/[pcre]/code/trunk/pcre_exec.c
ViewVC logotype

Diff of /code/trunk/pcre_exec.c

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

revision 231 by ph10, Tue Sep 11 11:15:33 2007 UTC revision 916 by ph10, Wed Feb 15 09:50:53 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 43  pattern matching using an NFA algorithm, Line 43  pattern matching using an NFA algorithm,
43  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK md             /* Block containing newline information */  #define NLBLOCK md             /* Block containing newline information */
# 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  /* 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.
86  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 95  static const char rep_max[] = { 0, 0, 0,
95    
96    
97    
98  #ifdef DEBUG  #ifdef PCRE_DEBUG
99  /*************************************************  /*************************************************
100  *        Debugging function to print chars       *  *        Debugging function to print chars       *
101  *************************************************/  *************************************************/
# Line 107  Returns:     nothing Line 113  Returns:     nothing
113  */  */
114    
115  static void  static void
116  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)
117  {  {
118  unsigned int c;  unsigned int c;
119  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 128  while (length-- > 0)
128  *          Match a back-reference                *  *          Match a back-reference                *
129  *************************************************/  *************************************************/
130    
131  /* 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
132  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
133    the length passed is zero. Note that in caseless UTF-8 mode, the number of
134    subject bytes matched may be different to the number of reference bytes.
135    
136  Arguments:  Arguments:
137    offset      index into the offset vector    offset      index into the offset vector
138    eptr        points into the subject    eptr        pointer into the subject
139    length      length to be matched    length      length of reference to be matched (number of bytes)
140    md          points to match data block    md          points to match data block
141    ims         the ims flags    caseless    TRUE if caseless
142    
143  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
144                  -1 no match
145                  -2 partial match; always given if at end subject
146  */  */
147    
148  static BOOL  static int
149  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,
150    unsigned long int ims)    BOOL caseless)
151  {  {
152  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
153    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
154    
155  #ifdef DEBUG  #ifdef PCRE_DEBUG
156  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
157    printf("matching subject <null>");    printf("matching subject <null>");
158  else  else
# Line 154  pchars(p, length, FALSE, md); Line 165  pchars(p, length, FALSE, md);
165  printf("\n");  printf("\n");
166  #endif  #endif
167    
168  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
169    case the length is passed as zero). */
170    
171  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
172    
173  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
174    properly if Unicode properties are supported. Otherwise, we can check only
175    ASCII characters. */
176    
177  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
178    {    {
179    while (length-- > 0)  #ifdef SUPPORT_UTF
180      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
181      if (md->utf)
182        {
183        /* Match characters up to the end of the reference. NOTE: the number of
184        bytes matched may differ, because there are some characters whose upper and
185        lower case versions code as different numbers of bytes. For example, U+023A
186        (2 bytes in UTF-8) is the upper case version of U+2C65 (3 bytes in UTF-8);
187        a sequence of 3 of the former uses 6 bytes, as does a sequence of two of
188        the latter. It is important, therefore, to check the length along the
189        reference, not along the subject (earlier code did this wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193          {
194          int c, d;
195          if (eptr >= md->end_subject) return -2;   /* Partial match */
196          GETCHARINC(c, eptr);
197          GETCHARINC(d, p);
198          if (c != d && c != UCD_OTHERCASE(d)) return -1;
199          }
200        }
201      else
202    #endif
203    #endif
204    
205      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
206      is no UCP support. */
207        {
208        while (length-- > 0)
209          {
210          if (eptr >= md->end_subject) return -2;   /* Partial match */
211          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
212          p++;
213          eptr++;
214          }
215        }
216    }    }
217    
218    /* In the caseful case, we can just compare the bytes, whether or not we
219    are in UTF-8 mode. */
220    
221  else  else
222    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
223      while (length-- > 0)
224        {
225        if (eptr >= md->end_subject) return -2;   /* Partial match */
226        if (*p++ != *eptr++) return -1;
227        }
228      }
229    
230  return TRUE;  return (int)(eptr - eptr_start);
231  }  }
232    
233    
# Line 219  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 278  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
278         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
279         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
280         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
281         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
282           RM61,  RM62, RM63, RM64, RM65, RM66 };
283    
284  /* 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
285  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
286  actuall used in this definition. */  actually used in this definition. */
287    
288  #ifndef NO_RECURSE  #ifndef NO_RECURSE
289  #define REGISTER register  #define REGISTER register
290    
291  #ifdef DEBUG  #ifdef PCRE_DEBUG
292  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
293    { \    { \
294    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
295    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
296    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
297    }    }
298  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 241  actuall used in this definition. */ Line 301  actuall used in this definition. */
301    return ra; \    return ra; \
302    }    }
303  #else  #else
304  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
305    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
306  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
307  #endif  #endif
308    
# Line 255  argument of match(), which never changes Line 315  argument of match(), which never changes
315    
316  #define REGISTER  #define REGISTER
317    
318  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
319    {\    {\
320    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
321      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
322    frame->Xwhere = rw; \    frame->Xwhere = rw; \
323    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
324    newframe->Xecode = rb;\    newframe->Xecode = rb;\
325    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
326    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
327    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
328    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
329    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
330    frame = newframe;\    frame = newframe;\
# Line 277  argument of match(), which never changes Line 336  argument of match(), which never changes
336    
337  #define RRETURN(ra)\  #define RRETURN(ra)\
338    {\    {\
339    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
340    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
341    (pcre_stack_free)(newframe);\    if (oldframe != &frame_zero) (PUBL(stack_free))(oldframe);\
342    if (frame != NULL)\    if (frame != NULL)\
343      {\      {\
344      rrc = ra;\      rrc = ra;\
# Line 296  typedef struct heapframe { Line 355  typedef struct heapframe {
355    
356    /* Function arguments that may change */    /* Function arguments that may change */
357    
358    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
359    const uschar *Xecode;    const pcre_uchar *Xecode;
360    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
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    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
369    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
370    const uschar *Xnext;  #endif
371    const uschar *Xpp;    PCRE_PUCHAR Xdata;
372    const uschar *Xprev;    PCRE_PUCHAR Xnext;
373    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
374      PCRE_PUCHAR Xprev;
375      PCRE_PUCHAR Xsaved_eptr;
376    
377    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
378    
# Line 321  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;
392    int Xctype;    int Xctype;
393    unsigned int Xfc;    unsigned int Xfc;
394    int Xfi;    int Xfi;
# Line 369  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    appear 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) RRETURN(PCRE_ERROR_PARTIAL); \
442        }
443    
444    #define SCHECK_PARTIAL()\
445      if (md->partial != 0 && eptr > md->start_used_ptr) \
446        { \
447        md->hitend = TRUE; \
448        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
449        }
450    
451  Performance note: It might be tempting to extract commonly used fields from the  
452  md structure (e.g. utf8, end_subject) into individual variables to improve  /* 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 383  Arguments: Line 461  Arguments:
461                   by encountering \K)                   by encountering \K)
462     offset_top  current top pointer     offset_top  current top pointer
463     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
464     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
465                   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  
466     rdepth      the recursion depth     rdepth      the recursion depth
467    
468  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
469                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
470                   a negative MATCH_xxx value for PRUNE, SKIP, etc
471                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
472                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
473  */  */
474    
475  static int  static int
476  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
477    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
478    int flags, unsigned int rdepth)    unsigned int rdepth)
479  {  {
480  /* 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,
481  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 410  so they can be ordinary variables in all Line 484  so they can be ordinary variables in all
484  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
485  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
486  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
487  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
488    
489  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
490    BOOL caseless;
491    int condcode;
492    
493  /* 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
494  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
495  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
496  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
497    the top-level on the stack rather than malloc-ing them all gives a performance
498    boost in many cases where there is not much "recursion". */
499    
500  #ifdef NO_RECURSE  #ifdef NO_RECURSE
501  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe frame_zero;
502    heapframe *frame = &frame_zero;
503  frame->Xprevframe = NULL;            /* Marks the top level */  frame->Xprevframe = NULL;            /* Marks the top level */
504    
505  /* Copy in the original argument variables */  /* Copy in the original argument variables */
# Line 429  frame->Xeptr = eptr; Line 508  frame->Xeptr = eptr;
508  frame->Xecode = ecode;  frame->Xecode = ecode;
509  frame->Xmstart = mstart;  frame->Xmstart = mstart;
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 444  HEAP_RECURSE: Line 521  HEAP_RECURSE:
521  #define ecode              frame->Xecode  #define ecode              frame->Xecode
522  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
523  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
524  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
525  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
526    
527  /* Ditto for the local variables */  /* Ditto for the local variables */
528    
529  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
530  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
531  #endif  #endif
532  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
533    #define codelink           frame->Xcodelink
534  #define data               frame->Xdata  #define data               frame->Xdata
535  #define next               frame->Xnext  #define next               frame->Xnext
536  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 467  HEAP_RECURSE: Line 543  HEAP_RECURSE:
543  #define condition          frame->Xcondition  #define condition          frame->Xcondition
544  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
545    
 #define original_ims       frame->Xoriginal_ims  
   
546  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
547  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
548  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
549  #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  
550  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
551  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
552  #endif  #endif
# Line 505  i, and fc and c, can be the same variabl Line 576  i, and fc and c, can be the same variabl
576  #define fi i  #define fi i
577  #define fc c  #define fc c
578    
579    /* Many of the following variables are used only in small blocks of the code.
580    My normal style of coding would have declared them within each of those blocks.
581    However, in order to accommodate the version of this code that uses an external
582    "stack" implemented on the heap, it is easier to declare them all here, so the
583    declarations can be cut out in a block. The only declarations within blocks
584    below are for variables that do not have to be preserved over a recursive call
585    to RMATCH(). */
586    
587    #ifdef SUPPORT_UTF
588    const pcre_uchar *charptr;
589    #endif
590    const pcre_uchar *callpat;
591    const pcre_uchar *data;
592    const pcre_uchar *next;
593    PCRE_PUCHAR       pp;
594    const pcre_uchar *prev;
595    PCRE_PUCHAR       saved_eptr;
596    
597    recursion_info new_recursive;
598    
599  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  BOOL cur_is_word;
 const uschar *charptr;             /* in small blocks of the code. My normal */  
 #endif                             /* style of coding would have declared    */  
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
600  BOOL condition;  BOOL condition;
601  BOOL prev_is_word;  BOOL prev_is_word;
602    
 unsigned long int original_ims;  
   
603  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
604  int prop_type;  int prop_type;
605  int prop_value;  int prop_value;
606  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
607  int oclength;  int oclength;
608  uschar occhars[8];  pcre_uchar occhars[6];
609  #endif  #endif
610    
611    int codelink;
612  int ctype;  int ctype;
613  int length;  int length;
614  int max;  int max;
# Line 547  int save_offset1, save_offset2, save_off Line 621  int save_offset1, save_offset2, save_off
621  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
622    
623  eptrblock newptrb;  eptrblock newptrb;
624    
625    /* There is a special fudge for calling match() in a way that causes it to
626    measure the size of its basic stack frame when the stack is being used for
627    recursion. The second argument (ecode) being NULL triggers this behaviour. It
628    cannot normally ever be NULL. The return is the negated value of the frame
629    size. */
630    
631    if (ecode == NULL)
632      {
633      if (rdepth == 0)
634        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
635      else
636        {
637        int len = (char *)&rdepth - (char *)eptr;
638        return (len > 0)? -len : len;
639        }
640      }
641  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
642    
643    /* To save space on the stack and in the heap frame, I have doubled up on some
644    of the local variables that are used only in localised parts of the code, but
645    still need to be preserved over recursive calls of match(). These macros define
646    the alternative names that are used. */
647    
648    #define allow_zero    cur_is_word
649    #define cbegroup      condition
650    #define code_offset   codelink
651    #define condassert    condition
652    #define matched_once  prev_is_word
653    #define foc           number
654    #define save_mark     data
655    
656  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
657  variables. */  variables. */
658    
# Line 568  TAIL_RECURSE: Line 672  TAIL_RECURSE:
672  /* 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
673  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
674  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()
675  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
676  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
677  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,
678  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
679    
680  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
681  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
682  #else  #else
683  utf8 = FALSE;  utf = FALSE;
684  #endif  #endif
685    
686  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
# Line 585  haven't exceeded the recursive call limi Line 689  haven't exceeded the recursive call limi
689  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
690  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
691    
 original_ims = ims;    /* Save for resetting on ')' */  
   
692  /* 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
693  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
694  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
695  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
696  When match() is called in other circumstances, don't add to the chain. The  
697  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
698  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
699  match(). */  to break infinite loops that match no characters. When match() is called in
700    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
701    NOT be used with tail recursion, because the memory block that is used is on
702    the stack, so a new one may be required for each match(). */
703    
704  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
705    {    {
706    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
707    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
708    eptrb = &newptrb;    eptrb = &newptrb;
709      md->match_function_type = 0;
710    }    }
711    
712  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 610  for (;;) Line 716  for (;;)
716    minimize = possessive = FALSE;    minimize = possessive = FALSE;
717    op = *ecode;    op = *ecode;
718    
   /* 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;  
   
719    switch(op)    switch(op)
720      {      {
721        case OP_MARK:
722        md->nomatch_mark = ecode + 2;
723        md->mark = NULL;    /* In case previously set by assertion */
724        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
725          eptrb, RM55);
726        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
727             md->mark == NULL) md->mark = ecode + 2;
728    
729        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
730        argument, and we must check whether that argument matches this MARK's
731        argument. It is passed back in md->start_match_ptr (an overloading of that
732        variable). If it does match, we reset that variable to the current subject
733        position and return MATCH_SKIP. Otherwise, pass back the return code
734        unaltered. */
735    
736        else if (rrc == MATCH_SKIP_ARG &&
737            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
738          {
739          md->start_match_ptr = eptr;
740          RRETURN(MATCH_SKIP);
741          }
742        RRETURN(rrc);
743    
744      case OP_FAIL:      case OP_FAIL:
745      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
746    
747      case OP_PRUNE:      /* COMMIT overrides PRUNE, SKIP, and THEN */
     RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,  
       ims, eptrb, flags, RM51);  
     if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
     RRETURN(MATCH_PRUNE);  
748    
749      case OP_COMMIT:      case OP_COMMIT:
750      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
751        ims, eptrb, flags, RM52);        eptrb, RM52);
752      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
753            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
754            rrc != MATCH_THEN)
755          RRETURN(rrc);
756      RRETURN(MATCH_COMMIT);      RRETURN(MATCH_COMMIT);
757    
758        /* PRUNE overrides THEN */
759    
760        case OP_PRUNE:
761        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
762          eptrb, RM51);
763        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
764        RRETURN(MATCH_PRUNE);
765    
766        case OP_PRUNE_ARG:
767        md->nomatch_mark = ecode + 2;
768        md->mark = NULL;    /* In case previously set by assertion */
769        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
770          eptrb, RM56);
771        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
772             md->mark == NULL) md->mark = ecode + 2;
773        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
774        RRETURN(MATCH_PRUNE);
775    
776        /* SKIP overrides PRUNE and THEN */
777    
778      case OP_SKIP:      case OP_SKIP:
779      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
780        ims, eptrb, flags, RM53);        eptrb, RM53);
781      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
782          RRETURN(rrc);
783      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
784      RRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
785    
786        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
787        nomatch_mark. There is a flag that disables this opcode when re-matching a
788        pattern that ended with a SKIP for which there was not a matching MARK. */
789    
790        case OP_SKIP_ARG:
791        if (md->ignore_skip_arg)
792          {
793          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
794          break;
795          }
796        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
797          eptrb, RM57);
798        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
799          RRETURN(rrc);
800    
801        /* Pass back the current skip name by overloading md->start_match_ptr and
802        returning the special MATCH_SKIP_ARG return code. This will either be
803        caught by a matching MARK, or get to the top, where it causes a rematch
804        with the md->ignore_skip_arg flag set. */
805    
806        md->start_match_ptr = ecode + 2;
807        RRETURN(MATCH_SKIP_ARG);
808    
809        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
810        the branch in which it occurs can be determined. Overload the start of
811        match pointer to do this. */
812    
813      case OP_THEN:      case OP_THEN:
814      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
815        ims, eptrb, flags, RM54);        eptrb, RM54);
816      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
817        md->start_match_ptr = ecode;
818      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
819    
820      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_THEN_ARG:
821      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
822      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
823      set from a previous iteration of this group, and be referred to by a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
824      reference inside the group.        md, eptrb, RM58);
825        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
826      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
827      values of the final offsets, in case they were set by a previous iteration      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
828      of the same bracket.      md->start_match_ptr = ecode;
829        RRETURN(MATCH_THEN);
830    
831        /* Handle an atomic group that does not contain any capturing parentheses.
832        This can be handled like an assertion. Prior to 8.13, all atomic groups
833        were handled this way. In 8.13, the code was changed as below for ONCE, so
834        that backups pass through the group and thereby reset captured values.
835        However, this uses a lot more stack, so in 8.20, atomic groups that do not
836        contain any captures generate OP_ONCE_NC, which can be handled in the old,
837        less stack intensive way.
838    
839        Check the alternative branches in turn - the matching won't pass the KET
840        for this kind of subpattern. If any one branch matches, we carry on as at
841        the end of a normal bracket, leaving the subject pointer, but resetting
842        the start-of-match value in case it was changed by \K. */
843    
844        case OP_ONCE_NC:
845        prev = ecode;
846        saved_eptr = eptr;
847        save_mark = md->mark;
848        do
849          {
850          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
851          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
852            {
853            mstart = md->start_match_ptr;
854            break;
855            }
856          if (rrc == MATCH_THEN)
857            {
858            next = ecode + GET(ecode,1);
859            if (md->start_match_ptr < next &&
860                (*ecode == OP_ALT || *next == OP_ALT))
861              rrc = MATCH_NOMATCH;
862            }
863    
864          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
865          ecode += GET(ecode,1);
866          md->mark = save_mark;
867          }
868        while (*ecode == OP_ALT);
869    
870        /* If hit the end of the group (which could be repeated), fail */
871    
872        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
873    
874        /* Continue as from after the group, updating the offsets high water
875        mark, since extracts may have been taken. */
876    
877        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
878    
879        offset_top = md->end_offset_top;
880        eptr = md->end_match_ptr;
881    
882        /* For a non-repeating ket, just continue at this level. This also
883        happens for a repeating ket if no characters were matched in the group.
884        This is the forcible breaking of infinite loops as implemented in Perl
885        5.005. */
886    
887        if (*ecode == OP_KET || eptr == saved_eptr)
888          {
889          ecode += 1+LINK_SIZE;
890          break;
891          }
892    
893        /* The repeating kets try the rest of the pattern or restart from the
894        preceding bracket, in the appropriate order. The second "call" of match()
895        uses tail recursion, to avoid using another stack frame. */
896    
897        if (*ecode == OP_KETRMIN)
898          {
899          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
900          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
901          ecode = prev;
902          goto TAIL_RECURSE;
903          }
904        else  /* OP_KETRMAX */
905          {
906          md->match_function_type = MATCH_CBEGROUP;
907          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
908          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
909          ecode += 1 + LINK_SIZE;
910          goto TAIL_RECURSE;
911          }
912        /* Control never gets here */
913    
914        /* Handle a capturing bracket, other than those that are possessive with an
915        unlimited repeat. If there is space in the offset vector, save the current
916        subject position in the working slot at the top of the vector. We mustn't
917        change the current values of the data slot, because they may be set from a
918        previous iteration of this group, and be referred to by a reference inside
919        the group. A failure to match might occur after the group has succeeded,
920        if something later on doesn't match. For this reason, we need to restore
921        the working value and also the values of the final offsets, in case they
922        were set by a previous iteration of the same bracket.
923    
924      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
925      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 667  for (;;) Line 930  for (;;)
930      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
931      offset = number << 1;      offset = number << 1;
932    
933  #ifdef DEBUG  #ifdef PCRE_DEBUG
934      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
935      printf("subject=");      printf("subject=");
936      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 680  for (;;) Line 943  for (;;)
943        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
944        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
945        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
946          save_mark = md->mark;
947    
948        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
949        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
950            (int)(eptr - md->start_subject);
951    
952        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
953          {          {
954          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
955            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
956          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
957            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
958    
959            /* If we backed up to a THEN, check whether it is within the current
960            branch by comparing the address of the THEN that is passed back with
961            the end of the branch. If it is within the current branch, and the
962            branch is one of two or more alternatives (it either starts or ends
963            with OP_ALT), we have reached the limit of THEN's action, so convert
964            the return code to NOMATCH, which will cause normal backtracking to
965            happen from now on. Otherwise, THEN is passed back to an outer
966            alternative. This implements Perl's treatment of parenthesized groups,
967            where a group not containing | does not affect the current alternative,
968            that is, (X) is NOT the same as (X|(*F)). */
969    
970            if (rrc == MATCH_THEN)
971              {
972              next = ecode + GET(ecode,1);
973              if (md->start_match_ptr < next &&
974                  (*ecode == OP_ALT || *next == OP_ALT))
975                rrc = MATCH_NOMATCH;
976              }
977    
978            /* Anything other than NOMATCH is passed back. */
979    
980            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
981          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
982          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
983            md->mark = save_mark;
984            if (*ecode != OP_ALT) break;
985          }          }
       while (*ecode == OP_ALT);  
986    
987        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
988        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
989        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
990        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
991    
992        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
993    
994          RRETURN(rrc);
995        }        }
996    
997      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 715  for (;;) Line 1005  for (;;)
1005      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1006      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1007    
1008      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1009      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
1010      recursive call to match() whatever happened. We can reduce stack usage by  
1011      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
1012      is set.*/      the result of a recursive call to match() whatever happened so it was
1013        possible to reduce stack usage by turning this into a tail recursion,
1014        except in the case of a possibly empty group. However, now that there is
1015        the possiblity of (*THEN) occurring in the final alternative, this
1016        optimization is no longer always possible.
1017    
1018        We can optimize if we know there are no (*THEN)s in the pattern; at present
1019        this is the best that can be done.
1020    
1021        MATCH_ONCE is returned when the end of an atomic group is successfully
1022        reached, but subsequent matching fails. It passes back up the tree (causing
1023        captured values to be reset) until the original atomic group level is
1024        reached. This is tested by comparing md->once_target with the start of the
1025        group. At this point, the return is converted into MATCH_NOMATCH so that
1026        previous backup points can be taken. */
1027    
1028        case OP_ONCE:
1029      case OP_BRA:      case OP_BRA:
1030      case OP_SBRA:      case OP_SBRA:
1031      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1032      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1033      for (;;)      for (;;)
1034        {        {
1035        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE) md->match_function_type = MATCH_CBEGROUP;
1036    
1037          /* If this is not a possibly empty group, and there are no (*THEN)s in
1038          the pattern, and this is the final alternative, optimize as described
1039          above. */
1040    
1041          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1042            {
1043            ecode += PRIV(OP_lengths)[*ecode];
1044            goto TAIL_RECURSE;
1045            }
1046    
1047          /* In all other cases, we have to make another call to match(). */
1048    
1049          save_mark = md->mark;
1050          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1051            RM2);
1052    
1053          /* See comment in the code for capturing groups above about handling
1054          THEN. */
1055    
1056          if (rrc == MATCH_THEN)
1057            {
1058            next = ecode + GET(ecode,1);
1059            if (md->start_match_ptr < next &&
1060                (*ecode == OP_ALT || *next == OP_ALT))
1061              rrc = MATCH_NOMATCH;
1062            }
1063    
1064          if (rrc != MATCH_NOMATCH)
1065          {          {
1066          if (flags == 0)    /* Not a possibly empty group */          if (rrc == MATCH_ONCE)
1067            {            {
1068            ecode += _pcre_OP_lengths[*ecode];            const pcre_uchar *scode = ecode;
1069            DPRINTF(("bracket 0 tail recursion\n"));            if (*scode != OP_ONCE)           /* If not at start, find it */
1070            goto TAIL_RECURSE;              {
1071                while (*scode == OP_ALT) scode += GET(scode, 1);
1072                scode -= GET(scode, 1);
1073                }
1074              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1075            }            }
1076            RRETURN(rrc);
1077            }
1078          ecode += GET(ecode, 1);
1079          md->mark = save_mark;
1080          if (*ecode != OP_ALT) break;
1081          }
1082    
1083          /* Possibly empty group; can't use tail recursion. */      RRETURN(MATCH_NOMATCH);
1084    
1085          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,      /* Handle possessive capturing brackets with an unlimited repeat. We come
1086            eptrb, flags, RM48);      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1087          RRETURN(rrc);      handled similarly to the normal case above. However, the matching is
1088        different. The end of these brackets will always be OP_KETRPOS, which
1089        returns MATCH_KETRPOS without going further in the pattern. By this means
1090        we can handle the group by iteration rather than recursion, thereby
1091        reducing the amount of stack needed. */
1092    
1093        case OP_CBRAPOS:
1094        case OP_SCBRAPOS:
1095        allow_zero = FALSE;
1096    
1097        POSSESSIVE_CAPTURE:
1098        number = GET2(ecode, 1+LINK_SIZE);
1099        offset = number << 1;
1100    
1101    #ifdef PCRE_DEBUG
1102        printf("start possessive bracket %d\n", number);
1103        printf("subject=");
1104        pchars(eptr, 16, TRUE, md);
1105        printf("\n");
1106    #endif
1107    
1108        if (offset < md->offset_max)
1109          {
1110          matched_once = FALSE;
1111          code_offset = (int)(ecode - md->start_code);
1112    
1113          save_offset1 = md->offset_vector[offset];
1114          save_offset2 = md->offset_vector[offset+1];
1115          save_offset3 = md->offset_vector[md->offset_end - number];
1116          save_capture_last = md->capture_last;
1117    
1118          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1119    
1120          /* Each time round the loop, save the current subject position for use
1121          when the group matches. For MATCH_MATCH, the group has matched, so we
1122          restart it with a new subject starting position, remembering that we had
1123          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1124          usual. If we haven't matched any alternatives in any iteration, check to
1125          see if a previous iteration matched. If so, the group has matched;
1126          continue from afterwards. Otherwise it has failed; restore the previous
1127          capture values before returning NOMATCH. */
1128    
1129          for (;;)
1130            {
1131            md->offset_vector[md->offset_end - number] =
1132              (int)(eptr - md->start_subject);
1133            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1134            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1135              eptrb, RM63);
1136            if (rrc == MATCH_KETRPOS)
1137              {
1138              offset_top = md->end_offset_top;
1139              eptr = md->end_match_ptr;
1140              ecode = md->start_code + code_offset;
1141              save_capture_last = md->capture_last;
1142              matched_once = TRUE;
1143              continue;
1144              }
1145    
1146            /* See comment in the code for capturing groups above about handling
1147            THEN. */
1148    
1149            if (rrc == MATCH_THEN)
1150              {
1151              next = ecode + GET(ecode,1);
1152              if (md->start_match_ptr < next &&
1153                  (*ecode == OP_ALT || *next == OP_ALT))
1154                rrc = MATCH_NOMATCH;
1155              }
1156    
1157            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1158            md->capture_last = save_capture_last;
1159            ecode += GET(ecode, 1);
1160            if (*ecode != OP_ALT) break;
1161            }
1162    
1163          if (!matched_once)
1164            {
1165            md->offset_vector[offset] = save_offset1;
1166            md->offset_vector[offset+1] = save_offset2;
1167            md->offset_vector[md->offset_end - number] = save_offset3;
1168          }          }
1169    
1170        /* For non-final alternatives, continue the loop for a NOMATCH result;        if (allow_zero || matched_once)
1171        otherwise return. */          {
1172            ecode += 1 + LINK_SIZE;
1173            break;
1174            }
1175    
1176        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        RRETURN(MATCH_NOMATCH);
1177          eptrb, flags, RM2);        }
1178        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
1179        /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1180        as a non-capturing bracket. */
1181    
1182        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1183        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1184    
1185        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1186    
1187        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1188        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1189    
1190        /* Non-capturing possessive bracket with unlimited repeat. We come here
1191        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1192        without the capturing complication. It is written out separately for speed
1193        and cleanliness. */
1194    
1195        case OP_BRAPOS:
1196        case OP_SBRAPOS:
1197        allow_zero = FALSE;
1198    
1199        POSSESSIVE_NON_CAPTURE:
1200        matched_once = FALSE;
1201        code_offset = (int)(ecode - md->start_code);
1202    
1203        for (;;)
1204          {
1205          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1206          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1207            eptrb, RM48);
1208          if (rrc == MATCH_KETRPOS)
1209            {
1210            offset_top = md->end_offset_top;
1211            eptr = md->end_match_ptr;
1212            ecode = md->start_code + code_offset;
1213            matched_once = TRUE;
1214            continue;
1215            }
1216    
1217          /* See comment in the code for capturing groups above about handling
1218          THEN. */
1219    
1220          if (rrc == MATCH_THEN)
1221            {
1222            next = ecode + GET(ecode,1);
1223            if (md->start_match_ptr < next &&
1224                (*ecode == OP_ALT || *next == OP_ALT))
1225              rrc = MATCH_NOMATCH;
1226            }
1227    
1228          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1229        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1230          if (*ecode != OP_ALT) break;
1231          }
1232    
1233        if (matched_once || allow_zero)
1234          {
1235          ecode += 1 + LINK_SIZE;
1236          break;
1237        }        }
1238        RRETURN(MATCH_NOMATCH);
1239    
1240      /* Control never reaches here. */      /* Control never reaches here. */
1241    
1242      /* Conditional group: compilation checked that there are no more than      /* Conditional group: compilation checked that there are no more than
1243      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
1244      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
1245      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. */  
1246    
1247      case OP_COND:      case OP_COND:
1248      case OP_SCOND:      case OP_SCOND:
1249      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      codelink = GET(ecode, 1);
1250    
1251        /* Because of the way auto-callout works during compile, a callout item is
1252        inserted between OP_COND and an assertion condition. */
1253    
1254        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1255        {        {
1256        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/        if (PUBL(callout) != NULL)
1257        condition = md->recursive != NULL &&          {
1258          (offset == RREF_ANY || offset == md->recursive->group_num);          PUBL(callout_block) cb;
1259        ecode += condition? 3 : GET(ecode, 1);          cb.version          = 2;   /* Version 1 of the callout block */
1260            cb.callout_number   = ecode[LINK_SIZE+2];
1261            cb.offset_vector    = md->offset_vector;
1262    #ifdef COMPILE_PCRE8
1263            cb.subject          = (PCRE_SPTR)md->start_subject;
1264    #else
1265            cb.subject          = (PCRE_SPTR16)md->start_subject;
1266    #endif
1267            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1268            cb.start_match      = (int)(mstart - md->start_subject);
1269            cb.current_position = (int)(eptr - md->start_subject);
1270            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1271            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1272            cb.capture_top      = offset_top/2;
1273            cb.capture_last     = md->capture_last;
1274            cb.callout_data     = md->callout_data;
1275            cb.mark             = md->nomatch_mark;
1276            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1277            if (rrc < 0) RRETURN(rrc);
1278            }
1279          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1280        }        }
1281    
1282      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      condcode = ecode[LINK_SIZE+1];
1283    
1284        /* Now see what the actual condition is */
1285    
1286        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1287          {
1288          if (md->recursive == NULL)                /* Not recursing => FALSE */
1289            {
1290            condition = FALSE;
1291            ecode += GET(ecode, 1);
1292            }
1293          else
1294            {
1295            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1296            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1297    
1298            /* If the test is for recursion into a specific subpattern, and it is
1299            false, but the test was set up by name, scan the table to see if the
1300            name refers to any other numbers, and test them. The condition is true
1301            if any one is set. */
1302    
1303            if (!condition && condcode == OP_NRREF)
1304              {
1305              pcre_uchar *slotA = md->name_table;
1306              for (i = 0; i < md->name_count; i++)
1307                {
1308                if (GET2(slotA, 0) == recno) break;
1309                slotA += md->name_entry_size;
1310                }
1311    
1312              /* Found a name for the number - there can be only one; duplicate
1313              names for different numbers are allowed, but not vice versa. First
1314              scan down for duplicates. */
1315    
1316              if (i < md->name_count)
1317                {
1318                pcre_uchar *slotB = slotA;
1319                while (slotB > md->name_table)
1320                  {
1321                  slotB -= md->name_entry_size;
1322                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1323                    {
1324                    condition = GET2(slotB, 0) == md->recursive->group_num;
1325                    if (condition) break;
1326                    }
1327                  else break;
1328                  }
1329    
1330                /* Scan up for duplicates */
1331    
1332                if (!condition)
1333                  {
1334                  slotB = slotA;
1335                  for (i++; i < md->name_count; i++)
1336                    {
1337                    slotB += md->name_entry_size;
1338                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1339                      {
1340                      condition = GET2(slotB, 0) == md->recursive->group_num;
1341                      if (condition) break;
1342                      }
1343                    else break;
1344                    }
1345                  }
1346                }
1347              }
1348    
1349            /* Chose branch according to the condition */
1350    
1351            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1352            }
1353          }
1354    
1355        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1356        {        {
1357        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1358        condition = offset < offset_top && md->offset_vector[offset] >= 0;        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1359        ecode += condition? 3 : GET(ecode, 1);  
1360          /* If the numbered capture is unset, but the reference was by name,
1361          scan the table to see if the name refers to any other numbers, and test
1362          them. The condition is true if any one is set. This is tediously similar
1363          to the code above, but not close enough to try to amalgamate. */
1364    
1365          if (!condition && condcode == OP_NCREF)
1366            {
1367            int refno = offset >> 1;
1368            pcre_uchar *slotA = md->name_table;
1369    
1370            for (i = 0; i < md->name_count; i++)
1371              {
1372              if (GET2(slotA, 0) == refno) break;
1373              slotA += md->name_entry_size;
1374              }
1375    
1376            /* Found a name for the number - there can be only one; duplicate names
1377            for different numbers are allowed, but not vice versa. First scan down
1378            for duplicates. */
1379    
1380            if (i < md->name_count)
1381              {
1382              pcre_uchar *slotB = slotA;
1383              while (slotB > md->name_table)
1384                {
1385                slotB -= md->name_entry_size;
1386                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1387                  {
1388                  offset = GET2(slotB, 0) << 1;
1389                  condition = offset < offset_top &&
1390                    md->offset_vector[offset] >= 0;
1391                  if (condition) break;
1392                  }
1393                else break;
1394                }
1395    
1396              /* Scan up for duplicates */
1397    
1398              if (!condition)
1399                {
1400                slotB = slotA;
1401                for (i++; i < md->name_count; i++)
1402                  {
1403                  slotB += md->name_entry_size;
1404                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1405                    {
1406                    offset = GET2(slotB, 0) << 1;
1407                    condition = offset < offset_top &&
1408                      md->offset_vector[offset] >= 0;
1409                    if (condition) break;
1410                    }
1411                  else break;
1412                  }
1413                }
1414              }
1415            }
1416    
1417          /* Chose branch according to the condition */
1418    
1419          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1420        }        }
1421    
1422      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      else if (condcode == OP_DEF)     /* DEFINE - always false */
1423        {        {
1424        condition = FALSE;        condition = FALSE;
1425        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1426        }        }
1427    
1428      /* The condition is an assertion. Call match() to evaluate it - setting      /* The condition is an assertion. Call match() to evaluate it - setting
1429      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
1430      assertion. */      an assertion. */
1431    
1432      else      else
1433        {        {
1434        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        md->match_function_type = MATCH_CONDASSERT;
1435            match_condassert, RM3);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1436        if (rrc == MATCH_MATCH)        if (rrc == MATCH_MATCH)
1437          {          {
1438            if (md->end_offset_top > offset_top)
1439              offset_top = md->end_offset_top;  /* Captures may have happened */
1440          condition = TRUE;          condition = TRUE;
1441          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1442          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1443          }          }
1444    
1445          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1446          assertion; it is therefore treated as NOMATCH. */
1447    
1448        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1449          {          {
1450          RRETURN(rrc);         /* Need braces because of following else */          RRETURN(rrc);         /* Need braces because of following else */
# Line 803  for (;;) Line 1452  for (;;)
1452        else        else
1453          {          {
1454          condition = FALSE;          condition = FALSE;
1455          ecode += GET(ecode, 1);          ecode += codelink;
1456          }          }
1457        }        }
1458    
1459      /* 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
1460      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
1461      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
1462      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
1463        exist, in which case we can just plough on. Note that, for compatibility
1464        with Perl, the | in a conditional group is NOT treated as creating two
1465        alternatives. If a THEN is encountered in the branch, it propagates out to
1466        the enclosing alternative (unless nested in a deeper set of alternatives,
1467        of course). */
1468    
1469      if (condition || *ecode == OP_ALT)      if (condition || *ecode == OP_ALT)
1470        {        {
1471        ecode += 1 + LINK_SIZE;        if (op != OP_SCOND)
       if (op == OP_SCOND)        /* Possibly empty group */  
1472          {          {
1473          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);          ecode += 1 + LINK_SIZE;
         RRETURN(rrc);  
         }  
       else                       /* Group must match something */  
         {  
         flags = 0;  
1474          goto TAIL_RECURSE;          goto TAIL_RECURSE;
1475          }          }
1476    
1477          md->match_function_type = MATCH_CBEGROUP;
1478          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1479          RRETURN(rrc);
1480        }        }
1481      else                         /* Condition false & no 2nd alternative */  
1482         /* Condition false & no alternative; continue after the group. */
1483    
1484        else
1485        {        {
1486        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1487        }        }
1488      break;      break;
1489    
1490    
1491      /* 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,
1492      recursion, we should restore the offsets appropriately and continue from      to close any currently open capturing brackets. */
     after the call. */  
1493    
1494      case OP_ACCEPT:      case OP_CLOSE:
1495      case OP_END:      number = GET2(ecode, 1);
1496      if (md->recursive != NULL && md->recursive->group_num == 0)      offset = number << 1;
1497    
1498    #ifdef PCRE_DEBUG
1499          printf("end bracket %d at *ACCEPT", number);
1500          printf("\n");
1501    #endif
1502    
1503        md->capture_last = number;
1504        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1505        {        {
1506        recursion_info *rec = md->recursive;        md->offset_vector[offset] =
1507        DPRINTF(("End of pattern in a (?0) recursion\n"));          md->offset_vector[md->offset_end - number];
1508        md->recursive = rec->prevrec;        md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1509        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;  
1510        }        }
1511        ecode += 1 + IMM2_SIZE;
1512        break;
1513    
     /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty  
     string - backtracking will then try other alternatives, if any. */  
1514    
1515      if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);      /* End of the pattern, either real or forced. */
1516    
1517        case OP_END:
1518        case OP_ACCEPT:
1519        case OP_ASSERT_ACCEPT:
1520    
1521        /* If we have matched an empty string, fail if not in an assertion and not
1522        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1523        is set and we have matched at the start of the subject. In both cases,
1524        backtracking will then try other alternatives, if any. */
1525    
1526        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1527             md->recursive == NULL &&
1528             (md->notempty ||
1529               (md->notempty_atstart &&
1530                 mstart == md->start_subject + md->start_offset)))
1531          RRETURN(MATCH_NOMATCH);
1532    
1533        /* Otherwise, we have a match. */
1534    
1535      md->end_match_ptr = eptr;           /* Record where we ended */      md->end_match_ptr = eptr;           /* Record where we ended */
1536      md->end_offset_top = offset_top;    /* and how many extracts were taken */      md->end_offset_top = offset_top;    /* and how many extracts were taken */
1537      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);  
1538    
1539      /* Change option settings */      /* For some reason, the macros don't work properly if an expression is
1540        given as the argument to RRETURN when the heap is in use. */
1541    
1542      case OP_OPT:      rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1543      ims = ecode[1];      RRETURN(rrc);
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1544    
1545      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1546      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,
1547      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
1548      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
1549      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1550        of a condition, we want to return immediately afterwards. The caller of
1551        this incarnation of the match() function will have set MATCH_CONDASSERT in
1552        md->match_function type, and one of these opcodes will be the first opcode
1553        that is processed. We use a local variable that is preserved over calls to
1554        match() to remember this case. */
1555    
1556      case OP_ASSERT:      case OP_ASSERT:
1557      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1558        save_mark = md->mark;
1559        if (md->match_function_type == MATCH_CONDASSERT)
1560          {
1561          condassert = TRUE;
1562          md->match_function_type = 0;
1563          }
1564        else condassert = FALSE;
1565    
1566      do      do
1567        {        {
1568        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1569          RM4);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1570        if (rrc == MATCH_MATCH) break;          {
1571            mstart = md->start_match_ptr;   /* In case \K reset it */
1572            break;
1573            }
1574    
1575          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1576          as NOMATCH. */
1577    
1578        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1579        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1580          md->mark = save_mark;
1581        }        }
1582      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1583    
1584      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1585    
1586      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1587    
1588      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1589    
1590      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1591      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 900  for (;;) Line 1595  for (;;)
1595      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1596      continue;      continue;
1597    
1598      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1599        PRUNE, or COMMIT means we must assume failure without checking subsequent
1600        branches. */
1601    
1602      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1603      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1604        save_mark = md->mark;
1605        if (md->match_function_type == MATCH_CONDASSERT)
1606          {
1607          condassert = TRUE;
1608          md->match_function_type = 0;
1609          }
1610        else condassert = FALSE;
1611    
1612      do      do
1613        {        {
1614        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1615          RM5);        md->mark = save_mark;
1616        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1617          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1618            {
1619            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1620            break;
1621            }
1622    
1623          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1624          as NOMATCH. */
1625    
1626        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1627        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1628        }        }
1629      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1630    
1631      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1632    
1633      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1634      continue;      continue;
# Line 925  for (;;) Line 1639  for (;;)
1639      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1640    
1641      case OP_REVERSE:      case OP_REVERSE:
1642  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1643      if (utf8)      if (utf)
1644        {        {
1645        i = GET(ecode, 1);        i = GET(ecode, 1);
1646        while (i-- > 0)        while (i-- > 0)
# Line 946  for (;;) Line 1660  for (;;)
1660        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1661        }        }
1662    
1663      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1664    
1665        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1666      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1667      break;      break;
1668    
# Line 956  for (;;) Line 1671  for (;;)
1671      function is able to force a failure. */      function is able to force a failure. */
1672    
1673      case OP_CALLOUT:      case OP_CALLOUT:
1674      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1675        {        {
1676        pcre_callout_block cb;        PUBL(callout_block) cb;
1677        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1678        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1679        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1680    #ifdef COMPILE_PCRE8
1681        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1682        cb.subject_length   = md->end_subject - md->start_subject;  #else
1683        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1684        cb.current_position = eptr - md->start_subject;  #endif
1685          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1686          cb.start_match      = (int)(mstart - md->start_subject);
1687          cb.current_position = (int)(eptr - md->start_subject);
1688        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1689        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1690        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1691        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1692        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1693        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1694          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1695        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1696        }        }
1697      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 981  for (;;) Line 1701  for (;;)
1701      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
1702      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1703    
1704      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1705      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
1706      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
1707      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
1708      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
1709      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
1710      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.  
1711    
1712      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
1713      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
1714      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1715        a lot, so he is not to blame for the current way it works. */
1716    
1717      case OP_RECURSE:      case OP_RECURSE:
1718        {        {
1719          recursion_info *ri;
1720          int recno;
1721    
1722        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1723        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1724          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1725    
1726          /* Check for repeating a recursion without advancing the subject pointer.
1727          This should catch convoluted mutual recursions. (Some simple cases are
1728          caught at compile time.) */
1729    
1730          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1731            if (recno == ri->group_num && eptr == ri->subject_position)
1732              RRETURN(PCRE_ERROR_RECURSELOOP);
1733    
1734        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1735    
1736          new_recursive.group_num = recno;
1737          new_recursive.subject_position = eptr;
1738        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1739        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1740    
1741        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1742    
1743        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1744    
1745        /* Now save the offset data. */        /* Now save the offset data */
1746    
1747        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1748        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1020  for (;;) Line 1750  for (;;)
1750        else        else
1751          {          {
1752          new_recursive.offset_save =          new_recursive.offset_save =
1753            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1754          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1755          }          }
   
1756        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1757              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1758    
1759        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1760        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1761          might be changed, so reset it before looping. */
1762    
1763        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1764        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1765        do        do
1766          {          {
1767          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1768            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1769          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1770            memcpy(md->offset_vector, new_recursive.offset_save,
1771                new_recursive.saved_max * sizeof(int));
1772            md->recursive = new_recursive.prevrec;
1773            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1774            {            {
1775            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1776            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1777              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1778            RRETURN(MATCH_MATCH);  
1779              /* Set where we got to in the subject, and reset the start in case
1780              it was changed by \K. This *is* propagated back out of a recursion,
1781              for Perl compatibility. */
1782    
1783              eptr = md->end_match_ptr;
1784              mstart = md->start_match_ptr;
1785              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1786            }            }
1787    
1788            /* PCRE does not allow THEN to escape beyond a recursion; it is treated
1789            as NOMATCH. */
1790    
1791          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1792            {            {
1793            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1794              if (new_recursive.offset_save != stacksave)
1795                (PUBL(free))(new_recursive.offset_save);
1796            RRETURN(rrc);            RRETURN(rrc);
1797            }            }
1798    
1799          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1800          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1801          }          }
1802        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1062  for (;;) Line 1804  for (;;)
1804        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1805        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1806        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1807          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1808        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1809        }        }
     /* 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);  
1810    
1811      /* Continue as from after the assertion, updating the offsets high water      RECURSION_MATCHED:
1812      mark, since extracts may have been taken. */      break;
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1813    
1814      /* 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
1815      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1148  for (;;) Line 1818  for (;;)
1818      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1819      break;      break;
1820    
1821      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1822      that it may occur zero times. It may repeat infinitely, or not at all -      indicating that it may occur zero times. It may repeat infinitely, or not
1823      i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1824      repeat limits are compiled as a number of copies, with the optional ones      with fixed upper repeat limits are compiled as a number of copies, with the
1825      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1826    
1827      case OP_BRAZERO:      case OP_BRAZERO:
1828        {      next = ecode + 1;
1829        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1830        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1831        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1832        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1833      break;      break;
1834    
1835      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1836        {      next = ecode + 1;
1837        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1838        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1839        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1840        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1841      break;      break;
1842    
1843        case OP_SKIPZERO:
1844        next = ecode+1;
1845        do next += GET(next,1); while (*next == OP_ALT);
1846        ecode = next + 1 + LINK_SIZE;
1847        break;
1848    
1849        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1850        here; just jump to the group, with allow_zero set TRUE. */
1851    
1852        case OP_BRAPOSZERO:
1853        op = *(++ecode);
1854        allow_zero = TRUE;
1855        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1856          goto POSSESSIVE_NON_CAPTURE;
1857    
1858      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1859    
1860      case OP_KET:      case OP_KET:
1861      case OP_KETRMIN:      case OP_KETRMIN:
1862      case OP_KETRMAX:      case OP_KETRMAX:
1863        case OP_KETRPOS:
1864      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1865    
1866      /* 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
1867      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1868      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1869    
1870      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1871        {        {
1872        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1873        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1874        }        }
1875      else saved_eptr = NULL;      else saved_eptr = NULL;
1876    
1877      /* 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
1878      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
1879      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1880        start in case it was changed by \K. */
1881      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1882          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1883          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1884        {        {
1885        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1886        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1887        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1888          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1889        }        }
1890    
1891      /* 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
1892      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1893      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
1894      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
1895      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
1896        the current subject position and start match pointer and give a MATCH
1897        return. */
1898    
1899      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1900            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1901        {        {
1902        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1903        offset = number << 1;        offset = number << 1;
1904    
1905  #ifdef DEBUG  #ifdef PCRE_DEBUG
1906        printf("end bracket %d", number);        printf("end bracket %d", number);
1907        printf("\n");        printf("\n");
1908  #endif  #endif
1909    
1910          /* Handle a recursively called group. */
1911    
1912          if (md->recursive != NULL && md->recursive->group_num == number)
1913            {
1914            md->end_match_ptr = eptr;
1915            md->start_match_ptr = mstart;
1916            RRETURN(MATCH_MATCH);
1917            }
1918    
1919          /* Deal with capturing */
1920    
1921        md->capture_last = number;        md->capture_last = number;
1922        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1923          {          {
1924            /* If offset is greater than offset_top, it means that we are
1925            "skipping" a capturing group, and that group's offsets must be marked
1926            unset. In earlier versions of PCRE, all the offsets were unset at the
1927            start of matching, but this doesn't work because atomic groups and
1928            assertions can cause a value to be set that should later be unset.
1929            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1930            part of the atomic group, but this is not on the final matching path,
1931            so must be unset when 2 is set. (If there is no group 2, there is no
1932            problem, because offset_top will then be 2, indicating no capture.) */
1933    
1934            if (offset > offset_top)
1935              {
1936              register int *iptr = md->offset_vector + offset_top;
1937              register int *iend = md->offset_vector + offset;
1938              while (iptr < iend) *iptr++ = -1;
1939              }
1940    
1941            /* Now make the extraction */
1942    
1943          md->offset_vector[offset] =          md->offset_vector[offset] =
1944            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1945          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1946          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1947          }          }
1948          }
1949    
1950        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1951        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1952        group. This is the forcible breaking of infinite loops as implemented in
1953        Perl 5.005. For a non-repeating atomic group that includes captures,
1954        establish a backup point by processing the rest of the pattern at a lower
1955        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1956        original OP_ONCE level, thereby bypassing intermediate backup points, but
1957        resetting any captures that happened along the way. */
1958    
1959        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1960          {
1961          if (*prev == OP_ONCE)
1962          {          {
1963          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1964          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1965          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1966          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;  
1967          }          }
1968          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1969          break;
1970        }        }
1971    
1972      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1973      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1974        at a time from the outer level, thus saving stack. */
     ims = original_ims;  
     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. */  
1975    
1976      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KETRPOS)
1977        {        {
1978        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
1979        break;        md->end_offset_top = offset_top;
1980          RRETURN(MATCH_KETRPOS);
1981        }        }
1982    
1983      /* 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
1984      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
1985      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
1986      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
1987        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
1988    
1989      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
1990        {        {
1991        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
1992        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1993        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
1994            {
1995            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
1996            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1997            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1998            RRETURN(MATCH_ONCE);
1999            }
2000          if (*prev >= OP_SBRA)    /* Could match an empty string */
2001          {          {
2002          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          md->match_function_type = MATCH_CBEGROUP;
2003            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2004          RRETURN(rrc);          RRETURN(rrc);
2005          }          }
2006        ecode = prev;        ecode = prev;
# Line 1286  for (;;) Line 2008  for (;;)
2008        }        }
2009      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2010        {        {
2011        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        if (*prev >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
2012          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2013          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2014        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2015          if (*prev == OP_ONCE)
2016            {
2017            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2018            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2019            md->once_target = prev;
2020            RRETURN(MATCH_ONCE);
2021            }
2022        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2023        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2024        }        }
2025      /* Control never gets here */      /* Control never gets here */
2026    
2027      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2028    
2029      case OP_CIRC:      case OP_CIRC:
2030      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2031    
2032      /* Start of subject assertion */      /* Start of subject assertion */
2033    
# Line 1315  for (;;) Line 2036  for (;;)
2036      ecode++;      ecode++;
2037      break;      break;
2038    
2039        /* Multiline mode: start of subject unless notbol, or after any newline. */
2040    
2041        case OP_CIRCM:
2042        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2043        if (eptr != md->start_subject &&
2044            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2045          RRETURN(MATCH_NOMATCH);
2046        ecode++;
2047        break;
2048    
2049      /* Start of match assertion */      /* Start of match assertion */
2050    
2051      case OP_SOM:      case OP_SOM:
# Line 1329  for (;;) Line 2060  for (;;)
2060      ecode++;      ecode++;
2061      break;      break;
2062    
2063      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2064      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2065    
2066      case OP_DOLL:      case OP_DOLLM:
2067      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2068        {        {
2069        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2070          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2071        else          if (eptr + 1 >= md->end_subject &&
2072          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              md->partial != 0 &&
2073        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2074        break;              NLBLOCK->nllen == 2 &&
2075                *eptr == NLBLOCK->nl[0])
2076              {
2077              md->hitend = TRUE;
2078              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2079              }
2080            RRETURN(MATCH_NOMATCH);
2081            }
2082        }        }
2083      else      else
2084        {        {
2085        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2086        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2087        }        }
2088        ecode++;
2089        break;
2090    
2091        /* Not multiline mode: assert before a terminating newline or before end of
2092        subject unless noteol is set. */
2093    
2094        case OP_DOLL:
2095        if (md->noteol) RRETURN(MATCH_NOMATCH);
2096        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2097    
2098      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2099    
2100      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2101    
2102      case OP_EOD:      case OP_EOD:
2103      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2104        SCHECK_PARTIAL();
2105      ecode++;      ecode++;
2106      break;      break;
2107    
2108      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2109    
2110      case OP_EODN:      case OP_EODN:
2111      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2112        if (eptr < md->end_subject &&
2113          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2114          {
2115          if (eptr + 1 >= md->end_subject &&
2116              md->partial != 0 &&
2117              NLBLOCK->nltype == NLTYPE_FIXED &&
2118              NLBLOCK->nllen == 2 &&
2119              *eptr == NLBLOCK->nl[0])
2120            {
2121            md->hitend = TRUE;
2122            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2123            }
2124        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2125          }
2126    
2127        /* Either at end of string or \n before end. */
2128    
2129        SCHECK_PARTIAL();
2130      ecode++;      ecode++;
2131      break;      break;
2132    
# Line 1380  for (;;) Line 2138  for (;;)
2138    
2139        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2140        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
2141        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2142          partial matching. */
2143    
2144  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2145        if (utf8)        if (utf)
2146          {          {
2147            /* Get status of previous character */
2148    
2149          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2150            {            {
2151            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2152            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2153              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2154            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2155    #ifdef SUPPORT_UCP
2156              if (md->use_ucp)
2157                {
2158                if (c == '_') prev_is_word = TRUE; else
2159                  {
2160                  int cat = UCD_CATEGORY(c);
2161                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2162                  }
2163                }
2164              else
2165    #endif
2166            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2167            }            }
2168          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2169            /* Get status of next character */
2170    
2171            if (eptr >= md->end_subject)
2172              {
2173              SCHECK_PARTIAL();
2174              cur_is_word = FALSE;
2175              }
2176            else
2177            {            {
2178            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2179    #ifdef SUPPORT_UCP
2180              if (md->use_ucp)
2181                {
2182                if (c == '_') cur_is_word = TRUE; else
2183                  {
2184                  int cat = UCD_CATEGORY(c);
2185                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2186                  }
2187                }
2188              else
2189    #endif
2190            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2191            }            }
2192          }          }
2193        else        else
2194    #endif
2195    
2196          /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2197          consistency with the behaviour of \w we do use it in this case. */
2198    
2199            {
2200            /* Get status of previous character */
2201    
2202            if (eptr == md->start_subject) prev_is_word = FALSE; else
2203              {
2204              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2205    #ifdef SUPPORT_UCP
2206              if (md->use_ucp)
2207                {
2208                c = eptr[-1];
2209                if (c == '_') prev_is_word = TRUE; else
2210                  {
2211                  int cat = UCD_CATEGORY(c);
2212                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2213                  }
2214                }
2215              else
2216    #endif
2217              prev_is_word = MAX_255(eptr[-1])
2218                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2219              }
2220    
2221            /* Get status of next character */
2222    
2223            if (eptr >= md->end_subject)
2224              {
2225              SCHECK_PARTIAL();
2226              cur_is_word = FALSE;
2227              }
2228            else
2229    #ifdef SUPPORT_UCP
2230            if (md->use_ucp)
2231              {
2232              c = *eptr;
2233              if (c == '_') cur_is_word = TRUE; else
2234                {
2235                int cat = UCD_CATEGORY(c);
2236                cur_is_word = (cat == ucp_L || cat == ucp_N);
2237                }
2238              }
2239            else
2240  #endif  #endif
2241            cur_is_word = MAX_255(*eptr)
2242        /* More streamlined when not in UTF-8 mode */            && ((md->ctypes[*eptr] & ctype_word) != 0);
   
         {  
         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);  
2243          }          }
2244    
2245        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1421  for (;;) Line 2253  for (;;)
2253      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
2254    
2255      case OP_ANY:      case OP_ANY:
2256      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2257        {      /* Fall through */
2258        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);  
2259        case OP_ALLANY:
2260        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2261          {                            /* not be updated before SCHECK_PARTIAL. */
2262          SCHECK_PARTIAL();
2263          RRETURN(MATCH_NOMATCH);
2264        }        }
2265      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2266      if (utf8)  #ifdef SUPPORT_UTF
2267        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2268    #endif
2269      ecode++;      ecode++;
2270      break;      break;
2271    
# Line 1435  for (;;) Line 2273  for (;;)
2273      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2274    
2275      case OP_ANYBYTE:      case OP_ANYBYTE:
2276      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2277          {                            /* not be updated before SCHECK_PARTIAL. */
2278          SCHECK_PARTIAL();
2279          RRETURN(MATCH_NOMATCH);
2280          }
2281        eptr++;
2282      ecode++;      ecode++;
2283      break;      break;
2284    
2285      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2286      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2287          {
2288          SCHECK_PARTIAL();
2289          RRETURN(MATCH_NOMATCH);
2290          }
2291      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2292      if (      if (
2293  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2294         c < 256 &&         c < 256 &&
2295  #endif  #endif
2296         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1453  for (;;) Line 2300  for (;;)
2300      break;      break;
2301    
2302      case OP_DIGIT:      case OP_DIGIT:
2303      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2304          {
2305          SCHECK_PARTIAL();
2306          RRETURN(MATCH_NOMATCH);
2307          }
2308      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2309      if (      if (
2310  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2311         c >= 256 ||         c > 255 ||
2312  #endif  #endif
2313         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2314         )         )
# Line 1466  for (;;) Line 2317  for (;;)
2317      break;      break;
2318    
2319      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2320      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2321          {
2322          SCHECK_PARTIAL();
2323          RRETURN(MATCH_NOMATCH);
2324          }
2325      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2326      if (      if (
2327  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2328         c < 256 &&         c < 256 &&
2329  #endif  #endif
2330         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1479  for (;;) Line 2334  for (;;)
2334      break;      break;
2335    
2336      case OP_WHITESPACE:      case OP_WHITESPACE:
2337      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2338          {
2339          SCHECK_PARTIAL();
2340          RRETURN(MATCH_NOMATCH);
2341          }
2342      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2343      if (      if (
2344  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2345         c >= 256 ||         c > 255 ||
2346  #endif  #endif
2347         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2348         )         )
# Line 1492  for (;;) Line 2351  for (;;)
2351      break;      break;
2352    
2353      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2354      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2355          {
2356          SCHECK_PARTIAL();
2357          RRETURN(MATCH_NOMATCH);
2358          }
2359      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2360      if (      if (
2361  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2362         c < 256 &&         c < 256 &&
2363  #endif  #endif
2364         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1505  for (;;) Line 2368  for (;;)
2368      break;      break;
2369    
2370      case OP_WORDCHAR:      case OP_WORDCHAR:
2371      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2372          {
2373          SCHECK_PARTIAL();
2374          RRETURN(MATCH_NOMATCH);
2375          }
2376      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2377      if (      if (
2378  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2379         c >= 256 ||         c > 255 ||
2380  #endif  #endif
2381         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2382         )         )
# Line 1518  for (;;) Line 2385  for (;;)
2385      break;      break;
2386    
2387      case OP_ANYNL:      case OP_ANYNL:
2388      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2389          {
2390          SCHECK_PARTIAL();
2391          RRETURN(MATCH_NOMATCH);
2392          }
2393      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2394      switch(c)      switch(c)
2395        {        {
2396        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2397    
2398        case 0x000d:        case 0x000d:
2399        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        if (eptr >= md->end_subject)
2400            {
2401            SCHECK_PARTIAL();
2402            }
2403          else if (*eptr == 0x0a) eptr++;
2404        break;        break;
2405    
2406        case 0x000a:        case 0x000a:
# Line 1542  for (;;) Line 2418  for (;;)
2418      break;      break;
2419    
2420      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2421      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2422          {
2423          SCHECK_PARTIAL();
2424          RRETURN(MATCH_NOMATCH);
2425          }
2426      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2427      switch(c)      switch(c)
2428        {        {
# Line 1572  for (;;) Line 2452  for (;;)
2452      break;      break;
2453    
2454      case OP_HSPACE:      case OP_HSPACE:
2455      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2456          {
2457          SCHECK_PARTIAL();
2458          RRETURN(MATCH_NOMATCH);
2459          }
2460      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2461      switch(c)      switch(c)
2462        {        {
# Line 1602  for (;;) Line 2486  for (;;)
2486      break;      break;
2487    
2488      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2489      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2490          {
2491          SCHECK_PARTIAL();
2492          RRETURN(MATCH_NOMATCH);
2493          }
2494      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2495      switch(c)      switch(c)
2496        {        {
# Line 1620  for (;;) Line 2508  for (;;)
2508      break;      break;
2509    
2510      case OP_VSPACE:      case OP_VSPACE:
2511      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2512          {
2513          SCHECK_PARTIAL();
2514          RRETURN(MATCH_NOMATCH);
2515          }
2516      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2517      switch(c)      switch(c)
2518        {        {
# Line 1643  for (;;) Line 2535  for (;;)
2535    
2536      case OP_PROP:      case OP_PROP:
2537      case OP_NOTPROP:      case OP_NOTPROP:
2538      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2539          {
2540          SCHECK_PARTIAL();
2541          RRETURN(MATCH_NOMATCH);
2542          }
2543      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2544        {        {
2545        int chartype, script;        const ucd_record *prop = GET_UCD(c);
       int category = _pcre_ucp_findprop(c, &chartype, &script);  
2546    
2547        switch(ecode[1])        switch(ecode[1])
2548          {          {
# Line 1656  for (;;) Line 2551  for (;;)
2551          break;          break;
2552    
2553          case PT_LAMP:          case PT_LAMP:
2554          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2555               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2556               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2557            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2558           break;          break;
2559    
2560          case PT_GC:          case PT_GC:
2561          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2562            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2563          break;          break;
2564    
2565          case PT_PC:          case PT_PC:
2566          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2567            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2568          break;          break;
2569    
2570          case PT_SC:          case PT_SC:
2571          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2572              RRETURN(MATCH_NOMATCH);
2573            break;
2574    
2575            /* These are specials */
2576    
2577            case PT_ALNUM:
2578            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2579                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2580              RRETURN(MATCH_NOMATCH);
2581            break;
2582    
2583            case PT_SPACE:    /* Perl space */
2584            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2585                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2586                   == (op == OP_NOTPROP))
2587              RRETURN(MATCH_NOMATCH);
2588            break;
2589    
2590            case PT_PXSPACE:  /* POSIX space */
2591            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2592                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2593                 c == CHAR_FF || c == CHAR_CR)
2594                   == (op == OP_NOTPROP))
2595              RRETURN(MATCH_NOMATCH);
2596            break;
2597    
2598            case PT_WORD:
2599            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2600                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2601                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2602            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2603          break;          break;
2604    
2605            /* This should never occur */
2606    
2607          default:          default:
2608          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2609          }          }
# Line 1689  for (;;) Line 2616  for (;;)
2616      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2617    
2618      case OP_EXTUNI:      case OP_EXTUNI:
2619      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2620          {
2621          SCHECK_PARTIAL();
2622          RRETURN(MATCH_NOMATCH);
2623          }
2624      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2625        if (UCD_CATEGORY(c) == ucp_M) RRETURN(MATCH_NOMATCH);
2626        while (eptr < md->end_subject)
2627        {        {
2628        int chartype, script;        int len = 1;
2629        int category = _pcre_ucp_findprop(c, &chartype, &script);        if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2630        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        if (UCD_CATEGORY(c) != ucp_M) break;
2631        while (eptr < md->end_subject)        eptr += len;
         {  
         int len = 1;  
         if (!utf8) c = *eptr; else  
           {  
           GETCHARLEN(c, eptr, len);  
           }  
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
         eptr += len;  
         }  
2632        }        }
2633        CHECK_PARTIAL();
2634      ecode++;      ecode++;
2635      break;      break;
2636  #endif  #endif
# Line 1721  for (;;) Line 2645  for (;;)
2645      loops). */      loops). */
2646    
2647      case OP_REF:      case OP_REF:
2648        {      case OP_REFI:
2649        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2650        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2651        ecode += 1 + IMM2_SIZE;
2652    
2653        /* If the reference is unset, set the length to be longer than the amount      /* If the reference is unset, there are two possibilities:
       of subject left; this ensures that every attempt at a match fails. We  
       can't just fail here, because of the possibility of quantifiers with zero  
       minima. */  
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2654    
2655        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2656        this ensures that every attempt at a match fails. We can't just fail
2657        here, because of the possibility of quantifiers with zero minima.
2658    
2659        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2660          {      so that the back reference matches an empty string.
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
2661    
2662          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2663          case OP_CRMINRANGE:      referenced subpattern. */
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2664    
2665          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2666          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2667          eptr += length;      else
2668          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2669    
2670        /* Set up for repetition, or handle the non-repeated case */
2671    
2672        switch (*ecode)
2673          {
2674          case OP_CRSTAR:
2675          case OP_CRMINSTAR:
2676          case OP_CRPLUS:
2677          case OP_CRMINPLUS:
2678          case OP_CRQUERY:
2679          case OP_CRMINQUERY:
2680          c = *ecode++ - OP_CRSTAR;
2681          minimize = (c & 1) != 0;
2682          min = rep_min[c];                 /* Pick up values from tables; */
2683          max = rep_max[c];                 /* zero for max => infinity */
2684          if (max == 0) max = INT_MAX;
2685          break;
2686    
2687          case OP_CRRANGE:
2688          case OP_CRMINRANGE:
2689          minimize = (*ecode == OP_CRMINRANGE);
2690          min = GET2(ecode, 1);
2691          max = GET2(ecode, 1 + IMM2_SIZE);
2692          if (max == 0) max = INT_MAX;
2693          ecode += 1 + 2 * IMM2_SIZE;
2694          break;
2695    
2696          default:               /* No repeat follows */
2697          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2698            {
2699            if (length == -2) eptr = md->end_subject;   /* Partial match */
2700            CHECK_PARTIAL();
2701            RRETURN(MATCH_NOMATCH);
2702          }          }
2703          eptr += length;
2704          continue;              /* With the main loop */
2705          }
2706    
2707        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2708        main loop. */      zero, just continue with the main loop. If the length is negative, it
2709        means the reference is unset in non-Java-compatible mode. If the minimum is
2710        zero, we can continue at the same level without recursion. For any other
2711        minimum, carrying on will result in NOMATCH. */
2712    
2713        if (length == 0) continue;      if (length == 0) continue;
2714        if (length < 0 && min == 0) continue;
2715    
2716        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2717        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2718        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2719    
2720        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2721          {
2722          int slength;
2723          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2724          {          {
2725          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2726          eptr += length;          CHECK_PARTIAL();
2727            RRETURN(MATCH_NOMATCH);
2728          }          }
2729          eptr += slength;
2730          }
2731    
2732        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2733        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2734    
2735        if (min == max) continue;      if (min == max) continue;
2736    
2737        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2738    
2739        if (minimize)      if (minimize)
2740          {
2741          for (fi = min;; fi++)
2742          {          {
2743          for (fi = min;; fi++)          int slength;
2744            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2745            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2746            if (fi >= max) RRETURN(MATCH_NOMATCH);
2747            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2748            {            {
2749            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2750            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2751            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2752            }            }
2753          /* Control never gets here */          eptr += slength;
2754          }          }
2755          /* Control never gets here */
2756          }
2757    
2758        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2759    
2760        else      else
2761          {
2762          pp = eptr;
2763          for (i = min; i < max; i++)
2764          {          {
2765          pp = eptr;          int slength;
2766          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)  
2767            {            {
2768            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2769            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2770            eptr -= length;  
2771              if (slength == -2 && md->partial != 0 &&
2772                  md->end_subject > md->start_used_ptr)
2773                {
2774                md->hitend = TRUE;
2775                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2776                }
2777              break;
2778            }            }
2779          RRETURN(MATCH_NOMATCH);          eptr += slength;
2780            }
2781    
2782          while (eptr >= pp)
2783            {
2784            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2785            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2786            eptr -= length;
2787          }          }
2788          RRETURN(MATCH_NOMATCH);
2789        }        }
2790      /* Control never gets here */      /* Control never gets here */
2791    
   
   
2792      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2793      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,
2794      and either the matching is caseful, or the characters are in the range      and either the matching is caseful, or the characters are in the range
# Line 1838  for (;;) Line 2803  for (;;)
2803      case OP_NCLASS:      case OP_NCLASS:
2804      case OP_CLASS:      case OP_CLASS:
2805        {        {
2806          /* The data variable is saved across frames, so the byte map needs to
2807          be stored there. */
2808    #define BYTE_MAP ((pcre_uint8 *)data)
2809        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2810        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2811    
2812        switch (*ecode)        switch (*ecode)
2813          {          {
# Line 1860  for (;;) Line 2828  for (;;)
2828          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2829          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2830          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2831          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2832          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2833          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2834          break;          break;
2835    
2836          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1872  for (;;) Line 2840  for (;;)
2840    
2841        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2842    
2843  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2844        /* UTF-8 mode */        if (utf)
       if (utf8)  
2845          {          {
2846          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2847            {            {
2848            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2849                {
2850                SCHECK_PARTIAL();
2851                RRETURN(MATCH_NOMATCH);
2852                }
2853            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2854            if (c > 255)            if (c > 255)
2855              {              {
2856              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2857              }              }
2858            else            else
2859              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2860            }            }
2861          }          }
2862        else        else
2863  #endif  #endif
2864        /* Not UTF-8 mode */        /* Not UTF mode */
2865          {          {
2866          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2867            {            {
2868            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2869                {
2870                SCHECK_PARTIAL();
2871                RRETURN(MATCH_NOMATCH);
2872                }
2873            c = *eptr++;            c = *eptr++;
2874            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2875              if (c > 255)
2876                {
2877                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2878                }
2879              else
2880    #endif
2881                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2882            }            }
2883          }          }
2884    
# Line 1912  for (;;) Line 2892  for (;;)
2892    
2893        if (minimize)        if (minimize)
2894          {          {
2895  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2896          /* UTF-8 mode */          if (utf)
         if (utf8)  
2897            {            {
2898            for (fi = min;; fi++)            for (fi = min;; fi++)
2899              {              {
2900              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2901              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2902              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2903                if (eptr >= md->end_subject)
2904                  {
2905                  SCHECK_PARTIAL();
2906                  RRETURN(MATCH_NOMATCH);
2907                  }
2908              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2909              if (c > 255)              if (c > 255)
2910                {                {
2911                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2912                }                }
2913              else              else
2914                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2915              }              }
2916            }            }
2917          else          else
2918  #endif  #endif
2919          /* Not UTF-8 mode */          /* Not UTF mode */
2920            {            {
2921            for (fi = min;; fi++)            for (fi = min;; fi++)
2922              {              {
2923              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2924              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2925              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2926                if (eptr >= md->end_subject)
2927                  {
2928                  SCHECK_PARTIAL();
2929                  RRETURN(MATCH_NOMATCH);
2930                  }
2931              c = *eptr++;              c = *eptr++;
2932              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2933                if (c > 255)
2934                  {
2935                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2936                  }
2937                else
2938    #endif
2939                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2940              }              }
2941            }            }
2942          /* Control never gets here */          /* Control never gets here */
# Line 1954  for (;;) Line 2948  for (;;)
2948          {          {
2949          pp = eptr;          pp = eptr;
2950    
2951  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2952          /* UTF-8 mode */          if (utf)
         if (utf8)  
2953            {            {
2954            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2955              {              {
2956              int len = 1;              int len = 1;
2957              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2958                  {
2959                  SCHECK_PARTIAL();
2960                  break;
2961                  }
2962              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2963              if (c > 255)              if (c > 255)
2964                {                {
2965                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2966                }                }
2967              else              else
2968                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2969              eptr += len;              eptr += len;
2970              }              }
2971            for (;;)            for (;;)
2972              {              {
2973              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2974              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2975              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2976              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1983  for (;;) Line 2978  for (;;)
2978            }            }
2979          else          else
2980  #endif  #endif
2981            /* Not UTF-8 mode */            /* Not UTF mode */
2982            {            {
2983            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2984              {              {
2985              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2986                  {
2987                  SCHECK_PARTIAL();
2988                  break;
2989                  }
2990              c = *eptr;              c = *eptr;
2991              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2992                if (c > 255)
2993                  {
2994                  if (op == OP_CLASS) break;
2995                  }
2996                else
2997    #endif
2998                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
2999              eptr++;              eptr++;
3000              }              }
3001            while (eptr >= pp)            while (eptr >= pp)
3002              {              {
3003              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3004              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3005              eptr--;              eptr--;
3006              }              }
# Line 2002  for (;;) Line 3008  for (;;)
3008    
3009          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3010          }          }
3011    #undef BYTE_MAP
3012        }        }
3013      /* Control never gets here */      /* Control never gets here */
3014    
3015    
3016      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3017      in UTF-8 mode, because that's the only time it is compiled. */      when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
3018        mode, because Unicode properties are supported in non-UTF-8 mode. */
3019    
3020  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3021      case OP_XCLASS:      case OP_XCLASS:
3022        {        {
3023        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2034  for (;;) Line 3042  for (;;)
3042          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3043          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3044          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3045          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3046          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3047          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3048          break;          break;
3049    
3050          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2048  for (;;) Line 3056  for (;;)
3056    
3057        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3058          {          {
3059          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3060          GETCHARINC(c, eptr);            {
3061          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3062              RRETURN(MATCH_NOMATCH);
3063              }
3064            GETCHARINCTEST(c, eptr);
3065            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3066          }          }
3067    
3068        /* If max == min we can continue with the main loop without the        /* If max == min we can continue with the main loop without the
# Line 2065  for (;;) Line 3077  for (;;)
3077          {          {
3078          for (fi = min;; fi++)          for (fi = min;; fi++)
3079            {            {
3080            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3081            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3082            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3083            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3084            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3085                SCHECK_PARTIAL();
3086                RRETURN(MATCH_NOMATCH);
3087                }
3088              GETCHARINCTEST(c, eptr);
3089              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3090            }            }
3091          /* Control never gets here */          /* Control never gets here */
3092          }          }
# Line 2082  for (;;) Line 3099  for (;;)
3099          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3100            {            {
3101            int len = 1;            int len = 1;
3102            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3103            GETCHARLEN(c, eptr, len);              {
3104            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3105                break;
3106                }
3107    #ifdef SUPPORT_UTF
3108              GETCHARLENTEST(c, eptr, len);
3109    #else
3110              c = *eptr;
3111    #endif
3112              if (!PRIV(xclass)(c, data, utf)) break;
3113            eptr += len;            eptr += len;
3114            }            }
3115          for(;;)          for(;;)
3116            {            {
3117            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3118            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3119            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3120            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3121              if (utf) BACKCHAR(eptr);
3122    #endif
3123            }            }
3124          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3125          }          }
# Line 2104  for (;;) Line 3131  for (;;)
3131      /* Match a single character, casefully */      /* Match a single character, casefully */
3132    
3133      case OP_CHAR:      case OP_CHAR:
3134  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3135      if (utf8)      if (utf)
3136        {        {
3137        length = 1;        length = 1;
3138        ecode++;        ecode++;
3139        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3140        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3141            {
3142            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3143            RRETURN(MATCH_NOMATCH);
3144            }
3145        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3146        }        }
3147      else      else
3148  #endif  #endif
3149        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3150        {        {
3151        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3152            {
3153            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3154            RRETURN(MATCH_NOMATCH);
3155            }
3156        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3157        ecode += 2;        ecode += 2;
3158        }        }
3159      break;      break;
3160    
3161      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3162        subject, give up immediately. */
3163    
3164        case OP_CHARI:
3165        if (eptr >= md->end_subject)
3166          {
3167          SCHECK_PARTIAL();
3168          RRETURN(MATCH_NOMATCH);
3169          }
3170    
3171      case OP_CHARNC:  #ifdef SUPPORT_UTF
3172  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3173        {        {
3174        length = 1;        length = 1;
3175        ecode++;        ecode++;
3176        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3177    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3178        /* 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
3179        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3180          fast lookup table. We know that there is at least one byte left in the
3181          subject. */
3182    
3183        if (fc < 128)        if (fc < 128)
3184          {          {
3185          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3186                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3187            ecode++;
3188            eptr++;
3189          }          }
3190    
3191        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3192          use the value of "length" to check for sufficient bytes left, because the
3193          other case of the character may have more or fewer bytes.  */
3194    
3195        else        else
3196          {          {
# Line 2158  for (;;) Line 3204  for (;;)
3204          if (fc != dc)          if (fc != dc)
3205            {            {
3206  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3207            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3208  #endif  #endif
3209              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3210            }            }
3211          }          }
3212        }        }
3213      else      else
3214  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3215    
3216      /* Non-UTF-8 mode */      /* Not UTF mode */
3217        {        {
3218        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3219        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3220          eptr++;
3221        ecode += 2;        ecode += 2;
3222        }        }
3223      break;      break;
# Line 2178  for (;;) Line 3225  for (;;)
3225      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3226    
3227      case OP_EXACT:      case OP_EXACT:
3228        case OP_EXACTI:
3229      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3230      ecode += 3;      ecode += 1 + IMM2_SIZE;
3231      goto REPEATCHAR;      goto REPEATCHAR;
3232    
3233      case OP_POSUPTO:      case OP_POSUPTO:
3234        case OP_POSUPTOI:
3235      possessive = TRUE;      possessive = TRUE;
3236      /* Fall through */      /* Fall through */
3237    
3238      case OP_UPTO:      case OP_UPTO:
3239        case OP_UPTOI:
3240      case OP_MINUPTO:      case OP_MINUPTO:
3241        case OP_MINUPTOI:
3242      min = 0;      min = 0;
3243      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3244      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3245      ecode += 3;      ecode += 1 + IMM2_SIZE;
3246      goto REPEATCHAR;      goto REPEATCHAR;
3247    
3248      case OP_POSSTAR:      case OP_POSSTAR:
3249        case OP_POSSTARI:
3250      possessive = TRUE;      possessive = TRUE;
3251      min = 0;      min = 0;
3252      max = INT_MAX;      max = INT_MAX;
# Line 2202  for (;;) Line 3254  for (;;)
3254      goto REPEATCHAR;      goto REPEATCHAR;
3255    
3256      case OP_POSPLUS:      case OP_POSPLUS:
3257        case OP_POSPLUSI:
3258      possessive = TRUE;      possessive = TRUE;
3259      min = 1;      min = 1;
3260      max = INT_MAX;      max = INT_MAX;
# Line 2209  for (;;) Line 3262  for (;;)
3262      goto REPEATCHAR;      goto REPEATCHAR;
3263    
3264      case OP_POSQUERY:      case OP_POSQUERY:
3265        case OP_POSQUERYI:
3266      possessive = TRUE;      possessive = TRUE;
3267      min = 0;      min = 0;
3268      max = 1;      max = 1;
# Line 2216  for (;;) Line 3270  for (;;)
3270      goto REPEATCHAR;      goto REPEATCHAR;
3271    
3272      case OP_STAR:      case OP_STAR:
3273        case OP_STARI:
3274      case OP_MINSTAR:      case OP_MINSTAR:
3275        case OP_MINSTARI:
3276      case OP_PLUS:      case OP_PLUS:
3277        case OP_PLUSI:
3278      case OP_MINPLUS:      case OP_MINPLUS:
3279        case OP_MINPLUSI:
3280      case OP_QUERY:      case OP_QUERY:
3281        case OP_QUERYI:
3282      case OP_MINQUERY:      case OP_MINQUERY:
3283      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3284        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3285      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3286      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3287      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3288      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3289    
3290      /* 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. */  
3291    
3292      REPEATCHAR:      REPEATCHAR:
3293  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3294      if (utf8)      if (utf)
3295        {        {
3296        length = 1;        length = 1;
3297        charptr = ecode;        charptr = ecode;
3298        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3299        ecode += length;        ecode += length;
3300    
3301        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2248  for (;;) Line 3305  for (;;)
3305          {          {
3306  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3307          unsigned int othercase;          unsigned int othercase;
3308          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3309              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3310            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3311          else oclength = 0;          else oclength = 0;
3312  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3313    
3314          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3315            {            {
3316            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3317                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3318  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3319            /* Need braces because of following else */            else if (oclength > 0 &&
3320            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3321                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3322    #endif  /* SUPPORT_UCP */
3323            else            else
3324              {              {
3325              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3326              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3327              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3328            }            }
3329    
3330          if (min == max) continue;          if (min == max) continue;
# Line 2276  for (;;) Line 3333  for (;;)
3333            {            {
3334            for (fi = min;; fi++)            for (fi = min;; fi++)
3335              {              {
3336              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3337              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3338              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3339              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3340                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3341  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3342              /* Need braces because of following else */              else if (oclength > 0 &&
3343              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3344                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3345    #endif  /* SUPPORT_UCP */
3346              else              else
3347                {                {
3348                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3349                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3350                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3351              }              }
3352            /* Control never gets here */            /* Control never gets here */
3353            }            }
# Line 2300  for (;;) Line 3357  for (;;)
3357            pp = eptr;            pp = eptr;
3358            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3359              {              {
3360              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3361              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3362  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3363              else if (oclength == 0) break;              else if (oclength > 0 &&
3364                         eptr <= md->end_subject - oclength &&
3365                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3366    #endif  /* SUPPORT_UCP */
3367              else              else
3368                {                {
3369                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3370                eptr += oclength;                break;
3371                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3372              }              }
3373    
3374            if (possessive) continue;            if (possessive) continue;
3375    
3376            for(;;)            for(;;)
3377             {              {
3378             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3379             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3380             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3381  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3382             eptr--;              eptr--;
3383             BACKCHAR(eptr);              BACKCHAR(eptr);
3384  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3385             eptr -= length;              eptr -= length;
3386  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3387             }              }
3388            }            }
3389          /* Control never gets here */          /* Control never gets here */
3390          }          }
# Line 2336  for (;;) Line 3394  for (;;)
3394        value of fc will always be < 128. */        value of fc will always be < 128. */
3395        }        }
3396      else      else
3397  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3398          /* 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);  
3399        fc = *ecode++;        fc = *ecode++;
       }  
3400    
3401      /* 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
3402      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
3403      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3404      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3405      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2356  for (;;) Line 3410  for (;;)
3410      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3411        max, eptr));        max, eptr));
3412    
3413      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3414        {        {
3415        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3416          /* fc must be < 128 if UTF is enabled. */
3417          foc = md->fcc[fc];
3418    #else
3419    #ifdef SUPPORT_UTF
3420    #ifdef SUPPORT_UCP
3421          if (utf && fc > 127)
3422            foc = UCD_OTHERCASE(fc);
3423    #else
3424          if (utf && fc > 127)
3425            foc = fc;
3426    #endif /* SUPPORT_UCP */
3427          else
3428    #endif /* SUPPORT_UTF */
3429            foc = TABLE_GET(fc, md->fcc, fc);
3430    #endif /* COMPILE_PCRE8 */
3431    
3432        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3433          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3434            if (eptr >= md->end_subject)
3435              {
3436              SCHECK_PARTIAL();
3437              RRETURN(MATCH_NOMATCH);
3438              }
3439            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3440            eptr++;
3441            }
3442        if (min == max) continue;        if (min == max) continue;
3443        if (minimize)        if (minimize)
3444          {          {
3445          for (fi = min;; fi++)          for (fi = min;; fi++)
3446            {            {
3447            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3448            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3449            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3450                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3451                {
3452                SCHECK_PARTIAL();
3453              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3454                }
3455              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3456              eptr++;
3457            }            }
3458          /* Control never gets here */          /* Control never gets here */
3459          }          }
# Line 2379  for (;;) Line 3462  for (;;)
3462          pp = eptr;          pp = eptr;
3463          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3464            {            {
3465            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3466                {
3467                SCHECK_PARTIAL();
3468                break;
3469                }
3470              if (fc != *eptr && foc != *eptr) break;
3471            eptr++;            eptr++;
3472            }            }
3473    
3474          if (possessive) continue;          if (possessive) continue;
3475    
3476          while (eptr >= pp)          while (eptr >= pp)
3477            {            {
3478            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3479            eptr--;            eptr--;
3480            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3481            }            }
# Line 2396  for (;;) Line 3486  for (;;)
3486    
3487      /* Caseful comparisons (includes all multi-byte characters) */      /* Caseful comparisons (includes all multi-byte characters) */
3488    
3489      else      else
3490        {        {
3491        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3492            {
3493            if (eptr >= md->end_subject)
3494              {
3495              SCHECK_PARTIAL();
3496              RRETURN(MATCH_NOMATCH);
3497              }
3498            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3499            }
3500    
3501        if (min == max) continue;        if (min == max) continue;
3502    
3503        if (minimize)        if (minimize)
3504          {          {
3505          for (fi = min;; fi++)          for (fi = min;; fi++)
3506            {            {
3507            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3508            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3509            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3510              if (eptr >= md->end_subject)
3511                {
3512                SCHECK_PARTIAL();
3513              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3514                }
3515              if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3516            }            }
3517          /* Control never gets here */          /* Control never gets here */
3518          }          }
# Line 2416  for (;;) Line 3521  for (;;)
3521          pp = eptr;          pp = eptr;
3522          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3523            {            {
3524            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3525                {
3526                SCHECK_PARTIAL();
3527                break;
3528                }
3529              if (fc != *eptr) break;
3530            eptr++;            eptr++;
3531            }            }
3532          if (possessive) continue;          if (possessive) continue;
3533    
3534          while (eptr >= pp)          while (eptr >= pp)
3535            {            {
3536            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3537            eptr--;            eptr--;
3538            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3539            }            }
# Line 2435  for (;;) Line 3546  for (;;)
3546      checking can be multibyte. */      checking can be multibyte. */
3547    
3548      case OP_NOT:      case OP_NOT:
3549      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3550        if (eptr >= md->end_subject)
3551          {
3552          SCHECK_PARTIAL();
3553          RRETURN(MATCH_NOMATCH);
3554          }
3555      ecode++;      ecode++;
3556      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
3557      if ((ims & PCRE_CASELESS) != 0)      if (op == OP_NOTI)         /* The caseless case */
3558        {        {
3559  #ifdef SUPPORT_UTF8        register unsigned int ch, och;
3560        if (c < 256)        ch = *ecode++;
3561  #endif  #ifdef COMPILE_PCRE8
3562        c = md->lcc[c];        /* ch must be < 128 if UTF is enabled. */
3563        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);        och = md->fcc[ch];
3564    #else