/[pcre]/code/trunk/pcre_exec.c
ViewVC logotype

Diff of /code/trunk/pcre_exec.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 91 by nigel, Sat Feb 24 21:41:34 2007 UTC revision 618 by ph10, Sat Jul 16 17:24:16 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
43  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
44    
45  #define NLBLOCK md           /* The block containing newline information */  #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK md             /* Block containing newline information */
50    #define PSSTART start_subject  /* Field containing processed string start */
51    #define PSEND   end_subject    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    /* Undefine some potentially clashing cpp symbols */
56    
57  /* Structure for building a chain of data that actually lives on the  #undef min
58  stack, for holding the values of the subject pointer at the start of each  #undef max
 subpattern, so as to detect when an empty string has been matched by a  
 subpattern - to break infinite loops. When NO_RECURSE is set, these blocks  
 are on the heap, not on the stack. */  
   
 typedef struct eptrblock {  
   struct eptrblock *epb_prev;  
   USPTR epb_saved_eptr;  
 } eptrblock;  
59    
60  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
61    of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
63    
64  #define match_condassert   0x01    /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_isgroup      0x02    /* Set if start of bracketed group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 68  defined PCRE_ERROR_xxx codes, which are Line 70  defined PCRE_ERROR_xxx codes, which are
70  #define MATCH_MATCH        1  #define MATCH_MATCH        1
71  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
72    
73    /* Special internal returns from the match() function. Make them sufficiently
74    negative to avoid the external error codes. */
75    
76    #define MATCH_ACCEPT       (-999)
77    #define MATCH_COMMIT       (-998)
78    #define MATCH_KETRPOS      (-997)
79    #define MATCH_ONCE         (-996)
80    #define MATCH_PRUNE        (-995)
81    #define MATCH_SKIP         (-994)
82    #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85    /* This is a convenience macro for code that occurs many times. */
86    
87    #define MRRETURN(ra) \
88      { \
89      md->mark = markptr; \
90      RRETURN(ra); \
91      }
92    
93  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
94  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
95  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 81  static const char rep_max[] = { 0, 0, 0, Line 103  static const char rep_max[] = { 0, 0, 0,
103    
104    
105    
106  #ifdef DEBUG  #ifdef PCRE_DEBUG
107  /*************************************************  /*************************************************
108  *        Debugging function to print chars       *  *        Debugging function to print chars       *
109  *************************************************/  *************************************************/
# Line 101  Returns:     nothing Line 123  Returns:     nothing
123  static void  static void
124  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)
125  {  {
126  int c;  unsigned int c;
127  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
128  while (length-- > 0)  while (length-- > 0)
129    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);
# Line 114  while (length-- > 0) Line 136  while (length-- > 0)
136  *          Match a back-reference                *  *          Match a back-reference                *
137  *************************************************/  *************************************************/
138    
139  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
140  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
141    the length passed is zero. Note that in caseless UTF-8 mode, the number of
142    subject bytes matched may be different to the number of reference bytes.
143    
144  Arguments:  Arguments:
145    offset      index into the offset vector    offset      index into the offset vector
146    eptr        points into the subject    eptr        pointer into the subject
147    length      length to be matched    length      length of reference to be matched (number of bytes)
148    md          points to match data block    md          points to match data block
149    ims         the ims flags    caseless    TRUE if caseless
150    
151  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
152  */  */
153    
154  static BOOL  static int
155  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register USPTR eptr, int length, match_data *md,
156    unsigned long int ims)    BOOL caseless)
157  {  {
158  USPTR p = md->start_subject + md->offset_vector[offset];  USPTR eptr_start = eptr;
159    register USPTR p = md->start_subject + md->offset_vector[offset];
160    
161  #ifdef DEBUG  #ifdef PCRE_DEBUG
162  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
163    printf("matching subject <null>");    printf("matching subject <null>");
164  else  else
# Line 146  pchars(p, length, FALSE, md); Line 171  pchars(p, length, FALSE, md);
171  printf("\n");  printf("\n");
172  #endif  #endif
173    
174  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
175    
176  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
177    
178  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
179    properly if Unicode properties are supported. Otherwise, we can check only
180    ASCII characters. */
181    
182  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
183    {    {
184    while (length-- > 0)  #ifdef SUPPORT_UTF8
185      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
186      if (md->utf8)
187        {
188        /* Match characters up to the end of the reference. NOTE: the number of
189        bytes matched may differ, because there are some characters whose upper and
190        lower case versions code as different numbers of bytes. For example, U+023A
191        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
192        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
193        the latter. It is important, therefore, to check the length along the
194        reference, not along the subject (earlier code did this wrong). */
195    
196        USPTR endptr = p + length;
197        while (p < endptr)
198          {
199          int c, d;
200          if (eptr >= md->end_subject) return -1;
201          GETCHARINC(c, eptr);
202          GETCHARINC(d, p);
203          if (c != d && c != UCD_OTHERCASE(d)) return -1;
204          }
205        }
206      else
207    #endif
208    #endif
209    
210      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
211      is no UCP support. */
212        {
213        if (eptr + length > md->end_subject) return -1;
214        while (length-- > 0)
215          { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }
216        }
217    }    }
218    
219    /* In the caseful case, we can just compare the bytes, whether or not we
220    are in UTF-8 mode. */
221    
222  else  else
223    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
224      if (eptr + length > md->end_subject) return -1;
225      while (length-- > 0) if (*p++ != *eptr++) return -1;
226      }
227    
228  return TRUE;  return eptr - eptr_start;
229  }  }
230    
231    
# Line 186  calls by keeping local variables that ne Line 251  calls by keeping local variables that ne
251  obtained from malloc() instead instead of on the stack. Macros are used to  obtained from malloc() instead instead of on the stack. Macros are used to
252  achieve this so that the actual code doesn't look very different to what it  achieve this so that the actual code doesn't look very different to what it
253  always used to.  always used to.
254    
255    The original heap-recursive code used longjmp(). However, it seems that this
256    can be very slow on some operating systems. Following a suggestion from Stan
257    Switzer, the use of longjmp() has been abolished, at the cost of having to
258    provide a unique number for each call to RMATCH. There is no way of generating
259    a sequence of numbers at compile time in C. I have given them names, to make
260    them stand out more clearly.
261    
262    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
263    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
264    tests. Furthermore, not using longjmp() means that local dynamic variables
265    don't have indeterminate values; this has meant that the frame size can be
266    reduced because the result can be "passed back" by straight setting of the
267    variable instead of being passed in the frame.
268  ****************************************************************************  ****************************************************************************
269  ***************************************************************************/  ***************************************************************************/
270    
271    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
272    below must be updated in sync.  */
273    
274    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
275           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
276           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
277           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
278           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
279           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
280           RM61,  RM62, RM63, RM64, RM65, RM66 };
281    
282  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
283  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
284    actually used in this definition. */
285    
286  #ifndef NO_RECURSE  #ifndef NO_RECURSE
287  #define REGISTER register  #define REGISTER register
288  #ifdef DEBUG  
289  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
290    #define RMATCH(ra,rb,rc,rd,re,rw) \
291    { \    { \
292    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
293    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1); \
294    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
295    }    }
296  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 208  versions and production versions. */ Line 299  versions and production versions. */
299    return ra; \    return ra; \
300    }    }
301  #else  #else
302  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
303    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1)
304  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
305  #endif  #endif
306    
307  #else  #else
308    
309    
310  /* These versions of the macros manage a private stack on the heap. Note  /* These versions of the macros manage a private stack on the heap. Note that
311  that the rd argument of RMATCH isn't actually used. It's the md argument of  the "rd" argument of RMATCH isn't actually used in this definition. It's the md
312  match(), which never changes. */  argument of match(), which never changes. */
313    
314  #define REGISTER  #define REGISTER
315    
316  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
317    {\    {\
318    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\
319    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
320      {\    frame->Xwhere = rw; \
321      newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
322      newframe->Xecode = rb;\    newframe->Xecode = rb;\
323      newframe->Xoffset_top = rc;\    newframe->Xmstart = mstart;\
324      newframe->Xims = re;\    newframe->Xmarkptr = markptr;\
325      newframe->Xeptrb = rf;\    newframe->Xoffset_top = rc;\
326      newframe->Xflags = rg;\    newframe->Xeptrb = re;\
327      newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
328      newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
329      frame = newframe;\    frame = newframe;\
330      DPRINTF(("restarting from line %d\n", __LINE__));\    DPRINTF(("restarting from line %d\n", __LINE__));\
331      goto HEAP_RECURSE;\    goto HEAP_RECURSE;\
332      }\    L_##rw:\
333    else\    DPRINTF(("jumped back to line %d\n", __LINE__));\
     {\  
     DPRINTF(("longjumped back to line %d\n", __LINE__));\  
     frame = md->thisframe;\  
     rx = frame->Xresult;\  
     }\  
334    }    }
335    
336  #define RRETURN(ra)\  #define RRETURN(ra)\
337    {\    {\
338    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
339    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
340    (pcre_stack_free)(newframe);\    (pcre_stack_free)(oldframe);\
341    if (frame != NULL)\    if (frame != NULL)\
342      {\      {\
343      frame->Xresult = ra;\      rrc = ra;\
344      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
345      }\      }\
346    return ra;\    return ra;\
347    }    }
# Line 269  typedef struct heapframe { Line 354  typedef struct heapframe {
354    
355    /* Function arguments that may change */    /* Function arguments that may change */
356    
357    const uschar *Xeptr;    USPTR Xeptr;
358    const uschar *Xecode;    const uschar *Xecode;
359      USPTR Xmstart;
360      USPTR Xmarkptr;
361    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
362    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
363    unsigned int Xrdepth;    unsigned int Xrdepth;
364    
365    /* Function local variables */    /* Function local variables */
366    
367    const uschar *Xcallpat;    USPTR Xcallpat;
368    const uschar *Xcharptr;  #ifdef SUPPORT_UTF8
369    const uschar *Xdata;    USPTR Xcharptr;
370    const uschar *Xnext;  #endif
371    const uschar *Xpp;    USPTR Xdata;
372    const uschar *Xprev;    USPTR Xnext;
373    const uschar *Xsaved_eptr;    USPTR Xpp;
374      USPTR Xprev;
375      USPTR Xsaved_eptr;
376    
377    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
378    
379    BOOL Xcur_is_word;    BOOL Xcur_is_word;
380    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
381    BOOL Xprev_is_word;    BOOL Xprev_is_word;
382    
   unsigned long int Xoriginal_ims;  
   
383  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
384    int Xprop_type;    int Xprop_type;
385    int Xprop_value;    int Xprop_value;
# Line 303  typedef struct heapframe { Line 387  typedef struct heapframe {
387    int Xprop_category;    int Xprop_category;
388    int Xprop_chartype;    int Xprop_chartype;
389    int Xprop_script;    int Xprop_script;
390    int *Xprop_test_variable;    int Xoclength;
391      uschar Xocchars[8];
392  #endif  #endif
393    
394      int Xcodelink;
395    int Xctype;    int Xctype;
396    int Xfc;    unsigned int Xfc;
397    int Xfi;    int Xfi;
398    int Xlength;    int Xlength;
399    int Xmax;    int Xmax;
# Line 321  typedef struct heapframe { Line 407  typedef struct heapframe {
407    
408    eptrblock Xnewptrb;    eptrblock Xnewptrb;
409    
410    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
411    
412    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
413    
414  } heapframe;  } heapframe;
415    
# Line 340  typedef struct heapframe { Line 425  typedef struct heapframe {
425  *         Match from current position            *  *         Match from current position            *
426  *************************************************/  *************************************************/
427    
428  /* On entry ecode points to the first opcode, and eptr to the first character  /* This function is called recursively in many circumstances. Whenever it
 in the subject string, while eptrb holds the value of eptr at the start of the  
 last bracketed group - used for breaking infinite loops matching zero-length  
 strings. This function is called recursively in many circumstances. Whenever it  
429  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
430  same response.  same response. */
431    
432    /* These macros pack up tests that are used for partial matching, and which
433    appears several times in the code. We set the "hit end" flag if the pointer is
434    at the end of the subject and also past the start of the subject (i.e.
435    something has been matched). For hard partial matching, we then return
436    immediately. The second one is used when we already know we are past the end of
437    the subject. */
438    
439    #define CHECK_PARTIAL()\
440      if (md->partial != 0 && eptr >= md->end_subject && \
441          eptr > md->start_used_ptr) \
442        { \
443        md->hitend = TRUE; \
444        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
445        }
446    
447    #define SCHECK_PARTIAL()\
448      if (md->partial != 0 && eptr > md->start_used_ptr) \
449        { \
450        md->hitend = TRUE; \
451        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
452        }
453    
454    
455  Performance note: It might be tempting to extract commonly used fields from the  /* Performance note: It might be tempting to extract commonly used fields from
456  md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf8, end_subject) into individual variables to improve
457  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
458  made performance worse.  made performance worse.
459    
460  Arguments:  Arguments:
461     eptr        pointer in subject     eptr        pointer to current character in subject
462     ecode       position in code     ecode       pointer to current position in compiled code
463       mstart      pointer to the current match start position (can be modified
464                     by encountering \K)
465       markptr     pointer to the most recent MARK name, or NULL
466     offset_top  current top pointer     offset_top  current top pointer
467     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
468     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
469                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_isgroup - this is the start of a bracketed group  
470     rdepth      the recursion depth     rdepth      the recursion depth
471    
472  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
473                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
474                   a negative MATCH_xxx value for PRUNE, SKIP, etc
475                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
476                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
477  */  */
478    
479  static int  static int
480  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
481    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
482    int flags, unsigned int rdepth)    unsigned int rdepth)
483  {  {
484  /* 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,
485  so they can be ordinary variables in all cases. Mark them with "register"  so they can be ordinary variables in all cases. Mark some of them with
486  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
487    
488  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
489  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
490  register unsigned int  c;  /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
491  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
492    
493    BOOL minimize, possessive; /* Quantifier options */
494    BOOL caseless;
495    int condcode;
496    
497  /* 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
498  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
499  heap storage. Set up the top-level frame here; others are obtained from the  heap storage. Set up the top-level frame here; others are obtained from the
500  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
501    
502  #ifdef NO_RECURSE  #ifdef NO_RECURSE
503  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));
504    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
505  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
506    
507  /* Copy in the original argument variables */  /* Copy in the original argument variables */
508    
509  frame->Xeptr = eptr;  frame->Xeptr = eptr;
510  frame->Xecode = ecode;  frame->Xecode = ecode;
511    frame->Xmstart = mstart;
512    frame->Xmarkptr = markptr;
513  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
514  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
515  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
516    
517  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 412  HEAP_RECURSE: Line 522  HEAP_RECURSE:
522    
523  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
524  #define ecode              frame->Xecode  #define ecode              frame->Xecode
525    #define mstart             frame->Xmstart
526    #define markptr            frame->Xmarkptr
527  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
528  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
529  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
530    
531  /* Ditto for the local variables */  /* Ditto for the local variables */
# Line 424  HEAP_RECURSE: Line 534  HEAP_RECURSE:
534  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
535  #endif  #endif
536  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
537    #define codelink           frame->Xcodelink
538  #define data               frame->Xdata  #define data               frame->Xdata
539  #define next               frame->Xnext  #define next               frame->Xnext
540  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 434  HEAP_RECURSE: Line 545  HEAP_RECURSE:
545    
546  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
547  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
548  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
549    
 #define original_ims       frame->Xoriginal_ims  
   
550  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
551  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
552  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
# Line 446  HEAP_RECURSE: Line 554  HEAP_RECURSE:
554  #define prop_category      frame->Xprop_category  #define prop_category      frame->Xprop_category
555  #define prop_chartype      frame->Xprop_chartype  #define prop_chartype      frame->Xprop_chartype
556  #define prop_script        frame->Xprop_script  #define prop_script        frame->Xprop_script
557  #define prop_test_variable frame->Xprop_test_variable  #define oclength           frame->Xoclength
558    #define occhars            frame->Xocchars
559  #endif  #endif
560    
561  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 470  HEAP_RECURSE: Line 579  HEAP_RECURSE:
579  get preserved during recursion in the normal way. In this environment, fi and  get preserved during recursion in the normal way. In this environment, fi and
580  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
581    
582  #else  #else         /* NO_RECURSE not defined */
583  #define fi i  #define fi i
584  #define fc c  #define fc c
585    
586    /* Many of the following variables are used only in small blocks of the code.
587  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  My normal style of coding would have declared them within each of those blocks.
588  const uschar *charptr;             /* in small blocks of the code. My normal */  However, in order to accommodate the version of this code that uses an external
589  #endif                             /* style of coding would have declared    */  "stack" implemented on the heap, it is easier to declare them all here, so the
590  const uschar *callpat;             /* them within each of those blocks.      */  declarations can be cut out in a block. The only declarations within blocks
591  const uschar *data;                /* However, in order to accommodate the   */  below are for variables that do not have to be preserved over a recursive call
592  const uschar *next;                /* version of this code that uses an      */  to RMATCH(). */
593  USPTR         pp;                  /* external "stack" implemented on the    */  
594  const uschar *prev;                /* heap, it is easier to declare them all */  #ifdef SUPPORT_UTF8
595  USPTR         saved_eptr;          /* here, so the declarations can be cut   */  const uschar *charptr;
596                                     /* out in a block. The only declarations  */  #endif
597  recursion_info new_recursive;      /* within blocks below are for variables  */  const uschar *callpat;
598                                     /* that do not have to be preserved over  */  const uschar *data;
599  BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  const uschar *next;
600    USPTR         pp;
601    const uschar *prev;
602    USPTR         saved_eptr;
603    
604    recursion_info new_recursive;
605    
606    BOOL cur_is_word;
607  BOOL condition;  BOOL condition;
 BOOL minimize;  
608  BOOL prev_is_word;  BOOL prev_is_word;
609    
 unsigned long int original_ims;  
   
610  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
611  int prop_type;  int prop_type;
612  int prop_value;  int prop_value;
# Line 501  int prop_fail_result; Line 614  int prop_fail_result;
614  int prop_category;  int prop_category;
615  int prop_chartype;  int prop_chartype;
616  int prop_script;  int prop_script;
617  int *prop_test_variable;  int oclength;
618    uschar occhars[8];
619  #endif  #endif
620    
621    int codelink;
622  int ctype;  int ctype;
623  int length;  int length;
624  int max;  int max;
# Line 516  int save_offset1, save_offset2, save_off Line 631  int save_offset1, save_offset2, save_off
631  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
632    
633  eptrblock newptrb;  eptrblock newptrb;
634  #endif  #endif     /* NO_RECURSE */
635    
636    /* To save space on the stack and in the heap frame, I have doubled up on some
637    of the local variables that are used only in localised parts of the code, but
638    still need to be preserved over recursive calls of match(). These macros define
639    the alternative names that are used. */
640    
641    #define allow_zero    cur_is_word
642    #define cbegroup      condition
643    #define code_offset   codelink
644    #define condassert    condition
645    #define matched_once  prev_is_word
646    
647  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
648  variables. */  variables. */
# Line 524  variables. */ Line 650  variables. */
650  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
651  prop_value = 0;  prop_value = 0;
652  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_variable = NULL;  
653  #endif  #endif
654    
655    
656  /* This label is used for tail recursion, which is used in a few cases even  /* This label is used for tail recursion, which is used in a few cases even
657  when NO_RECURSE is not defined, in order to reduce the amount of stack that is  when NO_RECURSE is not defined, in order to reduce the amount of stack that is
658  used. Thanks to Ian Taylor for noticing this possibility and sending the  used. Thanks to Ian Taylor for noticing this possibility and sending the
# Line 537  TAIL_RECURSE: Line 663  TAIL_RECURSE:
663  /* 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
664  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
665  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()
666  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
667  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
668  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,
669  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
670    
 /* 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 ')' */  
   
671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
672  utf8 = md->utf8;       /* Local copy of the flag */  utf8 = md->utf8;       /* Local copy of the flag */
673  #else  #else
674  utf8 = FALSE;  utf8 = FALSE;
675  #endif  #endif
676    
677  /* At the start of a bracketed group, add the current subject pointer to the  /* First check that we haven't called match() too many times, or that we
678  stack of such pointers, to be re-instated at the end of the group when we hit  haven't exceeded the recursive call limit. */
679  the closing ket. When match() is called in other circumstances, we don't add to  
680  this stack. */  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
681    if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
682    
683    /* At the start of a group with an unlimited repeat that may match an empty
684    string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
685    done this way to save having to use another function argument, which would take
686    up space on the stack. See also MATCH_CONDASSERT below.
687    
688    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
689    such remembered pointers, to be checked when we hit the closing ket, in order
690    to break infinite loops that match no characters. When match() is called in
691    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
692    NOT be used with tail recursion, because the memory block that is used is on
693    the stack, so a new one may be required for each match(). */
694    
695  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
696    {    {
   newptrb.epb_prev = eptrb;  
697    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
698      newptrb.epb_prev = eptrb;
699    eptrb = &newptrb;    eptrb = &newptrb;
700      md->match_function_type = 0;
701    }    }
702    
703  /* Now start processing the operations. */  /* Now start processing the opcodes. */
704    
705  for (;;)  for (;;)
706    {    {
707      minimize = possessive = FALSE;
708    op = *ecode;    op = *ecode;
709    minimize = FALSE;  
710      switch(op)
   /* 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;  
   
   /* Opening capturing bracket. If there is space in the offset vector, save  
   the current subject position in the working slot at the top of the vector. We  
   mustn't change the current values of the data slot, because they may be set  
   from a previous iteration of this group, and be referred to by a reference  
   inside the group.  
   
   If the bracket fails to match, we need to restore this value and also the  
   values of the final offsets, in case they were set by a previous iteration of  
   the same bracket.  
   
   If there isn't enough space in the offset vector, treat this as if it were a  
   non-capturing bracket. Don't worry about setting the flag for the error case  
   here; that is handled in the code for KET. */  
   
   if (op > OP_BRA)  
711      {      {
712      number = op - OP_BRA;      case OP_MARK:
713        markptr = ecode + 2;
714      /* For extended extraction brackets (large number), we have to fish out the      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
715      number from a dummy opcode at the start. */        eptrb, RM55);
716    
717      if (number > EXTRACT_BASIC_MAX)      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
718        number = GET2(ecode, 2+LINK_SIZE);      argument, and we must check whether that argument matches this MARK's
719        argument. It is passed back in md->start_match_ptr (an overloading of that
720        variable). If it does match, we reset that variable to the current subject
721        position and return MATCH_SKIP. Otherwise, pass back the return code
722        unaltered. */
723    
724        if (rrc == MATCH_SKIP_ARG &&
725            strcmp((char *)markptr, (char *)(md->start_match_ptr)) == 0)
726          {
727          md->start_match_ptr = eptr;
728          RRETURN(MATCH_SKIP);
729          }
730    
731        if (md->mark == NULL) md->mark = markptr;
732        RRETURN(rrc);
733    
734        case OP_FAIL:
735        MRRETURN(MATCH_NOMATCH);
736    
737        /* COMMIT overrides PRUNE, SKIP, and THEN */
738    
739        case OP_COMMIT:
740        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
741          eptrb, RM52);
742        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
743            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
744            rrc != MATCH_THEN)
745          RRETURN(rrc);
746        MRRETURN(MATCH_COMMIT);
747    
748        /* PRUNE overrides THEN */
749    
750        case OP_PRUNE:
751        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
752          eptrb, RM51);
753        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
754        MRRETURN(MATCH_PRUNE);
755    
756        case OP_PRUNE_ARG:
757        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
758          eptrb, RM56);
759        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
760        md->mark = ecode + 2;
761        RRETURN(MATCH_PRUNE);
762    
763        /* SKIP overrides PRUNE and THEN */
764    
765        case OP_SKIP:
766        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
767          eptrb, RM53);
768        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
769          RRETURN(rrc);
770        md->start_match_ptr = eptr;   /* Pass back current position */
771        MRRETURN(MATCH_SKIP);
772    
773        case OP_SKIP_ARG:
774        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
775          eptrb, RM57);
776        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
777          RRETURN(rrc);
778    
779        /* Pass back the current skip name by overloading md->start_match_ptr and
780        returning the special MATCH_SKIP_ARG return code. This will either be
781        caught by a matching MARK, or get to the top, where it is treated the same
782        as PRUNE. */
783    
784        md->start_match_ptr = ecode + 2;
785        RRETURN(MATCH_SKIP_ARG);
786    
787        /* For THEN (and THEN_ARG) we pass back the address of the bracket or
788        the alt that is at the start of the current branch. This makes it possible
789        to skip back past alternatives that precede the THEN within the current
790        branch. */
791    
792        case OP_THEN:
793        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
794          eptrb, RM54);
795        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
796        md->start_match_ptr = ecode - GET(ecode, 1);
797        MRRETURN(MATCH_THEN);
798    
799        case OP_THEN_ARG:
800        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],
801          offset_top, md, eptrb, RM58);
802        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
803        md->start_match_ptr = ecode - GET(ecode, 1);
804        md->mark = ecode + LINK_SIZE + 2;
805        RRETURN(MATCH_THEN);
806    
807        /* Handle a capturing bracket, other than those that are possessive with an
808        unlimited repeat. If there is space in the offset vector, save the current
809        subject position in the working slot at the top of the vector. We mustn't
810        change the current values of the data slot, because they may be set from a
811        previous iteration of this group, and be referred to by a reference inside
812        the group. A failure to match might occur after the group has succeeded,
813        if something later on doesn't match. For this reason, we need to restore
814        the working value and also the values of the final offsets, in case they
815        were set by a previous iteration of the same bracket.
816    
817        If there isn't enough space in the offset vector, treat this as if it were
818        a non-capturing bracket. Don't worry about setting the flag for the error
819        case here; that is handled in the code for KET. */
820    
821        case OP_CBRA:
822        case OP_SCBRA:
823        number = GET2(ecode, 1+LINK_SIZE);
824      offset = number << 1;      offset = number << 1;
825    
826  #ifdef DEBUG  #ifdef PCRE_DEBUG
827      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
828        printf("subject=");
829      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
830      printf("\n");      printf("\n");
831  #endif  #endif
# Line 622  for (;;) Line 838  for (;;)
838        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
839    
840        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
841        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
842            (int)(eptr - md->start_subject);
843    
844        do        for (;;)
845          {          {
846          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
847            match_isgroup);          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
848          if (rrc != MATCH_NOMATCH) RRETURN(rrc);            eptrb, RM1);
849            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
850            if (rrc != MATCH_NOMATCH &&
851                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
852              RRETURN(rrc);
853          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
854          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
855            if (*ecode != OP_ALT) break;
856          }          }
       while (*ecode == OP_ALT);  
857    
858        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
859        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
860        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
861        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
862    
863          /* At this point, rrc will be one of MATCH_ONCE, MATCH_NOMATCH, or
864          MATCH_THEN. */
865    
866          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
867          RRETURN(((rrc == MATCH_ONCE)? MATCH_ONCE:MATCH_NOMATCH));
868          }
869    
870        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
871        as a non-capturing bracket. */
872    
873        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
874        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
875    
876        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
877    
878        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
879        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
880    
881        /* Non-capturing or atomic group, except for possessive with unlimited
882        repeat. Loop for all the alternatives. When we get to the final alternative
883        within the brackets, we used to return the result of a recursive call to
884        match() whatever happened so it was possible to reduce stack usage by
885        turning this into a tail recursion, except in the case of a possibly empty
886        group. However, now that there is the possiblity of (*THEN) occurring in
887        the final alternative, this optimization is no longer possible.
888    
889        MATCH_ONCE is returned when the end of an atomic group is successfully
890        reached, but subsequent matching fails. It passes back up the tree (causing
891        captured values to be reset) until the original atomic group level is
892        reached. This is tested by comparing md->once_target with the start of the
893        group. At this point, the return is converted into MATCH_NOMATCH so that
894        previous backup points can be taken. */
895    
896        RRETURN(MATCH_NOMATCH);      case OP_ONCE:
897        case OP_BRA:
898        case OP_SBRA:
899        DPRINTF(("start non-capturing bracket\n"));
900    
901        for (;;)
902          {
903          if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
904          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
905            RM2);
906          if (rrc != MATCH_NOMATCH &&
907              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
908            {
909            if (rrc == MATCH_ONCE)
910              {
911              const uschar *scode = ecode;
912              if (*scode != OP_ONCE)           /* If not at start, find it */
913                {
914                while (*scode == OP_ALT) scode += GET(scode, 1);
915                scode -= GET(scode, 1);
916                }
917              if (md->once_target == scode) rrc = MATCH_NOMATCH;
918              }
919            RRETURN(rrc);
920            }
921          ecode += GET(ecode, 1);
922          if (*ecode != OP_ALT) break;
923        }        }
924        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
925        RRETURN(MATCH_NOMATCH);
926    
927      /* Insufficient room for saving captured contents */      /* Handle possessive capturing brackets with an unlimited repeat. We come
928        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
929        handled similarly to the normal case above. However, the matching is
930        different. The end of these brackets will always be OP_KETRPOS, which
931        returns MATCH_KETRPOS without going further in the pattern. By this means
932        we can handle the group by iteration rather than recursion, thereby
933        reducing the amount of stack needed. */
934    
935        case OP_CBRAPOS:
936        case OP_SCBRAPOS:
937        allow_zero = FALSE;
938    
939        POSSESSIVE_CAPTURE:
940        number = GET2(ecode, 1+LINK_SIZE);
941        offset = number << 1;
942    
943      else op = OP_BRA;  #ifdef PCRE_DEBUG
944      }      printf("start possessive bracket %d\n", number);
945        printf("subject=");
946        pchars(eptr, 16, TRUE, md);
947        printf("\n");
948    #endif
949    
950    /* Other types of node can be handled by a switch */      if (offset < md->offset_max)
951          {
952          matched_once = FALSE;
953          code_offset = ecode - md->start_code;
954    
955    switch(op)        save_offset1 = md->offset_vector[offset];
956      {        save_offset2 = md->offset_vector[offset+1];
957      case OP_BRA:     /* Non-capturing bracket: optimized */        save_offset3 = md->offset_vector[md->offset_end - number];
958      DPRINTF(("start bracket 0\n"));        save_capture_last = md->capture_last;
959    
960          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
961    
962          /* Each time round the loop, save the current subject position for use
963          when the group matches. For MATCH_MATCH, the group has matched, so we
964          restart it with a new subject starting position, remembering that we had
965          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
966          usual. If we haven't matched any alternatives in any iteration, check to
967          see if a previous iteration matched. If so, the group has matched;
968          continue from afterwards. Otherwise it has failed; restore the previous
969          capture values before returning NOMATCH. */
970    
971          for (;;)
972            {
973            md->offset_vector[md->offset_end - number] =
974              (int)(eptr - md->start_subject);
975            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
976            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
977              eptrb, RM63);
978            if (rrc == MATCH_KETRPOS)
979              {
980              offset_top = md->end_offset_top;
981              eptr = md->end_match_ptr;
982              ecode = md->start_code + code_offset;
983              save_capture_last = md->capture_last;
984              matched_once = TRUE;
985              continue;
986              }
987            if (rrc != MATCH_NOMATCH &&
988                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
989              RRETURN(rrc);
990            md->capture_last = save_capture_last;
991            ecode += GET(ecode, 1);
992            if (*ecode != OP_ALT) break;
993            }
994    
995      /* Loop for all the alternatives */        if (!matched_once)
996            {
997            md->offset_vector[offset] = save_offset1;
998            md->offset_vector[offset+1] = save_offset2;
999            md->offset_vector[md->offset_end - number] = save_offset3;
1000            }
1001    
1002          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
1003          if (allow_zero || matched_once)
1004            {
1005            ecode += 1 + LINK_SIZE;
1006            break;
1007            }
1008    
1009          RRETURN(MATCH_NOMATCH);
1010          }
1011    
1012        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1013        as a non-capturing bracket. */
1014    
1015        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1016        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1017    
1018        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1019    
1020        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1021        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1022    
1023        /* Non-capturing possessive bracket with unlimited repeat. We come here
1024        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1025        without the capturing complication. It is written out separately for speed
1026        and cleanliness. */
1027    
1028        case OP_BRAPOS:
1029        case OP_SBRAPOS:
1030        allow_zero = FALSE;
1031    
1032        POSSESSIVE_NON_CAPTURE:
1033        matched_once = FALSE;
1034        code_offset = ecode - md->start_code;
1035    
1036      for (;;)      for (;;)
1037        {        {
1038        /* When we get to the final alternative within the brackets, we would        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1039        return the result of a recursive call to match() whatever happened. We        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1040        can reduce stack usage by turning this into a tail recursion. */          eptrb, RM48);
1041          if (rrc == MATCH_KETRPOS)
1042        if (ecode[GET(ecode, 1)] != OP_ALT)          {
1043         {          offset_top = md->end_offset_top;
1044         ecode += 1 + LINK_SIZE;          eptr = md->end_match_ptr;
1045         flags = match_isgroup;          ecode = md->start_code + code_offset;
1046         DPRINTF(("bracket 0 tail recursion\n"));          matched_once = TRUE;
1047         goto TAIL_RECURSE;          continue;
1048         }          }
1049          if (rrc != MATCH_NOMATCH &&
1050        /* For non-final alternatives, continue the loop for a NOMATCH result;            (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1051        otherwise return. */          RRETURN(rrc);
   
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,  
         match_isgroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1052        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1053          if (*ecode != OP_ALT) break;
1054        }        }
1055    
1056        if (matched_once || allow_zero)
1057          {
1058          ecode += 1 + LINK_SIZE;
1059          break;
1060          }
1061        RRETURN(MATCH_NOMATCH);
1062    
1063      /* Control never reaches here. */      /* Control never reaches here. */
1064    
1065      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1066      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
1067      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
1068      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. */  
1069    
1070      case OP_COND:      case OP_COND:
1071      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      case OP_SCOND:
1072        codelink = GET(ecode, 1);
1073    
1074        /* Because of the way auto-callout works during compile, a callout item is
1075        inserted between OP_COND and an assertion condition. */
1076    
1077        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1078          {
1079          if (pcre_callout != NULL)
1080            {
1081            pcre_callout_block cb;
1082            cb.version          = 1;   /* Version 1 of the callout block */
1083            cb.callout_number   = ecode[LINK_SIZE+2];
1084            cb.offset_vector    = md->offset_vector;
1085            cb.subject          = (PCRE_SPTR)md->start_subject;
1086            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1087            cb.start_match      = (int)(mstart - md->start_subject);
1088            cb.current_position = (int)(eptr - md->start_subject);
1089            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1090            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1091            cb.capture_top      = offset_top/2;
1092            cb.capture_last     = md->capture_last;
1093            cb.callout_data     = md->callout_data;
1094            if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1095            if (rrc < 0) RRETURN(rrc);
1096            }
1097          ecode += _pcre_OP_lengths[OP_CALLOUT];
1098          }
1099    
1100        condcode = ecode[LINK_SIZE+1];
1101    
1102        /* Now see what the actual condition is */
1103    
1104        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1105          {
1106          if (md->recursive == NULL)                /* Not recursing => FALSE */
1107            {
1108            condition = FALSE;
1109            ecode += GET(ecode, 1);
1110            }
1111          else
1112            {
1113            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1114            condition =  (recno == RREF_ANY || recno == md->recursive->group_num);
1115    
1116            /* If the test is for recursion into a specific subpattern, and it is
1117            false, but the test was set up by name, scan the table to see if the
1118            name refers to any other numbers, and test them. The condition is true
1119            if any one is set. */
1120    
1121            if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
1122              {
1123              uschar *slotA = md->name_table;
1124              for (i = 0; i < md->name_count; i++)
1125                {
1126                if (GET2(slotA, 0) == recno) break;
1127                slotA += md->name_entry_size;
1128                }
1129    
1130              /* Found a name for the number - there can be only one; duplicate
1131              names for different numbers are allowed, but not vice versa. First
1132              scan down for duplicates. */
1133    
1134              if (i < md->name_count)
1135                {
1136                uschar *slotB = slotA;
1137                while (slotB > md->name_table)
1138                  {
1139                  slotB -= md->name_entry_size;
1140                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1141                    {
1142                    condition = GET2(slotB, 0) == md->recursive->group_num;
1143                    if (condition) break;
1144                    }
1145                  else break;
1146                  }
1147    
1148                /* Scan up for duplicates */
1149    
1150                if (!condition)
1151                  {
1152                  slotB = slotA;
1153                  for (i++; i < md->name_count; i++)
1154                    {
1155                    slotB += md->name_entry_size;
1156                    if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1157                      {
1158                      condition = GET2(slotB, 0) == md->recursive->group_num;
1159                      if (condition) break;
1160                      }
1161                    else break;
1162                    }
1163                  }
1164                }
1165              }
1166    
1167            /* Chose branch according to the condition */
1168    
1169            ecode += condition? 3 : GET(ecode, 1);
1170            }
1171          }
1172    
1173        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1174        {        {
1175        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1176        condition = (offset == CREF_RECURSE * 2)?        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1177          (md->recursive != NULL) :  
1178          (offset < offset_top && md->offset_vector[offset] >= 0);        /* If the numbered capture is unset, but the reference was by name,
1179        ecode += condition? (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1));        scan the table to see if the name refers to any other numbers, and test
1180        flags = match_isgroup;        them. The condition is true if any one is set. This is tediously similar
1181        goto TAIL_RECURSE;        to the code above, but not close enough to try to amalgamate. */
1182    
1183          if (!condition && condcode == OP_NCREF)
1184            {
1185            int refno = offset >> 1;
1186            uschar *slotA = md->name_table;
1187    
1188            for (i = 0; i < md->name_count; i++)
1189              {
1190              if (GET2(slotA, 0) == refno) break;
1191              slotA += md->name_entry_size;
1192              }
1193    
1194            /* Found a name for the number - there can be only one; duplicate names
1195            for different numbers are allowed, but not vice versa. First scan down
1196            for duplicates. */
1197    
1198            if (i < md->name_count)
1199              {
1200              uschar *slotB = slotA;
1201              while (slotB > md->name_table)
1202                {
1203                slotB -= md->name_entry_size;
1204                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1205                  {
1206                  offset = GET2(slotB, 0) << 1;
1207                  condition = offset < offset_top &&
1208                    md->offset_vector[offset] >= 0;
1209                  if (condition) break;
1210                  }
1211                else break;
1212                }
1213    
1214              /* Scan up for duplicates */
1215    
1216              if (!condition)
1217                {
1218                slotB = slotA;
1219                for (i++; i < md->name_count; i++)
1220                  {
1221                  slotB += md->name_entry_size;
1222                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1223                    {
1224                    offset = GET2(slotB, 0) << 1;
1225                    condition = offset < offset_top &&
1226                      md->offset_vector[offset] >= 0;
1227                    if (condition) break;
1228                    }
1229                  else break;
1230                  }
1231                }
1232              }
1233            }
1234    
1235          /* Chose branch according to the condition */
1236    
1237          ecode += condition? 3 : GET(ecode, 1);
1238          }
1239    
1240        else if (condcode == OP_DEF)     /* DEFINE - always false */
1241          {
1242          condition = FALSE;
1243          ecode += GET(ecode, 1);
1244        }        }
1245    
1246      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1247      the final argument TRUE causes it to stop at the end of an assertion. */      md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1248        an assertion. */
1249    
1250      else      else
1251        {        {
1252        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1253            match_condassert | match_isgroup);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1254        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1255          {          {
1256          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          condition = TRUE;
1257            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1258          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1259          }          }
1260        else if (rrc != MATCH_NOMATCH)        else if (rrc != MATCH_NOMATCH &&
1261                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1262          {          {
1263          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1264          }          }
1265        else ecode += GET(ecode, 1);        else
1266            {
1267            condition = FALSE;
1268            ecode += codelink;
1269            }
1270          }
1271    
1272        /* 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,
1273        we can use tail recursion to avoid using another stack frame. */      we used to use tail recursion to avoid using another stack frame, except
1274        when there was unlimited repeat of a possibly empty group. However, that
1275        strategy no longer works because of the possibilty of (*THEN) being
1276        encountered in the branch. A recursive call to match() is always required,
1277        unless the second alternative doesn't exist, in which case we can just
1278        plough on. */
1279    
1280        if (condition || *ecode == OP_ALT)
1281          {
1282          if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1283          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1284          if (rrc == MATCH_THEN && md->start_match_ptr == ecode)
1285            rrc = MATCH_NOMATCH;
1286          RRETURN(rrc);
1287          }
1288        else                         /* Condition false & no alternative */
1289          {
1290        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
1291        }        }
1292      /* Control never reaches here */      break;
1293    
1294    
1295        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1296        to close any currently open capturing brackets. */
1297    
1298        case OP_CLOSE:
1299        number = GET2(ecode, 1);
1300        offset = number << 1;
1301    
1302      /* Skip over conditional reference or large extraction number data if  #ifdef PCRE_DEBUG
1303      encountered. */        printf("end bracket %d at *ACCEPT", number);
1304          printf("\n");
1305    #endif
1306    
1307      case OP_CREF:      md->capture_last = number;
1308      case OP_BRANUMBER:      if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1309          {
1310          md->offset_vector[offset] =
1311            md->offset_vector[md->offset_end - number];
1312          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1313          if (offset_top <= offset) offset_top = offset + 2;
1314          }
1315      ecode += 3;      ecode += 3;
1316      break;      break;
1317    
     /* End of the pattern. If we are in a recursion, we should restore the  
     offsets appropriately and continue from after the call. */  
1318    
1319        /* End of the pattern, either real or forced. If we are in a recursion, we
1320        should restore the offsets appropriately, and if it's a top-level
1321        recursion, continue from after the call. */
1322    
1323        case OP_ACCEPT:
1324        case OP_ASSERT_ACCEPT:
1325      case OP_END:      case OP_END:
1326      if (md->recursive != NULL && md->recursive->group_num == 0)  
1327    /*
1328        if (md->recursive != NULL)
1329        {        {
1330        recursion_info *rec = md->recursive;        recursion_info *rec = md->recursive;
1331        DPRINTF(("End of pattern in a (?0) recursion\n"));  
1332        md->recursive = rec->prevrec;        md->recursive = rec->prevrec;
1333        memmove(md->offset_vector, rec->offset_save,  
1334          memmove(md->offset_vector, rec->offset_save,
1335          rec->saved_max * sizeof(int));          rec->saved_max * sizeof(int));
1336        md->start_match = rec->save_start;        offset_top = rec->save_offset_top;
1337        ims = original_ims;        if (rec->group_num == 0)
1338        ecode = rec->after_call;          {
1339        break;          ecode = rec->after_call;
1340            break;
1341            }
1342        }        }
1343    */
1344        /* Otherwise, if we have matched an empty string, fail if not in an
1345        assertion and if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1346        is set and we have matched at the start of the subject. In both cases,
1347        backtracking will then try other alternatives, if any. */
1348    
1349    /*    else */ if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1350    
1351             md->recursive == NULL &&
1352    
1353            (md->notempty ||
1354              (md->notempty_atstart &&
1355                mstart == md->start_subject + md->start_offset)))
1356          MRRETURN(MATCH_NOMATCH);
1357    
1358        /* Otherwise, we have a match. */
1359    
1360        md->end_match_ptr = eptr;           /* Record where we ended */
1361        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1362        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1363    
1364      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* For some reason, the macros don't work properly if an expression is
1365      string - backtracking will then try other alternatives, if any. */      given as the argument to MRRETURN when the heap is in use. */
1366    
1367      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1368      md->end_match_ptr = eptr;          /* Record where we ended */      MRRETURN(rrc);
     md->end_offset_top = offset_top;   /* and how many extracts were taken */  
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1369    
1370      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1371      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,
1372      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
1373      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
1374      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1375        of a condition, we want to return immediately afterwards. The caller of
1376        this incarnation of the match() function will have set MATCH_CONDASSERT in
1377        md->match_function type, and one of these opcodes will be the first opcode
1378        that is processed. We use a local variable that is preserved over calls to
1379        match() to remember this case. */
1380    
1381      case OP_ASSERT:      case OP_ASSERT:
1382      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1383        if (md->match_function_type == MATCH_CONDASSERT)
1384          {
1385          condassert = TRUE;
1386          md->match_function_type = 0;
1387          }
1388        else condassert = FALSE;
1389    
1390      do      do
1391        {        {
1392        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1393          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1394        if (rrc == MATCH_MATCH) break;          {
1395        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          mstart = md->start_match_ptr;   /* In case \K reset it */
1396            break;
1397            }
1398          if (rrc != MATCH_NOMATCH &&
1399              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1400            RRETURN(rrc);
1401        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1402        }        }
1403      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1404      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1405        if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1406    
1407      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1408    
1409      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1410    
1411      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1412      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 798  for (;;) Line 1416  for (;;)
1416      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1417      continue;      continue;
1418    
1419      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1420        PRUNE, or COMMIT means we must assume failure without checking subsequent
1421        branches. */
1422    
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1425        if (md->match_function_type == MATCH_CONDASSERT)
1426          {
1427          condassert = TRUE;
1428          md->match_function_type = 0;
1429          }
1430        else condassert = FALSE;
1431    
1432      do      do
1433        {        {
1434        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1435          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1436        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1437        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          {
1438            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1439            break;
1440            }
1441          if (rrc != MATCH_NOMATCH &&
1442              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1443            RRETURN(rrc);
1444        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1445        }        }
1446      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1447    
1448      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1449    
1450      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1451      continue;      continue;
1452    
# Line 826  for (;;) Line 1459  for (;;)
1459  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1460      if (utf8)      if (utf8)
1461        {        {
1462        c = GET(ecode,1);        i = GET(ecode, 1);
1463        for (i = 0; i < c; i++)        while (i-- > 0)
1464          {          {
1465          eptr--;          eptr--;
1466          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1467          BACKCHAR(eptr)          BACKCHAR(eptr);
1468          }          }
1469        }        }
1470      else      else
# Line 840  for (;;) Line 1473  for (;;)
1473      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1474    
1475        {        {
1476        eptr -= GET(ecode,1);        eptr -= GET(ecode, 1);
1477        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1478        }        }
1479    
1480      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1481    
1482        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1483      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1484      break;      break;
1485    
# Line 861  for (;;) Line 1495  for (;;)
1495        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1496        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1497        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1498        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1499        cb.start_match      = md->start_match - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1500        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1501        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1502        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1503        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1504        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1505        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1506        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1507        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1508        }        }
1509      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 878  for (;;) Line 1512  for (;;)
1512      /* Recursion either matches the current regex, or some subexpression. The      /* Recursion either matches the current regex, or some subexpression. The
1513      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
1514      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1515    
1516      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1517      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
1518      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
1519      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
1520      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
1521      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
1522      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.  
1523    
1524      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
1525      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
1526      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1527        a lot, so he is not to blame for the current way it works. */
1528    
1529      case OP_RECURSE:      case OP_RECURSE:
1530        {        {
1531        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1532        new_recursive.group_num = *callpat - OP_BRA;        new_recursive.group_num = (callpat == md->start_code)? 0 :
1533            GET2(callpat, 1 + LINK_SIZE);
       /* For extended extraction brackets (large number), we have to fish out  
       the number from a dummy opcode at the start. */  
   
       if (new_recursive.group_num > EXTRACT_BASIC_MAX)  
         new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);  
1534    
1535        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1536    
1537        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1538        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1539    
1540        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1541    
1542        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1543    
1544        /* Now save the offset data. */        /* Now save the offset data */
1545    
1546        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1547        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 926  for (;;) Line 1552  for (;;)
1552            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1553          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1554          }          }
   
1555        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1556              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1557        new_recursive.save_start = md->start_match;  
1558        md->start_match = eptr;        /* OK, now we can do the recursion. After processing each alternative,
1559          restore the offset data. If there were nested recursions, md->recursive
1560        /* OK, now we can do the recursion. For each top-level alternative we        might be changed, so reset it before looping. */
       restore the offset and recursion data. */  
1561    
1562        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1563          cbegroup = (*callpat >= OP_SBRA);
1564        do        do
1565          {          {
1566          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1567              eptrb, match_isgroup);          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1568          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1569            memcpy(md->offset_vector, new_recursive.offset_save,
1570                new_recursive.saved_max * sizeof(int));
1571            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1572            {            {
1573            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
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_MATCH);  
1578              /* Set where we got to in the subject, and reset the start in case
1579              it was changed by \K. This *is* propagated back out of a recursion,
1580              for Perl compatibility. */
1581    
1582              eptr = md->end_match_ptr;
1583              mstart = md->start_match_ptr;
1584              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1585            }            }
1586          else if (rrc != MATCH_NOMATCH)          else if (rrc != MATCH_NOMATCH &&
1587                    (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1588            {            {
1589            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1590              if (new_recursive.offset_save != stacksave)
1591                (pcre_free)(new_recursive.offset_save);
1592            RRETURN(rrc);            RRETURN(rrc);
1593            }            }
1594    
1595          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1596          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1597          }          }
1598        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 965  for (;;) Line 1601  for (;;)
1601        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1602        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1603          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1604        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1605        }        }
1606      /* Control never reaches here */  
1607        RECURSION_MATCHED:
1608      /* "Once" brackets are like assertion brackets except that after a match,      break;
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
1609    
1610      case OP_ONCE:      /* An alternation is the end of a branch; scan along to find the end of the
1611      prev = ecode;      bracketed group and go to there. */
     saved_eptr = eptr;  
1612    
1613      do      case OP_ALT:
1614        {      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1615        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,      break;
         eptrb, match_isgroup);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
1616    
1617      /* If hit the end of the group (which could be repeated), fail */      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1618        indicating that it may occur zero times. It may repeat infinitely, or not
1619        at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1620        with fixed upper repeat limits are compiled as a number of copies, with the
1621        optional ones preceded by BRAZERO or BRAMINZERO. */
1622    
1623        case OP_BRAZERO:
1624        next = ecode + 1;
1625        RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1626        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1627        do next += GET(next, 1); while (*next == OP_ALT);
1628        ecode = next + 1 + LINK_SIZE;
1629        break;
1630    
1631        case OP_BRAMINZERO:
1632        next = ecode + 1;
1633        do next += GET(next, 1); while (*next == OP_ALT);
1634        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1635        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1636        ecode++;
1637        break;
1638    
1639      if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);      case OP_SKIPZERO:
1640        next = ecode+1;
1641        do next += GET(next,1); while (*next == OP_ALT);
1642        ecode = next + 1 + LINK_SIZE;
1643        break;
1644    
1645        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1646        here; just jump to the group, with allow_zero set TRUE. */
1647    
1648        case OP_BRAPOSZERO:
1649        op = *(++ecode);
1650        allow_zero = TRUE;
1651        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1652          goto POSSESSIVE_NON_CAPTURE;
1653    
1654      /* Continue as from after the assertion, updating the offsets high water      /* End of a group, repeated or non-repeating. */
     mark, since extracts may have been taken. */  
1655    
1656      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_KET:
1657        case OP_KETRMIN:
1658      offset_top = md->end_offset_top;      case OP_KETRMAX:
1659      eptr = md->end_match_ptr;      case OP_KETRPOS:
1660        prev = ecode - GET(ecode, 1);
1661      /* For a non-repeating ket, just continue at this level. This also  
1662      happens for a repeating ket if no characters were matched in the group.      /* If this was a group that remembered the subject start, in order to break
1663      This is the forcible breaking of infinite loops as implemented in Perl      infinite repeats of empty string matches, retrieve the subject start from
1664      5.005. If there is an options reset, it will get obeyed in the normal      the chain. Otherwise, set it NULL. */
1665      course of events. */  
1666        if (*prev >= OP_SBRA || *prev == OP_ONCE)
1667      if (*ecode == OP_KET || eptr == saved_eptr)        {
1668        {        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1669        ecode += 1+LINK_SIZE;        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1670        break;        }
1671        }      else saved_eptr = NULL;
1672    
1673      /* The repeating kets try the rest of the pattern or restart from the      /* If we are at the end of an assertion group, stop matching and return
1674      preceding bracket, in the appropriate order. The second "call" of match()      MATCH_MATCH, but record the current high water mark for use by positive
1675      uses tail recursion, to avoid using another stack frame. We need to reset      assertions. We also need to record the match start in case it was changed
1676      any options that changed within the bracket before re-running it, so      by \K. */
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
   
     /* 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);  
     break;  
   
     /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating  
     that it may occur zero times. It may repeat infinitely, or not at all -  
     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, match_isgroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       do next += GET(next,1); while (*next == OP_ALT);  
       ecode = next + 1+LINK_SIZE;  
       }  
     break;  
   
     case OP_BRAMINZERO:  
       {  
       next = ecode+1;  
       do next += GET(next,1); while (*next == OP_ALT);  
       RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,  
         match_isgroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode++;  
       }  
     break;  
   
     /* End of a group, repeated or non-repeating. If we are at the end of  
     an assertion "group", stop matching and return MATCH_MATCH, but record the  
     current high water mark for use by positive assertions. Do this also  
     for the "once" (not-backup up) groups. */  
   
     case OP_KET:  
     case OP_KETRMIN:  
     case OP_KETRMAX:  
     prev = ecode - GET(ecode, 1);  
     saved_eptr = eptrb->epb_saved_eptr;  
   
     /* Back up the stack of bracket start pointers. */  
   
     eptrb = eptrb->epb_prev;  
1677    
1678      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1679          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
         *prev == OP_ONCE)  
1680        {        {
1681        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE */
1682        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1683        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1684          MRRETURN(MATCH_MATCH);
1685        }        }
1686    
1687      /* In all other cases except a conditional group we have to check the      /* For capturing groups we have to check the group number back at the start
1688      group number back at the start and if necessary complete handling an      and if necessary complete handling an extraction by setting the offsets and
1689      extraction by setting the offsets and bumping the high water mark. */      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1690        into group 0, so it won't be picked up here. Instead, we catch it when the
1691        OP_END is reached. Other recursion is handled here. We just have to record
1692        the current subject position and start match pointer and give a MATCH
1693        return. */
1694    
1695      if (*prev != OP_COND)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1696            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1697        {        {
1698        number = *prev - OP_BRA;        number = GET2(prev, 1+LINK_SIZE);
   
       /* For extended extraction brackets (large number), we have to fish out  
       the number from a dummy opcode at the start. */  
   
       if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);  
1699        offset = number << 1;        offset = number << 1;
1700    
1701  #ifdef DEBUG  #ifdef PCRE_DEBUG
1702        printf("end bracket %d", number);        printf("end bracket %d", number);
1703        printf("\n");        printf("\n");
1704  #endif  #endif
1705    
1706        /* Test for a numbered group. This includes groups called as a result        /* Handle a recursively called group. */
       of recursion. Note that whole-pattern recursion is coded as a recurse  
       into group 0, so it won't be picked up here. Instead, we catch it when  
       the OP_END is reached. */  
1707    
1708        if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1709          {          {
1710          md->capture_last = number;          md->end_match_ptr = eptr;
1711          if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1712            {          RRETURN(MATCH_MATCH);
1713            md->offset_vector[offset] =          }
1714              md->offset_vector[md->offset_end - number];  
1715            md->offset_vector[offset+1] = eptr - md->start_subject;        /* Deal with capturing */
1716            if (offset_top <= offset) offset_top = offset + 2;  
1717            }        md->capture_last = number;
1718          if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1719          /* Handle a recursively called group. Restore the offsets          {
1720          appropriately and continue from after the call. */          /* If offset is greater than offset_top, it means that we are
1721            "skipping" a capturing group, and that group's offsets must be marked
1722          if (md->recursive != NULL && md->recursive->group_num == number)          unset. In earlier versions of PCRE, all the offsets were unset at the
1723            {          start of matching, but this doesn't work because atomic groups and
1724            recursion_info *rec = md->recursive;          assertions can cause a value to be set that should later be unset.
1725            DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1726            md->recursive = rec->prevrec;          part of the atomic group, but this is not on the final matching path,
1727            md->start_match = rec->save_start;          so must be unset when 2 is set. (If there is no group 2, there is no
1728            memcpy(md->offset_vector, rec->offset_save,          problem, because offset_top will then be 2, indicating no capture.) */
1729              rec->saved_max * sizeof(int));  
1730            ecode = rec->after_call;          if (offset > offset_top)
1731            ims = original_ims;            {
1732            break;            register int *iptr = md->offset_vector + offset_top;
1733            }            register int *iend = md->offset_vector + offset;
1734              while (iptr < iend) *iptr++ = -1;
1735              }
1736    
1737            /* Now make the extraction */
1738    
1739            md->offset_vector[offset] =
1740              md->offset_vector[md->offset_end - number];
1741            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1742            if (offset_top <= offset) offset_top = offset + 2;
1743          }          }
1744        }        }
1745    
1746      /* Reset the value of the ims flags, in case they got changed during      /* For an ordinary non-repeating ket, just continue at this level. This
1747      the group. */      also happens for a repeating ket if no characters were matched in the
1748        group. This is the forcible breaking of infinite loops as implemented in
1749      ims = original_ims;      Perl 5.005. For a non-repeating atomic group, establish a backup point by
1750      DPRINTF(("ims reset to %02lx\n", ims));      processing the rest of the pattern at a lower level. If this results in a
1751        NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1752      /* For a non-repeating ket, just continue at this level. This also      bypassing intermediate backup points, but resetting any captures that
1753      happens for a repeating ket if no characters were matched in the group.      happened along the way. */
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
1754    
1755      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1756        {        {
1757        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1758            {
1759            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1760            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1761            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1762            RRETURN(MATCH_ONCE);
1763            }
1764          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1765        break;        break;
1766        }        }
1767    
1768      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1769      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1770      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
1771    
1772        if (*ecode == OP_KETRPOS)
1773          {
1774          md->end_match_ptr = eptr;
1775          md->end_offset_top = offset_top;
1776          RRETURN(MATCH_KETRPOS);
1777          }
1778    
1779        /* The normal repeating kets try the rest of the pattern or restart from
1780        the preceding bracket, in the appropriate order. In the second case, we can
1781        use tail recursion to avoid using another stack frame, unless we have an
1782        an atomic group or an unlimited repeat of a group that can match an empty
1783        string. */
1784    
1785      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1786        {        {
1787        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
1788        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1789          if (*prev == OP_ONCE)
1790            {
1791            RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
1792            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1793            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1794            RRETURN(MATCH_ONCE);
1795            }
1796          if (*prev >= OP_SBRA)    /* Could match an empty string */
1797            {
1798            md->match_function_type = MATCH_CBEGROUP;
1799            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1800            RRETURN(rrc);
1801            }
1802        ecode = prev;        ecode = prev;
       flags = match_isgroup;  
1803        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1804        }        }
1805      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1806        {        {
1807        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1808          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1809          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1810        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1811          if (*prev == OP_ONCE)
1812            {
1813            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
1814            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1815            md->once_target = prev;
1816            RRETURN(MATCH_ONCE);
1817            }
1818        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1819        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1820        }        }
1821      /* Control never gets here */      /* Control never gets here */
1822    
1823      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1824    
1825      case OP_CIRC:      case OP_CIRC:
1826      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1827      if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject ||  
            eptr < md->start_subject + md->nllen ||  
            !IS_NEWLINE(eptr - md->nllen)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
   
1828      /* Start of subject assertion */      /* Start of subject assertion */
1829    
1830      case OP_SOD:      case OP_SOD:
1831      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1832        ecode++;
1833        break;
1834    
1835        /* Multiline mode: start of subject unless notbol, or after any newline. */
1836    
1837        case OP_CIRCM:
1838        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1839        if (eptr != md->start_subject &&
1840            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1841          MRRETURN(MATCH_NOMATCH);
1842      ecode++;      ecode++;
1843      break;      break;
1844    
1845      /* Start of match assertion */      /* Start of match assertion */
1846    
1847      case OP_SOM:      case OP_SOM:
1848      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
1849      ecode++;      ecode++;
1850      break;      break;
1851    
1852      /* 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. */  
1853    
1854      case OP_DOLL:      case OP_SET_SOM:
1855      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
1856        {      ecode++;
1857        if (eptr < md->end_subject)      break;
1858          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
1859        else      /* Multiline mode: assert before any newline, or before end of subject
1860          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
1861        ecode++;  
1862        break;      case OP_DOLLM:
1863        }      if (eptr < md->end_subject)
1864          { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
1865      else      else
1866        {        {
1867        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1868        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
1869        }        }
1870        ecode++;
1871        break;
1872    
1873        /* Not multiline mode: assert before a terminating newline or before end of
1874        subject unless noteol is set. */
1875    
1876        case OP_DOLL:
1877        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1878        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1879    
1880      /* ... else fall through for endonly */      /* ... else fall through for endonly */
1881    
1882      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
1883    
1884      case OP_EOD:      case OP_EOD:
1885      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
1886        SCHECK_PARTIAL();
1887      ecode++;      ecode++;
1888      break;      break;
1889    
1890      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
1891    
1892      case OP_EODN:      case OP_EODN:
1893      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
1894          (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))      if (eptr < md->end_subject &&
1895        RRETURN(MATCH_NOMATCH);          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1896          MRRETURN(MATCH_NOMATCH);
1897    
1898        /* Either at end of string or \n before end. */
1899    
1900        SCHECK_PARTIAL();
1901      ecode++;      ecode++;
1902      break;      break;
1903    
# Line 1276  for (;;) Line 1909  for (;;)
1909    
1910        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
1911        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
1912        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
1913          partial matching. */
1914    
1915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1916        if (utf8)        if (utf8)
1917          {          {
1918            /* Get status of previous character */
1919    
1920          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
1921            {            {
1922            const uschar *lastptr = eptr - 1;            USPTR lastptr = eptr - 1;
1923            while((*lastptr & 0xc0) == 0x80) lastptr--;            while((*lastptr & 0xc0) == 0x80) lastptr--;
1924              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
1925            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
1926    #ifdef SUPPORT_UCP
1927              if (md->use_ucp)
1928                {
1929                if (c == '_') prev_is_word = TRUE; else
1930                  {
1931                  int cat = UCD_CATEGORY(c);
1932                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1933                  }
1934                }
1935              else
1936    #endif
1937            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1938            }            }
1939          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
1940            /* Get status of next character */
1941    
1942            if (eptr >= md->end_subject)
1943              {
1944              SCHECK_PARTIAL();
1945              cur_is_word = FALSE;
1946              }
1947            else
1948            {            {
1949            GETCHAR(c, eptr);            GETCHAR(c, eptr);
1950    #ifdef SUPPORT_UCP
1951              if (md->use_ucp)
1952                {
1953                if (c == '_') cur_is_word = TRUE; else
1954                  {
1955                  int cat = UCD_CATEGORY(c);
1956                  cur_is_word = (cat == ucp_L || cat == ucp_N);
1957                  }
1958                }
1959              else
1960    #endif
1961            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1962            }            }
1963          }          }
1964        else        else
1965  #endif  #endif
1966    
1967        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
1968          consistency with the behaviour of \w we do use it in this case. */
1969    
1970          {          {
1971          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
1972            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
1973          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
1974            ((md->ctypes[*eptr] & ctype_word) != 0);            {
1975              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
1976    #ifdef SUPPORT_UCP
1977              if (md->use_ucp)
1978                {
1979                c = eptr[-1];
1980                if (c == '_') prev_is_word = TRUE; else
1981                  {
1982                  int cat = UCD_CATEGORY(c);
1983                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1984                  }
1985                }
1986              else
1987    #endif
1988              prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
1989              }
1990    
1991            /* Get status of next character */
1992    
1993            if (eptr >= md->end_subject)
1994              {
1995              SCHECK_PARTIAL();
1996              cur_is_word = FALSE;
1997              }
1998            else
1999    #ifdef SUPPORT_UCP
2000            if (md->use_ucp)
2001              {
2002              c = *eptr;
2003              if (c == '_') cur_is_word = TRUE; else
2004                {
2005                int cat = UCD_CATEGORY(c);
2006                cur_is_word = (cat == ucp_L || cat == ucp_N);
2007                }
2008              }
2009            else
2010    #endif
2011            cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
2012          }          }
2013    
2014        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
2015    
2016        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
2017             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2018          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2019        }        }
2020      break;      break;
2021    
2022      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2023    
2024      case OP_ANY:      case OP_ANY:
2025      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
2026        /* Fall through */
2027    
2028        case OP_ALLANY:
2029        if (eptr++ >= md->end_subject)
2030        {        {
2031        if (eptr <= md->end_subject - md->nllen && IS_NEWLINE(eptr))        SCHECK_PARTIAL();
2032          RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2033        }        }
2034      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++;  
2035      ecode++;      ecode++;
2036      break;      break;
2037    
# Line 1332  for (;;) Line 2039  for (;;)
2039      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2040    
2041      case OP_ANYBYTE:      case OP_ANYBYTE:
2042      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr++ >= md->end_subject)
2043          {
2044          SCHECK_PARTIAL();
2045          MRRETURN(MATCH_NOMATCH);
2046          }
2047      ecode++;      ecode++;
2048      break;      break;
2049    
2050      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2051      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2052          {
2053          SCHECK_PARTIAL();
2054          MRRETURN(MATCH_NOMATCH);
2055          }
2056      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2057      if (      if (
2058  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1345  for (;;) Line 2060  for (;;)
2060  #endif  #endif
2061         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2062         )         )
2063        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2064      ecode++;      ecode++;
2065      break;      break;
2066    
2067      case OP_DIGIT:      case OP_DIGIT:
2068      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2069          {
2070          SCHECK_PARTIAL();
2071          MRRETURN(MATCH_NOMATCH);
2072          }
2073      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2074      if (      if (
2075  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1358  for (;;) Line 2077  for (;;)
2077  #endif  #endif
2078         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2079         )         )
2080        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2081      ecode++;      ecode++;
2082      break;      break;
2083    
2084      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2085      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2086          {
2087          SCHECK_PARTIAL();
2088          MRRETURN(MATCH_NOMATCH);
2089          }
2090      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2091      if (      if (
2092  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1371  for (;;) Line 2094  for (;;)
2094  #endif  #endif
2095         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2096         )         )
2097        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2098      ecode++;      ecode++;
2099      break;      break;
2100    
2101      case OP_WHITESPACE:      case OP_WHITESPACE:
2102      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2103          {
2104          SCHECK_PARTIAL();
2105          MRRETURN(MATCH_NOMATCH);
2106          }
2107      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2108      if (      if (
2109  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1384  for (;;) Line 2111  for (;;)
2111  #endif  #endif
2112         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2113         )         )
2114        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2115      ecode++;      ecode++;
2116      break;      break;
2117    
2118      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2119      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2120          {
2121          SCHECK_PARTIAL();
2122          MRRETURN(MATCH_NOMATCH);
2123          }
2124      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2125      if (      if (
2126  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1397  for (;;) Line 2128  for (;;)
2128  #endif  #endif
2129         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2130         )         )
2131        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2132      ecode++;      ecode++;
2133      break;      break;
2134    
2135      case OP_WORDCHAR:      case OP_WORDCHAR:
2136      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2137          {
2138          SCHECK_PARTIAL();
2139          MRRETURN(MATCH_NOMATCH);
2140          }
2141      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2142      if (      if (
2143  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1410  for (;;) Line 2145  for (;;)
2145  #endif  #endif
2146         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2147         )         )
2148        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2149        ecode++;
2150        break;
2151    
2152        case OP_ANYNL:
2153        if (eptr >= md->end_subject)
2154          {
2155          SCHECK_PARTIAL();
2156          MRRETURN(MATCH_NOMATCH);
2157          }
2158        GETCHARINCTEST(c, eptr);
2159        switch(c)
2160          {
2161          default: MRRETURN(MATCH_NOMATCH);
2162    
2163          case 0x000d:
2164          if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2165          break;
2166    
2167          case 0x000a:
2168          break;
2169    
2170          case 0x000b:
2171          case 0x000c:
2172          case 0x0085:
2173          case 0x2028:
2174          case 0x2029:
2175          if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
2176          break;
2177          }
2178        ecode++;
2179        break;
2180    
2181        case OP_NOT_HSPACE:
2182        if (eptr >= md->end_subject)
2183          {
2184          SCHECK_PARTIAL();
2185          MRRETURN(MATCH_NOMATCH);
2186          }
2187        GETCHARINCTEST(c, eptr);
2188        switch(c)
2189          {
2190          default: break;
2191          case 0x09:      /* HT */
2192          case 0x20:      /* SPACE */
2193          case 0xa0:      /* NBSP */
2194          case 0x1680:    /* OGHAM SPACE MARK */
2195          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2196          case 0x2000:    /* EN QUAD */
2197          case 0x2001:    /* EM QUAD */
2198          case 0x2002:    /* EN SPACE */
2199          case 0x2003:    /* EM SPACE */
2200          case 0x2004:    /* THREE-PER-EM SPACE */
2201          case 0x2005:    /* FOUR-PER-EM SPACE */
2202          case 0x2006:    /* SIX-PER-EM SPACE */
2203          case 0x2007:    /* FIGURE SPACE */
2204          case 0x2008:    /* PUNCTUATION SPACE */
2205          case 0x2009:    /* THIN SPACE */
2206          case 0x200A:    /* HAIR SPACE */
2207          case 0x202f:    /* NARROW NO-BREAK SPACE */
2208          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2209          case 0x3000:    /* IDEOGRAPHIC SPACE */
2210          MRRETURN(MATCH_NOMATCH);
2211          }
2212        ecode++;
2213        break;
2214    
2215        case OP_HSPACE:
2216        if (eptr >= md->end_subject)
2217          {
2218          SCHECK_PARTIAL();
2219          MRRETURN(MATCH_NOMATCH);
2220          }
2221        GETCHARINCTEST(c, eptr);
2222        switch(c)
2223          {
2224          default: MRRETURN(MATCH_NOMATCH);
2225          case 0x09:      /* HT */
2226          case 0x20:      /* SPACE */
2227          case 0xa0:      /* NBSP */
2228          case 0x1680:    /* OGHAM SPACE MARK */
2229          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2230          case 0x2000:    /* EN QUAD */
2231          case 0x2001:    /* EM QUAD */
2232          case 0x2002:    /* EN SPACE */
2233          case 0x2003:    /* EM SPACE */
2234          case 0x2004:    /* THREE-PER-EM SPACE */
2235          case 0x2005:    /* FOUR-PER-EM SPACE */
2236          case 0x2006:    /* SIX-PER-EM SPACE */
2237          case 0x2007:    /* FIGURE SPACE */
2238          case 0x2008:    /* PUNCTUATION SPACE */
2239          case 0x2009:    /* THIN SPACE */
2240          case 0x200A:    /* HAIR SPACE */
2241          case 0x202f:    /* NARROW NO-BREAK SPACE */
2242          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2243          case 0x3000:    /* IDEOGRAPHIC SPACE */
2244          break;
2245          }
2246        ecode++;
2247        break;
2248    
2249        case OP_NOT_VSPACE:
2250        if (eptr >= md->end_subject)
2251          {
2252          SCHECK_PARTIAL();
2253          MRRETURN(MATCH_NOMATCH);
2254          }
2255        GETCHARINCTEST(c, eptr);
2256        switch(c)
2257          {
2258          default: break;
2259          case 0x0a:      /* LF */
2260          case 0x0b:      /* VT */
2261          case 0x0c:      /* FF */
2262          case 0x0d:      /* CR */
2263          case 0x85:      /* NEL */
2264          case 0x2028:    /* LINE SEPARATOR */
2265          case 0x2029:    /* PARAGRAPH SEPARATOR */
2266          MRRETURN(MATCH_NOMATCH);
2267          }
2268        ecode++;
2269        break;
2270    
2271        case OP_VSPACE:
2272        if (eptr >= md->end_subject)
2273          {
2274          SCHECK_PARTIAL();
2275          MRRETURN(MATCH_NOMATCH);
2276          }
2277        GETCHARINCTEST(c, eptr);
2278        switch(c)
2279          {
2280          default: MRRETURN(MATCH_NOMATCH);
2281          case 0x0a:      /* LF */
2282          case 0x0b:      /* VT */
2283          case 0x0c:      /* FF */
2284          case 0x0d:      /* CR */
2285          case 0x85:      /* NEL */
2286          case 0x2028:    /* LINE SEPARATOR */
2287          case 0x2029:    /* PARAGRAPH SEPARATOR */
2288          break;
2289          }
2290      ecode++;      ecode++;
2291      break;      break;
2292    
# Line 1420  for (;;) Line 2296  for (;;)
2296    
2297      case OP_PROP:      case OP_PROP:
2298      case OP_NOTPROP:      case OP_NOTPROP:
2299      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2300          {
2301          SCHECK_PARTIAL();
2302          MRRETURN(MATCH_NOMATCH);
2303          }
2304      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2305        {        {
2306        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2307    
2308        switch(ecode[1])        switch(ecode[1])
2309          {          {
2310          case PT_ANY:          case PT_ANY:
2311          if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
2312          break;          break;
2313    
2314          case PT_LAMP:          case PT_LAMP:
2315          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2316               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2317               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2318            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2319           break;          break;
2320    
2321          case PT_GC:          case PT_GC:
2322          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
2323            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2324          break;          break;
2325    
2326          case PT_PC:          case PT_PC:
2327          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2328            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2329          break;          break;
2330    
2331          case PT_SC:          case PT_SC:
2332          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2333            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2334            break;
2335    
2336            /* These are specials */
2337    
2338            case PT_ALNUM:
2339            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2340                 _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2341              MRRETURN(MATCH_NOMATCH);
2342            break;
2343    
2344            case PT_SPACE:    /* Perl space */
2345            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2346                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2347                   == (op == OP_NOTPROP))
2348              MRRETURN(MATCH_NOMATCH);
2349            break;
2350    
2351            case PT_PXSPACE:  /* POSIX space */
2352            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2353                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2354                 c == CHAR_FF || c == CHAR_CR)
2355                   == (op == OP_NOTPROP))
2356              MRRETURN(MATCH_NOMATCH);
2357            break;
2358    
2359            case PT_WORD:
2360            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2361                 _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2362                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2363              MRRETURN(MATCH_NOMATCH);
2364          break;          break;
2365    
2366            /* This should never occur */
2367    
2368          default:          default:
2369          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
         break;  
2370          }          }
2371    
2372        ecode += 3;        ecode += 3;
# Line 1467  for (;;) Line 2377  for (;;)
2377      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2378    
2379      case OP_EXTUNI:      case OP_EXTUNI:
2380      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2381          {
2382          SCHECK_PARTIAL();
2383          MRRETURN(MATCH_NOMATCH);
2384          }
2385      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2386        {        {
2387        int chartype, script;        int category = UCD_CATEGORY(c);
2388        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);
       if (category == ucp_M) RRETURN(MATCH_NOMATCH);  
2389        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2390          {          {
2391          int len = 1;          int len = 1;
# Line 1480  for (;;) Line 2393  for (;;)
2393            {            {
2394            GETCHARLEN(c, eptr, len);            GETCHARLEN(c, eptr, len);
2395            }            }
2396          category = _pcre_ucp_findprop(c, &chartype, &script);          category = UCD_CATEGORY(c);
2397          if (category != ucp_M) break;          if (category != ucp_M) break;
2398          eptr += len;          eptr += len;
2399          }          }
# Line 1499  for (;;) Line 2412  for (;;)
2412      loops). */      loops). */
2413    
2414      case OP_REF:      case OP_REF:
2415        case OP_REFI:
2416        caseless = op == OP_REFI;
2417        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2418        ecode += 3;
2419    
2420        /* If the reference is unset, there are two possibilities:
2421    
2422        (a) In the default, Perl-compatible state, set the length negative;
2423        this ensures that every attempt at a match fails. We can't just fail
2424        here, because of the possibility of quantifiers with zero minima.
2425    
2426        (b) If the JavaScript compatibility flag is set, set the length to zero
2427        so that the back reference matches an empty string.
2428    
2429        Otherwise, set the length to the length of what was matched by the
2430        referenced subpattern. */
2431    
2432        if (offset >= offset_top || md->offset_vector[offset] < 0)
2433          length = (md->jscript_compat)? 0 : -1;
2434        else
2435          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2436    
2437        /* Set up for repetition, or handle the non-repeated case */
2438    
2439        switch (*ecode)
2440        {        {
2441        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */        case OP_CRSTAR:
2442        ecode += 3;                                 /* Advance past item */        case OP_CRMINSTAR:
2443          case OP_CRPLUS:
2444          case OP_CRMINPLUS:
2445          case OP_CRQUERY:
2446          case OP_CRMINQUERY:
2447          c = *ecode++ - OP_CRSTAR;
2448          minimize = (c & 1) != 0;
2449          min = rep_min[c];                 /* Pick up values from tables; */
2450          max = rep_max[c];                 /* zero for max => infinity */
2451          if (max == 0) max = INT_MAX;
2452          break;
2453    
2454        /* If the reference is unset, set the length to be longer than the amount        case OP_CRRANGE:
2455        of subject left; this ensures that every attempt at a match fails. We        case OP_CRMINRANGE:
2456        can't just fail here, because of the possibility of quantifiers with zero        minimize = (*ecode == OP_CRMINRANGE);
2457        minima. */        min = GET2(ecode, 1);
2458          max = GET2(ecode, 3);
2459        length = (offset >= offset_top || md->offset_vector[offset] < 0)?        if (max == 0) max = INT_MAX;
2460          md->end_subject - eptr + 1 :        ecode += 5;
2461          md->offset_vector[offset+1] - md->offset_vector[offset];        break;
2462    
2463        /* Set up for repetition, or handle the non-repeated case */        default:               /* No repeat follows */
2464          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2465            {
2466            CHECK_PARTIAL();
2467            MRRETURN(MATCH_NOMATCH);
2468            }
2469          eptr += length;
2470          continue;              /* With the main loop */
2471          }
2472    
2473        switch (*ecode)      /* Handle repeated back references. If the length of the reference is
2474        zero, just continue with the main loop. */
2475    
2476        if (length == 0) continue;
2477    
2478        /* First, ensure the minimum number of matches are present. We get back
2479        the length of the reference string explicitly rather than passing the
2480        address of eptr, so that eptr can be a register variable. */
2481    
2482        for (i = 1; i <= min; i++)
2483          {
2484          int slength;
2485          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2486          {          {
2487          case OP_CRSTAR:          CHECK_PARTIAL();
2488          case OP_CRMINSTAR:          MRRETURN(MATCH_NOMATCH);
         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;  
   
         case OP_CRRANGE:  
         case OP_CRMINRANGE:  
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
   
         default:               /* No repeat follows */  
         if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);  
         eptr += length;  
         continue;              /* With the main loop */  
         }  
   
       /* If the length of the reference is zero, just continue with the  
       main loop. */  
   
       if (length == 0) continue;  
   
       /* First, ensure the minimum number of matches are present. We get back  
       the length of the reference string explicitly rather than passing the  
       address of eptr, so that eptr can be a register variable. */  
   
       for (i = 1; i <= min; i++)  
         {  
         if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);  
         eptr += length;  
2489          }          }
2490          eptr += slength;
2491          }
2492    
2493        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2494        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2495    
2496        if (min == max) continue;      if (min == max) continue;
2497    
2498        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2499    
2500        if (minimize)      if (minimize)
2501          {
2502          for (fi = min;; fi++)
2503          {          {
2504          for (fi = min;; fi++)          int slength;
2505            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2506            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2507            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2508            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2509            {            {
2510            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2511            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2512            }            }
2513          /* Control never gets here */          eptr += slength;
2514          }          }
2515          /* Control never gets here */
2516          }
2517    
2518        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2519    
2520        else      else
2521          {
2522          pp = eptr;
2523          for (i = min; i < max; i++)
2524          {          {
2525          pp = eptr;          int slength;
2526          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)  
2527            {            {
2528            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2529            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2530            }            }
2531          RRETURN(MATCH_NOMATCH);          eptr += slength;
2532            }
2533          while (eptr >= pp)
2534            {
2535            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2536            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2537            eptr -= length;
2538          }          }
2539          MRRETURN(MATCH_NOMATCH);
2540        }        }
2541      /* Control never gets here */      /* Control never gets here */
2542    
   
   
2543      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2544      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,
2545      and either the matching is caseful, or the characters are in the range      and either the matching is caseful, or the characters are in the range
# Line 1656  for (;;) Line 2594  for (;;)
2594          {          {
2595          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2596            {            {
2597            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2598                {
2599                SCHECK_PARTIAL();
2600                MRRETURN(MATCH_NOMATCH);
2601                }
2602            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2603            if (c > 255)            if (c > 255)
2604              {              {
2605              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2606              }              }
2607            else            else
2608              {              {
2609              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2610              }              }
2611            }            }
2612          }          }
# Line 1674  for (;;) Line 2616  for (;;)
2616          {          {
2617          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2618            {            {
2619            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2620                {
2621                SCHECK_PARTIAL();
2622                MRRETURN(MATCH_NOMATCH);
2623                }
2624            c = *eptr++;            c = *eptr++;
2625            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2626            }            }
2627          }          }
2628    
# Line 1696  for (;;) Line 2642  for (;;)
2642            {            {
2643            for (fi = min;; fi++)            for (fi = min;; fi++)
2644              {              {
2645              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2646              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2647              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2648                if (eptr >= md->end_subject)
2649                  {
2650                  SCHECK_PARTIAL();
2651                  MRRETURN(MATCH_NOMATCH);
2652                  }
2653              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2654              if (c > 255)              if (c > 255)
2655                {                {
2656                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2657                }                }
2658              else              else
2659                {                {
2660                if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);                if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2661                }                }
2662              }              }
2663            }            }
# Line 1716  for (;;) Line 2667  for (;;)
2667            {            {
2668            for (fi = min;; fi++)            for (fi = min;; fi++)
2669              {              {
2670              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2671              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2672              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2673                if (eptr >= md->end_subject)
2674                  {
2675                  SCHECK_PARTIAL();
2676                  MRRETURN(MATCH_NOMATCH);
2677                  }
2678              c = *eptr++;              c = *eptr++;
2679              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2680              }              }
2681            }            }
2682          /* Control never gets here */          /* Control never gets here */
# Line 1739  for (;;) Line 2695  for (;;)
2695            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2696              {              {
2697              int len = 1;              int len = 1;
2698              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2699                  {
2700                  SCHECK_PARTIAL();
2701                  break;
2702                  }
2703              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2704              if (c > 255)              if (c > 255)
2705                {                {
# Line 1753  for (;;) Line 2713  for (;;)
2713              }              }
2714            for (;;)            for (;;)
2715              {              {
2716              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2717              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2718              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2719              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1765  for (;;) Line 2725  for (;;)
2725            {            {
2726            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2727              {              {
2728              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2729                  {
2730                  SCHECK_PARTIAL();
2731                  break;
2732                  }
2733              c = *eptr;              c = *eptr;
2734              if ((data[c/8] & (1 << (c&7))) == 0) break;              if ((data[c/8] & (1 << (c&7))) == 0) break;
2735              eptr++;              eptr++;
2736              }              }
2737            while (eptr >= pp)            while (eptr >= pp)
2738              {              {
2739              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2740              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2741              eptr--;              eptr--;
2742              }              }
2743            }            }
2744    
2745          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2746          }          }
2747        }        }
2748      /* Control never gets here */      /* Control never gets here */
2749    
2750    
2751      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2752      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
2753        mode, because Unicode properties are supported in non-UTF-8 mode. */
2754    
2755  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2756      case OP_XCLASS:      case OP_XCLASS:
# Line 1826  for (;;) Line 2791  for (;;)
2791    
2792        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2793          {          {
2794          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2795          GETCHARINC(c, eptr);            {
2796          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2797              MRRETURN(MATCH_NOMATCH);
2798              }
2799            GETCHARINCTEST(c, eptr);
2800            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2801          }          }
2802    
2803        /* If max == min we can continue with the main loop without the        /* If max == min we can continue with the main loop without the
# Line 1843  for (;;) Line 2812  for (;;)
2812          {          {
2813          for (fi = min;; fi++)          for (fi = min;; fi++)
2814            {            {
2815            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2816            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2817            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2818            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
2819            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
2820                SCHECK_PARTIAL();
2821                MRRETURN(MATCH_NOMATCH);
2822                }
2823              GETCHARINCTEST(c, eptr);
2824              if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2825            }            }
2826          /* Control never gets here */          /* Control never gets here */
2827          }          }
# Line 1860  for (;;) Line 2834  for (;;)
2834          for (i = min; i < max; i++)          for (i = min; i < max; i++)
2835            {            {
2836            int len = 1;            int len = 1;
2837            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
2838            GETCHARLEN(c, eptr, len);              {
2839                SCHECK_PARTIAL();
2840                break;
2841                }
2842              GETCHARLENTEST(c, eptr, len);
2843            if (!_pcre_xclass(c, data)) break;            if (!_pcre_xclass(c, data)) break;
2844            eptr += len;            eptr += len;
2845            }            }
2846          for(;;)          for(;;)
2847            {            {
2848            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2849            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2850            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2851            BACKCHAR(eptr)            if (utf8) BACKCHAR(eptr);
2852            }            }
2853          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2854          }          }
2855    
2856        /* Control never gets here */        /* Control never gets here */
# Line 1888  for (;;) Line 2866  for (;;)
2866        length = 1;        length = 1;
2867        ecode++;        ecode++;
2868        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2869        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2870        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
2871            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2872            MRRETURN(MATCH_NOMATCH);
2873            }
2874          while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
2875        }        }
2876      else      else
2877  #endif  #endif
2878    
2879      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2880        {        {
2881        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2882        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);          {
2883            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2884            MRRETURN(MATCH_NOMATCH);
2885            }
2886          if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
2887        ecode += 2;        ecode += 2;
2888        }        }
2889      break;      break;
2890    
2891      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2892    
2893      case OP_CHARNC:      case OP_CHARI:
2894  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2895      if (utf8)      if (utf8)
2896        {        {
# Line 1912  for (;;) Line 2898  for (;;)
2898        ecode++;        ecode++;
2899        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2900    
2901        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2902            {
2903            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2904            MRRETURN(MATCH_NOMATCH);
2905            }
2906    
2907        /* 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
2908        can use the fast lookup table. */        can use the fast lookup table. */
2909    
2910        if (fc < 128)        if (fc < 128)
2911          {          {
2912          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2913          }          }
2914    
2915        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character */
2916    
2917        else        else
2918          {          {
2919          int dc;          unsigned int dc;
2920          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
2921          ecode += length;          ecode += length;
2922    
# Line 1936  for (;;) Line 2926  for (;;)
2926          if (fc != dc)          if (fc != dc)
2927            {            {
2928  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2929            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
2930  #endif  #endif
2931              RRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
2932            }            }
2933          }          }
2934        }        }
# Line 1947  for (;;) Line 2937  for (;;)
2937    
2938      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2939        {        {
2940        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2941        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
2942            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2943            MRRETURN(MATCH_NOMATCH);
2944            }
2945          if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2946        ecode += 2;        ecode += 2;
2947        }        }
2948      break;      break;
2949    
2950      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
2951    
2952      case OP_EXACT:      case OP_EXACT:
2953        case OP_EXACTI:
2954      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
2955      ecode += 3;      ecode += 3;
2956      goto REPEATCHAR;      goto REPEATCHAR;
2957    
2958        case OP_POSUPTO:
2959        case OP_POSUPTOI:
2960        possessive = TRUE;
2961        /* Fall through */
2962    
2963      case OP_UPTO:      case OP_UPTO:
2964        case OP_UPTOI:
2965      case OP_MINUPTO:      case OP_MINUPTO:
2966        case OP_MINUPTOI:
2967      min = 0;      min = 0;
2968      max = GET2(ecode, 1);      max = GET2(ecode, 1);
2969      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
2970      ecode += 3;      ecode += 3;
2971      goto REPEATCHAR;      goto REPEATCHAR;
2972    
2973        case OP_POSSTAR:
2974        case OP_POSSTARI:
2975        possessive = TRUE;
2976        min = 0;
2977        max = INT_MAX;
2978        ecode++;
2979        goto REPEATCHAR;
2980    
2981        case OP_POSPLUS:
2982        case OP_POSPLUSI:
2983        possessive = TRUE;
2984        min = 1;
2985        max = INT_MAX;
2986        ecode++;
2987        goto REPEATCHAR;
2988    
2989        case OP_POSQUERY:
2990        case OP_POSQUERYI:
2991        possessive = TRUE;
2992        min = 0;
2993        max = 1;
2994        ecode++;
2995        goto REPEATCHAR;
2996    
2997      case OP_STAR:      case OP_STAR:
2998        case OP_STARI:
2999      case OP_MINSTAR:      case OP_MINSTAR:
3000        case OP_MINSTARI:
3001      case OP_PLUS:      case OP_PLUS:
3002        case OP_PLUSI:
3003      case OP_MINPLUS:      case OP_MINPLUS:
3004        case OP_MINPLUSI:
3005      case OP_QUERY:      case OP_QUERY:
3006        case OP_QUERYI:
3007      case OP_MINQUERY:      case OP_MINQUERY:
3008      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3009        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3010      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3011      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3012      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3013      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3014    
3015      /* 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. */  
3016    
3017      REPEATCHAR:      REPEATCHAR:
3018  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1991  for (;;) Line 3021  for (;;)
3021        length = 1;        length = 1;
3022        charptr = ecode;        charptr = ecode;
3023        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3024        ecode += length;        ecode += length;
3025    
3026        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1999  for (;;) Line 3028  for (;;)
3028    
3029        if (length > 1)        if (length > 1)
3030          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3031  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3032          int othercase;          unsigned int othercase;
3033          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3034              (othercase = _pcre_ucp_othercase(fc)) >= 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
              othercase >= 0)  
3035            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3036            else oclength = 0;
3037  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3038    
3039          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3040            {            {
3041            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3042            /* Need braces because of following else */              memcmp(eptr, charptr, length) == 0) eptr += length;
3043            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3044              else if (oclength > 0 &&
3045                       eptr <= md->end_subject - oclength &&
3046                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3047    #endif  /* SUPPORT_UCP */
3048            else            else
3049              {              {
3050              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3051              eptr += oclength;              MRRETURN(MATCH_NOMATCH);
3052              }              }
3053            }            }
3054    
# Line 2028  for (;;) Line 3058  for (;;)
3058            {            {
3059            for (fi = min;; fi++)            for (fi = min;; fi++)
3060              {              {
3061              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3062              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3063              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3064              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3065              /* Need braces because of following else */                memcmp(eptr, charptr, length) == 0) eptr += length;
3066              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3067                else if (oclength > 0 &&
3068                         eptr <= md->end_subject - oclength &&
3069                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3070    #endif  /* SUPPORT_UCP */
3071              else              else
3072                {                {
3073                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3074                eptr += oclength;                MRRETURN(MATCH_NOMATCH);
3075                }                }
3076              }              }
3077            /* Control never gets here */            /* Control never gets here */
3078            }            }
3079          else  
3080            else  /* Maximize */
3081            {            {
3082            pp = eptr;            pp = eptr;
3083            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3084              {              {
3085              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3086              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, length) == 0) eptr += length;
3087              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3088                else if (oclength > 0 &&
3089                         eptr <= md->end_subject - oclength &&
3090                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3091    #endif  /* SUPPORT_UCP */
3092              else              else
3093                {                {
3094                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3095                eptr += oclength;                break;
3096                }                }
3097              }              }
3098            while (eptr >= pp)  
3099             {            if (possessive) continue;
3100             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3101             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3102             eptr -= length;              {
3103             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3104            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3105                if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3106    #ifdef SUPPORT_UCP
3107                eptr--;
3108                BACKCHAR(eptr);
3109    #else   /* without SUPPORT_UCP */
3110                eptr -= length;
3111    #endif  /* SUPPORT_UCP */
3112                }
3113            }            }
3114          /* Control never gets here */          /* Control never gets here */
3115          }          }
# Line 2075  for (;;) Line 3122  for (;;)
3122  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3123    
3124      /* When not in UTF-8 mode, load a single-byte character. */      /* When not in UTF-8 mode, load a single-byte character. */
3125        {  
3126        if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);      fc = *ecode++;
       fc = *ecode++;  
       }  
3127    
3128      /* 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
3129      may not be in UTF-8 mode. The code is duplicated for the caseless and      may not be in UTF-8 mode. The code is duplicated for the caseless and
# Line 2092  for (;;) Line 3137  for (;;)
3137      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3138        max, eptr));        max, eptr));
3139    
3140      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3141        {        {
3142        fc = md->lcc[fc];        fc = md->lcc[fc];
3143        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3144          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3145            if (eptr >= md->end_subject)
3146              {
3147              SCHECK_PARTIAL();
3148              MRRETURN(MATCH_NOMATCH);
3149              }
3150            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3151            }
3152        if (min == max) continue;        if (min == max) continue;
3153        if (minimize)        if (minimize)
3154          {          {
3155          for (fi = min;; fi++)          for (fi = min;; fi++)
3156            {            {
3157            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3158            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3159            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3160                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3161              RRETURN(MATCH_NOMATCH);              {
3162                SCHECK_PARTIAL();
3163                MRRETURN(MATCH_NOMATCH);
3164                }
3165              if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3166            }            }
3167          /* Control never gets here */          /* Control never gets here */
3168          }          }
3169        else        else  /* Maximize */
3170          {          {
3171          pp = eptr;          pp = eptr;
3172          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3173            {            {
3174            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3175                {
3176                SCHECK_PARTIAL();
3177                break;
3178                }
3179              if (fc != md->lcc[*eptr]) break;
3180            eptr++;            eptr++;
3181            }            }
3182    
3183            if (possessive) continue;
3184    
3185          while (eptr >= pp)          while (eptr >= pp)
3186            {            {
3187            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3188            eptr--;            eptr--;
3189            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3190            }            }
3191          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3192          }          }
3193        /* Control never gets here */        /* Control never gets here */
3194        }        }
# Line 2133  for (;;) Line 3197  for (;;)
3197    
3198      else      else
3199        {        {
3200        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3201            {
3202            if (eptr >= md->end_subject)
3203              {
3204              SCHECK_PARTIAL();
3205              MRRETURN(MATCH_NOMATCH);
3206              }
3207            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3208            }
3209    
3210        if (min == max) continue;        if (min == max) continue;
3211    
3212        if (minimize)        if (minimize)
3213          {          {
3214          for (fi = min;; fi++)          for (fi = min;; fi++)
3215            {            {
3216            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3217            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3218            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3219              RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
3220                {
3221                SCHECK_PARTIAL();
3222                MRRETURN(MATCH_NOMATCH);
3223                }
3224              if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3225            }            }
3226          /* Control never gets here */          /* Control never gets here */
3227          }          }
3228        else        else  /* Maximize */
3229          {          {
3230          pp = eptr;          pp = eptr;
3231          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3232            {            {
3233            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3234                {
3235                SCHECK_PARTIAL();
3236                break;
3237                }
3238              if (fc != *eptr) break;
3239            eptr++;            eptr++;
3240            }            }
3241            if (possessive) continue;
3242    
3243          while (eptr >= pp)          while (eptr >= pp)
3244            {            {
3245            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3246            eptr--;            eptr--;
3247            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3248            }            }
3249          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3250          }          }
3251        }        }
3252      /* Control never gets here */      /* Control never gets here */
# Line 2169  for (;;) Line 3255  for (;;)
3255      checking can be multibyte. */      checking can be multibyte. */
3256    
3257      case OP_NOT:      case OP_NOT:
3258      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3259        if (eptr >= md->end_subject)
3260          {
3261          SCHECK_PARTIAL();
3262          MRRETURN(MATCH_NOMATCH);
3263          }
3264      ecode++;      ecode++;
3265      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3266      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3267        {        {
3268  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3269        if (c < 256)        if (c < 256)
3270  #endif  #endif
3271        c = md->lcc[c];        c = md->lcc[c];
3272        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3273        }        }
3274      else      else    /* Caseful */
3275        {        {
3276        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3277        }        }
3278      break;      break;
3279    
# Line 2194  for (;;) Line 3285  for (;;)
3285      about... */      about... */
3286    
3287      case OP_NOTEXACT:      case OP_NOTEXACT:
3288        case OP_NOTEXACTI:
3289      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3290      ecode += 3;      ecode += 3;
3291      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3292    
3293      case OP_NOTUPTO:      case OP_NOTUPTO:
3294        case OP_NOTUPTOI:
3295      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3296        case OP_NOTMINUPTOI:
3297        min = 0;
3298        max = GET2(ecode, 1);
3299        minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3300        ecode += 3;
3301        goto REPEATNOTCHAR;
3302    
3303        case OP_NOTPOSSTAR:
3304        case OP_NOTPOSSTARI:
3305        possessive = TRUE;
3306        min = 0;
3307        max = INT_MAX;
3308        ecode++;
3309        goto REPEATNOTCHAR;
3310    
3311        case OP_NOTPOSPLUS:
3312        case OP_NOTPOSPLUSI:
3313        possessive = TRUE;
3314        min = 1;
3315        max = INT_MAX;
3316        ecode++;
3317        goto REPEATNOTCHAR;
3318    
3319        case OP_NOTPOSQUERY:
3320        case OP_NOTPOSQUERYI:
3321        possessive = TRUE;
3322        min = 0;
3323        max = 1;
3324        ecode++;
3325        goto REPEATNOTCHAR;
3326    
3327        case OP_NOTPOSUPTO:
3328        case OP_NOTPOSUPTOI:
3329        possessive = TRUE;
3330      min = 0;      min = 0;
3331      max = GET2(ecode, 1);      max = GET2(ecode, 1);
     minimize = *ecode == OP_NOTMINUPTO;  
3332      ecode += 3;      ecode += 3;
3333      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3334    
3335      case OP_NOTSTAR:      case OP_NOTSTAR:
3336        case OP_NOTSTARI:
3337      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3338        case OP_NOTMINSTARI:
3339      case OP_NOTPLUS:      case OP_NOTPLUS:
3340        case OP_NOTPLUSI:
3341      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3342        case OP_NOTMINPLUSI:
3343      case OP_NOTQUERY:      case OP_NOTQUERY:
3344        case OP_NOTQUERYI:
3345      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3346      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3347        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3348      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3349      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3350      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3351      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3352    
3353      /* 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. */  
3354    
3355      REPEATNOTCHAR:      REPEATNOTCHAR:
     if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3356      fc = *ecode++;      fc = *ecode++;
3357    
3358      /* The code is duplicated for the caseless and caseful cases, for speed,      /* The code is duplicated for the caseless and caseful cases, for speed,
# Line 2237  for (;;) Line 3366  for (;;)
3366      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,
3367        max, eptr));        max, eptr));
3368    
3369      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3370        {        {
3371        fc = md->lcc[fc];        fc = md->lcc[fc];
3372    
# Line 2245  for (;;) Line 3374  for (;;)
3374        /* UTF-8 mode */        /* UTF-8 mode */
3375        if (utf8)        if (utf8)
3376          {          {
3377          register int d;          register unsigned int d;
3378          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3379            {            {
3380              if (eptr >= md->end_subject)
3381                {
3382                SCHECK_PARTIAL();
3383                MRRETURN(MATCH_NOMATCH);
3384                }
3385            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3386            if (d < 256) d = md->lcc[d];            if (d < 256) d = md->lcc[d];
3387            if (fc == d) RRETURN(MATCH_NOMATCH);            if (fc == d) MRRETURN(MATCH_NOMATCH);
3388            }            }
3389          }          }
3390        else        else
# Line 2259  for (;;) Line 3393  for (;;)
3393        /* Not UTF-8 mode */        /* Not UTF-8 mode */
3394          {          {
3395          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3396            if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            {
3397              if (eptr >= md->end_subject)
3398                {
3399                SCHECK_PARTIAL();
3400                MRRETURN(MATCH_NOMATCH);
3401                }
3402              if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3403              }
3404          }          }
3405    
3406        if (min == max) continue;        if (min == max) continue;
# Line 2270  for (;;) Line 3411  for (;;)
3411          /* UTF-8 mode */          /* UTF-8 mode */
3412          if (utf8)          if (utf8)
3413            {            {
3414            register int d;            register unsigned int d;
3415            for (fi = min;; fi++)            for (fi = min;; fi++)
3416              {              {
3417              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3418              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3419                if (fi >= max) MRRETURN(MATCH_NOMATCH);
3420                if (eptr >= md->end_subject)
3421                  {
3422                  SCHECK_PARTIAL();
3423                  MRRETURN(MATCH_NOMATCH);
3424                  }
3425              GETCHARINC(d, eptr);              GETCHARINC(d, eptr);
3426              if (d < 256) d = md->lcc[d];              if (d < 256) d = md->lcc[d];
3427              if (fi >= max || eptr >= md->end_subject || fc == d)              if (fc == d) MRRETURN(MATCH_NOMATCH);
               RRETURN(MATCH_NOMATCH);  
3428              }              }
3429            }            }
3430          else          else
# Line 2287  for (;;) Line 3433  for (;;)
3433            {            {
3434            for (fi = min;; fi++)            for (fi = min;; fi++)
3435              {              {
3436              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3437              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3438              if (fi >= max || eptr >= md->end_subject || fc == md->lcc[*eptr++])              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3439                RRETURN(MATCH_NOMATCH);              if (eptr >= md->end_subject)
3440                  {