/[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 529 by ph10, Mon May 31 17:28:08 2010 UTC revision 1046 by ph10, Tue Sep 25 16:27:58 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
# 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 */  /* Values for setting in md->match_function_type to indicate two special types
60    of call to match(). We do it this way to save on using another stack variable,
61    as stack usage is to be discouraged. */
62    
63  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
64  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
65    
66  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
67  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 73  negative to avoid the external error cod Line 74  negative to avoid the external error cod
74    
75  #define MATCH_ACCEPT       (-999)  #define MATCH_ACCEPT       (-999)
76  #define MATCH_COMMIT       (-998)  #define MATCH_COMMIT       (-998)
77  #define MATCH_PRUNE        (-997)  #define MATCH_KETRPOS      (-997)
78  #define MATCH_SKIP         (-996)  #define MATCH_ONCE         (-996)
79  #define MATCH_SKIP_ARG     (-995)  #define MATCH_PRUNE        (-995)
80  #define MATCH_THEN         (-994)  #define MATCH_SKIP         (-994)
81    #define MATCH_SKIP_ARG     (-993)
82  /* This is a convenience macro for code that occurs many times. */  #define MATCH_THEN         (-992)
   
 #define MRRETURN(ra) \  
   { \  
   md->mark = markptr; \  
   RRETURN(ra); \  
   }  
83    
84  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
85  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
# Line 117  Returns:     nothing Line 112  Returns:     nothing
112  */  */
113    
114  static void  static void
115  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
116  {  {
117  unsigned int c;  unsigned int c;
118  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
# Line 132  while (length-- > 0) Line 127  while (length-- > 0)
127  *          Match a back-reference                *  *          Match a back-reference                *
128  *************************************************/  *************************************************/
129    
130  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
131  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
132    the length passed is zero. Note that in caseless UTF-8 mode, the number of
133    subject bytes matched may be different to the number of reference bytes.
134    
135  Arguments:  Arguments:
136    offset      index into the offset vector    offset      index into the offset vector
137    eptr        points into the subject    eptr        pointer into the subject
138    length      length to be matched    length      length of reference to be matched (number of bytes)
139    md          points to match data block    md          points to match data block
140    ims         the ims flags    caseless    TRUE if caseless
141    
142  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
143                  -1 no match
144                  -2 partial match; always given if at end subject
145  */  */
146    
147  static BOOL  static int
148  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
149    unsigned long int ims)    BOOL caseless)
150  {  {
151  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
152    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
153    
154  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
155  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
# Line 164  pchars(p, length, FALSE, md); Line 164  pchars(p, length, FALSE, md);
164  printf("\n");  printf("\n");
165  #endif  #endif
166    
167  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
168    case the length is passed as zero). */
169    
170  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
171    
172  /* Separate the 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
173  properly if Unicode properties are supported. Otherwise, we can check only  properly if Unicode properties are supported. Otherwise, we can check only
174  ASCII characters. */  ASCII characters. */
175    
176  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
177    {    {
178  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
179  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
180    if (md->utf8)    if (md->utf)
181      {      {
182      USPTR endptr = eptr + length;      /* Match characters up to the end of the reference. NOTE: the number of
183      while (eptr < endptr)      data units matched may differ, because in UTF-8 there are some characters
184        {      whose upper and lower case versions code have different numbers of bytes.
185        int c, d;      For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
186        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
187        sequence of two of the latter. It is important, therefore, to check the
188        length along the reference, not along the subject (earlier code did this
189        wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193          {
194          unsigned int c, d;
195          const ucd_record *ur;
196          if (eptr >= md->end_subject) return -2;   /* Partial match */
197        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
198        GETCHARINC(d, p);        GETCHARINC(d, p);
199        if (c != d && c != UCD_OTHERCASE(d)) return FALSE;        ur = GET_UCD(d);
200          if (c != d && c != d + ur->other_case)
201            {
202            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
203            for (;;)
204              {
205              if (c < *pp) return -1;
206              if (c == *pp++) break;
207              }
208            }
209        }        }
210      }      }
211    else    else
# Line 193  if ((ims & PCRE_CASELESS) != 0) Line 214  if ((ims & PCRE_CASELESS) != 0)
214    
215    /* 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
216    is no UCP support. */    is no UCP support. */
217        {
218    while (length-- > 0)      while (length-- > 0)
219      { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }        {
220          if (eptr >= md->end_subject) return -2;   /* Partial match */
221          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
222          p++;
223          eptr++;
224          }
225        }
226    }    }
227    
228  /* In the caseful case, we can just compare the bytes, whether or not we  /* In the caseful case, we can just compare the bytes, whether or not we
229  are in UTF-8 mode. */  are in UTF-8 mode. */
230    
231  else  else
232    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
233      while (length-- > 0)
234        {
235        if (eptr >= md->end_subject) return -2;   /* Partial match */
236        if (*p++ != *eptr++) return -1;
237        }
238      }
239    
240  return TRUE;  return (int)(eptr - eptr_start);
241  }  }
242    
243    
# Line 256  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 289  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
289         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
290         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
291         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
292         RM61,  RM62 };         RM61,  RM62, RM63, RM64, RM65, RM66 };
293    
294  /* 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
295  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
# Line 266  actually used in this definition. */ Line 299  actually used in this definition. */
299  #define REGISTER register  #define REGISTER register
300    
301  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
302  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
303    { \    { \
304    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
305    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
306    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
307    }    }
308  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 278  actually used in this definition. */ Line 311  actually used in this definition. */
311    return ra; \    return ra; \
312    }    }
313  #else  #else
314  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
315    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
316  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
317  #endif  #endif
318    
# Line 292  argument of match(), which never changes Line 325  argument of match(), which never changes
325    
326  #define REGISTER  #define REGISTER
327    
328  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
329    {\    {\
330    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
331    frame->Xwhere = rw; \    if (newframe == NULL)\
332        {\
333        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
334        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
335        newframe->Xnextframe = NULL;\
336        frame->Xnextframe = newframe;\
337        }\
338      frame->Xwhere = rw;\
339    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
340    newframe->Xecode = rb;\    newframe->Xecode = rb;\
341    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
   newframe->Xmarkptr = markptr;\  
342    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
343    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
344    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
345    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
346    frame = newframe;\    frame = newframe;\
# Line 317  argument of match(), which never changes Line 354  argument of match(), which never changes
354    {\    {\
355    heapframe *oldframe = frame;\    heapframe *oldframe = frame;\
356    frame = oldframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(oldframe);\  
357    if (frame != NULL)\    if (frame != NULL)\
358      {\      {\
359      rrc = ra;\      rrc = ra;\
# Line 331  argument of match(), which never changes Line 367  argument of match(), which never changes
367    
368  typedef struct heapframe {  typedef struct heapframe {
369    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
370      struct heapframe *Xnextframe;
371    
372    /* Function arguments that may change */    /* Function arguments that may change */
373    
374    USPTR Xeptr;    PCRE_PUCHAR Xeptr;
375    const uschar *Xecode;    const pcre_uchar *Xecode;
376    USPTR Xmstart;    PCRE_PUCHAR Xmstart;
   USPTR Xmarkptr;  
377    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
378    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
379    unsigned int Xrdepth;    unsigned int Xrdepth;
380    
381    /* Function local variables */    /* Function local variables */
382    
383    USPTR Xcallpat;    PCRE_PUCHAR Xcallpat;
384  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
385    USPTR Xcharptr;    PCRE_PUCHAR Xcharptr;
386  #endif  #endif
387    USPTR Xdata;    PCRE_PUCHAR Xdata;
388    USPTR Xnext;    PCRE_PUCHAR Xnext;
389    USPTR Xpp;    PCRE_PUCHAR Xpp;
390    USPTR Xprev;    PCRE_PUCHAR Xprev;
391    USPTR Xsaved_eptr;    PCRE_PUCHAR Xsaved_eptr;
392    
393    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
394    
# Line 362  typedef struct heapframe { Line 396  typedef struct heapframe {
396    BOOL Xcondition;    BOOL Xcondition;
397    BOOL Xprev_is_word;    BOOL Xprev_is_word;
398    
   unsigned long int Xoriginal_ims;  
   
399  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
400    int Xprop_type;    int Xprop_type;
401    int Xprop_value;    int Xprop_value;
402    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
403    int Xoclength;    int Xoclength;
404    uschar Xocchars[8];    pcre_uchar Xocchars[6];
405  #endif  #endif
406    
407    int Xcodelink;    int Xcodelink;
# Line 414  returns a negative (error) response, the Line 443  returns a negative (error) response, the
443  same response. */  same response. */
444    
445  /* 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
446  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
447  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.
448  something has been matched). For hard partial matching, we then return  something has been matched). For hard partial matching, we then return
449  immediately. The second one is used when we already know we are past the end of  immediately. The second one is used when we already know we are past the end of
450  the subject. */  the subject. */
451    
452  #define CHECK_PARTIAL()\  #define CHECK_PARTIAL()\
453    if (md->partial != 0 && eptr >= md->end_subject && eptr > mstart)\    if (md->partial != 0 && eptr >= md->end_subject && \
454      {\        eptr > md->start_used_ptr) \
455      md->hitend = TRUE;\      { \
456      if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL);\      md->hitend = TRUE; \
457        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
458      }      }
459    
460  #define SCHECK_PARTIAL()\  #define SCHECK_PARTIAL()\
461    if (md->partial != 0 && eptr > mstart)\    if (md->partial != 0 && eptr > md->start_used_ptr) \
462      {\      { \
463      md->hitend = TRUE;\      md->hitend = TRUE; \
464      if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL);\      if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
465      }      }
466    
467    
468  /* Performance note: It might be tempting to extract commonly used fields from  /* Performance note: It might be tempting to extract commonly used fields from
469  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
470  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
471  made performance worse.  made performance worse.
472    
# Line 445  Arguments: Line 475  Arguments:
475     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
476     mstart      pointer to the current match start position (can be modified     mstart      pointer to the current match start position (can be modified
477                   by encountering \K)                   by encountering \K)
    markptr     pointer to the most recent MARK name, or NULL  
478     offset_top  current top pointer     offset_top  current top pointer
479     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
480     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
481                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    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  
482     rdepth      the recursion depth     rdepth      the recursion depth
483    
484  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
# Line 465  Returns:       MATCH_MATCH if matched Line 489  Returns:       MATCH_MATCH if matched
489  */  */
490    
491  static int  static int
492  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
493    const uschar *markptr, int offset_top, match_data *md, unsigned long int ims,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
494    eptrblock *eptrb, int flags, unsigned int rdepth)    unsigned int rdepth)
495  {  {
496  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
497  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 476  so they can be ordinary variables in all Line 500  so they can be ordinary variables in all
500  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
501  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
502  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
503  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
504    
505  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
506    BOOL caseless;
507  int condcode;  int condcode;
508    
509  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
510  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
511  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
512  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
513    the top-level on the stack rather than malloc-ing them all gives a performance
514    boost in many cases where there is not much "recursion". */
515    
516  #ifdef NO_RECURSE  #ifdef NO_RECURSE
517  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
518    
519  /* Copy in the original argument variables */  /* Copy in the original argument variables */
520    
521  frame->Xeptr = eptr;  frame->Xeptr = eptr;
522  frame->Xecode = ecode;  frame->Xecode = ecode;
523  frame->Xmstart = mstart;  frame->Xmstart = mstart;
 frame->Xmarkptr = markptr;  
524  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
525  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
526  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
527    
528  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 511  HEAP_RECURSE: Line 534  HEAP_RECURSE:
534  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
535  #define ecode              frame->Xecode  #define ecode              frame->Xecode
536  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
 #define markptr            frame->Xmarkptr  
537  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
538  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
539  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
540    
541  /* Ditto for the local variables */  /* Ditto for the local variables */
542    
543  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
544  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
545  #endif  #endif
546  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
# Line 537  HEAP_RECURSE: Line 557  HEAP_RECURSE:
557  #define condition          frame->Xcondition  #define condition          frame->Xcondition
558  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
559    
 #define original_ims       frame->Xoriginal_ims  
   
560  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
561  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
562  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
563  #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  
564  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
565  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
566  #endif  #endif
# Line 575  i, and fc and c, can be the same variabl Line 590  i, and fc and c, can be the same variabl
590  #define fi i  #define fi i
591  #define fc c  #define fc c
592    
593    /* Many of the following variables are used only in small blocks of the code.
594    My normal style of coding would have declared them within each of those blocks.
595    However, in order to accommodate the version of this code that uses an external
596    "stack" implemented on the heap, it is easier to declare them all here, so the
597    declarations can be cut out in a block. The only declarations within blocks
598    below are for variables that do not have to be preserved over a recursive call
599    to RMATCH(). */
600    
601    #ifdef SUPPORT_UTF
602    const pcre_uchar *charptr;
603    #endif
604    const pcre_uchar *callpat;
605    const pcre_uchar *data;
606    const pcre_uchar *next;
607    PCRE_PUCHAR       pp;
608    const pcre_uchar *prev;
609    PCRE_PUCHAR       saved_eptr;
610    
611    recursion_info new_recursive;
612    
613  #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().          */  
614  BOOL condition;  BOOL condition;
615  BOOL prev_is_word;  BOOL prev_is_word;
616    
 unsigned long int original_ims;  
   
617  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
618  int prop_type;  int prop_type;
619  int prop_value;  int prop_value;
620  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
621  int oclength;  int oclength;
622  uschar occhars[8];  pcre_uchar occhars[6];
623  #endif  #endif
624    
625  int codelink;  int codelink;
# Line 618  int save_offset1, save_offset2, save_off Line 635  int save_offset1, save_offset2, save_off
635  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
636    
637  eptrblock newptrb;  eptrblock newptrb;
638    
639    /* There is a special fudge for calling match() in a way that causes it to
640    measure the size of its basic stack frame when the stack is being used for
641    recursion. The second argument (ecode) being NULL triggers this behaviour. It
642    cannot normally ever be NULL. The return is the negated value of the frame
643    size. */
644    
645    if (ecode == NULL)
646      {
647      if (rdepth == 0)
648        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
649      else
650        {
651        int len = (char *)&rdepth - (char *)eptr;
652        return (len > 0)? -len : len;
653        }
654      }
655  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
656    
657    /* To save space on the stack and in the heap frame, I have doubled up on some
658    of the local variables that are used only in localised parts of the code, but
659    still need to be preserved over recursive calls of match(). These macros define
660    the alternative names that are used. */
661    
662    #define allow_zero    cur_is_word
663    #define cbegroup      condition
664    #define code_offset   codelink
665    #define condassert    condition
666    #define matched_once  prev_is_word
667    #define foc           number
668    #define save_mark     data
669    
670  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
671  variables. */  variables. */
672    
# Line 644  defined). However, RMATCH isn't like a f Line 691  defined). However, RMATCH isn't like a f
691  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,
692  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
693    
694  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
695  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
696  #else  #else
697  utf8 = FALSE;  utf = FALSE;
698  #endif  #endif
699    
700  /* 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 656  haven't exceeded the recursive call limi Line 703  haven't exceeded the recursive call limi
703  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
704  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
705    
 original_ims = ims;    /* Save for resetting on ')' */  
   
706  /* 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
707  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
708  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
709  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
710  When match() is called in other circumstances, don't add to the chain. The  
711  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
712  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
713  match(). */  to break infinite loops that match no characters. When match() is called in
714    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
715    NOT be used with tail recursion, because the memory block that is used is on
716    the stack, so a new one may be required for each match(). */
717    
718  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
719    {    {
720    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
721    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
722    eptrb = &newptrb;    eptrb = &newptrb;
723      md->match_function_type = 0;
724    }    }
725    
726  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 684  for (;;) Line 733  for (;;)
733    switch(op)    switch(op)
734      {      {
735      case OP_MARK:      case OP_MARK:
736      markptr = ecode + 2;      md->nomatch_mark = ecode + 2;
737      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      md->mark = NULL;    /* In case previously set by assertion */
738        ims, eptrb, flags, RM55);      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
739          eptrb, RM55);
740        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
741             md->mark == NULL) md->mark = ecode + 2;
742    
743      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
744      argument, and we must check whether that argument matches this MARK's      argument, and we must check whether that argument matches this MARK's
# Line 695  for (;;) Line 747  for (;;)
747      position and return MATCH_SKIP. Otherwise, pass back the return code      position and return MATCH_SKIP. Otherwise, pass back the return code
748      unaltered. */      unaltered. */
749    
750      if (rrc == MATCH_SKIP_ARG &&      else if (rrc == MATCH_SKIP_ARG &&
751          strcmp((char *)markptr, (char *)(md->start_match_ptr)) == 0)          STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
752        {        {
753        md->start_match_ptr = eptr;        md->start_match_ptr = eptr;
754        RRETURN(MATCH_SKIP);        RRETURN(MATCH_SKIP);
755        }        }
   
     if (md->mark == NULL) md->mark = markptr;  
756      RRETURN(rrc);      RRETURN(rrc);
757    
758      case OP_FAIL:      case OP_FAIL:
759      MRRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
760    
761        /* COMMIT overrides PRUNE, SKIP, and THEN */
762    
763      case OP_COMMIT:      case OP_COMMIT:
764      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
765        ims, eptrb, flags, RM52);        eptrb, RM52);
766      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
767      MRRETURN(MATCH_COMMIT);          rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
768            rrc != MATCH_THEN)
769          RRETURN(rrc);
770        RRETURN(MATCH_COMMIT);
771    
772        /* PRUNE overrides THEN */
773    
774      case OP_PRUNE:      case OP_PRUNE:
775      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
776        ims, eptrb, flags, RM51);        eptrb, RM51);
777      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
778      MRRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
779    
780      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
781      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      md->nomatch_mark = ecode + 2;
782        ims, eptrb, flags, RM56);      md->mark = NULL;    /* In case previously set by assertion */
783      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
784      md->mark = ecode + 2;        eptrb, RM56);
785        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
786             md->mark == NULL) md->mark = ecode + 2;
787        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
788      RRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
789    
790        /* SKIP overrides PRUNE and THEN */
791    
792      case OP_SKIP:      case OP_SKIP:
793      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
794        ims, eptrb, flags, RM53);        eptrb, RM53);
795      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
796          RRETURN(rrc);
797      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
798      MRRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
799    
800        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
801        nomatch_mark. There is a flag that disables this opcode when re-matching a
802        pattern that ended with a SKIP for which there was not a matching MARK. */
803    
804      case OP_SKIP_ARG:      case OP_SKIP_ARG:
805      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      if (md->ignore_skip_arg)
806        ims, eptrb, flags, RM57);        {
807      if (rrc != MATCH_NOMATCH) RRETURN(rrc);        ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
808          break;
809          }
810        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
811          eptrb, RM57);
812        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
813          RRETURN(rrc);
814    
815      /* Pass back the current skip name by overloading md->start_match_ptr and      /* Pass back the current skip name by overloading md->start_match_ptr and
816      returning the special MATCH_SKIP_ARG return code. This will either be      returning the special MATCH_SKIP_ARG return code. This will either be
817      caught by a matching MARK, or get to the top, where it is treated the same      caught by a matching MARK, or get to the top, where it causes a rematch
818      as PRUNE. */      with the md->ignore_skip_arg flag set. */
819    
820      md->start_match_ptr = ecode + 2;      md->start_match_ptr = ecode + 2;
821      RRETURN(MATCH_SKIP_ARG);      RRETURN(MATCH_SKIP_ARG);
822    
823        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
824        the branch in which it occurs can be determined. Overload the start of
825        match pointer to do this. */
826    
827      case OP_THEN:      case OP_THEN:
828      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
829        ims, eptrb, flags, RM54);        eptrb, RM54);
830      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831      MRRETURN(MATCH_THEN);      md->start_match_ptr = ecode;
832        RRETURN(MATCH_THEN);
833    
834      case OP_THEN_ARG:      case OP_THEN_ARG:
835      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode] + ecode[1], offset_top, md,      md->nomatch_mark = ecode + 2;
836        ims, eptrb, flags, RM58);      md->mark = NULL;    /* In case previously set by assertion */
837        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
838          md, eptrb, RM58);
839        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
840             md->mark == NULL) md->mark = ecode + 2;
841      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
842      md->mark = ecode + 2;      md->start_match_ptr = ecode;
843      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
844    
845      /* Handle a capturing bracket. If there is space in the offset vector, save      /* Handle an atomic group that does not contain any capturing parentheses.
846      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
847      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
848      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.
849      reference inside the group.      However, this uses a lot more stack, so in 8.20, atomic groups that do not
850        contain any captures generate OP_ONCE_NC, which can be handled in the old,
851      If the bracket fails to match, we need to restore this value and also the      less stack intensive way.
852      values of the final offsets, in case they were set by a previous iteration  
853      of the same bracket.      Check the alternative branches in turn - the matching won't pass the KET
854        for this kind of subpattern. If any one branch matches, we carry on as at
855        the end of a normal bracket, leaving the subject pointer, but resetting
856        the start-of-match value in case it was changed by \K. */
857    
858        case OP_ONCE_NC:
859        prev = ecode;
860        saved_eptr = eptr;
861        save_mark = md->mark;
862        do
863          {
864          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
865          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
866            {
867            mstart = md->start_match_ptr;
868            break;
869            }
870          if (rrc == MATCH_THEN)
871            {
872            next = ecode + GET(ecode,1);
873            if (md->start_match_ptr < next &&
874                (*ecode == OP_ALT || *next == OP_ALT))
875              rrc = MATCH_NOMATCH;
876            }
877    
878          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
879          ecode += GET(ecode,1);
880          md->mark = save_mark;
881          }
882        while (*ecode == OP_ALT);
883    
884        /* If hit the end of the group (which could be repeated), fail */
885    
886        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
887    
888        /* Continue as from after the group, updating the offsets high water
889        mark, since extracts may have been taken. */
890    
891        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
892    
893        offset_top = md->end_offset_top;
894        eptr = md->end_match_ptr;
895    
896        /* For a non-repeating ket, just continue at this level. This also
897        happens for a repeating ket if no characters were matched in the group.
898        This is the forcible breaking of infinite loops as implemented in Perl
899        5.005. */
900    
901        if (*ecode == OP_KET || eptr == saved_eptr)
902          {
903          ecode += 1+LINK_SIZE;
904          break;
905          }
906    
907        /* The repeating kets try the rest of the pattern or restart from the
908        preceding bracket, in the appropriate order. The second "call" of match()
909        uses tail recursion, to avoid using another stack frame. */
910    
911        if (*ecode == OP_KETRMIN)
912          {
913          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
914          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
915          ecode = prev;
916          goto TAIL_RECURSE;
917          }
918        else  /* OP_KETRMAX */
919          {
920          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
921          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
922          ecode += 1 + LINK_SIZE;
923          goto TAIL_RECURSE;
924          }
925        /* Control never gets here */
926    
927        /* Handle a capturing bracket, other than those that are possessive with an
928        unlimited repeat. If there is space in the offset vector, save the current
929        subject position in the working slot at the top of the vector. We mustn't
930        change the current values of the data slot, because they may be set from a
931        previous iteration of this group, and be referred to by a reference inside
932        the group. A failure to match might occur after the group has succeeded,
933        if something later on doesn't match. For this reason, we need to restore
934        the working value and also the values of the final offsets, in case they
935        were set by a previous iteration of the same bracket.
936    
937      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
938      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 792  for (;;) Line 956  for (;;)
956        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
957        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
958        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
959          save_mark = md->mark;
960    
961        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
962        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
963            (int)(eptr - md->start_subject);
964    
965        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
966          {          {
967          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
968            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
969          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
970            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
971    
972            /* If we backed up to a THEN, check whether it is within the current
973            branch by comparing the address of the THEN that is passed back with
974            the end of the branch. If it is within the current branch, and the
975            branch is one of two or more alternatives (it either starts or ends
976            with OP_ALT), we have reached the limit of THEN's action, so convert
977            the return code to NOMATCH, which will cause normal backtracking to
978            happen from now on. Otherwise, THEN is passed back to an outer
979            alternative. This implements Perl's treatment of parenthesized groups,
980            where a group not containing | does not affect the current alternative,
981            that is, (X) is NOT the same as (X|(*F)). */
982    
983            if (rrc == MATCH_THEN)
984              {
985              next = ecode + GET(ecode,1);
986              if (md->start_match_ptr < next &&
987                  (*ecode == OP_ALT || *next == OP_ALT))
988                rrc = MATCH_NOMATCH;
989              }
990    
991            /* Anything other than NOMATCH is passed back. */
992    
993            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
994          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
995          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
996            md->mark = save_mark;
997            if (*ecode != OP_ALT) break;
998          }          }
       while (*ecode == OP_ALT);  
999    
1000        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1001        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1002        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1003        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1004    
1005        if (rrc != MATCH_THEN) md->mark = markptr;        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1006        RRETURN(MATCH_NOMATCH);  
1007          RRETURN(rrc);
1008        }        }
1009    
1010      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 828  for (;;) Line 1018  for (;;)
1018      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1019      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1020    
1021      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1022      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
1023      recursive call to match() whatever happened. We can reduce stack usage by  
1024      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
1025      is set.*/      the result of a recursive call to match() whatever happened so it was
1026        possible to reduce stack usage by turning this into a tail recursion,
1027        except in the case of a possibly empty group. However, now that there is
1028        the possiblity of (*THEN) occurring in the final alternative, this
1029        optimization is no longer always possible.
1030    
1031        We can optimize if we know there are no (*THEN)s in the pattern; at present
1032        this is the best that can be done.
1033    
1034        MATCH_ONCE is returned when the end of an atomic group is successfully
1035        reached, but subsequent matching fails. It passes back up the tree (causing
1036        captured values to be reset) until the original atomic group level is
1037        reached. This is tested by comparing md->once_target with the start of the
1038        group. At this point, the return is converted into MATCH_NOMATCH so that
1039        previous backup points can be taken. */
1040    
1041        case OP_ONCE:
1042      case OP_BRA:      case OP_BRA:
1043      case OP_SBRA:      case OP_SBRA:
1044      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1045      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1046      for (;;)      for (;;)
1047        {        {
1048        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE)
1049            md->match_function_type = MATCH_CBEGROUP;
1050    
1051          /* If this is not a possibly empty group, and there are no (*THEN)s in
1052          the pattern, and this is the final alternative, optimize as described
1053          above. */
1054    
1055          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1056            {
1057            ecode += PRIV(OP_lengths)[*ecode];
1058            goto TAIL_RECURSE;
1059            }
1060    
1061          /* In all other cases, we have to make another call to match(). */
1062    
1063          save_mark = md->mark;
1064          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1065            RM2);
1066    
1067          /* See comment in the code for capturing groups above about handling
1068          THEN. */
1069    
1070          if (rrc == MATCH_THEN)
1071          {          {
1072          if (flags == 0)    /* Not a possibly empty group */          next = ecode + GET(ecode,1);
1073            if (md->start_match_ptr < next &&
1074                (*ecode == OP_ALT || *next == OP_ALT))
1075              rrc = MATCH_NOMATCH;
1076            }
1077    
1078          if (rrc != MATCH_NOMATCH)
1079            {
1080            if (rrc == MATCH_ONCE)
1081            {            {
1082            ecode += _pcre_OP_lengths[*ecode];            const pcre_uchar *scode = ecode;
1083            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
1084            goto TAIL_RECURSE;              {
1085                while (*scode == OP_ALT) scode += GET(scode, 1);
1086                scode -= GET(scode, 1);
1087                }
1088              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1089            }            }
1090            RRETURN(rrc);
1091            }
1092          ecode += GET(ecode, 1);
1093          md->mark = save_mark;
1094          if (*ecode != OP_ALT) break;
1095          }
1096    
1097          /* Possibly empty group; can't use tail recursion. */      RRETURN(MATCH_NOMATCH);
1098    
1099          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      /* Handle possessive capturing brackets with an unlimited repeat. We come
1100            eptrb, flags, RM48);      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1101          if (rrc == MATCH_NOMATCH) md->mark = markptr;      handled similarly to the normal case above. However, the matching is
1102          RRETURN(rrc);      different. The end of these brackets will always be OP_KETRPOS, which
1103        returns MATCH_KETRPOS without going further in the pattern. By this means
1104        we can handle the group by iteration rather than recursion, thereby
1105        reducing the amount of stack needed. */
1106    
1107        case OP_CBRAPOS:
1108        case OP_SCBRAPOS:
1109        allow_zero = FALSE;
1110    
1111        POSSESSIVE_CAPTURE:
1112        number = GET2(ecode, 1+LINK_SIZE);
1113        offset = number << 1;
1114    
1115    #ifdef PCRE_DEBUG
1116        printf("start possessive bracket %d\n", number);
1117        printf("subject=");
1118        pchars(eptr, 16, TRUE, md);
1119        printf("\n");
1120    #endif
1121    
1122        if (offset < md->offset_max)
1123          {
1124          matched_once = FALSE;
1125          code_offset = (int)(ecode - md->start_code);
1126    
1127          save_offset1 = md->offset_vector[offset];
1128          save_offset2 = md->offset_vector[offset+1];
1129          save_offset3 = md->offset_vector[md->offset_end - number];
1130          save_capture_last = md->capture_last;
1131    
1132          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1133    
1134          /* Each time round the loop, save the current subject position for use
1135          when the group matches. For MATCH_MATCH, the group has matched, so we
1136          restart it with a new subject starting position, remembering that we had
1137          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1138          usual. If we haven't matched any alternatives in any iteration, check to
1139          see if a previous iteration matched. If so, the group has matched;
1140          continue from afterwards. Otherwise it has failed; restore the previous
1141          capture values before returning NOMATCH. */
1142    
1143          for (;;)
1144            {
1145            md->offset_vector[md->offset_end - number] =
1146              (int)(eptr - md->start_subject);
1147            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1148            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1149              eptrb, RM63);
1150            if (rrc == MATCH_KETRPOS)
1151              {
1152              offset_top = md->end_offset_top;
1153              eptr = md->end_match_ptr;
1154              ecode = md->start_code + code_offset;
1155              save_capture_last = md->capture_last;
1156              matched_once = TRUE;
1157              continue;
1158              }
1159    
1160            /* See comment in the code for capturing groups above about handling
1161            THEN. */
1162    
1163            if (rrc == MATCH_THEN)
1164              {
1165              next = ecode + GET(ecode,1);
1166              if (md->start_match_ptr < next &&
1167                  (*ecode == OP_ALT || *next == OP_ALT))
1168                rrc = MATCH_NOMATCH;
1169              }
1170    
1171            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1172            md->capture_last = save_capture_last;
1173            ecode += GET(ecode, 1);
1174            if (*ecode != OP_ALT) break;
1175          }          }
1176    
1177        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (!matched_once)
1178        otherwise return. */          {
1179            md->offset_vector[offset] = save_offset1;
1180            md->offset_vector[offset+1] = save_offset2;
1181            md->offset_vector[md->offset_end - number] = save_offset3;
1182            }
1183    
1184        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        if (allow_zero || matched_once)
1185          eptrb, flags, RM2);          {
1186        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);          ecode += 1 + LINK_SIZE;
1187            break;
1188            }
1189    
1190          RRETURN(MATCH_NOMATCH);
1191          }
1192    
1193        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1194        as a non-capturing bracket. */
1195    
1196        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1197        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1198    
1199        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1200    
1201        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1202        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1203    
1204        /* Non-capturing possessive bracket with unlimited repeat. We come here
1205        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1206        without the capturing complication. It is written out separately for speed
1207        and cleanliness. */
1208    
1209        case OP_BRAPOS:
1210        case OP_SBRAPOS:
1211        allow_zero = FALSE;
1212    
1213        POSSESSIVE_NON_CAPTURE:
1214        matched_once = FALSE;
1215        code_offset = (int)(ecode - md->start_code);
1216    
1217        for (;;)
1218          {
1219          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1220          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1221            eptrb, RM48);
1222          if (rrc == MATCH_KETRPOS)
1223            {
1224            offset_top = md->end_offset_top;
1225            eptr = md->end_match_ptr;
1226            ecode = md->start_code + code_offset;
1227            matched_once = TRUE;
1228            continue;
1229            }
1230    
1231          /* See comment in the code for capturing groups above about handling
1232          THEN. */
1233    
1234          if (rrc == MATCH_THEN)
1235            {
1236            next = ecode + GET(ecode,1);
1237            if (md->start_match_ptr < next &&
1238                (*ecode == OP_ALT || *next == OP_ALT))
1239              rrc = MATCH_NOMATCH;
1240            }
1241    
1242          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1243        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1244          if (*ecode != OP_ALT) break;
1245          }
1246    
1247        if (matched_once || allow_zero)
1248          {
1249          ecode += 1 + LINK_SIZE;
1250          break;
1251        }        }
1252        RRETURN(MATCH_NOMATCH);
1253    
1254      /* Control never reaches here. */      /* Control never reaches here. */
1255    
1256      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1257      two branches. If the condition is false, skipping the first branch takes us      two branches. If the condition is false, skipping the first branch takes us
1258      past the end if there is only one branch, but that's OK because that is      past the end if there is only one branch, but that's OK because that is
1259      exactly what going to the ket would do. As there is only one branch to be      exactly what going to the ket would do. */
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1260    
1261      case OP_COND:      case OP_COND:
1262      case OP_SCOND:      case OP_SCOND:
1263      codelink= GET(ecode, 1);      codelink = GET(ecode, 1);
1264    
1265      /* 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
1266      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
1267    
1268      if (ecode[LINK_SIZE+1] == OP_CALLOUT)      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1269        {        {
1270        if (pcre_callout != NULL)        if (PUBL(callout) != NULL)
1271          {          {
1272          pcre_callout_block cb;          PUBL(callout_block) cb;
1273          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1274          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1275          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1276    #ifdef COMPILE_PCRE8
1277          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
1278          cb.subject_length   = md->end_subject - md->start_subject;  #else
1279          cb.start_match      = mstart - md->start_subject;          cb.subject          = (PCRE_SPTR16)md->start_subject;
1280          cb.current_position = eptr - md->start_subject;  #endif
1281            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1282            cb.start_match      = (int)(mstart - md->start_subject);
1283            cb.current_position = (int)(eptr - md->start_subject);
1284          cb.pattern_position = GET(ecode, LINK_SIZE + 3);          cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1285          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1286          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1287          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1288          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1289          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);          cb.mark             = md->nomatch_mark;
1290            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1291          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1292          }          }
1293        ecode += _pcre_OP_lengths[OP_CALLOUT];        ecode += PRIV(OP_lengths)[OP_CALLOUT];
1294        }        }
1295    
1296      condcode = ecode[LINK_SIZE+1];      condcode = ecode[LINK_SIZE+1];
# Line 917  for (;;) Line 1307  for (;;)
1307        else        else
1308          {          {
1309          int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/          int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1310          condition =  (recno == RREF_ANY || recno == md->recursive->group_num);          condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1311    
1312          /* If the test is for recursion into a specific subpattern, and it is          /* If the test is for recursion into a specific subpattern, and it is
1313          false, but the test was set up by name, scan the table to see if the          false, but the test was set up by name, scan the table to see if the
1314          name refers to any other numbers, and test them. The condition is true          name refers to any other numbers, and test them. The condition is true
1315          if any one is set. */          if any one is set. */
1316    
1317          if (!condition && condcode == OP_NRREF && recno != RREF_ANY)          if (!condition && condcode == OP_NRREF)
1318            {            {
1319            uschar *slotA = md->name_table;            pcre_uchar *slotA = md->name_table;
1320            for (i = 0; i < md->name_count; i++)            for (i = 0; i < md->name_count; i++)
1321              {              {
1322              if (GET2(slotA, 0) == recno) break;              if (GET2(slotA, 0) == recno) break;
# Line 939  for (;;) Line 1329  for (;;)
1329    
1330            if (i < md->name_count)            if (i < md->name_count)
1331              {              {
1332              uschar *slotB = slotA;              pcre_uchar *slotB = slotA;
1333              while (slotB > md->name_table)              while (slotB > md->name_table)
1334                {                {
1335                slotB -= md->name_entry_size;                slotB -= md->name_entry_size;
1336                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1337                  {                  {
1338                  condition = GET2(slotB, 0) == md->recursive->group_num;                  condition = GET2(slotB, 0) == md->recursive->group_num;
1339                  if (condition) break;                  if (condition) break;
# Line 959  for (;;) Line 1349  for (;;)
1349                for (i++; i < md->name_count; i++)                for (i++; i < md->name_count; i++)
1350                  {                  {
1351                  slotB += md->name_entry_size;                  slotB += md->name_entry_size;
1352                  if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1353                    {                    {
1354                    condition = GET2(slotB, 0) == md->recursive->group_num;                    condition = GET2(slotB, 0) == md->recursive->group_num;
1355                    if (condition) break;                    if (condition) break;
# Line 972  for (;;) Line 1362  for (;;)
1362    
1363          /* Chose branch according to the condition */          /* Chose branch according to the condition */
1364    
1365          ecode += condition? 3 : GET(ecode, 1);          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1366          }          }
1367        }        }
1368    
# Line 989  for (;;) Line 1379  for (;;)
1379        if (!condition && condcode == OP_NCREF)        if (!condition && condcode == OP_NCREF)
1380          {          {
1381          int refno = offset >> 1;          int refno = offset >> 1;
1382          uschar *slotA = md->name_table;          pcre_uchar *slotA = md->name_table;
1383    
1384          for (i = 0; i < md->name_count; i++)          for (i = 0; i < md->name_count; i++)
1385            {            {
# Line 1003  for (;;) Line 1393  for (;;)
1393    
1394          if (i < md->name_count)          if (i < md->name_count)
1395            {            {
1396            uschar *slotB = slotA;            pcre_uchar *slotB = slotA;
1397            while (slotB > md->name_table)            while (slotB > md->name_table)
1398              {              {
1399              slotB -= md->name_entry_size;              slotB -= md->name_entry_size;
1400              if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)              if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1401                {                {
1402                offset = GET2(slotB, 0) << 1;                offset = GET2(slotB, 0) << 1;
1403                condition = offset < offset_top &&                condition = offset < offset_top &&
# Line 1025  for (;;) Line 1415  for (;;)
1415              for (i++; i < md->name_count; i++)              for (i++; i < md->name_count; i++)
1416                {                {
1417                slotB += md->name_entry_size;                slotB += md->name_entry_size;
1418                if (strcmp((char *)slotA + 2, (char *)slotB + 2) == 0)                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1419                  {                  {
1420                  offset = GET2(slotB, 0) << 1;                  offset = GET2(slotB, 0) << 1;
1421                  condition = offset < offset_top &&                  condition = offset < offset_top &&
# Line 1040  for (;;) Line 1430  for (;;)
1430    
1431        /* Chose branch according to the condition */        /* Chose branch according to the condition */
1432    
1433        ecode += condition? 3 : GET(ecode, 1);        ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1434        }        }
1435    
1436      else if (condcode == OP_DEF)     /* DEFINE - always false */      else if (condcode == OP_DEF)     /* DEFINE - always false */
# Line 1050  for (;;) Line 1440  for (;;)
1440        }        }
1441    
1442      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1443      the final argument match_condassert causes it to stop at the end of an      md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1444      assertion. */      an assertion. */
1445    
1446      else      else
1447        {        {
1448        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1449            match_condassert, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1450        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1451          {          {
1452            if (md->end_offset_top > offset_top)
1453              offset_top = md->end_offset_top;  /* Captures may have happened */
1454          condition = TRUE;          condition = TRUE;
1455          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1456          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1457          }          }
1458    
1459          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1460          assertion; it is therefore treated as NOMATCH. */
1461    
1462        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1463          {          {
1464          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
# Line 1074  for (;;) Line 1470  for (;;)
1470          }          }
1471        }        }
1472    
1473      /* We are now at the branch that is to be obeyed. As there is only one,      /* We are now at the branch that is to be obeyed. As there is only one, can
1474      we can use tail recursion to avoid using another stack frame, except when      use tail recursion to avoid using another stack frame, except when there is
1475      match_cbegroup is required for an unlimited repeat of a possibly empty      unlimited repeat of a possibly empty group. In the latter case, a recursive
1476      group. If the second alternative doesn't exist, we can just plough on. */      call to match() is always required, unless the second alternative doesn't
1477        exist, in which case we can just plough on. Note that, for compatibility
1478        with Perl, the | in a conditional group is NOT treated as creating two
1479        alternatives. If a THEN is encountered in the branch, it propagates out to
1480        the enclosing alternative (unless nested in a deeper set of alternatives,
1481        of course). */
1482    
1483      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1484        {        {
1485        ecode += 1 + LINK_SIZE;        if (op != OP_SCOND)
       if (op == OP_SCOND)        /* Possibly empty group */  
         {  
         RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);  
         RRETURN(rrc);  
         }  
       else                       /* Group must match something */  
1486          {          {
1487          flags = 0;          ecode += 1 + LINK_SIZE;
1488          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1489          }          }
1490    
1491          md->match_function_type = MATCH_CBEGROUP;
1492          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1493          RRETURN(rrc);
1494        }        }
1495      else                         /* Condition false & no alternative */  
1496         /* Condition false & no alternative; continue after the group. */
1497    
1498        else
1499        {        {
1500        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1501        }        }
# Line 1117  for (;;) Line 1519  for (;;)
1519        {        {
1520        md->offset_vector[offset] =        md->offset_vector[offset] =
1521          md->offset_vector[md->offset_end - number];          md->offset_vector[md->offset_end - number];
1522        md->offset_vector[offset+1] = eptr - md->start_subject;        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1523        if (offset_top <= offset) offset_top = offset + 2;        if (offset_top <= offset) offset_top = offset + 2;
1524        }        }
1525      ecode += 3;      ecode += 1 + IMM2_SIZE;
1526      break;      break;
1527    
1528    
1529      /* End of the pattern, either real or forced. If we are in a top-level      /* End of the pattern, either real or forced. */
     recursion, we should restore the offsets appropriately and continue from  
     after the call. */  
1530    
     case OP_ACCEPT:  
1531      case OP_END:      case OP_END:
1532      if (md->recursive != NULL && md->recursive->group_num == 0)      case OP_ACCEPT:
1533        {      case OP_ASSERT_ACCEPT:
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       offset_top = rec->save_offset_top;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1534    
1535      /* 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
1536      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
1537      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,
1538      if any. */      backtracking will then try other alternatives, if any. */
1539    
1540      if (eptr == mstart &&      if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1541          (md->notempty ||           md->recursive == NULL &&
1542            (md->notempty_atstart &&           (md->notempty ||
1543              mstart == md->start_subject + md->start_offset)))             (md->notempty_atstart &&
1544        MRRETURN(MATCH_NOMATCH);               mstart == md->start_subject + md->start_offset)))
1545          RRETURN(MATCH_NOMATCH);
1546    
1547      /* Otherwise, we have a match. */      /* Otherwise, we have a match. */
1548    
# Line 1161  for (;;) Line 1551  for (;;)
1551      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1552    
1553      /* For some reason, the macros don't work properly if an expression is      /* For some reason, the macros don't work properly if an expression is
1554      given as the argument to MRRETURN when the heap is in use. */      given as the argument to RRETURN when the heap is in use. */
1555    
1556      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1557      MRRETURN(rrc);      RRETURN(rrc);
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1558    
1559      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1560      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,
1561      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
1562      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
1563      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1564        of a condition, we want to return immediately afterwards. The caller of
1565        this incarnation of the match() function will have set MATCH_CONDASSERT in
1566        md->match_function type, and one of these opcodes will be the first opcode
1567        that is processed. We use a local variable that is preserved over calls to
1568        match() to remember this case. */
1569    
1570      case OP_ASSERT:      case OP_ASSERT:
1571      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1572        save_mark = md->mark;
1573        if (md->match_function_type == MATCH_CONDASSERT)
1574          {
1575          condassert = TRUE;
1576          md->match_function_type = 0;
1577          }
1578        else condassert = FALSE;
1579    
1580      do      do
1581        {        {
1582        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
         RM4);  
1583        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1584          {          {
1585          mstart = md->start_match_ptr;   /* In case \K reset it */          mstart = md->start_match_ptr;   /* In case \K reset it */
1586          break;          break;
1587          }          }
1588          md->mark = save_mark;
1589    
1590          /* A COMMIT failure must fail the entire assertion, without trying any
1591          subsequent branches. */
1592    
1593          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1594    
1595          /* PCRE does not allow THEN to escape beyond an assertion; it
1596          is treated as NOMATCH. */
1597    
1598        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1599        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1600        }        }
1601      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1602      if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);  
1603        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1604    
1605      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1606    
1607      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1608    
1609      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1610      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 1215  for (;;) Line 1620  for (;;)
1620    
1621      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1622      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1623        save_mark = md->mark;
1624        if (md->match_function_type == MATCH_CONDASSERT)
1625          {
1626          condassert = TRUE;
1627          md->match_function_type = 0;
1628          }
1629        else condassert = FALSE;
1630    
1631      do      do
1632        {        {
1633        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1634          RM5);        md->mark = save_mark;
1635        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1636        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1637          {          {
1638          do ecode += GET(ecode,1); while (*ecode == OP_ALT);          do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1639          break;          break;
1640          }          }
1641    
1642          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1643          as NOMATCH. */
1644    
1645        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
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);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1651    
1652      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1653      continue;      continue;
# Line 1241  for (;;) Line 1658  for (;;)
1658      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1659    
1660      case OP_REVERSE:      case OP_REVERSE:
1661  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1662      if (utf8)      if (utf)
1663        {        {
1664        i = GET(ecode, 1);        i = GET(ecode, 1);
1665        while (i-- > 0)        while (i-- > 0)
1666          {          {
1667          eptr--;          eptr--;
1668          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1669          BACKCHAR(eptr);          BACKCHAR(eptr);
1670          }          }
1671        }        }
# Line 1259  for (;;) Line 1676  for (;;)
1676    
1677        {        {
1678        eptr -= GET(ecode, 1);        eptr -= GET(ecode, 1);
1679        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1680        }        }
1681    
1682      /* Save the earliest consulted character, then skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
# Line 1273  for (;;) Line 1690  for (;;)
1690      function is able to force a failure. */      function is able to force a failure. */
1691    
1692      case OP_CALLOUT:      case OP_CALLOUT:
1693      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1694        {        {
1695        pcre_callout_block cb;        PUBL(callout_block) cb;
1696        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1697        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1698        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1699    #ifdef COMPILE_PCRE8
1700        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1701        cb.subject_length   = md->end_subject - md->start_subject;  #else
1702        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1703        cb.current_position = eptr - md->start_subject;  #endif
1704          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1705          cb.start_match      = (int)(mstart - md->start_subject);
1706          cb.current_position = (int)(eptr - md->start_subject);
1707        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1708        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1709        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1710        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1711        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1712        if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1713          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1714        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1715        }        }
1716      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 1298  for (;;) Line 1720  for (;;)
1720      offset data is the offset to the starting bracket from the start of the      offset data is the offset to the starting bracket from the start of the
1721      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1722    
1723      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1724      save their starting points and reinstate them after the recursion. However,      re-instated afterwards. We don't know how many are started and not yet
1725      we don't know how many such there are (offset_top records the completed      finished (offset_top records the completed total) so we just have to save
1726      total) so we just have to save all the potential data. There may be up to      all the potential data. There may be up to 65535 such values, which is too
1727      65535 such values, which is too large to put on the stack, but using malloc      large to put on the stack, but using malloc for small numbers seems
1728      for small numbers seems expensive. As a compromise, the stack is used when      expensive. As a compromise, the stack is used when there are no more than
1729      there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc      REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
     is used. A problem is what to do if the malloc fails ... there is no way of  
     returning to the top level with an error. Save the top REC_STACK_SAVE_MAX  
     values on the stack, and accept that the rest may be wrong.  
1730    
1731      There are also other values that have to be saved. We use a chained      There are also other values that have to be saved. We use a chained
1732      sequence of blocks that actually live on the stack. Thanks to Robin Houston      sequence of blocks that actually live on the stack. Thanks to Robin Houston
1733      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1734        a lot, so he is not to blame for the current way it works. */
1735    
1736      case OP_RECURSE:      case OP_RECURSE:
1737        {        {
1738          recursion_info *ri;
1739          int recno;
1740    
1741        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1742        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1743          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1744    
1745          /* Check for repeating a recursion without advancing the subject pointer.
1746          This should catch convoluted mutual recursions. (Some simple cases are
1747          caught at compile time.) */
1748    
1749          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1750            if (recno == ri->group_num && eptr == ri->subject_position)
1751              RRETURN(PCRE_ERROR_RECURSELOOP);
1752    
1753        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1754    
1755          new_recursive.group_num = recno;
1756          new_recursive.subject_position = eptr;
1757        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1758        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1759    
1760        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1761    
1762        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1763    
1764        /* Now save the offset data. */        /* Now save the offset data */
1765    
1766        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1767        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1337  for (;;) Line 1769  for (;;)
1769        else        else
1770          {          {
1771          new_recursive.offset_save =          new_recursive.offset_save =
1772            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1773          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1774          }          }
   
1775        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1776              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_offset_top = offset_top;  
1777    
1778        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1779        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1780          might be changed, so reset it before looping. */
1781    
1782        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1783        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1784        do        do
1785          {          {
1786          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1787            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1788              md, eptrb, RM6);
1789            memcpy(md->offset_vector, new_recursive.offset_save,
1790                new_recursive.saved_max * sizeof(int));
1791            md->recursive = new_recursive.prevrec;
1792          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1793            {            {
1794            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1795            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1796              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1797            MRRETURN(MATCH_MATCH);  
1798              /* Set where we got to in the subject, and reset the start in case
1799              it was changed by \K. This *is* propagated back out of a recursion,
1800              for Perl compatibility. */
1801    
1802              eptr = md->end_match_ptr;
1803              mstart = md->start_match_ptr;
1804              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1805            }            }
1806          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1807            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1808            is treated as NOMATCH. */
1809    
1810            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1811                     rrc != MATCH_COMMIT)
1812            {            {
1813            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1814            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1815              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1816            RRETURN(rrc);            RRETURN(rrc);
1817            }            }
1818    
1819          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1820          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1821          }          }
1822        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1380  for (;;) Line 1824  for (;;)
1824        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1825        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1826        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1827          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1828        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
       }  
     /* 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, but resetting  
     the start-of-match value in case it was changed by \K. */  
   
     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)  /* Note: _not_ MATCH_ACCEPT */  
         {  
         mstart = md->start_match_ptr;  
         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));  
1829        }        }
1830    
1831      if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1832        {      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 */  
1833    
1834      /* An alternation is the end of a branch; scan along to find the end of the      /* An alternation is the end of a branch; scan along to find the end of the
1835      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1478  for (;;) Line 1845  for (;;)
1845      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1846    
1847      case OP_BRAZERO:      case OP_BRAZERO:
1848        {      next = ecode + 1;
1849        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1850        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1851        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1852        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1853      break;      break;
1854    
1855      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1856        {      next = ecode + 1;
1857        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1858        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1859        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1860        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1861      break;      break;
1862    
1863      case OP_SKIPZERO:      case OP_SKIPZERO:
1864        {      next = ecode+1;
1865        next = ecode+1;      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        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1870        here; just jump to the group, with allow_zero set TRUE. */
1871    
1872        case OP_BRAPOSZERO:
1873        op = *(++ecode);
1874        allow_zero = TRUE;
1875        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1876          goto POSSESSIVE_NON_CAPTURE;
1877    
1878      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1879    
1880      case OP_KET:      case OP_KET:
1881      case OP_KETRMIN:      case OP_KETRMIN:
1882      case OP_KETRMAX:      case OP_KETRMAX:
1883        case OP_KETRPOS:
1884      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1885    
1886      /* 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
1887      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1888      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1889    
1890      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1891        {        {
1892        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1893        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1894        }        }
1895      else saved_eptr = NULL;      else saved_eptr = NULL;
1896    
1897      /* If we are at the end of an assertion group or an atomic group, stop      /* If we are at the end of an assertion group or a non-capturing atomic
1898      matching and return MATCH_MATCH, but record the current high water mark for      group, stop matching and return MATCH_MATCH, but record the current high
1899      use by positive assertions. We also need to record the match start in case      water mark for use by positive assertions. We also need to record the match
1900      it was changed by \K. */      start in case it was changed by \K. */
1901    
1902      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1903          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||           *prev == OP_ONCE_NC)
         *prev == OP_ONCE)  
1904        {        {
1905        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1906        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1907        md->start_match_ptr = mstart;        md->start_match_ptr = mstart;
1908        MRRETURN(MATCH_MATCH);        RRETURN(MATCH_MATCH);         /* Sets md->mark */
1909        }        }
1910    
1911      /* 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
1912      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1913      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
1914      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
1915      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
1916        the current subject position and start match pointer and give a MATCH
1917        return. */
1918    
1919      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1920            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1921        {        {
1922        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1923        offset = number << 1;        offset = number << 1;
# Line 1554  for (;;) Line 1927  for (;;)
1927        printf("\n");        printf("\n");
1928  #endif  #endif
1929    
1930          /* Handle a recursively called group. */
1931    
1932          if (md->recursive != NULL && md->recursive->group_num == number)
1933            {
1934            md->end_match_ptr = eptr;
1935            md->start_match_ptr = mstart;
1936            RRETURN(MATCH_MATCH);
1937            }
1938    
1939          /* Deal with capturing */
1940    
1941        md->capture_last = number;        md->capture_last = number;
1942        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1943          {          {
1944            /* If offset is greater than offset_top, it means that we are
1945            "skipping" a capturing group, and that group's offsets must be marked
1946            unset. In earlier versions of PCRE, all the offsets were unset at the
1947            start of matching, but this doesn't work because atomic groups and
1948            assertions can cause a value to be set that should later be unset.
1949            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1950            part of the atomic group, but this is not on the final matching path,
1951            so must be unset when 2 is set. (If there is no group 2, there is no
1952            problem, because offset_top will then be 2, indicating no capture.) */
1953    
1954            if (offset > offset_top)
1955              {
1956              register int *iptr = md->offset_vector + offset_top;
1957              register int *iend = md->offset_vector + offset;
1958              while (iptr < iend) *iptr++ = -1;
1959              }
1960    
1961            /* Now make the extraction */
1962    
1963          md->offset_vector[offset] =          md->offset_vector[offset] =
1964            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1965          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1966          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1967          }          }
   
       /* Handle a recursively called group. Restore the offsets  
       appropriately and continue from after the call. */  
   
       if (md->recursive != NULL && md->recursive->group_num == number)  
         {  
         recursion_info *rec = md->recursive;  
         DPRINTF(("Recursion (%d) succeeded - continuing\n", number));  
         md->recursive = rec->prevrec;  
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         offset_top = rec->save_offset_top;  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
         }  
1968        }        }
1969    
1970      /* For both capturing and non-capturing groups, reset the value of the ims      /* For an ordinary non-repeating ket, just continue at this level. This
1971      flags, in case they got changed during the group. */      also happens for a repeating ket if no characters were matched in the
1972        group. This is the forcible breaking of infinite loops as implemented in
1973      ims = original_ims;      Perl 5.005. For a non-repeating atomic group that includes captures,
1974      DPRINTF(("ims reset to %02lx\n", ims));      establish a backup point by processing the rest of the pattern at a lower
1975        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1976      /* For a non-repeating ket, just continue at this level. This also      original OP_ONCE level, thereby bypassing intermediate backup points, but
1977      happens for a repeating ket if no characters were matched in the group.      resetting any captures that happened along the way. */
     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. */  
1978    
1979      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1980        {        {
1981        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1982            {
1983            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1984            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1985            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1986            RRETURN(MATCH_ONCE);
1987            }
1988          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1989        break;        break;
1990        }        }
1991    
1992      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1993      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1994      tail recursion to avoid using another stack frame, unless we have an      at a time from the outer level, thus saving stack. */
1995      unlimited repeat of a group that can match an empty string. */  
1996        if (*ecode == OP_KETRPOS)
1997          {
1998          md->end_match_ptr = eptr;
1999          md->end_offset_top = offset_top;
2000          RRETURN(MATCH_KETRPOS);
2001          }
2002    
2003      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      /* The normal repeating kets try the rest of the pattern or restart from
2004        the preceding bracket, in the appropriate order. In the second case, we can
2005        use tail recursion to avoid using another stack frame, unless we have an
2006        an atomic group or an unlimited repeat of a group that can match an empty
2007        string. */
2008    
2009      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2010        {        {
2011        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2012        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2013        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2014          {          {
2015          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2016            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2017            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2018            RRETURN(MATCH_ONCE);
2019            }
2020          if (*prev >= OP_SBRA)    /* Could match an empty string */
2021            {
2022            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2023          RRETURN(rrc);          RRETURN(rrc);
2024          }          }
2025        ecode = prev;        ecode = prev;
# Line 1619  for (;;) Line 2027  for (;;)
2027        }        }
2028      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2029        {        {
2030        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2031          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2032        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2033          if (*prev == OP_ONCE)
2034            {
2035            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2036            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2037            md->once_target = prev;
2038            RRETURN(MATCH_ONCE);
2039            }
2040        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2041        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2042        }        }
2043      /* Control never gets here */      /* Control never gets here */
2044    
2045      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2046    
2047      case OP_CIRC:      case OP_CIRC:
2048      if (md->notbol && eptr == md->start_subject) MRRETURN(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)))  
         MRRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2049    
2050      /* Start of subject assertion */      /* Start of subject assertion */
2051    
2052      case OP_SOD:      case OP_SOD:
2053      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);
2054        ecode++;
2055        break;
2056    
2057        /* Multiline mode: start of subject unless notbol, or after any newline. */
2058    
2059        case OP_CIRCM:
2060        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2061        if (eptr != md->start_subject &&
2062            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2063          RRETURN(MATCH_NOMATCH);
2064      ecode++;      ecode++;
2065      break;      break;
2066    
2067      /* Start of match assertion */      /* Start of match assertion */
2068    
2069      case OP_SOM:      case OP_SOM:
2070      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);
2071      ecode++;      ecode++;
2072      break;      break;
2073    
# Line 1662  for (;;) Line 2078  for (;;)
2078      ecode++;      ecode++;
2079      break;      break;
2080    
2081      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2082      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2083    
2084      case OP_DOLL:      case OP_DOLLM:
2085      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2086        {        {
2087        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2088          { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }          {
2089        else          if (md->partial != 0 &&
2090          { if (md->noteol) MRRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2091        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2092        break;              NLBLOCK->nllen == 2 &&
2093                *eptr == NLBLOCK->nl[0])
2094              {
2095              md->hitend = TRUE;
2096              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2097              }
2098            RRETURN(MATCH_NOMATCH);
2099            }
2100        }        }
2101      else      else
2102        {        {
2103        if (md->noteol) MRRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2104        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           MRRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2105        }        }
2106        ecode++;
2107        break;
2108    
2109        /* Not multiline mode: assert before a terminating newline or before end of
2110        subject unless noteol is set. */
2111    
2112        case OP_DOLL:
2113        if (md->noteol) RRETURN(MATCH_NOMATCH);
2114        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2115    
2116      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2117    
2118      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2119    
2120      case OP_EOD:      case OP_EOD:
2121      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2122        SCHECK_PARTIAL();
2123      ecode++;      ecode++;
2124      break;      break;
2125    
2126      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2127    
2128      case OP_EODN:      case OP_EODN:
2129      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2130        if (eptr < md->end_subject &&
2131          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2132        MRRETURN(MATCH_NOMATCH);        {
2133          if (md->partial != 0 &&
2134              eptr + 1 >= md->end_subject &&
2135              NLBLOCK->nltype == NLTYPE_FIXED &&
2136              NLBLOCK->nllen == 2 &&
2137              *eptr == NLBLOCK->nl[0])
2138            {
2139            md->hitend = TRUE;
2140            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2141            }
2142          RRETURN(MATCH_NOMATCH);
2143          }
2144    
2145        /* Either at end of string or \n before end. */
2146    
2147        SCHECK_PARTIAL();
2148      ecode++;      ecode++;
2149      break;      break;
2150    
# Line 1716  for (;;) Line 2159  for (;;)
2159        be "non-word" characters. Remember the earliest consulted character for        be "non-word" characters. Remember the earliest consulted character for
2160        partial matching. */        partial matching. */
2161    
2162  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2163        if (utf8)        if (utf)
2164          {          {
2165          /* Get status of previous character */          /* Get status of previous character */
2166    
2167          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2168            {            {
2169            USPTR lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2170            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2171            if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;            if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2172            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2173  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 1789  for (;;) Line 2232  for (;;)
2232              }              }
2233            else            else
2234  #endif  #endif
2235            prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);            prev_is_word = MAX_255(eptr[-1])
2236                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2237            }            }
2238    
2239          /* Get status of next character */          /* Get status of next character */
# Line 1812  for (;;) Line 2256  for (;;)
2256            }            }
2257          else          else
2258  #endif  #endif
2259          cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);          cur_is_word = MAX_255(*eptr)
2260              && ((md->ctypes[*eptr] & ctype_word) != 0);
2261          }          }
2262    
2263        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
2264    
2265        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
2266             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2267          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
2268        }        }
2269      break;      break;
2270    
2271      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2272        CRLF newlines and partial matching. */
2273    
2274      case OP_ANY:      case OP_ANY:
2275      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2276        if (md->partial != 0 &&
2277            eptr + 1 >= md->end_subject &&
2278            NLBLOCK->nltype == NLTYPE_FIXED &&
2279            NLBLOCK->nllen == 2 &&
2280            *eptr == NLBLOCK->nl[0])
2281          {
2282          md->hitend = TRUE;
2283          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2284          }
2285    
2286      /* Fall through */      /* Fall through */
2287    
2288        /* Match any single character whatsoever. */
2289    
2290      case OP_ALLANY:      case OP_ALLANY:
2291      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2292        {        {                            /* not be updated before SCHECK_PARTIAL. */
2293        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2294        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2295        }        }
2296      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      eptr++;
2297    #ifdef SUPPORT_UTF
2298        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2299    #endif
2300      ecode++;      ecode++;
2301      break;      break;
2302    
# Line 1843  for (;;) Line 2304  for (;;)
2304      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2305    
2306      case OP_ANYBYTE:      case OP_ANYBYTE:
2307      if (eptr++ >= md->end_subject)      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2308        {        {                            /* not be updated before SCHECK_PARTIAL. */
2309        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2310        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2311        }        }
2312        eptr++;
2313      ecode++;      ecode++;
2314      break;      break;
2315    
# Line 1855  for (;;) Line 2317  for (;;)
2317      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2318        {        {
2319        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2320        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2321        }        }
2322      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2323      if (      if (
2324  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2325         c < 256 &&         c < 256 &&
2326  #endif  #endif
2327         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2328         )         )
2329        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2330      ecode++;      ecode++;
2331      break;      break;
2332    
# Line 1872  for (;;) Line 2334  for (;;)
2334      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2335        {        {
2336        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2337        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2338        }        }
2339      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2340      if (      if (
2341  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2342         c >= 256 ||         c > 255 ||
2343  #endif  #endif
2344         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2345         )         )
2346        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2347      ecode++;      ecode++;
2348      break;      break;
2349    
# Line 1889  for (;;) Line 2351  for (;;)
2351      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2352        {        {
2353        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2354        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2355        }        }
2356      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2357      if (      if (
2358  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2359         c < 256 &&         c < 256 &&
2360  #endif  #endif
2361         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2362         )         )
2363        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2364      ecode++;      ecode++;
2365      break;      break;
2366    
# Line 1906  for (;;) Line 2368  for (;;)
2368      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2369        {        {
2370        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2371        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2372        }        }
2373      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2374      if (      if (
2375  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2376         c >= 256 ||         c > 255 ||
2377  #endif  #endif
2378         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2379         )         )
2380        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2381      ecode++;      ecode++;
2382      break;      break;
2383    
# Line 1923  for (;;) Line 2385  for (;;)
2385      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2386        {        {
2387        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2388        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2389        }        }
2390      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2391      if (      if (
2392  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2393         c < 256 &&         c < 256 &&
2394  #endif  #endif
2395         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2396         )         )
2397        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2398      ecode++;      ecode++;
2399      break;      break;
2400    
# Line 1940  for (;;) Line 2402  for (;;)
2402      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2403        {        {
2404        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2405        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2406        }        }
2407      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2408      if (      if (
2409  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2410         c >= 256 ||         c > 255 ||
2411  #endif  #endif
2412         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2413         )         )
2414        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2415      ecode++;      ecode++;
2416      break;      break;
2417    
# Line 1957  for (;;) Line 2419  for (;;)
2419      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2420        {        {
2421        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2422        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2423        }        }
2424      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2425      switch(c)      switch(c)
2426        {        {
2427        default: MRRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2428        case 0x000d:  
2429        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2430          if (eptr >= md->end_subject)
2431            {
2432            SCHECK_PARTIAL();
2433            }
2434          else if (*eptr == CHAR_LF) eptr++;
2435        break;        break;
2436    
2437        case 0x000a:        case CHAR_LF:
2438        break;        break;
2439    
2440        case 0x000b:        case CHAR_VT:
2441        case 0x000c:        case CHAR_FF:
2442        case 0x0085:        case CHAR_NEL:
2443    #ifndef EBCDIC
2444        case 0x2028:        case 0x2028:
2445        case 0x2029:        case 0x2029:
2446        if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);  #endif  /* Not EBCDIC */
2447          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2448        break;        break;
2449        }        }
2450      ecode++;      ecode++;
# Line 1985  for (;;) Line 2454  for (;;)
2454      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2455        {        {
2456        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2457        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2458        }        }
2459      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2460      switch(c)      switch(c)
2461        {        {
2462          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2463        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 */  
       MRRETURN(MATCH_NOMATCH);  
2464        }        }
2465      ecode++;      ecode++;
2466      break;      break;
# Line 2019  for (;;) Line 2469  for (;;)
2469      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2470        {        {
2471        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2472        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2473        }        }
2474      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2475      switch(c)      switch(c)
2476        {        {
2477        default: MRRETURN(MATCH_NOMATCH);        HSPACE_CASES: break;  /* Byte and multibyte cases */
2478        case 0x09:      /* HT */        default: RRETURN(MATCH_NOMATCH);
       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;  
2479        }        }
2480      ecode++;      ecode++;
2481      break;      break;
# Line 2053  for (;;) Line 2484  for (;;)
2484      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2485        {        {
2486        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2487        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2488        }        }
2489      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2490      switch(c)      switch(c)
2491        {        {
2492          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2493        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 */  
       MRRETURN(MATCH_NOMATCH);  
2494        }        }
2495      ecode++;      ecode++;
2496      break;      break;
# Line 2075  for (;;) Line 2499  for (;;)
2499      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2500        {        {
2501        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2502        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2503        }        }
2504      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2505      switch(c)      switch(c)
2506        {        {
2507        default: MRRETURN(MATCH_NOMATCH);        VSPACE_CASES: break;
2508        case 0x0a:      /* LF */        default: RRETURN(MATCH_NOMATCH);
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       break;  
2509        }        }
2510      ecode++;      ecode++;
2511      break;      break;
# Line 2102  for (;;) Line 2519  for (;;)
2519      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2520        {        {
2521        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2522        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2523        }        }
2524      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2525        {        {
2526          const pcre_uint32 *cp;
2527        const ucd_record *prop = GET_UCD(c);        const ucd_record *prop = GET_UCD(c);
2528    
2529        switch(ecode[1])        switch(ecode[1])
2530          {          {
2531          case PT_ANY:          case PT_ANY:
2532          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2533          break;          break;
2534    
2535          case PT_LAMP:          case PT_LAMP:
2536          if ((prop->chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2537               prop->chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2538               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2539            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2540          break;          break;
2541    
2542          case PT_GC:          case PT_GC:
2543          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2544            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2545          break;          break;
2546    
2547          case PT_PC:          case PT_PC:
2548          if ((ecode[2] != prop->chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2549            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2550          break;          break;
2551    
2552          case PT_SC:          case PT_SC:
2553          if ((ecode[2] != prop->script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2554            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2555          break;          break;
2556    
2557          /* These are specials */          /* These are specials */
2558    
2559          case PT_ALNUM:          case PT_ALNUM:
2560          if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||          if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2561               _pcre_ucp_gentype[prop->chartype] == ucp_N) == (op == OP_NOTPROP))               PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2562            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2563          break;          break;
2564    
2565          case PT_SPACE:    /* Perl space */          case PT_SPACE:    /* Perl space */
2566          if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||          if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2567               c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)               c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2568                 == (op == OP_NOTPROP))                 == (op == OP_NOTPROP))
2569            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2570          break;          break;
2571    
2572          case PT_PXSPACE:  /* POSIX space */          case PT_PXSPACE:  /* POSIX space */
2573          if ((_pcre_ucp_gentype[prop->chartype] == ucp_Z ||          if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2574               c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||               c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2575               c == CHAR_FF || c == CHAR_CR)               c == CHAR_FF || c == CHAR_CR)
2576                 == (op == OP_NOTPROP))                 == (op == OP_NOTPROP))
2577            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2578          break;          break;
2579    
2580          case PT_WORD:          case PT_WORD:
2581          if ((_pcre_ucp_gentype[prop->chartype] == ucp_L ||          if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2582               _pcre_ucp_gentype[prop->chartype] == ucp_N ||               PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2583               c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))               c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2584            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2585          break;          break;
2586    
2587            case PT_CLIST:
2588            cp = PRIV(ucd_caseless_sets) + prop->caseset;
2589            for (;;)
2590              {
2591              if (c < *cp)
2592                { if (op == OP_PROP) RRETURN(MATCH_NOMATCH); else break; }
2593              if (c == *cp++)
2594                { if (op == OP_PROP) break; else RRETURN(MATCH_NOMATCH); }
2595              }
2596            break;
2597    
2598          /* This should never occur */          /* This should never occur */
2599    
# Line 2183  for (;;) Line 2612  for (;;)
2612      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
2613        {        {
2614        SCHECK_PARTIAL();        SCHECK_PARTIAL();
2615        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2616        }        }
2617      GETCHARINCTEST(c, eptr);      else
2618        {        {
2619        int category = UCD_CATEGORY(c);        int lgb, rgb;
2620        if (category == ucp_M) MRRETURN(MATCH_NOMATCH);        GETCHARINCTEST(c, eptr);
2621          lgb = UCD_GRAPHBREAK(c);
2622        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2623          {          {
2624          int len = 1;          int len = 1;
2625          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2626            {          rgb = UCD_GRAPHBREAK(c);
2627            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2628            }          lgb = rgb;
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
2629          eptr += len;          eptr += len;
2630          }          }
2631        }        }
2632        CHECK_PARTIAL();
2633      ecode++;      ecode++;
2634      break;      break;
2635  #endif  #endif  /* SUPPORT_UCP */
2636    
2637    
2638      /* Match a back reference, possibly repeatedly. Look past the end of the      /* Match a back reference, possibly repeatedly. Look past the end of the
# Line 2215  for (;;) Line 2644  for (;;)
2644      loops). */      loops). */
2645    
2646      case OP_REF:      case OP_REF:
2647        {      case OP_REFI:
2648        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2649        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2650        ecode += 1 + IMM2_SIZE;
2651    
2652        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2653    
2654        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2655        than the amount of subject left; this ensures that every attempt at a      this ensures that every attempt at a match fails. We can't just fail
2656        match fails. We can't just fail here, because of the possibility of      here, because of the possibility of quantifiers with zero minima.
       quantifiers with zero minima.  
2657    
2658        (b) If the JavaScript compatibility flag is set, set the length to zero      (b) If the JavaScript compatibility flag is set, set the length to zero
2659        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2660    
2661        Otherwise, set the length to the length of what was matched by the      Otherwise, set the length to the length of what was matched by the
2662        referenced subpattern. */      referenced subpattern. */
2663    
2664        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2665          length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;        length = (md->jscript_compat)? 0 : -1;
2666        else      else
2667          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2668    
2669        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2670    
2671        switch (*ecode)      switch (*ecode)
2672          {        {
2673          case OP_CRSTAR:        case OP_CRSTAR:
2674          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2675          case OP_CRPLUS:        case OP_CRPLUS:
2676          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2677          case OP_CRQUERY:        case OP_CRQUERY:
2678          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2679          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2680          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2681          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2682          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2683          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2684          break;        break;
2685    
2686          case OP_CRRANGE:        case OP_CRRANGE:
2687          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2688          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2689          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2690          max = GET2(ecode, 3);        max = GET2(ecode, 1 + IMM2_SIZE);
2691          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2692          ecode += 5;        ecode += 1 + 2 * IMM2_SIZE;
2693          break;        break;
2694    
2695          default:               /* No repeat follows */        default:               /* No repeat follows */
2696          if (!match_ref(offset, eptr, length, md, ims))        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2697            {          {
2698            CHECK_PARTIAL();          if (length == -2) eptr = md->end_subject;   /* Partial match */
2699            MRRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2700            }          RRETURN(MATCH_NOMATCH);
         eptr += length;  
         continue;              /* With the main loop */  
2701          }          }
2702          eptr += length;
2703          continue;              /* With the main loop */
2704          }
2705    
2706        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2707        main loop. */      zero, just continue with the main loop. If the length is negative, it
2708        means the reference is unset in non-Java-compatible mode. If the minimum is
2709        zero, we can continue at the same level without recursion. For any other
2710        minimum, carrying on will result in NOMATCH. */
2711    
2712        if (length == 0) continue;      if (length == 0) continue;
2713        if (length < 0 && min == 0) continue;
2714    
2715        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2716        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2717        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2718    
2719        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2720          {
2721          int slength;
2722          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2723          {          {
2724          if (!match_ref(offset, eptr, length, md, ims))          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2725            {          CHECK_PARTIAL();
2726            CHECK_PARTIAL();          RRETURN(MATCH_NOMATCH);
           MRRETURN(MATCH_NOMATCH);  
           }  
         eptr += length;  
2727          }          }
2728          eptr += slength;
2729          }
2730    
2731        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2732        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2733    
2734        if (min == max) continue;      if (min == max) continue;
2735    
2736        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2737    
2738        if (minimize)      if (minimize)
2739          {
2740          for (fi = min;; fi++)
2741          {          {
2742          for (fi = min;; fi++)          int slength;
2743            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2744            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2745            if (fi >= max) RRETURN(MATCH_NOMATCH);
2746            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2747            {            {
2748            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2749            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2750            if (fi >= max) MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
           if (!match_ref(offset, eptr, length, md, ims))  
             {  
             CHECK_PARTIAL();  
             MRRETURN(MATCH_NOMATCH);  
             }  
           eptr += length;  
2751            }            }
2752          /* Control never gets here */          eptr += slength;
2753          }          }
2754          /* Control never gets here */
2755          }
2756    
2757        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2758    
2759        else      else
2760          {
2761          pp = eptr;
2762          for (i = min; i < max; i++)
2763          {          {
2764          pp = eptr;          int slength;
2765          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2766            {            {
2767            if (!match_ref(offset, eptr, length, md, ims))            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2768              the soft partial matching case. */
2769    
2770              if (slength == -2 && md->partial != 0 &&
2771                  md->end_subject > md->start_used_ptr)
2772              {              {
2773              CHECK_PARTIAL();              md->hitend = TRUE;
2774              break;              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2775              }              }
2776            eptr += length;            break;
           }  
         while (eptr >= pp)  
           {  
           RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);  
           if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
           eptr -= length;  
2777            }            }
2778          MRRETURN(MATCH_NOMATCH);          eptr += slength;
2779          }          }
2780    
2781          while (eptr >= pp)
2782            {
2783            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2784            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2785            eptr -= length;
2786            }
2787          RRETURN(MATCH_NOMATCH);
2788        }        }
2789      /* Control never gets here */      /* Control never gets here */
2790    
# Line 2355  for (;;) Line 2802  for (;;)
2802      case OP_NCLASS:      case OP_NCLASS:
2803      case OP_CLASS:      case OP_CLASS:
2804        {        {
2805          /* The data variable is saved across frames, so the byte map needs to
2806          be stored there. */
2807    #define BYTE_MAP ((pcre_uint8 *)data)
2808        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2809        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2810    
2811        switch (*ecode)        switch (*ecode)
2812          {          {
# Line 2377  for (;;) Line 2827  for (;;)
2827          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2828          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2829          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2830          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2831          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2832          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2833          break;          break;
2834    
2835          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2389  for (;;) Line 2839  for (;;)
2839    
2840        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2841    
2842  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2843        /* UTF-8 mode */        if (utf)
       if (utf8)  
2844          {          {
2845          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2846            {            {
2847            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
2848              {              {
2849              SCHECK_PARTIAL();              SCHECK_PARTIAL();
2850              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
2851              }              }
2852            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2853            if (c > 255)            if (c > 255)
2854              {              {
2855              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2856              }              }
2857            else            else
2858              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  
             }  
2859            }            }
2860          }          }
2861        else        else
2862  #endif  #endif
2863        /* Not UTF-8 mode */        /* Not UTF mode */
2864          {          {
2865          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2866            {            {
2867            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
2868              {              {
2869              SCHECK_PARTIAL();              SCHECK_PARTIAL();
2870              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
2871              }              }
2872            c = *eptr++;            c = *eptr++;
2873            if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2874              if (c > 255)
2875                {
2876                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2877                }
2878              else
2879    #endif
2880                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2881            }            }
2882          }          }
2883    
# Line 2437  for (;;) Line 2891  for (;;)
2891    
2892        if (minimize)        if (minimize)
2893          {          {
2894  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2895          /* UTF-8 mode */          if (utf)
         if (utf8)  
2896            {            {
2897            for (fi = min;; fi++)            for (fi = min;; fi++)
2898              {              {
2899              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2900              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2901              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2902              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
2903                {                {
2904                SCHECK_PARTIAL();                SCHECK_PARTIAL();
2905                MRRETURN(MATCH_NOMATCH);                RRETURN(MATCH_NOMATCH);
2906                }                }
2907              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2908              if (c > 255)              if (c > 255)
2909                {                {
2910                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2911                }                }
2912              else              else
2913                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  
               }  
2914              }              }
2915            }            }
2916          else          else
2917  #endif  #endif
2918          /* Not UTF-8 mode */          /* Not UTF mode */
2919            {            {
2920            for (fi = min;; fi++)            for (fi = min;; fi++)
2921              {              {
2922              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2923              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2924              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2925              if (eptr >= md->end_subject)              if (eptr >= md->end_subject)
2926                {                {
2927                SCHECK_PARTIAL();                SCHECK_PARTIAL();
2928                MRRETURN(MATCH_NOMATCH);                RRETURN(MATCH_NOMATCH);
2929                }                }
2930              c = *eptr++;              c = *eptr++;
2931              if ((data[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2932                if (c > 255)
2933                  {
2934                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2935                  }
2936                else
2937    #endif
2938                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2939              }              }
2940            }            }
2941          /* Control never gets here */          /* Control never gets here */
# Line 2489  for (;;) Line 2947  for (;;)
2947          {          {
2948          pp = eptr;          pp = eptr;
2949    
2950  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2951          /* UTF-8 mode */          if (utf)
         if (utf8)  
2952            {            {
2953            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2954              {              {
# Line 2507  for (;;) Line 2964  for (;;)
2964                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2965                }                }
2966              else              else
2967                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2968              eptr += len;              eptr += len;
2969              }              }
2970            for (;;)            for (;;)
2971              {              {
2972              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2973              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2974              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2975              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2522  for (;;) Line 2977  for (;;)
2977            }            }
2978          else          else
2979  #endif  #endif
2980            /* Not UTF-8 mode */            /* Not UTF mode */
2981            {            {
2982            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2983              {              {
# Line 2532  for (;;) Line 2987  for (;;)
2987                break;                break;
2988                }                }
2989              c = *eptr;              c = *eptr;
2990              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2991                if (c > 255)
2992                  {
2993                  if (op == OP_CLASS) break;
2994                  }
2995                else
2996    #endif
2997                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2998              eptr++;              eptr++;
2999              }              }
3000            while (eptr >= pp)            while (eptr >= pp)
3001              {              {
3002              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3003              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3004              eptr--;              eptr--;
3005              }              }
3006            }            }
3007    
3008          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3009          }          }
3010    #undef BYTE_MAP
3011        }        }
3012      /* Control never gets here */      /* Control never gets here */
3013    
# Line 2553  for (;;) Line 3016  for (;;)
3016      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
3017      mode, because Unicode properties are supported in non-UTF-8 mode. */      mode, because Unicode properties are supported in non-UTF-8 mode. */
3018    
3019  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3020      case OP_XCLASS:      case OP_XCLASS:
3021        {        {
3022        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2578  for (;;) Line 3041  for (;;)
3041          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3042          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3043          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3044          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3045          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3046          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3047          break;          break;
3048    
3049          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2595  for (;;) Line 3058  for (;;)
3058          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
3059            {            {
3060            SCHECK_PARTIAL();            SCHECK_PARTIAL();
3061            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3062            }            }
3063          GETCHARINCTEST(c, eptr);          GETCHARINCTEST(c, eptr);
3064          if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);          if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3065          }          }
3066    
3067        /* If max == min we can continue with the main loop without the        /* If max == min we can continue with the main loop without the
# Line 2613  for (;;) Line 3076  for (;;)
3076          {          {
3077          for (fi = min;; fi++)          for (fi = min;; fi++)
3078            {            {
3079            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3080            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3081            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3082            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3083              {              {
3084              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3085              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3086              }              }
3087            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
3088            if (!_pcre_xclass(c, data)) MRRETURN(MATCH_NOMATCH);            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3089            }            }
3090          /* Control never gets here */          /* Control never gets here */
3091          }          }
# Line 2640  for (;;) Line 3103  for (;;)
3103              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3104              break;              break;
3105              }              }
3106    #ifdef SUPPORT_UTF
3107            GETCHARLENTEST(c, eptr, len);            GETCHARLENTEST(c, eptr, len);
3108            if (!_pcre_xclass(c, data)) break;  #else
3109              c = *eptr;
3110    #endif
3111              if (!PRIV(xclass)(c, data, utf)) break;
3112            eptr += len;            eptr += len;
3113            }            }
3114          for(;;)          for(;;)
3115            {            {
3116            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3117            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3118            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3119            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3120              if (utf) BACKCHAR(eptr);
3121    #endif
3122            }            }
3123          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3124          }          }
3125    
3126        /* Control never gets here */        /* Control never gets here */
# Line 2661  for (;;) Line 3130  for (;;)
3130      /* Match a single character, casefully */      /* Match a single character, casefully */
3131    
3132      case OP_CHAR:      case OP_CHAR:
3133  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3134      if (utf8)      if (utf)
3135        {        {
3136        length = 1;        length = 1;
3137        ecode++;        ecode++;
# Line 2670  for (;;) Line 3139  for (;;)
3139        if (length > md->end_subject - eptr)        if (length > md->end_subject - eptr)
3140          {          {
3141          CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */          CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3142          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3143          }          }
3144        while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3145        }        }
3146      else      else
3147  #endif  #endif
3148        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3149        {        {
3150        if (md->end_subject - eptr < 1)        if (md->end_subject - eptr < 1)
3151          {          {
3152          SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */          SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3153          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3154          }          }
3155        if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3156        ecode += 2;        ecode += 2;
3157        }        }
3158      break;      break;
3159    
3160      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3161        subject, give up immediately. */
3162    
3163        case OP_CHARI:
3164        if (eptr >= md->end_subject)
3165          {
3166          SCHECK_PARTIAL();
3167          RRETURN(MATCH_NOMATCH);
3168          }
3169    
3170      case OP_CHARNC:  #ifdef SUPPORT_UTF
3171  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3172        {        {
3173        length = 1;        length = 1;
3174        ecode++;        ecode++;
3175        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3176    
       if (length > md->end_subject - eptr)  
         {  
         CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */  
         MRRETURN(MATCH_NOMATCH);  
         }  
   
3177        /* If the pattern character's value is < 128, we have only one byte, and        /* If the pattern character's value is < 128, we have only one byte, and
3178        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3179          fast lookup table. We know that there is at least one byte left in the
3180          subject. */
3181    
3182        if (fc < 128)        if (fc < 128)
3183          {          {
3184          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3185                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3186            ecode++;
3187            eptr++;
3188          }          }
3189    
3190        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3191          use the value of "length" to check for sufficient bytes left, because the
3192          other case of the character may have more or fewer bytes.  */
3193    
3194        else        else
3195          {          {
# Line 2729  for (;;) Line 3205  for (;;)
3205  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3206            if (dc != UCD_OTHERCASE(fc))            if (dc != UCD_OTHERCASE(fc))
3207  #endif  #endif
3208              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3209            }            }
3210          }          }
3211        }        }
3212      else      else
3213  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3214    
3215      /* Non-UTF-8 mode */      /* Not UTF mode */
3216        {        {
3217        if (md->end_subject - eptr < 1)        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3218          {            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3219          SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */        eptr++;
         MRRETURN(MATCH_NOMATCH);  
         }  
       if (md->lcc[ecode[1]] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);  
3220        ecode += 2;        ecode += 2;
3221        }        }
3222      break;      break;
# Line 2751  for (;;) Line 3224  for (;;)
3224      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3225    
3226      case OP_EXACT:      case OP_EXACT:
3227        case OP_EXACTI:
3228      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3229      ecode += 3;      ecode += 1 + IMM2_SIZE;
3230      goto REPEATCHAR;      goto REPEATCHAR;
3231    
3232      case OP_POSUPTO:      case OP_POSUPTO:
3233        case OP_POSUPTOI:
3234      possessive = TRUE;      possessive = TRUE;
3235      /* Fall through */      /* Fall through */
3236    
3237      case OP_UPTO:      case OP_UPTO:
3238        case OP_UPTOI:
3239      case OP_MINUPTO:      case OP_MINUPTO:
3240        case OP_MINUPTOI:
3241      min = 0;      min = 0;
3242      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3243      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3244      ecode += 3;      ecode += 1 + IMM2_SIZE;
3245      goto REPEATCHAR;      goto REPEATCHAR;
3246    
3247      case OP_POSSTAR:      case OP_POSSTAR:
3248        case OP_POSSTARI:
3249      possessive = TRUE;      possessive = TRUE;
3250      min = 0;      min = 0;
3251      max = INT_MAX;      max = INT_MAX;
# Line 2775  for (;;) Line 3253  for (;;)
3253      goto REPEATCHAR;      goto REPEATCHAR;
3254    
3255      case OP_POSPLUS:      case OP_POSPLUS:
3256        case OP_POSPLUSI:
3257      possessive = TRUE;      possessive = TRUE;
3258      min = 1;      min = 1;
3259      max = INT_MAX;      max = INT_MAX;
# Line 2782  for (;;) Line 3261  for (;;)
3261      goto REPEATCHAR;      goto REPEATCHAR;
3262    
3263      case OP_POSQUERY:      case OP_POSQUERY:
3264        case OP_POSQUERYI:
3265      possessive = TRUE;      possessive = TRUE;
3266      min = 0;      min = 0;
3267      max = 1;      max = 1;
# Line 2789  for (;;) Line 3269  for (;;)
3269      goto REPEATCHAR;      goto REPEATCHAR;
3270    
3271      case OP_STAR:      case OP_STAR:
3272        case OP_STARI:
3273      case OP_MINSTAR:      case OP_MINSTAR:
3274        case OP_MINSTARI:
3275      case OP_PLUS:      case OP_PLUS:
3276        case OP_PLUSI:
3277      case OP_MINPLUS:      case OP_MINPLUS:
3278        case OP_MINPLUSI:
3279      case OP_QUERY:      case OP_QUERY:
3280        case OP_QUERYI:
3281      case OP_MINQUERY:      case OP_MINQUERY:
3282      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3283        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3284      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
   
3285      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3286      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3287      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
# Line 2804  for (;;) Line 3289  for (;;)
3289      /* Common code for all repeated single-character matches. */      /* Common code for all repeated single-character matches. */
3290    
3291      REPEATCHAR:      REPEATCHAR:
3292  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3293      if (utf8)      if (utf)
3294        {        {
3295        length = 1;        length = 1;
3296        charptr = ecode;        charptr = ecode;
# Line 2819  for (;;) Line 3304  for (;;)
3304          {          {
3305  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3306          unsigned int othercase;          unsigned int othercase;
3307          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3308              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3309            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3310          else oclength = 0;          else oclength = 0;
3311  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3312    
3313          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3314            {            {
3315            if (eptr <= md->end_subject - length &&            if (eptr <= md->end_subject - length &&
3316              memcmp(eptr, charptr, length) == 0) eptr += length;              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3317  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3318            else if (oclength > 0 &&            else if (oclength > 0 &&
3319                     eptr <= md->end_subject - oclength &&                     eptr <= md->end_subject - oclength &&
3320                     memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                     memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3321  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3322            else            else
3323              {              {
3324              CHECK_PARTIAL();              CHECK_PARTIAL();
3325              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3326              }              }
3327            }            }
3328    
# Line 2847  for (;;) Line 3332  for (;;)
3332            {            {
3333            for (fi = min;; fi++)            for (fi = min;; fi++)
3334              {              {
3335              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3336              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3337              if (fi >= max) MRRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3338              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
3339                memcmp(eptr, charptr, length) == 0) eptr += length;                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3340  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3341              else if (oclength > 0 &&              else if (oclength > 0 &&
3342                       eptr <= md->end_subject - oclength &&                       eptr <= md->end_subject - oclength &&
3343                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3344  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3345              else              else
3346                {                {
3347                CHECK_PARTIAL();                CHECK_PARTIAL();
3348                MRRETURN(MATCH_NOMATCH);                RRETURN(MATCH_NOMATCH);
3349                }                }
3350              }              }
3351            /* Control never gets here */            /* Control never gets here */
# Line 2872  for (;;) Line 3357  for (;;)
3357            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3358              {              {
3359              if (eptr <= md->end_subject - length &&              if (eptr <= md->end_subject - length &&
3360                  memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3361  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3362              else if (oclength > 0 &&              else if (oclength > 0 &&
3363                       eptr <= md->end_subject - oclength &&                       eptr <= md->end_subject - oclength &&
3364                       memcmp(eptr, occhars, oclength) == 0) eptr += oclength;                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3365  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3366              else              else
3367                {                {
# Line 2889  for (;;) Line 3374  for (;;)
3374    
3375            for(;;)            for(;;)
3376              {              {
3377              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3378              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3379              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3380  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3381              eptr--;              eptr--;
3382              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2908  for (;;) Line 3393  for (;;)
3393        value of fc will always be < 128. */        value of fc will always be < 128. */
3394        }        }
3395      else      else
3396  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3397          /* When not in UTF-8 mode, load a single-byte character. */
3398          fc = *ecode++;
3399    
3400      /* 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
3401        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  
3402      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3403      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3404      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2924  for (;;) Line 3407  for (;;)
3407      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3408    
3409      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3410        max, eptr));        max, (char *)eptr));
3411    
3412      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3413        {        {
3414        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3415          /* fc must be < 128 if UTF is enabled. */
3416          foc = md->fcc[fc];
3417    #else
3418    #ifdef SUPPORT_UTF
3419    #ifdef SUPPORT_UCP
3420          if (utf && fc > 127)
3421            foc = UCD_OTHERCASE(fc);
3422    #else
3423          if (utf && fc > 127)
3424            foc = fc;
3425    #endif /* SUPPORT_UCP */
3426          else
3427    #endif /* SUPPORT_UTF */
3428            foc = TABLE_GET(fc, md->fcc, fc);
3429    #endif /* COMPILE_PCRE8 */
3430    
3431        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3432          {          {
3433          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
3434            {            {
3435            SCHECK_PARTIAL();            SCHECK_PARTIAL();
3436            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3437            }            }
3438          if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);          if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3439            eptr++;
3440          }          }
3441        if (min == max) continue;        if (min == max) continue;
3442        if (minimize)        if (minimize)
3443          {          {
3444          for (fi = min;; fi++)          for (fi = min;; fi++)
3445            {            {
3446            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3447            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3448            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3449            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3450              {              {
3451              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3452              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3453              }              }
3454            if (fc != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3455              eptr++;
3456            }            }
3457          /* Control never gets here */          /* Control never gets here */
3458          }          }
# Line 2965  for (;;) Line 3466  for (;;)
3466              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3467              break;              break;
3468              }              }
3469            if (fc != md->lcc[*eptr]) break;            if (fc != *eptr && foc != *eptr) break;
3470            eptr++;            eptr++;
3471            }            }
3472    
# Line 2973  for (;;) Line 3474  for (;;)
3474    
3475          while (eptr >= pp)          while (eptr >= pp)
3476            {            {
3477            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3478            eptr--;            eptr--;
3479            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3480            }            }
3481          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3482          }          }
3483        /* Control never gets here */        /* Control never gets here */
3484        }        }
# Line 2991  for (;;) Line 3492  for (;;)
3492          if (eptr >= md->end_subject)          if (eptr >= md->end_subject)
3493            {            {
3494            SCHECK_PARTIAL();            SCHECK_PARTIAL();
3495            MRRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
3496            }            }
3497          if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);          if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3498          }          }
3499    
3500        if (min == max) continue;        if (min == max) continue;
# Line 3002  for (;;) Line 3503  for (;;)
3503          {          {
3504          for (fi = min;; fi++)          for (fi = min;; fi++)
3505            {            {
3506            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3507            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3508            if (fi >= max) MRRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3509            if (eptr >= md->end_subject)            if (eptr >= md->end_subject)
3510              {              {
3511              SCHECK_PARTIAL();              SCHECK_PARTIAL();
3512              MRRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3513              }              }
3514            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3515            }            }
3516          /* Control never gets here */          /* Control never gets here */
3517          }          }
# Line 3031  for (;;) Line 3532  for (;;)
3532    
3533          while (eptr >= pp)          while (eptr >= pp)
3534            {            {
3535            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3536            eptr--;            eptr--;
3537            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3538            }            }
3539          MRRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3540          }          }
3541        }        }
3542      /* Control never gets here */      /* Control never gets here */
# Line 3044  for (;;) Line 3545  for (;;)
3545      checking can be multibyte. */      checking can be multibyte. */
3546    
3547      case OP_NOT:      case OP_NOT:
3548        case OP_NOTI:
3549      if (eptr >= md->end_subject)      if (eptr >= md->end_subject)
3550        {        {
3551        SCHECK_PARTIAL();        SCHECK_PARTIAL();
3552        MRRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
3553        }        }
3554      ecode++;  #ifdef SUPPORT_UTF
3555      GETCHARINCTEST(c, eptr);      if (utf)
     if ((ims & PCRE_CASELESS) != 0)  
3556        {        {
3557  #ifdef SUPPORT_UTF8        register unsigned int ch, och;
3558        if (c < 256)  
3559  #endif        ecode++;
3560        c = md->lcc[c];        GETCHARINC(ch, ecode);
3561        if (md->lcc[*ecode++] == c) MRRETURN(MATCH_NOMATCH);        GETCHARINC(c, eptr);
3562    
3563          if (op == OP_NOT)
3564            {
3565            if (ch == c) RRETURN(MATCH_NOMATCH);
3566            }
3567          else
3568            {
3569    #ifdef SUPPORT_UCP
3570            if (ch > 127)
3571              och = UCD_OTHERCASE(ch);
3572    #else
3573            if (ch > 127)
3574              och = ch;
3575    #endif /* SUPPORT_UCP */
3576            else
3577              och = TABLE_GET(ch, md->fcc, ch);
3578            if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3579            }
3580        }        }
3581      else      else
3582    #endif
3583        {        {
3584        if (*ecode++ == c) MRRETURN(MATCH_NOMATCH);        register unsigned int ch = ecode[1];
3585          c = *eptr++;
3586          if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3587            RRETURN(MATCH_NOMATCH);
3588          ecode += 2;
3589        }        }
3590      break;      break;
3591    
# Line 3073  for (;;) Line 3597  for (;;)
3597      about... */      about... */
3598    
3599      case OP_NOTEXACT:      case OP_NOTEXACT:
3600        case OP_NOTEXACTI:
3601      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3602      ecode += 3;      ecode += 1 + IMM2_SIZE;
3603      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3604    
3605      case OP_NOTUPTO:      case OP_NOTUPTO:
3606        case OP_NOTUPTOI:
3607      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3608        case OP_NOTMINUPTOI:
3609      min = 0;      min = 0;
3610      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3611      minimize = *ecode == OP_NOTMINUPTO;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3612      ecode += 3;      ecode += 1 + IMM2_SIZE;
3613      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3614    
3615      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
3616        case OP_NOTPOSSTARI:
3617      possessive = TRUE;      possessive = TRUE;
3618      min = 0;      min = 0;
3619      max = INT_MAX;      max = INT_MAX;
# Line 3093  for (;;) Line 3621  for (;;)
3621      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3622    
3623      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
3624        case OP_NOTPOSPLUSI:
3625      possessive = TRUE;      possessive = TRUE;
3626      min = 1;      min = 1;
3627      max = INT_MAX;      max = INT_MAX;
# Line 3100  for (;;) Line 3629  for (;;)
3629      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3630    
3631      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
3632        case OP_NOTPOSQUERYI:
3633      possessive = TRUE;      possessive = TRUE;
3634      min = 0;      min = 0;
3635      max = 1;      max = 1;
# Line 3107  for (;;) Line 3637  for (;;)
3637      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3638    
3639      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
3640        case OP_NOTPOSUPTOI:
3641      possessive = TRUE;      possessive = TRUE;
3642      min = 0;      min = 0;
3643      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3644      ecode += 3;      ecode += 1 + IMM2_SIZE;
3645      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3646    
3647      case OP_NOTSTAR:      case OP_NOTSTAR:
3648        case OP_NOTSTARI:
3649