/[pcre]/code/branches/pcre16/pcre_exec.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_exec.c

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

code/trunk/pcre_exec.c revision 406 by ph10, Mon Mar 23 12:05:43 2009 UTC code/branches/pcre16/pcre_exec.c revision 795 by zherczeg, Sat Dec 10 02:20:06 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2009 University of Cambridge             Copyright (c) 1997-2011 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 57  possible. There are also some static sup Line 57  possible. There are also some static sup
57  #undef min  #undef min
58  #undef max  #undef max
59    
60  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
61    of call to match(). We do it this way to save on using another stack variable,
62    as stack usage is to be discouraged. */
63    
64  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
65  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
66    
67  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
68  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 71  defined PCRE_ERROR_xxx codes, which are Line 73  defined PCRE_ERROR_xxx codes, which are
73  /* Special internal returns from the match() function. Make them sufficiently  /* Special internal returns from the match() function. Make them sufficiently
74  negative to avoid the external error codes. */  negative to avoid the external error codes. */
75    
76  #define MATCH_COMMIT       (-999)  #define MATCH_ACCEPT       (-999)
77  #define MATCH_PRUNE        (-998)  #define MATCH_COMMIT       (-998)
78  #define MATCH_SKIP         (-997)  #define MATCH_KETRPOS      (-997)
79  #define MATCH_THEN         (-996)  #define MATCH_ONCE         (-996)
80    #define MATCH_PRUNE        (-995)
81    #define MATCH_SKIP         (-994)
82    #define MATCH_SKIP_ARG     (-993)
83    #define MATCH_THEN         (-992)
84    
85    /* This is a convenience macro for code that occurs many times. */
86    
87    #define MRRETURN(ra) \
88      { \
89      md->mark = markptr; \
90      RRETURN(ra); \
91      }
92    
93  /* 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.
94  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 89  static const char rep_max[] = { 0, 0, 0, Line 103  static const char rep_max[] = { 0, 0, 0,
103    
104    
105    
106  #ifdef DEBUG  #ifdef PCRE_DEBUG
107  /*************************************************  /*************************************************
108  *        Debugging function to print chars       *  *        Debugging function to print chars       *
109  *************************************************/  *************************************************/
# Line 107  Returns:     nothing Line 121  Returns:     nothing
121  */  */
122    
123  static void  static void
124  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)
125  {  {
126  unsigned int c;  unsigned int c;
127  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 122  while (length-- > 0) Line 136  while (length-- > 0)
136  *          Match a back-reference                *  *          Match a back-reference                *
137  *************************************************/  *************************************************/
138    
139  /* 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
140  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
141    the length passed is zero. Note that in caseless UTF-8 mode, the number of
142    subject bytes matched may be different to the number of reference bytes.
143    
144  Arguments:  Arguments:
145    offset      index into the offset vector    offset      index into the offset vector
146    eptr        points into the subject    eptr        pointer into the subject
147    length      length to be matched    length      length of reference to be matched (number of bytes)
148    md          points to match data block    md          points to match data block
149    ims         the ims flags    caseless    TRUE if caseless
150    
151  Returns:      TRUE if matched  Returns:      < 0 if not matched, otherwise the number of subject bytes matched
152  */  */
153    
154  static BOOL  static int
155  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,
156    unsigned long int ims)    BOOL caseless)
157  {  {
158  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
159    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
160    
161  #ifdef DEBUG  #ifdef PCRE_DEBUG
162  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
163    printf("matching subject <null>");    printf("matching subject <null>");
164  else  else
# Line 154  pchars(p, length, FALSE, md); Line 171  pchars(p, length, FALSE, md);
171  printf("\n");  printf("\n");
172  #endif  #endif
173    
174  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible). */
175    
176  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
177    
178  /* 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
179  properly if Unicode properties are supported. Otherwise, we can check only  properly if Unicode properties are supported. Otherwise, we can check only
180  ASCII characters. */  ASCII characters. */
181    
182  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
183    {    {
184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
185  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
186    if (md->utf8)    if (md->utf)
187      {      {
188      USPTR endptr = eptr + length;      /* Match characters up to the end of the reference. NOTE: the number of
189      while (eptr < endptr)      bytes matched may differ, because there are some characters whose upper and
190        lower case versions code as different numbers of bytes. For example, U+023A
191        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
192        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
193        the latter. It is important, therefore, to check the length along the
194        reference, not along the subject (earlier code did this wrong). */
195    
196        PCRE_PUCHAR endptr = p + length;
197        while (p < endptr)
198        {        {
199        int c, d;        int c, d;
200          if (eptr >= md->end_subject) return -1;
201        GETCHARINC(c, eptr);        GETCHARINC(c, eptr);
202        GETCHARINC(d, p);        GETCHARINC(d, p);
203        if (c != d && c != UCD_OTHERCASE(d)) return FALSE;        if (c != d && c != UCD_OTHERCASE(d)) return -1;
204        }        }
205      }      }
206    else    else
# Line 183  if ((ims & PCRE_CASELESS) != 0) Line 209  if ((ims & PCRE_CASELESS) != 0)
209    
210    /* 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
211    is no UCP support. */    is no UCP support. */
212        {
213    while (length-- > 0)      if (eptr + length > md->end_subject) return -1;
214      { if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE; }      while (length-- > 0)
215          { if (md->lcc[*p++] != md->lcc[*eptr++]) return -1; }
216        }
217    }    }
218    
219  /* 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
220  are in UTF-8 mode. */  are in UTF-8 mode. */
221    
222  else  else
223    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
224      if (eptr + length > md->end_subject) return -1;
225      while (length-- > 0) if (*p++ != *eptr++) return -1;
226      }
227    
228  return TRUE;  return eptr - eptr_start;
229  }  }
230    
231    
# Line 245  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 276  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
276         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
277         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
278         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
279         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
280           RM61,  RM62, RM63, RM64, RM65, RM66 };
281    
282  /* 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
283  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
284  actuall used in this definition. */  actually used in this definition. */
285    
286  #ifndef NO_RECURSE  #ifndef NO_RECURSE
287  #define REGISTER register  #define REGISTER register
288    
289  #ifdef DEBUG  #ifdef PCRE_DEBUG
290  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
291    { \    { \
292    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
293    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1); \
294    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
295    }    }
296  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 267  actuall used in this definition. */ Line 299  actuall used in this definition. */
299    return ra; \    return ra; \
300    }    }
301  #else  #else
302  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
303    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,markptr,rc,rd,re,rdepth+1)
304  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
305  #endif  #endif
306    
# Line 281  argument of match(), which never changes Line 313  argument of match(), which never changes
313    
314  #define REGISTER  #define REGISTER
315    
316  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
317    {\    {\
318    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));\
319      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
320    frame->Xwhere = rw; \    frame->Xwhere = rw; \
321    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
322    newframe->Xecode = rb;\    newframe->Xecode = rb;\
323    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
324      newframe->Xmarkptr = markptr;\
325    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
326    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
327    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
328    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
329    frame = newframe;\    frame = newframe;\
# Line 303  argument of match(), which never changes Line 335  argument of match(), which never changes
335    
336  #define RRETURN(ra)\  #define RRETURN(ra)\
337    {\    {\
338    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
339    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
340    (pcre_stack_free)(newframe);\    (pcre_stack_free)(oldframe);\
341    if (frame != NULL)\    if (frame != NULL)\
342      {\      {\
343      rrc = ra;\      rrc = ra;\
# Line 322  typedef struct heapframe { Line 354  typedef struct heapframe {
354    
355    /* Function arguments that may change */    /* Function arguments that may change */
356    
357    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
358    const uschar *Xecode;    const pcre_uchar *Xecode;
359    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
360      PCRE_PUCHAR Xmarkptr;
361    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
362    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
363    unsigned int Xrdepth;    unsigned int Xrdepth;
364    
365    /* Function local variables */    /* Function local variables */
366    
367    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
368  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
369    const uschar *Xcharptr;    PCRE_PUCHAR Xcharptr;
370  #endif  #endif
371    const uschar *Xdata;    PCRE_PUCHAR Xdata;
372    const uschar *Xnext;    PCRE_PUCHAR Xnext;
373    const uschar *Xpp;    PCRE_PUCHAR Xpp;
374    const uschar *Xprev;    PCRE_PUCHAR Xprev;
375    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xsaved_eptr;
376    
377    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
378    
# Line 349  typedef struct heapframe { Line 380  typedef struct heapframe {
380    BOOL Xcondition;    BOOL Xcondition;
381    BOOL Xprev_is_word;    BOOL Xprev_is_word;
382    
   unsigned long int Xoriginal_ims;  
   
383  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
384    int Xprop_type;    int Xprop_type;
385    int Xprop_value;    int Xprop_value;
386    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
387    int Xoclength;    int Xoclength;
388    uschar Xocchars[8];    pcre_uchar Xocchars[6];
389  #endif  #endif
390    
391    int Xcodelink;    int Xcodelink;
# Line 398  typedef struct heapframe { Line 424  typedef struct heapframe {
424    
425  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
426  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
427  same response.  same response. */
428    
429    /* These macros pack up tests that are used for partial matching, and which
430    appears several times in the code. We set the "hit end" flag if the pointer is
431    at the end of the subject and also past the start of the subject (i.e.
432    something has been matched). For hard partial matching, we then return
433    immediately. The second one is used when we already know we are past the end of
434    the subject. */
435    
436    #define CHECK_PARTIAL()\
437      if (md->partial != 0 && eptr >= md->end_subject && \
438          eptr > md->start_used_ptr) \
439        { \
440        md->hitend = TRUE; \
441        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
442        }
443    
444  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
445  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
446        { \
447        md->hitend = TRUE; \
448        if (md->partial > 1) MRRETURN(PCRE_ERROR_PARTIAL); \
449        }
450    
451    
452    /* Performance note: It might be tempting to extract commonly used fields from
453    the md structure (e.g. utf, end_subject) into individual variables to improve
454  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
455  made performance worse.  made performance worse.
456    
# Line 410  Arguments: Line 459  Arguments:
459     ecode       pointer to current position in compiled code     ecode       pointer to current position in compiled code
460     mstart      pointer to the current match start position (can be modified     mstart      pointer to the current match start position (can be modified
461                   by encountering \K)                   by encountering \K)
462       markptr     pointer to the most recent MARK name, or NULL
463     offset_top  current top pointer     offset_top  current top pointer
464     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
465     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
466                   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  
467     rdepth      the recursion depth     rdepth      the recursion depth
468    
469  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
470                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
471                   a negative MATCH_xxx value for PRUNE, SKIP, etc
472                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
473                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
474  */  */
475    
476  static int  static int
477  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
478    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, const pcre_uchar *markptr, int offset_top,
479    int flags, unsigned int rdepth)    match_data *md, eptrblock *eptrb, unsigned int rdepth)
480  {  {
481  /* 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,
482  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 439  so they can be ordinary variables in all Line 485  so they can be ordinary variables in all
485  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
486  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
487  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
488  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
489    
490  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
491    BOOL caseless;
492  int condcode;  int condcode;
493    
494  /* 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
# Line 450  heap storage. Set up the top-level frame Line 497  heap storage. Set up the top-level frame
497  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  heap whenever RMATCH() does a "recursion". See the macro definitions above. */
498    
499  #ifdef NO_RECURSE  #ifdef NO_RECURSE
500  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)(pcre_stack_malloc)(sizeof(heapframe));
501    if (frame == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
502  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
503    
504  /* Copy in the original argument variables */  /* Copy in the original argument variables */
# Line 458  frame->Xprevframe = NULL;            /* Line 506  frame->Xprevframe = NULL;            /*
506  frame->Xeptr = eptr;  frame->Xeptr = eptr;
507  frame->Xecode = ecode;  frame->Xecode = ecode;
508  frame->Xmstart = mstart;  frame->Xmstart = mstart;
509    frame->Xmarkptr = markptr;
510  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
511  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
512  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
513    
514  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 473  HEAP_RECURSE: Line 520  HEAP_RECURSE:
520  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
521  #define ecode              frame->Xecode  #define ecode              frame->Xecode
522  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
523    #define markptr            frame->Xmarkptr
524  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
525  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
526  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
527    
528  /* Ditto for the local variables */  /* Ditto for the local variables */
529    
530  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
531  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
532  #endif  #endif
533  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
# Line 498  HEAP_RECURSE: Line 544  HEAP_RECURSE:
544  #define condition          frame->Xcondition  #define condition          frame->Xcondition
545  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
546    
 #define original_ims       frame->Xoriginal_ims  
   
547  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
548  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
549  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
550  #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  
551  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
552  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
553  #endif  #endif
# Line 536  i, and fc and c, can be the same variabl Line 577  i, and fc and c, can be the same variabl
577  #define fi i  #define fi i
578  #define fc c  #define fc c
579    
580    /* Many of the following variables are used only in small blocks of the code.
581    My normal style of coding would have declared them within each of those blocks.
582    However, in order to accommodate the version of this code that uses an external
583    "stack" implemented on the heap, it is easier to declare them all here, so the
584    declarations can be cut out in a block. The only declarations within blocks
585    below are for variables that do not have to be preserved over a recursive call
586    to RMATCH(). */
587    
588    #ifdef SUPPORT_UTF
589    const pcre_uchar *charptr;
590    #endif
591    const pcre_uchar *callpat;
592    const pcre_uchar *data;
593    const pcre_uchar *next;
594    PCRE_PUCHAR       pp;
595    const pcre_uchar *prev;
596    PCRE_PUCHAR       saved_eptr;
597    
598  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
599  const uschar *charptr;             /* in small blocks of the code. My normal */  
600  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
601  BOOL condition;  BOOL condition;
602  BOOL prev_is_word;  BOOL prev_is_word;
603    
 unsigned long int original_ims;  
   
604  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
605  int prop_type;  int prop_type;
606  int prop_value;  int prop_value;
607  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
608  int oclength;  int oclength;
609  uschar occhars[8];  pcre_uchar occhars[6];
610  #endif  #endif
611    
612  int codelink;  int codelink;
# Line 581  int stacksave[REC_STACK_SAVE_MAX]; Line 624  int stacksave[REC_STACK_SAVE_MAX];
624  eptrblock newptrb;  eptrblock newptrb;
625  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
626    
627    /* To save space on the stack and in the heap frame, I have doubled up on some
628    of the local variables that are used only in localised parts of the code, but
629    still need to be preserved over recursive calls of match(). These macros define
630    the alternative names that are used. */
631    
632    #define allow_zero    cur_is_word
633    #define cbegroup      condition
634    #define code_offset   codelink
635    #define condassert    condition
636    #define matched_once  prev_is_word
637    #define foc           number
638    
639  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
640  variables. */  variables. */
641    
# Line 600  TAIL_RECURSE: Line 655  TAIL_RECURSE:
655  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
656  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
657  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
658  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
659  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
660  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,
661  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
662    
663  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
664  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
665  #else  #else
666  utf8 = FALSE;  utf = FALSE;
667  #endif  #endif
668    
669  /* 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 617  haven't exceeded the recursive call limi Line 672  haven't exceeded the recursive call limi
672  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
673  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
674    
 original_ims = ims;    /* Save for resetting on ')' */  
   
675  /* 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
676  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
677  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
678  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
679  When match() is called in other circumstances, don't add to the chain. The  
680  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
681  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
682  match(). */  to break infinite loops that match no characters. When match() is called in
683    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
684    NOT be used with tail recursion, because the memory block that is used is on
685    the stack, so a new one may be required for each match(). */
686    
687  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
688    {    {
689    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
690    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
691    eptrb = &newptrb;    eptrb = &newptrb;
692      md->match_function_type = 0;
693    }    }
694    
695  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 642  for (;;) Line 699  for (;;)
699    minimize = possessive = FALSE;    minimize = possessive = FALSE;
700    op = *ecode;    op = *ecode;
701    
   /* For partial matching, remember if we ever hit the end of the subject after  
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > mstart)  
     md->hitend = TRUE;  
   
702    switch(op)    switch(op)
703      {      {
704        case OP_MARK:
705        markptr = ecode + 2;
706        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
707          eptrb, RM55);
708    
709        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
710        argument, and we must check whether that argument matches this MARK's
711        argument. It is passed back in md->start_match_ptr (an overloading of that
712        variable). If it does match, we reset that variable to the current subject
713        position and return MATCH_SKIP. Otherwise, pass back the return code
714        unaltered. */
715    
716        if (rrc == MATCH_SKIP_ARG &&
717            STRCMP_UC_UC(markptr, md->start_match_ptr) == 0)
718          {
719          md->start_match_ptr = eptr;
720          RRETURN(MATCH_SKIP);
721          }
722    
723        if (md->mark == NULL) md->mark = markptr;
724        RRETURN(rrc);
725    
726      case OP_FAIL:      case OP_FAIL:
727      RRETURN(MATCH_NOMATCH);      MRRETURN(MATCH_NOMATCH);
728    
729        /* COMMIT overrides PRUNE, SKIP, and THEN */
730    
731        case OP_COMMIT:
732        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
733          eptrb, RM52);
734        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
735            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
736            rrc != MATCH_THEN)
737          RRETURN(rrc);
738        MRRETURN(MATCH_COMMIT);
739    
740        /* PRUNE overrides THEN */
741    
742      case OP_PRUNE:      case OP_PRUNE:
743      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
744        ims, eptrb, flags, RM51);        eptrb, RM51);
745      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
746        MRRETURN(MATCH_PRUNE);
747    
748        case OP_PRUNE_ARG:
749        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
750          eptrb, RM56);
751        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
752        md->mark = ecode + 2;
753      RRETURN(MATCH_PRUNE);      RRETURN(MATCH_PRUNE);
754    
755      case OP_COMMIT:      /* SKIP overrides PRUNE and THEN */
     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,  
       ims, eptrb, flags, RM52);  
     if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
     RRETURN(MATCH_COMMIT);  
756    
757      case OP_SKIP:      case OP_SKIP:
758      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
759        ims, eptrb, flags, RM53);        eptrb, RM53);
760      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
761          RRETURN(rrc);
762      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
763      RRETURN(MATCH_SKIP);      MRRETURN(MATCH_SKIP);
764    
765        case OP_SKIP_ARG:
766        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
767          eptrb, RM57);
768        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
769          RRETURN(rrc);
770    
771        /* Pass back the current skip name by overloading md->start_match_ptr and
772        returning the special MATCH_SKIP_ARG return code. This will either be
773        caught by a matching MARK, or get to the top, where it is treated the same
774        as PRUNE. */
775    
776        md->start_match_ptr = ecode + 2;
777        RRETURN(MATCH_SKIP_ARG);
778    
779        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
780        the branch in which it occurs can be determined. Overload the start of
781        match pointer to do this. */
782    
783      case OP_THEN:      case OP_THEN:
784      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
785        ims, eptrb, flags, RM54);        eptrb, RM54);
786      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
787        md->start_match_ptr = ecode;
788        MRRETURN(MATCH_THEN);
789    
790        case OP_THEN_ARG:
791        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
792          md, eptrb, RM58);
793        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794        md->start_match_ptr = ecode;
795        md->mark = ecode + 2;
796      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
797    
798      /* Handle a capturing bracket. If there is space in the offset vector, save      /* Handle an atomic group that does not contain any capturing parentheses.
799      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
800      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
801      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.
802      reference inside the group.      However, this uses a lot more stack, so in 8.20, atomic groups that do not
803        contain any captures generate OP_ONCE_NC, which can be handled in the old,
804      If the bracket fails to match, we need to restore this value and also the      less stack intensive way.
805      values of the final offsets, in case they were set by a previous iteration  
806      of the same bracket.      Check the alternative branches in turn - the matching won't pass the KET
807        for this kind of subpattern. If any one branch matches, we carry on as at
808        the end of a normal bracket, leaving the subject pointer, but resetting
809        the start-of-match value in case it was changed by \K. */
810    
811        case OP_ONCE_NC:
812        prev = ecode;
813        saved_eptr = eptr;
814        do
815          {
816          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
817          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
818            {
819            mstart = md->start_match_ptr;
820            markptr = md->mark;
821            break;
822            }
823          if (rrc == MATCH_THEN)
824            {
825            next = ecode + GET(ecode,1);
826            if (md->start_match_ptr < next &&
827                (*ecode == OP_ALT || *next == OP_ALT))
828              rrc = MATCH_NOMATCH;
829            }
830    
831          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
832          ecode += GET(ecode,1);
833          }
834        while (*ecode == OP_ALT);
835    
836        /* If hit the end of the group (which could be repeated), fail */
837    
838        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
839    
840        /* Continue as from after the group, updating the offsets high water
841        mark, since extracts may have been taken. */
842    
843        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
844    
845        offset_top = md->end_offset_top;
846        eptr = md->end_match_ptr;
847    
848        /* For a non-repeating ket, just continue at this level. This also
849        happens for a repeating ket if no characters were matched in the group.
850        This is the forcible breaking of infinite loops as implemented in Perl
851        5.005. */
852    
853        if (*ecode == OP_KET || eptr == saved_eptr)
854          {
855          ecode += 1+LINK_SIZE;
856          break;
857          }
858    
859        /* The repeating kets try the rest of the pattern or restart from the
860        preceding bracket, in the appropriate order. The second "call" of match()
861        uses tail recursion, to avoid using another stack frame. */
862    
863        if (*ecode == OP_KETRMIN)
864          {
865          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
866          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
867          ecode = prev;
868          goto TAIL_RECURSE;
869          }
870        else  /* OP_KETRMAX */
871          {
872          md->match_function_type = MATCH_CBEGROUP;
873          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
874          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
875          ecode += 1 + LINK_SIZE;
876          goto TAIL_RECURSE;
877          }
878        /* Control never gets here */
879    
880        /* Handle a capturing bracket, other than those that are possessive with an
881        unlimited repeat. If there is space in the offset vector, save the current
882        subject position in the working slot at the top of the vector. We mustn't
883        change the current values of the data slot, because they may be set from a
884        previous iteration of this group, and be referred to by a reference inside
885        the group. A failure to match might occur after the group has succeeded,
886        if something later on doesn't match. For this reason, we need to restore
887        the working value and also the values of the final offsets, in case they
888        were set by a previous iteration of the same bracket.
889    
890      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
891      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 699  for (;;) Line 896  for (;;)
896      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
897      offset = number << 1;      offset = number << 1;
898    
899  #ifdef DEBUG  #ifdef PCRE_DEBUG
900      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
901      printf("subject=");      printf("subject=");
902      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 714  for (;;) Line 911  for (;;)
911        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
912    
913        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
914        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
915            (int)(eptr - md->start_subject);
916    
917        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
918          {          {
919          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
920            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
921          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
922            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
923    
924            /* If we backed up to a THEN, check whether it is within the current
925            branch by comparing the address of the THEN that is passed back with
926            the end of the branch. If it is within the current branch, and the
927            branch is one of two or more alternatives (it either starts or ends
928            with OP_ALT), we have reached the limit of THEN's action, so convert
929            the return code to NOMATCH, which will cause normal backtracking to
930            happen from now on. Otherwise, THEN is passed back to an outer
931            alternative. This implements Perl's treatment of parenthesized groups,
932            where a group not containing | does not affect the current alternative,
933            that is, (X) is NOT the same as (X|(*F)). */
934    
935            if (rrc == MATCH_THEN)
936              {
937              next = ecode + GET(ecode,1);
938              if (md->start_match_ptr < next &&
939                  (*ecode == OP_ALT || *next == OP_ALT))
940                rrc = MATCH_NOMATCH;
941              }
942    
943            /* Anything other than NOMATCH is passed back. */
944    
945            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
946          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
947          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
948            if (*ecode != OP_ALT) break;
949          }          }
       while (*ecode == OP_ALT);  
950    
951        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
952        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
953        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
954        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
955    
956        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
957    
958          if (md->mark == NULL) md->mark = markptr;
959          RRETURN(rrc);
960        }        }
961    
962      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 747  for (;;) Line 970  for (;;)
970      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
971      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
972    
973      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
974      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
975      recursive call to match() whatever happened. We can reduce stack usage by  
976      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
977      is set.*/      the result of a recursive call to match() whatever happened so it was
978        possible to reduce stack usage by turning this into a tail recursion,
979        except in the case of a possibly empty group. However, now that there is
980        the possiblity of (*THEN) occurring in the final alternative, this
981        optimization is no longer always possible.
982    
983        We can optimize if we know there are no (*THEN)s in the pattern; at present
984        this is the best that can be done.
985    
986        MATCH_ONCE is returned when the end of an atomic group is successfully
987        reached, but subsequent matching fails. It passes back up the tree (causing
988        captured values to be reset) until the original atomic group level is
989        reached. This is tested by comparing md->once_target with the start of the
990        group. At this point, the return is converted into MATCH_NOMATCH so that
991        previous backup points can be taken. */
992    
993        case OP_ONCE:
994      case OP_BRA:      case OP_BRA:
995      case OP_SBRA:      case OP_SBRA:
996      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
997      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
998      for (;;)      for (;;)
999        {        {
1000        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1001    
1002          /* If this is not a possibly empty group, and there are no (*THEN)s in
1003          the pattern, and this is the final alternative, optimize as described
1004          above. */
1005    
1006          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1007            {
1008            ecode += PRIV(OP_lengths)[*ecode];
1009            goto TAIL_RECURSE;
1010            }
1011    
1012          /* In all other cases, we have to make another call to match(). */
1013    
1014          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1015            RM2);
1016    
1017          /* See comment in the code for capturing groups above about handling
1018          THEN. */
1019    
1020          if (rrc == MATCH_THEN)
1021          {          {
1022          if (flags == 0)    /* Not a possibly empty group */          next = ecode + GET(ecode,1);
1023            if (md->start_match_ptr < next &&
1024                (*ecode == OP_ALT || *next == OP_ALT))
1025              rrc = MATCH_NOMATCH;
1026            }
1027    
1028          if (rrc != MATCH_NOMATCH)
1029            {
1030            if (rrc == MATCH_ONCE)
1031            {            {
1032            ecode += _pcre_OP_lengths[*ecode];            const pcre_uchar *scode = ecode;
1033            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
1034            goto TAIL_RECURSE;              {
1035                while (*scode == OP_ALT) scode += GET(scode, 1);
1036                scode -= GET(scode, 1);
1037                }
1038              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1039            }            }
1040            RRETURN(rrc);
1041            }
1042          ecode += GET(ecode, 1);
1043          if (*ecode != OP_ALT) break;
1044          }
1045    
1046          /* Possibly empty group; can't use tail recursion. */      if (md->mark == NULL) md->mark = markptr;
1047        RRETURN(MATCH_NOMATCH);
1048    
1049          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      /* Handle possessive capturing brackets with an unlimited repeat. We come
1050            eptrb, flags, RM48);      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1051          RRETURN(rrc);      handled similarly to the normal case above. However, the matching is
1052        different. The end of these brackets will always be OP_KETRPOS, which
1053        returns MATCH_KETRPOS without going further in the pattern. By this means
1054        we can handle the group by iteration rather than recursion, thereby
1055        reducing the amount of stack needed. */
1056    
1057        case OP_CBRAPOS:
1058        case OP_SCBRAPOS:
1059        allow_zero = FALSE;
1060    
1061        POSSESSIVE_CAPTURE:
1062        number = GET2(ecode, 1+LINK_SIZE);
1063        offset = number << 1;
1064    
1065    #ifdef PCRE_DEBUG
1066        printf("start possessive bracket %d\n", number);
1067        printf("subject=");
1068        pchars(eptr, 16, TRUE, md);
1069        printf("\n");
1070    #endif
1071    
1072        if (offset < md->offset_max)
1073          {
1074          matched_once = FALSE;
1075          code_offset = ecode - md->start_code;
1076    
1077          save_offset1 = md->offset_vector[offset];
1078          save_offset2 = md->offset_vector[offset+1];
1079          save_offset3 = md->offset_vector[md->offset_end - number];
1080          save_capture_last = md->capture_last;
1081    
1082          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1083    
1084          /* Each time round the loop, save the current subject position for use
1085          when the group matches. For MATCH_MATCH, the group has matched, so we
1086          restart it with a new subject starting position, remembering that we had
1087          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1088          usual. If we haven't matched any alternatives in any iteration, check to
1089          see if a previous iteration matched. If so, the group has matched;
1090          continue from afterwards. Otherwise it has failed; restore the previous
1091          capture values before returning NOMATCH. */
1092    
1093          for (;;)
1094            {
1095            md->offset_vector[md->offset_end - number] =
1096              (int)(eptr - md->start_subject);
1097            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1098            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1099              eptrb, RM63);
1100            if (rrc == MATCH_KETRPOS)
1101              {
1102              offset_top = md->end_offset_top;
1103              eptr = md->end_match_ptr;
1104              ecode = md->start_code + code_offset;
1105              save_capture_last = md->capture_last;
1106              matched_once = TRUE;
1107              continue;
1108              }
1109    
1110            /* See comment in the code for capturing groups above about handling
1111            THEN. */
1112    
1113            if (rrc == MATCH_THEN)
1114              {
1115              next = ecode + GET(ecode,1);
1116              if (md->start_match_ptr < next &&
1117                  (*ecode == OP_ALT || *next == OP_ALT))
1118                rrc = MATCH_NOMATCH;
1119              }
1120    
1121            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1122            md->capture_last = save_capture_last;
1123            ecode += GET(ecode, 1);
1124            if (*ecode != OP_ALT) break;
1125          }          }
1126    
1127        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (!matched_once)
1128        otherwise return. */          {
1129            md->offset_vector[offset] = save_offset1;
1130            md->offset_vector[offset+1] = save_offset2;
1131            md->offset_vector[md->offset_end - number] = save_offset3;
1132            }
1133    
1134        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        if (md->mark == NULL) md->mark = markptr;
1135          eptrb, flags, RM2);        if (allow_zero || matched_once)
1136        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);          {
1137            ecode += 1 + LINK_SIZE;
1138            break;
1139            }
1140    
1141          RRETURN(MATCH_NOMATCH);
1142          }
1143    
1144        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1145        as a non-capturing bracket. */
1146    
1147        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1148        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1149    
1150        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1151    
1152        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1153        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1154    
1155        /* Non-capturing possessive bracket with unlimited repeat. We come here
1156        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1157        without the capturing complication. It is written out separately for speed
1158        and cleanliness. */
1159    
1160        case OP_BRAPOS:
1161        case OP_SBRAPOS:
1162        allow_zero = FALSE;
1163    
1164        POSSESSIVE_NON_CAPTURE:
1165        matched_once = FALSE;
1166        code_offset = ecode - md->start_code;
1167    
1168        for (;;)
1169          {
1170          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1171          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1172            eptrb, RM48);
1173          if (rrc == MATCH_KETRPOS)
1174            {
1175            offset_top = md->end_offset_top;
1176            eptr = md->end_match_ptr;
1177            ecode = md->start_code + code_offset;
1178            matched_once = TRUE;
1179            continue;
1180            }
1181    
1182          /* See comment in the code for capturing groups above about handling
1183          THEN. */
1184    
1185          if (rrc == MATCH_THEN)
1186            {
1187            next = ecode + GET(ecode,1);
1188            if (md->start_match_ptr < next &&
1189                (*ecode == OP_ALT || *next == OP_ALT))
1190              rrc = MATCH_NOMATCH;
1191            }
1192    
1193          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1194        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1195          if (*ecode != OP_ALT) break;
1196        }        }
1197    
1198        if (matched_once || allow_zero)
1199          {
1200          ecode += 1 + LINK_SIZE;
1201          break;
1202          }
1203        RRETURN(MATCH_NOMATCH);
1204    
1205      /* Control never reaches here. */      /* Control never reaches here. */
1206    
1207      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1208      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
1209      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
1210      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. */  
1211    
1212      case OP_COND:      case OP_COND:
1213      case OP_SCOND:      case OP_SCOND:
1214      codelink= GET(ecode, 1);      codelink = GET(ecode, 1);
1215    
1216      /* 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
1217      inserted between OP_COND and an assertion condition. */      inserted between OP_COND and an assertion condition. */
# Line 803  for (;;) Line 1221  for (;;)
1221        if (pcre_callout != NULL)        if (pcre_callout != NULL)
1222          {          {
1223          pcre_callout_block cb;          pcre_callout_block cb;
1224          cb.version          = 1;   /* Version 1 of the callout block */          cb.version          = 2;   /* Version 1 of the callout block */
1225          cb.callout_number   = ecode[LINK_SIZE+2];          cb.callout_number   = ecode[LINK_SIZE+2];
1226          cb.offset_vector    = md->offset_vector;          cb.offset_vector    = md->offset_vector;
1227          cb.subject          = (PCRE_SPTR)md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
1228          cb.subject_length   = md->end_subject - md->start_subject;          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1229          cb.start_match      = mstart - md->start_subject;          cb.start_match      = (int)(mstart - md->start_subject);
1230          cb.current_position = eptr - md->start_subject;          cb.current_position = (int)(eptr - md->start_subject);
1231          cb.pattern_position = GET(ecode, LINK_SIZE + 3);          cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1232          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);          cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1233          cb.capture_top      = offset_top/2;          cb.capture_top      = offset_top/2;
1234          cb.capture_last     = md->capture_last;          cb.capture_last     = md->capture_last;
1235          cb.callout_data     = md->callout_data;          cb.callout_data     = md->callout_data;
1236          if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);          cb.mark             = (unsigned char *)markptr;
1237            if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1238          if (rrc < 0) RRETURN(rrc);          if (rrc < 0) RRETURN(rrc);
1239          }          }
1240        ecode += _pcre_OP_lengths[OP_CALLOUT];        ecode += PRIV(OP_lengths)[OP_CALLOUT];
1241        }        }
1242    
1243      condcode = ecode[LINK_SIZE+1];      condcode = ecode[LINK_SIZE+1];
1244    
1245      /* Now see what the actual condition is */      /* Now see what the actual condition is */
1246    
1247      if (condcode == OP_RREF)         /* Recursion test */      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1248        {        {
1249        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/        if (md->recursive == NULL)                /* Not recursing => FALSE */
1250        condition = md->recursive != NULL &&          {
1251          (offset == RREF_ANY || offset == md->recursive->group_num);          condition = FALSE;
1252        ecode += condition? 3 : GET(ecode, 1);          ecode += GET(ecode, 1);
1253            }
1254          else
1255            {
1256            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1257            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1258    
1259            /* If the test is for recursion into a specific subpattern, and it is
1260            false, but the test was set up by name, scan the table to see if the
1261            name refers to any other numbers, and test them. The condition is true
1262            if any one is set. */
1263    
1264            if (!condition && condcode == OP_NRREF)
1265              {
1266              pcre_uchar *slotA = md->name_table;
1267              for (i = 0; i < md->name_count; i++)
1268                {
1269                if (GET2(slotA, 0) == recno) break;
1270                slotA += md->name_entry_size;
1271                }
1272    
1273              /* Found a name for the number - there can be only one; duplicate
1274              names for different numbers are allowed, but not vice versa. First
1275              scan down for duplicates. */
1276    
1277              if (i < md->name_count)
1278                {
1279                pcre_uchar *slotB = slotA;
1280                while (slotB > md->name_table)
1281                  {
1282                  slotB -= md->name_entry_size;
1283                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1284                    {
1285                    condition = GET2(slotB, 0) == md->recursive->group_num;
1286                    if (condition) break;
1287                    }
1288                  else break;
1289                  }
1290    
1291                /* Scan up for duplicates */
1292    
1293                if (!condition)
1294                  {
1295                  slotB = slotA;
1296                  for (i++; i < md->name_count; i++)
1297                    {
1298                    slotB += md->name_entry_size;
1299                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1300                      {
1301                      condition = GET2(slotB, 0) == md->recursive->group_num;
1302                      if (condition) break;
1303                      }
1304                    else break;
1305                    }
1306                  }
1307                }
1308              }
1309    
1310            /* Chose branch according to the condition */
1311    
1312            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1313            }
1314        }        }
1315    
1316      else if (condcode == OP_CREF)    /* Group used test */      else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1317        {        {
1318        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1319        condition = offset < offset_top && md->offset_vector[offset] >= 0;        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1320        ecode += condition? 3 : GET(ecode, 1);  
1321          /* If the numbered capture is unset, but the reference was by name,
1322          scan the table to see if the name refers to any other numbers, and test
1323          them. The condition is true if any one is set. This is tediously similar
1324          to the code above, but not close enough to try to amalgamate. */
1325    
1326          if (!condition && condcode == OP_NCREF)
1327            {
1328            int refno = offset >> 1;
1329            pcre_uchar *slotA = md->name_table;
1330    
1331            for (i = 0; i < md->name_count; i++)
1332              {
1333              if (GET2(slotA, 0) == refno) break;
1334              slotA += md->name_entry_size;
1335              }
1336    
1337            /* Found a name for the number - there can be only one; duplicate names
1338            for different numbers are allowed, but not vice versa. First scan down
1339            for duplicates. */
1340    
1341            if (i < md->name_count)
1342              {
1343              pcre_uchar *slotB = slotA;
1344              while (slotB > md->name_table)
1345                {
1346                slotB -= md->name_entry_size;
1347                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1348                  {
1349                  offset = GET2(slotB, 0) << 1;
1350                  condition = offset < offset_top &&
1351                    md->offset_vector[offset] >= 0;
1352                  if (condition) break;
1353                  }
1354                else break;
1355                }
1356    
1357              /* Scan up for duplicates */
1358    
1359              if (!condition)
1360                {
1361                slotB = slotA;
1362                for (i++; i < md->name_count; i++)
1363                  {
1364                  slotB += md->name_entry_size;
1365                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1366                    {
1367                    offset = GET2(slotB, 0) << 1;
1368                    condition = offset < offset_top &&
1369                      md->offset_vector[offset] >= 0;
1370                    if (condition) break;
1371                    }
1372                  else break;
1373                  }
1374                }
1375              }
1376            }
1377    
1378          /* Chose branch according to the condition */
1379    
1380          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1381        }        }
1382    
1383      else if (condcode == OP_DEF)     /* DEFINE - always false */      else if (condcode == OP_DEF)     /* DEFINE - always false */
# Line 847  for (;;) Line 1387  for (;;)
1387        }        }
1388    
1389      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1390      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
1391      assertion. */      an assertion. */
1392    
1393      else      else
1394        {        {
1395        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1396            match_condassert, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1397        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1398          {          {
1399            if (md->end_offset_top > offset_top)
1400              offset_top = md->end_offset_top;  /* Captures may have happened */
1401          condition = TRUE;          condition = TRUE;
1402          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1403          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1404          }          }
1405    
1406          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1407          assertion; it is therefore treated as NOMATCH. */
1408    
1409        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1410          {          {
1411          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
# Line 871  for (;;) Line 1417  for (;;)
1417          }          }
1418        }        }
1419    
1420      /* 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
1421      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
1422      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
1423      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
1424        exist, in which case we can just plough on. Note that, for compatibility
1425        with Perl, the | in a conditional group is NOT treated as creating two
1426        alternatives. If a THEN is encountered in the branch, it propagates out to
1427        the enclosing alternative (unless nested in a deeper set of alternatives,
1428        of course). */
1429    
1430      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1431        {        {
1432        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 */  
1433          {          {
1434          flags = 0;          ecode += 1 + LINK_SIZE;
1435          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1436          }          }
1437    
1438          md->match_function_type = MATCH_CBEGROUP;
1439          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1440          RRETURN(rrc);
1441        }        }
1442      else                         /* Condition false & no alternative */  
1443         /* Condition false & no alternative; continue after the group. */
1444    
1445        else
1446        {        {
1447        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1448        }        }
1449      break;      break;
1450    
1451    
1452      /* End of the pattern, either real or forced. If we are in a top-level      /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1453      recursion, we should restore the offsets appropriately and continue from      to close any currently open capturing brackets. */
     after the call. */  
1454    
1455      case OP_ACCEPT:      case OP_CLOSE:
1456      case OP_END:      number = GET2(ecode, 1);
1457      if (md->recursive != NULL && md->recursive->group_num == 0)      offset = number << 1;
1458    
1459    #ifdef PCRE_DEBUG
1460          printf("end bracket %d at *ACCEPT", number);
1461          printf("\n");
1462    #endif
1463    
1464        md->capture_last = number;
1465        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1466        {        {
1467        recursion_info *rec = md->recursive;        md->offset_vector[offset] =
1468        DPRINTF(("End of pattern in a (?0) recursion\n"));          md->offset_vector[md->offset_end - number];
1469        md->recursive = rec->prevrec;        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1470        memmove(md->offset_vector, rec->offset_save,        if (offset_top <= offset) offset_top = offset + 2;
         rec->saved_max * sizeof(int));  
       mstart = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
1471        }        }
1472        ecode += 1 + IMM2_SIZE;
1473        break;
1474    
1475    
1476      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* End of the pattern, either real or forced. */
1477      string - backtracking will then try other alternatives, if any. */  
1478        case OP_END:
1479        case OP_ACCEPT:
1480        case OP_ASSERT_ACCEPT:
1481    
1482        /* If we have matched an empty string, fail if not in an assertion and not
1483        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1484        is set and we have matched at the start of the subject. In both cases,
1485        backtracking will then try other alternatives, if any. */
1486    
1487        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1488             md->recursive == NULL &&
1489             (md->notempty ||
1490               (md->notempty_atstart &&
1491                 mstart == md->start_subject + md->start_offset)))
1492          MRRETURN(MATCH_NOMATCH);
1493    
1494        /* Otherwise, we have a match. */
1495    
     if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);  
1496      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1497      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1498      md->start_match_ptr = mstart;       /* and the start (\K can modify) */      md->start_match_ptr = mstart;       /* and the start (\K can modify) */
     RRETURN(MATCH_MATCH);  
1499    
1500      /* Change option settings */      /* For some reason, the macros don't work properly if an expression is
1501        given as the argument to MRRETURN when the heap is in use. */
1502    
1503      case OP_OPT:      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1504      ims = ecode[1];      MRRETURN(rrc);
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1505    
1506      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1507      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,
1508      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
1509      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
1510      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1511        of a condition, we want to return immediately afterwards. The caller of
1512        this incarnation of the match() function will have set MATCH_CONDASSERT in
1513        md->match_function type, and one of these opcodes will be the first opcode
1514        that is processed. We use a local variable that is preserved over calls to
1515        match() to remember this case. */
1516    
1517      case OP_ASSERT:      case OP_ASSERT:
1518      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1519        if (md->match_function_type == MATCH_CONDASSERT)
1520          {
1521          condassert = TRUE;
1522          md->match_function_type = 0;
1523          }
1524        else condassert = FALSE;
1525    
1526      do      do
1527        {        {
1528        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1529          RM4);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1530        if (rrc == MATCH_MATCH) break;          {
1531            mstart = md->start_match_ptr;   /* In case \K reset it */
1532            markptr = md->mark;
1533            break;
1534            }
1535    
1536          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1537          as NOMATCH. */
1538    
1539        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1540        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1541        }        }
1542      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1543      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1544        if (*ecode == OP_KET) MRRETURN(MATCH_NOMATCH);
1545    
1546      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1547    
1548      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1549    
1550      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1551      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 964  for (;;) Line 1555  for (;;)
1555      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1556      continue;      continue;
1557    
1558      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1559        PRUNE, or COMMIT means we must assume failure without checking subsequent
1560        branches. */
1561    
1562      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1563      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1564        if (md->match_function_type == MATCH_CONDASSERT)
1565          {
1566          condassert = TRUE;
1567          md->match_function_type = 0;
1568          }
1569        else condassert = FALSE;
1570    
1571      do      do
1572        {        {
1573        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1574          RM5);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) MRRETURN(MATCH_NOMATCH);
1575        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1576            {
1577            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1578            break;
1579            }
1580    
1581          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1582          as NOMATCH. */
1583    
1584        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1585        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1586        }        }
1587      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1588    
1589      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1590    
1591      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1592      continue;      continue;
# Line 989  for (;;) Line 1597  for (;;)
1597      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1598    
1599      case OP_REVERSE:      case OP_REVERSE:
1600  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1601      if (utf8)      if (utf)
1602        {        {
1603        i = GET(ecode, 1);        i = GET(ecode, 1);
1604        while (i-- > 0)        while (i-- > 0)
1605          {          {
1606          eptr--;          eptr--;
1607          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1608          BACKCHAR(eptr);          BACKCHAR(eptr);
1609          }          }
1610        }        }
# Line 1007  for (;;) Line 1615  for (;;)
1615    
1616        {        {
1617        eptr -= GET(ecode, 1);        eptr -= GET(ecode, 1);
1618        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) MRRETURN(MATCH_NOMATCH);
1619        }        }
1620    
1621      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1622    
1623        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1624      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1625      break;      break;
1626    
# Line 1023  for (;;) Line 1632  for (;;)
1632      if (pcre_callout != NULL)      if (pcre_callout != NULL)
1633        {        {
1634        pcre_callout_block cb;        pcre_callout_block cb;
1635        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1636        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1637        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1638        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1639        cb.subject_length   = md->end_subject - md->start_subject;        cb.subject_length   = (int)(md->end_subject - md->start_subject);
1640        cb.start_match      = mstart - md->start_subject;        cb.start_match      = (int)(mstart - md->start_subject);
1641        cb.current_position = eptr - md->start_subject;        cb.current_position = (int)(eptr - md->start_subject);
1642        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1643        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1644        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1645        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1646        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1647        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = (unsigned char *)markptr;
1648          if ((rrc = (*pcre_callout)(&cb)) > 0) MRRETURN(MATCH_NOMATCH);
1649        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1650        }        }
1651      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 1045  for (;;) Line 1655  for (;;)
1655      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
1656      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1657    
1658      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1659      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
1660      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
1661      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
1662      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
1663      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
1664      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.  
1665    
1666      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
1667      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
1668      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1669        a lot, so he is not to blame for the current way it works. */
1670    
1671      case OP_RECURSE:      case OP_RECURSE:
1672        {        {
1673          recursion_info *ri;
1674          int recno;
1675    
1676        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1677        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1678          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1679    
1680          /* Check for repeating a recursion without advancing the subject pointer.
1681          This should catch convoluted mutual recursions. (Some simple cases are
1682          caught at compile time.) */
1683    
1684          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1685            if (recno == ri->group_num && eptr == ri->subject_position)
1686              RRETURN(PCRE_ERROR_RECURSELOOP);
1687    
1688        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1689    
1690          new_recursive.group_num = recno;
1691          new_recursive.subject_position = eptr;
1692        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1693        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1694    
1695        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1696    
1697        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1698    
1699        /* Now save the offset data. */        /* Now save the offset data */
1700    
1701        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1702        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1087  for (;;) Line 1707  for (;;)
1707            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));
1708          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1709          }          }
   
1710        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1711              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1712    
1713        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1714        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1715          might be changed, so reset it before looping. */
1716    
1717        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1718        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1719        do        do
1720          {          {
1721          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1722            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1723          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1724            memcpy(md->offset_vector, new_recursive.offset_save,
1725                new_recursive.saved_max * sizeof(int));
1726            md->recursive = new_recursive.prevrec;
1727            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1728            {            {
1729            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1730            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1731              (pcre_free)(new_recursive.offset_save);              (pcre_free)(new_recursive.offset_save);
1732            RRETURN(MATCH_MATCH);  
1733              /* Set where we got to in the subject, and reset the start in case
1734              it was changed by \K. This *is* propagated back out of a recursion,
1735              for Perl compatibility. */
1736    
1737              eptr = md->end_match_ptr;
1738              mstart = md->start_match_ptr;
1739              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1740            }            }
1741    
1742            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1743            as NOMATCH. */
1744    
1745          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1746            {            {
1747            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
# Line 1119  for (;;) Line 1751  for (;;)
1751            }            }
1752    
1753          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1754          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1755          }          }
1756        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1129  for (;;) Line 1759  for (;;)
1759        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1760        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1761          (pcre_free)(new_recursive.offset_save);          (pcre_free)(new_recursive.offset_save);
1762        RRETURN(MATCH_NOMATCH);        MRRETURN(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. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
1763        }        }
1764    
1765      if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1766        {      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 */  
1767    
1768      /* 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
1769      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1221  for (;;) Line 1779  for (;;)
1779      optional ones preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1780    
1781      case OP_BRAZERO:      case OP_BRAZERO:
1782        {      next = ecode + 1;
1783        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1784        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1785        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1786        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1787      break;      break;
1788    
1789      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1790        {      next = ecode + 1;
1791        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1792        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1793        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1794        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1795      break;      break;
1796    
1797      case OP_SKIPZERO:      case OP_SKIPZERO:
1798        {      next = ecode+1;
1799        next = ecode+1;      do next += GET(next,1); while (*next == OP_ALT);
1800        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1801      break;      break;
1802    
1803        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1804        here; just jump to the group, with allow_zero set TRUE. */
1805    
1806        case OP_BRAPOSZERO:
1807        op = *(++ecode);
1808        allow_zero = TRUE;
1809        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1810          goto POSSESSIVE_NON_CAPTURE;
1811    
1812      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1813    
1814      case OP_KET:      case OP_KET:
1815      case OP_KETRMIN:      case OP_KETRMIN:
1816      case OP_KETRMAX:      case OP_KETRMAX:
1817        case OP_KETRPOS:
1818      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1819    
1820      /* 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
1821      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1822      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1823    
1824      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1825        {        {
1826        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1827        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1828        }        }
1829      else saved_eptr = NULL;      else saved_eptr = NULL;
1830    
1831      /* If we are at the end of an assertion group, stop matching and return      /* If we are at the end of an assertion group or a non-capturing atomic
1832      MATCH_MATCH, but record the current high water mark for use by positive      group, stop matching and return MATCH_MATCH, but record the current high
1833      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1834        start in case it was changed by \K. */
1835      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1836          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1837          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1838        {        {
1839        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1840        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1841        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1842          MRRETURN(MATCH_MATCH);         /* Sets md->mark */
1843        }        }
1844    
1845      /* 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
1846      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1847      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
1848      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
1849      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
1850        the current subject position and start match pointer and give a MATCH
1851        return. */
1852    
1853      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1854            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1855        {        {
1856        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1857        offset = number << 1;        offset = number << 1;
1858    
1859  #ifdef DEBUG  #ifdef PCRE_DEBUG
1860        printf("end bracket %d", number);        printf("end bracket %d", number);
1861        printf("\n");        printf("\n");
1862  #endif  #endif
1863    
1864          /* Handle a recursively called group. */
1865    
1866          if (md->recursive != NULL && md->recursive->group_num == number)
1867            {
1868            md->end_match_ptr = eptr;
1869            md->start_match_ptr = mstart;
1870            RRETURN(MATCH_MATCH);
1871            }
1872    
1873          /* Deal with capturing */
1874    
1875        md->capture_last = number;        md->capture_last = number;
1876        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1877          {          {
1878            /* If offset is greater than offset_top, it means that we are
1879            "skipping" a capturing group, and that group's offsets must be marked
1880            unset. In earlier versions of PCRE, all the offsets were unset at the
1881            start of matching, but this doesn't work because atomic groups and
1882            assertions can cause a value to be set that should later be unset.
1883            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1884            part of the atomic group, but this is not on the final matching path,
1885            so must be unset when 2 is set. (If there is no group 2, there is no
1886            problem, because offset_top will then be 2, indicating no capture.) */
1887    
1888            if (offset > offset_top)
1889              {
1890              register int *iptr = md->offset_vector + offset_top;
1891              register int *iend = md->offset_vector + offset;
1892              while (iptr < iend) *iptr++ = -1;
1893              }
1894    
1895            /* Now make the extraction */
1896    
1897          md->offset_vector[offset] =          md->offset_vector[offset] =
1898            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1899          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1900          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1901          }          }
1902          }
1903    
1904        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1905        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1906        group. This is the forcible breaking of infinite loops as implemented in
1907        Perl 5.005. For a non-repeating atomic group that includes captures,
1908        establish a backup point by processing the rest of the pattern at a lower
1909        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1910        original OP_ONCE level, thereby bypassing intermediate backup points, but
1911        resetting any captures that happened along the way. */
1912    
1913        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1914          {
1915          if (*prev == OP_ONCE)
1916          {          {
1917          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1918          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1919          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1920          mstart = rec->save_start;          RRETURN(MATCH_ONCE);
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
1921          }          }
1922          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1923          break;
1924        }        }
1925    
1926      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1927      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1928        at a time from the outer level, thus saving stack. */
1929    
1930      ims = original_ims;      if (*ecode == OP_KETRPOS)
     DPRINTF(("ims reset to %02lx\n", ims));  
   
     /* 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)  
1931        {        {
1932        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1933        break;        md->end_offset_top = offset_top;
1934          RRETURN(MATCH_KETRPOS);
1935        }        }
1936    
1937      /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
1938      preceding bracket, in the appropriate order. In the second case, we can use      the preceding bracket, in the appropriate order. In the second case, we can
1939      tail recursion to avoid using another stack frame, unless we have an      use tail recursion to avoid using another stack frame, unless we have an
1940      unlimited repeat of a group that can match an empty string. */      an atomic group or an unlimited repeat of a group that can match an empty
1941        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
1942    
1943      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1944        {        {
1945        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1946        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1947        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
1948          {          {
1949          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1950            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1951            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1952            RRETURN(MATCH_ONCE);
1953            }
1954          if (*prev >= OP_SBRA)    /* Could match an empty string */
1955            {
1956            md->match_function_type = MATCH_CBEGROUP;
1957            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
1958          RRETURN(rrc);          RRETURN(rrc);
1959          }          }
1960        ecode = prev;        ecode = prev;
# Line 1360  for (;;) Line 1962  for (;;)
1962        }        }
1963      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
1964        {        {
1965        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1966          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
1967          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
1968        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1969          if (*prev == OP_ONCE)
1970            {
1971            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
1972            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1973            md->once_target = prev;
1974            RRETURN(MATCH_ONCE);
1975            }
1976        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
1977        goto TAIL_RECURSE;        goto TAIL_RECURSE;
1978        }        }
1979      /* Control never gets here */      /* Control never gets here */
1980    
1981      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
1982    
1983      case OP_CIRC:      case OP_CIRC:
1984      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
1985    
1986      /* Start of subject assertion */      /* Start of subject assertion */
1987    
1988      case OP_SOD:      case OP_SOD:
1989      if (eptr != md->start_subject) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject) MRRETURN(MATCH_NOMATCH);
1990        ecode++;
1991        break;
1992    
1993        /* Multiline mode: start of subject unless notbol, or after any newline. */
1994    
1995        case OP_CIRCM:
1996        if (md->notbol && eptr == md->start_subject) MRRETURN(MATCH_NOMATCH);
1997        if (eptr != md->start_subject &&
1998            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
1999          MRRETURN(MATCH_NOMATCH);
2000      ecode++;      ecode++;
2001      break;      break;
2002    
2003      /* Start of match assertion */      /* Start of match assertion */
2004    
2005      case OP_SOM:      case OP_SOM:
2006      if (eptr != md->start_subject + md->start_offset) RRETURN(MATCH_NOMATCH);      if (eptr != md->start_subject + md->start_offset) MRRETURN(MATCH_NOMATCH);
2007      ecode++;      ecode++;
2008      break;      break;
2009    
# Line 1403  for (;;) Line 2014  for (;;)
2014      ecode++;      ecode++;
2015      break;      break;
2016    
2017      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2018      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2019    
2020      case OP_DOLL:      case OP_DOLLM:
2021      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2022        {        { if (!IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH); }
       if (eptr < md->end_subject)  
         { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }  
       else  
         { if (md->noteol) RRETURN(MATCH_NOMATCH); }  
       ecode++;  
       break;  
       }  
2023      else      else
2024        {        {
2025        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) MRRETURN(MATCH_NOMATCH);
2026        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2027        }        }
2028        ecode++;
2029        break;
2030    
2031        /* Not multiline mode: assert before a terminating newline or before end of
2032        subject unless noteol is set. */
2033    
2034        case OP_DOLL:
2035        if (md->noteol) MRRETURN(MATCH_NOMATCH);
2036        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2037    
2038      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2039    
2040      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2041    
2042      case OP_EOD:      case OP_EOD:
2043      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) MRRETURN(MATCH_NOMATCH);
2044        SCHECK_PARTIAL();
2045      ecode++;      ecode++;
2046      break;      break;
2047    
2048      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2049    
2050      case OP_EODN:      case OP_EODN:
2051      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2052        if (eptr < md->end_subject &&
2053          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2054        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2055    
2056        /* Either at end of string or \n before end. */
2057    
2058        SCHECK_PARTIAL();
2059      ecode++;      ecode++;
2060      break;      break;
2061    
# Line 1452  for (;;) Line 2065  for (;;)
2065      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
2066        {        {
2067    
2068        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2069        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2070        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2071          partial matching. */
2072    
2073    #ifdef SUPPORT_UTF
2074          if (utf)
2075            {
2076            /* Get status of previous character */
2077    
2078            if (eptr == md->start_subject) prev_is_word = FALSE; else
2079              {
2080              PCRE_PUCHAR lastptr = eptr - 1;
2081              BACKCHAR(lastptr);
2082              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2083              GETCHAR(c, lastptr);
2084    #ifdef SUPPORT_UCP
2085              if (md->use_ucp)
2086                {
2087                if (c == '_') prev_is_word = TRUE; else
2088                  {
2089                  int cat = UCD_CATEGORY(c);
2090                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2091                  }
2092                }
2093              else
2094    #endif
2095              prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2096              }
2097    
2098            /* Get status of next character */
2099    
2100            if (eptr >= md->end_subject)
2101              {
2102              SCHECK_PARTIAL();
2103              cur_is_word = FALSE;
2104              }
2105            else
2106              {
2107              GETCHAR(c, eptr);
2108    #ifdef SUPPORT_UCP
2109              if (md->use_ucp)
2110                {
2111                if (c == '_') cur_is_word = TRUE; else
2112                  {
2113                  int cat = UCD_CATEGORY(c);
2114                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2115                  }
2116                }
2117              else
2118    #endif
2119              cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2120              }
2121            }
2122          else
2123    #endif
2124    
2125          /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2126          consistency with the behaviour of \w we do use it in this case. */
2127    
 #ifdef SUPPORT_UTF8  
       if (utf8)  
2128          {          {
2129            /* Get status of previous character */
2130    
2131          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2132            {            {
2133            const uschar *lastptr = eptr - 1;            if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2134            while((*lastptr & 0xc0) == 0x80) lastptr--;  #ifdef SUPPORT_UCP
2135            GETCHAR(c, lastptr);            if (md->use_ucp)
2136            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;              {
2137                c = eptr[-1];
2138                if (c == '_') prev_is_word = TRUE; else
2139                  {
2140                  int cat = UCD_CATEGORY(c);
2141                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2142                  }
2143                }
2144              else
2145    #endif
2146              prev_is_word = ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2147            }            }
2148          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2149            /* Get status of next character */
2150    
2151            if (eptr >= md->end_subject)
2152            {            {
2153            GETCHAR(c, eptr);            SCHECK_PARTIAL();
2154            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = FALSE;
2155            }            }
2156          }          else
2157        else  #ifdef SUPPORT_UCP
2158            if (md->use_ucp)
2159              {
2160              c = *eptr;
2161              if (c == '_') cur_is_word = TRUE; else
2162                {
2163                int cat = UCD_CATEGORY(c);
2164                cur_is_word = (cat == ucp_L || cat == ucp_N);
2165                }
2166              }
2167            else
2168  #endif  #endif
2169            cur_is_word = ((md->ctypes[*eptr] & ctype_word) != 0);
       /* More streamlined when not in UTF-8 mode */  
   
         {  
         prev_is_word = (eptr != md->start_subject) &&  
           ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
         cur_is_word = (eptr < md->end_subject) &&  
           ((md->ctypes[*eptr] & ctype_word) != 0);  
2170          }          }
2171    
2172        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
2173    
2174        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
2175             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
2176          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2177        }        }
2178      break;      break;
2179    
2180      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2181    
2182      case OP_ANY:      case OP_ANY:
2183      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);      if (IS_NEWLINE(eptr)) MRRETURN(MATCH_NOMATCH);
2184      /* Fall through */      /* Fall through */
2185    
2186      case OP_ALLANY:      case OP_ALLANY:
2187      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2188      if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;        {                            /* not be updated before SCHECK_PARTIAL. */
2189          SCHECK_PARTIAL();
2190          MRRETURN(MATCH_NOMATCH);
2191          }
2192        eptr++;
2193    #ifdef SUPPORT_UTF
2194        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2195    #endif
2196      ecode++;      ecode++;
2197      break;      break;
2198    
# Line 1508  for (;;) Line 2200  for (;;)
2200      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2201    
2202      case OP_ANYBYTE:      case OP_ANYBYTE:
2203      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2204          {                            /* not be updated before SCHECK_PARTIAL. */
2205          SCHECK_PARTIAL();
2206          MRRETURN(MATCH_NOMATCH);
2207          }
2208        eptr++;
2209      ecode++;      ecode++;
2210      break;      break;
2211    
2212      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2213      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2214          {
2215          SCHECK_PARTIAL();
2216          MRRETURN(MATCH_NOMATCH);
2217          }
2218      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2219      if (      if (
2220  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2221         c < 256 &&         c < 256 &&
2222  #endif  #endif
2223         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
2224         )         )
2225        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2226      ecode++;      ecode++;
2227      break;      break;
2228    
2229      case OP_DIGIT:      case OP_DIGIT:
2230      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2231          {
2232          SCHECK_PARTIAL();
2233          MRRETURN(MATCH_NOMATCH);
2234          }
2235      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2236      if (      if (
2237  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2238         c >= 256 ||         c > 255 ||
2239  #endif  #endif
2240         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2241         )         )
2242        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2243      ecode++;      ecode++;
2244      break;      break;
2245    
2246      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2247      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2248          {
2249          SCHECK_PARTIAL();
2250          MRRETURN(MATCH_NOMATCH);
2251          }
2252      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2253      if (      if (
2254  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2255         c < 256 &&         c < 256 &&
2256  #endif  #endif
2257         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
2258         )         )
2259        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2260      ecode++;      ecode++;
2261      break;      break;
2262    
2263      case OP_WHITESPACE:      case OP_WHITESPACE:
2264      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2265          {
2266          SCHECK_PARTIAL();
2267          MRRETURN(MATCH_NOMATCH);
2268          }
2269      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2270      if (      if (
2271  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2272         c >= 256 ||         c > 255 ||
2273  #endif  #endif
2274         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2275         )         )
2276        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2277      ecode++;      ecode++;
2278      break;      break;
2279    
2280      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2281      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2282          {
2283          SCHECK_PARTIAL();
2284          MRRETURN(MATCH_NOMATCH);
2285          }
2286      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2287      if (      if (
2288  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2289         c < 256 &&         c < 256 &&
2290  #endif  #endif
2291         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
2292         )         )
2293        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2294      ecode++;      ecode++;
2295      break;      break;
2296    
2297      case OP_WORDCHAR:      case OP_WORDCHAR:
2298      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2299          {
2300          SCHECK_PARTIAL();
2301          MRRETURN(MATCH_NOMATCH);
2302          }
2303      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2304      if (      if (
2305  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2306         c >= 256 ||         c > 255 ||
2307  #endif  #endif
2308         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2309         )         )
2310        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2311      ecode++;      ecode++;
2312      break;      break;
2313    
2314      case OP_ANYNL:      case OP_ANYNL:
2315      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2316          {
2317          SCHECK_PARTIAL();
2318          MRRETURN(MATCH_NOMATCH);
2319          }
2320      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2321      switch(c)      switch(c)
2322        {        {
2323        default: RRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2324    
2325        case 0x000d:        case 0x000d:
2326        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2327        break;        break;
# Line 1608  for (;;) Line 2334  for (;;)
2334        case 0x0085:        case 0x0085:
2335        case 0x2028:        case 0x2028:
2336        case 0x2029:        case 0x2029:
2337        if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);        if (md->bsr_anycrlf) MRRETURN(MATCH_NOMATCH);
2338        break;        break;
2339        }        }
2340      ecode++;      ecode++;
2341      break;      break;
2342    
2343      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2344      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2345          {
2346          SCHECK_PARTIAL();
2347          MRRETURN(MATCH_NOMATCH);
2348          }
2349      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2350      switch(c)      switch(c)
2351        {        {
# Line 1639  for (;;) Line 2369  for (;;)
2369        case 0x202f:    /* NARROW NO-BREAK SPACE */        case 0x202f:    /* NARROW NO-BREAK SPACE */
2370        case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */        case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */
2371        case 0x3000:    /* IDEOGRAPHIC SPACE */        case 0x3000:    /* IDEOGRAPHIC SPACE */
2372        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2373        }        }
2374      ecode++;      ecode++;
2375      break;      break;
2376    
2377      case OP_HSPACE:      case OP_HSPACE:
2378      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2379          {
2380          SCHECK_PARTIAL();
2381          MRRETURN(MATCH_NOMATCH);
2382          }
2383      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2384      switch(c)      switch(c)
2385        {        {
2386        default: RRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2387        case 0x09:      /* HT */        case 0x09:      /* HT */
2388        case 0x20:      /* SPACE */        case 0x20:      /* SPACE */
2389        case 0xa0:      /* NBSP */        case 0xa0:      /* NBSP */
# Line 1675  for (;;) Line 2409  for (;;)
2409      break;      break;
2410    
2411      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2412      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2413          {
2414          SCHECK_PARTIAL();
2415          MRRETURN(MATCH_NOMATCH);
2416          }
2417      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2418      switch(c)      switch(c)
2419        {        {
# Line 1687  for (;;) Line 2425  for (;;)
2425        case 0x85:      /* NEL */        case 0x85:      /* NEL */
2426        case 0x2028:    /* LINE SEPARATOR */        case 0x2028:    /* LINE SEPARATOR */
2427        case 0x2029:    /* PARAGRAPH SEPARATOR */        case 0x2029:    /* PARAGRAPH SEPARATOR */
2428        RRETURN(MATCH_NOMATCH);        MRRETURN(MATCH_NOMATCH);
2429        }        }
2430      ecode++;      ecode++;
2431      break;      break;
2432    
2433      case OP_VSPACE:      case OP_VSPACE:
2434      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2435          {
2436          SCHECK_PARTIAL();
2437          MRRETURN(MATCH_NOMATCH);
2438          }
2439      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2440      switch(c)      switch(c)
2441        {        {
2442        default: RRETURN(MATCH_NOMATCH);        default: MRRETURN(MATCH_NOMATCH);
2443        case 0x0a:      /* LF */        case 0x0a:      /* LF */
2444        case 0x0b:      /* VT */        case 0x0b:      /* VT */
2445        case 0x0c:      /* FF */        case 0x0c:      /* FF */
# Line 1716  for (;;) Line 2458  for (;;)
2458    
2459      case OP_PROP:      case OP_PROP:
2460      case OP_NOTPROP:      case OP_NOTPROP:
2461      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2462          {
2463          SCHECK_PARTIAL();
2464          MRRETURN(MATCH_NOMATCH);
2465          }
2466      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2467        {        {
2468        const ucd_record *prop = GET_UCD(c);        const ucd_record *prop = GET_UCD(c);
# Line 1724  for (;;) Line 2470  for (;;)
2470        switch(ecode[1])        switch(ecode[1])
2471          {          {
2472          case PT_ANY:          case PT_ANY:
2473          if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);          if (op == OP_NOTPROP) MRRETURN(MATCH_NOMATCH);
2474          break;          break;
2475    
2476          case PT_LAMP:          case PT_LAMP:
2477          if ((prop->chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2478               prop->chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2479               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2480            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2481           break;          break;
2482    
2483          case PT_GC:          case PT_GC:
2484          if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2485            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2486          break;          break;
2487    
2488          case PT_PC:          case PT_PC:
2489          if ((ecode[2] != prop->chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2490            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2491          break;          break;
2492    
2493          case PT_SC:          case PT_SC:
2494          if ((ecode[2] != prop->script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2495            RRETURN(MATCH_NOMATCH);            MRRETURN(MATCH_NOMATCH);
2496            break;
2497    
2498            /* These are specials */
2499    
2500            case PT_ALNUM:
2501            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2502                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2503              MRRETURN(MATCH_NOMATCH);
2504            break;
2505    
2506            case PT_SPACE:    /* Perl space */
2507            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2508                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2509                   == (op == OP_NOTPROP))
2510              MRRETURN(MATCH_NOMATCH);
2511            break;
2512    
2513            case PT_PXSPACE:  /* POSIX space */
2514            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2515                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2516                 c == CHAR_FF || c == CHAR_CR)
2517                   == (op == OP_NOTPROP))
2518              MRRETURN(MATCH_NOMATCH);
2519          break;          break;
2520    
2521            case PT_WORD:
2522            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2523                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2524                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2525              MRRETURN(MATCH_NOMATCH);
2526            break;
2527    
2528            /* This should never occur */
2529    
2530          default:          default:
2531          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2532          }          }
# Line 1761  for (;;) Line 2539  for (;;)
2539      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2540    
2541      case OP_EXTUNI:      case OP_EXTUNI:
2542      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2543          {
2544          SCHECK_PARTIAL();
2545          MRRETURN(MATCH_NOMATCH);
2546          }
2547      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2548        if (UCD_CATEGORY(c) == ucp_M) MRRETURN(MATCH_NOMATCH);
2549        while (eptr < md->end_subject)
2550        {        {
2551        int category = UCD_CATEGORY(c);        int len = 1;
2552        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2553        while (eptr < md->end_subject)        if (UCD_CATEGORY(c) != ucp_M) break;
2554          {        eptr += len;
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = UCD_CATEGORY(c);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2555        }        }
2556      ecode++;      ecode++;
2557      break;      break;
# Line 1792  for (;;) Line 2567  for (;;)
2567      loops). */      loops). */
2568    
2569      case OP_REF:      case OP_REF:
2570        {      case OP_REFI:
2571        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2572        ecode += 3;      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2573        ecode += 1 + IMM2_SIZE;
2574    
2575        /* If the reference is unset, there are two possibilities:      /* If the reference is unset, there are two possibilities:
2576    
2577        (a) In the default, Perl-compatible state, set the length to be longer      (a) In the default, Perl-compatible state, set the length negative;
2578        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
2579        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.  
2580    
2581        (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
2582        so that the back reference matches an empty string.      so that the back reference matches an empty string.
2583    
2584        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
2585        referenced subpattern. */      referenced subpattern. */
2586    
2587        if (offset >= offset_top || md->offset_vector[offset] < 0)      if (offset >= offset_top || md->offset_vector[offset] < 0)
2588          length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;        length = (md->jscript_compat)? 0 : -1;
2589        else      else
2590          length = md->offset_vector[offset+1] - md->offset_vector[offset];        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2591    
2592        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2593    
2594        switch (*ecode)      switch (*ecode)
2595          {        {
2596          case OP_CRSTAR:        case OP_CRSTAR:
2597          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2598          case OP_CRPLUS:        case OP_CRPLUS:
2599          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2600          case OP_CRQUERY:        case OP_CRQUERY:
2601          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2602          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2603          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2604          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2605          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2606          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2607          break;        break;
2608    
2609          case OP_CRRANGE:        case OP_CRRANGE:
2610          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2611          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2612          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2613          max = GET2(ecode, 3);        max = GET2(ecode, 1 + IMM2_SIZE);
2614          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2615          ecode += 5;        ecode += 1 + 2 * IMM2_SIZE;
2616          break;        break;
2617    
2618          default:               /* No repeat follows */        default:               /* No repeat follows */
2619          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2620          eptr += length;          {
2621          continue;              /* With the main loop */          CHECK_PARTIAL();
2622            MRRETURN(MATCH_NOMATCH);
2623          }          }
2624          eptr += length;
2625          continue;              /* With the main loop */
2626          }
2627    
2628        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2629        main loop. */      zero, just continue with the main loop. */
2630    
2631        if (length == 0) continue;      if (length == 0) continue;
2632    
2633        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2634        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2635        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2636    
2637        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2638          {
2639          int slength;
2640          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2641          {          {
2642          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          CHECK_PARTIAL();
2643          eptr += length;          MRRETURN(MATCH_NOMATCH);
2644          }          }
2645          eptr += slength;
2646          }
2647    
2648        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2649        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2650    
2651        if (min == max) continue;      if (min == max) continue;
2652    
2653        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2654    
2655        if (minimize)      if (minimize)
2656          {
2657          for (fi = min;; fi++)
2658          {          {
2659          for (fi = min;; fi++)          int slength;
2660            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2661            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2662            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2663            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2664            {            {
2665            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            CHECK_PARTIAL();
2666            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            MRRETURN(MATCH_NOMATCH);
           if (fi >= max || !match_ref(offset, eptr, length, md, ims))  
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2667            }            }
2668          /* Control never gets here */          eptr += slength;
2669          }          }
2670          /* Control never gets here */
2671          }
2672    
2673        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2674    
2675        else      else
2676          {
2677          pp = eptr;
2678          for (i = min; i < max; i++)
2679          {          {
2680          pp = eptr;          int slength;
2681          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
2682            {            {
2683            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            CHECK_PARTIAL();
2684            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            break;
           eptr -= length;  
2685            }            }
2686          RRETURN(MATCH_NOMATCH);          eptr += slength;
2687            }
2688          while (eptr >= pp)
2689            {
2690            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2691            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2692            eptr -= length;
2693          }          }
2694          MRRETURN(MATCH_NOMATCH);
2695        }        }
2696      /* Control never gets here */      /* Control never gets here */
2697    
   
   
2698      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2699      used when all the characters in the class have values in the range 0-255,      used when all the characters in the class have values in the range 0-255,
2700      and either the matching is caseful, or the characters are in the range      and either the matching is caseful, or the characters are in the range
# Line 1918  for (;;) Line 2709  for (;;)
2709      case OP_NCLASS:      case OP_NCLASS:
2710      case OP_CLASS:      case OP_CLASS:
2711        {        {
2712          /* The data variable is saved across frames, so the byte map needs to
2713          be stored there. */
2714    #define BYTE_MAP ((pcre_uint8 *)data)
2715        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2716        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2717    
2718        switch (*ecode)        switch (*ecode)
2719          {          {
# Line 1940  for (;;) Line 2734  for (;;)
2734          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2735          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2736          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2737          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2738          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2739          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2740          break;          break;
2741    
2742          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1952  for (;;) Line 2746  for (;;)
2746    
2747        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2748    
2749  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2750        /* UTF-8 mode */        if (utf)
       if (utf8)  
2751          {          {
2752          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2753            {            {
2754            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2755                {
2756                SCHECK_PARTIAL();
2757                MRRETURN(MATCH_NOMATCH);
2758                }
2759            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2760            if (c > 255)            if (c > 255)
2761              {              {
2762              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2763              }              }
2764            else            else
2765              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2766            }            }
2767          }          }
2768        else        else
2769  #endif  #endif
2770        /* Not UTF-8 mode */        /* Not UTF mode */
2771          {          {
2772          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2773            {            {
2774            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2775                {
2776                SCHECK_PARTIAL();
2777                MRRETURN(MATCH_NOMATCH);
2778                }
2779            c = *eptr++;            c = *eptr++;
2780            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2781              if (c > 255)
2782                {
2783                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2784                }
2785              else
2786    #endif
2787                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2788            }            }
2789          }          }
2790    
# Line 1992  for (;;) Line 2798  for (;;)
2798    
2799        if (minimize)        if (minimize)
2800          {          {
2801  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2802          /* UTF-8 mode */          if (utf)
         if (utf8)  
2803            {            {
2804            for (fi = min;; fi++)            for (fi = min;; fi++)
2805              {              {
2806              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2807              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2808              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2809                if (eptr >= md->end_subject)
2810                  {
2811                  SCHECK_PARTIAL();
2812                  MRRETURN(MATCH_NOMATCH);
2813                  }
2814              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2815              if (c > 255)              if (c > 255)
2816                {                {
2817                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2818                }                }
2819              else              else
2820                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2821              }              }
2822            }            }
2823          else          else
2824  #endif  #endif
2825          /* Not UTF-8 mode */          /* Not UTF mode */
2826            {            {
2827            for (fi = min;; fi++)            for (fi = min;; fi++)
2828              {              {
2829              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2830              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2831              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
2832                if (eptr >= md->end_subject)
2833                  {
2834                  SCHECK_PARTIAL();
2835                  MRRETURN(MATCH_NOMATCH);
2836                  }
2837              c = *eptr++;              c = *eptr++;
2838              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2839                if (c > 255)
2840                  {
2841                  if (op == OP_CLASS) MRRETURN(MATCH_NOMATCH);
2842                  }
2843                else
2844    #endif
2845                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) MRRETURN(MATCH_NOMATCH);
2846              }              }
2847            }            }
2848          /* Control never gets here */          /* Control never gets here */
# Line 2034  for (;;) Line 2854  for (;;)
2854          {          {
2855          pp = eptr;          pp = eptr;
2856    
2857  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2858          /* UTF-8 mode */          if (utf)
         if (utf8)  
2859            {            {
2860            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2861              {              {
2862              int len = 1;              int len = 1;
2863              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2864                  {
2865                  SCHECK_PARTIAL();
2866                  break;
2867                  }
2868              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2869              if (c > 255)              if (c > 255)
2870                {                {
2871                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2872                }                }
2873              else              else
2874                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2875              eptr += len;              eptr += len;
2876              }              }
2877            for (;;)            for (;;)
2878              {              {
2879              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2880              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2881              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2882              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 2063  for (;;) Line 2884  for (;;)
2884            }            }
2885          else          else
2886  #endif  #endif
2887            /* Not UTF-8 mode */            /* Not UTF mode */
2888            {            {
2889            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2890              {              {
2891              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2892                  {
2893                  SCHECK_PARTIAL();
2894                  break;
2895                  }
2896              c = *eptr;              c = *eptr;
2897              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2898                if (c > 255)
2899                  {
2900                  if (op == OP_CLASS) break;
2901                  }
2902                else
2903    #endif
2904                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2905              eptr++;              eptr++;
2906              }              }
2907            while (eptr >= pp)            while (eptr >= pp)
2908              {              {
2909              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
2910              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2911              eptr--;              eptr--;
2912              }              }
2913            }            }
2914    
2915          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
2916          }          }
2917    #undef BYTE_MAP
2918        }        }
2919      /* Control never gets here */      /* Control never gets here */
2920    
# Line 2090  for (;;) Line 2923  for (;;)
2923      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
2924      mode, because Unicode properties are supported in non-UTF-8 mode. */      mode, because Unicode properties are supported in non-UTF-8 mode. */
2925    
2926  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2927      case OP_XCLASS:      case OP_XCLASS:
2928        {        {
2929        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2115  for (;;) Line 2948  for (;;)
2948          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2949          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2950          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2951          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2952          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2953          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2954          break;          break;
2955    
2956          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2129  for (;;) Line 2962  for (;;)
2962    
2963        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
2964          {          {
2965          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
2966              {
2967              SCHECK_PARTIAL();
2968              MRRETURN(MATCH_NOMATCH);
2969              }
2970          GETCHARINCTEST(c, eptr);          GETCHARINCTEST(c, eptr);
2971          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);          if (!PRIV(xclass)(c, data, utf)) MRRETURN(MATCH_NOMATCH);
2972          }          }
2973    
2974        /* 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 2146  for (;;) Line 2983  for (;;)
2983          {          {
2984          for (fi = min;; fi++)          for (fi = min;; fi++)
2985            {            {
2986            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
2987            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2988            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) MRRETURN(MATCH_NOMATCH);
2989              if (eptr >= md->end_subject)
2990                {
2991                SCHECK_PARTIAL();
2992                MRRETURN(MATCH_NOMATCH);
2993                }
2994            GETCHARINCTEST(c, eptr);            GETCHARINCTEST(c, eptr);
2995            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            if (!PRIV(xclass)(c, data, utf)) MRRETURN(MATCH_NOMATCH);
2996            }            }
2997          /* Control never gets here */          /* Control never gets here */
2998          }          }
# Line 2163  for (;;) Line 3005  for (;;)
3005          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3006            {            {
3007            int len = 1;            int len = 1;
3008            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3009                {
3010                SCHECK_PARTIAL();
3011                break;
3012                }
3013    #ifdef SUPPORT_UTF
3014            GETCHARLENTEST(c, eptr, len);            GETCHARLENTEST(c, eptr, len);
3015            if (!_pcre_xclass(c, data)) break;  #else
3016              c = *eptr;
3017    #endif
3018              if (!PRIV(xclass)(c, data, utf)) break;
3019            eptr += len;            eptr += len;
3020            }            }
3021          for(;;)          for(;;)
3022            {            {
3023            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3024            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3025            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3026            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3027              if (utf) BACKCHAR(eptr);
3028    #endif
3029            }            }
3030          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3031          }          }
3032    
3033        /* Control never gets here */        /* Control never gets here */
# Line 2185  for (;;) Line 3037  for (;;)
3037      /* Match a single character, casefully */      /* Match a single character, casefully */
3038    
3039      case OP_CHAR:      case OP_CHAR:
3040  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3041      if (utf8)      if (utf)
3042        {        {
3043        length = 1;        length = 1;
3044        ecode++;        ecode++;
3045        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3046        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3047        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3048            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3049            MRRETURN(MATCH_NOMATCH);
3050            }
3051          while (length-- > 0) if (*ecode++ != *eptr++) MRRETURN(MATCH_NOMATCH);
3052        }        }
3053      else      else
3054  #endif  #endif
3055        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3056        {        {
3057        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3058        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);          {
3059            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3060            MRRETURN(MATCH_NOMATCH);
3061            }
3062          if (ecode[1] != *eptr++) MRRETURN(MATCH_NOMATCH);
3063        ecode += 2;        ecode += 2;
3064        }        }
3065      break;      break;
3066    
3067      /* Match a single character, caselessly */      /* Match a single character, caselessly */
3068    
3069      case OP_CHARNC:      case OP_CHARI:
3070  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3071      if (utf8)      if (utf)
3072        {        {
3073        length = 1;        length = 1;
3074        ecode++;        ecode++;
3075        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3076    
3077        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3078            {
3079            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3080            MRRETURN(MATCH_NOMATCH);
3081            }
3082    
3083        /* 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
3084        can use the fast lookup table. */        can use the fast lookup table. */
3085    
3086        if (fc < 128)        if (fc < 128)
3087          {          {
3088          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[*ecode++] != md->lcc[*eptr++]) MRRETURN(MATCH_NOMATCH);
3089          }          }
3090    
3091        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character */
# Line 2241  for (;;) Line 3104  for (;;)
3104  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3105            if (dc != UCD_OTHERCASE(fc))            if (dc != UCD_OTHERCASE(fc))
3106  #endif  #endif
3107              RRETURN(MATCH_NOMATCH);              MRRETURN(MATCH_NOMATCH);
3108            }            }
3109          }          }
3110        }        }
3111      else      else
3112  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3113    
3114      /* Non-UTF-8 mode */      /* Not UTF mode */
3115        {        {
3116        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3117        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3118            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3119            MRRETURN(MATCH_NOMATCH);
3120            }
3121          if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3122              != TABLE_GET(*eptr, md->lcc, *eptr)) MRRETURN(MATCH_NOMATCH);
3123          eptr++;
3124        ecode += 2;        ecode += 2;
3125        }        }
3126      break;      break;
# Line 2259  for (;;) Line 3128  for (;;)
3128      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3129    
3130      case OP_EXACT:      case OP_EXACT:
3131        case OP_EXACTI:
3132      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3133      ecode += 3;      ecode += 1 + IMM2_SIZE;
3134      goto REPEATCHAR;      goto REPEATCHAR;
3135    
3136      case OP_POSUPTO:      case OP_POSUPTO:
3137        case OP_POSUPTOI:
3138      possessive = TRUE;      possessive = TRUE;
3139      /* Fall through */      /* Fall through */
3140    
3141      case OP_UPTO:      case OP_UPTO:
3142        case OP_UPTOI:
3143      case OP_MINUPTO:      case OP_MINUPTO:
3144        case OP_MINUPTOI:
3145      min = 0;      min = 0;
3146      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3147      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3148      ecode += 3;      ecode += 1 + IMM2_SIZE;
3149      goto REPEATCHAR;      goto REPEATCHAR;
3150    
3151      case OP_POSSTAR:      case OP_POSSTAR:
3152        case OP_POSSTARI:
3153      possessive = TRUE;      possessive = TRUE;
3154      min = 0;      min = 0;
3155      max = INT_MAX;      max = INT_MAX;
# Line 2283  for (;;) Line 3157  for (;;)
3157      goto REPEATCHAR;      goto REPEATCHAR;
3158    
3159      case OP_POSPLUS:      case OP_POSPLUS:
3160        case OP_POSPLUSI:
3161      possessive = TRUE;      possessive = TRUE;
3162      min = 1;      min = 1;
3163      max = INT_MAX;      max = INT_MAX;
# Line 2290  for (;;) Line 3165  for (;;)
3165      goto REPEATCHAR;      goto REPEATCHAR;
3166    
3167      case OP_POSQUERY:      case OP_POSQUERY:
3168        case OP_POSQUERYI:
3169      possessive = TRUE;      possessive = TRUE;
3170      min = 0;      min = 0;
3171      max = 1;      max = 1;
# Line 2297  for (;;) Line 3173  for (;;)
3173      goto REPEATCHAR;      goto REPEATCHAR;
3174    
3175      case OP_STAR:      case OP_STAR:
3176        case OP_STARI:
3177      case OP_MINSTAR:      case OP_MINSTAR:
3178        case OP_MINSTARI:
3179      case OP_PLUS:      case OP_PLUS:
3180        case OP_PLUSI:
3181      case OP_MINPLUS:      case OP_MINPLUS:
3182        case OP_MINPLUSI:
3183      case OP_QUERY:      case OP_QUERY:
3184        case OP_QUERYI:
3185      case OP_MINQUERY:      case OP_MINQUERY:
3186      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3187        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3188      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3189      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3190      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3191      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3192    
3193      /* Common code for all repeated single-character matches. We can give      /* Common code for all repeated single-character matches. */
     up quickly if there are fewer than the minimum number of characters left in  
     the subject. */  
3194    
3195      REPEATCHAR:      REPEATCHAR:
3196  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3197      if (utf8)      if (utf)
3198        {        {
3199        length = 1;        length = 1;
3200        charptr = ecode;        charptr = ecode;
3201        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3202        ecode += length;        ecode += length;
3203    
3204        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2329  for (;;) Line 3208  for (;;)
3208          {          {
3209  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3210          unsigned int othercase;          unsigned int othercase;
3211          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3212              (othercase = UCD_OTHERCASE(fc)) != fc)              (othercase = UCD_OTHERCASE(fc)) != fc)
3213            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3214          else oclength = 0;          else oclength = 0;
3215  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3216    
3217          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3218            {            {
3219            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3220                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3221  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3222            /* Need braces because of following else */            else if (oclength > 0 &&
3223            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3224                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3225    #endif  /* SUPPORT_UCP */
3226            else            else
3227              {              {
3228              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3229              eptr += oclength;              MRRETURN(MATCH_NOMATCH);
3230              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3231            }            }
3232    
3233          if (min == max) continue;          if (min == max) continue;
# Line 2357  for (;;) Line 3236  for (;;)
3236            {            {
3237            for (fi = min;; fi++)            for (fi = min;; fi++)
3238              {              {
3239              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3240              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3241              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) MRRETURN(MATCH_NOMATCH);
3242              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3243                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3244  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3245              /* Need braces because of following else */              else if (oclength > 0 &&
3246              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3247                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3248    #endif  /* SUPPORT_UCP */
3249              else              else
3250                {                {
3251                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3252                eptr += oclength;                MRRETURN(MATCH_NOMATCH);
3253                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3254              }              }
3255            /* Control never gets here */            /* Control never gets here */
3256            }            }
# Line 2381  for (;;) Line 3260  for (;;)
3260            pp = eptr;            pp = eptr;
3261            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3262              {              {
3263              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3264              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3265  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3266              else if (oclength == 0) break;              else if (oclength > 0 &&
3267                         eptr <= md->end_subject - oclength &&
3268                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3269    #endif  /* SUPPORT_UCP */
3270              else              else
3271                {                {
3272                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3273                eptr += oclength;                break;
3274                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3275              }              }
3276    
3277            if (possessive) continue;            if (possessive) continue;
3278    
3279            for(;;)            for(;;)
3280             {              {
3281             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3282             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3283             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { MRRETURN(MATCH_NOMATCH); }
3284  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3285             eptr--;              eptr--;
3286             BACKCHAR(eptr);              BACKCHAR(eptr);
3287  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3288             eptr -= length;              eptr -= length;
3289  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3290             }              }
3291            }            }
3292          /* Control never gets here */          /* Control never gets here */
3293          }          }
# Line 2417  for (;;) Line 3297  for (;;)
3297        value of fc will always be < 128. */        value of fc will always be < 128. */
3298        }        }
3299      else      else
3300  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3301          /* When not in UTF-8 mode, load a single-byte character. */
     /* When not in UTF-8 mode, load a single-byte character. */  
       {  
       if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3302        fc = *ecode++;        fc = *ecode++;
       }  
3303    
3304      /* The value of fc at this point is always less than 256, though we may or      /* The value of fc at this point is always one character, though we may
3305      may not be in UTF-8 mode. The code is duplicated for the caseless and      or may not be in UTF mode. The code is duplicated for the caseless and
3306      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3307      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3308      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2437  for (;;) Line 3313  for (;;)
3313      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3314        max, eptr));        max, eptr));
3315    
3316      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3317        {        {
3318        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3319          /* fc must be < 128 */
3320          foc = md->fcc[fc];
3321    #else
3322    #ifdef SUPPORT_UTF
3323    #ifdef SUPPORT_UCP
3324          if (utf && fc > 127)
3325            foc = UCD_OTHERCASE(fc);
3326    #else
3327          if (utf && fc > 127)
3328            foc = fc;
3329    #endif /* SUPPORT_UCP */
3330          else
3331    #endif /* SUPPORT_UTF */
3332            foc = TABLE_GET(fc, md->fcc, fc);
3333    #endif /* COMPILE_PCRE8 */
3334    
3335        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3336          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3337            if (eptr >= md->end_subject)
3338              {
3339              SCHECK_PARTIAL();
3340              MRRETURN(MATCH_NOMATCH);
3341              }
3342            if (fc != *eptr && foc != *eptr) MRRETURN(MATCH_NOMATCH);
3343            eptr++;
3344            }
3345        if (min == max) continue;        if (min == max) continue;
3346        if (minimize)        if (minimize)
3347          {          {
3348          for (fi = min;; fi++)          for (fi = min;; fi++)
3349            {            {
3350            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3351            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3352            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3353                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3354              RRETURN(MATCH_NOMATCH);              {
3355                SCHECK_PARTIAL();
3356                MRRETURN(MATCH_NOMATCH);
3357                }
3358              if (fc != *eptr && foc != *eptr) MRRETURN(MATCH_NOMATCH);
3359              eptr++;
3360            }            }
3361          /* Control never gets here */          /* Control never gets here */
3362          }          }
# Line 2460  for (;;) Line 3365  for (;;)
3365          pp = eptr;          pp = eptr;
3366          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3367            {            {
3368            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3369                {
3370                SCHECK_PARTIAL();
3371                break;
3372                }
3373              if (fc != *eptr && foc != *eptr) break;
3374            eptr++;            eptr++;
3375            }            }
3376    
3377          if (possessive) continue;          if (possessive) continue;
3378    
3379          while (eptr >= pp)          while (eptr >= pp)
3380            {            {
3381            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3382            eptr--;            eptr--;
3383            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3384            }            }
3385          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3386          }          }
3387        /* Control never gets here */        /* Control never gets here */
3388        }        }
# Line 2479  for (;;) Line 3391  for (;;)
3391    
3392      else      else
3393        {        {
3394        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3395            {
3396            if (eptr >= md->end_subject)
3397              {
3398              SCHECK_PARTIAL();
3399              MRRETURN(MATCH_NOMATCH);
3400              }
3401            if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3402            }
3403    
3404        if (min == max) continue;        if (min == max) continue;
3405    
3406        if (minimize)        if (minimize)
3407          {          {
3408          for (fi = min;; fi++)          for (fi = min;; fi++)
3409            {            {
3410            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3411            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3412            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) MRRETURN(MATCH_NOMATCH);
3413              RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
3414                {
3415                SCHECK_PARTIAL();
3416                MRRETURN(MATCH_NOMATCH);
3417                }
3418              if (fc != *eptr++) MRRETURN(MATCH_NOMATCH);
3419            }            }
3420          /* Control never gets here */          /* Control never gets here */
3421          }          }
# Line 2497  for (;;) Line 3424  for (;;)
3424          pp = eptr;          pp = eptr;
3425          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3426            {            {
3427            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3428                {
3429                SCHECK_PARTIAL();
3430                break;
3431                }
3432              if (fc != *eptr) break;
3433            eptr++;            eptr++;
3434            }            }
3435          if (possessive) continue;          if (possessive) continue;
3436    
3437          while (eptr >= pp)          while (eptr >= pp)
3438            {            {
3439            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3440            eptr--;            eptr--;
3441            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3442            }            }
3443          RRETURN(MATCH_NOMATCH);          MRRETURN(MATCH_NOMATCH);
3444          }          }
3445        }        }
3446      /* Control never gets here */      /* Control never gets here */
# Line 2516  for (;;) Line 3449  for (;;)
3449      checking can be multibyte. */      checking can be multibyte. */
3450    
3451      case OP_NOT:      case OP_NOT:
3452      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3453        if (eptr >= md->end_subject)
3454          {
3455          SCHECK_PARTIAL();
3456          MRRETURN(MATCH_NOMATCH);
3457          }
3458      ecode++;      ecode++;
3459      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);