/[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 443 by ph10, Sun Sep 13 16:00:08 2009 UTC revision 1397 by ph10, Mon Nov 11 18:33:23 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) 1997-2009 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_KETRPOS      (-998)
91  #define MATCH_SKIP         (-997)  #define MATCH_ONCE         (-997)
92  #define MATCH_THEN         (-996)  /* The next 5 must be kept together and in sequence so that a test that checks
93    for any one of them can use a range. */
94    #define MATCH_COMMIT       (-996)
95    #define MATCH_PRUNE        (-995)
96    #define MATCH_SKIP         (-994)
97    #define MATCH_SKIP_ARG     (-993)
98    #define MATCH_THEN         (-992)
99    #define MATCH_BACKTRACK_MAX MATCH_THEN
100    #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101    
102  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
103  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
# Line 84  because the offset vector is always a mu Line 107  because the offset vector is always a mu
107    
108  /* Min and max values for the common repeats; for the maxima, 0 => infinity */  /* Min and max values for the common repeats; for the maxima, 0 => infinity */
109    
110  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, };
111  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, };
   
112    
113    #ifdef PCRE_DEBUG
 #ifdef DEBUG  
114  /*************************************************  /*************************************************
115  *        Debugging function to print chars       *  *        Debugging function to print chars       *
116  *************************************************/  *************************************************/
# Line 107  Returns:     nothing Line 128  Returns:     nothing
128  */  */
129    
130  static void  static void
131  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132  {  {
133  unsigned int c;  pcre_uint32 c;
134    BOOL utf = md->utf;
135  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136  while (length-- > 0)  while (length-- > 0)
137    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138  }  }
139  #endif  #endif
140    
# Line 122  while (length-- > 0) Line 144  while (length-- > 0)
144  *          Match a back-reference                *  *          Match a back-reference                *
145  *************************************************/  *************************************************/
146    
147  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
148  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
149    the length passed is zero. Note that in caseless UTF-8 mode, the number of
150    subject bytes matched may be different to the number of reference bytes.
151    
152  Arguments:  Arguments:
153    offset      index into the offset vector    offset      index into the offset vector
154    eptr        points into the subject    eptr        pointer into the subject
155    length      length to be matched    length      length of reference to be matched (number of bytes)
156    md          points to match data block    md          points to match data block
157    ims         the ims flags    caseless    TRUE if caseless
158    
159  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
160                  -1 no match
161                  -2 partial match; always given if at end subject
162  */  */
163    
164  static BOOL  static int
165  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166    unsigned long int ims)    BOOL caseless)
167  {  {
168  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
169    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170    #ifdef SUPPORT_UTF
171    BOOL utf = md->utf;
172    #endif
173    
174  #ifdef DEBUG  #ifdef PCRE_DEBUG
175  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
176    printf("matching subject <null>");    printf("matching subject <null>");
177  else  else
# Line 154  pchars(p, length, FALSE, md); Line 184  pchars(p, length, FALSE, md);
184  printf("\n");  printf("\n");
185  #endif  #endif
186    
187  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
188    case the length is passed as zero). */
189    
190  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
191    
192  /* Separate the caseless case for speed. In UTF-8 mode we can only do this  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193  properly if Unicode properties are supported. Otherwise, we can check only  properly if Unicode properties are supported. Otherwise, we can check only
194  ASCII characters. */  ASCII characters. */
195    
196  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
197    {    {
198  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
199  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
200    if (md->utf8)    if (utf)
201      {      {
202      USPTR endptr = eptr + length;      /* Match characters up to the end of the reference. NOTE: the number of
203      while (eptr < endptr)      data units matched may differ, because in UTF-8 there are some characters
204        {      whose upper and lower case versions code have different numbers of bytes.
205        int c, d;      For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207        sequence of two of the latter. It is important, therefore, to check the
208        length along the reference, not along the subject (earlier code did this
209        wrong). */
210    
211        PCRE_PUCHAR endptr = p + length;
212        while (p < endptr)
213          {
214          pcre_uint32 c, d;
215          const ucd_record *ur;
216          if (eptr >= md->end_subject) return -2;   /* Partial match */
217        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
218        GETCHARINC(d, p);        GETCHARINC(d, p);
219        if (c != d && c != UCD_OTHERCASE(d)) return FALSE;        ur = GET_UCD(d);
220          if (c != d && c != d + ur->other_case)
221            {
222            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223            for (;;)
224              {
225              if (c < *pp) return -1;
226              if (c == *pp++) break;
227              }
228            }
229        }        }
230      }      }
231    else    else
# Line 183  if ((ims & PCRE_CASELESS) != 0) Line 234  if ((ims & PCRE_CASELESS) != 0)
234    
235    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there    /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236    is no UCP support. */    is no UCP support. */
237        {
238    while (length-- > 0)      while (length-- > 0)
239      { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }        {
240          pcre_uint32 cc, cp;
241          if (eptr >= md->end_subject) return -2;   /* Partial match */
242          cc = RAWUCHARTEST(eptr);
243          cp = RAWUCHARTEST(p);
244          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245          p++;
246          eptr++;
247          }
248        }
249    }    }
250    
251  /* In the caseful case, we can just compare the bytes, whether or not we  /* In the caseful case, we can just compare the bytes, whether or not we
252  are in UTF-8 mode. */  are in UTF-8 mode. */
253    
254  else  else
255    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
256      while (length-- > 0)
257        {
258        if (eptr >= md->end_subject) return -2;   /* Partial match */
259        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260        }
261      }
262    
263  return TRUE;  return (int)(eptr - eptr_start);
264  }  }
265    
266    
# Line 245  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 311  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
311         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315           RM61,  RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316    
317  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
318  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
319  actuall used in this definition. */  actually used in this definition. */
320    
321  #ifndef NO_RECURSE  #ifndef NO_RECURSE
322  #define REGISTER register  #define REGISTER register
323    
324  #ifdef DEBUG  #ifdef PCRE_DEBUG
325  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
326    { \    { \
327    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
328    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
330    }    }
331  #define RRETURN(ra) \  #define RRETURN(ra) \
332    { \    { \
333    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
334    return ra; \    return ra; \
335    }    }
336  #else  #else
337  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
338    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
340  #endif  #endif
341    
# Line 281  argument of match(), which never changes Line 348  argument of match(), which never changes
348    
349  #define REGISTER  #define REGISTER
350    
351  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
352    {\    {\
353    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
354    frame->Xwhere = rw; \    if (newframe == NULL)\
355        {\
356        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358        newframe->Xnextframe = NULL;\
359        frame->Xnextframe = newframe;\
360        }\
361      frame->Xwhere = rw;\
362    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
363    newframe->Xecode = rb;\    newframe->Xecode = rb;\
364    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
365    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
366    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
367    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
368    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
369    frame = newframe;\    frame = newframe;\
# Line 303  argument of match(), which never changes Line 375  argument of match(), which never changes
375    
376  #define RRETURN(ra)\  #define RRETURN(ra)\
377    {\    {\
378    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
379    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
380    if (frame != NULL)\    if (frame != NULL)\
381      {\      {\
382      rrc = ra;\      rrc = ra;\
# Line 319  argument of match(), which never changes Line 390  argument of match(), which never changes
390    
391  typedef struct heapframe {  typedef struct heapframe {
392    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
393      struct heapframe *Xnextframe;
394    
395    /* Function arguments that may change */    /* Function arguments that may change */
396    
397    USPTR Xeptr;    PCRE_PUCHAR Xeptr;
398    const uschar *Xecode;    const pcre_uchar *Xecode;
399    USPTR Xmstart;    PCRE_PUCHAR Xmstart;
400    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
401    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
402    unsigned int Xrdepth;    unsigned int Xrdepth;
403    
404    /* Function local variables */    /* Function local variables */
405    
406    USPTR Xcallpat;    PCRE_PUCHAR Xcallpat;
407  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
408    USPTR Xcharptr;    PCRE_PUCHAR Xcharptr;
409  #endif  #endif
410    USPTR Xdata;    PCRE_PUCHAR Xdata;
411    USPTR Xnext;    PCRE_PUCHAR Xnext;
412    USPTR Xpp;    PCRE_PUCHAR Xpp;
413    USPTR Xprev;    PCRE_PUCHAR Xprev;
414    USPTR Xsaved_eptr;    PCRE_PUCHAR Xsaved_eptr;
415    
416    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
417    
# Line 349  typedef struct heapframe { Line 419  typedef struct heapframe {
419    BOOL Xcondition;    BOOL Xcondition;
420    BOOL Xprev_is_word;    BOOL Xprev_is_word;
421    
   unsigned long int Xoriginal_ims;  
   
422  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
423    int Xprop_type;    int Xprop_type;
424    int Xprop_value;    unsigned int Xprop_value;
425    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
426    int Xoclength;    int Xoclength;
427    uschar Xocchars[8];    pcre_uchar Xocchars[6];
428  #endif  #endif
429    
430    int Xcodelink;    int Xcodelink;
# Line 369  typedef struct heapframe { Line 434  typedef struct heapframe {
434    int Xlength;    int Xlength;
435    int Xmax;    int Xmax;
436    int Xmin;    int Xmin;
437    int Xnumber;    unsigned int Xnumber;
438    int Xoffset;    int Xoffset;
439    int Xop;    unsigned int Xop;
440    int Xsave_capture_last;    pcre_int32 Xsave_capture_last;
441    int Xsave_offset1, Xsave_offset2, Xsave_offset3;    int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442    int Xstacksave[REC_STACK_SAVE_MAX];    int Xstacksave[REC_STACK_SAVE_MAX];
443    
# Line 401  returns a negative (error) response, the Line 466  returns a negative (error) response, the
466  same response. */  same response. */
467    
468  /* These macros pack up tests that are used for partial matching, and which  /* These macros pack up tests that are used for partial matching, and which
469  appears several times in the code. We set the "hit end" flag if the pointer is  appear several times in the code. We set the "hit end" flag if the pointer is
470  at the end of the subject and also past the start of the subject (i.e.  at the end of the subject and also past the start of the subject (i.e.
471  something has been matched). For hard partial matching, we then return  something has been matched). For hard partial matching, we then return
472  immediately. The second one is used when we already know we are past the end of  immediately. The second one is used when we already know we are past the end of
473  the subject. */  the subject. */
474    
475  #define CHECK_PARTIAL()\  #define CHECK_PARTIAL()\
476    if (md->partial != 0 && eptr >= md->end_subject && eptr > mstart)\    if (md->partial != 0 && eptr >= md->end_subject && \
477      {\        eptr > md->start_used_ptr) \
478      md->hitend = TRUE;\      { \
479      if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);\      md->hitend = TRUE; \
480        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481      }      }
482    
483  #define SCHECK_PARTIAL()\  #define SCHECK_PARTIAL()\
484    if (md->partial && eptr > mstart)\    if (md->partial != 0 && eptr > md->start_used_ptr) \
485      {\      { \
486      md->hitend = TRUE;\      md->hitend = TRUE; \
487      if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);\      if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488      }      }
489    
490    
491  /* Performance note: It might be tempting to extract commonly used fields from  /* Performance note: It might be tempting to extract commonly used fields from
492  the md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf, end_subject) into individual variables to improve
493  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
494  made performance worse.  made performance worse.
495    
# Line 434  Arguments: Line 500  Arguments:
500                   by encountering \K)                   by encountering \K)
501     offset_top  current top pointer     offset_top  current top pointer
502     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
503     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
504                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
505     rdepth      the recursion depth     rdepth      the recursion depth
506    
507  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
508                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
509                   a negative MATCH_xxx value for PRUNE, SKIP, etc
510                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
511                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
512  */  */
513    
514  static int  static int
515  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517    int flags, unsigned int rdepth)    unsigned int rdepth)
518  {  {
519  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
520  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 460  so they can be ordinary variables in all Line 522  so they can be ordinary variables in all
522    
523  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
524  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
525  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
526  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
527    
528  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
529    BOOL caseless;
530  int condcode;  int condcode;
531    
532  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
533  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
534  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
535  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536    the top-level on the stack rather than malloc-ing them all gives a performance
537    boost in many cases where there is not much "recursion". */
538    
539  #ifdef NO_RECURSE  #ifdef NO_RECURSE
540  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
541    
542  /* Copy in the original argument variables */  /* Copy in the original argument variables */
543    
# Line 481  frame->Xeptr = eptr; Line 545  frame->Xeptr = eptr;
545  frame->Xecode = ecode;  frame->Xecode = ecode;
546  frame->Xmstart = mstart;  frame->Xmstart = mstart;
547  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
548  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
549  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
550    
551  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 496  HEAP_RECURSE: Line 558  HEAP_RECURSE:
558  #define ecode              frame->Xecode  #define ecode              frame->Xecode
559  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
560  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
561  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
562  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
563    
564  /* Ditto for the local variables */  /* Ditto for the local variables */
565    
566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
567  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
568  #endif  #endif
569  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
# Line 520  HEAP_RECURSE: Line 580  HEAP_RECURSE:
580  #define condition          frame->Xcondition  #define condition          frame->Xcondition
581  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
582    
 #define original_ims       frame->Xoriginal_ims  
   
583  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
584  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
585  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
586  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
587  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
588  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
589  #endif  #endif
# Line 558  i, and fc and c, can be the same variabl Line 613  i, and fc and c, can be the same variabl
613  #define fi i  #define fi i
614  #define fc c  #define fc c
615    
616    /* Many of the following variables are used only in small blocks of the code.
617    My normal style of coding would have declared them within each of those blocks.
618    However, in order to accommodate the version of this code that uses an external
619    "stack" implemented on the heap, it is easier to declare them all here, so the
620    declarations can be cut out in a block. The only declarations within blocks
621    below are for variables that do not have to be preserved over a recursive call
622    to RMATCH(). */
623    
624    #ifdef SUPPORT_UTF
625    const pcre_uchar *charptr;
626    #endif
627    const pcre_uchar *callpat;
628    const pcre_uchar *data;
629    const pcre_uchar *next;
630    PCRE_PUCHAR       pp;
631    const pcre_uchar *prev;
632    PCRE_PUCHAR       saved_eptr;
633    
634    recursion_info new_recursive;
635    
636  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  BOOL cur_is_word;
 const uschar *charptr;             /* in small blocks of the code. My normal */  
 #endif                             /* style of coding would have declared    */  
 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().          */  
637  BOOL condition;  BOOL condition;
638  BOOL prev_is_word;  BOOL prev_is_word;
639    
 unsigned long int original_ims;  
   
640  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
641  int prop_type;  int prop_type;
642  int prop_value;  unsigned int prop_value;
643  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
644  int oclength;  int oclength;
645  uschar occhars[8];  pcre_uchar occhars[6];
646  #endif  #endif
647    
648  int codelink;  int codelink;
# Line 593  int ctype; Line 650  int ctype;
650  int length;  int length;
651  int max;  int max;
652  int min;  int min;
653  int number;  unsigned int number;
654  int offset;  int offset;
655  int op;  unsigned int op;
656  int save_capture_last;  pcre_int32 save_capture_last;
657  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
658  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
659    
660  eptrblock newptrb;  eptrblock newptrb;
661    
662    /* There is a special fudge for calling match() in a way that causes it to
663    measure the size of its basic stack frame when the stack is being used for
664    recursion. The second argument (ecode) being NULL triggers this behaviour. It
665    cannot normally ever be NULL. The return is the negated value of the frame
666    size. */
667    
668    if (ecode == NULL)
669      {
670      if (rdepth == 0)
671        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672      else
673        {
674        int len = (char *)&rdepth - (char *)eptr;
675        return (len > 0)? -len : len;
676        }
677      }
678  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
679    
680    /* To save space on the stack and in the heap frame, I have doubled up on some
681    of the local variables that are used only in localised parts of the code, but
682    still need to be preserved over recursive calls of match(). These macros define
683    the alternative names that are used. */
684    
685    #define allow_zero    cur_is_word
686    #define cbegroup      condition
687    #define code_offset   codelink
688    #define condassert    condition
689    #define matched_once  prev_is_word
690    #define foc           number
691    #define save_mark     data
692    
693  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
694  variables. */  variables. */
695    
# Line 622  TAIL_RECURSE: Line 709  TAIL_RECURSE:
709  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
710  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
711  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
714  complicated macro. It has to be used in one particular way. This shouldn't,  complicated macro. It has to be used in one particular way. This shouldn't,
715  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
716    
717  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
718  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
719  #else  #else
720  utf8 = FALSE;  utf = FALSE;
721  #endif  #endif
722    
723  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
# Line 639  haven't exceeded the recursive call limi Line 726  haven't exceeded the recursive call limi
726  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728    
 original_ims = ims;    /* Save for resetting on ')' */  
   
729  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
730  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
732  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
733  When match() is called in other circumstances, don't add to the chain. The  
734  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
735  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
736  match(). */  to break infinite loops that match no characters. When match() is called in
737    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738    NOT be used with tail recursion, because the memory block that is used is on
739    the stack, so a new one may be required for each match(). */
740    
741  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
742    {    {
743    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
744    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
745    eptrb = &newptrb;    eptrb = &newptrb;
746      md->match_function_type = 0;
747    }    }
748    
749  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 666  for (;;) Line 755  for (;;)
755    
756    switch(op)    switch(op)
757      {      {
758        case OP_MARK:
759        md->nomatch_mark = ecode + 2;
760        md->mark = NULL;    /* In case previously set by assertion */
761        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762          eptrb, RM55);
763        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764             md->mark == NULL) md->mark = ecode + 2;
765    
766        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767        argument, and we must check whether that argument matches this MARK's
768        argument. It is passed back in md->start_match_ptr (an overloading of that
769        variable). If it does match, we reset that variable to the current subject
770        position and return MATCH_SKIP. Otherwise, pass back the return code
771        unaltered. */
772    
773        else if (rrc == MATCH_SKIP_ARG &&
774            STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775          {
776          md->start_match_ptr = eptr;
777          RRETURN(MATCH_SKIP);
778          }
779        RRETURN(rrc);
780    
781      case OP_FAIL:      case OP_FAIL:
782      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
783    
784        case OP_COMMIT:
785        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786          eptrb, RM52);
787        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788        RRETURN(MATCH_COMMIT);
789    
790      case OP_PRUNE:      case OP_PRUNE:
791      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792        ims, eptrb, flags, RM51);        eptrb, RM51);
793      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794      RRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
795    
796      case OP_COMMIT:      case OP_PRUNE_ARG:
797      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      md->nomatch_mark = ecode + 2;
798        ims, eptrb, flags, RM52);      md->mark = NULL;    /* In case previously set by assertion */
799        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800          eptrb, RM56);
801        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
802             md->mark == NULL) md->mark = ecode + 2;
803      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
804      RRETURN(MATCH_COMMIT);      RRETURN(MATCH_PRUNE);
805    
806      case OP_SKIP:      case OP_SKIP:
807      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
808        ims, eptrb, flags, RM53);        eptrb, RM53);
809      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
810      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
811      RRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
812    
813        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
814        nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
815        not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
816        that failed and any that precede it (either they also failed, or were not
817        triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
818        SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
819        set to the count of the one that failed. */
820    
821        case OP_SKIP_ARG:
822        md->skip_arg_count++;
823        if (md->skip_arg_count <= md->ignore_skip_arg)
824          {
825          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
826          break;
827          }
828        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
829          eptrb, RM57);
830        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831    
832        /* Pass back the current skip name by overloading md->start_match_ptr and
833        returning the special MATCH_SKIP_ARG return code. This will either be
834        caught by a matching MARK, or get to the top, where it causes a rematch
835        with md->ignore_skip_arg set to the value of md->skip_arg_count. */
836    
837        md->start_match_ptr = ecode + 2;
838        RRETURN(MATCH_SKIP_ARG);
839    
840        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
841        the branch in which it occurs can be determined. Overload the start of
842        match pointer to do this. */
843    
844      case OP_THEN:      case OP_THEN:
845      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
846        ims, eptrb, flags, RM54);        eptrb, RM54);
847        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
848        md->start_match_ptr = ecode;
849        RRETURN(MATCH_THEN);
850    
851        case OP_THEN_ARG:
852        md->nomatch_mark = ecode + 2;
853        md->mark = NULL;    /* In case previously set by assertion */
854        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
855          md, eptrb, RM58);
856        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
857             md->mark == NULL) md->mark = ecode + 2;
858      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859        md->start_match_ptr = ecode;
860      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
861    
862      /* Handle a capturing bracket. If there is space in the offset vector, save      /* Handle an atomic group that does not contain any capturing parentheses.
863      the current subject position in the working slot at the top of the vector.      This can be handled like an assertion. Prior to 8.13, all atomic groups
864      We mustn't change the current values of the data slot, because they may be      were handled this way. In 8.13, the code was changed as below for ONCE, so
865      set from a previous iteration of this group, and be referred to by a      that backups pass through the group and thereby reset captured values.
866      reference inside the group.      However, this uses a lot more stack, so in 8.20, atomic groups that do not
867        contain any captures generate OP_ONCE_NC, which can be handled in the old,
868      If the bracket fails to match, we need to restore this value and also the      less stack intensive way.
869      values of the final offsets, in case they were set by a previous iteration  
870      of the same bracket.      Check the alternative branches in turn - the matching won't pass the KET
871        for this kind of subpattern. If any one branch matches, we carry on as at
872        the end of a normal bracket, leaving the subject pointer, but resetting
873        the start-of-match value in case it was changed by \K. */
874    
875        case OP_ONCE_NC:
876        prev = ecode;
877        saved_eptr = eptr;
878        save_mark = md->mark;
879        do
880          {
881          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
882          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
883            {
884            mstart = md->start_match_ptr;
885            break;
886            }
887          if (rrc == MATCH_THEN)
888            {
889            next = ecode + GET(ecode,1);
890            if (md->start_match_ptr < next &&
891                (*ecode == OP_ALT || *next == OP_ALT))
892              rrc = MATCH_NOMATCH;
893            }
894    
895          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
896          ecode += GET(ecode,1);
897          md->mark = save_mark;
898          }
899        while (*ecode == OP_ALT);
900    
901        /* If hit the end of the group (which could be repeated), fail */
902    
903        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
904    
905        /* Continue as from after the group, updating the offsets high water
906        mark, since extracts may have been taken. */
907    
908        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
909    
910        offset_top = md->end_offset_top;
911        eptr = md->end_match_ptr;
912    
913        /* For a non-repeating ket, just continue at this level. This also
914        happens for a repeating ket if no characters were matched in the group.
915        This is the forcible breaking of infinite loops as implemented in Perl
916        5.005. */
917    
918        if (*ecode == OP_KET || eptr == saved_eptr)
919          {
920          ecode += 1+LINK_SIZE;
921          break;
922          }
923    
924        /* The repeating kets try the rest of the pattern or restart from the
925        preceding bracket, in the appropriate order. The second "call" of match()
926        uses tail recursion, to avoid using another stack frame. */
927    
928        if (*ecode == OP_KETRMIN)
929          {
930          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
931          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
932          ecode = prev;
933          goto TAIL_RECURSE;
934          }
935        else  /* OP_KETRMAX */
936          {
937          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
938          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
939          ecode += 1 + LINK_SIZE;
940          goto TAIL_RECURSE;
941          }
942        /* Control never gets here */
943    
944        /* Handle a capturing bracket, other than those that are possessive with an
945        unlimited repeat. If there is space in the offset vector, save the current
946        subject position in the working slot at the top of the vector. We mustn't
947        change the current values of the data slot, because they may be set from a
948        previous iteration of this group, and be referred to by a reference inside
949        the group. A failure to match might occur after the group has succeeded,
950        if something later on doesn't match. For this reason, we need to restore
951        the working value and also the values of the final offsets, in case they
952        were set by a previous iteration of the same bracket.
953    
954      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
955      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 713  for (;;) Line 960  for (;;)
960      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
961      offset = number << 1;      offset = number << 1;
962    
963  #ifdef DEBUG  #ifdef PCRE_DEBUG
964      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
965      printf("subject=");      printf("subject=");
966      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 726  for (;;) Line 973  for (;;)
973        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
974        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
975        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
976          save_mark = md->mark;
977    
978        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
979        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
980            (int)(eptr - md->start_subject);
981    
982        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
983          {          {
984          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
986          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
987            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
988    
989            /* If we backed up to a THEN, check whether it is within the current
990            branch by comparing the address of the THEN that is passed back with
991            the end of the branch. If it is within the current branch, and the
992            branch is one of two or more alternatives (it either starts or ends
993            with OP_ALT), we have reached the limit of THEN's action, so convert
994            the return code to NOMATCH, which will cause normal backtracking to
995            happen from now on. Otherwise, THEN is passed back to an outer
996            alternative. This implements Perl's treatment of parenthesized groups,
997            where a group not containing | does not affect the current alternative,
998            that is, (X) is NOT the same as (X|(*F)). */
999    
1000            if (rrc == MATCH_THEN)
1001              {
1002              next = ecode + GET(ecode,1);
1003              if (md->start_match_ptr < next &&
1004                  (*ecode == OP_ALT || *next == OP_ALT))
1005                rrc = MATCH_NOMATCH;
1006              }
1007    
1008            /* Anything other than NOMATCH is passed back. */
1009    
1010            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1011          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1012          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1013            md->mark = save_mark;
1014            if (*ecode != OP_ALT) break;
1015          }          }
       while (*ecode == OP_ALT);  
1016    
1017        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1018        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1019        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1020        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1021    
1022        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1023    
1024          RRETURN(rrc);
1025        }        }
1026    
1027      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 761  for (;;) Line 1035  for (;;)
1035      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037    
1038      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1039      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
1040      recursive call to match() whatever happened. We can reduce stack usage by  
1041      turning this into a tail recursion, except in the case when match_cbegroup      When we get to the final alternative within the brackets, we used to return
1042      is set.*/      the result of a recursive call to match() whatever happened so it was
1043        possible to reduce stack usage by turning this into a tail recursion,
1044        except in the case of a possibly empty group. However, now that there is
1045        the possiblity of (*THEN) occurring in the final alternative, this
1046        optimization is no longer always possible.
1047    
1048        We can optimize if we know there are no (*THEN)s in the pattern; at present
1049        this is the best that can be done.
1050    
1051        MATCH_ONCE is returned when the end of an atomic group is successfully
1052        reached, but subsequent matching fails. It passes back up the tree (causing
1053        captured values to be reset) until the original atomic group level is
1054        reached. This is tested by comparing md->once_target with the start of the
1055        group. At this point, the return is converted into MATCH_NOMATCH so that
1056        previous backup points can be taken. */
1057    
1058        case OP_ONCE:
1059      case OP_BRA:      case OP_BRA:
1060      case OP_SBRA:      case OP_SBRA:
1061      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1062      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1063      for (;;)      for (;;)
1064        {        {
1065        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE)
1066            md->match_function_type = MATCH_CBEGROUP;
1067    
1068          /* If this is not a possibly empty group, and there are no (*THEN)s in
1069          the pattern, and this is the final alternative, optimize as described
1070          above. */
1071    
1072          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073            {
1074            ecode += PRIV(OP_lengths)[*ecode];
1075            goto TAIL_RECURSE;
1076            }
1077    
1078          /* In all other cases, we have to make another call to match(). */
1079    
1080          save_mark = md->mark;
1081          save_capture_last = md->capture_last;
1082          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1083            RM2);
1084    
1085          /* See comment in the code for capturing groups above about handling
1086          THEN. */
1087    
1088          if (rrc == MATCH_THEN)
1089            {
1090            next = ecode + GET(ecode,1);
1091            if (md->start_match_ptr < next &&
1092                (*ecode == OP_ALT || *next == OP_ALT))
1093              rrc = MATCH_NOMATCH;
1094            }
1095    
1096          if (rrc != MATCH_NOMATCH)
1097          {          {
1098          if (flags == 0)    /* Not a possibly empty group */          if (rrc == MATCH_ONCE)
1099            {            {
1100            ecode += _pcre_OP_lengths[*ecode];            const pcre_uchar *scode = ecode;
1101            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
1102            goto TAIL_RECURSE;              {
1103                while (*scode == OP_ALT) scode += GET(scode, 1);
1104                scode -= GET(scode, 1);
1105                }
1106              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1107            }            }
1108            RRETURN(rrc);
1109            }
1110          ecode += GET(ecode, 1);
1111          md->mark = save_mark;
1112          if (*ecode != OP_ALT) break;
1113          md->capture_last = save_capture_last;
1114          }
1115    
1116        RRETURN(MATCH_NOMATCH);
1117    
1118          /* Possibly empty group; can't use tail recursion. */      /* Handle possessive capturing brackets with an unlimited repeat. We come
1119        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120        handled similarly to the normal case above. However, the matching is
1121        different. The end of these brackets will always be OP_KETRPOS, which
1122        returns MATCH_KETRPOS without going further in the pattern. By this means
1123        we can handle the group by iteration rather than recursion, thereby
1124        reducing the amount of stack needed. */
1125    
1126        case OP_CBRAPOS:
1127        case OP_SCBRAPOS:
1128        allow_zero = FALSE;
1129    
1130          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      POSSESSIVE_CAPTURE:
1131            eptrb, flags, RM48);      number = GET2(ecode, 1+LINK_SIZE);
1132          RRETURN(rrc);      offset = number << 1;
1133    
1134    #ifdef PCRE_DEBUG
1135        printf("start possessive bracket %d\n", number);
1136        printf("subject=");
1137        pchars(eptr, 16, TRUE, md);
1138        printf("\n");
1139    #endif
1140    
1141        if (offset < md->offset_max)
1142          {
1143          matched_once = FALSE;
1144          code_offset = (int)(ecode - md->start_code);
1145    
1146          save_offset1 = md->offset_vector[offset];
1147          save_offset2 = md->offset_vector[offset+1];
1148          save_offset3 = md->offset_vector[md->offset_end - number];
1149          save_capture_last = md->capture_last;
1150    
1151          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1152    
1153          /* Each time round the loop, save the current subject position for use
1154          when the group matches. For MATCH_MATCH, the group has matched, so we
1155          restart it with a new subject starting position, remembering that we had
1156          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157          usual. If we haven't matched any alternatives in any iteration, check to
1158          see if a previous iteration matched. If so, the group has matched;
1159          continue from afterwards. Otherwise it has failed; restore the previous
1160          capture values before returning NOMATCH. */
1161    
1162          for (;;)
1163            {
1164            md->offset_vector[md->offset_end - number] =
1165              (int)(eptr - md->start_subject);
1166            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168              eptrb, RM63);
1169            if (rrc == MATCH_KETRPOS)
1170              {
1171              offset_top = md->end_offset_top;
1172              eptr = md->end_match_ptr;
1173              ecode = md->start_code + code_offset;
1174              save_capture_last = md->capture_last;
1175              matched_once = TRUE;
1176              mstart = md->start_match_ptr;    /* In case \K changed it */
1177              continue;
1178              }
1179    
1180            /* See comment in the code for capturing groups above about handling
1181            THEN. */
1182    
1183            if (rrc == MATCH_THEN)
1184              {
1185              next = ecode + GET(ecode,1);
1186              if (md->start_match_ptr < next &&
1187                  (*ecode == OP_ALT || *next == OP_ALT))
1188                rrc = MATCH_NOMATCH;
1189              }
1190    
1191            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1192            md->capture_last = save_capture_last;
1193            ecode += GET(ecode, 1);
1194            if (*ecode != OP_ALT) break;
1195          }          }
1196    
1197        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (!matched_once)
1198        otherwise return. */          {
1199            md->offset_vector[offset] = save_offset1;
1200            md->offset_vector[offset+1] = save_offset2;
1201            md->offset_vector[md->offset_end - number] = save_offset3;
1202            }
1203    
1204          if (allow_zero || matched_once)
1205            {
1206            ecode += 1 + LINK_SIZE;
1207            break;
1208            }
1209    
1210          RRETURN(MATCH_NOMATCH);
1211          }
1212    
1213        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1214        as a non-capturing bracket. */
1215    
1216        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1217        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1218    
1219        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1220    
1221        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1222        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1223    
1224        /* Non-capturing possessive bracket with unlimited repeat. We come here
1225        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1226        without the capturing complication. It is written out separately for speed
1227        and cleanliness. */
1228    
1229        case OP_BRAPOS:
1230        case OP_SBRAPOS:
1231        allow_zero = FALSE;
1232    
1233        POSSESSIVE_NON_CAPTURE:
1234        matched_once = FALSE;
1235        code_offset = (int)(ecode - md->start_code);
1236        save_capture_last = md->capture_last;
1237    
1238        for (;;)
1239          {
1240          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1241          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1242            eptrb, RM48);
1243          if (rrc == MATCH_KETRPOS)
1244            {
1245            offset_top = md->end_offset_top;
1246            eptr = md->end_match_ptr;
1247            ecode = md->start_code + code_offset;
1248            matched_once = TRUE;
1249            mstart = md->start_match_ptr;   /* In case \K reset it */
1250            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        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
         eptrb, flags, RM2);  
       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
1265        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1266          if (*ecode != OP_ALT) break;
1267          md->capture_last = save_capture_last;
1268          }
1269    
1270        if (matched_once || allow_zero)
1271          {
1272          ecode += 1 + LINK_SIZE;
1273          break;
1274        }        }
1275        RRETURN(MATCH_NOMATCH);
1276    
1277      /* Control never reaches here. */      /* Control never reaches here. */
1278    
1279      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than two
1280      two branches. If the condition is false, skipping the first branch takes us      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      past the end of the item if there is only one branch, but that's exactly
1282      exactly what going to the ket would do. As there is only one branch to be      what we want. */
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1283    
1284      case OP_COND:      case OP_COND:
1285      case OP_SCOND:      case OP_SCOND:
1286      codelink= GET(ecode, 1);  
1287        /* The variable codelink will be added to ecode when the condition is
1288        false, to get to the second branch. Setting it to the offset to the ALT
1289        or KET, then incrementing ecode achieves this effect. We now have ecode
1290        pointing to the condition or callout. */
1291    
1292        codelink = GET(ecode, 1);   /* Offset to the second branch */
1293        ecode += 1 + LINK_SIZE;     /* From this opcode */
1294    
1295      /* Because of the way auto-callout works during compile, a callout item is      /* Because of the way auto-callout works during compile, a callout item is
1296      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
1297    
1298      if (ecode[LINK_SIZE+1] == OP_CALLOUT)      if (*ecode == OP_CALLOUT)
1299        {        {
1300        if (pcre_callout != NULL)        if (PUBL(callout) != NULL)
1301          {          {
1302          pcre_callout_block cb;          PUBL(callout_block) cb;
1303          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1304          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[1];
1305          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1306    #if defined COMPILE_PCRE8
1307          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
1308          cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1309          cb.start_match      = mstart - md->start_subject;          cb.subject          = (PCRE_SPTR16)md->start_subject;
1310          cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1311          cb.pattern_position = GET(ecode, LINK_SIZE + 3);          cb.subject          = (PCRE_SPTR32)md->start_subject;
1312          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);  #endif
1313            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1314            cb.start_match      = (int)(mstart - md->start_subject);
1315            cb.current_position = (int)(eptr - md->start_subject);
1316            cb.pattern_position = GET(ecode, 2);
1317            cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1318          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1319          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last & CAPLMASK;
1320            /* Internal change requires this for API compatibility. */
1321            if (cb.capture_last == 0) cb.capture_last = -1;
1322          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1323          if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);          cb.mark             = md->nomatch_mark;
1324            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1325          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1326          }          }
       ecode += _pcre_OP_lengths[OP_CALLOUT];  
       }  
1327    
1328      condcode = ecode[LINK_SIZE+1];        /* Advance ecode past the callout, so it now points to the condition. We
1329          must adjust codelink so that the value of ecode+codelink is unchanged. */
1330    
1331      /* Now see what the actual condition is */        ecode += PRIV(OP_lengths)[OP_CALLOUT];
1332          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
     if (condcode == OP_RREF)         /* Recursion test */  
       {  
       offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/  
       condition = md->recursive != NULL &&  
         (offset == RREF_ANY || offset == md->recursive->group_num);  
       ecode += condition? 3 : GET(ecode, 1);  
1333        }        }
1334    
1335      else if (condcode == OP_CREF)    /* Group used test */      /* Test the various possible conditions */
1336    
1337        condition = FALSE;
1338        switch(condcode = *ecode)
1339        {        {
1340        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        case OP_RREF:         /* Numbered group recursion test */
1341          if (md->recursive != NULL)     /* Not recursing => FALSE */
1342            {
1343            unsigned int recno = GET2(ecode, 1);   /* Recursion group number*/
1344            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1345            }
1346          break;
1347    
1348          case OP_DNRREF:       /* Duplicate named group recursion test */
1349          if (md->recursive != NULL)
1350            {
1351            int count = GET2(ecode, 1 + IMM2_SIZE);
1352            pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1353            while (count-- > 0)
1354              {
1355              unsigned int recno = GET2(slot, 0);
1356              condition = recno == md->recursive->group_num;
1357              if (condition) break;
1358              slot += md->name_entry_size;
1359              }
1360            }
1361          break;
1362    
1363          case OP_CREF:         /* Numbered group used test */
1364          offset = GET2(ecode, 1) << 1;  /* Doubled ref number */
1365        condition = offset < offset_top && md->offset_vector[offset] >= 0;        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1366        ecode += condition? 3 : GET(ecode, 1);        break;
       }  
1367    
1368      else if (condcode == OP_DEF)     /* DEFINE - always false */        case OP_DNCREF:      /* Duplicate named group used test */
1369        {          {
1370        condition = FALSE;          int count = GET2(ecode, 1 + IMM2_SIZE);
1371        ecode += GET(ecode, 1);          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1372        }          while (count-- > 0)
1373              {
1374              offset = GET2(slot, 0) << 1;
1375              condition = offset < offset_top && md->offset_vector[offset] >= 0;
1376              if (condition) break;
1377              slot += md->name_entry_size;
1378              }
1379            }
1380          break;
1381    
1382      /* The condition is an assertion. Call match() to evaluate it - setting        case OP_DEF:     /* DEFINE - always false */
1383      the final argument match_condassert causes it to stop at the end of an        break;
     assertion. */  
1384    
1385      else        /* The condition is an assertion. Call match() to evaluate it - setting
1386        {        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end
1387        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        of an assertion. */
1388            match_condassert, RM3);  
1389          default:
1390          md->match_function_type = MATCH_CONDASSERT;
1391          RMATCH(eptr, ecode, offset_top, md, NULL, RM3);
1392        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1393          {          {
1394            if (md->end_offset_top > offset_top)
1395              offset_top = md->end_offset_top;  /* Captures may have happened */
1396          condition = TRUE;          condition = TRUE;
1397          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);  
1398            /* Advance ecode past the assertion to the start of the first branch,
1399            but adjust it so that the general choosing code below works. */
1400    
1401            ecode += GET(ecode, 1);
1402          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1403            ecode += 1 + LINK_SIZE - PRIV(OP_lengths)[condcode];
1404          }          }
1405    
1406          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1407          assertion; it is therefore treated as NOMATCH. Any other return is an
1408          error. */
1409    
1410        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1411          {          {
1412          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
1413          }          }
1414        else        break;
         {  
         condition = FALSE;  
         ecode += codelink;  
         }  
1415        }        }
1416    
1417      /* We are now at the branch that is to be obeyed. As there is only one,      /* Choose branch according to the condition */
1418      we can use tail recursion to avoid using another stack frame, except when  
1419      match_cbegroup is required for an unlimited repeat of a possibly empty      ecode += condition? PRIV(OP_lengths)[condcode] : codelink;
     group. If the second alternative doesn't exist, we can just plough on. */  
1420    
1421      if (condition || *ecode == OP_ALT)      /* We are now at the branch that is to be obeyed. As there is only one, we
1422        can use tail recursion to avoid using another stack frame, except when
1423        there is unlimited repeat of a possibly empty group. In the latter case, a
1424        recursive call to match() is always required, unless the second alternative
1425        doesn't exist, in which case we can just plough on. Note that, for
1426        compatibility with Perl, the | in a conditional group is NOT treated as
1427        creating two alternatives. If a THEN is encountered in the branch, it
1428        propagates out to the enclosing alternative (unless nested in a deeper set
1429        of alternatives, of course). */
1430    
1431        if (condition || ecode[-(1+LINK_SIZE)] == OP_ALT)
1432        {        {
1433        ecode += 1 + LINK_SIZE;        if (op != OP_SCOND)
       if (op == OP_SCOND)        /* Possibly empty group */  
1434          {          {
         RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);  
         RRETURN(rrc);  
         }  
       else                       /* Group must match something */  
         {  
         flags = 0;  
1435          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1436          }          }
1437    
1438          md->match_function_type = MATCH_CBEGROUP;
1439          RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);
1440          RRETURN(rrc);
1441        }        }
1442      else                         /* Condition false & no alternative */  
1443         /* Condition false & no alternative; continue after the group. */
1444    
1445        else
1446        {        {
       ecode += 1 + LINK_SIZE;  
1447        }        }
1448      break;      break;
1449    
1450    
1451      /* End of the pattern, either real or forced. If we are in a top-level      /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1452      recursion, we should restore the offsets appropriately and continue from      to close any currently open capturing brackets. */
     after the call. */  
1453    
1454      case OP_ACCEPT:      case OP_CLOSE:
1455      case OP_END:      number = GET2(ecode, 1);   /* Must be less than 65536 */
1456      if (md->recursive != NULL && md->recursive->group_num == 0)      offset = number << 1;
1457    
1458    #ifdef PCRE_DEBUG
1459          printf("end bracket %d at *ACCEPT", number);
1460          printf("\n");
1461    #endif
1462    
1463        md->capture_last = (md->capture_last & OVFLMASK) | number;
1464        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1465        {        {
1466        recursion_info *rec = md->recursive;        md->offset_vector[offset] =
1467        DPRINTF(("End of pattern in a (?0) recursion\n"));          md->offset_vector[md->offset_end - number];
1468        md->recursive = rec->prevrec;        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1469        memmove(md->offset_vector, rec->offset_save,        if (offset_top <= offset) offset_top = offset + 2;
         rec->saved_max * sizeof(int));  
       mstart = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
1470        }        }
1471        ecode += 1 + IMM2_SIZE;
1472        break;
1473    
1474    
1475        /* End of the pattern, either real or forced. */
1476    
1477        case OP_END:
1478        case OP_ACCEPT:
1479        case OP_ASSERT_ACCEPT:
1480    
1481      /* Otherwise, if we have matched an empty string, fail if PCRE_NOTEMPTY is      /* If we have matched an empty string, fail if not in an assertion and not
1482      set, or if PCRE_NOTEMPTY_ATSTART is set and we have matched at the start of      in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1483      the subject. In both cases, backtracking will then try other alternatives,      is set and we have matched at the start of the subject. In both cases,
1484      if any. */      backtracking will then try other alternatives, if any. */
1485    
1486      if (eptr == mstart &&      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1487          (md->notempty ||           md->recursive == NULL &&
1488            (md->notempty_atstart &&           (md->notempty ||
1489              mstart == md->start_subject + md->start_offset)))             (md->notempty_atstart &&
1490                 mstart == md->start_subject + md->start_offset)))
1491        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1492    
1493      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
# Line 946  for (;;) Line 1495  for (;;)
1495      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1496      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1497      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
     RRETURN(MATCH_MATCH);  
1498    
1499      /* Change option settings */      /* For some reason, the macros don't work properly if an expression is
1500        given as the argument to RRETURN when the heap is in use. */
1501    
1502      case OP_OPT:      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1503      ims = ecode[1];      RRETURN(rrc);
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1504    
1505      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1506      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,
1507      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
1508      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
1509      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1510        of a condition, we want to return immediately afterwards. The caller of
1511        this incarnation of the match() function will have set MATCH_CONDASSERT in
1512        md->match_function type, and one of these opcodes will be the first opcode
1513        that is processed. We use a local variable that is preserved over calls to
1514        match() to remember this case. */
1515    
1516      case OP_ASSERT:      case OP_ASSERT:
1517      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1518        save_mark = md->mark;
1519        if (md->match_function_type == MATCH_CONDASSERT)
1520          {
1521          condassert = TRUE;
1522          md->match_function_type = 0;
1523          }
1524        else condassert = FALSE;
1525    
1526        /* Loop for each branch */
1527    
1528      do      do
1529        {        {
1530        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1531          RM4);  
1532        if (rrc == MATCH_MATCH) break;        /* A match means that the assertion is true; break out of the loop
1533        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        that matches its alternatives. */
1534    
1535          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1536            {
1537            mstart = md->start_match_ptr;   /* In case \K reset it */
1538            break;
1539            }
1540    
1541          /* If not matched, restore the previous mark setting. */
1542    
1543          md->mark = save_mark;
1544    
1545          /* See comment in the code for capturing groups above about handling
1546          THEN. */
1547    
1548          if (rrc == MATCH_THEN)
1549            {
1550            next = ecode + GET(ecode,1);
1551            if (md->start_match_ptr < next &&
1552                (*ecode == OP_ALT || *next == OP_ALT))
1553              rrc = MATCH_NOMATCH;
1554            }
1555    
1556          /* Anything other than NOMATCH causes the entire assertion to fail,
1557          passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1558          uncaptured THEN, which means they take their normal effect. This
1559          consistent approach does not always have exactly the same effect as in
1560          Perl. */
1561    
1562          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1563        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1564        }        }
1565      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);   /* Continue for next alternative */
1566    
1567        /* If we have tried all the alternative branches, the assertion has
1568        failed. If not, we broke out after a match. */
1569    
1570      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1571    
1572      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1573    
1574      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1575    
1576      /* Continue from after the assertion, updating the offsets high water      /* Continue from after a successful assertion, updating the offsets high
1577      mark, since extracts may have been taken during the assertion. */      water mark, since extracts may have been taken during the assertion. */
1578    
1579      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1580      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1581      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1582      continue;      continue;
1583    
1584      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match for the assertion to
1585        succeed. */
1586    
1587      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1588      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1589        save_mark = md->mark;
1590        if (md->match_function_type == MATCH_CONDASSERT)
1591          {
1592          condassert = TRUE;
1593          md->match_function_type = 0;
1594          }
1595        else condassert = FALSE;
1596    
1597        /* Loop for each alternative branch. */
1598    
1599      do      do
1600        {        {
1601        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1602          RM5);        md->mark = save_mark;   /* Always restore the mark setting */
1603        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);  
1604        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        switch(rrc)
1605            {
1606            case MATCH_MATCH:            /* A successful match means */
1607            case MATCH_ACCEPT:           /* the assertion has failed. */
1608            RRETURN(MATCH_NOMATCH);
1609    
1610            case MATCH_NOMATCH:          /* Carry on with next branch */
1611            break;
1612    
1613            /* See comment in the code for capturing groups above about handling
1614            THEN. */
1615    
1616            case MATCH_THEN:
1617            next = ecode + GET(ecode,1);
1618            if (md->start_match_ptr < next &&
1619                (*ecode == OP_ALT || *next == OP_ALT))
1620              {
1621              rrc = MATCH_NOMATCH;
1622              break;
1623              }
1624            /* Otherwise fall through. */
1625    
1626            /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1627            assertion to fail to match, without considering any more alternatives.
1628            Failing to match means the assertion is true. This is a consistent
1629            approach, but does not always have the same effect as in Perl. */
1630    
1631            case MATCH_COMMIT:
1632            case MATCH_SKIP:
1633            case MATCH_SKIP_ARG:
1634            case MATCH_PRUNE:
1635            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1636            goto NEG_ASSERT_TRUE;   /* Break out of alternation loop */
1637    
1638            /* Anything else is an error */
1639    
1640            default:
1641            RRETURN(rrc);
1642            }
1643    
1644          /* Continue with next branch */
1645    
1646        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1647        }        }
1648      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1649    
1650      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* All branches in the assertion failed to match. */
1651    
1652      ecode += 1 + LINK_SIZE;      NEG_ASSERT_TRUE:
1653        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1654        ecode += 1 + LINK_SIZE;                /* Continue with current branch */
1655      continue;      continue;
1656    
1657      /* Move the subject pointer back. This occurs only at the start of      /* Move the subject pointer back. This occurs only at the start of
# Line 1012  for (;;) Line 1660  for (;;)
1660      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1661    
1662      case OP_REVERSE:      case OP_REVERSE:
1663  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1664      if (utf8)      if (utf)
1665        {        {
1666        i = GET(ecode, 1);        i = GET(ecode, 1);
1667        while (i-- > 0)        while (i-- > 0)
# Line 1044  for (;;) Line 1692  for (;;)
1692      function is able to force a failure. */      function is able to force a failure. */
1693    
1694      case OP_CALLOUT:      case OP_CALLOUT:
1695      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1696        {        {
1697        pcre_callout_block cb;        PUBL(callout_block) cb;
1698        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1699        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1700        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1701    #if defined COMPILE_PCRE8
1702        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1703        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1704        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1705        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1706          cb.subject          = (PCRE_SPTR32)md->start_subject;
1707    #endif
1708          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1709          cb.start_match      = (int)(mstart - md->start_subject);
1710          cb.current_position = (int)(eptr - md->start_subject);
1711        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1712        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1713        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1714        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1715          /* Internal change requires this for API compatibility. */
1716          if (cb.capture_last == 0) cb.capture_last = -1;
1717        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1718        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1719          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1720        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1721        }        }
1722      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 1069  for (;;) Line 1726  for (;;)
1726      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
1727      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1728    
1729      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1730      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
1731      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
1732      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
1733      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
1734      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
1735      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.  
1736    
1737      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
1738      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
1739      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1740        a lot, so he is not to blame for the current way it works. */
1741    
1742      case OP_RECURSE:      case OP_RECURSE:
1743        {        {
1744          recursion_info *ri;
1745          unsigned int recno;
1746    
1747        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1748        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1749          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1750    
1751          /* Check for repeating a recursion without advancing the subject pointer.
1752          This should catch convoluted mutual recursions. (Some simple cases are
1753          caught at compile time.) */
1754    
1755          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1756            if (recno == ri->group_num && eptr == ri->subject_position)
1757              RRETURN(PCRE_ERROR_RECURSELOOP);
1758    
1759        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1760    
1761          new_recursive.group_num = recno;
1762          new_recursive.saved_capture_last = md->capture_last;
1763          new_recursive.subject_position = eptr;
1764        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1765        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1766    
1767        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1768    
1769        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1770    
1771        /* Now save the offset data. */        /* Now save the offset data */
1772    
1773        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1774        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1108  for (;;) Line 1776  for (;;)
1776        else        else
1777          {          {
1778          new_recursive.offset_save =          new_recursive.offset_save =
1779            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1780          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1781          }          }
   
1782        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1783              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1784    
1785        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1786        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1787          recursions, md->recursive might be changed, so reset it before looping.
1788          */
1789    
1790        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1791        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1792        do        do
1793          {          {
1794          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1795            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1796          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1797            memcpy(md->offset_vector, new_recursive.offset_save,
1798                new_recursive.saved_max * sizeof(int));
1799            md->capture_last = new_recursive.saved_capture_last;
1800            md->recursive = new_recursive.prevrec;
1801            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1802            {            {
1803            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1804            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1805              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1806            RRETURN(MATCH_MATCH);  
1807              /* Set where we got to in the subject, and reset the start in case
1808              it was changed by \K. This *is* propagated back out of a recursion,
1809              for Perl compatibility. */
1810    
1811              eptr = md->end_match_ptr;
1812              mstart = md->start_match_ptr;
1813              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1814            }            }
1815          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1816            /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1817            recursion; they cause a NOMATCH for the entire recursion. These codes
1818            are defined in a range that can be tested for. */
1819    
1820            if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1821              RRETURN(MATCH_NOMATCH);
1822    
1823            /* Any return code other than NOMATCH is an error. */
1824    
1825            if (rrc != MATCH_NOMATCH)
1826            {            {
1827            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1828            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1829              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1830            RRETURN(rrc);            RRETURN(rrc);
1831            }            }
1832    
1833          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1834          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1835          }          }
1836        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1152  for (;;) Line 1838  for (;;)
1838        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1839        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1840        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1841          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1842        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1843        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(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));  
       }  
1844    
1845      if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1846        {      break;
       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 */  
1847    
1848      /* 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
1849      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1245  for (;;) Line 1859  for (;;)
1859      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1860    
1861      case OP_BRAZERO:      case OP_BRAZERO:
1862        {      next = ecode + 1;
1863        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1864        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1865        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1866        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1867      break;      break;
1868    
1869      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1870        {      next = ecode + 1;
1871        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1872        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1873        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1874        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1875      break;      break;
1876    
1877      case OP_SKIPZERO:      case OP_SKIPZERO:
1878        {      next = ecode+1;
1879        next = ecode+1;      do next += GET(next,1); while (*next == OP_ALT);
1880        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1881      break;      break;
1882    
1883        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1884        here; just jump to the group, with allow_zero set TRUE. */
1885    
1886        case OP_BRAPOSZERO:
1887        op = *(++ecode);
1888        allow_zero = TRUE;
1889        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1890          goto POSSESSIVE_NON_CAPTURE;
1891    
1892      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1893    
1894      case OP_KET:      case OP_KET:
1895      case OP_KETRMIN:      case OP_KETRMIN:
1896      case OP_KETRMAX:      case OP_KETRMAX:
1897        case OP_KETRPOS:
1898      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1899    
1900      /* 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
1901      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1902      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1903    
1904      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1905        {        {
1906        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1907        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1908        }        }
1909      else saved_eptr = NULL;      else saved_eptr = NULL;
1910    
1911      /* 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
1912      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
1913      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1914        start in case it was changed by \K. */
1915      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1916          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1917          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1918        {        {
1919        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1920        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1921        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1922          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1923        }        }
1924    
1925      /* 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
1926      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1927      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
1928      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
1929      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
1930        the current subject position and start match pointer and give a MATCH
1931        return. */
1932    
1933      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1934            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1935        {        {
1936        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1937        offset = number << 1;        offset = number << 1;
1938    
1939  #ifdef DEBUG  #ifdef PCRE_DEBUG
1940        printf("end bracket %d", number);        printf("end bracket %d", number);
1941        printf("\n");        printf("\n");
1942  #endif  #endif
1943    
1944        md->capture_last = number;        /* Handle a recursively called group. */
1945        if (offset >= md->offset_max) md->offset_overflow = TRUE; else  
1946          if (md->recursive != NULL && md->recursive->group_num == number)
1947          {          {
1948          md->offset_vector[offset] =          md->end_match_ptr = eptr;
1949            md->offset_vector[md->offset_end - number];          md->start_match_ptr = mstart;
1950          md->offset_vector[offset+1] = eptr - md->start_subject;          RRETURN(MATCH_MATCH);
         if (offset_top <= offset) offset_top = offset + 2;  
1951          }          }
1952    
1953        /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
       appropriately and continue from after the call. */  
1954    
1955        if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
1956          if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1957          {          {
1958          recursion_info *rec = md->recursive;          /* If offset is greater than offset_top, it means that we are
1959          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          "skipping" a capturing group, and that group's offsets must be marked
1960          md->recursive = rec->prevrec;          unset. In earlier versions of PCRE, all the offsets were unset at the
1961          mstart = rec->save_start;          start of matching, but this doesn't work because atomic groups and
1962          memcpy(md->offset_vector, rec->offset_save,          assertions can cause a value to be set that should later be unset.
1963            rec->saved_max * sizeof(int));          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1964          ecode = rec->after_call;          part of the atomic group, but this is not on the final matching path,
1965          ims = original_ims;          so must be unset when 2 is set. (If there is no group 2, there is no
1966          break;          problem, because offset_top will then be 2, indicating no capture.) */
1967          }  
1968        }          if (offset > offset_top)
1969              {
1970              register int *iptr = md->offset_vector + offset_top;
1971              register int *iend = md->offset_vector + offset;
1972              while (iptr < iend) *iptr++ = -1;
1973              }
1974    
1975      /* For both capturing and non-capturing groups, reset the value of the ims          /* Now make the extraction */
     flags, in case they got changed during the group. */  
1976    
1977      ims = original_ims;          md->offset_vector[offset] =
1978      DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1979            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1980            if (offset_top <= offset) offset_top = offset + 2;
1981            }
1982          }
1983    
1984      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1985      happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1986      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1987      5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group that includes captures,
1988      course of events. */      establish a backup point by processing the rest of the pattern at a lower
1989        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1990        original OP_ONCE level, thereby bypassing intermediate backup points, but
1991        resetting any captures that happened along the way. */
1992    
1993      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1994        {        {
1995        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1996            {
1997            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1998            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1999            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2000            RRETURN(MATCH_ONCE);
2001            }
2002          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2003        break;        break;
2004        }        }
2005    
2006      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2007      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
2008      tail recursion to avoid using another stack frame, unless we have an      at a time from the outer level, thus saving stack. */
2009      unlimited repeat of a group that can match an empty string. */  
2010        if (*ecode == OP_KETRPOS)
2011          {
2012          md->start_match_ptr = mstart;    /* In case \K reset it */
2013          md->end_match_ptr = eptr;
2014          md->end_offset_top = offset_top;
2015          RRETURN(MATCH_KETRPOS);
2016          }
2017    
2018      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      /* The normal repeating kets try the rest of the pattern or restart from
2019        the preceding bracket, in the appropriate order. In the second case, we can
2020        use tail recursion to avoid using another stack frame, unless we have an
2021        an atomic group or an unlimited repeat of a group that can match an empty
2022        string. */
2023    
2024      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2025        {        {
2026        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2027        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2028        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2029            {
2030            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2031            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2032            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2033            RRETURN(MATCH_ONCE);
2034            }
2035          if (*prev >= OP_SBRA)    /* Could match an empty string */
2036          {          {
2037          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2038          RRETURN(rrc);          RRETURN(rrc);
2039          }          }
2040        ecode = prev;        ecode = prev;
# Line 1384  for (;;) Line 2042  for (;;)
2042        }        }
2043      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2044        {        {
2045        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2046          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2047        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2048          if (*prev == OP_ONCE)
2049            {
2050            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2051            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2052            md->once_target = prev;
2053            RRETURN(MATCH_ONCE);
2054            }
2055        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2056        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2057        }        }
2058      /* Control never gets here */      /* Control never gets here */
2059    
2060      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2061    
2062      case OP_CIRC:      case OP_CIRC:
2063      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 */  
2064    
2065      /* Start of subject assertion */      /* Start of subject assertion */
2066    
# Line 1413  for (;;) Line 2069  for (;;)
2069      ecode++;      ecode++;
2070      break;      break;
2071    
2072        /* Multiline mode: start of subject unless notbol, or after any newline. */
2073    
2074        case OP_CIRCM:
2075        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2076        if (eptr != md->start_subject &&
2077            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2078          RRETURN(MATCH_NOMATCH);
2079        ecode++;
2080        break;
2081    
2082      /* Start of match assertion */      /* Start of match assertion */
2083    
2084      case OP_SOM:      case OP_SOM:
# Line 1427  for (;;) Line 2093  for (;;)
2093      ecode++;      ecode++;
2094      break;      break;
2095    
2096      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2097      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2098    
2099      case OP_DOLL:      case OP_DOLLM:
2100      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2101        {        {
2102        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2103          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2104        else          if (md->partial != 0 &&
2105          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2106        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2107        break;              NLBLOCK->nllen == 2 &&
2108                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2109              {
2110              md->hitend = TRUE;
2111              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2112              }
2113            RRETURN(MATCH_NOMATCH);
2114            }
2115        }        }
2116      else      else
2117        {        {
2118        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2119        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2120        }        }
2121        ecode++;
2122        break;
2123    
2124        /* Not multiline mode: assert before a terminating newline or before end of
2125        subject unless noteol is set. */
2126    
2127        case OP_DOLL:
2128        if (md->noteol) RRETURN(MATCH_NOMATCH);
2129        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2130    
2131      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2132    
2133      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2134    
2135      case OP_EOD:      case OP_EOD:
2136      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2137        SCHECK_PARTIAL();
2138      ecode++;      ecode++;
2139      break;      break;
2140    
2141      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2142    
2143      case OP_EODN:      case OP_EODN:
2144      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2145        if (eptr < md->end_subject &&
2146          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2147          {
2148          if (md->partial != 0 &&
2149              eptr + 1 >= md->end_subject &&
2150              NLBLOCK->nltype == NLTYPE_FIXED &&
2151              NLBLOCK->nllen == 2 &&
2152              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2153            {
2154            md->hitend = TRUE;
2155            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2156            }
2157        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2158          }
2159    
2160        /* Either at end of string or \n before end. */
2161    
2162        SCHECK_PARTIAL();
2163      ecode++;      ecode++;
2164      break;      break;
2165    
# Line 1481  for (;;) Line 2174  for (;;)
2174        be "non-word" characters. Remember the earliest consulted character for        be "non-word" characters. Remember the earliest consulted character for
2175        partial matching. */        partial matching. */
2176    
2177  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2178        if (utf8)        if (utf)
2179          {          {
2180            /* Get status of previous character */
2181    
2182          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2183            {            {
2184            USPTR lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2185            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2186            if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;            if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2187            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2188    #ifdef SUPPORT_UCP
2189              if (md->use_ucp)
2190                {
2191                if (c == '_') prev_is_word = TRUE; else
2192                  {
2193                  int cat = UCD_CATEGORY(c);
2194                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2195                  }
2196                }
2197              else
2198    #endif
2199            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2200            }            }
2201    
2202            /* Get status of next character */
2203    
2204          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
2205            {            {
2206            SCHECK_PARTIAL();            SCHECK_PARTIAL();
# Line 1500  for (;;) Line 2209  for (;;)
2209          else          else
2210            {            {
2211            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2212    #ifdef SUPPORT_UCP
2213              if (md->use_ucp)
2214                {
2215                if (c == '_') cur_is_word = TRUE; else
2216                  {
2217                  int cat = UCD_CATEGORY(c);
2218                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2219                  }
2220                }
2221              else
2222    #endif
2223            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2224            }            }
2225          }          }
2226        else        else
2227  #endif  #endif
2228    
2229        /* Not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2230          consistency with the behaviour of \w we do use it in this case. */
2231    
2232          {          {
2233            /* Get status of previous character */
2234    
2235          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2236            {            {
2237            if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;            if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2238            prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);  #ifdef SUPPORT_UCP
2239              if (md->use_ucp)
2240                {
2241                c = eptr[-1];
2242                if (c == '_') prev_is_word = TRUE; else
2243                  {
2244                  int cat = UCD_CATEGORY(c);
2245                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2246                  }
2247                }
2248              else
2249    #endif
2250              prev_is_word = MAX_255(eptr[-1])
2251                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2252            }            }
2253    
2254            /* Get status of next character */
2255    
2256          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
2257            {            {
2258            SCHECK_PARTIAL();            SCHECK_PARTIAL();
2259            cur_is_word = FALSE;            cur_is_word = FALSE;
2260            }            }
2261          else cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);          else
2262    #ifdef SUPPORT_UCP
2263            if (md->use_ucp)
2264              {
2265              c = *eptr;
2266              if (c == '_') cur_is_word = TRUE; else
2267                {
2268                int cat = UCD_CATEGORY(c);
2269                cur_is_word = (cat == ucp_L || cat == ucp_N);
2270                }
2271              }
2272            else
2273    #endif
2274            cur_is_word = MAX_255(*eptr)
2275              && ((md->ctypes[*eptr] & ctype_word) != 0);
2276          }          }
2277    
2278        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1530  for (;;) Line 2283  for (;;)
2283        }        }
2284      break;      break;
2285    
2286      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2287        CRLF newlines and partial matching. */
2288    
2289      case OP_ANY:      case OP_ANY:
2290      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2291        if (md->partial != 0 &&
2292            eptr + 1 >= md->end_subject &&
2293            NLBLOCK->nltype == NLTYPE_FIXED &&
2294            NLBLOCK->nllen == 2 &&
2295            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2296          {
2297          md->hitend = TRUE;
2298          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2299          }
2300    
2301      /* Fall through */      /* Fall through */
2302    
2303        /* Match any single character whatsoever. */
2304    
2305      case OP_ALLANY:      case OP_ALLANY:
2306      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2307        {        {                            /* not be updated before SCHECK_PARTIAL. */
2308        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2309        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2310        }        }
2311      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      eptr++;
2312    #ifdef SUPPORT_UTF
2313        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2314    #endif
2315      ecode++;      ecode++;
2316      break;      break;
2317    
# Line 1550  for (;;) Line 2319  for (;;)
2319      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2320    
2321      case OP_ANYBYTE:      case OP_ANYBYTE:
2322      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2323        {        {                            /* not be updated before SCHECK_PARTIAL. */
2324        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2325        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2326        }        }
2327        eptr++;
2328      ecode++;      ecode++;
2329      break;      break;
2330    
# Line 1566  for (;;) Line 2336  for (;;)
2336        }        }
2337      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2338      if (      if (
2339  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2340         c < 256 &&         c < 256 &&
2341  #endif  #endif
2342         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1583  for (;;) Line 2353  for (;;)
2353        }        }
2354      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2355      if (      if (
2356  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2357         c >= 256 ||         c > 255 ||
2358  #endif  #endif
2359         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2360         )         )
# Line 1600  for (;;) Line 2370  for (;;)
2370        }        }
2371      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2372      if (      if (
2373  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2374         c < 256 &&         c < 256 &&
2375  #endif  #endif
2376         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1617  for (;;) Line 2387  for (;;)
2387        }        }
2388      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2389      if (      if (
2390  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2391         c >= 256 ||         c > 255 ||
2392  #endif  #endif
2393         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2394         )         )
# Line 1634  for (;;) Line 2404  for (;;)
2404        }        }
2405      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2406      if (      if (
2407  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2408         c < 256 &&         c < 256 &&
2409  #endif  #endif
2410         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1651  for (;;) Line 2421  for (;;)
2421        }        }
2422      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2423      if (      if (
2424  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2425         c >= 256 ||         c > 255 ||
2426  #endif  #endif
2427         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2428         )         )
# Line 1670  for (;;) Line 2440  for (;;)
2440      switch(c)      switch(c)
2441        {        {
2442        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2443        case 0x000d:  
2444        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2445          if (eptr >= md->end_subject)
2446            {
2447            SCHECK_PARTIAL();
2448            }
2449          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2450        break;        break;
2451    
2452        case 0x000a:        case CHAR_LF:
2453        break;        break;
2454    
2455        case 0x000b:        case CHAR_VT:
2456        case 0x000c:        case CHAR_FF:
2457        case 0x0085:        case CHAR_NEL:
2458    #ifndef EBCDIC
2459        case 0x2028:        case 0x2028:
2460        case 0x2029:        case 0x2029:
2461    #endif  /* Not EBCDIC */
2462        if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);        if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2463        break;        break;
2464        }        }
# Line 1697  for (;;) Line 2474  for (;;)
2474      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2475      switch(c)      switch(c)
2476        {        {
2477          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2478        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);  
2479        }        }
2480      ecode++;      ecode++;
2481      break;      break;
# Line 1731  for (;;) Line 2489  for (;;)
2489      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2490      switch(c)      switch(c)
2491        {        {
2492          HSPACE_CASES: break;  /* Byte and multibyte cases */
2493        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;  
2494        }        }
2495      ecode++;      ecode++;
2496      break;      break;
# Line 1765  for (;;) Line 2504  for (;;)
2504      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2505      switch(c)      switch(c)
2506        {        {
2507          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2508        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);  
2509        }        }
2510      ecode++;      ecode++;
2511      break;      break;
# Line 1787  for (;;) Line 2519  for (;;)
2519      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2520      switch(c)      switch(c)
2521        {        {
2522          VSPACE_CASES: break;
2523        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;  
2524        }        }
2525      ecode++;      ecode++;
2526      break;      break;
# Line 1813  for (;;) Line 2538  for (;;)
2538        }        }
2539      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2540        {        {
2541          const pcre_uint32 *cp;
2542        const ucd_record *prop = GET_UCD(c);        const ucd_record *prop = GET_UCD(c);
2543    
2544        switch(ecode[1])        switch(ecode[1])
# Line 1826  for (;;) Line 2552  for (;;)
2552               prop->chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2553               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2554            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2555           break;          break;
2556    
2557          case PT_GC:          case PT_GC:
2558          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2559            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2560          break;          break;
2561    
# Line 1843  for (;;) Line 2569  for (;;)
2569            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2570          break;          break;
2571    
2572          default:          /* These are specials */
2573          RRETURN(PCRE_ERROR_INTERNAL);  
2574          }          case PT_ALNUM:
2575            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2576                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2577              RRETURN(MATCH_NOMATCH);
2578            break;
2579    
2580            /* Perl space used to exclude VT, but from Perl 5.18 it is included,
2581            which means that Perl space and POSIX space are now identical. PCRE
2582            was changed at release 8.34. */
2583    
2584            case PT_SPACE:    /* Perl space */
2585            case PT_PXSPACE:  /* POSIX space */
2586            switch(c)
2587              {
2588              HSPACE_CASES:
2589              VSPACE_CASES:
2590              if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2591              break;
2592    
2593              default:
2594              if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z) ==
2595                (op == OP_NOTPROP)) RRETURN(MATCH_NOMATCH);
2596              break;
2597              }
2598            break;
2599    
2600            case PT_WORD:
2601            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2602                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2603                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2604              RRETURN(MATCH_NOMATCH);
2605            break;
2606    
2607            case PT_CLIST:
2608            cp = PRIV(ucd_caseless_sets) + ecode[2];
2609            for (;;)
2610              {
2611              if (c < *cp)
2612                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2613              if (c == *cp++)
2614                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2615              }
2616            break;
2617    
2618            case PT_UCNC:
2619            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2620                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2621                 c >= 0xe000) == (op == OP_NOTPROP))
2622              RRETURN(MATCH_NOMATCH);
2623            break;
2624    
2625            /* This should never occur */
2626    
2627            default:
2628            RRETURN(PCRE_ERROR_INTERNAL);
2629            }
2630    
2631        ecode += 3;        ecode += 3;
2632        }        }
# Line 1860  for (;;) Line 2641  for (;;)
2641        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2642        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2643        }        }
2644      GETCHARINCTEST(c, eptr);      else
2645        {        {
2646        int category = UCD_CATEGORY(c);        int lgb, rgb;
2647        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        GETCHARINCTEST(c, eptr);
2648          lgb = UCD_GRAPHBREAK(c);
2649        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2650          {          {
2651          int len = 1;          int len = 1;
2652          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2653            {          rgb = UCD_GRAPHBREAK(c);
2654            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2655            }          lgb = rgb;
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
2656          eptr += len;          eptr += len;
2657          }          }
2658        }        }
2659        CHECK_PARTIAL();
2660      ecode++;      ecode++;
2661      break;      break;
2662  #endif  #endif  /* SUPPORT_UCP */
2663    
2664    
2665      /* 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 1887  for (;;) Line 2668  for (;;)
2668      similar code to character type repeats - written out again for speed.      similar code to character type repeats - written out again for speed.
2669      However, if the referenced string is the empty string, always treat      However, if the referenced string is the empty string, always treat
2670      it as matched, any number of times (otherwise there could be infinite      it as matched, any number of times (otherwise there could be infinite
2671      loops). */      loops). If the reference is unset, there are two possibilities:
2672    
2673      case OP_REF:      (a) In the default, Perl-compatible state, set the length negative;
2674        {      this ensures that every attempt at a match fails. We can't just fail
2675        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      here, because of the possibility of quantifiers with zero minima.
       ecode += 3;  
2676    
2677        /* If the reference is unset, there are two possibilities:      (b) If the JavaScript compatibility flag is set, set the length to zero
2678        so that the back reference matches an empty string.
2679    
2680        (a) In the default, Perl-compatible state, set the length to be longer      Otherwise, set the length to the length of what was matched by the
2681        than the amount of subject left; this ensures that every attempt at a      referenced subpattern.
       match fails. We can't just fail here, because of the possibility of  
       quantifiers with zero minima.  
2682    
2683        (b) If the JavaScript compatibility flag is set, set the length to zero      The OP_REF and OP_REFI opcodes are used for a reference to a numbered group
2684        so that the back reference matches an empty string.      or to a non-duplicated named group. For a duplicated named group, OP_DNREF
2685        and OP_DNREFI are used. In this case we must scan the list of groups to
2686        which the name refers, and use the first one that is set. */
2687    
2688        Otherwise, set the length to the length of what was matched by the      case OP_DNREF:
2689        referenced subpattern. */      case OP_DNREFI:
2690        caseless = op == OP_DNREFI;
2691          {
2692          int count = GET2(ecode, 1+IMM2_SIZE);
2693          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
2694          ecode += 1 + 2*IMM2_SIZE;
2695    
2696        if (offset >= offset_top || md->offset_vector[offset] < 0)        while (count-- > 0)
2697          length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;          {
2698            offset = GET2(slot, 0) << 1;
2699            if (offset < offset_top && md->offset_vector[offset] >= 0) break;
2700            slot += md->name_entry_size;
2701            }
2702          if (count < 0)
2703            length = (md->jscript_compat)? 0 : -1;
2704        else        else
2705          length = md->offset_vector[offset+1] - md->offset_vector[offset];          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2706          }
2707        goto REF_REPEAT;
2708    
2709        case OP_REF:
2710        case OP_REFI:
2711        caseless = op == OP_REFI;
2712        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2713        ecode += 1 + IMM2_SIZE;
2714        if (offset >= offset_top || md->offset_vector[offset] < 0)
2715          length = (md->jscript_compat)? 0 : -1;
2716        else
2717          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2718    
2719        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2720    
2721        switch (*ecode)      REF_REPEAT:
2722        switch (*ecode)
2723          {
2724          case OP_CRSTAR:
2725          case OP_CRMINSTAR:
2726          case OP_CRPLUS:
2727          case OP_CRMINPLUS:
2728          case OP_CRQUERY:
2729          case OP_CRMINQUERY:
2730          c = *ecode++ - OP_CRSTAR;
2731          minimize = (c & 1) != 0;
2732          min = rep_min[c];                 /* Pick up values from tables; */
2733          max = rep_max[c];                 /* zero for max => infinity */
2734          if (max == 0) max = INT_MAX;
2735          break;
2736    
2737          case OP_CRRANGE:
2738          case OP_CRMINRANGE:
2739          minimize = (*ecode == OP_CRMINRANGE);
2740          min = GET2(ecode, 1);
2741          max = GET2(ecode, 1 + IMM2_SIZE);
2742          if (max == 0) max = INT_MAX;
2743          ecode += 1 + 2 * IMM2_SIZE;
2744          break;
2745    
2746          default:               /* No repeat follows */
2747          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2748          {          {
2749          case OP_CRSTAR:          if (length == -2) eptr = md->end_subject;   /* Partial match */
2750          case OP_CRMINSTAR:          CHECK_PARTIAL();
2751          case OP_CRPLUS:          RRETURN(MATCH_NOMATCH);
2752          case OP_CRMINPLUS:          }
2753          case OP_CRQUERY:        eptr += length;
2754          case OP_CRMINQUERY:        continue;              /* With the main loop */
2755          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;  
2756    
2757          case OP_CRRANGE:      /* Handle repeated back references. If the length of the reference is
2758          case OP_CRMINRANGE:      zero, just continue with the main loop. If the length is negative, it
2759          minimize = (*ecode == OP_CRMINRANGE);      means the reference is unset in non-Java-compatible mode. If the minimum is
2760          min = GET2(ecode, 1);      zero, we can continue at the same level without recursion. For any other
2761          max = GET2(ecode, 3);      minimum, carrying on will result in NOMATCH. */
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2762    
2763          default:               /* No repeat follows */      if (length == 0) continue;
2764          if (!match_ref(offset, eptr, length, md, ims))      if (length < 0 && min == 0) continue;
2765            {  
2766            CHECK_PARTIAL();      /* First, ensure the minimum number of matches are present. We get back
2767            RRETURN(MATCH_NOMATCH);      the length of the reference string explicitly rather than passing the
2768            }      address of eptr, so that eptr can be a register variable. */
2769          eptr += length;  
2770          continue;              /* With the main loop */      for (i = 1; i <= min; i++)
2771          {
2772          int slength;
2773          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2774            {
2775            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2776            CHECK_PARTIAL();
2777            RRETURN(MATCH_NOMATCH);
2778          }          }
2779          eptr += slength;
2780          }
2781    
2782        /* If the length of the reference is zero, just continue with the      /* If min = max, continue at the same level without recursion.
2783        main loop. */      They are not both allowed to be zero. */
2784    
2785        if (length == 0) continue;      if (min == max) continue;
2786    
2787        /* First, ensure the minimum number of matches are present. We get back      /* If minimizing, keep trying and advancing the pointer */
       the length of the reference string explicitly rather than passing the  
       address of eptr, so that eptr can be a register variable. */  
2788    
2789        for (i = 1; i <= min; i++)      if (minimize)
2790          {
2791          for (fi = min;; fi++)
2792          {          {
2793          if (!match_ref(offset, eptr, length, md, ims))          int slength;
2794            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2795            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2796            if (fi >= max) RRETURN(MATCH_NOMATCH);
2797            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2798            {            {
2799              if (slength == -2) eptr = md->end_subject;   /* Partial match */
2800            CHECK_PARTIAL();            CHECK_PARTIAL();
2801            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2802            }            }
2803          eptr += length;          eptr += slength;
2804          }          }
2805          /* Control never gets here */
2806          }
2807    
2808        /* If min = max, continue at the same level without recursion.      /* If maximizing, find the longest string and work backwards */
       They are not both allowed to be zero. */  
   
       if (min == max) continue;  
   
       /* If minimizing, keep trying and advancing the pointer */  
2809    
2810        if (minimize)      else
2811          {
2812          pp = eptr;
2813          for (i = min; i < max; i++)
2814          {          {
2815          for (fi = min;; fi++)          int slength;
2816            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2817            {            {
2818            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2819            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2820            if (fi >= max) RRETURN(MATCH_NOMATCH);  
2821            if (!match_ref(offset, eptr, length, md, ims))            if (slength == -2 && md->partial != 0 &&
2822                  md->end_subject > md->start_used_ptr)
2823              {              {
2824              CHECK_PARTIAL();              md->hitend = TRUE;
2825              RRETURN(MATCH_NOMATCH);              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2826              }              }
2827            eptr += length;            break;
2828            }            }
2829          /* Control never gets here */          eptr += slength;
2830          }          }
2831    
2832        /* If maximizing, find the longest string and work backwards */        while (eptr >= pp)
   
       else  
2833          {          {
2834          pp = eptr;          RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2835          for (i = min; i < max; i++)          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2836            {          eptr -= length;
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
           {  
           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);  
           if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
           eptr -= length;  
           }  
         RRETURN(MATCH_NOMATCH);  
2837          }          }
2838          RRETURN(MATCH_NOMATCH);
2839        }        }
2840      /* Control never gets here */      /* Control never gets here */
2841    
# Line 2026  for (;;) Line 2853  for (;;)
2853      case OP_NCLASS:      case OP_NCLASS:
2854      case OP_CLASS:      case OP_CLASS:
2855        {        {
2856          /* The data variable is saved across frames, so the byte map needs to
2857          be stored there. */
2858    #define BYTE_MAP ((pcre_uint8 *)data)
2859        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2860        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2861    
2862        switch (*ecode)        switch (*ecode)
2863          {          {
# Line 2037  for (;;) Line 2867  for (;;)
2867          case OP_CRMINPLUS:          case OP_CRMINPLUS:
2868          case OP_CRQUERY:          case OP_CRQUERY:
2869          case OP_CRMINQUERY:          case OP_CRMINQUERY:
2870            case OP_CRPOSSTAR:
2871            case OP_CRPOSPLUS:
2872            case OP_CRPOSQUERY:
2873          c = *ecode++ - OP_CRSTAR;          c = *ecode++ - OP_CRSTAR;
2874          minimize = (c & 1) != 0;          if (c < OP_CRPOSSTAR - OP_CRSTAR) minimize = (c & 1) != 0;
2875            else possessive = TRUE;
2876          min = rep_min[c];                 /* Pick up values from tables; */          min = rep_min[c];                 /* Pick up values from tables; */
2877          max = rep_max[c];                 /* zero for max => infinity */          max = rep_max[c];                 /* zero for max => infinity */
2878          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
# Line 2046  for (;;) Line 2880  for (;;)
2880    
2881          case OP_CRRANGE:          case OP_CRRANGE:
2882          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2883            case OP_CRPOSRANGE:
2884          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2885            possessive = (*ecode == OP_CRPOSRANGE);
2886          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2887          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2888          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2889          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2890          break;          break;
2891    
2892          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2060  for (;;) Line 2896  for (;;)
2896    
2897        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2898    
2899  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2900        /* UTF-8 mode */        if (utf)
       if (utf8)  
2901          {          {
2902          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2903            {            {
# Line 2077  for (;;) Line 2912  for (;;)
2912              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2913              }              }
2914            else            else
2915              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2916            }            }
2917          }          }
2918        else        else
2919  #endif  #endif
2920        /* Not UTF-8 mode */        /* Not UTF mode */
2921          {          {
2922          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2923            {            {
# Line 2094  for (;;) Line 2927  for (;;)
2927              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
2928              }              }
2929            c = *eptr++;            c = *eptr++;
2930            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2931              if (c > 255)
2932                {
2933                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2934                }
2935              else
2936    #endif
2937                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2938            }            }
2939          }          }
2940    
# Line 2108  for (;;) Line 2948  for (;;)
2948    
2949        if (minimize)        if (minimize)
2950          {          {
2951  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2952          /* UTF-8 mode */          if (utf)
         if (utf8)  
2953            {            {
2954            for (fi = min;; fi++)            for (fi = min;; fi++)
2955              {              {
2956              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2957              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2958              if (fi >= max) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2959              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2128  for (;;) Line 2967  for (;;)
2967                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2968                }                }
2969              else              else
2970                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2971              }              }
2972            }            }
2973          else          else
2974  #endif  #endif
2975          /* Not UTF-8 mode */          /* Not UTF mode */
2976            {            {
2977            for (fi = min;; fi++)            for (fi = min;; fi++)
2978              {              {
2979              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2980              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2981              if (fi >= max) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2982              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
# Line 2148  for (;;) Line 2985  for (;;)
2985                RRETURN(MATCH_NOMATCH);                RRETURN(MATCH_NOMATCH);
2986                }                }
2987              c = *eptr++;              c = *eptr++;
2988              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2989                if (c > 255)
2990                  {
2991                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2992                  }
2993                else
2994    #endif
2995                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2996              }              }
2997            }            }
2998          /* Control never gets here */          /* Control never gets here */
# Line 2160  for (;;) Line 3004  for (;;)
3004          {          {
3005          pp = eptr;          pp = eptr;
3006    
3007  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3008          /* UTF-8 mode */          if (utf)
         if (utf8)  
3009            {            {
3010            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3011              {              {
3012              int len = 1;              int len = 1;
3013              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3014                  {
3015                  SCHECK_PARTIAL();
3016                  break;
3017                  }
3018              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
3019              if (c > 255)              if (c > 255)
3020                {                {
3021                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3022                }                }
3023              else              else
3024                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3025              eptr += len;              eptr += len;
3026              }              }
3027    
3028              if (possessive) continue;    /* No backtracking */
3029    
3030            for (;;)            for (;;)
3031              {              {
3032              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3033              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3034              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3035              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2189  for (;;) Line 3037  for (;;)
3037            }            }
3038          else          else
3039  #endif  #endif
3040            /* Not UTF-8 mode */            /* Not UTF mode */
3041            {            {
3042            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3043              {              {
3044              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3045                  {
3046                  SCHECK_PARTIAL();
3047                  break;
3048                  }
3049              c = *eptr;              c = *eptr;
3050              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3051                if (c > 255)
3052                  {
3053                  if (op == OP_CLASS) break;
3054                  }
3055                else
3056    #endif
3057                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3058              eptr++;              eptr++;
3059              }              }
3060    
3061              if (possessive) continue;    /* No backtracking */
3062    
3063            while (eptr >= pp)            while (eptr >= pp)
3064              {              {
3065              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3066              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3067              eptr--;              eptr--;
3068              }              }
# Line 2208  for (;;) Line 3070  for (;;)
3070    
3071          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3072          }          }
3073    #undef BYTE_MAP
3074        }        }
3075      /* Control never gets here */      /* Control never gets here */
3076    
# Line 2216  for (;;) Line 3079  for (;;)
3079      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
3080      mode, because Unicode properties are supported in non-UTF-8 mode. */      mode, because Unicode properties are supported in non-UTF-8 mode. */
3081    
3082  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3083      case OP_XCLASS:      case OP_XCLASS:
3084        {        {
3085        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2230  for (;;) Line 3093  for (;;)
3093          case OP_CRMINPLUS:          case OP_CRMINPLUS:
3094          case OP_CRQUERY:          case OP_CRQUERY:
3095          case OP_CRMINQUERY:          case OP_CRMINQUERY:
3096            case OP_CRPOSSTAR:
3097            case OP_CRPOSPLUS:
3098            case OP_CRPOSQUERY:
3099          c = *ecode++ - OP_CRSTAR;          c = *ecode++ - OP_CRSTAR;
3100          minimize = (c & 1) != 0;          if (c < OP_CRPOSSTAR - OP_CRSTAR) minimize = (c & 1) != 0;
3101            else possessive = TRUE;
3102          min = rep_min[c];                 /* Pick up values from tables; */          min = rep_min[c];                 /* Pick up values from tables; */
3103          max = rep_max[c];                 /* zero for max => infinity */          max = rep_max[c];                 /* zero for max => infinity */
3104          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
# Line 2239  for (;;) Line 3106  for (;;)
3106    
3107          case OP_CRRANGE:          case OP_CRRANGE:
3108          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3109            case OP_CRPOSRANGE:
3110          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3111            possessive = (*ecode == OP_CRPOSRANGE);
3112          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3113          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3114          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3115          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3116          break;          break;
3117    
3118          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2261  for (;;) Line 3130  for (;;)
3130            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3131            }            }
3132          GETCHARINCTEST(c, eptr);          GETCHARINCTEST(c, eptr);
3133          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);          if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3134          }          }
3135    
3136        /* 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 2276  for (;;) Line 3145  for (;;)
3145          {          {
3146          for (fi = min;; fi++)          for (fi = min;; fi++)
3147            {            {
3148            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3149            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3150            if (fi >= max) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3151            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 2285  for (;;) Line 3154  for (;;)
3154              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3155              }              }
3156            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3157            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3158            }            }
3159          /* Control never gets here */          /* Control never gets here */
3160          }          }
# Line 2298  for (;;) Line 3167  for (;;)
3167          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3168            {            {
3169            int len = 1;            int len = 1;
3170            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3171                {
3172                SCHECK_PARTIAL();
3173                break;
3174                }
3175    #ifdef SUPPORT_UTF
3176            GETCHARLENTEST(c, eptr, len);            GETCHARLENTEST(c, eptr, len);
3177            if (!_pcre_xclass(c, data)) break;  #else
3178              c = *eptr;
3179    #endif
3180              if (!PRIV(xclass)(c, data, utf)) break;
3181            eptr += len;            eptr += len;
3182            }            }
3183    
3184            if (possessive) continue;    /* No backtracking */
3185    
3186          for(;;)          for(;;)
3187            {            {
3188            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3189            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3190            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3191            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3192              if (utf) BACKCHAR(eptr);
3193    #endif
3194            }            }
3195          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3196          }          }
# Line 2320  for (;;) Line 3202  for (;;)
3202      /* Match a single character, casefully */      /* Match a single character, casefully */
3203    
3204      case OP_CHAR:      case OP_CHAR:
3205  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3206      if (utf8)      if (utf)
3207        {        {
3208        length = 1;        length = 1;
3209        ecode++;        ecode++;
# Line 2331  for (;;) Line 3213  for (;;)
3213          CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */          CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3214          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3215          }          }
3216        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3217        }        }
3218      else      else
3219  #endif  #endif
3220        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3221        {        {
3222        if (md->end_subject - eptr < 1)        if (md->end_subject - eptr < 1)
3223          {          {
# Line 2348  for (;;) Line 3229  for (;;)
3229        }        }
3230      break;      break;
3231    
3232      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3233        subject, give up immediately. */
3234    
3235        case OP_CHARI:
3236        if (eptr >= md->end_subject)
3237          {
3238          SCHECK_PARTIAL();
3239          RRETURN(MATCH_NOMATCH);
3240          }
3241    
3242      case OP_CHARNC:  #ifdef SUPPORT_UTF
3243  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3244        {        {
3245        length = 1;        length = 1;
3246        ecode++;        ecode++;
3247        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3248    
       if (length > md->end_subject - eptr)  
         {  
         CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */  
         RRETURN(MATCH_NOMATCH);  
         }  
   
3249        /* 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
3250        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3251          fast lookup table. We know that there is at least one byte left in the
3252          subject. */
3253    
3254        if (fc < 128)        if (fc < 128)
3255          {          {
3256          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3257            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3258            ecode++;
3259            eptr++;
3260          }          }
3261    
3262        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3263          use the value of "length" to check for sufficient bytes left, because the
3264          other case of the character may have more or fewer bytes.  */
3265    
3266        else        else
3267          {          {
3268          unsigned int dc;          pcre_uint32 dc;
3269          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3270          ecode += length;          ecode += length;
3271    
# Line 2393  for (;;) Line 3282  for (;;)
3282          }          }
3283        }        }
3284      else      else
3285  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3286    
3287      /* Non-UTF-8 mode */      /* Not UTF mode */
3288        {        {
3289        if (md->end_subject - eptr < 1)        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3290          {            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3291          SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */        eptr++;
         RRETURN(MATCH_NOMATCH);  
         }  
       if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);  
3292        ecode += 2;        ecode += 2;
3293        }        }
3294      break;      break;
# Line 2410  for (;;) Line 3296  for (;;)
3296      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3297    
3298      case OP_EXACT:      case OP_EXACT:
3299        case OP_EXACTI:
3300      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3301      ecode += 3;      ecode += 1 + IMM2_SIZE;
3302      goto REPEATCHAR;      goto REPEATCHAR;
3303    
3304      case OP_POSUPTO:      case OP_POSUPTO:
3305        case OP_POSUPTOI:
3306      possessive = TRUE;      possessive = TRUE;
3307      /* Fall through */      /* Fall through */
3308    
3309      case OP_UPTO:      case OP_UPTO:
3310        case OP_UPTOI:
3311      case OP_MINUPTO:      case OP_MINUPTO:
3312        case OP_MINUPTOI:
3313      min = 0;      min = 0;
3314      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3315      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3316      ecode += 3;      ecode += 1 + IMM2_SIZE;
3317      goto REPEATCHAR;      goto REPEATCHAR;
3318    
3319      case OP_POSSTAR:      case OP_POSSTAR:
3320        case OP_POSSTARI:
3321      possessive = TRUE;      possessive = TRUE;
3322      min = 0;      min = 0;
3323      max = INT_MAX;      max = INT_MAX;
# Line 2434  for (;;) Line 3325  for (;;)
3325      goto REPEATCHAR;      goto REPEATCHAR;
3326    
3327      case OP_POSPLUS:      case OP_POSPLUS:
3328        case OP_POSPLUSI:
3329      possessive = TRUE;      possessive = TRUE;
3330      min = 1;      min = 1;
3331      max = INT_MAX;      max = INT_MAX;
# Line 2441  for (;;) Line 3333  for (;;)
3333      goto REPEATCHAR;      goto REPEATCHAR;
3334    
3335      case OP_POSQUERY:      case OP_POSQUERY:
3336        case OP_POSQUERYI:
3337      possessive = TRUE;      possessive = TRUE;
3338      min = 0;      min = 0;
3339      max = 1;      max = 1;
# Line 2448  for (;;) Line 3341  for (;;)
3341      goto REPEATCHAR;      goto REPEATCHAR;
3342    
3343      case OP_STAR:      case OP_STAR:
3344        case OP_STARI:
3345      case OP_MINSTAR:      case OP_MINSTAR:
3346        case OP_MINSTARI:
3347      case OP_PLUS:      case OP_PLUS:
3348        case OP_PLUSI:
3349      case OP_MINPLUS:      case OP_MINPLUS:
3350        case OP_MINPLUSI:
3351      case OP_QUERY:      case OP_QUERY:
3352        case OP_QUERYI:
3353      case OP_MINQUERY:      case OP_MINQUERY:
3354      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3355        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3356      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
   
3357      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3358      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3359      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3360    
3361      /* Common code for all repeated single-character matches. */      /* Common code for all repeated single-character matches. We first check
3362        for the minimum number of characters. If the minimum equals the maximum, we
3363        are done. Otherwise, if minimizing, check the rest of the pattern for a
3364        match; if there isn't one, advance up to the maximum, one character at a
3365        time.
3366    
3367        If maximizing, advance up to the maximum number of matching characters,
3368        until eptr is past the end of the maximum run. If possessive, we are
3369        then done (no backing up). Otherwise, match at this position; anything
3370        other than no match is immediately returned. For nomatch, back up one
3371        character, unless we are matching \R and the last thing matched was
3372        \r\n, in which case, back up two bytes. When we reach the first optional
3373        character position, we can save stack by doing a tail recurse.
3374    
3375        The various UTF/non-UTF and caseful/caseless cases are handled separately,
3376        for speed. */
3377    
3378      REPEATCHAR:      REPEATCHAR:
3379  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3380      if (utf8)      if (utf)
3381        {        {
3382        length = 1;        length = 1;
3383        charptr = ecode;        charptr = ecode;
# Line 2477  for (;;) Line 3390  for (;;)
3390        if (length > 1)        if (length > 1)
3391          {          {
3392  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3393          unsigned int othercase;          pcre_uint32 othercase;
3394          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3395              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3396            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3397          else oclength = 0;          else oclength = 0;
3398  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3399    
3400          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3401            {            {
3402            if (eptr <= md->end_subject - length &&            if (eptr <= md->end_subject - length &&
3403              memcmp(eptr, charptr, length) == 0) eptr += length;              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3404  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3405            else if (oclength > 0 &&            else if (oclength > 0 &&
3406                     eptr <= md->end_subject - oclength &&                     eptr <= md->end_subject - oclength &&
3407                     memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                     memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3408  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3409            else            else
3410              {              {
# Line 2506  for (;;) Line 3419  for (;;)
3419            {            {
3420            for (fi = min;; fi++)            for (fi = min;; fi++)
3421              {              {
3422              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3423              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3424              if (fi >= max) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3425              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
3426                memcmp(eptr, charptr, length) == 0) eptr += length;                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3427  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3428              else if (oclength > 0 &&              else if (oclength > 0 &&
3429                       eptr <= md->end_subject - oclength &&                       eptr <= md->end_subject - oclength &&
3430                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3431  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3432              else              else
3433                {                {
# Line 2531  for (;;) Line 3444  for (;;)
3444            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3445              {              {
3446              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
3447                  memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3448  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3449              else if (oclength > 0 &&              else if (oclength > 0 &&
3450                       eptr <= md->end_subject - oclength &&                       eptr <= md->end_subject - oclength &&
3451                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3452  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3453              else break;              else
3454                  {
3455                  CHECK_PARTIAL();
3456                  break;
3457                  }
3458              }              }
3459    
3460            if (possessive) continue;            if (possessive) continue;    /* No backtracking */
   
3461            for(;;)            for(;;)
3462              {              {
3463              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              if (eptr == pp) goto TAIL_RECURSE;
3464                RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3465              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
             if (eptr == pp) { RRETURN(MATCH_NOMATCH); }  
3466  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3467              eptr--;              eptr--;
3468              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2563  for (;;) Line 3479  for (;;)
3479        value of fc will always be < 128. */        value of fc will always be < 128. */
3480        }        }
3481      else      else
3482  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3483          /* When not in UTF-8 mode, load a single-byte character. */
3484          fc = *ecode++;
3485    
3486      /* When not in UTF-8 mode, load a single-byte character. */      /* The value of fc at this point is always one character, though we may
3487        or may not be in UTF mode. The code is duplicated for the caseless and
     fc = *ecode++;  
   
     /* The value of fc at this point is always less than 256, though we may or  
     may not be in UTF-8 mode. The code is duplicated for the caseless and  
3488      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3489      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3490      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2579  for (;;) Line 3493  for (;;)
3493      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3494    
3495      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3496        max, eptr));        max, (char *)eptr));
3497    
3498      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3499        {        {
3500        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3501          /* fc must be < 128 if UTF is enabled. */
3502          foc = md->fcc[fc];
3503    #else
3504    #ifdef SUPPORT_UTF
3505    #ifdef SUPPORT_UCP
3506          if (utf && fc > 127)
3507            foc = UCD_OTHERCASE(fc);
3508    #else
3509          if (utf && fc > 127)
3510            foc = fc;
3511    #endif /* SUPPORT_UCP */
3512          else
3513    #endif /* SUPPORT_UTF */
3514            foc = TABLE_GET(fc, md->fcc, fc);
3515    #endif /* COMPILE_PCRE8 */
3516    
3517        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3518          {          {
3519            pcre_uint32 cc;                 /* Faster than pcre_uchar */
3520          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
3521            {            {
3522            SCHECK_PARTIAL();            SCHECK_PARTIAL();
3523            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3524            }            }
3525          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          cc = RAWUCHARTEST(eptr);
3526            if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);
3527            eptr++;
3528          }          }
3529        if (min == max) continue;        if (min == max) continue;
3530        if (minimize)        if (minimize)
3531          {          {
3532          for (fi = min;; fi++)          for (fi = min;; fi++)
3533            {            {
3534            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            pcre_uint32 cc;               /* Faster than pcre_uchar */
3535              RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3536            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3537            if (fi >= max) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3538            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
# Line 2606  for (;;) Line 3540  for (;;)
3540              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3541              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3542              }              }
3543            if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            cc = RAWUCHARTEST(eptr);
3544              if (fc != cc && foc != cc) RRETURN(MATCH_NOMATCH);