/[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 85 by nigel, Sat Feb 24 21:41:13 2007 UTC revision 623 by ph10, Tue Jul 19 09:58:42 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) 1997-2005 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
43  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK md             /* Block containing newline information */
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;  
   const uschar *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 const uschar *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  const uschar *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 169  return TRUE; Line 234  return TRUE;
234  ****************************************************************************  ****************************************************************************
235                     RECURSION IN THE match() FUNCTION                     RECURSION IN THE match() FUNCTION
236    
237  The match() function is highly recursive. Some regular expressions can cause  The match() function is highly recursive, though not every recursive call
238  it to recurse thousands of times. I was writing for Unix, so I just let it  increases the recursive depth. Nevertheless, some regular expressions can cause
239  call itself recursively. This uses the stack for saving everything that has  it to recurse to a great depth. I was writing for Unix, so I just let it call
240  to be saved for a recursive call. On Unix, the stack can be large, and this  itself recursively. This uses the stack for saving everything that has to be
241  works fine.  saved for a recursive call. On Unix, the stack can be large, and this works
242    fine.
243  It turns out that on non-Unix systems there are problems with programs that  
244  use a lot of stack. (This despite the fact that every last chip has oodles  It turns out that on some non-Unix-like systems there are problems with
245  of memory these days, and techniques for extending the stack have been known  programs that use a lot of stack. (This despite the fact that every last chip
246  for decades.) So....  has oodles of memory these days, and techniques for extending the stack have
247    been known for decades.) So....
248    
249  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
250  calls by keeping local variables that need to be preserved in blocks of memory  calls by keeping local variables that need to be preserved in blocks of memory
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  /* These versions of the macros use the stack, as normal */  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
275           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
276           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
277           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
278           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
279           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
280           RM61,  RM62, RM63 };
281    
282    /* These versions of the macros use the stack, as normal. There are debugging
283    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  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg)  
289    #ifdef PCRE_DEBUG
290    #define RMATCH(ra,rb,rc,rd,re,rw) \
291      { \
292      printf("match() called in line %d\n", __LINE__); \
293      rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1); \
294      printf("to line %d\n", __LINE__); \
295      }
296    #define RRETURN(ra) \
297      { \
298      printf("match() returned %d from line %d ", ra, __LINE__); \
299      return ra; \
300      }
301    #else
302    #define RMATCH(ra,rb,rc,rd,re,rw) \
303      rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1)
304  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
305    #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->Xprevframe = frame;\    newframe->Xrdepth = frame->Xrdepth + 1;\
328      frame = newframe;\    newframe->Xprevframe = frame;\
329      DPRINTF(("restarting from line %d\n", __LINE__));\    frame = newframe;\
330      goto HEAP_RECURSE;\    DPRINTF(("restarting from line %d\n", __LINE__));\
331      }\    goto HEAP_RECURSE;\
332    else\    L_##rw:\
333      {\    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 250  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;
363    int Xflags;    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;
386    int Xprop_fail_result;    int Xprop_fail_result;
387    int Xprop_category;    int Xoclength;
388    int Xprop_chartype;    uschar Xocchars[8];
   int Xprop_othercase;  
   int Xprop_test_against;  
   int *Xprop_test_variable;  
389  #endif  #endif
390    
391      int Xcodelink;
392    int Xctype;    int Xctype;
393    int Xfc;    unsigned int Xfc;
394    int Xfi;    int Xfi;
395    int Xlength;    int Xlength;
396    int Xmax;    int Xmax;
# Line 301  typedef struct heapframe { Line 404  typedef struct heapframe {
404    
405    eptrblock Xnewptrb;    eptrblock Xnewptrb;
406    
407    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
408    
409    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
410    
411  } heapframe;  } heapframe;
412    
# Line 320  typedef struct heapframe { Line 422  typedef struct heapframe {
422  *         Match from current position            *  *         Match from current position            *
423  *************************************************/  *************************************************/
424    
425  /* On entry ecode points to the first opcode, and eptr to the first character  /* This function is called recursively in many circumstances. Whenever it
 in the subject string, while eptrb holds the value of eptr at the start of the  
 last bracketed group - used for breaking infinite loops matching zero-length  
 strings. This function is called recursively in many circumstances. Whenever it  
426  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
427  same response.  same response. */
428    
429    /* These macros pack up tests that are used for partial matching, and which
430    appears several times in the code. We set the "hit end" flag if the pointer is
431    at the end of the subject and also past the start of the subject (i.e.
432    something has been matched). For hard partial matching, we then return
433    immediately. The second one is used when we already know we are past the end of
434    the subject. */
435    
436    #define CHECK_PARTIAL()\
437      if (md->partial != 0 && eptr >= md->end_subject && \
438          eptr > md->start_used_ptr) \
439        { \
440        md->hitend = TRUE; \
441        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
442        }
443    
444    #define SCHECK_PARTIAL()\
445      if (md->partial != 0 && eptr > md->start_used_ptr) \
446        { \
447        md->hitend = TRUE; \
448        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
449        }
450    
451    
452  Performance note: It might be tempting to extract commonly used fields from the  /* Performance note: It might be tempting to extract commonly used fields from
453  md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf8, end_subject) into individual variables to improve
454  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
455  made performance worse.  made performance worse.
456    
457  Arguments:  Arguments:
458     eptr        pointer in subject     eptr        pointer to current character in subject
459     ecode       position in code     ecode       pointer to current position in compiled code
460       mstart      pointer to the current match start position (can be modified
461                     by encountering \K)
462       markptr     pointer to the most recent MARK name, or NULL
463     offset_top  current top pointer     offset_top  current top pointer
464     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
465     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
466                   brackets - for testing for empty matches                   brackets - for testing for empty matches
467     flags       can contain     rdepth      the recursion depth
                  match_condassert - this is an assertion condition  
                  match_isgroup - this is the start of a bracketed group  
468    
469  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
470                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
471                   a negative MATCH_xxx value for PRUNE, SKIP, etc
472                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
473                   (e.g. stopped by recursion limit)                   (e.g. stopped by repeated call or recursion limit)
474  */  */
475    
476  static int  static int
477  match(REGISTER const uschar *eptr, REGISTER const uschar *ecode,  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
478    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    const uschar *markptr, int offset_top, match_data *md, eptrblock *eptrb,
479    int flags)    unsigned int rdepth)
480  {  {
481  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
482  so they can be ordinary variables in all cases. Mark them with "register"  so they can be ordinary variables in all cases. Mark some of them with
483  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
484    
485  register int  rrc;    /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
486  register int  i;      /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
487  register int  c;      /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
488  register BOOL utf8;   /* Local copy of UTF-8 flag for speed */  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */
489    
490    BOOL minimize, possessive; /* Quantifier options */
491    BOOL caseless;
492    int condcode;
493    
494  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
495  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame" which is obtained from
# Line 370  heap storage. Set up the top-level frame Line 497  heap storage. Set up the top-level frame
497  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
498    
499  #ifdef NO_RECURSE  #ifdef NO_RECURSE
500  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));
501    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
502  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
503    
504  /* Copy in the original argument variables */  /* Copy in the original argument variables */
505    
506  frame->Xeptr = eptr;  frame->Xeptr = eptr;
507  frame->Xecode = ecode;  frame->Xecode = ecode;
508    frame->Xmstart = mstart;
509    frame->Xmarkptr = markptr;
510  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
511  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
512  frame->Xflags = flags;  frame->Xrdepth = rdepth;
513    
514  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
515    
# Line 390  HEAP_RECURSE: Line 519  HEAP_RECURSE:
519    
520  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
521  #define ecode              frame->Xecode  #define ecode              frame->Xecode
522    #define mstart             frame->Xmstart
523    #define markptr            frame->Xmarkptr
524  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
525  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
526  #define flags              frame->Xflags  #define rdepth             frame->Xrdepth
527    
528  /* Ditto for the local variables */  /* Ditto for the local variables */
529    
# Line 401  HEAP_RECURSE: Line 531  HEAP_RECURSE:
531  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
532  #endif  #endif
533  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
534    #define codelink           frame->Xcodelink
535  #define data               frame->Xdata  #define data               frame->Xdata
536  #define next               frame->Xnext  #define next               frame->Xnext
537  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 411  HEAP_RECURSE: Line 542  HEAP_RECURSE:
542    
543  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
544  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
545  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
546    
 #define original_ims       frame->Xoriginal_ims  
   
547  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
548  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
549    #define prop_value         frame->Xprop_value
550  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
551  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
552  #define prop_chartype      frame->Xprop_chartype  #define occhars            frame->Xocchars
 #define prop_othercase     frame->Xprop_othercase  
 #define prop_test_against  frame->Xprop_test_against  
 #define prop_test_variable frame->Xprop_test_variable  
553  #endif  #endif
554    
555  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 447  HEAP_RECURSE: Line 573  HEAP_RECURSE:
573  get preserved during recursion in the normal way. In this environment, fi and  get preserved during recursion in the normal way. In this environment, fi and
574  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
575    
576  #else  #else         /* NO_RECURSE not defined */
577  #define fi i  #define fi i
578  #define fc c  #define fc c
579    
580    /* Many of the following variables are used only in small blocks of the code.
581  #ifdef SUPPORT_UTF8                /* Many of these variables are used ony */  My normal style of coding would have declared them within each of those blocks.
582  const uschar *charptr;             /* small blocks of the code. My normal  */  However, in order to accommodate the version of this code that uses an external
583  #endif                             /* style of coding would have declared  */  "stack" implemented on the heap, it is easier to declare them all here, so the
584  const uschar *callpat;             /* them within each of those blocks.    */  declarations can be cut out in a block. The only declarations within blocks
585  const uschar *data;                /* However, in order to accommodate the */  below are for variables that do not have to be preserved over a recursive call
586  const uschar *next;                /* version of this code that uses an    */  to RMATCH(). */
587  const uschar *pp;                  /* external "stack" implemented on the  */  
588  const uschar *prev;                /* heap, it is easier to declare them   */  #ifdef SUPPORT_UTF8
589  const uschar *saved_eptr;          /* all here, so the declarations can    */  const uschar *charptr;
590                                     /* be cut out in a block. The only      */  #endif
591  recursion_info new_recursive;      /* declarations within blocks below are */  const uschar *callpat;
592                                     /* for variables that do not have to    */  const uschar *data;
593  BOOL cur_is_word;                  /* be preserved over a recursive call   */  const uschar *next;
594  BOOL condition;                    /* to RMATCH().                         */  USPTR         pp;
595  BOOL minimize;  const uschar *prev;
596    USPTR         saved_eptr;
597    
598    recursion_info new_recursive;
599    
600    BOOL cur_is_word;
601    BOOL condition;
602  BOOL prev_is_word;  BOOL prev_is_word;
603    
 unsigned long int original_ims;  
   
604  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
605  int prop_type;  int prop_type;
606    int prop_value;
607  int prop_fail_result;  int prop_fail_result;
608  int prop_category;  int oclength;
609  int prop_chartype;  uschar occhars[8];
 int prop_othercase;  
 int prop_test_against;  
 int *prop_test_variable;  
610  #endif  #endif
611    
612    int codelink;
613  int ctype;  int ctype;
614  int length;  int length;
615  int max;  int max;
# Line 493  int save_offset1, save_offset2, save_off Line 622  int save_offset1, save_offset2, save_off
622  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
623    
624  eptrblock newptrb;  eptrblock newptrb;
625  #endif  #endif     /* NO_RECURSE */
626    
627    /* To save space on the stack and in the heap frame, I have doubled up on some
628    of the local variables that are used only in localised parts of the code, but
629    still need to be preserved over recursive calls of match(). These macros define
630    the alternative names that are used. */
631    
632    #define allow_zero    cur_is_word
633    #define cbegroup      condition
634    #define code_offset   codelink
635    #define condassert    condition
636    #define matched_once  prev_is_word
637    
638  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
639  variables. */  variables. */
640    
641  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
642    prop_value = 0;
643  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_against = 0;  
 prop_test_variable = NULL;  
644  #endif  #endif
645    
 /* OK, now we can get on with the real code of the function. Recursion is  
 specified by the macros RMATCH and RRETURN. When NO_RECURSE is *not* defined,  
 these just turn into a recursive call to match() and a "return", respectively.  
 However, RMATCH isn't like a function call because it's quite a complicated  
 macro. It has to be used in one particular way. This shouldn't, however, impact  
 performance when true recursion is being used. */  
646    
647  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  /* This label is used for tail recursion, which is used in a few cases even
648    when NO_RECURSE is not defined, in order to reduce the amount of stack that is
649    used. Thanks to Ian Taylor for noticing this possibility and sending the
650    original patch. */
651    
652    TAIL_RECURSE:
653    
654    /* OK, now we can get on with the real code of the function. Recursive calls
655    are specified by the macro RMATCH and RRETURN is used to return. When
656    NO_RECURSE is *not* defined, these just turn into a recursive call to match()
657    and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
658    defined). However, RMATCH isn't like a function call because it's quite a
659    complicated macro. It has to be used in one particular way. This shouldn't,
660    however, impact performance when true recursion is being used. */
661    
662  original_ims = ims;    /* Save for resetting on ')' */  #ifdef SUPPORT_UTF8
663  utf8 = md->utf8;       /* Local copy of the flag */  utf8 = md->utf8;       /* Local copy of the flag */
664    #else
665    utf8 = FALSE;
666    #endif
667    
668    /* First check that we haven't called match() too many times, or that we
669    haven't exceeded the recursive call limit. */
670    
671    if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
672    if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
673    
674  /* At the start of a bracketed group, add the current subject pointer to the  /* At the start of a group with an unlimited repeat that may match an empty
675  stack of such pointers, to be re-instated at the end of the group when we hit  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
676  the closing ket. When match() is called in other circumstances, we don't add to  done this way to save having to use another function argument, which would take
677  this stack. */  up space on the stack. See also MATCH_CONDASSERT below.
678    
679    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
680    such remembered pointers, to be checked when we hit the closing ket, in order
681    to break infinite loops that match no characters. When match() is called in
682    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
683    NOT be used with tail recursion, because the memory block that is used is on
684    the stack, so a new one may be required for each match(). */
685    
686  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
687    {    {
   newptrb.epb_prev = eptrb;  
688    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
689      newptrb.epb_prev = eptrb;
690    eptrb = &newptrb;    eptrb = &newptrb;
691      md->match_function_type = 0;
692    }    }
693    
694  /* Now start processing the operations. */  /* Now start processing the opcodes. */
695    
696  for (;;)  for (;;)
697    {    {
698      minimize = possessive = FALSE;
699    op = *ecode;    op = *ecode;
700    minimize = FALSE;  
701      switch(op)
702    /* For partial matching, remember if we ever hit the end of the subject after      {
703    matching at least one subject character. */      case OP_MARK:
704        markptr = ecode + 2;
705        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
706          eptrb, RM55);
707    
708        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
709        argument, and we must check whether that argument matches this MARK's
710        argument. It is passed back in md->start_match_ptr (an overloading of that
711        variable). If it does match, we reset that variable to the current subject
712        position and return MATCH_SKIP. Otherwise, pass back the return code
713        unaltered. */
714    
715        if (rrc == MATCH_SKIP_ARG &&
716            strcmp((char *)markptr, (char *)(md->start_match_ptr)) == 0)
717          {
718          md->start_match_ptr = eptr;
719          RRETURN(MATCH_SKIP);
720          }
721    
722        if (md->mark == NULL) md->mark = markptr;
723        RRETURN(rrc);
724    
725        case OP_FAIL:
726        MRRETURN(MATCH_NOMATCH);
727    
728        /* COMMIT overrides PRUNE, SKIP, and THEN */
729    
730        case OP_COMMIT:
731        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
732          eptrb, RM52);
733        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
734            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
735            rrc != MATCH_THEN)
736          RRETURN(rrc);
737        MRRETURN(MATCH_COMMIT);
738    
739    if (md->partial &&      /* PRUNE overrides THEN */
       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. */  
740    
741    if (op > OP_BRA)      case OP_PRUNE:
742      {      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
743      number = op - OP_BRA;        eptrb, RM51);
744        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
745        MRRETURN(MATCH_PRUNE);
746    
747        case OP_PRUNE_ARG:
748        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
749          eptrb, RM56);
750        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
751        md->mark = ecode + 2;
752        RRETURN(MATCH_PRUNE);
753    
754        /* SKIP overrides PRUNE and THEN */
755    
756        case OP_SKIP:
757        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
758          eptrb, RM53);
759        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
760          RRETURN(rrc);
761        md->start_match_ptr = eptr;   /* Pass back current position */
762        MRRETURN(MATCH_SKIP);
763    
764      /* For extended extraction brackets (large number), we have to fish out the      case OP_SKIP_ARG:
765      number from a dummy opcode at the start. */      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,
766          eptrb, RM57);
767        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
768          RRETURN(rrc);
769    
770      if (number > EXTRACT_BASIC_MAX)      /* Pass back the current skip name by overloading md->start_match_ptr and
771        number = GET2(ecode, 2+LINK_SIZE);      returning the special MATCH_SKIP_ARG return code. This will either be
772        caught by a matching MARK, or get to the top, where it is treated the same
773        as PRUNE. */
774    
775        md->start_match_ptr = ecode + 2;
776        RRETURN(MATCH_SKIP_ARG);
777    
778        /* For THEN (and THEN_ARG) we pass back the address of the bracket or
779        the alt that is at the start of the current branch. This makes it possible
780        to skip back past alternatives that precede the THEN within the current
781        branch. */
782    
783        case OP_THEN:
784        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
785          eptrb, RM54);
786        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
787        md->start_match_ptr = ecode - GET(ecode, 1);
788        MRRETURN(MATCH_THEN);
789    
790        case OP_THEN_ARG:
791        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1+LINK_SIZE],
792          offset_top, md, eptrb, RM58);
793        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794        md->start_match_ptr = ecode - GET(ecode, 1);
795        md->mark = ecode + LINK_SIZE + 2;
796        RRETURN(MATCH_THEN);
797    
798        /* Handle a capturing bracket, other than those that are possessive with an
799        unlimited repeat. If there is space in the offset vector, save the current
800        subject position in the working slot at the top of the vector. We mustn't
801        change the current values of the data slot, because they may be set from a
802        previous iteration of this group, and be referred to by a reference inside
803        the group. A failure to match might occur after the group has succeeded,
804        if something later on doesn't match. For this reason, we need to restore
805        the working value and also the values of the final offsets, in case they
806        were set by a previous iteration of the same bracket.
807    
808        If there isn't enough space in the offset vector, treat this as if it were
809        a non-capturing bracket. Don't worry about setting the flag for the error
810        case here; that is handled in the code for KET. */
811    
812        case OP_CBRA:
813        case OP_SCBRA:
814        number = GET2(ecode, 1+LINK_SIZE);
815      offset = number << 1;      offset = number << 1;
816    
817  #ifdef DEBUG  #ifdef PCRE_DEBUG
818      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
819        printf("subject=");
820      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
821      printf("\n");      printf("\n");
822  #endif  #endif
# Line 582  for (;;) Line 829  for (;;)
829        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
830    
831        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
832        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
833            (int)(eptr - md->start_subject);
834    
835        do        for (;;)
836          {          {
837          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
838            match_isgroup);          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
839          if (rrc != MATCH_NOMATCH) RRETURN(rrc);            eptrb, RM1);
840            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
841            if (rrc != MATCH_NOMATCH &&
842                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
843              RRETURN(rrc);
844          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
845          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
846            if (*ecode != OP_ALT) break;
847          }          }
       while (*ecode == OP_ALT);  
848    
849        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
850        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
851        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
852        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
853    
854          /* At this point, rrc will be one of MATCH_ONCE, MATCH_NOMATCH, or
855          MATCH_THEN. */
856    
857          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
858          RRETURN(((rrc == MATCH_ONCE)? MATCH_ONCE:MATCH_NOMATCH));
859          }
860    
861        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
862        as a non-capturing bracket. */
863    
864        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
865        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
866    
867        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
868    
869        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
870        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
871    
872        /* Non-capturing or atomic group, except for possessive with unlimited
873        repeat. Loop for all the alternatives. When we get to the final alternative
874        within the brackets, we used to return the result of a recursive call to
875        match() whatever happened so it was possible to reduce stack usage by
876        turning this into a tail recursion, except in the case of a possibly empty
877        group. However, now that there is the possiblity of (*THEN) occurring in
878        the final alternative, this optimization is no longer possible.
879    
880        MATCH_ONCE is returned when the end of an atomic group is successfully
881        reached, but subsequent matching fails. It passes back up the tree (causing
882        captured values to be reset) until the original atomic group level is
883        reached. This is tested by comparing md->once_target with the start of the
884        group. At this point, the return is converted into MATCH_NOMATCH so that
885        previous backup points can be taken. */
886    
887        RRETURN(MATCH_NOMATCH);      case OP_ONCE:
888        case OP_BRA:
889        case OP_SBRA:
890        DPRINTF(("start non-capturing bracket\n"));
891    
892        for (;;)
893          {
894          if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
895          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, eptrb,
896            RM2);
897          if (rrc != MATCH_NOMATCH &&
898              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
899            {
900            if (rrc == MATCH_ONCE)
901              {
902              const uschar *scode = ecode;
903              if (*scode != OP_ONCE)           /* If not at start, find it */
904                {
905                while (*scode == OP_ALT) scode += GET(scode, 1);
906                scode -= GET(scode, 1);
907                }
908              if (md->once_target == scode) rrc = MATCH_NOMATCH;
909              }
910            RRETURN(rrc);
911            }
912          ecode += GET(ecode, 1);
913          if (*ecode != OP_ALT) break;
914        }        }
915        if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
916        RRETURN(MATCH_NOMATCH);
917    
918      /* Insufficient room for saving captured contents */      /* Handle possessive capturing brackets with an unlimited repeat. We come
919        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
920      else op = OP_BRA;      handled similarly to the normal case above. However, the matching is
921      }      different. The end of these brackets will always be OP_KETRPOS, which
922        returns MATCH_KETRPOS without going further in the pattern. By this means
923        we can handle the group by iteration rather than recursion, thereby
924        reducing the amount of stack needed. */
925    
926        case OP_CBRAPOS:
927        case OP_SCBRAPOS:
928        allow_zero = FALSE;
929    
930        POSSESSIVE_CAPTURE:
931        number = GET2(ecode, 1+LINK_SIZE);
932        offset = number << 1;
933    
934    /* Other types of node can be handled by a switch */  #ifdef PCRE_DEBUG
935        printf("start possessive bracket %d\n", number);
936        printf("subject=");
937        pchars(eptr, 16, TRUE, md);
938        printf("\n");
939    #endif
940    
941    switch(op)      if (offset < md->offset_max)
     {  
     case OP_BRA:     /* Non-capturing bracket: optimized */  
     DPRINTF(("start bracket 0\n"));  
     do  
942        {        {
943        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        matched_once = FALSE;
944          match_isgroup);        code_offset = ecode - md->start_code;
945        if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
946          save_offset1 = md->offset_vector[offset];
947          save_offset2 = md->offset_vector[offset+1];
948          save_offset3 = md->offset_vector[md->offset_end - number];
949          save_capture_last = md->capture_last;
950    
951          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
952    
953          /* Each time round the loop, save the current subject position for use
954          when the group matches. For MATCH_MATCH, the group has matched, so we
955          restart it with a new subject starting position, remembering that we had
956          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
957          usual. If we haven't matched any alternatives in any iteration, check to
958          see if a previous iteration matched. If so, the group has matched;
959          continue from afterwards. Otherwise it has failed; restore the previous
960          capture values before returning NOMATCH. */
961    
962          for (;;)
963            {
964            md->offset_vector[md->offset_end - number] =
965              (int)(eptr - md->start_subject);
966            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
967            RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
968              eptrb, RM63);
969            if (rrc == MATCH_KETRPOS)
970              {
971              offset_top = md->end_offset_top;
972              eptr = md->end_match_ptr;
973              ecode = md->start_code + code_offset;
974              save_capture_last = md->capture_last;
975              matched_once = TRUE;
976              continue;
977              }
978            if (rrc != MATCH_NOMATCH &&
979                (rrc != MATCH_THEN || md->start_match_ptr != ecode))
980              RRETURN(rrc);
981            md->capture_last = save_capture_last;
982            ecode += GET(ecode, 1);
983            if (*ecode != OP_ALT) break;
984            }
985    
986          if (!matched_once)
987            {
988            md->offset_vector[offset] = save_offset1;
989            md->offset_vector[offset+1] = save_offset2;
990            md->offset_vector[md->offset_end - number] = save_offset3;
991            }
992    
993          if (rrc != MATCH_THEN && md->mark == NULL) md->mark = markptr;
994          if (allow_zero || matched_once)
995            {
996            ecode += 1 + LINK_SIZE;
997            break;
998            }
999    
1000          RRETURN(MATCH_NOMATCH);
1001          }
1002    
1003        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1004        as a non-capturing bracket. */
1005    
1006        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1007        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1008    
1009        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1010    
1011        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1012        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1013    
1014        /* Non-capturing possessive bracket with unlimited repeat. We come here
1015        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1016        without the capturing complication. It is written out separately for speed
1017        and cleanliness. */
1018    
1019        case OP_BRAPOS:
1020        case OP_SBRAPOS:
1021        allow_zero = FALSE;
1022    
1023        POSSESSIVE_NON_CAPTURE:
1024        matched_once = FALSE;
1025        code_offset = ecode - md->start_code;
1026    
1027        for (;;)
1028          {
1029          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1030          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
1031            eptrb, RM48);
1032          if (rrc == MATCH_KETRPOS)
1033            {
1034            offset_top = md->end_offset_top;
1035            eptr = md->end_match_ptr;
1036            ecode = md->start_code + code_offset;
1037            matched_once = TRUE;
1038            continue;
1039            }
1040          if (rrc != MATCH_NOMATCH &&
1041              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1042            RRETURN(rrc);
1043        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1044          if (*ecode != OP_ALT) break;
1045        }        }
1046      while (*ecode == OP_ALT);  
1047      DPRINTF(("bracket 0 failed\n"));      if (matched_once || allow_zero)
1048          {
1049          ecode += 1 + LINK_SIZE;
1050          break;
1051          }
1052      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1053    
1054        /* Control never reaches here. */
1055    
1056      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1057      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1058      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1059      exactly what going to the ket would do. */      exactly what going to the ket would do. */
1060    
1061      case OP_COND:      case OP_COND:
1062      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      case OP_SCOND:
1063        codelink = GET(ecode, 1);
1064    
1065        /* Because of the way auto-callout works during compile, a callout item is
1066        inserted between OP_COND and an assertion condition. */
1067    
1068        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1069          {
1070          if (pcre_callout != NULL)
1071            {
1072            pcre_callout_block cb;
1073            cb.version          = 1;   /* Version 1 of the callout block */
1074            cb.callout_number   = ecode[LINK_SIZE+2];
1075            cb.offset_vector    = md->offset_vector;
1076            cb.subject          = (PCRE_SPTR)md->start_subject;
1077            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1078            cb.start_match      = (int)(mstart - md->start_subject);
1079            cb.current_position = (int)(eptr - md->start_subject);
1080            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1081            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1082            cb.capture_top      = offset_top/2;
1083            cb.capture_last     = md->capture_last;
1084            cb.callout_data     = md->callout_data;
1085            if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1086            if (rrc < 0) RRETURN(rrc);
1087            }
1088          ecode += _pcre_OP_lengths[OP_CALLOUT];
1089          }
1090    
1091        condcode = ecode[LINK_SIZE+1];
1092    
1093        /* Now see what the actual condition is */
1094    
1095        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1096          {
1097          if (md->recursive == NULL)                /* Not recursing => FALSE */
1098            {
1099            condition = FALSE;
1100            ecode += GET(ecode, 1);
1101            }
1102          else
1103            {
1104            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1105            condition =  (recno == RREF_ANY || recno == md->recursive->group_num);
1106    
1107            /* If the test is for recursion into a specific subpattern, and it is
1108            false, but the test was set up by name, scan the table to see if the
1109            name refers to any other numbers, and test them. The condition is true
1110            if any one is set. */
1111    
1112            if (!condition && condcode == OP_NRREF && recno != RREF_ANY)
1113              {
1114              uschar *slotA = md->name_table;
1115              for (i = 0; i < md->name_count; i++)
1116                {
1117                if (GET2(slotA, 0) == recno) break;
1118                slotA += md->name_entry_size;
1119                }
1120    
1121              /* Found a name for the number - there can be only one; duplicate
1122              names for different numbers are allowed, but not vice versa. First
1123              scan down for duplicates. */
1124    
1125              if (i < md->name_count)
1126                {
1127                uschar *slotB = slotA;
1128                while (slotB > md->name_table)
1129                  {
1130                  slotB -= md->name_entry_size;
1131                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1132                    {
1133                    condition = GET2(slotB, 0) == md->recursive->group_num;
1134                    if (condition) break;
1135                    }
1136                  else break;
1137                  }
1138    
1139                /* Scan up for duplicates */
1140    
1141                if (!condition)
1142                  {
1143                  slotB = slotA;
1144                  for (i++; i < md->name_count; i++)
1145                    {
1146                    slotB += md->name_entry_size;
1147                    if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1148                      {
1149                      condition = GET2(slotB, 0) == md->recursive->group_num;
1150                      if (condition) break;
1151                      }
1152                    else break;
1153                    }
1154                  }
1155                }
1156              }
1157    
1158            /* Chose branch according to the condition */
1159    
1160            ecode += condition? 3 : GET(ecode, 1);
1161            }
1162          }
1163    
1164        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1165        {        {
1166        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1167        condition = (offset == CREF_RECURSE * 2)?        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1168          (md->recursive != NULL) :  
1169          (offset < offset_top && md->offset_vector[offset] >= 0);        /* If the numbered capture is unset, but the reference was by name,
1170        RMATCH(rrc, eptr, ecode + (condition?        scan the table to see if the name refers to any other numbers, and test
1171          (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1))),        them. The condition is true if any one is set. This is tediously similar
1172          offset_top, md, ims, eptrb, match_isgroup);        to the code above, but not close enough to try to amalgamate. */
1173        RRETURN(rrc);  
1174          if (!condition && condcode == OP_NCREF)
1175            {
1176            int refno = offset >> 1;
1177            uschar *slotA = md->name_table;
1178    
1179            for (i = 0; i < md->name_count; i++)
1180              {
1181              if (GET2(slotA, 0) == refno) break;
1182              slotA += md->name_entry_size;
1183              }
1184    
1185            /* Found a name for the number - there can be only one; duplicate names
1186            for different numbers are allowed, but not vice versa. First scan down
1187            for duplicates. */
1188    
1189            if (i < md->name_count)
1190              {
1191              uschar *slotB = slotA;
1192              while (slotB > md->name_table)
1193                {
1194                slotB -= md->name_entry_size;
1195                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1196                  {
1197                  offset = GET2(slotB, 0) << 1;
1198                  condition = offset < offset_top &&
1199                    md->offset_vector[offset] >= 0;
1200                  if (condition) break;
1201                  }
1202                else break;
1203                }
1204    
1205              /* Scan up for duplicates */
1206    
1207              if (!condition)
1208                {
1209                slotB = slotA;
1210                for (i++; i < md->name_count; i++)
1211                  {
1212                  slotB += md->name_entry_size;
1213                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)
1214                    {
1215                    offset = GET2(slotB, 0) << 1;
1216                    condition = offset < offset_top &&
1217                      md->offset_vector[offset] >= 0;
1218                    if (condition) break;
1219                    }
1220                  else break;
1221                  }
1222                }
1223              }
1224            }
1225    
1226          /* Chose branch according to the condition */
1227    
1228          ecode += condition? 3 : GET(ecode, 1);
1229          }
1230    
1231        else if (condcode == OP_DEF)     /* DEFINE - always false */
1232          {
1233          condition = FALSE;
1234          ecode += GET(ecode, 1);
1235        }        }
1236    
1237      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1238      the final argument 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
1239        an assertion. */
1240    
1241      else      else
1242        {        {
1243        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1244            match_condassert | match_isgroup);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1245        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1246          {          {
1247          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          if (md->end_offset_top > offset_top)
1248              offset_top = md->end_offset_top;  /* Captures may have happened */
1249            condition = TRUE;
1250            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1251          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1252          }          }
1253        else if (rrc != MATCH_NOMATCH)        else if (rrc != MATCH_NOMATCH &&
1254                  (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1255          {          {
1256          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1257          }          }
1258        else ecode += GET(ecode, 1);        else
1259        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          {
1260          match_isgroup);          condition = FALSE;
1261            ecode += codelink;
1262            }
1263          }
1264    
1265        /* We are now at the branch that is to be obeyed. As there is only one,
1266        we used to use tail recursion to avoid using another stack frame, except
1267        when there was unlimited repeat of a possibly empty group. However, that
1268        strategy no longer works because of the possibilty of (*THEN) being
1269        encountered in the branch. A recursive call to match() is always required,
1270        unless the second alternative doesn't exist, in which case we can just
1271        plough on. */
1272    
1273        if (condition || *ecode == OP_ALT)
1274          {
1275          if (op == OP_SCOND) md->match_function_type = MATCH_CBEGROUP;
1276          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1277          if (rrc == MATCH_THEN && md->start_match_ptr == ecode)
1278            rrc = MATCH_NOMATCH;
1279        RRETURN(rrc);        RRETURN(rrc);
1280        }        }
1281      /* Control never reaches here */      else                         /* Condition false & no alternative */
1282          {
1283          ecode += 1 + LINK_SIZE;
1284          }
1285        break;
1286    
1287    
1288        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1289        to close any currently open capturing brackets. */
1290    
1291      /* Skip over conditional reference or large extraction number data if      case OP_CLOSE:
1292      encountered. */      number = GET2(ecode, 1);
1293        offset = number << 1;
1294    
1295    #ifdef PCRE_DEBUG
1296          printf("end bracket %d at *ACCEPT", number);
1297          printf("\n");
1298    #endif
1299    
1300      case OP_CREF:      md->capture_last = number;
1301      case OP_BRANUMBER:      if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1302          {
1303          md->offset_vector[offset] =
1304            md->offset_vector[md->offset_end - number];
1305          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1306          if (offset_top <= offset) offset_top = offset + 2;
1307          }
1308      ecode += 3;      ecode += 3;
1309      break;      break;
1310    
1311      /* End of the pattern. If we are in a recursion, we should restore the  
1312      offsets appropriately and continue from after the call. */      /* End of the pattern, either real or forced. */
1313    
1314      case OP_END:      case OP_END:
1315      if (md->recursive != NULL && md->recursive->group_num == 0)      case OP_ACCEPT:
1316        {      case OP_ASSERT_ACCEPT:
1317        recursion_info *rec = md->recursive;  
1318        DPRINTF(("Hit the end in a (?0) recursion\n"));      /* If we have matched an empty string, fail if not in an assertion and not
1319        md->recursive = rec->prevrec;      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1320        memmove(md->offset_vector, rec->offset_save,      is set and we have matched at the start of the subject. In both cases,
1321          rec->saved_max * sizeof(int));      backtracking will then try other alternatives, if any. */
1322        md->start_match = rec->save_start;  
1323        ims = original_ims;      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1324        ecode = rec->after_call;           md->recursive == NULL &&
1325        break;           (md->notempty ||
1326        }             (md->notempty_atstart &&
1327                 mstart == md->start_subject + md->start_offset)))
1328          MRRETURN(MATCH_NOMATCH);
1329    
1330        /* Otherwise, we have a match. */
1331    
1332        md->end_match_ptr = eptr;           /* Record where we ended */
1333        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1334        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1335    
1336      /* 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
1337      string - backtracking will then try other alternatives, if any. */      given as the argument to MRRETURN when the heap is in use. */
1338    
1339      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1340      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;  
1341    
1342      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1343      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1344      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1345      start of each branch to move the current point backwards, so the code at      start of each branch to move the current point backwards, so the code at
1346      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1347        of a condition, we want to return immediately afterwards. The caller of
1348        this incarnation of the match() function will have set MATCH_CONDASSERT in
1349        md->match_function type, and one of these opcodes will be the first opcode
1350        that is processed. We use a local variable that is preserved over calls to
1351        match() to remember this case. */
1352    
1353      case OP_ASSERT:      case OP_ASSERT:
1354      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1355        if (md->match_function_type == MATCH_CONDASSERT)
1356          {
1357          condassert = TRUE;
1358          md->match_function_type = 0;
1359          }
1360        else condassert = FALSE;
1361    
1362      do      do
1363        {        {
1364        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1365          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1366        if (rrc == MATCH_MATCH) break;          {
1367        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          mstart = md->start_match_ptr;   /* In case \K reset it */
1368            break;
1369            }
1370          if (rrc != MATCH_NOMATCH &&
1371              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1372            RRETURN(rrc);
1373        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1374        }        }
1375      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1376      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1377        if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1378    
1379      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1380    
1381      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1382    
1383      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1384      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 738  for (;;) Line 1388  for (;;)
1388      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1389      continue;      continue;
1390    
1391      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1392        PRUNE, or COMMIT means we must assume failure without checking subsequent
1393        branches. */
1394    
1395      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1396      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1397        if (md->match_function_type == MATCH_CONDASSERT)
1398          {
1399          condassert = TRUE;
1400          md->match_function_type = 0;
1401          }
1402        else condassert = FALSE;
1403    
1404      do      do
1405        {        {
1406        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1407          match_isgroup);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1408        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1409        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          {
1410            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1411            break;
1412            }
1413          if (rrc != MATCH_NOMATCH &&
1414              (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1415            RRETURN(rrc);
1416        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1417        }        }
1418      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1419    
1420      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1421    
1422      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1423      continue;      continue;
1424    
# Line 766  for (;;) Line 1431  for (;;)
1431  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1432      if (utf8)      if (utf8)
1433        {        {
1434        c = GET(ecode,1);        i = GET(ecode, 1);
1435        for (i = 0; i < c; i++)        while (i-- > 0)
1436          {          {
1437          eptr--;          eptr--;
1438          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1439          BACKCHAR(eptr)          BACKCHAR(eptr);
1440          }          }
1441        }        }
1442      else      else
# Line 780  for (;;) Line 1445  for (;;)
1445      /* 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 */
1446    
1447        {        {
1448        eptr -= GET(ecode,1);        eptr -= GET(ecode, 1);
1449        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1450        }        }
1451    
1452      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1453    
1454        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1455      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1456      break;      break;
1457    
# Line 800  for (;;) Line 1466  for (;;)
1466        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 1;   /* Version 1 of the callout block */
1467        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1468        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1469        cb.subject          = (const char *)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1470        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1471        cb.start_match      = md->start_match - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1472        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1473        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1474        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1475        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1476        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1477        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1478        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1479        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1480        }        }
1481      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 818  for (;;) Line 1484  for (;;)
1484      /* Recursion either matches the current regex, or some subexpression. The      /* Recursion either matches the current regex, or some subexpression. The
1485      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
1486      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1487    
1488      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1489      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
1490      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
1491      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
1492      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
1493      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
1494      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.  
1495    
1496      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
1497      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
1498      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1499        a lot, so he is not to blame for the current way it works. */
1500    
1501      case OP_RECURSE:      case OP_RECURSE:
1502        {        {
1503        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1504        new_recursive.group_num = *callpat - OP_BRA;        new_recursive.group_num = (callpat == md->start_code)? 0 :
1505            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);  
1506    
1507        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1508    
1509        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1510        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1511    
1512        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1513    
1514        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1515    
1516        /* Now save the offset data. */        /* Now save the offset data */
1517    
1518        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1519        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 866  for (;;) Line 1524  for (;;)
1524            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1525          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1526          }          }
   
1527        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1528              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
1529        new_recursive.save_start = md->start_match;  
1530        md->start_match = eptr;        /* OK, now we can do the recursion. After processing each alternative,
1531          restore the offset data. If there were nested recursions, md->recursive
1532        /* 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. */  
1533    
1534        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1535          cbegroup = (*callpat >= OP_SBRA);
1536        do        do
1537          {          {
1538          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1539              eptrb, match_isgroup);          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1540          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1541            memcpy(md->offset_vector, new_recursive.offset_save,
1542                new_recursive.saved_max * sizeof(int));
1543            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1544            {            {
1545              DPRINTF(("Recursion matched\n"));
1546            md->recursive = new_recursive.prevrec;            md->recursive = new_recursive.prevrec;
1547            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1548              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1549            RRETURN(MATCH_MATCH);  
1550              /* Set where we got to in the subject, and reset the start in case
1551              it was changed by \K. This *is* propagated back out of a recursion,
1552              for Perl compatibility. */
1553    
1554              eptr = md->end_match_ptr;
1555              mstart = md->start_match_ptr;
1556              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1557              }
1558            else if (rrc != MATCH_NOMATCH &&
1559                    (rrc != MATCH_THEN || md->start_match_ptr != ecode))
1560              {
1561              DPRINTF(("Recursion gave error %d\n", rrc));
1562              if (new_recursive.offset_save != stacksave)
1563                (pcre_free)(new_recursive.offset_save);
1564              RRETURN(rrc);
1565            }            }
         else if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1566    
1567          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1568          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1569          }          }
1570        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 900  for (;;) Line 1573  for (;;)
1573        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1574        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1575          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1576        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
1577        }        }
1578      /* Control never reaches here */  
1579        RECURSION_MATCHED:
1580      /* "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. */  
   
     case OP_ONCE:  
       {  
       prev = ecode;  
       saved_eptr = eptr;  
1581    
1582        do      /* An alternation is the end of a branch; scan along to find the end of the
1583          {      bracketed group and go to there. */
         RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
           eptrb, match_isgroup);  
         if (rrc == MATCH_MATCH) break;  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         ecode += GET(ecode,1);  
         }  
       while (*ecode == OP_ALT);  
1584    
1585        /* If hit the end of the group (which could be repeated), fail */      case OP_ALT:
1586        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1587        break;
1588    
1589        if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1590        indicating that it may occur zero times. It may repeat infinitely, or not
1591        at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1592        with fixed upper repeat limits are compiled as a number of copies, with the
1593        optional ones preceded by BRAZERO or BRAMINZERO. */
1594    
1595        case OP_BRAZERO:
1596        next = ecode + 1;
1597        RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1598        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1599        do next += GET(next, 1); while (*next == OP_ALT);
1600        ecode = next + 1 + LINK_SIZE;
1601        break;
1602    
1603        case OP_BRAMINZERO:
1604        next = ecode + 1;
1605        do next += GET(next, 1); while (*next == OP_ALT);
1606        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1607        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1608        ecode++;
1609        break;
1610    
1611        /* Continue as from after the assertion, updating the offsets high water      case OP_SKIPZERO:
1612        mark, since extracts may have been taken. */      next = ecode+1;
1613        do next += GET(next,1); while (*next == OP_ALT);
1614        ecode = next + 1 + LINK_SIZE;
1615        break;
1616    
1617        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1618        here; just jump to the group, with allow_zero set TRUE. */
1619    
1620        case OP_BRAPOSZERO:
1621        op = *(++ecode);
1622        allow_zero = TRUE;
1623        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1624          goto POSSESSIVE_NON_CAPTURE;
1625    
1626        do ecode += GET(ecode,1); while (*ecode == OP_ALT);      /* End of a group, repeated or non-repeating. */
1627    
1628        offset_top = md->end_offset_top;      case OP_KET:
1629        eptr = md->end_match_ptr;      case OP_KETRMIN:
1630        case OP_KETRMAX:
1631        case OP_KETRPOS:
1632        prev = ecode - GET(ecode, 1);
1633    
1634        /* If this was a group that remembered the subject start, in order to break
1635        infinite repeats of empty string matches, retrieve the subject start from
1636        the chain. Otherwise, set it NULL. */
1637    
1638        if (*prev >= OP_SBRA || *prev == OP_ONCE)
1639          {
1640          saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1641          eptrb = eptrb->epb_prev;              /* Backup to previous group */
1642          }
1643        else saved_eptr = NULL;
1644    
1645        /* If we are at the end of an assertion group, stop matching and return
1646        MATCH_MATCH, but record the current high water mark for use by positive
1647        assertions. We also need to record the match start in case it was changed
1648        by \K. */
1649    
1650        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
1651            *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT)
1652          {
1653          md->end_match_ptr = eptr;      /* For ONCE */
1654          md->end_offset_top = offset_top;
1655          md->start_match_ptr = mstart;
1656          MRRETURN(MATCH_MATCH);
1657          }
1658    
1659        /* For capturing groups we have to check the group number back at the start
1660        and if necessary complete handling an extraction by setting the offsets and
1661        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1662        into group 0, so it won't be picked up here. Instead, we catch it when the
1663        OP_END is reached. Other recursion is handled here. We just have to record
1664        the current subject position and start match pointer and give a MATCH
1665        return. */
1666    
1667        if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1668            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1669          {
1670          number = GET2(prev, 1+LINK_SIZE);
1671          offset = number << 1;
1672    
1673    #ifdef PCRE_DEBUG
1674          printf("end bracket %d", number);
1675          printf("\n");
1676    #endif
1677    
1678        /* For a non-repeating ket, just continue at this level. This also        /* Handle a recursively called group. */
       happens for a repeating ket if no characters were matched in the group.  
       This is the forcible breaking of infinite loops as implemented in Perl  
       5.005. If there is an options reset, it will get obeyed in the normal  
       course of events. */  
1679    
1680        if (*ecode == OP_KET || eptr == saved_eptr)        if (md->recursive != NULL && md->recursive->group_num == number)
1681          {          {
1682          ecode += 1+LINK_SIZE;          md->end_match_ptr = eptr;
1683          break;          md->start_match_ptr = mstart;
1684            RRETURN(MATCH_MATCH);
1685          }          }
1686    
1687        /* The repeating kets try the rest of the pattern or restart from the        /* Deal with capturing */
       preceding bracket, in the appropriate order. We need to reset any options  
       that changed within the bracket before re-running it, so check the next  
       opcode. */  
1688    
1689        if (ecode[1+LINK_SIZE] == OP_OPT)        md->capture_last = number;
1690          if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1691          {          {
1692          ims = (ims & ~PCRE_IMS) | ecode[4];          /* If offset is greater than offset_top, it means that we are
1693          DPRINTF(("ims set to %02lx at group repeat\n", ims));          "skipping" a capturing group, and that group's offsets must be marked
1694            unset. In earlier versions of PCRE, all the offsets were unset at the
1695            start of matching, but this doesn't work because atomic groups and
1696            assertions can cause a value to be set that should later be unset.
1697            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1698            part of the atomic group, but this is not on the final matching path,
1699            so must be unset when 2 is set. (If there is no group 2, there is no
1700            problem, because offset_top will then be 2, indicating no capture.) */
1701    
1702            if (offset > offset_top)
1703              {
1704              register int *iptr = md->offset_vector + offset_top;
1705              register int *iend = md->offset_vector + offset;
1706              while (iptr < iend) *iptr++ = -1;
1707              }
1708    
1709            /* Now make the extraction */
1710    
1711            md->offset_vector[offset] =
1712              md->offset_vector[md->offset_end - number];
1713            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1714            if (offset_top <= offset) offset_top = offset + 2;
1715          }          }
1716          }
1717    
1718        if (*ecode == OP_KETRMIN)      /* For an ordinary non-repeating ket, just continue at this level. This
1719          {      also happens for a repeating ket if no characters were matched in the
1720          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);      group. This is the forcible breaking of infinite loops as implemented in
1721          if (rrc != MATCH_NOMATCH) RRETURN(rrc);      Perl 5.005. For a non-repeating atomic group, establish a backup point by
1722          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);      processing the rest of the pattern at a lower level. If this results in a
1723          if (rrc != MATCH_NOMATCH) RRETURN(rrc);      NOMATCH return, pass MATCH_ONCE back to the original OP_ONCE level, thereby
1724          }      bypassing intermediate backup points, but resetting any captures that
1725        else  /* OP_KETRMAX */      happened along the way. */
1726    
1727        if (*ecode == OP_KET || eptr == saved_eptr)
1728          {
1729          if (*prev == OP_ONCE)
1730          {          {
1731          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1732          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1733          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1734          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RRETURN(MATCH_ONCE);
1735          }          }
1736        }        ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1737      RRETURN(MATCH_NOMATCH);        break;
   
     /* 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;  
1738        }        }
1739      break;  
1740        /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1741        and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1742        at a time from the outer level, thus saving stack. */
1743    
1744        if (*ecode == OP_KETRPOS)
1745          {
1746          md->end_match_ptr = eptr;
1747          md->end_offset_top = offset_top;
1748          RRETURN(MATCH_KETRPOS);
1749          }
1750    
1751        /* The normal repeating kets try the rest of the pattern or restart from
1752        the preceding bracket, in the appropriate order. In the second case, we can
1753        use tail recursion to avoid using another stack frame, unless we have an
1754        an atomic group or an unlimited repeat of a group that can match an empty
1755        string. */
1756    
1757      case OP_BRAMINZERO:      if (*ecode == OP_KETRMIN)
1758        {        {
1759        next = ecode+1;        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
       do next += GET(next,1); while (*next == OP_ALT);  
       RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,  
         match_isgroup);  
1760        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1761        ecode++;        if (*prev == OP_ONCE)
       }  
     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;  
   
       if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
           *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||  
           *prev == OP_ONCE)  
         {  
         md->end_match_ptr = eptr;      /* For ONCE */  
         md->end_offset_top = offset_top;  
         RRETURN(MATCH_MATCH);  
         }  
   
       /* In all other cases except a conditional group we have to check the  
       group number back at the start and if necessary complete handling an  
       extraction by setting the offsets and bumping the high water mark. */  
   
       if (*prev != OP_COND)  
         {  
         number = *prev - OP_BRA;  
   
         /* 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);  
         offset = number << 1;  
   
 #ifdef DEBUG  
         printf("end bracket %d", number);  
         printf("\n");  
 #endif  
   
         /* Test for a numbered group. This includes groups called as a result  
         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. */  
   
         if (number > 0)  
           {  
           md->capture_last = number;  
           if (offset >= md->offset_max) md->offset_overflow = TRUE; else  
             {  
             md->offset_vector[offset] =  
               md->offset_vector[md->offset_end - number];  
             md->offset_vector[offset+1] = eptr - md->start_subject;  
             if (offset_top <= offset) offset_top = offset + 2;  
             }  
   
           /* Handle a recursively called group. Restore the offsets  
           appropriately and continue from after the call. */  
   
           if (md->recursive != NULL && md->recursive->group_num == number)  
             {  
             recursion_info *rec = md->recursive;  
             DPRINTF(("Recursion (%d) succeeded - continuing\n", number));  
             md->recursive = rec->prevrec;  
             md->start_match = rec->save_start;  
             memcpy(md->offset_vector, rec->offset_save,  
               rec->saved_max * sizeof(int));  
             ecode = rec->after_call;  
             ims = original_ims;  
             break;  
             }  
           }  
         }  
   
       /* Reset the value of the ims flags, in case they got changed during  
       the group. */  
   
       ims = original_ims;  
       DPRINTF(("ims reset to %02lx\n", ims));  
   
       /* For a non-repeating ket, just continue at this level. This also  
       happens for a repeating ket if no characters were matched in the group.  
       This is the forcible breaking of infinite loops as implemented in Perl  
       5.005. If there is an options reset, it will get obeyed in the normal  
       course of events. */  
   
       if (*ecode == OP_KET || eptr == saved_eptr)  
         {  
         ecode += 1 + LINK_SIZE;  
         break;  
         }  
   
       /* The repeating kets try the rest of the pattern or restart from the  
       preceding bracket, in the appropriate order. */  
   
       if (*ecode == OP_KETRMIN)  
1762          {          {
1763          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
1764          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1765          }          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1766        else  /* OP_KETRMAX */          RRETURN(MATCH_ONCE);
1767            }
1768          if (*prev >= OP_SBRA)    /* Could match an empty string */
1769            {
1770            md->match_function_type = MATCH_CBEGROUP;
1771            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1772            RRETURN(rrc);
1773            }
1774          ecode = prev;
1775          goto TAIL_RECURSE;
1776          }
1777        else  /* OP_KETRMAX */
1778          {
1779          if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1780          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1781          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1782          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1783          if (*prev == OP_ONCE)
1784          {          {
1785          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1786          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1787          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          md->once_target = prev;
1788          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RRETURN(MATCH_ONCE);
1789          }          }
1790          ecode += 1 + LINK_SIZE;
1791          goto TAIL_RECURSE;
1792        }        }
1793        /* Control never gets here */
1794    
1795      RRETURN(MATCH_NOMATCH);      /* Not multiline mode: start of subject assertion, unless notbol. */
   
     /* Start of subject unless notbol, or after internal newline if multiline */  
1796    
1797      case OP_CIRC:      case OP_CIRC:
1798      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1799      if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject && eptr[-1] != NEWLINE)  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
   
1800      /* Start of subject assertion */      /* Start of subject assertion */
1801    
1802      case OP_SOD:      case OP_SOD:
1803      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1804        ecode++;
1805        break;
1806    
1807        /* Multiline mode: start of subject unless notbol, or after any newline. */
1808    
1809        case OP_CIRCM:
1810        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1811        if (eptr != md->start_subject &&
1812            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1813          MRRETURN(MATCH_NOMATCH);
1814      ecode++;      ecode++;
1815      break;      break;
1816    
1817      /* Start of match assertion */      /* Start of match assertion */
1818    
1819      case OP_SOM:      case OP_SOM:
1820      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
1821      ecode++;      ecode++;
1822      break;      break;
1823    
1824      /* 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. */  
1825    
1826      case OP_DOLL:      case OP_SET_SOM:
1827      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
1828        {      ecode++;
1829        if (eptr < md->end_subject)      break;
1830          { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }  
1831        else      /* Multiline mode: assert before any newline, or before end of subject
1832          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
1833        ecode++;  
1834        break;      case OP_DOLLM:
1835        }      if (eptr < md->end_subject)
1836          { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
1837      else      else
1838        {        {
1839        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1840        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr < md->end_subject - 1 ||  
            (eptr == md->end_subject - 1 && *eptr != NEWLINE))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
1841        }        }
1842      /* ... else fall through */      ecode++;
1843        break;
1844    
1845        /* Not multiline mode: assert before a terminating newline or before end of
1846        subject unless noteol is set. */
1847    
1848        case OP_DOLL:
1849        if (md->noteol) MRRETURN(MATCH_NOMATCH);
1850        if (!md->endonly) goto ASSERT_NL_OR_EOS;
1851    
1852        /* ... else fall through for endonly */
1853    
1854      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
1855    
1856      case OP_EOD:      case OP_EOD:
1857      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
1858        SCHECK_PARTIAL();
1859      ecode++;      ecode++;
1860      break;      break;
1861    
1862      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
1863    
1864      case OP_EODN:      case OP_EODN:
1865      if (eptr < md->end_subject - 1 ||      ASSERT_NL_OR_EOS:
1866         (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject &&
1867            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
1868          MRRETURN(MATCH_NOMATCH);
1869    
1870        /* Either at end of string or \n before end. */
1871    
1872        SCHECK_PARTIAL();
1873      ecode++;      ecode++;
1874      break;      break;
1875    
# Line 1206  for (;;) Line 1881  for (;;)
1881    
1882        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
1883        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
1884        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
1885          partial matching. */
1886    
1887  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1888        if (utf8)        if (utf8)
1889          {          {
1890            /* Get status of previous character */
1891    
1892          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
1893            {            {
1894            const uschar *lastptr = eptr - 1;            USPTR lastptr = eptr - 1;
1895            while((*lastptr & 0xc0) == 0x80) lastptr--;            while((*lastptr & 0xc0) == 0x80) lastptr--;
1896              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
1897            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
1898    #ifdef SUPPORT_UCP
1899              if (md->use_ucp)
1900                {
1901                if (c == '_') prev_is_word = TRUE; else
1902                  {
1903                  int cat = UCD_CATEGORY(c);
1904                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1905                  }
1906                }
1907              else
1908    #endif
1909            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1910            }            }
1911          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
1912            /* Get status of next character */
1913    
1914            if (eptr >= md->end_subject)
1915              {
1916              SCHECK_PARTIAL();
1917              cur_is_word = FALSE;
1918              }
1919            else
1920            {            {
1921            GETCHAR(c, eptr);            GETCHAR(c, eptr);
1922    #ifdef SUPPORT_UCP
1923              if (md->use_ucp)
1924                {
1925                if (c == '_') cur_is_word = TRUE; else
1926                  {
1927                  int cat = UCD_CATEGORY(c);
1928                  cur_is_word = (cat == ucp_L || cat == ucp_N);
1929                  }
1930                }
1931              else
1932    #endif
1933            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1934            }            }
1935          }          }
1936        else        else
1937  #endif  #endif
1938    
1939        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
1940          consistency with the behaviour of \w we do use it in this case. */
1941    
1942          {          {
1943          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
1944            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
1945          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
1946            ((md->ctypes[*eptr] & ctype_word) != 0);            {
1947              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
1948    #ifdef SUPPORT_UCP
1949              if (md->use_ucp)
1950                {
1951                c = eptr[-1];
1952                if (c == '_') prev_is_word = TRUE; else
1953                  {
1954                  int cat = UCD_CATEGORY(c);
1955                  prev_is_word = (cat == ucp_L || cat == ucp_N);
1956                  }
1957                }
1958              else
1959    #endif
1960              prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
1961              }
1962    
1963            /* Get status of next character */
1964    
1965            if (eptr >= md->end_subject)
1966              {
1967              SCHECK_PARTIAL();
1968              cur_is_word = FALSE;
1969              }
1970            else
1971    #ifdef SUPPORT_UCP
1972            if (md->use_ucp)
1973              {
1974              c = *eptr;
1975              if (c == '_') cur_is_word = TRUE; else
1976                {
1977                int cat = UCD_CATEGORY(c);
1978                cur_is_word = (cat == ucp_L || cat == ucp_N);
1979                }
1980              }
1981            else
1982    #endif
1983            cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
1984          }          }
1985    
1986        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
1987    
1988        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
1989             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
1990          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
1991        }        }
1992      break;      break;
1993    
1994      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
1995    
1996      case OP_ANY:      case OP_ANY:
1997      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
1998        RRETURN(MATCH_NOMATCH);      /* Fall through */
1999      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);  
2000  #ifdef SUPPORT_UTF8      case OP_ALLANY:
2001      if (utf8)      if (eptr++ >= md->end_subject)
2002        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;        {
2003  #endif        SCHECK_PARTIAL();
2004          MRRETURN(MATCH_NOMATCH);
2005          }
2006        if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2007      ecode++;      ecode++;
2008      break;      break;
2009    
# Line 1261  for (;;) Line 2011  for (;;)
2011      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2012    
2013      case OP_ANYBYTE:      case OP_ANYBYTE:
2014      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr++ >= md->end_subject)
2015          {
2016          SCHECK_PARTIAL();
2017          MRRETURN(MATCH_NOMATCH);
2018          }
2019      ecode++;      ecode++;
2020      break;      break;
2021    
2022      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2023      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2024          {
2025          SCHECK_PARTIAL();
2026          MRRETURN(MATCH_NOMATCH);
2027          }
2028      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2029      if (      if (
2030  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1274  for (;;) Line 2032  for (;;)
2032  #endif  #endif
2033         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2034         )         )
2035        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2036      ecode++;      ecode++;
2037      break;      break;
2038    
2039      case OP_DIGIT:      case OP_DIGIT:
2040      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2041          {
2042          SCHECK_PARTIAL();
2043          MRRETURN(MATCH_NOMATCH);
2044          }
2045      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2046      if (      if (
2047  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1287  for (;;) Line 2049  for (;;)
2049  #endif  #endif
2050         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2051         )         )
2052        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2053      ecode++;      ecode++;
2054      break;      break;
2055    
2056      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2057      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2058          {
2059          SCHECK_PARTIAL();
2060          MRRETURN(MATCH_NOMATCH);
2061          }
2062      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2063      if (      if (
2064  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1300  for (;;) Line 2066  for (;;)
2066  #endif  #endif
2067         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2068         )         )
2069        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2070      ecode++;      ecode++;
2071      break;      break;
2072    
2073      case OP_WHITESPACE:      case OP_WHITESPACE:
2074      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2075          {
2076          SCHECK_PARTIAL();
2077          MRRETURN(MATCH_NOMATCH);
2078          }
2079      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2080      if (      if (
2081  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1313  for (;;) Line 2083  for (;;)
2083  #endif  #endif
2084         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2085         )         )
2086        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2087      ecode++;      ecode++;
2088      break;      break;
2089    
2090      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2091      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2092          {
2093          SCHECK_PARTIAL();
2094          MRRETURN(MATCH_NOMATCH);
2095          }
2096      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2097      if (      if (
2098  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1326  for (;;) Line 2100  for (;;)
2100  #endif  #endif
2101         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2102         )         )
2103        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2104      ecode++;      ecode++;
2105      break;      break;
2106    
2107      case OP_WORDCHAR:      case OP_WORDCHAR:
2108      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2109          {
2110          SCHECK_PARTIAL();
2111          MRRETURN(MATCH_NOMATCH);
2112          }
2113      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2114      if (      if (
2115  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1339  for (;;) Line 2117  for (;;)
2117  #endif  #endif
2118         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2119         )         )
2120        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2121        ecode++;
2122        break;
2123    
2124        case OP_ANYNL:
2125        if (eptr >= md->end_subject)
2126          {
2127          SCHECK_PARTIAL();
2128          MRRETURN(MATCH_NOMATCH);
2129          }
2130        GETCHARINCTEST(c, eptr);
2131        switch(c)
2132          {
2133          default: MRRETURN(MATCH_NOMATCH);
2134    
2135          case 0x000d:
2136          if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2137          break;
2138    
2139          case 0x000a:
2140          break;
2141    
2142          case 0x000b:
2143          case 0x000c:
2144          case 0x0085:
2145          case 0x2028:
2146          case 0x2029:
2147          if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
2148          break;
2149          }
2150        ecode++;
2151        break;
2152    
2153        case OP_NOT_HSPACE:
2154        if (eptr >= md->end_subject)
2155          {
2156          SCHECK_PARTIAL();
2157          MRRETURN(MATCH_NOMATCH);
2158          }
2159        GETCHARINCTEST(c, eptr);
2160        switch(c)
2161          {
2162          default: break;
2163          case 0x09:      /* HT */
2164          case 0x20:      /* SPACE */
2165          case 0xa0:      /* NBSP */
2166          case 0x1680:    /* OGHAM SPACE MARK */
2167          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2168          case 0x2000:    /* EN QUAD */
2169          case 0x2001:    /* EM QUAD */
2170          case 0x2002:    /* EN SPACE */
2171          case 0x2003:    /* EM SPACE */
2172          case 0x2004:    /* THREE-PER-EM SPACE */
2173          case 0x2005:    /* FOUR-PER-EM SPACE */
2174          case 0x2006:    /* SIX-PER-EM SPACE */
2175          case 0x2007:    /* FIGURE SPACE */
2176          case 0x2008:    /* PUNCTUATION SPACE */
2177          case 0x2009:    /* THIN SPACE */
2178          case 0x200A:    /* HAIR SPACE */
2179          case 0x202f:    /* NARROW NO-BREAK SPACE */
2180          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2181          case 0x3000:    /* IDEOGRAPHIC SPACE */
2182          MRRETURN(MATCH_NOMATCH);
2183          }
2184        ecode++;
2185        break;
2186    
2187        case OP_HSPACE:
2188        if (eptr >= md->end_subject)
2189          {
2190          SCHECK_PARTIAL();
2191          MRRETURN(MATCH_NOMATCH);
2192          }
2193        GETCHARINCTEST(c, eptr);
2194        switch(c)
2195          {
2196          default: MRRETURN(MATCH_NOMATCH);
2197          case 0x09:      /* HT */
2198          case 0x20:      /* SPACE */
2199          case 0xa0:      /* NBSP */
2200          case 0x1680:    /* OGHAM SPACE MARK */
2201          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2202          case 0x2000:    /* EN QUAD */
2203          case 0x2001:    /* EM QUAD */
2204          case 0x2002:    /* EN SPACE */
2205          case 0x2003:    /* EM SPACE */
2206          case 0x2004:    /* THREE-PER-EM SPACE */
2207          case 0x2005:    /* FOUR-PER-EM SPACE */
2208          case 0x2006:    /* SIX-PER-EM SPACE */
2209          case 0x2007:    /* FIGURE SPACE */
2210          case 0x2008:    /* PUNCTUATION SPACE */
2211          case 0x2009:    /* THIN SPACE */
2212          case 0x200A:    /* HAIR SPACE */
2213          case 0x202f:    /* NARROW NO-BREAK SPACE */
2214          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2215          case 0x3000:    /* IDEOGRAPHIC SPACE */
2216          break;
2217          }
2218        ecode++;
2219        break;
2220    
2221        case OP_NOT_VSPACE:
2222        if (eptr >= md->end_subject)
2223          {
2224          SCHECK_PARTIAL();
2225          MRRETURN(MATCH_NOMATCH);
2226          }
2227        GETCHARINCTEST(c, eptr);
2228        switch(c)
2229          {
2230          default: break;
2231          case 0x0a:      /* LF */
2232          case 0x0b:      /* VT */
2233          case 0x0c:      /* FF */
2234          case 0x0d:      /* CR */
2235          case 0x85:      /* NEL */
2236          case 0x2028:    /* LINE SEPARATOR */
2237          case 0x2029:    /* PARAGRAPH SEPARATOR */
2238          MRRETURN(MATCH_NOMATCH);
2239          }
2240        ecode++;
2241        break;
2242    
2243        case OP_VSPACE:
2244        if (eptr >= md->end_subject)
2245          {
2246          SCHECK_PARTIAL();
2247          MRRETURN(MATCH_NOMATCH);
2248          }
2249        GETCHARINCTEST(c, eptr);
2250        switch(c)
2251          {
2252          default: MRRETURN(MATCH_NOMATCH);
2253          case 0x0a:      /* LF */
2254          case 0x0b:      /* VT */
2255          case 0x0c:      /* FF */
2256          case 0x0d:      /* CR */
2257          case 0x85:      /* NEL */
2258          case 0x2028:    /* LINE SEPARATOR */
2259          case 0x2029:    /* PARAGRAPH SEPARATOR */
2260          break;
2261          }
2262      ecode++;      ecode++;
2263      break;      break;
2264    
# Line 1349  for (;;) Line 2268  for (;;)
2268    
2269      case OP_PROP:      case OP_PROP:
2270      case OP_NOTPROP:      case OP_NOTPROP:
2271      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2272          {
2273          SCHECK_PARTIAL();
2274          MRRETURN(MATCH_NOMATCH);
2275          }
2276      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2277        {        {
2278        int chartype, rqdtype;        const ucd_record *prop = GET_UCD(c);
       int othercase;  
       int category = _pcre_ucp_findchar(c, &chartype, &othercase);  
2279    
2280        rqdtype = *(++ecode);        switch(ecode[1])
       ecode++;  
   
       if (rqdtype >= 128)  
         {  
         if ((rqdtype - 128 != category) == (op == OP_PROP))  
           RRETURN(MATCH_NOMATCH);  
         }  
       else  
2281          {          {
2282          if ((rqdtype != chartype) == (op == OP_PROP))          case PT_ANY:
2283            RRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
2284            break;
2285    
2286            case PT_LAMP:
2287            if ((prop->chartype == ucp_Lu ||
2288                 prop->chartype == ucp_Ll ||
2289                 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2290              MRRETURN(MATCH_NOMATCH);
2291            break;
2292    
2293            case PT_GC:
2294            if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
2295              MRRETURN(MATCH_NOMATCH);
2296            break;
2297    
2298            case PT_PC:
2299            if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2300              MRRETURN(MATCH_NOMATCH);
2301            break;
2302    
2303            case PT_SC:
2304            if ((ecode[2] != prop->script) == (op == OP_PROP))
2305              MRRETURN(MATCH_NOMATCH);
2306            break;
2307    
2308            /* These are specials */
2309    
2310            case PT_ALNUM:
2311            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2312                 _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2313              MRRETURN(MATCH_NOMATCH);
2314            break;
2315    
2316            case PT_SPACE:    /* Perl space */
2317            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2318                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2319                   == (op == OP_NOTPROP))
2320              MRRETURN(MATCH_NOMATCH);
2321            break;
2322    
2323            case PT_PXSPACE:  /* POSIX space */
2324            if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2325                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2326                 c == CHAR_FF || c == CHAR_CR)
2327                   == (op == OP_NOTPROP))
2328              MRRETURN(MATCH_NOMATCH);
2329            break;
2330    
2331            case PT_WORD:
2332            if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2333                 _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2334                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2335              MRRETURN(MATCH_NOMATCH);
2336            break;
2337    
2338            /* This should never occur */
2339    
2340            default:
2341            RRETURN(PCRE_ERROR_INTERNAL);
2342          }          }
2343    
2344          ecode += 3;
2345        }        }
2346      break;      break;
2347    
# Line 1376  for (;;) Line 2349  for (;;)
2349      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2350    
2351      case OP_EXTUNI:      case OP_EXTUNI:
2352      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2353          {
2354          SCHECK_PARTIAL();
2355          MRRETURN(MATCH_NOMATCH);
2356          }
2357      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2358        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2359        while (eptr < md->end_subject)
2360        {        {
2361        int chartype;        int len = 1;
2362        int othercase;        if (!utf8) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2363        int category = _pcre_ucp_findchar(c, &chartype, &othercase);        if (UCD_CATEGORY(c) != ucp_M) break;
2364        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        eptr += len;
       while (eptr < md->end_subject)  
         {  
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = _pcre_ucp_findchar(c, &chartype, &othercase);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2365        }        }
2366      ecode++;      ecode++;
2367      break;      break;
# Line 1409  for (;;) Line 2377  for (;;)
2377      loops). */      loops). */
2378    
2379      case OP_REF:      case OP_REF:
2380        {      case OP_REFI:
2381        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2382        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2383        ecode += 3;
2384    
2385        /* If the reference is unset, set the length to be longer than the amount      /* If the reference is unset, there are two possibilities:
       of subject left; this ensures that every attempt at a match fails. We  
       can't just fail here, because of the possibility of quantifiers with zero  
       minima. */  
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2386    
2387        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2388        this ensures that every attempt at a match fails. We can't just fail
2389        here, because of the possibility of quantifiers with zero minima.
2390    
2391        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2392          {      so that the back reference matches an empty string.
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
2393    
2394          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2395          case OP_CRMINRANGE:      referenced subpattern. */
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2396    
2397          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2398          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2399          eptr += length;      else
2400          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2401    
2402        /* Set up for repetition, or handle the non-repeated case */
2403    
2404        switch (*ecode)
2405          {
2406          case OP_CRSTAR:
2407          case OP_CRMINSTAR:
2408          case OP_CRPLUS:
2409          case OP_CRMINPLUS:
2410          case OP_CRQUERY:
2411          case OP_CRMINQUERY:
2412          c = *ecode++ - OP_CRSTAR;
2413          minimize = (c & 1) != 0;
2414          min = rep_min[c];                 /* Pick up values from tables; */
2415          max = rep_max[c];                 /* zero for max => infinity */
2416          if (max == 0) max = INT_MAX;
2417          break;
2418    
2419          case OP_CRRANGE:
2420          case OP_CRMINRANGE:
2421          minimize = (*ecode == OP_CRMINRANGE);
2422          min = GET2(ecode, 1);
2423          max = GET2(ecode, 3);
2424          if (max == 0) max = INT_MAX;
2425          ecode += 5;
2426          break;
2427    
2428          default:               /* No repeat follows */
2429          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2430            {
2431            CHECK_PARTIAL();
2432            MRRETURN(MATCH_NOMATCH);
2433          }          }
2434          eptr += length;
2435          continue;              /* With the main loop */
2436          }
2437    
2438        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2439        main loop. */      zero, just continue with the main loop. */
2440    
2441        if (length == 0) continue;      if (length == 0) continue;
2442    
2443        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2444        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2445        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2446    
2447        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2448          {
2449          int slength;
2450          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2451          {          {
2452          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2453          eptr += length;          MRRETURN(MATCH_NOMATCH);
2454          }          }
2455          eptr += slength;
2456          }
2457    
2458        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2459        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2460    
2461        if (min == max) continue;      if (min == max) continue;
2462    
2463        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2464    
2465        if (minimize)      if (minimize)
2466          {
2467          for (fi = min;; fi++)
2468          {          {
2469          for (fi = min;; fi++)          int slength;
2470            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2471            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2472            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2473            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2474            {            {
2475            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2476            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2477            }            }
2478          /* Control never gets here */          eptr += slength;
2479          }          }
2480          /* Control never gets here */
2481          }
2482    
2483        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2484    
2485        else      else
2486          {
2487          pp = eptr;
2488          for (i = min; i < max; i++)
2489          {          {
2490          pp = eptr;          int slength;
2491          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)  
2492            {            {
2493            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2494            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2495            }            }
2496          RRETURN(MATCH_NOMATCH);          eptr += slength;
2497            }
2498          while (eptr >= pp)
2499            {
2500            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2501            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2502            eptr -= length;
2503          }          }
2504          MRRETURN(MATCH_NOMATCH);
2505        }        }
2506      /* Control never gets here */      /* Control never gets here */
2507    
   
   
2508      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2509      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,
2510      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 1566  for (;;) Line 2559  for (;;)
2559          {          {
2560          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2561            {            {
2562            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2563                {
2564                SCHECK_PARTIAL();
2565                MRRETURN(MATCH_NOMATCH);
2566                }
2567            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2568            if (c > 255)            if (c > 255)
2569              {              {
2570              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2571              }              }
2572            else            else
2573              {              {
2574              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2575              }              }
2576            }            }
2577          }          }
# Line 1584  for (;;) Line 2581  for (;;)
2581          {          {
2582          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2583            {            {
2584            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2585                {
2586                SCHECK_PARTIAL();
2587                MRRETURN(MATCH_NOMATCH);
2588                }
2589            c = *eptr++;            c = *eptr++;
2590            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2591            }            }
2592          }          }
2593    
# Line 1606  for (;;) Line 2607  for (;;)
2607            {            {
2608            for (fi = min;; fi++)            for (fi = min;; fi++)
2609              {              {
2610              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2611              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2612              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2613                if (eptr >= md->end_subject)
2614                  {
2615                  SCHECK_PARTIAL();
2616                  MRRETURN(MATCH_NOMATCH);
2617                  }
2618              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2619              if (c > 255)              if (c > 255)
2620                {                {
2621                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2622                }                }
2623              else              else
2624                {                {
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 1626  for (;;) Line 2632  for (;;)
2632            {            {
2633            for (fi = min;; fi++)            for (fi = min;; fi++)
2634              {              {
2635              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2636              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2637              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2638                if (eptr >= md->end_subject)
2639                  {
2640                  SCHECK_PARTIAL();
2641                  MRRETURN(MATCH_NOMATCH);
2642                  }
2643              c = *eptr++;              c = *eptr++;
2644              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2645              }              }
2646            }            }
2647          /* Control never gets here */          /* Control never gets here */
# Line 1649  for (;;) Line 2660  for (;;)
2660            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2661              {              {
2662              int len = 1;              int len = 1;
2663              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2664                  {
2665                  SCHECK_PARTIAL();
2666                  break;
2667                  }
2668              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2669              if (c > 255)              if (c > 255)
2670                {                {
# Line 1663  for (;;) Line 2678  for (;;)
2678              }              }
2679            for (;;)            for (;;)
2680              {              {
2681              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2682              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2683              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2684              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1675  for (;;) Line 2690  for (;;)
2690            {            {
2691            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2692              {              {
2693              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2694                  {
2695                  SCHECK_PARTIAL();
2696                  break;
2697                  }
2698              c = *eptr;              c = *eptr;
2699              if ((data[c/8] & (1 << (c&7))) == 0) break;              if ((data[c/8] & (1 << (c&7))) == 0) break;
2700              eptr++;              eptr++;
2701              }              }
2702            while (eptr >= pp)            while (eptr >= pp)
2703              {              {
2704              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
             eptr--;  
2705              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2706                eptr--;
2707              }              }
2708            }            }
2709    
2710          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2711          }          }
2712        }        }
2713      /* Control never gets here */      /* Control never gets here */
2714    
2715    
2716      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2717      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
2718        mode, because Unicode properties are supported in non-UTF-8 mode. */
2719    
2720  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2721      case OP_XCLASS:      case OP_XCLASS:
# Line 1736  for (;;) Line 2756  for (;;)
2756    
2757        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2758          {          {
2759          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2760          GETCHARINC(c, eptr);            {
2761          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2762              MRRETURN(MATCH_NOMATCH);
2763              }
2764            GETCHARINCTEST(c, eptr);
2765            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2766          }          }
2767    
2768        /* 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 1753  for (;;) Line 2777  for (;;)
2777          {          {
2778          for (fi = min;; fi++)          for (fi = min;; fi++)
2779            {            {
2780            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2781            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2782            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2783            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
2784            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
2785                SCHECK_PARTIAL();
2786                MRRETURN(MATCH_NOMATCH);
2787                }
2788              GETCHARINCTEST(c, eptr);
2789              if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);
2790            }            }
2791          /* Control never gets here */          /* Control never gets here */
2792          }          }
# Line 1770  for (;;) Line 2799  for (;;)
2799          for (i = min; i < max; i++)          for (i = min; i < max; i++)
2800            {            {
2801            int len = 1;            int len = 1;
2802            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
2803            GETCHARLEN(c, eptr, len);              {
2804                SCHECK_PARTIAL();
2805                break;
2806                }
2807              GETCHARLENTEST(c, eptr, len);
2808            if (!_pcre_xclass(c, data)) break;            if (!_pcre_xclass(c, data)) break;
2809            eptr += len;            eptr += len;
2810            }            }
2811          for(;;)          for(;;)
2812            {            {
2813            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
2814            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2815            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
2816            BACKCHAR(eptr)            if (utf8) BACKCHAR(eptr);
2817            }            }
2818          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2819          }          }
2820    
2821        /* Control never gets here */        /* Control never gets here */
# Line 1798  for (;;) Line 2831  for (;;)
2831        length = 1;        length = 1;
2832        ecode++;        ecode++;
2833        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2834        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2835        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
2836            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2837            MRRETURN(MATCH_NOMATCH);
2838            }
2839          while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
2840        }        }
2841      else      else
2842  #endif  #endif
2843    
2844      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2845        {        {
2846        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2847        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);          {
2848            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2849            MRRETURN(MATCH_NOMATCH);
2850            }
2851          if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
2852        ecode += 2;        ecode += 2;
2853        }        }
2854      break;      break;
2855    
2856      /* Match a single character, caselessly */      /* Match a single character, caselessly */
2857    
2858      case OP_CHARNC:      case OP_CHARI:
2859  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2860      if (utf8)      if (utf8)
2861        {        {
# Line 1822  for (;;) Line 2863  for (;;)
2863        ecode++;        ecode++;
2864        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
2865    
2866        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
2867            {
2868            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
2869            MRRETURN(MATCH_NOMATCH);
2870            }
2871    
2872        /* 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
2873        can use the fast lookup table. */        can use the fast lookup table. */
2874    
2875        if (fc < 128)        if (fc < 128)
2876          {          {
2877          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2878          }          }
2879    
2880        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character */
2881    
2882        else        else
2883          {          {
2884          int dc;          unsigned int dc;
2885          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
2886          ecode += length;          ecode += length;
2887    
2888          /* If we have Unicode property support, we can use it to test the other          /* If we have Unicode property support, we can use it to test the other
2889          case of the character, if there is one. The result of _pcre_ucp_findchar() is          case of the character, if there is one. */
         < 0 if the char isn't found, and othercase is returned as zero if there  
         isn't one. */  
2890    
2891          if (fc != dc)          if (fc != dc)
2892            {            {
2893  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2894            int chartype;            if (dc != UCD_OTHERCASE(fc))
           int othercase;  
           if (_pcre_ucp_findchar(fc, &chartype, &othercase) < 0 || dc != othercase)  
2895  #endif  #endif
2896              RRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
2897            }            }
2898          }          }
2899        }        }
# Line 1861  for (;;) Line 2902  for (;;)
2902    
2903      /* Non-UTF-8 mode */      /* Non-UTF-8 mode */
2904        {        {
2905        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
2906        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
2907            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
2908            MRRETURN(MATCH_NOMATCH);
2909            }
2910          if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
2911        ecode += 2;        ecode += 2;
2912        }        }
2913      break;      break;
2914    
2915      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
2916    
2917      case OP_EXACT:      case OP_EXACT:
2918        case OP_EXACTI:
2919      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
2920      ecode += 3;      ecode += 3;
2921      goto REPEATCHAR;      goto REPEATCHAR;
2922    
2923        case OP_POSUPTO:
2924        case OP_POSUPTOI:
2925        possessive = TRUE;
2926        /* Fall through */
2927    
2928      case OP_UPTO:      case OP_UPTO:
2929        case OP_UPTOI:
2930      case OP_MINUPTO:      case OP_MINUPTO:
2931        case OP_MINUPTOI:
2932      min = 0;      min = 0;
2933      max = GET2(ecode, 1);      max = GET2(ecode, 1);
2934      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
2935      ecode += 3;      ecode += 3;
2936      goto REPEATCHAR;      goto REPEATCHAR;
2937    
2938        case OP_POSSTAR:
2939        case OP_POSSTARI:
2940        possessive = TRUE;
2941        min = 0;
2942        max = INT_MAX;
2943        ecode++;
2944        goto REPEATCHAR;
2945    
2946        case OP_POSPLUS:
2947        case OP_POSPLUSI:
2948        possessive = TRUE;
2949        min = 1;
2950        max = INT_MAX;
2951        ecode++;
2952        goto REPEATCHAR;
2953    
2954        case OP_POSQUERY:
2955        case OP_POSQUERYI:
2956        possessive = TRUE;
2957        min = 0;
2958        max = 1;
2959        ecode++;
2960        goto REPEATCHAR;
2961    
2962      case OP_STAR:      case OP_STAR:
2963        case OP_STARI:
2964      case OP_MINSTAR:      case OP_MINSTAR:
2965        case OP_MINSTARI:
2966      case OP_PLUS:      case OP_PLUS:
2967        case OP_PLUSI:
2968      case OP_MINPLUS:      case OP_MINPLUS:
2969        case OP_MINPLUSI:
2970      case OP_QUERY:      case OP_QUERY:
2971        case OP_QUERYI:
2972      case OP_MINQUERY:      case OP_MINQUERY:
2973      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
2974        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
2975      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
2976      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
2977      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
2978      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
2979    
2980      /* 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. */  
2981    
2982      REPEATCHAR:      REPEATCHAR:
2983  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1905  for (;;) Line 2986  for (;;)
2986        length = 1;        length = 1;
2987        charptr = ecode;        charptr = ecode;
2988        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
2989        ecode += length;        ecode += length;
2990    
2991        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1913  for (;;) Line 2993  for (;;)
2993    
2994        if (length > 1)        if (length > 1)
2995          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
2996  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2997          int othercase;          unsigned int othercase;
2998          int chartype;          if (op >= OP_STARI &&     /* Caseless */
2999          if ((ims & PCRE_CASELESS) != 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
              _pcre_ucp_findchar(fc, &chartype, &othercase) >= 0 &&  
              othercase > 0)  
3000            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = _pcre_ord2utf8(othercase, occhars);
3001            else oclength = 0;
3002  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3003    
3004          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3005            {            {
3006            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3007            /* Need braces because of following else */              memcmp(eptr, charptr, length) == 0) eptr += length;
3008            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3009              else if (oclength > 0 &&
3010                       eptr <= md->end_subject - oclength &&
3011                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3012    #endif  /* SUPPORT_UCP */
3013            else            else
3014              {              {
3015              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3016              eptr += oclength;              MRRETURN(MATCH_NOMATCH);
3017              }              }
3018            }            }
3019    
# Line 1943  for (;;) Line 3023  for (;;)
3023            {            {
3024            for (fi = min;; fi++)            for (fi = min;; fi++)
3025              {              {
3026              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3027              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3028              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3029              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3030              /* Need braces because of following else */                memcmp(eptr, charptr, length) == 0) eptr += length;
3031              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3032                else if (oclength > 0 &&
3033                         eptr <= md->end_subject - oclength &&
3034                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3035    #endif  /* SUPPORT_UCP */
3036              else              else
3037                {                {
3038                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3039                eptr += oclength;                MRRETURN(MATCH_NOMATCH);
3040                }                }
3041              }              }
3042            /* Control never gets here */            /* Control never gets here */
3043            }            }
3044          else  
3045            else  /* Maximize */
3046            {            {
3047            pp = eptr;            pp = eptr;
3048            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3049              {              {
3050              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3051              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, length) == 0) eptr += length;
3052              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3053                else if (oclength > 0 &&
3054                         eptr <= md->end_subject - oclength &&
3055                         memcmp(eptr, occhars, oclength) == 0) eptr += oclength;
3056    #endif  /* SUPPORT_UCP */
3057              else              else
3058                {                {
3059                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3060                eptr += oclength;                break;
3061                }                }
3062              }              }
3063            while (eptr >= pp)  
3064             {            if (possessive) continue;
3065             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3066             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3067             eptr -= length;              {
3068             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3069            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3070                if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3071    #ifdef SUPPORT_UCP
3072                eptr--;
3073                BACKCHAR(eptr);
3074    #else   /* without SUPPORT_UCP */
3075                eptr -= length;
3076    #endif  /* SUPPORT_UCP */
3077                }
3078            }            }
3079          /* Control never gets here */          /* Control never gets here */
3080          }          }
# Line 1990  for (;;) Line 3087  for (;;)
3087  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3088    
3089      /* When not in UTF-8 mode, load a single-byte character. */      /* When not in UTF-8 mode, load a single-byte character. */
3090        {  
3091        if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);      fc = *ecode++;
       fc = *ecode++;  
       }  
3092    
3093      /* 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
3094      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 2007  for (;;) Line 3102  for (;;)
3102      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3103        max, eptr));        max, eptr));
3104    
3105      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3106        {        {
3107        fc = md->lcc[fc];        fc = md->lcc[fc];
3108        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3109          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3110            if (eptr >= md->end_subject)
3111              {
3112              SCHECK_PARTIAL();
3113              MRRETURN(MATCH_NOMATCH);
3114              }
3115            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3116            }
3117        if (min == max) continue;        if (min == max) continue;
3118        if (minimize)        if (minimize)
3119          {          {
3120          for (fi = min;; fi++)          for (fi = min;; fi++)
3121            {            {
3122            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3123            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3124            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3125                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3126              RRETURN(MATCH_NOMATCH);              {
3127                SCHECK_PARTIAL();
3128                MRRETURN(MATCH_NOMATCH);
3129                }
3130              if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3131            }            }
3132          /* Control never gets here */          /* Control never gets here */
3133          }          }
3134        else        else  /* Maximize */
3135          {          {
3136          pp = eptr;          pp = eptr;
3137          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3138            {            {
3139            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3140                {
3141                SCHECK_PARTIAL();
3142                break;
3143                }
3144              if (fc != md->lcc[*eptr]) break;
3145            eptr++;            eptr++;
3146            }            }
3147    
3148            if (possessive) continue;
3149    
3150          while (eptr >= pp)          while (eptr >= pp)
3151            {            {
3152            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3153            eptr--;            eptr--;
3154            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3155            }            }
3156          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3157          }          }
3158        /* Control never gets here */        /* Control never gets here */
3159        }        }
# Line 2048  for (;;) Line 3162  for (;;)
3162    
3163      else      else
3164        {        {
3165        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3166            {
3167            if (eptr >= md->end_subject)
3168              {
3169              SCHECK_PARTIAL();
3170              MRRETURN(MATCH_NOMATCH);
3171              }
3172            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3173            }
3174    
3175        if (min == max) continue;        if (min == max) continue;
3176    
3177        if (minimize)        if (minimize)
3178          {          {
3179          for (fi = min;; fi++)          for (fi = min;; fi++)
3180            {            {
3181            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3182            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3183            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3184              RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
3185                {
3186                SCHECK_PARTIAL();
3187                MRRETURN(MATCH_NOMATCH);
3188                }
3189              if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3190            }            }
3191          /* Control never gets here */          /* Control never gets here */
3192          }          }
3193        else        else  /* Maximize */
3194          {          {
3195          pp = eptr;          pp = eptr;
3196          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3197            {            {
3198            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3199                {
3200                SCHECK_PARTIAL();
3201                break;
3202                }
3203              if (fc != *eptr) break;
3204            eptr++;            eptr++;
3205            }            }
3206            if (possessive) continue;
3207    
3208          while (eptr >= pp)          while (eptr >= pp)
3209            {            {
3210            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3211            eptr--;            eptr--;
3212            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3213            }            }
3214          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3215          }          }
3216        }        }
3217      /* Control never gets here */      /* Control never gets here */
# Line 2084  for (;;) Line 3220  for (;;)
3220      checking can be multibyte. */      checking can be multibyte. */
3221    
3222      case OP_NOT:      case OP_NOT:
3223      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3224        if (eptr >= md->end_subject)
3225          {
3226          SCHECK_PARTIAL();
3227          MRRETURN(MATCH_NOMATCH);
3228          }
3229      ecode++;      ecode++;
3230      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3231      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3232        {        {
3233  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3234        if (c < 256)        if (c < 256)
3235  #endif  #endif
3236        c = md->lcc[c];        c = md->lcc[c];
3237        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);
3238        }        }
3239      else      else    /* Caseful */
3240        {        {
3241        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);
3242        }        }
3243      break;      break;
3244    
# Line 2109  for (;;) Line 3250  for (;;)
3250      about... */      about... */
3251    
3252      case OP_NOTEXACT:      case OP_NOTEXACT:
3253        case OP_NOTEXACTI:
3254      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3255      ecode += 3;      ecode += 3;
3256      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3257    
3258      case OP_NOTUPTO:      case OP_NOTUPTO:
3259        case OP_NOTUPTOI:
3260      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3261        case OP_NOTMINUPTOI:
3262        min = 0;
3263        max = GET2(ecode, 1);
3264        minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3265        ecode += 3;
3266        goto REPEATNOTCHAR;
3267    
3268        case OP_NOTPOSSTAR:
3269        case OP_NOTPOSSTARI:
3270        possessive = TRUE;
3271        min = 0;
3272        max = INT_MAX;
3273        ecode++;
3274        goto REPEATNOTCHAR;
3275    
3276        case OP_NOTPOSPLUS:
3277        case OP_NOTPOSPLUSI:
3278        possessive = TRUE;
3279        min = 1;
3280        max = INT_MAX;
3281        ecode++;
3282        goto REPEATNOTCHAR;
3283    
3284        case OP_NOTPOSQUERY:
3285        case OP_NOTPOSQUERYI:
3286        possessive = TRUE;
3287        min = 0;
3288        max = 1;
3289        ecode++;
3290        goto REPEATNOTCHAR;
3291    
3292        case OP_NOTPOSUPTO:
3293        case OP_NOTPOSUPTOI:
3294        possessive = TRUE;
3295      min = 0;      min = 0;
3296      max = GET2(ecode, 1);      max = GET2(ecode, 1);
     minimize = *ecode == OP_NOTMINUPTO;  
3297      ecode += 3;      ecode += 3;
3298      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3299    
3300      case OP_NOTSTAR:      case OP_NOTSTAR:
3301        case OP_NOTSTARI:
3302      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
3303        case OP_NOTMINSTARI:
3304      case OP_NOTPLUS:      case OP_NOTPLUS:
3305        case OP_NOTPLUSI:
3306      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
3307        case OP_NOTMINPLUSI:
3308      case OP_NOTQUERY:      case OP_NOTQUERY:
3309        case OP_NOTQUERYI:
3310      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
3311      c = *ecode++ - OP_NOTSTAR;      case OP_NOTMINQUERYI:
3312        c = *ecode++ - ((op >= OP_NOTSTARI)? OP_NOTSTARI: OP_NOTSTAR);
3313      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3314      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3315      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3316      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3317    
3318      /* 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. */  
3319    
3320      REPEATNOTCHAR:      REPEATNOTCHAR:
     if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3321      fc = *ecode++;      fc = *ecode++;
3322    
3323      /* 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 2152  for (;;) Line 3331  for (;;)
3331      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,
3332        max, eptr));        max, eptr));
3333    
3334      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_NOTSTARI)     /* Caseless */
3335        {        {
3336        fc = md->lcc[fc];        fc = md->lcc[fc];
3337    
# Line 2160  for (;;) Line 3339  for (;;)
3339        /* UTF-8 mode */        /* UTF-8 mode */
3340        if (utf8)        if (utf8)
3341          {          {
3342          register int d;          register unsigned int d;
3343          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3344            {            {
3345              if (eptr >= md->end_subject)
3346                {
3347                SCHECK_PARTIAL();
3348                MRRETURN(MATCH_NOMATCH);
3349                }
3350            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3351            if (d < 256) d = md->lcc[d];            if (d < 256) d = md->lcc[d];
3352            if (fc == d) RRETURN(MATCH_NOMATCH);            if (fc == d) MRRETURN(MATCH_NOMATCH);
3353            }            }
3354          }          }
3355        else        else
# Line 2174  for (;;) Line 3358  for (;;)
3358        /* Not UTF-8 mode */        /* Not UTF-8 mode */
3359          {          {
3360          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3361            if (fc == md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            {
3362              if (eptr >= md->end_subject)
3363                {
3364                SCHECK_PARTIAL();
3365                MRRETURN(MATCH_NOMATCH);
3366                }
3367              if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3368              }
3369          }          }
3370    
3371        if (min == max) continue;        if (min == max) continue;
# Line 2185  for (;;) Line 3376  for (;;)
3376          /* UTF-8 mode */          /* UTF-8 mode */
3377          if (utf8)          if (utf8)
3378            {            {
3379            register int d;            register unsigned int d;
3380            for (fi = min;; fi++)            for (fi = min;; fi++)
3381              {              {
3382              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM28);
3383              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3384                if (fi >= max) MRRETURN(MATCH_NOMATCH);
3385                if (eptr >= md->end_subject)
3386                  {
3387                  SCHECK_PARTIAL();
3388                  MRRETURN(MATCH_NOMATCH);
3389                  }
3390              GETCHARINC(d, eptr);              GETCHARINC(d, eptr);
3391              if (d < 256) d = md->lcc[d];              if (d < 256) d = md->lcc[d];
3392              if (fi >= max || eptr >= md->end_subject || fc == d)              if (fc == d) MRRETURN(MATCH_NOMATCH);
               RRETURN(MATCH_NOMATCH);  
3393              }              }
3394            }            }
3395          else          else
# Line 2202  for (;;) Line 3398  for (;;)
3398            {            {
3399            for (fi = min;; fi++)            for (fi = min;; fi++)
3400              {              {
3401              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM29);
3402              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3403              if (fi >= max || eptr >= md->end_subject || fc == md->lcc[*eptr++])              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3404                RRETURN(MATCH_NOMATCH);              if (eptr >= md->end_subject)
3405                  {
3406                  SCHECK_PARTIAL();
3407                  MRRETURN(MATCH_NOMATCH);
3408                  }
3409                if (fc == md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3410              }              }
3411            }            }
3412          /* Control never gets here */          /* Control never gets here */
# Line 2221  for (;;) Line 3422  for (;;)
3422          /* UTF-8 mode */          /* UTF-8 mode */
3423          if (utf8)          if (utf8)
3424            {            {
3425            register int d;            register unsigned int d;
3426            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3427              {              {
3428              int len = 1;              int len = 1;
3429              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3430                  {
3431                  SCHECK_PARTIAL();
3432                  break;
3433                  }
3434              GETCHARLEN(d, eptr, len);              GETCHARLEN(d, eptr, len);
3435              if (d < 256) d = md->lcc[d];              if (d < 256) d = md->lcc[d];
3436              if (fc == d) break;              if (fc == d) break;
3437              eptr += len;              eptr += len;
3438              }              }
3439            for(;;)          if (possessive) continue;
3440            for(;;)
3441              {              {
3442              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM30);
3443              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3444              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3445              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2245  for (;;) Line 3451  for (;;)
3451            {            {
3452            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3453              {              {
3454              if (eptr >= md->end_subject || fc == md->lcc[*eptr]) break;              if (eptr >= md->end_subject)
3455                  {
3456                  SCHECK_PARTIAL();
3457                  break;
3458                  }
3459                if (fc == md->lcc[*eptr]) break;
3460              eptr++;              eptr++;
3461              }              }
3462              if (possessive) continue;
3463            while (eptr >= pp)            while (eptr >= pp)
3464              {              {
3465              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM31);
3466              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3467              eptr--;              eptr--;
3468              }              }
3469            }            }
3470    
3471          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3472          }          }
3473        /* Control never gets here */        /* Control never gets here */
3474        }        }
# Line 2269  for (;;) Line 3481  for (;;)
3481        /* UTF-8 mode */        /* UTF-8 mode */
3482        if (utf8)        if (utf8)
3483          {          {
3484          register int d;          register unsigned int d;
3485          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3486            {            {
3487              if (eptr >= md->end_subject)
3488                {
3489                SCHECK_PARTIAL();
3490                MRRETURN(MATCH_NOMATCH);
3491                }
3492            GETCHARINC(d, eptr);            GETCHARINC(d, eptr);
3493            if (fc == d) RRETURN(MATCH_NOMATCH);            if (fc == d) MRRETURN(MATCH_NOMATCH);
3494            }            }
3495          }          }
3496        else        else
# Line 2281  for (;;) Line 3498  for (;;)
3498        /* Not UTF-8 mode */        /* Not UTF-8 mode */
3499          {          {
3500          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3501            if (fc == *eptr++) RRETURN(MATCH_NOMATCH);            {
3502              if (eptr >= md->end_subject)
3503                {
3504                SCHECK_PARTIAL();
3505                MRRETURN(MATCH_NOMATCH);
3506                }
3507              if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
3508              }
3509          }          }
3510    
3511        if (min == max) continue;        if (min == max) continue;
# Line 2292  for (;;) Line 3516  for (;;)
3516          /* UTF-8 mode */          /* UTF-8 mode */
3517          if (utf8)          if (utf8)
3518            {            {
3519            register int d;            register unsigned int d;
3520            for (fi = min;; fi++)            for (fi = min;; fi++)
3521              {              {
3522              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM32);
3523              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3524                if (fi >= max) MRRETURN(MATCH_NOMATCH);
3525                if (eptr >= md->end_subject)
3526                  {
3527                  SCHECK_PARTIAL();
3528                  MRRETURN(MATCH_NOMATCH);
3529                  }
3530              GETCHARINC(d, eptr);              GETCHARINC(d, eptr);
3531              if (fi >= max || eptr >= md->end_subject || fc == d)              if (fc == d) MRRETURN(MATCH_NOMATCH);
               RRETURN(MATCH_NOMATCH);  
3532              }              }
3533            }            }
3534          else          else
# Line 2308  for (;;) Line 3537  for (;;)
3537            {            {
3538            for (fi = min;; fi++)            for (fi = min;; fi++)
3539              {              {
3540              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM33);
3541              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3542              if (fi >= max || eptr >= md->end_subject || fc == *eptr++)              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3543                RRETURN(MATCH_NOMATCH);              if (eptr >= md->end_subject)
3544                  {
3545                  SCHECK_PARTIAL();
3546                  MRRETURN(MATCH_NOMATCH);
3547                  }
3548                if (fc == *eptr++) MRRETURN(MATCH_NOMATCH);
3549              }              }
3550            }            }
3551          /* Control never gets here */          /* Control never gets here */
# Line 2327  for (;;) Line 3561  for (;;)
3561          /* UTF-8 mode */          /* UTF-8 mode */
3562          if (utf8)          if (utf8)
3563            {            {
3564            register int d;            register unsigned int d;
3565            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3566              {              {
3567              int len = 1;              int len = 1;
3568              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3569                  {
3570                  SCHECK_PARTIAL();
3571                  break;
3572                  }
3573              GETCHARLEN(d, eptr, len);              GETCHARLEN(d, eptr, len);
3574              if (fc == d) break;              if (fc == d) break;
3575              eptr += len;              eptr += len;
3576              }              }
3577              if (possessive) continue;
3578            for(;;)            for(;;)
3579              {              {
3580              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM34);
3581             &