/[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 298 by ph10, Thu Jan 10 17:09:12 2008 UTC revision 1265 by ph10, Sun Mar 3 10:42:46 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. */
# Line 57  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  /* Flag bits for the match() function */  /* The md->capture_last field uses the lower 16 bits for the last captured
60    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    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  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
78  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
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 71  defined PCRE_ERROR_xxx codes, which are Line 86  defined PCRE_ERROR_xxx codes, which are
86  /* Special internal returns from the match() function. Make them sufficiently  /* Special internal returns from the match() function. Make them sufficiently
87  negative to avoid the external error codes. */  negative to avoid the external error codes. */
88    
89  #define MATCH_COMMIT       (-999)  #define MATCH_ACCEPT       (-999)
90  #define MATCH_PRUNE        (-998)  #define MATCH_COMMIT       (-998)
91  #define MATCH_SKIP         (-997)  #define MATCH_KETRPOS      (-997)
92  #define MATCH_THEN         (-996)  #define MATCH_ONCE         (-996)
93    #define MATCH_PRUNE        (-995)
94    #define MATCH_SKIP         (-994)
95    #define MATCH_SKIP_ARG     (-993)
96    #define MATCH_THEN         (-992)
97    
98  /* 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.
99  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,
# Line 87  because the offset vector is always a mu Line 106  because the offset vector is always a mu
106  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
107  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
108    
109    #ifdef PCRE_DEBUG
   
 #ifdef DEBUG  
110  /*************************************************  /*************************************************
111  *        Debugging function to print chars       *  *        Debugging function to print chars       *
112  *************************************************/  *************************************************/
# Line 107  Returns:     nothing Line 124  Returns:     nothing
124  */  */
125    
126  static void  static void
127  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)
128  {  {
129  unsigned int c;  pcre_uint32 c;
130    BOOL utf = md->utf;
131  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;
132  while (length-- > 0)  while (length-- > 0)
133    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);
134  }  }
135  #endif  #endif
136    
# Line 122  while (length-- > 0) Line 140  while (length-- > 0)
140  *          Match a back-reference                *  *          Match a back-reference                *
141  *************************************************/  *************************************************/
142    
143  /* 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
144  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
145    the length passed is zero. Note that in caseless UTF-8 mode, the number of
146    subject bytes matched may be different to the number of reference bytes.
147    
148  Arguments:  Arguments:
149    offset      index into the offset vector    offset      index into the offset vector
150    eptr        points into the subject    eptr        pointer into the subject
151    length      length to be matched    length      length of reference to be matched (number of bytes)
152    md          points to match data block    md          points to match data block
153    ims         the ims flags    caseless    TRUE if caseless
154    
155  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
156                  -1 no match
157                  -2 partial match; always given if at end subject
158  */  */
159    
160  static BOOL  static int
161  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,
162    unsigned long int ims)    BOOL caseless)
163  {  {
164  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
165    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
166    #ifdef SUPPORT_UTF
167    BOOL utf = md->utf;
168    #endif
169    
170  #ifdef DEBUG  #ifdef PCRE_DEBUG
171  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
172    printf("matching subject <null>");    printf("matching subject <null>");
173  else  else
# Line 154  pchars(p, length, FALSE, md); Line 180  pchars(p, length, FALSE, md);
180  printf("\n");  printf("\n");
181  #endif  #endif
182    
183  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
184    case the length is passed as zero). */
185    
186  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
187    
188  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
189    properly if Unicode properties are supported. Otherwise, we can check only
190    ASCII characters. */
191    
192  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
193    {    {
194    while (length-- > 0)  #ifdef SUPPORT_UTF
195      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
196      if (utf)
197        {
198        /* Match characters up to the end of the reference. NOTE: the number of
199        data units matched may differ, because in UTF-8 there are some characters
200        whose upper and lower case versions code have different numbers of bytes.
201        For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
202        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
203        sequence of two of the latter. It is important, therefore, to check the
204        length along the reference, not along the subject (earlier code did this
205        wrong). */
206    
207        PCRE_PUCHAR endptr = p + length;
208        while (p < endptr)
209          {
210          pcre_uint32 c, d;
211          const ucd_record *ur;
212          if (eptr >= md->end_subject) return -2;   /* Partial match */
213          GETCHARINC(c, eptr);
214          GETCHARINC(d, p);
215          ur = GET_UCD(d);
216          if (c != d && c != d + ur->other_case)
217            {
218            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
219            for (;;)
220              {
221              if (c < *pp) return -1;
222              if (c == *pp++) break;
223              }
224            }
225          }
226        }
227      else
228    #endif
229    #endif
230    
231      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
232      is no UCP support. */
233        {
234        while (length-- > 0)
235          {
236          pcre_uint32 cc, cp;
237          if (eptr >= md->end_subject) return -2;   /* Partial match */
238          cc = RAWUCHARTEST(eptr);
239          cp = RAWUCHARTEST(p);
240          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
241          p++;
242          eptr++;
243          }
244        }
245    }    }
246    
247    /* In the caseful case, we can just compare the bytes, whether or not we
248    are in UTF-8 mode. */
249    
250  else  else
251    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
252      while (length-- > 0)
253        {
254        if (eptr >= md->end_subject) return -2;   /* Partial match */
255        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
256        }
257      }
258    
259  return TRUE;  return (int)(eptr - eptr_start);
260  }  }
261    
262    
# Line 219  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 307  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
307         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
308         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
309         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
310         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
311           RM61,  RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
312    
313  /* 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
314  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
315  actuall used in this definition. */  actually used in this definition. */
316    
317  #ifndef NO_RECURSE  #ifndef NO_RECURSE
318  #define REGISTER register  #define REGISTER register
319    
320  #ifdef DEBUG  #ifdef PCRE_DEBUG
321  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
322    { \    { \
323    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
324    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
325    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
326    }    }
327  #define RRETURN(ra) \  #define RRETURN(ra) \
328    { \    { \
329    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
330    return ra; \    return ra; \
331    }    }
332  #else  #else
333  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
334    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
335  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
336  #endif  #endif
337    
# Line 255  argument of match(), which never changes Line 344  argument of match(), which never changes
344    
345  #define REGISTER  #define REGISTER
346    
347  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
348    {\    {\
349    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
350    frame->Xwhere = rw; \    if (newframe == NULL)\
351        {\
352        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
353        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
354        newframe->Xnextframe = NULL;\
355        frame->Xnextframe = newframe;\
356        }\
357      frame->Xwhere = rw;\
358    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
359    newframe->Xecode = rb;\    newframe->Xecode = rb;\
360    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
361    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
362    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
363    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
364    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
365    frame = newframe;\    frame = newframe;\
# Line 277  argument of match(), which never changes Line 371  argument of match(), which never changes
371    
372  #define RRETURN(ra)\  #define RRETURN(ra)\
373    {\    {\
374    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
375    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
376    if (frame != NULL)\    if (frame != NULL)\
377      {\      {\
378      rrc = ra;\      rrc = ra;\
# Line 293  argument of match(), which never changes Line 386  argument of match(), which never changes
386    
387  typedef struct heapframe {  typedef struct heapframe {
388    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
389      struct heapframe *Xnextframe;
390    
391    /* Function arguments that may change */    /* Function arguments that may change */
392    
393    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
394    const uschar *Xecode;    const pcre_uchar *Xecode;
395    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
396    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
397    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
398    unsigned int Xrdepth;    unsigned int Xrdepth;
399    
400    /* Function local variables */    /* Function local variables */
401    
402    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
403    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
404    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
405    const uschar *Xnext;  #endif
406    const uschar *Xpp;    PCRE_PUCHAR Xdata;
407    const uschar *Xprev;    PCRE_PUCHAR Xnext;
408    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
409      PCRE_PUCHAR Xprev;
410      PCRE_PUCHAR Xsaved_eptr;
411    
412    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
413    
# Line 321  typedef struct heapframe { Line 415  typedef struct heapframe {
415    BOOL Xcondition;    BOOL Xcondition;
416    BOOL Xprev_is_word;    BOOL Xprev_is_word;
417    
   unsigned long int Xoriginal_ims;  
   
418  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
419    int Xprop_type;    int Xprop_type;
420    int Xprop_value;    unsigned int Xprop_value;
421    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
422    int Xoclength;    int Xoclength;
423    uschar Xocchars[8];    pcre_uchar Xocchars[6];
424  #endif  #endif
425    
426      int Xcodelink;
427    int Xctype;    int Xctype;
428    unsigned int Xfc;    unsigned int Xfc;
429    int Xfi;    int Xfi;
430    int Xlength;    int Xlength;
431    int Xmax;    int Xmax;
432    int Xmin;    int Xmin;
433    int Xnumber;    unsigned int Xnumber;
434    int Xoffset;    int Xoffset;
435    int Xop;    unsigned int Xop;
436    int Xsave_capture_last;    pcre_int32 Xsave_capture_last;
437    int Xsave_offset1, Xsave_offset2, Xsave_offset3;    int Xsave_offset1, Xsave_offset2, Xsave_offset3;
438    int Xstacksave[REC_STACK_SAVE_MAX];    int Xstacksave[REC_STACK_SAVE_MAX];
439    
# Line 369  typedef struct heapframe { Line 459  typedef struct heapframe {
459    
460  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
461  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
462  same response.  same response. */
463    
464  Performance note: It might be tempting to extract commonly used fields from the  /* These macros pack up tests that are used for partial matching, and which
465  md structure (e.g. utf8, end_subject) into individual variables to improve  appear several times in the code. We set the "hit end" flag if the pointer is
466    at the end of the subject and also past the start of the subject (i.e.
467    something has been matched). For hard partial matching, we then return
468    immediately. The second one is used when we already know we are past the end of
469    the subject. */
470    
471    #define CHECK_PARTIAL()\
472      if (md->partial != 0 && eptr >= md->end_subject && \
473          eptr > md->start_used_ptr) \
474        { \
475        md->hitend = TRUE; \
476        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
477        }
478    
479    #define SCHECK_PARTIAL()\
480      if (md->partial != 0 && eptr > md->start_used_ptr) \
481        { \
482        md->hitend = TRUE; \
483        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
484        }
485    
486    
487    /* Performance note: It might be tempting to extract commonly used fields from
488    the md structure (e.g. utf, end_subject) into individual variables to improve
489  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
490  made performance worse.  made performance worse.
491    
# Line 383  Arguments: Line 496  Arguments:
496                   by encountering \K)                   by encountering \K)
497     offset_top  current top pointer     offset_top  current top pointer
498     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
499     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
500                   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  
501     rdepth      the recursion depth     rdepth      the recursion depth
502    
503  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
504                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
505                   a negative MATCH_xxx value for PRUNE, SKIP, etc
506                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
507                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
508  */  */
509    
510  static int  static int
511  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
512    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
513    int flags, unsigned int rdepth)    unsigned int rdepth)
514  {  {
515  /* 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,
516  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 409  so they can be ordinary variables in all Line 518  so they can be ordinary variables in all
518    
519  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
520  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
521  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
522  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
523    
524  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
525    BOOL caseless;
526    int condcode;
527    
528  /* 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
529  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
530  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
531  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
532    the top-level on the stack rather than malloc-ing them all gives a performance
533    boost in many cases where there is not much "recursion". */
534    
535  #ifdef NO_RECURSE  #ifdef NO_RECURSE
536  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
537    
538  /* Copy in the original argument variables */  /* Copy in the original argument variables */
539    
# Line 429  frame->Xeptr = eptr; Line 541  frame->Xeptr = eptr;
541  frame->Xecode = ecode;  frame->Xecode = ecode;
542  frame->Xmstart = mstart;  frame->Xmstart = mstart;
543  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
544  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
545  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
546    
547  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 444  HEAP_RECURSE: Line 554  HEAP_RECURSE:
554  #define ecode              frame->Xecode  #define ecode              frame->Xecode
555  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
556  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
557  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
558  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
559    
560  /* Ditto for the local variables */  /* Ditto for the local variables */
561    
562  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
563  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
564  #endif  #endif
565  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
566    #define codelink           frame->Xcodelink
567  #define data               frame->Xdata  #define data               frame->Xdata
568  #define next               frame->Xnext  #define next               frame->Xnext
569  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 467  HEAP_RECURSE: Line 576  HEAP_RECURSE:
576  #define condition          frame->Xcondition  #define condition          frame->Xcondition
577  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
578    
 #define original_ims       frame->Xoriginal_ims  
   
579  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
580  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
581  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
582  #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  
583  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
584  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
585  #endif  #endif
# Line 505  i, and fc and c, can be the same variabl Line 609  i, and fc and c, can be the same variabl
609  #define fi i  #define fi i
610  #define fc c  #define fc c
611    
612    /* Many of the following variables are used only in small blocks of the code.
613    My normal style of coding would have declared them within each of those blocks.
614    However, in order to accommodate the version of this code that uses an external
615    "stack" implemented on the heap, it is easier to declare them all here, so the
616    declarations can be cut out in a block. The only declarations within blocks
617    below are for variables that do not have to be preserved over a recursive call
618    to RMATCH(). */
619    
620    #ifdef SUPPORT_UTF
621    const pcre_uchar *charptr;
622    #endif
623    const pcre_uchar *callpat;
624    const pcre_uchar *data;
625    const pcre_uchar *next;
626    PCRE_PUCHAR       pp;
627    const pcre_uchar *prev;
628    PCRE_PUCHAR       saved_eptr;
629    
630  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
631  const uschar *charptr;             /* in small blocks of the code. My normal */  
632  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
633  BOOL condition;  BOOL condition;
634  BOOL prev_is_word;  BOOL prev_is_word;
635    
 unsigned long int original_ims;  
   
636  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
637  int prop_type;  int prop_type;
638  int prop_value;  unsigned int prop_value;
639  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
640  int oclength;  int oclength;
641  uschar occhars[8];  pcre_uchar occhars[6];
642  #endif  #endif
643    
644    int codelink;
645  int ctype;  int ctype;
646  int length;  int length;
647  int max;  int max;
648  int min;  int min;
649  int number;  unsigned int number;
650  int offset;  int offset;
651  int op;  unsigned int op;
652  int save_capture_last;  pcre_int32 save_capture_last;
653  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
654  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
655    
656  eptrblock newptrb;  eptrblock newptrb;
657    
658    /* There is a special fudge for calling match() in a way that causes it to
659    measure the size of its basic stack frame when the stack is being used for
660    recursion. The second argument (ecode) being NULL triggers this behaviour. It
661    cannot normally ever be NULL. The return is the negated value of the frame
662    size. */
663    
664    if (ecode == NULL)
665      {
666      if (rdepth == 0)
667        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
668      else
669        {
670        int len = (char *)&rdepth - (char *)eptr;
671        return (len > 0)? -len : len;
672        }
673      }
674  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
675    
676    /* To save space on the stack and in the heap frame, I have doubled up on some
677    of the local variables that are used only in localised parts of the code, but
678    still need to be preserved over recursive calls of match(). These macros define
679    the alternative names that are used. */
680    
681    #define allow_zero    cur_is_word
682    #define cbegroup      condition
683    #define code_offset   codelink
684    #define condassert    condition
685    #define matched_once  prev_is_word
686    #define foc           number
687    #define save_mark     data
688    
689  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
690  variables. */  variables. */
691    
# Line 568  TAIL_RECURSE: Line 705  TAIL_RECURSE:
705  /* 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
706  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
707  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()
708  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
709  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
710  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,
711  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
712    
713  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
714  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
715  #else  #else
716  utf8 = FALSE;  utf = FALSE;
717  #endif  #endif
718    
719  /* 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
# Line 585  haven't exceeded the recursive call limi Line 722  haven't exceeded the recursive call limi
722  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
723  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
724    
 original_ims = ims;    /* Save for resetting on ')' */  
   
725  /* 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
726  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
727  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
728  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
729  When match() is called in other circumstances, don't add to the chain. The  
730  match_cbegroup flag must NOT be used with tail recursion, because the memory  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
731  block that is used is on the stack, so a new one may be required for each  such remembered pointers, to be checked when we hit the closing ket, in order
732  match(). */  to break infinite loops that match no characters. When match() is called in
733    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
734    NOT be used with tail recursion, because the memory block that is used is on
735    the stack, so a new one may be required for each match(). */
736    
737  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
738    {    {
739    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
740    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
741    eptrb = &newptrb;    eptrb = &newptrb;
742      md->match_function_type = 0;
743    }    }
744    
745  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 610  for (;;) Line 749  for (;;)
749    minimize = possessive = FALSE;    minimize = possessive = FALSE;
750    op = *ecode;    op = *ecode;
751    
   /* 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 > mstart)  
     md->hitend = TRUE;  
   
752    switch(op)    switch(op)
753      {      {
754        case OP_MARK:
755        md->nomatch_mark = ecode + 2;
756        md->mark = NULL;    /* In case previously set by assertion */
757        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
758          eptrb, RM55);
759        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
760             md->mark == NULL) md->mark = ecode + 2;
761    
762        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
763        argument, and we must check whether that argument matches this MARK's
764        argument. It is passed back in md->start_match_ptr (an overloading of that
765        variable). If it does match, we reset that variable to the current subject
766        position and return MATCH_SKIP. Otherwise, pass back the return code
767        unaltered. */
768    
769        else if (rrc == MATCH_SKIP_ARG &&
770            STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
771          {
772          md->start_match_ptr = eptr;
773          RRETURN(MATCH_SKIP);
774          }
775        RRETURN(rrc);
776    
777      case OP_FAIL:      case OP_FAIL:
778      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
779    
780      case OP_PRUNE:      /* COMMIT overrides PRUNE, SKIP, and THEN */
     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,  
       ims, eptrb, flags, RM51);  
     if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
     RRETURN(MATCH_PRUNE);  
781    
782      case OP_COMMIT:      case OP_COMMIT:
783      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
784        ims, eptrb, flags, RM52);        eptrb, RM52);
785      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
786            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
787            rrc != MATCH_THEN)
788          RRETURN(rrc);
789      RRETURN(MATCH_COMMIT);      RRETURN(MATCH_COMMIT);
790    
791        /* PRUNE overrides THEN */
792    
793        case OP_PRUNE:
794        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
795          eptrb, RM51);
796        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
797        RRETURN(MATCH_PRUNE);
798    
799        case OP_PRUNE_ARG:
800        md->nomatch_mark = ecode + 2;
801        md->mark = NULL;    /* In case previously set by assertion */
802        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
803          eptrb, RM56);
804        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
805             md->mark == NULL) md->mark = ecode + 2;
806        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
807        RRETURN(MATCH_PRUNE);
808    
809        /* SKIP overrides PRUNE and THEN */
810    
811      case OP_SKIP:      case OP_SKIP:
812      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
813        ims, eptrb, flags, RM53);        eptrb, RM53);
814      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
815          RRETURN(rrc);
816      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
817      RRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
818    
819        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
820        nomatch_mark. There is a flag that disables this opcode when re-matching a
821        pattern that ended with a SKIP for which there was not a matching MARK. */
822    
823        case OP_SKIP_ARG:
824        if (md->ignore_skip_arg)
825          {
826          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
827          break;
828          }
829        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
830          eptrb, RM57);
831        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
832          RRETURN(rrc);
833    
834        /* Pass back the current skip name by overloading md->start_match_ptr and
835        returning the special MATCH_SKIP_ARG return code. This will either be
836        caught by a matching MARK, or get to the top, where it causes a rematch
837        with the md->ignore_skip_arg flag set. */
838    
839        md->start_match_ptr = ecode + 2;
840        RRETURN(MATCH_SKIP_ARG);
841    
842        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
843        the branch in which it occurs can be determined. Overload the start of
844        match pointer to do this. */
845    
846      case OP_THEN:      case OP_THEN:
847      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
848        ims, eptrb, flags, RM54);        eptrb, RM54);
849      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
850        md->start_match_ptr = ecode;
851      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
852    
853      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_THEN_ARG:
854      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
855      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
856      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,
857      reference inside the group.        md, eptrb, RM58);
858        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
859      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
860      values of the final offsets, in case they were set by a previous iteration      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
861      of the same bracket.      md->start_match_ptr = ecode;
862        RRETURN(MATCH_THEN);
863    
864        /* Handle an atomic group that does not contain any capturing parentheses.
865        This can be handled like an assertion. Prior to 8.13, all atomic groups
866        were handled this way. In 8.13, the code was changed as below for ONCE, so
867        that backups pass through the group and thereby reset captured values.
868        However, this uses a lot more stack, so in 8.20, atomic groups that do not
869        contain any captures generate OP_ONCE_NC, which can be handled in the old,
870        less stack intensive way.
871    
872        Check the alternative branches in turn - the matching won't pass the KET
873        for this kind of subpattern. If any one branch matches, we carry on as at
874        the end of a normal bracket, leaving the subject pointer, but resetting
875        the start-of-match value in case it was changed by \K. */
876    
877        case OP_ONCE_NC:
878        prev = ecode;
879        saved_eptr = eptr;
880        save_mark = md->mark;
881        do
882          {
883          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
884          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
885            {
886            mstart = md->start_match_ptr;
887            break;
888            }
889          if (rrc == MATCH_THEN)
890            {
891            next = ecode + GET(ecode,1);
892            if (md->start_match_ptr < next &&
893                (*ecode == OP_ALT || *next == OP_ALT))
894              rrc = MATCH_NOMATCH;
895            }
896    
897          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
898          ecode += GET(ecode,1);
899          md->mark = save_mark;
900          }
901        while (*ecode == OP_ALT);
902    
903        /* If hit the end of the group (which could be repeated), fail */
904    
905        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
906    
907        /* Continue as from after the group, updating the offsets high water
908        mark, since extracts may have been taken. */
909    
910        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
911    
912        offset_top = md->end_offset_top;
913        eptr = md->end_match_ptr;
914    
915        /* For a non-repeating ket, just continue at this level. This also
916        happens for a repeating ket if no characters were matched in the group.
917        This is the forcible breaking of infinite loops as implemented in Perl
918        5.005. */
919    
920        if (*ecode == OP_KET || eptr == saved_eptr)
921          {
922          ecode += 1+LINK_SIZE;
923          break;
924          }
925    
926        /* The repeating kets try the rest of the pattern or restart from the
927        preceding bracket, in the appropriate order. The second "call" of match()
928        uses tail recursion, to avoid using another stack frame. */
929    
930        if (*ecode == OP_KETRMIN)
931          {
932          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
933          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
934          ecode = prev;
935          goto TAIL_RECURSE;
936          }
937        else  /* OP_KETRMAX */
938          {
939          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
940          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
941          ecode += 1 + LINK_SIZE;
942          goto TAIL_RECURSE;
943          }
944        /* Control never gets here */
945    
946        /* Handle a capturing bracket, other than those that are possessive with an
947        unlimited repeat. If there is space in the offset vector, save the current
948        subject position in the working slot at the top of the vector. We mustn't
949        change the current values of the data slot, because they may be set from a
950        previous iteration of this group, and be referred to by a reference inside
951        the group. A failure to match might occur after the group has succeeded,
952        if something later on doesn't match. For this reason, we need to restore
953        the working value and also the values of the final offsets, in case they
954        were set by a previous iteration of the same bracket.
955    
956      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
957      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 667  for (;;) Line 962  for (;;)
962      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
963      offset = number << 1;      offset = number << 1;
964    
965  #ifdef DEBUG  #ifdef PCRE_DEBUG
966      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
967      printf("subject=");      printf("subject=");
968      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 680  for (;;) Line 975  for (;;)
975        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
976        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
977        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
978          save_mark = md->mark;
979    
980        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
981        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
982            (int)(eptr - md->start_subject);
983    
984        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
985          {          {
986          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
987            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
988          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
989            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
990    
991            /* If we backed up to a THEN, check whether it is within the current
992            branch by comparing the address of the THEN that is passed back with
993            the end of the branch. If it is within the current branch, and the
994            branch is one of two or more alternatives (it either starts or ends
995            with OP_ALT), we have reached the limit of THEN's action, so convert
996            the return code to NOMATCH, which will cause normal backtracking to
997            happen from now on. Otherwise, THEN is passed back to an outer
998            alternative. This implements Perl's treatment of parenthesized groups,
999            where a group not containing | does not affect the current alternative,
1000            that is, (X) is NOT the same as (X|(*F)). */
1001    
1002            if (rrc == MATCH_THEN)
1003              {
1004              next = ecode + GET(ecode,1);
1005              if (md->start_match_ptr < next &&
1006                  (*ecode == OP_ALT || *next == OP_ALT))
1007                rrc = MATCH_NOMATCH;
1008              }
1009    
1010            /* Anything other than NOMATCH is passed back. */
1011    
1012            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1013          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1014          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1015            md->mark = save_mark;
1016            if (*ecode != OP_ALT) break;
1017          }          }
       while (*ecode == OP_ALT);  
1018    
1019        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1020        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1021        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1022        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1023    
1024        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1025    
1026          RRETURN(rrc);
1027        }        }
1028    
1029      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 715  for (;;) Line 1037  for (;;)
1037      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1038      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1039    
1040      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1041      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion, except in the case when match_cbegroup  
     is set.*/  
1042    
1043        When we get to the final alternative within the brackets, we used to return
1044        the result of a recursive call to match() whatever happened so it was
1045        possible to reduce stack usage by turning this into a tail recursion,
1046        except in the case of a possibly empty group. However, now that there is
1047        the possiblity of (*THEN) occurring in the final alternative, this
1048        optimization is no longer always possible.
1049    
1050        We can optimize if we know there are no (*THEN)s in the pattern; at present
1051        this is the best that can be done.
1052    
1053        MATCH_ONCE is returned when the end of an atomic group is successfully
1054        reached, but subsequent matching fails. It passes back up the tree (causing
1055        captured values to be reset) until the original atomic group level is
1056        reached. This is tested by comparing md->once_target with the start of the
1057        group. At this point, the return is converted into MATCH_NOMATCH so that
1058        previous backup points can be taken. */
1059    
1060        case OP_ONCE:
1061      case OP_BRA:      case OP_BRA:
1062      case OP_SBRA:      case OP_SBRA:
1063      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1064      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1065      for (;;)      for (;;)
1066        {        {
1067        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE)
1068          {          md->match_function_type = MATCH_CBEGROUP;
         if (flags == 0)    /* Not a possibly empty group */  
           {  
           ecode += _pcre_OP_lengths[*ecode];  
           DPRINTF(("bracket 0 tail recursion\n"));  
           goto TAIL_RECURSE;  
           }  
1069    
1070          /* Possibly empty group; can't use tail recursion. */        /* If this is not a possibly empty group, and there are no (*THEN)s in
1071          the pattern, and this is the final alternative, optimize as described
1072          above. */
1073    
1074          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1075            eptrb, flags, RM48);          {
1076          RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1077            goto TAIL_RECURSE;
1078          }          }
1079    
1080        /* 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. */  
1081    
1082        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1083          eptrb, flags, RM2);        save_capture_last = md->capture_last;
1084        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1085            RM2);
1086    
1087          /* See comment in the code for capturing groups above about handling
1088          THEN. */
1089    
1090          if (rrc == MATCH_THEN)
1091            {
1092            next = ecode + GET(ecode,1);
1093            if (md->start_match_ptr < next &&
1094                (*ecode == OP_ALT || *next == OP_ALT))
1095              rrc = MATCH_NOMATCH;
1096            }
1097    
1098          if (rrc != MATCH_NOMATCH)
1099            {
1100            if (rrc == MATCH_ONCE)
1101              {
1102              const pcre_uchar *scode = ecode;
1103              if (*scode != OP_ONCE)           /* If not at start, find it */
1104                {
1105                while (*scode == OP_ALT) scode += GET(scode, 1);
1106                scode -= GET(scode, 1);
1107                }
1108              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1109              }
1110            RRETURN(rrc);
1111            }
1112        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1113          md->mark = save_mark;
1114          if (*ecode != OP_ALT) break;
1115          md->capture_last = save_capture_last;
1116        }        }
     /* Control never reaches here. */  
1117    
1118      /* Conditional group: compilation checked that there are no more than      RRETURN(MATCH_NOMATCH);
     two branches. If the condition is false, skipping the first branch takes us  
     past the end if there is only one branch, but that's OK because that is  
     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. */  
1119    
1120      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1121      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1122      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1123        {      different. The end of these brackets will always be OP_KETRPOS, which
1124        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1125        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1126          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
1127        ecode += condition? 3 : GET(ecode, 1);  
1128        }      case OP_CBRAPOS:
1129        case OP_SCBRAPOS:
1130        allow_zero = FALSE;
1131    
1132      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      POSSESSIVE_CAPTURE:
1133        {      number = GET2(ecode, 1+LINK_SIZE);
1134        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      offset = number << 1;
       condition = offset < offset_top && md->offset_vector[offset] >= 0;  
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1135    
1136      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */  #ifdef PCRE_DEBUG
1137        printf("start possessive bracket %d\n", number);
1138        printf("subject=");
1139        pchars(eptr, 16, TRUE, md);
1140        printf("\n");
1141    #endif
1142    
1143        if (offset < md->offset_max)
1144        {        {
1145        condition = FALSE;        matched_once = FALSE;
1146        ecode += GET(ecode, 1);        code_offset = (int)(ecode - md->start_code);
       }  
1147    
1148      /* The condition is an assertion. Call match() to evaluate it - setting        save_offset1 = md->offset_vector[offset];
1149      the final argument match_condassert causes it to stop at the end of an        save_offset2 = md->offset_vector[offset+1];
1150      assertion. */        save_offset3 = md->offset_vector[md->offset_end - number];
1151          save_capture_last = md->capture_last;
1152    
1153      else        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1154        {  
1155        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        /* Each time round the loop, save the current subject position for use
1156            match_condassert, RM3);        when the group matches. For MATCH_MATCH, the group has matched, so we
1157        if (rrc == MATCH_MATCH)        restart it with a new subject starting position, remembering that we had
1158          {        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1159          condition = TRUE;        usual. If we haven't matched any alternatives in any iteration, check to
1160          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        see if a previous iteration matched. If so, the group has matched;
1161          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        continue from afterwards. Otherwise it has failed; restore the previous
1162          capture values before returning NOMATCH. */
1163    
1164          for (;;)
1165            {
1166            md->offset_vector[md->offset_end - number] =
1167              (int)(eptr - md->start_subject);
1168            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1169            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1170              eptrb, RM63);
1171            if (rrc == MATCH_KETRPOS)
1172              {
1173              offset_top = md->end_offset_top;
1174              eptr = md->end_match_ptr;
1175              ecode = md->start_code + code_offset;
1176              save_capture_last = md->capture_last;
1177              matched_once = TRUE;
1178              continue;
1179              }
1180    
1181            /* See comment in the code for capturing groups above about handling
1182            THEN. */
1183    
1184            if (rrc == MATCH_THEN)
1185              {
1186              next = ecode + GET(ecode,1);
1187              if (md->start_match_ptr < next &&
1188                  (*ecode == OP_ALT || *next == OP_ALT))
1189                rrc = MATCH_NOMATCH;
1190              }
1191    
1192            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1193            md->capture_last = save_capture_last;
1194            ecode += GET(ecode, 1);
1195            if (*ecode != OP_ALT) break;
1196          }          }
1197        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1198          if (!matched_once)
1199          {          {
1200          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1201            md->offset_vector[offset+1] = save_offset2;
1202            md->offset_vector[md->offset_end - number] = save_offset3;
1203          }          }
1204        else  
1205          if (allow_zero || matched_once)
1206          {          {
1207          condition = FALSE;          ecode += 1 + LINK_SIZE;
1208          ecode += GET(ecode, 1);          break;
1209          }          }
1210    
1211          RRETURN(MATCH_NOMATCH);
1212        }        }
1213    
1214      /* We are now at the branch that is to be obeyed. As there is only one,      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1215      we can use tail recursion to avoid using another stack frame, except when      as a non-capturing bracket. */
     match_cbegroup is required for an unlimited repeat of a possibly empty  
     group. If the second alternative doesn't exist, we can just plough on. */  
1216    
1217      if (condition || *ecode == OP_ALT)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1218        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1219    
1220        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1221    
1222        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1223        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1224    
1225        /* Non-capturing possessive bracket with unlimited repeat. We come here
1226        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1227        without the capturing complication. It is written out separately for speed
1228        and cleanliness. */
1229    
1230        case OP_BRAPOS:
1231        case OP_SBRAPOS:
1232        allow_zero = FALSE;
1233    
1234        POSSESSIVE_NON_CAPTURE:
1235        matched_once = FALSE;
1236        code_offset = (int)(ecode - md->start_code);
1237        save_capture_last = md->capture_last;
1238    
1239        for (;;)
1240        {        {
1241        ecode += 1 + LINK_SIZE;        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1242        if (op == OP_SCOND)        /* Possibly empty group */        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1243          {          eptrb, RM48);
1244          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);        if (rrc == MATCH_KETRPOS)
1245          RRETURN(rrc);          {
1246          }          offset_top = md->end_offset_top;
1247        else                       /* Group must match something */          eptr = md->end_match_ptr;
1248          {          ecode = md->start_code + code_offset;
1249          flags = 0;          matched_once = TRUE;
1250          goto TAIL_RECURSE;          continue;
1251            }
1252    
1253          /* See comment in the code for capturing groups above about handling
1254          THEN. */
1255    
1256          if (rrc == MATCH_THEN)
1257            {
1258            next = ecode + GET(ecode,1);
1259            if (md->start_match_ptr < next &&
1260                (*ecode == OP_ALT || *next == OP_ALT))
1261              rrc = MATCH_NOMATCH;
1262          }          }
1263    
1264          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1265          ecode += GET(ecode, 1);
1266          if (*ecode != OP_ALT) break;
1267          md->capture_last = save_capture_last;
1268        }        }
1269      else                         /* Condition false & no 2nd alternative */  
1270        if (matched_once || allow_zero)
1271        {        {
1272        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1273          break;
1274        }        }
1275      break;      RRETURN(MATCH_NOMATCH);
1276    
1277        /* Control never reaches here. */
1278    
1279        /* Conditional group: compilation checked that there are no more than
1280        two branches. If the condition is false, skipping the first branch takes us
1281        past the end if there is only one branch, but that's OK because that is
1282        exactly what going to the ket would do. */
1283    
1284      /* End of the pattern, either real or forced. If we are in a top-level      case OP_COND:
1285      recursion, we should restore the offsets appropriately and continue from      case OP_SCOND:
1286      after the call. */      codelink = GET(ecode, 1);
1287    
1288      case OP_ACCEPT:      /* Because of the way auto-callout works during compile, a callout item is
1289      case OP_END:      inserted between OP_COND and an assertion condition. */
1290      if (md->recursive != NULL && md->recursive->group_num == 0)  
1291        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1292        {        {
1293        recursion_info *rec = md->recursive;        if (PUBL(callout) != NULL)
1294        DPRINTF(("End of pattern in a (?0) recursion\n"));          {
1295        md->recursive = rec->prevrec;          PUBL(callout_block) cb;
1296        memmove(md->offset_vector, rec->offset_save,          cb.version          = 2;   /* Version 1 of the callout block */
1297          rec->saved_max * sizeof(int));          cb.callout_number   = ecode[LINK_SIZE+2];
1298        mstart = rec->save_start;          cb.offset_vector    = md->offset_vector;
1299        ims = original_ims;  #if defined COMPILE_PCRE8
1300        ecode = rec->after_call;          cb.subject          = (PCRE_SPTR)md->start_subject;
1301        break;  #elif defined COMPILE_PCRE16
1302            cb.subject          = (PCRE_SPTR16)md->start_subject;
1303    #elif defined COMPILE_PCRE32
1304            cb.subject          = (PCRE_SPTR32)md->start_subject;
1305    #endif
1306            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1307            cb.start_match      = (int)(mstart - md->start_subject);
1308            cb.current_position = (int)(eptr - md->start_subject);
1309            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1310            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1311            cb.capture_top      = offset_top/2;
1312            cb.capture_last     = md->capture_last & CAPLMASK;
1313            /* Internal change requires this for API compatibility. */
1314            if (cb.capture_last == 0) cb.capture_last = -1;
1315            cb.callout_data     = md->callout_data;
1316            cb.mark             = md->nomatch_mark;
1317            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1318            if (rrc < 0) RRETURN(rrc);
1319            }
1320          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1321          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1322        }        }
1323    
1324      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      condcode = ecode[LINK_SIZE+1];
     string - backtracking will then try other alternatives, if any. */  
   
     if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);  
     md->end_match_ptr = eptr;           /* Record where we ended */  
     md->end_offset_top = offset_top;    /* and how many extracts were taken */  
     md->start_match_ptr = mstart;       /* and the start (\K can modify) */  
     RRETURN(MATCH_MATCH);  
1325    
1326      /* Change option settings */      /* Now see what the actual condition is */
1327    
1328      case OP_OPT:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1329      ims = ecode[1];        {
1330      ecode += 2;        if (md->recursive == NULL)                /* Not recursing => FALSE */
1331      DPRINTF(("ims set to %02lx\n", ims));          {
1332      break;          condition = FALSE;
1333            ecode += GET(ecode, 1);
1334            }
1335          else
1336            {
1337            unsigned int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1338            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1339    
1340            /* If the test is for recursion into a specific subpattern, and it is
1341            false, but the test was set up by name, scan the table to see if the
1342            name refers to any other numbers, and test them. The condition is true
1343            if any one is set. */
1344    
1345            if (!condition && condcode == OP_NRREF)
1346              {
1347              pcre_uchar *slotA = md->name_table;
1348              for (i = 0; i < md->name_count; i++)
1349                {
1350                if (GET2(slotA, 0) == recno) break;
1351                slotA += md->name_entry_size;
1352                }
1353    
1354              /* Found a name for the number - there can be only one; duplicate
1355              names for different numbers are allowed, but not vice versa. First
1356              scan down for duplicates. */
1357    
1358              if (i < md->name_count)
1359                {
1360                pcre_uchar *slotB = slotA;
1361                while (slotB > md->name_table)
1362                  {
1363                  slotB -= md->name_entry_size;
1364                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1365                    {
1366                    condition = GET2(slotB, 0) == md->recursive->group_num;
1367                    if (condition) break;
1368                    }
1369                  else break;
1370                  }
1371    
1372                /* Scan up for duplicates */
1373    
1374                if (!condition)
1375                  {
1376                  slotB = slotA;
1377                  for (i++; i < md->name_count; i++)
1378                    {
1379                    slotB += md->name_entry_size;
1380                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1381                      {
1382                      condition = GET2(slotB, 0) == md->recursive->group_num;
1383                      if (condition) break;
1384                      }
1385                    else break;
1386                    }
1387                  }
1388                }
1389              }
1390    
1391            /* Chose branch according to the condition */
1392    
1393            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1394            }
1395          }
1396    
1397        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1398          {
1399          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1400          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1401    
1402          /* If the numbered capture is unset, but the reference was by name,
1403          scan the table to see if the name refers to any other numbers, and test
1404          them. The condition is true if any one is set. This is tediously similar
1405          to the code above, but not close enough to try to amalgamate. */
1406    
1407          if (!condition && condcode == OP_NCREF)
1408            {
1409            unsigned int refno = offset >> 1;
1410            pcre_uchar *slotA = md->name_table;
1411    
1412            for (i = 0; i < md->name_count; i++)
1413              {
1414              if (GET2(slotA, 0) == refno) break;
1415              slotA += md->name_entry_size;
1416              }
1417    
1418            /* Found a name for the number - there can be only one; duplicate names
1419            for different numbers are allowed, but not vice versa. First scan down
1420            for duplicates. */
1421    
1422            if (i < md->name_count)
1423              {
1424              pcre_uchar *slotB = slotA;
1425              while (slotB > md->name_table)
1426                {
1427                slotB -= md->name_entry_size;
1428                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1429                  {
1430                  offset = GET2(slotB, 0) << 1;
1431                  condition = offset < offset_top &&
1432                    md->offset_vector[offset] >= 0;
1433                  if (condition) break;
1434                  }
1435                else break;
1436                }
1437    
1438              /* Scan up for duplicates */
1439    
1440              if (!condition)
1441                {
1442                slotB = slotA;
1443                for (i++; i < md->name_count; i++)
1444                  {
1445                  slotB += md->name_entry_size;
1446                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1447                    {
1448                    offset = GET2(slotB, 0) << 1;
1449                    condition = offset < offset_top &&
1450                      md->offset_vector[offset] >= 0;
1451                    if (condition) break;
1452                    }
1453                  else break;
1454                  }
1455                }
1456              }
1457            }
1458    
1459          /* Chose branch according to the condition */
1460    
1461          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1462          }
1463    
1464        else if (condcode == OP_DEF)     /* DEFINE - always false */
1465          {
1466          condition = FALSE;
1467          ecode += GET(ecode, 1);
1468          }
1469    
1470        /* The condition is an assertion. Call match() to evaluate it - setting
1471        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1472        an assertion. */
1473    
1474        else
1475          {
1476          md->match_function_type = MATCH_CONDASSERT;
1477          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1478          if (rrc == MATCH_MATCH)
1479            {
1480            if (md->end_offset_top > offset_top)
1481              offset_top = md->end_offset_top;  /* Captures may have happened */
1482            condition = TRUE;
1483            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1484            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1485            }
1486    
1487          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1488          assertion; it is therefore treated as NOMATCH. */
1489    
1490          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1491            {
1492            RRETURN(rrc);         /* Need braces because of following else */
1493            }
1494          else
1495            {
1496            condition = FALSE;
1497            ecode += codelink;
1498            }
1499          }
1500    
1501        /* We are now at the branch that is to be obeyed. As there is only one, can
1502        use tail recursion to avoid using another stack frame, except when there is
1503        unlimited repeat of a possibly empty group. In the latter case, a recursive
1504        call to match() is always required, unless the second alternative doesn't
1505        exist, in which case we can just plough on. Note that, for compatibility
1506        with Perl, the | in a conditional group is NOT treated as creating two
1507        alternatives. If a THEN is encountered in the branch, it propagates out to
1508        the enclosing alternative (unless nested in a deeper set of alternatives,
1509        of course). */
1510    
1511        if (condition || *ecode == OP_ALT)
1512          {
1513          if (op != OP_SCOND)
1514            {
1515            ecode += 1 + LINK_SIZE;
1516            goto TAIL_RECURSE;
1517            }
1518    
1519          md->match_function_type = MATCH_CBEGROUP;
1520          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1521          RRETURN(rrc);
1522          }
1523    
1524         /* Condition false & no alternative; continue after the group. */
1525    
1526        else
1527          {
1528          ecode += 1 + LINK_SIZE;
1529          }
1530        break;
1531    
1532    
1533        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1534        to close any currently open capturing brackets. */
1535    
1536        case OP_CLOSE:
1537        number = GET2(ecode, 1);   /* Must be less than 65536 */
1538        offset = number << 1;
1539    
1540    #ifdef PCRE_DEBUG
1541          printf("end bracket %d at *ACCEPT", number);
1542          printf("\n");
1543    #endif
1544    
1545        md->capture_last = (md->capture_last & OVFLMASK) | number;
1546        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1547          {
1548          md->offset_vector[offset] =
1549            md->offset_vector[md->offset_end - number];
1550          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1551          if (offset_top <= offset) offset_top = offset + 2;
1552          }
1553        ecode += 1 + IMM2_SIZE;
1554        break;
1555    
1556    
1557        /* End of the pattern, either real or forced. */
1558    
1559        case OP_END:
1560        case OP_ACCEPT:
1561        case OP_ASSERT_ACCEPT:
1562    
1563        /* If we have matched an empty string, fail if not in an assertion and not
1564        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1565        is set and we have matched at the start of the subject. In both cases,
1566        backtracking will then try other alternatives, if any. */
1567    
1568        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1569             md->recursive == NULL &&
1570             (md->notempty ||
1571               (md->notempty_atstart &&
1572                 mstart == md->start_subject + md->start_offset)))
1573          RRETURN(MATCH_NOMATCH);
1574    
1575        /* Otherwise, we have a match. */
1576    
1577        md->end_match_ptr = eptr;           /* Record where we ended */
1578        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1579        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1580    
1581        /* For some reason, the macros don't work properly if an expression is
1582        given as the argument to RRETURN when the heap is in use. */
1583    
1584        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1585        RRETURN(rrc);
1586    
1587      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1588      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
1589      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1590      start of each branch to move the current point backwards, so the code at      start of each branch to move the current point backwards, so the code at
1591      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1592        of a condition, we want to return immediately afterwards. The caller of
1593        this incarnation of the match() function will have set MATCH_CONDASSERT in
1594        md->match_function type, and one of these opcodes will be the first opcode
1595        that is processed. We use a local variable that is preserved over calls to
1596        match() to remember this case. */
1597    
1598      case OP_ASSERT:      case OP_ASSERT:
1599      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1600        save_mark = md->mark;
1601        if (md->match_function_type == MATCH_CONDASSERT)
1602          {
1603          condassert = TRUE;
1604          md->match_function_type = 0;
1605          }
1606        else condassert = FALSE;
1607    
1608      do      do
1609        {        {
1610        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1611          RM4);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1612        if (rrc == MATCH_MATCH) break;          {
1613            mstart = md->start_match_ptr;   /* In case \K reset it */
1614            break;
1615            }
1616          md->mark = save_mark;
1617    
1618          /* A COMMIT failure must fail the entire assertion, without trying any
1619          subsequent branches. */
1620    
1621          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1622    
1623          /* PCRE does not allow THEN to escape beyond an assertion; it
1624          is treated as NOMATCH. */
1625    
1626        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1627        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1628        }        }
1629      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1630    
1631      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1632    
1633      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1634    
1635      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1636    
1637      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1638      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 900  for (;;) Line 1642  for (;;)
1642      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1643      continue;      continue;
1644    
1645      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1646        PRUNE, or COMMIT means we must assume failure without checking subsequent
1647        branches. */
1648    
1649      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1650      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1651        save_mark = md->mark;
1652        if (md->match_function_type == MATCH_CONDASSERT)
1653          {
1654          condassert = TRUE;
1655          md->match_function_type = 0;
1656          }
1657        else condassert = FALSE;
1658    
1659      do      do
1660        {        {
1661        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1662          RM5);        md->mark = save_mark;
1663        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1664          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1665            {
1666            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1667            break;
1668            }
1669    
1670          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1671          as NOMATCH. */
1672    
1673        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1674        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1675        }        }
1676      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1677    
1678      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1679    
1680      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1681      continue;      continue;
# Line 925  for (;;) Line 1686  for (;;)
1686      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1687    
1688      case OP_REVERSE:      case OP_REVERSE:
1689  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1690      if (utf8)      if (utf)
1691        {        {
1692        i = GET(ecode, 1);        i = GET(ecode, 1);
1693        while (i-- > 0)        while (i-- > 0)
# Line 946  for (;;) Line 1707  for (;;)
1707        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1708        }        }
1709    
1710      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1711    
1712        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1713      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1714      break;      break;
1715    
# Line 956  for (;;) Line 1718  for (;;)
1718      function is able to force a failure. */      function is able to force a failure. */
1719    
1720      case OP_CALLOUT:      case OP_CALLOUT:
1721      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1722        {        {
1723        pcre_callout_block cb;        PUBL(callout_block) cb;
1724        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1725        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1726        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1727    #if defined COMPILE_PCRE8
1728        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1729        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1730        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1731        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1732          cb.subject          = (PCRE_SPTR32)md->start_subject;
1733    #endif
1734          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1735          cb.start_match      = (int)(mstart - md->start_subject);
1736          cb.current_position = (int)(eptr - md->start_subject);
1737        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1738        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1739        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1740        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1741          /* Internal change requires this for API compatibility. */
1742          if (cb.capture_last == 0) cb.capture_last = -1;
1743        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1744        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1745          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1746        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1747        }        }
1748      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 981  for (;;) Line 1752  for (;;)
1752      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
1753      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1754    
1755      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1756      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
1757      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
1758      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
1759      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
1760      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
1761      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.  
1762    
1763      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
1764      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
1765      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1766        a lot, so he is not to blame for the current way it works. */
1767    
1768      case OP_RECURSE:      case OP_RECURSE:
1769        {        {
1770          recursion_info *ri;
1771          unsigned int recno;
1772    
1773        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1774        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1775          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1776    
1777          /* Check for repeating a recursion without advancing the subject pointer.
1778          This should catch convoluted mutual recursions. (Some simple cases are
1779          caught at compile time.) */
1780    
1781          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1782            if (recno == ri->group_num && eptr == ri->subject_position)
1783              RRETURN(PCRE_ERROR_RECURSELOOP);
1784    
1785        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1786    
1787          new_recursive.group_num = recno;
1788          new_recursive.saved_capture_last = md->capture_last;
1789          new_recursive.subject_position = eptr;
1790        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1791        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1792    
1793        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1794    
1795        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1796    
1797        /* Now save the offset data. */        /* Now save the offset data */
1798    
1799        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1800        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1020  for (;;) Line 1802  for (;;)
1802        else        else
1803          {          {
1804          new_recursive.offset_save =          new_recursive.offset_save =
1805            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1806          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1807          }          }
   
1808        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1809              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1810    
1811        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1812        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1813          recursions, md->recursive might be changed, so reset it before looping.
1814          */
1815    
1816        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1817        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1818        do        do
1819          {          {
1820          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1821            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1822          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1823            memcpy(md->offset_vector, new_recursive.offset_save,
1824                new_recursive.saved_max * sizeof(int));
1825            md->capture_last = new_recursive.saved_capture_last;
1826            md->recursive = new_recursive.prevrec;
1827            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1828            {            {
1829            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1830            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1831              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1832            RRETURN(MATCH_MATCH);  
1833              /* Set where we got to in the subject, and reset the start in case
1834              it was changed by \K. This *is* propagated back out of a recursion,
1835              for Perl compatibility. */
1836    
1837              eptr = md->end_match_ptr;
1838              mstart = md->start_match_ptr;
1839              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1840            }            }
1841          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1842            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1843            is treated as NOMATCH. */
1844    
1845            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1846                     rrc != MATCH_COMMIT)
1847            {            {
1848            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1849              if (new_recursive.offset_save != stacksave)
1850                (PUBL(free))(new_recursive.offset_save);
1851            RRETURN(rrc);            RRETURN(rrc);
1852            }            }
1853    
1854          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1855          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1856          }          }
1857        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1062  for (;;) Line 1859  for (;;)
1859        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1860        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1861        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1862          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1863        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1864        }        }
     /* Control never reaches here */  
1865    
1866      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1867      the point in the subject string is not moved back. Thus there can never be      break;
     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(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) 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. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1868    
1869      /* 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
1870      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1148  for (;;) Line 1873  for (;;)
1873      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1874      break;      break;
1875    
1876      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1877      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
1878      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
1879      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
1880      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1881    
1882      case OP_BRAZERO:      case OP_BRAZERO:
1883        {      next = ecode + 1;
1884        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1885        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1886        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1887        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1888      break;      break;
1889    
1890      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1891        {      next = ecode + 1;
1892        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1893        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1894        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1895        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
1896        ecode++;      break;
1897        }  
1898        case OP_SKIPZERO:
1899        next = ecode+1;
1900        do next += GET(next,1); while (*next == OP_ALT);
1901        ecode = next + 1 + LINK_SIZE;
1902      break;      break;
1903    
1904        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1905        here; just jump to the group, with allow_zero set TRUE. */
1906    
1907        case OP_BRAPOSZERO:
1908        op = *(++ecode);
1909        allow_zero = TRUE;
1910        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1911          goto POSSESSIVE_NON_CAPTURE;
1912    
1913      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1914    
1915      case OP_KET:      case OP_KET:
1916      case OP_KETRMIN:      case OP_KETRMIN:
1917      case OP_KETRMAX:      case OP_KETRMAX:
1918        case OP_KETRPOS:
1919      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1920    
1921      /* 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
1922      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1923      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1924    
1925      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1926        {        {
1927        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1928        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1929        }        }
1930      else saved_eptr = NULL;      else saved_eptr = NULL;
1931    
1932      /* 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
1933      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
1934      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1935        start in case it was changed by \K. */
1936      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1937          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1938          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1939        {        {
1940        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1941        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1942        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1943          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1944        }        }
1945    
1946      /* 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
1947      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1948      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
1949      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
1950      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
1951        the current subject position and start match pointer and give a MATCH
1952        return. */
1953    
1954      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1955            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1956        {        {
1957        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1958        offset = number << 1;        offset = number << 1;
1959    
1960  #ifdef DEBUG  #ifdef PCRE_DEBUG
1961        printf("end bracket %d", number);        printf("end bracket %d", number);
1962        printf("\n");        printf("\n");
1963  #endif  #endif
1964    
1965        md->capture_last = number;        /* Handle a recursively called group. */
1966        if (offset >= md->offset_max) md->offset_overflow = TRUE; else  
1967          if (md->recursive != NULL && md->recursive->group_num == number)
1968          {          {
1969          md->offset_vector[offset] =          md->end_match_ptr = eptr;
1970            md->offset_vector[md->offset_end - number];          md->start_match_ptr = mstart;
1971          md->offset_vector[offset+1] = eptr - md->start_subject;          RRETURN(MATCH_MATCH);
         if (offset_top <= offset) offset_top = offset + 2;  
1972          }          }
1973    
1974        /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
       appropriately and continue from after the call. */  
1975    
1976        if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
1977          if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1978          {          {
1979          recursion_info *rec = md->recursive;          /* If offset is greater than offset_top, it means that we are
1980          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          "skipping" a capturing group, and that group's offsets must be marked
1981          md->recursive = rec->prevrec;          unset. In earlier versions of PCRE, all the offsets were unset at the
1982          mstart = rec->save_start;          start of matching, but this doesn't work because atomic groups and
1983          memcpy(md->offset_vector, rec->offset_save,          assertions can cause a value to be set that should later be unset.
1984            rec->saved_max * sizeof(int));          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1985          ecode = rec->after_call;          part of the atomic group, but this is not on the final matching path,
1986          ims = original_ims;          so must be unset when 2 is set. (If there is no group 2, there is no
1987          break;          problem, because offset_top will then be 2, indicating no capture.) */
         }  
       }  
1988    
1989      /* For both capturing and non-capturing groups, reset the value of the ims          if (offset > offset_top)
1990      flags, in case they got changed during the group. */            {
1991              register int *iptr = md->offset_vector + offset_top;
1992              register int *iend = md->offset_vector + offset;
1993              while (iptr < iend) *iptr++ = -1;
1994              }
1995    
1996      ims = original_ims;          /* Now make the extraction */
     DPRINTF(("ims reset to %02lx\n", ims));  
1997    
1998      /* For a non-repeating ket, just continue at this level. This also          md->offset_vector[offset] =
1999      happens for a repeating ket if no characters were matched in the group.            md->offset_vector[md->offset_end - number];
2000      This is the forcible breaking of infinite loops as implemented in Perl          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2001      5.005. If there is an options reset, it will get obeyed in the normal          if (offset_top <= offset) offset_top = offset + 2;
2002      course of events. */          }
2003          }
2004    
2005        /* For an ordinary non-repeating ket, just continue at this level. This
2006        also happens for a repeating ket if no characters were matched in the
2007        group. This is the forcible breaking of infinite loops as implemented in
2008        Perl 5.005. For a non-repeating atomic group that includes captures,
2009        establish a backup point by processing the rest of the pattern at a lower
2010        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2011        original OP_ONCE level, thereby bypassing intermediate backup points, but
2012        resetting any captures that happened along the way. */
2013    
2014      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
2015        {        {
2016        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
2017            {
2018            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2019            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2020            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2021            RRETURN(MATCH_ONCE);
2022            }
2023          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2024        break;        break;
2025        }        }
2026    
2027      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2028      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
2029      tail recursion to avoid using another stack frame, unless we have an      at a time from the outer level, thus saving stack. */
     unlimited repeat of a group that can match an empty string. */  
2030    
2031      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      if (*ecode == OP_KETRPOS)
2032          {
2033          md->end_match_ptr = eptr;
2034          md->end_offset_top = offset_top;
2035          RRETURN(MATCH_KETRPOS);
2036          }
2037    
2038        /* The normal repeating kets try the rest of the pattern or restart from
2039        the preceding bracket, in the appropriate order. In the second case, we can
2040        use tail recursion to avoid using another stack frame, unless we have an
2041        an atomic group or an unlimited repeat of a group that can match an empty
2042        string. */
2043    
2044      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2045        {        {
2046        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2047        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2048        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2049            {
2050            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2051            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2052            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2053            RRETURN(MATCH_ONCE);
2054            }
2055          if (*prev >= OP_SBRA)    /* Could match an empty string */
2056          {          {
2057          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2058          RRETURN(rrc);          RRETURN(rrc);
2059          }          }
2060        ecode = prev;        ecode = prev;
# Line 1286  for (;;) Line 2062  for (;;)
2062        }        }
2063      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2064        {        {
2065        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2066          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2067        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2068          if (*prev == OP_ONCE)
2069            {
2070            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2071            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2072            md->once_target = prev;
2073            RRETURN(MATCH_ONCE);
2074            }
2075        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2076        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2077        }        }
2078      /* Control never gets here */      /* Control never gets here */
2079    
2080      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2081    
2082      case OP_CIRC:      case OP_CIRC:
2083      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 */  
2084    
2085      /* Start of subject assertion */      /* Start of subject assertion */
2086    
# Line 1315  for (;;) Line 2089  for (;;)
2089      ecode++;      ecode++;
2090      break;      break;
2091    
2092        /* Multiline mode: start of subject unless notbol, or after any newline. */
2093    
2094        case OP_CIRCM:
2095        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2096        if (eptr != md->start_subject &&
2097            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2098          RRETURN(MATCH_NOMATCH);
2099        ecode++;
2100        break;
2101    
2102      /* Start of match assertion */      /* Start of match assertion */
2103    
2104      case OP_SOM:      case OP_SOM:
# Line 1329  for (;;) Line 2113  for (;;)
2113      ecode++;      ecode++;
2114      break;      break;
2115    
2116      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2117      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2118    
2119      case OP_DOLL:      case OP_DOLLM:
2120      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2121        {        {
2122        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2123          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2124        else          if (md->partial != 0 &&
2125          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2126        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2127        break;              NLBLOCK->nllen == 2 &&
2128                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2129              {
2130              md->hitend = TRUE;
2131              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2132              }
2133            RRETURN(MATCH_NOMATCH);
2134            }
2135        }        }
2136      else      else
2137        {        {
2138        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2139        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2140        }        }
2141        ecode++;
2142        break;
2143    
2144        /* Not multiline mode: assert before a terminating newline or before end of
2145        subject unless noteol is set. */
2146    
2147        case OP_DOLL:
2148        if (md->noteol) RRETURN(MATCH_NOMATCH);
2149        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2150    
2151      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2152    
2153      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2154    
2155      case OP_EOD:      case OP_EOD:
2156      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2157        SCHECK_PARTIAL();
2158      ecode++;      ecode++;
2159      break;      break;
2160    
2161      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2162    
2163      case OP_EODN:      case OP_EODN:
2164      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2165        if (eptr < md->end_subject &&
2166          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2167          {
2168          if (md->partial != 0 &&
2169              eptr + 1 >= md->end_subject &&
2170              NLBLOCK->nltype == NLTYPE_FIXED &&
2171              NLBLOCK->nllen == 2 &&
2172              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2173            {
2174            md->hitend = TRUE;
2175            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2176            }
2177        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2178          }
2179    
2180        /* Either at end of string or \n before end. */
2181    
2182        SCHECK_PARTIAL();
2183      ecode++;      ecode++;
2184      break;      break;
2185    
# Line 1380  for (;;) Line 2191  for (;;)
2191    
2192        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2193        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
2194        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2195          partial matching. */
2196    
2197  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2198        if (utf8)        if (utf)
2199          {          {
2200            /* Get status of previous character */
2201    
2202          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2203            {            {
2204            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2205            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2206              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2207            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2208    #ifdef SUPPORT_UCP
2209              if (md->use_ucp)
2210                {
2211                if (c == '_') prev_is_word = TRUE; else
2212                  {
2213                  int cat = UCD_CATEGORY(c);
2214                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2215                  }
2216                }
2217              else
2218    #endif
2219            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2220            }            }
2221          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2222            /* Get status of next character */
2223    
2224            if (eptr >= md->end_subject)
2225              {
2226              SCHECK_PARTIAL();
2227              cur_is_word = FALSE;
2228              }
2229            else
2230            {            {
2231            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2232    #ifdef SUPPORT_UCP
2233              if (md->use_ucp)
2234                {
2235                if (c == '_') cur_is_word = TRUE; else
2236                  {
2237                  int cat = UCD_CATEGORY(c);
2238                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2239                  }
2240                }
2241              else
2242    #endif
2243            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2244            }            }
2245          }          }
2246        else        else
2247  #endif  #endif
2248    
2249        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2250          consistency with the behaviour of \w we do use it in this case. */
2251    
2252          {          {
2253          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2254            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2255          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2256            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2257              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2258    #ifdef SUPPORT_UCP
2259              if (md->use_ucp)
2260                {
2261                c = eptr[-1];
2262                if (c == '_') prev_is_word = TRUE; else
2263                  {
2264                  int cat = UCD_CATEGORY(c);
2265                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2266                  }
2267                }
2268              else
2269    #endif
2270              prev_is_word = MAX_255(eptr[-1])
2271                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2272              }
2273    
2274            /* Get status of next character */
2275    
2276            if (eptr >= md->end_subject)
2277              {
2278              SCHECK_PARTIAL();
2279              cur_is_word = FALSE;
2280              }
2281            else
2282    #ifdef SUPPORT_UCP
2283            if (md->use_ucp)
2284              {
2285              c = *eptr;
2286              if (c == '_') cur_is_word = TRUE; else
2287                {
2288                int cat = UCD_CATEGORY(c);
2289                cur_is_word = (cat == ucp_L || cat == ucp_N);
2290                }
2291              }
2292            else
2293    #endif
2294            cur_is_word = MAX_255(*eptr)
2295              && ((md->ctypes[*eptr] & ctype_word) != 0);
2296          }          }
2297    
2298        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1418  for (;;) Line 2303  for (;;)
2303        }        }
2304      break;      break;
2305    
2306      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2307        CRLF newlines and partial matching. */
2308    
2309      case OP_ANY:      case OP_ANY:
2310      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2311        if (md->partial != 0 &&
2312            eptr + 1 >= md->end_subject &&
2313            NLBLOCK->nltype == NLTYPE_FIXED &&
2314            NLBLOCK->nllen == 2 &&
2315            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2316        {        {
2317        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);        md->hitend = TRUE;
2318          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2319          }
2320    
2321        /* Fall through */
2322    
2323        /* Match any single character whatsoever. */
2324    
2325        case OP_ALLANY:
2326        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2327          {                            /* not be updated before SCHECK_PARTIAL. */
2328          SCHECK_PARTIAL();
2329          RRETURN(MATCH_NOMATCH);
2330        }        }
2331      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2332      if (utf8)  #ifdef SUPPORT_UTF
2333        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2334    #endif
2335      ecode++;      ecode++;
2336      break;      break;
2337    
# Line 1435  for (;;) Line 2339  for (;;)
2339      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2340    
2341      case OP_ANYBYTE:      case OP_ANYBYTE:
2342      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2343          {                            /* not be updated before SCHECK_PARTIAL. */
2344          SCHECK_PARTIAL();
2345          RRETURN(MATCH_NOMATCH);
2346          }
2347        eptr++;
2348      ecode++;      ecode++;
2349      break;      break;
2350    
2351      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2352      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2353          {
2354          SCHECK_PARTIAL();
2355          RRETURN(MATCH_NOMATCH);
2356          }
2357      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2358      if (      if (
2359  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2360         c < 256 &&         c < 256 &&
2361  #endif  #endif
2362         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1453  for (;;) Line 2366  for (;;)
2366      break;      break;
2367    
2368      case OP_DIGIT:      case OP_DIGIT:
2369      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2370          {
2371          SCHECK_PARTIAL();
2372          RRETURN(MATCH_NOMATCH);
2373          }
2374      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2375      if (      if (
2376  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2377         c >= 256 ||         c > 255 ||
2378  #endif  #endif
2379         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2380         )         )
# Line 1466  for (;;) Line 2383  for (;;)
2383      break;      break;
2384    
2385      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2386      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2387          {
2388          SCHECK_PARTIAL();
2389          RRETURN(MATCH_NOMATCH);
2390          }
2391      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2392      if (      if (
2393  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2394         c < 256 &&         c < 256 &&
2395  #endif  #endif
2396         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1479  for (;;) Line 2400  for (;;)
2400      break;      break;
2401    
2402      case OP_WHITESPACE:      case OP_WHITESPACE:
2403      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2404          {
2405          SCHECK_PARTIAL();
2406          RRETURN(MATCH_NOMATCH);
2407          }
2408      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2409      if (      if (
2410  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2411         c >= 256 ||         c > 255 ||
2412  #endif  #endif
2413         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2414         )         )
# Line 1492  for (;;) Line 2417  for (;;)
2417      break;      break;
2418    
2419      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2420      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2421          {
2422          SCHECK_PARTIAL();
2423          RRETURN(MATCH_NOMATCH);
2424          }
2425      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2426      if (      if (
2427  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2428         c < 256 &&         c < 256 &&
2429  #endif  #endif
2430         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1505  for (;;) Line 2434  for (;;)
2434      break;      break;
2435    
2436      case OP_WORDCHAR:      case OP_WORDCHAR:
2437      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2438          {
2439          SCHECK_PARTIAL();
2440          RRETURN(MATCH_NOMATCH);
2441          }
2442      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2443      if (      if (
2444  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2445         c >= 256 ||         c > 255 ||
2446  #endif  #endif
2447         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2448         )         )
# Line 1518  for (;;) Line 2451  for (;;)
2451      break;      break;
2452    
2453      case OP_ANYNL:      case OP_ANYNL:
2454      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2455          {
2456          SCHECK_PARTIAL();
2457          RRETURN(MATCH_NOMATCH);
2458          }
2459      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2460      switch(c)      switch(c)
2461        {        {
2462        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2463        case 0x000d:  
2464        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2465          if (eptr >= md->end_subject)
2466            {
2467            SCHECK_PARTIAL();
2468            }
2469          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2470        break;        break;
2471    
2472        case 0x000a:        case CHAR_LF:
2473        break;        break;
2474    
2475        case 0x000b:        case CHAR_VT:
2476        case 0x000c:        case CHAR_FF:
2477        case 0x0085:        case CHAR_NEL:
2478    #ifndef EBCDIC
2479        case 0x2028:        case 0x2028:
2480        case 0x2029:        case 0x2029:
2481    #endif  /* Not EBCDIC */
2482        if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);        if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2483        break;        break;
2484        }        }
# Line 1542  for (;;) Line 2486  for (;;)
2486      break;      break;
2487    
2488      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2489      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2490          {
2491          SCHECK_PARTIAL();
2492          RRETURN(MATCH_NOMATCH);
2493          }
2494      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2495      switch(c)      switch(c)
2496        {        {
2497          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2498        default: break;        default: break;
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       RRETURN(MATCH_NOMATCH);  
2499        }        }
2500      ecode++;      ecode++;
2501      break;      break;
2502    
2503      case OP_HSPACE:      case OP_HSPACE:
2504      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2505          {
2506          SCHECK_PARTIAL();
2507          RRETURN(MATCH_NOMATCH);
2508          }
2509      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2510      switch(c)      switch(c)
2511        {        {
2512          HSPACE_CASES: break;  /* Byte and multibyte cases */
2513        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       break;  
2514        }        }
2515      ecode++;      ecode++;
2516      break;      break;
2517    
2518      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2519      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2520          {
2521          SCHECK_PARTIAL();
2522          RRETURN(MATCH_NOMATCH);
2523          }
2524      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2525      switch(c)      switch(c)
2526        {        {
2527          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2528        default: break;        default: break;
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       RRETURN(MATCH_NOMATCH);  
2529        }        }
2530      ecode++;      ecode++;
2531      break;      break;
2532    
2533      case OP_VSPACE:      case OP_VSPACE:
2534      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2535          {
2536          SCHECK_PARTIAL();
2537          RRETURN(MATCH_NOMATCH);
2538          }
2539      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2540      switch(c)      switch(c)
2541        {        {
2542          VSPACE_CASES: break;
2543        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       break;  
2544        }        }
2545      ecode++;      ecode++;
2546      break;      break;
# Line 1643  for (;;) Line 2551  for (;;)
2551    
2552      case OP_PROP:      case OP_PROP:
2553      case OP_NOTPROP:      case OP_NOTPROP:
2554      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2555          {
2556          SCHECK_PARTIAL();
2557          RRETURN(MATCH_NOMATCH);
2558          }
2559      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2560        {        {
2561        int chartype, script;        const pcre_uint32 *cp;
2562        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2563    
2564        switch(ecode[1])        switch(ecode[1])
2565          {          {
# Line 1656  for (;;) Line 2568  for (;;)
2568          break;          break;
2569    
2570          case PT_LAMP:          case PT_LAMP:
2571          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2572               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2573               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2574            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2575           break;          break;
2576    
2577          case PT_GC:          case PT_GC:
2578          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2579            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2580          break;          break;
2581    
2582          case PT_PC:          case PT_PC:
2583          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2584            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2585          break;          break;
2586    
2587          case PT_SC:          case PT_SC:
2588          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2589            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2590          break;          break;
2591    
2592            /* These are specials */
2593    
2594            case PT_ALNUM:
2595            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2596                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2597              RRETURN(MATCH_NOMATCH);
2598            break;
2599    
2600            case PT_SPACE:    /* Perl space */
2601            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2602                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2603                   == (op == OP_NOTPROP))
2604              RRETURN(MATCH_NOMATCH);
2605            break;
2606    
2607            case PT_PXSPACE:  /* POSIX space */
2608            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2609                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2610                 c == CHAR_FF || c == CHAR_CR)
2611                   == (op == OP_NOTPROP))
2612              RRETURN(MATCH_NOMATCH);
2613            break;
2614    
2615            case PT_WORD:
2616            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2617                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2618                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2619              RRETURN(MATCH_NOMATCH);
2620            break;
2621    
2622            case PT_CLIST:
2623            cp = PRIV(ucd_caseless_sets) + ecode[2];
2624            for (;;)
2625              {
2626              if (c < *cp)
2627                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2628              if (c == *cp++)
2629                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2630              }
2631            break;
2632    
2633            case PT_UCNC:
2634            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2635                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2636                 c >= 0xe000) == (op == OP_NOTPROP))
2637              RRETURN(MATCH_NOMATCH);
2638            break;
2639    
2640            /* This should never occur */
2641    
2642          default:          default:
2643          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2644          }          }
# Line 1689  for (;;) Line 2651  for (;;)
2651      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2652    
2653      case OP_EXTUNI:      case OP_EXTUNI:
2654      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2655      GETCHARINCTEST(c, eptr);        {
2656          SCHECK_PARTIAL();
2657          RRETURN(MATCH_NOMATCH);
2658          }
2659        else
2660        {        {
2661        int chartype, script;        int lgb, rgb;
2662        int category = _pcre_ucp_findprop(c, &chartype, &script);        GETCHARINCTEST(c, eptr);
2663        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        lgb = UCD_GRAPHBREAK(c);
2664        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2665          {          {
2666          int len = 1;          int len = 1;
2667          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2668            {          rgb = UCD_GRAPHBREAK(c);
2669            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2670            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2671          eptr += len;          eptr += len;
2672          }          }
2673        }        }
2674        CHECK_PARTIAL();
2675      ecode++;      ecode++;
2676      break;      break;
2677  #endif  #endif  /* SUPPORT_UCP */
2678    
2679    
2680      /* 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 1721  for (;;) Line 2686  for (;;)
2686      loops). */      loops). */
2687    
2688      case OP_REF:      case OP_REF:
2689        {      case OP_REFI:
2690        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2691        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2692        ecode += 1 + IMM2_SIZE;
2693    
2694        /* If the reference is unset, set the length to be longer than the amount      /* If the reference is unset, there are two possibilities:
       of subject left; this ensures that every attempt at a match fails. We  
       can't just fail here, because of the possibility of quantifiers with zero  
       minima. */  
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2695    
2696        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2697        this ensures that every attempt at a match fails. We can't just fail
2698        here, because of the possibility of quantifiers with zero minima.
2699    
2700        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2701          {      so that the back reference matches an empty string.
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
2702    
2703          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2704          case OP_CRMINRANGE:      referenced subpattern. */
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2705    
2706          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2707          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2708          eptr += length;      else
2709          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2710    
2711        /* Set up for repetition, or handle the non-repeated case */
2712    
2713        switch (*ecode)
2714          {
2715          case OP_CRSTAR:
2716          case OP_CRMINSTAR:
2717          case OP_CRPLUS:
2718          case OP_CRMINPLUS:
2719          case OP_CRQUERY:
2720          case OP_CRMINQUERY:
2721          c = *ecode++ - OP_CRSTAR;
2722          minimize = (c & 1) != 0;
2723          min = rep_min[c];                 /* Pick up values from tables; */
2724          max = rep_max[c];                 /* zero for max => infinity */
2725          if (max == 0) max = INT_MAX;
2726          break;
2727    
2728          case OP_CRRANGE:
2729          case OP_CRMINRANGE:
2730          minimize = (*ecode == OP_CRMINRANGE);
2731          min = GET2(ecode, 1);
2732          max = GET2(ecode, 1 + IMM2_SIZE);
2733          if (max == 0) max = INT_MAX;
2734          ecode += 1 + 2 * IMM2_SIZE;
2735          break;
2736    
2737          default:               /* No repeat follows */
2738          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2739            {
2740            if (length == -2) eptr = md->end_subject;   /* Partial match */
2741            CHECK_PARTIAL();
2742            RRETURN(MATCH_NOMATCH);
2743          }          }
2744          eptr += length;
2745          continue;              /* With the main loop */
2746          }
2747    
2748        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2749        main loop. */      zero, just continue with the main loop. If the length is negative, it
2750        means the reference is unset in non-Java-compatible mode. If the minimum is
2751        zero, we can continue at the same level without recursion. For any other
2752        minimum, carrying on will result in NOMATCH. */
2753    
2754        if (length == 0) continue;      if (length == 0) continue;
2755        if (length < 0 && min == 0) continue;
2756    
2757        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2758        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2759        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2760    
2761        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2762          {
2763          int slength;
2764          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2765          {          {
2766          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2767          eptr += length;          CHECK_PARTIAL();
2768            RRETURN(MATCH_NOMATCH);
2769          }          }
2770          eptr += slength;
2771          }
2772    
2773        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2774        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2775    
2776        if (min == max) continue;      if (min == max) continue;
2777    
2778        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2779    
2780        if (minimize)      if (minimize)
2781          {
2782          for (fi = min;; fi++)
2783          {          {
2784          for (fi = min;; fi++)          int slength;
2785            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2786            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2787            if (fi >= max) RRETURN(MATCH_NOMATCH);
2788            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2789            {            {
2790            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2791            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2792            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2793            }            }
2794          /* Control never gets here */          eptr += slength;
2795          }          }
2796          /* Control never gets here */
2797          }
2798    
2799        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2800    
2801        else      else
2802          {
2803          pp = eptr;
2804          for (i = min; i < max; i++)
2805          {          {
2806          pp = eptr;          int slength;
2807          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)  
2808            {            {
2809            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2810            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2811            eptr -= length;  
2812              if (slength == -2 && md->partial != 0 &&
2813                  md->end_subject > md->start_used_ptr)
2814                {
2815                md->hitend = TRUE;
2816                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2817                }
2818              break;
2819            }            }
2820          RRETURN(MATCH_NOMATCH);          eptr += slength;
2821            }
2822    
2823          while (eptr >= pp)
2824            {
2825            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2826            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2827            eptr -= length;
2828          }          }
2829          RRETURN(MATCH_NOMATCH);
2830        }        }
2831      /* Control never gets here */      /* Control never gets here */
2832    
   
   
2833      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2834      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,
2835      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 1838  for (;;) Line 2844  for (;;)
2844      case OP_NCLASS:      case OP_NCLASS:
2845      case OP_CLASS:      case OP_CLASS:
2846        {        {
2847          /* The data variable is saved across frames, so the byte map needs to
2848          be stored there. */
2849    #define BYTE_MAP ((pcre_uint8 *)data)
2850        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2851        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2852    
2853        switch (*ecode)        switch (*ecode)
2854          {          {
# Line 1860  for (;;) Line 2869  for (;;)
2869          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2870          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2871          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2872          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2873          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2874          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2875          break;          break;
2876    
2877          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1872  for (;;) Line 2881  for (;;)
2881    
2882        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2883    
2884  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2885        /* UTF-8 mode */        if (utf)
       if (utf8)  
2886          {          {
2887          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2888            {            {
2889            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2890                {
2891                SCHECK_PARTIAL();
2892                RRETURN(MATCH_NOMATCH);
2893                }
2894            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2895            if (c > 255)            if (c > 255)
2896              {              {
2897              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2898              }              }
2899            else            else
2900              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2901            }            }
2902          }          }
2903        else        else
2904  #endif  #endif
2905        /* Not UTF-8 mode */        /* Not UTF mode */
2906          {          {
2907          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2908            {            {
2909            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2910                {
2911                SCHECK_PARTIAL();
2912                RRETURN(MATCH_NOMATCH);
2913                }
2914            c = *eptr++;            c = *eptr++;
2915            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2916              if (c > 255)
2917                {
2918                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2919                }
2920              else
2921    #endif
2922                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2923            }            }
2924          }          }
2925    
# Line 1912  for (;;) Line 2933  for (;;)
2933    
2934        if (minimize)        if (minimize)
2935          {          {
2936  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2937          /* UTF-8 mode */          if (utf)
         if (utf8)  
2938            {            {
2939            for (fi = min;; fi++)            for (fi = min;; fi++)
2940              {              {
2941              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2942              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2943              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2944                if (eptr >= md->end_subject)
2945                  {
2946                  SCHECK_PARTIAL();
2947                  RRETURN(MATCH_NOMATCH);
2948                  }
2949              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2950              if (c > 255)              if (c > 255)
2951                {                {
2952                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2953                }                }
2954              else              else
2955                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2956              }              }
2957            }            }
2958          else          else
2959  #endif  #endif
2960          /* Not UTF-8 mode */          /* Not UTF mode */
2961            {            {
2962            for (fi = min;; fi++)            for (fi = min;; fi++)
2963              {              {
2964              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2965              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2966              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2967                if (eptr >= md->end_subject)
2968                  {
2969                  SCHECK_PARTIAL();
2970                  RRETURN(MATCH_NOMATCH);
2971                  }
2972              c = *eptr++;              c = *eptr++;
2973              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2974                if (c > 255)
2975                  {
2976                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2977                  }
2978                else
2979    #endif
2980                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2981              }              }
2982            }            }
2983          /* Control never gets here */          /* Control never gets here */
# Line 1954  for (;;) Line 2989  for (;;)
2989          {          {
2990          pp = eptr;          pp = eptr;
2991    
2992  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2993          /* UTF-8 mode */          if (utf)
         if (utf8)  
2994            {            {
2995            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2996              {              {
2997              int len = 1;              int len = 1;
2998              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2999                  {
3000                  SCHECK_PARTIAL();
3001                  break;
3002                  }
3003              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
3004              if (c > 255)              if (c > 255)
3005                {                {
3006                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3007                }                }
3008              else              else
3009                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3010              eptr += len;              eptr += len;
3011              }              }
3012            for (;;)            for (;;)
3013              {              {
3014              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3015              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3016              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3017              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1983  for (;;) Line 3019  for (;;)
3019            }            }
3020          else          else
3021  #endif  #endif
3022            /* Not UTF-8 mode */            /* Not UTF mode */
3023            {            {
3024            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3025              {              {
3026              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3027                  {
3028                  SCHECK_PARTIAL();
3029                  break;
3030                  }
3031              c = *eptr;              c = *eptr;
3032              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3033                if (c > 255)
3034                  {
3035                  if (op == OP_CLASS) break;
3036                  }
3037                else
3038    #endif
3039                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3040              eptr++;              eptr++;
3041              }              }
3042            while (eptr >= pp)            while (eptr >= pp)
3043              {              {
3044              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3045              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3046              eptr--;              eptr--;
3047              }              }
# Line 2002  for (;;) Line 3049  for (;;)
3049    
3050          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3051          }          }
3052    #undef BYTE_MAP
3053        }        }
3054      /* Control never gets here */      /* Control never gets here */
3055    
3056    
3057      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3058      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
3059        mode, because Unicode properties are supported in non-UTF-8 mode. */
3060    
3061  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3062      case OP_XCLASS:      case OP_XCLASS:
3063        {        {
3064        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2034  for (;;) Line 3083  for (;;)
3083          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3084          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3085          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3086          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3087          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3088          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3089          break;          break;
3090    
3091          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2048  for (;;) Line 3097  for (;;)
3097    
3098        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3099          {          {
3100          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3101          GETCHARINC(c, eptr);            {
3102          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3103              RRETURN(MATCH_NOMATCH);
3104              }
3105            GETCHARINCTEST(c, eptr);
3106            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3107          }          }
3108    
3109        /* 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 2065  for (;;) Line 3118  for (;;)
3118          {          {
3119          for (fi = min;; fi++)          for (fi = min;; fi++)
3120            {            {
3121            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3122            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3123            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3124            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3125            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3126                SCHECK_PARTIAL();
3127                RRETURN(MATCH_NOMATCH);
3128                }
3129              GETCHARINCTEST(c, eptr);
3130              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3131            }            }
3132          /* Control never gets here */          /* Control never gets here */
3133          }          }
# Line 2082  for (;;) Line 3140  for (;;)
3140          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3141            {            {
3142            int len = 1;            int len = 1;
3143            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3144            GETCHARLEN(c, eptr, len);              {
3145            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3146                break;
3147                }
3148    #ifdef SUPPORT_UTF
3149              GETCHARLENTEST(c, eptr, len);
3150    #else
3151              c = *eptr;
3152    #endif
3153              if (!PRIV(xclass)(c, data, utf)) break;
3154            eptr += len;            eptr += len;
3155            }            }
3156          for(;;)          for(;;)
3157            {            {
3158            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3159            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3160            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3161            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3162              if (utf) BACKCHAR(eptr);
3163    #endif
3164            }            }
3165          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3166          }          }
# Line 2104  for (;;) Line 3172  for (;;)
3172      /* Match a single character, casefully */      /* Match a single character, casefully */
3173    
3174      case OP_CHAR:      case OP_CHAR:
3175  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3176      if (utf8)      if (utf)
3177        {        {
3178        length = 1;        length = 1;
3179        ecode++;        ecode++;
3180        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3181        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3182        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3183            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3184            RRETURN(MATCH_NOMATCH);
3185            }
3186          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3187        }        }
3188      else      else
3189  #endif  #endif
3190        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3191        {        {
3192        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3193            {
3194            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3195            RRETURN(MATCH_NOMATCH);
3196            }
3197        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3198        ecode += 2;        ecode += 2;
3199        }        }
3200      break;      break;
3201    
3202      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3203        subject, give up immediately. */
3204    
3205        case OP_CHARI:
3206        if (eptr >= md->end_subject)
3207          {
3208          SCHECK_PARTIAL();
3209          RRETURN(MATCH_NOMATCH);
3210          }
3211    
3212      case OP_CHARNC:  #ifdef SUPPORT_UTF
3213  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3214        {        {
3215        length = 1;        length = 1;
3216        ecode++;        ecode++;
3217        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3218    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3219        /* 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
3220        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3221          fast lookup table. We know that there is at least one byte left in the
3222          subject. */
3223    
3224        if (fc < 128)        if (fc < 128)
3225          {          {
3226          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3227            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3228            ecode++;
3229            eptr++;
3230          }          }
3231    
3232        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3233          use the value of "length" to check for sufficient bytes left, because the
3234          other case of the character may have more or fewer bytes.  */
3235    
3236        else        else
3237          {          {
3238          unsigned int dc;          pcre_uint32 dc;
3239          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3240          ecode += length;          ecode += length;
3241    
# Line 2158  for (;;) Line 3245  for (;;)
3245          if (fc != dc)          if (fc != dc)
3246            {            {
3247  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3248            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3249  #endif  #endif
3250              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3251            }            }
3252          }          }
3253        }        }
3254      else      else
3255  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3256    
3257      /* Non-UTF-8 mode */      /* Not UTF mode */
3258        {        {
3259        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3260        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3261          eptr++;
3262        ecode += 2;        ecode += 2;
3263        }        }
3264      break;      break;
# Line 2178  for (;;) Line 3266  for (;;)
3266      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3267    
3268      case OP_EXACT:      case OP_EXACT:
3269        case OP_EXACTI:
3270      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3271      ecode += 3;      ecode += 1 + IMM2_SIZE;
3272      goto REPEATCHAR;      goto REPEATCHAR;
3273    
3274      case OP_POSUPTO:      case OP_POSUPTO:
3275        case OP_POSUPTOI:
3276      possessive = TRUE;      possessive = TRUE;
3277      /* Fall through */      /* Fall through */
3278    
3279      case OP_UPTO:      case OP_UPTO:
3280        case OP_UPTOI:
3281      case OP_MINUPTO:      case OP_MINUPTO:
3282        case OP_MINUPTOI:
3283      min = 0;      min = 0;
3284      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3285      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3286      ecode += 3;      ecode += 1 + IMM2_SIZE;
3287      goto REPEATCHAR;      goto REPEATCHAR;
3288    
3289      case OP_POSSTAR:      case OP_POSSTAR:
3290        case OP_POSSTARI:
3291      possessive = TRUE;      possessive = TRUE;
3292      min = 0;      min = 0;
3293      max = INT_MAX;      max = INT_MAX;
# Line 2202  for (;;) Line 3295  for (;;)
3295      goto REPEATCHAR;      goto REPEATCHAR;
3296    
3297      case OP_POSPLUS:      case OP_POSPLUS:
3298        case OP_POSPLUSI:
3299      possessive = TRUE;      possessive = TRUE;
3300      min = 1;      min = 1;
3301      max = INT_MAX;      max = INT_MAX;
# Line 2209  for (;;) Line 3303  for (;;)
3303      goto REPEATCHAR;      goto REPEATCHAR;
3304    
3305      case OP_POSQUERY:      case OP_POSQUERY:
3306        case OP_POSQUERYI:
3307      possessive = TRUE;      possessive = TRUE;
3308      min = 0;      min = 0;
3309      max = 1;      max = 1;
# Line 2216  for (;;) Line 3311  for (;;)
3311      goto REPEATCHAR;      goto REPEATCHAR;
3312    
3313      case OP_STAR:      case OP_STAR:
3314        case OP_STARI:
3315      case OP_MINSTAR:      case OP_MINSTAR:
3316        case OP_MINSTARI:
3317      case OP_PLUS:      case OP_PLUS:
3318        case OP_PLUSI:
3319      case OP_MINPLUS:      case OP_MINPLUS:
3320        case OP_MINPLUSI:
3321      case OP_QUERY:      case OP_QUERY:
3322        case OP_QUERYI:
3323      case OP_MINQUERY:      case OP_MINQUERY:
3324      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3325        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3326      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3327      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3328      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3329      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3330    
3331      /* 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. */  
3332    
3333      REPEATCHAR:      REPEATCHAR:
3334  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3335      if (utf8)      if (utf)
3336        {        {
3337        length = 1;        length = 1;
3338        charptr = ecode;        charptr = ecode;
3339        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3340        ecode += length;        ecode += length;
3341    
3342        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2247  for (;;) Line 3345  for (;;)
3345        if (length > 1)        if (length > 1)
3346          {          {
3347  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3348          unsigned int othercase;          pcre_uint32 othercase;
3349          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3350              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3351            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3352          else oclength = 0;          else oclength = 0;
3353  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3354    
3355          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3356            {            {
3357            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3358                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3359  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3360            /* Need braces because of following else */            else if (oclength > 0 &&
3361            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3362                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3363    #endif  /* SUPPORT_UCP */
3364            else            else
3365              {              {
3366              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3367              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3368              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3369            }            }
3370    
3371          if (min == max) continue;          if (min == max) continue;
# Line 2276  for (;;) Line 3374  for (;;)
3374            {            {
3375            for (fi = min;; fi++)            for (fi = min;; fi++)
3376              {              {
3377              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3378              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3379              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3380              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3381                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3382  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3383              /* Need braces because of following else */              else if (oclength > 0 &&
3384              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3385                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3386    #endif  /* SUPPORT_UCP */
3387              else              else
3388                {                {
3389                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3390                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3391                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3392              }              }
3393            /* Control never gets here */            /* Control never gets here */
3394            }            }
# Line 2300  for (;;) Line 3398  for (;;)
3398            pp = eptr;            pp = eptr;
3399            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3400              {              {
3401              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3402              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3403  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3404              else if (oclength == 0) break;              else if (oclength > 0 &&
3405                         eptr <= md->end_subject - oclength &&
3406                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3407    #endif  /* SUPPORT_UCP */
3408              else              else
3409                {                {
3410                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3411                eptr += oclength;                break;
3412                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3413              }              }
3414    
3415            if (possessive) continue;            if (possessive) continue;
3416    
3417            for(;;)            for(;;)
3418             {              {
3419             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3420             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3421             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3422  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3423             eptr--;              eptr--;
3424             BACKCHAR(eptr);              BACKCHAR(eptr);
3425  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3426             eptr -= length;              eptr -= length;
3427  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3428             }              }
3429            }            }
3430          /* Control never gets here */          /* Control never gets here */
3431          }          }
# Line 2336  for (;;) Line 3435  for (;;)
3435        value of fc will always be < 128. */        value of fc will always be < 128. */
3436        }        }
3437      else      else
3438  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3439          /* 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);  
3440        fc = *ecode++;        fc = *ecode++;
       }  
3441    
3442      /* 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
3443      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
3444      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3445      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3446      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2354  for (;;) Line 3449  for (;;)
3449      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3450    
3451      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3452        max, eptr));        max, (char *)eptr));
3453    
3454      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3455        {        {
3456        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3457          /* fc must be < 128 if UTF is enabled. */
3458          foc = md->fcc[fc];
3459    #else
3460    #ifdef SUPPORT_UTF
3461    #ifdef SUPPORT_UCP
3462          if (utf && fc > 127)
3463            foc = UCD_OTHERCASE(fc);
3464    #else
3465          if (utf && fc > 127)
3466            foc = fc;
3467    #endif /* SUPPORT_UCP */
3468          else
3469    #endif /* SUPPORT_UTF */
3470            foc = TABLE_GET(fc, md->fcc, fc);
3471    #endif /* COMPILE_PCRE8 */
3472    
3473        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3474          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3475            pcre_uint32 cc;                 /* Faster than pcre_uchar */
3476            if (eptr >= md->end_subject)
3477              {
3478              SCHECK_PARTIAL();
3479              RRETURN(MATCH_NOMATCH);
3480              }
3481            cc = RAWUCHARTEST(eptr);
3482            if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3483            eptr++;
3484            }
3485        if (min == max) continue;        if (min == max) continue;
3486        if (minimize)        if (minimize)
3487          {          {
3488          for (fi = min;; fi++)          for (fi = min;; fi++)
3489            {            {
3490            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            pcre_uint32 cc;               /* Faster than pcre_uchar */
3491              RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3492            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3493            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3494                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3495                {
3496                SCHECK_PARTIAL();
3497              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3498                }
3499              cc = RAWUCHARTEST(eptr);
3500              if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3501              eptr++;
3502            }            }
3503          /* Control never gets here */          /* Control never gets here */
3504          }          }
# Line 2379  for (;;) Line 3507  for (;;)
3507          pp = eptr;          pp = eptr;
3508          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3509            {            {
3510            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            pcre_uint32 cc;               /* Faster than pcre_uchar */
3511              if (eptr >= md->end_subject)
3512                {
3513                SCHECK_PARTIAL();
3514                break;
3515                }
3516              cc = RAWUCHARTEST(eptr);
3517              if (fc != cc && foc != cc) break;
3518            eptr++;            eptr++;
3519            }            }
3520    
3521          if (possessive) continue;          if (possessive) continue;
3522    
3523          while (eptr >= pp)          while (eptr >= pp)
3524            {            {
3525            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3526            eptr--;            eptr--;
3527            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3528            }            }
# Line 2398  for (;;) Line 3535  for (;;)
3535    
3536      else      else
3537        {        {
3538        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3539            {
3540            if (eptr >= md->end_subject)
3541              {
3542              SCHECK_PARTIAL();
3543              RRETURN(MATCH_NOMATCH);
3544              }
3545            if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3546            }
3547    
3548        if (min == max) continue;        if (min == max) continue;
3549    
3550        if (minimize)        if (minimize)
3551          {          {
3552          for (fi = min;; fi++)          for (fi = min;; fi++)
3553            {            {
3554            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3555            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3556            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3557              if (eptr >= md->end_subject)
3558                {
3559                SCHECK_PARTIAL();
3560              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3561                }
3562              if (fc != RAWUCHARINCTEST(eptr)) RRETURN(MATCH_NOMATCH);
3563            }            }
3564          /* Control never gets here */          /* Control never gets here */
3565          }          }
# Line 2416  for (;;) Line 3568  for (;;)
3568          pp = eptr;          pp = eptr;
3569          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3570            {            {
3571            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3572                {
3573                SCHECK_PARTIAL();
3574                break;
3575                }
3576              if (fc != RAWUCHARTEST(eptr)) break;
3577            eptr++;            eptr++;
3578            }            }
3579          if (possessive) continue;          if (possessive) continue;
3580    
3581          while (eptr >= pp)          while (eptr >= pp)
3582            {            {