/[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 891 by ph10, Wed Jan 18 16:26:18 2012 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) 1997-2012 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  /* 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.
86  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,
87  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 95  static const char rep_max[] = { 0, 0, 0,
95    
96    
97    
98  #ifdef DEBUG  #ifdef PCRE_DEBUG
99  /*************************************************  /*************************************************
100  *        Debugging function to print chars       *  *        Debugging function to print chars       *
101  *************************************************/  *************************************************/
# Line 99  Returns:     nothing Line 113  Returns:     nothing
113  */  */
114    
115  static void  static void
116  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
117  {  {
118  int c;  unsigned int c;
119  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;
120  while (length-- > 0)  while (length-- > 0)
121    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 128  while (length-- > 0)
128  *          Match a back-reference                *  *          Match a back-reference                *
129  *************************************************/  *************************************************/
130    
131  /* 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
132  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
133    the length passed is zero. Note that in caseless UTF-8 mode, the number of
134    subject bytes matched may be different to the number of reference bytes.
135    
136  Arguments:  Arguments:
137    offset      index into the offset vector    offset      index into the offset vector
138    eptr        points into the subject    eptr        pointer into the subject
139    length      length to be matched    length      length of reference to be matched (number of bytes)
140    md          points to match data block    md          points to match data block
141    ims         the ims flags    caseless    TRUE if caseless
142    
143  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
144  */  */
145    
146  static BOOL  static int
147  match_ref(int offset, register const uschar *eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
148    unsigned long int ims)    BOOL caseless)
149  {  {
150  const uschar *p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
151    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
152    
153  #ifdef DEBUG  #ifdef PCRE_DEBUG
154  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
155    printf("matching subject <null>");    printf("matching subject <null>");
156  else  else
# Line 146  pchars(p, length, FALSE, md); Line 163  pchars(p, length, FALSE, md);
163  printf("\n");  printf("\n");
164  #endif  #endif
165    
166  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
167    
168  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
169    
170  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
171    properly if Unicode properties are supported. Otherwise, we can check only
172    ASCII characters. */
173    
174  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
175    {    {
176    while (length-- > 0)  #ifdef SUPPORT_UTF
177      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
178      if (md->utf)
179        {
180        /* Match characters up to the end of the reference. NOTE: the number of
181        bytes matched may differ, because there are some characters whose upper and
182        lower case versions code as different numbers of bytes. For example, U+023A
183        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
184        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
185        the latter. It is important, therefore, to check the length along the
186        reference, not along the subject (earlier code did this wrong). */
187    
188        PCRE_PUCHAR endptr = p + length;
189        while (p < endptr)
190          {
191          int c, d;
192          if (eptr >= md->end_subject) return -1;
193          GETCHARINC(c, eptr);
194          GETCHARINC(d, p);
195          if (c != d && c != UCD_OTHERCASE(d)) return -1;
196          }
197        }
198      else
199    #endif
200    #endif
201    
202      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
203      is no UCP support. */
204        {
205        if (eptr + length > md->end_subject) return -1;
206        while (length-- > 0)
207          {
208          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
209          p++;
210          eptr++;
211          }
212        }
213    }    }
214    
215    /* In the caseful case, we can just compare the bytes, whether or not we
216    are in UTF-8 mode. */
217    
218  else  else
219    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
220      if (eptr + length > md->end_subject) return -1;
221      while (length-- > 0) if (*p++ != *eptr++) return -1;
222      }
223    
224  return TRUE;  return (int)(eptr - eptr_start);
225  }  }
226    
227    
# Line 169  return TRUE; Line 230  return TRUE;
230  ****************************************************************************  ****************************************************************************
231                     RECURSION IN THE match() FUNCTION                     RECURSION IN THE match() FUNCTION
232    
233  The match() function is highly recursive. Some regular expressions can cause  The match() function is highly recursive, though not every recursive call
234  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
235  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
236  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
237  works fine.  saved for a recursive call. On Unix, the stack can be large, and this works
238    fine.
239  It turns out that on non-Unix systems there are problems with programs that  
240  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
241  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
242  for decades.) So....  has oodles of memory these days, and techniques for extending the stack have
243    been known for decades.) So....
244    
245  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
246  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
247  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
248  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
249  always used to.  always used to.
250    
251    The original heap-recursive code used longjmp(). However, it seems that this
252    can be very slow on some operating systems. Following a suggestion from Stan
253    Switzer, the use of longjmp() has been abolished, at the cost of having to
254    provide a unique number for each call to RMATCH. There is no way of generating
255    a sequence of numbers at compile time in C. I have given them names, to make
256    them stand out more clearly.
257    
258    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
259    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
260    tests. Furthermore, not using longjmp() means that local dynamic variables
261    don't have indeterminate values; this has meant that the frame size can be
262    reduced because the result can be "passed back" by straight setting of the
263    variable instead of being passed in the frame.
264  ****************************************************************************  ****************************************************************************
265  ***************************************************************************/  ***************************************************************************/
266    
267    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
268    below must be updated in sync.  */
269    
270  /* These versions of the macros use the stack, as normal */  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
271           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
272           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
273           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
274           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
275           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
276           RM61,  RM62, RM63, RM64, RM65, RM66 };
277    
278    /* These versions of the macros use the stack, as normal. There are debugging
279    versions and production versions. Note that the "rw" argument of RMATCH isn't
280    actually used in this definition. */
281    
282  #ifndef NO_RECURSE  #ifndef NO_RECURSE
283  #define REGISTER register  #define REGISTER register
284  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg)  
285    #ifdef PCRE_DEBUG
286    #define RMATCH(ra,rb,rc,rd,re,rw) \
287      { \
288      printf("match() called in line %d\n", __LINE__); \
289      rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
290      printf("to line %d\n", __LINE__); \
291      }
292    #define RRETURN(ra) \
293      { \
294      printf("match() returned %d from line %d ", ra, __LINE__); \
295      return ra; \
296      }
297    #else
298    #define RMATCH(ra,rb,rc,rd,re,rw) \
299      rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
300  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
301    #endif
302    
303  #else  #else
304    
305    
306  /* 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
307  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
308  match(), which never changes. */  argument of match(), which never changes. */
309    
310  #define REGISTER  #define REGISTER
311    
312  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
313    {\    {\
314    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
315    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
316      {\    frame->Xwhere = rw; \
317      newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
318      newframe->Xecode = rb;\    newframe->Xecode = rb;\
319      newframe->Xoffset_top = rc;\    newframe->Xmstart = mstart;\
320      newframe->Xims = re;\    newframe->Xoffset_top = rc;\
321      newframe->Xeptrb = rf;\    newframe->Xeptrb = re;\
322      newframe->Xflags = rg;\    newframe->Xrdepth = frame->Xrdepth + 1;\
323      newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
324      frame = newframe;\    frame = newframe;\
325      DPRINTF(("restarting from line %d\n", __LINE__));\    DPRINTF(("restarting from line %d\n", __LINE__));\
326      goto HEAP_RECURSE;\    goto HEAP_RECURSE;\
327      }\    L_##rw:\
328    else\    DPRINTF(("jumped back to line %d\n", __LINE__));\
     {\  
     DPRINTF(("longjumped back to line %d\n", __LINE__));\  
     frame = md->thisframe;\  
     rx = frame->Xresult;\  
     }\  
329    }    }
330    
331  #define RRETURN(ra)\  #define RRETURN(ra)\
332    {\    {\
333    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
334    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
335    (pcre_stack_free)(newframe);\    (PUBL(stack_free))(oldframe);\
336    if (frame != NULL)\    if (frame != NULL)\
337      {\      {\
338      frame->Xresult = ra;\      rrc = ra;\
339      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
340      }\      }\
341    return ra;\    return ra;\
342    }    }
# Line 250  typedef struct heapframe { Line 349  typedef struct heapframe {
349    
350    /* Function arguments that may change */    /* Function arguments that may change */
351    
352    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
353    const uschar *Xecode;    const pcre_uchar *Xecode;
354      PCRE_PUCHAR Xmstart;
355    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
356    eptrblock *Xeptrb;    eptrblock *Xeptrb;
357    int Xflags;    unsigned int Xrdepth;
358    
359    /* Function local variables */    /* Function local variables */
360    
361    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
362    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
363    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
364    const uschar *Xnext;  #endif
365    const uschar *Xpp;    PCRE_PUCHAR Xdata;
366    const uschar *Xprev;    PCRE_PUCHAR Xnext;
367    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
368      PCRE_PUCHAR Xprev;
369      PCRE_PUCHAR Xsaved_eptr;
370    
371    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
372    
373    BOOL Xcur_is_word;    BOOL Xcur_is_word;
374    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
375    BOOL Xprev_is_word;    BOOL Xprev_is_word;
376    
   unsigned long int Xoriginal_ims;  
   
377  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
378    int Xprop_type;    int Xprop_type;
379      int Xprop_value;
380    int Xprop_fail_result;    int Xprop_fail_result;
381    int Xprop_category;    int Xoclength;
382    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_othercase;  
   int Xprop_test_against;  
   int *Xprop_test_variable;  
383  #endif  #endif
384    
385      int Xcodelink;
386    int Xctype;    int Xctype;
387    int Xfc;    unsigned int Xfc;
388    int Xfi;    int Xfi;
389    int Xlength;    int Xlength;
390    int Xmax;    int Xmax;
# Line 301  typedef struct heapframe { Line 398  typedef struct heapframe {
398    
399    eptrblock Xnewptrb;    eptrblock Xnewptrb;
400    
401    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
402    
403    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
404    
405  } heapframe;  } heapframe;
406    
# Line 320  typedef struct heapframe { Line 416  typedef struct heapframe {
416  *         Match from current position            *  *         Match from current position            *
417  *************************************************/  *************************************************/
418    
419  /* 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  
420  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
421  same response.  same response. */
422    
423    /* These macros pack up tests that are used for partial matching, and which
424    appear several times in the code. We set the "hit end" flag if the pointer is
425    at the end of the subject and also past the start of the subject (i.e.
426    something has been matched). For hard partial matching, we then return
427    immediately. The second one is used when we already know we are past the end of
428    the subject. */
429    
430    #define CHECK_PARTIAL()\
431      if (md->partial != 0 && eptr >= md->end_subject && \
432          eptr > md->start_used_ptr) \
433        { \
434        md->hitend = TRUE; \
435        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
436        }
437    
438    #define SCHECK_PARTIAL()\
439      if (md->partial != 0 && eptr > md->start_used_ptr) \
440        { \
441        md->hitend = TRUE; \
442        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
443        }
444    
445  Performance note: It might be tempting to extract commonly used fields from the  
446  md structure (e.g. utf8, end_subject) into individual variables to improve  /* Performance note: It might be tempting to extract commonly used fields from
447    the md structure (e.g. utf, end_subject) into individual variables to improve
448  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
449  made performance worse.  made performance worse.
450    
451  Arguments:  Arguments:
452     eptr        pointer in subject     eptr        pointer to current character in subject
453     ecode       position in code     ecode       pointer to current position in compiled code
454       mstart      pointer to the current match start position (can be modified
455                     by encountering \K)
456     offset_top  current top pointer     offset_top  current top pointer
457     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
458     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
459                   brackets - for testing for empty matches                   brackets - for testing for empty matches
460     flags       can contain     rdepth      the recursion depth
                  match_condassert - this is an assertion condition  
                  match_isgroup - this is the start of a bracketed group  
461    
462  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
463                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
464                   a negative MATCH_xxx value for PRUNE, SKIP, etc
465                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
466                   (e.g. stopped by recursion limit)                   (e.g. stopped by repeated call or recursion limit)
467  */  */
468    
469  static int  static int
470  match(REGISTER const uschar *eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
471    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
472    int flags)    unsigned int rdepth)
473  {  {
474  /* 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,
475  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
476  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
477    
478  register int  rrc;    /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
479  register int  i;      /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
480  register int  c;      /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
481  register BOOL utf8;   /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
482    
483    BOOL minimize, possessive; /* Quantifier options */
484    BOOL caseless;
485    int condcode;
486    
487  /* 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
488  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 490  heap storage. Set up the top-level frame
490  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
491    
492  #ifdef NO_RECURSE  #ifdef NO_RECURSE
493  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));
494    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
495  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
496    
497  /* Copy in the original argument variables */  /* Copy in the original argument variables */
498    
499  frame->Xeptr = eptr;  frame->Xeptr = eptr;
500  frame->Xecode = ecode;  frame->Xecode = ecode;
501    frame->Xmstart = mstart;
502  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
503  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
504  frame->Xflags = flags;  frame->Xrdepth = rdepth;
505    
506  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
507    
# Line 390  HEAP_RECURSE: Line 511  HEAP_RECURSE:
511    
512  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
513  #define ecode              frame->Xecode  #define ecode              frame->Xecode
514    #define mstart             frame->Xmstart
515  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
516  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
517  #define flags              frame->Xflags  #define rdepth             frame->Xrdepth
518    
519  /* Ditto for the local variables */  /* Ditto for the local variables */
520    
521  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
522  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
523  #endif  #endif
524  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
525    #define codelink           frame->Xcodelink
526  #define data               frame->Xdata  #define data               frame->Xdata
527  #define next               frame->Xnext  #define next               frame->Xnext
528  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 411  HEAP_RECURSE: Line 533  HEAP_RECURSE:
533    
534  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
535  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
536  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
537    
 #define original_ims       frame->Xoriginal_ims  
   
538  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
539  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
540    #define prop_value         frame->Xprop_value
541  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
542  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
543  #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  
544  #endif  #endif
545    
546  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 447  HEAP_RECURSE: Line 564  HEAP_RECURSE:
564  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
565  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
566    
567  #else  #else         /* NO_RECURSE not defined */
568  #define fi i  #define fi i
569  #define fc c  #define fc c
570    
571    /* Many of the following variables are used only in small blocks of the code.
572    My normal style of coding would have declared them within each of those blocks.
573    However, in order to accommodate the version of this code that uses an external
574    "stack" implemented on the heap, it is easier to declare them all here, so the
575    declarations can be cut out in a block. The only declarations within blocks
576    below are for variables that do not have to be preserved over a recursive call
577    to RMATCH(). */
578    
579  #ifdef SUPPORT_UTF8                /* Many of these variables are used ony */  #ifdef SUPPORT_UTF
580  const uschar *charptr;             /* small blocks of the code. My normal  */  const pcre_uchar *charptr;
581  #endif                             /* style of coding would have declared  */  #endif
582  const uschar *callpat;             /* them within each of those blocks.    */  const pcre_uchar *callpat;
583  const uschar *data;                /* However, in order to accommodate the */  const pcre_uchar *data;
584  const uschar *next;                /* version of this code that uses an    */  const pcre_uchar *next;
585  const uschar *pp;                  /* external "stack" implemented on the  */  PCRE_PUCHAR       pp;
586  const uschar *prev;                /* heap, it is easier to declare them   */  const pcre_uchar *prev;
587  const uschar *saved_eptr;          /* all here, so the declarations can    */  PCRE_PUCHAR       saved_eptr;
588                                     /* be cut out in a block. The only      */  
589  recursion_info new_recursive;      /* declarations within blocks below are */  recursion_info new_recursive;
                                    /* for variables that do not have to    */  
 BOOL cur_is_word;                  /* be preserved over a recursive call   */  
 BOOL condition;                    /* to RMATCH().                         */  
 BOOL minimize;  
 BOOL prev_is_word;  
590    
591  unsigned long int original_ims;  BOOL cur_is_word;
592    BOOL condition;
593    BOOL prev_is_word;
594    
595  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
596  int prop_type;  int prop_type;
597    int prop_value;
598  int prop_fail_result;  int prop_fail_result;
599  int prop_category;  int oclength;
600  int prop_chartype;  pcre_uchar occhars[6];
 int prop_othercase;  
 int prop_test_against;  
 int *prop_test_variable;  
601  #endif  #endif
602    
603    int codelink;
604  int ctype;  int ctype;
605  int length;  int length;
606  int max;  int max;
# Line 493  int save_offset1, save_offset2, save_off Line 613  int save_offset1, save_offset2, save_off
613  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
614    
615  eptrblock newptrb;  eptrblock newptrb;
616  #endif  #endif     /* NO_RECURSE */
617    
618    /* To save space on the stack and in the heap frame, I have doubled up on some
619    of the local variables that are used only in localised parts of the code, but
620    still need to be preserved over recursive calls of match(). These macros define
621    the alternative names that are used. */
622    
623    #define allow_zero    cur_is_word
624    #define cbegroup      condition
625    #define code_offset   codelink
626    #define condassert    condition
627    #define matched_once  prev_is_word
628    #define foc           number
629    #define save_mark     data
630    
631  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
632  variables. */  variables. */
633    
634  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
635    prop_value = 0;
636  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_against = 0;  
 prop_test_variable = NULL;  
637  #endif  #endif
638    
 /* 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. */  
639    
640  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
641    when NO_RECURSE is not defined, in order to reduce the amount of stack that is
642    used. Thanks to Ian Taylor for noticing this possibility and sending the
643    original patch. */
644    
645    TAIL_RECURSE:
646    
647    /* OK, now we can get on with the real code of the function. Recursive calls
648    are specified by the macro RMATCH and RRETURN is used to return. When
649    NO_RECURSE is *not* defined, these just turn into a recursive call to match()
650    and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
651    defined). However, RMATCH isn't like a function call because it's quite a
652    complicated macro. It has to be used in one particular way. This shouldn't,
653    however, impact performance when true recursion is being used. */
654    
655    #ifdef SUPPORT_UTF
656    utf = md->utf;       /* Local copy of the flag */
657    #else
658    utf = FALSE;
659    #endif
660    
661    /* First check that we haven't called match() too many times, or that we
662    haven't exceeded the recursive call limit. */
663    
664  original_ims = ims;    /* Save for resetting on ')' */  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
665  utf8 = md->utf8;       /* Local copy of the flag */  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
666    
667  /* 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
668  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
669  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
670  this stack. */  up space on the stack. See also MATCH_CONDASSERT below.
671    
672    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
673    such remembered pointers, to be checked when we hit the closing ket, in order
674    to break infinite loops that match no characters. When match() is called in
675    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
676    NOT be used with tail recursion, because the memory block that is used is on
677    the stack, so a new one may be required for each match(). */
678    
679  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
680    {    {
   newptrb.epb_prev = eptrb;  
681    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
682      newptrb.epb_prev = eptrb;
683    eptrb = &newptrb;    eptrb = &newptrb;
684      md->match_function_type = 0;
685    }    }
686    
687  /* Now start processing the operations. */  /* Now start processing the opcodes. */
688    
689  for (;;)  for (;;)
690    {    {
691      minimize = possessive = FALSE;
692    op = *ecode;    op = *ecode;
   minimize = FALSE;  
693    
694    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
695    matching at least one subject character. */      {
696        case OP_MARK:
697        md->nomatch_mark = ecode + 2;
698        md->mark = NULL;    /* In case previously set by assertion */
699        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
700          eptrb, RM55);
701        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
702             md->mark == NULL) md->mark = ecode + 2;
703    
704        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
705        argument, and we must check whether that argument matches this MARK's
706        argument. It is passed back in md->start_match_ptr (an overloading of that
707        variable). If it does match, we reset that variable to the current subject
708        position and return MATCH_SKIP. Otherwise, pass back the return code
709        unaltered. */
710    
711        else if (rrc == MATCH_SKIP_ARG &&
712            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
713          {
714          md->start_match_ptr = eptr;
715          RRETURN(MATCH_SKIP);
716          }
717        RRETURN(rrc);
718    
719        case OP_FAIL:
720        RRETURN(MATCH_NOMATCH);
721    
722    if (md->partial &&      /* COMMIT overrides PRUNE, SKIP, and 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. */  
723    
724    if (op > OP_BRA)      case OP_COMMIT:
725      {      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
726      number = op - OP_BRA;        eptrb, RM52);
727        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
728            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
729            rrc != MATCH_THEN)
730          RRETURN(rrc);
731        RRETURN(MATCH_COMMIT);
732    
733        /* PRUNE overrides THEN */
734    
735        case OP_PRUNE:
736        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
737          eptrb, RM51);
738        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
739        RRETURN(MATCH_PRUNE);
740    
741        case OP_PRUNE_ARG:
742        md->nomatch_mark = ecode + 2;
743        md->mark = NULL;    /* In case previously set by assertion */
744        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
745          eptrb, RM56);
746        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
747             md->mark == NULL) md->mark = ecode + 2;
748        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
749        RRETURN(MATCH_PRUNE);
750    
751        /* SKIP overrides PRUNE and THEN */
752    
753        case OP_SKIP:
754        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
755          eptrb, RM53);
756        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
757          RRETURN(rrc);
758        md->start_match_ptr = eptr;   /* Pass back current position */
759        RRETURN(MATCH_SKIP);
760    
761        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
762        nomatch_mark. There is a flag that disables this opcode when re-matching a
763        pattern that ended with a SKIP for which there was not a matching MARK. */
764    
765        case OP_SKIP_ARG:
766        if (md->ignore_skip_arg)
767          {
768          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
769          break;
770          }
771        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
772          eptrb, RM57);
773        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
774          RRETURN(rrc);
775    
776        /* Pass back the current skip name by overloading md->start_match_ptr and
777        returning the special MATCH_SKIP_ARG return code. This will either be
778        caught by a matching MARK, or get to the top, where it causes a rematch
779        with the md->ignore_skip_arg flag set. */
780    
781        md->start_match_ptr = ecode + 2;
782        RRETURN(MATCH_SKIP_ARG);
783    
784        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
785        the branch in which it occurs can be determined. Overload the start of
786        match pointer to do this. */
787    
788        case OP_THEN:
789        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
790          eptrb, RM54);
791        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
792        md->start_match_ptr = ecode;
793        RRETURN(MATCH_THEN);
794    
795        case OP_THEN_ARG:
796        md->nomatch_mark = ecode + 2;
797        md->mark = NULL;    /* In case previously set by assertion */
798        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
799          md, eptrb, RM58);
800        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
801             md->mark == NULL) md->mark = ecode + 2;
802        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
803        md->start_match_ptr = ecode;
804        RRETURN(MATCH_THEN);
805    
806        /* Handle an atomic group that does not contain any capturing parentheses.
807        This can be handled like an assertion. Prior to 8.13, all atomic groups
808        were handled this way. In 8.13, the code was changed as below for ONCE, so
809        that backups pass through the group and thereby reset captured values.
810        However, this uses a lot more stack, so in 8.20, atomic groups that do not
811        contain any captures generate OP_ONCE_NC, which can be handled in the old,
812        less stack intensive way.
813    
814        Check the alternative branches in turn - the matching won't pass the KET
815        for this kind of subpattern. If any one branch matches, we carry on as at
816        the end of a normal bracket, leaving the subject pointer, but resetting
817        the start-of-match value in case it was changed by \K. */
818    
819        case OP_ONCE_NC:
820        prev = ecode;
821        saved_eptr = eptr;
822        save_mark = md->mark;
823        do
824          {
825          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
826          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
827            {
828            mstart = md->start_match_ptr;
829            break;
830            }
831          if (rrc == MATCH_THEN)
832            {
833            next = ecode + GET(ecode,1);
834            if (md->start_match_ptr < next &&
835                (*ecode == OP_ALT || *next == OP_ALT))
836              rrc = MATCH_NOMATCH;
837            }
838    
839          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
840          ecode += GET(ecode,1);
841          md->mark = save_mark;
842          }
843        while (*ecode == OP_ALT);
844    
845        /* If hit the end of the group (which could be repeated), fail */
846    
847        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
848    
849        /* Continue as from after the group, updating the offsets high water
850        mark, since extracts may have been taken. */
851    
852        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
853    
854        offset_top = md->end_offset_top;
855        eptr = md->end_match_ptr;
856    
857        /* For a non-repeating ket, just continue at this level. This also
858        happens for a repeating ket if no characters were matched in the group.
859        This is the forcible breaking of infinite loops as implemented in Perl
860        5.005. */
861    
862        if (*ecode == OP_KET || eptr == saved_eptr)
863          {
864          ecode += 1+LINK_SIZE;
865          break;
866          }
867    
868      /* For extended extraction brackets (large number), we have to fish out the      /* The repeating kets try the rest of the pattern or restart from the
869      number from a dummy opcode at the start. */      preceding bracket, in the appropriate order. The second "call" of match()
870        uses tail recursion, to avoid using another stack frame. */
871    
872        if (*ecode == OP_KETRMIN)
873          {
874          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
875          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
876          ecode = prev;
877          goto TAIL_RECURSE;
878          }
879        else  /* OP_KETRMAX */
880          {
881          md->match_function_type = MATCH_CBEGROUP;
882          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
883          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
884          ecode += 1 + LINK_SIZE;
885          goto TAIL_RECURSE;
886          }
887        /* Control never gets here */
888    
889      if (number > EXTRACT_BASIC_MAX)      /* Handle a capturing bracket, other than those that are possessive with an
890        number = GET2(ecode, 2+LINK_SIZE);      unlimited repeat. If there is space in the offset vector, save the current
891        subject position in the working slot at the top of the vector. We mustn't
892        change the current values of the data slot, because they may be set from a
893        previous iteration of this group, and be referred to by a reference inside
894        the group. A failure to match might occur after the group has succeeded,
895        if something later on doesn't match. For this reason, we need to restore
896        the working value and also the values of the final offsets, in case they
897        were set by a previous iteration of the same bracket.
898    
899        If there isn't enough space in the offset vector, treat this as if it were
900        a non-capturing bracket. Don't worry about setting the flag for the error
901        case here; that is handled in the code for KET. */
902    
903        case OP_CBRA:
904        case OP_SCBRA:
905        number = GET2(ecode, 1+LINK_SIZE);
906      offset = number << 1;      offset = number << 1;
907    
908  #ifdef DEBUG  #ifdef PCRE_DEBUG
909      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
910        printf("subject=");
911      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
912      printf("\n");      printf("\n");
913  #endif  #endif
# Line 580  for (;;) Line 918  for (;;)
918        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
919        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
920        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
921          save_mark = md->mark;
922    
923        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
924        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
925            (int)(eptr - md->start_subject);
926    
927        do        for (;;)
928          {          {
929          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
930            match_isgroup);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
931              eptrb, RM1);
932            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
933    
934            /* If we backed up to a THEN, check whether it is within the current
935            branch by comparing the address of the THEN that is passed back with
936            the end of the branch. If it is within the current branch, and the
937            branch is one of two or more alternatives (it either starts or ends
938            with OP_ALT), we have reached the limit of THEN's action, so convert
939            the return code to NOMATCH, which will cause normal backtracking to
940            happen from now on. Otherwise, THEN is passed back to an outer
941            alternative. This implements Perl's treatment of parenthesized groups,
942            where a group not containing | does not affect the current alternative,
943            that is, (X) is NOT the same as (X|(*F)). */
944    
945            if (rrc == MATCH_THEN)
946              {
947              next = ecode + GET(ecode,1);
948              if (md->start_match_ptr < next &&
949                  (*ecode == OP_ALT || *next == OP_ALT))
950                rrc = MATCH_NOMATCH;
951              }
952    
953            /* Anything other than NOMATCH is passed back. */
954    
955          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
956          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
957          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
958            md->mark = save_mark;
959            if (*ecode != OP_ALT) break;
960          }          }
       while (*ecode == OP_ALT);  
961    
962        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
963        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
964        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
965        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
966    
967        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
968    
969          RRETURN(rrc);
970        }        }
971    
972      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
973        as a non-capturing bracket. */
974    
975      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
976      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
977    
978    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
979    
980    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
981      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
982      case OP_BRA:     /* Non-capturing bracket: optimized */  
983      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
984      do      repeat and ONCE group with no captures. Loop for all the alternatives.
985    
986        When we get to the final alternative within the brackets, we used to return
987        the result of a recursive call to match() whatever happened so it was
988        possible to reduce stack usage by turning this into a tail recursion,
989        except in the case of a possibly empty group. However, now that there is
990        the possiblity of (*THEN) occurring in the final alternative, this
991        optimization is no longer always possible.
992    
993        We can optimize if we know there are no (*THEN)s in the pattern; at present
994        this is the best that can be done.
995    
996        MATCH_ONCE is returned when the end of an atomic group is successfully
997        reached, but subsequent matching fails. It passes back up the tree (causing
998        captured values to be reset) until the original atomic group level is
999        reached. This is tested by comparing md->once_target with the start of the
1000        group. At this point, the return is converted into MATCH_NOMATCH so that
1001        previous backup points can be taken. */
1002    
1003        case OP_ONCE:
1004        case OP_BRA:
1005        case OP_SBRA:
1006        DPRINTF(("start non-capturing bracket\n"));
1007    
1008        for (;;)
1009        {        {
1010        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1011          match_isgroup);  
1012        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        /* If this is not a possibly empty group, and there are no (*THEN)s in
1013          the pattern, and this is the final alternative, optimize as described
1014          above. */
1015    
1016          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1017            {
1018            ecode += PRIV(OP_lengths)[*ecode];
1019            goto TAIL_RECURSE;
1020            }
1021    
1022          /* In all other cases, we have to make another call to match(). */
1023    
1024          save_mark = md->mark;
1025          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1026            RM2);
1027    
1028          /* See comment in the code for capturing groups above about handling
1029          THEN. */
1030    
1031          if (rrc == MATCH_THEN)
1032            {
1033            next = ecode + GET(ecode,1);
1034            if (md->start_match_ptr < next &&
1035                (*ecode == OP_ALT || *next == OP_ALT))
1036              rrc = MATCH_NOMATCH;
1037            }
1038    
1039          if (rrc != MATCH_NOMATCH)
1040            {
1041            if (rrc == MATCH_ONCE)
1042              {
1043              const pcre_uchar *scode = ecode;
1044              if (*scode != OP_ONCE)           /* If not at start, find it */
1045                {
1046                while (*scode == OP_ALT) scode += GET(scode, 1);
1047                scode -= GET(scode, 1);
1048                }
1049              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1050              }
1051            RRETURN(rrc);
1052            }
1053        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1054          md->mark = save_mark;
1055          if (*ecode != OP_ALT) break;
1056        }        }
1057      while (*ecode == OP_ALT);  
     DPRINTF(("bracket 0 failed\n"));  
1058      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1059    
1060      /* Conditional group: compilation checked that there are no more than      /* Handle possessive capturing brackets with an unlimited repeat. We come
1061      two branches. If the condition is false, skipping the first branch takes us      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1062      past the end if there is only one branch, but that's OK because that is      handled similarly to the normal case above. However, the matching is
1063      exactly what going to the ket would do. */      different. The end of these brackets will always be OP_KETRPOS, which
1064        returns MATCH_KETRPOS without going further in the pattern. By this means
1065        we can handle the group by iteration rather than recursion, thereby
1066        reducing the amount of stack needed. */
1067    
1068        case OP_CBRAPOS:
1069        case OP_SCBRAPOS:
1070        allow_zero = FALSE;
1071    
1072      case OP_COND:      POSSESSIVE_CAPTURE:
1073      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      number = GET2(ecode, 1+LINK_SIZE);
1074        {      offset = number << 1;
       offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */  
       condition = (offset == CREF_RECURSE * 2)?  
         (md->recursive != NULL) :  
         (offset < offset_top && md->offset_vector[offset] >= 0);  
       RMATCH(rrc, eptr, ecode + (condition?  
         (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1))),  
         offset_top, md, ims, eptrb, match_isgroup);  
       RRETURN(rrc);  
       }  
1075    
1076      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1077      the final argument TRUE causes it to stop at the end of an assertion. */      printf("start possessive bracket %d\n", number);
1078        printf("subject=");
1079        pchars(eptr, 16, TRUE, md);
1080        printf("\n");
1081    #endif
1082    
1083      else      if (offset < md->offset_max)
1084        {        {
1085        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1086            match_condassert | match_isgroup);        code_offset = (int)(ecode - md->start_code);
1087        if (rrc == MATCH_MATCH)  
1088          save_offset1 = md->offset_vector[offset];
1089          save_offset2 = md->offset_vector[offset+1];
1090          save_offset3 = md->offset_vector[md->offset_end - number];
1091          save_capture_last = md->capture_last;
1092    
1093          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1094    
1095          /* Each time round the loop, save the current subject position for use
1096          when the group matches. For MATCH_MATCH, the group has matched, so we
1097          restart it with a new subject starting position, remembering that we had
1098          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1099          usual. If we haven't matched any alternatives in any iteration, check to
1100          see if a previous iteration matched. If so, the group has matched;
1101          continue from afterwards. Otherwise it has failed; restore the previous
1102          capture values before returning NOMATCH. */
1103    
1104          for (;;)
1105            {
1106            md->offset_vector[md->offset_end - number] =
1107              (int)(eptr - md->start_subject);
1108            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1109            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1110              eptrb, RM63);
1111            if (rrc == MATCH_KETRPOS)
1112              {
1113              offset_top = md->end_offset_top;
1114              eptr = md->end_match_ptr;
1115              ecode = md->start_code + code_offset;
1116              save_capture_last = md->capture_last;
1117              matched_once = TRUE;
1118              continue;
1119              }
1120    
1121            /* See comment in the code for capturing groups above about handling
1122            THEN. */
1123    
1124            if (rrc == MATCH_THEN)
1125              {
1126              next = ecode + GET(ecode,1);
1127              if (md->start_match_ptr < next &&
1128                  (*ecode == OP_ALT || *next == OP_ALT))
1129                rrc = MATCH_NOMATCH;
1130              }
1131    
1132            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1133            md->capture_last = save_capture_last;
1134            ecode += GET(ecode, 1);
1135            if (*ecode != OP_ALT) break;
1136            }
1137    
1138          if (!matched_once)
1139          {          {
1140          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          md->offset_vector[offset] = save_offset1;
1141          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          md->offset_vector[offset+1] = save_offset2;
1142            md->offset_vector[md->offset_end - number] = save_offset3;
1143          }          }
1144        else if (rrc != MATCH_NOMATCH)  
1145          if (allow_zero || matched_once)
1146          {          {
1147          RRETURN(rrc);         /* Need braces because of following else */          ecode += 1 + LINK_SIZE;
1148            break;
1149          }          }
1150        else ecode += GET(ecode, 1);  
1151        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        RRETURN(MATCH_NOMATCH);
         match_isgroup);  
       RRETURN(rrc);  
1152        }        }
     /* Control never reaches here */  
1153    
1154      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1155      encountered. */      as a non-capturing bracket. */
1156    
1157      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1158      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1159    
1160      /* End of the pattern. If we are in a recursion, we should restore the      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     offsets appropriately and continue from after the call. */  
1161    
1162      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1163      if (md->recursive != NULL && md->recursive->group_num == 0)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("Hit the end in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       md->start_match = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1164    
1165      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1166      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1167        without the capturing complication. It is written out separately for speed
1168        and cleanliness. */
1169    
1170      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1171      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1172      md->end_offset_top = offset_top;   /* and how many extracts were taken */      allow_zero = FALSE;
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1173    
1174      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1175      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1176      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      code_offset = (int)(ecode - md->start_code);
     start of each branch to move the current point backwards, so the code at  
     this level is identical to the lookahead case. */  
1177    
1178      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1179        {        {
1180        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1181          match_isgroup);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1182        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1183          if (rrc == MATCH_KETRPOS)
1184            {
1185            offset_top = md->end_offset_top;
1186            eptr = md->end_match_ptr;
1187            ecode = md->start_code + code_offset;
1188            matched_once = TRUE;
1189            continue;
1190            }
1191    
1192          /* See comment in the code for capturing groups above about handling
1193          THEN. */
1194    
1195          if (rrc == MATCH_THEN)
1196            {
1197            next = ecode + GET(ecode,1);
1198            if (md->start_match_ptr < next &&
1199                (*ecode == OP_ALT || *next == OP_ALT))
1200              rrc = MATCH_NOMATCH;
1201            }
1202    
1203        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1204        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1205          if (*ecode != OP_ALT) break;
1206        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1207    
1208      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1209          {
1210          ecode += 1 + LINK_SIZE;
1211          break;
1212          }
1213        RRETURN(MATCH_NOMATCH);
1214    
1215      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1216    
1217      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1218      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1219        past the end if there is only one branch, but that's OK because that is
1220        exactly what going to the ket would do. */
1221    
1222      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1223      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1224      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1225    
1226      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1227        inserted between OP_COND and an assertion condition. */
1228    
1229      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1230        {        {
1231        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (PUBL(callout) != NULL)
1232          match_isgroup);          {
1233        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1234        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1235        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1236            cb.offset_vector    = md->offset_vector;
1237    #ifdef COMPILE_PCRE8
1238            cb.subject          = (PCRE_SPTR)md->start_subject;
1239    #else
1240            cb.subject          = (PCRE_SPTR16)md->start_subject;
1241    #endif
1242            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1243            cb.start_match      = (int)(mstart - md->start_subject);
1244            cb.current_position = (int)(eptr - md->start_subject);
1245            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1246            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1247            cb.capture_top      = offset_top/2;
1248            cb.capture_last     = md->capture_last;
1249            cb.callout_data     = md->callout_data;
1250            cb.mark             = md->nomatch_mark;
1251            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1252            if (rrc < 0) RRETURN(rrc);
1253            }
1254          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1255        }        }
     while (*ecode == OP_ALT);  
   
     if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);  
1256    
1257      ecode += 1 + LINK_SIZE;      condcode = ecode[LINK_SIZE+1];
     continue;  
1258    
1259      /* Move the subject pointer back. This occurs only at the start of      /* Now see what the actual condition is */
     each branch of a lookbehind assertion. If we are too close to the start to  
     move back, this match function fails. When working with UTF-8 we move  
     back a number of characters, not bytes. */  
1260    
1261      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1262        {        {
1263        c = GET(ecode,1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
       for (i = 0; i < c; i++)  
1264          {          {
1265          eptr--;          condition = FALSE;
1266          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          ecode += GET(ecode, 1);
         BACKCHAR(eptr)  
1267          }          }
1268        }        else
1269      else          {
1270  #endif          int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1271            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1272    
1273      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */          /* If the test is for recursion into a specific subpattern, and it is
1274            false, but the test was set up by name, scan the table to see if the
1275            name refers to any other numbers, and test them. The condition is true
1276            if any one is set. */
1277    
1278        {          if (!condition && condcode == OP_NRREF)
1279        eptr -= GET(ecode,1);            {
1280        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);            pcre_uchar *slotA = md->name_table;
1281        }            for (i = 0; i < md->name_count; i++)
1282                {
1283                if (GET2(slotA, 0) == recno) break;
1284                slotA += md->name_entry_size;
1285                }
1286    
1287      /* Skip to next op code */            /* Found a name for the number - there can be only one; duplicate
1288              names for different numbers are allowed, but not vice versa. First
1289              scan down for duplicates. */
1290    
1291              if (i < md->name_count)
1292                {
1293                pcre_uchar *slotB = slotA;
1294                while (slotB > md->name_table)
1295                  {
1296                  slotB -= md->name_entry_size;
1297                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1298                    {
1299                    condition = GET2(slotB, 0) == md->recursive->group_num;
1300                    if (condition) break;
1301                    }
1302                  else break;
1303                  }
1304    
1305                /* Scan up for duplicates */
1306    
1307                if (!condition)
1308                  {
1309                  slotB = slotA;
1310                  for (i++; i < md->name_count; i++)
1311                    {
1312                    slotB += md->name_entry_size;
1313                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1314                      {
1315                      condition = GET2(slotB, 0) == md->recursive->group_num;
1316                      if (condition) break;
1317                      }
1318                    else break;
1319                    }
1320                  }
1321                }
1322              }
1323    
1324            /* Chose branch according to the condition */
1325    
1326            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1327            }
1328          }
1329    
1330        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1331          {
1332          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1333          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1334    
1335          /* If the numbered capture is unset, but the reference was by name,
1336          scan the table to see if the name refers to any other numbers, and test
1337          them. The condition is true if any one is set. This is tediously similar
1338          to the code above, but not close enough to try to amalgamate. */
1339    
1340          if (!condition && condcode == OP_NCREF)
1341            {
1342            int refno = offset >> 1;
1343            pcre_uchar *slotA = md->name_table;
1344    
1345            for (i = 0; i < md->name_count; i++)
1346              {
1347              if (GET2(slotA, 0) == refno) break;
1348              slotA += md->name_entry_size;
1349              }
1350    
1351            /* Found a name for the number - there can be only one; duplicate names
1352            for different numbers are allowed, but not vice versa. First scan down
1353            for duplicates. */
1354    
1355            if (i < md->name_count)
1356              {
1357              pcre_uchar *slotB = slotA;
1358              while (slotB > md->name_table)
1359                {
1360                slotB -= md->name_entry_size;
1361                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1362                  {
1363                  offset = GET2(slotB, 0) << 1;
1364                  condition = offset < offset_top &&
1365                    md->offset_vector[offset] >= 0;
1366                  if (condition) break;
1367                  }
1368                else break;
1369                }
1370    
1371              /* Scan up for duplicates */
1372    
1373              if (!condition)
1374                {
1375                slotB = slotA;
1376                for (i++; i < md->name_count; i++)
1377                  {
1378                  slotB += md->name_entry_size;
1379                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1380                    {
1381                    offset = GET2(slotB, 0) << 1;
1382                    condition = offset < offset_top &&
1383                      md->offset_vector[offset] >= 0;
1384                    if (condition) break;
1385                    }
1386                  else break;
1387                  }
1388                }
1389              }
1390            }
1391    
1392          /* Chose branch according to the condition */
1393    
1394          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1395          }
1396    
1397        else if (condcode == OP_DEF)     /* DEFINE - always false */
1398          {
1399          condition = FALSE;
1400          ecode += GET(ecode, 1);
1401          }
1402    
1403        /* The condition is an assertion. Call match() to evaluate it - setting
1404        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1405        an assertion. */
1406    
1407        else
1408          {
1409          md->match_function_type = MATCH_CONDASSERT;
1410          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1411          if (rrc == MATCH_MATCH)
1412            {
1413            if (md->end_offset_top > offset_top)
1414              offset_top = md->end_offset_top;  /* Captures may have happened */
1415            condition = TRUE;
1416            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1417            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1418            }
1419    
1420          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1421          assertion; it is therefore treated as NOMATCH. */
1422    
1423          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1424            {
1425            RRETURN(rrc);         /* Need braces because of following else */
1426            }
1427          else
1428            {
1429            condition = FALSE;
1430            ecode += codelink;
1431            }
1432          }
1433    
1434        /* We are now at the branch that is to be obeyed. As there is only one, can
1435        use tail recursion to avoid using another stack frame, except when there is
1436        unlimited repeat of a possibly empty group. In the latter case, a recursive
1437        call to match() is always required, unless the second alternative doesn't
1438        exist, in which case we can just plough on. Note that, for compatibility
1439        with Perl, the | in a conditional group is NOT treated as creating two
1440        alternatives. If a THEN is encountered in the branch, it propagates out to
1441        the enclosing alternative (unless nested in a deeper set of alternatives,
1442        of course). */
1443    
1444        if (condition || *ecode == OP_ALT)
1445          {
1446          if (op != OP_SCOND)
1447            {
1448            ecode += 1 + LINK_SIZE;
1449            goto TAIL_RECURSE;
1450            }
1451    
1452          md->match_function_type = MATCH_CBEGROUP;
1453          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1454          RRETURN(rrc);
1455          }
1456    
1457         /* Condition false & no alternative; continue after the group. */
1458    
1459        else
1460          {
1461          ecode += 1 + LINK_SIZE;
1462          }
1463        break;
1464    
1465    
1466        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1467        to close any currently open capturing brackets. */
1468    
1469        case OP_CLOSE:
1470        number = GET2(ecode, 1);
1471        offset = number << 1;
1472    
1473    #ifdef PCRE_DEBUG
1474          printf("end bracket %d at *ACCEPT", number);
1475          printf("\n");
1476    #endif
1477    
1478        md->capture_last = number;
1479        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1480          {
1481          md->offset_vector[offset] =
1482            md->offset_vector[md->offset_end - number];
1483          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1484          if (offset_top <= offset) offset_top = offset + 2;
1485          }
1486        ecode += 1 + IMM2_SIZE;
1487        break;
1488    
1489    
1490        /* End of the pattern, either real or forced. */
1491    
1492        case OP_END:
1493        case OP_ACCEPT:
1494        case OP_ASSERT_ACCEPT:
1495    
1496        /* If we have matched an empty string, fail if not in an assertion and not
1497        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1498        is set and we have matched at the start of the subject. In both cases,
1499        backtracking will then try other alternatives, if any. */
1500    
1501        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1502             md->recursive == NULL &&
1503             (md->notempty ||
1504               (md->notempty_atstart &&
1505                 mstart == md->start_subject + md->start_offset)))
1506          RRETURN(MATCH_NOMATCH);
1507    
1508        /* Otherwise, we have a match. */
1509    
1510        md->end_match_ptr = eptr;           /* Record where we ended */
1511        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1512        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1513    
1514        /* For some reason, the macros don't work properly if an expression is
1515        given as the argument to RRETURN when the heap is in use. */
1516    
1517        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1518        RRETURN(rrc);
1519    
1520        /* Assertion brackets. Check the alternative branches in turn - the
1521        matching won't pass the KET for an assertion. If any one branch matches,
1522        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1523        start of each branch to move the current point backwards, so the code at
1524        this level is identical to the lookahead case. When the assertion is part
1525        of a condition, we want to return immediately afterwards. The caller of
1526        this incarnation of the match() function will have set MATCH_CONDASSERT in
1527        md->match_function type, and one of these opcodes will be the first opcode
1528        that is processed. We use a local variable that is preserved over calls to
1529        match() to remember this case. */
1530    
1531        case OP_ASSERT:
1532        case OP_ASSERTBACK:
1533        save_mark = md->mark;
1534        if (md->match_function_type == MATCH_CONDASSERT)
1535          {
1536          condassert = TRUE;
1537          md->match_function_type = 0;
1538          }
1539        else condassert = FALSE;
1540    
1541        do
1542          {
1543          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1544          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1545            {
1546            mstart = md->start_match_ptr;   /* In case \K reset it */
1547            break;
1548            }
1549    
1550          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1551          as NOMATCH. */
1552    
1553          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1554          ecode += GET(ecode, 1);
1555          md->mark = save_mark;
1556          }
1557        while (*ecode == OP_ALT);
1558    
1559        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1560    
1561        /* If checking an assertion for a condition, return MATCH_MATCH. */
1562    
1563        if (condassert) RRETURN(MATCH_MATCH);
1564    
1565        /* Continue from after the assertion, updating the offsets high water
1566        mark, since extracts may have been taken during the assertion. */
1567    
1568        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1569        ecode += 1 + LINK_SIZE;
1570        offset_top = md->end_offset_top;
1571        continue;
1572    
1573        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1574        PRUNE, or COMMIT means we must assume failure without checking subsequent
1575        branches. */
1576    
1577        case OP_ASSERT_NOT:
1578        case OP_ASSERTBACK_NOT:
1579        save_mark = md->mark;
1580        if (md->match_function_type == MATCH_CONDASSERT)
1581          {
1582          condassert = TRUE;
1583          md->match_function_type = 0;
1584          }
1585        else condassert = FALSE;
1586    
1587        do
1588          {
1589          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1590          md->mark = save_mark;
1591          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1592          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1593            {
1594            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1595            break;
1596            }
1597    
1598          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1599          as NOMATCH. */
1600    
1601          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1602          ecode += GET(ecode,1);
1603          }
1604        while (*ecode == OP_ALT);
1605    
1606        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1607    
1608        ecode += 1 + LINK_SIZE;
1609        continue;
1610    
1611        /* Move the subject pointer back. This occurs only at the start of
1612        each branch of a lookbehind assertion. If we are too close to the start to
1613        move back, this match function fails. When working with UTF-8 we move
1614        back a number of characters, not bytes. */
1615    
1616        case OP_REVERSE:
1617    #ifdef SUPPORT_UTF
1618        if (utf)
1619          {
1620          i = GET(ecode, 1);
1621          while (i-- > 0)
1622            {
1623            eptr--;
1624            if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1625            BACKCHAR(eptr);
1626            }
1627          }
1628        else
1629    #endif
1630    
1631        /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1632    
1633          {
1634          eptr -= GET(ecode, 1);
1635          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1636          }
1637    
1638        /* Save the earliest consulted character, then skip to next op code */
1639    
1640        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1641      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1642      break;      break;
1643    
# Line 794  for (;;) Line 1646  for (;;)
1646      function is able to force a failure. */      function is able to force a failure. */
1647    
1648      case OP_CALLOUT:      case OP_CALLOUT:
1649      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1650        {        {
1651        pcre_callout_block cb;        PUBL(callout_block) cb;
1652        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1653        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1654        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1655        cb.subject          = (const char *)md->start_subject;  #ifdef COMPILE_PCRE8
1656        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1657        cb.start_match      = md->start_match - md->start_subject;  #else
1658        cb.current_position = eptr - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1659    #endif
1660          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1661          cb.start_match      = (int)(mstart - md->start_subject);
1662          cb.current_position = (int)(eptr - md->start_subject);
1663        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1664        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1665        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1666        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1667        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1668        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1669          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1670        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1671        }        }
1672      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 819  for (;;) Line 1676  for (;;)
1676      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
1677      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1678    
1679      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1680      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
1681      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
1682      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
1683      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
1684      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
1685      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.  
1686    
1687      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
1688      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
1689      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1690        a lot, so he is not to blame for the current way it works. */
1691    
1692      case OP_RECURSE:      case OP_RECURSE:
1693        {        {
1694        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1695        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1696    
1697        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1698        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1699            GET2(callpat, 1 + LINK_SIZE);
1700    
1701        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1702          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1703          caught at compile time.) */
1704    
1705          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1706            if (recno == ri->group_num && eptr == ri->subject_position)
1707              RRETURN(PCRE_ERROR_RECURSELOOP);
1708    
1709        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1710    
1711          new_recursive.group_num = recno;
1712          new_recursive.subject_position = eptr;
1713        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1714        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1715    
1716        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1717    
1718        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1719    
1720        /* Now save the offset data. */        /* Now save the offset data */
1721    
1722        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1723        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 863  for (;;) Line 1725  for (;;)
1725        else        else
1726          {          {
1727          new_recursive.offset_save =          new_recursive.offset_save =
1728            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1729          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1730          }          }
   
1731        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1732              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1733    
1734        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1735        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1736          might be changed, so reset it before looping. */
1737    
1738        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1739          cbegroup = (*callpat >= OP_SBRA);
1740        do        do
1741          {          {
1742          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1743              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1744          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1745            memcpy(md->offset_vector, new_recursive.offset_save,
1746                new_recursive.saved_max * sizeof(int));
1747            md->recursive = new_recursive.prevrec;
1748            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1749              {
1750              DPRINTF(("Recursion matched\n"));
1751              if (new_recursive.offset_save != stacksave)
1752                (PUBL(free))(new_recursive.offset_save);
1753    
1754              /* Set where we got to in the subject, and reset the start in case
1755              it was changed by \K. This *is* propagated back out of a recursion,
1756              for Perl compatibility. */
1757    
1758              eptr = md->end_match_ptr;
1759              mstart = md->start_match_ptr;
1760              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1761              }
1762    
1763            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1764            as NOMATCH. */
1765    
1766            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1767            {            {
1768            md->recursive = new_recursive.prevrec;            DPRINTF(("Recursion gave error %d\n", rrc));
1769            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1770              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1771            RRETURN(MATCH_MATCH);            RRETURN(rrc);
1772            }            }
         else if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1773    
1774          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1775          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1776          }          }
1777        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 899  for (;;) Line 1779  for (;;)
1779        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1780        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1781        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1782          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1783        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1784        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
       {  
       prev = ecode;  
       saved_eptr = eptr;  
   
       do  
         {  
         RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
           eptrb, match_isgroup);  
         if (rrc == MATCH_MATCH) break;  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         ecode += GET(ecode,1);  
         }  
       while (*ecode == OP_ALT);  
   
       /* If hit the end of the group (which could be repeated), fail */  
   
       if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
       /* Continue as from after the assertion, updating the offsets high water  
       mark, since extracts may have been taken. */  
   
       do ecode += GET(ecode,1); while (*ecode == OP_ALT);  
   
       offset_top = md->end_offset_top;  
       eptr = md->end_match_ptr;  
1785    
1786        /* For a non-repeating ket, just continue at this level. This also      RECURSION_MATCHED:
1787        happens for a repeating ket if no characters were matched in the group.      break;
       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. We need to reset any options  
       that changed within the bracket before re-running it, so check the next  
       opcode. */  
   
       if (ecode[1+LINK_SIZE] == OP_OPT)  
         {  
         ims = (ims & ~PCRE_IMS) | ecode[4];  
         DPRINTF(("ims set to %02lx at group repeat\n", ims));  
         }  
   
       if (*ecode == OP_KETRMIN)  
         {  
         RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         }  
       else  /* OP_KETRMAX */  
         {  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         }  
       }  
     RRETURN(MATCH_NOMATCH);  
1788    
1789      /* An alternation is the end of a branch; scan along to find the end of the      /* An alternation is the end of a branch; scan along to find the end of the
1790      bracketed group and go to there. */      bracketed group and go to there. */
# Line 985  for (;;) Line 1793  for (;;)
1793      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1794      break;      break;
1795    
1796      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1797      that it may occur zero times. It may repeat infinitely, or not at all -      indicating that it may occur zero times. It may repeat infinitely, or not
1798      i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1799      repeat limits are compiled as a number of copies, with the optional ones      with fixed upper repeat limits are compiled as a number of copies, with the
1800      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1801    
1802      case OP_BRAZERO:      case OP_BRAZERO:
1803        {      next = ecode + 1;
1804        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1805        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1806        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1807        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1808      break;      break;
1809    
1810      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1811        {      next = ecode + 1;
1812        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1813        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1814        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1815          match_isgroup);      ecode++;
1816        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1817        ecode++;  
1818        }      case OP_SKIPZERO:
1819        next = ecode+1;
1820        do next += GET(next,1); while (*next == OP_ALT);
1821        ecode = next + 1 + LINK_SIZE;
1822      break;      break;
1823    
1824      /* End of a group, repeated or non-repeating. If we are at the end of      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1825      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1826      current high water mark for use by positive assertions. Do this also  
1827      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1828        op = *(++ecode);
1829        allow_zero = TRUE;
1830        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1831          goto POSSESSIVE_NON_CAPTURE;
1832    
1833        /* End of a group, repeated or non-repeating. */
1834    
1835      case OP_KET:      case OP_KET:
1836      case OP_KETRMIN:      case OP_KETRMIN:
1837      case OP_KETRMAX:      case OP_KETRMAX:
1838        {      case OP_KETRPOS:
1839        prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
       saved_eptr = eptrb->epb_saved_eptr;  
1840    
1841        /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1842        infinite repeats of empty string matches, retrieve the subject start from
1843        the chain. Otherwise, set it NULL. */
1844    
1845        eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1846          {
1847        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1848            *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1849            *prev == OP_ONCE)        }
1850          {      else saved_eptr = NULL;
         md->end_match_ptr = eptr;      /* For ONCE */  
         md->end_offset_top = offset_top;  
         RRETURN(MATCH_MATCH);  
         }  
1851    
1852        /* In all other cases except a conditional group we have to check the      /* If we are at the end of an assertion group or a non-capturing atomic
1853        group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1854        extraction by setting the offsets and bumping the high water mark. */      water mark for use by positive assertions. We also need to record the match
1855        start in case it was changed by \K. */
1856    
1857        if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1858          {           *prev == OP_ONCE_NC)
1859          number = *prev - OP_BRA;        {
1860          md->end_match_ptr = eptr;      /* For ONCE_NC */
1861          md->end_offset_top = offset_top;
1862          md->start_match_ptr = mstart;
1863          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1864          }
1865    
1866          /* For extended extraction brackets (large number), we have to fish out      /* For capturing groups we have to check the group number back at the start
1867          the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1868        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1869        into group 0, so it won't be picked up here. Instead, we catch it when the
1870        OP_END is reached. Other recursion is handled here. We just have to record
1871        the current subject position and start match pointer and give a MATCH
1872        return. */
1873    
1874          if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1875          offset = number << 1;          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1876          {
1877          number = GET2(prev, 1+LINK_SIZE);
1878          offset = number << 1;
1879    
1880  #ifdef DEBUG  #ifdef PCRE_DEBUG
1881          printf("end bracket %d", number);        printf("end bracket %d", number);
1882          printf("\n");        printf("\n");
1883  #endif  #endif
1884    
1885          /* Test for a numbered group. This includes groups called as a result        /* Handle a recursively called group. */
         of recursion. Note that whole-pattern recursion is coded as a recurse  
         into group 0, so it won't be picked up here. Instead, we catch it when  
         the OP_END is reached. */  
1886    
1887          if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1888            {          {
1889            md->capture_last = number;          md->end_match_ptr = eptr;
1890            if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1891              {          RRETURN(MATCH_MATCH);
1892              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;  
             }  
1893    
1894            /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
           appropriately and continue from after the call. */  
1895    
1896            if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = number;
1897              {        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1898              recursion_info *rec = md->recursive;          {
1899              DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          /* If offset is greater than offset_top, it means that we are
1900              md->recursive = rec->prevrec;          "skipping" a capturing group, and that group's offsets must be marked
1901              md->start_match = rec->save_start;          unset. In earlier versions of PCRE, all the offsets were unset at the
1902              memcpy(md->offset_vector, rec->offset_save,          start of matching, but this doesn't work because atomic groups and
1903                rec->saved_max * sizeof(int));          assertions can cause a value to be set that should later be unset.
1904              ecode = rec->after_call;          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1905              ims = original_ims;          part of the atomic group, but this is not on the final matching path,
1906              break;          so must be unset when 2 is set. (If there is no group 2, there is no
1907              }          problem, because offset_top will then be 2, indicating no capture.) */
1908    
1909            if (offset > offset_top)
1910              {
1911              register int *iptr = md->offset_vector + offset_top;
1912              register int *iend = md->offset_vector + offset;
1913              while (iptr < iend) *iptr++ = -1;
1914            }            }
         }  
1915    
1916        /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
       the group. */  
1917    
1918        ims = original_ims;          md->offset_vector[offset] =
1919        DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1920            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1921            if (offset_top <= offset) offset_top = offset + 2;
1922            }
1923          }
1924    
1925        /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1926        happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1927        This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1928        5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group that includes captures,
1929        course of events. */      establish a backup point by processing the rest of the pattern at a lower
1930        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1931        original OP_ONCE level, thereby bypassing intermediate backup points, but
1932        resetting any captures that happened along the way. */
1933    
1934        if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1935          {
1936          if (*prev == OP_ONCE)
1937          {          {
1938          ecode += 1 + LINK_SIZE;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1939          break;          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1940            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1941            RRETURN(MATCH_ONCE);
1942          }          }
1943          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1944          break;
1945          }
1946    
1947        /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1948        preceding bracket, in the appropriate order. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1949        at a time from the outer level, thus saving stack. */
1950    
1951        if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRPOS)
1952          {
1953          md->end_match_ptr = eptr;
1954          md->end_offset_top = offset_top;
1955          RRETURN(MATCH_KETRPOS);
1956          }
1957    
1958        /* The normal repeating kets try the rest of the pattern or restart from
1959        the preceding bracket, in the appropriate order. In the second case, we can
1960        use tail recursion to avoid using another stack frame, unless we have an
1961        an atomic group or an unlimited repeat of a group that can match an empty
1962        string. */
1963    
1964        if (*ecode == OP_KETRMIN)
1965          {
1966          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1967          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1968          if (*prev == OP_ONCE)
1969          {          {
1970          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);  
1971          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1972            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1973            RRETURN(MATCH_ONCE);
1974          }          }
1975        else  /* OP_KETRMAX */        if (*prev >= OP_SBRA)    /* Could match an empty string */
1976          {          {
1977          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          md->match_function_type = MATCH_CBEGROUP;
1978          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1979          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          RRETURN(rrc);
1980            }
1981          ecode = prev;
1982          goto TAIL_RECURSE;
1983          }
1984        else  /* OP_KETRMAX */
1985          {
1986          if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1987          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1988          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1989          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1990          if (*prev == OP_ONCE)
1991            {
1992            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1993          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1994            md->once_target = prev;
1995            RRETURN(MATCH_ONCE);
1996          }          }
1997          ecode += 1 + LINK_SIZE;
1998          goto TAIL_RECURSE;
1999        }        }
2000        /* Control never gets here */
2001    
2002      RRETURN(MATCH_NOMATCH);      /* Not multiline mode: start of subject assertion, unless notbol. */
   
     /* Start of subject unless notbol, or after internal newline if multiline */  
2003    
2004      case OP_CIRC:      case OP_CIRC:
2005      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject && eptr[-1] != NEWLINE)  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2006    
2007      /* Start of subject assertion */      /* Start of subject assertion */
2008    
# Line 1149  for (;;) Line 2011  for (;;)
2011      ecode++;      ecode++;
2012      break;      break;
2013    
2014        /* Multiline mode: start of subject unless notbol, or after any newline. */
2015    
2016        case OP_CIRCM:
2017        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2018        if (eptr != md->start_subject &&
2019            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2020          RRETURN(MATCH_NOMATCH);
2021        ecode++;
2022        break;
2023    
2024      /* Start of match assertion */      /* Start of match assertion */
2025    
2026      case OP_SOM:      case OP_SOM:
# Line 1156  for (;;) Line 2028  for (;;)
2028      ecode++;      ecode++;
2029      break;      break;
2030    
2031      /* 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. */  
2032    
2033      case OP_DOLL:      case OP_SET_SOM:
2034      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2035        {      ecode++;
2036        if (eptr < md->end_subject)      break;
2037          { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }  
2038        else      /* Multiline mode: assert before any newline, or before end of subject
2039          { if (md->noteol) RRETURN(MATCH_NOMATCH); }      unless noteol is set. */
2040        ecode++;  
2041        break;      case OP_DOLLM:
2042        }      if (eptr < md->end_subject)
2043          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
2044      else      else
2045        {        {
2046        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2047        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr < md->end_subject - 1 ||  
            (eptr == md->end_subject - 1 && *eptr != NEWLINE))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2048        }        }
2049      /* ... else fall through */      ecode++;
2050        break;
2051    
2052        /* Not multiline mode: assert before a terminating newline or before end of
2053        subject unless noteol is set. */
2054    
2055        case OP_DOLL:
2056        if (md->noteol) RRETURN(MATCH_NOMATCH);
2057        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2058    
2059        /* ... else fall through for endonly */
2060    
2061      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2062    
2063      case OP_EOD:      case OP_EOD:
2064      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2065        SCHECK_PARTIAL();
2066      ecode++;      ecode++;
2067      break;      break;
2068    
2069      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2070    
2071      case OP_EODN:      case OP_EODN:
2072      if (eptr < md->end_subject - 1 ||      ASSERT_NL_OR_EOS:
2073         (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject &&
2074            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2075          RRETURN(MATCH_NOMATCH);
2076    
2077        /* Either at end of string or \n before end. */
2078    
2079        SCHECK_PARTIAL();
2080      ecode++;      ecode++;
2081      break;      break;
2082    
# Line 1204  for (;;) Line 2086  for (;;)
2086      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
2087        {        {
2088    
2089        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2090        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
2091        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2092          partial matching. */
2093    
2094    #ifdef SUPPORT_UTF
2095          if (utf)
2096            {
2097            /* Get status of previous character */
2098    
2099            if (eptr == md->start_subject) prev_is_word = FALSE; else
2100              {
2101              PCRE_PUCHAR lastptr = eptr - 1;
2102              BACKCHAR(lastptr);
2103              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2104              GETCHAR(c, lastptr);
2105    #ifdef SUPPORT_UCP
2106              if (md->use_ucp)
2107                {
2108                if (c == '_') prev_is_word = TRUE; else
2109                  {
2110                  int cat = UCD_CATEGORY(c);
2111                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2112                  }
2113                }
2114              else
2115    #endif
2116              prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2117              }
2118    
2119            /* Get status of next character */
2120    
2121            if (eptr >= md->end_subject)
2122              {
2123              SCHECK_PARTIAL();
2124              cur_is_word = FALSE;
2125              }
2126            else
2127              {
2128              GETCHAR(c, eptr);
2129    #ifdef SUPPORT_UCP
2130              if (md->use_ucp)
2131                {
2132                if (c == '_') cur_is_word = TRUE; else
2133                  {
2134                  int cat = UCD_CATEGORY(c);
2135                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2136                  }
2137                }
2138              else
2139    #endif
2140              cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2141              }
2142            }
2143          else
2144    #endif
2145    
2146          /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2147          consistency with the behaviour of \w we do use it in this case. */
2148    
2149            {
2150            /* Get status of previous character */
2151    
2152            if (eptr == md->start_subject) prev_is_word = FALSE; else
2153              {
2154              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2155    #ifdef SUPPORT_UCP
2156              if (md->use_ucp)
2157                {
2158                c = eptr[-1];
2159                if (c == '_') prev_is_word = TRUE; else
2160                  {
2161                  int cat = UCD_CATEGORY(c);
2162                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2163                  }
2164                }
2165              else
2166    #endif
2167              prev_is_word = MAX_255(eptr[-1])
2168                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2169              }
2170    
2171            /* Get status of next character */
2172    
2173  #ifdef SUPPORT_UTF8          if (eptr >= md->end_subject)
       if (utf8)  
         {  
         if (eptr == md->start_subject) prev_is_word = FALSE; else  
2174            {            {
2175            const uschar *lastptr = eptr - 1;            SCHECK_PARTIAL();
2176            while((*lastptr & 0xc0) == 0x80) lastptr--;            cur_is_word = FALSE;
           GETCHAR(c, lastptr);  
           prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;  
2177            }            }
2178          if (eptr >= md->end_subject) cur_is_word = FALSE; else          else
2179    #ifdef SUPPORT_UCP
2180            if (md->use_ucp)
2181            {            {
2182            GETCHAR(c, eptr);            c = *eptr;
2183            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            if (c == '_') cur_is_word = TRUE; else
2184                {
2185                int cat = UCD_CATEGORY(c);
2186                cur_is_word = (cat == ucp_L || cat == ucp_N);
2187                }
2188            }            }
2189          }          else
       else  
2190  #endif  #endif
2191            cur_is_word = MAX_255(*eptr)
2192        /* More streamlined when not in UTF-8 mode */            && ((md->ctypes[*eptr] & ctype_word) != 0);
   
         {  
         prev_is_word = (eptr != md->start_subject) &&  
           ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
         cur_is_word = (eptr < md->end_subject) &&  
           ((md->ctypes[*eptr] & ctype_word) != 0);  
2193          }          }
2194    
2195        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1247  for (;;) Line 2203  for (;;)
2203      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2204    
2205      case OP_ANY:      case OP_ANY:
2206      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2207        /* Fall through */
2208    
2209        case OP_ALLANY:
2210        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2211          {                            /* not be updated before SCHECK_PARTIAL. */
2212          SCHECK_PARTIAL();
2213        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2214      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);        }
2215  #ifdef SUPPORT_UTF8      eptr++;
2216      if (utf8)  #ifdef SUPPORT_UTF
2217        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2218  #endif  #endif
2219      ecode++;      ecode++;
2220      break;      break;
# Line 1261  for (;;) Line 2223  for (;;)
2223      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2224    
2225      case OP_ANYBYTE:      case OP_ANYBYTE:
2226      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2227          {                            /* not be updated before SCHECK_PARTIAL. */
2228          SCHECK_PARTIAL();
2229          RRETURN(MATCH_NOMATCH);
2230          }
2231        eptr++;
2232      ecode++;      ecode++;
2233      break;      break;
2234    
2235      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2236      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2237          {
2238          SCHECK_PARTIAL();
2239          RRETURN(MATCH_NOMATCH);
2240          }
2241      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2242      if (      if (
2243  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2244         c < 256 &&         c < 256 &&
2245  #endif  #endif
2246         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1279  for (;;) Line 2250  for (;;)
2250      break;      break;
2251    
2252      case OP_DIGIT:      case OP_DIGIT:
2253      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2254          {
2255          SCHECK_PARTIAL();
2256          RRETURN(MATCH_NOMATCH);
2257          }
2258      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2259      if (      if (
2260  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2261         c >= 256 ||         c > 255 ||
2262  #endif  #endif
2263         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2264         )         )
# Line 1292  for (;;) Line 2267  for (;;)
2267      break;      break;
2268    
2269      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2270      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2271          {
2272          SCHECK_PARTIAL();
2273          RRETURN(MATCH_NOMATCH);
2274          }
2275      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2276      if (      if (
2277  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2278         c < 256 &&         c < 256 &&
2279  #endif  #endif
2280         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1305  for (;;) Line 2284  for (;;)
2284      break;      break;
2285    
2286      case OP_WHITESPACE:      case OP_WHITESPACE:
2287      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2288          {
2289          SCHECK_PARTIAL();
2290          RRETURN(MATCH_NOMATCH);
2291          }
2292      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2293      if (      if (
2294  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2295         c >= 256 ||         c > 255 ||
2296  #endif  #endif
2297         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2298         )         )
# Line 1318  for (;;) Line 2301  for (;;)
2301      break;      break;
2302    
2303      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2304      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2305          {
2306          SCHECK_PARTIAL();
2307          RRETURN(MATCH_NOMATCH);
2308          }
2309      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2310      if (      if (
2311  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2312         c < 256 &&         c < 256 &&
2313  #endif  #endif
2314         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1331  for (;;) Line 2318  for (;;)
2318      break;      break;
2319    
2320      case OP_WORDCHAR:      case OP_WORDCHAR:
2321      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2322          {
2323          SCHECK_PARTIAL();
2324          RRETURN(MATCH_NOMATCH);
2325          }
2326      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2327      if (      if (
2328  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2329         c >= 256 ||         c > 255 ||
2330  #endif  #endif
2331         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2332         )         )
# Line 1343  for (;;) Line 2334  for (;;)
2334      ecode++;      ecode++;
2335      break;      break;
2336    
2337        case OP_ANYNL:
2338        if (eptr >= md->end_subject)
2339          {
2340          SCHECK_PARTIAL();
2341          RRETURN(MATCH_NOMATCH);
2342          }
2343        GETCHARINCTEST(c, eptr);
2344        switch(c)
2345          {
2346          default: RRETURN(MATCH_NOMATCH);
2347    
2348          case 0x000d:
2349          if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2350          break;
2351    
2352          case 0x000a:
2353          break;
2354    
2355          case 0x000b:
2356          case 0x000c:
2357          case 0x0085:
2358          case 0x2028:
2359          case 0x2029:
2360          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2361          break;
2362          }
2363        ecode++;
2364        break;
2365    
2366        case OP_NOT_HSPACE:
2367        if (eptr >= md->end_subject)
2368          {
2369          SCHECK_PARTIAL();
2370          RRETURN(MATCH_NOMATCH);
2371          }
2372        GETCHARINCTEST(c, eptr);
2373        switch(c)
2374          {
2375          default: break;
2376          case 0x09:      /* HT */
2377          case 0x20:      /* SPACE */
2378          case 0xa0:      /* NBSP */
2379          case 0x1680:    /* OGHAM SPACE MARK */
2380          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2381          case 0x2000:    /* EN QUAD */
2382          case 0x2001:    /* EM QUAD */
2383          case 0x2002:    /* EN SPACE */
2384          case 0x2003:    /* EM SPACE */
2385          case 0x2004:    /* THREE-PER-EM SPACE */
2386          case 0x2005:    /* FOUR-PER-EM SPACE */
2387          case 0x2006:    /* SIX-PER-EM SPACE */
2388          case 0x2007:    /* FIGURE SPACE */
2389          case 0x2008:    /* PUNCTUATION SPACE */
2390          case 0x2009:    /* THIN SPACE */
2391          case 0x200A:    /* HAIR SPACE */
2392          case 0x202f:    /* NARROW NO-BREAK SPACE */
2393          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2394          case 0x3000:    /* IDEOGRAPHIC SPACE */
2395          RRETURN(MATCH_NOMATCH);
2396          }
2397        ecode++;
2398        break;
2399    
2400        case OP_HSPACE:
2401        if (eptr >= md->end_subject)
2402          {
2403          SCHECK_PARTIAL();
2404          RRETURN(MATCH_NOMATCH);
2405          }
2406        GETCHARINCTEST(c, eptr);
2407        switch(c)
2408          {
2409          default: RRETURN(MATCH_NOMATCH);
2410          case 0x09:      /* HT */
2411          case 0x20:      /* SPACE */
2412          case 0xa0:      /* NBSP */
2413          case 0x1680:    /* OGHAM SPACE MARK */
2414          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2415          case 0x2000:    /* EN QUAD */
2416          case 0x2001:    /* EM QUAD */
2417          case 0x2002:    /* EN SPACE */
2418          case 0x2003:    /* EM SPACE */
2419          case 0x2004:    /* THREE-PER-EM SPACE */
2420          case 0x2005:    /* FOUR-PER-EM SPACE */
2421          case 0x2006:    /* SIX-PER-EM SPACE */
2422          case 0x2007:    /* FIGURE SPACE */
2423          case 0x2008:    /* PUNCTUATION SPACE */
2424          case 0x2009:    /* THIN SPACE */
2425          case 0x200A:    /* HAIR SPACE */
2426          case 0x202f:    /* NARROW NO-BREAK SPACE */
2427          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2428          case 0x3000:    /* IDEOGRAPHIC SPACE */
2429          break;
2430          }
2431        ecode++;
2432        break;
2433    
2434        case OP_NOT_VSPACE:
2435        if (eptr >= md->end_subject)
2436          {
2437          SCHECK_PARTIAL();
2438          RRETURN(MATCH_NOMATCH);
2439          }
2440        GETCHARINCTEST(c, eptr);
2441        switch(c)
2442          {
2443          default: break;
2444          case 0x0a:      /* LF */
2445          case 0x0b:      /* VT */
2446          case 0x0c:      /* FF */
2447          case 0x0d:      /* CR */
2448          case 0x85:      /* NEL */
2449          case 0x2028:    /* LINE SEPARATOR */
2450          case 0x2029:    /* PARAGRAPH SEPARATOR */
2451          RRETURN(MATCH_NOMATCH);
2452          }
2453        ecode++;
2454        break;
2455    
2456        case OP_VSPACE:
2457        if (eptr >= md->end_subject)
2458          {
2459          SCHECK_PARTIAL();
2460          RRETURN(MATCH_NOMATCH);
2461          }
2462        GETCHARINCTEST(c, eptr);
2463        switch(c)
2464          {
2465          default: RRETURN(MATCH_NOMATCH);
2466          case 0x0a:      /* LF */
2467          case 0x0b:      /* VT */
2468          case 0x0c:      /* FF */
2469          case 0x0d:      /* CR */
2470          case 0x85:      /* NEL */
2471          case 0x2028:    /* LINE SEPARATOR */
2472          case 0x2029:    /* PARAGRAPH SEPARATOR */
2473          break;
2474          }
2475        ecode++;
2476        break;
2477    
2478  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2479      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2480      if the support is in the binary; otherwise a compile-time error occurs. */      if the support is in the binary; otherwise a compile-time error occurs. */
2481    
2482      case OP_PROP:      case OP_PROP:
2483      case OP_NOTPROP:      case OP_NOTPROP:
2484      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2485          {
2486          SCHECK_PARTIAL();
2487          RRETURN(MATCH_NOMATCH);
2488          }
2489      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2490        {        {
2491        int chartype, rqdtype;        const ucd_record *prop = GET_UCD(c);
       int othercase;  
       int category = _pcre_ucp_findchar(c, &chartype, &othercase);  
   
       rqdtype = *(++ecode);  
       ecode++;  
2492    
2493        if (rqdtype >= 128)        switch(ecode[1])
2494          {          {
2495          if ((rqdtype - 128 != category) == (op == OP_PROP))          case PT_ANY:
2496            if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2497            break;
2498    
2499            case PT_LAMP:
2500            if ((prop->chartype == ucp_Lu ||
2501                 prop->chartype == ucp_Ll ||
2502                 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2503            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2504          }          break;
2505        else  
2506          {          case PT_GC:
2507          if ((rqdtype != chartype) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2508              RRETURN(MATCH_NOMATCH);
2509            break;
2510    
2511            case PT_PC:
2512            if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2513              RRETURN(MATCH_NOMATCH);
2514            break;
2515    
2516            case PT_SC:
2517            if ((ecode[2] != prop->script) == (op == OP_PROP))
2518              RRETURN(MATCH_NOMATCH);
2519            break;
2520    
2521            /* These are specials */
2522    
2523            case PT_ALNUM:
2524            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2525                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2526              RRETURN(MATCH_NOMATCH);
2527            break;
2528    
2529            case PT_SPACE:    /* Perl space */
2530            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2531                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2532                   == (op == OP_NOTPROP))
2533              RRETURN(MATCH_NOMATCH);
2534            break;
2535    
2536            case PT_PXSPACE:  /* POSIX space */
2537            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2538                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2539                 c == CHAR_FF || c == CHAR_CR)
2540                   == (op == OP_NOTPROP))
2541            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2542            break;
2543    
2544            case PT_WORD:
2545            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2546                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2547                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2548              RRETURN(MATCH_NOMATCH);
2549            break;
2550    
2551            /* This should never occur */
2552    
2553            default:
2554            RRETURN(PCRE_ERROR_INTERNAL);
2555          }          }
2556    
2557          ecode += 3;
2558        }        }
2559      break;      break;
2560    
# Line 1376  for (;;) Line 2562  for (;;)
2562      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2563    
2564      case OP_EXTUNI:      case OP_EXTUNI:
2565      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2566          {
2567          SCHECK_PARTIAL();
2568          RRETURN(MATCH_NOMATCH);
2569          }
2570      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2571        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2572        while (eptr < md->end_subject)
2573        {        {
2574        int chartype;        int len = 1;
2575        int othercase;        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2576        int category = _pcre_ucp_findchar(c, &chartype, &othercase);        if (UCD_CATEGORY(c) != ucp_M) break;
2577        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;  
         }  
2578        }        }
2579      ecode++;      ecode++;
2580      break;      break;
# Line 1409  for (;;) Line 2590  for (;;)
2590      loops). */      loops). */
2591    
2592      case OP_REF:      case OP_REF:
2593        {      case OP_REFI:
2594        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2595        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2596        ecode += 1 + IMM2_SIZE;
2597    
2598        /* 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];  
2599    
2600        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2601        this ensures that every attempt at a match fails. We can't just fail
2602        here, because of the possibility of quantifiers with zero minima.
2603    
2604        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2605          {      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;  
2606    
2607          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2608          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;  
2609    
2610          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2611          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2612          eptr += length;      else
2613          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2614    
2615        /* Set up for repetition, or handle the non-repeated case */
2616    
2617        switch (*ecode)
2618          {
2619          case OP_CRSTAR:
2620          case OP_CRMINSTAR:
2621          case OP_CRPLUS:
2622          case OP_CRMINPLUS:
2623          case OP_CRQUERY:
2624          case OP_CRMINQUERY:
2625          c = *ecode++ - OP_CRSTAR;
2626          minimize = (c & 1) != 0;
2627          min = rep_min[c];                 /* Pick up values from tables; */
2628          max = rep_max[c];                 /* zero for max => infinity */
2629          if (max == 0) max = INT_MAX;
2630          break;
2631    
2632          case OP_CRRANGE:
2633          case OP_CRMINRANGE:
2634          minimize = (*ecode == OP_CRMINRANGE);
2635          min = GET2(ecode, 1);
2636          max = GET2(ecode, 1 + IMM2_SIZE);
2637          if (max == 0) max = INT_MAX;
2638          ecode += 1 + 2 * IMM2_SIZE;
2639          break;
2640    
2641          default:               /* No repeat follows */
2642          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2643            {
2644            CHECK_PARTIAL();
2645            RRETURN(MATCH_NOMATCH);
2646          }          }
2647          eptr += length;
2648          continue;              /* With the main loop */
2649          }
2650    
2651        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2652        main loop. */      zero, just continue with the main loop. If the length is negative, it
2653        means the reference is unset in non-Java-compatible mode. If the minimum is
2654        zero, we can continue at the same level without recursion. For any other
2655        minimum, carrying on will result in NOMATCH. */
2656    
2657        if (length == 0) continue;      if (length == 0) continue;
2658        if (length < 0 && min == 0) continue;
2659    
2660        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2661        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2662        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2663    
2664        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2665          {
2666          int slength;
2667          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2668          {          {
2669          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2670          eptr += length;          RRETURN(MATCH_NOMATCH);
2671          }          }
2672          eptr += slength;
2673          }
2674    
2675        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2676        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2677    
2678        if (min == max) continue;      if (min == max) continue;
2679    
2680        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2681    
2682        if (minimize)      if (minimize)
2683          {
2684          for (fi = min;; fi++)
2685          {          {
2686          for (fi = min;; fi++)          int slength;
2687            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2688            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2689            if (fi >= max) RRETURN(MATCH_NOMATCH);
2690            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2691            {            {
2692            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            CHECK_PARTIAL();
2693            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            RRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2694            }            }
2695          /* Control never gets here */          eptr += slength;
2696          }          }
2697          /* Control never gets here */
2698          }
2699    
2700        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2701    
2702        else      else
2703          {
2704          pp = eptr;
2705          for (i = min; i < max; i++)
2706          {          {
2707          pp = eptr;          int slength;
2708          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2709            {            {
2710            if (!match_ref(offset, eptr, length, md, ims)) break;            CHECK_PARTIAL();
2711            eptr += length;            break;
           }  
         while (eptr >= pp)  
           {  
           RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
           if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
           eptr -= length;  
2712            }            }
2713          RRETURN(MATCH_NOMATCH);          eptr += slength;
2714            }
2715          while (eptr >= pp)
2716            {
2717            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2718            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2719            eptr -= length;
2720          }          }
2721          RRETURN(MATCH_NOMATCH);
2722        }        }
2723      /* Control never gets here */      /* Control never gets here */
2724    
   
   
2725      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2726      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,
2727      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 1526  for (;;) Line 2736  for (;;)
2736      case OP_NCLASS:      case OP_NCLASS:
2737      case OP_CLASS:      case OP_CLASS:
2738        {        {
2739          /* The data variable is saved across frames, so the byte map needs to
2740          be stored there. */
2741    #define BYTE_MAP ((pcre_uint8 *)data)
2742        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2743        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2744    
2745        switch (*ecode)        switch (*ecode)
2746          {          {
# Line 1548  for (;;) Line 2761  for (;;)
2761          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2762          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2763          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2764          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2765          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2766          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2767          break;          break;
2768    
2769          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1560  for (;;) Line 2773  for (;;)
2773    
2774        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2775    
2776  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2777        /* UTF-8 mode */        if (utf)
       if (utf8)  
2778          {          {
2779          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2780            {            {
2781            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2782                {
2783                SCHECK_PARTIAL();
2784                RRETURN(MATCH_NOMATCH);
2785                }
2786            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2787            if (c > 255)            if (c > 255)
2788              {              {
2789              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2790              }              }
2791            else            else
2792              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2793            }            }
2794          }          }
2795        else        else
2796  #endif  #endif
2797        /* Not UTF-8 mode */        /* Not UTF mode */
2798          {          {
2799          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2800            {            {
2801            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2802                {
2803                SCHECK_PARTIAL();
2804                RRETURN(MATCH_NOMATCH);
2805                }
2806            c = *eptr++;            c = *eptr++;
2807            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2808              if (c > 255)
2809                {
2810                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2811                }
2812              else
2813    #endif
2814                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2815            }            }
2816          }          }
2817    
# Line 1600  for (;;) Line 2825  for (;;)
2825    
2826        if (minimize)        if (minimize)
2827          {          {
2828  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2829          /* UTF-8 mode */          if (utf)
         if (utf8)  
2830            {            {
2831            for (fi = min;; fi++)            for (fi = min;; fi++)
2832              {              {
2833              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2834              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2835              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2836                if (eptr >= md->end_subject)
2837                  {
2838                  SCHECK_PARTIAL();
2839                  RRETURN(MATCH_NOMATCH);
2840                  }
2841              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2842              if (c > 255)              if (c > 255)
2843                {                {
2844                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2845                }                }
2846              else              else
2847                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2848              }              }
2849            }            }
2850          else          else
2851  #endif  #endif
2852          /* Not UTF-8 mode */          /* Not UTF mode */
2853            {            {
2854            for (fi = min;; fi++)            for (fi = min;; fi++)
2855              {              {
2856              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2857              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2858              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2859                if (eptr >= md->end_subject)
2860                  {
2861                  SCHECK_PARTIAL();
2862                  RRETURN(MATCH_NOMATCH);
2863                  }
2864              c = *eptr++;              c = *eptr++;
2865              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2866                if (c > 255)
2867                  {
2868                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2869                  }
2870                else
2871    #endif
2872                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2873              }              }
2874            }            }
2875          /* Control never gets here */          /* Control never gets here */
# Line 1642  for (;;) Line 2881  for (;;)
2881          {          {
2882          pp = eptr;          pp = eptr;
2883    
2884  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2885          /* UTF-8 mode */          if (utf)
         if (utf8)  
2886            {            {
2887            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2888              {              {
2889              int len = 1;              int len = 1;
2890              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2891                  {
2892                  SCHECK_PARTIAL();
2893                  break;
2894                  }
2895              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2896              if (c > 255)              if (c > 255)
2897                {                {
2898                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2899                }                }
2900              else              else
2901                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2902              eptr += len;              eptr += len;
2903              }              }
2904            for (;;)            for (;;)
2905              {              {
2906              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2907              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2908              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2909              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1671  for (;;) Line 2911  for (;;)
2911            }            }
2912          else          else
2913  #endif  #endif
2914            /* Not UTF-8 mode */            /* Not UTF mode */
2915            {            {
2916            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2917              {              {
2918              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2919                  {
2920                  SCHECK_PARTIAL();
2921                  break;
2922                  }
2923              c = *eptr;              c = *eptr;
2924              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2925                if (c > 255)
2926                  {
2927                  if (op == OP_CLASS) break;
2928                  }
2929                else
2930    #endif
2931                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2932              eptr++;              eptr++;
2933              }              }
2934            while (eptr >= pp)            while (eptr >= pp)
2935              {              {
2936              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
             eptr--;  
2937              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2938                eptr--;
2939              }              }
2940            }            }
2941    
2942          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2943          }          }
2944    #undef BYTE_MAP
2945        }        }
2946      /* Control never gets here */      /* Control never gets here */
2947    
2948    
2949      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
2950      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
2951        mode, because Unicode properties are supported in non-UTF-8 mode. */
2952    
2953  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2954      case OP_XCLASS:      case OP_XCLASS:
2955        {        {
2956        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1722  for (;;) Line 2975  for (;;)
2975          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2976          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2977          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2978          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2979          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2980          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2981          break;          break;
2982    
2983          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1736  for (;;) Line 2989  for (;;)
2989    
2990        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2991          {          {
2992          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2993          GETCHARINC(c, eptr);            {
2994          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
2995              RRETURN(MATCH_NOMATCH);
2996              }
2997            GETCHARINCTEST(c, eptr);
2998            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
2999          }          }
3000    
3001        /* 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 3010  for (;;)
3010          {          {
3011          for (fi = min;; fi++)          for (fi = min;; fi++)
3012            {            {
3013            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3014            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3015            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3016            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3017            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3018                SCHECK_PARTIAL();
3019                RRETURN(MATCH_NOMATCH);
3020                }
3021              GETCHARINCTEST(c, eptr);
3022              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3023            }            }
3024          /* Control never gets here */          /* Control never gets here */
3025          }          }
# Line 1770  for (;;) Line 3032  for (;;)
3032          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3033            {            {
3034            int len = 1;            int len = 1;
3035            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3036            GETCHARLEN(c, eptr, len);              {
3037            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3038                break;
3039                }
3040    #ifdef SUPPORT_UTF
3041              GETCHARLENTEST(c, eptr, len);
3042    #else
3043              c = *eptr;
3044    #endif
3045              if (!PRIV(xclass)(c, data, utf)) break;
3046            eptr += len;            eptr += len;
3047            }            }
3048          for(;;)          for(;;)
3049            {            {
3050            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3051            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3052            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3053            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3054              if (utf) BACKCHAR(eptr);
3055    #endif
3056            }            }
3057          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3058          }          }
# Line 1792  for (;;) Line 3064  for (;;)
3064      /* Match a single character, casefully */      /* Match a single character, casefully */
3065    
3066      case OP_CHAR:      case OP_CHAR:
3067  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3068      if (utf8)      if (utf)
3069        {        {
3070        length = 1;        length = 1;
3071        ecode++;        ecode++;
3072        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3073        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3074            {
3075            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3076            RRETURN(MATCH_NOMATCH);
3077            }
3078        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3079        }        }
3080      else      else
3081  #endif  #endif
3082        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3083        {        {
3084        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3085            {
3086            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3087            RRETURN(MATCH_NOMATCH);
3088            }
3089        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3090        ecode += 2;        ecode += 2;
3091        }        }
3092      break;      break;
3093    
3094      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3095        subject, give up immediately. */
3096    
3097        case OP_CHARI:
3098        if (eptr >= md->end_subject)
3099          {
3100          SCHECK_PARTIAL();
3101          RRETURN(MATCH_NOMATCH);
3102          }
3103    
3104      case OP_CHARNC:  #ifdef SUPPORT_UTF
3105  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3106        {        {
3107        length = 1;        length = 1;
3108        ecode++;        ecode++;
3109        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3110    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3111        /* 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
3112        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3113          fast lookup table. We know that there is at least one byte left in the
3114          subject. */
3115    
3116        if (fc < 128)        if (fc < 128)
3117          {          {
3118          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3119                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3120            ecode++;
3121            eptr++;
3122          }          }
3123    
3124        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3125          use the value of "length" to check for sufficient bytes left, because the
3126          other case of the character may have more or fewer bytes.  */
3127    
3128        else        else
3129          {          {
3130          int dc;          unsigned int dc;
3131          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3132          ecode += length;          ecode += length;
3133    
3134          /* 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
3135          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. */  
3136    
3137          if (fc != dc)          if (fc != dc)
3138            {            {
3139  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3140            int chartype;            if (dc != UCD_OTHERCASE(fc))
           int othercase;  
           if (_pcre_ucp_findchar(fc, &chartype, &othercase) < 0 || dc != othercase)  
3141  #endif  #endif
3142              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3143            }            }
3144          }          }
3145        }        }
3146      else      else
3147  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3148    
3149      /* Non-UTF-8 mode */      /* Not UTF mode */
3150        {        {
3151        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3152        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3153          eptr++;
3154        ecode += 2;        ecode += 2;
3155        }        }
3156      break;      break;
3157    
3158      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3159    
3160      case OP_EXACT:      case OP_EXACT:
3161        case OP_EXACTI:
3162      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3163      ecode += 3;      ecode += 1 + IMM2_SIZE;
3164      goto REPEATCHAR;      goto REPEATCHAR;
3165    
3166        case OP_POSUPTO:
3167        case OP_POSUPTOI:
3168        possessive = TRUE;
3169        /* Fall through */
3170    
3171      case OP_UPTO:      case OP_UPTO:
3172        case OP_UPTOI:
3173      case OP_MINUPTO:      case OP_MINUPTO:
3174        case OP_MINUPTOI:
3175      min = 0;      min = 0;
3176      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3177      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3178      ecode += 3;      ecode += 1 + IMM2_SIZE;
3179        goto REPEATCHAR;
3180    
3181        case OP_POSSTAR:
3182        case OP_POSSTARI:
3183        possessive = TRUE;
3184        min = 0;
3185        max = INT_MAX;
3186        ecode++;
3187        goto REPEATCHAR;
3188    
3189        case OP_POSPLUS:
3190        case OP_POSPLUSI:
3191        possessive = TRUE;
3192        min = 1;
3193        max = INT_MAX;
3194        ecode++;
3195        goto REPEATCHAR;
3196    
3197        case OP_POSQUERY:
3198        case OP_POSQUERYI:
3199        possessive = TRUE;
3200        min = 0;
3201        max = 1;
3202        ecode++;
3203      goto REPEATCHAR;      goto REPEATCHAR;
3204    
3205      case OP_STAR:      case OP_STAR:
3206        case OP_STARI:
3207      case OP_MINSTAR:      case OP_MINSTAR:
3208        case OP_MINSTARI:
3209      case OP_PLUS:      case OP_PLUS:
3210        case OP_PLUSI:
3211      case OP_MINPLUS:      case OP_MINPLUS:
3212        case OP_MINPLUSI:
3213      case OP_QUERY:      case OP_QUERY:
3214        case OP_QUERYI:
3215      case OP_MINQUERY:      case OP_MINQUERY:
3216      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3217        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3218      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3219      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3220      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3221      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3222    
3223      /* 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. */  
3224    
3225      REPEATCHAR:      REPEATCHAR:
3226  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3227      if (utf8)      if (utf)
3228        {        {
3229        length = 1;        length = 1;
3230        charptr = ecode;        charptr = ecode;
3231        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3232        ecode += length;        ecode += length;
3233    
3234        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1913  for (;;) Line 3236  for (;;)
3236    
3237        if (length > 1)        if (length > 1)
3238          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3239  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3240          int othercase;          unsigned int othercase;
3241          int chartype;          if (op >= OP_STARI &&     /* Caseless */
3242          if ((ims & PCRE_CASELESS) != 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
3243               _pcre_ucp_findchar(fc, &chartype, &othercase) >= 0 &&            oclength = PRIV(ord2utf)(othercase, occhars);
3244               othercase > 0)          else oclength = 0;
           oclength = _pcre_ord2utf8(othercase, occhars);  
3245  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3246    
3247          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3248            {            {
3249            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3250            /* Need braces because of following else */              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3251            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3252              else if (oclength > 0 &&
3253                       eptr <= md->end_subject - oclength &&
3254                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3255    #endif  /* SUPPORT_UCP */
3256            else            else
3257              {              {
3258              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3259              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3260              }              }
3261            }            }
3262    
# Line 1943  for (;;) Line 3266  for (;;)
3266            {            {
3267            for (fi = min;; fi++)            for (fi = min;; fi++)
3268              {              {
3269              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3270              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3271              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3272              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3273              /* Need braces because of following else */                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3274              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3275                else if (oclength > 0 &&
3276                         eptr <= md->end_subject - oclength &&
3277                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3278    #endif  /* SUPPORT_UCP */
3279              else              else
3280                {                {
3281                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3282                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3283                }                }
3284              }              }
3285            /* Control never gets here */            /* Control never gets here */
3286            }            }
3287          else  
3288            else  /* Maximize */
3289            {            {
3290            pp = eptr;            pp = eptr;
3291            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3292              {              {
3293              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3294              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3295              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3296                else if (oclength > 0 &&
3297                         eptr <= md->end_subject - oclength &&
3298                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3299    #endif  /* SUPPORT_UCP */
3300              else              else
3301                {                {
3302                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3303                eptr += oclength;                break;
3304                }                }
3305              }              }
3306            while (eptr >= pp)  
3307             {            if (possessive) continue;
3308             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3309             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3310             eptr -= length;              {
3311             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3312            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3313                if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3314    #ifdef SUPPORT_UCP
3315                eptr--;
3316                BACKCHAR(eptr);
3317    #else   /* without SUPPORT_UCP */
3318                eptr -= length;
3319    #endif  /* SUPPORT_UCP */
3320                }
3321            }            }
3322          /* Control never gets here */          /* Control never gets here */
3323          }          }
# Line 1987  for (;;) Line 3327  for (;;)
3327        value of fc will always be < 128. */        value of fc will always be < 128. */
3328        }        }
3329      else      else
3330  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3331          /* When not in UTF-8 mode, load a single-byte character. */
     /* When not in UTF-8 mode, load a single-byte character. */  
       {  
       if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3332        fc = *ecode++;        fc = *ecode++;
       }  
3333    
3334      /* 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 one character, though we may
3335      may not be in UTF-8 mode. The code is duplicated for the caseless and      or may not be in UTF mode. The code is duplicated for the caseless and
3336      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3337      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3338      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2007  for (;;) Line 3343  for (;;)
3343      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3344        max, eptr));        max, eptr));
3345    
3346      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3347        {        {
3348        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3349          /* fc must be < 128 if UTF is enabled. */
3350          foc = md->fcc[fc];
3351    #else
3352    #ifdef SUPPORT_UTF
3353    #ifdef SUPPORT_UCP
3354          if (utf && fc > 127)
3355            foc = UCD_OTHERCASE(fc);
3356    #else
3357          if (utf && fc > 127)
3358            foc = fc;
3359    #endif /* SUPPORT_UCP */
3360          else
3361    #endif /* SUPPORT_UTF */
3362            foc = TABLE_GET(fc, md->fcc, fc);
3363    #endif /* COMPILE_PCRE8 */
3364    
3365        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3366          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3367            if (eptr >= md->end_subject)
3368              {
3369              SCHECK_PARTIAL();
3370              RRETURN(MATCH_NOMATCH);
3371              }
3372            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3373            eptr++;
3374            }
3375        if (min == max) continue;        if (min == max) continue;
3376        if (minimize)        if (minimize)
3377          {          {
3378          for (fi = min;; fi++)          for (fi = min;; fi++)
3379            {            {
3380            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3381            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3382            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3383                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3384                {
3385                SCHECK_PARTIAL();
3386              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3387                }
3388              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3389              eptr++;
3390            }            }
3391          /* Control never gets here */          /* Control never gets here */
3392          }          }
3393        else        else  /* Maximize */
3394          {          {
3395          pp = eptr;          pp = eptr;
3396          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3397            {            {
3398            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3399                {
3400                SCHECK_PARTIAL();
3401                break;
3402                }
3403              if (fc != *eptr && foc != *eptr) break;
3404            eptr++;            eptr++;
3405            }            }
3406    
3407            if (possessive) continue;
3408    
3409          while (eptr >= pp)          while (eptr >= pp)
3410            {            {
3411            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3412            eptr--;            eptr--;
3413            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3414            }            }
# Line 2048  for (;;) Line 3421  for (;;)
3421    
3422      else      else
3423        {        {
3424        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3425            {
3426            if (eptr >= md->end_subject)
3427              {
3428              SCHECK_PARTIAL();
3429              RRETURN(MATCH_NOMATCH);
3430              }
3431            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3432            }
3433    
3434        if (min == max) continue;        if (min == max) continue;
3435    
3436        if (minimize)        if (minimize)
3437          {          {
3438          for (fi = min;; fi++)          for (fi = min;; fi++)
3439            {            {
3440            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3441            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3442            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3443              if (eptr >= md->end_subject)
3444                {
3445                SCHECK_PARTIAL();
3446              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3447                }
3448              if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3449            }            }
3450          /* Control never gets here */          /* Control never gets here */
3451          }          }
3452        else        else  /* Maximize */
3453          {          {
3454          pp = eptr;          pp = eptr;
3455          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3456            {            {
3457            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3458                {
3459                SCHECK_PARTIAL();
3460                break;
3461                }
3462              if (fc != *eptr) break;
3463            eptr++;            eptr++;
3464            }            }
3465            if (possessive) continue;
3466    
3467          while (eptr >= pp)          while (eptr >= pp)
3468            {            {
3469            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3470            eptr--;            eptr--;
3471            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3472            }            }
# Line 2084  for (;;) Line 3479  for (;;)
3479      checking can be multibyte. */      checking can be multibyte. */
3480    
3481      case OP_NOT:      case OP_NOT:
3482      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3483        if (eptr >= md->end_subject)
3484          {
3485          SCHECK_PARTIAL();
3486          RRETURN(MATCH_NOMATCH);
3487          }
3488      ecode++;      ecode++;
3489      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3490      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3491        {        {
3492  #ifdef SUPPORT_UTF8        register int ch, och;
3493        if (c < 256)        ch = *ecode++;
3494  #endif  #ifdef COMPILE_PCRE8
3495        c = md->lcc[c];        /* ch must be < 128 if UTF is enabled. */
3496        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        och = md->fcc[ch];
3497    #else
3498    #ifdef SUPPORT_UTF
3499    #ifdef SUPPORT_UCP
3500          if (utf && ch > 127)
3501            och = UCD_OTHERCASE(ch);
3502    #else
3503          if (utf && ch > 127)
3504            och = ch;
3505    #endif /* SUPPORT_UCP */
3506          else
3507    #endif /* SUPPORT_UTF */
3508            och = TABLE_GET(ch, md->fcc, ch);
3509    #endif /* COMPILE_PCRE8 */
3510          if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3511        }        }
3512      else      else    /* Caseful */
3513        {