subject)
2294          {
2295          SCHECK_PARTIAL();
2296          MRRETURN(MATCH_NOMATCH);
2297          }
2298        GETCHARINCTEST(c, eptr);
2299        switch(c)
2300          {
2301          default: MRRETURN(MATCH_NOMATCH);
2302          case 0x09:      /* HT */
2303          case 0x20:      /* SPACE */
2304          case 0xa0:      /* NBSP */
2305          case 0x1680:    /* OGHAM SPACE MARK */
2306          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2307          case 0x2000:    /* EN QUAD */
2308          case 0x2001:    /* EM QUAD */
2309          case 0x2002:    /* EN SPACE */
2310          case 0x2003:    /* EM SPACE */
2311          case 0x2004:    /* THREE-PER-EM SPACE */
2312          case 0x2005:    /* FOUR-PER-EM SPACE */
2313          case 0x2006:    /* SIX-PER-EM SPACE */
2314          case 0x2007:    /* FIGURE SPACE */
2315          case 0x2008:    /* PUNCTUATION SPACE */
2316          case 0x2009:    /* THIN SPACE */
2317          case 0x200A:    /* HAIR SPACE */
2318          case 0x202f:    /* NARROW NO-BREAK SPACE */
2319          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2320          case 0x3000:    /* IDEOGRAPHIC SPACE */
2321          break;
2322          }
2323        ecode++;
2324        break;
2325    
2326        case OP_NOT_VSPACE:
2327        if (eptr >= md->end_subject)
2328          {
2329          SCHECK_PARTIAL();
2330          MRRETURN(MATCH_NOMATCH);
2331          }
2332        GETCHARINCTEST(c, eptr);
2333        switch(c)
2334          {
2335          default: break;
2336          case 0x0a:      /* LF */
2337          case 0x0b:      /* VT */
2338          case 0x0c:      /* FF */
2339          case 0x0d:      /* CR */
2340          case 0x85:      /* NEL */
2341          case 0x2028:    /* LINE SEPARATOR */
2342          case 0x2029:    /* PARAGRAPH SEPARATOR */
2343          MRRETURN(MATCH_NOMATCH);
2344          }
2345        ecode++;
2346        break;
2347    
2348        case OP_VSPACE:
2349        if (eptr >= md->end_subject)
2350          {
2351          SCHECK_PARTIAL();
2352          MRRETURN(MATCH_NOMATCH);
2353          }
2354        GETCHARINCTEST(c, eptr);
2355        switch(c)
2356          {
2357          default: MRRETURN(MATCH_NOMATCH);
2358          case 0x0a:      /* LF */
2359          case 0x0b:      /* VT */
2360          case 0x0c:      /* FF */
2361          case 0x0d:      /* CR */
2362          case 0x85:      /* NEL */
2363          case 0x2028:    /* LINE SEPARATOR */
2364          case 0x2029:    /* PARAGRAPH SEPARATOR */
2365          break;
2366          }
2367      ecode++;      ecode++;
2368      break;      break;
2369    
# Line 1420  for (;;) Line 2373  for (;;)
2373    
2374      case OP_PROP:      case OP_PROP:
2375      case OP_NOTPROP:      case OP_NOTPROP:
2376      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2377          {
2378          SCHECK_PARTIAL();
2379          MRRETURN(MATCH_NOMATCH);
2380          }
2381      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2382        {        {
2383        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2384    
2385        switch(ecode[1])        switch(ecode[1])
2386          {          {
2387          case PT_ANY:          case PT_ANY:
2388          if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
2389          break;          break;
2390    
2391          case PT_LAMP:          case PT_LAMP:
2392          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2393               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2394               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2395            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2396           break;          break;
2397    
2398          case PT_GC:          case PT_GC:
2399          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
2400            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2401          break;          break;
2402    
2403          case PT_PC:          case PT_PC:
2404          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2405            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2406          break;          break;
2407    
2408          case PT_SC:          case PT_SC:
2409          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2410            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2411          break;          break;
2412    
2413          default:          /* These are specials */
2414          RRETURN(PCRE_ERROR_INTERNAL);  
2415            case PT_ALNUM:
2416            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2417                 _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2418              MRRETURN(MATCH_NOMATCH);
2419            break;
2420    
2421            case PT_SPACE:    /* Perl space */
2422            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2423                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2424                   == (op == OP_NOTPROP))
2425              MRRETURN(MATCH_NOMATCH);
2426            break;
2427    
2428            case PT_PXSPACE:  /* POSIX space */
2429            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2430                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2431                 c == CHAR_FF || c == CHAR_CR)
2432                   == (op == OP_NOTPROP))
2433              MRRETURN(MATCH_NOMATCH);
2434            break;
2435    
2436            case PT_WORD:
2437            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2438                 _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2439                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2440              MRRETURN(MATCH_NOMATCH);
2441          break;          break;
2442    
2443            /* This should never occur */
2444    
2445            default:
2446            RRETURN(PCRE_ERROR_INTERNAL);
2447          }          }
2448    
2449        ecode += 3;        ecode += 3;
# Line 1467  for (;;) Line 2454  for (;;)
2454      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2455    
2456      case OP_EXTUNI:      case OP_EXTUNI:
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        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2464        while (eptr < md->end_subject)
2465        {        {
2466        int chartype, script;        int len = 1;
2467        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2468        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2469        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;  
         }  
2470        }        }
2471      ecode++;      ecode++;
2472      break;      break;
# Line 1499  for (;;) Line 2482  for (;;)
2482      loops). */      loops). */
2483    
2484      case OP_REF:      case OP_REF:
2485        {      case OP_REFI:
2486        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2487        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2488        ecode += 3;
2489    
2490        /* 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];  
2491    
2492        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2493        this ensures that every attempt at a match fails. We can't just fail
2494        here, because of the possibility of quantifiers with zero minima.
2495    
2496        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2497          {      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;  
2498    
2499          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2500          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;  
2501    
2502          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2503          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2504          eptr += length;      else
2505          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2506    
2507        /* Set up for repetition, or handle the non-repeated case */
2508    
2509        switch (*ecode)
2510          {
2511          case OP_CRSTAR:
2512          case OP_CRMINSTAR:
2513          case OP_CRPLUS:
2514          case OP_CRMINPLUS:
2515          case OP_CRQUERY:
2516          case OP_CRMINQUERY:
2517          c = *ecode++ - OP_CRSTAR;
2518          minimize = (c & 1) != 0;
2519          min = rep_min[c];                 /* Pick up values from tables; */
2520          max = rep_max[c];                 /* zero for max => infinity */
2521          if (max == 0) max = INT_MAX;
2522          break;
2523    
2524          case OP_CRRANGE:
2525          case OP_CRMINRANGE:
2526          minimize = (*ecode == OP_CRMINRANGE);
2527          min = GET2(ecode, 1);
2528          max = GET2(ecode, 3);
2529          if (max == 0) max = INT_MAX;
2530          ecode += 5;
2531          break;
2532    
2533          default:               /* No repeat follows */
2534          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2535            {
2536            CHECK_PARTIAL();
2537            MRRETURN(MATCH_NOMATCH);
2538          }          }
2539          eptr += length;
2540          continue;              /* With the main loop */
2541          }
2542    
2543        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2544        main loop. */      zero, just continue with the main loop. */
2545    
2546        if (length == 0) continue;      if (length == 0) continue;
2547    
2548        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2549        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2550        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2551    
2552        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2553          {
2554          int slength;
2555          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2556          {          {
2557          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2558          eptr += length;          MRRETURN(MATCH_NOMATCH);
2559          }          }
2560          eptr += slength;
2561          }
2562    
2563        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2564        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2565    
2566        if (min == max) continue;      if (min == max) continue;
2567    
2568        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2569    
2570        if (minimize)      if (minimize)
2571          {
2572          for (fi = min;; fi++)
2573          {          {
2574          for (fi = min;; fi++)          int slength;
2575            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2576            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2577            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2578            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2579            {            {
2580            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2581            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2582            }            }
2583          /* Control never gets here */          eptr += slength;
2584          }          }
2585          /* Control never gets here */
2586          }
2587    
2588        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2589    
2590        else      else
2591          {
2592          pp = eptr;
2593          for (i = min; i < max; i++)
2594          {          {
2595          pp = eptr;          int slength;
2596          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2597            {            {
2598            if (!match_ref(offset, eptr, length, md, ims)) break;            CHECK_PARTIAL();
2599            eptr += length;            break;
           }  
         while (eptr >= pp)  
           {  
           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
           if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
           eptr -= length;  
2600            }            }
2601          RRETURN(MATCH_NOMATCH);          eptr += slength;
2602            }
2603          while (eptr >= pp)
2604            {
2605            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2606            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2607            eptr -= length;
2608          }          }
2609          MRRETURN(MATCH_NOMATCH);
2610        }        }
2611      /* Control never gets here */      /* Control never gets here */
2612    
   
   
2613      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2614      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,
2615      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 2664  for (;;)
2664          {          {
2665          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2666            {            {
2667            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2668                {
2669                SCHECK_PARTIAL();
2670                MRRETURN(MATCH_NOMATCH);
2671                }
2672            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2673            if (c > 255)            if (c > 255)
2674              {              {
2675              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2676              }              }
2677            else            else
2678              {              {
2679              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2680              }              }
2681            }            }
2682          }          }
# Line 1674  for (;;) Line 2686  for (;;)
2686          {          {
2687          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2688            {            {
2689            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2690                {
2691                SCHECK_PARTIAL();
2692                MRRETURN(MATCH_NOMATCH);
2693                }
2694            c = *eptr++;            c = *eptr++;
2695            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2696            }            }
2697          }          }
2698    
# Line 1696  for (;;) Line 2712  for (;;)
2712            {            {
2713            for (fi = min;; fi++)            for (fi = min;; fi++)
2714              {              {
2715              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2716              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2717              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2718                if (eptr >= md->end_subject)
2719                  {
2720                  SCHECK_PARTIAL();
2721                  MRRETURN(MATCH_NOMATCH);
2722                  }
2723              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2724              if (c > 255)              if (c > 255)
2725                {                {
2726                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2727                }                }
2728              else              else
2729                {                {
2730                if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);                if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2731                }                }
2732              }              }
2733            }            }
# Line 1716  for (;;) Line 2737  for (;;)
2737            {            {
2738            for (fi = min;; fi++)            for (fi = min;; fi++)
2739              {              {
2740              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2741              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2742              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2743                if (eptr >= md->end_subject)
2744                  {
2745                  SCHECK_PARTIAL();
2746                  MRRETURN(MATCH_NOMATCH);
2747                  }
2748              c = *eptr++;              c = *eptr++;
2749              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2750              }              }
2751            }            }
2752          /* Control never gets here */          /* Control never gets here */
# Line 1739  for (;;) Line 2765  for (;;)
2765            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2766              {              {
2767              int len = 1;              int len = 1;
2768              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2769                  {
2770                  SCHECK_PARTIAL();
2771                  break;
2772                  }
2773              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2774              if (c > 255)              if (c > 255)
2775                {                {
# Line 1753  for (;;) Line 2783  for (;;)
2783              }              }
2784            for (;;)            for (;;)
2785              {              {
2786              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2787              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2788              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2789              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1765  for (;;) Line 2795  for (;;)
2795            {            {
2796            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2797              {              {
2798              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2799                  {
2800                  SCHECK_PARTIAL();
2801                  break;
2802                  }
2803              c = *eptr;              c = *eptr;
2804              if ((data[c/8] & (1 << (c&7))) == 0) break;              if ((data[c/8] & (1 << (c&7))) == 0) break;
2805              eptr++;              eptr++;
2806              }              }
2807            while (eptr >= pp)            while (eptr >= pp)
2808              {              {
2809              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2810              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2811              eptr--;              eptr--;
2812              }              }
2813            }            }
2814    
2815          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2816          }          }
2817        }        }
2818      /* Control never gets here */      /* Control never gets here */
2819    
2820    
2821      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2822      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
2823        mode, because Unicode properties are supported in non-UTF-8 mode. */
2824    
2825  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2826      case OP_XCLASS:      case OP_XCLASS:
# Line 1826  for (;;) Line 2861  for (;;)
2861    
2862        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2863          {          {
2864          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2865          GETCHARINC(c, eptr);            {
2866          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2867              MRRETURN(MATCH_NOMATCH);
2868              }
2869            GETCHARINCTEST(c, eptr);
2870            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2871          }          }
2872    
2873        /* 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 2882  for (;;)
2882          {          {
2883          for (fi = min;; fi++)          for (fi = min;; fi++)
2884            {            {
2885            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2886            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2887            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2888            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
2889            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
2890                SCHECK_PARTIAL();
2891                MRRETURN(MATCH_NOMATCH);
2892                }
2893              GETCHARINCTEST(c, eptr);
2894              if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2895            }            }
2896          /* Control never gets here */          /* Control never gets here */
2897          }          }
# Line 1860  for (;;) Line 2904  for (;;)
2904          for (i = min; i < max; i++)          for (i = min; i < max; i++)
2905            {            {
2906            int len = 1;            int len = 1;
2907            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
2908            GETCHARLEN(c, eptr, len);              {
2909                SCHECK_PARTIAL();
2910                break;
2911                }
2912              GETCHARLENTEST(c, eptr, len);
2913            if (!_pcre_xclass(c, data)) break;            if (!_pcre_xclass(c, data)) break;
2914            eptr += len;            eptr += len;
2915            }            }
2916          for(;;)          for(;;)
2917            {            {
2918            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2919            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2920            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2921            BACKCHAR(eptr)            if (utf8) BACKCHAR(eptr);
2922            }            }
2923          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2924          }          }
2925    
2926        /* Control never gets here */        /* Control never gets here */
# Line 1888  for (;;) Line 2936  for (;;)
2936        length = 1;        length = 1;
2937        ecode++;        ecode++;
2938        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2939        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2940        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
2941            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2942            MRRETURN(MATCH_NOMATCH);
2943            }
2944          while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
2945        }        }
2946      else      else
2947  #endif  #endif
2948    
2949      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2950        {        {
2951        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2952        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);          {
2953            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2954            MRRETURN(MATCH_NOMATCH);
2955            }
2956          if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
2957        ecode += 2;        ecode += 2;
2958        }        }
2959      break;      break;
2960    
2961      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2962    
2963      case OP_CHARNC:      case OP_CHARI:
2964  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2965      if (utf8)      if (utf8)
2966        {        {
# Line 1912  for (;;) Line 2968  for (;;)
2968        ecode++;        ecode++;
2969        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2970    
2971        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2972            {
2973            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2974            MRRETURN(MATCH_NOMATCH);
2975            }
2976    
2977        /* 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
2978        can use the fast lookup table. */        can use the fast lookup table. */
2979    
2980        if (fc < 128)        if (fc < 128)
2981          {          {
2982          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2983          }          }
2984    
2985        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character */
2986    
2987        else        else
2988          {          {
2989          int dc;          unsigned int dc;
2990          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
2991          ecode += length;          ecode += length;
2992    
# Line 1936  for (;;) Line 2996  for (;;)
2996          if (fc != dc)          if (fc != dc)
2997            {            {
2998  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2999            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3000  #endif  #endif
3001              RRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3002            }            }
3003          }          }
3004        }        }
# Line 1947  for (;;) Line 3007  for (;;)
3007    
3008      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
3009        {        {
3010        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3011        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3012            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3013            MRRETURN(MATCH_NOMATCH);
3014            }
3015          if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3016        ecode += 2;        ecode += 2;
3017        }        }
3018      break;      break;
3019    
3020      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3021    
3022      case OP_EXACT:      case OP_EXACT:
3023        case OP_EXACTI:
3024      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3025      ecode += 3;      ecode += 3;
3026      goto REPEATCHAR;      goto REPEATCHAR;
3027    
3028        case OP_POSUPTO:
3029        case OP_POSUPTOI:
3030        possessive = TRUE;
3031        /* Fall through */
3032    
3033      case OP_UPTO:      case OP_UPTO:
3034        case OP_UPTOI:
3035      case OP_MINUPTO:      case OP_MINUPTO:
3036        case OP_MINUPTOI:
3037      min = 0;      min = 0;
3038      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3039      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3040      ecode += 3;      ecode += 3;
3041      goto REPEATCHAR;      goto REPEATCHAR;
3042    
3043        case OP_POSSTAR:
3044        case OP_POSSTARI:
3045        possessive = TRUE;
3046        min = 0;
3047        max = INT_MAX;
3048        ecode++;
3049        goto REPEATCHAR;
3050    
3051        case OP_POSPLUS:
3052        case OP_POSPLUSI:
3053        possessive = TRUE;
3054        min = 1;
3055        max = INT_MAX;
3056        ecode++;
3057        goto REPEATCHAR;
3058    
3059        case OP_POSQUERY:
3060        case OP_POSQUERYI:
3061        possessive = TRUE;
3062        min = 0;
3063        max = 1;
3064        ecode++;
3065        goto REPEATCHAR;
3066    
3067      case OP_STAR:      case OP_STAR:
3068        case OP_STARI:
3069      case OP_MINSTAR:      case OP_MINSTAR:
3070        case OP_MINSTARI:
3071      case OP_PLUS:      case OP_PLUS:
3072        case OP_PLUSI:
3073      case OP_MINPLUS:      case OP_MINPLUS:
3074        case OP_MINPLUSI:
3075      case OP_QUERY:      case OP_QUERY:
3076        case OP_QUERYI:
3077      case OP_MINQUERY:      case OP_MINQUERY:
3078      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3079        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3080      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3081      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3082      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3083      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3084    
3085      /* 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. */  
3086    
3087      REPEATCHAR:      REPEATCHAR:
3088  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1991  for (;;) Line 3091  for (;;)
3091        length = 1;        length = 1;
3092        charptr = ecode;        charptr = ecode;
3093        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3094        ecode += length;        ecode += length;
3095    
3096        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1999  for (;;) Line 3098  for (;;)
3098    
3099        if (length > 1)        if (length > 1)
3100          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3101  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3102          int othercase;          unsigned int othercase;
3103          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3104              (othercase = _pcre_ucp_othercase(fc)) >= 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
              othercase >= 0)  
3105            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3106            else oclength = 0;
3107  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3108    
3109          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3110            {            {
3111            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3112            /* Need braces because of following else */              memcmp(eptr, charptr, length) == 0) eptr += length;
3113            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3114              else if (oclength > 0 &&
3115                       eptr <= md->end_subject - oclength &&
3116                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3117    #endif  /* SUPPORT_UCP */
3118            else            else
3119              {              {
3120              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3121              eptr += oclength;              MRRETURN(MATCH_NOMATCH);
3122              }              }
3123            }            }
3124    
# Line 2028  for (;;) Line 3128  for (;;)
3128            {            {
3129            for (fi = min;; fi++)            for (fi = min;; fi++)
3130              {              {
3131              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3132              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3133              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3134              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3135              /* Need braces because of following else */                memcmp(eptr, charptr, length) == 0) eptr += length;
3136              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3137                else if (oclength > 0 &&
3138                         eptr <= md->end_subject - oclength &&
3139                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3140    #endif  /* SUPPORT_UCP */
3141              else              else
3142                {                {
3143                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3144                eptr += oclength;                MRRETURN(MATCH_NOMATCH);
3145                }                }
3146              }              }
3147            /* Control never gets here */            /* Control never gets here */
3148            }            }
3149          else  
3150            else  /* Maximize */
3151            {            {
3152            pp = eptr;            pp = eptr;
3153            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3154              {              {
3155              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3156              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, length) == 0) eptr += length;
3157              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3158                else if (oclength > 0 &&
3159                         eptr <= md->end_subject - oclength &&
3160                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3161    #endif  /* SUPPORT_UCP */
3162              else              else
3163                {                {
3164                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3165                eptr += oclength;                break;
3166                }                }
3167              }              }
3168            while (eptr >= pp)  
3169             {            if (possessive) continue;
3170             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3171             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3172             eptr -= length;              {
3173             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3174            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3175                if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3176    #ifdef SUPPORT_UCP
3177                eptr--;
3178                BACKCHAR(eptr);
3179    #else   /* without SUPPORT_UCP */
3180                eptr -= length;
3181    #endif  /* SUPPORT_UCP */
3182                }
3183            }            }
3184          /* Control never gets here */          /* Control never gets here */
3185          }          }
# Line 2075  for (;;) Line 3192  for (;;)
3192  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3193    
3194      /* When not in UTF-8 mode, load a single-byte character. */      /* When not in UTF-8 mode, load a single-byte character. */
3195        {  
3196        if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);      fc = *ecode++;
       fc = *ecode++;  
       }  
3197    
3198      /* 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
3199      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 3207  for (;;)
3207      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3208        max, eptr));        max, eptr));
3209    
3210      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3211        {        {
3212        fc = md->lcc[fc];        fc = md->lcc[fc];
3213        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3214          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3215            if (eptr >= md->end_subject)
3216              {
3217              SCHECK_PARTIAL();
3218              MRRETURN(MATCH_NOMATCH);
3219              }
3220            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3221            }
3222        if (min == max) continue;        if (min == max) continue;
3223        if (minimize)        if (minimize)
3224          {          {
3225          for (fi = min;; fi++)          for (fi = min;; fi++)
3226            {            {
3227            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3228            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3229            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3230                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3231              RRETURN(MATCH_NOMATCH);              {
3232                SCHECK_PARTIAL();
3233                MRRETURN(MATCH_NOMATCH);
3234                }
3235              if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3236            }            }
3237          /* Control never gets here */          /* Control never gets here */
3238          }          }
3239        else        else  /* Maximize */
3240          {          {
3241          pp = eptr;          pp = eptr;
3242          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3243            {            {
3244            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3245                {
3246                SCHECK_PARTIAL();
3247                break;
3248                }
3249              if (fc != md->lcc[*eptr]) break;
3250            eptr++;            eptr++;
3251            }            }
3252    
3253            if (possessive) continue;
3254    
3255          while (eptr >= pp)          while (eptr >= pp)
3256            {            {
3257            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3258            eptr--;            eptr--;
3259            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3260            }            }
3261          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3262          }          }
3263        /* Control never gets here */        /* Control never gets here */
3264        }        }
# Line 2133  for (;;) Line 3267  for (;;)
3267    
3268      else      else
3269        {        {
3270        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3271            {
3272            if (eptr >= md->end_subject)
3273              {
3274              SCHECK_PARTIAL();
3275              MRRETURN(MATCH_NOMATCH);
3276              }
3277            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3278            }
3279    
3280        if (min == max) continue;        if (min == max) continue;
3281    
3282        if (minimize)        if (minimize)
3283          {          {
3284          for (fi = min;; fi++)          for (fi = min;; fi++)
3285            {            {
3286            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3287            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3288            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3289              RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
3290                {
3291                SCHECK_PARTIAL();
3292                MRRETURN(MATCH_NOMATCH);
3293                }
3294              if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3295            }            }
3296          /* Control never gets here */          /* Control never gets here */
3297          }          }
3298        else        else  /* Maximize */
3299          {          {
3300          pp = eptr;          pp = eptr;
3301          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3302            {            {
3303            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3304                {
3305                SCHECK_PARTIAL();
3306                break;
3307                }
3308              if (fc != *eptr) break;
3309            eptr++;            eptr++;
3310            }            }
3311            if (possessive) continue;
3312    
3313          while (eptr >= pp)          while (eptr >= pp)
3314            {            {
3315            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3316            eptr--;            eptr--;
3317            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3318            }            }
3319          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3320          }          }
3321        }        }
3322      /* Control never gets here */      /* Control never gets here */
# Line 2169  for (;;) Line 3325  for (;;)
3325      checking can be multibyte. */      checking can be multibyte. */
3326    
3327      case OP_NOT:      case OP_NOT:
3328      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3329        if (eptr >= md->end_subject)
3330          {
3331          SCHECK_PARTIAL();
3332          MRRETURN(MATCH_NOMATCH);
3333          }
3334      ecode++;      ecode++;
3335      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3336      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3337        {        {
3338  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3339        if (c < 256)        if (c < 256)
3340  #endif  #endif
3341        c = md->lcc[c];        c = md->lcc[c];
3342        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3343        }        }
3344      else      else    /* Caseful */
3345        {        {
3346        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3347        }        }
3348      break;      break;
3349    
# Line 2194  for (;;) Line 3355  for (;;)
3355      about... */      about... */
3356    
3357      case OP_NOTEXACT:      case OP_NOTEXACT:
3358        case OP_NOTEXACTI:
3359      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3360      ecode += 3;      ecode += 3;
3361      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3362    
3363      case OP_NOTUPTO:      case OP_NOTUPTO:
3364        case OP_NOTUPTOI:
3365      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3366        case OP_NOTMINUPTOI:
3367        min = 0;
3368        max = GET2(ecode, 1);
3369        minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3370        ecode += 3;
3371        goto REPEATNOTCHAR;
3372    
3373        case OP_NOTPOSSTAR:
3374        case OP_NOTPOSSTARI:
3375        possessive = TRUE;
3376        min = 0;
3377        max = INT_MAX;
3378        ecode++;
3379        goto REPEATNOTCHAR;
3380    
3381        case OP_NOTPOSPLUS:
3382        case OP_NOTPOSPLUSI:
3383        possessive = TRUE;
3384        min = 1;
3385        max = INT_MAX;
3386        ecode++;
3387        goto REPEATNOTCHAR;
3388    
3389        case OP_NOTPOSQUERY:
3390        case OP_NOTPOSQUERYI:
3391        possessive = TRUE;
3392        min = 0;
3393        max = 1;
3394        ecode++;
3395        goto REPEATNOTCHAR;
3396    
3397        case OP_NOTPOSUPTO:
3398        case OP_NOTPOSUPTOI:
3399        possessive = TRUE;
3400      min = 0;      min = 0;
3401      max = GET2(ecode, 1);      max = GET2(ecode, 1);
     minimize = *ecode == OP_NOTMINUPTO;  
3402      ecode += 3;      ecode += 3;
3403      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3404    
3405      case OP_NOTSTAR:      case OP_NOTSTAR:
3406        case OP_NOTSTARI:
3407      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3408        case OP_NOTMINSTARI:
3409      case OP_NOTPLUS:      case OP_NOTPLUS:
3410        case OP_NOTPLUSI:
3411      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3412        case OP_NOTMINPLUSI:
3413      case OP_NOTQUERY:      case OP_NOTQUERY:
3414        case OP_NOTQUERYI:
3415      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3416      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3417        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3418      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3419      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3420      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3421      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3422    
3423      /* 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. */  
3424