/[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 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 630 by ph10, Fri Jul 22 10:00:10 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) 1997-2007 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    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK md             /* Block containing newline information */  #define NLBLOCK md             /* Block containing newline information */
50  #define PSSTART start_subject  /* Field containing processed string start */  #define PSSTART start_subject  /* Field containing processed string start */
51  #define PSEND   end_subject    /* Field containing processed string end */  #define PSEND   end_subject    /* Field containing processed string end */
# Line 53  possible. There are also some static sup Line 57  possible. There are also some static sup
57  #undef min  #undef min
58  #undef max  #undef max
59    
60  /* The chain of eptrblocks for tail recursions uses memory in stack workspace,  /* Values for setting in md->match_function_type to indicate two special types
61  obtained at top level, the size of which is defined by EPTR_WORK_SIZE. */  of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
 #define EPTR_WORK_SIZE (1000)  
   
 /* Flag bits for the match() function */  
63    
64  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
 #define match_tail_recursed  0x04  /* Tail recursive call */  
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 70  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 83  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 116  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 148  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 188  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 210  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 271  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    
# Line 295  typedef struct heapframe { Line 380  typedef struct heapframe {
380    BOOL Xcondition;    BOOL Xcondition;
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;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
387    int Xoclength;    int Xoclength;
388    uschar Xocchars[8];    uschar Xocchars[8];
389  #endif  #endif
390    
391      int Xcodelink;
392    int Xctype;    int Xctype;
393    unsigned int Xfc;    unsigned int Xfc;
394    int Xfi;    int Xfi;
# Line 323  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 344  typedef struct heapframe { Line 424  typedef struct heapframe {
424    
425  /* This function is called recursively in many circumstances. Whenever it  /* 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  Performance note: It might be tempting to extract commonly used fields from the  
452  md structure (e.g. utf8, end_subject) into individual variables to improve  /* Performance note: It might be tempting to extract commonly used fields from
453    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 to current character in subject     eptr        pointer to current character in subject
459     ecode       pointer to current position in compiled 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_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
                  match_tail_recursed - this is a tail_recursed group  
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 some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 387  register unsigned int c;   /* Character Line 488  register unsigned int c;   /* Character
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 */  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
# Line 394  heap storage. Set up the top-level frame Line 497  heap storage. Set up the top-level frame
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 415  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 427  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 439  HEAP_RECURSE: Line 544  HEAP_RECURSE:
544  #define condition          frame->Xcondition  #define condition          frame->Xcondition
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
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
551  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
552  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
553  #endif  #endif
# Line 477  i, and fc and c, can be the same variabl Line 577  i, and fc and c, can be the same variabl
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;
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;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
608  int oclength;  int oclength;
609  uschar occhars[8];  uschar occhars[8];
610  #endif  #endif
611    
612    int codelink;
613  int ctype;  int ctype;
614  int length;  int length;
615  int max;  int max;
# Line 521  int stacksave[REC_STACK_SAVE_MAX]; Line 624  int stacksave[REC_STACK_SAVE_MAX];
624  eptrblock newptrb;  eptrblock newptrb;
625  #endif     /* NO_RECURSE */  #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. */
640    
# Line 540  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    /* First check that we haven't called match() too many times, or that we
669    haven't exceeded the recursive call limit. */
670    
671    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  /* At the start of a group with an unlimited repeat that may match an empty
675  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
676  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
677  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
678  When match() is called in other circumstances, don't add to the chain. If this  
679  is a tail recursion, use a block from the workspace, as the one on the stack is  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
680  already used. */  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_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
687    {    {
688    eptrblock *p;    newptrb.epb_saved_eptr = eptr;
689    if ((flags & match_tail_recursed) != 0)    newptrb.epb_prev = eptrb;
690      {    eptrb = &newptrb;
691      if (md->eptrn >= EPTR_WORK_SIZE) RRETURN(PCRE_ERROR_NULLWSLIMIT);    md->match_function_type = 0;
     p = md->eptrchain + md->eptrn++;  
     }  
   else p = &newptrb;  
   p->epb_saved_eptr = eptr;  
   p->epb_prev = eptrb;  
   eptrb = p;  
692    }    }
693    
694  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 588  for (;;) Line 698  for (;;)
698    minimize = possessive = FALSE;    minimize = possessive = FALSE;
699    op = *ecode;    op = *ecode;
700    
   /* For partial matching, remember if we ever hit the end of the subject after  
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > md->start_match)  
     md->hitend = TRUE;  
   
701    switch(op)    switch(op)
702      {      {
703      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
704      the current subject position in the working slot at the top of the vector.      markptr = ecode + 2;
705      We mustn't change the current values of the data slot, because they may be      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
706      set from a previous iteration of this group, and be referred to by a        eptrb, RM55);
707      reference inside the group.  
708        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
709      If the bracket fails to match, we need to restore this value and also the      argument, and we must check whether that argument matches this MARK's
710      values of the final offsets, in case they were set by a previous iteration      argument. It is passed back in md->start_match_ptr (an overloading of that
711      of the same bracket.      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 bracket or
779        the alt that is at the start of the current branch. This makes it possible
780        to skip back past alternatives that precede the THEN within the current
781        branch. */
782    
783        case OP_THEN:
784        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
785          eptrb, RM54);
786        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
787        md->start_match_ptr = ecode - GET(ecode, 1);
788        MRRETURN(MATCH_THEN);
789    
790        case OP_THEN_ARG:
791        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],
792          offset_top, md, eptrb, RM58);
793        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794        md->start_match_ptr = ecode - GET(ecode, 1);
795        md->mark = ecode + LINK_SIZE + 2;
796        RRETURN(MATCH_THEN);
797    
798        /* Handle a capturing bracket, other than those that are possessive with an
799        unlimited repeat. If there is space in the offset vector, save the current
800        subject position in the working slot at the top of the vector. We mustn't
801        change the current values of the data slot, because they may be set from a
802        previous iteration of this group, and be referred to by a reference inside
803        the group. A failure to match might occur after the group has succeeded,
804        if something later on doesn't match. For this reason, we need to restore
805        the working value and also the values of the final offsets, in case they
806        were set by a previous iteration of the same bracket.
807    
808      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
809      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 617  for (;;) Line 814  for (;;)
814      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
815      offset = number << 1;      offset = number << 1;
816    
817  #ifdef DEBUG  #ifdef PCRE_DEBUG
818      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
819      printf("subject=");      printf("subject=");
820      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 632  for (;;) Line 829  for (;;)
829        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
830    
831        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
832        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
833            (int)(eptr - md->start_subject);
834    
835        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
836          {          {
837          RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
838            ims, eptrb, flags);          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
839          if (rrc != MATCH_NOMATCH) RRETURN(rrc);            eptrb, RM1);
840            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
841            if (rrc != MATCH_NOMATCH &&
842                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
843              RRETURN(rrc);
844          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
845          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
846            if (*ecode != OP_ALT) break;
847          }          }
       while (*ecode == OP_ALT);  
848    
849        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
850        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
851        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
852        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
853    
854        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE, MATCH_NOMATCH, or
855          MATCH_THEN. */
856    
857          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
858          RRETURN(((rrc == MATCH_ONCE)? MATCH_ONCE:MATCH_NOMATCH));
859        }        }
860    
861      /* Insufficient room for saving captured contents. Treat as a non-capturing      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
862      bracket. */      as a non-capturing bracket. */
863    
864        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
865        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
866    
867      DPRINTF(("insufficient capture room: treat as non-capturing\n"));      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
868    
869      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
870      final alternative within the brackets, we would return the result of a      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion. */  
871    
872        /* Non-capturing or atomic group, except for possessive with unlimited
873        repeat. Loop for all the alternatives. When we get to the final alternative
874        within the brackets, we used to return the result of a recursive call to
875        match() whatever happened so it was possible to reduce stack usage by
876        turning this into a tail recursion, except in the case of a possibly empty
877        group. However, now that there is the possiblity of (*THEN) occurring in
878        the final alternative, this optimization is no longer possible.
879    
880        MATCH_ONCE is returned when the end of an atomic group is successfully
881        reached, but subsequent matching fails. It passes back up the tree (causing
882        captured values to be reset) until the original atomic group level is
883        reached. This is tested by comparing md->once_target with the start of the
884        group. At this point, the return is converted into MATCH_NOMATCH so that
885        previous backup points can be taken. */
886    
887        case OP_ONCE:
888      case OP_BRA:      case OP_BRA:
889      case OP_SBRA:      case OP_SBRA:
890      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
891      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
892      for (;;)      for (;;)
893        {        {
894        if (ecode[GET(ecode, 1)] != OP_ALT)        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
895          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
896            RM2);
897          if (rrc != MATCH_NOMATCH &&
898              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
899          {          {
900          ecode += _pcre_OP_lengths[*ecode];          if (rrc == MATCH_ONCE)
901          flags |= match_tail_recursed;            {
902          DPRINTF(("bracket 0 tail recursion\n"));            const uschar *scode = ecode;
903          goto TAIL_RECURSE;            if (*scode != OP_ONCE)           /* If not at start, find it */
904                {
905                while (*scode == OP_ALT) scode += GET(scode, 1);
906                scode -= GET(scode, 1);
907                }
908              if (md->once_target == scode) rrc = MATCH_NOMATCH;
909              }
910            RRETURN(rrc);
911          }          }
912          ecode += GET(ecode, 1);
913          if (*ecode != OP_ALT) break;
914          }
915        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
916        RRETURN(MATCH_NOMATCH);
917    
918        /* For non-final alternatives, continue the loop for a NOMATCH result;      /* Handle possessive capturing brackets with an unlimited repeat. We come
919        otherwise return. */      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
920        handled similarly to the normal case above. However, the matching is
921        different. The end of these brackets will always be OP_KETRPOS, which
922        returns MATCH_KETRPOS without going further in the pattern. By this means
923        we can handle the group by iteration rather than recursion, thereby
924        reducing the amount of stack needed. */
925    
926        case OP_CBRAPOS:
927        case OP_SCBRAPOS:
928        allow_zero = FALSE;
929    
930        RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      POSSESSIVE_CAPTURE:
931          eptrb, flags);      number = GET2(ecode, 1+LINK_SIZE);
932        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      offset = number << 1;
933    
934    #ifdef PCRE_DEBUG
935        printf("start possessive bracket %d\n", number);
936        printf("subject=");
937        pchars(eptr, 16, TRUE, md);
938        printf("\n");
939    #endif
940    
941        if (offset < md->offset_max)
942          {
943          matched_once = FALSE;
944          code_offset = ecode - md->start_code;
945    
946          save_offset1 = md->offset_vector[offset];
947          save_offset2 = md->offset_vector[offset+1];
948          save_offset3 = md->offset_vector[md->offset_end - number];
949          save_capture_last = md->capture_last;
950    
951          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
952    
953          /* Each time round the loop, save the current subject position for use
954          when the group matches. For MATCH_MATCH, the group has matched, so we
955          restart it with a new subject starting position, remembering that we had
956          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
957          usual. If we haven't matched any alternatives in any iteration, check to
958          see if a previous iteration matched. If so, the group has matched;
959          continue from afterwards. Otherwise it has failed; restore the previous
960          capture values before returning NOMATCH. */
961    
962          for (;;)
963            {
964            md->offset_vector[md->offset_end - number] =
965              (int)(eptr - md->start_subject);
966            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
967            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
968              eptrb, RM63);
969            if (rrc == MATCH_KETRPOS)
970              {
971              offset_top = md->end_offset_top;
972              eptr = md->end_match_ptr;
973              ecode = md->start_code + code_offset;
974              save_capture_last = md->capture_last;
975              matched_once = TRUE;
976              continue;
977              }
978            if (rrc != MATCH_NOMATCH &&
979                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
980              RRETURN(rrc);
981            md->capture_last = save_capture_last;
982            ecode += GET(ecode, 1);
983            if (*ecode != OP_ALT) break;
984            }
985    
986          if (!matched_once)
987            {
988            md->offset_vector[offset] = save_offset1;
989            md->offset_vector[offset+1] = save_offset2;
990            md->offset_vector[md->offset_end - number] = save_offset3;
991            }
992    
993          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
994          if (allow_zero || matched_once)
995            {
996            ecode += 1 + LINK_SIZE;
997            break;
998            }
999    
1000          RRETURN(MATCH_NOMATCH);
1001          }
1002    
1003        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1004        as a non-capturing bracket. */
1005    
1006        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1007        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1008    
1009        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1010    
1011        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1012        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1013    
1014        /* Non-capturing possessive bracket with unlimited repeat. We come here
1015        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1016        without the capturing complication. It is written out separately for speed
1017        and cleanliness. */
1018    
1019        case OP_BRAPOS:
1020        case OP_SBRAPOS:
1021        allow_zero = FALSE;
1022    
1023        POSSESSIVE_NON_CAPTURE:
1024        matched_once = FALSE;
1025        code_offset = ecode - md->start_code;
1026    
1027        for (;;)
1028          {
1029          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1030          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1031            eptrb, RM48);
1032          if (rrc == MATCH_KETRPOS)
1033            {
1034            offset_top = md->end_offset_top;
1035            eptr = md->end_match_ptr;
1036            ecode = md->start_code + code_offset;
1037            matched_once = TRUE;
1038            continue;
1039            }
1040          if (rrc != MATCH_NOMATCH &&
1041              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1042            RRETURN(rrc);
1043        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1044          if (*ecode != OP_ALT) break;
1045          }
1046    
1047        if (matched_once || allow_zero)
1048          {
1049          ecode += 1 + LINK_SIZE;
1050          break;
1051        }        }
1052        RRETURN(MATCH_NOMATCH);
1053    
1054      /* Control never reaches here. */      /* Control never reaches here. */
1055    
1056      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1057      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
1058      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
1059      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. */  
1060    
1061      case OP_COND:      case OP_COND:
1062      case OP_SCOND:      case OP_SCOND:
1063      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      codelink = GET(ecode, 1);
1064    
1065        /* Because of the way auto-callout works during compile, a callout item is
1066        inserted between OP_COND and an assertion condition. */
1067    
1068        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1069        {        {
1070        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/        if (pcre_callout != NULL)
1071        condition = md->recursive != NULL &&          {
1072          (offset == RREF_ANY || offset == md->recursive->group_num);          pcre_callout_block cb;
1073        ecode += condition? 3 : GET(ecode, 1);          cb.version          = 1;   /* Version 1 of the callout block */
1074            cb.callout_number   = ecode[LINK_SIZE+2];
1075            cb.offset_vector    = md->offset_vector;
1076            cb.subject          = (PCRE_SPTR)md->start_subject;
1077            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1078            cb.start_match      = (int)(mstart - md->start_subject);
1079            cb.current_position = (int)(eptr - md->start_subject);
1080            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1081            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1082            cb.capture_top      = offset_top/2;
1083            cb.capture_last     = md->capture_last;
1084            cb.callout_data     = md->callout_data;
1085            if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1086            if (rrc < 0) RRETURN(rrc);
1087            }
1088          ecode += _pcre_OP_lengths[OP_CALLOUT];
1089          }
1090    
1091        condcode = ecode[LINK_SIZE+1];
1092    
1093        /* Now see what the actual condition is */
1094    
1095        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1096          {
1097          if (md->recursive == NULL)                /* Not recursing => FALSE */
1098            {
1099            condition = FALSE;
1100            ecode += GET(ecode, 1);
1101            }
1102          else
1103            {
1104            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1105            condition =  (recno == RREF_ANY || recno == md->recursive->group_num);
1106    
1107            /* If the test is for recursion into a specific subpattern, and it is
1108            false, but the test was set up by name, scan the table to see if the
1109            name refers to any other numbers, and test them. The condition is true
1110            if any one is set. */
1111    
1112            if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
1113              {
1114              uschar *slotA = md->name_table;
1115              for (i = 0; i < md->name_count; i++)
1116                {
1117                if (GET2(slotA, 0) == recno) break;
1118                slotA += md->name_entry_size;
1119                }
1120    
1121              /* Found a name for the number - there can be only one; duplicate
1122              names for different numbers are allowed, but not vice versa. First
1123              scan down for duplicates. */
1124    
1125              if (i < md->name_count)
1126                {
1127                uschar *slotB = slotA;
1128                while (slotB > md->name_table)
1129                  {
1130                  slotB -= md->name_entry_size;
1131                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1132                    {
1133                    condition = GET2(slotB, 0) == md->recursive->group_num;
1134                    if (condition) break;
1135                    }
1136                  else break;
1137                  }
1138    
1139                /* Scan up for duplicates */
1140    
1141                if (!condition)
1142                  {
1143                  slotB = slotA;
1144                  for (i++; i < md->name_count; i++)
1145                    {
1146                    slotB += md->name_entry_size;
1147                    if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1148                      {
1149                      condition = GET2(slotB, 0) == md->recursive->group_num;
1150                      if (condition) break;
1151                      }
1152                    else break;
1153                    }
1154                  }
1155                }
1156              }
1157    
1158            /* Chose branch according to the condition */
1159    
1160            ecode += condition? 3 : GET(ecode, 1);
1161            }
1162        }        }
1163    
1164      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1165        {        {
1166        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1167        condition = offset < offset_top && md->offset_vector[offset] >= 0;        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1168    
1169          /* If the numbered capture is unset, but the reference was by name,
1170          scan the table to see if the name refers to any other numbers, and test
1171          them. The condition is true if any one is set. This is tediously similar
1172          to the code above, but not close enough to try to amalgamate. */
1173    
1174          if (!condition && condcode == OP_NCREF)
1175            {
1176            int refno = offset >> 1;
1177            uschar *slotA = md->name_table;
1178    
1179            for (i = 0; i < md->name_count; i++)
1180              {
1181              if (GET2(slotA, 0) == refno) break;
1182              slotA += md->name_entry_size;
1183              }
1184    
1185            /* Found a name for the number - there can be only one; duplicate names
1186            for different numbers are allowed, but not vice versa. First scan down
1187            for duplicates. */
1188    
1189            if (i < md->name_count)
1190              {
1191              uschar *slotB = slotA;
1192              while (slotB > md->name_table)
1193                {
1194                slotB -= md->name_entry_size;
1195                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1196                  {
1197                  offset = GET2(slotB, 0) << 1;
1198                  condition = offset < offset_top &&
1199                    md->offset_vector[offset] >= 0;
1200                  if (condition) break;
1201                  }
1202                else break;
1203                }
1204    
1205              /* Scan up for duplicates */
1206    
1207              if (!condition)
1208                {
1209                slotB = slotA;
1210                for (i++; i < md->name_count; i++)
1211                  {
1212                  slotB += md->name_entry_size;
1213                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1214                    {
1215                    offset = GET2(slotB, 0) << 1;
1216                    condition = offset < offset_top &&
1217                      md->offset_vector[offset] >= 0;
1218                    if (condition) break;
1219                    }
1220                  else break;
1221                  }
1222                }
1223              }
1224            }
1225    
1226          /* Chose branch according to the condition */
1227    
1228        ecode += condition? 3 : GET(ecode, 1);        ecode += condition? 3 : GET(ecode, 1);
1229        }        }
1230    
1231      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      else if (condcode == OP_DEF)     /* DEFINE - always false */
1232        {        {
1233        condition = FALSE;        condition = FALSE;
1234        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1235        }        }
1236    
1237      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1238      the final argument match_condassert causes it to stop at the end of an      md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1239      assertion. */      an assertion. */
1240    
1241      else      else
1242        {        {
1243        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1244            match_condassert);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1245        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1246          {          {
1247            if (md->end_offset_top > offset_top)
1248              offset_top = md->end_offset_top;  /* Captures may have happened */
1249          condition = TRUE;          condition = TRUE;
1250          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1251          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1252          }          }
1253        else if (rrc != MATCH_NOMATCH)        else if (rrc != MATCH_NOMATCH &&
1254                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1255          {          {
1256          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1257          }          }
1258        else        else
1259          {          {
1260          condition = FALSE;          condition = FALSE;
1261          ecode += GET(ecode, 1);          ecode += codelink;
1262          }          }
1263        }        }
1264    
1265      /* We are now at the branch that is to be obeyed. As there is only one,      /* We are now at the branch that is to be obeyed. As there is only one,
1266      we can use tail recursion to avoid using another stack frame. If the second      we used to use tail recursion to avoid using another stack frame, except
1267      alternative doesn't exist, we can just plough on. */      when there was unlimited repeat of a possibly empty group. However, that
1268        strategy no longer works because of the possibilty of (*THEN) being
1269        encountered in the branch. A recursive call to match() is always required,
1270        unless the second alternative doesn't exist, in which case we can just
1271        plough on. */
1272    
1273      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1274        {        {
1275        ecode += 1 + LINK_SIZE;        if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1276        flags = match_tail_recursed | ((op == OP_SCOND)? match_cbegroup : 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1277        goto TAIL_RECURSE;        if (rrc == MATCH_THEN && md->start_match_ptr == ecode)
1278            rrc = MATCH_NOMATCH;
1279          RRETURN(rrc);
1280        }        }
1281      else      else                         /* Condition false & no alternative */
1282        {        {
1283        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1284        }        }
1285      break;      break;
1286    
1287    
1288      /* End of the pattern. If we are in a top-level recursion, we should      /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1289      restore the offsets appropriately and continue from after the call. */      to close any currently open capturing brackets. */
1290    
1291      case OP_END:      case OP_CLOSE:
1292      if (md->recursive != NULL && md->recursive->group_num == 0)      number = GET2(ecode, 1);
1293        offset = number << 1;
1294    
1295    #ifdef PCRE_DEBUG
1296          printf("end bracket %d at *ACCEPT", number);
1297          printf("\n");
1298    #endif
1299    
1300        md->capture_last = number;
1301        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1302        {        {
1303        recursion_info *rec = md->recursive;        md->offset_vector[offset] =
1304        DPRINTF(("End of pattern in a (?0) recursion\n"));          md->offset_vector[md->offset_end - number];
1305        md->recursive = rec->prevrec;        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1306        memmove(md->offset_vector, rec->offset_save,        if (offset_top <= offset) offset_top = offset + 2;
         rec->saved_max * sizeof(int));  
       md->start_match = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
1307        }        }
1308        ecode += 3;
1309        break;
1310    
     /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty  
     string - backtracking will then try other alternatives, if any. */  
1311    
1312      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* End of the pattern, either real or forced. */
1313      md->end_match_ptr = eptr;          /* Record where we ended */  
1314      md->end_offset_top = offset_top;   /* and how many extracts were taken */      case OP_END:
1315      RRETURN(MATCH_MATCH);      case OP_ACCEPT:
1316        case OP_ASSERT_ACCEPT:
1317      /* Change option settings */  
1318        /* If we have matched an empty string, fail if not in an assertion and not
1319      case OP_OPT:      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1320      ims = ecode[1];      is set and we have matched at the start of the subject. In both cases,
1321      ecode += 2;      backtracking will then try other alternatives, if any. */
1322      DPRINTF(("ims set to %02lx\n", ims));  
1323      break;      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1324             md->recursive == NULL &&
1325             (md->notempty ||
1326               (md->notempty_atstart &&
1327                 mstart == md->start_subject + md->start_offset)))
1328          MRRETURN(MATCH_NOMATCH);
1329    
1330        /* Otherwise, we have a match. */
1331    
1332        md->end_match_ptr = eptr;           /* Record where we ended */
1333        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1334        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1335    
1336        /* For some reason, the macros don't work properly if an expression is
1337        given as the argument to MRRETURN when the heap is in use. */
1338    
1339        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1340        MRRETURN(rrc);
1341    
1342      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1343      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,
1344      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
1345      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
1346      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1347        of a condition, we want to return immediately afterwards. The caller of
1348        this incarnation of the match() function will have set MATCH_CONDASSERT in
1349        md->match_function type, and one of these opcodes will be the first opcode
1350        that is processed. We use a local variable that is preserved over calls to
1351        match() to remember this case. */
1352    
1353      case OP_ASSERT:      case OP_ASSERT:
1354      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1355        if (md->match_function_type == MATCH_CONDASSERT)
1356          {
1357          condassert = TRUE;
1358          md->match_function_type = 0;
1359          }
1360        else condassert = FALSE;
1361    
1362      do      do
1363        {        {
1364        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1365        if (rrc == MATCH_MATCH) break;        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1366        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          {
1367            mstart = md->start_match_ptr;   /* In case \K reset it */
1368            markptr = md->mark;
1369            break;
1370            }
1371          if (rrc != MATCH_NOMATCH &&
1372              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1373            RRETURN(rrc);
1374        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1375        }        }
1376      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1377      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1378        if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1379    
1380      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1381    
1382      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1383    
1384      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1385      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 822  for (;;) Line 1389  for (;;)
1389      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1390      continue;      continue;
1391    
1392      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1393        PRUNE, or COMMIT means we must assume failure without checking subsequent
1394        branches. */
1395    
1396      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1397      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1398        if (md->match_function_type == MATCH_CONDASSERT)
1399          {
1400          condassert = TRUE;
1401          md->match_function_type = 0;
1402          }
1403        else condassert = FALSE;
1404    
1405      do      do
1406        {        {
1407        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1408        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1409        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1410            {
1411            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1412            break;
1413            }
1414          if (rrc != MATCH_NOMATCH &&
1415              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1416            RRETURN(rrc);
1417        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1418        }        }
1419      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1420    
1421      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1422    
1423      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1424      continue;      continue;
# Line 853  for (;;) Line 1436  for (;;)
1436        while (i-- > 0)        while (i-- > 0)
1437          {          {
1438          eptr--;          eptr--;
1439          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1440          BACKCHAR(eptr)          BACKCHAR(eptr);
1441          }          }
1442        }        }
1443      else      else
# Line 864  for (;;) Line 1447  for (;;)
1447    
1448        {        {
1449        eptr -= GET(ecode, 1);        eptr -= GET(ecode, 1);
1450        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1451        }        }
1452    
1453      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1454    
1455        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1456      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1457      break;      break;
1458    
# Line 884  for (;;) Line 1468  for (;;)
1468        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1469        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1470        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1471        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1472        cb.start_match      = md->start_match - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1473        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1474        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1475        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1476        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1477        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1478        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1479        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1480        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1481        }        }
1482      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 902  for (;;) Line 1486  for (;;)
1486      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
1487      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1488    
1489      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1490      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
1491      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
1492      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
1493      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
1494      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
1495      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.  
1496    
1497      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
1498      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
1499      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1500        a lot, so he is not to blame for the current way it works. */
1501    
1502      case OP_RECURSE:      case OP_RECURSE:
1503        {        {
# Line 928  for (;;) Line 1510  for (;;)
1510        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1511        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1512    
1513        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1514    
1515        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1516    
1517        /* Now save the offset data. */        /* Now save the offset data */
1518    
1519        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1520        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 944  for (;;) Line 1525  for (;;)
1525            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1526          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1527          }          }
   
1528        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1529              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1530    
1531        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1532        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1533          might be changed, so reset it before looping. */
1534    
1535        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1536        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1537        do        do
1538          {          {
1539          RMATCH(rrc, eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1540            md, ims, eptrb, flags);          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1541          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1542            memcpy(md->offset_vector, new_recursive.offset_save,
1543                new_recursive.saved_max * sizeof(int));
1544            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1545            {            {
1546            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
1547            md->recursive = new_recursive.prevrec;            md->recursive = new_recursive.prevrec;
1548            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1549              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1550            RRETURN(MATCH_MATCH);  
1551              /* Set where we got to in the subject, and reset the start in case
1552              it was changed by \K. This *is* propagated back out of a recursion,
1553              for Perl compatibility. */
1554    
1555              eptr = md->end_match_ptr;
1556              mstart = md->start_match_ptr;
1557              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1558            }            }
1559          else if (rrc != MATCH_NOMATCH)          else if (rrc != MATCH_NOMATCH &&
1560                    (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1561            {            {
1562            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1563              if (new_recursive.offset_save != stacksave)
1564                (pcre_free)(new_recursive.offset_save);
1565            RRETURN(rrc);            RRETURN(rrc);
1566            }            }
1567    
1568          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1569          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1570          }          }
1571        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 984  for (;;) Line 1574  for (;;)
1574        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1575        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1576          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1577        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1578        }        }
     /* Control never reaches here */  
1579    
1580      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1581      the point in the subject string is not moved back. Thus there can never be      break;
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
1582    
1583      case OP_ONCE:      /* An alternation is the end of a branch; scan along to find the end of the
1584      prev = ecode;      bracketed group and go to there. */
     saved_eptr = eptr;  
1585    
1586      do      case OP_ALT:
1587        {      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1588        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,      break;
         eptrb, 0);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
1589    
1590      /* If hit the end of the group (which could be repeated), fail */      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1591        indicating that it may occur zero times. It may repeat infinitely, or not
1592        at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1593        with fixed upper repeat limits are compiled as a number of copies, with the
1594        optional ones preceded by BRAZERO or BRAMINZERO. */
1595    
1596      if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);      case OP_BRAZERO:
1597        next = ecode + 1;
1598        RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1599        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1600        do next += GET(next, 1); while (*next == OP_ALT);
1601        ecode = next + 1 + LINK_SIZE;
1602        break;
1603    
1604      /* Continue as from after the assertion, updating the offsets high water      case OP_BRAMINZERO:
1605      mark, since extracts may have been taken. */      next = ecode + 1;
1606        do next += GET(next, 1); while (*next == OP_ALT);
1607        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1608        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1609        ecode++;
1610        break;
1611    
1612      do ecode += GET(ecode, 1); while (*ecode == OP_ALT);      case OP_SKIPZERO:
1613        next = ecode+1;
1614      offset_top = md->end_offset_top;      do next += GET(next,1); while (*next == OP_ALT);
1615      eptr = md->end_match_ptr;      ecode = next + 1 + LINK_SIZE;
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_cbegroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
   
     /* An alternation is the end of a branch; scan along to find the end of the  
     bracketed group and go to there. */  
   
     case OP_ALT:  
     do ecode += GET(ecode,1); while (*ecode == OP_ALT);  
1616      break;      break;
1617    
1618      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1619      that it may occur zero times. It may repeat infinitely, or not at all -      here; just jump to the group, with allow_zero set TRUE. */
     i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper  
     repeat limits are compiled as a number of copies, with the optional ones  
     preceded by BRAZERO or BRAMINZERO. */  
   
     case OP_BRAZERO:  
       {  
       next = ecode+1;  
       RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       do next += GET(next,1); while (*next == OP_ALT);  
       ecode = next + 1 + LINK_SIZE;  
       }  
     break;  
1620    
1621      case OP_BRAMINZERO:      case OP_BRAPOSZERO:
1622        {      op = *(++ecode);
1623        next = ecode+1;      allow_zero = TRUE;
1624        do next += GET(next, 1); while (*next == OP_ALT);      if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1625        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        goto POSSESSIVE_NON_CAPTURE;
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode++;  
       }  
     break;  
1626    
1627      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1628    
1629      case OP_KET:      case OP_KET:
1630      case OP_KETRMIN:      case OP_KETRMIN:
1631      case OP_KETRMAX:      case OP_KETRMAX:
1632        case OP_KETRPOS:
1633      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1634    
1635      /* If this was a group that remembered the subject start, in order to break      /* If this was a group that remembered the subject start, in order to break
1636      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1637      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1638    
1639      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1640        {        {
1641        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1642        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
# Line 1116  for (;;) Line 1645  for (;;)
1645    
1646      /* If we are at the end of an assertion group, stop matching and return      /* If we are at the end of an assertion group, stop matching and return
1647      MATCH_MATCH, but record the current high water mark for use by positive      MATCH_MATCH, but record the current high water mark for use by positive
1648      assertions. Do this also for the "once" (atomic) groups. */      assertions. We also need to record the match start in case it was changed
1649        by \K. */
1650    
1651      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1652          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1653        {        {
1654        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1655        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1656        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1657          MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1658        }        }
1659    
1660      /* For capturing groups we have to check the group number back at the start      /* For capturing groups we have to check the group number back at the start
1661      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1662      bumping the high water mark. Note that whole-pattern recursion is coded as      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1663      a recurse into group 0, so it won't be picked up here. Instead, we catch it      into group 0, so it won't be picked up here. Instead, we catch it when the
1664      when the OP_END is reached. Other recursion is handled here. */      OP_END is reached. Other recursion is handled here. We just have to record
1665        the current subject position and start match pointer and give a MATCH
1666        return. */
1667    
1668      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1669            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1670        {        {
1671        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1672        offset = number << 1;        offset = number << 1;
1673    
1674  #ifdef DEBUG  #ifdef PCRE_DEBUG
1675        printf("end bracket %d", number);        printf("end bracket %d", number);
1676        printf("\n");        printf("\n");
1677  #endif  #endif
1678    
1679          /* Handle a recursively called group. */
1680    
1681          if (md->recursive != NULL && md->recursive->group_num == number)
1682            {
1683            md->end_match_ptr = eptr;
1684            md->start_match_ptr = mstart;
1685            RRETURN(MATCH_MATCH);
1686            }
1687    
1688          /* Deal with capturing */
1689    
1690        md->capture_last = number;        md->capture_last = number;
1691        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1692          {          {
1693            /* If offset is greater than offset_top, it means that we are
1694            "skipping" a capturing group, and that group's offsets must be marked
1695            unset. In earlier versions of PCRE, all the offsets were unset at the
1696            start of matching, but this doesn't work because atomic groups and
1697            assertions can cause a value to be set that should later be unset.
1698            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1699            part of the atomic group, but this is not on the final matching path,
1700            so must be unset when 2 is set. (If there is no group 2, there is no
1701            problem, because offset_top will then be 2, indicating no capture.) */
1702    
1703            if (offset > offset_top)
1704              {
1705              register int *iptr = md->offset_vector + offset_top;
1706              register int *iend = md->offset_vector + offset;
1707              while (iptr < iend) *iptr++ = -1;
1708              }
1709    
1710            /* Now make the extraction */
1711    
1712          md->offset_vector[offset] =          md->offset_vector[offset] =
1713            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1714          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1715          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1716          }          }
1717          }
1718    
1719        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1720        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1721        group. This is the forcible breaking of infinite loops as implemented in
1722        Perl 5.005. For a non-repeating atomic group, establish a backup point by
1723        processing the rest of the pattern at a lower level. If this results in a
1724        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1725        bypassing intermediate backup points, but resetting any captures that
1726        happened along the way. */
1727    
1728        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1729          {
1730          if (*prev == OP_ONCE)
1731          {          {
1732          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1733          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1734          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1735          md->start_match = rec->save_start;          RRETURN(MATCH_ONCE);
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
1736          }          }
1737          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1738          break;
1739        }        }
1740    
1741      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1742      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1743        at a time from the outer level, thus saving stack. */
     ims = original_ims;  
     DPRINTF(("ims reset to %02lx\n", ims));  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
1744    
1745      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KETRPOS)
1746        {        {
1747        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1748        break;        md->end_offset_top = offset_top;
1749          RRETURN(MATCH_KETRPOS);
1750        }        }
1751    
1752      /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
1753      preceding bracket, in the appropriate order. In the second case, we can use      the preceding bracket, in the appropriate order. In the second case, we can
1754      tail recursion to avoid using another stack frame. */      use tail recursion to avoid using another stack frame, unless we have an
1755        an atomic group or an unlimited repeat of a group that can match an empty
1756      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      string. */
1757    
1758      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1759        {        {
1760        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1761        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1762          if (*prev == OP_ONCE)
1763            {
1764            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1765            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1766            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1767            RRETURN(MATCH_ONCE);
1768            }
1769          if (*prev >= OP_SBRA)    /* Could match an empty string */
1770            {
1771            md->match_function_type = MATCH_CBEGROUP;
1772            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1773            RRETURN(rrc);
1774            }
1775        ecode = prev;        ecode = prev;
       flags |= match_tail_recursed;  
1776        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1777        }        }
1778      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1779        {        {
1780        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, flags);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1781          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1782          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1783        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1784          if (*prev == OP_ONCE)
1785            {
1786            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1787            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1788            md->once_target = prev;
1789            RRETURN(MATCH_ONCE);
1790            }
1791        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_tail_recursed;  
1792        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1793        }        }
1794      /* Control never gets here */      /* Control never gets here */
1795    
1796      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1797    
1798      case OP_CIRC:      case OP_CIRC:
1799      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 || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
1800    
1801      /* Start of subject assertion */      /* Start of subject assertion */
1802    
1803      case OP_SOD:      case OP_SOD:
1804      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1805        ecode++;
1806        break;
1807    
1808        /* Multiline mode: start of subject unless notbol, or after any newline. */
1809    
1810        case OP_CIRCM:
1811        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1812        if (eptr != md->start_subject &&
1813            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1814          MRRETURN(MATCH_NOMATCH);
1815      ecode++;      ecode++;
1816      break;      break;
1817    
1818      /* Start of match assertion */      /* Start of match assertion */
1819    
1820      case OP_SOM:      case OP_SOM:
1821      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
1822      ecode++;      ecode++;
1823      break;      break;
1824    
1825      /* 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. */  
1826    
1827      case OP_DOLL:      case OP_SET_SOM:
1828      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
1829        {      ecode++;
1830        if (eptr < md->end_subject)      break;
1831          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
1832        else      /* Multiline mode: assert before any newline, or before end of subject
1833          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
1834        ecode++;  
1835        break;      case OP_DOLLM:
1836        }      if (eptr < md->end_subject)
1837          { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
1838      else      else
1839        {        {
1840        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1841        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
1842        }        }
1843        ecode++;
1844        break;
1845    
1846        /* Not multiline mode: assert before a terminating newline or before end of
1847        subject unless noteol is set. */
1848    
1849        case OP_DOLL:
1850        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1851        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1852    
1853      /* ... else fall through for endonly */      /* ... else fall through for endonly */
1854    
1855      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
1856    
1857      case OP_EOD:      case OP_EOD:
1858      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
1859        SCHECK_PARTIAL();
1860      ecode++;      ecode++;
1861      break;      break;
1862    
1863      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
1864    
1865      case OP_EODN:      case OP_EODN:
1866      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
1867        if (eptr < md->end_subject &&
1868          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1869        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1870    
1871        /* Either at end of string or \n before end. */
1872    
1873        SCHECK_PARTIAL();
1874      ecode++;      ecode++;
1875      break;      break;
1876    
# Line 1290  for (;;) Line 1882  for (;;)
1882    
1883        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
1884        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
1885        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
1886          partial matching. */
1887    
1888  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1889        if (utf8)        if (utf8)
1890          {          {
1891            /* Get status of previous character */
1892    
1893          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
1894            {            {
1895            const uschar *lastptr = eptr - 1;            USPTR lastptr = eptr - 1;
1896            while((*lastptr & 0xc0) == 0x80) lastptr--;            while((*lastptr & 0xc0) == 0x80) lastptr--;
1897              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
1898            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
1899    #ifdef SUPPORT_UCP
1900              if (md->use_ucp)
1901                {
1902                if (c == '_') prev_is_word = TRUE; else
1903                  {
1904                  int cat = UCD_CATEGORY(c);
1905                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1906                  }
1907                }
1908              else
1909    #endif
1910            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1911            }            }
1912          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
1913            /* Get status of next character */
1914    
1915            if (eptr >= md->end_subject)
1916              {
1917              SCHECK_PARTIAL();
1918              cur_is_word = FALSE;
1919              }
1920            else
1921            {            {
1922            GETCHAR(c, eptr);            GETCHAR(c, eptr);
1923    #ifdef SUPPORT_UCP
1924              if (md->use_ucp)
1925                {
1926                if (c == '_') cur_is_word = TRUE; else
1927                  {
1928                  int cat = UCD_CATEGORY(c);
1929                  cur_is_word = (cat == ucp_L || cat == ucp_N);
1930                  }
1931                }
1932              else
1933    #endif
1934            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1935            }            }
1936          }          }
1937        else        else
1938  #endif  #endif
1939    
1940        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
1941          consistency with the behaviour of \w we do use it in this case. */
1942    
1943          {          {
1944          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
1945            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
1946          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
1947            ((md->ctypes[*eptr] & ctype_word) != 0);            {
1948              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
1949    #ifdef SUPPORT_UCP
1950              if (md->use_ucp)
1951                {
1952                c = eptr[-1];
1953                if (c == '_') prev_is_word = TRUE; else
1954                  {
1955                  int cat = UCD_CATEGORY(c);
1956                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1957                  }
1958                }
1959              else
1960    #endif
1961              prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
1962              }
1963    
1964            /* Get status of next character */
1965    
1966            if (eptr >= md->end_subject)
1967              {
1968              SCHECK_PARTIAL();
1969              cur_is_word = FALSE;
1970              }
1971            else
1972    #ifdef SUPPORT_UCP
1973            if (md->use_ucp)
1974              {
1975              c = *eptr;
1976              if (c == '_') cur_is_word = TRUE; else
1977                {
1978                int cat = UCD_CATEGORY(c);
1979                cur_is_word = (cat == ucp_L || cat == ucp_N);
1980                }
1981              }
1982            else
1983    #endif
1984            cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
1985          }          }
1986    
1987        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
1988    
1989        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
1990             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
1991          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
1992        }        }
1993      break;      break;
1994    
1995      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
1996    
1997      case OP_ANY:      case OP_ANY:
1998      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
1999        /* Fall through */
2000    
2001        case OP_ALLANY:
2002        if (eptr++ >= md->end_subject)
2003        {        {
2004        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);        SCHECK_PARTIAL();
2005          MRRETURN(MATCH_NOMATCH);
2006        }        }
2007      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
     if (utf8)  
       while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;  
2008      ecode++;      ecode++;
2009      break;      break;
2010    
# Line 1345  for (;;) Line 2012  for (;;)
2012      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2013    
2014      case OP_ANYBYTE:      case OP_ANYBYTE:
2015      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr++ >= md->end_subject)
2016          {
2017          SCHECK_PARTIAL();
2018          MRRETURN(MATCH_NOMATCH);
2019          }
2020      ecode++;      ecode++;
2021      break;      break;
2022    
2023      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2024      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2025          {
2026          SCHECK_PARTIAL();
2027          MRRETURN(MATCH_NOMATCH);
2028          }
2029      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2030      if (      if (
2031  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1358  for (;;) Line 2033  for (;;)
2033  #endif  #endif
2034         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2035         )         )
2036        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2037      ecode++;      ecode++;
2038      break;      break;
2039    
2040      case OP_DIGIT:      case OP_DIGIT:
2041      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2042          {
2043          SCHECK_PARTIAL();
2044          MRRETURN(MATCH_NOMATCH);
2045          }
2046      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2047      if (      if (
2048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1371  for (;;) Line 2050  for (;;)
2050  #endif  #endif
2051         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2052         )         )
2053        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2054      ecode++;      ecode++;
2055      break;      break;
2056    
2057      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2058      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2059          {
2060          SCHECK_PARTIAL();
2061          MRRETURN(MATCH_NOMATCH);
2062          }
2063      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2064      if (      if (
2065  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1384  for (;;) Line 2067  for (;;)
2067  #endif  #endif
2068         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2069         )         )
2070        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2071      ecode++;      ecode++;
2072      break;      break;
2073    
2074      case OP_WHITESPACE:      case OP_WHITESPACE:
2075      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2076          {
2077          SCHECK_PARTIAL();
2078          MRRETURN(MATCH_NOMATCH);
2079          }
2080      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2081      if (      if (
2082  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1397  for (;;) Line 2084  for (;;)
2084  #endif  #endif
2085         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2086         )         )
2087        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2088      ecode++;      ecode++;
2089      break;      break;
2090    
2091      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2092      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2093          {
2094          SCHECK_PARTIAL();
2095          MRRETURN(MATCH_NOMATCH);
2096          }
2097      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2098      if (      if (
2099  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1410  for (;;) Line 2101  for (;;)
2101  #endif  #endif
2102         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2103         )         )
2104        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2105      ecode++;      ecode++;
2106      break;      break;
2107    
2108      case OP_WORDCHAR:      case OP_WORDCHAR:
2109      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2110          {
2111          SCHECK_PARTIAL();
2112          MRRETURN(MATCH_NOMATCH);
2113          }
2114      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2115      if (      if (
2116  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1423  for (;;) Line 2118  for (;;)
2118  #endif  #endif
2119         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2120         )         )
2121        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2122      ecode++;      ecode++;
2123      break;      break;
2124    
2125      case OP_ANYNL:      case OP_ANYNL:
2126      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2127          {
2128          SCHECK_PARTIAL();
2129          MRRETURN(MATCH_NOMATCH);
2130          }
2131      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2132      switch(c)      switch(c)
2133        {        {
2134        default: RRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2135    
2136        case 0x000d:        case 0x000d:
2137        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2138        break;        break;
2139    
2140        case 0x000a:        case 0x000a:
2141          break;
2142    
2143        case 0x000b:        case 0x000b:
2144        case 0x000c:        case 0x000c:
2145        case 0x0085:        case 0x0085:
2146        case 0x2028:        case 0x2028:
2147        case 0x2029:        case 0x2029:
2148          if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
2149          break;
2150          }
2151        ecode++;
2152        break;
2153    
2154        case OP_NOT_HSPACE:
2155        if (eptr >= md->end_subject)
2156          {
2157          SCHECK_PARTIAL();
2158          MRRETURN(MATCH_NOMATCH);
2159          }
2160        GETCHARINCTEST(c, eptr);
2161        switch(c)
2162          {
2163          default: break;
2164          case 0x09:      /* HT */
2165          case 0x20:      /* SPACE */
2166          case 0xa0:      /* NBSP */
2167          case 0x1680:    /* OGHAM SPACE MARK */
2168          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2169          case 0x2000:    /* EN QUAD */
2170          case 0x2001:    /* EM QUAD */
2171          case 0x2002:    /* EN SPACE */
2172          case 0x2003:    /* EM SPACE */
2173          case 0x2004:    /* THREE-PER-EM SPACE */
2174          case 0x2005:    /* FOUR-PER-EM SPACE */
2175          case 0x2006:    /* SIX-PER-EM SPACE */
2176          case 0x2007:    /* FIGURE SPACE */
2177          case 0x2008:    /* PUNCTUATION SPACE */
2178          case 0x2009:    /* THIN SPACE */
2179          case 0x200A:    /* HAIR SPACE */
2180          case 0x202f:    /* NARROW NO-BREAK SPACE */
2181          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2182          case 0x3000:    /* IDEOGRAPHIC SPACE */
2183          MRRETURN(MATCH_NOMATCH);
2184          }
2185        ecode++;
2186        break;
2187    
2188        case OP_HSPACE:
2189        if (eptr >= md->end_subject)
2190          {
2191          SCHECK_PARTIAL();
2192          MRRETURN(MATCH_NOMATCH);
2193          }
2194        GETCHARINCTEST(c, eptr);
2195        switch(c)
2196          {
2197          default: MRRETURN(MATCH_NOMATCH);
2198          case 0x09:      /* HT */
2199          case 0x20:      /* SPACE */
2200          case 0xa0:      /* NBSP */
2201          case 0x1680:    /* OGHAM SPACE MARK */
2202          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2203          case 0x2000:    /* EN QUAD */
2204          case 0x2001:    /* EM QUAD */
2205          case 0x2002:    /* EN SPACE */
2206          case 0x2003:    /* EM SPACE */
2207          case 0x2004:    /* THREE-PER-EM SPACE */
2208          case 0x2005:    /* FOUR-PER-EM SPACE */
2209          case 0x2006:    /* SIX-PER-EM SPACE */
2210          case 0x2007:    /* FIGURE SPACE */
2211          case 0x2008:    /* PUNCTUATION SPACE */
2212          case 0x2009:    /* THIN SPACE */
2213          case 0x200A:    /* HAIR SPACE */
2214          case 0x202f:    /* NARROW NO-BREAK SPACE */
2215          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2216          case 0x3000:    /* IDEOGRAPHIC SPACE */
2217          break;
2218          }
2219        ecode++;
2220        break;
2221    
2222        case OP_NOT_VSPACE:
2223        if (eptr >= md->end_subject)
2224          {
2225          SCHECK_PARTIAL();
2226          MRRETURN(MATCH_NOMATCH);
2227          }
2228        GETCHARINCTEST(c, eptr);
2229        switch(c)
2230          {
2231          default: break;
2232          case 0x0a:      /* LF */
2233          case 0x0b:      /* VT */
2234          case 0x0c:      /* FF */
2235          case 0x0d:      /* CR */
2236          case 0x85:      /* NEL */
2237          case 0x2028:    /* LINE SEPARATOR */
2238          case 0x2029:    /* PARAGRAPH SEPARATOR */
2239          MRRETURN(MATCH_NOMATCH);
2240          }
2241        ecode++;
2242        break;
2243    
2244        case OP_VSPACE:
2245        if (eptr >= md->end_subject)
2246          {
2247          SCHECK_PARTIAL();
2248          MRRETURN(MATCH_NOMATCH);
2249          }
2250        GETCHARINCTEST(c, eptr);
2251        switch(c)
2252          {
2253          default: MRRETURN(MATCH_NOMATCH);
2254          case 0x0a:      /* LF */
2255          case 0x0b:      /* VT */
2256          case 0x0c:      /* FF */
2257          case 0x0d:      /* CR */
2258          case 0x85:      /* NEL */
2259          case 0x2028:    /* LINE SEPARATOR */
2260          case 0x2029:    /* PARAGRAPH SEPARATOR */
2261        break;        break;
2262        }        }
2263      ecode++;      ecode++;
# Line 1453  for (;;) Line 2269  for (;;)
2269    
2270      case OP_PROP:      case OP_PROP:
2271      case OP_NOTPROP:      case OP_NOTPROP:
2272      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2273          {
2274          SCHECK_PARTIAL();
2275          MRRETURN(MATCH_NOMATCH);
2276          }
2277      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2278        {        {
2279        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2280    
2281        switch(ecode[1])        switch(ecode[1])
2282          {          {
2283          case PT_ANY:          case PT_ANY:
2284          if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
2285          break;          break;
2286    
2287          case PT_LAMP:          case PT_LAMP:
2288          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2289               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2290               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2291            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2292           break;          break;
2293    
2294          case PT_GC:          case PT_GC:
2295          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
2296            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2297          break;          break;
2298    
2299          case PT_PC:          case PT_PC:
2300          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2301            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2302          break;          break;
2303    
2304          case PT_SC:          case PT_SC:
2305          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2306            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2307            break;
2308    
2309            /* These are specials */
2310    
2311            case PT_ALNUM:
2312            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2313                 _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2314              MRRETURN(MATCH_NOMATCH);
2315          break;          break;
2316    
2317            case PT_SPACE:    /* Perl space */
2318            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2319                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2320                   == (op == OP_NOTPROP))
2321              MRRETURN(MATCH_NOMATCH);
2322            break;
2323    
2324            case PT_PXSPACE:  /* POSIX space */
2325            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2326                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2327                 c == CHAR_FF || c == CHAR_CR)
2328                   == (op == OP_NOTPROP))
2329              MRRETURN(MATCH_NOMATCH);
2330            break;
2331    
2332            case PT_WORD:
2333            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2334                 _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2335                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2336              MRRETURN(MATCH_NOMATCH);
2337            break;
2338    
2339            /* This should never occur */
2340    
2341          default:          default:
2342          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2343          }          }
# Line 1499  for (;;) Line 2350  for (;;)
2350      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2351    
2352      case OP_EXTUNI:      case OP_EXTUNI:
2353      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2354          {
2355          SCHECK_PARTIAL();
2356          MRRETURN(MATCH_NOMATCH);
2357          }
2358      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2359        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2360        while (eptr < md->end_subject)
2361        {        {
2362        int chartype, script;        int len = 1;
2363        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2364        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2365        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;  
         }  
2366        }        }
2367      ecode++;      ecode++;
2368      break;      break;
# Line 1531  for (;;) Line 2378  for (;;)
2378      loops). */      loops). */
2379    
2380      case OP_REF:      case OP_REF:
2381        {      case OP_REFI:
2382        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2383        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2384        ecode += 3;
2385    
2386        /* 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];  
2387    
2388        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2389        this ensures that every attempt at a match fails. We can't just fail
2390        here, because of the possibility of quantifiers with zero minima.
2391    
2392        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2393          {      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;  
2394    
2395          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2396          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;  
2397    
2398          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2399          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2400          eptr += length;      else
2401          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2402    
2403        /* Set up for repetition, or handle the non-repeated case */
2404    
2405        switch (*ecode)
2406          {
2407          case OP_CRSTAR:
2408          case OP_CRMINSTAR:
2409          case OP_CRPLUS:
2410          case OP_CRMINPLUS:
2411          case OP_CRQUERY:
2412          case OP_CRMINQUERY:
2413          c = *ecode++ - OP_CRSTAR;
2414          minimize = (c & 1) != 0;
2415          min = rep_min[c];                 /* Pick up values from tables; */
2416          max = rep_max[c];                 /* zero for max => infinity */
2417          if (max == 0) max = INT_MAX;
2418          break;
2419    
2420          case OP_CRRANGE:
2421          case OP_CRMINRANGE:
2422          minimize = (*ecode == OP_CRMINRANGE);
2423          min = GET2(ecode, 1);
2424          max = GET2(ecode, 3);
2425          if (max == 0) max = INT_MAX;
2426          ecode += 5;
2427          break;
2428    
2429          default:               /* No repeat follows */
2430          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2431            {
2432            CHECK_PARTIAL();
2433            MRRETURN(MATCH_NOMATCH);
2434          }          }
2435          eptr += length;
2436          continue;              /* With the main loop */
2437          }
2438    
2439        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2440        main loop. */      zero, just continue with the main loop. */
2441    
2442        if (length == 0) continue;      if (length == 0) continue;
2443    
2444        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2445        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2446        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2447    
2448        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2449          {
2450          int slength;
2451          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2452          {          {
2453          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2454          eptr += length;          MRRETURN(MATCH_NOMATCH);
2455          }          }
2456          eptr += slength;
2457          }
2458    
2459        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2460        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2461    
2462        if (min == max) continue;      if (min == max) continue;
2463    
2464        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2465    
2466        if (minimize)      if (minimize)
2467          {
2468          for (fi = min;; fi++)
2469          {          {
2470          for (fi = min;; fi++)          int slength;
2471            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2472            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2473            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2474            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2475            {            {
2476            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2477            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2478            }            }
2479          /* Control never gets here */          eptr += slength;
2480          }          }
2481          /* Control never gets here */
2482          }
2483    
2484        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2485    
2486        else      else
2487          {
2488          pp = eptr;
2489          for (i = min; i < max; i++)
2490          {          {
2491          pp = eptr;          int slength;
2492          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
2493            {            {
2494            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2495            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2496            }            }
2497          RRETURN(MATCH_NOMATCH);          eptr += slength;
2498            }
2499          while (eptr >= pp)
2500            {
2501            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2502            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2503            eptr -= length;
2504          }          }
2505          MRRETURN(MATCH_NOMATCH);
2506        }        }
2507      /* Control never gets here */      /* Control never gets here */
2508    
   
   
2509      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2510      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,
2511      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 1688  for (;;) Line 2560  for (;;)
2560          {          {
2561          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2562            {            {
2563            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2564                {
2565                SCHECK_PARTIAL();
2566                MRRETURN(MATCH_NOMATCH);
2567                }
2568            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2569            if (c > 255)            if (c > 255)
2570              {              {
2571              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2572              }              }
2573            else            else
2574              {              {
2575              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2576              }              }
2577            }            }
2578          }          }
# Line 1706  for (;;) Line 2582  for (;;)
2582          {          {
2583          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2584            {            {
2585            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2586                {
2587                SCHECK_PARTIAL();
2588                MRRETURN(MATCH_NOMATCH);
2589                }
2590            c = *eptr++;            c = *eptr++;
2591            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2592            }            }
2593          }          }
2594    
# Line 1728  for (;;) Line 2608  for (;;)
2608            {            {
2609            for (fi = min;; fi++)            for (fi = min;; fi++)
2610              {              {
2611              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2612              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2613              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2614                if (eptr >= md->end_subject)
2615                  {
2616                  SCHECK_PARTIAL();
2617                  MRRETURN(MATCH_NOMATCH);
2618                  }
2619              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2620              if (c > 255)              if (c > 255)
2621                {                {
2622                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2623                }                }
2624              else              else
2625                {                {
2626                if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);                if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2627                }                }
2628              }              }
2629            }            }
# Line 1748  for (;;) Line 2633  for (;;)
2633            {            {
2634            for (fi = min;; fi++)            for (fi = min;; fi++)
2635              {              {
2636              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2637              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2638              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2639                if (eptr >= md->end_subject)
2640                  {
2641                  SCHECK_PARTIAL();
2642                  MRRETURN(MATCH_NOMATCH);
2643                  }
2644              c = *eptr++;              c = *eptr++;
2645              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2646              }              }
2647            }            }
2648          /* Control never gets here */          /* Control never gets here */
# Line 1771  for (;;) Line 2661  for (;;)
2661            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2662              {              {
2663              int len = 1;              int len = 1;
2664              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2665                  {
2666                  SCHECK_PARTIAL();
2667                  break;
2668                  }
2669              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2670              if (c > 255)              if (c > 255)
2671                {                {
# Line 1785  for (;;) Line 2679  for (;;)
2679              }              }
2680            for (;;)            for (;;)
2681              {              {
2682              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2683              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2684              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2685              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1797  for (;;) Line 2691  for (;;)
2691            {            {
2692            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2693              {              {
2694              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2695                  {
2696                  SCHECK_PARTIAL();
2697                  break;
2698                  }
2699              c = *eptr;              c = *eptr;
2700              if ((data[c/8] & (1 << (c&7))) == 0) break;              if ((data[c/8] & (1 << (c&7))) == 0) break;
2701              eptr++;              eptr++;
2702              }              }
2703            while (eptr >= pp)            while (eptr >= pp)
2704              {              {
2705              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2706              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2707              eptr--;              eptr--;
2708              }              }
2709            }            }
2710    
2711          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2712          }          }
2713        }        }
2714      /* Control never gets here */      /* Control never gets here */
2715    
2716    
2717      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2718      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
2719        mode, because Unicode properties are supported in non-UTF-8 mode. */
2720    
2721  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2722      case OP_XCLASS:      case OP_XCLASS:
# Line 1858  for (;;) Line 2757  for (;;)
2757    
2758        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2759          {          {
2760          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2761          GETCHARINC(c, eptr);            {
2762          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2763              MRRETURN(MATCH_NOMATCH);
2764              }
2765            GETCHARINCTEST(c, eptr);
2766            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2767          }          }
2768    
2769        /* 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 1875  for (;;) Line 2778  for (;;)
2778          {          {
2779          for (fi = min;; fi++)          for (fi = min;; fi++)
2780            {            {
2781            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2782            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2783            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2784            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
2785            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
2786                SCHECK_PARTIAL();
2787                MRRETURN(MATCH_NOMATCH);
2788                }
2789              GETCHARINCTEST(c, eptr);
2790              if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2791            }            }
2792          /* Control never gets here */          /* Control never gets here */
2793          }          }
# Line 1892  for (;;) Line 2800  for (;;)
2800          for (i = min; i < max; i++)          for (i = min; i < max; i++)
2801            {            {
2802            int len = 1;            int len = 1;
2803            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
2804            GETCHARLEN(c, eptr, len);              {
2805                SCHECK_PARTIAL();
2806                break;
2807                }
2808              GETCHARLENTEST(c, eptr, len);
2809            if (!_pcre_xclass(c, data)) break;            if (!_pcre_xclass(c, data)) break;
2810            eptr += len;            eptr += len;
2811            }            }
2812          for(;;)          for(;;)
2813            {            {
2814            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2815            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2816            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2817            BACKCHAR(eptr)            if (utf8) BACKCHAR(eptr);
2818            }            }
2819          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2820          }          }
2821    
2822        /* Control never gets here */        /* Control never gets here */
# Line 1920  for (;;) Line 2832  for (;;)
2832        length = 1;        length = 1;
2833        ecode++;        ecode++;
2834        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2835        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2836        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
2837            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2838            MRRETURN(MATCH_NOMATCH);
2839            }
2840          while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
2841        }        }
2842      else      else
2843  #endif  #endif
2844    
2845      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2846        {        {
2847        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2848        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);          {
2849            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2850            MRRETURN(MATCH_NOMATCH);
2851            }
2852          if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
2853        ecode += 2;        ecode += 2;
2854        }        }
2855      break;      break;
2856    
2857      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2858    
2859      case OP_CHARNC:      case OP_CHARI:
2860  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2861      if (utf8)      if (utf8)
2862        {        {
# Line 1944  for (;;) Line 2864  for (;;)
2864        ecode++;        ecode++;
2865        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2866    
2867        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2868            {
2869            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2870            MRRETURN(MATCH_NOMATCH);
2871            }
2872    
2873        /* 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
2874        can use the fast lookup table. */        can use the fast lookup table. */
2875    
2876        if (fc < 128)        if (fc < 128)
2877          {          {
2878          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2879          }          }
2880    
2881        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character */
# Line 1968  for (;;) Line 2892  for (;;)
2892          if (fc != dc)          if (fc != dc)
2893            {            {
2894  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2895            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
2896  #endif  #endif
2897              RRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
2898            }            }
2899          }          }
2900        }        }
# Line 1979  for (;;) Line 2903  for (;;)
2903    
2904      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2905        {        {
2906        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2907        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
2908            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2909            MRRETURN(MATCH_NOMATCH);
2910            }
2911          if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2912        ecode += 2;        ecode += 2;
2913        }        }
2914      break;      break;
# Line 1988  for (;;) Line 2916  for (;;)
2916      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
2917    
2918      case OP_EXACT:      case OP_EXACT:
2919        case OP_EXACTI:
2920      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
2921      ecode += 3;      ecode += 3;
2922      goto REPEATCHAR;      goto REPEATCHAR;
2923    
2924      case OP_POSUPTO:      case OP_POSUPTO:
2925        case OP_POSUPTOI:
2926      possessive = TRUE;      possessive = TRUE;
2927      /* Fall through */      /* Fall through */
2928    
2929      case OP_UPTO:      case OP_UPTO:
2930        case OP_UPTOI:
2931      case OP_MINUPTO:      case OP_MINUPTO:
2932        case OP_MINUPTOI:
2933      min = 0;      min = 0;
2934      max = GET2(ecode, 1);      max = GET2(ecode, 1);
2935      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
2936      ecode += 3;      ecode += 3;
2937      goto REPEATCHAR;      goto REPEATCHAR;
2938    
2939      case OP_POSSTAR:      case OP_POSSTAR:
2940        case OP_POSSTARI:
2941      possessive = TRUE;      possessive = TRUE;
2942      min = 0;      min = 0;
2943      max = INT_MAX;      max = INT_MAX;
# Line 2012  for (;;) Line 2945  for (;;)
2945      goto REPEATCHAR;      goto REPEATCHAR;
2946    
2947      case OP_POSPLUS:      case OP_POSPLUS:
2948        case OP_POSPLUSI:
2949      possessive = TRUE;      possessive = TRUE;
2950      min = 1;      min = 1;
2951      max = INT_MAX;      max = INT_MAX;
# Line 2019  for (;;) Line 2953  for (;;)
2953      goto REPEATCHAR;      goto REPEATCHAR;
2954    
2955      case OP_POSQUERY:      case OP_POSQUERY:
2956        case OP_POSQUERYI:
2957      possessive = TRUE;      possessive = TRUE;
2958      min = 0;      min = 0;
2959      max = 1;      max = 1;
# Line 2026  for (;;) Line 2961  for (;;)
2961      goto REPEATCHAR;      goto REPEATCHAR;
2962    
2963      case OP_STAR:      case OP_STAR:
2964        case OP_STARI:
2965      case OP_MINSTAR:      case OP_MINSTAR:
2966        case OP_MINSTARI:
2967      case OP_PLUS:      case OP_PLUS:
2968        case OP_PLUSI:
2969      case OP_MINPLUS:      case OP_MINPLUS:
2970        case OP_MINPLUSI:
2971      case OP_QUERY:      case OP_QUERY:
2972        case OP_QUERYI:
2973      case OP_MINQUERY:      case OP_MINQUERY:
2974      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
2975        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
2976      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
2977      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
2978      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
2979      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
2980    
2981      /* 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. */  
2982    
2983      REPEATCHAR:      REPEATCHAR:
2984  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2048  for (;;) Line 2987  for (;;)
2987        length = 1;        length = 1;
2988        charptr = ecode;        charptr = ecode;
2989        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
2990        ecode += length;        ecode += length;
2991    
2992        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2058  for (;;) Line 2996  for (;;)
2996          {          {
2997  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2998          unsigned int othercase;          unsigned int othercase;
2999          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3000              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3001            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3002          else oclength = 0;          else oclength = 0;
3003  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3004    
3005          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3006            {            {
3007            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3008                memcmp(eptr, charptr, length) == 0) eptr += length;
3009  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3010            /* Need braces because of following else */            else if (oclength > 0 &&
3011            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3012                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3013    #endif  /* SUPPORT_UCP */
3014            else            else
3015              {              {
3016              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3017              eptr += oclength;              MRRETURN(MATCH_NOMATCH);
3018              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3019            }            }
3020    
3021          if (min == max) continue;          if (min == max) continue;
# Line 2086  for (;;) Line 3024  for (;;)
3024            {            {
3025            for (fi = min;; fi++)            for (fi = min;; fi++)
3026              {              {
3027              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3028              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3029              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3030              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3031                  memcmp(eptr, charptr, length) == 0) eptr += length;
3032  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3033              /* Need braces because of following else */              else if (oclength > 0 &&
3034              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3035                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3036    #endif  /* SUPPORT_UCP */
3037              else              else
3038                {                {
3039                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3040                eptr += oclength;                MRRETURN(MATCH_NOMATCH);
3041                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3042              }              }
3043            /* Control never gets here */            /* Control never gets here */
3044            }            }
# Line 2110  for (;;) Line 3048  for (;;)
3048            pp = eptr;            pp = eptr;
3049            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3050              {              {
3051              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3052              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, length) == 0) eptr += length;
3053  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3054              else if (oclength == 0) break;              else if (oclength > 0 &&
3055                         eptr <= md->end_subject - oclength &&
3056                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3057    #endif  /* SUPPORT_UCP */
3058              else              else
3059                {                {
3060                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3061                eptr += oclength;                break;
3062                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3063              }              }
3064    
3065            if (possessive) continue;            if (possessive) continue;
3066    
3067            for(;;)            for(;;)
3068             {              {
3069             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3070             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3071             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3072  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3073             eptr--;              eptr--;
3074             BACKCHAR(eptr);              BACKCHAR(eptr);
3075  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3076             eptr -= length;              eptr -= length;
3077  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3078             }              }
3079            }            }
3080          /* Control never gets here */          /* Control never gets here */
3081          }          }
# Line 2149  for (;;) Line 3088  for (;;)
3088  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3089    
3090      /* When not in UTF-8 mode, load a single-byte character. */      /* When not in UTF-8 mode, load a single-byte character. */
3091        {  
3092        if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);      fc = *ecode++;
       fc = *ecode++;  
       }  
3093    
3094      /* 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
3095      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 2166  for (;;) Line 3103  for (;;)
3103      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3104        max, eptr));        max, eptr));
3105    
3106      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3107        {        {
3108        fc = md->lcc[fc];        fc = md->lcc[fc];
3109        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3110          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3111            if (eptr >= md->end_subject)
3112              {
3113              SCHECK_PARTIAL();
3114              MRRETURN(MATCH_NOMATCH);
3115              }
3116            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3117            }
3118        if (min == max) continue;        if (min == max) continue;
3119        if (minimize)        if (minimize)
3120          {          {
3121          for (fi = min;; fi++)          for (fi = min;; fi++)
3122            {            {
3123            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3124            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3125            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3126                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3127              RRETURN(MATCH_NOMATCH);              {
3128                SCHECK_PARTIAL();
3129                MRRETURN(MATCH_NOMATCH);
3130                }
3131              if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3132            }            }
3133          /* Control never gets here */          /* Control never gets here */
3134          }          }
# Line 2189  for (;;) Line 3137  for (;;)
3137          pp = eptr;          pp = eptr;
3138          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3139            {            {
3140            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3141                {
3142                SCHECK_PARTIAL();
3143                break;
3144                }
3145              if (fc != md->lcc[*eptr]) break;
3146            eptr++;            eptr++;
3147            }            }
3148    
3149          if (possessive) continue;          if (possessive) continue;
3150    
3151          while (eptr >= pp)          while (eptr >= pp)
3152            {            {
3153            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3154            eptr--;            eptr--;
3155            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3156            }            }
3157          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3158          }          }
3159        /* Control never gets here */        /* Control never gets here */
3160        }        }
# Line 2208  for (;;) Line 3163  for (;;)
3163    
3164      else      else
3165        {        {
3166        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3167            {
3168            if (eptr >= md->end_subject)
3169              {
3170              SCHECK_PARTIAL();
3171              MRRETURN(MATCH_NOMATCH);
3172              }
3173            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3174            }
3175    
3176        if (min == max) continue;        if (min == max) continue;
3177    
3178        if (minimize)        if (minimize)
3179          {          {
3180          for (fi = min;; fi++)          for (fi = min;; fi++)
3181            {            {
3182            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3183            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3184            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3185              RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
3186                {
3187                SCHECK_PARTIAL();
3188                MRRETURN(MATCH_NOMATCH);
3189                }
3190              if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3191            }            }
3192          /* Control never gets here */          /* Control never gets here */
3193          }          }
# Line 2226  for (;;) Line 3196  for (;;)
3196          pp = eptr;          pp = eptr;
3197          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3198            {            {
3199            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3200                {
3201                SCHECK_PARTIAL();
3202                break;
3203                }
3204              if (fc != *eptr) break;
3205            eptr++;            eptr++;
3206            }            }
3207          if (possessive) continue;          if (possessive) continue;
3208    
3209          while (eptr >= pp)          while (eptr >= pp)
3210            {            {
3211            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3212            eptr--;            eptr--;
3213            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3214            }            }
3215          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3216          }          }
3217        }        }
3218      /* Control never gets here */      /* Control never gets here */
# Line 2245  for (;;) Line 3221  for (;;)
3221      checking can be multibyte. */      checking can be multibyte. */
3222    
3223      case OP_NOT:      case OP_NOT:
3224      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3225        if (eptr >= md->end_subject)
3226          {
3227          SCHECK_PARTIAL();
3228          MRRETURN(MATCH_NOMATCH);
3229          }
3230      ecode++;      ecode++;
3231      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3232      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3233        {        {
3234  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3235        if (c < 256)        if (c < 256)
3236  #endif  #endif
3237        c = md->lcc[c];        c = md->lcc[c];
3238        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3239        }        }
3240      else      else    /* Caseful */
3241        {        {
3242        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3243        }        }
3244      break;      break;
3245    
# Line 2270  for (;;) Line 3251  for (;;)
3251      about... */      about... */
3252    
3253      case OP_NOTEXACT:      case OP_NOTEXACT:
3254        case OP_NOTEXACTI:
3255      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3256      ecode += 3;      ecode += 3;
3257      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3258    
3259      case OP_NOTUPTO:      case OP_NOTUPTO:
3260        case OP_NOTUPTOI:
3261      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3262        case OP_NOTMINUPTOI:
3263      min = 0;      min = 0;
3264      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3265      minimize = *ecode == OP_NOTMINUPTO;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3266      ecode += 3;      ecode += 3;
3267      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3268    
3269      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
3270        case OP_NOTPOSSTARI:
3271      possessive = TRUE;      possessive = TRUE;
3272      min = 0;      min = 0;
3273      max = INT_MAX;      max = INT_MAX;
# Line 2290  for (;;) Line 3275  for (;;)
3275      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3276    
3277      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
3278        case OP_NOTPOSPLUSI:
3279      possessive = TRUE;      possessive = TRUE;
3280      min = 1;      min = 1;
3281      max = INT_MAX;      max = INT_MAX;
# Line 2297  for (;;) Line 3283  for (;;)
3283      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3284    
3285      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
3286        case OP_NOTPOSQUERYI:
3287      possessive = TRUE;      possessive = TRUE;
3288      min = 0;      min = 0;
3289      max = 1;      max = 1;
# Line 2304  for (;;) Line 3291  for (;;)
3291      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3292    
3293      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
3294        case OP_NOTPOSUPTOI:
3295      possessive = TRUE;      possessive = TRUE;
3296      min = 0;      min = 0;
3297      max = GET2(ecode, 1);      max = GET2(ecode, 1);
# Line 2311  for (;;) Line 3299  for (;;)
3299      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3300    
3301      case OP_NOTSTAR:      case OP_NOTSTAR:
3302        case OP_NOTSTARI:
3303      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3304        case OP_NOTMINSTARI:
3305      case OP_NOTPLUS:      case OP_NOTPLUS:
3306        case OP_NOTPLUSI:
3307      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3308        case OP_NOTMINPLUSI:
3309      case OP_NOTQUERY:      case OP_NOTQUERY:
3310        case OP_NOTQUERYI:
3311      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3312      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3313        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3314      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3315      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3316      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3317      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3318    
3319      /* 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. */  
3320    
3321      REPEATNOTCHAR:      REPEATNOTCHAR:
     if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3322      fc = *ecode++;      fc = *ecode++;
3323    
3324      /* The code is duplicated for the caseless and caseful cases, for speed,      /* The code is duplicated for the caseless and caseful cases, for speed,
# Line 2341  for (;;) Line 3332  for (;;)
3332      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3333        max, eptr));        max, eptr));
3334    
3335      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3336        {        {
3337        fc = md->lcc[fc];        fc = md->lcc[fc];
3338    
# Line 2352  for (;;) Line 3343  for (;;)
3343          register unsigned int d;          register unsigned int d;
3344          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3345            {            {
3346              if (eptr >= md->end_subject)
3347                {
3348                SCHECK_PARTIAL();
3349                MRRETURN(MATCH_NOMATCH);
3350                }
3351            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3352            if (d < 256) d = md->lcc[d];            if (d < 256) d = md->lcc[d];
3353            if (fc == d) RRETURN(MATCH_NOMATCH);            if (fc == d) MRRETURN(MATCH_NOMATCH);
3354            }            }
3355          }          }
3356        else        else
# Line 2363  for (;;) Line 3359  for (;;)
3359        /* Not UTF-8 mode */        /* Not UTF-8 mode */
3360          {          {
3361          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3362            if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            {
3363              if (eptr >= md->end_subject)
3364                {
3365                SCHECK_PARTIAL();
3366                MRRETURN(MATCH_NOMATCH);
3367                }
3368              if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3369              }
3370          }          }
3371    
3372        if (min == max) continue;        if (min == max) continue;
# Line 2377  for (;;) Line 3380  for (;;)
3380            register unsigned int d;            register unsigned int d;
3381            for (fi = min;; fi++)            for (fi = min;; fi++)
3382              {              {
3383              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3384              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3385                if (fi >= max) MRRETURN(MATCH_NOMATCH);
3386                if (eptr >= md->end_subject)
3387                  {
3388                  SCHECK_PARTIAL();
3389                  MRRETURN(MATCH_NOMATCH);
3390                  }
3391              GETCHARINC(d, eptr);              GETCHARINC(d, eptr);
3392              if (d < 256) d = md->lcc[d];              if (d < 256) d = md->lcc[d];
3393              if (fi >= max || eptr >= md->end_subject || fc == d)              if (fc == d) MRRETURN(MATCH_NOMATCH);
               RRETURN(MATCH_NOMATCH);  
3394              }              }
3395            }            }
3396          else          else
# Line 2391  for (;;) Line 3399  for (;;)
3399            {            {
3400            for (fi = min;; fi++)            for (fi = min;; fi++)
3401              {              {
3402              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3403              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3404              if (fi >= max || eptr >= md->end_subject || fc == md->lcc[*eptr++])              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3405                RRETURN(MATCH_NOMATCH);              if (eptr >= md->end_subject)
3406                  {
3407                  SCHECK_PARTIAL();
3408                  MRRETURN(MATCH_NOMATCH);
3409                  }
3410                if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3411              }              }
3412            }            }
3413          /* Control never gets here */          /* Control never gets here */
# Line 2414  for (;;) Line 3427  for (;;)
3427            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3428              {              {
3429              int len = 1;              int len = 1;
3430              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3431                  {
3432                  SCHECK_PARTIAL();
3433                  break;
3434                  }
3435              GETCHARLEN(d, eptr, len);              GETCHARLEN(d, eptr, len);
3436              if (d < 256) d = md->lcc[d];              if (d < 256) d = md->lcc[d];
3437              if (fc == d) break;              if (fc == d) break;
# Line 2423  for (;;) Line 3440  for (;;)
3440          if (possessive) continue;          if (possessive) continue;
3441          for(;;)          for(;;)
3442              {              {
3443              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3444              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3445              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3446              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2435  for (;;) Line 3452  for (;;)
3452            {            {
3453            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3454              {              {
3455              if (eptr >= md->end_subject || fc == md->lcc[*eptr]) break;              if (eptr >= md->end_subject)
3456                  {
3457                  SCHECK_PARTIAL();
3458                  break;
3459                  }
3460                if (fc == md->lcc[*eptr]) break;
3461              eptr++;              eptr++;
3462              }              }
3463            if (possessive) continue;            if (possessive) continue;
3464            while (eptr >= pp)            while (eptr >= pp)
3465              {              {
3466              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3467              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3468              eptr--;              eptr--;
3469              }              }
3470            }            }
3471    
3472          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3473          }          }
3474        /* Control never gets here */        /* Control never gets here */
3475        }        }
# Line 2463  for (;;) Line 3485  for (;;)
3485          register unsigned int d;          register unsigned int d;
3486          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3487            {            {
3488              if (eptr >= md->end_subject)
3489                {
3490                SCHECK_PARTIAL();
3491                MRRETURN(MATCH_NOMATCH);
3492                }
3493            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3494            if (fc == d) RRETURN(MATCH_NOMATCH);            if (fc == d) MRRETURN(MATCH_NOMATCH);
3495            }            }
3496          }          }
3497        else        else
# Line 2472  for (;;) Line 3499  for (;;)
3499        /* Not UTF-8 mode */        /* Not UTF-8 mode */
3500          {          {
3501          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3502            if (fc == *eptr++) RRETURN(MATCH_NOMATCH);            {
3503              if (eptr >= md->end_subject)
3504                {
3505                SCHECK_PARTIAL();
3506                MRRETURN(MATCH_NOMATCH);
3507                }
3508              if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
3509              }
3510          }          }
3511    
3512        if (min == max) continue;        if (min == max) continue;
# Line 2486  for (;;) Line 3520  for (;;)
3520            register unsigned int d;            register unsigned int d;
3521            for (fi = min;; fi++)            for (fi = min;; fi++)
3522              {              {
3523              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3524              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3525                if (fi >= max) MRRETURN(MATCH_NOMATCH);
3526                if (eptr >= md->end_subject)
3527                  {
3528                  SCHECK_PARTIAL();
3529                  MRRETURN(MATCH_NOMATCH);
3530                  }
3531              GETCHARINC(d, eptr);              GETCHARINC(d, eptr);
3532              if (fi >= max || eptr >= md->end_subject || fc == d)              if (fc == d) MRRETURN(MATCH_NOMATCH);
               RRETURN(MATCH_NOMATCH);  
3533              }              }
3534            }            }
3535          else          else
# Line 2499  for (;;) Line 3538  for (;;)
3538            {            {
3539            for (fi = min;; fi++)            for (fi = min;; fi++)
3540              {              {
3541              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
3542              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3543              if (fi >= max || eptr >= md->end_subject || fc == *eptr++)              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3544                RRETURN(MATCH_NOMATCH);              if (eptr >= md->end_subject)
3545                  {
3546                  SCHECK_PARTIAL();
3547                  MRRETURN(MATCH_NOMATCH);
3548                  }
3549                if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
3550              }              }
3551            }            }
3552          /* Control never gets here */          /* Control never gets here */
# Line 2522  for (;;) Line 3566  for (;;)
3566