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

Diff of /code/trunk/pcre_exec.c

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

revision 85 by nigel, Sat Feb 24 21:41:13 2007 UTC revision 1047 by zherczeg, Fri Sep 28 15:06:38 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 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  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
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
43    
44    #ifdef HAVE_CONFIG_H
45    #include "config.h"
46    #endif
47    
48    #define NLBLOCK md             /* Block containing newline information */
49    #define PSSTART start_subject  /* Field containing processed string start */
50    #define PSEND   end_subject    /* Field containing processed string end */
51    
52  #include "pcre_internal.h"  #include "pcre_internal.h"
53    
54    /* Undefine some potentially clashing cpp symbols */
55    
56  /* Structure for building a chain of data that actually lives on the  #undef min
57  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;  
58    
59  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
60    of call to match(). We do it this way to save on using another stack variable,
61    as stack usage is to be discouraged. */
62    
63  #define match_condassert   0x01    /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
64  #define match_isgroup      0x02    /* Set if start of bracketed group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
65    
66  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
67  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 69  defined PCRE_ERROR_xxx codes, which are
69  #define MATCH_MATCH        1  #define MATCH_MATCH        1
70  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
71    
72    /* Special internal returns from the match() function. Make them sufficiently
73    negative to avoid the external error codes. */
74    
75    #define MATCH_ACCEPT       (-999)
76    #define MATCH_COMMIT       (-998)
77    #define MATCH_KETRPOS      (-997)
78    #define MATCH_ONCE         (-996)
79    #define MATCH_PRUNE        (-995)
80    #define MATCH_SKIP         (-994)
81    #define MATCH_SKIP_ARG     (-993)
82    #define MATCH_THEN         (-992)
83    
84  /* 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.
85  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,
86  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 94  static const char rep_max[] = { 0, 0, 0,
94    
95    
96    
97  #ifdef DEBUG  #ifdef PCRE_DEBUG
98  /*************************************************  /*************************************************
99  *        Debugging function to print chars       *  *        Debugging function to print chars       *
100  *************************************************/  *************************************************/
# Line 99  Returns:     nothing Line 112  Returns:     nothing
112  */  */
113    
114  static void  static void
115  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)
116  {  {
117  int c;  unsigned int c;
118  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;
119  while (length-- > 0)  while (length-- > 0)
120    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 127  while (length-- > 0)
127  *          Match a back-reference                *  *          Match a back-reference                *
128  *************************************************/  *************************************************/
129    
130  /* 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
131  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
132    the length passed is zero. Note that in caseless UTF-8 mode, the number of
133    subject bytes matched may be different to the number of reference bytes.
134    
135  Arguments:  Arguments:
136    offset      index into the offset vector    offset      index into the offset vector
137    eptr        points into the subject    eptr        pointer into the subject
138    length      length to be matched    length      length of reference to be matched (number of bytes)
139    md          points to match data block    md          points to match data block
140    ims         the ims flags    caseless    TRUE if caseless
141    
142  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
143                  -1 no match
144                  -2 partial match; always given if at end subject
145  */  */
146    
147  static BOOL  static int
148  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,
149    unsigned long int ims)    BOOL caseless)
150  {  {
151  const uschar *p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
152    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
153    
154  #ifdef DEBUG  #ifdef PCRE_DEBUG
155  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
156    printf("matching subject <null>");    printf("matching subject <null>");
157  else  else
# Line 146  pchars(p, length, FALSE, md); Line 164  pchars(p, length, FALSE, md);
164  printf("\n");  printf("\n");
165  #endif  #endif
166    
167  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
168    case the length is passed as zero). */
169    
170  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
171    
172  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
173    properly if Unicode properties are supported. Otherwise, we can check only
174    ASCII characters. */
175    
176  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
177    {    {
178    while (length-- > 0)  #ifdef SUPPORT_UTF
179      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
180      if (md->utf)
181        {
182        /* Match characters up to the end of the reference. NOTE: the number of
183        data units matched may differ, because in UTF-8 there are some characters
184        whose upper and lower case versions code have different numbers of bytes.
185        For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
186        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
187        sequence of two of the latter. It is important, therefore, to check the
188        length along the reference, not along the subject (earlier code did this
189        wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193          {
194          unsigned int c, d;
195          const ucd_record *ur;
196          if (eptr >= md->end_subject) return -2;   /* Partial match */
197          GETCHARINC(c, eptr);
198          GETCHARINC(d, p);
199          ur = GET_UCD(d);
200          if (c != d && c != d + ur->other_case)
201            {
202            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
203            for (;;)
204              {
205              if (c < *pp) return -1;
206              if (c == *pp++) break;
207              }
208            }
209          }
210        }
211      else
212    #endif
213    #endif
214    
215      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
216      is no UCP support. */
217        {
218        while (length-- > 0)
219          {
220          if (eptr >= md->end_subject) return -2;   /* Partial match */
221          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
222          p++;
223          eptr++;
224          }
225        }
226    }    }
227    
228    /* In the caseful case, we can just compare the bytes, whether or not we
229    are in UTF-8 mode. */
230    
231  else  else
232    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
233      while (length-- > 0)
234        {
235        if (eptr >= md->end_subject) return -2;   /* Partial match */
236        if (*p++ != *eptr++) return -1;
237        }
238      }
239    
240  return TRUE;  return (int)(eptr - eptr_start);
241  }  }
242    
243    
# Line 169  return TRUE; Line 246  return TRUE;
246  ****************************************************************************  ****************************************************************************
247                     RECURSION IN THE match() FUNCTION                     RECURSION IN THE match() FUNCTION
248    
249  The match() function is highly recursive. Some regular expressions can cause  The match() function is highly recursive, though not every recursive call
250  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
251  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
252  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
253  works fine.  saved for a recursive call. On Unix, the stack can be large, and this works
254    fine.
255  It turns out that on non-Unix systems there are problems with programs that  
256  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
257  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
258  for decades.) So....  has oodles of memory these days, and techniques for extending the stack have
259    been known for decades.) So....
260    
261  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive  There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
262  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
263  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
264  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
265  always used to.  always used to.
266    
267    The original heap-recursive code used longjmp(). However, it seems that this
268    can be very slow on some operating systems. Following a suggestion from Stan
269    Switzer, the use of longjmp() has been abolished, at the cost of having to
270    provide a unique number for each call to RMATCH. There is no way of generating
271    a sequence of numbers at compile time in C. I have given them names, to make
272    them stand out more clearly.
273    
274    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
275    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
276    tests. Furthermore, not using longjmp() means that local dynamic variables
277    don't have indeterminate values; this has meant that the frame size can be
278    reduced because the result can be "passed back" by straight setting of the
279    variable instead of being passed in the frame.
280  ****************************************************************************  ****************************************************************************
281  ***************************************************************************/  ***************************************************************************/
282    
283    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
284    below must be updated in sync.  */
285    
286  /* These versions of the macros use the stack, as normal */  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
287           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
288           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
289           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
290           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
291           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
292           RM61,  RM62, RM63, RM64, RM65, RM66 };
293    
294    /* These versions of the macros use the stack, as normal. There are debugging
295    versions and production versions. Note that the "rw" argument of RMATCH isn't
296    actually used in this definition. */
297    
298  #ifndef NO_RECURSE  #ifndef NO_RECURSE
299  #define REGISTER register  #define REGISTER register
300  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg)  
301    #ifdef PCRE_DEBUG
302    #define RMATCH(ra,rb,rc,rd,re,rw) \
303      { \
304      printf("match() called in line %d\n", __LINE__); \
305      rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
306      printf("to line %d\n", __LINE__); \
307      }
308    #define RRETURN(ra) \
309      { \
310      printf("match() returned %d from line %d ", ra, __LINE__); \
311      return ra; \
312      }
313    #else
314    #define RMATCH(ra,rb,rc,rd,re,rw) \
315      rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
316  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
317    #endif
318    
319  #else  #else
320    
321    
322  /* 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
323  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
324  match(), which never changes. */  argument of match(), which never changes. */
325    
326  #define REGISTER  #define REGISTER
327    
328  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
329    {\    {\
330    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
331    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\  
332      {\      {\
333      DPRINTF(("longjumped back to line %d\n", __LINE__));\      newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
334      frame = md->thisframe;\      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
335      rx = frame->Xresult;\      newframe->Xnextframe = NULL;\
336        frame->Xnextframe = newframe;\
337      }\      }\
338      frame->Xwhere = rw;\
339      newframe->Xeptr = ra;\
340      newframe->Xecode = rb;\
341      newframe->Xmstart = mstart;\
342      newframe->Xoffset_top = rc;\
343      newframe->Xeptrb = re;\
344      newframe->Xrdepth = frame->Xrdepth + 1;\
345      newframe->Xprevframe = frame;\
346      frame = newframe;\
347      DPRINTF(("restarting from line %d\n", __LINE__));\
348      goto HEAP_RECURSE;\
349      L_##rw:\
350      DPRINTF(("jumped back to line %d\n", __LINE__));\
351    }    }
352    
353  #define RRETURN(ra)\  #define RRETURN(ra)\
354    {\    {\
355    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
356    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
357    if (frame != NULL)\    if (frame != NULL)\
358      {\      {\
359      frame->Xresult = ra;\      rrc = ra;\
360      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
361      }\      }\
362    return ra;\    return ra;\
363    }    }
# Line 247  match(), which never changes. */ Line 367  match(), which never changes. */
367    
368  typedef struct heapframe {  typedef struct heapframe {
369    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
370      struct heapframe *Xnextframe;
371    
372    /* Function arguments that may change */    /* Function arguments that may change */
373    
374    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
375    const uschar *Xecode;    const pcre_uchar *Xecode;
376      PCRE_PUCHAR Xmstart;
377    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
378    eptrblock *Xeptrb;    eptrblock *Xeptrb;
379    int Xflags;    unsigned int Xrdepth;
380    
381    /* Function local variables */    /* Function local variables */
382    
383    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
384    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
385    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
386    const uschar *Xnext;  #endif
387    const uschar *Xpp;    PCRE_PUCHAR Xdata;
388    const uschar *Xprev;    PCRE_PUCHAR Xnext;
389    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
390      PCRE_PUCHAR Xprev;
391      PCRE_PUCHAR Xsaved_eptr;
392    
393    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
394    
395    BOOL Xcur_is_word;    BOOL Xcur_is_word;
396    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
397    BOOL Xprev_is_word;    BOOL Xprev_is_word;
398    
   unsigned long int Xoriginal_ims;  
   
399  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
400    int Xprop_type;    int Xprop_type;
401      int Xprop_value;
402    int Xprop_fail_result;    int Xprop_fail_result;
403    int Xprop_category;    int Xoclength;
404    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_othercase;  
   int Xprop_test_against;  
   int *Xprop_test_variable;  
405  #endif  #endif
406    
407      int Xcodelink;
408    int Xctype;    int Xctype;
409    int Xfc;    unsigned int Xfc;
410    int Xfi;    int Xfi;
411    int Xlength;    int Xlength;
412    int Xmax;    int Xmax;
# Line 301  typedef struct heapframe { Line 420  typedef struct heapframe {
420    
421    eptrblock Xnewptrb;    eptrblock Xnewptrb;
422    
423    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
424    
425    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
426    
427  } heapframe;  } heapframe;
428    
# Line 320  typedef struct heapframe { Line 438  typedef struct heapframe {
438  *         Match from current position            *  *         Match from current position            *
439  *************************************************/  *************************************************/
440    
441  /* 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  
442  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
443  same response.  same response. */
444    
445    /* These macros pack up tests that are used for partial matching, and which
446    appear several times in the code. We set the "hit end" flag if the pointer is
447    at the end of the subject and also past the start of the subject (i.e.
448    something has been matched). For hard partial matching, we then return
449    immediately. The second one is used when we already know we are past the end of
450    the subject. */
451    
452    #define CHECK_PARTIAL()\
453      if (md->partial != 0 && eptr >= md->end_subject && \
454          eptr > md->start_used_ptr) \
455        { \
456        md->hitend = TRUE; \
457        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
458        }
459    
460    #define SCHECK_PARTIAL()\
461      if (md->partial != 0 && eptr > md->start_used_ptr) \
462        { \
463        md->hitend = TRUE; \
464        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
465        }
466    
467    
468  Performance note: It might be tempting to extract commonly used fields from the  /* Performance note: It might be tempting to extract commonly used fields from
469  md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf, end_subject) into individual variables to improve
470  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
471  made performance worse.  made performance worse.
472    
473  Arguments:  Arguments:
474     eptr        pointer in subject     eptr        pointer to current character in subject
475     ecode       position in code     ecode       pointer to current position in compiled code
476       mstart      pointer to the current match start position (can be modified
477                     by encountering \K)
478     offset_top  current top pointer     offset_top  current top pointer
479     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
480     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
481                   brackets - for testing for empty matches                   brackets - for testing for empty matches
482     flags       can contain     rdepth      the recursion depth
                  match_condassert - this is an assertion condition  
                  match_isgroup - this is the start of a bracketed group  
483    
484  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
485                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
486                   a negative MATCH_xxx value for PRUNE, SKIP, etc
487                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
488                   (e.g. stopped by recursion limit)                   (e.g. stopped by repeated call or recursion limit)
489  */  */
490    
491  static int  static int
492  match(REGISTER const uschar *eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
493    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
494    int flags)    unsigned int rdepth)
495  {  {
496  /* 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,
497  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
498  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
499    
500  register int  rrc;    /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
501  register int  i;      /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
502  register int  c;      /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
503  register BOOL utf8;   /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
504    
505    BOOL minimize, possessive; /* Quantifier options */
506    BOOL caseless;
507    int condcode;
508    
509  /* 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
510  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
511  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
512  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
513    the top-level on the stack rather than malloc-ing them all gives a performance
514    boost in many cases where there is not much "recursion". */
515    
516  #ifdef NO_RECURSE  #ifdef NO_RECURSE
517  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
518    
519  /* Copy in the original argument variables */  /* Copy in the original argument variables */
520    
521  frame->Xeptr = eptr;  frame->Xeptr = eptr;
522  frame->Xecode = ecode;  frame->Xecode = ecode;
523    frame->Xmstart = mstart;
524  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
525  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
526  frame->Xflags = flags;  frame->Xrdepth = rdepth;
527    
528  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
529    
# Line 390  HEAP_RECURSE: Line 533  HEAP_RECURSE:
533    
534  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
535  #define ecode              frame->Xecode  #define ecode              frame->Xecode
536    #define mstart             frame->Xmstart
537  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
538  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
539  #define flags              frame->Xflags  #define rdepth             frame->Xrdepth
540    
541  /* Ditto for the local variables */  /* Ditto for the local variables */
542    
543  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
544  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
545  #endif  #endif
546  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
547    #define codelink           frame->Xcodelink
548  #define data               frame->Xdata  #define data               frame->Xdata
549  #define next               frame->Xnext  #define next               frame->Xnext
550  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 411  HEAP_RECURSE: Line 555  HEAP_RECURSE:
555    
556  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
557  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
558  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
559    
 #define original_ims       frame->Xoriginal_ims  
   
560  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
561  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
562    #define prop_value         frame->Xprop_value
563  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
564  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
565  #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  
566  #endif  #endif
567    
568  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 447  HEAP_RECURSE: Line 586  HEAP_RECURSE:
586  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
587  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
588    
589  #else  #else         /* NO_RECURSE not defined */
590  #define fi i  #define fi i
591  #define fc c  #define fc c
592    
593    /* Many of the following variables are used only in small blocks of the code.
594    My normal style of coding would have declared them within each of those blocks.
595    However, in order to accommodate the version of this code that uses an external
596    "stack" implemented on the heap, it is easier to declare them all here, so the
597    declarations can be cut out in a block. The only declarations within blocks
598    below are for variables that do not have to be preserved over a recursive call
599    to RMATCH(). */
600    
601  #ifdef SUPPORT_UTF8                /* Many of these variables are used ony */  #ifdef SUPPORT_UTF
602  const uschar *charptr;             /* small blocks of the code. My normal  */  const pcre_uchar *charptr;
603  #endif                             /* style of coding would have declared  */  #endif
604  const uschar *callpat;             /* them within each of those blocks.    */  const pcre_uchar *callpat;
605  const uschar *data;                /* However, in order to accommodate the */  const pcre_uchar *data;
606  const uschar *next;                /* version of this code that uses an    */  const pcre_uchar *next;
607  const uschar *pp;                  /* external "stack" implemented on the  */  PCRE_PUCHAR       pp;
608  const uschar *prev;                /* heap, it is easier to declare them   */  const pcre_uchar *prev;
609  const uschar *saved_eptr;          /* all here, so the declarations can    */  PCRE_PUCHAR       saved_eptr;
                                    /* be cut out in a block. The only      */  
 recursion_info new_recursive;      /* declarations within blocks below are */  
                                    /* 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;  
610    
611  unsigned long int original_ims;  recursion_info new_recursive;
612    
613    BOOL cur_is_word;
614    BOOL condition;
615    BOOL prev_is_word;
616    
617  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
618  int prop_type;  int prop_type;
619    int prop_value;
620  int prop_fail_result;  int prop_fail_result;
621  int prop_category;  int oclength;
622  int prop_chartype;  pcre_uchar occhars[6];
 int prop_othercase;  
 int prop_test_against;  
 int *prop_test_variable;  
623  #endif  #endif
624    
625    int codelink;
626  int ctype;  int ctype;
627  int length;  int length;
628  int max;  int max;
# Line 493  int save_offset1, save_offset2, save_off Line 635  int save_offset1, save_offset2, save_off
635  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
636    
637  eptrblock newptrb;  eptrblock newptrb;
638  #endif  
639    /* There is a special fudge for calling match() in a way that causes it to
640    measure the size of its basic stack frame when the stack is being used for
641    recursion. The second argument (ecode) being NULL triggers this behaviour. It
642    cannot normally ever be NULL. The return is the negated value of the frame
643    size. */
644    
645    if (ecode == NULL)
646      {
647      if (rdepth == 0)
648        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
649      else
650        {
651        int len = (char *)&rdepth - (char *)eptr;
652        return (len > 0)? -len : len;
653        }
654      }
655    #endif     /* NO_RECURSE */
656    
657    /* To save space on the stack and in the heap frame, I have doubled up on some
658    of the local variables that are used only in localised parts of the code, but
659    still need to be preserved over recursive calls of match(). These macros define
660    the alternative names that are used. */
661    
662    #define allow_zero    cur_is_word
663    #define cbegroup      condition
664    #define code_offset   codelink
665    #define condassert    condition
666    #define matched_once  prev_is_word
667    #define foc           number
668    #define save_mark     data
669    
670  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
671  variables. */  variables. */
672    
673  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
674    prop_value = 0;
675  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_against = 0;  
 prop_test_variable = NULL;  
676  #endif  #endif
677    
 /* 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. */  
678    
679  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
680    when NO_RECURSE is not defined, in order to reduce the amount of stack that is
681    used. Thanks to Ian Taylor for noticing this possibility and sending the
682    original patch. */
683    
684    TAIL_RECURSE:
685    
686    /* OK, now we can get on with the real code of the function. Recursive calls
687    are specified by the macro RMATCH and RRETURN is used to return. When
688    NO_RECURSE is *not* defined, these just turn into a recursive call to match()
689    and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
690    defined). However, RMATCH isn't like a function call because it's quite a
691    complicated macro. It has to be used in one particular way. This shouldn't,
692    however, impact performance when true recursion is being used. */
693    
694  original_ims = ims;    /* Save for resetting on ')' */  #ifdef SUPPORT_UTF
695  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
696    #else
697    utf = FALSE;
698    #endif
699    
700    /* First check that we haven't called match() too many times, or that we
701    haven't exceeded the recursive call limit. */
702    
703    if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
704    if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
705    
706  /* 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
707  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
708  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
709  this stack. */  up space on the stack. See also MATCH_CONDASSERT below.
710    
711    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
712    such remembered pointers, to be checked when we hit the closing ket, in order
713    to break infinite loops that match no characters. When match() is called in
714    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
715    NOT be used with tail recursion, because the memory block that is used is on
716    the stack, so a new one may be required for each match(). */
717    
718  if ((flags & match_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
719    {    {
   newptrb.epb_prev = eptrb;  
720    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
721      newptrb.epb_prev = eptrb;
722    eptrb = &newptrb;    eptrb = &newptrb;
723      md->match_function_type = 0;
724    }    }
725    
726  /* Now start processing the operations. */  /* Now start processing the opcodes. */
727    
728  for (;;)  for (;;)
729    {    {
730      minimize = possessive = FALSE;
731    op = *ecode;    op = *ecode;
   minimize = FALSE;  
732    
733    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
734    matching at least one subject character. */      {
735        case OP_MARK:
736        md->nomatch_mark = ecode + 2;
737        md->mark = NULL;    /* In case previously set by assertion */
738        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
739          eptrb, RM55);
740        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
741             md->mark == NULL) md->mark = ecode + 2;
742    
743        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
744        argument, and we must check whether that argument matches this MARK's
745        argument. It is passed back in md->start_match_ptr (an overloading of that
746        variable). If it does match, we reset that variable to the current subject
747        position and return MATCH_SKIP. Otherwise, pass back the return code
748        unaltered. */
749    
750        else if (rrc == MATCH_SKIP_ARG &&
751            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
752          {
753          md->start_match_ptr = eptr;
754          RRETURN(MATCH_SKIP);
755          }
756        RRETURN(rrc);
757    
758        case OP_FAIL:
759        RRETURN(MATCH_NOMATCH);
760    
761    if (md->partial &&      /* COMMIT overrides PRUNE, SKIP, and THEN */
       eptr >= md->end_subject &&  
       eptr > md->start_match)  
     md->hitend = TRUE;  
   
   /* Opening capturing bracket. If there is space in the offset vector, save  
   the current subject position in the working slot at the top of the vector. We  
   mustn't change the current values of the data slot, because they may be set  
   from a previous iteration of this group, and be referred to by a reference  
   inside the group.  
   
   If the bracket fails to match, we need to restore this value and also the  
   values of the final offsets, in case they were set by a previous iteration of  
   the same bracket.  
   
   If there isn't enough space in the offset vector, treat this as if it were a  
   non-capturing bracket. Don't worry about setting the flag for the error case  
   here; that is handled in the code for KET. */  
762    
763    if (op > OP_BRA)      case OP_COMMIT:
764      {      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
765      number = op - OP_BRA;        eptrb, RM52);
766        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
767            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
768            rrc != MATCH_THEN)
769          RRETURN(rrc);
770        RRETURN(MATCH_COMMIT);
771    
772        /* PRUNE overrides THEN */
773    
774        case OP_PRUNE:
775        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
776          eptrb, RM51);
777        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
778        RRETURN(MATCH_PRUNE);
779    
780        case OP_PRUNE_ARG:
781        md->nomatch_mark = ecode + 2;
782        md->mark = NULL;    /* In case previously set by assertion */
783        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
784          eptrb, RM56);
785        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
786             md->mark == NULL) md->mark = ecode + 2;
787        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
788        RRETURN(MATCH_PRUNE);
789    
790        /* SKIP overrides PRUNE and THEN */
791    
792        case OP_SKIP:
793        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
794          eptrb, RM53);
795        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
796          RRETURN(rrc);
797        md->start_match_ptr = eptr;   /* Pass back current position */
798        RRETURN(MATCH_SKIP);
799    
800        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
801        nomatch_mark. There is a flag that disables this opcode when re-matching a
802        pattern that ended with a SKIP for which there was not a matching MARK. */
803    
804        case OP_SKIP_ARG:
805        if (md->ignore_skip_arg)
806          {
807          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
808          break;
809          }
810        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
811          eptrb, RM57);
812        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
813          RRETURN(rrc);
814    
815        /* Pass back the current skip name by overloading md->start_match_ptr and
816        returning the special MATCH_SKIP_ARG return code. This will either be
817        caught by a matching MARK, or get to the top, where it causes a rematch
818        with the md->ignore_skip_arg flag set. */
819    
820        md->start_match_ptr = ecode + 2;
821        RRETURN(MATCH_SKIP_ARG);
822    
823        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
824        the branch in which it occurs can be determined. Overload the start of
825        match pointer to do this. */
826    
827        case OP_THEN:
828        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
829          eptrb, RM54);
830        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831        md->start_match_ptr = ecode;
832        RRETURN(MATCH_THEN);
833    
834        case OP_THEN_ARG:
835        md->nomatch_mark = ecode + 2;
836        md->mark = NULL;    /* In case previously set by assertion */
837        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
838          md, eptrb, RM58);
839        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
840             md->mark == NULL) md->mark = ecode + 2;
841        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
842        md->start_match_ptr = ecode;
843        RRETURN(MATCH_THEN);
844    
845        /* Handle an atomic group that does not contain any capturing parentheses.
846        This can be handled like an assertion. Prior to 8.13, all atomic groups
847        were handled this way. In 8.13, the code was changed as below for ONCE, so
848        that backups pass through the group and thereby reset captured values.
849        However, this uses a lot more stack, so in 8.20, atomic groups that do not
850        contain any captures generate OP_ONCE_NC, which can be handled in the old,
851        less stack intensive way.
852    
853        Check the alternative branches in turn - the matching won't pass the KET
854        for this kind of subpattern. If any one branch matches, we carry on as at
855        the end of a normal bracket, leaving the subject pointer, but resetting
856        the start-of-match value in case it was changed by \K. */
857    
858        case OP_ONCE_NC:
859        prev = ecode;
860        saved_eptr = eptr;
861        save_mark = md->mark;
862        do
863          {
864          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
865          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
866            {
867            mstart = md->start_match_ptr;
868            break;
869            }
870          if (rrc == MATCH_THEN)
871            {
872            next = ecode + GET(ecode,1);
873            if (md->start_match_ptr < next &&
874                (*ecode == OP_ALT || *next == OP_ALT))
875              rrc = MATCH_NOMATCH;
876            }
877    
878          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
879          ecode += GET(ecode,1);
880          md->mark = save_mark;
881          }
882        while (*ecode == OP_ALT);
883    
884        /* If hit the end of the group (which could be repeated), fail */
885    
886        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
887    
888        /* Continue as from after the group, updating the offsets high water
889        mark, since extracts may have been taken. */
890    
891        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
892    
893        offset_top = md->end_offset_top;
894        eptr = md->end_match_ptr;
895    
896        /* For a non-repeating ket, just continue at this level. This also
897        happens for a repeating ket if no characters were matched in the group.
898        This is the forcible breaking of infinite loops as implemented in Perl
899        5.005. */
900    
901      /* For extended extraction brackets (large number), we have to fish out the      if (*ecode == OP_KET || eptr == saved_eptr)
902      number from a dummy opcode at the start. */        {
903          ecode += 1+LINK_SIZE;
904          break;
905          }
906    
907        /* The repeating kets try the rest of the pattern or restart from the
908        preceding bracket, in the appropriate order. The second "call" of match()
909        uses tail recursion, to avoid using another stack frame. */
910    
911        if (*ecode == OP_KETRMIN)
912          {
913          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
914          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
915          ecode = prev;
916          goto TAIL_RECURSE;
917          }
918        else  /* OP_KETRMAX */
919          {
920          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
921          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
922          ecode += 1 + LINK_SIZE;
923          goto TAIL_RECURSE;
924          }
925        /* Control never gets here */
926    
927      if (number > EXTRACT_BASIC_MAX)      /* Handle a capturing bracket, other than those that are possessive with an
928        number = GET2(ecode, 2+LINK_SIZE);      unlimited repeat. If there is space in the offset vector, save the current
929        subject position in the working slot at the top of the vector. We mustn't
930        change the current values of the data slot, because they may be set from a
931        previous iteration of this group, and be referred to by a reference inside
932        the group. A failure to match might occur after the group has succeeded,
933        if something later on doesn't match. For this reason, we need to restore
934        the working value and also the values of the final offsets, in case they
935        were set by a previous iteration of the same bracket.
936    
937        If there isn't enough space in the offset vector, treat this as if it were
938        a non-capturing bracket. Don't worry about setting the flag for the error
939        case here; that is handled in the code for KET. */
940    
941        case OP_CBRA:
942        case OP_SCBRA:
943        number = GET2(ecode, 1+LINK_SIZE);
944      offset = number << 1;      offset = number << 1;
945    
946  #ifdef DEBUG  #ifdef PCRE_DEBUG
947      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
948        printf("subject=");
949      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
950      printf("\n");      printf("\n");
951  #endif  #endif
# Line 580  for (;;) Line 956  for (;;)
956        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
957        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
958        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
959          save_mark = md->mark;
960    
961        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
962        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
963            (int)(eptr - md->start_subject);
964    
965        do        for (;;)
966          {          {
967          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
968            match_isgroup);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
969              eptrb, RM1);
970            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
971    
972            /* If we backed up to a THEN, check whether it is within the current
973            branch by comparing the address of the THEN that is passed back with
974            the end of the branch. If it is within the current branch, and the
975            branch is one of two or more alternatives (it either starts or ends
976            with OP_ALT), we have reached the limit of THEN's action, so convert
977            the return code to NOMATCH, which will cause normal backtracking to
978            happen from now on. Otherwise, THEN is passed back to an outer
979            alternative. This implements Perl's treatment of parenthesized groups,
980            where a group not containing | does not affect the current alternative,
981            that is, (X) is NOT the same as (X|(*F)). */
982    
983            if (rrc == MATCH_THEN)
984              {
985              next = ecode + GET(ecode,1);
986              if (md->start_match_ptr < next &&
987                  (*ecode == OP_ALT || *next == OP_ALT))
988                rrc = MATCH_NOMATCH;
989              }
990    
991            /* Anything other than NOMATCH is passed back. */
992    
993          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
994          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
995          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
996            md->mark = save_mark;
997            if (*ecode != OP_ALT) break;
998          }          }
       while (*ecode == OP_ALT);  
999    
1000        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1001        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1002        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1003        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1004    
1005        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1006    
1007          RRETURN(rrc);
1008        }        }
1009    
1010      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1011        as a non-capturing bracket. */
1012    
1013      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1014      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1015    
1016    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1017    
1018    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1019      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1020      case OP_BRA:     /* Non-capturing bracket: optimized */  
1021      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
1022      do      repeat and ONCE group with no captures. Loop for all the alternatives.
1023    
1024        When we get to the final alternative within the brackets, we used to return
1025        the result of a recursive call to match() whatever happened so it was
1026        possible to reduce stack usage by turning this into a tail recursion,
1027        except in the case of a possibly empty group. However, now that there is
1028        the possiblity of (*THEN) occurring in the final alternative, this
1029        optimization is no longer always possible.
1030    
1031        We can optimize if we know there are no (*THEN)s in the pattern; at present
1032        this is the best that can be done.
1033    
1034        MATCH_ONCE is returned when the end of an atomic group is successfully
1035        reached, but subsequent matching fails. It passes back up the tree (causing
1036        captured values to be reset) until the original atomic group level is
1037        reached. This is tested by comparing md->once_target with the start of the
1038        group. At this point, the return is converted into MATCH_NOMATCH so that
1039        previous backup points can be taken. */
1040    
1041        case OP_ONCE:
1042        case OP_BRA:
1043        case OP_SBRA:
1044        DPRINTF(("start non-capturing bracket\n"));
1045    
1046        for (;;)
1047        {        {
1048        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        if (op >= OP_SBRA || op == OP_ONCE)
1049          match_isgroup);          md->match_function_type = MATCH_CBEGROUP;
1050        if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1051          /* If this is not a possibly empty group, and there are no (*THEN)s in
1052          the pattern, and this is the final alternative, optimize as described
1053          above. */
1054    
1055          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1056            {
1057            ecode += PRIV(OP_lengths)[*ecode];
1058            goto TAIL_RECURSE;
1059            }
1060    
1061          /* In all other cases, we have to make another call to match(). */
1062    
1063          save_mark = md->mark;
1064          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1065            RM2);
1066    
1067          /* See comment in the code for capturing groups above about handling
1068          THEN. */
1069    
1070          if (rrc == MATCH_THEN)
1071            {
1072            next = ecode + GET(ecode,1);
1073            if (md->start_match_ptr < next &&
1074                (*ecode == OP_ALT || *next == OP_ALT))
1075              rrc = MATCH_NOMATCH;
1076            }
1077    
1078          if (rrc != MATCH_NOMATCH)
1079            {
1080            if (rrc == MATCH_ONCE)
1081              {
1082              const pcre_uchar *scode = ecode;
1083              if (*scode != OP_ONCE)           /* If not at start, find it */
1084                {
1085                while (*scode == OP_ALT) scode += GET(scode, 1);
1086                scode -= GET(scode, 1);
1087                }
1088              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1089              }
1090            RRETURN(rrc);
1091            }
1092        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1093          md->mark = save_mark;
1094          if (*ecode != OP_ALT) break;
1095        }        }
1096      while (*ecode == OP_ALT);  
     DPRINTF(("bracket 0 failed\n"));  
1097      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1098    
1099      /* Conditional group: compilation checked that there are no more than      /* Handle possessive capturing brackets with an unlimited repeat. We come
1100      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
1101      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
1102      exactly what going to the ket would do. */      different. The end of these brackets will always be OP_KETRPOS, which
1103        returns MATCH_KETRPOS without going further in the pattern. By this means
1104        we can handle the group by iteration rather than recursion, thereby
1105        reducing the amount of stack needed. */
1106    
1107        case OP_CBRAPOS:
1108        case OP_SCBRAPOS:
1109        allow_zero = FALSE;
1110    
1111      case OP_COND:      POSSESSIVE_CAPTURE:
1112      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      number = GET2(ecode, 1+LINK_SIZE);
1113        {      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);  
       }  
1114    
1115      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1116      the final argument TRUE causes it to stop at the end of an assertion. */      printf("start possessive bracket %d\n", number);
1117        printf("subject=");
1118        pchars(eptr, 16, TRUE, md);
1119        printf("\n");
1120    #endif
1121    
1122      else      if (offset < md->offset_max)
1123        {        {
1124        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1125            match_condassert | match_isgroup);        code_offset = (int)(ecode - md->start_code);
1126        if (rrc == MATCH_MATCH)  
1127          save_offset1 = md->offset_vector[offset];
1128          save_offset2 = md->offset_vector[offset+1];
1129          save_offset3 = md->offset_vector[md->offset_end - number];
1130          save_capture_last = md->capture_last;
1131    
1132          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1133    
1134          /* Each time round the loop, save the current subject position for use
1135          when the group matches. For MATCH_MATCH, the group has matched, so we
1136          restart it with a new subject starting position, remembering that we had
1137          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1138          usual. If we haven't matched any alternatives in any iteration, check to
1139          see if a previous iteration matched. If so, the group has matched;
1140          continue from afterwards. Otherwise it has failed; restore the previous
1141          capture values before returning NOMATCH. */
1142    
1143          for (;;)
1144            {
1145            md->offset_vector[md->offset_end - number] =
1146              (int)(eptr - md->start_subject);
1147            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1148            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1149              eptrb, RM63);
1150            if (rrc == MATCH_KETRPOS)
1151              {
1152              offset_top = md->end_offset_top;
1153              eptr = md->end_match_ptr;
1154              ecode = md->start_code + code_offset;
1155              save_capture_last = md->capture_last;
1156              matched_once = TRUE;
1157              continue;
1158              }
1159    
1160            /* See comment in the code for capturing groups above about handling
1161            THEN. */
1162    
1163            if (rrc == MATCH_THEN)
1164              {
1165              next = ecode + GET(ecode,1);
1166              if (md->start_match_ptr < next &&
1167                  (*ecode == OP_ALT || *next == OP_ALT))
1168                rrc = MATCH_NOMATCH;
1169              }
1170    
1171            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1172            md->capture_last = save_capture_last;
1173            ecode += GET(ecode, 1);
1174            if (*ecode != OP_ALT) break;
1175            }
1176    
1177          if (!matched_once)
1178          {          {
1179          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          md->offset_vector[offset] = save_offset1;
1180          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          md->offset_vector[offset+1] = save_offset2;
1181            md->offset_vector[md->offset_end - number] = save_offset3;
1182          }          }
1183        else if (rrc != MATCH_NOMATCH)  
1184          if (allow_zero || matched_once)
1185          {          {
1186          RRETURN(rrc);         /* Need braces because of following else */          ecode += 1 + LINK_SIZE;
1187            break;
1188          }          }
1189        else ecode += GET(ecode, 1);  
1190        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        RRETURN(MATCH_NOMATCH);
         match_isgroup);  
       RRETURN(rrc);  
1191        }        }
     /* Control never reaches here */  
1192    
1193      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1194      encountered. */      as a non-capturing bracket. */
1195    
1196      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1197      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1198    
1199      /* 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. */  
1200    
1201      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1202      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;  
       }  
1203    
1204      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1205      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1206        without the capturing complication. It is written out separately for speed
1207        and cleanliness. */
1208    
1209      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1210      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1211      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;  
1212    
1213      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1214      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1215      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. */  
1216    
1217      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1218        {        {
1219        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1220          match_isgroup);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1221        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1222          if (rrc == MATCH_KETRPOS)
1223            {
1224            offset_top = md->end_offset_top;
1225            eptr = md->end_match_ptr;
1226            ecode = md->start_code + code_offset;
1227            matched_once = TRUE;
1228            continue;
1229            }
1230    
1231          /* See comment in the code for capturing groups above about handling
1232          THEN. */
1233    
1234          if (rrc == MATCH_THEN)
1235            {
1236            next = ecode + GET(ecode,1);
1237            if (md->start_match_ptr < next &&
1238                (*ecode == OP_ALT || *next == OP_ALT))
1239              rrc = MATCH_NOMATCH;
1240            }
1241    
1242        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1243        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1244          if (*ecode != OP_ALT) break;
1245        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1246    
1247      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1248          {
1249          ecode += 1 + LINK_SIZE;
1250          break;
1251          }
1252        RRETURN(MATCH_NOMATCH);
1253    
1254      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1255    
1256      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1257      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1258        past the end if there is only one branch, but that's OK because that is
1259        exactly what going to the ket would do. */
1260    
1261      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1262      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1263      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1264    
1265      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1266        inserted between OP_COND and an assertion condition. */
1267    
1268      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1269        {        {
1270        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (PUBL(callout) != NULL)
1271          match_isgroup);          {
1272        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1273        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1274        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1275            cb.offset_vector    = md->offset_vector;
1276    #ifdef COMPILE_PCRE8
1277            cb.subject          = (PCRE_SPTR)md->start_subject;
1278    #else
1279            cb.subject          = (PCRE_SPTR16)md->start_subject;
1280    #endif
1281            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1282            cb.start_match      = (int)(mstart - md->start_subject);
1283            cb.current_position = (int)(eptr - md->start_subject);
1284            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1285            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1286            cb.capture_top      = offset_top/2;
1287            cb.capture_last     = md->capture_last;
1288            cb.callout_data     = md->callout_data;
1289            cb.mark             = md->nomatch_mark;
1290            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1291            if (rrc < 0) RRETURN(rrc);
1292            }
1293          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1294        }        }
     while (*ecode == OP_ALT);  
1295    
1296      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      condcode = ecode[LINK_SIZE+1];
1297    
1298      ecode += 1 + LINK_SIZE;      /* Now see what the actual condition is */
     continue;  
   
     /* Move the subject pointer back. This occurs only at the start of  
     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. */  
1299    
1300      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1301        {        {
1302        c = GET(ecode,1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
1303        for (i = 0; i < c; i++)          {
1304            condition = FALSE;
1305            ecode += GET(ecode, 1);
1306            }
1307          else
1308            {
1309            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1310            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1311    
1312            /* If the test is for recursion into a specific subpattern, and it is
1313            false, but the test was set up by name, scan the table to see if the
1314            name refers to any other numbers, and test them. The condition is true
1315            if any one is set. */
1316    
1317            if (!condition && condcode == OP_NRREF)
1318              {
1319              pcre_uchar *slotA = md->name_table;
1320              for (i = 0; i < md->name_count; i++)
1321                {
1322                if (GET2(slotA, 0) == recno) break;
1323                slotA += md->name_entry_size;
1324                }
1325    
1326              /* Found a name for the number - there can be only one; duplicate
1327              names for different numbers are allowed, but not vice versa. First
1328              scan down for duplicates. */
1329    
1330              if (i < md->name_count)
1331                {
1332                pcre_uchar *slotB = slotA;
1333                while (slotB > md->name_table)
1334                  {
1335                  slotB -= md->name_entry_size;
1336                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1337                    {
1338                    condition = GET2(slotB, 0) == md->recursive->group_num;
1339                    if (condition) break;
1340                    }
1341                  else break;
1342                  }
1343    
1344                /* Scan up for duplicates */
1345    
1346                if (!condition)
1347                  {
1348                  slotB = slotA;
1349                  for (i++; i < md->name_count; i++)
1350                    {
1351                    slotB += md->name_entry_size;
1352                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1353                      {
1354                      condition = GET2(slotB, 0) == md->recursive->group_num;
1355                      if (condition) break;
1356                      }
1357                    else break;
1358                    }
1359                  }
1360                }
1361              }
1362    
1363            /* Chose branch according to the condition */
1364    
1365            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1366            }
1367          }
1368    
1369        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1370          {
1371          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1372          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1373    
1374          /* If the numbered capture is unset, but the reference was by name,
1375          scan the table to see if the name refers to any other numbers, and test
1376          them. The condition is true if any one is set. This is tediously similar
1377          to the code above, but not close enough to try to amalgamate. */
1378    
1379          if (!condition && condcode == OP_NCREF)
1380            {
1381            int refno = offset >> 1;
1382            pcre_uchar *slotA = md->name_table;
1383    
1384            for (i = 0; i < md->name_count; i++)
1385              {
1386              if (GET2(slotA, 0) == refno) break;
1387              slotA += md->name_entry_size;
1388              }
1389    
1390            /* Found a name for the number - there can be only one; duplicate names
1391            for different numbers are allowed, but not vice versa. First scan down
1392            for duplicates. */
1393    
1394            if (i < md->name_count)
1395              {
1396              pcre_uchar *slotB = slotA;
1397              while (slotB > md->name_table)
1398                {
1399                slotB -= md->name_entry_size;
1400                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1401                  {
1402                  offset = GET2(slotB, 0) << 1;
1403                  condition = offset < offset_top &&
1404                    md->offset_vector[offset] >= 0;
1405                  if (condition) break;
1406                  }
1407                else break;
1408                }
1409    
1410              /* Scan up for duplicates */
1411    
1412              if (!condition)
1413                {
1414                slotB = slotA;
1415                for (i++; i < md->name_count; i++)
1416                  {
1417                  slotB += md->name_entry_size;
1418                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1419                    {
1420                    offset = GET2(slotB, 0) << 1;
1421                    condition = offset < offset_top &&
1422                      md->offset_vector[offset] >= 0;
1423                    if (condition) break;
1424                    }
1425                  else break;
1426                  }
1427                }
1428              }
1429            }
1430    
1431          /* Chose branch according to the condition */
1432    
1433          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1434          }
1435    
1436        else if (condcode == OP_DEF)     /* DEFINE - always false */
1437          {
1438          condition = FALSE;
1439          ecode += GET(ecode, 1);
1440          }
1441    
1442        /* The condition is an assertion. Call match() to evaluate it - setting
1443        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1444        an assertion. */
1445    
1446        else
1447          {
1448          md->match_function_type = MATCH_CONDASSERT;
1449          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1450          if (rrc == MATCH_MATCH)
1451            {
1452            if (md->end_offset_top > offset_top)
1453              offset_top = md->end_offset_top;  /* Captures may have happened */
1454            condition = TRUE;
1455            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1456            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1457            }
1458    
1459          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1460          assertion; it is therefore treated as NOMATCH. */
1461    
1462          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1463            {
1464            RRETURN(rrc);         /* Need braces because of following else */
1465            }
1466          else
1467            {
1468            condition = FALSE;
1469            ecode += codelink;
1470            }
1471          }
1472    
1473        /* We are now at the branch that is to be obeyed. As there is only one, can
1474        use tail recursion to avoid using another stack frame, except when there is
1475        unlimited repeat of a possibly empty group. In the latter case, a recursive
1476        call to match() is always required, unless the second alternative doesn't
1477        exist, in which case we can just plough on. Note that, for compatibility
1478        with Perl, the | in a conditional group is NOT treated as creating two
1479        alternatives. If a THEN is encountered in the branch, it propagates out to
1480        the enclosing alternative (unless nested in a deeper set of alternatives,
1481        of course). */
1482    
1483        if (condition || *ecode == OP_ALT)
1484          {
1485          if (op != OP_SCOND)
1486            {
1487            ecode += 1 + LINK_SIZE;
1488            goto TAIL_RECURSE;
1489            }
1490    
1491          md->match_function_type = MATCH_CBEGROUP;
1492          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1493          RRETURN(rrc);
1494          }
1495    
1496         /* Condition false & no alternative; continue after the group. */
1497    
1498        else
1499          {
1500          ecode += 1 + LINK_SIZE;
1501          }
1502        break;
1503    
1504    
1505        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1506        to close any currently open capturing brackets. */
1507    
1508        case OP_CLOSE:
1509        number = GET2(ecode, 1);
1510        offset = number << 1;
1511    
1512    #ifdef PCRE_DEBUG
1513          printf("end bracket %d at *ACCEPT", number);
1514          printf("\n");
1515    #endif
1516    
1517        md->capture_last = number;
1518        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1519          {
1520          md->offset_vector[offset] =
1521            md->offset_vector[md->offset_end - number];
1522          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1523          if (offset_top <= offset) offset_top = offset + 2;
1524          }
1525        ecode += 1 + IMM2_SIZE;
1526        break;
1527    
1528    
1529        /* End of the pattern, either real or forced. */
1530    
1531        case OP_END:
1532        case OP_ACCEPT:
1533        case OP_ASSERT_ACCEPT:
1534    
1535        /* If we have matched an empty string, fail if not in an assertion and not
1536        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1537        is set and we have matched at the start of the subject. In both cases,
1538        backtracking will then try other alternatives, if any. */
1539    
1540        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1541             md->recursive == NULL &&
1542             (md->notempty ||
1543               (md->notempty_atstart &&
1544                 mstart == md->start_subject + md->start_offset)))
1545          RRETURN(MATCH_NOMATCH);
1546    
1547        /* Otherwise, we have a match. */
1548    
1549        md->end_match_ptr = eptr;           /* Record where we ended */
1550        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1551        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1552    
1553        /* For some reason, the macros don't work properly if an expression is
1554        given as the argument to RRETURN when the heap is in use. */
1555    
1556        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1557        RRETURN(rrc);
1558    
1559        /* Assertion brackets. Check the alternative branches in turn - the
1560        matching won't pass the KET for an assertion. If any one branch matches,
1561        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1562        start of each branch to move the current point backwards, so the code at
1563        this level is identical to the lookahead case. When the assertion is part
1564        of a condition, we want to return immediately afterwards. The caller of
1565        this incarnation of the match() function will have set MATCH_CONDASSERT in
1566        md->match_function type, and one of these opcodes will be the first opcode
1567        that is processed. We use a local variable that is preserved over calls to
1568        match() to remember this case. */
1569    
1570        case OP_ASSERT:
1571        case OP_ASSERTBACK:
1572        save_mark = md->mark;
1573        if (md->match_function_type == MATCH_CONDASSERT)
1574          {
1575          condassert = TRUE;
1576          md->match_function_type = 0;
1577          }
1578        else condassert = FALSE;
1579    
1580        do
1581          {
1582          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1583          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1584            {
1585            mstart = md->start_match_ptr;   /* In case \K reset it */
1586            break;
1587            }
1588          md->mark = save_mark;
1589    
1590          /* A COMMIT failure must fail the entire assertion, without trying any
1591          subsequent branches. */
1592    
1593          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1594    
1595          /* PCRE does not allow THEN to escape beyond an assertion; it
1596          is treated as NOMATCH. */
1597    
1598          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1599          ecode += GET(ecode, 1);
1600          }
1601        while (*ecode == OP_ALT);
1602    
1603        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1604    
1605        /* If checking an assertion for a condition, return MATCH_MATCH. */
1606    
1607        if (condassert) RRETURN(MATCH_MATCH);
1608    
1609        /* Continue from after the assertion, updating the offsets high water
1610        mark, since extracts may have been taken during the assertion. */
1611    
1612        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1613        ecode += 1 + LINK_SIZE;
1614        offset_top = md->end_offset_top;
1615        continue;
1616    
1617        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1618        PRUNE, or COMMIT means we must assume failure without checking subsequent
1619        branches. */
1620    
1621        case OP_ASSERT_NOT:
1622        case OP_ASSERTBACK_NOT:
1623        save_mark = md->mark;
1624        if (md->match_function_type == MATCH_CONDASSERT)
1625          {
1626          condassert = TRUE;
1627          md->match_function_type = 0;
1628          }
1629        else condassert = FALSE;
1630    
1631        do
1632          {
1633          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1634          md->mark = save_mark;
1635          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1636          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1637            {
1638            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1639            break;
1640            }
1641    
1642          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1643          as NOMATCH. */
1644    
1645          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1646          ecode += GET(ecode,1);
1647          }
1648        while (*ecode == OP_ALT);
1649    
1650        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1651    
1652        ecode += 1 + LINK_SIZE;
1653        continue;
1654    
1655        /* Move the subject pointer back. This occurs only at the start of
1656        each branch of a lookbehind assertion. If we are too close to the start to
1657        move back, this match function fails. When working with UTF-8 we move
1658        back a number of characters, not bytes. */
1659    
1660        case OP_REVERSE:
1661    #ifdef SUPPORT_UTF
1662        if (utf)
1663          {
1664          i = GET(ecode, 1);
1665          while (i-- > 0)
1666          {          {
1667          eptr--;          eptr--;
1668          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1669          BACKCHAR(eptr)          BACKCHAR(eptr);
1670          }          }
1671        }        }
1672      else      else
# Line 780  for (;;) Line 1675  for (;;)
1675      /* 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 */
1676    
1677        {        {
1678        eptr -= GET(ecode,1);        eptr -= GET(ecode, 1);
1679        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1680        }        }
1681    
1682      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1683    
1684        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1685      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1686      break;      break;
1687    
# Line 794  for (;;) Line 1690  for (;;)
1690      function is able to force a failure. */      function is able to force a failure. */
1691    
1692      case OP_CALLOUT:      case OP_CALLOUT:
1693      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1694        {        {
1695        pcre_callout_block cb;        PUBL(callout_block) cb;
1696        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1697        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1698        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1699        cb.subject          = (const char *)md->start_subject;  #ifdef COMPILE_PCRE8
1700        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1701        cb.start_match      = md->start_match - md->start_subject;  #else
1702        cb.current_position = eptr - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1703    #endif
1704          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1705          cb.start_match      = (int)(mstart - md->start_subject);
1706          cb.current_position = (int)(eptr - md->start_subject);
1707        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1708        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1709        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1710        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1711        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1712        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1713          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1714        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1715        }        }
1716      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 819  for (;;) Line 1720  for (;;)
1720      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
1721      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1722    
1723      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1724      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
1725      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
1726      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
1727      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
1728      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
1729      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.  
1730    
1731      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
1732      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
1733      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1734        a lot, so he is not to blame for the current way it works. */
1735    
1736      case OP_RECURSE:      case OP_RECURSE:
1737        {        {
1738        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1739        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1740    
1741        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1742        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1743            GET2(callpat, 1 + LINK_SIZE);
1744    
1745        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1746          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1747          caught at compile time.) */
1748    
1749          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1750            if (recno == ri->group_num && eptr == ri->subject_position)
1751              RRETURN(PCRE_ERROR_RECURSELOOP);
1752    
1753        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1754    
1755          new_recursive.group_num = recno;
1756          new_recursive.subject_position = eptr;
1757        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1758        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1759    
1760        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1761    
1762        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1763    
1764        /* Now save the offset data. */        /* Now save the offset data */
1765    
1766        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1767        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 863  for (;;) Line 1769  for (;;)
1769        else        else
1770          {          {
1771          new_recursive.offset_save =          new_recursive.offset_save =
1772            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1773          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1774          }          }
   
1775        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1776              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1777    
1778        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1779        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1780          might be changed, so reset it before looping. */
1781    
1782        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1783          cbegroup = (*callpat >= OP_SBRA);
1784        do        do
1785          {          {
1786          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1787              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1788          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1789            memcpy(md->offset_vector, new_recursive.offset_save,
1790                new_recursive.saved_max * sizeof(int));
1791            md->recursive = new_recursive.prevrec;
1792            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1793              {
1794              DPRINTF(("Recursion matched\n"));
1795              if (new_recursive.offset_save != stacksave)
1796                (PUBL(free))(new_recursive.offset_save);
1797    
1798              /* Set where we got to in the subject, and reset the start in case
1799              it was changed by \K. This *is* propagated back out of a recursion,
1800              for Perl compatibility. */
1801    
1802              eptr = md->end_match_ptr;
1803              mstart = md->start_match_ptr;
1804              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1805              }
1806    
1807            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1808            is treated as NOMATCH. */
1809    
1810            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1811                     rrc != MATCH_COMMIT)
1812            {            {
1813            md->recursive = new_recursive.prevrec;            DPRINTF(("Recursion gave error %d\n", rrc));
1814            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1815              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1816            RRETURN(MATCH_MATCH);            RRETURN(rrc);
1817            }            }
         else if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1818    
1819          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1820          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1821          }          }
1822        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 899  for (;;) Line 1824  for (;;)
1824        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1825        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1826        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1827          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1828        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1829        }        }
     /* 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;  
1830    
1831        /* For a non-repeating ket, just continue at this level. This also      RECURSION_MATCHED:
1832        happens for a repeating ket if no characters were matched in the group.      break;
       This is the forcible breaking of infinite loops as implemented in Perl  
       5.005. If there is an options reset, it will get obeyed in the normal  
       course of events. */  
   
       if (*ecode == OP_KET || eptr == saved_eptr)  
         {  
         ecode += 1+LINK_SIZE;  
         break;  
         }  
   
       /* The repeating kets try the rest of the pattern or restart from the  
       preceding bracket, in the appropriate order. We need to reset any options  
       that changed within the bracket before re-running it, so check the next  
       opcode. */  
   
       if (ecode[1+LINK_SIZE] == OP_OPT)  
         {  
         ims = (ims & ~PCRE_IMS) | ecode[4];  
         DPRINTF(("ims set to %02lx at group repeat\n", ims));  
         }  
   
       if (*ecode == OP_KETRMIN)  
         {  
         RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         }  
       else  /* OP_KETRMAX */  
         {  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         }  
       }  
     RRETURN(MATCH_NOMATCH);  
1833    
1834      /* 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
1835      bracketed group and go to there. */      bracketed group and go to there. */
# Line 985  for (;;) Line 1838  for (;;)
1838      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1839      break;      break;
1840    
1841      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1842      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
1843      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
1844      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
1845      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1846    
1847      case OP_BRAZERO:      case OP_BRAZERO:
1848        {      next = ecode + 1;
1849        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1850        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1851        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1852        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1853      break;      break;
1854    
1855      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1856        {      next = ecode + 1;
1857        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1858        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1859        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1860          match_isgroup);      ecode++;
1861        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1862        ecode++;  
1863        }      case OP_SKIPZERO:
1864        next = ecode+1;
1865        do next += GET(next,1); while (*next == OP_ALT);
1866        ecode = next + 1 + LINK_SIZE;
1867      break;      break;
1868    
1869      /* 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
1870      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1871      current high water mark for use by positive assertions. Do this also  
1872      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1873        op = *(++ecode);
1874        allow_zero = TRUE;
1875        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1876          goto POSSESSIVE_NON_CAPTURE;
1877    
1878        /* End of a group, repeated or non-repeating. */
1879    
1880      case OP_KET:      case OP_KET:
1881      case OP_KETRMIN:      case OP_KETRMIN:
1882      case OP_KETRMAX:      case OP_KETRMAX:
1883        {      case OP_KETRPOS:
1884        prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
       saved_eptr = eptrb->epb_saved_eptr;  
1885    
1886        /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1887        infinite repeats of empty string matches, retrieve the subject start from
1888        the chain. Otherwise, set it NULL. */
1889    
1890        eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1891          {
1892        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1893            *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1894            *prev == OP_ONCE)        }
1895          {      else saved_eptr = NULL;
         md->end_match_ptr = eptr;      /* For ONCE */  
         md->end_offset_top = offset_top;  
         RRETURN(MATCH_MATCH);  
         }  
1896    
1897        /* 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
1898        group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1899        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
1900        start in case it was changed by \K. */
1901    
1902        if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1903          {           *prev == OP_ONCE_NC)
1904          number = *prev - OP_BRA;        {
1905          md->end_match_ptr = eptr;      /* For ONCE_NC */
1906          md->end_offset_top = offset_top;
1907          md->start_match_ptr = mstart;
1908          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1909          }
1910    
1911          /* 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
1912          the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1913        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1914        into group 0, so it won't be picked up here. Instead, we catch it when the
1915        OP_END is reached. Other recursion is handled here. We just have to record
1916        the current subject position and start match pointer and give a MATCH
1917        return. */
1918    
1919          if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1920          offset = number << 1;          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1921          {
1922          number = GET2(prev, 1+LINK_SIZE);
1923          offset = number << 1;
1924    
1925  #ifdef DEBUG  #ifdef PCRE_DEBUG
1926          printf("end bracket %d", number);        printf("end bracket %d", number);
1927          printf("\n");        printf("\n");
1928  #endif  #endif
1929    
1930          /* 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. */  
1931    
1932          if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1933            {          {
1934            md->capture_last = number;          md->end_match_ptr = eptr;
1935            if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1936              {          RRETURN(MATCH_MATCH);
1937              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;  
             }  
1938    
1939            /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
           appropriately and continue from after the call. */  
1940    
1941            if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = number;
1942              {        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1943              recursion_info *rec = md->recursive;          {
1944              DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          /* If offset is greater than offset_top, it means that we are
1945              md->recursive = rec->prevrec;          "skipping" a capturing group, and that group's offsets must be marked
1946              md->start_match = rec->save_start;          unset. In earlier versions of PCRE, all the offsets were unset at the
1947              memcpy(md->offset_vector, rec->offset_save,          start of matching, but this doesn't work because atomic groups and
1948                rec->saved_max * sizeof(int));          assertions can cause a value to be set that should later be unset.
1949              ecode = rec->after_call;          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1950              ims = original_ims;          part of the atomic group, but this is not on the final matching path,
1951              break;          so must be unset when 2 is set. (If there is no group 2, there is no
1952              }          problem, because offset_top will then be 2, indicating no capture.) */
1953    
1954            if (offset > offset_top)
1955              {
1956              register int *iptr = md->offset_vector + offset_top;
1957              register int *iend = md->offset_vector + offset;
1958              while (iptr < iend) *iptr++ = -1;
1959            }            }
         }  
1960    
1961        /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
       the group. */  
1962    
1963        ims = original_ims;          md->offset_vector[offset] =
1964        DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1965            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1966            if (offset_top <= offset) offset_top = offset + 2;
1967            }
1968          }
1969    
1970        /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1971        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
1972        This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1973        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,
1974        course of events. */      establish a backup point by processing the rest of the pattern at a lower
1975        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1976        original OP_ONCE level, thereby bypassing intermediate backup points, but
1977        resetting any captures that happened along the way. */
1978    
1979        if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1980          {
1981          if (*prev == OP_ONCE)
1982          {          {
1983          ecode += 1 + LINK_SIZE;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1984          break;          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1985            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1986            RRETURN(MATCH_ONCE);
1987          }          }
1988          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1989          break;
1990          }
1991    
1992        /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1993        preceding bracket, in the appropriate order. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1994        at a time from the outer level, thus saving stack. */
1995    
1996        if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRPOS)
1997          {
1998          md->end_match_ptr = eptr;
1999          md->end_offset_top = offset_top;
2000          RRETURN(MATCH_KETRPOS);
2001          }
2002    
2003        /* The normal repeating kets try the rest of the pattern or restart from
2004        the preceding bracket, in the appropriate order. In the second case, we can
2005        use tail recursion to avoid using another stack frame, unless we have an
2006        an atomic group or an unlimited repeat of a group that can match an empty
2007        string. */
2008    
2009        if (*ecode == OP_KETRMIN)
2010          {
2011          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2012          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2013          if (*prev == OP_ONCE)
2014          {          {
2015          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);  
2016          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2017            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2018            RRETURN(MATCH_ONCE);
2019          }          }
2020        else  /* OP_KETRMAX */        if (*prev >= OP_SBRA)    /* Could match an empty string */
2021          {          {
2022          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2023          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RRETURN(rrc);
2024          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          }
2025          ecode = prev;
2026          goto TAIL_RECURSE;
2027          }
2028        else  /* OP_KETRMAX */
2029          {
2030          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2031          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2032          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2033          if (*prev == OP_ONCE)
2034            {
2035            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2036          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2037            md->once_target = prev;
2038            RRETURN(MATCH_ONCE);
2039          }          }
2040          ecode += 1 + LINK_SIZE;
2041          goto TAIL_RECURSE;
2042        }        }
2043        /* Control never gets here */
2044    
2045      RRETURN(MATCH_NOMATCH);      /* Not multiline mode: start of subject assertion, unless notbol. */
   
     /* Start of subject unless notbol, or after internal newline if multiline */  
2046    
2047      case OP_CIRC:      case OP_CIRC:
2048      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 */  
2049    
2050      /* Start of subject assertion */      /* Start of subject assertion */
2051    
# Line 1149  for (;;) Line 2054  for (;;)
2054      ecode++;      ecode++;
2055      break;      break;
2056    
2057        /* Multiline mode: start of subject unless notbol, or after any newline. */
2058    
2059        case OP_CIRCM:
2060        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2061        if (eptr != md->start_subject &&
2062            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2063          RRETURN(MATCH_NOMATCH);
2064        ecode++;
2065        break;
2066    
2067      /* Start of match assertion */      /* Start of match assertion */
2068    
2069      case OP_SOM:      case OP_SOM:
# Line 1156  for (;;) Line 2071  for (;;)
2071      ecode++;      ecode++;
2072      break;      break;
2073    
2074      /* 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. */  
2075    
2076      case OP_DOLL:      case OP_SET_SOM:
2077      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2078        ecode++;
2079        break;
2080    
2081        /* Multiline mode: assert before any newline, or before end of subject
2082        unless noteol is set. */
2083    
2084        case OP_DOLLM:
2085        if (eptr < md->end_subject)
2086        {        {
2087        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2088          { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }          {
2089        else          if (md->partial != 0 &&
2090          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2091        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2092        break;              NLBLOCK->nllen == 2 &&
2093                *eptr == NLBLOCK->nl[0])
2094              {
2095              md->hitend = TRUE;
2096              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2097              }
2098            RRETURN(MATCH_NOMATCH);
2099            }
2100        }        }
2101      else      else
2102        {        {
2103        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2104        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr < md->end_subject - 1 ||  
            (eptr == md->end_subject - 1 && *eptr != NEWLINE))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2105        }        }
2106      /* ... else fall through */      ecode++;
2107        break;
2108    
2109        /* Not multiline mode: assert before a terminating newline or before end of
2110        subject unless noteol is set. */
2111    
2112        case OP_DOLL:
2113        if (md->noteol) RRETURN(MATCH_NOMATCH);
2114        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2115    
2116        /* ... else fall through for endonly */
2117    
2118      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2119    
2120      case OP_EOD:      case OP_EOD:
2121      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2122        SCHECK_PARTIAL();
2123      ecode++;      ecode++;
2124      break;      break;
2125    
2126      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2127    
2128      case OP_EODN:      case OP_EODN:
2129      if (eptr < md->end_subject - 1 ||      ASSERT_NL_OR_EOS:
2130         (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject &&
2131            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2132          {
2133          if (md->partial != 0 &&
2134              eptr + 1 >= md->end_subject &&
2135              NLBLOCK->nltype == NLTYPE_FIXED &&
2136              NLBLOCK->nllen == 2 &&
2137              *eptr == NLBLOCK->nl[0])
2138            {
2139            md->hitend = TRUE;
2140            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2141            }
2142          RRETURN(MATCH_NOMATCH);
2143          }
2144    
2145        /* Either at end of string or \n before end. */
2146    
2147        SCHECK_PARTIAL();
2148      ecode++;      ecode++;
2149      break;      break;
2150    
# Line 1206  for (;;) Line 2156  for (;;)
2156    
2157        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2158        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
2159        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2160          partial matching. */
2161    
2162  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2163        if (utf8)        if (utf)
2164          {          {
2165            /* Get status of previous character */
2166    
2167          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2168            {            {
2169            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2170            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2171              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2172            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2173    #ifdef SUPPORT_UCP
2174              if (md->use_ucp)
2175                {
2176                if (c == '_') prev_is_word = TRUE; else
2177                  {
2178                  int cat = UCD_CATEGORY(c);
2179                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2180                  }
2181                }
2182              else
2183    #endif
2184            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2185            }            }
2186          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2187            /* Get status of next character */
2188    
2189            if (eptr >= md->end_subject)
2190              {
2191              SCHECK_PARTIAL();
2192              cur_is_word = FALSE;
2193              }
2194            else
2195            {            {
2196            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2197    #ifdef SUPPORT_UCP
2198              if (md->use_ucp)
2199                {
2200                if (c == '_') cur_is_word = TRUE; else
2201                  {
2202                  int cat = UCD_CATEGORY(c);
2203                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2204                  }
2205                }
2206              else
2207    #endif
2208            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2209            }            }
2210          }          }
2211        else        else
2212  #endif  #endif
2213    
2214        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2215          consistency with the behaviour of \w we do use it in this case. */
2216    
2217          {          {
2218          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2219            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2220          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2221            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2222              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2223    #ifdef SUPPORT_UCP
2224              if (md->use_ucp)
2225                {
2226                c = eptr[-1];
2227                if (c == '_') prev_is_word = TRUE; else
2228                  {
2229                  int cat = UCD_CATEGORY(c);
2230                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2231                  }
2232                }
2233              else
2234    #endif
2235              prev_is_word = MAX_255(eptr[-1])
2236                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2237              }
2238    
2239            /* Get status of next character */
2240    
2241            if (eptr >= md->end_subject)
2242              {
2243              SCHECK_PARTIAL();
2244              cur_is_word = FALSE;
2245              }
2246            else
2247    #ifdef SUPPORT_UCP
2248            if (md->use_ucp)
2249              {
2250              c = *eptr;
2251              if (c == '_') cur_is_word = TRUE; else
2252                {
2253                int cat = UCD_CATEGORY(c);
2254                cur_is_word = (cat == ucp_L || cat == ucp_N);
2255                }
2256              }
2257            else
2258    #endif
2259            cur_is_word = MAX_255(*eptr)
2260              && ((md->ctypes[*eptr] & ctype_word) != 0);
2261          }          }
2262    
2263        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1244  for (;;) Line 2268  for (;;)
2268        }        }
2269      break;      break;
2270    
2271      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2272        CRLF newlines and partial matching. */
2273    
2274      case OP_ANY:      case OP_ANY:
2275      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2276        if (md->partial != 0 &&
2277            eptr + 1 >= md->end_subject &&
2278            NLBLOCK->nltype == NLTYPE_FIXED &&
2279            NLBLOCK->nllen == 2 &&
2280            *eptr == NLBLOCK->nl[0])
2281          {
2282          md->hitend = TRUE;
2283          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2284          }
2285    
2286        /* Fall through */
2287    
2288        /* Match any single character whatsoever. */
2289    
2290        case OP_ALLANY:
2291        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2292          {                            /* not be updated before SCHECK_PARTIAL. */
2293          SCHECK_PARTIAL();
2294        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2295      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);        }
2296  #ifdef SUPPORT_UTF8      eptr++;
2297      if (utf8)  #ifdef SUPPORT_UTF
2298        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2299  #endif  #endif
2300      ecode++;      ecode++;
2301      break;      break;
# Line 1261  for (;;) Line 2304  for (;;)
2304      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2305    
2306      case OP_ANYBYTE:      case OP_ANYBYTE:
2307      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2308          {                            /* not be updated before SCHECK_PARTIAL. */
2309          SCHECK_PARTIAL();
2310          RRETURN(MATCH_NOMATCH);
2311          }
2312        eptr++;
2313      ecode++;      ecode++;
2314      break;      break;
2315    
2316      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2317      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2318          {
2319          SCHECK_PARTIAL();
2320          RRETURN(MATCH_NOMATCH);
2321          }
2322      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2323      if (      if (
2324  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2325         c < 256 &&         c < 256 &&
2326  #endif  #endif
2327         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1279  for (;;) Line 2331  for (;;)
2331      break;      break;
2332    
2333      case OP_DIGIT:      case OP_DIGIT:
2334      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2335          {
2336          SCHECK_PARTIAL();
2337          RRETURN(MATCH_NOMATCH);
2338          }
2339      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2340      if (      if (
2341  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2342         c >= 256 ||         c > 255 ||
2343  #endif  #endif
2344         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2345         )         )
# Line 1292  for (;;) Line 2348  for (;;)
2348      break;      break;
2349    
2350      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2351      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2352          {
2353          SCHECK_PARTIAL();
2354          RRETURN(MATCH_NOMATCH);
2355          }
2356      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2357      if (      if (
2358  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2359         c < 256 &&         c < 256 &&
2360  #endif  #endif
2361         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1305  for (;;) Line 2365  for (;;)
2365      break;      break;
2366    
2367      case OP_WHITESPACE:      case OP_WHITESPACE:
2368      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2369          {
2370          SCHECK_PARTIAL();
2371          RRETURN(MATCH_NOMATCH);
2372          }
2373      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2374      if (      if (
2375  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2376         c >= 256 ||         c > 255 ||
2377  #endif  #endif
2378         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2379         )         )
# Line 1318  for (;;) Line 2382  for (;;)
2382      break;      break;
2383    
2384      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2385      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2386          {
2387          SCHECK_PARTIAL();
2388          RRETURN(MATCH_NOMATCH);
2389          }
2390      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2391      if (      if (
2392  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2393         c < 256 &&         c < 256 &&
2394  #endif  #endif
2395         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1331  for (;;) Line 2399  for (;;)
2399      break;      break;
2400    
2401      case OP_WORDCHAR:      case OP_WORDCHAR:
2402      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2403          {
2404          SCHECK_PARTIAL();
2405          RRETURN(MATCH_NOMATCH);
2406          }
2407      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2408      if (      if (
2409  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2410         c >= 256 ||         c > 255 ||
2411  #endif  #endif
2412         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2413         )         )
# Line 1343  for (;;) Line 2415  for (;;)
2415      ecode++;      ecode++;
2416      break;      break;
2417    
2418        case OP_ANYNL:
2419        if (eptr >= md->end_subject)
2420          {
2421          SCHECK_PARTIAL();
2422          RRETURN(MATCH_NOMATCH);
2423          }
2424        GETCHARINCTEST(c, eptr);
2425        switch(c)
2426          {
2427          default: RRETURN(MATCH_NOMATCH);
2428    
2429          case CHAR_CR:
2430          if (eptr >= md->end_subject)
2431            {
2432            SCHECK_PARTIAL();
2433            }
2434          else if (*eptr == CHAR_LF) eptr++;
2435          break;
2436    
2437          case CHAR_LF:
2438          break;
2439    
2440          case CHAR_VT:
2441          case CHAR_FF:
2442          case CHAR_NEL:
2443    #ifndef EBCDIC
2444          case 0x2028:
2445          case 0x2029:
2446    #endif  /* Not EBCDIC */
2447          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2448          break;
2449          }
2450        ecode++;
2451        break;
2452    
2453        case OP_NOT_HSPACE:
2454        if (eptr >= md->end_subject)
2455          {
2456          SCHECK_PARTIAL();
2457          RRETURN(MATCH_NOMATCH);
2458          }
2459        GETCHARINCTEST(c, eptr);
2460        switch(c)
2461          {
2462          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2463          default: break;
2464          }
2465        ecode++;
2466        break;
2467    
2468        case OP_HSPACE:
2469        if (eptr >= md->end_subject)
2470          {
2471          SCHECK_PARTIAL();
2472          RRETURN(MATCH_NOMATCH);
2473          }
2474        GETCHARINCTEST(c, eptr);
2475        switch(c)
2476          {
2477          HSPACE_CASES: break;  /* Byte and multibyte cases */
2478          default: RRETURN(MATCH_NOMATCH);
2479          }
2480        ecode++;
2481        break;
2482    
2483        case OP_NOT_VSPACE:
2484        if (eptr >= md->end_subject)
2485          {
2486          SCHECK_PARTIAL();
2487          RRETURN(MATCH_NOMATCH);
2488          }
2489        GETCHARINCTEST(c, eptr);
2490        switch(c)
2491          {
2492          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2493          default: break;
2494          }
2495        ecode++;
2496        break;
2497    
2498        case OP_VSPACE:
2499        if (eptr >= md->end_subject)
2500          {
2501          SCHECK_PARTIAL();
2502          RRETURN(MATCH_NOMATCH);
2503          }
2504        GETCHARINCTEST(c, eptr);
2505        switch(c)
2506          {
2507          VSPACE_CASES: break;
2508          default: RRETURN(MATCH_NOMATCH);
2509          }
2510        ecode++;
2511        break;
2512    
2513  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2514      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2515      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. */
2516    
2517      case OP_PROP:      case OP_PROP:
2518      case OP_NOTPROP:      case OP_NOTPROP:
2519      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2520          {
2521          SCHECK_PARTIAL();
2522          RRETURN(MATCH_NOMATCH);
2523          }
2524      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2525        {        {
2526        int chartype, rqdtype;        const pcre_uint32 *cp;
2527        int othercase;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findchar(c, &chartype, &othercase);  
   
       rqdtype = *(++ecode);  
       ecode++;  
2528    
2529        if (rqdtype >= 128)        switch(ecode[1])
2530          {          {
2531          if ((rqdtype - 128 != category) == (op == OP_PROP))          case PT_ANY:
2532            if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2533            break;
2534    
2535            case PT_LAMP:
2536            if ((prop->chartype == ucp_Lu ||
2537                 prop->chartype == ucp_Ll ||
2538                 prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2539            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2540          }          break;
2541        else  
2542          {          case PT_GC:
2543          if ((rqdtype != chartype) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2544              RRETURN(MATCH_NOMATCH);
2545            break;
2546    
2547            case PT_PC:
2548            if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2549              RRETURN(MATCH_NOMATCH);
2550            break;
2551    
2552            case PT_SC:
2553            if ((ecode[2] != prop->script) == (op == OP_PROP))
2554              RRETURN(MATCH_NOMATCH);
2555            break;
2556    
2557            /* These are specials */
2558    
2559            case PT_ALNUM:
2560            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2561                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2562              RRETURN(MATCH_NOMATCH);
2563            break;
2564    
2565            case PT_SPACE:    /* Perl space */
2566            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2567                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2568                   == (op == OP_NOTPROP))
2569            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2570            break;
2571    
2572            case PT_PXSPACE:  /* POSIX space */
2573            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2574                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2575                 c == CHAR_FF || c == CHAR_CR)
2576                   == (op == OP_NOTPROP))
2577              RRETURN(MATCH_NOMATCH);
2578            break;
2579    
2580            case PT_WORD:
2581            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2582                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2583                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2584              RRETURN(MATCH_NOMATCH);
2585            break;
2586    
2587            case PT_CLIST:
2588            cp = PRIV(ucd_caseless_sets) + prop->caseset;
2589            for (;;)
2590              {
2591              if (c < *cp)
2592                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2593              if (c == *cp++)
2594                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2595              }
2596            break;
2597    
2598            /* This should never occur */
2599    
2600            default:
2601            RRETURN(PCRE_ERROR_INTERNAL);
2602          }          }
2603    
2604          ecode += 3;
2605        }        }
2606      break;      break;
2607    
# Line 1376  for (;;) Line 2609  for (;;)
2609      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2610    
2611      case OP_EXTUNI:      case OP_EXTUNI:
2612      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
2613        {        {
2614        int chartype;        SCHECK_PARTIAL();
2615        int othercase;        RRETURN(MATCH_NOMATCH);
2616        int category = _pcre_ucp_findchar(c, &chartype, &othercase);        }
2617        if (category == ucp_M) RRETURN(MATCH_NOMATCH);      else
2618          {
2619          int lgb, rgb;
2620          GETCHARINCTEST(c, eptr);
2621          lgb = UCD_GRAPHBREAK(c);
2622        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2623          {          {
2624          int len = 1;          int len = 1;
2625          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2626            {          rgb = UCD_GRAPHBREAK(c);
2627            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2628            }          lgb = rgb;
         category = _pcre_ucp_findchar(c, &chartype, &othercase);  
         if (category != ucp_M) break;  
2629          eptr += len;          eptr += len;
2630          }          }
2631        }        }
2632        CHECK_PARTIAL();
2633      ecode++;      ecode++;
2634      break;      break;
2635  #endif  #endif  /* SUPPORT_UCP */
2636    
2637    
2638      /* Match a back reference, possibly repeatedly. Look past the end of the      /* Match a back reference, possibly repeatedly. Look past the end of the
# Line 1409  for (;;) Line 2644  for (;;)
2644      loops). */      loops). */
2645    
2646      case OP_REF:      case OP_REF:
2647        {      case OP_REFI:
2648        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2649        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2650        ecode += 1 + IMM2_SIZE;
2651    
2652        /* 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];  
2653    
2654        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2655        this ensures that every attempt at a match fails. We can't just fail
2656        here, because of the possibility of quantifiers with zero minima.
2657    
2658        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2659          {      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;  
2660    
2661          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2662          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;  
2663    
2664          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2665          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2666          eptr += length;      else
2667          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2668    
2669        /* Set up for repetition, or handle the non-repeated case */
2670    
2671        switch (*ecode)
2672          {
2673          case OP_CRSTAR:
2674          case OP_CRMINSTAR:
2675          case OP_CRPLUS:
2676          case OP_CRMINPLUS:
2677          case OP_CRQUERY:
2678          case OP_CRMINQUERY:
2679          c = *ecode++ - OP_CRSTAR;
2680          minimize = (c & 1) != 0;
2681          min = rep_min[c];                 /* Pick up values from tables; */
2682          max = rep_max[c];                 /* zero for max => infinity */
2683          if (max == 0) max = INT_MAX;
2684          break;
2685    
2686          case OP_CRRANGE:
2687          case OP_CRMINRANGE:
2688          minimize = (*ecode == OP_CRMINRANGE);
2689          min = GET2(ecode, 1);
2690          max = GET2(ecode, 1 + IMM2_SIZE);
2691          if (max == 0) max = INT_MAX;
2692          ecode += 1 + 2 * IMM2_SIZE;
2693          break;
2694    
2695          default:               /* No repeat follows */
2696          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2697            {
2698            if (length == -2) eptr = md->end_subject;   /* Partial match */
2699            CHECK_PARTIAL();
2700            RRETURN(MATCH_NOMATCH);
2701          }          }
2702          eptr += length;
2703          continue;              /* With the main loop */
2704          }
2705    
2706        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2707        main loop. */      zero, just continue with the main loop. If the length is negative, it
2708        means the reference is unset in non-Java-compatible mode. If the minimum is
2709        zero, we can continue at the same level without recursion. For any other
2710        minimum, carrying on will result in NOMATCH. */
2711    
2712        if (length == 0) continue;      if (length == 0) continue;
2713        if (length < 0 && min == 0) continue;
2714    
2715        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2716        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2717        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2718    
2719        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2720          {
2721          int slength;
2722          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2723          {          {
2724          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2725          eptr += length;          CHECK_PARTIAL();
2726            RRETURN(MATCH_NOMATCH);
2727          }          }
2728          eptr += slength;
2729          }
2730    
2731        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2732        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2733    
2734        if (min == max) continue;      if (min == max) continue;
2735    
2736        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2737    
2738        if (minimize)      if (minimize)
2739          {
2740          for (fi = min;; fi++)
2741          {          {
2742          for (fi = min;; fi++)          int slength;
2743            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2744            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2745            if (fi >= max) RRETURN(MATCH_NOMATCH);
2746            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2747            {            {
2748            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2749            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2750            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2751            }            }
2752          /* Control never gets here */          eptr += slength;
2753          }          }
2754          /* Control never gets here */
2755          }
2756    
2757        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2758    
2759        else      else
2760          {
2761          pp = eptr;
2762          for (i = min; i < max; i++)
2763          {          {
2764          pp = eptr;          int slength;
2765          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)  
2766            {            {
2767            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2768            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2769            eptr -= length;  
2770              if (slength == -2 && md->partial != 0 &&
2771                  md->end_subject > md->start_used_ptr)
2772                {
2773                md->hitend = TRUE;
2774                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2775                }
2776              break;
2777            }            }
2778          RRETURN(MATCH_NOMATCH);          eptr += slength;
2779          }          }
2780    
2781          while (eptr >= pp)
2782            {
2783            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2784            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2785            eptr -= length;
2786            }
2787          RRETURN(MATCH_NOMATCH);
2788        }        }
2789      /* Control never gets here */      /* Control never gets here */
2790    
   
   
2791      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2792      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,
2793      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 2802  for (;;)
2802      case OP_NCLASS:      case OP_NCLASS:
2803      case OP_CLASS:      case OP_CLASS:
2804        {        {
2805          /* The data variable is saved across frames, so the byte map needs to
2806          be stored there. */
2807    #define BYTE_MAP ((pcre_uint8 *)data)
2808        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2809        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2810    
2811        switch (*ecode)        switch (*ecode)
2812          {          {
# Line 1548  for (;;) Line 2827  for (;;)
2827          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2828          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2829          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2830          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2831          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2832          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2833          break;          break;
2834    
2835          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1560  for (;;) Line 2839  for (;;)
2839    
2840        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2841    
2842  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2843        /* UTF-8 mode */        if (utf)
       if (utf8)  
2844          {          {
2845          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2846            {            {
2847            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2848                {
2849                SCHECK_PARTIAL();
2850                RRETURN(MATCH_NOMATCH);
2851                }
2852            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2853            if (c > 255)            if (c > 255)
2854              {              {
2855              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2856              }              }
2857            else            else
2858              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2859            }            }
2860          }          }
2861        else        else
2862  #endif  #endif
2863        /* Not UTF-8 mode */        /* Not UTF mode */
2864          {          {
2865          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2866            {            {
2867            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2868                {
2869                SCHECK_PARTIAL();
2870                RRETURN(MATCH_NOMATCH);
2871                }
2872            c = *eptr++;            c = *eptr++;
2873            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2874              if (c > 255)
2875                {
2876                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2877                }
2878              else
2879    #endif
2880                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2881            }            }
2882          }          }
2883    
# Line 1600  for (;;) Line 2891  for (;;)
2891    
2892        if (minimize)        if (minimize)
2893          {          {
2894  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2895          /* UTF-8 mode */          if (utf)
         if (utf8)  
2896            {            {
2897            for (fi = min;; fi++)            for (fi = min;; fi++)
2898              {              {
2899              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2900              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2901              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2902                if (eptr >= md->end_subject)
2903                  {
2904                  SCHECK_PARTIAL();
2905                  RRETURN(MATCH_NOMATCH);
2906                  }
2907              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2908              if (c > 255)              if (c > 255)
2909                {                {
2910                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2911                }                }
2912              else              else
2913                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2914              }              }
2915            }            }
2916          else          else
2917  #endif  #endif
2918          /* Not UTF-8 mode */          /* Not UTF mode */
2919            {            {
2920            for (fi = min;; fi++)            for (fi = min;; fi++)
2921              {              {
2922              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2923              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2924              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2925                if (eptr >= md->end_subject)
2926                  {
2927                  SCHECK_PARTIAL();
2928                  RRETURN(MATCH_NOMATCH);
2929                  }
2930              c = *eptr++;              c = *eptr++;
2931              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2932                if (c > 255)
2933                  {
2934                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2935                  }
2936                else
2937    #endif
2938                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2939              }              }
2940            }            }
2941          /* Control never gets here */          /* Control never gets here */
# Line 1642  for (;;) Line 2947  for (;;)
2947          {          {
2948          pp = eptr;          pp = eptr;
2949    
2950  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2951          /* UTF-8 mode */          if (utf)
         if (utf8)  
2952            {            {
2953            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2954              {              {
2955              int len = 1;              int len = 1;
2956              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2957                  {
2958                  SCHECK_PARTIAL();
2959                  break;
2960                  }
2961              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2962              if (c > 255)              if (c > 255)
2963                {                {
2964                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2965                }                }
2966              else              else
2967                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2968              eptr += len;              eptr += len;
2969              }              }
2970            for (;;)            for (;;)
2971              {              {
2972              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2973              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2974              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2975              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1671  for (;;) Line 2977  for (;;)
2977            }            }
2978          else          else
2979  #endif  #endif
2980            /* Not UTF-8 mode */            /* Not UTF mode */
2981            {            {
2982            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2983              {              {
2984              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2985                  {
2986                  SCHECK_PARTIAL();
2987                  break;
2988                  }
2989              c = *eptr;              c = *eptr;
2990              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2991                if (c > 255)
2992                  {
2993                  if (op == OP_CLASS) break;
2994                  }
2995                else
2996    #endif
2997                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2998              eptr++;              eptr++;
2999              }              }
3000            while (eptr >= pp)            while (eptr >= pp)
3001              {              {
3002              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
             eptr--;  
3003              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3004                eptr--;
3005              }              }
3006            }            }
3007    
3008          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3009          }          }
3010    #undef BYTE_MAP
3011        }        }
3012      /* Control never gets here */      /* Control never gets here */
3013    
3014    
3015      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3016      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
3017        mode, because Unicode properties are supported in non-UTF-8 mode. */
3018    
3019  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3020      case OP_XCLASS:      case OP_XCLASS:
3021        {        {
3022        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1722  for (;;) Line 3041  for (;;)
3041          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3042          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3043          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3044          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3045          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3046          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3047          break;          break;
3048    
3049          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1736  for (;;) Line 3055  for (;;)
3055    
3056        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3057          {          {
3058          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3059          GETCHARINC(c, eptr);            {
3060          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3061              RRETURN(MATCH_NOMATCH);
3062              }
3063            GETCHARINCTEST(c, eptr);
3064            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3065          }          }
3066    
3067        /* 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 3076  for (;;)
3076          {          {
3077          for (fi = min;; fi++)          for (fi = min;; fi++)
3078            {            {
3079            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3080            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3081            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3082            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3083            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3084                SCHECK_PARTIAL();
3085                RRETURN(MATCH_NOMATCH);
3086                }
3087              GETCHARINCTEST(c, eptr);
3088              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3089            }            }
3090          /* Control never gets here */          /* Control never gets here */
3091          }          }
# Line 1770  for (;;) Line 3098  for (;;)
3098          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3099            {            {
3100            int len = 1;            int len = 1;
3101            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3102            GETCHARLEN(c, eptr, len);              {
3103            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3104                break;
3105                }
3106    #ifdef SUPPORT_UTF
3107              GETCHARLENTEST(c, eptr, len);
3108    #else
3109              c = *eptr;
3110    #endif
3111              if (!PRIV(xclass)(c, data, utf)) break;
3112            eptr += len;            eptr += len;
3113            }            }
3114          for(;;)          for(;;)
3115            {            {
3116            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3117            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3118            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3119            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3120              if (utf) BACKCHAR(eptr);
3121    #endif
3122            }            }
3123          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3124          }          }
# Line 1792  for (;;) Line 3130  for (;;)
3130      /* Match a single character, casefully */      /* Match a single character, casefully */
3131    
3132      case OP_CHAR:      case OP_CHAR:
3133  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3134      if (utf8)      if (utf)
3135        {        {
3136        length = 1;        length = 1;
3137        ecode++;        ecode++;
3138        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3139        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3140            {
3141            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3142            RRETURN(MATCH_NOMATCH);
3143            }
3144        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3145        }        }
3146      else      else
3147  #endif  #endif
3148        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3149        {        {
3150        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3151            {
3152            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3153            RRETURN(MATCH_NOMATCH);
3154            }
3155        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3156        ecode += 2;        ecode += 2;
3157        }        }
3158      break;      break;
3159    
3160      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3161        subject, give up immediately. */
3162    
3163      case OP_CHARNC:      case OP_CHARI:
3164  #ifdef SUPPORT_UTF8      if (eptr >= md->end_subject)
3165      if (utf8)        {
3166          SCHECK_PARTIAL();
3167          RRETURN(MATCH_NOMATCH);
3168          }
3169    
3170    #ifdef SUPPORT_UTF
3171        if (utf)
3172        {        {
3173        length = 1;        length = 1;
3174        ecode++;        ecode++;
3175        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3176    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3177        /* 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
3178        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3179          fast lookup table. We know that there is at least one byte left in the
3180          subject. */
3181    
3182        if (fc < 128)        if (fc < 128)
3183          {          {
3184          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3185                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3186            ecode++;
3187            eptr++;
3188          }          }
3189    
3190        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3191          use the value of "length" to check for sufficient bytes left, because the
3192          other case of the character may have more or fewer bytes.  */
3193    
3194        else        else
3195          {          {
3196          int dc;          unsigned int dc;
3197          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3198          ecode += length;          ecode += length;
3199    
3200          /* 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
3201          case of the character, if there is one. The result of _pcre_ucp_findchar() is          case of the character, if there is one. */
         < 0 if the char isn't found, and othercase is returned as zero if there  
         isn't one. */  
3202    
3203          if (fc != dc)          if (fc != dc)
3204            {            {
3205  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3206            int chartype;            if (dc != UCD_OTHERCASE(fc))
           int othercase;  
           if (_pcre_ucp_findchar(fc, &chartype, &othercase) < 0 || dc != othercase)  
3207  #endif  #endif
3208              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3209            }            }
3210          }          }
3211        }        }
3212      else      else
3213  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3214    
3215      /* Non-UTF-8 mode */      /* Not UTF mode */
3216        {        {
3217        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3218        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3219          eptr++;
3220        ecode += 2;        ecode += 2;
3221        }        }
3222      break;      break;
3223    
3224      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3225    
3226      case OP_EXACT:      case OP_EXACT:
3227        case OP_EXACTI:
3228      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3229      ecode += 3;      ecode += 1 + IMM2_SIZE;
3230      goto REPEATCHAR;      goto REPEATCHAR;
3231    
3232        case OP_POSUPTO:
3233        case OP_POSUPTOI:
3234        possessive = TRUE;
3235        /* Fall through */
3236    
3237      case OP_UPTO:      case OP_UPTO:
3238        case OP_UPTOI:
3239      case OP_MINUPTO:      case OP_MINUPTO:
3240        case OP_MINUPTOI:
3241      min = 0;      min = 0;
3242      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3243      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3244      ecode += 3;      ecode += 1 + IMM2_SIZE;
3245        goto REPEATCHAR;
3246    
3247        case OP_POSSTAR:
3248        case OP_POSSTARI:
3249        possessive = TRUE;
3250        min = 0;
3251        max = INT_MAX;
3252        ecode++;
3253        goto REPEATCHAR;
3254    
3255        case OP_POSPLUS:
3256        case OP_POSPLUSI:
3257        possessive = TRUE;
3258        min = 1;
3259        max = INT_MAX;
3260        ecode++;
3261        goto REPEATCHAR;
3262    
3263        case OP_POSQUERY:
3264        case OP_POSQUERYI:
3265        possessive = TRUE;
3266        min = 0;
3267        max = 1;
3268        ecode++;
3269      goto REPEATCHAR;      goto REPEATCHAR;
3270    
3271      case OP_STAR:      case OP_STAR:
3272        case OP_STARI:
3273      case OP_MINSTAR:      case OP_MINSTAR:
3274        case OP_MINSTARI:
3275      case OP_PLUS:      case OP_PLUS:
3276        case OP_PLUSI:
3277      case OP_MINPLUS:      case OP_MINPLUS:
3278        case OP_MINPLUSI:
3279      case OP_QUERY:      case OP_QUERY:
3280        case OP_QUERYI:
3281      case OP_MINQUERY:      case OP_MINQUERY:
3282      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3283        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3284      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3285      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3286      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3287      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3288    
3289      /* 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. */  
3290    
3291      REPEATCHAR:      REPEATCHAR:
3292  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3293      if (utf8)      if (utf)
3294        {        {
3295        length = 1;        length = 1;
3296        charptr = ecode;        charptr = ecode;
3297        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3298        ecode += length;        ecode += length;
3299    
3300        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1913  for (;;) Line 3302  for (;;)
3302    
3303        if (length > 1)        if (length > 1)
3304          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3305  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3306          int othercase;          unsigned int othercase;
3307          int chartype;          if (op >= OP_STARI &&     /* Caseless */
3308          if ((ims & PCRE_CASELESS) != 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
3309               _pcre_ucp_findchar(fc, &chartype, &othercase) >= 0 &&            oclength = PRIV(ord2utf)(othercase, occhars);
3310               othercase > 0)          else oclength = 0;
           oclength = _pcre_ord2utf8(othercase, occhars);  
3311  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3312    
3313          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3314            {            {
3315            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3316            /* Need braces because of following else */              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3317            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3318              else if (oclength > 0 &&
3319                       eptr <= md->end_subject - oclength &&
3320                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3321    #endif  /* SUPPORT_UCP */
3322            else            else
3323              {              {
3324              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3325              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3326              }              }
3327            }            }
3328    
# Line 1943  for (;;) Line 3332  for (;;)
3332            {            {
3333            for (fi = min;; fi++)            for (fi = min;; fi++)
3334              {              {
3335              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3336              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3337              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3338              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3339              /* Need braces because of following else */                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3340              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3341                else if (oclength > 0 &&
3342                         eptr <= md->end_subject - oclength &&
3343                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3344    #endif  /* SUPPORT_UCP */
3345              else              else
3346                {                {
3347                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3348                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3349                }                }
3350              }              }
3351            /* Control never gets here */            /* Control never gets here */
3352            }            }
3353          else  
3354            else  /* Maximize */
3355            {            {
3356            pp = eptr;            pp = eptr;
3357            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3358              {              {
3359              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3360              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3361              else if (oclength == 0) break;  #ifdef SUPPORT_UCP
3362                else if (oclength > 0 &&
3363                         eptr <= md->end_subject - oclength &&
3364                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3365    #endif  /* SUPPORT_UCP */
3366              else              else
3367                {                {
3368                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3369                eptr += oclength;                break;
3370                }                }
3371              }              }
3372            while (eptr >= pp)  
3373             {            if (possessive) continue;
3374             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);  
3375             if (rrc != MATCH_NOMATCH) RRETURN(rrc);            for(;;)
3376             eptr -= length;              {
3377             }              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3378            RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3379                if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3380    #ifdef SUPPORT_UCP
3381                eptr--;
3382                BACKCHAR(eptr);
3383    #else   /* without SUPPORT_UCP */
3384                eptr -= length;
3385    #endif  /* SUPPORT_UCP */
3386                }
3387            }            }
3388          /* Control never gets here */          /* Control never gets here */
3389          }          }
# Line 1987  for (;;) Line 3393  for (;;)
3393        value of fc will always be < 128. */        value of fc will always be < 128. */
3394        }        }
3395      else      else
3396  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3397          /* 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);  
3398        fc = *ecode++;        fc = *ecode++;
       }  
3399    
3400      /* 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
3401      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
3402      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3403      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3404      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2005  for (;;) Line 3407  for (;;)
3407      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3408    
3409      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3410        max, eptr));        max, (char *)eptr));
3411    
3412      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3413        {        {
3414        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3415          /* fc must be < 128 if UTF is enabled. */
3416          foc = md->fcc[fc];
3417    #else
3418    #ifdef SUPPORT_UTF
3419    #ifdef SUPPORT_UCP
3420          if (utf && fc > 127)
3421            foc = UCD_OTHERCASE(fc);
3422    #else
3423          if (utf && fc > 127)
3424            foc = fc;
3425    #endif /* SUPPORT_UCP */
3426          else
3427    #endif /* SUPPORT_UTF */
3428            foc = TABLE_GET(fc, md->fcc, fc);
3429    #endif /* COMPILE_PCRE8 */
3430    
3431        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3432          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3433            if (eptr >= md->end_subject)
3434              {
3435              SCHECK_PARTIAL();
3436              RRETURN(MATCH_NOMATCH);
3437              }
3438            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3439            eptr++;
3440            }
3441        if (min == max) continue;        if (min == max) continue;
3442        if (minimize)        if (minimize)
3443          {          {
3444          for (fi = min;; fi++)          for (fi = min;; fi++)
3445            {            {
3446            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3447            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3448            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3449                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3450                {
3451                SCHECK_PARTIAL();
3452              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3453                }
3454              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3455              eptr++;
3456            }            }
3457          /* Control never gets here */          /* Control never gets here */
3458          }          }
3459        else        else  /* Maximize */
3460          {          {
3461          pp = eptr;          pp = eptr;
3462          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3463            {            {
3464            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3465                {
3466                SCHECK_PARTIAL();
3467                break;
3468                }