/[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 77 by nigel, Sat Feb 24 21:40:45 2007 UTC revision 933 by ph10, Sat Feb 25 12:18:23 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 the number of subject bytes matched
144                  -1 no match
145                  -2 partial match; always given if at end subject
146  */  */
147    
148  static BOOL  static int
149  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,
150    unsigned long int ims)    BOOL caseless)
151  {  {
152  const uschar *p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
153    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
154    
155  #ifdef DEBUG  #ifdef PCRE_DEBUG
156  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
157    printf("matching subject <null>");    printf("matching subject <null>");
158  else  else
# Line 146  pchars(p, length, FALSE, md); Line 165  pchars(p, length, FALSE, md);
165  printf("\n");  printf("\n");
166  #endif  #endif
167    
168  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
169    case the length is passed as zero). */
170    
171  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
172    
173  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
174    properly if Unicode properties are supported. Otherwise, we can check only
175    ASCII characters. */
176    
177  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
178    {    {
179    while (length-- > 0)  #ifdef SUPPORT_UTF
180      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
181      if (md->utf)
182        {
183        /* Match characters up to the end of the reference. NOTE: the number of
184        bytes matched may differ, because there are some characters whose upper and
185        lower case versions code as different numbers of bytes. For example, U+023A
186        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
187        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
188        the latter. It is important, therefore, to check the length along the
189        reference, not along the subject (earlier code did this wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193          {
194          int c, d;
195          if (eptr >= md->end_subject) return -2;   /* Partial match */
196          GETCHARINC(c, eptr);
197          GETCHARINC(d, p);
198          if (c != d && c != UCD_OTHERCASE(d)) return -1;
199          }
200        }
201      else
202    #endif
203    #endif
204    
205      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
206      is no UCP support. */
207        {
208        while (length-- > 0)
209          {
210          if (eptr >= md->end_subject) return -2;   /* Partial match */
211          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
212          p++;
213          eptr++;
214          }
215        }
216    }    }
217    
218    /* In the caseful case, we can just compare the bytes, whether or not we
219    are in UTF-8 mode. */
220    
221  else  else
222    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
223      while (length-- > 0)
224        {
225        if (eptr >= md->end_subject) return -2;   /* Partial match */
226        if (*p++ != *eptr++) return -1;
227        }
228      }
229    
230  return TRUE;  return (int)(eptr - eptr_start);
231  }  }
232    
233    
# Line 169  return TRUE; Line 236  return TRUE;
236  ****************************************************************************  ****************************************************************************
237                     RECURSION IN THE match() FUNCTION                     RECURSION IN THE match() FUNCTION
238    
239  The match() function is highly recursive. Some regular expressions can cause  The match() function is highly recursive, though not every recursive call
240  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
241  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
242  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
243  works fine.  saved for a recursive call. On Unix, the stack can be large, and this works
244    fine.
245  It turns out that on non-Unix systems there are problems with programs that  
246  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
247  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
248  for decades.) So....  has oodles of memory these days, and techniques for extending the stack have
249    been known for decades.) So....
250    
251  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
252  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
253  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
254  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
255  always used to.  always used to.
256    
257    The original heap-recursive code used longjmp(). However, it seems that this
258    can be very slow on some operating systems. Following a suggestion from Stan
259    Switzer, the use of longjmp() has been abolished, at the cost of having to
260    provide a unique number for each call to RMATCH. There is no way of generating
261    a sequence of numbers at compile time in C. I have given them names, to make
262    them stand out more clearly.
263    
264    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
265    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
266    tests. Furthermore, not using longjmp() means that local dynamic variables
267    don't have indeterminate values; this has meant that the frame size can be
268    reduced because the result can be "passed back" by straight setting of the
269    variable instead of being passed in the frame.
270  ****************************************************************************  ****************************************************************************
271  ***************************************************************************/  ***************************************************************************/
272    
273    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
274    below must be updated in sync.  */
275    
276  /* These versions of the macros use the stack, as normal */  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
277           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
278           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
279           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
280           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
281           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
282           RM61,  RM62, RM63, RM64, RM65, RM66 };
283    
284    /* These versions of the macros use the stack, as normal. There are debugging
285    versions and production versions. Note that the "rw" argument of RMATCH isn't
286    actually used in this definition. */
287    
288  #ifndef NO_RECURSE  #ifndef NO_RECURSE
289  #define REGISTER register  #define REGISTER register
290  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg)  
291    #ifdef PCRE_DEBUG
292    #define RMATCH(ra,rb,rc,rd,re,rw) \
293      { \
294      printf("match() called in line %d\n", __LINE__); \
295      rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
296      printf("to line %d\n", __LINE__); \
297      }
298    #define RRETURN(ra) \
299      { \
300      printf("match() returned %d from line %d ", ra, __LINE__); \
301      return ra; \
302      }
303    #else
304    #define RMATCH(ra,rb,rc,rd,re,rw) \
305      rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
306  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
307    #endif
308    
309  #else  #else
310    
311    
312  /* 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
313  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
314  match(), which never changes. */  argument of match(), which never changes. */
315    
316  #define REGISTER  #define REGISTER
317    
318  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
319    {\    {\
320    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
321    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL)\
     {\  
     newframe->Xeptr = ra;\  
     newframe->Xecode = rb;\  
     newframe->Xoffset_top = rc;\  
     newframe->Xims = re;\  
     newframe->Xeptrb = rf;\  
     newframe->Xflags = rg;\  
     newframe->Xprevframe = frame;\  
     frame = newframe;\  
     DPRINTF(("restarting from line %d\n", __LINE__));\  
     goto HEAP_RECURSE;\  
     }\  
   else\  
322      {\      {\
323      DPRINTF(("longjumped back to line %d\n", __LINE__));\      newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
324      frame = md->thisframe;\      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
325      rx = frame->Xresult;\      newframe->Xnextframe = NULL;\
326        frame->Xnextframe = newframe;\
327      }\      }\
328      frame->Xwhere = rw;\
329      newframe->Xeptr = ra;\
330      newframe->Xecode = rb;\
331      newframe->Xmstart = mstart;\
332      newframe->Xoffset_top = rc;\
333      newframe->Xeptrb = re;\
334      newframe->Xrdepth = frame->Xrdepth + 1;\
335      newframe->Xprevframe = frame;\
336      frame = newframe;\
337      DPRINTF(("restarting from line %d\n", __LINE__));\
338      goto HEAP_RECURSE;\
339      L_##rw:\
340      DPRINTF(("jumped back to line %d\n", __LINE__));\
341    }    }
342    
343  #define RRETURN(ra)\  #define RRETURN(ra)\
344    {\    {\
345    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
346    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
347    if (frame != NULL)\    if (frame != NULL)\
348      {\      {\
349      frame->Xresult = ra;\      rrc = ra;\
350      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
351      }\      }\
352    return ra;\    return ra;\
353    }    }
# Line 247  match(), which never changes. */ Line 357  match(), which never changes. */
357    
358  typedef struct heapframe {  typedef struct heapframe {
359    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
360      struct heapframe *Xnextframe;
361    
362    /* Function arguments that may change */    /* Function arguments that may change */
363    
364    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
365    const uschar *Xecode;    const pcre_uchar *Xecode;
366      PCRE_PUCHAR Xmstart;
367    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
368    eptrblock *Xeptrb;    eptrblock *Xeptrb;
369    int Xflags;    unsigned int Xrdepth;
370    
371    /* Function local variables */    /* Function local variables */
372    
373    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
374    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
375    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
376    const uschar *Xnext;  #endif
377    const uschar *Xpp;    PCRE_PUCHAR Xdata;
378    const uschar *Xprev;    PCRE_PUCHAR Xnext;
379    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
380      PCRE_PUCHAR Xprev;
381      PCRE_PUCHAR Xsaved_eptr;
382    
383    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
384    
385    BOOL Xcur_is_word;    BOOL Xcur_is_word;
386    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
387    BOOL Xprev_is_word;    BOOL Xprev_is_word;
388    
   unsigned long int Xoriginal_ims;  
   
389  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
390    int Xprop_type;    int Xprop_type;
391      int Xprop_value;
392    int Xprop_fail_result;    int Xprop_fail_result;
393    int Xprop_category;    int Xoclength;
394    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_othercase;  
   int Xprop_test_against;  
   int *Xprop_test_variable;  
395  #endif  #endif
396    
397      int Xcodelink;
398    int Xctype;    int Xctype;
399    int Xfc;    unsigned int Xfc;
400    int Xfi;    int Xfi;
401    int Xlength;    int Xlength;
402    int Xmax;    int Xmax;
# Line 301  typedef struct heapframe { Line 410  typedef struct heapframe {
410    
411    eptrblock Xnewptrb;    eptrblock Xnewptrb;
412    
413    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
414    
415    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
416    
417  } heapframe;  } heapframe;
418    
# Line 320  typedef struct heapframe { Line 428  typedef struct heapframe {
428  *         Match from current position            *  *         Match from current position            *
429  *************************************************/  *************************************************/
430    
431  /* 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  
432  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
433  same response.  same response. */
434    
435    /* These macros pack up tests that are used for partial matching, and which
436    appear several times in the code. We set the "hit end" flag if the pointer is
437    at the end of the subject and also past the start of the subject (i.e.
438    something has been matched). For hard partial matching, we then return
439    immediately. The second one is used when we already know we are past the end of
440    the subject. */
441    
442    #define CHECK_PARTIAL()\
443      if (md->partial != 0 && eptr >= md->end_subject && \
444          eptr > md->start_used_ptr) \
445        { \
446        md->hitend = TRUE; \
447        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
448        }
449    
450  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
451  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
452        { \
453        md->hitend = TRUE; \
454        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
455        }
456    
457    
458    /* Performance note: It might be tempting to extract commonly used fields from
459    the md structure (e.g. utf, end_subject) into individual variables to improve
460  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
461  made performance worse.  made performance worse.
462    
463  Arguments:  Arguments:
464     eptr        pointer in subject     eptr        pointer to current character in subject
465     ecode       position in code     ecode       pointer to current position in compiled code
466       mstart      pointer to the current match start position (can be modified
467                     by encountering \K)
468     offset_top  current top pointer     offset_top  current top pointer
469     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
470     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
471                   brackets - for testing for empty matches                   brackets - for testing for empty matches
472     flags       can contain     rdepth      the recursion depth
                  match_condassert - this is an assertion condition  
                  match_isgroup - this is the start of a bracketed group  
473    
474  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
475                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
476                   a negative MATCH_xxx value for PRUNE, SKIP, etc
477                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
478                   (e.g. stopped by recursion limit)                   (e.g. stopped by repeated call or recursion limit)
479  */  */
480    
481  static int  static int
482  match(REGISTER const uschar *eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
483    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
484    int flags)    unsigned int rdepth)
485  {  {
486  /* 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,
487  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
488  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
489    
490  register int  rrc;    /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
491  register int  i;      /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
492  register int  c;      /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
493  register BOOL utf8;   /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
494    
495    BOOL minimize, possessive; /* Quantifier options */
496    BOOL caseless;
497    int condcode;
498    
499  /* 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
500  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
501  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
502  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
503    the top-level on the stack rather than malloc-ing them all gives a performance
504    boost in many cases where there is not much "recursion". */
505    
506  #ifdef NO_RECURSE  #ifdef NO_RECURSE
507  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
508    
509  /* Copy in the original argument variables */  /* Copy in the original argument variables */
510    
511  frame->Xeptr = eptr;  frame->Xeptr = eptr;
512  frame->Xecode = ecode;  frame->Xecode = ecode;
513    frame->Xmstart = mstart;
514  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
515  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
516  frame->Xflags = flags;  frame->Xrdepth = rdepth;
517    
518  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
519    
# Line 390  HEAP_RECURSE: Line 523  HEAP_RECURSE:
523    
524  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
525  #define ecode              frame->Xecode  #define ecode              frame->Xecode
526    #define mstart             frame->Xmstart
527  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
528  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
529  #define flags              frame->Xflags  #define rdepth             frame->Xrdepth
530    
531  /* Ditto for the local variables */  /* Ditto for the local variables */
532    
533  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
534  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
535  #endif  #endif
536  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
537    #define codelink           frame->Xcodelink
538  #define data               frame->Xdata  #define data               frame->Xdata
539  #define next               frame->Xnext  #define next               frame->Xnext
540  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 411  HEAP_RECURSE: Line 545  HEAP_RECURSE:
545    
546  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
547  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
548  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
549    
 #define original_ims       frame->Xoriginal_ims  
   
550  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
551  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
552    #define prop_value         frame->Xprop_value
553  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
554  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
555  #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  
556  #endif  #endif
557    
558  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 447  HEAP_RECURSE: Line 576  HEAP_RECURSE:
576  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
577  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
578    
579  #else  #else         /* NO_RECURSE not defined */
580  #define fi i  #define fi i
581  #define fc c  #define fc c
582    
583    /* Many of the following variables are used only in small blocks of the code.
584    My normal style of coding would have declared them within each of those blocks.
585    However, in order to accommodate the version of this code that uses an external
586    "stack" implemented on the heap, it is easier to declare them all here, so the
587    declarations can be cut out in a block. The only declarations within blocks
588    below are for variables that do not have to be preserved over a recursive call
589    to RMATCH(). */
590    
591  #ifdef SUPPORT_UTF8                /* Many of these variables are used ony */  #ifdef SUPPORT_UTF
592  const uschar *charptr;             /* small blocks of the code. My normal  */  const pcre_uchar *charptr;
593  #endif                             /* style of coding would have declared  */  #endif
594  const uschar *callpat;             /* them within each of those blocks.    */  const pcre_uchar *callpat;
595  const uschar *data;                /* However, in order to accommodate the */  const pcre_uchar *data;
596  const uschar *next;                /* version of this code that uses an    */  const pcre_uchar *next;
597  const uschar *pp;                  /* external "stack" implemented on the  */  PCRE_PUCHAR       pp;
598  const uschar *prev;                /* heap, it is easier to declare them   */  const pcre_uchar *prev;
599  const uschar *saved_eptr;          /* all here, so the declarations can    */  PCRE_PUCHAR       saved_eptr;
600                                     /* be cut out in a block. The only      */  
601  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;  
602    
603  unsigned long int original_ims;  BOOL cur_is_word;
604    BOOL condition;
605    BOOL prev_is_word;
606    
607  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
608  int prop_type;  int prop_type;
609    int prop_value;
610  int prop_fail_result;  int prop_fail_result;
611  int prop_category;  int oclength;
612  int prop_chartype;  pcre_uchar occhars[6];
 int prop_othercase;  
 int prop_test_against;  
 int *prop_test_variable;  
613  #endif  #endif
614    
615    int codelink;
616  int ctype;  int ctype;
617  int length;  int length;
618  int max;  int max;
# Line 493  int save_offset1, save_offset2, save_off Line 625  int save_offset1, save_offset2, save_off
625  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
626    
627  eptrblock newptrb;  eptrblock newptrb;
628  #endif  
629    /* There is a special fudge for calling match() in a way that causes it to
630    measure the size of its basic stack frame when the stack is being used for
631    recursion. The second argument (ecode) being NULL triggers this behaviour. It
632    cannot normally ever be NULL. The return is the negated value of the frame
633    size. */
634    
635    if (ecode == NULL)
636      {
637      if (rdepth == 0)
638        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
639      else
640        {
641        int len = (char *)&rdepth - (char *)eptr;
642        return (len > 0)? -len : len;
643        }
644      }
645    #endif     /* NO_RECURSE */
646    
647    /* To save space on the stack and in the heap frame, I have doubled up on some
648    of the local variables that are used only in localised parts of the code, but
649    still need to be preserved over recursive calls of match(). These macros define
650    the alternative names that are used. */
651    
652    #define allow_zero    cur_is_word
653    #define cbegroup      condition
654    #define code_offset   codelink
655    #define condassert    condition
656    #define matched_once  prev_is_word
657    #define foc           number
658    #define save_mark     data
659    
660  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
661  variables. */  variables. */
662    
663  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
664    prop_value = 0;
665  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_against = 0;  
 prop_test_variable = NULL;  
666  #endif  #endif
667    
 /* 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. */  
668    
669  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
670    when NO_RECURSE is not defined, in order to reduce the amount of stack that is
671    used. Thanks to Ian Taylor for noticing this possibility and sending the
672    original patch. */
673    
674    TAIL_RECURSE:
675    
676    /* OK, now we can get on with the real code of the function. Recursive calls
677    are specified by the macro RMATCH and RRETURN is used to return. When
678    NO_RECURSE is *not* defined, these just turn into a recursive call to match()
679    and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
680    defined). However, RMATCH isn't like a function call because it's quite a
681    complicated macro. It has to be used in one particular way. This shouldn't,
682    however, impact performance when true recursion is being used. */
683    
684    #ifdef SUPPORT_UTF
685    utf = md->utf;       /* Local copy of the flag */
686    #else
687    utf = FALSE;
688    #endif
689    
690  original_ims = ims;    /* Save for resetting on ')' */  /* First check that we haven't called match() too many times, or that we
691  utf8 = md->utf8;       /* Local copy of the flag */  haven't exceeded the recursive call limit. */
692    
693    if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
694    if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
695    
696  /* 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
697  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
698  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
699  this stack. */  up space on the stack. See also MATCH_CONDASSERT below.
700    
701    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
702    such remembered pointers, to be checked when we hit the closing ket, in order
703    to break infinite loops that match no characters. When match() is called in
704    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
705    NOT be used with tail recursion, because the memory block that is used is on
706    the stack, so a new one may be required for each match(). */
707    
708  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
709    {    {
   newptrb.epb_prev = eptrb;  
710    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
711      newptrb.epb_prev = eptrb;
712    eptrb = &newptrb;    eptrb = &newptrb;
713      md->match_function_type = 0;
714    }    }
715    
716  /* Now start processing the operations. */  /* Now start processing the opcodes. */
717    
718  for (;;)  for (;;)
719    {    {
720      minimize = possessive = FALSE;
721    op = *ecode;    op = *ecode;
   minimize = FALSE;  
722    
723    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
724    matching at least one subject character. */      {
725        case OP_MARK:
726        md->nomatch_mark = ecode + 2;
727        md->mark = NULL;    /* In case previously set by assertion */
728        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
729          eptrb, RM55);
730        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
731             md->mark == NULL) md->mark = ecode + 2;
732    
733        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
734        argument, and we must check whether that argument matches this MARK's
735        argument. It is passed back in md->start_match_ptr (an overloading of that
736        variable). If it does match, we reset that variable to the current subject
737        position and return MATCH_SKIP. Otherwise, pass back the return code
738        unaltered. */
739    
740        else if (rrc == MATCH_SKIP_ARG &&
741            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
742          {
743          md->start_match_ptr = eptr;
744          RRETURN(MATCH_SKIP);
745          }
746        RRETURN(rrc);
747    
748    if (md->partial &&      case OP_FAIL:
749        eptr >= md->end_subject &&      RRETURN(MATCH_NOMATCH);
       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. */  
750    
751    if (op > OP_BRA)      /* COMMIT overrides PRUNE, SKIP, and THEN */
752      {  
753      number = op - OP_BRA;      case OP_COMMIT:
754        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
755          eptrb, RM52);
756        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
757            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
758            rrc != MATCH_THEN)
759          RRETURN(rrc);
760        RRETURN(MATCH_COMMIT);
761    
762        /* PRUNE overrides THEN */
763    
764        case OP_PRUNE:
765        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
766          eptrb, RM51);
767        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
768        RRETURN(MATCH_PRUNE);
769    
770        case OP_PRUNE_ARG:
771        md->nomatch_mark = ecode + 2;
772        md->mark = NULL;    /* In case previously set by assertion */
773        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
774          eptrb, RM56);
775        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
776             md->mark == NULL) md->mark = ecode + 2;
777        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
778        RRETURN(MATCH_PRUNE);
779    
780        /* SKIP overrides PRUNE and THEN */
781    
782        case OP_SKIP:
783        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
784          eptrb, RM53);
785        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
786          RRETURN(rrc);
787        md->start_match_ptr = eptr;   /* Pass back current position */
788        RRETURN(MATCH_SKIP);
789    
790        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
791        nomatch_mark. There is a flag that disables this opcode when re-matching a
792        pattern that ended with a SKIP for which there was not a matching MARK. */
793    
794        case OP_SKIP_ARG:
795        if (md->ignore_skip_arg)
796          {
797          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
798          break;
799          }
800        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
801          eptrb, RM57);
802        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
803          RRETURN(rrc);
804    
805        /* Pass back the current skip name by overloading md->start_match_ptr and
806        returning the special MATCH_SKIP_ARG return code. This will either be
807        caught by a matching MARK, or get to the top, where it causes a rematch
808        with the md->ignore_skip_arg flag set. */
809    
810        md->start_match_ptr = ecode + 2;
811        RRETURN(MATCH_SKIP_ARG);
812    
813        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
814        the branch in which it occurs can be determined. Overload the start of
815        match pointer to do this. */
816    
817        case OP_THEN:
818        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
819          eptrb, RM54);
820        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
821        md->start_match_ptr = ecode;
822        RRETURN(MATCH_THEN);
823    
824        case OP_THEN_ARG:
825        md->nomatch_mark = ecode + 2;
826        md->mark = NULL;    /* In case previously set by assertion */
827        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
828          md, eptrb, RM58);
829        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
830             md->mark == NULL) md->mark = ecode + 2;
831        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
832        md->start_match_ptr = ecode;
833        RRETURN(MATCH_THEN);
834    
835        /* Handle an atomic group that does not contain any capturing parentheses.
836        This can be handled like an assertion. Prior to 8.13, all atomic groups
837        were handled this way. In 8.13, the code was changed as below for ONCE, so
838        that backups pass through the group and thereby reset captured values.
839        However, this uses a lot more stack, so in 8.20, atomic groups that do not
840        contain any captures generate OP_ONCE_NC, which can be handled in the old,
841        less stack intensive way.
842    
843        Check the alternative branches in turn - the matching won't pass the KET
844        for this kind of subpattern. If any one branch matches, we carry on as at
845        the end of a normal bracket, leaving the subject pointer, but resetting
846        the start-of-match value in case it was changed by \K. */
847    
848        case OP_ONCE_NC:
849        prev = ecode;
850        saved_eptr = eptr;
851        save_mark = md->mark;
852        do
853          {
854          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
855          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
856            {
857            mstart = md->start_match_ptr;
858            break;
859            }
860          if (rrc == MATCH_THEN)
861            {
862            next = ecode + GET(ecode,1);
863            if (md->start_match_ptr < next &&
864                (*ecode == OP_ALT || *next == OP_ALT))
865              rrc = MATCH_NOMATCH;
866            }
867    
868          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
869          ecode += GET(ecode,1);
870          md->mark = save_mark;
871          }
872        while (*ecode == OP_ALT);
873    
874        /* If hit the end of the group (which could be repeated), fail */
875    
876        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
877    
878        /* Continue as from after the group, updating the offsets high water
879        mark, since extracts may have been taken. */
880    
881        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
882    
883        offset_top = md->end_offset_top;
884        eptr = md->end_match_ptr;
885    
886        /* For a non-repeating ket, just continue at this level. This also
887        happens for a repeating ket if no characters were matched in the group.
888        This is the forcible breaking of infinite loops as implemented in Perl
889        5.005. */
890    
891        if (*ecode == OP_KET || eptr == saved_eptr)
892          {
893          ecode += 1+LINK_SIZE;
894          break;
895          }
896    
897        /* The repeating kets try the rest of the pattern or restart from the
898        preceding bracket, in the appropriate order. The second "call" of match()
899        uses tail recursion, to avoid using another stack frame. */
900    
901      /* For extended extraction brackets (large number), we have to fish out the      if (*ecode == OP_KETRMIN)
902      number from a dummy opcode at the start. */        {
903          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
904          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
905          ecode = prev;
906          goto TAIL_RECURSE;
907          }
908        else  /* OP_KETRMAX */
909          {
910          md->match_function_type = MATCH_CBEGROUP;
911          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
912          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
913          ecode += 1 + LINK_SIZE;
914          goto TAIL_RECURSE;
915          }
916        /* Control never gets here */
917    
918      if (number > EXTRACT_BASIC_MAX)      /* Handle a capturing bracket, other than those that are possessive with an
919        number = GET2(ecode, 2+LINK_SIZE);      unlimited repeat. If there is space in the offset vector, save the current
920        subject position in the working slot at the top of the vector. We mustn't
921        change the current values of the data slot, because they may be set from a
922        previous iteration of this group, and be referred to by a reference inside
923        the group. A failure to match might occur after the group has succeeded,
924        if something later on doesn't match. For this reason, we need to restore
925        the working value and also the values of the final offsets, in case they
926        were set by a previous iteration of the same bracket.
927    
928        If there isn't enough space in the offset vector, treat this as if it were
929        a non-capturing bracket. Don't worry about setting the flag for the error
930        case here; that is handled in the code for KET. */
931    
932        case OP_CBRA:
933        case OP_SCBRA:
934        number = GET2(ecode, 1+LINK_SIZE);
935      offset = number << 1;      offset = number << 1;
936    
937  #ifdef DEBUG  #ifdef PCRE_DEBUG
938      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
939        printf("subject=");
940      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
941      printf("\n");      printf("\n");
942  #endif  #endif
# Line 580  for (;;) Line 947  for (;;)
947        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
948        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
949        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
950          save_mark = md->mark;
951    
952        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
953        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
954            (int)(eptr - md->start_subject);
955    
956        do        for (;;)
957          {          {
958          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
959            match_isgroup);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
960              eptrb, RM1);
961            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
962    
963            /* If we backed up to a THEN, check whether it is within the current
964            branch by comparing the address of the THEN that is passed back with
965            the end of the branch. If it is within the current branch, and the
966            branch is one of two or more alternatives (it either starts or ends
967            with OP_ALT), we have reached the limit of THEN's action, so convert
968            the return code to NOMATCH, which will cause normal backtracking to
969            happen from now on. Otherwise, THEN is passed back to an outer
970            alternative. This implements Perl's treatment of parenthesized groups,
971            where a group not containing | does not affect the current alternative,
972            that is, (X) is NOT the same as (X|(*F)). */
973    
974            if (rrc == MATCH_THEN)
975              {
976              next = ecode + GET(ecode,1);
977              if (md->start_match_ptr < next &&
978                  (*ecode == OP_ALT || *next == OP_ALT))
979                rrc = MATCH_NOMATCH;
980              }
981    
982            /* Anything other than NOMATCH is passed back. */
983    
984          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
985          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
986          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
987            md->mark = save_mark;
988            if (*ecode != OP_ALT) break;
989          }          }
       while (*ecode == OP_ALT);  
990    
991        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
992        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
993        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
994        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
995    
996        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
997    
998          RRETURN(rrc);
999        }        }
1000    
1001      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1002        as a non-capturing bracket. */
1003    
1004      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1005      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1006    
1007    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1008    
1009    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1010      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1011      case OP_BRA:     /* Non-capturing bracket: optimized */  
1012      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
1013      do      repeat and ONCE group with no captures. Loop for all the alternatives.
1014    
1015        When we get to the final alternative within the brackets, we used to return
1016        the result of a recursive call to match() whatever happened so it was
1017        possible to reduce stack usage by turning this into a tail recursion,
1018        except in the case of a possibly empty group. However, now that there is
1019        the possiblity of (*THEN) occurring in the final alternative, this
1020        optimization is no longer always possible.
1021    
1022        We can optimize if we know there are no (*THEN)s in the pattern; at present
1023        this is the best that can be done.
1024    
1025        MATCH_ONCE is returned when the end of an atomic group is successfully
1026        reached, but subsequent matching fails. It passes back up the tree (causing
1027        captured values to be reset) until the original atomic group level is
1028        reached. This is tested by comparing md->once_target with the start of the
1029        group. At this point, the return is converted into MATCH_NOMATCH so that
1030        previous backup points can be taken. */
1031    
1032        case OP_ONCE:
1033        case OP_BRA:
1034        case OP_SBRA:
1035        DPRINTF(("start non-capturing bracket\n"));
1036    
1037        for (;;)
1038        {        {
1039        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;
1040          match_isgroup);  
1041        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        /* If this is not a possibly empty group, and there are no (*THEN)s in
1042          the pattern, and this is the final alternative, optimize as described
1043          above. */
1044    
1045          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1046            {
1047            ecode += PRIV(OP_lengths)[*ecode];
1048            goto TAIL_RECURSE;
1049            }
1050    
1051          /* In all other cases, we have to make another call to match(). */
1052    
1053          save_mark = md->mark;
1054          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1055            RM2);
1056    
1057          /* See comment in the code for capturing groups above about handling
1058          THEN. */
1059    
1060          if (rrc == MATCH_THEN)
1061            {
1062            next = ecode + GET(ecode,1);
1063            if (md->start_match_ptr < next &&
1064                (*ecode == OP_ALT || *next == OP_ALT))
1065              rrc = MATCH_NOMATCH;
1066            }
1067    
1068          if (rrc != MATCH_NOMATCH)
1069            {
1070            if (rrc == MATCH_ONCE)
1071              {
1072              const pcre_uchar *scode = ecode;
1073              if (*scode != OP_ONCE)           /* If not at start, find it */
1074                {
1075                while (*scode == OP_ALT) scode += GET(scode, 1);
1076                scode -= GET(scode, 1);
1077                }
1078              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1079              }
1080            RRETURN(rrc);
1081            }
1082        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1083          md->mark = save_mark;
1084          if (*ecode != OP_ALT) break;
1085        }        }
1086      while (*ecode == OP_ALT);  
     DPRINTF(("bracket 0 failed\n"));  
1087      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1088    
1089      /* Conditional group: compilation checked that there are no more than      /* Handle possessive capturing brackets with an unlimited repeat. We come
1090      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
1091      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
1092      exactly what going to the ket would do. */      different. The end of these brackets will always be OP_KETRPOS, which
1093        returns MATCH_KETRPOS without going further in the pattern. By this means
1094        we can handle the group by iteration rather than recursion, thereby
1095        reducing the amount of stack needed. */
1096    
1097        case OP_CBRAPOS:
1098        case OP_SCBRAPOS:
1099        allow_zero = FALSE;
1100    
1101      case OP_COND:      POSSESSIVE_CAPTURE:
1102      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      number = GET2(ecode, 1+LINK_SIZE);
1103        {      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);  
       }  
1104    
1105      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1106      the final argument TRUE causes it to stop at the end of an assertion. */      printf("start possessive bracket %d\n", number);
1107        printf("subject=");
1108        pchars(eptr, 16, TRUE, md);
1109        printf("\n");
1110    #endif
1111    
1112      else      if (offset < md->offset_max)
1113        {        {
1114        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1115            match_condassert | match_isgroup);        code_offset = (int)(ecode - md->start_code);
1116        if (rrc == MATCH_MATCH)  
1117          save_offset1 = md->offset_vector[offset];
1118          save_offset2 = md->offset_vector[offset+1];
1119          save_offset3 = md->offset_vector[md->offset_end - number];
1120          save_capture_last = md->capture_last;
1121    
1122          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1123    
1124          /* Each time round the loop, save the current subject position for use
1125          when the group matches. For MATCH_MATCH, the group has matched, so we
1126          restart it with a new subject starting position, remembering that we had
1127          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1128          usual. If we haven't matched any alternatives in any iteration, check to
1129          see if a previous iteration matched. If so, the group has matched;
1130          continue from afterwards. Otherwise it has failed; restore the previous
1131          capture values before returning NOMATCH. */
1132    
1133          for (;;)
1134            {
1135            md->offset_vector[md->offset_end - number] =
1136              (int)(eptr - md->start_subject);
1137            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1138            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1139              eptrb, RM63);
1140            if (rrc == MATCH_KETRPOS)
1141              {
1142              offset_top = md->end_offset_top;
1143              eptr = md->end_match_ptr;
1144              ecode = md->start_code + code_offset;
1145              save_capture_last = md->capture_last;
1146              matched_once = TRUE;
1147              continue;
1148              }
1149    
1150            /* See comment in the code for capturing groups above about handling
1151            THEN. */
1152    
1153            if (rrc == MATCH_THEN)
1154              {
1155              next = ecode + GET(ecode,1);
1156              if (md->start_match_ptr < next &&
1157                  (*ecode == OP_ALT || *next == OP_ALT))
1158                rrc = MATCH_NOMATCH;
1159              }
1160    
1161            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1162            md->capture_last = save_capture_last;
1163            ecode += GET(ecode, 1);
1164            if (*ecode != OP_ALT) break;
1165            }
1166    
1167          if (!matched_once)
1168          {          {
1169          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          md->offset_vector[offset] = save_offset1;
1170          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          md->offset_vector[offset+1] = save_offset2;
1171            md->offset_vector[md->offset_end - number] = save_offset3;
1172          }          }
1173        else if (rrc != MATCH_NOMATCH)  
1174          if (allow_zero || matched_once)
1175          {          {
1176          RRETURN(rrc);         /* Need braces because of following else */          ecode += 1 + LINK_SIZE;
1177            break;
1178          }          }
1179        else ecode += GET(ecode, 1);  
1180        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        RRETURN(MATCH_NOMATCH);
         match_isgroup);  
       RRETURN(rrc);  
1181        }        }
     /* Control never reaches here */  
1182    
1183      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1184      encountered. */      as a non-capturing bracket. */
1185    
1186      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1187      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1188    
1189      /* 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. */  
1190    
1191      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1192      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;  
       }  
1193    
1194      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1195      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1196        without the capturing complication. It is written out separately for speed
1197        and cleanliness. */
1198    
1199      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1200      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1201      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;  
1202    
1203      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1204      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1205      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. */  
1206    
1207      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1208        {        {
1209        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1210          match_isgroup);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1211        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1212          if (rrc == MATCH_KETRPOS)
1213            {
1214            offset_top = md->end_offset_top;
1215            eptr = md->end_match_ptr;
1216            ecode = md->start_code + code_offset;
1217            matched_once = TRUE;
1218            continue;
1219            }
1220    
1221          /* See comment in the code for capturing groups above about handling
1222          THEN. */
1223    
1224          if (rrc == MATCH_THEN)
1225            {
1226            next = ecode + GET(ecode,1);
1227            if (md->start_match_ptr < next &&
1228                (*ecode == OP_ALT || *next == OP_ALT))
1229              rrc = MATCH_NOMATCH;
1230            }
1231    
1232        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1233        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1234          if (*ecode != OP_ALT) break;
1235        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1236    
1237      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1238          {
1239          ecode += 1 + LINK_SIZE;
1240          break;
1241          }
1242        RRETURN(MATCH_NOMATCH);
1243    
1244      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1245    
1246      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1247      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1248        past the end if there is only one branch, but that's OK because that is
1249        exactly what going to the ket would do. */
1250    
1251      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1252      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1253      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1254    
1255      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1256        inserted between OP_COND and an assertion condition. */
1257    
1258      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1259        {        {
1260        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (PUBL(callout) != NULL)
1261          match_isgroup);          {
1262        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1263        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1264        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1265            cb.offset_vector    = md->offset_vector;
1266    #ifdef COMPILE_PCRE8
1267            cb.subject          = (PCRE_SPTR)md->start_subject;
1268    #else
1269            cb.subject          = (PCRE_SPTR16)md->start_subject;
1270    #endif
1271            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1272            cb.start_match      = (int)(mstart - md->start_subject);
1273            cb.current_position = (int)(eptr - md->start_subject);
1274            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1275            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1276            cb.capture_top      = offset_top/2;
1277            cb.capture_last     = md->capture_last;
1278            cb.callout_data     = md->callout_data;
1279            cb.mark             = md->nomatch_mark;
1280            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1281            if (rrc < 0) RRETURN(rrc);
1282            }
1283          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1284        }        }
     while (*ecode == OP_ALT);  
   
     if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);  
1285    
1286      ecode += 1 + LINK_SIZE;      condcode = ecode[LINK_SIZE+1];
     continue;  
1287    
1288      /* 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. */  
1289    
1290      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1291        {        {
1292        c = GET(ecode,1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
       for (i = 0; i < c; i++)  
1293          {          {
1294          eptr--;          condition = FALSE;
1295          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          ecode += GET(ecode, 1);
         BACKCHAR(eptr)  
1296          }          }
1297        }        else
1298      else          {
1299            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1300            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1301    
1302            /* If the test is for recursion into a specific subpattern, and it is
1303            false, but the test was set up by name, scan the table to see if the
1304            name refers to any other numbers, and test them. The condition is true
1305            if any one is set. */
1306    
1307            if (!condition && condcode == OP_NRREF)
1308              {
1309              pcre_uchar *slotA = md->name_table;
1310              for (i = 0; i < md->name_count; i++)
1311                {
1312                if (GET2(slotA, 0) == recno) break;
1313                slotA += md->name_entry_size;
1314                }
1315    
1316              /* Found a name for the number - there can be only one; duplicate
1317              names for different numbers are allowed, but not vice versa. First
1318              scan down for duplicates. */
1319    
1320              if (i < md->name_count)
1321                {
1322                pcre_uchar *slotB = slotA;
1323                while (slotB > md->name_table)
1324                  {
1325                  slotB -= md->name_entry_size;
1326                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1327                    {
1328                    condition = GET2(slotB, 0) == md->recursive->group_num;
1329                    if (condition) break;
1330                    }
1331                  else break;
1332                  }
1333    
1334                /* Scan up for duplicates */
1335    
1336                if (!condition)
1337                  {
1338                  slotB = slotA;
1339                  for (i++; i < md->name_count; i++)
1340                    {
1341                    slotB += md->name_entry_size;
1342                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1343                      {
1344                      condition = GET2(slotB, 0) == md->recursive->group_num;
1345                      if (condition) break;
1346                      }
1347                    else break;
1348                    }
1349                  }
1350                }
1351              }
1352    
1353            /* Chose branch according to the condition */
1354    
1355            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1356            }
1357          }
1358    
1359        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1360          {
1361          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1362          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1363    
1364          /* If the numbered capture is unset, but the reference was by name,
1365          scan the table to see if the name refers to any other numbers, and test
1366          them. The condition is true if any one is set. This is tediously similar
1367          to the code above, but not close enough to try to amalgamate. */
1368    
1369          if (!condition && condcode == OP_NCREF)
1370            {
1371            int refno = offset >> 1;
1372            pcre_uchar *slotA = md->name_table;
1373    
1374            for (i = 0; i < md->name_count; i++)
1375              {
1376              if (GET2(slotA, 0) == refno) break;
1377              slotA += md->name_entry_size;
1378              }
1379    
1380            /* Found a name for the number - there can be only one; duplicate names
1381            for different numbers are allowed, but not vice versa. First scan down
1382            for duplicates. */
1383    
1384            if (i < md->name_count)
1385              {
1386              pcre_uchar *slotB = slotA;
1387              while (slotB > md->name_table)
1388                {
1389                slotB -= md->name_entry_size;
1390                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1391                  {
1392                  offset = GET2(slotB, 0) << 1;
1393                  condition = offset < offset_top &&
1394                    md->offset_vector[offset] >= 0;
1395                  if (condition) break;
1396                  }
1397                else break;
1398                }
1399    
1400              /* Scan up for duplicates */
1401    
1402              if (!condition)
1403                {
1404                slotB = slotA;
1405                for (i++; i < md->name_count; i++)
1406                  {
1407                  slotB += md->name_entry_size;
1408                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1409                    {
1410                    offset = GET2(slotB, 0) << 1;
1411                    condition = offset < offset_top &&
1412                      md->offset_vector[offset] >= 0;
1413                    if (condition) break;
1414                    }
1415                  else break;
1416                  }
1417                }
1418              }
1419            }
1420    
1421          /* Chose branch according to the condition */
1422    
1423          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1424          }
1425    
1426        else if (condcode == OP_DEF)     /* DEFINE - always false */
1427          {
1428          condition = FALSE;
1429          ecode += GET(ecode, 1);
1430          }
1431    
1432        /* The condition is an assertion. Call match() to evaluate it - setting
1433        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1434        an assertion. */
1435    
1436        else
1437          {
1438          md->match_function_type = MATCH_CONDASSERT;
1439          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1440          if (rrc == MATCH_MATCH)
1441            {
1442            if (md->end_offset_top > offset_top)
1443              offset_top = md->end_offset_top;  /* Captures may have happened */
1444            condition = TRUE;
1445            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1446            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1447            }
1448    
1449          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1450          assertion; it is therefore treated as NOMATCH. */
1451    
1452          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1453            {
1454            RRETURN(rrc);         /* Need braces because of following else */
1455            }
1456          else
1457            {
1458            condition = FALSE;
1459            ecode += codelink;
1460            }
1461          }
1462    
1463        /* We are now at the branch that is to be obeyed. As there is only one, can
1464        use tail recursion to avoid using another stack frame, except when there is
1465        unlimited repeat of a possibly empty group. In the latter case, a recursive
1466        call to match() is always required, unless the second alternative doesn't
1467        exist, in which case we can just plough on. Note that, for compatibility
1468        with Perl, the | in a conditional group is NOT treated as creating two
1469        alternatives. If a THEN is encountered in the branch, it propagates out to
1470        the enclosing alternative (unless nested in a deeper set of alternatives,
1471        of course). */
1472    
1473        if (condition || *ecode == OP_ALT)
1474          {
1475          if (op != OP_SCOND)
1476            {
1477            ecode += 1 + LINK_SIZE;
1478            goto TAIL_RECURSE;
1479            }
1480    
1481          md->match_function_type = MATCH_CBEGROUP;
1482          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1483          RRETURN(rrc);
1484          }
1485    
1486         /* Condition false & no alternative; continue after the group. */
1487    
1488        else
1489          {
1490          ecode += 1 + LINK_SIZE;
1491          }
1492        break;
1493    
1494    
1495        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1496        to close any currently open capturing brackets. */
1497    
1498        case OP_CLOSE:
1499        number = GET2(ecode, 1);
1500        offset = number << 1;
1501    
1502    #ifdef PCRE_DEBUG
1503          printf("end bracket %d at *ACCEPT", number);
1504          printf("\n");
1505    #endif
1506    
1507        md->capture_last = number;
1508        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1509          {
1510          md->offset_vector[offset] =
1511            md->offset_vector[md->offset_end - number];
1512          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1513          if (offset_top <= offset) offset_top = offset + 2;
1514          }
1515        ecode += 1 + IMM2_SIZE;
1516        break;
1517    
1518    
1519        /* End of the pattern, either real or forced. */
1520    
1521        case OP_END:
1522        case OP_ACCEPT:
1523        case OP_ASSERT_ACCEPT:
1524    
1525        /* If we have matched an empty string, fail if not in an assertion and not
1526        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1527        is set and we have matched at the start of the subject. In both cases,
1528        backtracking will then try other alternatives, if any. */
1529    
1530        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1531             md->recursive == NULL &&
1532             (md->notempty ||
1533               (md->notempty_atstart &&
1534                 mstart == md->start_subject + md->start_offset)))
1535          RRETURN(MATCH_NOMATCH);
1536    
1537        /* Otherwise, we have a match. */
1538    
1539        md->end_match_ptr = eptr;           /* Record where we ended */
1540        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1541        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1542    
1543        /* For some reason, the macros don't work properly if an expression is
1544        given as the argument to RRETURN when the heap is in use. */
1545    
1546        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1547        RRETURN(rrc);
1548    
1549        /* Assertion brackets. Check the alternative branches in turn - the
1550        matching won't pass the KET for an assertion. If any one branch matches,
1551        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1552        start of each branch to move the current point backwards, so the code at
1553        this level is identical to the lookahead case. When the assertion is part
1554        of a condition, we want to return immediately afterwards. The caller of
1555        this incarnation of the match() function will have set MATCH_CONDASSERT in
1556        md->match_function type, and one of these opcodes will be the first opcode
1557        that is processed. We use a local variable that is preserved over calls to
1558        match() to remember this case. */
1559    
1560        case OP_ASSERT:
1561        case OP_ASSERTBACK:
1562        save_mark = md->mark;
1563        if (md->match_function_type == MATCH_CONDASSERT)
1564          {
1565          condassert = TRUE;
1566          md->match_function_type = 0;
1567          }
1568        else condassert = FALSE;
1569    
1570        do
1571          {
1572          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1573          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1574            {
1575            mstart = md->start_match_ptr;   /* In case \K reset it */
1576            break;
1577            }
1578    
1579          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1580          as NOMATCH. */
1581    
1582          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1583          ecode += GET(ecode, 1);
1584          md->mark = save_mark;
1585          }
1586        while (*ecode == OP_ALT);
1587    
1588        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1589    
1590        /* If checking an assertion for a condition, return MATCH_MATCH. */
1591    
1592        if (condassert) RRETURN(MATCH_MATCH);
1593    
1594        /* Continue from after the assertion, updating the offsets high water
1595        mark, since extracts may have been taken during the assertion. */
1596    
1597        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1598        ecode += 1 + LINK_SIZE;
1599        offset_top = md->end_offset_top;
1600        continue;
1601    
1602        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1603        PRUNE, or COMMIT means we must assume failure without checking subsequent
1604        branches. */
1605    
1606        case OP_ASSERT_NOT:
1607        case OP_ASSERTBACK_NOT:
1608        save_mark = md->mark;
1609        if (md->match_function_type == MATCH_CONDASSERT)
1610          {
1611          condassert = TRUE;
1612          md->match_function_type = 0;
1613          }
1614        else condassert = FALSE;
1615    
1616        do
1617          {
1618          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1619          md->mark = save_mark;
1620          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1621          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1622            {
1623            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1624            break;
1625            }
1626    
1627          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1628          as NOMATCH. */
1629    
1630          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1631          ecode += GET(ecode,1);
1632          }
1633        while (*ecode == OP_ALT);
1634    
1635        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1636    
1637        ecode += 1 + LINK_SIZE;
1638        continue;
1639    
1640        /* Move the subject pointer back. This occurs only at the start of
1641        each branch of a lookbehind assertion. If we are too close to the start to
1642        move back, this match function fails. When working with UTF-8 we move
1643        back a number of characters, not bytes. */
1644    
1645        case OP_REVERSE:
1646    #ifdef SUPPORT_UTF
1647        if (utf)
1648          {
1649          i = GET(ecode, 1);
1650          while (i-- > 0)
1651            {
1652            eptr--;
1653            if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1654            BACKCHAR(eptr);
1655            }
1656          }
1657        else
1658  #endif  #endif
1659    
1660      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1661    
1662        {        {
1663        eptr -= GET(ecode,1);        eptr -= GET(ecode, 1);
1664        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1665        }        }
1666    
1667      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1668    
1669        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1670      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1671      break;      break;
1672    
# Line 794  for (;;) Line 1675  for (;;)
1675      function is able to force a failure. */      function is able to force a failure. */
1676    
1677      case OP_CALLOUT:      case OP_CALLOUT:
1678      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1679        {        {
1680        pcre_callout_block cb;        PUBL(callout_block) cb;
1681        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1682        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1683        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1684        cb.subject          = (const char *)md->start_subject;  #ifdef COMPILE_PCRE8
1685        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1686        cb.start_match      = md->start_match - md->start_subject;  #else
1687        cb.current_position = eptr - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1688    #endif
1689          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1690          cb.start_match      = (int)(mstart - md->start_subject);
1691          cb.current_position = (int)(eptr - md->start_subject);
1692        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1693        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1694        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1695        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1696        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1697        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1698          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1699        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1700        }        }
1701      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 819  for (;;) Line 1705  for (;;)
1705      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
1706      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1707    
1708      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1709      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
1710      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
1711      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
1712      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
1713      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
1714      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.  
1715    
1716      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
1717      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
1718      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1719        a lot, so he is not to blame for the current way it works. */
1720    
1721      case OP_RECURSE:      case OP_RECURSE:
1722        {        {
1723        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1724        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1725    
1726        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1727        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1728            GET2(callpat, 1 + LINK_SIZE);
1729    
1730        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1731          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1732          caught at compile time.) */
1733    
1734          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1735            if (recno == ri->group_num && eptr == ri->subject_position)
1736              RRETURN(PCRE_ERROR_RECURSELOOP);
1737    
1738        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1739    
1740          new_recursive.group_num = recno;
1741          new_recursive.subject_position = eptr;
1742        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1743        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1744    
1745        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1746    
1747        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1748    
1749        /* Now save the offset data. */        /* Now save the offset data */
1750    
1751        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1752        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 863  for (;;) Line 1754  for (;;)
1754        else        else
1755          {          {
1756          new_recursive.offset_save =          new_recursive.offset_save =
1757            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1758          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1759          }          }
   
1760        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1761              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1762    
1763        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1764        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1765          might be changed, so reset it before looping. */
1766    
1767        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1768          cbegroup = (*callpat >= OP_SBRA);
1769        do        do
1770          {          {
1771          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1772              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1773          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1774            memcpy(md->offset_vector, new_recursive.offset_save,
1775                new_recursive.saved_max * sizeof(int));
1776            md->recursive = new_recursive.prevrec;
1777            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1778            {            {
1779            md->recursive = new_recursive.prevrec;            DPRINTF(("Recursion matched\n"));
1780            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1781              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1782            RRETURN(MATCH_MATCH);  
1783              /* Set where we got to in the subject, and reset the start in case
1784              it was changed by \K. This *is* propagated back out of a recursion,
1785              for Perl compatibility. */
1786    
1787              eptr = md->end_match_ptr;
1788              mstart = md->start_match_ptr;
1789              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1790              }
1791    
1792            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1793            as NOMATCH. */
1794    
1795            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1796              {
1797              DPRINTF(("Recursion gave error %d\n", rrc));
1798              if (new_recursive.offset_save != stacksave)
1799                (PUBL(free))(new_recursive.offset_save);
1800              RRETURN(rrc);
1801            }            }
         else if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1802    
1803          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1804          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1805          }          }
1806        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 899  for (;;) Line 1808  for (;;)
1808        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1809        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1810        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1811          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1812        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1813        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
       {  
       prev = ecode;  
       saved_eptr = eptr;  
   
       do  
         {  
         RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
           eptrb, match_isgroup);  
         if (rrc == MATCH_MATCH) break;  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         ecode += GET(ecode,1);  
         }  
       while (*ecode == OP_ALT);  
   
       /* If hit the end of the group (which could be repeated), fail */  
   
       if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
       /* Continue as from after the assertion, updating the offsets high water  
       mark, since extracts may have been taken. */  
   
       do ecode += GET(ecode,1); while (*ecode == OP_ALT);  
   
       offset_top = md->end_offset_top;  
       eptr = md->end_match_ptr;  
   
       /* For a non-repeating ket, just continue at this level. This also  
       happens for a repeating ket if no characters were matched in the group.  
       This is the forcible breaking of infinite loops as implemented in Perl  
       5.005. If there is an options reset, it will get obeyed in the normal  
       course of events. */  
   
       if (*ecode == OP_KET || eptr == saved_eptr)  
         {  
         ecode += 1+LINK_SIZE;  
         break;  
         }  
   
       /* The repeating kets try the rest of the pattern or restart from the  
       preceding bracket, in the appropriate order. 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));  
         }  
1814    
1815        if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1816          {      break;
         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);  
1817    
1818      /* 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
1819      bracketed group and go to there. */      bracketed group and go to there. */
# Line 985  for (;;) Line 1822  for (;;)
1822      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1823      break;      break;
1824    
1825      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1826      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
1827      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
1828      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
1829      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1830    
1831      case OP_BRAZERO:      case OP_BRAZERO:
1832        {      next = ecode + 1;
1833        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1834        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1835        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1836        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1837      break;      break;
1838    
1839      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1840        {      next = ecode + 1;
1841        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1842        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1843        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1844          match_isgroup);      ecode++;
1845        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1846        ecode++;  
1847        }      case OP_SKIPZERO:
1848        next = ecode+1;
1849        do next += GET(next,1); while (*next == OP_ALT);
1850        ecode = next + 1 + LINK_SIZE;
1851      break;      break;
1852    
1853      /* 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
1854      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1855      current high water mark for use by positive assertions. Do this also  
1856      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1857        op = *(++ecode);
1858        allow_zero = TRUE;
1859        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1860          goto POSSESSIVE_NON_CAPTURE;
1861    
1862        /* End of a group, repeated or non-repeating. */
1863    
1864      case OP_KET:      case OP_KET:
1865      case OP_KETRMIN:      case OP_KETRMIN:
1866      case OP_KETRMAX:      case OP_KETRMAX:
1867        {      case OP_KETRPOS:
1868        prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
       saved_eptr = eptrb->epb_saved_eptr;  
1869    
1870        /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1871        infinite repeats of empty string matches, retrieve the subject start from
1872        the chain. Otherwise, set it NULL. */
1873    
1874        eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1875          {
1876        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1877            *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1878            *prev == OP_ONCE)        }
1879          {      else saved_eptr = NULL;
         md->end_match_ptr = eptr;      /* For ONCE */  
         md->end_offset_top = offset_top;  
         RRETURN(MATCH_MATCH);  
         }  
1880    
1881        /* 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
1882        group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1883        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
1884        start in case it was changed by \K. */
1885    
1886        if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1887          {           *prev == OP_ONCE_NC)
1888          number = *prev - OP_BRA;        {
1889          md->end_match_ptr = eptr;      /* For ONCE_NC */
1890          md->end_offset_top = offset_top;
1891          md->start_match_ptr = mstart;
1892          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1893          }
1894    
1895          /* 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
1896          the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1897        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1898        into group 0, so it won't be picked up here. Instead, we catch it when the
1899        OP_END is reached. Other recursion is handled here. We just have to record
1900        the current subject position and start match pointer and give a MATCH
1901        return. */
1902    
1903          if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1904          offset = number << 1;          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1905          {
1906          number = GET2(prev, 1+LINK_SIZE);
1907          offset = number << 1;
1908    
1909  #ifdef DEBUG  #ifdef PCRE_DEBUG
1910          printf("end bracket %d", number);        printf("end bracket %d", number);
1911          printf("\n");        printf("\n");
1912  #endif  #endif
1913    
1914          /* 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. */  
1915    
1916          if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1917            {          {
1918            md->capture_last = number;          md->end_match_ptr = eptr;
1919            if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1920              {          RRETURN(MATCH_MATCH);
1921              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;  
             }  
1922    
1923            /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
           appropriately and continue from after the call. */  
1924    
1925            if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = number;
1926              {        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1927              recursion_info *rec = md->recursive;          {
1928              DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          /* If offset is greater than offset_top, it means that we are
1929              md->recursive = rec->prevrec;          "skipping" a capturing group, and that group's offsets must be marked
1930              md->start_match = rec->save_start;          unset. In earlier versions of PCRE, all the offsets were unset at the
1931              memcpy(md->offset_vector, rec->offset_save,          start of matching, but this doesn't work because atomic groups and
1932                rec->saved_max * sizeof(int));          assertions can cause a value to be set that should later be unset.
1933              ecode = rec->after_call;          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1934              ims = original_ims;          part of the atomic group, but this is not on the final matching path,
1935              break;          so must be unset when 2 is set. (If there is no group 2, there is no
1936              }          problem, because offset_top will then be 2, indicating no capture.) */
1937    
1938            if (offset > offset_top)
1939              {
1940              register int *iptr = md->offset_vector + offset_top;
1941              register int *iend = md->offset_vector + offset;
1942              while (iptr < iend) *iptr++ = -1;
1943            }            }
         }  
1944    
1945        /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
       the group. */  
1946    
1947        ims = original_ims;          md->offset_vector[offset] =
1948        DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1949            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1950            if (offset_top <= offset) offset_top = offset + 2;
1951            }
1952          }
1953    
1954        /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1955        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
1956        This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1957        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,
1958        course of events. */      establish a backup point by processing the rest of the pattern at a lower
1959        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1960        original OP_ONCE level, thereby bypassing intermediate backup points, but
1961        resetting any captures that happened along the way. */
1962    
1963        if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1964          {
1965          if (*prev == OP_ONCE)
1966          {          {
1967          ecode += 1 + LINK_SIZE;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1968          break;          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1969            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1970            RRETURN(MATCH_ONCE);
1971          }          }
1972          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1973          break;
1974          }
1975    
1976        /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1977        and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1978        at a time from the outer level, thus saving stack. */
1979    
1980        if (*ecode == OP_KETRPOS)
1981          {
1982          md->end_match_ptr = eptr;
1983          md->end_offset_top = offset_top;
1984          RRETURN(MATCH_KETRPOS);
1985          }
1986    
1987        /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
1988        preceding bracket, in the appropriate order. */      the preceding bracket, in the appropriate order. In the second case, we can
1989        use tail recursion to avoid using another stack frame, unless we have an
1990        an atomic group or an unlimited repeat of a group that can match an empty
1991        string. */
1992    
1993        if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1994          {
1995          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1996          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1997          if (*prev == OP_ONCE)
1998          {          {
1999          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);  
2000          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2001            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2002            RRETURN(MATCH_ONCE);
2003          }          }
2004        else  /* OP_KETRMAX */        if (*prev >= OP_SBRA)    /* Could match an empty string */
2005          {          {
2006          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          md->match_function_type = MATCH_CBEGROUP;
2007          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2008          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          RRETURN(rrc);
2009            }
2010          ecode = prev;
2011          goto TAIL_RECURSE;
2012          }
2013        else  /* OP_KETRMAX */
2014          {
2015          if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
2016          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2017          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2018          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2019          if (*prev == OP_ONCE)
2020            {
2021            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2022          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2023            md->once_target = prev;
2024            RRETURN(MATCH_ONCE);
2025          }          }
2026          ecode += 1 + LINK_SIZE;
2027          goto TAIL_RECURSE;
2028        }        }
2029        /* Control never gets here */
2030    
2031      RRETURN(MATCH_NOMATCH);      /* Not multiline mode: start of subject assertion, unless notbol. */
   
     /* Start of subject unless notbol, or after internal newline if multiline */  
2032    
2033      case OP_CIRC:      case OP_CIRC:
2034      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 */  
2035    
2036      /* Start of subject assertion */      /* Start of subject assertion */
2037    
# Line 1149  for (;;) Line 2040  for (;;)
2040      ecode++;      ecode++;
2041      break;      break;
2042    
2043        /* Multiline mode: start of subject unless notbol, or after any newline. */
2044    
2045        case OP_CIRCM:
2046        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2047        if (eptr != md->start_subject &&
2048            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2049          RRETURN(MATCH_NOMATCH);
2050        ecode++;
2051        break;
2052    
2053      /* Start of match assertion */      /* Start of match assertion */
2054    
2055      case OP_SOM:      case OP_SOM:
# Line 1156  for (;;) Line 2057  for (;;)
2057      ecode++;      ecode++;
2058      break;      break;
2059    
2060      /* 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. */  
2061    
2062      case OP_DOLL:      case OP_SET_SOM:
2063      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2064        ecode++;
2065        break;
2066    
2067        /* Multiline mode: assert before any newline, or before end of subject
2068        unless noteol is set. */
2069    
2070        case OP_DOLLM:
2071        if (eptr < md->end_subject)
2072        {        {
2073        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2074          { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }          {
2075        else          if (md->partial != 0 &&
2076          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2077        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2078        break;              NLBLOCK->nllen == 2 &&
2079                *eptr == NLBLOCK->nl[0])
2080              {
2081              md->hitend = TRUE;
2082              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2083              }
2084            RRETURN(MATCH_NOMATCH);
2085            }
2086        }        }
2087      else      else
2088        {        {
2089        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2090        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr < md->end_subject - 1 ||  
            (eptr == md->end_subject - 1 && *eptr != NEWLINE))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2091        }        }
2092      /* ... else fall through */      ecode++;
2093        break;
2094    
2095        /* Not multiline mode: assert before a terminating newline or before end of
2096        subject unless noteol is set. */
2097    
2098        case OP_DOLL:
2099        if (md->noteol) RRETURN(MATCH_NOMATCH);
2100        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2101    
2102        /* ... else fall through for endonly */
2103    
2104      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2105    
2106      case OP_EOD:      case OP_EOD:
2107      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2108        SCHECK_PARTIAL();
2109      ecode++;      ecode++;
2110      break;      break;
2111    
2112      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2113    
2114      case OP_EODN:      case OP_EODN:
2115      if (eptr < md->end_subject - 1 ||      ASSERT_NL_OR_EOS:
2116         (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject &&
2117            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2118          {
2119          if (md->partial != 0 &&
2120              eptr + 1 >= md->end_subject &&
2121              NLBLOCK->nltype == NLTYPE_FIXED &&
2122              NLBLOCK->nllen == 2 &&
2123              *eptr == NLBLOCK->nl[0])
2124            {
2125            md->hitend = TRUE;
2126            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2127            }
2128          RRETURN(MATCH_NOMATCH);
2129          }
2130    
2131        /* Either at end of string or \n before end. */
2132    
2133        SCHECK_PARTIAL();
2134      ecode++;      ecode++;
2135      break;      break;
2136    
# Line 1206  for (;;) Line 2142  for (;;)
2142    
2143        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2144        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
2145        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2146          partial matching. */
2147    
2148  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2149        if (utf8)        if (utf)
2150          {          {
2151            /* Get status of previous character */
2152    
2153          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2154            {            {
2155            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2156            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2157              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2158            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2159    #ifdef SUPPORT_UCP
2160              if (md->use_ucp)
2161                {
2162                if (c == '_') prev_is_word = TRUE; else
2163                  {
2164                  int cat = UCD_CATEGORY(c);
2165                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2166                  }
2167                }
2168              else
2169    #endif
2170            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2171            }            }
2172          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2173            /* Get status of next character */
2174    
2175            if (eptr >= md->end_subject)
2176              {
2177              SCHECK_PARTIAL();
2178              cur_is_word = FALSE;
2179              }
2180            else
2181            {            {
2182            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2183    #ifdef SUPPORT_UCP
2184              if (md->use_ucp)
2185                {
2186                if (c == '_') cur_is_word = TRUE; else
2187                  {
2188                  int cat = UCD_CATEGORY(c);
2189                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2190                  }
2191                }
2192              else
2193    #endif
2194            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2195            }            }
2196          }          }
2197        else        else
2198  #endif  #endif
2199    
2200        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2201          consistency with the behaviour of \w we do use it in this case. */
2202    
2203          {          {
2204          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2205            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2206          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2207            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2208              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2209    #ifdef SUPPORT_UCP
2210              if (md->use_ucp)
2211                {
2212                c = eptr[-1];
2213                if (c == '_') prev_is_word = TRUE; else
2214                  {
2215                  int cat = UCD_CATEGORY(c);
2216                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2217                  }
2218                }
2219              else
2220    #endif
2221              prev_is_word = MAX_255(eptr[-1])
2222                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2223              }
2224    
2225            /* Get status of next character */
2226    
2227            if (eptr >= md->end_subject)
2228              {
2229              SCHECK_PARTIAL();
2230              cur_is_word = FALSE;
2231              }
2232            else
2233    #ifdef SUPPORT_UCP
2234            if (md->use_ucp)
2235              {
2236              c = *eptr;
2237              if (c == '_') cur_is_word = TRUE; else
2238                {
2239                int cat = UCD_CATEGORY(c);
2240                cur_is_word = (cat == ucp_L || cat == ucp_N);
2241                }
2242              }
2243            else
2244    #endif
2245            cur_is_word = MAX_255(*eptr)
2246              && ((md->ctypes[*eptr] & ctype_word) != 0);
2247          }          }
2248    
2249        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1242  for (;;) Line 2252  for (;;)
2252             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2253          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2254        }        }
2255      break;      break;
2256    
2257        /* Match any single character type except newline; have to take care with
2258        CRLF newlines and partial matching. */
2259    
2260        case OP_ANY:
2261        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2262        if (md->partial != 0 &&
2263            eptr + 1 >= md->end_subject &&
2264            NLBLOCK->nltype == NLTYPE_FIXED &&
2265            NLBLOCK->nllen == 2 &&
2266            *eptr == NLBLOCK->nl[0])
2267          {
2268          md->hitend = TRUE;
2269          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2270          }
2271    
2272      /* Match a single character type; inline for speed */      /* Fall through */
2273    
2274      case OP_ANY:      /* Match any single character whatsoever. */
2275      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)  
2276        case OP_ALLANY:
2277        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2278          {                            /* not be updated before SCHECK_PARTIAL. */
2279          SCHECK_PARTIAL();
2280        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2281      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);        }
2282  #ifdef SUPPORT_UTF8      eptr++;
2283      if (utf8)  #ifdef SUPPORT_UTF
2284        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2285  #endif  #endif
2286      ecode++;      ecode++;
2287      break;      break;
# Line 1261  for (;;) Line 2290  for (;;)
2290      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2291    
2292      case OP_ANYBYTE:      case OP_ANYBYTE:
2293      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2294          {                            /* not be updated before SCHECK_PARTIAL. */
2295          SCHECK_PARTIAL();
2296          RRETURN(MATCH_NOMATCH);
2297          }
2298        eptr++;
2299      ecode++;      ecode++;
2300      break;      break;
2301    
2302      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2303      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2304          {
2305          SCHECK_PARTIAL();
2306          RRETURN(MATCH_NOMATCH);
2307          }
2308      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2309      if (      if (
2310  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2311         c < 256 &&         c < 256 &&
2312  #endif  #endif
2313         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1279  for (;;) Line 2317  for (;;)
2317      break;      break;
2318    
2319      case OP_DIGIT:      case OP_DIGIT:
2320      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2321          {
2322          SCHECK_PARTIAL();
2323          RRETURN(MATCH_NOMATCH);
2324          }
2325      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2326      if (      if (
2327  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2328         c >= 256 ||         c > 255 ||
2329  #endif  #endif
2330         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2331         )         )
# Line 1292  for (;;) Line 2334  for (;;)
2334      break;      break;
2335    
2336      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2337      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2338          {
2339          SCHECK_PARTIAL();
2340          RRETURN(MATCH_NOMATCH);
2341          }
2342      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2343      if (      if (
2344  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2345         c < 256 &&         c < 256 &&
2346  #endif  #endif
2347         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1305  for (;;) Line 2351  for (;;)
2351      break;      break;
2352    
2353      case OP_WHITESPACE:      case OP_WHITESPACE:
2354      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2355          {
2356          SCHECK_PARTIAL();
2357          RRETURN(MATCH_NOMATCH);
2358          }
2359      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2360      if (      if (
2361  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2362         c >= 256 ||         c > 255 ||
2363  #endif  #endif
2364         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2365         )         )
# Line 1318  for (;;) Line 2368  for (;;)
2368      break;      break;
2369    
2370      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2371      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2372          {
2373          SCHECK_PARTIAL();
2374          RRETURN(MATCH_NOMATCH);
2375          }
2376      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2377      if (      if (
2378  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2379         c < 256 &&         c < 256 &&
2380  #endif  #endif
2381         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1331  for (;;) Line 2385  for (;;)
2385      break;      break;
2386    
2387      case OP_WORDCHAR:      case OP_WORDCHAR:
2388      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2389          {
2390          SCHECK_PARTIAL();
2391          RRETURN(MATCH_NOMATCH);
2392          }
2393      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2394      if (      if (
2395  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2396         c >= 256 ||         c > 255 ||
2397  #endif  #endif
2398         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2399         )         )
# Line 1343  for (;;) Line 2401  for (;;)
2401      ecode++;      ecode++;
2402      break;      break;
2403    
2404        case OP_ANYNL:
2405        if (eptr >= md->end_subject)
2406          {
2407          SCHECK_PARTIAL();
2408          RRETURN(MATCH_NOMATCH);
2409          }
2410        GETCHARINCTEST(c, eptr);
2411        switch(c)
2412          {
2413          default: RRETURN(MATCH_NOMATCH);
2414    
2415          case 0x000d:
2416          if (eptr >= md->end_subject)
2417            {
2418            SCHECK_PARTIAL();
2419            }
2420          else if (*eptr == 0x0a) eptr++;
2421          break;
2422    
2423          case 0x000a:
2424          break;
2425    
2426          case 0x000b:
2427          case 0x000c:
2428          case 0x0085:
2429          case 0x2028:
2430          case 0x2029:
2431          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2432          break;
2433          }
2434        ecode++;
2435        break;
2436    
2437        case OP_NOT_HSPACE:
2438        if (eptr >= md->end_subject)
2439          {
2440          SCHECK_PARTIAL();
2441          RRETURN(MATCH_NOMATCH);
2442          }
2443        GETCHARINCTEST(c, eptr);
2444        switch(c)
2445          {
2446          default: break;
2447          case 0x09:      /* HT */
2448          case 0x20:      /* SPACE */
2449          case 0xa0:      /* NBSP */
2450          case 0x1680:    /* OGHAM SPACE MARK */
2451          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2452          case 0x2000:    /* EN QUAD */
2453          case 0x2001:    /* EM QUAD */
2454          case 0x2002:    /* EN SPACE */
2455          case 0x2003:    /* EM SPACE */
2456          case 0x2004:    /* THREE-PER-EM SPACE */
2457          case 0x2005:    /* FOUR-PER-EM SPACE */
2458          case 0x2006:    /* SIX-PER-EM SPACE */
2459          case 0x2007:    /* FIGURE SPACE */
2460          case 0x2008:    /* PUNCTUATION SPACE */
2461          case 0x2009:    /* THIN SPACE */
2462          case 0x200A:    /* HAIR SPACE */
2463          case 0x202f:    /* NARROW NO-BREAK SPACE */
2464          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2465          case 0x3000:    /* IDEOGRAPHIC SPACE */
2466          RRETURN(MATCH_NOMATCH);
2467          }
2468        ecode++;
2469        break;
2470    
2471        case OP_HSPACE:
2472        if (eptr >= md->end_subject)
2473          {
2474          SCHECK_PARTIAL();
2475          RRETURN(MATCH_NOMATCH);
2476          }
2477        GETCHARINCTEST(c, eptr);
2478        switch(c)
2479          {
2480          default: RRETURN(MATCH_NOMATCH);
2481          case 0x09:      /* HT */
2482          case 0x20:      /* SPACE */
2483          case 0xa0:      /* NBSP */
2484          case 0x1680:    /* OGHAM SPACE MARK */
2485          case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */
2486          case 0x2000:    /* EN QUAD */
2487          case 0x2001:    /* EM QUAD */
2488          case 0x2002:    /* EN SPACE */
2489          case 0x2003:    /* EM SPACE */
2490          case 0x2004:    /* THREE-PER-EM SPACE */
2491          case 0x2005:    /* FOUR-PER-EM SPACE */
2492          case 0x2006:    /* SIX-PER-EM SPACE */
2493          case 0x2007:    /* FIGURE SPACE */
2494          case 0x2008:    /* PUNCTUATION SPACE */
2495          case 0x2009:    /* THIN SPACE */
2496          case 0x200A:    /* HAIR SPACE */
2497          case 0x202f:    /* NARROW NO-BREAK SPACE */
2498          case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2499          case 0x3000:    /* IDEOGRAPHIC SPACE */
2500          break;
2501          }
2502        ecode++;
2503        break;
2504    
2505        case OP_NOT_VSPACE:
2506        if (eptr >= md->end_subject)
2507          {
2508          SCHECK_PARTIAL();
2509          RRETURN(MATCH_NOMATCH);
2510          }
2511        GETCHARINCTEST(c, eptr);
2512        switch(c)
2513          {
2514          default: break;
2515          case 0x0a:      /* LF */
2516          case 0x0b:      /* VT */
2517          case 0x0c:      /* FF */
2518          case 0x0d:      /* CR */
2519          case 0x85:      /* NEL */
2520          case 0x2028:    /* LINE SEPARATOR */
2521          case 0x2029:    /* PARAGRAPH SEPARATOR */
2522          RRETURN(MATCH_NOMATCH);
2523          }
2524        ecode++;
2525        break;
2526    
2527        case OP_VSPACE:
2528        if (eptr >= md->end_subject)
2529          {
2530          SCHECK_PARTIAL();
2531          RRETURN(MATCH_NOMATCH);
2532          }
2533        GETCHARINCTEST(c, eptr);
2534        switch(c)
2535          {
2536          default: RRETURN(MATCH_NOMATCH);
2537          case 0x0a:      /* LF */
2538          case 0x0b:      /* VT */
2539          case 0x0c:      /* FF */
2540          case 0x0d:      /* CR */
2541          case 0x85:      /* NEL */
2542          case 0x2028:    /* LINE SEPARATOR */
2543          case 0x2029:    /* PARAGRAPH SEPARATOR */
2544          break;
2545          }
2546        ecode++;
2547        break;
2548    
2549  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2550      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2551      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. */
2552    
2553      case OP_PROP:      case OP_PROP:
2554      case OP_NOTPROP:      case OP_NOTPROP:
2555      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2556          {
2557          SCHECK_PARTIAL();
2558          RRETURN(MATCH_NOMATCH);
2559          }
2560      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2561        {        {
2562        int chartype, rqdtype;        const ucd_record *prop = GET_UCD(c);
       int othercase;  
       int category = ucp_findchar(c, &chartype, &othercase);  
2563    
2564        rqdtype = *(++ecode);        switch(ecode[1])
       ecode++;  
   
       if (rqdtype >= 128)  
2565          {          {
2566          if ((rqdtype - 128 != category) == (op == OP_PROP))          case PT_ANY:
2567            if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2568            break;
2569    
2570            case PT_LAMP:
2571            if ((prop->chartype == ucp_Lu ||
2572                 prop->chartype == ucp_Ll ||
2573                 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2574            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2575          }          break;
2576        else  
2577          {          case PT_GC:
2578          if ((rqdtype != chartype) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2579              RRETURN(MATCH_NOMATCH);
2580            break;
2581    
2582            case PT_PC:
2583            if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2584            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2585            break;
2586    
2587            case PT_SC:
2588            if ((ecode[2] != prop->script) == (op == OP_PROP))
2589              RRETURN(MATCH_NOMATCH);
2590            break;
2591    
2592            /* These are specials */
2593    
2594            case PT_ALNUM:
2595            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2596                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2597              RRETURN(MATCH_NOMATCH);
2598            break;
2599    
2600            case PT_SPACE:    /* Perl space */
2601            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2602                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2603                   == (op == OP_NOTPROP))
2604              RRETURN(MATCH_NOMATCH);
2605            break;
2606    
2607            case PT_PXSPACE:  /* POSIX space */
2608            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2609                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2610                 c == CHAR_FF || c == CHAR_CR)
2611                   == (op == OP_NOTPROP))
2612              RRETURN(MATCH_NOMATCH);
2613            break;
2614    
2615            case PT_WORD:
2616            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2617                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2618                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2619              RRETURN(MATCH_NOMATCH);
2620            break;
2621    
2622            /* This should never occur */
2623    
2624            default:
2625            RRETURN(PCRE_ERROR_INTERNAL);
2626          }          }
2627    
2628          ecode += 3;
2629        }        }
2630      break;      break;
2631    
# Line 1376  for (;;) Line 2633  for (;;)
2633      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2634    
2635      case OP_EXTUNI:      case OP_EXTUNI:
2636      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2637          {
2638          SCHECK_PARTIAL();
2639          RRETURN(MATCH_NOMATCH);
2640          }
2641      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2642        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2643        while (eptr < md->end_subject)
2644        {        {
2645        int chartype;        int len = 1;
2646        int othercase;        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2647        int category = ucp_findchar(c, &chartype, &othercase);        if (UCD_CATEGORY(c) != ucp_M) break;
2648        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 = ucp_findchar(c, &chartype, &othercase);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2649        }        }
2650        CHECK_PARTIAL();
2651      ecode++;      ecode++;
2652      break;      break;
2653  #endif  #endif
# Line 1409  for (;;) Line 2662  for (;;)
2662      loops). */      loops). */
2663    
2664      case OP_REF:      case OP_REF:
2665        {      case OP_REFI:
2666        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2667        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2668        ecode += 1 + IMM2_SIZE;
2669    
2670        /* 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];  
2671    
2672        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2673        this ensures that every attempt at a match fails. We can't just fail
2674        here, because of the possibility of quantifiers with zero minima.
2675    
2676        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2677          {      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;  
2678    
2679          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2680          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;  
2681    
2682          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2683          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2684          eptr += length;      else
2685          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2686    
2687        /* Set up for repetition, or handle the non-repeated case */
2688    
2689        switch (*ecode)
2690          {
2691          case OP_CRSTAR:
2692          case OP_CRMINSTAR:
2693          case OP_CRPLUS:
2694          case OP_CRMINPLUS:
2695          case OP_CRQUERY:
2696          case OP_CRMINQUERY:
2697          c = *ecode++ - OP_CRSTAR;
2698          minimize = (c & 1) != 0;
2699          min = rep_min[c];                 /* Pick up values from tables; */
2700          max = rep_max[c];                 /* zero for max => infinity */
2701          if (max == 0) max = INT_MAX;
2702          break;
2703    
2704          case OP_CRRANGE:
2705          case OP_CRMINRANGE:
2706          minimize = (*ecode == OP_CRMINRANGE);
2707          min = GET2(ecode, 1);
2708          max = GET2(ecode, 1 + IMM2_SIZE);
2709          if (max == 0) max = INT_MAX;
2710          ecode += 1 + 2 * IMM2_SIZE;
2711          break;
2712    
2713          default:               /* No repeat follows */
2714          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2715            {
2716            if (length == -2) eptr = md->end_subject;   /* Partial match */
2717            CHECK_PARTIAL();
2718            RRETURN(MATCH_NOMATCH);
2719          }          }
2720          eptr += length;
2721          continue;              /* With the main loop */
2722          }
2723    
2724        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2725        main loop. */      zero, just continue with the main loop. If the length is negative, it
2726        means the reference is unset in non-Java-compatible mode. If the minimum is
2727        zero, we can continue at the same level without recursion. For any other
2728        minimum, carrying on will result in NOMATCH. */
2729    
2730        if (length == 0) continue;      if (length == 0) continue;
2731        if (length < 0 && min == 0) continue;
2732    
2733        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2734        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2735        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2736    
2737        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2738          {
2739          int slength;
2740          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2741          {          {
2742          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2743          eptr += length;          CHECK_PARTIAL();
2744            RRETURN(MATCH_NOMATCH);
2745          }          }
2746          eptr += slength;
2747          }
2748    
2749        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2750        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2751    
2752        if (min == max) continue;      if (min == max) continue;
2753    
2754        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2755    
2756        if (minimize)      if (minimize)
2757          {
2758          for (fi = min;; fi++)
2759          {          {
2760          for (fi = min;; fi++)          int slength;
2761            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2762            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2763            if (fi >= max) RRETURN(MATCH_NOMATCH);
2764            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2765            {            {
2766            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2767            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2768            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2769            }            }
2770          /* Control never gets here */          eptr += slength;
2771          }          }
2772          /* Control never gets here */
2773          }
2774    
2775        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2776    
2777        else      else
2778          {
2779          pp = eptr;
2780          for (i = min; i < max; i++)
2781          {          {
2782          pp = eptr;          int slength;
2783          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
2784            {            {
2785            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2786            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2787            eptr -= length;  
2788              if (slength == -2 && md->partial != 0 &&
2789                  md->end_subject > md->start_used_ptr)
2790                {
2791                md->hitend = TRUE;
2792                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2793                }
2794              break;
2795            }            }
2796          RRETURN(MATCH_NOMATCH);          eptr += slength;
2797            }
2798    
2799          while (eptr >= pp)
2800            {
2801            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2802            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2803            eptr -= length;
2804          }          }
2805          RRETURN(MATCH_NOMATCH);
2806        }        }
2807      /* Control never gets here */      /* Control never gets here */
2808    
   
   
2809      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2810      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,
2811      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 2820  for (;;)
2820      case OP_NCLASS:      case OP_NCLASS:
2821      case OP_CLASS:      case OP_CLASS:
2822        {        {
2823          /* The data variable is saved across frames, so the byte map needs to
2824          be stored there. */
2825    #define BYTE_MAP ((pcre_uint8 *)data)
2826        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2827        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2828    
2829        switch (*ecode)        switch (*ecode)
2830          {          {
# Line 1548  for (;;) Line 2845  for (;;)
2845          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2846          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2847          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2848          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2849          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2850          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2851          break;          break;
2852    
2853          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1560  for (;;) Line 2857  for (;;)
2857    
2858        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2859    
2860  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2861        /* UTF-8 mode */        if (utf)
       if (utf8)  
2862          {          {
2863          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2864            {            {
2865            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2866                {
2867                SCHECK_PARTIAL();
2868                RRETURN(MATCH_NOMATCH);
2869                }
2870            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2871            if (c > 255)            if (c > 255)
2872              {              {
2873              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2874              }              }
2875            else            else
2876              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2877            }            }
2878          }          }
2879        else        else
2880  #endif  #endif
2881        /* Not UTF-8 mode */        /* Not UTF mode */
2882          {          {
2883          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2884            {            {
2885            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2886                {
2887                SCHECK_PARTIAL();
2888                RRETURN(MATCH_NOMATCH);
2889                }
2890            c = *eptr++;            c = *eptr++;
2891            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2892              if (c > 255)
2893                {
2894                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2895                }
2896              else
2897    #endif
2898                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2899            }            }
2900          }          }
2901    
# Line 1600  for (;;) Line 2909  for (;;)
2909    
2910        if (minimize)        if (minimize)
2911          {          {
2912  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2913          /* UTF-8 mode */          if (utf)
         if (utf8)  
2914            {            {
2915            for (fi = min;; fi++)            for (fi = min;; fi++)
2916              {              {
2917              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2918              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2919              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2920                if (eptr >= md->end_subject)
2921                  {
2922                  SCHECK_PARTIAL();
2923                  RRETURN(MATCH_NOMATCH);
2924                  }
2925              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2926              if (c > 255)              if (c > 255)
2927                {                {
2928                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2929                }                }
2930              else              else
2931                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2932              }              }
2933            }            }
2934          else          else
2935  #endif  #endif
2936          /* Not UTF-8 mode */          /* Not UTF mode */
2937            {            {
2938            for (fi = min;; fi++)            for (fi = min;; fi++)
2939              {              {
2940              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2941              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2942              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2943                if (eptr >= md->end_subject)
2944                  {
2945                  SCHECK_PARTIAL();
2946                  RRETURN(MATCH_NOMATCH);
2947                  }
2948              c = *eptr++;              c = *eptr++;
2949              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2950                if (c > 255)
2951                  {
2952                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2953                  }
2954                else
2955    #endif
2956                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2957              }              }
2958            }            }
2959          /* Control never gets here */          /* Control never gets here */
# Line 1642  for (;;) Line 2965  for (;;)
2965          {          {
2966          pp = eptr;          pp = eptr;
2967    
2968  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2969          /* UTF-8 mode */          if (utf)
         if (utf8)  
2970            {            {
2971            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2972              {              {
2973              int len = 1;              int len = 1;
2974              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2975                  {
2976                  SCHECK_PARTIAL();
2977                  break;
2978                  }
2979              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2980              if (c > 255)              if (c > 255)
2981                {                {
2982                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2983                }                }
2984              else              else
2985                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2986              eptr += len;              eptr += len;
2987              }              }
2988            for (;;)            for (;;)
2989              {              {
2990              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2991              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2992              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2993              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1671  for (;;) Line 2995  for (;;)
2995            }            }
2996          else          else
2997  #endif  #endif
2998            /* Not UTF-8 mode */            /* Not UTF mode */
2999            {            {
3000            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3001              {              {
3002              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3003                  {
3004                  SCHECK_PARTIAL();
3005                  break;
3006                  }
3007              c = *eptr;              c = *eptr;
3008              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3009                if (c > 255)
3010                  {
3011                  if (op == OP_CLASS) break;
3012                  }
3013                else
3014    #endif
3015                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3016              eptr++;              eptr++;
3017              }              }
3018            while (eptr >= pp)            while (eptr >= pp)
3019              {              {
3020              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
             eptr--;  
3021              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3022                eptr--;
3023              }              }
3024            }            }
3025    
3026          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3027          }          }
3028    #undef BYTE_MAP
3029        }        }
3030      /* Control never gets here */      /* Control never gets here */
3031    
3032    
3033      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3034      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
3035        mode, because Unicode properties are supported in non-UTF-8 mode. */
3036    
3037  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3038      case OP_XCLASS:      case OP_XCLASS:
3039        {        {
3040        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1722  for (;;) Line 3059  for (;;)
3059          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3060          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3061          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3062          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3063          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3064          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3065          break;          break;
3066    
3067          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1736  for (;;) Line 3073  for (;;)
3073    
3074        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3075          {          {
3076          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3077          GETCHARINC(c, eptr);            {
3078          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3079              RRETURN(MATCH_NOMATCH);
3080              }
3081            GETCHARINCTEST(c, eptr);
3082            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3083          }          }
3084    
3085        /* 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 3094  for (;;)
3094          {          {
3095          for (fi = min;; fi++)          for (fi = min;; fi++)
3096            {            {
3097            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3098            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3099            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3100            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3101            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3102                SCHECK_PARTIAL();
3103                RRETURN(MATCH_NOMATCH);
3104                }
3105              GETCHARINCTEST(c, eptr);
3106              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3107            }            }
3108          /* Control never gets here */          /* Control never gets here */
3109          }          }
# Line 1770  for (;;) Line 3116  for (;;)
3116          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3117            {            {
3118            int len = 1;            int len = 1;
3119            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3120            GETCHARLEN(c, eptr, len);              {
3121            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3122                break;
3123                }
3124    #ifdef SUPPORT_UTF
3125              GETCHARLENTEST(c, eptr, len);
3126    #else
3127              c = *eptr;
3128    #endif
3129              if (!PRIV(xclass)(c, data, utf)) break;
3130            eptr += len;            eptr += len;
3131            }            }
3132          for(;;)          for(;;)
3133            {            {
3134            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3135            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3136            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3137            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3138              if (utf) BACKCHAR(eptr);
3139    #endif
3140            }            }
3141          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3142          }          }
# Line 1792  for (;;) Line 3148  for (;;)
3148      /* Match a single character, casefully */      /* Match a single character, casefully */
3149    
3150      case OP_CHAR:      case OP_CHAR:
3151  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3152      if (utf8)      if (utf)
3153        {        {
3154        length = 1;        length = 1;
3155        ecode++;        ecode++;
3156        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3157        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3158            {
3159            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3160            RRETURN(MATCH_NOMATCH);
3161            }
3162        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3163        }        }
3164      else      else
3165  #endif  #endif
3166        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3167        {        {
3168        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3169            {
3170            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3171            RRETURN(MATCH_NOMATCH);
3172            }
3173        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3174        ecode += 2;        ecode += 2;
3175        }        }
3176      break;      break;
3177    
3178      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3179        subject, give up immediately. */
3180    
3181        case OP_CHARI:
3182        if (eptr >= md->end_subject)
3183          {
3184          SCHECK_PARTIAL();
3185          RRETURN(MATCH_NOMATCH);
3186          }
3187    
3188      case OP_CHARNC:  #ifdef SUPPORT_UTF
3189  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3190        {        {
3191        length = 1;        length = 1;
3192        ecode++;        ecode++;
3193        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3194    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3195        /* 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
3196        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3197          fast lookup table. We know that there is at least one byte left in the
3198          subject. */
3199    
3200        if (fc < 128)        if (fc < 128)
3201          {          {
3202          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3203                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3204            ecode++;
3205            eptr++;
3206          }          }
3207    
3208        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3209          use the value of "length" to check for sufficient bytes left, because the
3210          other case of the character may have more or fewer bytes.  */
3211    
3212        else        else
3213          {          {
3214          int dc;          unsigned int dc;
3215          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3216          ecode += length;          ecode += length;
3217    
3218          /* 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
3219          case of the character, if there is one. The result of 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. */  
3220    
3221          if (fc != dc)          if (fc != dc)
3222            {            {
3223  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3224            int chartype;            if (dc != UCD_OTHERCASE(fc))
           int othercase;  
           if (ucp_findchar(fc, &chartype, &othercase) < 0 || dc != othercase)  
3225  #endif  #endif
3226              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3227            }            }
3228          }          }
3229        }        }
3230      else      else
3231  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3232    
3233      /* Non-UTF-8 mode */      /* Not UTF mode */
3234        {        {
3235        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3236        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3237          eptr++;
3238        ecode += 2;        ecode += 2;
3239        }        }
3240      break;      break;
3241    
3242      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3243    
3244      case OP_EXACT:      case OP_EXACT:
3245        case OP_EXACTI:
3246      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3247      ecode += 3;      ecode += 1 + IMM2_SIZE;
3248      goto REPEATCHAR;      goto REPEATCHAR;
3249    
3250        case OP_POSUPTO:
3251        case OP_POSUPTOI:
3252        possessive = TRUE;
3253        /* Fall through */
3254    
3255      case OP_UPTO:      case OP_UPTO:
3256        case OP_UPTOI:
3257      case OP_MINUPTO:      case OP_MINUPTO:
3258        case OP_MINUPTOI:
3259      min = 0;      min = 0;
3260      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3261      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3262      ecode += 3;      ecode += 1 + IMM2_SIZE;
3263        goto REPEATCHAR;
3264    
3265        case OP_POSSTAR:
3266        case OP_POSSTARI:
3267        possessive = TRUE;
3268        min = 0;
3269        max = INT_MAX;
3270        ecode++;
3271        goto REPEATCHAR;
3272    
3273        case OP_POSPLUS:
3274        case OP_POSPLUSI:
3275        possessive = TRUE;
3276        min = 1;
3277        max = INT_MAX;
3278        ecode++;
3279        goto REPEATCHAR;
3280    
3281        case OP_POSQUERY:
3282        case OP_POSQUERYI:
3283        possessive = TRUE;
3284        min = 0;
3285        max = 1;
3286        ecode++;
3287      goto REPEATCHAR;      goto REPEATCHAR;
3288    
3289      case OP_STAR:      case OP_STAR:
3290        case OP_STARI:
3291      case OP_MINSTAR:      case OP_MINSTAR:
3292        case OP_MINSTARI:
3293      case OP_PLUS:      case OP_PLUS:
3294        case OP_PLUSI:
3295      case OP_MINPLUS:      case OP_MINPLUS:
3296        case OP_MINPLUSI:
3297      case OP_QUERY:      case OP_QUERY:
3298        case OP_QUERYI:
3299      case OP_MINQUERY:      case OP_MINQUERY:
3300      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3301        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3302      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3303      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3304      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3305      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3306    
3307      /* 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. */  
3308    
3309      REPEATCHAR:      REPEATCHAR:
3310  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3311      if (utf8)      if (utf)
3312        {        {
3313        length = 1;        length = 1;
3314        charptr = ecode;        charptr = ecode;
3315        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3316        ecode += length;        ecode += length;
3317    
3318        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1913  for (;;) Line 3320  for (;;)
3320    
3321        if (length > 1)        if (length > 1)
3322          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3323  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3324          int othercase;          unsigned int othercase;
3325          int chartype;          if (op >= OP_STARI &&     /* Caseless */
3326          if ((ims & PCRE_CASELESS) != 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
3327               ucp_findchar(fc, &chartype, &othercase) >= 0 &&            oclength = PRIV(ord2utf)(othercase, occhars);
3328               othercase > 0)          else oclength = 0;
           oclength = _pcre_ord2utf8(othercase, occhars);  
3329  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3330    
3331          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3332            {            {
3333            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3334            /* Need braces because of following else */              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3335            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3336              else if (oclength > 0 &&
3337                       eptr <= md->end_subject - oclength &&
3338                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3339    #endif  /* SUPPORT_UCP */
3340            else            else
3341              {              {
3342              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3343              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3344              }              }
3345            }            }
3346    
# Line 1943  for (;;) Line 3350  for (;;)
3350            {            {
3351            for (fi = min;; fi++)            for (fi = min;; fi++)
3352              {              {
3353              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3354              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3355              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3356              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3357              /* Need braces because of following else */                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3358              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3359                else if (oclength > 0 &&
3360                         eptr <= md->end_subject - oclength &&
3361                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3362    #endif  /* SUPPORT_UCP */
3363              else              else
3364                {                {
3365                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3366                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3367                }                }
3368              }              }
3369            /* Control never gets here */            /* Control never gets here */
3370            }            }
3371          else  
3372            else  /* Maximize */
3373            {            {
3374            pp = eptr;            pp = eptr;
3375            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3376              {              {
3377              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3378              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3379              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3380                else if (oclength > 0 &&
3381                         eptr <= md->end_subject - oclength &&
3382                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3383    #endif  /* SUPPORT_UCP */
3384              else              else
3385                {                {
3386                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3387                eptr += oclength;                break;
3388                }                }
3389              }              }
3390            while (eptr >= pp)  
3391             {            if (possessive) continue;
3392             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3393             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3394             eptr -= length;              {
3395             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3396            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3397                if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3398    #ifdef SUPPORT_UCP
3399                eptr--;
3400                BACKCHAR(eptr);
3401    #else   /* without SUPPORT_UCP */
3402                eptr -= length;
3403    #endif  /* SUPPORT_UCP */
3404                }
3405            }            }
3406          /* Control never gets here */          /* Control never gets here */
3407          }          }
# Line 1987  for (;;) Line 3411  for (;;)
3411        value of fc will always be < 128. */        value of fc will always be < 128. */
3412        }        }
3413      else      else
3414  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3415          /* 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);  
3416        fc = *ecode++;        fc = *ecode++;
       }  
3417    
3418      /* 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
3419      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
3420      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3421      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3422      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2007  for (;;) Line 3427  for (;;)
3427      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3428        max, eptr));        max, eptr));
3429    
3430      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3431        {        {
3432        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3433          /* fc must be < 128 if UTF is enabled. */
3434          foc = md->fcc[fc];
3435    #else
3436    #ifdef SUPPORT_UTF
3437    #ifdef SUPPORT_UCP
3438          if (utf && fc > 127)
3439            foc = UCD_OTHERCASE(fc);
3440    #else
3441          if (utf && fc > 127)
3442            foc = fc;
3443    #endif /* SUPPORT_UCP */
3444          else
3445    #endif /* SUPPORT_UTF */
3446            foc = TABLE_GET(fc, md->fcc, fc);
3447    #endif /* COMPILE_PCRE8 */
3448    
3449        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3450          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3451            if (eptr >= md->end_subject)
3452              {
3453              SCHECK_PARTIAL();
3454              RRETURN(MATCH_NOMATCH);
3455              }
3456            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3457            eptr++;
3458            }
3459        if (min == max) continue;        if (min == max) continue;
3460        if (minimize)        if (minimize)
3461          {          {
3462          for (fi = min;; fi++)          for (fi = min;; fi++)
3463            {            {
3464            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3465            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3466            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3467                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3468                {
3469                SCHECK_PARTIAL();
3470              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3471                }
3472              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3473              eptr++;
3474            }            }
3475          /* Control never gets here */          /* Control never gets here */
3476          }          }
3477        else        else  /* Maximize */
3478          {          {
3479          pp = eptr;          pp = eptr;
3480          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3481            {            {
3482            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3483                {
3484                SCHECK_PARTIAL();
3485                break;
3486