/[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 150 by ph10, Tue Apr 17 08:22:40 2007 UTC revision 1284 by ph10, Fri Mar 15 11:54:58 2013 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) 1997-2007 University of Cambridge             Copyright (c) 1997-2013 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 */  #define NLBLOCK md             /* Block containing newline information */
49  #define PSSTART start_subject  /* Field containing processed string start */  #define PSSTART start_subject  /* Field containing processed string start */
50  #define PSEND   end_subject    /* Field containing processed string end */  #define PSEND   end_subject    /* Field containing processed string end */
# Line 53  possible. There are also some static sup Line 56  possible. There are also some static sup
56  #undef min  #undef min
57  #undef max  #undef max
58    
59  /* The chain of eptrblocks for tail recursions uses memory in stack workspace,  /* The md->capture_last field uses the lower 16 bits for the last captured
60  obtained at top level, the size of which is defined by EPTR_WORK_SIZE. */  substring (which can never be greater than 65535) and a bit in the top half
61    to mean "capture vector overflowed". This odd way of doing things was
62  #define EPTR_WORK_SIZE (1000)  implemented when it was realized that preserving and restoring the overflow bit
63    whenever the last capture number was saved/restored made for a neater
64    interface, and doing it this way saved on (a) another variable, which would
65    have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66    separate set of save/restore instructions. The following defines are used in
67    implementing this. */
68    
69    #define CAPLMASK    0x0000ffff    /* The bits used for last_capture */
70    #define OVFLMASK    0xffff0000    /* The bits used for the overflow flag */
71    #define OVFLBIT     0x00010000    /* The bit that is set for overflow */
72    
73    /* Values for setting in md->match_function_type to indicate two special types
74    of call to match(). We do it this way to save on using another stack variable,
75    as stack usage is to be discouraged. */
76    
77  /* Flag bits for the match() function */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
78    #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
 #define match_condassert     0x01  /* Called to check a condition assertion */  
 #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  
 #define match_tail_recursed  0x04  /* Tail recursive call */  
79    
80  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
81  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 70  defined PCRE_ERROR_xxx codes, which are Line 83  defined PCRE_ERROR_xxx codes, which are
83  #define MATCH_MATCH        1  #define MATCH_MATCH        1
84  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
85    
86    /* Special internal returns from the match() function. Make them sufficiently
87    negative to avoid the external error codes. */
88    
89    #define MATCH_ACCEPT       (-999)
90    #define MATCH_KETRPOS      (-998)
91    #define MATCH_ONCE         (-997)
92    /* The next 5 must be kept together and in sequence so that a test that checks
93    for any one of them can use a range. */
94    #define MATCH_COMMIT       (-996)
95    #define MATCH_PRUNE        (-995)
96    #define MATCH_SKIP         (-994)
97    #define MATCH_SKIP_ARG     (-993)
98    #define MATCH_THEN         (-992)
99    #define MATCH_BACKTRACK_MAX MATCH_THEN
100    #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101    
102  /* 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.
103  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,
104  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 81  because the offset vector is always a mu Line 110  because the offset vector is always a mu
110  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
111  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
112    
113    #ifdef PCRE_DEBUG
   
 #ifdef DEBUG  
114  /*************************************************  /*************************************************
115  *        Debugging function to print chars       *  *        Debugging function to print chars       *
116  *************************************************/  *************************************************/
# Line 101  Returns:     nothing Line 128  Returns:     nothing
128  */  */
129    
130  static void  static void
131  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)
132  {  {
133  unsigned int c;  pcre_uint32 c;
134    BOOL utf = md->utf;
135  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;
136  while (length-- > 0)  while (length-- > 0)
137    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138  }  }
139  #endif  #endif
140    
# Line 116  while (length-- > 0) Line 144  while (length-- > 0)
144  *          Match a back-reference                *  *          Match a back-reference                *
145  *************************************************/  *************************************************/
146    
147  /* 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
148  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
149    the length passed is zero. Note that in caseless UTF-8 mode, the number of
150    subject bytes matched may be different to the number of reference bytes.
151    
152  Arguments:  Arguments:
153    offset      index into the offset vector    offset      index into the offset vector
154    eptr        points into the subject    eptr        pointer into the subject
155    length      length to be matched    length      length of reference to be matched (number of bytes)
156    md          points to match data block    md          points to match data block
157    ims         the ims flags    caseless    TRUE if caseless
158    
159  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
160                  -1 no match
161                  -2 partial match; always given if at end subject
162  */  */
163    
164  static BOOL  static int
165  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166    unsigned long int ims)    BOOL caseless)
167  {  {
168  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
169    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170    #ifdef SUPPORT_UTF
171    BOOL utf = md->utf;
172    #endif
173    
174  #ifdef DEBUG  #ifdef PCRE_DEBUG
175  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
176    printf("matching subject <null>");    printf("matching subject <null>");
177  else  else
# Line 148  pchars(p, length, FALSE, md); Line 184  pchars(p, length, FALSE, md);
184  printf("\n");  printf("\n");
185  #endif  #endif
186    
187  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
188    case the length is passed as zero). */
189    
190  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
191    
192  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193    properly if Unicode properties are supported. Otherwise, we can check only
194    ASCII characters. */
195    
196  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
197    {    {
198    while (length-- > 0)  #ifdef SUPPORT_UTF
199      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
200      if (utf)
201        {
202        /* Match characters up to the end of the reference. NOTE: the number of
203        data units matched may differ, because in UTF-8 there are some characters
204        whose upper and lower case versions code have different numbers of bytes.
205        For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207        sequence of two of the latter. It is important, therefore, to check the
208        length along the reference, not along the subject (earlier code did this
209        wrong). */
210    
211        PCRE_PUCHAR endptr = p + length;
212        while (p < endptr)
213          {
214          pcre_uint32 c, d;
215          const ucd_record *ur;
216          if (eptr >= md->end_subject) return -2;   /* Partial match */
217          GETCHARINC(c, eptr);
218          GETCHARINC(d, p);
219          ur = GET_UCD(d);
220          if (c != d && c != d + ur->other_case)
221            {
222            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223            for (;;)
224              {
225              if (c < *pp) return -1;
226              if (c == *pp++) break;
227              }
228            }
229          }
230        }
231      else
232    #endif
233    #endif
234    
235      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236      is no UCP support. */
237        {
238        while (length-- > 0)
239          {
240          pcre_uint32 cc, cp;
241          if (eptr >= md->end_subject) return -2;   /* Partial match */
242          cc = RAWUCHARTEST(eptr);
243          cp = RAWUCHARTEST(p);
244          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245          p++;
246          eptr++;
247          }
248        }
249    }    }
250    
251    /* In the caseful case, we can just compare the bytes, whether or not we
252    are in UTF-8 mode. */
253    
254  else  else
255    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
256      while (length-- > 0)
257        {
258        if (eptr >= md->end_subject) return -2;   /* Partial match */
259        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260        }
261      }
262    
263  return TRUE;  return (int)(eptr - eptr_start);
264  }  }
265    
266    
# Line 188  calls by keeping local variables that ne Line 286  calls by keeping local variables that ne
286  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
287  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
288  always used to.  always used to.
289    
290    The original heap-recursive code used longjmp(). However, it seems that this
291    can be very slow on some operating systems. Following a suggestion from Stan
292    Switzer, the use of longjmp() has been abolished, at the cost of having to
293    provide a unique number for each call to RMATCH. There is no way of generating
294    a sequence of numbers at compile time in C. I have given them names, to make
295    them stand out more clearly.
296    
297    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
298    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
299    tests. Furthermore, not using longjmp() means that local dynamic variables
300    don't have indeterminate values; this has meant that the frame size can be
301    reduced because the result can be "passed back" by straight setting of the
302    variable instead of being passed in the frame.
303  ****************************************************************************  ****************************************************************************
304  ***************************************************************************/  ***************************************************************************/
305    
306    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307    below must be updated in sync.  */
308    
309    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
310           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315           RM61,  RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316    
317  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
318  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
319    actually used in this definition. */
320    
321  #ifndef NO_RECURSE  #ifndef NO_RECURSE
322  #define REGISTER register  #define REGISTER register
323  #ifdef DEBUG  
324  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
325    #define RMATCH(ra,rb,rc,rd,re,rw) \
326    { \    { \
327    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
328    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
330    }    }
331  #define RRETURN(ra) \  #define RRETURN(ra) \
332    { \    { \
333    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
334    return ra; \    return ra; \
335    }    }
336  #else  #else
337  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
338    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
340  #endif  #endif
341    
342  #else  #else
343    
344    
345  /* 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
346  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
347  match(), which never changes. */  argument of match(), which never changes. */
348    
349  #define REGISTER  #define REGISTER
350    
351  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
352    {\    {\
353    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
354    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->Xrdepth = frame->Xrdepth + 1;\  
     newframe->Xprevframe = frame;\  
     frame = newframe;\  
     DPRINTF(("restarting from line %d\n", __LINE__));\  
     goto HEAP_RECURSE;\  
     }\  
   else\  
355      {\      {\
356      DPRINTF(("longjumped back to line %d\n", __LINE__));\      newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357      frame = md->thisframe;\      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358      rx = frame->Xresult;\      newframe->Xnextframe = NULL;\
359        frame->Xnextframe = newframe;\
360      }\      }\
361      frame->Xwhere = rw;\
362      newframe->Xeptr = ra;\
363      newframe->Xecode = rb;\
364      newframe->Xmstart = mstart;\
365      newframe->Xoffset_top = rc;\
366      newframe->Xeptrb = re;\
367      newframe->Xrdepth = frame->Xrdepth + 1;\
368      newframe->Xprevframe = frame;\
369      frame = newframe;\
370      DPRINTF(("restarting from line %d\n", __LINE__));\
371      goto HEAP_RECURSE;\
372      L_##rw:\
373      DPRINTF(("jumped back to line %d\n", __LINE__));\
374    }    }
375    
376  #define RRETURN(ra)\  #define RRETURN(ra)\
377    {\    {\
378    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
379    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
380    if (frame != NULL)\    if (frame != NULL)\
381      {\      {\
382      frame->Xresult = ra;\      rrc = ra;\
383      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
384      }\      }\
385    return ra;\    return ra;\
386    }    }
# Line 268  match(), which never changes. */ Line 390  match(), which never changes. */
390    
391  typedef struct heapframe {  typedef struct heapframe {
392    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
393      struct heapframe *Xnextframe;
394    
395    /* Function arguments that may change */    /* Function arguments that may change */
396    
397    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
398    const uschar *Xecode;    const pcre_uchar *Xecode;
399      PCRE_PUCHAR Xmstart;
400    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
401    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
402    unsigned int Xrdepth;    unsigned int Xrdepth;
403    
404    /* Function local variables */    /* Function local variables */
405    
406    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
407    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
408    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
409    const uschar *Xnext;  #endif
410    const uschar *Xpp;    PCRE_PUCHAR Xdata;
411    const uschar *Xprev;    PCRE_PUCHAR Xnext;
412    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
413      PCRE_PUCHAR Xprev;
414      PCRE_PUCHAR Xsaved_eptr;
415    
416    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
417    
# Line 295  typedef struct heapframe { Line 419  typedef struct heapframe {
419    BOOL Xcondition;    BOOL Xcondition;
420    BOOL Xprev_is_word;    BOOL Xprev_is_word;
421    
   unsigned long int Xoriginal_ims;  
   
422  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
423    int Xprop_type;    int Xprop_type;
424    int Xprop_value;    unsigned int Xprop_value;
425    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
426    int Xoclength;    int Xoclength;
427    uschar Xocchars[8];    pcre_uchar Xocchars[6];
428  #endif  #endif
429    
430      int Xcodelink;
431    int Xctype;    int Xctype;
432    unsigned int Xfc;    unsigned int Xfc;
433    int Xfi;    int Xfi;
434    int Xlength;    int Xlength;
435    int Xmax;    int Xmax;
436    int Xmin;    int Xmin;
437    int Xnumber;    unsigned int Xnumber;
438    int Xoffset;    int Xoffset;
439    int Xop;    unsigned int Xop;
440    int Xsave_capture_last;    pcre_int32 Xsave_capture_last;
441    int Xsave_offset1, Xsave_offset2, Xsave_offset3;    int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442    int Xstacksave[REC_STACK_SAVE_MAX];    int Xstacksave[REC_STACK_SAVE_MAX];
443    
444    eptrblock Xnewptrb;    eptrblock Xnewptrb;
445    
446    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
447    
448    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
449    
450  } heapframe;  } heapframe;
451    
# Line 344  typedef struct heapframe { Line 463  typedef struct heapframe {
463    
464  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
465  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
466  same response.  same response. */
467    
468    /* These macros pack up tests that are used for partial matching, and which
469    appear several times in the code. We set the "hit end" flag if the pointer is
470    at the end of the subject and also past the start of the subject (i.e.
471    something has been matched). For hard partial matching, we then return
472    immediately. The second one is used when we already know we are past the end of
473    the subject. */
474    
475    #define CHECK_PARTIAL()\
476      if (md->partial != 0 && eptr >= md->end_subject && \
477          eptr > md->start_used_ptr) \
478        { \
479        md->hitend = TRUE; \
480        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481        }
482    
483  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
484  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
485        { \
486        md->hitend = TRUE; \
487        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488        }
489    
490    
491    /* Performance note: It might be tempting to extract commonly used fields from
492    the md structure (e.g. utf, end_subject) into individual variables to improve
493  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
494  made performance worse.  made performance worse.
495    
496  Arguments:  Arguments:
497     eptr        pointer to current character in subject     eptr        pointer to current character in subject
498     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
499       mstart      pointer to the current match start position (can be modified
500                     by encountering \K)
501     offset_top  current top pointer     offset_top  current top pointer
502     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
503     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
504                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
                  match_tail_recursed - this is a tail_recursed group  
505     rdepth      the recursion depth     rdepth      the recursion depth
506    
507  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
508                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
509                   a negative MATCH_xxx value for PRUNE, SKIP, etc
510                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
511                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
512  */  */
513    
514  static int  static int
515  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517    int flags, unsigned int rdepth)    unsigned int rdepth)
518  {  {
519  /* 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,
520  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 383  so they can be ordinary variables in all Line 522  so they can be ordinary variables in all
522    
523  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
524  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
525  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
526  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
527    
528  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
529    BOOL caseless;
530    int condcode;
531    
532  /* 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
533  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
534  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
535  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536    the top-level on the stack rather than malloc-ing them all gives a performance
537    boost in many cases where there is not much "recursion". */
538    
539  #ifdef NO_RECURSE  #ifdef NO_RECURSE
540  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
541    
542  /* Copy in the original argument variables */  /* Copy in the original argument variables */
543    
544  frame->Xeptr = eptr;  frame->Xeptr = eptr;
545  frame->Xecode = ecode;  frame->Xecode = ecode;
546    frame->Xmstart = mstart;
547  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
548  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
549  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
550    
551  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 415  HEAP_RECURSE: Line 556  HEAP_RECURSE:
556    
557  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
558  #define ecode              frame->Xecode  #define ecode              frame->Xecode
559    #define mstart             frame->Xmstart
560  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
561  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
562  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
563    
564  /* Ditto for the local variables */  /* Ditto for the local variables */
565    
566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
567  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
568  #endif  #endif
569  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
570    #define codelink           frame->Xcodelink
571  #define data               frame->Xdata  #define data               frame->Xdata
572  #define next               frame->Xnext  #define next               frame->Xnext
573  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 439  HEAP_RECURSE: Line 580  HEAP_RECURSE:
580  #define condition          frame->Xcondition  #define condition          frame->Xcondition
581  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
582    
 #define original_ims       frame->Xoriginal_ims  
   
583  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
584  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
585  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
586  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
587  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
588  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
589  #endif  #endif
# Line 477  i, and fc and c, can be the same variabl Line 613  i, and fc and c, can be the same variabl
613  #define fi i  #define fi i
614  #define fc c  #define fc c
615    
616    /* Many of the following variables are used only in small blocks of the code.
617    My normal style of coding would have declared them within each of those blocks.
618    However, in order to accommodate the version of this code that uses an external
619    "stack" implemented on the heap, it is easier to declare them all here, so the
620    declarations can be cut out in a block. The only declarations within blocks
621    below are for variables that do not have to be preserved over a recursive call
622    to RMATCH(). */
623    
624  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  #ifdef SUPPORT_UTF
625  const uschar *charptr;             /* in small blocks of the code. My normal */  const pcre_uchar *charptr;
626  #endif                             /* style of coding would have declared    */  #endif
627  const uschar *callpat;             /* them within each of those blocks.      */  const pcre_uchar *callpat;
628  const uschar *data;                /* However, in order to accommodate the   */  const pcre_uchar *data;
629  const uschar *next;                /* version of this code that uses an      */  const pcre_uchar *next;
630  USPTR         pp;                  /* external "stack" implemented on the    */  PCRE_PUCHAR       pp;
631  const uschar *prev;                /* heap, it is easier to declare them all */  const pcre_uchar *prev;
632  USPTR         saved_eptr;          /* here, so the declarations can be cut   */  PCRE_PUCHAR       saved_eptr;
633                                     /* out in a block. The only declarations  */  
634  recursion_info new_recursive;      /* within blocks below are for variables  */  recursion_info new_recursive;
635                                     /* that do not have to be preserved over  */  
636  BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  BOOL cur_is_word;
637  BOOL condition;  BOOL condition;
638  BOOL prev_is_word;  BOOL prev_is_word;
639    
 unsigned long int original_ims;  
   
640  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
641  int prop_type;  int prop_type;
642  int prop_value;  unsigned int prop_value;
643  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
644  int oclength;  int oclength;
645  uschar occhars[8];  pcre_uchar occhars[6];
646  #endif  #endif
647    
648    int codelink;
649  int ctype;  int ctype;
650  int length;  int length;
651  int max;  int max;
652  int min;  int min;
653  int number;  unsigned int number;
654  int offset;  int offset;
655  int op;  unsigned int op;
656  int save_capture_last;  pcre_int32 save_capture_last;
657  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
658  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
659    
660  eptrblock newptrb;  eptrblock newptrb;
661    
662    /* There is a special fudge for calling match() in a way that causes it to
663    measure the size of its basic stack frame when the stack is being used for
664    recursion. The second argument (ecode) being NULL triggers this behaviour. It
665    cannot normally ever be NULL. The return is the negated value of the frame
666    size. */
667    
668    if (ecode == NULL)
669      {
670      if (rdepth == 0)
671        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672      else
673        {
674        int len = (char *)&rdepth - (char *)eptr;
675        return (len > 0)? -len : len;
676        }
677      }
678  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
679    
680    /* To save space on the stack and in the heap frame, I have doubled up on some
681    of the local variables that are used only in localised parts of the code, but
682    still need to be preserved over recursive calls of match(). These macros define
683    the alternative names that are used. */
684    
685    #define allow_zero    cur_is_word
686    #define cbegroup      condition
687    #define code_offset   codelink
688    #define condassert    condition
689    #define matched_once  prev_is_word
690    #define foc           number
691    #define save_mark     data
692    
693  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
694  variables. */  variables. */
695    
# Line 540  TAIL_RECURSE: Line 709  TAIL_RECURSE:
709  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
710  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
711  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
714  complicated macro. It has to be used in one particular way. This shouldn't,  complicated macro. It has to be used in one particular way. This shouldn't,
715  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
716    
717    #ifdef SUPPORT_UTF
718    utf = md->utf;       /* Local copy of the flag */
719    #else
720    utf = FALSE;
721    #endif
722    
723  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
724  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
725    
726  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728    
 original_ims = ims;    /* Save for resetting on ')' */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
729  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
730  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
732  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
733  When match() is called in other circumstances, don't add to the chain. If this  
734  is a tail recursion, use a block from the workspace, as the one on the stack is  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735  already used. */  such remembered pointers, to be checked when we hit the closing ket, in order
736    to break infinite loops that match no characters. When match() is called in
737    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738    NOT be used with tail recursion, because the memory block that is used is on
739    the stack, so a new one may be required for each match(). */
740    
741  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
742    {    {
743    eptrblock *p;    newptrb.epb_saved_eptr = eptr;
744    if ((flags & match_tail_recursed) != 0)    newptrb.epb_prev = eptrb;
745      {    eptrb = &newptrb;
746      if (md->eptrn >= EPTR_WORK_SIZE) RRETURN(PCRE_ERROR_NULLWSLIMIT);    md->match_function_type = 0;
     p = md->eptrchain + md->eptrn++;  
     }  
   else p = &newptrb;  
   p->epb_saved_eptr = eptr;  
   p->epb_prev = eptrb;  
   eptrb = p;  
747    }    }
748    
749  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 588  for (;;) Line 753  for (;;)
753    minimize = possessive = FALSE;    minimize = possessive = FALSE;
754    op = *ecode;    op = *ecode;
755    
   /* For partial matching, remember if we ever hit the end of the subject after  
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > md->start_match)  
     md->hitend = TRUE;  
   
756    switch(op)    switch(op)
757      {      {
758      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
759      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
760      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
761      set from a previous iteration of this group, and be referred to by a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762      reference inside the group.        eptrb, RM55);
763        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
765      values of the final offsets, in case they were set by a previous iteration  
766      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767        argument, and we must check whether that argument matches this MARK's
768        argument. It is passed back in md->start_match_ptr (an overloading of that
769        variable). If it does match, we reset that variable to the current subject
770        position and return MATCH_SKIP. Otherwise, pass back the return code
771        unaltered. */
772    
773        else if (rrc == MATCH_SKIP_ARG &&
774            STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775          {
776          md->start_match_ptr = eptr;
777          RRETURN(MATCH_SKIP);
778          }
779        RRETURN(rrc);
780    
781        case OP_FAIL:
782        RRETURN(MATCH_NOMATCH);
783    
784        case OP_COMMIT:
785        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786          eptrb, RM52);
787        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788        RRETURN(MATCH_COMMIT);
789    
790        case OP_PRUNE:
791        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792          eptrb, RM51);
793        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794        RRETURN(MATCH_PRUNE);
795    
796        case OP_PRUNE_ARG:
797        md->nomatch_mark = ecode + 2;
798        md->mark = NULL;    /* In case previously set by assertion */
799        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800          eptrb, RM56);
801        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
802             md->mark == NULL) md->mark = ecode + 2;
803        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
804        RRETURN(MATCH_PRUNE);
805    
806        case OP_SKIP:
807        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
808          eptrb, RM53);
809        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
810        md->start_match_ptr = eptr;   /* Pass back current position */
811        RRETURN(MATCH_SKIP);
812    
813        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
814        nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
815        not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
816        that failed and any that preceed it (either they also failed, or were not
817        triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
818        SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
819        set to the count of the one that failed. */
820    
821        case OP_SKIP_ARG:
822        md->skip_arg_count++;
823        if (md->skip_arg_count <= md->ignore_skip_arg)
824          {
825          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
826          break;
827          }
828        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
829          eptrb, RM57);
830        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831    
832        /* Pass back the current skip name by overloading md->start_match_ptr and
833        returning the special MATCH_SKIP_ARG return code. This will either be
834        caught by a matching MARK, or get to the top, where it causes a rematch
835        with md->ignore_skip_arg set to the value of md->skip_arg_count. */
836    
837        md->start_match_ptr = ecode + 2;
838        RRETURN(MATCH_SKIP_ARG);
839    
840        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
841        the branch in which it occurs can be determined. Overload the start of
842        match pointer to do this. */
843    
844        case OP_THEN:
845        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
846          eptrb, RM54);
847        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
848        md->start_match_ptr = ecode;
849        RRETURN(MATCH_THEN);
850    
851        case OP_THEN_ARG:
852        md->nomatch_mark = ecode + 2;
853        md->mark = NULL;    /* In case previously set by assertion */
854        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
855          md, eptrb, RM58);
856        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
857             md->mark == NULL) md->mark = ecode + 2;
858        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859        md->start_match_ptr = ecode;
860        RRETURN(MATCH_THEN);
861    
862        /* Handle an atomic group that does not contain any capturing parentheses.
863        This can be handled like an assertion. Prior to 8.13, all atomic groups
864        were handled this way. In 8.13, the code was changed as below for ONCE, so
865        that backups pass through the group and thereby reset captured values.
866        However, this uses a lot more stack, so in 8.20, atomic groups that do not
867        contain any captures generate OP_ONCE_NC, which can be handled in the old,
868        less stack intensive way.
869    
870        Check the alternative branches in turn - the matching won't pass the KET
871        for this kind of subpattern. If any one branch matches, we carry on as at
872        the end of a normal bracket, leaving the subject pointer, but resetting
873        the start-of-match value in case it was changed by \K. */
874    
875        case OP_ONCE_NC:
876        prev = ecode;
877        saved_eptr = eptr;
878        save_mark = md->mark;
879        do
880          {
881          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
882          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
883            {
884            mstart = md->start_match_ptr;
885            break;
886            }
887          if (rrc == MATCH_THEN)
888            {
889            next = ecode + GET(ecode,1);
890            if (md->start_match_ptr < next &&
891                (*ecode == OP_ALT || *next == OP_ALT))
892              rrc = MATCH_NOMATCH;
893            }
894    
895          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
896          ecode += GET(ecode,1);
897          md->mark = save_mark;
898          }
899        while (*ecode == OP_ALT);
900    
901        /* If hit the end of the group (which could be repeated), fail */
902    
903        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
904    
905        /* Continue as from after the group, updating the offsets high water
906        mark, since extracts may have been taken. */
907    
908        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
909    
910        offset_top = md->end_offset_top;
911        eptr = md->end_match_ptr;
912    
913        /* For a non-repeating ket, just continue at this level. This also
914        happens for a repeating ket if no characters were matched in the group.
915        This is the forcible breaking of infinite loops as implemented in Perl
916        5.005. */
917    
918        if (*ecode == OP_KET || eptr == saved_eptr)
919          {
920          ecode += 1+LINK_SIZE;
921          break;
922          }
923    
924        /* The repeating kets try the rest of the pattern or restart from the
925        preceding bracket, in the appropriate order. The second "call" of match()
926        uses tail recursion, to avoid using another stack frame. */
927    
928        if (*ecode == OP_KETRMIN)
929          {
930          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
931          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
932          ecode = prev;
933          goto TAIL_RECURSE;
934          }
935        else  /* OP_KETRMAX */
936          {
937          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
938          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
939          ecode += 1 + LINK_SIZE;
940          goto TAIL_RECURSE;
941          }
942        /* Control never gets here */
943    
944        /* Handle a capturing bracket, other than those that are possessive with an
945        unlimited repeat. If there is space in the offset vector, save the current
946        subject position in the working slot at the top of the vector. We mustn't
947        change the current values of the data slot, because they may be set from a
948        previous iteration of this group, and be referred to by a reference inside
949        the group. A failure to match might occur after the group has succeeded,
950        if something later on doesn't match. For this reason, we need to restore
951        the working value and also the values of the final offsets, in case they
952        were set by a previous iteration of the same bracket.
953    
954      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
955      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 617  for (;;) Line 960  for (;;)
960      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
961      offset = number << 1;      offset = number << 1;
962    
963  #ifdef DEBUG  #ifdef PCRE_DEBUG
964      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
965      printf("subject=");      printf("subject=");
966      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 630  for (;;) Line 973  for (;;)
973        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
974        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
975        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
976          save_mark = md->mark;
977    
978        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
979        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
980            (int)(eptr - md->start_subject);
981    
982        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
983          {          {
984          RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985            ims, eptrb, flags);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
986              eptrb, RM1);
987            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
988    
989            /* If we backed up to a THEN, check whether it is within the current
990            branch by comparing the address of the THEN that is passed back with
991            the end of the branch. If it is within the current branch, and the
992            branch is one of two or more alternatives (it either starts or ends
993            with OP_ALT), we have reached the limit of THEN's action, so convert
994            the return code to NOMATCH, which will cause normal backtracking to
995            happen from now on. Otherwise, THEN is passed back to an outer
996            alternative. This implements Perl's treatment of parenthesized groups,
997            where a group not containing | does not affect the current alternative,
998            that is, (X) is NOT the same as (X|(*F)). */
999    
1000            if (rrc == MATCH_THEN)
1001              {
1002              next = ecode + GET(ecode,1);
1003              if (md->start_match_ptr < next &&
1004                  (*ecode == OP_ALT || *next == OP_ALT))
1005                rrc = MATCH_NOMATCH;
1006              }
1007    
1008            /* Anything other than NOMATCH is passed back. */
1009    
1010          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1011          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1012          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1013            md->mark = save_mark;
1014            if (*ecode != OP_ALT) break;
1015          }          }
       while (*ecode == OP_ALT);  
1016    
1017        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1018        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1019        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1020        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1021    
1022        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1023    
1024          RRETURN(rrc);
1025        }        }
1026    
1027      /* Insufficient room for saving captured contents. Treat as a non-capturing      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1028      bracket. */      as a non-capturing bracket. */
1029    
1030        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1031        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1032    
1033      DPRINTF(("insufficient capture room: treat as non-capturing\n"));      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1034    
1035      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036      final alternative within the brackets, we would return the result of a      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037      recursive call to match() whatever happened. We can reduce stack usage by  
1038      turning this into a tail recursion. */      /* Non-capturing or atomic group, except for possessive with unlimited
1039        repeat and ONCE group with no captures. Loop for all the alternatives.
1040    
1041        When we get to the final alternative within the brackets, we used to return
1042        the result of a recursive call to match() whatever happened so it was
1043        possible to reduce stack usage by turning this into a tail recursion,
1044        except in the case of a possibly empty group. However, now that there is
1045        the possiblity of (*THEN) occurring in the final alternative, this
1046        optimization is no longer always possible.
1047    
1048        We can optimize if we know there are no (*THEN)s in the pattern; at present
1049        this is the best that can be done.
1050    
1051        MATCH_ONCE is returned when the end of an atomic group is successfully
1052        reached, but subsequent matching fails. It passes back up the tree (causing
1053        captured values to be reset) until the original atomic group level is
1054        reached. This is tested by comparing md->once_target with the start of the
1055        group. At this point, the return is converted into MATCH_NOMATCH so that
1056        previous backup points can be taken. */
1057    
1058        case OP_ONCE:
1059      case OP_BRA:      case OP_BRA:
1060      case OP_SBRA:      case OP_SBRA:
1061      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1062      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1063      for (;;)      for (;;)
1064        {        {
1065        if (ecode[GET(ecode, 1)] != OP_ALT)        if (op >= OP_SBRA || op == OP_ONCE)
1066            md->match_function_type = MATCH_CBEGROUP;
1067    
1068          /* If this is not a possibly empty group, and there are no (*THEN)s in
1069          the pattern, and this is the final alternative, optimize as described
1070          above. */
1071    
1072          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073          {          {
1074          ecode += _pcre_OP_lengths[*ecode];          ecode += PRIV(OP_lengths)[*ecode];
         flags |= match_tail_recursed;  
         DPRINTF(("bracket 0 tail recursion\n"));  
1075          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1076          }          }
1077    
1078        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* In all other cases, we have to make another call to match(). */
       otherwise return. */  
1079    
1080        RMATCH(rrc, eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1081          eptrb, flags);        save_capture_last = md->capture_last;
1082        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1083            RM2);
1084    
1085          /* See comment in the code for capturing groups above about handling
1086          THEN. */
1087    
1088          if (rrc == MATCH_THEN)
1089            {
1090            next = ecode + GET(ecode,1);
1091            if (md->start_match_ptr < next &&
1092                (*ecode == OP_ALT || *next == OP_ALT))
1093              rrc = MATCH_NOMATCH;
1094            }
1095    
1096          if (rrc != MATCH_NOMATCH)
1097            {
1098            if (rrc == MATCH_ONCE)
1099              {
1100              const pcre_uchar *scode = ecode;
1101              if (*scode != OP_ONCE)           /* If not at start, find it */
1102                {
1103                while (*scode == OP_ALT) scode += GET(scode, 1);
1104                scode -= GET(scode, 1);
1105                }
1106              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1107              }
1108            RRETURN(rrc);
1109            }
1110        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1111        }        md->mark = save_mark;
1112      /* Control never reaches here. */        if (*ecode != OP_ALT) break;
1113          md->capture_last = save_capture_last;
1114          }
1115    
1116        RRETURN(MATCH_NOMATCH);
1117    
1118        /* Handle possessive capturing brackets with an unlimited repeat. We come
1119        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120        handled similarly to the normal case above. However, the matching is
1121        different. The end of these brackets will always be OP_KETRPOS, which
1122        returns MATCH_KETRPOS without going further in the pattern. By this means
1123        we can handle the group by iteration rather than recursion, thereby
1124        reducing the amount of stack needed. */
1125    
1126        case OP_CBRAPOS:
1127        case OP_SCBRAPOS:
1128        allow_zero = FALSE;
1129    
1130      /* Conditional group: compilation checked that there are no more than      POSSESSIVE_CAPTURE:
1131      two branches. If the condition is false, skipping the first branch takes us      number = GET2(ecode, 1+LINK_SIZE);
1132      past the end if there is only one branch, but that's OK because that is      offset = number << 1;
     exactly what going to the ket would do. As there is only one branch to be  
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1133    
1134      case OP_COND:  #ifdef PCRE_DEBUG
1135      case OP_SCOND:      printf("start possessive bracket %d\n", number);
1136      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      printf("subject=");
1137        {      pchars(eptr, 16, TRUE, md);
1138        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      printf("\n");
1139        condition = md->recursive != NULL &&  #endif
         (offset == RREF_ANY || offset == md->recursive->group_num);  
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1140    
1141      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      if (offset < md->offset_max)
1142        {        {
1143        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        matched_once = FALSE;
1144        condition = offset < offset_top && md->offset_vector[offset] >= 0;        code_offset = (int)(ecode - md->start_code);
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1145    
1146      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */        save_offset1 = md->offset_vector[offset];
1147        {        save_offset2 = md->offset_vector[offset+1];
1148        condition = FALSE;        save_offset3 = md->offset_vector[md->offset_end - number];
1149        ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
       }  
1150    
1151      /* The condition is an assertion. Call match() to evaluate it - setting        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
     the final argument match_condassert causes it to stop at the end of an  
     assertion. */  
1152    
1153      else        /* Each time round the loop, save the current subject position for use
1154        {        when the group matches. For MATCH_MATCH, the group has matched, so we
1155        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        restart it with a new subject starting position, remembering that we had
1156            match_condassert);        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157        if (rrc == MATCH_MATCH)        usual. If we haven't matched any alternatives in any iteration, check to
1158          {        see if a previous iteration matched. If so, the group has matched;
1159          condition = TRUE;        continue from afterwards. Otherwise it has failed; restore the previous
1160          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        capture values before returning NOMATCH. */
1161          while (*ecode == OP_ALT) ecode += GET(ecode, 1);  
1162          for (;;)
1163            {
1164            md->offset_vector[md->offset_end - number] =
1165              (int)(eptr - md->start_subject);
1166            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168              eptrb, RM63);
1169            if (rrc == MATCH_KETRPOS)
1170              {
1171              offset_top = md->end_offset_top;
1172              eptr = md->end_match_ptr;
1173              ecode = md->start_code + code_offset;
1174              save_capture_last = md->capture_last;
1175              matched_once = TRUE;
1176              continue;
1177              }
1178    
1179            /* See comment in the code for capturing groups above about handling
1180            THEN. */
1181    
1182            if (rrc == MATCH_THEN)
1183              {
1184              next = ecode + GET(ecode,1);
1185              if (md->start_match_ptr < next &&
1186                  (*ecode == OP_ALT || *next == OP_ALT))
1187                rrc = MATCH_NOMATCH;
1188              }
1189    
1190            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1191            md->capture_last = save_capture_last;
1192            ecode += GET(ecode, 1);
1193            if (*ecode != OP_ALT) break;
1194          }          }
1195        else if (rrc != MATCH_NOMATCH)  
1196          if (!matched_once)
1197          {          {
1198          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1199            md->offset_vector[offset+1] = save_offset2;
1200            md->offset_vector[md->offset_end - number] = save_offset3;
1201          }          }
1202        else  
1203          if (allow_zero || matched_once)
1204          {          {
1205          condition = FALSE;          ecode += 1 + LINK_SIZE;
1206          ecode += GET(ecode, 1);          break;
1207          }          }
       }  
1208    
1209      /* We are now at the branch that is to be obeyed. As there is only one,        RRETURN(MATCH_NOMATCH);
     we can use tail recursion to avoid using another stack frame. If the second  
     alternative doesn't exist, we can just plough on. */  
   
     if (condition || *ecode == OP_ALT)  
       {  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed | ((op == OP_SCOND)? match_cbegroup : 0);  
       goto TAIL_RECURSE;  
       }  
     else  
       {  
       ecode += 1 + LINK_SIZE;  
1210        }        }
     break;  
   
1211    
1212      /* End of the pattern. If we are in a top-level recursion, we should      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213      restore the offsets appropriately and continue from after the call. */      as a non-capturing bracket. */
1214    
1215      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216      if (md->recursive != NULL && md->recursive->group_num == 0)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern 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;  
       }  
1217    
1218      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     string - backtracking will then try other alternatives, if any. */  
1219    
1220      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221      md->end_match_ptr = eptr;          /* Record where we ended */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     md->end_offset_top = offset_top;   /* and how many extracts were taken */  
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1222    
1223      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1224      matching won't pass the KET for an assertion. If any one branch matches,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1225      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1226      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1227      this level is identical to the lookahead case. */  
1228        case OP_BRAPOS:
1229        case OP_SBRAPOS:
1230        allow_zero = FALSE;
1231    
1232        POSSESSIVE_NON_CAPTURE:
1233        matched_once = FALSE;
1234        code_offset = (int)(ecode - md->start_code);
1235        save_capture_last = md->capture_last;
1236    
1237      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1238        {        {
1239        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240        if (rrc == MATCH_MATCH) break;        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1241            eptrb, RM48);
1242          if (rrc == MATCH_KETRPOS)
1243            {
1244            offset_top = md->end_offset_top;
1245            eptr = md->end_match_ptr;
1246            ecode = md->start_code + code_offset;
1247            matched_once = TRUE;
1248            continue;
1249            }
1250    
1251          /* See comment in the code for capturing groups above about handling
1252          THEN. */
1253    
1254          if (rrc == MATCH_THEN)
1255            {
1256            next = ecode + GET(ecode,1);
1257            if (md->start_match_ptr < next &&
1258                (*ecode == OP_ALT || *next == OP_ALT))
1259              rrc = MATCH_NOMATCH;
1260            }
1261    
1262        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1263        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1264          if (*ecode != OP_ALT) break;
1265          md->capture_last = save_capture_last;
1266        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1267    
1268      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1269          {
1270          ecode += 1 + LINK_SIZE;
1271          break;
1272          }
1273        RRETURN(MATCH_NOMATCH);
1274    
1275      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1276    
1277      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1278      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1279        past the end if there is only one branch, but that's OK because that is
1280        exactly what going to the ket would do. */
1281    
1282      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1283      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1284      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1285    
1286      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1287        inserted between OP_COND and an assertion condition. */
1288    
1289      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1290        {        {
1291        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0);        if (PUBL(callout) != NULL)
1292        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          {
1293        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          PUBL(callout_block) cb;
1294            cb.version          = 2;   /* Version 1 of the callout block */
1295            cb.callout_number   = ecode[LINK_SIZE+2];
1296            cb.offset_vector    = md->offset_vector;
1297    #if defined COMPILE_PCRE8
1298            cb.subject          = (PCRE_SPTR)md->start_subject;
1299    #elif defined COMPILE_PCRE16
1300            cb.subject          = (PCRE_SPTR16)md->start_subject;
1301    #elif defined COMPILE_PCRE32
1302            cb.subject          = (PCRE_SPTR32)md->start_subject;
1303    #endif
1304            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1305            cb.start_match      = (int)(mstart - md->start_subject);
1306            cb.current_position = (int)(eptr - md->start_subject);
1307            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1308            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1309            cb.capture_top      = offset_top/2;
1310            cb.capture_last     = md->capture_last & CAPLMASK;
1311            /* Internal change requires this for API compatibility. */
1312            if (cb.capture_last == 0) cb.capture_last = -1;
1313            cb.callout_data     = md->callout_data;
1314            cb.mark             = md->nomatch_mark;
1315            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1316            if (rrc < 0) RRETURN(rrc);
1317            }
1318          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1319          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1320          }
1321    
1322        condcode = ecode[LINK_SIZE+1];
1323    
1324        /* Now see what the actual condition is */
1325    
1326        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1327          {
1328          if (md->recursive == NULL)                /* Not recursing => FALSE */
1329            {
1330            condition = FALSE;
1331            ecode += GET(ecode, 1);
1332            }
1333          else
1334            {
1335            unsigned int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1336            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1337    
1338            /* If the test is for recursion into a specific subpattern, and it is
1339            false, but the test was set up by name, scan the table to see if the
1340            name refers to any other numbers, and test them. The condition is true
1341            if any one is set. */
1342    
1343            if (!condition && condcode == OP_NRREF)
1344              {
1345              pcre_uchar *slotA = md->name_table;
1346              for (i = 0; i < md->name_count; i++)
1347                {
1348                if (GET2(slotA, 0) == recno) break;
1349                slotA += md->name_entry_size;
1350                }
1351    
1352              /* Found a name for the number - there can be only one; duplicate
1353              names for different numbers are allowed, but not vice versa. First
1354              scan down for duplicates. */
1355    
1356              if (i < md->name_count)
1357                {
1358                pcre_uchar *slotB = slotA;
1359                while (slotB > md->name_table)
1360                  {
1361                  slotB -= md->name_entry_size;
1362                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1363                    {
1364                    condition = GET2(slotB, 0) == md->recursive->group_num;
1365                    if (condition) break;
1366                    }
1367                  else break;
1368                  }
1369    
1370                /* Scan up for duplicates */
1371    
1372                if (!condition)
1373                  {
1374                  slotB = slotA;
1375                  for (i++; i < md->name_count; i++)
1376                    {
1377                    slotB += md->name_entry_size;
1378                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1379                      {
1380                      condition = GET2(slotB, 0) == md->recursive->group_num;
1381                      if (condition) break;
1382                      }
1383                    else break;
1384                    }
1385                  }
1386                }
1387              }
1388    
1389            /* Chose branch according to the condition */
1390    
1391            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1392            }
1393          }
1394    
1395        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1396          {
1397          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1398          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1399    
1400          /* If the numbered capture is unset, but the reference was by name,
1401          scan the table to see if the name refers to any other numbers, and test
1402          them. The condition is true if any one is set. This is tediously similar
1403          to the code above, but not close enough to try to amalgamate. */
1404    
1405          if (!condition && condcode == OP_NCREF)
1406            {
1407            unsigned int refno = offset >> 1;
1408            pcre_uchar *slotA = md->name_table;
1409    
1410            for (i = 0; i < md->name_count; i++)
1411              {
1412              if (GET2(slotA, 0) == refno) break;
1413              slotA += md->name_entry_size;
1414              }
1415    
1416            /* Found a name for the number - there can be only one; duplicate names
1417            for different numbers are allowed, but not vice versa. First scan down
1418            for duplicates. */
1419    
1420            if (i < md->name_count)
1421              {
1422              pcre_uchar *slotB = slotA;
1423              while (slotB > md->name_table)
1424                {
1425                slotB -= md->name_entry_size;
1426                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1427                  {
1428                  offset = GET2(slotB, 0) << 1;
1429                  condition = offset < offset_top &&
1430                    md->offset_vector[offset] >= 0;
1431                  if (condition) break;
1432                  }
1433                else break;
1434                }
1435    
1436              /* Scan up for duplicates */
1437    
1438              if (!condition)
1439                {
1440                slotB = slotA;
1441                for (i++; i < md->name_count; i++)
1442                  {
1443                  slotB += md->name_entry_size;
1444                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1445                    {
1446                    offset = GET2(slotB, 0) << 1;
1447                    condition = offset < offset_top &&
1448                      md->offset_vector[offset] >= 0;
1449                    if (condition) break;
1450                    }
1451                  else break;
1452                  }
1453                }
1454              }
1455            }
1456    
1457          /* Chose branch according to the condition */
1458    
1459          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1460          }
1461    
1462        else if (condcode == OP_DEF)     /* DEFINE - always false */
1463          {
1464          condition = FALSE;
1465          ecode += GET(ecode, 1);
1466          }
1467    
1468        /* The condition is an assertion. Call match() to evaluate it - setting
1469        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1470        an assertion. */
1471    
1472        else
1473          {
1474          md->match_function_type = MATCH_CONDASSERT;
1475          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1476          if (rrc == MATCH_MATCH)
1477            {
1478            if (md->end_offset_top > offset_top)
1479              offset_top = md->end_offset_top;  /* Captures may have happened */
1480            condition = TRUE;
1481            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1482            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1483            }
1484    
1485          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1486          assertion; it is therefore treated as NOMATCH. */
1487    
1488          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1489            {
1490            RRETURN(rrc);         /* Need braces because of following else */
1491            }
1492          else
1493            {
1494            condition = FALSE;
1495            ecode += codelink;
1496            }
1497          }
1498    
1499        /* We are now at the branch that is to be obeyed. As there is only one, can
1500        use tail recursion to avoid using another stack frame, except when there is
1501        unlimited repeat of a possibly empty group. In the latter case, a recursive
1502        call to match() is always required, unless the second alternative doesn't
1503        exist, in which case we can just plough on. Note that, for compatibility
1504        with Perl, the | in a conditional group is NOT treated as creating two
1505        alternatives. If a THEN is encountered in the branch, it propagates out to
1506        the enclosing alternative (unless nested in a deeper set of alternatives,
1507        of course). */
1508    
1509        if (condition || *ecode == OP_ALT)
1510          {
1511          if (op != OP_SCOND)
1512            {
1513            ecode += 1 + LINK_SIZE;
1514            goto TAIL_RECURSE;
1515            }
1516    
1517          md->match_function_type = MATCH_CBEGROUP;
1518          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1519          RRETURN(rrc);
1520          }
1521    
1522         /* Condition false & no alternative; continue after the group. */
1523    
1524        else
1525          {
1526          ecode += 1 + LINK_SIZE;
1527          }
1528        break;
1529    
1530    
1531        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1532        to close any currently open capturing brackets. */
1533    
1534        case OP_CLOSE:
1535        number = GET2(ecode, 1);   /* Must be less than 65536 */
1536        offset = number << 1;
1537    
1538    #ifdef PCRE_DEBUG
1539          printf("end bracket %d at *ACCEPT", number);
1540          printf("\n");
1541    #endif
1542    
1543        md->capture_last = (md->capture_last & OVFLMASK) | number;
1544        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1545          {
1546          md->offset_vector[offset] =
1547            md->offset_vector[md->offset_end - number];
1548          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1549          if (offset_top <= offset) offset_top = offset + 2;
1550          }
1551        ecode += 1 + IMM2_SIZE;
1552        break;
1553    
1554    
1555        /* End of the pattern, either real or forced. */
1556    
1557        case OP_END:
1558        case OP_ACCEPT:
1559        case OP_ASSERT_ACCEPT:
1560    
1561        /* If we have matched an empty string, fail if not in an assertion and not
1562        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1563        is set and we have matched at the start of the subject. In both cases,
1564        backtracking will then try other alternatives, if any. */
1565    
1566        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1567             md->recursive == NULL &&
1568             (md->notempty ||
1569               (md->notempty_atstart &&
1570                 mstart == md->start_subject + md->start_offset)))
1571          RRETURN(MATCH_NOMATCH);
1572    
1573        /* Otherwise, we have a match. */
1574    
1575        md->end_match_ptr = eptr;           /* Record where we ended */
1576        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1577        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1578    
1579        /* For some reason, the macros don't work properly if an expression is
1580        given as the argument to RRETURN when the heap is in use. */
1581    
1582        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1583        RRETURN(rrc);
1584    
1585        /* Assertion brackets. Check the alternative branches in turn - the
1586        matching won't pass the KET for an assertion. If any one branch matches,
1587        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1588        start of each branch to move the current point backwards, so the code at
1589        this level is identical to the lookahead case. When the assertion is part
1590        of a condition, we want to return immediately afterwards. The caller of
1591        this incarnation of the match() function will have set MATCH_CONDASSERT in
1592        md->match_function type, and one of these opcodes will be the first opcode
1593        that is processed. We use a local variable that is preserved over calls to
1594        match() to remember this case. */
1595    
1596        case OP_ASSERT:
1597        case OP_ASSERTBACK:
1598        save_mark = md->mark;
1599        if (md->match_function_type == MATCH_CONDASSERT)
1600          {
1601          condassert = TRUE;
1602          md->match_function_type = 0;
1603          }
1604        else condassert = FALSE;
1605    
1606        do
1607          {
1608          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1609          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1610            {
1611            mstart = md->start_match_ptr;   /* In case \K reset it */
1612            break;
1613            }
1614          md->mark = save_mark;
1615    
1616          /* A COMMIT failure must fail the entire assertion, without trying any
1617          subsequent branches. */
1618    
1619          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1620    
1621          /* PCRE does not allow THEN to escape beyond an assertion; it
1622          is treated as NOMATCH. */
1623    
1624          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1625          ecode += GET(ecode, 1);
1626          }
1627        while (*ecode == OP_ALT);
1628    
1629        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1630    
1631        /* If checking an assertion for a condition, return MATCH_MATCH. */
1632    
1633        if (condassert) RRETURN(MATCH_MATCH);
1634    
1635        /* Continue from after the assertion, updating the offsets high water
1636        mark, since extracts may have been taken during the assertion. */
1637    
1638        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1639        ecode += 1 + LINK_SIZE;
1640        offset_top = md->end_offset_top;
1641        continue;
1642    
1643        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1644        PRUNE, or COMMIT means we must assume failure without checking subsequent
1645        branches. */
1646    
1647        case OP_ASSERT_NOT:
1648        case OP_ASSERTBACK_NOT:
1649        save_mark = md->mark;
1650        if (md->match_function_type == MATCH_CONDASSERT)
1651          {
1652          condassert = TRUE;
1653          md->match_function_type = 0;
1654          }
1655        else condassert = FALSE;
1656    
1657        do
1658          {
1659          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1660          md->mark = save_mark;
1661          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1662          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1663            {
1664            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1665            break;
1666            }
1667    
1668          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1669          as NOMATCH. */
1670    
1671          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1672        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1673        }        }
1674      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1675    
1676      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1677    
1678      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1679      continue;      continue;
# Line 846  for (;;) Line 1684  for (;;)
1684      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1685    
1686      case OP_REVERSE:      case OP_REVERSE:
1687  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1688      if (utf8)      if (utf)
1689        {        {
1690        i = GET(ecode, 1);        i = GET(ecode, 1);
1691        while (i-- > 0)        while (i-- > 0)
1692          {          {
1693          eptr--;          eptr--;
1694          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1695          BACKCHAR(eptr)          BACKCHAR(eptr);
1696          }          }
1697        }        }
1698      else      else
# Line 867  for (;;) Line 1705  for (;;)
1705        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1706        }        }
1707    
1708      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1709    
1710        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1711      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1712      break;      break;
1713    
# Line 877  for (;;) Line 1716  for (;;)
1716      function is able to force a failure. */      function is able to force a failure. */
1717    
1718      case OP_CALLOUT:      case OP_CALLOUT:
1719      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1720        {        {
1721        pcre_callout_block cb;        PUBL(callout_block) cb;
1722        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1723        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1724        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1725    #if defined COMPILE_PCRE8
1726        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1727        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1728        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1729        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1730          cb.subject          = (PCRE_SPTR32)md->start_subject;
1731    #endif
1732          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1733          cb.start_match      = (int)(mstart - md->start_subject);
1734          cb.current_position = (int)(eptr - md->start_subject);
1735        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1736        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1737        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1738        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1739          /* Internal change requires this for API compatibility. */
1740          if (cb.capture_last == 0) cb.capture_last = -1;
1741        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1742        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1743          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1744        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1745        }        }
1746      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 902  for (;;) Line 1750  for (;;)
1750      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
1751      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1752    
1753      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1754      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
1755      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
1756      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
1757      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
1758      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
1759      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.  
1760    
1761      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
1762      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
1763      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1764        a lot, so he is not to blame for the current way it works. */
1765    
1766      case OP_RECURSE:      case OP_RECURSE:
1767        {        {
1768          recursion_info *ri;
1769          unsigned int recno;
1770    
1771        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1772        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1773          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1774    
1775          /* Check for repeating a recursion without advancing the subject pointer.
1776          This should catch convoluted mutual recursions. (Some simple cases are
1777          caught at compile time.) */
1778    
1779          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1780            if (recno == ri->group_num && eptr == ri->subject_position)
1781              RRETURN(PCRE_ERROR_RECURSELOOP);
1782    
1783        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1784    
1785          new_recursive.group_num = recno;
1786          new_recursive.saved_capture_last = md->capture_last;
1787          new_recursive.subject_position = eptr;
1788        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1789        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1790    
1791        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1792    
1793        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1794    
1795        /* Now save the offset data. */        /* Now save the offset data */
1796    
1797        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1798        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 941  for (;;) Line 1800  for (;;)
1800        else        else
1801          {          {
1802          new_recursive.offset_save =          new_recursive.offset_save =
1803            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1804          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1805          }          }
   
1806        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1807              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1808    
1809        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1810        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1811          recursions, md->recursive might be changed, so reset it before looping.
1812          */
1813    
1814        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1815        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1816        do        do
1817          {          {
1818          RMATCH(rrc, eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1819            md, ims, eptrb, flags);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1820          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1821            memcpy(md->offset_vector, new_recursive.offset_save,
1822                new_recursive.saved_max * sizeof(int));
1823            md->capture_last = new_recursive.saved_capture_last;
1824            md->recursive = new_recursive.prevrec;
1825            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1826            {            {
1827            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1828            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1829              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1830            RRETURN(MATCH_MATCH);  
1831              /* Set where we got to in the subject, and reset the start in case
1832              it was changed by \K. This *is* propagated back out of a recursion,
1833              for Perl compatibility. */
1834    
1835              eptr = md->end_match_ptr;
1836              mstart = md->start_match_ptr;
1837              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1838            }            }
1839          else if (rrc != MATCH_NOMATCH)  
1840            /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1841            recursion; they are treated as NOMATCH. These codes are defined in a
1842            range that can be tested for. Any other return code is an error. */
1843    
1844            else if (rrc != MATCH_NOMATCH &&
1845                     (rrc < MATCH_BACKTRACK_MIN || rrc > MATCH_BACKTRACK_MAX))
1846            {            {
1847            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1848              if (new_recursive.offset_save != stacksave)
1849                (PUBL(free))(new_recursive.offset_save);
1850            RRETURN(rrc);            RRETURN(rrc);
1851            }            }
1852    
1853          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1854          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1855          }          }
1856        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 983  for (;;) Line 1858  for (;;)
1858        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1859        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1860        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1861          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1862        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1863        }        }
     /* 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, 0);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
1864    
1865      if (ecode[1+LINK_SIZE] == OP_OPT)      RECURSION_MATCHED:
1866        {      break;
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_cbegroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = match_tail_recursed;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1867    
1868      /* 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
1869      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1070  for (;;) Line 1872  for (;;)
1872      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1873      break;      break;
1874    
1875      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1876      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
1877      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
1878      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
1879      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1880    
1881      case OP_BRAZERO:      case OP_BRAZERO:
1882        {      next = ecode + 1;
1883        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1884        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1885        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1886        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1887      break;      break;
1888    
1889      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1890        {      next = ecode + 1;
1891        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1892        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1893        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1894        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1895      break;      break;
1896    
1897        case OP_SKIPZERO:
1898        next = ecode+1;
1899        do next += GET(next,1); while (*next == OP_ALT);
1900        ecode = next + 1 + LINK_SIZE;
1901        break;
1902    
1903        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1904        here; just jump to the group, with allow_zero set TRUE. */
1905    
1906        case OP_BRAPOSZERO:
1907        op = *(++ecode);
1908        allow_zero = TRUE;
1909        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1910          goto POSSESSIVE_NON_CAPTURE;
1911    
1912      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1913    
1914      case OP_KET:      case OP_KET:
1915      case OP_KETRMIN:      case OP_KETRMIN:
1916      case OP_KETRMAX:      case OP_KETRMAX:
1917        case OP_KETRPOS:
1918      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1919    
1920      /* If this was a group that remembered the subject start, in order to break      /* If this was a group that remembered the subject start, in order to break
1921      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1922      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1923    
1924      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1925        {        {
1926        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1927        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1928        }        }
1929      else saved_eptr = NULL;      else saved_eptr = NULL;
1930    
1931      /* If we are at the end of an assertion group, stop matching and return      /* If we are at the end of an assertion group or a non-capturing atomic
1932      MATCH_MATCH, but record the current high water mark for use by positive      group, stop matching and return MATCH_MATCH, but record the current high
1933      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1934        start in case it was changed by \K. */
1935      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1936          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1937          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1938        {        {
1939        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1940        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1941        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1942          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1943        }        }
1944    
1945      /* For capturing groups we have to check the group number back at the start      /* For capturing groups we have to check the group number back at the start
1946      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1947      bumping the high water mark. Note that whole-pattern recursion is coded as      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1948      a recurse into group 0, so it won't be picked up here. Instead, we catch it      into group 0, so it won't be picked up here. Instead, we catch it when the
1949      when the OP_END is reached. Other recursion is handled here. */      OP_END is reached. Other recursion is handled here. We just have to record
1950        the current subject position and start match pointer and give a MATCH
1951        return. */
1952    
1953      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1954            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1955        {        {
1956        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1957        offset = number << 1;        offset = number << 1;
1958    
1959  #ifdef DEBUG  #ifdef PCRE_DEBUG
1960        printf("end bracket %d", number);        printf("end bracket %d", number);
1961        printf("\n");        printf("\n");
1962  #endif  #endif
1963    
1964        md->capture_last = number;        /* Handle a recursively called group. */
1965        if (offset >= md->offset_max) md->offset_overflow = TRUE; else  
1966          if (md->recursive != NULL && md->recursive->group_num == number)
1967          {          {
1968          md->offset_vector[offset] =          md->end_match_ptr = eptr;
1969            md->offset_vector[md->offset_end - number];          md->start_match_ptr = mstart;
1970          md->offset_vector[offset+1] = eptr - md->start_subject;          RRETURN(MATCH_MATCH);
         if (offset_top <= offset) offset_top = offset + 2;  
1971          }          }
1972    
1973        /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
       appropriately and continue from after the call. */  
1974    
1975        if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
1976          if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1977          {          {
1978          recursion_info *rec = md->recursive;          /* If offset is greater than offset_top, it means that we are
1979          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          "skipping" a capturing group, and that group's offsets must be marked
1980          md->recursive = rec->prevrec;          unset. In earlier versions of PCRE, all the offsets were unset at the
1981          md->start_match = rec->save_start;          start of matching, but this doesn't work because atomic groups and
1982          memcpy(md->offset_vector, rec->offset_save,          assertions can cause a value to be set that should later be unset.
1983            rec->saved_max * sizeof(int));          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1984          ecode = rec->after_call;          part of the atomic group, but this is not on the final matching path,
1985          ims = original_ims;          so must be unset when 2 is set. (If there is no group 2, there is no
1986          break;          problem, because offset_top will then be 2, indicating no capture.) */
         }  
       }  
1987    
1988      /* For both capturing and non-capturing groups, reset the value of the ims          if (offset > offset_top)
1989      flags, in case they got changed during the group. */            {
1990              register int *iptr = md->offset_vector + offset_top;
1991              register int *iend = md->offset_vector + offset;
1992              while (iptr < iend) *iptr++ = -1;
1993              }
1994    
1995      ims = original_ims;          /* Now make the extraction */
     DPRINTF(("ims reset to %02lx\n", ims));  
1996    
1997      /* For a non-repeating ket, just continue at this level. This also          md->offset_vector[offset] =
1998      happens for a repeating ket if no characters were matched in the group.            md->offset_vector[md->offset_end - number];
1999      This is the forcible breaking of infinite loops as implemented in Perl          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2000      5.005. If there is an options reset, it will get obeyed in the normal          if (offset_top <= offset) offset_top = offset + 2;
2001      course of events. */          }
2002          }
2003    
2004        /* For an ordinary non-repeating ket, just continue at this level. This
2005        also happens for a repeating ket if no characters were matched in the
2006        group. This is the forcible breaking of infinite loops as implemented in
2007        Perl 5.005. For a non-repeating atomic group that includes captures,
2008        establish a backup point by processing the rest of the pattern at a lower
2009        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2010        original OP_ONCE level, thereby bypassing intermediate backup points, but
2011        resetting any captures that happened along the way. */
2012    
2013      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
2014        {        {
2015        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
2016            {
2017            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2018            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2019            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2020            RRETURN(MATCH_ONCE);
2021            }
2022          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2023        break;        break;
2024        }        }
2025    
2026      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2027      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2028      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
2029    
2030        if (*ecode == OP_KETRPOS)
2031          {
2032          md->end_match_ptr = eptr;
2033          md->end_offset_top = offset_top;
2034          RRETURN(MATCH_KETRPOS);
2035          }
2036    
2037      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      /* The normal repeating kets try the rest of the pattern or restart from
2038        the preceding bracket, in the appropriate order. In the second case, we can
2039        use tail recursion to avoid using another stack frame, unless we have an
2040        an atomic group or an unlimited repeat of a group that can match an empty
2041        string. */
2042    
2043      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2044        {        {
2045        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2046        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2047          if (*prev == OP_ONCE)
2048            {
2049            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2050            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2051            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2052            RRETURN(MATCH_ONCE);
2053            }
2054          if (*prev >= OP_SBRA)    /* Could match an empty string */
2055            {
2056            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2057            RRETURN(rrc);
2058            }
2059        ecode = prev;        ecode = prev;
       flags |= match_tail_recursed;  
2060        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2061        }        }
2062      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2063        {        {
2064        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, flags);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2065          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2066        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2067          if (*prev == OP_ONCE)
2068            {
2069            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2070            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2071            md->once_target = prev;
2072            RRETURN(MATCH_ONCE);
2073            }
2074        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = match_tail_recursed;  
2075        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2076        }        }
2077      /* Control never gets here */      /* Control never gets here */
2078    
2079      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2080    
2081      case OP_CIRC:      case OP_CIRC:
2082      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 == md->end_subject || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2083    
2084      /* Start of subject assertion */      /* Start of subject assertion */
2085    
# Line 1232  for (;;) Line 2088  for (;;)
2088      ecode++;      ecode++;
2089      break;      break;
2090    
2091        /* Multiline mode: start of subject unless notbol, or after any newline. */
2092    
2093        case OP_CIRCM:
2094        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2095        if (eptr != md->start_subject &&
2096            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2097          RRETURN(MATCH_NOMATCH);
2098        ecode++;
2099        break;
2100    
2101      /* Start of match assertion */      /* Start of match assertion */
2102    
2103      case OP_SOM:      case OP_SOM:
# Line 1239  for (;;) Line 2105  for (;;)
2105      ecode++;      ecode++;
2106      break;      break;
2107    
2108      /* 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. */  
2109    
2110      case OP_DOLL:      case OP_SET_SOM:
2111      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2112        ecode++;
2113        break;
2114    
2115        /* Multiline mode: assert before any newline, or before end of subject
2116        unless noteol is set. */
2117    
2118        case OP_DOLLM:
2119        if (eptr < md->end_subject)
2120        {        {
2121        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2122          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2123        else          if (md->partial != 0 &&
2124          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2125        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2126        break;              NLBLOCK->nllen == 2 &&
2127                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2128              {
2129              md->hitend = TRUE;
2130              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2131              }
2132            RRETURN(MATCH_NOMATCH);
2133            }
2134        }        }
2135      else      else
2136        {        {
2137        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2138        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2139        }        }
2140        ecode++;
2141        break;
2142    
2143        /* Not multiline mode: assert before a terminating newline or before end of
2144        subject unless noteol is set. */
2145    
2146        case OP_DOLL:
2147        if (md->noteol) RRETURN(MATCH_NOMATCH);
2148        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2149    
2150      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2151    
2152      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2153    
2154      case OP_EOD:      case OP_EOD:
2155      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2156        SCHECK_PARTIAL();
2157      ecode++;      ecode++;
2158      break;      break;
2159    
2160      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2161    
2162      case OP_EODN:      case OP_EODN:
2163      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2164        if (eptr < md->end_subject &&
2165          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2166          {
2167          if (md->partial != 0 &&
2168              eptr + 1 >= md->end_subject &&
2169              NLBLOCK->nltype == NLTYPE_FIXED &&
2170              NLBLOCK->nllen == 2 &&
2171              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2172            {
2173            md->hitend = TRUE;
2174            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2175            }
2176        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2177          }
2178    
2179        /* Either at end of string or \n before end. */
2180    
2181        SCHECK_PARTIAL();
2182      ecode++;      ecode++;
2183      break;      break;
2184    
# Line 1290  for (;;) Line 2190  for (;;)
2190    
2191        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2192        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
2193        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2194          partial matching. */
2195    
2196  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2197        if (utf8)        if (utf)
2198          {          {
2199            /* Get status of previous character */
2200    
2201          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2202            {            {
2203            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2204            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2205              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2206            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2207    #ifdef SUPPORT_UCP
2208              if (md->use_ucp)
2209                {
2210                if (c == '_') prev_is_word = TRUE; else
2211                  {
2212                  int cat = UCD_CATEGORY(c);
2213                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2214                  }
2215                }
2216              else
2217    #endif
2218            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2219            }            }
2220          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2221            /* Get status of next character */
2222    
2223            if (eptr >= md->end_subject)
2224              {
2225              SCHECK_PARTIAL();
2226              cur_is_word = FALSE;
2227              }
2228            else
2229            {            {
2230            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2231    #ifdef SUPPORT_UCP
2232              if (md->use_ucp)
2233                {
2234                if (c == '_') cur_is_word = TRUE; else
2235                  {
2236                  int cat = UCD_CATEGORY(c);
2237                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2238                  }
2239                }
2240              else
2241    #endif
2242            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2243            }            }
2244          }          }
2245        else        else
2246  #endif  #endif
2247    
2248        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2249          consistency with the behaviour of \w we do use it in this case. */
2250    
2251          {          {
2252          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2253            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2254          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2255            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2256              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2257    #ifdef SUPPORT_UCP
2258              if (md->use_ucp)
2259                {
2260                c = eptr[-1];
2261                if (c == '_') prev_is_word = TRUE; else
2262                  {
2263                  int cat = UCD_CATEGORY(c);
2264                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2265                  }
2266                }
2267              else
2268    #endif
2269              prev_is_word = MAX_255(eptr[-1])
2270                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2271              }
2272    
2273            /* Get status of next character */
2274    
2275            if (eptr >= md->end_subject)
2276              {
2277              SCHECK_PARTIAL();
2278              cur_is_word = FALSE;
2279              }
2280            else
2281    #ifdef SUPPORT_UCP
2282            if (md->use_ucp)
2283              {
2284              c = *eptr;
2285              if (c == '_') cur_is_word = TRUE; else
2286                {
2287                int cat = UCD_CATEGORY(c);
2288                cur_is_word = (cat == ucp_L || cat == ucp_N);
2289                }
2290              }
2291            else
2292    #endif
2293            cur_is_word = MAX_255(*eptr)
2294              && ((md->ctypes[*eptr] & ctype_word) != 0);
2295          }          }
2296    
2297        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1328  for (;;) Line 2302  for (;;)
2302        }        }
2303      break;      break;
2304    
2305      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2306        CRLF newlines and partial matching. */
2307    
2308      case OP_ANY:      case OP_ANY:
2309      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2310        if (md->partial != 0 &&
2311            eptr + 1 >= md->end_subject &&
2312            NLBLOCK->nltype == NLTYPE_FIXED &&
2313            NLBLOCK->nllen == 2 &&
2314            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2315        {        {
2316        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);        md->hitend = TRUE;
2317          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2318          }
2319    
2320        /* Fall through */
2321    
2322        /* Match any single character whatsoever. */
2323    
2324        case OP_ALLANY:
2325        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2326          {                            /* not be updated before SCHECK_PARTIAL. */
2327          SCHECK_PARTIAL();
2328          RRETURN(MATCH_NOMATCH);
2329        }        }
2330      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2331      if (utf8)  #ifdef SUPPORT_UTF
2332        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2333    #endif
2334      ecode++;      ecode++;
2335      break;      break;
2336    
# Line 1345  for (;;) Line 2338  for (;;)
2338      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2339    
2340      case OP_ANYBYTE:      case OP_ANYBYTE:
2341      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2342          {                            /* not be updated before SCHECK_PARTIAL. */
2343          SCHECK_PARTIAL();
2344          RRETURN(MATCH_NOMATCH);
2345          }
2346        eptr++;
2347      ecode++;      ecode++;
2348      break;      break;
2349    
2350      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
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_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1363  for (;;) Line 2365  for (;;)
2365      break;      break;
2366    
2367      case OP_DIGIT:      case OP_DIGIT:
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_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2379         )         )
# Line 1376  for (;;) Line 2382  for (;;)
2382      break;      break;
2383    
2384      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
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_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1389  for (;;) Line 2399  for (;;)
2399      break;      break;
2400    
2401      case OP_WHITESPACE:      case OP_WHITESPACE:
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_space) == 0         (md->ctypes[c] & ctype_space) == 0
2413         )         )
# Line 1402  for (;;) Line 2416  for (;;)
2416      break;      break;
2417    
2418      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2419      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2420          {
2421          SCHECK_PARTIAL();
2422          RRETURN(MATCH_NOMATCH);
2423          }
2424      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2425      if (      if (
2426  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2427         c < 256 &&         c < 256 &&
2428  #endif  #endif
2429         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1415  for (;;) Line 2433  for (;;)
2433      break;      break;
2434    
2435      case OP_WORDCHAR:      case OP_WORDCHAR:
2436      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2437          {
2438          SCHECK_PARTIAL();
2439          RRETURN(MATCH_NOMATCH);
2440          }
2441      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2442      if (      if (
2443  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2444         c >= 256 ||         c > 255 ||
2445  #endif  #endif
2446         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2447         )         )
# Line 1428  for (;;) Line 2450  for (;;)
2450      break;      break;
2451    
2452      case OP_ANYNL:      case OP_ANYNL:
2453      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2454          {
2455          SCHECK_PARTIAL();
2456          RRETURN(MATCH_NOMATCH);
2457          }
2458      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2459      switch(c)      switch(c)
2460        {        {
2461        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2462        case 0x000d:  
2463        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2464          if (eptr >= md->end_subject)
2465            {
2466            SCHECK_PARTIAL();
2467            }
2468          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2469          break;
2470    
2471          case CHAR_LF:
2472        break;        break;
2473        case 0x000a:  
2474        case 0x000b:        case CHAR_VT:
2475        case 0x000c:        case CHAR_FF:
2476        case 0x0085:        case CHAR_NEL:
2477    #ifndef EBCDIC
2478        case 0x2028:        case 0x2028:
2479        case 0x2029:        case 0x2029:
2480    #endif  /* Not EBCDIC */
2481          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2482        break;        break;
2483        }        }
2484      ecode++;      ecode++;
2485      break;      break;
2486    
2487        case OP_NOT_HSPACE:
2488        if (eptr >= md->end_subject)
2489          {
2490          SCHECK_PARTIAL();
2491          RRETURN(MATCH_NOMATCH);
2492          }
2493        GETCHARINCTEST(c, eptr);
2494        switch(c)
2495          {
2496          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2497          default: break;
2498          }
2499        ecode++;
2500        break;
2501    
2502        case OP_HSPACE:
2503        if (eptr >= md->end_subject)
2504          {
2505          SCHECK_PARTIAL();
2506          RRETURN(MATCH_NOMATCH);
2507          }
2508        GETCHARINCTEST(c, eptr);
2509        switch(c)
2510          {
2511          HSPACE_CASES: break;  /* Byte and multibyte cases */
2512          default: RRETURN(MATCH_NOMATCH);
2513          }
2514        ecode++;
2515        break;
2516    
2517        case OP_NOT_VSPACE:
2518        if (eptr >= md->end_subject)
2519          {
2520          SCHECK_PARTIAL();
2521          RRETURN(MATCH_NOMATCH);
2522          }
2523        GETCHARINCTEST(c, eptr);
2524        switch(c)
2525          {
2526          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2527          default: break;
2528          }
2529        ecode++;
2530        break;
2531    
2532        case OP_VSPACE:
2533        if (eptr >= md->end_subject)
2534          {
2535          SCHECK_PARTIAL();
2536          RRETURN(MATCH_NOMATCH);
2537          }
2538        GETCHARINCTEST(c, eptr);
2539        switch(c)
2540          {
2541          VSPACE_CASES: break;
2542          default: RRETURN(MATCH_NOMATCH);
2543          }
2544        ecode++;
2545        break;
2546    
2547  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2548      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2549      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. */
2550    
2551      case OP_PROP:      case OP_PROP:
2552      case OP_NOTPROP:      case OP_NOTPROP:
2553      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2554          {
2555          SCHECK_PARTIAL();
2556          RRETURN(MATCH_NOMATCH);
2557          }
2558      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2559        {        {
2560        int chartype, script;        const pcre_uint32 *cp;
2561        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2562    
2563        switch(ecode[1])        switch(ecode[1])
2564          {          {
# Line 1466  for (;;) Line 2567  for (;;)
2567          break;          break;
2568    
2569          case PT_LAMP:          case PT_LAMP:
2570          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2571               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2572               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2573            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2574           break;          break;
2575    
2576          case PT_GC:          case PT_GC:
2577          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2578            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2579          break;          break;
2580    
2581          case PT_PC:          case PT_PC:
2582          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2583            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2584          break;          break;
2585    
2586          case PT_SC:          case PT_SC:
2587          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2588              RRETURN(MATCH_NOMATCH);
2589            break;
2590    
2591            /* These are specials */
2592    
2593            case PT_ALNUM:
2594            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2595                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2596              RRETURN(MATCH_NOMATCH);
2597            break;
2598    
2599            case PT_SPACE:    /* Perl space */
2600            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2601                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2602                   == (op == OP_NOTPROP))
2603              RRETURN(MATCH_NOMATCH);
2604            break;
2605    
2606            case PT_PXSPACE:  /* POSIX space */
2607            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2608                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2609                 c == CHAR_FF || c == CHAR_CR)
2610                   == (op == OP_NOTPROP))
2611            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2612          break;          break;
2613    
2614            case PT_WORD:
2615            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2616                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2617                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2618              RRETURN(MATCH_NOMATCH);
2619            break;
2620    
2621            case PT_CLIST:
2622            cp = PRIV(ucd_caseless_sets) + ecode[2];
2623            for (;;)
2624              {
2625              if (c < *cp)
2626                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2627              if (c == *cp++)
2628                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2629              }
2630            break;
2631    
2632            case PT_UCNC:
2633            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2634                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2635                 c >= 0xe000) == (op == OP_NOTPROP))
2636              RRETURN(MATCH_NOMATCH);
2637            break;
2638    
2639            /* This should never occur */
2640    
2641          default:          default:
2642          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2643          }          }
# Line 1499  for (;;) Line 2650  for (;;)
2650      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2651    
2652      case OP_EXTUNI:      case OP_EXTUNI:
2653      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2654      GETCHARINCTEST(c, eptr);        {
2655          SCHECK_PARTIAL();
2656          RRETURN(MATCH_NOMATCH);
2657          }
2658        else
2659        {        {
2660        int chartype, script;        int lgb, rgb;
2661        int category = _pcre_ucp_findprop(c, &chartype, &script);        GETCHARINCTEST(c, eptr);
2662        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        lgb = UCD_GRAPHBREAK(c);
2663        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2664          {          {
2665          int len = 1;          int len = 1;
2666          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2667            {          rgb = UCD_GRAPHBREAK(c);
2668            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2669            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2670          eptr += len;          eptr += len;
2671          }          }
2672        }        }
2673        CHECK_PARTIAL();
2674      ecode++;      ecode++;
2675      break;      break;
2676  #endif  #endif  /* SUPPORT_UCP */
2677    
2678    
2679      /* 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 1530  for (;;) Line 2684  for (;;)
2684      it as matched, any number of times (otherwise there could be infinite      it as matched, any number of times (otherwise there could be infinite
2685      loops). */      loops). */
2686    
2687      case OP_REF:      case OP_REF:
2688        {      case OP_REFI:
2689        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2690        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2691        ecode += 1 + IMM2_SIZE;
2692    
2693        /* If the reference is unset, there are two possibilities:
2694    
2695        (a) In the default, Perl-compatible state, set the length negative;
2696        this ensures that every attempt at a match fails. We can't just fail
2697        here, because of the possibility of quantifiers with zero minima.
2698    
2699        (b) If the JavaScript compatibility flag is set, set the length to zero
2700        so that the back reference matches an empty string.
2701    
2702        /* If the reference is unset, set the length to be longer than the amount      Otherwise, set the length to the length of what was matched by the
2703        of subject left; this ensures that every attempt at a match fails. We      referenced subpattern. */
       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];  
2704    
2705        /* Set up for repetition, or handle the non-repeated case */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2706          length = (md->jscript_compat)? 0 : -1;
2707        else
2708          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2709    
2710        switch (*ecode)      /* Set up for repetition, or handle the non-repeated case */
         {  
         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;  
2711    
2712          case OP_CRRANGE:      switch (*ecode)
2713          case OP_CRMINRANGE:        {
2714          minimize = (*ecode == OP_CRMINRANGE);        case OP_CRSTAR:
2715          min = GET2(ecode, 1);        case OP_CRMINSTAR:
2716          max = GET2(ecode, 3);        case OP_CRPLUS:
2717          if (max == 0) max = INT_MAX;        case OP_CRMINPLUS:
2718          ecode += 5;        case OP_CRQUERY:
2719          break;        case OP_CRMINQUERY:
2720          c = *ecode++ - OP_CRSTAR;
2721          minimize = (c & 1) != 0;
2722          min = rep_min[c];                 /* Pick up values from tables; */
2723          max = rep_max[c];                 /* zero for max => infinity */
2724          if (max == 0) max = INT_MAX;
2725          break;
2726    
2727          default:               /* No repeat follows */        case OP_CRRANGE:
2728          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        case OP_CRMINRANGE:
2729          eptr += length;        minimize = (*ecode == OP_CRMINRANGE);
2730          continue;              /* With the main loop */        min = GET2(ecode, 1);
2731          max = GET2(ecode, 1 + IMM2_SIZE);
2732          if (max == 0) max = INT_MAX;
2733          ecode += 1 + 2 * IMM2_SIZE;
2734          break;
2735    
2736          default:               /* No repeat follows */
2737          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2738            {
2739            if (length == -2) eptr = md->end_subject;   /* Partial match */
2740            CHECK_PARTIAL();
2741            RRETURN(MATCH_NOMATCH);
2742          }          }
2743          eptr += length;
2744          continue;              /* With the main loop */
2745          }
2746    
2747        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2748        main loop. */      zero, just continue with the main loop. If the length is negative, it
2749        means the reference is unset in non-Java-compatible mode. If the minimum is
2750        zero, we can continue at the same level without recursion. For any other
2751        minimum, carrying on will result in NOMATCH. */
2752    
2753        if (length == 0) continue;      if (length == 0) continue;
2754        if (length < 0 && min == 0) continue;
2755    
2756        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2757        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2758        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2759    
2760        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2761          {
2762          int slength;
2763          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2764          {          {
2765          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2766          eptr += length;          CHECK_PARTIAL();
2767            RRETURN(MATCH_NOMATCH);
2768          }          }
2769          eptr += slength;
2770          }
2771    
2772        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2773        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2774    
2775        if (min == max) continue;      if (min == max) continue;
2776    
2777        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2778    
2779        if (minimize)      if (minimize)
2780          {
2781          for (fi = min;; fi++)
2782          {          {
2783          for (fi = min;; fi++)          int slength;
2784            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2785            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2786            if (fi >= max) RRETURN(MATCH_NOMATCH);
2787            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2788            {            {
2789            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2790            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2791            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2792            }            }
2793          /* Control never gets here */          eptr += slength;
2794          }          }
2795          /* Control never gets here */
2796          }
2797    
2798        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2799    
2800        else      else
2801          {
2802          pp = eptr;
2803          for (i = min; i < max; i++)
2804          {          {
2805          pp = eptr;          int slength;
2806          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)  
2807            {            {
2808            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2809            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2810            eptr -= length;  
2811              if (slength == -2 && md->partial != 0 &&
2812                  md->end_subject > md->start_used_ptr)
2813                {
2814                md->hitend = TRUE;
2815                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2816                }
2817              break;
2818            }            }
2819          RRETURN(MATCH_NOMATCH);          eptr += slength;
2820            }
2821    
2822          while (eptr >= pp)
2823            {
2824            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2825            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2826            eptr -= length;
2827          }          }
2828          RRETURN(MATCH_NOMATCH);
2829        }        }
2830      /* Control never gets here */      /* Control never gets here */
2831    
   
   
2832      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2833      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,
2834      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 1648  for (;;) Line 2843  for (;;)
2843      case OP_NCLASS:      case OP_NCLASS:
2844      case OP_CLASS:      case OP_CLASS:
2845        {        {
2846          /* The data variable is saved across frames, so the byte map needs to
2847          be stored there. */
2848    #define BYTE_MAP ((pcre_uint8 *)data)
2849        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2850        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2851    
2852        switch (*ecode)        switch (*ecode)
2853          {          {
# Line 1670  for (;;) Line 2868  for (;;)
2868          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2869          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2870          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2871          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2872          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2873          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2874          break;          break;
2875    
2876          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1682  for (;;) Line 2880  for (;;)
2880    
2881        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2882    
2883  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2884        /* UTF-8 mode */        if (utf)
       if (utf8)  
2885          {          {
2886          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2887            {            {
2888            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2889                {
2890                SCHECK_PARTIAL();
2891                RRETURN(MATCH_NOMATCH);
2892                }
2893            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2894            if (c > 255)            if (c > 255)
2895              {              {
2896              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2897              }              }
2898            else            else
2899              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2900            }            }
2901          }          }
2902        else        else
2903  #endif  #endif
2904        /* Not UTF-8 mode */        /* Not UTF mode */
2905          {          {
2906          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2907            {            {
2908            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2909                {
2910                SCHECK_PARTIAL();
2911                RRETURN(MATCH_NOMATCH);
2912                }
2913            c = *eptr++;            c = *eptr++;
2914            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2915              if (c > 255)
2916                {
2917                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2918                }
2919              else
2920    #endif
2921                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2922            }            }
2923          }          }
2924    
# Line 1722  for (;;) Line 2932  for (;;)
2932    
2933        if (minimize)        if (minimize)
2934          {          {
2935  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2936          /* UTF-8 mode */          if (utf)
         if (utf8)  
2937            {            {
2938            for (fi = min;; fi++)            for (fi = min;; fi++)
2939              {              {
2940              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2941              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2942              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2943                if (eptr >= md->end_subject)
2944                  {
2945                  SCHECK_PARTIAL();
2946                  RRETURN(MATCH_NOMATCH);
2947                  }
2948              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2949              if (c > 255)              if (c > 255)
2950                {                {
2951                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2952                }                }
2953              else              else
2954                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2955              }              }
2956            }            }
2957          else          else
2958  #endif  #endif
2959          /* Not UTF-8 mode */          /* Not UTF mode */
2960            {            {
2961            for (fi = min;; fi++)            for (fi = min;; fi++)
2962              {              {
2963              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2964              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2965              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2966                if (eptr >= md->end_subject)
2967                  {
2968                  SCHECK_PARTIAL();
2969                  RRETURN(MATCH_NOMATCH);
2970                  }
2971              c = *eptr++;              c = *eptr++;
2972              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2973                if (c > 255)
2974                  {
2975                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2976                  }
2977                else
2978    #endif
2979                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2980              }              }
2981            }            }
2982          /* Control never gets here */          /* Control never gets here */
# Line 1764  for (;;) Line 2988  for (;;)
2988          {          {
2989          pp = eptr;          pp = eptr;
2990    
2991  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2992          /* UTF-8 mode */          if (utf)
         if (utf8)  
2993            {            {
2994            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2995              {              {
2996              int len = 1;              int len = 1;
2997              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2998                  {
2999                  SCHECK_PARTIAL();
3000                  break;
3001                  }
3002              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
3003              if (c > 255)              if (c > 255)
3004                {                {
3005                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3006                }                }
3007              else              else
3008                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3009              eptr += len;              eptr += len;
3010              }              }
3011            for (;;)            for (;;)
3012              {              {
3013              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3014              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3015              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3016              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1793  for (;;) Line 3018  for (;;)
3018            }            }
3019          else          else
3020  #endif  #endif
3021            /* Not UTF-8 mode */            /* Not UTF mode */
3022            {            {
3023            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3024              {              {
3025              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3026                  {
3027                  SCHECK_PARTIAL();
3028                  break;
3029                  }
3030              c = *eptr;              c = *eptr;
3031              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3032                if (c > 255)
3033                  {
3034                  if (op == OP_CLASS) break;
3035                  }
3036                else
3037    #endif
3038                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3039              eptr++;              eptr++;
3040              }              }
3041            while (eptr >= pp)            while (eptr >= pp)
3042              {              {
3043              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3044              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3045              eptr--;              eptr--;
3046              }              }
# Line 1812  for (;;) Line 3048  for (;;)
3048    
3049          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3050          }          }
3051    #undef BYTE_MAP
3052        }        }
3053      /* Control never gets here */      /* Control never gets here */
3054    
3055    
3056      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3057      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
3058        mode, because Unicode properties are supported in non-UTF-8 mode. */
3059    
3060  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3061      case OP_XCLASS:      case OP_XCLASS:
3062        {        {
3063        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1844  for (;;) Line 3082  for (;;)
3082          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3083          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3084          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3085          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3086          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3087          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3088          break;          break;
3089    
3090          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1858  for (;;) Line 3096  for (;;)
3096    
3097        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3098          {          {
3099          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3100          GETCHARINC(c, eptr);            {
3101          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3102              RRETURN(MATCH_NOMATCH);
3103              }
3104            GETCHARINCTEST(c, eptr);
3105            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3106          }          }
3107    
3108        /* 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 1875  for (;;) Line 3117  for (;;)
3117          {          {
3118          for (fi = min;; fi++)          for (fi = min;; fi++)
3119            {            {
3120            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3121            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3122            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3123            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3124            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3125                SCHECK_PARTIAL();
3126                RRETURN(MATCH_NOMATCH);
3127                }
3128              GETCHARINCTEST(c, eptr);
3129              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3130            }            }
3131          /* Control never gets here */          /* Control never gets here */
3132          }          }
# Line 1892  for (;;) Line 3139  for (;;)
3139          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3140            {            {
3141            int len = 1;            int len = 1;
3142            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3143            GETCHARLEN(c, eptr, len);              {
3144            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3145                break;
3146                }
3147    #ifdef SUPPORT_UTF
3148              GETCHARLENTEST(c, eptr, len);
3149    #else
3150              c = *eptr;
3151    #endif
3152              if (!PRIV(xclass)(c, data, utf)) break;
3153            eptr += len;            eptr += len;
3154            }            }
3155          for(;;)          for(;;)
3156            {            {
3157            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3158            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3159            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3160            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3161              if (utf) BACKCHAR(eptr);
3162    #endif
3163            }            }
3164          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3165          }          }
# Line 1914  for (;;) Line 3171  for (;;)
3171      /* Match a single character, casefully */      /* Match a single character, casefully */
3172    
3173      case OP_CHAR:      case OP_CHAR:
3174  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3175      if (utf8)      if (utf)
3176        {        {
3177        length = 1;        length = 1;
3178        ecode++;        ecode++;
3179        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3180        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3181        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3182            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3183            RRETURN(MATCH_NOMATCH);
3184            }
3185          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3186        }        }
3187      else      else
3188  #endif  #endif
3189        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3190        {        {
3191        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3192            {
3193            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3194            RRETURN(MATCH_NOMATCH);
3195            }
3196        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3197        ecode += 2;        ecode += 2;
3198        }        }
3199      break;      break;
3200    
3201      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3202        subject, give up immediately. */
3203    
3204        case OP_CHARI:
3205        if (eptr >= md->end_subject)
3206          {
3207          SCHECK_PARTIAL();
3208          RRETURN(MATCH_NOMATCH);
3209          }
3210    
3211      case OP_CHARNC:  #ifdef SUPPORT_UTF
3212  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3213        {        {
3214        length = 1;        length = 1;
3215        ecode++;        ecode++;
3216        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3217    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3218        /* 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
3219        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3220          fast lookup table. We know that there is at least one byte left in the
3221          subject. */
3222    
3223        if (fc < 128)        if (fc < 128)
3224          {          {
3225          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3226            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3227            ecode++;
3228            eptr++;
3229          }          }
3230    
3231        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3232          use the value of "length" to check for sufficient bytes left, because the
3233          other case of the character may have more or fewer bytes.  */
3234    
3235        else        else
3236          {          {
3237          unsigned int dc;          pcre_uint32 dc;
3238          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3239          ecode += length;          ecode += length;
3240    
# Line 1968  for (;;) Line 3244  for (;;)
3244          if (fc != dc)          if (fc != dc)
3245            {            {
3246  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3247            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3248  #endif  #endif
3249              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3250            }            }
3251          }          }
3252        }        }
3253      else      else
3254  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3255    
3256      /* Non-UTF-8 mode */      /* Not UTF mode */
3257        {        {
3258        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3259        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3260          eptr++;
3261        ecode += 2;        ecode += 2;
3262        }        }
3263      break;      break;
# Line 1988  for (;;) Line 3265  for (;;)
3265      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3266    
3267      case OP_EXACT:      case OP_EXACT:
3268        case OP_EXACTI:
3269      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3270      ecode += 3;      ecode += 1 + IMM2_SIZE;
3271      goto REPEATCHAR;      goto REPEATCHAR;
3272    
3273      case OP_POSUPTO:      case OP_POSUPTO:
3274        case OP_POSUPTOI:
3275      possessive = TRUE;      possessive = TRUE;
3276      /* Fall through */      /* Fall through */
3277    
3278      case OP_UPTO:      case OP_UPTO:
3279        case OP_UPTOI:
3280      case OP_MINUPTO:      case OP_MINUPTO:
3281        case OP_MINUPTOI:
3282      min = 0;      min = 0;
3283      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3284      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3285      ecode += 3;      ecode += 1 + IMM2_SIZE;
3286      goto REPEATCHAR;      goto REPEATCHAR;
3287    
3288      case OP_POSSTAR:      case OP_POSSTAR:
3289        case OP_POSSTARI:
3290      possessive = TRUE;      possessive = TRUE;
3291      min = 0;      min = 0;
3292      max = INT_MAX;      max = INT_MAX;
# Line 2012  for (;;) Line 3294  for (;;)
3294      goto REPEATCHAR;      goto REPEATCHAR;
3295    
3296      case OP_POSPLUS:      case OP_POSPLUS:
3297        case OP_POSPLUSI:
3298      possessive = TRUE;      possessive = TRUE;
3299      min = 1;      min = 1;
3300      max = INT_MAX;      max = INT_MAX;
# Line 2019  for (;;) Line 3302  for (;;)
3302      goto REPEATCHAR;      goto REPEATCHAR;
3303    
3304      case OP_POSQUERY:      case OP_POSQUERY:
3305        case OP_POSQUERYI:
3306      possessive = TRUE;      possessive = TRUE;
3307      min = 0;      min = 0;
3308      max = 1;      max = 1;
# Line 2026  for (;;) Line 3310  for (;;)
3310      goto REPEATCHAR;      goto REPEATCHAR;
3311    
3312      case OP_STAR:      case OP_STAR:
3313        case OP_STARI:
3314      case OP_MINSTAR:      case OP_MINSTAR:
3315        case OP_MINSTARI:
3316      case OP_PLUS:      case OP_PLUS:
3317        case OP_PLUSI:
3318      case OP_MINPLUS:      case OP_MINPLUS:
3319        case OP_MINPLUSI:
3320      case OP_QUERY:      case OP_QUERY:
3321        case OP_QUERYI:
3322      case OP_MINQUERY:      case OP_MINQUERY:
3323      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3324        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3325      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3326      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3327      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3328      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3329    
3330      /* 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. */  
3331    
3332      REPEATCHAR:      REPEATCHAR:
3333  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3334      if (utf8)      if (utf)
3335        {        {
3336        length = 1;        length = 1;
3337        charptr = ecode;        charptr = ecode;
3338        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3339        ecode += length;        ecode += length;
3340    
3341        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2057  for (;;) Line 3344  for (;;)
3344        if (length > 1)        if (length > 1)
3345          {          {
3346  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3347          unsigned int othercase;          pcre_uint32 othercase;
3348          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3349              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3350            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3351          else oclength = 0;          else oclength = 0;
3352  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3353    
3354          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3355            {            {
3356            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3357                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3358  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3359            /* Need braces because of following else */            else if (oclength > 0 &&
3360            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3361                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3362    #endif  /* SUPPORT_UCP */
3363            else            else
3364              {              {
3365              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3366              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3367              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3368            }            }
3369    
3370          if (min == max) continue;          if (min == max) continue;
# Line 2086  for (;;) Line 3373  for (;;)
3373            {            {
3374            for (fi = min;; fi++)            for (fi = min;; fi++)
3375              {              {
3376              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3377              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3378              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3379              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3380                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3381  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3382              /* Need braces because of following else */              else if (oclength > 0 &&
3383              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3384                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3385    #endif  /* SUPPORT_UCP */
3386              else              else
3387                {                {
3388                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3389                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3390                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3391              }              }
3392            /* Control never gets here */            /* Control never gets here */
3393            }            }
# Line 2110  for (;;) Line 3397  for (;;)
3397            pp = eptr;            pp = eptr;
3398            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3399              {              {
3400              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3401              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3402  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3403              else if (oclength == 0) break;              else if (oclength > 0 &&
3404                         eptr <= md->end_subject - oclength &&
3405                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3406    #endif  /* SUPPORT_UCP */
3407              else              else
3408                {                {
3409                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3410                eptr += oclength;                break;
3411                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3412              }              }
3413    
3414            if (possessive) continue;            if (possessive) continue;
3415    
3416            for(;;)            for(;;)
3417             {              {
3418             RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3419             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3420             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3421  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3422             eptr--;              eptr--;
3423             BACKCHAR(eptr);              BACKCHAR(eptr);
3424  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3425             eptr -= length;              eptr -= length;
3426  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3427             }              }
3428            }            }
3429          /* Control never gets here */          /* Control never gets here */
3430          }          }
# Line 2146  for (;;) Line 3434  for (;;)
3434        value of fc will always be < 128. */        value of fc will always be < 128. */
3435        }        }
3436      else      else
3437  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3438          /* 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);  
3439        fc = *ecode++;        fc = *ecode++;
       }  
3440    
3441      /* 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
3442      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
3443      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3444      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3445      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2164  for (;;) Line 3448  for (;;)
3448      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3449    
3450      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3451        max, eptr));        max, (char *)eptr));
3452    
3453      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3454        {        {
3455        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3456          /* fc must be < 128 if UTF is enabled. */
3457          foc = md->fcc[fc];
3458    #else
3459    #ifdef SUPPORT_UTF
3460    #ifdef SUPPORT_UCP
3461          if (utf && fc > 127)
3462            foc = UCD_OTHERCASE(fc);
3463    #else
3464          if (utf && fc > 127)
3465            foc = fc;
3466    #endif /* SUPPORT_UCP */
3467          else
3468    #endif /* SUPPORT_UTF */
3469            foc = TABLE_GET(fc, md->fcc, fc);
3470    #endif /* COMPILE_PCRE8 */
3471    
3472        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3473          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3474            pcre_uint32 cc;                 /* Faster than pcre_uchar */
3475            if (eptr >= md->end_subject)
3476              {
3477              SCHECK_PARTIAL();
3478              RRETURN(MATCH_NOMATCH);
3479              }
3480            cc = RAWUCHARTEST(eptr);
3481            if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3482            eptr++;
3483            }
3484        if (min == max) continue;        if (min == max) continue;
3485        if (minimize)        if (minimize)
3486          {          {
3487          for (fi = min;; fi++)          for (fi = min;; fi++)
3488            {            {
3489            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            pcre_uint32 cc;               /* Faster than pcre_uchar */
3490              RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3491            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3492            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3493                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3494                {
3495                SCHECK_PARTIAL();
3496              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3497                }
3498              cc = RAWUCHARTEST(eptr);
3499              if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3500              eptr++;
3501            }            }
3502          /* Control never gets here */          /* Control never gets here */
3503          }          }
# Line 2189  for (;;) Line 3506  for (;;)
3506          pp = eptr;          pp = eptr;
3507          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3508            {            {
3509            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            pcre_uint32 cc;               /* Faster than pcre_uchar */
3510              if (eptr >= md->end_subject)
3511                {
3512                SCHECK_PARTIAL();
3513                break;
3514                }
3515              cc = RAWUCHARTEST(eptr);
3516              if (fc != cc && foc != cc) break;
3517            eptr++;            eptr++;
3518            }            }
3519